FUNDAMENTAL SOLUTIONS IN ELASTODYNAMICS

This work is a compilation of fundamental solutions (or Green’s functions) for
classical or canonical problems in elastodynamics, presented with a common
format and notation. These formulas describe the displacements and stresses
elicited by transient and harmonic sources in solid elastic media such as full
spaces, half-spaces, and strata and plates in both two and three dimensions, using
the three major coordinate systems (Cartesian, cylindrical, and spherical). Such
formulas are useful for numerical methods and practical application to prob-
lems of wave propagation in elasticity, soil dynamics, earthquake engineering,
mechanical vibration, or geophysics. Together with the plots of the response
functions, this work should serve as a valuable reference to engineers and sci-
entists alike. These formulas were heretofore found only scattered throughout
the literature. The solutions are tabulated without proof, but giving reference
to appropriate modern papers and books containing full derivations. Most for-
mulas in the book have been programmed and tested within the MATLAB
environment, and the programs thus developed are both listed and available
for free download.

Eduardo Kausel earned his first professional degree in 1967, graduating as a
civil engineer from the University of Chile, and then worked at Chile’s National
Electricity Company. In 1969 he carried out postgraduate studies at the Tech-
nical University in Darmstadt. He earned his Master of Science (1972) and
Doctor of Science (1974) degrees from MIT. Following graduation, Dr. Kausel
worked at Stone and Webster Engineering Corporation in Boston, and then
joined the MIT faculty in 1978, where he has remained since. He is a regis-
tered professional engineer in the State of Massachusetts, is a senior member
of various professional organizations (ASCE, SSA, EERI, IACMG), and has
extensive experience as a consulting engineer.
Among the honors he has received are a 1989 Japanese Government
Research Award for Foreign Specialists from the Science and Technology
Agency, a 1992 Honorary Faculty Membership in Epsilon Chi, the 1994 Konrad
Zuse Guest Professorship at the University of Hamburg in Germany, the
Humboldt Prize from the German Government in 2000, and the 2001 MIT-
CEE Award for Conspicuously Effective Teaching.
Dr. Kausel is best known for his work on dynamic soil–structure interaction,
and for his very successful Green’s functions (fundamental solutions) for the
dynamic analysis of layered media, which are incorporated in a now widely
used program. Dr. Kausel is the author of more than 150 technical papers
and reports in the areas of structural dynamics, earthquake engineering, and
computational mechanics.
i

ii

Fundamental Solutions
in Elastodynamics
A Compendium

EDUARDO KAUSEL
Massachusetts Institute of Technology

iii

  
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo

Cambridge University Press
The Edinburgh Building, Cambridge  , UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521855709

© Eduardo Kausel 2006

This publication is in copyright. Subject to statutory exception and to the provision of
relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.

First published in print format

- ---- eBook (NetLibrary)
- --- eBook (NetLibrary)

- ---- hardback
- --- hardback

Cambridge University Press has no responsibility for the persistence or accuracy of s
for external or third-party internet websites referred to in this publication, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.

6 Dipoles in cylindrical coordinates 32 SECTION II: FULL SPACE PROBLEMS 3.1 Cartesian coordinates 13 1. stresses.3.4 Strains.4.3. Two-dimensional problems in full.3 Coordinate systems and differential operators 4 1.2 Cylindrical coordinates 6 1.1 Fundamental identities and definitions 35 3.3.4.2 Anti-plane line load (SH waves) 35 3.4 In-plane line load (SV-P waves) 38 v . Fundamentals 1 1. and the elastic wave equation 13 1.2 Cylindrical coordinates 16 1.4. homogeneous spaces 35 3.1 Cartesian coordinates 4 1.4 Seismic moments (double couples with no net resultant) 30 2.3 Spherical coordinates 21 2.5 Blast loads (explosive line and point sources) 31 2.3 Torsional point sources 30 2. Dipoles 27 2.2 Sign convention 4 1.3 SH line load in an orthotropic space 36 3.1 Notation and table of symbols 1 1.Contents Preface page ix SECTION I: PRELIMINARIES 1.3 Spherical coordinates 9 1.2 Line dipoles 29 2.1 Point dipoles or doublets: single couples and tensile crack sources 27 2.

5 Half-plane.3 Plate with mixed boundary conditions subjected to SV-P line source 90 7.1 Solution using the method of images 89 7. Three-dimensional problems in homogeneous half-spaces 78 6.6 Line blast source: suddenly applied pressure 43 3.5 Torsional point source 55 4.3 Tension cracks 52 4.1 Half-plane. or vice versa 73 5. Three-dimensional problems in full.3 Half-plane.8 Spherical cavity subjected to arbitrary pressure 59 4.2 Normal mode solution 87 7.2 3-D half-space. homogeneous spaces 48 4.3 3-D half-space.2. Two-dimensional problems in homogeneous half-spaces 69 5. SH line source and receiver anywhere 69 5. suddenly applied vertical point load on its surface (Pekeris-Mooney’s problem) 78 6. line blast load applied in its interior (Garvin’s problem) 76 6.2 Point load (Stokes problem) 48 4.1 Solution using the method of images 91 7.2 Stratum subjected to SH line source 88 7. SV-P source and receiver at surface (Lamb’s problem) 71 5.4 Half-plane.7 Cylindrical cavity subjected to pulsating pressure 44 4. SV-P source on surface.9 Spatially harmonic line source (21/2-D problem) 63 SECTION III: HALF-SPACE PROBLEMS 5. suddenly applied horizontal point load on its surface (Chao’s problem) 81 6.1 Plate subjected to SH line source 87 7.3.5 Dipoles in plane strain 40 3.3. receiver at interior.2 Normal mode solution 92 .6 Torsional point source with vertical axis 57 4.2 SH line load in an orthotropic half-plane 70 5.1. Two-dimensional problems in homogeneous plates and strata 87 7.2 Normal mode solution 90 7. buried torsional point source with vertical axis 83 SECTION IV: PLATES AND STRATA 7.2.vi Contents 3.1 Fundamental identities and definitions 48 4.7 Point blast source 58 4.1 3-D half-space.1.1 Solution using the method of images 87 7.4 Double couples (seismic moments) 54 4.

3.4.4 Elastic wave equation in Cartesian coordinates 104 8.1 Differential equation 185 11.5 Scalar Helmholtz equation in cylindrical coordinates 108 8.2 Cylindrically stratified media 136 9.9 Vector Helmholtz equation in spherical coordinates 118 8.2.1 Horizontally stratified media 130 9.2 Radially layered system 164 10.2. Stiffness matrix method for layered media 140 10.2.1.2 Numerical computation of stiffness matrices 150 10.10 Elastic wave equation in spherical coordinates 120 9.4.3 Expansion of source and displacements into spherical harmonics 181 10.Contents vii SECTION V: ANALYTICAL AND NUMERICAL METHODS Read me first 97 8.8 Scalar Helmholtz equation in spherical coordinates 117 8.1 Summary of method 141 10.3 Vector Helmholtz equation in Cartesian coordinates 102 8.3 Stiffness matrix method in cylindrical coordinates 159 10.3 Summary of computation 151 10.2 Radially stratified media 114 8.1 Summary of results 98 8.1 Horizontally stratified media.1 Horizontally layered system 160 10.4.7.1 Properties and use of impedance matrices 180 10.2 Stiffness matrix method in Cartesian coordinates 142 10.1. Solution to the Helmholtz and wave equations 98 8.3 Spherical coordinates 137 10.1 Cartesian coordinates 125 9.4.1 Horizontally stratified media 112 8.2 Cylindrical coordinates 130 9.7 Elastic wave equation in cylindrical coordinates 110 8.2.6 Vector Helmholtz equation in cylindrical coordinates 109 8.2 Asymmetry 180 10. Basic properties of mathematical functions 185 11.2 Scalar Helmholtz equation in Cartesian coordinates 101 8. plane strain 105 8.1 Bessel functions 185 11.1 Analytic continuation in the layers 148 10.4.7.4 Rigid body spheroidal modes 182 SECTION VI: APPENDICES 11.2.4 Stiffness matrix method for layered spheres 175 10.3.2 Recurrence relations 185 . Integral transform method 125 9.

4.4.1 Differential equation 187 11.viii Contents 11.4 Wronskians 186 11.4.3.5 Orthogonality conditions 186 11. Brief listing of integral transforms 194 12.3 Recurrence relations 189 11.4.3 Derivatives 186 11.3.1 Fourier transforms 194 12.4.4.4 Orthogonality of co-latitude matrix 193 11.2 Recurrence relations 192 11.2.6 Leibniz rule for the derivative of a product of two functions 193 12.3.1 Differential equation 191 11.1.4 Associated Legendre functions (spheroidal harmonics) 191 11.1.2 Rodrigues’s formula 190 11.2.6 Expansion in Legendre series 191 11.3 Spherical Hankel transforms 197 13.2 Hankel transforms 196 12.3.2 Spherical Bessel functions 187 11.4 Recurrence relations 190 11.2. MATLAB programs 198 .3.5 Expansion of arbitrary function in spheroidal harmonics 193 11.3 Trigonometric expansion (x = cos φ) 190 11.5 Orthogonality condition 190 11.2 Trigonometric representations 188 11.3 Orthogonality conditions 192 11.1.3.6 Useful integrals 187 11.1 Differential equation 189 11.1.3 Legendre polynomials 189 11.

for example. but making no effort at establishing the original sources of the derivations or. Sometimes. constitute invaluable tools for a large class of numerical solution techniques for wave propagation problems in elasticity. while we anguished initially at the choice of symbols for the angles in spherical coordinates. Also. elastic media. and sign convention. providing a historical account of ix . and in both two and three dimensions. these solutions can be extended to spatially distributed sources and/or sources with an arbitrary variation in time. or they include only partial results. for that matter. paper. say only the solution in the frequency domain or for some particular value of Poisson’s ratio. and plates of infinite lateral extent. which is often used to obtain the solution to wave propagation problems in finite bodies of irregular shape. it provides consistency between spherical and cylindrical coordinates and eases the transition between one and the other system. while the sources range from point and line forces to torques. or so-called Green’s func- tions. The bodies considered are full spaces. Thus. These fundamental solutions. but displacements up. and pressure pulses. We tabulate these solutions herein without proof.Preface We present in this work a collection of fundamental solutions. Although this contravenes the common notation. and so forth. By appropriate convolutions. each author. Examples are the Boundary Integral (or element) Method (BIM). or scaling the displays in unusual ways or using too small a scale. With this in mind. soil dynamics. and no single book was found to deal with them all in one place. which should greatly facilitate the application of these fundamental solutions. The solutions included herein are found scattered throughout the literature. or book uses sign conventions and symbols that differ from one another. for some classical or canonical problems in elastodynamics. taking forces to be positive down. coordinate systems. as their name implies. In addition. we decided in the end to use θ for the azimuth and φ for the polar angle. even while working with the Green’s functions for a full space. published results are also displayed in unconventional manners. seismic moments. or geophysics. giving reference to appropriate mod- ern papers and books containing full derivations. in both the frequency domain and the time domain. Such formulas provide the dynamic response functions for transient point sources acting within isotropic. we use throughout a consistent notation. earthquake engineering. it was felt that a compendium of the known solutions in a common format would serve a useful purpose. half-spaces.

the tables may not necessarily be complete in that solutions for some additional clas- sical problems. Also. not readily available. September 2005 . we give no references. or for purely pragmatic reasons. Also. either because an appropriate reference was not known to us. the possibility always exists that errors may remain undetected in some of these formulas. In some cases. we have made every effort at checking the formulas themselves for correctness and dimensional consistency. in which case we have developed the formulae ourselves using established methods. Yet. If the reader should find any such errors. Eduardo Kausel Cambridge. If and when these are brought to our attention. or important extensions to these. and provide plots of response functions that could be used to verify the correctness of a particular implementation. we have programmed most formulas within the MATLAB or other program- ming environment. Nonetheless. may exist of which we may be unaware. we shall be thankful if they are brought to our attention. we do not claim to have discovered new formulas. Finally. we shall be happy to consider them with proper credit when preparing a revised version of this work.x Preface these solutions. recogniz- ing that these are all classical problems.

Hn (kr ) First and second Hankel functions of order n (Bessel functions of the third kind) Hn 3×3 displacement matrix.and P-wave velocities b = bj eˆ j Body load vector CR Rayleigh-wave velocity Cn 3×3 Bessel matrix.k = ∂gij /∂xk Derivative with respect to the receiver location  gij.k = ∂gij /∂ xk Derivative with respect to the source location Gij Green’s function for the frequency domain response due to a dipole (1) (2) hn (kR).3) (1) (2) Hn .7) 1 .4) (2) F(1) n . Hn As Hn above. assembled with first and second Hankel functions Hm 3×3 spherical Bessel matrix. Fn As Fn above. assembled with first and second Hankel functions F Fourier transform operator F−1 Inverse Fourier transform operator gij Green’s function for the frequency domain response in direction i due to a unit load in direction j gij.2) gj = gij eˆ i Green’s function vector for the frequency domain response due to a unit load in direction j Fn 3×3 traction matrix. cylindrical coordinates (see Table 10.1 Notation and table of symbols Except where noted. the following symbols will be used consistently throughout this work: a = β /α Ratio of S. cylindrical coordinates and flat layers (see Table 10. hn (kR) First and second spherical Hankel functions of order n (1) (2) Hn (kr ). spherical coordinates (see Table 10. SECTION I: PRELIMINARIES 1 Fundamentals 1. cylindrical coordinates (see Table 10.

3) s = 1 − (kS /k) 2 Dimensionless vertical wavenumber for S waves t Time t P = r/α P-wave arrival time t S = r/β S-wave arrival time . √ but vertical component multiplied by −i = − −1 p  Pressure (positive when compressive) p = 1 − (kP /k) 2 Dimensionless vertical wavenumber for P waves P Load amplitude Pm Legendre function (polynomial) of the first kind Pmn Associated Legendre function of the first kind Qmn Associated Legendre function of the second kind R Source–receiver distance in 3-D space r Source–receiver distance in 2-D space.3. y. or seismic moment Mij Displacement in direction i due to seismic moment with axis j. Table 10.6) k Radial wavenumber kP = ω/α P wavenumber kS = ω/β S wavenumber kz  Vertical wavenumber kp = k2 − kP2 Vertical wavenumber for P waves. Yn (kr ) Bessel functions of the first and second kind J 3×3 spherical orthogonality condition (Section 9. j. eˆ 3 ˆı. eˆ 2 . 10. θ. Orthogonal unit basis vectors in Cartesian coordinates Jn (kr ). 3 or coordinates x. cylindrical layers Lnm 3×3 Spheroidal (co-latitude) matrix (see Tables 10. 2. flat layers  kα = kP2 − kz2 Radial wavenumber for P waves. z Cylindrical coordinates (see Fig. or Heaviside function ⎩2 √ 1 t ≥ tα i = −1 Imaginary unit (non-italicized) i. tˆ Orthogonal unit basis vectors in spherical coordinates R. tˆ. z ˆ kˆ ≡ eˆ 1 .8) M Intensity of moment. sˆ . cylindrical layers  kβ = kS2 − kz2 Radial wavenumber for S waves. k Sub-indices for the numbers 1. torque.7. j.2 Fundamentals ⎧ ⎨0 t < tα H(t − tα ) = 1 t = tα Unit step function.2) rˆ .θ Spherical coordinates (see Fig. flat layers  ks = k2 − kS2 Vertical wavenumber for S waves. 1. p Load vector p˜ Load vector in frequency–wavenumber domain p As above. kˆ Orthogonal unit basis vectors in cylindrical coordinates rˆ . φ. or range r. 1.

uθ . Displacement in direction x (or i) due to force in direction z (or j) Uij .1 Notation and table of symbols 3 t R = r/CR Rayleigh-wave arrival time T Torque Tn Azimuthal matrix Tij Displacement in direction i due to a unit torque with axis j. t ≥ tS  1. uz ≡ u1 . v. u Displacement vector u˜ Displacement vector in frequency-wavenumber domain u As above. uy . etc. δ(t − t S ) = = ∞. 2.1) x . i = j λ Lame´ constant λ + 2µ = ρ α 2 Constrained modulus µ = ρβ 2 Shear modulus ν Poisson’s ratio τ = tβ/r = t/tS Dimensionless time ρ Mass density θ Azimuth in cylindrical and spherical coordinates θi Angle between R and the ith axis (γ i = cos θ i ). w Displacement components in cylindrical coordinates uR . a Helmholtz shear potential) X = X(t) Inverse Fourier transform of χ (ω) (or the convolution of the latter with an arbitrary time function) ψ = ψ(ω) Dimensionless component function of Green’s functions (in later chapters. uz ≡ u. 3 Dilatational Helmholtz potential χ = χ(ω) Dimensionless component function of Green’s functions (in later chapters. i = j δi j = Kronecker delta 0. i = 1. a Helmholtz shear potential)  = (t) Inverse Fourier transform of ψ(ω) (or the convolution of the latter with an arbitrary time function) . t = t S Dirac-delta singularity function dt ⎩ 0.1. u3 Displacement components in Cartesian coordinates ur . u2 . 1. z ≡ x1 . uθ Displacement components in spherical coordinates uxz or uij . z Cartesian coordinates of the source α = β 2(1 − ν)/(1 − 2ν) P-wave velocity √ β = µ/ρ S-wave velocity γi Direction cosine of R with ith axis (see Cartesian coordinates) ⎧ t < tS dH(t − t S ) ⎨ 0. uij Green’s function for the time domain response in direction i due to a unit load in direction j x. x2 . y . y. √ but vertical component multiplied by −i = − −1 ux . uφ . x3 Cartesian coordinates (see Fig.

and that z = x3 is up.. are always defined positive in the positive coordinate directions. Point and line sources will usually – but not always – be located at the origin of coordi- nates. 1.e. we shall assume that x = x1 and y = y2 lie in a horizontal plane.1) −∞ −∞  2  +∞  +∞ 1 f (t. either in cylindrical or in spherical coordinates. the underlying Fourier transform pairs from frequency– wavenumber domain to the space–time domain are of the form  +∞  +∞ f (ω. this convention facilitates the conversion between Cartesian and either cylindrical or spherical coordinates. which is perpendicular to the plane of wave propagation. For two- dimensional (plane strain) problems. it provides a conve- nient x–y reference system when working in horizontal planes (i.3 Coordinate systems and differential operators We choose Cartesian coordinates in three-dimensional space forming a right-handed sys- tem. x)] (1. Nonetheless. and plots of displacements are always shown upright (i. and we denote these indifferently as either x. In most cases. k) = f (t.1a)  Source–receiver distance R = x 2 + y2 + z2 (1. x1 . relate to the principles of radiation. and the anti-plane (or SH) compo- nents will be in the horizontal direction y (or x2 ). in a bird’s-eye view). x2 . Also. on the other. z or x1 . and casting them in the frequency domain.4 Fundamentals Ψ Helmholtz vector potential (shear) ω Frequency (rad/s) S = ωr/β = ωtS Dimensionless frequency for S (shear) waves P = ωr/α = ωtP Dimensionless frequency for P (dilatational) waves 1. such as displacements and forces. in turn. x3 . y. boundedness at infinity. x3 ). never reversed or upside down). k)ei(ωt−kx) dω dk = F −1 [ f (ω. These. Wave propagation in the frequency–wavenumber domain will assume a dependence of the form exp i(ωt − kx). this convention calls for the use of second (cylindrical or spherical) Hankel functions when formulating wave propagation problems in infinite media.e.e.3.1 Cartesian coordinates a) Three-dimensional space (Fig. 1. When this is not the case. the in-plane (or SV-P) components will be contained in the vertical plane defined by x and z (i. it will be indicated explicitly.2 Sign convention Component of vectors. and causality. k)] (1. x) = f (ω.2) 2π −∞ −∞ Important consequences of this transformation convention concern the direction of posi- tive wave propagation and decay.3a) . x)e−i(ωt−kx) dt dx = F [ f (t. 1. and the location of poles for the dynamic system under consideration. you may rotate these systems to suit your convenience.. On the one hand.. that is.

1.3b) ∂ xi R ∂γi 1 First derivatives of γi.7) ∂x ∂y ∂z . are tensor bases.3e) γi γi = γ12 + γ22 + γ32 = 1 Nabla operator ∂ ∂ ∂ ∇ = ˆi + jˆ + kˆ (1. that is. SH) Figure 1. For example. or dyads. j k = (3γi γ j γk − γi δ jk − γ j δki − γk δi j ) (1.3 Coordinate systems and differential operators 5 z (or x3) z (or x3) Receiver Receiver R y (or x2) r θz θx ⊗ x (or x1) ⊗ x (or x1 ) Source Source Direction cosines γi a) 3-D b) 2-D (SV-P. two distinct projections of this tensor are ∂ux ˆ ∂u y ˆ ∂uz ˆ ∂u kˆ · ∇u = i+ j+ k= (1. (1. j = = (δi j − γi γ j ) (1.3d) R2 direction cosines Implied summations δii = δ11 + δ22 + δ33 = 3. etc.5) where the products of the form ˆi ˆi.4) ∂x ∂y ∂z Gradient of vector ∂ux ˆ ˆ ∂u y ˆ ˆ ∂uz ˆ ˆ ∂ux ˆ ˆ ∂u y ˆ ˆ ∂uz ˆ ˆ ∂ux ˆ ˆ ∂u y ˆ ˆ ∂uz ˆ ˆ ∇u = ii + ij + ik + ji + jj + jk + ki + kj + kk ∂x ∂x ∂x ∂y ∂y ∂y ∂z ∂z ∂z (1..6) ∂z ∂z ∂z ∂z and ∂uz ˆ ∂uz ˆ ∂uz ˆ (∇u) · kˆ = i+ j+ k = ∇uz (1.3c) ∂xj R direction cosines 1 Second derivatives of γi. ∇u is a tensor. ∂R xi Direction cosines of R γi = cos θi = = (1.1: Cartesian coordinates.

1b)  Source−receiver distance r = x 2 + z2 (1. (1.14e) γi γi = γ12 + γ32 = 1 (1.6 Fundamentals Divergence ∂ux ∂u y ∂uz ∇ ·u = + + (1.14f) 1.14c) γ3 = cos θ3 = cos θz (1. (Fig.11) ∂z ∂ x ∂z ∂ y ∂z2 Laplacian  ∂2 ∂2 ∂2 ∇ 2 u = ∇ · ∇u = ∇ 2 ux ˆi + ∇ 2 u y jˆ + ∇ 2 uz kˆ = + 2+ 2 u (1.12) ∂x 2 ∂y ∂z Wave equation (λ + µ)∇∇ · u + µ∇ 2 u + b = ρ u¨ (1.8) ∂x ∂y ∂z Curl    ∂uz ∂u y ∂ux ∂uz ˆ ∂u y ∂ux ˆ ∇ ×u= − ˆi + − j+ − k (1.15b) Azimuth tan θ = y/x (1. 1.14b) γ2 = 0. (1.3.10) ∂ ux 2 ∂ uy 2 ∂ uz ˆ ∂ ux 2 ∂ uy ∂ uz ˆ 2 2 2 ∇∇ · u = + + i+ + + j ∂ x2 ∂ x ∂ y ∂ x ∂z ∂y ∂x ∂ y2 ∂ y ∂z ∂ 2 ux ∂ 2uy ∂ 2 uz ˆ + + + k (1.14d) Implied summations δii = δ11 + δ33 = 2. (1.15c) .13) b) Two-dimensional space (x–z).2 Cylindrical coordinates  Source–receiver distance R= x 2 + y2 + z2 (1.15a)  Range r= x 2 + y2 (1.9) ∂y ∂z ∂z ∂x ∂x ∂y Curl of curl ∇ × ∇ × u = ∇∇ · u − ∇ · ∇u (1.14a) Direction cosines of r γ1 = cos θ1 = sin θz.

y Source ⊗ x θ r r cos θ r cos θ Direction cosines γ1 = √ = (1.15h) kˆ = kˆ (1.15i) Conversion between cylindrical and Cartesian coordinates u = ux ˆi + u y jˆ + uzkˆ = ur rˆ + uθ tˆ + uzkˆ (1.2: Cylindrical coordinates.3 Coordinate systems and differential operators 7 kˆ tˆ Receiver rˆ R z Figure 1. the dot product.18) uz 0 0 1 uz Nabla operator ∂ 1 ∂ ∂ ∇ = rˆ + tˆ + kˆ (1.15d) r 2 + z2 R r sin θ r sin θ γ2 = √ = (1.17) uz 0 0 1 uz ⎡ ⎤ ⎡ ⎤⎡ ⎤ ux cos θ −sin θ 0 ur ⎣ u y ⎦ = ⎣ sin θ cos θ 0 ⎦ ⎣ uθ ⎦ (1.1.19) ∂r r ∂θ ∂z Allowing the symbol ⊗ to stand for the scalar product.15g) tˆ = −ˆi sin θ + jˆ cos θ (1.15f) r 2 + z2 R Basis vectors rˆ = ˆi cos θ + jˆ sin θ (1. or the cross product.16) ⎡ ⎤ ⎡ ⎤⎡ ⎤ ur cos θ sin θ 0 ux ⎣ uθ ⎦ = ⎣ −sin θ cos θ 0 ⎦ ⎣ uy ⎦ (1. and considering that .15e) r 2 + z2 R z z γ3 = √ = (1.

or dyads.20) ∂θ ∂θ ∂θ we can write a generic nabla operation on a vector u in cylindrical coordinates as    ∂ 1 ∂ ∂ ∇ ⊗ u = rˆ + tˆ + kˆ ⊗ rˆ ur + tˆ uθ + kˆ uz ∂r r ∂θ ∂z ∂ur ∂uθ ∂uz = rˆ ⊗ rˆ + rˆ ⊗ tˆ + rˆ ⊗ kˆ ∂r ∂r ∂r   1 ∂ur ∂uθ ∂uz + − uθ tˆ ⊗ rˆ + + ur tˆ ⊗ tˆ + tˆ ⊗ kˆ r ∂θ ∂θ ∂θ ∂ur ˆ ∂uθ ˆ ∂uz ˆ + k ⊗ rˆ + k ⊗ tˆ + k ⊗ kˆ (1. etc.22) ∂z ∂z ∂z where the products of the form rˆ rˆ . = −ˆr. dot. are tensor bases.26) ∂z r ∂r r ∂θ ∂z . and cross products. =0 (1..21) ∂z ∂z ∂z Specializing this expression to the scalar. we obtain Gradient ∂ur ∂uθ ∂uz ˆ ∇u = rˆ rˆ + rˆ tˆ + rˆ k ∂r ∂r ∂r   1 ∂ur ∂uθ ∂uz ˆ + − uθ tˆ rˆ + + ur tˆ tˆ + tˆ k r ∂θ ∂θ ∂θ ∂ur ˆ ∂uθ ˆ ∂uz ˆ ˆ + k rˆ + k tˆ + kk (1.24) r ∂θ ∂z ∂z ∂r r ∂r ∂θ Curl of curl ∇ × ∇ × u = ∇∇ · u − ∇ · ∇u (1.23) ∂r r ∂θ ∂z r ∂r ∂θ ∂z Curl    1 ∂uz ∂uθ ∂ur ∂uz 1 ∂(r uθ ) ∂ur ˆ ∇ ×u= − rˆ + − tˆ + − k (1. Divergence  ∂ur 1 ∂uθ ∂uz 1 ∂(r ur ) ∂uθ ∂(r uz) ∇ ·u = + + ur + = + + (1.25) ∂ 1 ∂(r ur ) 1 ∂uθ ∂uz 1 ∂ 1 ∂(r ur ) 1 ∂uθ ∂uz ∇∇ · u = rˆ + + + tˆ + + ∂r r ∂r r ∂θ ∂z r ∂θ r ∂r r ∂θ ∂z ∂ 1 ∂(r ur ) 1 ∂uθ ∂uz + kˆ + + (1.8 Fundamentals ∂ rˆ ∂ tˆ ∂ kˆ = tˆ.

30g) . 1.30f) 1 − γ32  γ3 = cos φ ⇒ sin φ = 1 − γ32 (1. un = un (r. but only on r and z. vn .3.30b) Azimuth tan θ = y/x. v ≡ uθ .29c) n=0 sin nθ in which u ≡ ur . 0 ≤ θ ≤ 2π (1. Also. w ≡ uz.3 Coordinate systems and differential operators 9 Laplacian ∂2 1 ∂ 1 ∂2 ∂2 ∇2 = ∇ · ∇ = + + 2 2+ 2 ∂r 2 r ∂r r ∂θ ∂z 2  ∂ ur 1 ∂ur 1 ∂ 2 ur ∂uθ ∂ 2 ur ∇ 2 u = ∇ · ∇u = rˆ + + 2 − ur − 2 + ∂r 2 r ∂r r ∂θ 2 ∂θ ∂z 2 2  ∂ uθ 1 ∂uθ 1 ∂ 2 uθ ∂ur ∂ 2 uθ + tˆ + + 2 − uθ + 2 + ∂r 2 r ∂r r ∂θ 2 ∂θ ∂z 2 2 ∂ u z 1 ∂u z 1 ∂ 2 u z ∂ 2 u z + kˆ + + 2 + (1.1.30d) γ1 Direction cosines γ1 = sin φ cos θ ⇒ cos θ =  (1.30c) Polar angle φ = arccos (z/R).29b) n=0 cos nθ ∞  cos nθ w= wn (1.30e) 1 − γ32 γ2 γ2 = sin φ sin θ ⇒ sin θ =  (1. and either the lower or the upper element in the paren- theses must be used.) Wave equation (λ + µ)∇∇ · u + µ∇ · ∇u + b = ρ u¨ (1.30a)  Range r= x2 + y2 (1. as may be necessary.29a) n=0 sin nθ ∞  −sin nθ v= vn (1. 0≤φ≤π (1.3 Spherical coordinates  Source–receiver distance R= x 2 + y2 + z2 (1. z). that is.28) Expansion of a vector in Fourier series in the azimuth ∞  cos nθ u= un (1. tˆ. which do not depend on θ.27) ∂r 2 r ∂r r ∂θ 2 ∂z2 (Note: ∂ 2 /∂θ 2 in ∇ 2 acts both on the components of u and on the basis vectors rˆ . and so forth. wn are the coefficients of the Fourier series. un .

33) uθ −sin θ cos θ 0 uz ⎡ ⎤ ⎡ ⎤⎡ ⎤ ux sin φ cos θ cos φ cos θ −sin θ uR ⎣ u y ⎦ = ⎣ sin φ sin θ cos φ sin θ cos θ ⎦ ⎣ uφ ⎦ (1.32) ⎡ ⎤ ⎡ ⎤⎡ ⎤ uR sin φ cos θ sin φ sin θ cos φ ux ⎣ uφ ⎦ = ⎣cos φ cos θ cos φ sin θ −sin φ ⎦ ⎣ u y ⎦ (1. = −ˆr.31a) tˆ = −ˆi sin θ + jˆ cos θ Tangential (1.31b) sˆ = ˆi cos φ cos θ + jˆ cos φ sin θ − kˆ sin φ Meridional (1.36b) ∂θ ∂θ ∂θ . = −(sin φ rˆ + cos φ sˆ ) (1.35) ∂R R ∂φ R sin φ ∂θ Allowing the symbol ⊗ to stand for the scalar product.3: Spherical coordinates.31c) Conversion between spherical and Cartesian coordinates u = ux ˆi + u y jˆ + uz kˆ = u R rˆ + uθ tˆ + uφ sˆ (1. the dot product. (1. = 0.36a) ∂φ ∂φ ∂φ ∂ rˆ ∂ sˆ ∂ tˆ = sin φ tˆ. or the cross product.34) uz cos φ −sin φ 0 uθ Nabla operator ∂ 1 ∂ 1 ∂ ∇ = rˆ + sˆ + tˆ (1. y r θ x Basis vectors rˆ = ˆi sin φ cos θ + jˆ sin φ sin θ + kˆ cos φ Radial (1.10 Fundamentals z rˆ tˆ φ R sˆ Figure 1. = cos φ tˆ. and considering that ∂ rˆ ∂ sˆ ∂ tˆ = sˆ .

40) ∂R R R ∂φ .38) ∂θ where the products of the form rˆ rˆ . dot. and cross products.3 Coordinate systems and differential operators 11 we can write a generic nabla operation on a vector u in spherical coordinates as  ∂ 1 ∂ 1 ∂ ∇ ⊗ u = rˆ + sˆ + tˆ ⊗ (ˆr u R + sˆ uφ + tˆ uθ ) ∂R R ∂φ R sin φ ∂θ ∂u R ∂uφ ∂uθ = rˆ ⊗ rˆ + rˆ ⊗ sˆ + rˆ ⊗ tˆ ∂R ∂R ∂R   1 ∂u R ∂uφ ∂uθ + − uφ sˆ ⊗ rˆ + + u R sˆ ⊗ sˆ + sˆ ⊗ tˆ R ∂φ ∂φ ∂φ   1 ∂u R ∂uφ + − uθ sin φ tˆ ⊗ rˆ + − uθ cos φ tˆ ⊗ sˆ R sin φ ∂θ ∂θ  ∂uθ + + u R sin φ + uφ cos φ tˆ ⊗ tˆ (1.. etc. are tensor bases.39) ∂R R R ∂φ R R sin φ ∂θ Curl 1 ∂uθ 1 ∂uφ uθ ∇ ×u = − + cot φ rˆ R ∂φ R sin φ ∂θ R 1 ∂u R ∂uθ uθ + − − sˆ R sin φ ∂θ ∂R R ∂uφ uφ 1 ∂u R + + − tˆ (1.1. Divergence ∂u R 2 1 ∂uφ uφ 1 ∂uθ ∇ ·u= + uR + + cot φ + (1. or dyads.37) ∂θ Specializing this expression to the scalar. we obtain Gradient ∂u R ∂uφ ∂uθ ∇u = rˆ rˆ + rˆ sˆ + rˆ tˆ ∂R ∂R ∂R   1 ∂u R ∂uφ ∂uθ + − uφ sˆ rˆ + + ur sˆ sˆ + sˆ tˆ R ∂φ ∂φ ∂φ   1 ∂u R ∂uφ + − uθ sin φ tˆ rˆ + − uθ cos φ tˆ sˆ R sin φ ∂θ ∂θ  ∂uθ + + u R sin φ + uφ cos φ tˆ tˆ (1.

45) .) Wave equation (λ + µ)∇∇ · u + µ∇ · ∇u = ρ u¨ (1. ∂/∂θ in ∇ 2 act both on the components of u and on the basis vectors.41)  ∂ 2uR 2 ∂u R u R ∇∇ · u = + − ∂R 2 R ∂R R    1 ∂ ∂uφ uφ ∂uφ uφ + − + cot φ − R ∂φ ∂ R R ∂R R  1 ∂ ∂uθ uθ + − rˆ R sin φ ∂θ ∂ R R   1 ∂ ∂u R 2u R 1 ∂uφ 1 ∂uφ uφ + + + + cos φ − R ∂φ ∂ R R R ∂φ R sin φ ∂φ sin φ  1 ∂ ∂uθ + − uθ cot φ sˆ R sin φ ∂θ ∂φ 1 ∂ ∂u R 2u R 1 ∂uφ uφ 1 ∂uθ + + + + cot φ + tˆ (1.42) R sin φ ∂θ ∂ R R R ∂φ R R sin φ ∂θ Laplacian ∂2 2 ∂ 1 ∂2 cot φ ∂ 1 ∂2 ∇2 = + + + + (1.44) sin φ ∂θ (Note: ∂/∂φ.12 Fundamentals Curl of curl ∇ × ∇ × u = ∇∇ · u − ∇ · ∇u (1.43) ∂ R2 R ∂ R R2 ∂φ 2 R2 ∂φ 2 2 ∂θ R sin φ 2  ∂ 2uR 2 ∂u R 1 ∂ 2uR ∂uφ ∇ · ∇u = + + −2 − 2u R ∂ R2 R ∂r R2 ∂φ 2 ∂φ  ∂u R 1 ∂ 2uR 2 ∂uθ + cot φ − 2uφ + − rˆ ∂φ sin2 φ ∂θ 2 sin φ ∂θ 2  ∂ uφ 2 ∂uφ 1 ∂ 2 uφ ∂u R ∂uφ + + + 2 +2 + cot φ ∂ R2 R ∂R R ∂φ 2 ∂φ ∂φ  2 1 ∂ uφ ∂uθ + − 2 cos φ − uφ sˆ sin2 φ ∂θ 2 ∂θ 2   2 ∂ uθ 2 ∂uθ 1 ∂ 2 uθ ∂uθ 1 ∂ uθ + + + 2 + cot φ + − uθ ∂R 2 R ∂R R ∂φ 2 ∂φ sin φ ∂θ 2 2 2 ∂ + (u R + uφ cot φ) tˆ (1.

46) In matrix notation.50) ⎪ ⎪ ∂y ∂z ∂x ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ∂ ∂ ∂ ⎪ ⎩ 0 0 0 ⎪ ⎭ ∂z ∂y ∂x which can be abbreviated as ∂ ∂ ∂ L = Lx + Ly + Lz (1. εz = ∂x ∂y ∂z ∂uz ∂u y ∂ux ∂uz ∂u y ∂ux ε yz = + . and the elastic wave equation 1.4 Strains. ε yx = εxy (1.4 Strains.48) ε = Lu = strain–displacement relation (1.51) ∂x ∂y ∂z . and the elastic wave equation 13 σz σyz σxz dz σ zx Figure 1.4: Stresses (and strains) in Cartesian coordi- σ xy nates.4. this can be written as u = [ux uy u z] T = displacement vector (1. εzx = εxz.1. stresses.49) ⎧ ⎫ ⎪ ∂ ∂ ∂ ⎪ ⎪ ⎪ ⎪ ∂y ⎪ 0 0 0 ⎪ ∂x ⎪ ∂z ⎪ ⎪ ⎪ ⎨ ⎪ ∂ ∂ ∂ ⎬ LT = 0 0 0 (1.47) T ε = [εx εy εz ε yz εxz εxy ] = strain vector (1. stresses. σyx σy σzy σx dx dy 1. εzx = + . εxy = + ∂y ∂z ∂z ∂x ∂x ∂y εzy = ε yz.1 Cartesian coordinates Strains ∂ux ∂u y ∂uz εx = . εy = .

54) For a fully anisotropic medium.58) . Lz = (1. σzx = σxz. this matrix is ⎧ ⎫ ⎪ ⎪ λ + 2µ λ λ 0 0 0⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ λ ⎪ λ + 2µ λ 0 0 0⎪ ⎪ ⎪ ⎪ ⎨ ⎬ λ λ λ + 2µ 0 0 0 D= (1. σ yx = σxy (1.52) ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ 0 0 1⎪⎪ ⎪ ⎪ 0 1 0⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪0 ⎪ 0 1⎪⎪ ⎪ ⎪ ⎪0 ⎪ 0 0⎪⎪ ⎪ ⎪ ⎪1 ⎪ 0 0⎪⎪ ⎪ ⎪ ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 1 0 1 0 0 0 0 0 Stresses σ = [σx σy σz σ yz σxz σxy ]T = stress vector (1. the symmetric constitutive matrix is ⎧ ⎫ ⎪ ⎪ d11 d12 d13 d14 d15 d16 ⎪ ⎪ ⎪ ⎪d ⎪ ⎪ ⎪ ⎪ 12 d22 d23 d24 d25 d26 ⎪ ⎪ ⎪ ⎨ ⎪ ⎬ d13 d23 d33 d34 d35 d36 D= (1. and µ = shear modulus.57) The stresses.53) σ = Dε ε = constitutive law (1. εvol = εx + ε y + εz σ yz = µ ε yz. In this case.14 Fundamentals where by inspection ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎪ ⎪ 1 0 0⎪⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 1 0⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬ 0 0 0 0 0 0 0 0 1 Lx = . Ly = . the stress–strain rela- tionship is σ j = 2µ ε j + λ εvol .55) ⎪ ⎪ d14 d24 d34 d44 d45 d46 ⎪ ⎪ ⎪ ⎪d ⎪ ⎪ ⎪ ⎪ 15 d25 d35 d45 d55 d56 ⎪ ⎪ ⎪ ⎩ ⎪ ⎭ d16 d26 d36 d46 d56 d66 whereas for an isotropic medium. and body loads satisfy the dynamic equilibrium equation b − ρ u¨ + LT σ = 0 (1. inertial loads. σxy = µ εxy σzy = σ yz.56) ⎪ ⎪ 0 0 0 µ 0 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0 0 µ 0 ⎪ ⎪ ⎪ ⎩ ⎪ ⎭ 0 0 0 0 0 µ in which λ = Lame´ constant. σzx = µ εzx .

D yy = 0 λ + 2µ 0 . and the elastic wave equation 15 Stresses in principal surfaces The stresses (or tractions) acting in planes perpendicular to the x.62b) .4 Strains. D yy = d26 d22 d24 . z axes.61a) ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ d16 d12 d14 ⎬ ⎨ d15 d14 d13 ⎬ ⎨ d56 d46 d36 ⎬ Dxy = d66 d26 d46 . y. can be expressed symbolically as ∂u ∂u ∂u sx = [σx σ yx σzx ]T = LTx σ = LTx DLu = Dxx + Dxy + Dx z ∂x ∂y ∂z ∂u ∂u ∂u s y = [σxy σy σzy ]T = LTy σ = LTy DLu = D yx + D yy + Dy z ∂x ∂y ∂z ∂u ∂u ∂u sz = [σxz σ yz σz]T = LzT σ = LzT DLu = Dzx + Dzy + Dz z ∂x ∂y ∂z (1.59) which involve the 3×3 material matrices Di j = LiT DL j .60) These matrices are not symmetric. but D ji = DiTj . which are needed in the formulation for layered media. D yz = Dzy T = 0 0 λ ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 0 µ 0 0 0 µ 0 (1. Dzz = 0 µ 0 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 µ 0 0 µ 0 0 λ + 2µ (1.62a) ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨0 λ 0⎬ ⎨ 0 0 λ⎬ ⎨0 0 0⎬ Dxy = DTyx = µ 0 0 . D yz = d25 d24 d23 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ d56 d25 d45 d55 d45 d35 d45 d44 d34 (1. stresses. Dxz = Dzx T = 0 0 0 . Dxz = d56 d46 d36 . z (1. For a fully anisotropic medium. they are ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ d11 d16 d15 ⎬ ⎨ d66 d26 d46 ⎬ ⎨ d55 d45 d35 ⎬ Dxx = d16 d66 d56 . i.61b) For an isotropic medium. they are ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨λ + 2µ 0 0⎬ ⎨µ 0 0⎬ ⎨µ 0 0 ⎬ Dxx = 0 µ 0 . j = x. y. Dzz = d45 d44 d34 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ d15 d56 d55 d46 d24 d44 d35 d34 d33 (1.1.

the wave equation in Cartesian coordinates is ⎧ ⎫ ⎧ ⎫ ⎨ λ + 2µ 0 0⎬ 2 ∂ u ⎨ µ 0 0⎬ 2 ∂ u ρ u¨ = b + 0 µ 0 + 0 λ + 2µ 0 ⎩ ⎭ ∂ x2 ⎩ ⎭ ∂ y2 0 0 µ 0 0 µ ⎧ ⎫ ⎧ ⎫ ⎨µ 0 0 ⎬ 2 ∂ u ⎨ 0 λ+µ 0⎬ 2 ∂ u + 0 µ 0 + λ+µ 0 0 ⎩ ⎭ ∂z2 ⎩ ⎭ ∂x ∂y 0 0 λ + 2µ 0 0 0 ⎧ ⎫ ⎧ ⎫ ⎨0 0 0 ⎬ 2 ⎨ 0 0 λ + µ⎬ 2 ∂ u ∂ u + 0 0 λ+µ + 0 0 0 (1. b = {bx by bz}T (1.4. y.66) .16 Fundamentals Wave equation Combining the equations of equilibrium and stress–strain relations.63) where b = b(x. Expanding the last term. t) is the body load vector.64) For an isotropic medium. the equations decompose into two uncoupled problems. εθ = + . the SV-P problem (involving ux and uz ) and the SH problem (involving uy ). we obtain ∂ 2u ∂ 2u ∂ 2u ∂ 2u ∂ 2u b + Dxx + D yy 2 + Dzz 2 + (Dxy + D yx ) + (D yz + Dzy ) ∂ x2 ∂y ∂z ∂x ∂y ∂ y ∂z ∂ 2u + (Dxz + Dzx ) = ρ u¨ ∂ x ∂z (1.65) ⎩ ⎭ ∂ y ∂z ⎩ ⎭ ∂ x ∂z 0 λ+µ 0 λ+µ 0 0 The equations for plane strain problems are obtained from the above by discarding all partial derivatives with respect to (say) y.2 Cylindrical coordinates Strains ∂ur ur ∂uθ ∂uz εr = . εz = ∂r r r ∂θ ∂z ∂uθ ∂uz ∂ur ∂uz ∂ur ∂uθ uθ εθ z = + . εr z = + . 1. we obtain immediately b − ρ u¨ + LT DLu = 0. εr θ = + − ∂z r ∂θ ∂z ∂r r ∂θ ∂r r (1. As a result. z.

73) D is the same as for Cartesian coordinates. σr z = µ εr z.72) σ = Dε ε = constitutive law (1. For an anisotropic medium. σθr = σr θ (1.68) ε = Lε u = strain–displacement relation (1. stresses. so this defines a cylindrical type of anisotropy. For isotropic media. θ. we have σ j = 2µ ε j + λ εvol .1. and the elastic wave equation 17 σz σθ z dθ σrz σzr σθ r Figure 1.69) ⎧ ⎫ ⎪ ∂ 1 ∂ 1 ∂ ⎪ ⎪ ⎪ 0 0 ⎪ ⎪ ⎪ ⎪ ∂r r ∂z r ∂θ ⎪⎪ ⎪ ⎨ ⎪ ⎬ 1 ∂ ∂ ∂ 1 LεT = 0 0 0 − (1. we assume here that the elements dij of D are independent of the azimuth θ for any range r.67) T ε = [εr εθ εz εθ z εr z εr θ ] = strain vector (1. although its physical meaning is somewhat different. εvol = εr + εθ + εz σθ z = µ εθ z. σr θ = µ εr θ σzθ = σθ z.70) ⎪ ⎪ r ∂θ ∂z ∂r r⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ∂ 1 ∂ ∂ ⎪ ⎪ ⎩ 0 0 0 ⎭ ∂z r ∂θ ∂r which can be abbreviated as ∂ 1 ∂ ∂ 1 Lε = Lr + Lθ + Lz + L1 (1.4 Strains. dz σr σθ z σθ r dr u = [ur uθ u z] = displacement vector (1. j = r. σzr = σr z.71) ∂r r ∂θ ∂z r Stresses σ = [σr σθ σz σθ z σr z σr θ ]T = stress vector (1.5: Stresses (and strains) in cylindri- σrθ cal coordinates.74) . z.

Also. and Lz ≡ Lz are the same as in Cartesian coordinates. the operator matrices are ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎪ ⎪ 1 0 0⎪ ⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 1 0⎪⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ ⎪ ⎪ 1 0 0⎪⎪ ⎪ ⎪ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬ ⎨ ⎬ 0 0 0 0 0 0 0 0 1 0 0 0 Lr = . observe that Lσ = Lε . Lz = .77) ∂r r ∂θ ∂z r In full.76) ⎪ ⎪ r ∂θ ∂z ∂r r⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ∂ 1 ∂ ∂ 1 ⎪ ⎪ ⎩ 0 0 + 0 ⎭ ∂z r ∂θ ∂r r which can be written compactly as ∂ 1 ∂ ∂ 1 Lσ = Lr + Lθ + Lz + (Lr − L1 ) (1. Lθ ≡ Ly . Stresses in principal surfaces The stresses in radial.79) . bT = [br bθ bz]T = body load vector (1. L1 = ⎪ ⎪ 0 0 0⎪ ⎪ ⎪ ⎪ 0 0 1⎪⎪ ⎪ ⎪ 0 1 0⎪⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪0 0 1⎪ ⎪ ⎪ ⎪ ⎪0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 1 0 0⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0⎪⎪ ⎪ ⎪ ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 1 0 1 0 0 0 0 0 0 −1 0 (1.75) with ⎧ ⎫ ⎪ ∂ 1 1 ∂ 1 ∂ ⎪ ⎪ ⎪ + − 0 0 ⎪ ⎪ ⎪ ⎪ ∂r r r ∂z r ∂θ ⎪⎪ ⎪ ⎨ ⎪ ⎬ 1 ∂ ∂ ∂ 2 LσT = 0 0 0 + (1.78) Notice that Lr ≡ Lx . azimuthal and vertical surfaces are sr = [σr σr θ σr z]T = LrT σ ∂u 1 ∂u ∂u u = Drr + Dr θ + Dr z + Dr 1 ∂r r ∂θ ∂z r sθ = [σr θ σθ σzθ ]T = LθT σ ∂u 1 ∂u ∂u u = Dθr + Dθ θ + Dθ z + Dθ1 ∂r r ∂θ ∂z r sz = [σr z σθ z σz]T = LzT σ ∂u 1 ∂u ∂u u = Dzr + Dzθ + Dzz + Dz1 ∂r r ∂θ ∂z r (1. Lθ = . while L1 is new.18 Fundamentals The dynamic equilibrium equation is b − ρ u¨ + LσT σ = 0.

Dθ1 . are identical to those in Cartesian coor- dinates. Dz1 and their transposes. Dθ z = Dzθ = 0 0 λ T ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 0 µ 0 0 0 µ 0 (1.81b) ⎧ ⎫ ⎧ ⎫ ⎨λ 0 0⎬ ⎨ 0 −µ 0 ⎬ Dr 1 = D1r = 0 −µ 0 . Dzz = 0 µ 0 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 µ 0 0 µ 0 0 λ + 2µ (1. with the replacement x → r . Dr z = Dzr = 0 T 0 0 . Dθ θ = 0 λ + 2µ 0 . these matrices are ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ d11 d16 d15 ⎬ ⎨ d66 d26 d46 ⎬ ⎨ d55 d45 d35 ⎬ Drr = d16 d66 d56 . For a cylindrically isotropic material. Dr 1 = d26 −d66 0 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ d56 d25 d45 d55 d45 d35 d25 −d56 0 (1. the matrices are ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ λ + 2µ 0 0⎬ ⎨µ 0 0⎬ ⎨µ 0 0 ⎬ Drr = 0 µ 0 . but Dβα = DTαβ . Dθ1 = d22 −d26 0 . Dθ θ = d26 d22 d24 . we have the three new matrices Dr 1 . Dr z = d56 d46 d36 . stresses. the matrices Drr etc. Dz1 = d24 −d46 0 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ d45 d44 d34 d24 −d46 0 d23 −d36 0 (1.4 Strains.80d) ⎩ 26 ⎭ 0 0 0 Observe that these matrices are generally not symmetric. For a cylindrically anisotropic medium. In addition. Dzz = d45 d44 d34 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ d15 d56 d55 d46 d24 d44 d35 d34 d33 (1. y → θ. z → z.80b) ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ d56 d46 d36 ⎬ ⎨ d26 −d66 0⎬ ⎨ d25 −d56 0⎬ Dθ z = d25 d24 d23 .80c) ⎧ ⎫ ⎨ d22 −d26 0⎬ D11 = −d d66 0 (1.1. and the elastic wave equation 19 For an isotropic medium.81c) ⎩ ⎭ ⎩ ⎭ 0 0 0 0 0 0 .81a) ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨0 λ 0⎬ ⎨0 0 λ⎬ ⎨ 0 0 0⎬ Dr θ = Dθr T = µ 0 0 . Dθ1 = D1θ = λ + 2µ 0 0 T (1.80a) ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ d16 d12 d14 ⎬ ⎨ d15 d14 d13 ⎬ ⎨ d12 −d16 0⎬ Dr θ = d66 d26 d46 .

81d) ⎩ ⎭ ⎩ ⎭ λ 0 0 0 0 0 Wave equation b − ρ u¨ + LσT DLε u = 0.82) in which b is the vector of body loads.20 Fundamentals ⎧ ⎫ ⎧ ⎫ ⎨0 0 0⎬ ⎨ λ + 2µ 0 0⎬ Dz1 = D1z T = 0 0 0 . this results in the wave equation in cylindrical coordinates: ⎧ ⎫ ⎧ ⎫ ⎨ λ + 2µ 0 0 ⎬ 2 ⎨µ 0 0⎬ ∂ u 1 ∂u 1 ∂ 2u ρ u¨ = b + 0 µ 0 + + 0 λ + 2µ 0 ⎩ ⎭ ∂r 2 r ∂r ⎩ ⎭ r 2 ∂θ 2 0 0 µ 0 0 µ ⎧ ⎫ ⎧ ⎫ ⎨µ 0 0 ⎬ 2 ∂ u ⎨ 0 λ + µ 0⎬ 1 ∂ 2u + 0 µ 0 + λ+µ 0 0 ⎩ ⎭ ∂z2 ⎩ ⎭ r ∂r ∂θ 0 0 λ + 2µ 0 0 0 ⎧ ⎫ ⎧ ⎫ ⎨ 0 0 λ + µ⎬ 2 ⎨0 0 0 ⎬ ∂ u 1 ∂ 2u + 0 0 0 + 0 0 λ+µ ⎩ ⎭ ∂r ∂z ⎩ ⎭ r ∂z∂θ λ+µ 0 0 0 λ+µ 0 ⎧ ⎫ ⎧ ⎫ ⎨ −(λ + 3µ) 0 ⎬ 1 ∂u ⎨ 0 0 0 0⎬ 1 ∂u + λ + 3µ 0 0 + 0 0 0 ⎩ ⎭ r ∂θ ⎩ 2 ⎭ r ∂z 0 0 0 λ+µ 0 0 ⎧ ⎫ ⎨ λ + 2µ 0 0⎬ u − 0 µ 0 (1.84) ⎩ ⎭ r2 0 0 0 . Expansion of the second term yields ∂ 2u 1 ∂ 2u ∂ 2u 1 ∂ 2u b − ρ u¨ + Drr + Dθ θ 2 2 + Dzz 2 + (Dr θ + Dθr ) ∂r 2 r ∂θ ∂z r ∂r ∂θ ∂ 2u 1 ∂ 2u 1 ∂u 1 ∂u + (Dr z + Dzr ) + (Dθ z + Dzθ ) + Drr + (Dθ1 − D1θ ) 2 ∂r ∂z r ∂z∂θ r ∂r r ∂θ 1 ∂u u + (Dr z + Dz1 − D1z) − D11 2 = 0 r ∂z r (1. bT = [br bθ bz]T (1.83) For an isotropic medium. D11 = 0 µ 0 (1.

90) ε = Lε u Strain–displacement relation (1.89) T σ = [σ R σφ σθ σφθ σ Rθ σ Rφ ] Stress vector (1.87) b = [bR bφ bθ ]T Body load vector (1. stresses. εvol = ε R + εφ + εθ σφθ = µ εφθ .85) σ j = 2µ ε j + λ εvol .92) LTσ σ + b = ρ u¨ Dynamic equilibrium equation (1. and the elastic wave equation 21 1. sin φ R ∂θ ∂R R ∂u R ∂uφ uφ ε Rφ = + − R ∂φ ∂R R (1. φ. σ Rφ = µ ε Rφ σθ φ = σφθ .88) ε = [ε R εφ εθ εφθ ε Rθ ε Rφ ]T Strain vector (1.91) σ = Dε Constitutive law (1. j = R. θ. R R sin φ R ∂θ 1 ∂uφ ∂uθ uθ εφθ = + − cot φ sin φ R ∂θ R ∂φ R 1 ∂u R ∂uθ uθ ε Rθ = + − . σφ R = σ Rφ (1. ∂R uR ∂uφ εφ = + R R ∂φ uR uφ 1 ∂uθ εθ = + cot φ + .3 Spherical coordinates Strains and stresses ∂u R εR = .86) u = [u R uφ uθ ]T Displacement vector (1.4 Strains.93) . σ Rθ = µ ε Rθ .1.4. σθ R = σ Rθ .

97) ∂R R∂φ R sin φ ∂ θ R R with matrices ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎪ ⎪1 0 0⎪ ⎪ ⎪ ⎪0 0 0⎪ ⎪ ⎪ ⎪0 0 0⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 1 0⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0⎪ ⎪ ⎪ ⎪ ⎪ ⎨0 ⎪ ⎪ ⎪ ⎪ ⎪ 0 0⎬ ⎨0 0 0⎬ ⎨0 0 1⎬ LR = .6: Stresses (and strains) in spherical σθφ σφθ coordinates.95) The two differential operators Lε and Lσ can be written as ∂ 1 ∂ 1 ∂ 1 cot φ Lε = L R + Lφ + Lθ + L1 + L2 (1.94) ⎪ ⎪ R∂φ R R sin φ ∂ θ ∂R R⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ∂ 1 ∂ cot φ ∂ ⎪ ⎪ ⎪ ⎩ 0 1 1 ⎪ ⎭ 0 − − 0 R sin φ ∂ θ R∂φ R ∂R R ⎧ ⎫ ⎪ ∂ 2 1 1 1 ∂ 1 ∂ cot φ ⎪ ⎪ ⎪ + − − + ⎪ R ⎪ 0 ⎪∂ R R ⎪ R R R sin φ ∂ θ R∂φ ⎪ ⎪ ⎪ ⎨ ⎪ ⎬ 1 ∂ cot φ cot φ 1 ∂ ∂ 3 LσT = 0 + − 0 + ⎪ ⎪ R∂φ R R R sin φ ∂ θ ∂R R ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 1 ∂ 1 ∂ 2 cot φ ∂ 3 ⎪ ⎪ ⎩ 0 0 + + 0 ⎭ R sin φ ∂ θ R∂φ R ∂R R (1.22 Fundamentals σφ σRθ σR σaφ Figure 1.98a) ⎪0 ⎪ 0 0⎪⎪ ⎪0 ⎪ 0 1⎪⎪ ⎪0 ⎪ 1 0⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪0 1⎪⎪ ⎪ ⎪0 0⎪⎪ ⎪ ⎪1 0⎪⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 1 0 1 0 0 0 0 0 . Lθ = (1. Lφ = . dθ σRθ σφR dφ σθ R dR where ⎧ ∂ 1 1 1 ∂ 1 ∂ ⎫ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ ∂R R R R sin φ ∂ θ R∂φ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ⎪ ⎬ 1 ∂ cot φ 1 ∂ ∂ 1 Lε = T 0 0 − (1.96) ∂R R∂φ R sin φ ∂ θ R R ∂ 1 ∂ 1 ∂ 1 cot φ Lσ = L R + Lφ + Lθ + (2L R − L1 ) + (Lφ − L2 ) (1.

98b) ⎪ ⎪ 0 0 0 ⎪ ⎪ ⎪ ⎪ 0 0 −1 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪0 ⎪ −1 ⎪ ⎪ ⎪0 0 ⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎩ ⎭ ⎩ ⎭ 0 −1 0 0 0 0 Stresses in principal surfaces s R = [σ R σRφ σ R θ ]T = LTR σ = LTR D Lε u ∂ 1 ∂ 1 ∂ = D RR + DRφ + DRθ ∂R R∂ φ R sin φ ∂ θ 1 cot φ + DR1 + DR2 u R R sφ = [σφ R σφ σφθ ]T = LφT σ = LφT D Lε u ∂ 1 ∂ 1 ∂ = Dφ R + Dφφ + Dφ θ ∂R R∂ φ R sin φ ∂ θ 1 cot φ + Dφ1 + Dφ2 u R R sθ = [σθ R σθ φ σθ ]T = LθT σ = LθT D Lε u ∂ 1 ∂ 1 ∂ = Dθ R + Dθ φ + Dθ θ ∂R R∂ φ R sin φ ∂ θ 1 cot φ + Dθ1 + Dθ2 u R R (1. φ.100b) ⎩ ⎭ ⎩ ⎭ d35 d34 d33 d56 d25 d45 . Observe that these matrices are generally not symmetric.β = R. with α. 2. stresses. θ.1.4 Strains. and the elastic wave equation 23 ⎧ ⎫ ⎧ ⎫ ⎪ ⎪ 0 0 0 ⎪ ⎪ ⎪ ⎪ 0 0 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 1 0 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 0 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎨1 ⎪ ⎪ ⎪ 0 0 ⎬ ⎨0 1 0 ⎬ L1 = . For a spherically anisotropic material. but Dβα = DTαβ . Dφ φ = d26 d22 d24 (1. DRφ = d66 d26 d46 (1. L2 = (1.100a) ⎩ ⎭ ⎩ ⎭ d15 d56 d55 d46 d24 d44 ⎧ ⎫ ⎧ ⎫ ⎨ d55 d45 d35 ⎬ ⎨ d16 d12 d14 ⎬ Dθ θ = d45 d44 d34 . the matrices are ⎧ ⎫ ⎧ ⎫ ⎨ d11 d16 d15 ⎬ ⎨ d66 d26 d46 ⎬ D RR = d16 d66 d56 .99) which uses the definition Dαβ = LTα DLβ. 1.

Dφθ = 0 0 λ . Dφ θ = d25 d24 d23 (1.100f) ⎩ ⎭ ⎩ 22 ⎭ 0 d35 −d45 d24 + d34 −d46 −d45 ⎧ ⎫ ⎧ ⎫ ⎨ 0 d36 −d46 ⎬ ⎨ d25 + d35 −d56 −d55 ⎬ Dφ2 = 0 d23 −d24 .100e) ⎩ ⎭ ⎩ ⎭ 0 −d35 d45 d25 + d35 −d56 −d55 ⎧ ⎫ ⎧ ⎫ ⎨ 0 d13 −d14 ⎬ ⎨ d26 + d36 −d66 −d56 ⎬ Dr 2 = 0 d36 −d46 . D22 = 0 d33 −d34 ⎩ ⎭ ⎩ ⎭ −(d25 + d35 ) d56 d55 0 −d34 d44 (1.100g) ⎩ ⎭ ⎩ ⎭ 0 d34 −d44 d23 + d33 −d36 −d35 ⎧ ⎫ ⎨ 0 d35 −d45 ⎬ Dθ2 = 0 d34 −d44 (1.24 Fundamentals ⎧ ⎫ ⎧ ⎫ ⎨ d15 d14 d13 ⎬ ⎨ d56 d46 d36 ⎬ DRθ = d56 d46 d36 .100d) ⎧ ⎫ ⎧ ⎫ ⎨ 0 d23 + d33 −(d24 + d34 ) ⎬ ⎨ d12 + d13 −d16 −d15 ⎬ D12 = 0 −d36 d46 .101c) .101b) ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 0 0 µ 0 µ 0 0 ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ 2λ 0 0 ⎬ ⎨ 0 −µ 0 ⎬ ⎨ 0 0 −µ ⎬ D R1 = 0 −µ 0 .100c) ⎩ ⎭ ⎩ ⎭ d55 d45 d35 d45 d44 d34 ⎧ ⎫ ⎧ ⎫ ⎨d22 + 2d23 + d33 −(d26 + d36 ) −(d25 + d35 )⎬ ⎨0 0 0 ⎬ D11 = −(d26 + d36 ) d66 d56 . Dθ θ = 0 µ 0 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 µ 0 0 µ 0 0 λ + 2µ (1. D Rθ = 0 0 0 (1. Dθ1 = 0 0 0 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 −µ 0 0 0 2(λ + µ) 0 0 (1. Dφ φ = 0 λ + 2µ 0 . Dφ1 = 2(λ + µ) 0 0 . Dr 1 = d26 + d36 −d66 −d56 (1. the material matrices are ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨λ + 2µ 0 0⎬ ⎨µ 0 0⎬ ⎨µ 0 0 ⎬ D RR = 0 µ 0 .100h) ⎩ ⎭ 0 d33 −d34 For an isotropic medium. Dφ1 = d + d23 −d26 −d25 (1. Dθ1 = d24 + d34 −d46 −d45 (1.101a) ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨0 λ 0⎬ ⎨0 0 0⎬ ⎨0 0 λ⎬ D Rφ = µ 0 0 .

the wave equation is ∂2 1 ∂2 1 ∂2 1 ∂2 ρ u¨ = b + D RR + Dφφ 2 + Dθ θ + (D Rφ + DφR) ∂ R2 R ∂ φ2 R2 sin2 φ ∂ θ 2 R ∂ R∂ φ 1 ∂2 1 ∂2 + (D Rθ + DθR) + (Dφ θ + Dθ φ ) 2 R sin φ ∂ R ∂ θ R sin φ ∂ φ ∂ θ 1 ∂ cot φ ∂ + (2D RR + D R1 − D1R) + (Dφ R + D R2 − D2R) R∂ R R ∂R 1 ∂ cot φ ∂ + (D Rφ + Dφ1 − D1φ ) + (Dφφ + Dφ2 − D2φ ) R2 ∂ φ R2 ∂ φ 1 ∂ cot φ ∂ + (D Rθ + Dθ1 − D1θ ) + (Dθ2 − D2θ ) 2 R2 sin φ ∂ θ R sin φ ∂ θ 1 cot φ + (D R1 − D11 ) + (D R2 + Dφ1 − D12 − D21 ) R2 R2 1 cot2 φ − Dφ2 + (Dφ2 − D22 ) R2 sin φ 2 R2 (1. D12 = 0 0 0 . stresses. Dθ2 = 0 0 −µ (1. and the elastic wave equation 25 ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨0 λ 0⎬ ⎨0 0 0 ⎬ ⎨0 0 0 ⎬ D R2 = 0 0 0 .1.103) .101d) ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 0 0 0 −µ 0 λ + 2µ 0 ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ 4(λ + µ) 0 0 ⎬ ⎨ 0 2(λ + µ) 0 ⎬ ⎨0 0 0⎬ D11 = 0 µ 0 .102) For any arbitrary. spherically anisotropic material with Dαβ = LαT DLβ .101e) Wave equation The dynamic equilibrium equation can be written as LσT D Lε u + b = ρ u¨ (1. D22 = 0 λ + 2µ 0 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ 0 0 µ 0 0 0 0 0 µ (1.4 Strains. Dφ2 = 0 λ 0 .

26 Fundamentals For an isotropic material.104) R2 ⎭ µ R sin φ 2 2 0 0 0 0 −2µ . this results in the elastic wave equation in spherical coordinates: ⎧⎡ ⎤ ⎨ λ + 2µ 0 0  ∂ 2 2 ∂ 2 ρ u¨ = b + ⎣ 0 µ 0⎦ + − ⎩ ∂ R2 R ∂ R R2 0 0 µ ⎡ ⎤ µ 0 0  1 ∂2 cot φ ∂ + ⎣ 0 λ + 2µ 0⎦ 2 2 + R ∂φ R ∂φ 2 0 0 µ ⎡ ⎤ ⎡ ⎤ µ 0 0 0 λ+µ 0 1 ∂2 ⎦ 1 ∂ 2 + ⎣0 µ 0 ⎦ + ⎣λ + µ 0 0 ∂θ ∂ ∂φ λ + 2µ R sin φ 2 2 2 R R 0 0 0 0 0 ⎡ ⎤ ⎡ ⎤ 0 0 λ+µ 0 0 0 ⎣ ⎦ 1 ∂2 1 ∂2 + 0 0 0 + ⎣0 0 λ + µ⎦ 2 R sin φ ∂ R ∂θ R sin φ ∂φ ∂θ λ+µ 0 0 0 λ+µ 0 ⎡ ⎤ ⎡ ⎤ 0 λ+µ 0 0 −(λ + 3µ) 0 cot φ ∂ 1 ∂ + ⎣0 0 0⎦ + ⎣2(λ + 2µ) 0 0⎦ 2 R ∂R R ∂φ 0 0 0 0 0 0 ⎡ ⎤ 0 0 −(λ + 3µ) 1 ∂ +⎣ 0 0 · ⎦ R 2 sin φ ∂θ 2(λ + 2µ) 0 0 ⎡ ⎤ ⎡ ⎤ 0 0 0 0 −(λ + 3µ) 0 cot φ ∂ cot φ + ⎣0 0 −(λ + 3µ) ⎦ + ⎣0 0 0⎦ R sin φ ∂θ 2 R2 0 λ + 3µ 0 0 0 0 ⎡ ⎤ ⎡ ⎤ ⎫ cot2 φ ⎬ 0 0 0 0 0 0 1 + ⎣0 −λ 0⎦ + ⎣0 −2µ 0 ⎦ u (1.

a) = 2Pa ∇  g j · aˆ + O(a 2 ) (2.2 Dipoles In most cases.1 Point dipoles or doublets: single couples and tensile crack sources Let g j (r. r + a) − g j (r. ω) be the Green’s function for a receiver placed at location r due to an impulsive or harmonic source. Additional solutions for single couples. t) or g j (r. r . r ) = g j (r.3) in which the primes indicates that the gradient must be taken with respect to the source point. we obtain g j (r. and bimoments (moments of moments) can be obtained by differentiation. x2 . r . r ) ± eˆ 1 + eˆ 2 + eˆ 3 · [x1 eˆ 1 + x2 eˆ 2 + x3 eˆ 3 ] + · · · ∂ x1 ∂ x2 ∂ x3 = g j (r. x3 ) ± x1 + x2 + x3 + ··· ∂ x1 ∂ x2 ∂ x3 ∂ gj ∂ gj ∂ gj = g j (r. not the receiver.1) Writing a = a aˆ = x1 eˆ 1 + x2 eˆ 2 + x3 eˆ 3 and expanding in Taylor series. r − a)] (2. r ) ± (∇  g j ) · a + O(a 2 ) (2. Together. apply two such loads in opposite directions at some small distance 2a apart in the neighborhood of r . 2. (moments or torques). r ± a) = g j (r. double couples. Next. Define a = a aˆ to be a vector connecting the midpoint between the points of application of these forces to one of these. and let aˆ be a unit vector in that direction. x2 ± x2 . r ) = g j (r. and j identifies the direction of the force. we provide solutions for point forces only. x1 ± x1 .1.2) so u(r. Applying the previous 27 . r . This is done as described in the following. x3 ± x3 ) ∂ gj ∂ gj ∂ gj = g j (r. x1 . a point load P = 1 acting at location r in some principal direction j. as shown in Figure 2. tensile crack sources. this pair of forces causes a displacement field u = P [g j (r.

the depth of the source – but not the lateral position – must be considered.e. the Green’s functions depend only on the relative location of the receiver with respect to the source.. It should be noted that the displacement field elicited by single couples is not rotationally symmetric with respect to the axis of the couple.1: From point forces to point dipoles. we obtain the Green’s functions for the nine possible simple force dipoles as ∂ gj ∂ gj ∂ gj ∂ gj G jk = ∇  g j · eˆ k =  e1 + ˆ  e2 + ˆ  e3 · ek = ˆ ˆ ∂ x1 ∂ x2 ∂ x3 ∂ xk ∂ = [g1 j eˆ 1 + g2 j eˆ 2 + g3 j eˆ 3 ] ≡ gi j.keˆ i eˆ k) · (αmeˆ m) = M αk gi j. the Green’s functions for point dipoles are obtained from the gradient of the Green’s functions for point forces with respect to the location of the source (a tensor with nine components). The three dipoles with equal indices G( j j) (or Gxx . we obtain in the limit a → 0 the Green’s function for simple point dipoles. r ) = g j (r − r ). the directions of the dipoles are then defined by the right-hand rule. chosen so that the forces in the first octant are always pointing in the positive direction. the other six are single couples.e. Dipoles in a homogeneous space In the case of a fully homogeneous and unbounded space. or doublets.5) ∂ xk Observe that the dipoles constitute a third order tensor Gi jk.2 shows the nine simple dipoles. k indicate the directions of the forces and the moment arm.keˆ i (2. which switches the locations of the source and the receiver. The indices j.6) By contrast. not even in a fully homogeneous space.28 Dipoles y P Dipole â P â r − r′ Figure 2.) are tensile cracks.keˆ i = Gi jkeˆ i (2. in which case we could write G jk = Gi jkeˆ i = −∇g j (r) · eˆ k = −gi j. projected along the direction of the moment arm.keˆ i eˆ k (2.4) that is. in a body with physical boundaries or material discontinuities. For example. Figure 2. that is. Equivalent forms can then be obtained by recourse to the reciprocity principle. . the location of the source is relevant.. Receiver r′ r x equation to the two neighboring sources. Choosing the direction of the moment arm to coincide in turn with the three principal directions (i. In this situation. in the case of a plate of infinite lateral extent. M = 1). the gradient with respect to the source is simply the negative of the gradient with respect to the receiver. etc. g j (r. aˆ = eˆ k) and setting the strength of the dipoles to unity (i. and defining the strength M of the dipole to be M = 2Pa. as G = M ∇  g j · aˆ = M (gi j.

2 Line dipoles 29 z z z y y y x x x Gxx Gxy Gxz z z z y y y x x x Gyx Gyy Gyz z z z y y y x x x Gzx Gzy Gzz Figure 2. z z z y y Tz y Ty x x x Tx Figure 2.2: The nine simple dipoles. and using the Green’s functions for line sources in place of those for point forces. .2 Line dipoles These are dipoles formed by line sources.2.3: Pure torsional moments. 2. They are obtained from the preceding equations by setting to zero the partial derivatives with respect to the line direction.

2. M yy .8a) k=1  3   M y = My Mzx = My (Gxz + Gzx ) = My (gkx. These planes have normals in directions x.x − gkx. M yz ≡ Mzy = G yz + Gzy (2.y )eˆ k (2. zx. Observe that the diagonal terms of the moment tensor.y + gky.7a) k=1  3   T y = 12 Ty (Gxz − Gzx ) = 12 Ty (gkx.x + gkx. it is necessary to add appro- priately two of the Green’s functions for simple moments listed previously while taking into account the right-hand rule. xy.4 Seismic moments (double couples with no net resultant) To obtain the Green’s functions for seismic moments.z − gkz.y )eˆ k (2.x )eˆ k (2. etc.9b) Mxy ≡ M yx = Gxy + G yx (2. z.z )eˆ k (2. these torsional sources may (or may not) produce rotationally symmetric displacement fields. We thus obtain the Green’s functions for torsional point moments as  3   Tx = 12 Tx (Gzy − G yz) = 12 Tx (gkz. .8c) k=1 that is.30 Dipoles 2.y − gky.7c) k=1 in which Tx . When this is done. Mzz.e.z + gkz. which are double couples with no net resulting force or moment.7b) k=1  3   Tz = 12 Tz(G yx − Gxy ) = 12 Tz (gky.x )eˆ k (2. Mxx .9a) Mzx ≡ Mxz = Gzx + Gxz (2. i.. respectively.3 Torsional point sources To obtain the Green’s functions for conventional torques.z )eˆ k (2.. are twice as large as the simple dipoles for tension cracks.8b) k=1  3   Mz = MzMxy = Mz(G yx + Gxy ) = Mz (gky. y. give the strength of the torsional moments. Depending on the geometry and axis of application.9c) are the response functions for double couples of unit strength lying in planes yz. we again combine appropriately two of the simple moments previously listed. the Green’s functions for double couples are found to be  3   Mx = Mx M yz = Mx (Gzy + G yz) = Mx (gkz.

5 Blast loads (explosive line and point sources) Linelike and pointlike blast loads represent the limiting cases of cylindrical and spherical cavities of infinitesimal radius subjected to singularly large impulsive or harmonic com- pressive pressures. this gives an overpressure  δ(r ) p= → pr dr dθ = 1 2-D (2.z eˆ x + gzx.x + gzz.13) 4 β 1 Kausel.x + gky.x + gxz. 1998. E.). Blast loads versus point loads: the missing factor. 2 (Feb. Journal of Engineering Mechanics.11) 4π r 2 V The Green’s functions for line and point blast loads can be obtained from the Green’s functions for line and point forces by superposition of the three tensile cracks.z eˆ k Point blast source in 3-D 4µ k=1  3 α 2 = [Gxx + G yy + Gzz] (2. Vol. No.5 Blast loads (explosive line and point sources) 31 z z z y y y x x x Mx My Mz Figure 2. The strength of the source is defined so that when the overpressure is integrated over the volume of the cavity.2. except that they must be modified by a factor to compensate for the extraction of the infinitesimal particle that fills the cavity1 : λ + 2µ ∂ gx ∂ gz B2D =  +  µ ∂x ∂z λ + 2µ          = gxx.10) 2π r A  δ(r ) p= → pr2 dr dθ dφ = 1 3-D (2. In two and three dimensions.y + gkz.z eˆ z Line blast source in 2-D µ  2 α = [Gxx + Gzz] (2. 2. pp. 243–244.4: Double couples (seismic moments with no net resultant). one obtains a unit value. . ASCE. 124.12) β 3(λ + 2µ) ∂ gx ∂ gy ∂ gz B3D =  +  +  4µ ∂x ∂y ∂z 3(λ + 2µ) 3      = gkx..

Mxz. z. Now.. r  = 0) at depth z are expressed in cylindrical coordinates. Thus. here we have assumed the medium to be laterally homogeneous. the gradient with respect to the receiver is the negative of the gradient with respect to the source. aˆ = kˆ (2. the derivatives with respect to the source can be changed into derivatives with respect to the receiver. z. When we substitute the Green’s functions for horizontal and vertical loads given previously into the expression for the gradient together with the sign change described. z.18) in which the products rˆ rˆ . aˆ = jˆ = rˆ sin θ + tˆ cos θ. etc.. the Green’s functions for point loads in the vertical z direction (n = 0) and in the horizontal x and y directions (both n = 1) are of the form gz = U(r. z. and if we also express the appropriate moment arm direction aˆ in cylindrical coordinates as one of aˆ = ˆi = rˆ cos θ − tˆ sin θ. z )kˆ Vertical z load (2. are tensor bases. they have the general form  ∞    c −sn ˆ c g= un n rˆ + vn t + wn n kˆ (2. It follows that in the computation of the Green’s functions for dipoles. z ) cos θ rˆ + v(r. 2. Myy .16) g y = u(r. Mzz = 1) and for single couples (Mxy . viz.32 Dipoles Observe that in the case of a homogeneous full space. rˆ tˆ. = 1) listed below. z ) sin θ kˆ Horizontal y load (2. ∂g/∂ xi = −∂g/∂ xi . z.. z.17) The gradient ∇g for any of these Green’s functions with respect to the receiver location can be shown to be given by       ∂ u cn ∂ v −sn ∂ w cn ˆ ∇ g = rˆ ˆ r + ˆ+ t k ∂ r sn ∂r cn ∂ r sn    nu − v −sn u − nv cn nw −sn ˆ + tˆ rˆ + tˆ + k r cn r sn r cn    ∂ u cn ∂ v −sn ∂ w cn ˆ +kˆ rˆ + tˆ + k ∂ z sn ∂ z cn ∂ z sn (2. but that can be inhomogeneous and/or bounded in the vertical direction. z. or dyads. z ) sin θ rˆ + v(r. z ) (−sin θ) tˆ + w(r. etc. In particular.6 Dipoles in cylindrical coordinates Consider a point load acting in a cylindrically homogeneous and unbounded medium.19) we obtain after some algebra the Green’s functions for unit crack sources (Mxx .14) n=0 sn cn sn in which cn = cos nθ and sn = sin nθ. except for the terms involving the vertical derivatives. z. z ) cos θ tˆ + w(r. which also use . so the dependence of the Green’s functions within a given plane is only a function of the relative horizontal position of the receiver and the source.15) gx = u(r. When the Green’s functions for a point load acting on the cylindrical axis (i.e. z ) cos θ kˆ Horizontal x load (2. it suffices for us to reverse the signs of the terms other than those for the vertical derivatives. z ) rˆ + W(r.

22d) ∂r r   ∂u ∂v ∂w Gxz = +  cos θ rˆ −  sin θ tˆ +  cos θ kˆ (2.22e) ∂z ∂z ∂z   ∂U U ∂W Gzx = − cos θ rˆ − sin θ tˆ + cos θ kˆ (2. W = components of Green’s function for a vertical point load (n = 0) (2.22h) ∂r r ∂r Observe that the terms with derivatives with respect to the vertical source location z have positive sign.6 Dipoles in cylindrical coordinates 33 the following definitions: u.20) U.21)   1 ∂u u − v ∂w w ˆ Gxx = − + rˆ + + k 2 ∂r r ∂r r     ∂u u − v ∂v u−v ∂w w + − cos 2θ rˆ − + sin 2θ tˆ + − cos 2θ kˆ ∂r r ∂r r ∂r r (2.22f) ∂r r ∂r   ∂u ∂v ∂w G yz = +  sin θ rˆ +  cos θ tˆ +  sin θ kˆ (2. w = components of Green’s function for a horizontal point load (n = 1). (2.22a)   1 ∂u u − v ∂w w ˆ G yy = − + rˆ + + k 2 ∂r r ∂r r     ∂u u − v ∂v u−v ∂w w − − cos 2θ rˆ + + sin 2θ tˆ − − cos 2θ kˆ ∂r r ∂r r ∂r r (2. seismic double . v.22g) ∂z ∂z ∂z   ∂U U ∂W Gzy = − sin θ rˆ + cos θ tˆ + sin θ kˆ (2.22b)   ∂U ∂W ˆ Gzz = + rˆ +  k ∂z ∂z     1 ∂v u−v ∂u u − v ∂v u−v Gxy = − − − tˆ + − rˆ sin 2θ + + cos 2θ tˆ 2 ∂r r ∂r r ∂r r   ∂w w + − sin 2θ kˆ (2. the Green’s functions for point torsional moments. Also.22c) ∂r r    1 ∂v u−v ∂u u − v ∂v u−v G yx = − − tˆ + − rˆ sin 2θ + + cos 2θ tˆ 2 ∂r r ∂r r ∂r r   ∂w w + − sin 2θ kˆ (2.2.

23f)    3(λ + 2µ) ∂u u − v ∂U ∂w w ∂ W ˆ B3D = − + −  rˆ + + −  k (2. v.23g) 4µ ∂r r ∂z ∂r r ∂z in which u. respectively. When considered in combination. provides a third independent solu- tion for a point source on the axis with azimuthal variation n = 1. Here. The reason for including it here is that this dipole of dipoles. that is. and blast loads in cylindrical coordinates are     1 ∂u ∂U ∂v U ∂w ∂ W Tx = −  + sin θ rˆ +  + cos θ tˆ +  + sin θ kˆ (2.23c) 2 ∂r r     ∂u ∂U ∂v U ∂w ∂ W Mx = +  − sin θ rˆ +  − cos θ tˆ +  − sin θ kˆ (2. together with gx (response to horizontal point load) and Gxz (dipole about y).23) r ∂r r ∂r ∂r r . such as in the formulation of transmitting boundaries in cylindrical coordinates via boundary elements. Finally.23b) 2 ∂z ∂r ∂z r ∂z ∂r  1 ∂v u−v Tz = − − tˆ (2. of the form [cos θ − sin θ cos θ].23e) ∂z ∂r ∂z r ∂z ∂r     ∂u u − v ∂v u−v ∂w w Mz = − − sin 2θ rˆ + + cos 2θ tˆ + − sin 2θ kˆ ∂r r ∂r r ∂r r (2. w and U. The Green’s function for the torsional bimoment is     1 ∂v u−v ∂ ∂v u−v u=− − cos θ rˆ − − sin θ tˆ (2. W are again the components of the Green’s function in cylindrical coordinates for horizontal and vertical point loads. these three solutions allow inter- esting applications. but replacing the two forces by torsional moments. we take two equal and opposite torsional moments Tz in close proximity with moment arm along the y direction. we provide also the expression for one of the several bimoments (moments of moments) that can be constructed.23a) 2 ∂z ∂r ∂z r ∂z ∂r     1 ∂u ∂U ∂v U ∂w ∂ W Ty = +  + cos θ rˆ −  + sin θ tˆ +  + cos θ kˆ (2.23d) ∂z ∂r ∂z r ∂z ∂r     ∂u ∂U ∂v U ∂w ∂ W My = +  − cos θ rˆ −  − sin θ tˆ +  − cos θ kˆ (2.2.34 Dipoles couples. similar in con- cept to the eighth dipole in Figure 2.

γi. pp.1 Fundamental identities and definitions r r  tP = . 5. 5. see Fig. 285. r = x 2 + z2 (3.7) ∂r 2π µr t 2 − t 2 3/2 S 1 Graff.2.1) α β ωr ωr β 1 − 2ν P = . = γk (3. ω) = − H (S ) (3. The receiver is at x. SECTION II: FULL SPACE PROBLEMS 3 Two-dimensional problems in full. a= = (3.. homogeneous space.5) ∂r 4µr 1 Time domain. 5.4) 4µ 0 ∂g yy i S (2) = H (S ) (3. However. and eq. tS = .2 Anti-plane line load (SH waves)1 A unit impulsive SH line source acts at x = 0.2. 1975.6) 2π µ t 2 − t 2 S ∂u yy tS2 H(t − tS ) =   (3.21 should be conjugated to conform to our sign convention.k = = (δik − γi γk) . 288. 35 . impulse response 1 H(t − tS ) u yy (r. The response is given in Cartesian coordinates. both formulas are missing a divisor 2πµ. eqs. homogeneous spaces 3. Wave motion in elastic solids.2) α β α 2(1 − ν) xi ∂ cos θi 1 ∂ f (r ) ∂f γi = cos θi = . Ohio State University Press.3) r ∂ xk r ∂ xk ∂r 3. z = 0 in direction y within an infinite.40.21. S = . K. 3. t) =  (3. Frequency domain i (2) g yy (r.2.2. F. z.

8) ∂t 2 ∂ x2 ∂z ∂u y ∂u y τxy = µx .36 Two-dimensional problems in full.3 SH line load in an orthotropic space Consider an orthotropic elastic space with shear moduli µx . λz.1: Plane strain loads. imaginary part. . t) = Pδ(x)δ(z)δ(t) (3. homogeneous spaces z z y y x x a) Anti-plane (SH) line load b) In-plane (SV-P) line loads Figure 3. τ yz = µz (3.2a: Green’s function for anti-plane line load in full space. Solid line. µz in the plane of wave prop- agation. 3. The wave equation (including the source term) and the shearing stresses are ∂ 2uy ∂ 2uy ∂ 2uy ρ − µx − µz 2 = b(x. dashed line.2 0. by means of which the wave equation and stresses can be reduced to the equivalent 0.1 -0. z.9) ∂x ∂z Using stretching factors λx .2 0 2 4 6 8 10 ωr β Figure 3. real part.1 µ Ω S g yy 0 -0. define the scaled coordinates x˜ = λx x and z˜ = λz z.

ρ˜ = .2b: Impulse response function for anti-plane line load in full space. To obtain the stresses. z˜ = z .4 0.12) λ x λz λ x λz √ Choosing λx = 1.  ˜S = . µ˜ = µx µz.1 0 0 2 4 6 8 10 tβ r Figure 3. ˜ z)δ(t) ˜ (3.3. (3. t). β˜ = = = βx . isotropic form 2 ∂ 2uy ∂ uy ∂ 2uy ρ˜ 2 − µ˜ + = b(x.13) µz µx !    µz b˜ = b = Pδ(x)δ(˜z)δ(t) observe that δ(˜z) d z˜ = δ(z) dz = 1 (3. . this implies λz = µx /µz.5 0.17b) ∂z ∂z ∂ r˜ r˜ ∂ r˜ where.  ˜ S ) or u y = u yy (˜r . ˜ t) = Pδ(x)δ( ˜ z. b˜ = (3.2 0. τ˜ yz = µ˜ = τ yz (3. (3. t˜S = (3.15) µ˜ µx µx µ ! ! µx 2 µ˜ µx ωr˜ r˜ r˜ = x 2 + z. of course. observe that ∂u y ∂ r˜ ∂u y x ∂u y τxy = µx = µx = µx . ρ˜ = ρ (3.17a) ∂x ∂ x ∂ r˜ r˜ ∂ r˜ ∂u y ∂ r˜ ∂u y z ∂u y τ yz = µz = µz = µx .3 ρβ ru yy 0. The factor µx in τ yz is not a misprint. in which case the equivalent parameters are ! ! µx √ µz x˜ = x.14) µx b˜ P 1 1 = δ(x) δ(˜z) δ(t) → Green’s functions have factor in lieu of (3.3 SH line load in an orthotropic space 37 0. u y = g yy (˜r .11) ∂ x˜ λx µx ∂ z˜ λz µz where √ ρ b µ˜ = µ x µ z.10) ∂t ∂ x˜ 2 ∂ z˜2 ! ! ∂u y 1 µz ∂u y 1 µx τ˜ xy = µ˜ = τxy .16) µz ρ˜ ρ βx βx The equations of the previous section for an isotropic medium can now be applied using the equivalent properties defined above.

z.e. z. j = 1. No.20) ∂ xk µ ∂r ∂r r r $ (2)  2 (2)  % i H1 (S ) β H1 (P ) (2) ψ= − − H0 (S ) (3.4 In-plane line load (SV-P waves) Unit horizontal and vertical impulsive line sources act at x = 0.  and χ . sin θ = γz. The receiver is at x.. Note: The pairs of functions ψ. and Abascal. homogeneous space.19) µ    ∂gi j 1 ∂ψ ∂ψ 2χ χ = γk δi j + − γi γ j + [δik γ j + δ jk γi ] (3. homogeneous spaces z gθ gr Figure 3..38 Two-dimensional problems in full. Px = 1 x 3. 128–134. X defined below are Fourier transforms of each other. z = 0 in directions j = x. The response is given along the Cartesian directions i = x. z (3. 3. i. J. i. ω) = ψ δi j + χ γi γ j . 3 or x. Vol. Engineering Analysis.22) 4 α ∂ψ χ i (2) = + S H1 (S ) (3. respectively.24) ∂r 4r α r Cylindrical components 1 gr x = (ψ + χ ) (cos θ) (3. z and the cylindrical directions r. R. cos θ = θ γx . pp. 28. χ = F (X) (3. ψ = F () . 1.25) µ 1 gθ x = ψ (−sin θ) (3. eqs.18) 2 Frequency domain 1" # gi j (r.21) 4 S α P $ % i β 2 (2) (2) χ= H2 (P ) − H2 (S ) (3. cylindrical coordinates Here. On fundamental solutions for the boundary integral equations in static and dynamic elasticity. 20. . θ.3: Cartesian vs.26) µ 2 Dominguez. 1984.23) ∂r r 4r   ∂χ i β 2 (2) (2) χ = P H1 (P ) − S H1 (S ) − 2 (3. within an infinite.

p. eqs. 293. for a load in direction α. and Sneddon. Graff.31) 2π ⎩ t2 − t 2 tS 2 α tP 2 ⎭ S ⎧ ⎡  ⎤  1 ⎨ β 2⎣ 1 2 t 2 − tP2 ⎦ H(t − tP ) X=  + 2π ⎩ α t2 − t2 tP2 P ⎡  ⎤ ⎫ 1 2 t 2 − tS2 ⎬ − ⎣ + ⎦ H(t − tS ) (3.28) µ Time domain.32) 2 tS ⎭ t 2 − tS2 ∂ 1 tS2 X =   H (t − tS ) + (3. 3 Eason.37) µ 1 uθ z =  (cos θ) (3. K. Ohio State University Press.38) µ Figures 3. The generation of waves in an infinite elastic solid by a variable body. impulse response3 1" # ui j (r. . F.27) µ 1 gθ α = ψ (− sin(θ − α)) (3. Philosophical Transactions of the Royal Society of London. excluding the variation with θ.34) ∂r 2πr α t 2 − tP2 t 2 − tS2 r Cylindrical components Load in direction x 1 ur x = ( + X) (cos θ) (3.33) ∂r 2πr t 2 − t 2 3/2 r S $ % 2 ∂X 1 β 2 tP tS2 2X =  3/2 H (t − tP ) −  3/2 H (t − tS ) − (3.. J. N. Vol. Fulton.30) ∂ xk µ ∂r ∂r r r ⎧⎡  ⎤  ⎫  2 t 2 − t 2 1 ⎨⎣ 1 t 2 − tS2 ⎦ H(t − tS ) − β P ⎬ =  + H(t − tP ) (3. pp. 1956. 248.3. 5.35) µ 1 uθ x =  (− sin θ) (3. 575–607. t) =  δi j + Xγi γ j (3. 1975. Wave motion in elastic solids.36) µ Load in direction z 1 ur z = ( + X) (sin θ) (3.76–77.29) µ    ∂ui j 1 ∂ ∂X 2X X = γk δi j + − γi γ j + [δik γ j + δ jk γi ] (3.2. G.. 1 gr α = (ψ + χ) (cos(θ − α)) (3... I.4 In-plane line load (SV-P waves) 39 and more generally.4a–d show the displacement functions.

together with their derivatives.2 0. homogeneous spaces 0. are defined in section 3.5 Dipoles in plane strain Note: Solution for SV-P dipoles given in the frequency domain only. simply replace the functions ψ. Solid line: real part.1 -0.3 0.4a: Radial component of Green’s function for in-plane line load in full space with ν = 0. To obtain the time domain solution.3.4b: Tangential component of Green’s function for in-plane line load in full space with ν = 0.3. Dashed line: imaginary part.1 µ g rx 0 -0. to convert from Cartesian 0.k.4.1 µ gθ x 0 -0. χ by .1 -0.2 0 5 10 15 20 ωr β Figure 3.k. Dashed line: imaginary part. Strains (and stresses) Use the expressions given previously for the derivatives gi j. X. Solid line: real part.40 Two-dimensional problems in full.2 0 5 10 15 20 ωr β Figure 3. ui j. Varies as −sin θ with respect to load direc- tion.2 0. 3. These functions. respectively. .3 0. Varies as cos θ with respect to load direction. Also.

3.5 Dipoles in plane strain 41

0.6

0.4

ρβ ru rx

0.2

0
0 1 2 3 4 5

r
Figure 3.4c: Radial component of impulse response function for in-plane line
load in full space with ν = 0.3. Varies as cos θ with respect to load direction.

to polar coordinates we have used 2 γx γz = 2 sin θ cos θ = sin 2θ and γx2 − γz2 = cos2 θ −
sin2 θ = cos 2θ, see Fig. 3.3.

Single dipoles (Fig. 2.2)

Anti-plane dipoles (SH waves)


⎪ ∂g yy ˆ i S (2)
⎪ −γ j ∂r j = −γ j 4µ H1 (S )j
⎪ ˆ

G yj = ∂u yy ˆ t H (t − tS ) ˆ
2 j = x, z (3.39)

⎪ −γ j j= S  j

⎪ ∂r
⎩ 2π µr t 2 − t 2 S

0.6

0.4

ρβ ru θ x 0.2

0

-0.2
0 1 2 3 4 5

r
Figure 3.4d: Tangential component of impulse response function for in-plane line
load in full space with ν = 0.3. Varies as −sin θ with respect to load direction.

42 Two-dimensional problems in full, homogeneous spaces

In-plane dipoles (SV-P waves)
 
∂gxx ˆ ∂gzx ˆ
Gxx = − i+ k
∂x ∂x



1 ∂ψ ∂χ 2 2χ 2 ˆ ∂χ 2χ χ ˆ
=− γx + γx + γz i + γz − γx2 + k
µ ∂r ∂r r ∂r r r



1 ∂ψ ∂χ χ ∂ψ ∂χ χ ∂ψ χ
=− + + + cos 2θ + − rˆ − sin 2θ + tˆ
2µ ∂r ∂r r ∂r ∂r r ∂r r
(3.40a)
 
∂gxz ˆ ∂gzz ˆ
Gzx = − i+ k
∂x ∂x
 

 
1 ∂χ 2χ χ ˆ ∂ψ ∂χ 2χ
=− γz − γx2 + i + γx + − γz2 kˆ
µ ∂r r r ∂r ∂r r


 
1 ∂ψ ∂χ χ ∂ψ χ ∂ψ χ
=− sin 2θ + − rˆ + − + cos 2θ + tˆ (3.40b)
2µ ∂r ∂r r ∂r r ∂r r
 
∂gxx ˆ ∂gzx ˆ
Gxz = − i+ k
∂z ∂z

 

1 ∂ψ ∂χ 2χ ∂χ 2χ χ ˆ
=− γz + − γx2 ˆi + γx − γz2 + k
µ ∂r ∂r r ∂r r r


  
1 ∂ψ ∂χ χ ∂ψ χ ∂ψ χ
=− sin 2θ + − rˆ + − − + cos 2θ + tˆ
2µ ∂r ∂r r ∂r r ∂r r
(3.40c)
 
∂gxz ˆ ∂gzz ˆ
Gzz = − i+ k
∂z ∂z
 


1 ∂χ 2χ χ ˆ ∂ψ ∂χ 2 2χ 2 ˆ
=− γx − γz2 + i + γz + γz + γx k
µ ∂r r r ∂r ∂r r



1 ∂ψ ∂χ χ ∂ψ ∂χ χ ∂ψ χ
=− + + − cos 2θ + − rˆ + sin 2θ + tˆ
2µ ∂r ∂r r ∂r ∂r r ∂r r
(3.40d)
Double couple (Fig. 2.4)

M y = Gxz + Gzx


 
1 ∂ψ ∂χ 2χ χ ˆ ∂ψ ∂χ 2χ χ ˆ
=− γz +2 − γx2 + i + γx +2 − γz2 + k
µ ∂r ∂r r r ∂r ∂r r r



1 ∂ψ ∂χ χ ∂ψ χ
=− sin 2θ + − rˆ + cos 2θ + tˆ (3.41)
2µ ∂r ∂r r ∂r r

3.6 Line blast source: suddenly applied pressure 43

Torsional line source (Fig. 2.3)
1
Ty = (Gxz − Gzx )
2

'
1 ∂ψ χ & ˆ
=− − γzi − γx kˆ
2µ ∂r r


1 ∂ψ χ
=− − − tˆ (3.42)
2µ ∂r r
Note: T y is a vector along the y direction (i.e., into the paper), which implies a tor-
sional moment that is positive in the clockwise direction, that is, in the −tˆ tangential
direction.

Frequency domain
i (2)
Ty = S H1 (S )tˆ (3.43)
8µr

Time domain, impulse response
1 tS2
Ty =   H (t − tS ) tˆ (3.44)
4π µr t 2 − t 2 3/2
S

Blast load (compressive pressure)
See also the next Section 3.6, and the explanation for the factor (α/β)2 in Section 2.5.
 2
α
B2D = (Gxx + Gzz)
β


1 α 2 ∂ψ ∂χ χ & ˆ '
=− + + γx i + γzkˆ
µ β ∂r ∂r r


1 α 2 ∂ψ ∂χ χ
=− + + rˆ (3.45)
µ β ∂r ∂r r

Frequency domain
i (2)
B2D = − P H1 (P )ˆr (3.46)
4µr

Time domain, impulse response
1 tP2
B2D = −   H (t − tP ) rˆ (3.47)
2π µr t 2 − t 2 3/2
P

3.6 Line blast source: suddenly applied pressure
A line blast source acts within an infinitesimally small cylindrical cavity. The compressive
pressure of the line source has an implied space–time variation p(r, t) = H(t) δ(r )/2πr ,

44 Two-dimensional problems in full, homogeneous spaces

which, when integrated over the infinitesimal area of the cavity at times t > 0, gives a
blast source of unit strength. This line blast source produces only radial displacements
that consist of cylindrically propagating P waves. The response given is for the radial
displacement, positive away from the source.

Frequency domain
The displacement field caused by a harmonically pulsating line of pressure was shown in
the previous Section 3.5 to be given by

P (2) ωr
ur = H (P ), P = (3.48)
4 i µr 1 α

which is formally the frequency response function that corresponds to an impulsive line
of pressure. Observe, however, that in a strict sense, this expression does not have an
inverse Fourier transform, because its amplitude grows with frequency, and it thus fails
the Dirichlet conditions. Hence, the impulse response given previously is true only in the
sense of a distribution. To obtain the frequency response for a suddenly applied pressure
(i.e., a step load), we divide by iω and obtain

1 (2)
Gr (ω, r ) = − H (P ) (3.49)
4µα 1

∂Gr 1 & (2) (2)
'
= H1 (P ) − P H0 (P ) (3.50)
∂r 4 µ αr

Time domain, unit step load response
1 t r
Ur (t, R) =  H(t − tP ), tP = (3.51)
2π µr t 2 − t 2 α
P

⎡ ⎤
∂Ur t ⎣
tP2 1 ⎦ H(t − tP )
=  3/2 −  (3.52)
∂r 2π µr 2 t 2 − tP2 t 2 − tP2

3.7 Cylindrical cavity subjected to pulsating pressure4
A cylindrical cavity of radius r0 is subjected
( +∞ to an internal compressive internal pressure
p(t) with Fourier transform p(ω) = −∞ p(t) e−iωt dt, which elicits radial displacements
consisting of cylindrically propagating P waves. Response given is for the radial displace-
ment, positive away from the source.

4 Eringen, A. C. and Suhubi, S. S. (1975), Elastodynamics, Academic Press, Vol II, pp. 490–492.

3.7 Cylindrical cavity subjected to pulsating pressure 45

r0
Figure 3.5: Cylindrical cavity subjected to pressure.

p

Frequency domain
The displacement field caused by a harmonically pulsating pressure p(ω) at a distance
r ≥ r0 is
(2)
r0 p(ω) H1 (P )
gr (ω) = ,
µ 2H (2) (0 ) − a −2 0 H (2) ( 0 )
1 0

α ωr ωr0
a −2 = , P = , 0 = (3.53)
β α α
which depends on Poisson’s ratio because of the factor a.
Observe that if p(ω) = 1/πr02 , and we consider the limit of a vanishingly small cavity,
(2) (2)
i.e. r0 → 0, then 2H1 (0 ) → 4i/(π 0 ) = 4i r/(πr0 )/ P , P H0 (0 ) → 0. In this case,
the above formula converges to that of the line of pressure given in the previous section.

Time domain
No closed-form expression exists. However, a purely numerical solution is still possible by
Fourier synthesis:
 +∞
1
ur (r, r0 , t) = gr (ω) eiωt dω
2π −∞
 +∞ (2)
r0 p(ω) H1 (P )eiωt
= dω (3.54)
2π µ −∞
(2)
2H1 (0 ) − a −2 0 H0 ( 0 )
(2)

but this equation is surprisingly difficult to evaluate numerically, especially when the pres-
sure changes abruptly, because in that case the integrand falls off rather slowly. Indeed,
for an impulsive pressure p(ω) = 1, it can be shown that gr (ω) behaves asymptotically as
! 
α a2 r0 eiωt
gr (ω) eiωt = (3.55)
µ r iω

dashed line.2 µ rg r 0 -0. which facilitates comparisons. use the asymptotic expansion for the Hankel functions beginning at an appropriately large frequency ω0 (say. and ρ the mass density.2 -0. 5 Papoulis. A strategy to evaluate the impulse response would be to work in delayed time t  = t − t0 ≥ 0. imaginary part. 1962. for which accurate and efficient polynomial approximations exist:  +ω0  (2)  r0 H1 (P ) eiωt0 ur (r. that is. 0 > 2π. r0 . A.5 we infer that the magnitude of the response immediately behind the wave front is (see also Fig.46 Two-dimensional problems in full. t0+ ) = lim iω gr (ω) eiωt0 = = (3. and express the contribution of the two tails in terms of the sine integral.57) 2ρα r π Observe that the kernel of the above integral tends to a constant value at zero frequency. the denominator is never zero for any frequency or Poisson’s ratio. r0 . real part.b show the results for an impulsive pressure of strength p πr02 = δ(t) at a dimensionless distance r/r0 = 5 in a full space with Poisson’s ratio ν = 0.6. From this asymptotic structure and the properties of Fourier transforms.30.6a: Cylindrical cavity in full space with ν = 0. Figures 3. The Fourier integral and its applications. which means that the integrand has no real poles. 3. t) = eiω(t−t0 ) dω 2π µ −ω0 2H (2) (0 ) − a −2 0 H (2) (0 ) 1 0 ! 1 r0 2 + 1 − Si (ω0 (t − t0 )) (3. This pressure is consistent with that of a line of pressure in Section 3. . ω0 = 2π α/r0 ). Also.4 0. McGraw-Hill. homogeneous spaces 0.56) ω→∞ µ r ρα r with t0+ the instant immediately after passage of the wave front. Solid line.6a. where t0 = (r − r0 )/α is the time of arrival of P waves at the receiver station..3: frequency response at r/r0 = 5 for pressure pπr02 = 1. with t  = t − t0 .6b) ! !   α a 2 r0 1 r0 u(r.4 0 10 20 30 ωr α Figure 3.

. 3.7 Cylindrical cavity subjected to pulsating pressure 47 2.0 1.6b: Cylindrical cavity in full space with ν = 0. and the jump amplitude behind the wave front obtained previously from the asymptotic properties of Fourier transforms.0 0.6b agrees with the formula !  ρβr 2 1 r0 1 β r 3/2 2 ρα = = 1.5 0 1 2 3 tα r Figure 3.9 (3. Observe that the magnitude of the discontinuity in Fig.5 0 -0.5 ρβ r 2 ur 1. Impulse response at r/r0 = 5 for pressure pπr02 = δ(t).58) πr0 r π α r0 This expression is obtained by multiplying the displacement scaling factor.3. the applied pressure.3.

eqs. 1. in any coordinate direction. tS = . Engineering Analysis. 21. J. 128–134.1) α β ωR ωR β 1 − 2ν P = .. pp.2) α β α 2(1 − ν) xi ∂ cos θi 1 ∂ f (R) ∂f γi = cos θi = . within an infinite. γi. y.1 Fundamental identities and definitions R R  tP = . The receiver is at x1 .k = = (δik − γi γk). z) (4. a= = (4. i. 48 . 2. R. homogeneous spaces 4.6) α P P S S  2     β 3i 3 3i 3 χ = e−iP 1− − 2 − e−iS 1 − − 2 (4. 1984.4) 4π µR     ∂gi j 1 ∂ψ ψ ∂χ 3χ χ = γk − δi j + − γi γ j + [δik γ j + δ jk γi ] (4. ω) = {ψ δi j + χ γi γ j }.4 Three-dimensional problems in full. = γk (4.3) R ∂ xk R ∂ xk ∂R 4. x3 . R = x 2 + y2 + z2 (4. x2 .7) α P P S S 1 Dominguez. No.2 Point load (Stokes problem) Unit impulsive point sources act at the origin. On fundamental solutions for the boundary integral equations in static and dynamic elasticity. Vol. j = 1. 3. 3 (or x. 20. homogeneous space.5) ∂ xk 4π µR ∂R R ∂R R R  2     β i 1 i 1 ψ = e−iP + 2 + e−iS 1 − − 2 (4. and Abascal. Frequency domain1 1 gi j (R. S = .

3. Solid -0.4. full space with ν = 0.5 0.5 0 5 10 15 20 ωR β .5 2.2a: Green’s function for χ (ω).5 χ Figure 4.0 1.5 1. -0.5 line. -1.3.0 tβ R 1. full space with ν = 0.3.5 -1.5 ψ Figure 4. -1.5 0. real (imaginary) part.0 0 0.1b: Impulse response func- tion for (t). full space with ν = 0.5 (dashed) line.0 0. Solid (dashed) -0.2 Point load (Stokes problem) 49 1.5 ΨR 0 β Figure 4. real (imaginary) part.5 0 5 10 15 20 ωR β 1.1a: Green’s function for ψ(ω).

5 2.0 1.10e) 4π µR ψ gθ y = cos θ . (4.10g) 4π µR . (4. 0 -1 0 0.10c) 4π µR ψ +χ g Ry = sin φ sin θ . homogeneous spaces 4 3 2 ΧR β Figure 4.10b) 4π µR ψ gθ x = (−sin θ). Green’s functions due to forces in directions x.8) ∂R R α P P S S $  2 % ∂χ 1 6i 6 β 6i 6 = e−iS 3 + iS − − 2 − e−iP 3 + iP − − 2 (4. (4.10f) 4π µR ψ +χ g Rz = cos φ.3.0 tβ R $  2 % ∂ψ 1 β 2i 2 2i 2 = e−iP 1− − 2 − e−iS 1 + iS − − 2 (4.50 Three-dimensional problems in full.9) ∂R R S S α P P Spherical coordinates.2b: Impulse response func- 1 tion for X(t).10a) 4π µR ψ gφx = cos φ cos θ .10d) 4π µR ψ gφy = cos φ sin θ . (4. (4. (4. full space with ν = 0. z ψ +χ g Rx = sin φ cos θ .5 1. y. (4.

(4. H.13e) 4π µR  uθ y = cos θ .12) ∂ xk 4π µR ∂R R ∂R R R Spherical coordinates. and Richards. W.13i) Impulse response   β 2 H(t − tP ) H(t − tS )  = F −1 [ψ(ω)] = δ(t − tS ) − t − (4. (4. eq.13b) 4π µR  uθ x = (−sin θ).13a) 4π µR  uφx = cos φ cos θ .4. 4. p.13h) 4π µR uθ z = 0 (4. z:  +X u Rx = sin φ cos θ . y.23. I.13g) 4π µR  uφz = (−sin φ).13d) 4π µR  uφy = cos φ sin θ . K.11) 4π µR     ∂ui j 1 ∂  ∂X 3X X = γk − δi j + − γi γ j + [δik γ j + δ jk γi ] (4. (4.13f) 4π µR  +X u Rz = cos φ.. forces in directions x. Freeman. 73.14) α 2 tP 2 tS 2 Aki.10i) Time domain2 1 ui j = {δi j + X γi γ j } (4. 1980.2 Point load (Stokes problem) 51 ψ gφz = (−sin φ). (4. (4. Quantitative seismology: theory and methods. Vol. G. (4. 1980. (4.10h) 4π µR gθ z = 0 (4. (4. (4.13c) 4π µR  +X u Ry = sin φ sin θ . . P.

Response functions are listed only in the frequency domain. homogeneous spaces  2 β X = F −1 [χ(ω)] = δ(t − tP ) − δ(t − tS ) α   β 2 H(t − tP ) H(t − tS ) + 3t − (4.3 Tension cracks A tension point dipole is applied at the origin. tS ) = 2 τ P(t − τ ) dτ (4.22) ∂R β ∂t β  ∂ Q(t. tP .21) ∂R α ∂t α ∂ P(t − tS ) 1 ∂ P(t − tS ) ˙ − tS ) P(t =− =− (4. simply replace ψ. X . (4. for time domain solutions.20) α ∂ P(t − tP ) 1 ∂ P(t − tP ) ˙ − tP ) P(t =− =− . tS ) (4. χ by the .2.23) R ∂ 1 1 ∂ P(t − tS ) = (X − Q) − (4. tS ) β2 tS tP 2 tS = 2 P(t − tS ) − P(t − tP ) − τ P(t − τ ) dτ ∂R R β α R tP 1 = (Q − X ) (4.24) ∂R R β ∂t  2 ∂X β 1 ∂ P(t − tP ) 1 ∂ P(t − tS ) Q− X =− + +3 ∂R α α ∂t β ∂t R  2 Q − X ∂ β 1 ∂ P(t − tP ) =2 − − (4. tP .17) ∂R α ∂tP β ∂tS Arbitrary forcing function δ(x) δ(y) δ(z) P(t)  β 2 tS Q(t. tP . tP .19)  2 β X= P(t − tP ) − P(t − tS ) + 3 Q(t. see Fig.16) ∂R α ∂tP β ∂tS ∂X 1 ∂X 1 ∂X = + (4.25) R ∂R α α ∂t 4.15) α tP2 tS2 ∂ 1 ∂ψ 1 ∂ψ = + (4.52 Three-dimensional problems in full. 2.18) R tP  = P(t − tS ) − Q(t. tS ) (4.

replacing indices as appropriate. In spherical coordinates.27a) 4π µR ∂R ∂R R R R 1 ∂ψ ψ χ 1 Gφxx = − − + sin 2φ cos2 θ (4.29b) 4π µR ∂ R R R Gθ zz = 0 (4.2:   1 ∂ψ ψ χ ∂χ 3χ χ Gxx = −gi x. the factors in brackets are as follows: Frequency domain $ % ∂ψ ψ χ 1 β 2 −iP 6i 6 6i 6 − + = e 2− − 2 − e−iS 3 + i S − − 2 ∂R R R R α P P S S (4.27b) 4π µR ∂ R R R 2 1 ∂ψ ψ χ  1  Gθ xx = − − + − 2 sin φ sin 2θ (4.28c) 4π µR ∂ R R R 2 Crack zz   1 ∂ψ ψ ∂χ 2χ χ GRzz = − − + − cos2 φ + (4. The functions in brack- ets needed in these expressions are listed at the end of this section. the displacement components are: Crack xx   1 ∂ψ ∂χ ψ 2χ χ GRxx = − + − − sin2 φ cos2 θ + (4.30) .3 Tension cracks 53 given earlier in Section 4.4.26) and similar expressions for yy.28a) 4π µR ∂R R ∂R R R 1 ∂ψ ψ χ 1 Gφyy = − − + sin 2φ sin2 θ (4.28b) 4π µR ∂ R R R 2 1 ∂ψ ψ χ 1 Gθ yy = − − + sin φ sin 2θ (4. zz.29c) In the preceding expressions.x eˆ i = − − + δi x γx + − γi γx2 + γi eˆ i 4π µR ∂R R R ∂R R R (4.27c) 4π µR ∂ R R R Crack yy   1 ∂ψ ψ ∂χ 2χ χ GRyy = − − + − sin2 φ sin2 θ + (4.29a) 4π µR ∂R R ∂R R R 1 ∂ψ ψ χ  1  Gφzz = − − + − 2 sin 2φ (4.

the last term vanishes. for a time domain solution.36)  2 ∂  ∂X 2X M(t − tS ) β M(t − tP ) 1 ∂ M(t − tP ) Q − + − =3 − 4 + −9 ∂R R ∂R R R α R α ∂t R (4. χ by their Fourier transforms . Mkk = 2Gkk we recover twice the response functions for tension cracks. tS ) = τ M(t − τ ) dτ (4.37) 4.32) Time domain Source function = δ(x) δ(y) δ(z) M(t) (4.35) ∂R R R α R R β ∂t R  2 ∂X 3X M(t − tS ) 1 ∂ M(t − tS ) β M(t − tP ) 1 ∂ M(t − tP ) Q − = 6 + − 6 + − 15 ∂R R R β ∂t α R α ∂t R (4. See previous Section 4. 2.k + gik. we have Cartesian coordinates M jk = Mi jkeˆ i = − (gi j. Response func- tions are listed only in the frequency domain.3 for the expressions in brackets needed in either case. homogeneous spaces $  2 % ∂χ 3χ 1 15 i 15 β 15 i 15 − = e−iS 6 + i S − − 2 − e−iP 6 + i P − − 2 ∂R R R S S α P P (4. .54 Three-dimensional problems in full.4 Double couples (seismic moments) A seismic moment (point source) of unit strength is applied at the origin.4. X. For seismic moments.38) Observe that when j = k.34) R2 tP  2 ∂  X β M(t − tP ) M(t − tS ) 1 ∂ M(t − tS ) Q − + =2 − 3 + +6 (4. tP . With reference to Fig. simply replace ψ.33)  β2 tS Q(t.31)   2 % ∂ψ ψ ∂χ 2χ 1 −iS 9i 9 β 9i 9 − + − = e 3− − − e−iP 4 + i P − − 2 ∂R R ∂R R R S 2S α P P (4. j ) eˆ i   1 ∂ψ ψ χ ∂χ 3χ 2χ =− − + (δi j γk + δikγ j ) + − 2γi γ j γk + δ jkγi eˆ i 4π µR ∂R R R ∂R R R (4.

χ by their .39b) 4π µR ∂ R R R 1 ∂ψ ψ χ Mθ x = − − + cos φ cos θ (4. for a time domain solution.40c) 4π µR ∂ R R R Double couple in plane xy (normal z) 1 ∂ψ ψ ∂χ 2χ MRz = − − + − sin2 φ sin 2θ (4. Also.5 Torsional point source A torsional point source of unit strength is applied at the origin.41a) 4π µR ∂ R R ∂R R 1 ∂ψ ψ χ 1 Mφz = − − + sin 2φ sin 2θ (4.40b) 4π µR ∂ R R R 1 ∂ψ ψ χ Mθ y = − − + (−cos φ sin θ) (4.41c) 4π µR ∂ R R R 4. Response functions are listed only in the frequency domain. observe that the Mjk are displacements due to a seismic moment. replace ψ.40a) 4π µR ∂ R R ∂R R 1 ∂ψ ψ χ Mφy = − − + cos 2φ cos θ (4.39c) 4π µR ∂ R R R Double couple in plane xz (normal y) 1 ∂ψ ψ ∂χ 2χ MRy = − − + − sin 2φ cos θ (4. Double couple in plane yz (normal x) 1 ∂ψ ψ ∂χ 2χ MRx = − − + − sin 2φ sin θ (4.39a) 4π µR ∂ R R ∂R R 1 ∂ψ ψ χ Mφx = − − + cos 2φ sin θ (4.5 Torsional point source 55 Spherical coordinates Note: Instead of the double indices jk identifying the planes in which the double couples lie. and not moments.4.41b) 4π µR ∂ R R R 2 1 ∂ψ ψ χ Mθ z = − − + sin φ cos 2θ (4. we use in the following a single index identifying the direction of the normal to these planes.

1 ∂ψ ψ χ T jk = Ti jkeˆ i = − − − (δikγ j − δi j γk) eˆ i (4. or use the expressions listed below for an arbitrary source. we use in the following a single index identifying the direction of the nor- mal to these planes. Torsional moment in plane yz (normal x) TRx = 0 (4. xy for a positive (right-handed) torsional moment.45b) 8π µR ∂ R R R 1 ∂ψ ψ χ Tθ x = − − − (−cos φ cos θ) (4. homogeneous spaces Fourier transforms .45a) 1 ∂ψ ψ χ Tφx = − − − (−sin θ) (4.47a) 1 ∂ψ ψ χ Tθ z = − − − sin φ (4.46a) 1 ∂ψ ψ χ Tφy = − − − cos θ (4.56 Three-dimensional problems in full. With reference to Fig. 2. and observe that the reversed sequence jk = zy. zx.47b) 8π µR ∂ R R R See also the next section. observe that the Tjk are the displacements due to unit torsional moments.43) ∂R R R R ∂  X M(t − tS ) 1 ∂ M(t − tS ) − − =− + (4. X. and j = k gives zero.44) ∂R R R R β ∂t Spherical coordinates Note: Instead of the double indices jk identifying the planes in which the torsional moment lies.3 we have: Cartesian coordinates Note: Choose jk = yz.46b) 8π µR ∂ R R R 1 ∂ψ ψ χ Tθ y = − − − (−cos φ sin θ) (4. . yx leads to a negative moment.46c) 8π µR ∂ R R R Torsional moment in plane xy (normal z) TRz = Tφz = 0 (4.42) 8π µR ∂ R R R ∂ψ ψ χ 1 − − = − e−iS (1 + i S ) (4. Also. xz.45c) 8π µR ∂ R R R Torsional moment in plane xz (normal y) TRy = 0 (4.

52) 8π µ R 2 ∂τ Do notice. Particle motions are tangential to horizontal circles.6 Torsional point source with vertical axis Elicits SH waves that propagate spherically.50) 8π µ R 2 Time domain. because the source depends in addition on some other parameters. Response given in spherical coordinates.49) R R Frequency domain sin φ uθ = Mz(ω) (1 + i S ) e−iS (4. however. the dimensionless response uθ (τ ) R 2 is not invariant with respect to R. Preliminary definitions ωR R tβ S = .3: Torque about vertical axis. x2 x1 4. tS = . for .51) 8π µ R2 βR ∂t or in dimensionless time sin φ ∂ uθ = Mz(τ − 1) + Mz(τ − 1) (4. τ= (4. τ alone is not sufficient to fully define the response.6 Torsional point source with vertical axis 57 r Mz φ R Figure 4. For example. Thus. that in the dimensionless form.4. cos φ = (4.48) β β R r z sin φ = . arbitrary causal variation of source with time sin φ Mz(t − tS ) 1 ∂ Mz(t − tS ) uθ = + (4.

0 ≤ t ≤ td = 1.7 Point blast source Such a source elicits solely P waves that propagate spherically. Observe that this response is not a delayed.10 0.10 0 1 2 3 4 tβ τ= R Figure 4. Since the total wave energy flux must be constant. but both are needed to satisfy the wave equation in spherical coordinates. Figure 4.4: Response due to torsional pulse Mz = sin π t/td .5 (τs = 1 is the arrival time). the formulas above imply that the wave field is entirely contained within two concentric spheres of constant separation βtd . a source Mz = sin π(t/td ) = sin π(τ/τd ). 4. but only one. The response for such a source would be sin φ π µR 2 uθ (τ. with td being the duration. τd ) = sin(τ − 1) + cos(τ − 1) [H (τ − 1) − H (τ + τd − 1)] 8π τd (4. This can have practical consequences. this means that energy is gradually being transferred from one wavelet to the other.05 µ R 2uy 0 -0.53) The two Heaviside terms arise from the discontinuities in the source. homogeneous spaces 0. for example. One implication is that the shape of the torsional wave pulse evolves as it moves into the far field.5R/β. and that within this region it consists of two distinct wavelets that decay at different rates with increasing distance to the source. although it is not a priori obvious how this occurs. The explanation is that the individual wavelets cannot exist independently of one another.05 -0. there are not two wavelets. and looks different there from what it does in the near field. in the interpretation of cross-hole tests used to measure shear-wave velocities by means of torsional sources. More generally. the derivative term contains a factor π/τd = π R/(βtd ). scaled replica of the source. Particle motions are in the radial direction. . for an arbitrary torsional source M(t) of finite duration td . that is.58 Three-dimensional problems in full. which changes the relative mix of the two terms.4 shows the response due to a single pulse Mz = sin π(τ/τd ) when τd = 1.

except for the wave speed. the 3-D case is amenable to closed-form solution. 124.3 4. Vol.8 Spherical cavity subjected to arbitrary pressure 59 A singularly large pressure transient acts within an infinitesimally small. Journal of Engineering Mechanics. In contrast to the difficult-to-solve 2-D case of a cylindrical cavity. which. if the solution to this problem were found by superposition of the three tensile cracks in an infinite space. Particle motions are in the radial direction. By contrast.55) 16π µ R R α ∂t α Observe that.). spherical cavity. 0 = (4. 1998.56) 4µ 1 + i 0 − (0 α/2β)2 α Notice that the factor in square brackets is identical to the transfer function due to support motion of a single degree of freedom oscillator with natural frequency ωn . arbitrary causal variation of source with time: 3 b(t − tP ) 1 ∂b(t − tP ) R uR = + . Remarkably. . The source has an implied space–time variation p(R. a) Motion of cavity wall Frequency domain Frequency response function   R0 1 + i 0 ω R0 g0 (ω) = p(ω) . when integrated over the volume of the cavity. the particle’s effect does not vanish as its size shrinks to zero. This is because of the effect of the infinitesimally small particle that fills the cavity. 2 (Feb. P = (4. on Poisson’s ratio) but solely on the shear modulus µ. t) = b(t) δ(R)/4π R 2 . Contrary to intuition. pp. The response is given in spherical coordinates. 243–244. fraction 3 Kausel. reduces to the source strength b(t) used here.54) 16π µ R 2 α Time domain. the same solution would be obtained. the response is very similar to that of a torsional point source. the scaling factor does not depend on the constrained modulus λ + 2µ (i.e.8 Spherical cavity subjected to arbitrary pressure A harmonic or transient pressure acts within a spherical cavity of radius R0 . Blast loads versus point loads: the missing factor. and the scaling factor.. tP = (4. Frequency domain 3 ωR u R = b(ω) (1 + i P ) e−i P . No.4.. This pressure elicits P waves that propagate spherically. E. but with a factor 1/4(λ + 2µ) instead of 3/16µ. the lack of variation with polar angle. which arises when the solution is found using the dipoles for tensile cracks.

p of viscous critical damping ξ . it is remarkable that the resonant frequency should depend only on the shear modulus.61) 0 . Time domain Impulse response function R0 ωn2 −ξ ωn t u0 (t) = e cos (ωd t − φ) H(t) 4µ ωd  1 = 2(1 − ν) e−ξ ωn t cos (ωd t − φ) H(t) (4. homogeneous spaces R0 Figure 4. Also.59) 2ρβ √ ν λ 1 − 2ν sin φ = = . so it is independent of Poisson’s ratio. R0 ) = u0 ∗ p = u0 (τ ) p(t − τ ) dτ (4.60 Three-dimensional problems in full. not the dilatational (constrained) modulus. and damped frequency ωd : ω 1 + 2ξ i 1 + i 0 ωn  2 =  2 (4. the smaller the critical damping and thus the larger the response will be. ξ= . cos φ = (4.58) R0 α R0 2(1 − ν) Observe that the higher the Poisson’s ratio. The above analogy leads us immediately to the impulse response function u0 (t) given below.57) 1 0 α ω ω 1 + i 0 − 1 + 2ξ i − 4 β ωn ωn This implies an equivalent oscillator with the following parameters: 2β β ωn = . R0 α   2 2β β 2β ωd = ωn 1 − ξ 2 = 1− =  (4.5: Spherical cavity subjected to pressure.60) 1−ν λ + 2µ 1−ν Response to arbitrary pressure  t u R(t.

0 = . 4.4.67) α from which we obtain the response given below due to a unit impulsive pressure. For a harmonic pressure of the form p(ω)V0 = 1. that is. in which V0 = 43 π R03 is the volume of the cavity (i. as before. Time domain We obtain this response by considering a fictitious blast point source in an infinite medium beginning to act at an earlier time t0 = R0 /α. and g0 (ω) the frequency response function at the cavity wall.62) 4µR 2 1 + i 0 − (0 α/2β)2 (2) with h1 the second spherical Hankel function of first order.6a.66) 16π µ R R α ∂t with R − R0 t = (4. A plot of this response function is given in Fig..  3 b(t) 1 ∂b(t) 1 + = 2(1 − ν)e−ξ ωn t cos [ωd t − φ] H(t) (4.63) 16π µR 2 1 + i 0 − (0 α/2β)2 Setting in this formula 0 = 0.8 Spherical cavity subjected to arbitrary pressure 61 b) Response beyond the cavity Frequency domain Frequency response function (2) h1 (P ) g R (ω) = g0 (ω) (2) h1 (0 )  2 R0 1 + iP −i(R/R0 −1)0 ω R0 ωR = g0 (ω) e . with the same strength as a point blast load).64) 16π µ R0 R0 α ∂t 2ρβ The solution of this differential equation is  16π µ R20 b(t) = 2(1 − ν)e−ξ ωn t sin(ωd t) H(t) (4.e. and matching the displacement that it causes at R = R0 with the displacement produced by the impulsive pressure on the cavity wall. the frequency response function is   3 1 + i P g R (ω) = e−i(P −0 ) (4.65) 3 2ρβ The impulse response function at points beyond the cavity wall is then 3 1 b(t − t  ) 1 ∂b(t − t  ) uR = + (4. P = R 1 + i0 α α   1 1 + i P = p(ω)R03 e−i(R/R0 −1)0 (4. . we recover the point blast source.

1 µ R 2gR 0 -0.2 0.1 0 5 10 15 20 ωR α Figure 4.3 0. real part. .1 -0.6a: Spherical cavity of radius R0 in a full space with ν = 0. Notice that the phase lag does not enter into the sine term. imaginary part.45 sub- jected to a harmonic pressure p 43 π R03 = 1.45 subjected to an impulsive pressure p 43 π R03 = δ(t). in which case we recover the impulse response function at the cavity wall. By an appropriate 0.62 Three-dimensional problems in full.68) Observe that t  = 0 for R = R0 .6b: Spherical cavity of radius R0 in a full space with ν = 0. Solid line. homogeneous spaces 0. Response at R/R0 = 2. Response at R/R0 = 2. dashed line.1 ρβ R 3uR 0 -0. Impulse response function    2(1 − ν) −ξ ωn (t−t  ) R0 R0 u R(t) = e cos [ωd (t − t  ) − φ] − 1 − 2ρβ R R  × sin [ωd (t − t  )] H(t − t  ) (4.2 0 2 4 6 8 tα R Figure 4.

Earthquake Engineering and Structural Dynamics. 2000. Green’s functions for two-and-a-half dimensional elastodynamic problems. Kausel. 4. and Pei. D. 1998..9 Spatially harmonic line source (21/2-D problem)4 Consider an infinite. p = δ(t)/ 43 π R03 .70) A plot of this response function is given in Fig.5-D problem. The magnitude of the discontinuity at the wave front is obtained by setting t = t  . which gives 1 R0 u R(t  ) = (4. Vol. homogeneous three-dimensional space subjected at the origin of coor- dinates to a harmonic. A. 6 Pedersen. Journal of Engineering Mechanics... 1169–1183. j = 1. 1994.2 (4. the impulse response function is    3 2(1 − ν) −ξ ωn (t−t  ) R0 u R(t) = 2 e cos [ωd (t − t  ) − φ] − 1 − 8πρβ R0 R R  × sin [ωd (t − t  )] H(t − t  ) (4. including the response to moving loads. S. 3.). Three-dimensional scattering by two- dimensional topographies.6b.69) ρα R which is similar to that of the cylindrical cavity (Section 3. i = 1. 5 Papageorgiou. or x. Bulletin of the Seismological Society of America. A discrete wavenumber boundary element method for study of the 3-D response of 2-D scatterers. 1093–1097. the response to loads with arbitrary spatial variation in z can be obtained. 84. we can also obtain the response due to a pressure with an arbitrary variation in time. ´ F. Vol. and E. i. A. j = (δi j − γi γ j ) First derivatives of γi . 10 (Oct. 4. i. and Campillo. pp. γi γi = γ12 + γ22 = 1 Implied summations (4.e.2 Direction cosines in the x. M. J.7.6 Preliminary definitions 1 A= Amplitude (4. . No.75) r 4 Tadeu. as illustrated in Fig. Sanchez-Sesma.71) 4 iρω2   r = x 2 + y2 ≡ x12 + x22 Source–receiver distance (4. 2. Vol. 619–638.5 By an appropriate inverse Fourier transformation. pp. 126..9 Spatially harmonic line source (21/2-D problem) 63 convolution. A. 27.7). H.. pp.z).72) ∂r xi γi = = .y.y (4.73) ∂ xi r plane δii = δ11 + δ22 = 2. 4.74) 1 γi. spatially sinusoidal line load of the form b j (x) = δ(x)δ(y) exp i(ωt − kzz) acting in the jth spatial direction ( j = 1.4. In the special case of an impulsive pressure of unit strength p V = δ(t). except that it decays faster with increasing R. This is often referred to as the 2.

or in full 1 gxx = A ks2 H0β − B 1 + γx2 B 2 (4. homogeneous spaces y Figure 4.64 Three-dimensional problems in full.79) β   kα = k2P − k2z .85) ∂ x2 ∂y Green’s functions Cartesian coordinates The displacement in direction i due to a unit load applied in direction j.7: Spatially harmonic line source in a full 3-D space.l = −kβ γl H1β Derivative of Hankel function (4. kβ = k2S − k2z More wavenumbers (4.86a) r 1 g yy = A ks2 H0β − B 1 + γ y2 B 2 (4.83) in the z direction Bn = kβn Hnβ − kαn Hnα Bn functions (see end of section (4.81) H0β.76) r2 kz Wavenumber in z direction (4. x z 1 γi.2) H0β.z = −i kz H0β Derivative of Hankel function (4. can be written compactly as gi j = A(δi j k2s H0β + B0.77) ω kP = Wavenumber for P waves (4.86b) r .82) in the plane (l = 1.80) (2) (2) Hnα = Hn (kα r ).i j ). Hnβ = Hn (kβ r ) Shorthand for Hankel functions (4.78) α ω kS = Wavenumber for S waves (4. which has an implied factor exp i(ωt − kzz).84) for definitions and properties) ∂2 ∂2 ∇ˆ 2 = + 2 − k2z Laplacian (4. j k = (3γi γ j γk − γi δ jk − γ j δki − γk δi j ) Second derivatives (4.

4. Also. so that cos θ = γx . U.e. we recover the Green function B2-D for a line blast source of Section 3. tangential. gr y = u (sin θ) . γz = 0. gθ z = 0 (4. w..87b) r w = Ai kz B1 . but multiplied by µ/(λ + 2µ).87d)   W = A kS2 H0β − k2z B0 (4. since the space is homogeneous.86d) gxz = gzx = i kz γx AB 1 (4. using for this purpose the expressions given in Section 2.87a) r  1 v = A k2S H0β − B1 (4. W as  1 u = A k2S H0β − B1 + B2 . .88c) These can be used in turn to obtain the Green’s functions for 2. All other Green’s functions for dipoles are as in section 2. and seismic moments. gr z = U (4. Thus. Similarly. because there is no such thing as a 2. the Green’s function for a 2.87c) U = Ai kz B1 (4. Thus. (4.88b) gzx = w (cos θ) . the above expression is for a full solid. Observe also that we have changed the axial derivatives with respect to the source into axial derivatives with respect to the receiver.86c) gxy = g yx = γx γ y AB 2 (4. a Fourier transform could be used to recover the point blast source B3-D of section 4. and axial components of the Green’s functions are then gr x = u (cos θ) .7.5-D dipoles. so they need not be repeated. but multiplied by 4µ/3(λ + 2µ).88a) gθ x = v (−sin θ) . (4. torsional moments. define the parameters u.6.6. ∂/∂z = −∂/∂z. y plane. gθ y = v (cos θ) .87e) The radial.9 Spatially harmonic line source (21/2-D problem) 65   gzz = A ks2 H0β − k2z B 0 (4. ν. with θ being the angle between r and x.5.86e) g yz = gzy = i kz γ y AB 1 (4. when the above expression is specialized to the case of a line load (kz = 0 ⇒ kβ = kS ). sin θ = γ y . gzz = W (4.89) ∂r r ∂z ∂r r ∂z except that we have omitted here the factor 3(λ+ 2µ)/4µ associated with the presence of an infinitesimal cavity within which the blast load acts.5-D blast load in a homogeneous full space is    ∂u u − v ∂U ∂w w ∂ W ˆ B 2. gzy = w (sin θ) .5-D = − + + rˆ + + + k (4.86f) Cylindrical coordinates Let r be the source-receiver distance in the x. i. including the infinitesimal particle. For example.5-D cavity.

i = x.92c) εxy z = −2 i kz γx γ y AB2 (4.91f) r 2  4 εVol l = γl A −k2S kβ H1β + k2z B1 + B2 − B3 (4.y + gzl.92e) 2  1 ε zyz = 2 γ y A − k2S kβ H1β + k2z B1 (4. z (4.92f) 2  2 εVol z = i kz A −k2S H0β + k2z B0 + B1 − B2 (4. j + δ jl H0β. l = x.66 Three-dimensional problems in full. y  2 1 εlxx = γl A B2 − k2S kβ H1β δxl + B2 − γx2 B3 (4.90a) εil j = gil.91a) r r  2 1 εlyy = γl A B2 − k2S kβ H1β δ yl + B2 − γ y2 B3 (4. homogeneous spaces Strain components (l = x.i ) + 2AB0.91b) r r εzz l = γl k2z AB1 (4.92d)  1 εxz z = 2 γx A − k2S kβ H1β + k2z B1 (4.91c)  1 1 εlxy = 2A B2 − k2S kβ H1β (δxl γ y + δ yl γx ) − γx γ y γl B3 (4.i = k2S A(δil H0β.z = A k H0β + ∇ˆ 2 B0 (4.i jl .92b) r  2  εzz = i kz A −kS H0β + k2z B0 z (4. l = z  1 εxx z = i kz A B1 − γx2 B2 (4.92g) r . y.90c) ∂ xl S a) Strains for loads in the plane.x + g yl.(i) = k2S Aδil H0β.91g) r b) Strain for axial loads.(ii)l .91d) r 2  1 1 εlxz = 2i kz A B1 − k2S H0β δxl − γx γl B2 (4.91e) r 2  1 1 εlyz = 2i kz A B1 − k2S H0β δ yl − γ y γl B2 (4.90b) ∂  2  εVol l = gxl. i = j (4. y.92a) r  1 ε zyy = i kz A B1 − γ y2 B2 (4.(i) + AB 0. j + g jl. z = direction of load ) ε(ii) l = gil.

102) r  ∂ 2 4 ∇ˆ B0 = γi k2z B1 + B2 − B3 (4.i j k = (γi δ jk + γ j δki + γk δi j ) B2 − γi γ j γk B3 (4.93a) n n Bn = kβn Hn(2) (kβ r ) − kαn Hn(2) (kα r ) with knα ≡ (kα ) .93b) Properties dBn n Bn = = Bn − Bn+1 Recursion equation (4.96) 1 B0.101) r  4 B0.4.98) r 2 ∇ˆ 2 B0 = −k2z B0 − B1 + B2 (4. B0. B0.100) ∂ xi r r 1 B0.i j j = γi B2 − B3 (4.95) r B0. B0.i j = − B1 δi j + γi γ j B2 .103) ∂ xi r  ∂ 2 2 ∇ˆ B0 = i kz k2z B0 + B1 − B2 (4.z = −i kz B0 (4. knβ ≡ (kβ ) (4.i = γi Bn = γi Bn − Bn+1 (4.99) r  ∂ 1 1 B1 = −γi B2 (4.ii = − B1 + B2 .104) ∂z r .i = −γi B1 .97) r 2 B0.9 Spatially harmonic line source (21/2-D problem) 67 The Bn functions Definitions (2) (2) B0 = H0 (kβ r ) − H0 (kα r ) (4.zz = −k2z B0 (4.94) dr r n  Bn.i z = i kz γi B1 (4.

68 .

The origin of coordinates is taken at the free surface. The displacement at any other arbitrary point (x. 2004. t2 = (5. Preliminary definitions  ωr1 r1 r1 = (x − x0 )2 + (z − z0 )2 . and Kausel. 130. 2 = . z. z0 ) in the interior of an elastic half-space. ω) = − H0 (1 ) + H0 (2 ) (5. 1211–1222. Journal of Engineering Mechanics. The source is impulsive in time.3) 4µ Time domain ⎡ ⎤ 1 ⎣ H (t − t1 ) H (t − t2 ) ⎦ u yy (x. Impulse response of elastic half-space in the wavenumber–time domain. t1 = (5. 69 .4) 2π µ t 2 − t12 t 2 − t22 Mixed wavenumber–time domain1 A half-plane is subjected to a spatially harmonic anti-plane SH source at the surface with horizontal wavenumber k. 1 = . pp. Vol.2) β β Frequency domain i & (2) (2) ' g yy (r1 . The exact solution for the 1 Park.1 Half-plane. z) is obtained from the full space solution by the method of images. r2 . SECTION III: HALF-SPACE PROBLEMS 5 Two-dimensional problems in homogeneous half-spaces 5. 10. E. J.. t) =  +  (5. No.1) β β  ωr2 r2 r2 = (x − x0 )2 + (z + z0 )2 . SH line source and receiver anywhere A unit impulsive anti-plane line load is applied at some arbitrary point (x0 .

r˜2 = (x − x0 )2 + (z + z0 )2 (5.  +∞ 1 u y (x. The final equations are (for further details.9) βx βx . z. receiver at x. z.70 Two-dimensional problems in homogeneous half-spaces z x0 . apply the formulas above in Section 5. z. r2 . t) e−ikx dk (5. t˜ j = . impulse response function at depth z is   py (k) u yy (k. t) = J0 kβ t 2 − tz2 H (t − tz).8) µz µz ωr˜ j r˜ j  ˜j= . After reducing the problem to the isotropic form.5) ρβ in which J0 is the Bessel function of the first kind. t1 .2 SH line source in an orthotropic half-plane Scale the vertical coordinate as done in Section 3. tz = z/β (5.1. 2 (5. t2 . z0 . j = 1. see Section 3. z0 Figure 5. z r1 x0. z. − z 0 r2 x x. Hence.3 for the full space.6) −∞ is the spatial Fourier transform of the anti-plane load. and  +∞ py (k) = py (x) eikx dx (5. observing that scaled vertical coordinates must be used to define r1 .7) 2π −∞ 5.1: SH source in a half-space at depth x0 .3) ! ! µx µx r˜1 = (x − x0 )2 + (z − z0 )2 . t) = u yy (k. Method of images.

617. Academic Press.11) 2π µx t 2 − t˜ 2 t 2 − t˜ 2 1 2 5. Note: eq.13a) π µ |x| ⎪ ⎪ √ ⎪ ⎪ ⎪ ⎪ − τ2 − 1 ⎪ ⎩ √ √ τ >1 (2τ 2 − 1)2 − 4τ 2 τ 2 − a 2 τ 2 − 1 ⎧ ⎪ ⎪ 0 τ <a ⎪ ⎪ ⎪ ⎪ √ ⎪ ⎨ −(2τ 2 − 1)2 τ 2 − a 2 Pβ a≤τ ≤1 uzz = (2τ − 1) + 16τ (τ 2 − a 2 )(1 − τ 2 ) 2 4 4 (5. 7. C. The source has dimensions [F][T]/[L] = [impulse]/[length in y direction]. z.13d) Pβ 1 u yy = √ H (τ − 1) (5. SV-P source and receiver at surface (Lamb’s problem)2 An impulsive in-plane line source P is applied on the surface of a lower half-space. A.13b) π µ |x| ⎪ ⎪ √ ⎪ ⎪ ⎪ ⎪ − τ 2 − a2 ⎪ ⎩ √ √ τ >1 (2τ 2 − 1)2 − 4τ 2 τ 2 − a 2 τ 2 − 1 ⎧ √ √ ⎪ ⎪ 2τ (2τ 2 − 1) τ 2 − a 2 1 − τ 2 ⎪ ⎨ a≤τ ≤1 Pβ (2τ 2 − 1)4 + 16τ 4 (τ 2 − a 2 )(1 − τ 2 ) uxz = (5.7 in that book has an error that affects the coefficient of the Dirac delta term in uxz . δ = Dirac delta (5. reverse the sign of the coupling terms. . a= = . r˜2 . 2 Eringen. t) =  +  (5. This has been corrected here. SV-P source and receiver at surface (Lamb’s problem) 71 i & (2) ' g yy (˜r1 . and Suhubi. Also. 1975.3 Half-plane.16. For an upper half-space z > 0.13c) π µx ⎪ ⎪ π(2τR2 − 1)3 ⎪ ⎩   δ(τ − τR ) else 4 1 − 4τR2 + 8τR6 (1 − a 2 ) uzx = −uxz (5. S.12) |x| CR α 2(1 − ν) ⎧ ⎪ ⎪ 0 τ <a ⎪ ⎪ ⎪ ⎪ √ ⎪ ⎨ 4τ (1 − τ ) τ 2 − a 2 2 2 Pβ a≤τ ≤1 uxx = (2τ 2 − 1)4 + 16τ 4 (τ 2 − a 2 )(1 − τ 2 ) (5. τR = .5.13e) π µ|x| τ 2 − 1 Note: Observe that x in uxz has no absolute sign. (β/x)δ(τ − τR ) = sgn x δ(t − tR ). II.3 Half-plane. Vol.10) 4µx ⎡ ⎤ 1 ⎣ H (t − t˜1 ) H (t − t˜2 ) ⎦ u yy (x.. tβ β β 1 − 2ν τ= . Elastodynamics. p. and displacements are observed there. ω) = − H0 ( ˜ 1 ) + H0(2) ( ˜ 2) (5. S.

5 0 -0.5 1.5 tβ τ= r -1.0 0 0.0 ρβ ruxx 0. -0.5 tβ τ= r -1.0 1.5 0 Figure 5.0 0 0.0 .5 2.10 ρβ ruxx 0.0 ρβ ru xx 0.5 1. ν = 0.5 2.0 1.2: Lamb’s problem in 2-D.25.5 2.72 Two-dimensional problems in homogeneous half-spaces 1.0 0.0 1.5 1.0 1.05 0 tβ τ= r 0 0.

receiver at interior. SV-P source on surface.20c)   Im τ 2 − a 2 ≤ 0. or vice versa. in which case the formulas of Section 5. or vice versa 73 5.4 Half-plane. or vice versa3 An impulsive line load is applied on the surface of an elastic half-space.20d) that is. Chapter 11. a = = (5. Finally. Elsevier. . Re qβ2 + a 2 ≥ 0 (5.20a)   Re qα2 + a 2 ≥ 0.3 must be used. it follows that   qα = i τ sin θz + cos θz τ 2 − a 2 . qβ2 + 1 = τ cos θz + i τ sin θz τ 2 − 1 (5. τ + i qβ sin θz − cos θz 1 + qβ2 = 0 (5.15) where q denotes a dimensionless parameter. sin θz = .19) These parameters must satisfy Reqa ≥ 0. the critical dimensionless time is  cos(θz − θc ) θz ≤ θc τc = with θc = arcsin a = critical angle (5. This√is most easily accomplished in the computer by defining τα = conj τ 2 − a 2 and τβ = conj τ 2 − 1. = i sin θz + √ = √ ∂τ τ 2 − a2 τ 2 − a2 ∂τ τ2 − 1 τ2 − 1 (5.18)   ∂qα τ cos θz q2 + a 2 ∂qβ τ cos θz qβ + 1 2 = i sin θz + √ = √ α . r = x 2 + z2 .16) with the constraint that the dimensionless time τ must remain real. Re qβ2 + 1 ≥ 0 (5. receiver at interior. the terms involving square roots in time must have a negative imaginary part √ when τ is small. From here. L. Solution fails if z = 0. Im τ 2 − 1 ≤ 0 (5.5. qβ = i τ sin θz + cos θz τ 2 − 1 (5.17)     qα2 + a 2 = τ cos θz + i τ sin θz τ 2 − a 2 . The Cagniard–De Hoop paths for P and S waves are defined by   τ + i qα sin θz − cos θz a 2 + qα2 = 0. W. SV-P source on surface. Preliminary definitions βt |x| |z|  β 1 − 2ν τ= . (1979).21) 1 θ z > θc 3 Pilant. cos θz = .20b)   Re qα2 + 1 ≥ 0. Reqβ ≥ 0 (5. and displacements are observed at an interior point.14) r r r α 2(1 − ν)   R(q2 ) = (1 + 2q2 )2 − 4q2 q2 + a 2 q2 + 1 = Rayleigh function (5. Elastic waves in the earth.4 Half-plane.

74 Two-dimensional problems in homogeneous half-spaces

The ensuing formulas for displacements at z = 0 apply to:
r A lower half-space z < 0, with the load applied on the surface at x = 0, z = 0 and the
displacements observed in its interior at z < 0.
r An upper half-space with the load applied on the surface at x = 0, z = 0 and the
displacements observed in its interior at z > 0.
The formulas also may be used for either a lower or an upper half-space in which the load
is applied in its interior at x = 0, z = 0, and the displacements are observed on the surface
z = 0. In this case, replace uxz → −uzx and uzx → −uxz, that is, exchange the coupling terms
and reverse their signs (i.e., apply the reciprocity and symmetry principles, and reverse the
signs to take account of the location of a receiver to the right of the point of application
of the buried load). Observe that in general, uxz = −uzx , i.e., the coupling terms are not
antisymmetric.
⎧   ⎫
β ⎨ 2q2 1 + q2 ∂q
⎪ 1 + 2qβ2 1 + qβ2 ∂q ⎪

α β
uxx = Re − α   α H (τ − a) +   H (τ − τc ) (5.22a)
π µr ⎪
⎩ R qα2 ∂τ R qβ2 ∂τ ⎪

⎧    ⎫
β ⎨2q 1 + q2 a 2 + q2 ∂q qβ 1 + 2qβ2 ∂q ⎬
α α α α β
uzx = Im   H (τ − a) −   H (τ − τc ) sgn(x)sgn(z)
π µr ⎩ R qα2 ∂τ R qβ2 ∂τ ⎭

(5.22b)
⎧    ⎫

⎨ q 1 + 2q2 ∂q ⎬
β α α
2qβ a + qβ 1 + qβ ∂qβ
2 2 2

uxz = Im  α H (τ − a) −   H (τ − τc ) sgn(x)sgn(z)
π µr ⎩ R q2 ∂τ R q2 ∂τ ⎭
α β

(5.22c)
⎧  ⎫
β ⎨ 1 + 2q2  a 2 + q2 ∂q 2qβ2 a 2 + qβ2 ∂qβ ⎬
α α α
uzz = Re   H (τ − a) −   H (τ − τc ) (5.22d)
π µr ⎩ R qα 2 ∂τ R qβ2 ∂τ ⎭

Motions on epicentral line (x = 0)

τ= (5.23)
|z|
 2 
Dα (τ ) = τ 2 − a 2 + 12 − τ (τ 2 − a 2 ) τ 2 − a 2 + 1 (5.24)
  
1 2
Dβ (τ ) = τ − 2
2
− τ (τ − 1) τ 2 − 1 + a 2
2
(5.25)
$   √ √ %
β τ τ2 − 2
2 1
τ τ 2 − a2 τ 2 − a2 + 1
uxx (0, z, t) = √ H (τ − 1) − H (τ − a) (5.26a)
2π µ |z| Dβ τ − 1
2 Dα
$   √ √ %
β τ 2 τ 2 − a 2 + 12 τ τ 2 − 1 τ 2 − 1 + a2
uzz(0, z, t) = √ H (τ − a) − H (τ − 1)
2π µ |z| Dα τ 2 − a 2 Dβ (τ )
(5.26b)

uzx = uxz = 0 (5.26c)

5.4 Half-plane, SV-P source on surface receiver at interior, or vice versa 75

2
ρβ ruxx

1

0

x = 100
-1 x = 10
x=1 tβ
x=0 τ=
r
-2
0 0.5 1.0 1.5 2.0
2
ρβ ruzz

1

Figure 5.3: Displacements at surface due
0 to impulsive line loads at depth h = 1, ν =
0.25 (Lamb’s problem).
x = 100
-1 x = 10
x=1
x=0 tβ
τ=
r
-2
0 0.5 1.0 1.5 2.0
2
ρβ ruxz
x = 100
x = 10
x=1
1 x=0

0


τ=
r
-1
0 0.5 1.0 1.5 2.0

76 Two-dimensional problems in homogeneous half-spaces

4
µ rux

x=1
2 x=10
x=100

0


-2 τ=
r
Figure 5.4: Horizontal (top) and vertical
0 0.5 1.0 1.5 2.0
(bottom) displacements due to line blast
source at depth h = 1, ν = 0.25 (Garvin’s
2
µ ruz problem).

-2

x=1
x=10
x=100
-6


τ=
r
-10
0 0.5 1.0 1.5 2.0

5.5 Half-plane, line blast load applied in its interior (Garvin’s problem)4
A line blast source acts within an infinitesimally small cavity at x = 0 and some depth
h = |z| below the free surface. The singularly large pressure source is a step load in time
and of magnitude such that in the limit of a vanishingly small cylindrical cavity, the integral
of the pressure over the cross section is unity. Using the same notation as for the line source
in the Section 5.4, it is found that the displacements on the surface are
$  %
1 2qα 1 + qα2 ∂qα
ux (x, 0, t) = Im sgn(x) H (τ − a) (5.27a)
π µr R (qα2 ) ∂τ
$  %
−1 1 + 2qα2 ∂qα
uz(x, 0, t) = Re sgn(z) H (τ − a) (5.27b)
π µr R (qα ) ∂τ
2

4 Garvin, W. W. (1956), Exact transient solution of the buried line source problem, Proceedings of the Royal
Society of London, Series A, Mathematical and Physical Sciences, Vol. 234, No. 1199 (Mar.), pp. 528–541.

5.5 Half-plane, line blast load applied in its interior (Garvin’s problem) 77

This formula is valid for both an upper (z > 0) and a lower (z < 0) half-space, with z being
the coordinate of the source relative to the surface. For a lower half-space and at large
times, these displacements converge asymptotically to the static solution
2(1 − ν)
ux (x, 0, ∞) = sin θz (5.28a)
πµ r
2(1 − ν)
u y (x, 0, ∞) = cos θz (5.28b)
πµ r

6 Three-dimensional problems in
homogeneous half-spaces

6.1 3-D half-space, suddenly applied vertical point load on its surface
(Pekeris-Mooney’s problem)1
A vertical point source that varies as a step function in time is applied onto the surface.
Displacements are also observed on the surface at a range r.

Definitions

r = x 2 + y2 (6.1)
 2
β 1 − 2ν tβ
a2 = = , τ= , (6.2)
α 2(1 − ν) r

⎨ 1 t > t0
H (t − t0 ) = 12 t = t0 = Heaviside step function (6.3)

0 t < t0
 π/2

K(k) =  , (6.4)
0 1 − k2 sin2 θ
 π/2

(n, k) =  (6.5)
0 (1 + n sin2 θ) 1 − k2 sin2 θ
K and  are complete elliptical integrals of the first and third kind, respectively.
Consider the Rayleigh function
 
R(ξ 2 ) = (1 − 2ξ 2 )2 + 4 ξ 2 − 1 ξ 2 − a 2 = 0 (6.6)
in which ξ = β/cis a dimensionless
 wave slowness, with c a wave velocity. Multiplying
by (1 − 2ξ 2 )2 − 4 ξ 2 − 1 ξ 2 − a 2 and dividing by ξ 2 , one obtains the bicubic (i.e., cubic
in ξ 2 ) equation
1 − 8ξ 2 + 8ξ 4 (3 − 2a 2 ) − 16ξ 6 (1 − a 2 ) = 0 (6.7)

1 Eringen, A. C. and Suhubi, S. S., 1975, Elastodynamics, Academic Press, Vol. II, p. 748. Solution up to ν =
0.2631.

78

i. ξ32 ]. When ν < ν0 = 0.6.e.1 3-D half-space. ξ1 = ξ2 . a) ν < 0. while ξ3 ≡ β/CR is the actual true root. Vol. 473–491.11a) ⎧ ⎪ ⎪ 0 τ <a ⎪ ⎪  2    ⎪ ⎪ 1 − 2ξ12 a 2 − ξ12 1 A3 ⎪ ⎨ P(1 − ν) 8 Re ξ 2 − ξ 2 ξ 2 − ξ 2   Q − Q−1  +  2 −4 a<τ <1 uzz = 1 2 1 3 1 1 ξ3 − τ 2 16π µr ⎪ ⎪ ⎪ ⎪ 2A3   ⎪ ⎪  H ξ3 − τ − 8 τ >1 ⎪ ⎩ ξ2 − τ2 3 (6.10) 1−a 2 a − ξi 2 1 − 4ξ3 + 8(1 − a 2 )ξ36 2 Observe that n1 > 0. ni = 2 . Mooney presents the solution for arbitrary values of Poisson’s ratio. the false roots are repeated. 64. ξ22 . When ν = ν0 .5]. suddenly applied vertical point load on its surface 79 which has three roots [ξ12 . all three roots are real and satisfy 0 < ξ12 < ξ22 < a 2 < 1 < ξ32 .2631 ⎧ ⎪ ⎪0 ⎡ τ <a ⎪ ⎪ ⎤ ⎪ ⎪ ⎪ ⎪1 )3 Ai P(1 − ν) ⎨ ⎣1 − * * ⎦ a<τ <1 uzz = 2 i=1 *τ 2 − ξ 2 * (6.8b) in which  2 * *    1 − 2ξi2 *a 2 − ξ 2 * 1 − 2ξi2 1 − ξi2 i Ai =  2   . −1 + H (τ − ξ3 ) τ >1 ⎪ ⎪ 2K(k ) Bi i k ) ⎩ (1 − a 2 )3/2 i=1 τ 2 − ξ32 (6.2631 (ξ 1 . pp. but n3 < 0. It defines the point beyond which the false roots turn complex.2631. k) a<τ <1 Pτ ⎨ (1 − a ) 2 3/2 ur z = i=1 8π 2 µr ⎪ ⎪   ⎪ ⎪ k−1 3 2π C ⎪ ⎪ −1 − (n . ..9) 4 ξi − ξ j ξi − ξk 2 2 2 ξi − ξ j ξi − ξk 2 2  2 3 τ 2 − a2 1 − a2 2ξ3 − 1 k2 = . ξ 2 are complex conjugates)2 ur z not available (6. M. The transition value ν 0 is the root of the discriminant D(ν) = 32ν 3 − 16ν 2 + 21ν − 5 = 0 in the interval [0 ≤ ν ≤ 0. Bulletin of the Seismological Society of America. Bi =  2  2  ξi = ξ j = ξk (6.11b) 2 Mooney.. n2 > 0. b) ν > 0. No.8a) 2πµr ⎪ ⎪ i ⎪ ⎪ A3 ⎪ ⎪ 1−  H (ξ3 − τ ) τ > 1 ⎪ ⎪ ⎩ ξ32 − τ 2 ⎧ ⎪ ⎪ 0   τ <a ⎪ ⎪ ⎪ ⎪ 1  3 ⎪ ⎪ 2K(k) − Bi (k2 ni . C= (6. but does so only for the vertical component. 2. Some numerical solutions for Lamb’s problem. the first two of which are non-physical solutions of the rationalized Rayleigh function. 1974. H.

0 µ ru rz 0.0 in which A3 is as before. ν = 0.2 Figure 6.6 tβ τ= r -1. 0. replace Q1 → 1/Q1 . it is not enough to simply always add the square root term. it behooves one to check the absolute value of Q1 . because the proper sign of the square root depends also on the complex quadrant that z lies in. Hence.√If |Q1 | > 1. -0.1b: Vertical displacement at surface due to suddenly applied verti- cal point load. Thus.5 1.25 (Lamb–Pekeris– Mooney problem). Observe that ξ12 is complex and that Q2 (τ ) = 1 + 2z − z2 + z satisfies Q1 Q2 = 1.5 2.2 tβ τ= r -0.0 1. that is. ν = 0. and  a 2 − ξ12 Q1 (τ ) = 1 + 2z + z2 + z.5 1.2 µ ru zz -0.1a: Radial displacement at sur- face due to suddenly applied vertical point load.12) τ 2 − a2 provided that |Q1 | < 1.2 0 0. the test |Q1 | < 1 implicitly assigns the proper sign to the difference Q1 − Q−1 1 . and replace it with its reciprocal as needed.80 Three-dimensional problems in homogeneous half-spaces 1. However.0 .25 (Lamb–Pekeris– Mooney problem). which may change with time τ .5 2.0 1. Q−1 1 = Q2 . 0. z= (6.0 0 0.6 Figure 6.

6.2 3-D half-space, suddenly applied horizontal point load on its surface 81

6.2 3-D half-space, suddenly applied horizontal point load on its surface
(Chao’s problem)3
A horizontal point source that varies as a step function in time is applied onto the surface.
Displacements are also observed on the surface at a range r.
v = 0.25 only (solutions for other values of Poisson’s ratio unknown).

√  √   √ 
a= 1
3
3, ξ12 = 14 , ξ22 = 1
4
3− 3 , ξ32 = 14 3 + 3 (6.13)
  
√ √ √
C1 = 3
4
3, C2 = 18 6 3 + 10, C3 = 18 6 3 − 10, r = x 2 + y2 (6.14)



⎪ 0 ⎤ τ <a
⎪ ⎡




⎪ C C2 C3
P ⎨τ2 ⎣ 1 − − ⎦ a<τ <1
ur x = τ 2 − ξ12 τ 2 − ξ22 ξ32 − τ 2 (6.15a)
2π µr ⎪


⎪ 2τ 2 C3

⎪ 1−  [1 − H (τ − ξ3 )] τ >1


⎩ ξ32 − τ 2



⎪ 0
    τ < a
−3P ⎨ 1 4
uθ x = − C1 τ 2 − ξ2 − C
2 τ 2 − ξ2 + C
3 ξ 2
− τ 2 a<τ <1
8π µr ⎪

2 3

1 2 3

1 − 83 C3 ξ32 − τ 2 [1 − H(τ − ξ3 )] τ >1
(6.15b)

uzx = −uxz : see vertical load (6.15c)

The actual Cartesian displacements produced by a horizontal load in direction x are then

ux = ur x cos2 θ − uθ x sin2 θ, u y = (ur x + uθ x ) sin θ cos θ, uz = uzx sin θ (6.16)

in which θ is the azimuth.
Observation:
√ Experiments
√ demonstrate√ that the  coefficients Ci are of the form C1 =
3
64
6 D1 , −C2 = 64
2
6 D2 , −C3 = 64 2
6 D3 , Di = |1 − 2ξi2 |/[4(ξi2 − ξ 2j )(ξi2 − ξk2 )]. If the
coefficients in front of the Di were all equal to the same constant, it would mean that this
constant would depend only on Poisson’s ratio. Unfortunately, this is not the case, so these
expressions do not help in generalizing the expressions above for arbitrary Poisson’s ratio.
In addition, the sign of C3 in uθx is opposite to that in urx .

3 Chao, C. C., 1960, Dynamical response of an elastic half-space to tangential surface loadings, Journal of
Applied Mechanics, Vol. 27 (Sept.) pp. 559–567.

82 Three-dimensional problems in homogeneous half-spaces

0.2
µ rurx

0.1

0 Figure 6.2a: Radial displacement at
surface due to suddenly applied hor-
izontal point load, ν = 0.25 (Chao’s
-0.1 problem). Varies as cos θ.

-0.2

τ=
r
-0.3
0 0.5 1.0 1.5 2.0

Displacements on the epicentral axis (r = 0, z = 0)
At all times, uxx (0, z, t) = ur x (0, z, t), uθ x (0, z, t) = −ur x (0, z, t), uz(0, z, t) = 0. Define the
auxiliary functions
 
τ τ 2 − 13 τ 2 + 23
f (τ ) =  2   (6.17)
2τ 2 + 13 − 4τ τ 2 − 13 τ 2 + 2
3
 
τ 2 (τ 2 − 1) 2τ τ 2 − 2
3
− (2τ 2 − 1)
g(τ ) =  (6.18)
(2τ 2 − 1)2 − 4τ (τ 2 − 1) τ 2 − 23

0.07
µ ruθ x

0
Figure 6.2b: Tangential displacement at
surface due to suddenly applied hor-
izontal point load, ν = 0.25 (Chao’s
problem). Varies as −sin θ.
-0.07


τ=
r
-0.14
0 0.5 1.0 1.5 2.0

6.3 3-D half-space, buried torsional point source with vertical axis 83

z

φ2

h R2

h (x, y, z)
φ1
Mz(t) R1

(x0, y0, z0)
z0

x
Figure 6.3: Torsional point source in elastic half-space.

in which τ = tβ/|z|. The horizontal displacement on the epicentral axis is then

τ <a
P ⎨
0
uxx (0, z, t) = − f (τ ) a<τ <1 (6.19)
2π µ|z| ⎩ 1 2
2
(τ + 1) − f (τ ) + g(τ ) 1 < τ

As written, each function in the last expression grows as τ 2 , but the sum tends to the
constant value 5/8. Cancellation errors can be avoided by using asymptotic expansions for
each term.

6.3 3-D half-space, buried torsional point source with vertical axis
Consider a torsional point source with vertical axis (i.e., perpendicular to the free surface)
applied at an arbitrary point (x0 , y0 , z0 ). The response is observed at any other arbitrary
point (x, y, z). The distance from the point of application of the torque to the free surface
is h ≥ 0. If the origin is taken at the free surface, then z0 = −h. This problem is similar to
that of a torsional point source in a full space. Its solution is obtained by application of the
method of virtual images, as shown in Fig. 6.3.

Note: If the axis of the torque is not vertical, then reflections at the surface may partially
convert to waves other than SH waves, in which case the solution provided below breaks
down. On the other hand, a transversely anisotropic (or cross-isotropic) half-space can be
considered by recourse to the scaling method presented earlier for anti-plane loads in a
2-D full space.

84 Three-dimensional problems in homogeneous half-spaces

0.3

0.1

-0.1

ω

-0.3
0 0.5 1.0 1.5 2.0
Figure 6.4a: Tangential displacement) at (r, z) = (2, −1) due to unit har-
monic torsional source at z0 = −2 in half-space with µ = 1, β = 1. Solid
line, real part; dashed line, imaginary part.

Preliminary definitions

R1 = (x − x0 )2 + (y − y0 )2 + (z − z0 )2 (6.20)

R2 = (x − x0 )2 + (y − y0 )2 + (z − z0 − 2h)2 (6.21)

 2
z − z0 R1
sin φ1 = 1− , sin φ2 = sin φ1 (6.22)
R1 R2
ω Rj Rj
j = , tj = , j= 1, 2 (6.23)
β β

0.075

2 1
0.050
2

0.025

0

-0.025

t
-0.050
0 2 4 6
Figure 6.4b: Tangential displacement) at (r, z) = (2, −1) due to single
torsional sine pulse of duration td = 1 at z0 = −2 in half-space with µ =
1, β = 1.

6.3 3-D half-space, buried torsional point source with vertical axis 85

Frequency domain


1 sin φ1 sin φ2
uθ = Mz(ω) (1 + i 1 ) e−i1 + (1 + i 2 ) e−i2 (6.24)
8π µ R12 R22

Time domain, arbitrary causal variation of source with time


1 sin φ1 Mz(t − t1 ) 1 ∂ Mz(t − t1 )
uθ = +
8π µ R1 R1 β ∂t


sin φ2 Mz(t − t2 ) 1 ∂ Mz(t − t2 )
+ + (6.25)
R2 R2 β ∂t
These solutions are valid for both an upper and a lower half-space. Figure 6.4 illustrates
these response functions for a source at depth h = 2, and a receiver at depth d = 0.4h
and range r = 0.8h. The origin is at the surface of the half-space with µ = 1, β = 1,
and the torsional source is a single sine pulse of the form M(t) = sin π ttd , 0 ≤ t ≤ td = 1.
Observe that the response consists of a direct pulse and a reflected pulse, and is dominated
by the wavelet in the derivative of the moment. The wavelet in M is apparent only close
to the source.

86 .

Both external surfaces are traction-free.2 Normal mode solution x − x0 z z0 ωh ξ= . z.5) h h h β  κ j = S − π 2 j 2 .6) φ j = cos jπ ζ.3) 4µ j=−∞ ⎡ ⎤ 1  ∞ ⎣H (t − t1 j ) H (t − t2 j ) ⎦ u yy (t. z0 is the point of application of the load. 7. where x0 . x0 . . x. x0 . then r1 j = r2 j . in which case each pair in the summations consists of identical terms. z0 ) = − + (7. 2 j = . 2. ζ = . with 0 ≤ z0 ≤ h.1.2) β β β β ∞ &  ' i (2) (2) g yy (ω. x.1. z.1 Plate subjected to SH line source Consider a homogeneous plate of infinite length and arbitrary thickness h subjected to a harmonic exp(iωt) or impulsive δ(t) line source of the form b(x. t1 j = . ζ0 = . see Figure 7.4) 2π µ j=−∞ t 2 − t12j t 2 − t22j Observe that when z0 = 0. 1. z. j= 0.7) $ % i e−iS |ξ |  ∞ cos(π jζ ) cos(π jζ0 ) e−iκ j |ξ | g yy (ω. . t2 j = (7.1 Solution using the method of images   r1 j = (x − x0 )2 + (z − z0 − 2 j h)2 . S = (7.1) ωr1 j r1 j ωr2 j r2 j 1 j = . Im (κ j ) ≤ 0 Dimensionless modal wavenumber 2 (7. Normal mode (7.1. 7. x0 . SECTION IV: PLATES AND STRATA 7 Two-dimensional problems in homogeneous plates and strata 7. z0 ) =  +  (7. z) = δ(x − x0 ) δ(z − z0 ). r2 j = (x − x0 )2 + (z + z0 − 2 j h)2 (7. The origin of coordinates is at the bottom surface. . x. z0 ) = − H 0 (1 j ) + H 0 (2 j ) (7.8) µ 2S j=1 κj 87 .

50 0 1 2 3 4 ωh 2πβ Figure 7.3. Green’s function at (x. the index of the first neglected mode may be established by setting a threshold of exponential decay. Observe that the high modes have negative imaginary characteristic wavenumber (They are evanescent modes).25 µ gyy 0 -0.1: Plate subjected to SH line source. Also. say. 7.e. The mode for j = 0. upper surface). and ξ is the shortest range (source–receiver distance).2 Stratum subjected to SH line source Consider a homogeneous stratum of infinite length and arbitrary thickness h subjected to a harmonic exp(iωt) or impulsive δ(t) line source of the form b(x. is referred to as the rigid body mode. so these decay exponentially in x.. except near the source. with 0 ≤ z0 ≤ h. For small x.50 0. where x0 . 0.75h Figure 7. The origin of coordinates is at the rigid bottom interface.75h) due to SH line load at elevation z0 = h (i.88 Two-dimensional problems in homogeneous plates and strata h h 0.25 -0. 0.2a: Homogeneous plate. z) = δ(x − x0 ) δ(z − z0 ). . exp(−iκ j ξ ) < ε. where displacements are zero. where ε is some small number. the summation can safely be truncated above j  S /π . see Figure 7. which carries weight 1/2. κ j = 0 at each of the cutoff frequencies of the plate. Hence. z0 is the point of application of the load. z) = (h. so the last mode to change from imaginary to real contributes the most. The upper surface is traction-free.

7.9) ωr1 j r1 j ωr2 j r2 j 1 j = . 2 j = . so each pair in the summations cancels out. x0 .10) β β β β i  ∞ & ' (2) (2) g yy (ω. 0. z. t2 j = (7.75h Figure 7.0 0.2b: Figure Homogeneous plate.7. x. t1 j = .2 Stratum subjected to SH line source 89 1. because then r1 j = r2 j . r2 j = (x − x0 )2 + (z + z0 − 2 j h)2 (7. z. z) = (h.11) 4µ j=−∞ ⎡ ⎤ 1  ∞ H (t − t1 j ) H (t − t2 j ) ⎦ u yy (t.4 0. z0 ) = − (−1) j H0 (1 j ) − H0 (2 j ) (7.6 ρβ hu yy 0. impulse response function at (x.3: Stratum subjected to SH line source. h h 0. z0 ) = (−1) j ⎣  −  (7.1 Solution using the method of images   r1 j = (x − x0 )2 + (z − z0 − 2 j h)2 .75h) due to SH line load at elevation z0 = h. x0 .8 0. x.2.12) 2π µ j=−∞ t 2 − t12j t 2 − t22j Observe that the response is null when z0 = 0. .2 0 0 5 10 15 20 βt τ= h Figure 7.

z0 ) = − (7..5 . 0.75h) due to SH line load at elevation z0 = h (i. with 0 ≤ z0 ≤ h.e.13) h h h β l = 12 (2 j − 1) = j − 12 . . Im(κ j ) ≤ 0 Modal wavenumber (7. real part.2. Also. say. upper surface). Hence.ζ = . see Figure 7. z0 is the point of application of the load. the summation can safely be truncated above l  S /π.25 -0. that is. . ζ0 = . exp(−iκ j ξ ) < ε.50 0. κ j = 0 at each of the cutoff frequencies of the stratum. S = (7.50 0 1 2 3 4 5 ωh 2πβ Figure 7. where ε is some small number. The plate has mixed boundary conditions. Green’s function at (x. 7.14)  κ j = 2S − π 2l 2 . Modal indices (7. Solid line. imaginary part.2 Normal mode solution x − x0 z z0 ωh ξ= . 7. z.15) φ j = (−1) sin lπ ξ j Normal mode (7. but . 3 .17) µ j=1 κj Observe again that the high modes are evanescent and thus decay rapidly in x.16) i ∞ sin(π lζ ) sin(π lζ0 ) e−iκ j |ξ | g yy (ω. For small x. either the normal or the tangential displacements are restrained. The origin of coordinates is at the bottom surface. z) = (h. dashed line.90 Two-dimensional problems in homogeneous plates and strata 0. z) = δ(x − x0 ) δ(z − z0 ). x.3 Plate with mixed boundary conditions subjected to SV-P line source Consider a homogeneous plate of infinite length and arbitrary thickness h subjected to a harmonic exp(iωt) or impulsive δ(t) in-plane (SV-P) line source of the form b(x. where x0 . the index of the first neglected mode may be established by setting a threshold of exponential decay. x0 . 2. j = 1. and ξ is the shortest range (source–receiver distance).4a: Homogeneous stratum. so the last mode to change from imaginary to real contributes the most.25 µ gyy 0 -0.

z). Note: Other boundary combinations. Denoting a free condition as F.75h) due to SH line load at elevation z0 = h. the frequency response functions and the impulse response functions for an in-plane line load in a homogeneous full space. Also. 7. CF-CF. and listing first the tangential (horizontal) and then the normal (vertical) boundary condition. Uzz(x. z1 j = z − z0 − 2 j h. γnz = z j /rnj . Gzx (x. z).75h 0. z).1 Solution using the method of images Let Gxx (x.25h Figure 7. define the receiver coordinates relative to the location of the image sources as x j = x − x0 . Uxz(x. impulse response function at (x.18)  rnj = x 2j + znj 2 . with the bottom pair of conditions being listed first. z) and Uxx (x. z).19) h h 0.5: Plate with mixed boundary conditions. z2 j = z + z0 − 2 j h (7.3 Plate with mixed boundary conditions subjected to SV-P line source 91 0.3. Uzx (x. then a plate with mixed boundary conditions can have any of the following four combinations: FC-FC. FC-CF.25 -0. cannot be solved by the methods used to obtain the formulas that follow.25 ρβ huyy 0 -0. CF-FC. z).4.50 0.4b: Homogeneous stratum.50 0 5 10 15 20 βt τ= h Figure 7. 0. z) be. Gxz(x. respectively. . n = 1. Gzz(x. z) = (h.7. see Section 3. γnx = x j /rnj . z). such as FF-FF (Mindlin plate) or CC-FF (stratum). and a constrained condition as C. not both at the same time on the same surface. 2 (7.

z0 ) = [Ui x (x j . z0 ) = [Gi x (x j . z. z2 j )] (7. z1 j ) ∓ Ui z(x j . z (7. x. z (7. z. z0 ) = (−1) j [Gi x (x j .2 Normal mode solution Preliminary definitions x − x0 z z0 ωh ωh ξ= . z0 ) = [Ui z(x j . x0 . z2 j )].22b) j=−∞  ∞ ui x (t. z.25)  κP j = 2P − π 2 j 2 . x0 . z.22a) j=−∞  ∞ gi z(ω.24) h h h α β Definitions for cases a). x0 . i = x. b)  κS j = 2S − π 2 j 2 . z.92 Two-dimensional problems in homogeneous plates and strata Using the upper sign for FC-FC and the lower sign for CF-CF. z0 ) = [Gi z(x j . z1 j ) ± Gi x (x j . z0 ) = (−1) j [Ui z(x. x. S = (7. z1 j ) ± Ui x (x j . z2 j )] (7. x. z0 ) = (−1) j [Gi z(x j . z1 j ) ∓ Ui z(x.3. z. z2 j )]. z1 j ) ∓ Gi z(x j . z2 j )] (7.26) a) FC-FC $ % ∞  1 j 2 π 2 −iκS j |ξ | uxx = 1  e−iP |ξ | 2 P + κP j e−iκP j |ξ | + e cos( jπ ζ ) cos( jπ ζ0 ) iµ 2S j=1 κS j (7.20a) j=−∞  ∞ gi z(ω. z0 ) = (−1) j [Ui x (x j . x0 . z1 j ) ± Ui x (x j . x0 . x.20b) j=−∞  ∞ ui x (t. x0 . the formulas are  ∞ gi x (ω. z.23a) j=−∞  ∞ ui z(t. z. x0 . i = x.23b) j=−∞ 7. the formu- las are  ∞ gi x (ω. z2 j )] (7. x. x. z1 j ) ∓ Gi z(x j .27a) . x. z2 j )] (7. ζ0 = . x. Im (κP j ) ≤ 0 Modal wavenumber for P waves (7. z2 j )] (7.21b) j=−∞ For a mixed boundary condition FC-CF (upper sign) and CF-FC (lower sign). Im (κS j ) ≤ 0 Modal wavenumber for S waves (7.21a) j=−∞  ∞ ui z(t. x0 . P = . z1 j ) ± Gi x (x j . ζ = .

30)  κPl = 2P − π 2l 2 .29)  κSl = 2S − π 2l 2 . Im (κPl ) ≤ 0 Modal wavenumber for P waves (7.32d) iµ S j=1 κPl 2 .28b) µ S j=1 2 sgn(ξ )  ∞   uxz = − jπ e−iκP j |ξ | − e−iκS j |ξ | sin( jπ ζ ) cos( jπ ζ0 ) (7. Im (κSl ) ≤ 0 Modal wavenumber for S waves (7.27c) µ 2S j=1 ∞ 2 2 1  j π −iκP j |ξ | −iκS j |ξ | uzz = e + κ S j e sin(π jζ ) sin(π jζ0 ) (7. 52 .32b) µ 2S j=1 sgn(ξ )  ∞   uxz = + lπ e−iκPl |ξ | − e−iκSl |ξ | cos(lπ ζ ) sin(lπ ζ0 ) (7.31) c) FC-CF ∞ 1  l 2 π 2 −iκSl |ξ | uxx = κPl e−iκPl |ξ | + e cos(πlζ ) cos(πlζ0 ) (7.3 Plate with mixed boundary conditions subjected to SV-P line source 93 sgn(ξ )  ∞   uzx = − jπ e−iκP j |ξ | − e−iκS j |ξ | sin( jπ ζ ) cos( jπ ζ0 ) (7.27b) µ 2S j=1 sgn(ξ )  ∞   uxz = + jπ e−iκP j |ξ | − e−iκS j |ξ | cos( jπ ζ ) sin( jπ ζ0 ) (7.28d) Definitions for cases c) and d) l = 12 (2 j − 1) = 12 .32a) iµ S j=1 2 κSl sgn(ξ )  ∞   uzx = − lπ e−iκPl |ξ | − e−iκSl |ξ | sin(lπ ζ ) cos(lπ ζ0 ) (7.27d) iµ 2S j=1 κP j b) CF-CF ∞ 1  j 2 π 2 −iκS j |ξ | uxx = κP j e−iκP j |ξ | + e sin(π jζ ) sin(π jζ0 ) (7. 32 .28c) µ S j=1 2 $ % ∞ 2 2  1 j π uzz = 1  e−iS |ξ | 2 S + e−iκP j |ξ | + κS j e−iκS j |ξ | cos( jπ ζ ) cos( jπ ζ0 ) iµ 2S j=1 κP j (7. · · · (7.32c) µ 2S j=1 ∞ 2 2 1  l π −iκPl |ξ | uzz = e + κSl e−iκSl |ξ | sin(πlζ ) sin(πlζ0 ) (7.7.28a) iµ S j=1 2 κS j sgn(ξ )  ∞   uzx = + jπ e−iκP j |ξ | − e−iκS j |ξ | cos( jπ ζ ) sin( jπ ζ0 ) (7.

Solid line. dashed line. d) CF-FC ∞ 1  l 2 π 2 −iκSl |ξ | uxx = κPl e−iκPl |ξ | + e sin(πlζ ) sin(πlζ0 ) (7. real part. imaginary part.25h. 0. Bottom: time domain.6a: Homogeneous plate with mixed boundary conditions FC-CF. Horizontal response at (x.33b) µ S j=1 2 .05 -0.1 0 2 4 6 8 10 βt τ= h Figure 7.75h) due to horizontal SVP line load at elevation z0 = 0. ν = 0.1 ρβ huxx -0. Top: frequency domain.05 µgxx 0 -0.33a) iµ S j=1 2 κ Sl sgn(ξ )  ∞   uzx = + lπ e−iκPl |ξ | − e−iκSl |ξ | cos(lπ ζ ) sin(lπ ζ0 ) (7.10 0. z) = (h.25.10 0 1 2 3 4 5 ωh 2πβ 0.94 Two-dimensional problems in homogeneous plates and strata 0.

1 µ gzx 0 -0.25.7.33d) iµ S j=1 2 κ Pl Note: The modal solution is much faster and more accurate than the frequency domain solution with the method of images. 7. ν = 0. 0.6a.1 -0.6b: Homogeneous plate with mixed boundary conditions FC-CF. sgn(ξ )  ∞   uxz = − lπ e−iκPl |ξ | − e−iκSl |ξ | sin(lπ ζ ) cos(lπ ζ0 ) (7.2 0.33c) µ 2S j=1 ∞ 2 2 1  l π −iκPl |ξ | uzz = e + κSl e−iκSl |ξ | cos(πlζ ) cos(πlζo) (7.z) = (h. Solid line.6b. 7. real part.2 0 1 2 3 4 5 ωh 2πβ 0. Vertical response at (x. Bottom: time domain. Top: frequency domain. imaginary part.1 0 2 4 6 8 10 βt τ= h Figure 7.6c shows the results for a plate . dashed line.25h.75h) due to horizontal SVP line load at elevation z0 = 0.1 ρβ huzx -0.3 Plate with mixed boundary conditions subjected to SV-P line source 95 0. Figures 7.

25.05 µ gzz 0 -0. for Poison’s ratio ν = 0.25. the normal mode solution uses 20 modes. real part.1 0 2 4 6 8 10 βt τ= h Figure 7. but with some low amplitude ripples that decrease as the number of images is increased.6c: Homogeneous plate with mixed boundary conditions FC-CF. ν = 0. The frequency domain solution with the method of images and 50 images gives similar results to the modal method. with mixed boundary conditions FC (bottom) and CF (top).10 0 1 2 3 4 5 ωh 2πβ 0.96 Two-dimensional problems in homogeneous plates and strata 0.1 ρβ huzz -0. dashed line. Bottom: time domain. Top: frequency domain. . The time domain solution uses solely the mirror images whose arrivals at the receiver fall within the observation window.z) = (h. Vertical response at (x.05 -0.25h. Solid line.75h) due to vertical SVP line load at elevation z0 = 0. imaginary part.10 0. 0.

cylindrical. each of which depends on one coordinate only (whether or not the systems are layered). Thus. of which the first two present the theoretical frame- work and equations needed to obtain the practical numerical methods and formulations included in the third. such as finite and infinite plates. and less so for their own utility. or even infinite media. Finally. and then proceeds to give full derivations to these equations and to the wave equation in all three coordinate systems. or spherical coordinates (also called by some the spectral element method). Chapter 8 begins with a summary of the solutions to the scalar and vector Helmholtz equations in three-dimensional space. This greatly simplifies applications to layered media. SECTION V: ANALYTICAL AND NUMERICAL METHODS Read me first Section V consists of three chapters. Unlike most books on the theory of elasticity. Chapter 10 contains the detailed equations needed to implement the stiffness matrix method for layered media in Cartesian. Examples are included that show what purposes it serves and how the method is used. Examples are also given to illustrate the application of these concepts. A concise description is presented in the introduction and in Section 10. Nonetheless. the reader may wish to either browse lightly Chapters 8 and 9 – especially the example problems in Chapter 9 – or skip these altogether at first and return later only as may be needed to clarify matters. and manage to express the final results as products of matrices. we decided to include this material herein in full length because no readily available reference exists containing the detailed derivation of the powerful stiffness matrix method for all the three major coor- dinate systems.1. we use matrix algebra throughout. Chapter 9 makes a compact introduction to the integral transform methods commonly used to analyze stratified media. This is a powerful numerical tool that allows solving dynamically loaded laminated systems consisting of arbitrarily thick layers. or spheres. Chapters 8 and 9 are for the most part meant as a reference to Chapter 10. Thus. rods. 97 .

z) = c1 eikx x + c2 e−ikx x c3 eiky y + c4 e−iky y c5 eikz z + c6 e−ikz z (8. but the corresponding solutions for both the Helmholtz and wave equations in cylindrical and (especially) spherical coordinates are much less com- mon in the literature.6)    1 ∂ψ 1 Ψ = ψ kˆ + 2 ∇ + ∇ × χ kˆ (8. we present in the ensuing the detailed solutions in the frequency– wavenumber domain for the following three equations: Scalar Helmholtz equation ∇ 2 + kP2 = 0 (8. which is restricted to separable coor- dinate systems and to appropriate constitutive properties. We begin with a summary of the results in all three coordinate systems. and the parameters aj . and pro- vide thereafter the detailed solutions. All of these carry an implied exponential factor eiωt .3) We solve these equations by separation of variables. equations.8 Solution to the Helmholtz and wave equations The solution to the wave equations in Cartesian coordinates can be found in most books on the theory of elasticity.4)       ψ(x. such as isotropy and transverse isotropy. Thus.8) kP ∂z kS ∂z 98 . bj .5)     χ(x.2) Wave equation.1 Summary of results Cartesian coordinates     (x.1) Vector Helmholtz equation ∇ · ∇Ψ + kS2 Ψ = 0 (8. cj are arbitrary integration constants. isotropic medium (λ + 2µ) ∇∇ · u − µ∇ × ∇ × u = ρ u¨ (8. z) = a1 eikx x + a2 e−ikx x a3 eiky y + a4 e−iky y a5 eikz z + a6 e−ikz z (8.7) kS ∂z kS 1 ∂ 1 ∂ 2χ 1 ∂ψ ˆ 1 ∂ 1 ∂ 2χ 1 ∂ψ ˆ u= + 2 + i+ + 2 − j kP ∂ x kS ∂ x ∂z kS ∂ y kP ∂ y kS ∂ y ∂z kS ∂ x  2 1 ∂ 1 ∂ χ + + 2 kS2 χ + 2 kˆ (8. and nota- tion needed later on. y. to establish the fundamental methods. 8. z) = b 1 eikx x + b 2 e−ikx x b 3 eiky y + b4 e−iky y b5 eikz z + b6 e−ikz z (8. y. y.

horizontally layered media ¯ u(x.17)    1 ∂ψ 1 Ψ = ψ kˆ + 2 ∇ + ∇ × χ kˆ (8. z. c = phase velocity (8.1 Summary of results 99 Plane strain.19) kα ∂z kβ ∂z    cos nθ − sin nθ cos nθ Tn = diag (either the top or bottom elements) sin nθ cos nθ sin nθ (8. θ. ω) = [ux uy −i uz]T = u e−ikx (overbar for imaginary factor) (8.12) αk α βk β k ≡ kx = horizontal wavenumber.15)    ψ(r. z) = [ur uθ uz]T = Tn Cn u (8. z.18) kS ∂z kS     1 ∂ 1 ∂ 2χ 1 ∂ψ 1 ∂ 1 ∂ 2χ 1 ∂ψ u= + 2 + rˆ + + 2 − tˆ kα ∂r kβ ∂r ∂z kβ r ∂θ kα r ∂θ kβ r ∂θ ∂z kβ ∂r    1 ∂ 1 ∂ 2χ + + 2 kS2 χ + 2 kˆ (8. z) = (b 1 cos nθ + b 2 sin nθ) b 3 Jn (kα r ) + b4 Yn (kα r ) b5 eikz z + b6 e−ikz z (8. s= 1− . θ.20) Horizontally layered system (Observe the similarities with Cartesian coordinates.13) ⎧ ⎫ ⎧ ⎫ ⎨ 1 0 −s ⎬ ⎨1 0 s⎬ R1 = 0 1 0 .11) !  ω 2 !  c 2  2  2 ω c p= 1− = 1− .24) αk βk . ω) = R1 E−1 z a1 + R2 Eza2 (a1 . a2 = 3 × 1 vectors of constants) (8.16)    χ (r.) u(r.21) u(k. z) = (c1 cos nθ + c2 sin nθ) c3 Jn (kα r ) + c4 Yn (kα r ) c5 eikz z + c6 e−ikz z (8. s= 1− = 1− (8.23) !  ω 2  2 ω p= 1− .10) " # Ez = diag ekpz e ksz eksz (8.8.9) u(k. k = radial wavenumber (8. ω) = R1 E−1 z a1 + R2 Eza2 (a1 . a2 : 3 × 1 vectors of constants) (8. R2 = 0 1 0 (8. z. θ.14) ⎩ ⎭ ⎩ ⎭ −p 0 1 p 0 1 Cylindrical coordinates    (r. z) = (a1 cos nθ + a2 sin nθ) a3 Jn (kα r ) + a4 Yn (kα r ) a5 eikz z + a6 e−ikz z (8. θ.22) " # Ez = diag e kpz e ksz e ksz (8.

28) ⎧ ⎫ ⎪ ⎪   (1) kz  (1)  ⎪⎪ ⎪ (1)  Hβn ⎪ ⎪ Hαn n Hβn ⎪ ⎪ ⎪ ⎪ ⎪ kβ r kβ ⎪ ⎪ ⎪ ⎨ ⎪ (1)   (1) ⎬ H(1) = Hαn (1)  kz Hβn . Hβn = Hn(2) (kβ r ). and are thus not truly independent solutions to the Bessel equation. when the argument of the Bessel functions is complex. uθ −i uz]T = Tn u e−ikz z (overbar for imaginary factor) (8. (2) Hαn = d(kα r ) d(kα r ) (8.100 Solution to the Helmholtz and wave equations ⎧ ⎫ ⎪ ⎪ J n Jn 0⎪⎪ ⎨ n kr ⎬ d Jn (kr ) Cn = kr n Jn Jn 0 . Hβn = . (8. kz. (1) Hαn = .31)   (1)   (2) (1) (2) (1)  dHβn (2)  dHβn Hβn = Hn(1) (kβ r ). a2 : 3 × 1 vectors of constants) (8.26) ⎩ ⎭ ⎩ ⎭ −p 0 1 p 0 1 Cylindrically layered system ¯ θ. Jn = (8. kβ = kS2 − kz2 . . Hβn = d(kβ r ) d(kβ r ) (8. Jn and Yn are virtually proportional. (kz = axial wavenumber) (8.27) u(r. z.30)   (1) dHαn   (2) dHαn (1) Hαn = Hn(1) (ka r ). ω) = H(1) n a1 + H(2) n a2 (a1 .25) ⎪ ⎪ ⎪ ⎪ d(kr ) ⎩ 0 0 J ⎭ n ⎧ ⎫ ⎧ ⎫ ⎨ 1 0 −s ⎬ ⎨1 0 s⎬ R1 = 0 1 0 . we could also have used the standard Bessel and Neumann functions Jn .29b) n ⎪ ⎪ n Hβn n ⎪ ⎪ ⎪ k α r kβ kβ r ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ − kz H (2) (2) ⎪ ⎪ ⎩ 0 H ⎭ kα αn βn   kα = kP2 − kz2 .29a) n ⎪ ⎪ n Hβn n ⎪ ⎪ ⎪ kα r kβ kβ r ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ kz (1) ⎪ (1) ⎪ ⎪ ⎩ − Hαn 0 Hβn ⎭ kα ⎧ ⎫ ⎪ ⎪   (2) kz  (2)  ⎪⎪ ⎪ (2)  Hβn ⎪ ⎪ Hαn n Hβn ⎪ ⎪ ⎪ ⎪ ⎪ kβ r kβ ⎪ ⎪ ⎪ ⎨ ⎪ (2)   (2) ⎬ H(2) = Hαn (2)  H kz βn (8. ω) = [ur u(r. and especially when it is an imaginary number. R2 = 0 1 0 (8. Yn in place of the Hankel functions. However. (2) Hαn = Hn(2) (ka r ).32) Note: In the matrices Hn above.

44) with parameter kP = ω/α. hm (zS ) (8. z) = eikx x eiky y eikz z (8. θ ) = (a1 cos nθ + a2 sin nθ) (a3 jm(kR) + a4 ym(kR)) (a5 Pmn + a6 Qmn ) (8. (8. z) is a scalar function of the coordinates. a2 : 3 × 1 vectors of constants) (8. hm (zP ). zS = kS R (8. φ.35)  1 ∂ Rψ 1 Ψ = R ψ rˆ + 2 ∇ + ∇ × (Rχ rˆ ) (8.45) ∂ x2 ∂ y2 ∂z2 where (x. y. h Sm = h(1) (2) m (zS ). In Cartesian coordinates. φ. hm (zS ) .33) ψ(R.42)     H(1) m = Hm h(1) (1) m (zP ).41) h Pm = h(1) (2) m (zP ).37)     cos nθ cos nθ −sin nθ Tn = diag (either the top or bottom elements) sin nθ sin nθ cos nθ (8. Making an ansatz (x.46) . ω) = Tn Lnm H(1) m a1 + Hm a2 (2) (a1 .2 Scalar Helmholtz equation in Cartesian coordinates 101 Spherical coordinates (R. y. hm (zS ). φ.8.36) k ∂R k   u(R.38) ⎧ n ⎫ ⎪ ⎪ P 0 0 ⎪ ⎪ ⎪ m ⎪ n ⎪ ⎪ ⎨ 0 d Pmn n Pm ⎪⎪ ⎬ Lnm = dφ sin φ .2 Scalar Helmholtz equation in Cartesian coordinates Consider the Helmholtz equation involving a scalar function : ∇ 2 + k2P = 0 (8. H(2) m = Hm h(2) (2) m (zP ). θ ) = (c1 cos nθ + c2 sin nθ) (c3 jm(kR) + c4 ym(kR)) (c5 Pmn + c6 Qmn ) (8.39) ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ n Pmn d Pmn ⎪ ⎪ ⎪ ⎩ 0 ⎪ ⎭ sin φ dφ ⎧ ⎫ ⎪ ⎪ dh Pm h Sm ⎪ ⎪ ⎪ m(m + 1) 0 ⎪ ⎪ ⎪ ⎪ ⎨ dzP zS ⎪ ⎬ Hm = h Pm 1 d(zS h Sm) .34) χ (R. θ. jm(zP ) or ym(zP ). jm(zS ) or ym(zS ) (8. (8.40) ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ zP zS dzS ⎪ ⎪ ⎪ ⎩ 0 ⎪ 0 h Sm ⎭ zP = kP R. this is expressed as ∂ 2 ∂ 2 ∂ 2 + + + k2P = 0 (8. θ ) = (b1 cos nθ + b 2 sin nθ) (b 3 jm(kR) + b4 ym(kR)) (b5 Pmn + b6 Qmn ) (8. φ.43) 8.

49) 8. ky . we make use of some well-known vector operation identities.102 Solution to the Helmholtz and wave equations and substituting into the Helmholtz equation.48) in which the ai are arbitrary constants.55) To prove that this is indeed the solution. ∇ · ∇Ψ = ∇(∇ · Ψ) − ∇ × ∇ × Ψ = −∇ × ∇ × Ψ (8. Hence. so the solution can only contain two independent functions.. and considering only waves that propagate and/or decay in the positive x direction. only two of the three compo- nents of Ψ are independent. z.54) = ψ kˆ + 2 ∇ + ∇ × χ kˆ kS ∂z kS in which ψ. ∇ · ∇Ψ + kS2 Ψ = 0 (8.50) with parameter ks = ω/β.52) Hence ∇ × ∇ × Ψ = kS2 Ψ (8. namely ∇ × ∇ f = 0. kz. y. ω) = a1 eikx x + a2 e−ikx x a3 eiky y + a4 e−iky y a5 eikz z + a6 e−ikz z (8.53) This equation admits solutions of the form Ψ = Ψ1 + Ψ2  1 ∂ψ 1   (8. ω) = a1 eikz z + a2 e−ikz z e−ikx x . For two-dimensional problems in which the wave velocity α does not change in hori- zontal planes (i.e. In combination with the implied factor eiωt .47) The general solution to the scalar Helmholtz equation is then     (x. ∇ · ∇ × f = 0. for a laterally homogeneous medium).3 Vector Helmholtz equation in Cartesian coordinates Consider the vector Helmholtz equation ∇ 2 Ψ + kS2 Ψ = 0 (8. we find that it is satisfied if kx2 + k2y + kz2 = kP2 (8. More precisely. z. χ are solutions to the scalar Helmholtz equations ∇ 2 ψ + kS2 ψ = 0 and ∇ 2 χ + kS2 χ = 0 (8. this solution for represents compressional (P) waves that propagate in the (locally homogeneous) medium with velocity α. which are true for any scalar and vector . which means that it satisfies the gauge condition ∇ · Ψ = 0. kz = kP2 − kx2 (8. ˆ We shall assume Ψ to be solenoidal.51) which involves the vector Ψ = ψx ˆi + ψ y jˆ + ψzk. and for real wavenumbers kx . the solution simplifies to    (x. However.

z. Thus.8.63) χ(x. It remains to verify the gauge condition:      1 ∂ψ 1 ∇ · Ψ = ∇ · ψ kˆ + 2 ∇ · ∇ + ∇ · ∇ × χ kˆ kS ∂z kS   1 &   ' = ∇ · ψ kˆ + 2 ∇ · ∇ × ∇ × ψ kˆ + kˆ ∇ 2 ψ kS     = ∇ · ψ kˆ − ∇ · ψ kˆ = 0 (8.60) &  ' ∇ × kˆ ∇ 2 χ + kS2 χ = 0 (8. y. z. the second term gives  ∂χ k S ∇ × Ψ2 = ∇ − kˆ ∇ 2 χ (8.56) ∂y ∂x ∂ψ ˆ ∂ψ ˆ ∇ × ∇ × Ψ1 = ∇ × i− j ∂y ∂x  ∂ψ =∇ − kˆ ∇ 2 ψ (8.58) ∂z kS ∂z which results in the condition ∇ 2 ψ + kS2 ψ = 0.57) ∂z Since Ψ1 must satisfy ∇ × ∇ × Ψ1 = kS2 Ψ1 . Similarly. respectively. the solution provided does indeed satisfy the vector Helmholtz equation.59) ∂z so    ∂χ k S ∇ × ∇ × Ψ2 = ∇ × ∇ − ∇ × kˆ ∇ 2 χ = kS3 Ψ2 ∂z     = −∇ × kˆ ∇ 2 χ = kS2 ∇ × kˆ χ (8.61) which implies ∇ 2 χ + kS2 χ = 0. y. using for this purpose the expressions provided in the previous section:     ψ(x. ω) = b1 eikx x + b2 e−ikx x b3 eiky y + b4 e−iky y b5 eikz z + b6 e−ikz z  ik x      (8. the solution to the vector Helmholtz equation can now be formed with the solutions to the two scalar Helmholtz equations for ψ and χ . ω) = c1 e x + c2 e−ikx x c3 eiky y + c4 e−iky y c5 eikz z + c6 e−ikz z . then   ∂ψ 1 ∂ψ ∇ − kˆ ∇ 2 ψ = kS2 ψ kˆ + 2 ∇ (8.62) Hence.3 Vector Helmholtz equation in Cartesian coordinates 103 functions f and f. Consider the first term      1 ∂ψ ∇ × Ψ1 = ∇ × ψ kˆ + 2 ∇ × ∇ = ∇ × ψ kˆ kS ∂z ∂ψ ˆ ∂ψ ˆ = i− j (8.

ω) = b1 eikz z + b2 e−ikz z e−ikx x  ik z  −ikz z −ikx x kz = kS2 − kx2 (8. of which one depends on the other two (because of the condition ∇ · Ω = 0). Hence.66) Defining ε = ∇ · u = volumetric strain (8. the solution simplifies to   %  ψ(x. the wave equation can be written as (λ + 2µ) ∇ε − 2µ∇ × Ω = −ρω2 u (8. kz.65) χ(x. the solutions to these two equations are ε = . we obtain after straightforward algebra (λ + 2µ) ∇ 2 ε = ρ ε¨ and µ∇ 2 Ω = ρ Ω ¨ (8.75) .72) From the previous sections. In the 2-D case and considering only waves propagating and/or decaying in the positive x direction.4 Elastic wave equation in Cartesian coordinates Consider the elastic wave equation for an isotropic medium. α = (λ + 2µ)/ρ ⇒ P waves (8. ∇ 2 χ + kS2 χ = 0 (8. kP = ω/α. and for real wavenumbers kx . (λ + 2µ) ∇∇ · u − µ∇ × ∇ × u = ρ u¨ (8. z. ∇ 2 ψ + kS2 ψ = 0.74) kS ∂z kS Also. The latter consists of three scalar wave equations.67) Ω = 12 ∇ × u = rotation vector. these reduce to the two Helmholtz equations  ∇ 2 ε + kP2 ε = 0.71)  ∇ 2 Ω + kS2 Ω = 0. ky .104 Solution to the Helmholtz and wave equations in which the bi . these two solutions represent shear (S) waves that propagate in the medium with velocity β. z. ω) = c1 e + c2 e z e 8.73)    1 ∂ψ 1 2Ω ≡ Ψ = ψ kˆ + 2 ∇ + ∇ × χ kˆ (8. and kx2 + k 2y + kz2 = kS2 (8. β= µ/ρ ⇒ S waves (8.64) In combination with the implied factor eiωt . for harmonic motion. ∇ 2 + kP2 = 0.69) Applying in turn a gradient and then a curl to the above equation (but not both). for harmonic motion. ci are arbitrary constants.70) These constitute a scalar and a vector wave equation. kS = ω/β.68) we can write the elastic wave equation as (λ + 2µ) ∇ε − 2µ∇ × Ω = ρ u¨ (8. which satisfies ∇ · Ω = 0 (8.

77) kP ∂ x ∂y ∂z Because the solutions for . and the vertical wavenumbers are   kzα = kP2 − kx2 − k 2y and kzβ = kS2 − kx2 − k 2y (8. using rescaling factors kP for and kS for ψ and χ .1 Horizontally stratified media. Aψ are arbitrary integration constants. Aχ .4.81) in which the A .79) ±ikx x ±iky y ±ikzβ z χ =e e e Aχ (8. the displacement vector reduces to  1 ∂ 1 ∂ 2χ ˆ 1 ∂ψ ˆ 1 ∂ 1 ∂ 2χ u= + 2 i− j+ + 2 kS2 χ + 2 kˆ (8. plane strain We now restrict our analysis to plane-strain waves in a horizontally stratified medium defined in the two-dimensional x. χ .8. in which case the derivatives with respect to y must be discarded. Thus. these solutions can be written as = e±ikx x e±iky y e±ikzα z A (8. Hence.76) kS kS ∂z kS kP Carrying out the above differentiation operations. we obtain   1 1 ∂ 2χ ∂ψ ˆ 1 ∂ 2χ ∂ψ ˆ 1 ∂ 2χ ˆ u= 2 + i+ − j + kS χ + k kS kS ∂ x ∂z ∂y kS ∂ y ∂z ∂x kS ∂z2   1 ∂ ˆ ∂ ˆ ∂ ˆ − 2 i+ j+ k (8. we can write the displacement vector as 1 ∂ 1 ∂ 2χ 1 ∂ψ ˆ u= + 2 + i kP ∂ x kS ∂ x ∂z kS ∂ y 1 ∂ 1 ∂ 2χ 1 ∂ψ ˆ + + 2 − j kP ∂ y kS ∂ y ∂z kS ∂ x  1 ∂ 1 ∂ 2χ + + 2 kS2 χ + 2 kˆ (8.83) kP ∂ x kS ∂ x ∂z kS ∂ x kP ∂z kS ∂z . we can rescale these by any other constants. χ.4 Elastic wave equation in Cartesian coordinates 105 that is.78) kP ∂z kS ∂z in which the solutions for the scalar potentials . z space. ∇ × Ψ ∇ u= − 2 kS2 kP    1 1 ∂ψ 1 1 = 2 ∇ × ψ kˆ + 2 ∇ + ∇ × χ kˆ − 2 ∇ (8.80) ±ikx x ±iky y ±ikzβ z ψ=e e e Aψ (8.82) 8. In compact notation. and ψ contain arbitrary constants of integration. and ψ from the earlier sections must be used.

We .90) kS   2  k   k   uz = p A1 ekpz − A2 e−kpz + B1 eksz + B2 e−ksz e−ikx (8. we consider only waves that propagate or decay in the positive x direction.85)   −ksz −ikx ψ = C1 e + C2 e ksz e (8. and have identified this modification by adding an overbar to the displacement vector.91) kP kS Rescaling once more the (arbitrary) integration constants. s= 1− (8.88) kα kβ Introducing these potentials into the equation for the displacement vector. ω) = u ⎩ y ⎭ −i uz ⎧ ⎫⎧ ⎫ ⎪ ⎪ ekpz ⎪ ⎪ ⎪ ⎪ a1 ⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎧ ⎫⎪ ⎪ ⎪ e ksz ⎪ ⎪ ⎪ ⎪ ⎪ a2 ⎪⎪ ⎪ ⎨ 1 0 −s 1 0 s ⎬ ⎨ ksz ⎪ ⎬⎨ ⎪ ⎬ e a3 = 0 1 0 0 1 0 −kpz e−ikx ⎩ ⎭⎪ e ⎪ ⎪ a ⎪ −p 0 1 p 0 1 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 4 ⎪ ⎪ ⎪ ⎪ ⎪ e−ksz ⎪ ⎪ ⎪ ⎪ a5 ⎪⎪ ⎪ ⎩ ⎭⎪ ⎪ ⎩ ⎪ ⎭ e−ksz a6 (8. the term in ψ corresponds to horizontally polarized shear waves (SH waves). z. we can write this in matrix form as ⎧ ⎫ ⎨ ux ⎬ ¯ u(x.87) that is. From this point on.92) You will observe that we have added an imaginary factor in the vertical component.89) kP kS k   uy = i C1 eksz + C2 e−ksz e−ikx (8.84)   −ksz −ikx χ = B1 e + B2 e ksz e (8. in which case the potentials can be written in the form   = A1 ekpz + A2 e−kpz e−ikx . k ≡ kx (8. and drop for simplicity the sub-index for the horizontal wavenumber kx .86) in which we have defined   kp = i kzα = k2 − kP2 . ks = i kzβ = k2 − kS2 (8. !  ω 2  2 ω p= 1− .106 Solution to the Helmholtz and wave equations The terms in . χ are vertically polarized shear and pressure waves (SV-P waves). we obtain after brief algebra the solution to the vector wave equation in the frequency–horizontal- wavenumber domain for Cartesian coordinates as   2  k   k   ux = −i A1 ekpz + A2 e−kpz + s B1 eksz − B2 e−ksz e−ikx (8.

z. this greatly simplifies the transition between one and the other coordinate system. we use the material matrices Dzr . ω) e−ikx ⎧ ⎫ ⎨ u˜ x ⎬ u(k. z.95) The two boxes above provide the solution for horizontally stratified media in Cartesian coordinates. ω) = u˜ y = R1 E−1 z a1 + R2 Eza2 ⎩ ⎭ −i u˜ z = displcement in frequency–wavenumber domain a1 . carry out the required differentiations. introduce a convenience factor − i. ω) = σ˜ yz = 2kµ −Q1 E−1 z a1 + Q2 Eza2 ⎩ ⎭ −iσ˜ z ⎧ ⎫ ⎧ ⎫ ⎨ 2 p 0 −(1 + s 2 ) ⎬ ⎨ 2p 0 1 + s2 ⎬ Q1 = 0 s 0 . and obtain  T s¯ (x.94) ∂x ∂z To evaluate the stresses elicited by the displacement field previously found.4.93) Stresses in horizontal surfaces We provide also the solution for stresses in horizontal planes. R2 = 0 1 0 ⎩ ⎭ ⎩ ⎭ −p 0 1 p 0 1 (8. We can now write the displacement field compactly as ¯ u(x.1. z.8. expressed in the frequency–horizontal-wavenumber domain. for an isotropic medium.4 Elastic wave equation in Cartesian coordinates 107 have done this is for reasons of convenience: r To avoid imaginary factors in the matrix of coefficients. etc. as will be seen in Chapter 10. z. ω) = u(k. ω) = σxz σ yz −iσz = s e−ikx ⎧ ⎫ ⎨ σ˜ xz ⎬   s (k. . for these are necessary to enforce equilibrium across layer interfaces. Q2 = 0 s 0 ⎩ ⎭ ⎩ ⎭ −(1 + s 2 ) 0 2s 1 + s2 0 2s (8.. z. a2 = arbitrary 3 × 1 vectors of integration constants (wave amplitudes) " ±kpz # E±1 z = diag e e±ksz e±ksz ⎧ ⎫ ⎧ ⎫ ⎨ 1 0 −s ⎬ ⎨1 0 s⎬ R1 = 0 1 0 . From Section 1. these stresses are  T ∂u ∂u sz = σxz σ yz σz = Dzx + Dzz (8. r To obtain expressions that are virtually identical to those in cylindrical coordinates. r To obtain symmetric stiffness matrices for layered media.

c2 are constants that may depend on the parameter n. we use separation of variables. we obtain  1 ∂2 F 1 ∂F n2 1 ∂2G + + kP2 − 2 F = − (8. kz . this is expressed as ∂ 2 1 ∂ 1 ∂ 2 ∂ 2 + + 2 2 + + kP2 = 0 (8. the solution to these two equations is F(r ) = c3 Hn(1) (kα r ) + c4 Hn(2) (kα r ) Hankel functions (8.e.104) and G(z) = c5 ei kz z + c6 e−i kz z (8. n must be an integer.101) ∂r 2 r ∂r r ∂2G + kz2 G = 0 ⇒ 1-D Helmholtz equation (8. To satisfy continuity in the azimuth at the transition from 2π to 0. c4 .102) ∂z2  Defining kα = kP2 − kz2 .99) ∂θ 2 Substituting into the differential equation. This implies ∂ 2 = −n2 (8. c5 . and kP . non-negative number.100) F ∂r 2 r ∂r r G ∂z2 The left-hand side is a function only of r. c6 are constants that may depend on the parameters n. vertical) wavenumber. In cylindrical coordinates. To solve this problem..103) or F(r ) = c3 Jn (kα r ) + c4 Yn (kα r ) Bessel functions (8. It follows that in cylindrical coordinates. which we recognize later as the square of the axial (i.96) with parameter kP = ω/α. z) is a scalar function of the coordinates. each term must equal a constant kz2 .105) in which c3 .98) where c1 . particular solutions to the Helmholtz equation exist .5 Scalar Helmholtz equation in cylindrical coordinates Consider the Helmholtz equation involving a scalar function : ∇ 2 + kP2 = 0 (8.97) ∂r 2 r ∂r r ∂θ ∂z2 where (r. θ. z) = F(r ) (c1 cos nθ + c2 sin nθ) G(z) (8.108 Solution to the Helmholtz and wave equations 8. and taking the terms in G to the right-hand side. dividing by FG. the right-hand side is a function only of z. It follows that ∂2 F 1 ∂F n2 + + kP2 − kz2 − 2 F = 0 ⇒ Bessel equation (8. and begin by assuming a harmonic variation in the azimuth of the form (R. θ. Hence.

so the solution can only contain two independent functions. means that it satisfies the gauge condition ∇ · Ψ = 0. 8.8. z) = (c1 cos nθ + c2 sin nθ) (c3 Jn (kα r ) + c4 Yn (kα r )) c5 eikz z + c6 e−ikz z (8.6 Vector Helmholtz equation in cylindrical coordinates 109 that are the form   (r.107) with parameter kS = ω/β. only two of the three compo- nents of Ψ are independent. More precisely.106)  in which kα = kP2 − kz2 . χ are solutions to the scalar Helmholtz equations ∇ 2 ψ + kS2 ψ = 0 and ∇ 2 χ + kS2 χ = 0 (8. However. which which involves the vector Ψ = ψr rˆ + ψθ tˆ + ψzk. and kz is the axial wavenumber.114) ∂z .113) r ∂θ ∂r 1 ∂ψ ∂ψ ∇ × ∇ × Ψ1 = ∇ × rˆ − tˆ r ∂θ ∂r  ∂ψ =∇ − kˆ ∇ 2 ψ (8. θ. namely ∇ × ∇ f = 0. Consider the first term      1 ∂ψ ∇ × Ψ1 = ∇ × ψ kˆ + 2 ∇ × ∇ = ∇ × ψ kˆ kS ∂z 1 ∂ψ ∂ψ = rˆ − tˆ (8. Hence. we make use of some well-known vector operation identities.109) Hence ∇ × ∇ × Ψ = kS2 Ψ (8.111) = ψ kˆ + 2 ∇ + ∇ × χ kˆ kS ∂z kS in which ψ.110) This equation admits solutions of the form Ψ = Ψ1 + Ψ 2    1 ∂ψ 1 (8. respectively.6 Vector Helmholtz equation in cylindrical coordinates Consider the vector Helmholtz equation ∇ 2 Ψ + kS2 Ψ = 0 (8.112) To prove that this is indeed the solution. which are true for any scalar and vector functions f and f. ∇ · ∇ × f = 0. ∇ · ∇Ψ = ∇(∇ · Ψ) − ∇ × ∇ × Ψ = −∇ × ∇ × Ψ (8.108) ˆ We shall assume to be solenoidal. ∇ · ∇Ψ + kS2 Ψ = 0 (8.

126) .124) Ω= 1 2 ∇ × u = rotation vector.119) Hence. the solution to the vector Helmholtz equation can now be formed with the solutions to the two scalar Helmholtz equations for ψ and χ . then   ∂ψ 1 ∂ψ ∇ − kˆ ∇ 2 ψ = kS2 ψ kˆ + 2 ∇ (8. z) = (d1 cos nθ + d2 sin nθ) (d3 Jn (kβ r ) + d4 Yn (kβ r )) d5 eikz z + d6 e−ikz z (8.115) ∂z kS ∂z which results in the condition ∇ 2 ψ + kS2 ψ = 0.116) ∂z so    ∂χ kS ∇ × ∇ × Ψ2 = ∇ × ∇ − ∇ × kˆ ∇ 2 χ = kS3 Ψ2 ∂z     = −∇ × kˆ ∇ 2 χ = kS2 ∇ × kˆ χ (8.118) which implies ∇ χ + χ = 0. θ. and  kβ = kS2 − kz2 (8. the second term gives  ∂χ kS ∇ × Ψ2 = ∇ − kˆ ∇ 2 χ (8. Thus.123) Defining ε = ∇ · u = volumetric strain (8.117) &  ' ∇ × kˆ ∇ 2 χ + kS2 χ = 0 (8. using for this purpose the expressions provided in the previous section:   ψ(r.122) 8. z) = (c1 cos nθ + c2 sin nθ) (c3 Jn (kβ r ) + c4 Yn (kβ r )) c5 eikz z + c6 e−ikz z (8.121) in which the ci .120)   χ (r.110 Solution to the Helmholtz and wave equations and since Ψ1 must satisfy ∇ × ∇ × Ψ1 = kS2 Ψ1 . di are arbitrary constants. It remains to verify the gauge condition:      1 ∂ψ 1 ∇ · Ψ = ∇ · ψ kˆ + 2 ∇ · ∇ + ∇ · ∇ × χ kˆ kS ∂z kS   1 &   ' = ∇ · ψ kˆ + 2 ∇ · ∇ × ∇ × ψ kˆ + kˆ ∇ 2 ψ kS     = ∇ · ψ kˆ − ∇ · ψ kˆ = 0 (8.7 Elastic wave equation in cylindrical coordinates Consider the elastic wave equation for an isotropic medium (λ + 2µ) ∇∇ · u − µ∇ × ∇ × u = ρ u¨ (8.125) we can write the elastic wave equation as (λ + 2µ) ∇ε − 2µ∇ × Ω = ρ u¨ (8. θ. Similarly. which satisfies ∇ · Ω = 0 (8. the solution provided does indeed satisfy the vector 2 kS2 Helmholtz equation.

of which one depends on the other two (because of the condition ∇ · Ω = 0). we obtain after straightforward algebra (λ + 2µ) ∇ 2 ε = ρ ε¨ and µ∇ 2 Ω = ρ Ω ¨ (8. the solutions to these two equations are ε = . kP = ω/α.134) Since the solutions for . we can write the displacement vector as   1 ∂ 1 ∂ 2χ 1 ∂ψ u= + 2 + rˆ kα ∂r kβ ∂r ∂z kβ r ∂θ   1 ∂ 1 ∂ 2χ 1 ∂ψ + + 2 − tˆ kα r ∂θ kβ r ∂θ ∂z kβ ∂r    1 ∂ 1 ∂ 2χ + + 2 kS2 χ + 2 kˆ (8. we have   1 1 ∂ 2χ 1 ∂ψ 1 ∂ 2χ ∂ψ 1 ∂ 2χ ˆ 1 u= 2 + rˆ + − tˆ + kS χ + k − 2 ∇ kS kS ∂r ∂z r ∂θ kS r ∂θ ∂z ∂r kS ∂z2 kP (8.133) kS kS ∂z kS kP and using the previous developments in the solution of the vector Helmholtz equation.128)  ∇ 2 Ω + kS2 Ω = 0.7 Elastic wave equation in cylindrical coordinates 111 Applying in turn a gradient and then a curl to the above equation (but not both). β = µ/ρ ⇒ S waves (8. ∇ × Ψ ∇ u= − 2 kS2 kP    1 1 ∂ψ 1 1 = 2 ∇ × ψ kˆ + 2 ∇ + ∇ × χ kˆ − 2 ∇ (8. ∇ 2 χ + kS2 χ = 0 (8. for harmonic motion.135) kα ∂z kβ ∂z Observe that kS2 χ = − ∇ 2 χ . ∇ 2 ψ + kS2 ψ = 0. for harmonic motion. and ψ contain arbitrary constants of integration.132) that is. χ.130)    1 ∂ψ 1 2Ω ≡ Ψ = ψ kˆ + 2 ∇ + ∇ × χ kˆ (8. kS = ω/β. kS3 /kβ2 . and kS2 /kβ . these reduce to the two Helmholtz equations  ∇ 2 ε + kP2 ε = 0.8. we can rescale these by any other constant. α = (λ + 2µ)/ρ ⇒ P waves (8. which vanishes at zero frequency. the wave equation can be written as (λ + 2µ) ∇ε − 2µ∇ × Ω = −ρω2 u (8. . The latter consists of three scalar wave equations. applying rescaling factors kP2 /kα .127) These constitute a scalar and a vector wave equation. ∇ 2 + kP2 = 0.129) From the previous sections. Hence. Hence.131) kS ∂z kS Also.

7. 8.. we choose to express these indirectly in term of parameters p. p = 1 − (8. For convenience.142) αk and   2 ω ks = k2 − kS2 = i kzβ .140) that is   kzα = kP2 − k2 and kzβ = kS2 − k2 (8.136) sin nθ  cos nθ χ= Hβn e∓ikz β z (Aχ ) (8.138) cos nθ in which A . Hαn . the lower is for cases where this field is known to be antisymmetric with respect to this plane. in the second.. which requires kα = kβ ≡ k (8. In the first case. We consider each case in turn. kzα . i.e. compatibility and equilibrium of adjoining layers require the radial wavenumber to be common to all layers. and either the upper or lower trigonometric terms in parenthesis are used.1 Horizontally stratified media In this case. the solutions for the scalar potentials . because of the spatial symmetry of the loads).e. and ψ for waves that propagate in either the positive or negative z direction can be written in compact form as  cos nθ = Hαn e∓ikz α z (A ) (8. respectively. Aψ are (conveniently scaled) integration constants.141) where kzα . χ .112 Solution to the Helmholtz and wave equations Now. s= 1− (8. we must make a distinction between horizontally and cylindrically stratified media. The upper terms are called for in situations where the displacement field is known to be symmetric with respect to the vertical x. i. kzβ are the vertical wavenumbers for P and S waves. kzβ are the vertical wavenumbers for P and S waves. s as  !  ω 2 kp = k2 − kP2 = i kzα . kβ r .143) βk . z plane (say. the radial wavenumber k must be common to all layers. Aχ . Hβn is a short- hand for any of the Bessel functions of order n and argument kα r. from the earlier section on the scalar Helmholtz equation. At this point.139) It follows that   k = kP2 − kzα2 = kS2 − kzβ 2 (8. Non-symmetric fields can be obtained by linear combinations of these two cases. it is the vertical wavenumber that must be common.137) sin nθ  − sin nθ ψ= Hβn e∓ikz β z (−Aψ ) (8. which in this case differ from one another.

uR = Hn e A ∓ s e∓ksz Aχ + Hn e∓ksz Aψ (8.145) cos nθ kr  cos nθ "  # uz = Hn ∓ p e∓kpz A + e∓ksz Aχ (8.7 Elastic wave equation in cylindrical coordinates 113 Introducing the solutions for the scalar potentials . ω) = u˜ θ = R1 E−1 z a1 + R2 Eza2 ⎩ ⎭ u˜ z a1 . z) = [ ur uθ uz ]T = Tn Cn u ⎧ ⎫ ⎨ u˜ r ⎬ u(k. ψ into the solution for displace- ments and making use of the above parameters for the vertical wavenumbers. expressed in compact matrix form as u(r.148) ∂r r ∂θ ∂z r .147) Stresses in horizontal surfaces We provide also the solution for stresses in horizontal planes. From Section 1. ⎧ n ⎫ ⎪ ⎪  0⎪ ⎪ ⎨ Jn kr Jn ⎬ d Jn (kr ) Cn = n  ⎪ Jn Jn 0 ⎪ . R2 = 0 1 0 ⎩ ⎭ ⎩ ⎭ −p 0 1 p 0 1 (8. θ. these stresses are & 'T ∂u 1 ∂u ∂u u sz = σr z σθ z σz = Dzr + Dzθ + Dzz + Dz1 (8.146) sin nθ in which Hn is shorthand for any of the Bessel functions of order n and argument kr . and Hn = dHn /d(kr ). χ.4.8. after some algebra we find expressions of the form  cos nθ +   ∓kpz  n . Jn = d(kr ) ⎪ kr ⎩ ⎪ ⎭ 0 0 Jn ⎧ ⎫ ⎧ ⎫ ⎨ 1 0 −s ⎬ ⎨1 0 s⎬ R1 = 0 1 0 . we obtain the solution for displacements in hor- izontally stratified media for cylindrical coordinates in the frequency–radial-wavenumber domain.144) sin nθ kr    − sin nθ Hn  ∓kpz  uθ = n e A ∓ s e∓ksz Aχ + Hn e∓ksz Aψ (8. Combining the solutions for both negative and positive vertical wavenumbers and choosing Hn ≡ Jn (kr ). for these are necessary to enforce equilibrium across layer interfaces.2. a2 = arbitrary 3 × 1 vectors of integration constants (wave amplitudes)    cos nθ − sin nθ cos nθ Tn = diag (upper or lower elements used) sin nθ cos nθ sin nθ " # E±1 z = diag e ±kpz e±ksz e±ksz . z.

Q2 = 0 s 0 ⎩ ⎭ ⎩ ⎭ −(1 + s 2 ) 0 2s 1 + s2 0 2s (8. and ψ for waves that propagate in the positive z direc- tion (i.e.. Hence. we obtain    cos nθ  Hβn kz  uR = Hαn A + n Aψ + Hβn Aχ e−ikz z (8.. Introducing the solu- tions for the scalar potentials .149) Observe that u. for an isotropic medium. kβ r.151) sin nθ kβ r kβ    − sin nθ Hαn  kz Hβn uθ = n A + Hβn Aψ + n Aχ e−ikz z (8.153) sin nθ kα Combining all three into a matrix equation. changing Aχ into i Aχ (a trivial change in an arbitrary constant).150) d(kα r ) d(kβ r ) with Hαn . z. positive kz ) into the solution for displacements in the expression for u. ω) = σ˜ θ z = −Q1 E−1 z a1 + Q2 Eza2 ⎩ ⎭ σ˜ z ⎧ ⎫ ⎧ ⎫ ⎨ 2p 0 −(1 + s 2 ) ⎬ ⎨ 2p 0 1 + s2 ⎬ Q1 = 0 s 0 . the radial wavenumbers for P and S waves will be distinct.114 Solution to the Helmholtz and wave equations To evaluate the stresses elicited by the particular solution previously found. s do not depend explicitly on the azimuthal index n. and take into account the differential equation for the Bessel functions. χ . the vertical wavenumber kz must be common to P and S waves. and defining  dHαn  dHβn Hαn = Hβn = (8. we obtain .152) cos nθ kα r kβ kβ r    cos nθ kz uz = −i Hβn A + i Hβn Aχ e−ikz z (8.. carry out the required differentia- tions.7.2 Radially stratified media In this case. The final result is sz = kµ Tn Cn s ⎧ ⎫ ⎨ σ˜ r z ⎬ s(k. carrying out the differentiation operations. i. kα = kβ . Hβn being any of the Bessel functions of order n and argument kα r. we use the material matrices Dzr . 8. etc.e.

considering either the upper or lower trigonometric element. H(1) n = Hn Hn (kα r ).7 Elastic wave equation in cylindrical coordinates 115 ⎧ ⎫ ⎪ ⎪ cos nθ ⎪ ⎪ ⎪ ⎪ 0 0 ⎪ ⎪ ⎧ ⎫ ⎪ ⎪ sin nθ ⎪ ⎪ ⎨ uR ⎬ ⎪ ⎨  ⎪ ⎬ − sin nθ uθ = 0 0 ⎩ ⎭ ⎪ ⎪ cos nθ ⎪ uz ⎪ ⎪ ⎪ ⎪  ⎪ ⎪ ⎪ cos nθ ⎪⎪ ⎪ ⎪ ⎩ 0 0 ⎭ sin nθ ⎧ Hβn kz  ⎪ ⎫ ⎪ ⎪  Hαn n H ⎪ ⎪ ⎪ kβ βn ⎪ ⎪⎧ ⎫ ⎪ ⎪ kβ r ⎪ ⎪ ⎨ ⎬ ⎨ A ⎬ × Hαn  kz Hβn A e−ikz z (8. 2 d(kα r )   (i) dHβn (i) (i) Hβn = Hn(i) (kβ r ). ω) = u˜ θ n a1 + Hn a2 = H(1) (2) ⎩ ⎭ −i u˜ z a1 . −i uz = Tn u e uθ ⎧ ⎫ ⎨ u˜ r ⎬ u(r. Hn (kβ r ) . a2 = arbitrary vectors of integration constants ⎧ ⎫ ⎪ Hβn kz ⎪ ⎪ (Hαn ) ⎪  n (Hβn )  ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ kβ r kβ ⎪ ⎪ ⎨ H kz Hβn ⎬ αn  Hn = n (Hβn ) n . z) = ur u(r. kβ = kS2 − kz2 . we conclude that a solution to the elastic wave equation in cylindrical coordinates for a radially layered system is of the form (overbar is a reminder of imaginary factor. Hn (kβ r ) (1) (1) H(2) (2) (2)   kα = kP2 − kz2 . Hβn = d(kβ r )    cos nθ − sin nθ cos nθ Tn = diag (upper or lower elements used) sin nθ cos nθ sin nθ (8.8. n.154) ⎪ ⎪ n kα r Hβn n kβ kβ r ⎪ ⎪ ⎩ ψ⎭ ⎪ ⎪ ⎪ Aχ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ −i kz Hαn 0 i Hβn ⎭ kα Multiplying the third component by −i. i = 1.155) . tilde refers to frequency–wavenumber domain)  T −ikz z ¯ θ. kz = axial wavenumber   dHαn (i) (i) Hαn = Hn(i) (ka r ). kz. (i) Hαn = . ⎪ ⎪ k α r k β βk r ⎪ ⎪ ⎪ ⎪ kz ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ − Hαn 0 Hβn ⎪ ⎭ kα     n = Hn Hn (kα r ). and taking a linear combination of the first and second Hankel functions.

Yn (kβ r ) might be used). Hn (kβ r ) . Hβn (1) (1) (2) (2) that are shorthand for the Bessel functions Hn (ka r ).116 Solution to the Helmholtz and wave equations This is the solution for the displacements in a radially layered system in cylindrical coor- dinates. Jn (ka r ). The result is s¯ r = [ σr σr θ −iσr z ]T = Tn s e−ikz z ⎧ ⎫ σ˜ r ⎬ ⎨ s(r. Hn (kβ r ). or Hn (ka r ). take into account the differential equation for the Bessel functions.4.156) " # which are defined in terms of a generic matrix Fn = fi j with element functions Hαn . carry out the required differentiation. Stresses in cylindrical surfaces From Section 1. The elements of this matrix are $   2      %  kz Hαn n 2 f11 = −kα λ 1 + Hαn + 2µ + 1− Hαn kα kα r kα r  2nµ  Hβn f12 = Hβn − r kβ r $   2  %  Hβn n f13 = −2kzµ + 1− Hβn kβ r kβ r 2nµ  Hαn f21 = Hαn − r kα r . n. expressed in the frequency–vertical-wavenumber domain. which in this case is the radial–axial shearing stress. ω) = σ˜ r θ = F(1) n a1 + Fn a2 (2) ⎩ ⎭ −iσ˜ r z     (1) (1) (1) (2) (2) (2) Fn = Fn Hn (kα r ). the stresses in cylindrical surfaces are given by ∂u 1 ∂u ∂u u sr = [ σr σr θ σr z ]T = Drr + Dr θ + Dr z + Dr 1 ∂r r ∂θ ∂z r To evaluate the stresses elicited by the particular solution previously found. and re-apply a factor −i to the third component. we must first remove the imaginary factor in the vertical component of the displacement vector. Fn = Fn Hn (kα r ). u = diag{1 1 +i}u. Hn (kβ r ) (alternatively. kz. ¯ Thereafter.2. we use the material matrices for an isotropic medium. that is. Jn (kβ r ) and Yn (ka r ). Hn (kβ r ) ⎧ ⎫ ⎨ f11 f12 f13 ⎬ Fn = f21 f22 f23 ⎩ ⎭ f31 f32 f33 (8.

we use separation of variables. θ ) is a scalar function of the spherical coordinates.8 Scalar Helmholtz equation in spherical coordinates 117    2    Hβn n f22 = −kβ µ 2 + 1−2 Hβn kβ r kβ r 2n µ  Hβn f23 = kz Hβn − kβ r kβ r  f31 = −2 kz µ Hαn Hβn f32 = −kzn µ kβ r   2  kz  f33 = kβ µ 1 − Hβn kβ (8. the right-hand side is a function of φ only. n must be an integer. and θ is the azimuth or longitude.162) F ∂ R2 R∂R G ∂φ 2 ∂φ sin φ 2 The left-hand side is a function of R only. non-negative number. This implies ∂ 2 = −n2 (8. we obtain R2 ∂ 2 F 2 ∂F 1 ∂2G ∂G n2 G + + k2 F = − + cot φ − (8.8. φ. It follows that ∂2 F 2 ∂F m(m + 1) + + k2 − F =0 (spherical Bessel equation) (8. To satisfy continuity in the azimuth at the transition from 2π to 0.157) 8. dividing by FG. each side must equal a constant.159) ∂ R2 R∂R R ∂φ 2 R2 ∂φ R2 sin2 φ ∂θ 2 where (R. R is the radius.161) ∂θ 2 Substituting into the differential equation. and begin by assuming a harmonic variation in the azimuth of the form (R. multiplying by R2 . c2 are constants that may depend on the parameter n. θ ) = F(R) G(φ) (c1 cos nθ + c2 sin nθ) (8. and taking the terms in G to the right-hand side.8 Scalar Helmholtz equation in spherical coordinates Consider the Helmholtz equation involving a scalar function : ∇ 2 + k2 = 0 (8. Hence. which we choose to be the parameter m(m + 1).163) ∂R 2 R∂R R2 .160) where c1 . To solve this problem. φ.158) which in spherical coordinates is expressed as ∂ 2 2 ∂ 1 ∂ 2 cot φ ∂ 1 ∂ 2 + + 2 + + + k2 = 0 (8. φ the polar angle or co-latitude.

∇ · ∇Ψ = ∇(∇ · Ψ) − ∇ × ∇ × Ψ = −∇ × ∇ × Ψ (8.167) in which again c3 .172) .e. θ ) = (c1 cos nθ + c2 sin nθ) c3 hm (k R) + c4 hm (k R) (c5 Pmn + c6 Qmn ) (8. only two of the three compo- nents of Ψ are independent. so the solution can only contain two independent functions.166) and G(φ) = c5 Pmn + c6 Qmn (assoc.164) ∂φ 2 ∂φ sin2 φ The solution to these two equations is F(R) = c3 h(1) m (k R) + c4 hm (k R) (2) (spherical Hankel functions) (8. c6 will be zero. to satisfy the radiation condition at R = ∞. so for problems that include the full sphere. to avoid the singularity of the spherical Hankel functions (or yn ) at R = 0. π (i.169) or more precisely ∇ · ∇Ψ + k2 Ψ = 0 (8. We shall assume Ψ to be solenoidal. whereas media that include the origin may need to be formulated in terms of the conventional spherical Bessel functions jn .171) Hence ∇ × ∇ × Ψ = k2 Ψ (8.. However. Legendre functions) (8. because Qnm is singular at φ = 0. which means that it satisfies the gauge condition ∇ · Ψ = 0.118 Solution to the Helmholtz and wave equations ∂2G ∂G n2 + cot φ + m(m + 1) − G=0 (assoc. particular solutions to the Helmholtz equation exist that are the form   (1) (2) (R. it must be excluded on physical grounds. For infinite media. 8. at the north and south poles). c6 are constants that may depend on the parameters m. φ. c4 . c3 may also be zero. Legendre equation) (8. n It follows that in spherical coordinates.170) which involves the vector Ψ = ψr rˆ + ψφ sˆ + ψθ tˆ.168) In most cases. Hence. c5 .165) or F(R) = c3 jm(k R) + c4 ym(k R) (spherical Bessel functions) (8.9 Vector Helmholtz equation in spherical coordinates Consider the vector Helmholtz equation ∇ 2 Ψ + k2 Ψ = 0 (8.

which are true for any scalar and vector functions f and f.179)    ∇ × R rˆ ∇ 2 χ + k2 χ =0 (8. the solution provided does indeed satisfy the vector 2 2 Helmholtz equation.174) To prove that this is indeed the solution. Similarly. Consider the first term:  1 ∂ Rψ ∇ × Ψ1 = ∇ × (R ψ rˆ ) + 2 ∇ × ∇ = ∇ × (R ψ rˆ ) k ∂R 1 ∂ψ ∂ψ = sˆ − tˆ (8.181) .175) sin φ ∂θ ∂φ 1 ∂ψ ∂ψ ∇ × ∇ × Ψ1 = ∇ × sˆ − tˆ sin φ ∂θ ∂φ  ∂ Rψ =∇ − Rˆr ∇ 2 ψ (8. Thus.8. It remains to verify the gauge condition:  1 ∂ Rψ 1 ∇ · Ψ = ∇ · (R ψ rˆ ) + 2 ∇ · ∇ + ∇ · ∇ × (Rχ rˆ ) k ∂R k 1   = ∇ · (R ψ rˆ ) + ∇ · ∇ × ∇ × (R ψ rˆ ) + R rˆ ∇ 2 ψ k2 = ∇ · (R ψ rˆ ) − ∇ · (R ψ rˆ ) = 0 (8. the second term gives  ∂ Rχ k∇ × Ψ2 = ∇ − R rˆ ∇ 2 χ (8. then   ∂ Rψ 1 ∂ Rψ ∇ − R rˆ ∇ 2 ψ = k2 Rψ rˆ + 2 ∇ (8. respectively. we make use of some well-known vector operation identities.9 Vector Helmholtz equation in spherical coordinates 119 This equation admits a solution of the form Ψ = Ψ1 + Ψ 2  1 ∂ Rψ 1 (8.173) = R ψ rˆ + 2 ∇ + ∇ × (Rχ rˆ ) k ∂R k in which ∇ 2 ψ + k2 ψ = 0 and ∇ 2 χ + k2 χ = 0 (8. namely ∇ × ∇ f = 0.180) which implies ∇ χ + k χ = 0.177) ∂R k ∂R which results in the condition ∇ 2 ψ + k2 ψ = 0.176) ∂R and because Ψ1 must satisfy ∇ × ∇ × Ψ1 = k2 Ψ1 .178) ∂R so  ∂ Rχ   k∇ × ∇ × Ψ2 = ∇ × ∇ − ∇ × R rˆ ∇ 2 χ = k3 Ψ2 ∂R   = −∇ × R rˆ ∇ 2 χ = k2 ∇ × (R rˆ χ ) (8. ∇ · ∇ × f = 0.

190) From the previous sections. φ.191)  1 ∂ Rψ 1 2Ω ≡ Ψ = R ψ rˆ + 2 ∇ + ∇ × (Rχ rˆ ) (8.186) we can write the elastic wave equation as (λ + 2µ) ∇ε − 2µ∇ × Ω = ρ u¨ (8. α= (λ + 2µ)/ρ ⇒ P waves (8. of which one depends on the other two (because of the condition ∇ · Ω = 0). β = µ/ρ ⇒ S waves (8. these reduce to the two Helmholtz equations  ∇ 2 ε + kP2 ε = 0. φ. we obtain after straightforward algebra (λ + 2µ) ∇ 2 ε = ρ ε¨ and µ∇ 2 Ω = ρ Ω ¨ (8. kS = ω/β.189)  ∇ 2 Ω + kS2 Ω = 0. ∇ 2 ψ + k2 ψ = 0. which satisfies ∇ ·Ω = 0 (8.185) Ω = 12 ∇ × u = rotation vector.10 Elastic wave equation in spherical coordinates Consider the elastic wave equation for an isotropic medium (λ + 2µ) ∇∇ · u − µ∇ × ∇ × u = ρ u¨ (8. ∇ 2 + k2 = 0.120 Solution to the Helmholtz and wave equations Hence.188) These constitute a scalar and a vector wave equation. The latter consists of three scalar wave equations. for harmonic motion. θ ) = (d1 cos nθ + d2 sin nθ) d3 h(1) m (k R) + d4 h(2) m (k R) d5 Pmn + d6 Qmn (8.187) Applying in turn a gradient and then a curl to the above equation (but not both).183) 8. kP = ω/α. the solutions to these two equations are ε = .192) kS ∂R kS . the solution to the vector Helmholtz equation can now be formed with the solutions to the two scalar Helmholtz equations for ψ and χ . ∇ 2 χ + k2 χ = 0 (8. Hence. using for this purpose the expressions provided in the previous section:    ψ(R. θ ) = (c1 cos nθ + c2 sin nθ) c3 h(1) m (k R) + c4 hm (k R) c5 Pm + c6 Qm (2) n n (8.184) Defining ε = ∇ · u = volumetric strain (8.182)    χ (R.

kS2 into them.194) kS kS kP and using the developments in the previous section  1 1 ∂ 2 (Rχ ) 1 ∂ψ 1 ∂ 2 (Rχ) u = 2 kS Rχ + rˆ + + sˆ kS kS ∂ R 2 sin φ ∂θ kS R ∂ R ∂φ  1 ∂ 2 (Rχ ) ∂ψ 1 + − tˆ − 2 ∇ (8. we can incorporate the divisors kP2 . χ .196) kP R sin φ ∂θ kS R sin φ ∂ R ∂θ ∂φ Observe that kS2 Rχ = −R ∇ 2 χ .8. and write the displacement vector in matrix form as  ∂ 1 ∂ 2 (Rχ ) u= + + kS2 Rχ rˆ kP ∂ R kS ∂ R2 1 ∂ 1 ∂ 2 (Rχ ) 1 ∂ψ + + + sˆ kP R ∂φ kS R ∂ R ∂φ sin φ ∂θ 1 ∂ 1 ∂ 2 (Rχ) ∂ψ + + − tˆ (8. P-waves) eliciting particle motions that are normal to the P-wave front (the surface = constant.10 Elastic wave equation in spherical coordinates 121 Also. and ψ can be written in compact form as  cos nθ = Pmn hPm A (8.199) cos nθ .197) sin nθ  cos nθ χ = Pmn hSm Aχ (8. ∇ × Ψ ∇ u= − 2 kS2 kP 1 1 1 = 2 ∇ × (R ψ rˆ ) + ∇ × ∇ × (Rχ rˆ ) − 2 ∇ (8. the wave equation can be written as (λ + 2µ) ∇ε − 2µ∇ × Ω = −ρω2 u (8. the solutions for the scalar potentials . and two transverse solenoidal fields consisting of secondary waves (shear waves. whose normal is ∇ ).193) that is.198) sin nθ  − sin nθ ψ= Pmn hSm (−Aψ ) (8.195) kS R sin φ ∂ R ∂θ ∂φ kP This expression shows that the displacement field consists of a longitudinal irrotational field consisting of primary waves (pressure waves. χ. S-waves) eliciting particle motions that are contained in the tangent plane to the S-wave front (the surfaces ψ = constant and/or χ = constant). for harmonic motion. from the section on the scalar Helmholtz equation. Now. as the SV wave. Since the solutions for ψ. and contain arbitrary constants of integration. which vanishes at zero frequency. the shear motion component in vertical planes. The shear motion component in horizontal planes is referred to as the SH wave.

(hm (kS R)). Introducing these solutions into the expression for u. carrying out the differentiation operations. Aχ . (2) (2) but especially for the second spherical Hankel functions hm (kP R). Pmn (φ) are the associated Legendre func- tions of order m and rank n (the Qmn functions are excluded on physical grounds). Aψ are integration constants. we obtain    cos nθ dhPm m(m + 1) n u¯ R = Pmn A + Pm hSm Aχ (8.122 Solution to the Helmholtz and wave equations in which A . z plane (say.201) sin nθ dφ kP R dφ kS R d R sin φ    − sin nθ n n d(Rh Sm) d Pmn u¯ θ = Pmn hPm A + Pmn Aχ + hSm Aψ cos nθ kP R sin φ kS R sin φ dR dφ (8.204) ⎩ φ⎭ ⎪ ⎪ dφ kP R dφ kS R d R sin φ ⎪ ⎪ ⎩ ⎭ u¯ θ ⎪ ⎪ ⎪ Aψ ⎪ ⎪ ⎪ n n Pmn 1 d(RhSm) d Pmn ⎪ ⎪ ⎪ n Pm hPm ⎩ hSm ⎪ ⎭ sin φ kP R sin φ kS R d R dφ which can be factored into ⎧ n ⎫⎧ ⎫ ⎪ Pm 0 0 ⎪ ⎪ dhPm hSm ⎪ ⎧ ⎫ ⎪ ⎪ ⎪ ⎪ ⎪ m(m + 1) 0 ⎪ ⎪ ⎪ ⎪⎪ ⎪ ⎪ kP d R ⎪ kS R ⎪ ⎪ ⎪ ⎧ ⎫ ⎨ u¯ R ⎬ ⎨ d Pmn n Pmn ⎬⎨ ⎬ ⎨ A ⎬ = Tn 0 hPm 1 d(RhSm) u¯ ⎩ φ⎭ ⎪ dφ sin φ ⎪ ⎪ 0 ⎪ A ⎩ χ⎭ (8. The upper factors are called for in situations where the displacement field is known to be symmetric with respect to the vertical x. n.202) Defining also     cos nθ cos nθ − sin nθ Tn = diag (8.205) u¯ θ ⎪ ⎪ ⎪⎪ ⎪ ⎪ kP R kS R d R ⎪ ⎪ Aψ ⎪ ⎪ n Pmn d Pmn ⎪ ⎪ ⎪ ⎪ ⎩ 0 ⎭⎪⎩ ⎪ ⎭ sin φ dφ 0 0 hSm .200) sin nθ kP d R kS R    cos nθd Pmn hPm d Pmn 1 d(Rh Sm) n u¯ φ = A + Aχ + Pmn hSm Aψ (8. and adding an overbar to remind us that this is a particular solution for given indices m. hPm (hSm) is shorthand for any of the spherical Bessel functions of order m and argument kP R (kS R). Non- symmetric fields can be obtained by linear combinations of these two cases.203) sin nθ sin nθ cos nθ we obtain ⎧ ⎫ ⎪ dhPm hSm ⎪ ⎪ ⎪ Pmn m(m + 1)Pmn 0 ⎪ ⎪ ⎧ ⎫ ⎪ ⎪ kP d R kS R ⎪ ⎪ ⎧ ⎫ ⎪ ⎪ ⎪ ⎪ ⎨ u¯ R ⎬ ⎨ d Pn h d Pmn 1 d(RhSm) n Pmn ⎬ ⎨ A ⎬ m Pm u¯ = Tn hSm Aχ (8. because of the spatial symmetry of the loads) the lower ones are for cases where this field is known to be antisymmetric with respect to this plane. and either the upper or lower trigonometric factors in parenthesis are used.

and Pmn are the associated Legendre functions of order m and rank n.4. φ.3. Thus. ¯ u(R. θ.207) Introducing into this expression the material matrices for an isotropic medium given in Sec- tion 1. hPm = hm(zP ).10 Elastic wave equation in spherical coordinates 123 Hence. hm is any of the spherical Bessel functions. from Section 1.8 shows a list of the first few spheroidal (or co-latitude) matrices L.4. On the other hand. Table 10. hSm = hm(zS ). we obtain the stresses in spherical surfaces. zS = kS R.8.206) in which zP = kP R.3. ω) = Tn Lnm Hm a. a particular solution to the elastic wave equation in spherical coordinates is as given in the box below. It should be mentioned that these operations require tedious algebraic .   Tn = diag cos nθ cos nθ − sin nθ or   Tn = diag sin nθ sin nθ cos nθ ⎧ ⎫ ⎪ ⎪ Pn 0 0 ⎪ ⎪ ⎪ m ⎪ ⎪ ⎪ ⎪ ⎨ d Pmn n Pmn ⎪ ⎬ 0 Lnm = dφ sin φ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ n Pmn d Pmn ⎪ ⎪ ⎪ ⎩ 0 ⎪ ⎭ sin φ dφ ⎧ ⎫ ⎪ ⎪ dhPm hSm ⎪ ⎪ ⎪ ⎪ m(m + 1) 0 ⎪ ⎪ ⎪ ⎨ dzP zS ⎪ ⎬ Hm = h Pm 1 d(zS hSm) ⎪ ⎪ 0 ⎪ ⎪ ⎪ zP ⎪ zS dzS ⎪ ⎪ ⎪ ⎩ 0 ⎪ 0 hSm ⎭ (8. n≤m a = vector of integration constants. we have succeeded in separating the variations of the displacements in the radial direction from those in longitude and co-latitude. and carrying out the required differentiation operations. the stress in spherical surfaces associated with a displacement field u¯ is s¯ R = [ σ R σ R φ σ R θ ]T = LTR u¯ ∂ 1 ∂ 1 ∂ 1 cot φ = DR R + DRφ + DRθ + DR1 + DR2 Tn Lnm Hm a ∂R R∂ φ R sin φ ∂ θ r R (8.

m+1 − (m − 1) zS zS  2µ hPm f21 = −kP hP.m+1 − (m + 1) (m − 1) zS zS  2µ hSm f12 = −kS m (m + 1) hS.m+1 − (m − 1) zS (8.208) .m+1 + m (m − 1) zP zP  2µ hSm f22 = −kS µ hm − hS. The final result is  T s¯ R = σ R σRφ σRθ = Tn Lnm Fm a. ⎧ ⎫ ⎨ f11 f12 0 ⎬ Fm = f21 f22 0 ⎩ ⎭ 0 0 f33  2µ hPm f11 = −kP (λ + 2µ) hm − 2h P.m+1 − (m − 1) zP zP  hSm f33 = −kS µ hS.124 Solution to the Helmholtz and wave equations manipulations – including taking into account the differential equations for the spherical Bessel and associated Legendre functions.

In particular. the inversion must be carried out numerically. As we have seen in Section 1. the wave field) in space–time. 9.. an inverse integral transformation is applied. In addition. the method consists in carrying out an appropriate integral transform on the vector wave equation – including the source term – which changes the problem from a set of partial differential equations in the space-time domain to a system of coupled linear equations in the frequency–wavenumber domain. In a nutshell.. the method is exact. homogeneous. In most other (more complicated) cases. laterally unbounded system subjected to a source (or body load) b acting at some location.1) ∂ y ∂z ∂ x ∂z 125 . We provide a brief introduction to this method while picking up the fundamental tools needed for the powerful stiffness matrix method described in the next chapter.g. and layered media formulated in Cartesian. the system may be either finite or infinitely deep in the vertical direction. cylindrical. After solving the latter for the displacements.1 Cartesian coordinates Consider a horizontally layered. but only in the simplest of problems (e. it can be used to solve problems of sources in unbounded. and we illustrate these concepts by means of various examples.9 Integral transform method The integral transform method provides the most general framework for the analytical and numerical treatment of elastodynamic problems in unbounded continua. the wave equation in Cartesian coordinates for such a system is given by ∂ 2u ∂ 2u ∂ 2u ∂ 2u Dxx + D yy 2 + Dzz 2 + (Dxy + D yx ) ∂ x2 ∂y ∂z ∂x ∂y ∂ 2u ∂ 2u + (D yz + Dzy ) + (Dxz + Dzx ) + b = ρ u¨ (9. provided that the media exhibit specific geometric and material regularities. which returns the sought-after displacements (i.e. In principle.4.1. Pekeris’s or Chao’s problem) are the inverse transforms amenable to exact evaluation via contour integration. or spherical coordinates.

(9.5) ∂z ∂z in which ⎧ ⎫ ⎨ 0 λ + µ 0⎬ Bxy = λ + µ 0 0 . (9. y. we proceed to obtain the displacements in space–time from the inverse Fourier transformation  2  +∞  +∞ 1 u(x. y.6a) ⎩ ⎭ 0 0 0 ⎧ ⎫ ⎨ 0 0 −(λ + µ)⎬ Bxz = 0 0 0 . (9.3) 2π −∞ −∞ For reasons to be seen later. ky . we obtain the wave equation in the wavenumber domain: ∂ 2 u˜   ∂ u˜ Dzz − i kx (Dxz + Dzx ) + ky (D yz + Dzy ) ∂z2 ∂z   − kx2 Dxx + kx ky (Dxy + D yx ) + k2y D yy u˜ + b˜ = ρ u¨˜ (9. t)e−i(kx x+ky y) dkx dky u(k (9. t) and b˜ = b˜ (kx . it is convenient at this point to introduce the modified.2) in which u˜ = u˜ (kx . After solving this one-dimensional wave propagation problem in u(z. ky . Applying a Fourier transform in x. z. and kx . ky .4b) ⎪ ⎩ ⎪ ⎪ ⎭⎩ ˜ ⎭ ⎩ ˜ ⎪ ⎪ ⎪ ⎭ −i bz −i bz The Fourier-transformed wave equation is then   ∂u ∂ 2u kx2 Dxx + kx ky Bxy + k2y D yy u + (kx Bxz + ky B yz) − Dzz 2 + ρ u¨ = b (9. t) are the spatial Fourier transforms of u and b. z.6c) ⎩ ⎭ 0 λ+µ 0 . ˜ t) for a dense set of horizontal wavenumbers.4a) ⎩ ⎭⎩ y⎭ ⎩ ⎭ −i u˜ z −i u˜ z ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎪ ⎪ ˜ ˜ ⎨1 ⎬⎪⎨ bx ⎪⎬ ⎪⎨ bx ⎪ ⎬ b= 1 b˜ y = b˜ y (9.6b) ⎩ ⎭ λ+µ 0 0 ⎧ ⎫ ⎨0 0 0 ⎬ B yz = 0 0 −(λ + µ) (9.126 Integral transform method in which the Dαβ do not depend on x. transformed displacement and load vectors ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎨1 ⎬ ⎨ u˜ x ⎬ ⎨ u˜ x ⎬ u= 1 u˜ = u˜ y . y. t) = ˜ x . z. ky are the two horizontal wavenumbers. z.

ω) = R1 E−1 z a1 + R2 Eza2 .e. Additionally. namely..7) ∂z ∂z which involves only real coefficient matrices. s= 1− = 1− αk α βk β (k ≡ kx = horizontal wavenumber.10) ⎧ ⎫ ⎧ ⎫ ⎨ 1 0 −s ⎬ ⎨1 0 s⎬ R1 = 0 1 0 . no variation in the y direction).9) " kpz ksz ksz# Ez = diag e e e !  ω 2 !  c 2  ω 2  2 c p= 1− = 1− . a2 3 × 1 vectors of constants) (9. one for SV-P waves and one for SH waves. c = phase velocity) (9. ω) = u e−ikx = R1 E−1 z a1 + R2 Eza2 e (9..e. the Fourier- transformed wave equation reduces to ∂u ∂ 2u b − kx2 Dxx u − kx Bxz + Dzz 2 = ρ u¨ (9.8) For harmonic motions u¨ = −ω2 u and in the absence of sources (b = 0.1 Cartesian coordinates 127 In particular. (a1 . it will be found that this equation is exactly the same as in cylindrical coordinates. Their equations are         µ 0 ∂2 u˜ x 0 −(λ + µ) ∂ u˜ x − kx 0 λ + 2µ ∂z −iu˜ z 2 λ+µ 0 ∂z −iu˜ z      2   λ + 2µ 0 u˜ x ˜ bx ∂ u˜ x − kx2 + =ρ 2 0 µ −iu˜ z −ib˜ z ∂t −iu˜ z ∂ 2 u˜ y ∂ 2 u˜ y b˜ y + µ − µkx2 u˜ y = ρ 2 ∂z 2 ∂t (9. z. the transformed equation of motion admits solutions of the form u(k. and that it leads to symmetric stiffness matrices for layered systems.11) ⎩ ⎭ ⎩ ⎭ −p 0 1 p 0 1 which corresponds to displacements of the form (overbar a reminder of factor −i in uz )   −ikx ¯ u(x. for a problem in plane strain (i. z.12) This solution agrees perfectly with the solution obtained earlier by separation of variables via the solution to the Helmholtz equation. i. a homogeneous equation). . The above equation involves two uncoupled systems that should be solved separately.9. R2 = 0 1 0 (9.

j = 1.15) ∂z 2 ∂z *z=0 √ with β = µ/ρ the shear wave velocity. φ|z=h = 0 (9.13)  whose Fourier transform in x is simply b˜ y = δ(z − z ). source). B are constants. z. anti-plane line load Consider a homogeneous stratum of total thickness h whose lower and upper surfaces are free and fixed. we proceed to divide it by the shear modulus and consider first the homogeneous problem in which there is no load term (i. u˜ z. As we have already observed. we write its solution as u˜ y = φ(z). = 0. Source Example 9. respectively (Fig.16) in which A.14) ∂z where for notational simplicity we have written the horizontal wavenumber simply as kx ≡ k. Bxz is such that the equation for transverse displace- ment u˜ y is uncoupled from the in-plane degrees of freedom u˜ x . Dzz.1: Stratum subjected to SH line source.17) 2 . This is an eigenvalue prob- 2 2 2 2 lem in kβ whose solution is φ = A cos kβ z + B sin kβ z (9. To solve this equation together with the boundary conditions at the two external surfaces. ω) = δ(x) δ(z − z ) (9.1: Homogeneous stratum subjected to a harmonic. 3. 9. this implies a body load of the form by (x. . and kβ = ω /β − k . 2. we deduce π B=0 and cos kβ h = 0 ⇒ kβ h = (2 j − 1). the governing wave equation in the frequency–wavenumber domain is  2  ∂ 2 u˜ y µk − ρω2 u˜ y − µ 2 = δ(z − z ) (9.128 Integral transform method Receiver Figure 9. Hence. . From the boundary conditions. To distinguish this special case. the structure of the matrices Dxx . so the differential equation and its boundary conditions are * ∂ 2φ ∂φ ** + kβ2 φ = 0.. This (admittedly upside down) stratum is subjected to a harmonic anti-plane (SH) line load of frequency ω at x = 0 and some arbitrary elevation z . . Taking the origin of coordinates at the lower (free) surface.1).e. (9.

. .19) 2 ω1 in which ω1 = πβ/2h is the fundamental (or first cutoff) frequency of the stratum. Observe that for any given frequency ω.1 Cartesian coordinates 129 Writing kj for the values of k that satisfy the characteristic equation for kβ . which gives  ∞     h h µ Aj k2 − k2j φ j φk dz = φk δ(z − z ) dz = φk(z ) (9. We now return to our original problem. z. .9.21) j=1 ∂z which. after carrying out the differentiation and using the definition of the modal wavenumber kj . ω) = Aj φ j (z) (9.24) 0 2 . 2. j = 1. (9. Substituting this expansion into the original differential equation. the solution of the homogeneous equation is found to be (2 j − 1)π z φ j = Aj cos (9. there exist only a finite number of real eigenvalues kj that represent propagating modes. This ensures that the wave field satisfies the radiation and boundedness conditions at infinity (for positive x). and the φ j (z) are the normal propagation modes of the stratum in anti-plane shear with characteristic wavenumbers kj . .22) j=1 We next multiply both sides by another distinct mode φk(z) and integrate over the thick- ness of the stratum.20) j=1 with as yet undetermined participation factors Aj . . To solve the inhomogeneous wave propagation problem. we make use of modal superposition and express the displacement in terms of the normal modes:  ∞ u˜ y (k. of which we choose the one whose imaginary part is negative if kj is purely imaginary. At each frequency there exist two solutions.23) j=1 0 0 but from the orthogonality of the modes  h 1 φ j φk dz = hδ jk (9. while new modes emerge as ω rises above ω1 in ratios 1:3:5:7. we obtain  ∞   ∂ 2φ j Aj µ k2 − ρω2 φ j − µ 2 = δ(z − z ) (9. reduces to  ∞   µ Aj k2 − k2j φ j = δ(z − z ) (9. or whose real part is positive if kj is purely real. 3.18) 2h  2  2 ωh π(2 j − 1) kj h =± − β 2  2 π ω =± − (2 j − 1)2 . .

ω) = φ j (z ) φ j (z) dk (9. the above formula converges fast to the correct value. the transform is over the axial coordinate. in the neighborhood of the cutoff frequencies. 9. for any given frequency there exists only a finite number of modes with real wavenumbers. the spatial transform is in horizontal planes. the medium is assumed to be homogeneous and unbounded in the transformed direction. we first apply a Fourier transform in time.25) µh k2 − k2j µh k2 − k2j Hence.2 Cylindrical coordinates From Section 1. In the first case. stratum.e.2. 9. where there is a singularity.2.130 Integral transform method so that  2 φk(z ) 2 cos π (2 2h j−1)z Aj =   =   (9. if the medium has damping). evanescent) to real (i. and then apply a Hankel transform in r and a . In either case..4.g..28) r 2 ∂θ r ∂z r We now must distinguish between systems of flat layers (half-space. Hence.e. Again. which changes u¨ → −ω2 u.1 Horizontally stratified media To obtain the wave equation in the frequency–wavenumber domain. except of course. the solution to our original problem follows from the inverse Fourier transform  +∞ −ikx 2 ∞ 1 e u y (x.) and systems of cylindrical layers (e. the wave equation in cylindrical coordinates is ∂ 2u 1 ∂ 2u ∂ 2u 1 ∂ 2u b − ρ u¨ + Drr + Dθ θ 2 2 + Dzz 2 + (Dr θ + Dθr ) ∂r 2 r ∂θ ∂z r ∂r ∂θ ∂ 2u 1 ∂ 2u 1 ∂u + (Dr z + Dzr ) + (Dθz + Dzθ ) + Drr ∂r ∂z r ∂z∂θ r ∂r 1 ∂u 1 ∂u u + (Dθ1 − D1θ ) + (Dr z + Dz1 − D1z) − D11 2 = 0 (9..26) µh j=1 2π −∞ k2 − k2j Carrying out this transformation by means of contour integration. while in the latter. we obtain finally π(2 j−1)z 1 ∞ cos cos π (22h j−1)z uy = 2h e−ikj |x| (9. Observe also that at each cutoff frequency.27) iµ j=1 kj h This is the exact solution to this problem. etc. one more modal wavenumber changes from imaginary (i. the remainder are evanescent modes that quickly attenuate with distance from the source. at x = 0. which involve distinct integral transforms. a situation that represents a resonant condition. At this frequency. Hence. propagating). laminated cylinders). the last mode to become real contributes the most to the response because its denominator in the summation is small. the modal wavenumber passes through zero (or attains a small value.

t) = eiω t Tn k Cn u˜ n (k. (9. θ. z. This corresponds to the following operations (boxed equations are those commonly used in a solution): 1) Carry out a forward Fourier–Bessel transform on the source term. . z. z. θ. are symmetric with respect to the x axis.32) 0 0 −∞ In these expressions ⎧ ⎪ 1 ⎨ n=0 an = 2π (9. whereas a torsional load about a vertical axis calls for the lower set with n = 0. a point load in the x direction calls for the upper set with n = 1. whereas a load in the y direction involves the lower set with n = 1.34) sin nθ cos nθ sin nθ ⎧  ⎫ ⎪ ⎨ Jn n J kr n 0⎪⎬  Cn = kr Jn n Jn 0 (9. By contrast.29) 0 0 −∞ which admits the formal inversion  +∞ ∞  ∞ 1 b(r. and obtain the solution in the space– time domain (or the frequency domain. The upper one is used when the loads. t) = eiω t Tn k Cn b˜ n (k. the elements Jn = Jn (kr) of the matrix .  ∞  2π  +∞ b˜ n (k. θ. 3) Carry out an inverse Fourier–Bessel transform.31) 2π −∞ n=0 0 which again admits the formal inversion  ∞  2π  +∞ u˜ n (k.. t) e−iω t dt dθ dr (9. z. ω) dk dω (9.30) 2π −∞ n=0 0 2) Compute the solution in the frequency–wavenumber domain for the variation of dis- placements in z by solving the uncoupled equations for SV-P and SH waves.9. and therefore the displacements. a Fourier–Bessel transform). in which case the inverse Fourier transform over frequencies can be dispensed with):  +∞  ∞  ∞ 1 u(r. On the other hand. ω) = an r Cn Tn u(r. the lower one is used when loads and displacements are antisymmetric with respect to this axis. ω) = an r Cn Tn b(r.2 Cylindrical coordinates 131 Fourier transform in θ (i. non-symmetric loads would call for a combination of these sets together with a summation over many (or even infinitely many) n. z.35) ⎪ ⎩ ⎪ ⎭ 0 0 Jn The diagonal matrix Tn is formed with either the upper or the lower set of functions. More general. z. z. ω) dk dω (9. For example. a vertical load requires the upper set with n = 0. θ.e.33) ⎪ ⎩ 1 n = 0 π     cos nθ −sin nθ cos nθ Tn = diag . t) e−iω t dt dθ dr (9. z.

42) . and not simply r. t) = Tn Cn f(k.39) 0 Transformed wave equation Applying a spatial Fourier–Bessel transform to the wave equation.132 Integral transform method Cn are the cylindrical Bessel functions of the first kind and order n. the orthogonality of the Fourier series in θ must be used to complete the proof of inversion. Observe that the differentiation of the Bessel functions is carried out with respect to the product kr. θ. A brief summary of the fundamental properties of these functions is given in the Appendix. i. assum- ing that u and b are of the form u(r.40) in which f = [ f1 f2 f3 ]T .36) ⎪ ⎩ ⎪ ⎭ 0 ⎪ ⎩ ⎪ ⎪ ⎭⎩ ⎪ ⎭ F3 0 0 Jn f3 can be written in the fully diagonal form ⎧1 ⎫ ⎧ ⎫⎧ 1 ⎫ ⎨ 2 (F1 + F2 ) ⎪ ⎪ ⎬  ∞ ⎪ ⎨ Jn−1 0 0⎬ ⎪⎪⎨ 2 ( f1 + f2 ) ⎪ ⎬ 1 (F2 − F1 ) = r 0 Jn+1 0 1 ( f − f1 ) dr 2 2 (9. This involves the identity  2π Tn T j dθ = π δ(nj) (1 + δn0 δ j0 ) = δ(n) j /an . ⎧1 ⎫ ⎧ ⎫⎧ 1 ⎫ ⎨ 2 ( f1 + f2 ) ⎪ ⎪ ⎬  ∞ ⎪ ⎨ Jn−1 0 0⎪ ⎬⎪⎨ 2 (F1 + F2 ) ⎪ ⎬ 1 ( f − f1 ) = 2 2 k 0 Jn+1 0 1 (F2 − F1 ) dk (9. the forward transform ⎧ ⎫ ⎧  ⎫⎧ ⎫ ⎪ ⎨ F1 ⎪ ⎬  ∞ ⎪ ⎨ Jn n kr Jn 0⎪ ⎬⎪⎨ f1 ⎪ ⎬ F2 = r krn Jn Jn 0 f2 dr (9.38) ⎪ ⎩ ⎪ ⎭ 0 ⎪ ⎩ ⎪ ⎭⎪⎩ 2 ⎪ ⎭ f3 0 0 Jn F3 Also. θ. one obtains the three expressions (after lengthy and tedious algebra. n. The proof of the inversion is based on the fact that for arbitrary functions fj (r). or equivalently. n. t).37) ⎪ ⎩ 2 ⎪ ⎭ 0 ⎪ ⎩ ⎪ ⎭⎩⎪ ⎪ ⎭ F3 0 0 Jn f3 which is a self-reciprocating Hankel transform. z. t) (9. b(r. z. it may not be at all obvious that the set of integral transform equations given above for u and u˜ indeed constitute a direct and inverse Fourier–Bessel transform pair. (9.41) ⎪ ⎩ ∂  ∂  ⎪ ⎭ ∂z2 ∂z −k f1 + ∂z f3 u¨ = Tn Cn f¨ (9. t) = Tn Cn q(k. To the uninitiated reader.e. z. ∇ · ∇u = Tn Cn f − k2 f (9.. z. which requires also consideration of the differential equation for the Bessel functions) ⎧  ∂ ⎫ ⎨ −k k f1 − ∂z f3 ⎪ ⎪ ⎬  ∂2 ∇∇·u = Tn Cn 0 .

Plane layers in cylindrical vs. among other things. the same solution strategies used to find solutions to f(z. Cartesian coordinates The transformed equations in frequency–wavenumber space for horizontally homoge- neous systems were found to be exactly the same in cylindrical and in Cartesian coordinates.43) which implies ⎡ ⎤ ⎡ ⎤ ⎡  ⎤ ⎡ ⎤⎡ ⎤ q1 µ 0 0 f1 0 0 −(λ + µ) f1 ⎢ ⎥ ⎢ ⎥ ⎢  ⎥ ⎢ ⎥⎢ ⎥ ⎣ q2 ⎦ + ⎣ 0 µ 0 ⎦ ⎣ f2 ⎦ − k ⎣ 0 0 0 ⎦ ⎣ f2 ⎦ q3 0 λ + 2µ 0 f3 λ+µ 0 0 f3 ⎡ ⎤⎡ ⎤ ⎡ ⎤ λ + 2µ 0 0 f1 f¨1 2⎢ ⎥⎢ ⎥ ⎢ ⎥ −k ⎣ 0 µ 0 ⎦ ⎣ f2 ⎦ = ρ ⎣ f¨2 ⎦ (9. Writing these two equations separately. the wave equation (λ + µ)∇∇ · u + µ∇ · ∇u + b = ρ u¨ in cylindrical coordinates reduces to ⎧⎡ ⎤ ⎡  ⎤ ⎡ ⎤⎫ ∂ ⎪ ⎨ q1 −k k f1 − ∂z f3  2 2 f1 ⎪ ⎬ ⎢ ⎥ ⎢ ⎥ ∂ ∂ Tn Cn ⎣ q2 ⎦ + (λ + µ) ⎣ 0 ⎦+µ − k2 − ρ 2 ⎣ f2 ⎦ = 0 ⎪ ⎩ q   ∂z2 ∂t ⎪ 3 ∂ ∂ −k f1 + ∂z f3 f3 ⎭ ∂z (9. and so also are the solutions of the eigenvalue problems for guided waves in horizontally layered media. For example.44) 0 0 µ f3 f¨3 This equation in turn separates into a pair of uncoupled equations – a 2-vector equation for the SV-P degrees of freedom. Love waves elicited by anti-plane line loads have the same dispersion and wave propagation characteristics as torsional waves in that same medium.9. that the solution in the frequency–wavenumber domain is identical in both. This implies. and a scalar equation for the SH degree of freedom – but these are no longer plane waves.2 Cylindrical coordinates 133 Hence. t) in a Cartesian system can be used for a cylindrical system. Observe that they are identical to the equation of motion for a plane-strain system in Cartesian coordinates expressed in the horizontal wavenumber √ domain – provided the latter is modified by application of the factor −i = − −1 to the vertical components of displacements and loads.45) These are the equations of motion in cylindrical coordinates expressed in the radial– azimuthal wavenumber domain. as will be illustrated later on. Hence. we obtain 2 q1 µ 0 ∂ f1 0 −(λ + µ) ∂ f1 + −k q3 0 λ + 2µ ∂z 2 f3 λ+µ 0 ∂z f3 λ + 2µ 0 f1 ∂ 2 f1 − k2 =ρ 2 0 µ f3 ∂t f3  2 ∂ ∂ f2 2 q2 + µ − k2 f2 = ρ 2 ∂z2 ∂t (9. Furthermore. these wave propagation modes are independent of the azimuthal index n. except that the latter decay as they spread out – a result of the inverse .

Hence.46) 2π R2 When this body load is integrated over a volume enclosing the load. we see that the equation for tan- gential displacement u˜ θ is uncoupled from the in-plane degrees of freedom u˜ r . The Fourier–Bessel transform for this load reduces then to ⎧  ⎫⎧ ⎫  ∞ ⎪ ⎨ J0 ⎪ ⎬⎪⎨ 0 ⎪ ⎬ Mt   b˜ n = δ(z − z ) r J δ(r − R) dr 2πR 2 0 ⎪ ⎩ 0 ⎪ ⎭⎪⎩ ⎪ ⎭ J0 0 ⎧ ⎫ ⎧ ⎫ ⎨ 0 ⎪ ⎪ ⎬ ⎨0⎪ ⎪ ⎬ Mt M δ(z − z ) J0 (kR) = − J1 (kR) δ(z − z ) 1 t = (9. integration over θ simply produces a factor 2π that cancels out with the coef- ficient a0 in front of the transform integral. Example 9. Taking the origin of coordinates at the lower (free) surface.134 Integral transform method z Torsional Receiver Figure 9. Br z. This could be achieved by Fourier-transforming the 2-D plane-strain solutions (both for SV-P and SH line loads) into frequency–wavenumber space. we conclude that the transform involves only the n = 0 azimuthal term. 0}. Recognizing that this load is constant with θ and also antisymmetric with respect to the x axis. ω) = δ(z − z ) δ(r − R) (9.47) 2πR ⎪ ⎩ ⎪ ⎭ 2π R ⎩ ⎪ ⎪ ⎭ 0 0 Observing the structure of the matrices Drr . this implies a body load in the tangential direction of the form Mt bθ (r. u˜ z. it produces a net tor- sional moment Mt . so T0 = diag{0.48) ∂z 2π R . Dzz. This (inverted) stratum is subjected to a harmonic torsional ring load of radius R and frequency ω applied at some arbitrary elevation z . Hence. a property that is sometimes referred to as the inversion of the descent of dimensions. and that the lower component in the matrix Tn must be used. A corollary is that it is possible – at least in principle – to obtain the solution for point loads in cylindrical coordinates from the plane-strain solutions for line loads in that same layered medium.2: Stratum subjected to torsional ring source source. h z z’ x Fourier–Bessel transform – as will be shown in the example that follows. respectively.2: Homogeneous stratum subjected to a torsional ring load Consider a homogeneous stratum of total thickness h whose lower and upper surfaces are free and fixed. and inverting back into 3-D space via a Fourier–Bessel transformation. 1. the governing wave equation in the frequency–wavenumber domain is  2  ∂ 2 u˜ θ Mt µ k − ρω2 u˜ θ − µ 2 = − J1 (kR) δ(z − z ) (9.

the solution in the frequency–wavenumber domain. then Im(kj ) < 0 for any frequency.56) R→0 R 0 k2 − k2j 4i 1 Finally Mt  ∞ π(2 j − 1)z π (2 j − 1)z (2) uθ (x.9.54) π Rµh j=1 −∞ k2 − k2j Now. ω) = Aj φ j (z) (9. j = 1. ω) = φ j (z ) φ j (z) dk (9. or the real part is positive if purely real. assuming that the soil has a small amount of damping (which can be arbitrarily small). and so also are the boundary conditions. ω) = cos cos kj H1 (kj r ) (9. In particular. the integral divided by R converges to  1 ∞ k J1 (kR)J1 (kr ) π kj (2) lim dk = H (kj r ) (9. which allows evaluating the Hankel transform as (see Appendix)  ∞ $ π (2) k J1 (kR)J1 (kr ) H1 (kj R) J1 (kj r ) r ≤ R dk = 2i Im(kj ) < 0 (9.52) 2π R µh k2 − k2j where again the roots kj are chosen so that the imaginary part is negative.53) π Rµh j=1 0 k2 − k2j that is  Mt  ∞ +∞ k J1 (kR)J1 (kr ) uθ (x. which is based on the homogeneous solution for normal modes. . is identical: (2 j − 1)π z φ j = Aj cos (9. 3. the solution to the torsional ring load follows from the inverse Hankel transform  ∞ Mt  ∞ J1 (kR) J0 (kr ) uθ (x. 2.55) 0 k2 − kj 2 π (2) J1 (kj R) H (kj r ) r ≥ R 2i 1 with which we have obtained the complete solution.2 Cylindrical coordinates 135 Other than the scaling factor before δ(z − z ) on the right-hand side. this equation is identical to that in the example of an anti-plane line load given previously. (9. Hence. in the limit of a torsional point load R → 0.57) 4iµh j=1 2h 2h which is an exact solution for the wave field caused by a torsional point load applied in the interior of an elastic stratum. . .49) 2h  2  2 ωh π (2 j − 1) kj h = ± − β 2  2 π ω =± − (2 j − 1)2 . Hence. ω) = − φ j (z ) φ j (z) k dk (9.51) j=1  Mt 2 φk(z ) Aj = − J1 (kR)   (9.50) 2 ω1  ∞ u˜ θ (k. . z.

59) 2π −∞ n=0  2π  +∞ u˜ n = an Tn u(r. ω) = H(1) n a1 + Hn a2 (1) (a1 .e. θ. a2 : 3 × 1 vectors of arbitrary constants) (9. H(2) = (2) Hαn  (2)  kz Hβn . θ. n ⎪ ⎪ n Hβn n ⎪ n ⎪ n Hβn n ⎪ ⎪ ⎪ k αr kβ kβ r ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ k αr kβ kβ r ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪− kz H(1) ⎩ 0 H (1) ⎪ ⎭ ⎪− kz H(2) ⎩ 0 H (2) ⎪ ⎭ kα αn βn kα αn βn (9. θ. an = 2π (9. kz. ω) = Tn u˜ n e−ikz z dkz (9. we obtain the one-dimensional system of differential equations in the radial direction. ⎧ ⎡ ⎤ ⎡ ⎤ ⎨ λ + 2µ 0 0 λ + 2µ −n(λ + µ) 0 ∂2 1 ∂ b + ρω2 I + ⎣ 0 µ 0⎦ 2 + ⎣n(λ + µ) µ 0⎦ ⎩ ∂r r ∂r 0 0 µ 0 0 µ ⎡ ⎤ ⎡ ⎤ λ + 2µ + n µ 2 −n(λ + 3µ) 0 µ 0 0 ⎢ ⎥ 1 ⎢ ⎥ − ⎣ −n(λ + 3µ) µ + n2 (λ + 2µ) 0 ⎦ 2 −⎣0 µ 0 ⎦ kz2 r 0 0 nµ 2 0 0 λ + 2µ ⎡ ⎤ ⎡ ⎤ ⎫ kz ⎬ 0 0 1 0 0 0 ∂ +⎣ 0 0 0 ⎦ (λ + µ) kz − ⎣ 0 0 n ⎦ (λ + µ) u=0 (9.58) −∞ ⎪ ⎪ 1 0 ⎩ n = 0 π  +∞ $ ∞ % 1 b(r. and applying factors −i = − −1 to each of the vertical components so that u = [ u˜ r u˜ θ −i u˜ z ] and b = [ b˜ r b˜ θ −i b˜ z ]. Applying the stated Fourier transforms√ in θ. z. and t to the wave equation.60) 0 −∞  $ % 1 +∞ ∞ u(r.61) 2π −∞ n=0 in which again the boxed equations are those commonly used. z.136 Integral transform method 9.62) ∂r r ⎭ −1 0 0 1 n 0 In the absence of body loads (i. z. ω) e−ikz zdz dθ (9. ω) e −ikz z dz dθ . this equation can be shown to admit solutions of the form u(r.64) . z. θ. collecting terms.2.63) ⎧ (1) ⎫ ⎧ (2) ⎫ ⎪ ⎪  (1)  Hβn kz  (1)  ⎪ ⎪ ⎪ ⎪  (2)  Hβn kz  (2)  ⎪ ⎪ ⎪ ⎪ Hαn n Hβn ⎪ ⎪ ⎪ ⎪ Hαn n Hβn ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ kβ r kβ ⎪ ⎪ kβ r kβ ⎪ (1) ⎬ ⎨ (2) ⎬ H(1) = (1) Hαn  (1)  kz Hβn . b = 0).. z. but now we perform instead a Fourier transform in the axial and azimuthal directions: ⎧  2π  +∞ ⎪ ⎪ 1 ⎨ n=0 ˜bn = an Tn b(r.2 Cylindrically stratified media The starting point in this case is again the wave equation in cylindrical coordinates. ω) = Tn b˜ n e−ikz z dkz (9.

Hαn are the Hankel functions of the first and second kind of order n.7. we postpone examples of application until the next chapter on the stiffness matrix method in Section 10. when internal stresses in the transformed frequency–wavenumber space are com- puted. z. kβ = kS2 − kz2 . n a1 + Hn a2 e (2) (9.3 Spherical coordinates 137   kα = kP2 − kz2 . In the interest of brevity and to avoid repetition. Hαn = (9. Hβn = (9. The above solution corresponds to displacements of the form   −ikz z ¯ θ.9. kz = axial wavenumber (9.67) d(kβ r ) d(kβ r ) (1) (2) in which Hαn . except that these are not convenient because they become quasi-linearly dependent when the argument is complex. Hβn = Hn(2) (kβ r ).65)   (1)  dHαn (1)   (2)  dHαn (2) (1) Hαn = Hn(1) (kα r ). Hβn = . (2) Hαn = Hn(2) (kα r ). the Hankel functions can also be changed into Bessel functions.3. so they need not be repeated. by appropriate choices for the constants a1 . β indicate the argument type.2. Hαn = . Thus.3 Spherical coordinates From Section 1. The added sub-indices α.3. the wave equation for isotropic media in spherical coordinates is ⎧⎡ ⎤ ⎨ λ + 2µ · ·  ∂ 2 2 ∂ 2 ρ u¨ = b + ⎣ · µ ·⎦ + − 2 ⎩ ∂R 2 R∂R R · · µ ⎡ ⎤ µ · ·  1 ∂2 cot φ ∂ + ⎣ · λ + 2µ ·⎦ + R2 ∂φ 2 R2 ∂φ · · µ ⎡ ⎤ ⎡ ⎤ µ · · 0 λ+µ · ⎣ ⎦ 1 ∂2 1 ∂2 + · µ · + ⎣λ + µ 0 ·⎦ ∂θ R ∂ R ∂φ · · λ + 2µ R sin φ 2 2 2 · · 0 ⎡ ⎤ ⎡ ⎤ 0 · λ+µ 0 · · 1 ∂2 1 ∂2 +⎣ · 0 · ⎦ +⎣· 0 λ + µ⎦ 2 R sin φ ∂ R ∂θ R sin φ ∂φ ∂θ λ+µ · 0 · λ+µ 0 ⎡ ⎤ ⎡ ⎤ 0 λ+µ · 0 −(λ + 3µ) · cot φ ∂ 1 ∂ +⎣· 0 · ⎦ + ⎣ 2(λ + 2µ) 0 ·⎦ 2 R ∂R R ∂φ · · 0 · · 0 ( f or mula continued on next page) . and especially imaginary. a2 . ω) = Tn u e−ikz z = Tn H(1) u(r.68) This solution agrees with the results obtained earlier by different means in Section 8. Clearly. it will be found that they too coincide with those in Section 8.7.2. 9.66) d(kα r ) d(kα r ) (1) (2) (1) (2)  (1)  dHβn  (2)  dHβn Hβn = Hn(1) (kβ r ).4.

φ.72) m=0 n=0 which again admits the formal inversion  π  2π u˜ mn (R. θ. and ω.74) 0 .138 Integral transform method (continued) ⎡ ⎤ 0 · −(λ + 3µ) 1 ∂ +⎣ · 0 · ⎦ R2 sin φ ∂θ 2(λ + 2µ) · 0 ⎡ ⎤ ⎡ ⎤ 0 · · 0 −(λ + 3µ) · cot φ ∂ cot φ +⎣· 0 −(λ + 3µ) ⎦ 2 +⎣· 0 ·⎦ R sin φ ∂θ R2 · λ + 3µ 0 · · 0 ⎡ ⎤ ⎡ ⎤ ⎫ · · · · cot2 φ ⎬ 0 0 1 + ⎣ · −λ · ⎦ + ⎣ · −2µ · ⎦ u (9.71) m=0 n=0 2) For each m.73) 0 0 The proof of these inversion formulas is obtained by considering the orthogonality prop- erties of Fourier series in the azimuth and of the spheroidal matrix in co-latitude. the solution of this equation follows along the following formal steps (the matrices involved are defined in Section 8. ω) = Tn Lnmu˜ mn (9. Thus. 3) Obtain the solution in the space domain from the series solution  ∞  m u(R.10. obtain the solution u˜ mn (R) in the frequency–wavenumber domain by solving the one-dimensional system of equations in the radial coordinate R. φ. and tabulated in Section 10.70) 0 0 which admits the formal inversion  ∞  m b= Tn Lnmb˜ mn (9. instead of attempting to re-derive the relevant equations in the transformed domain for this case. θ. we shall simply rely on the results already obtained in Section 8. ω) dθ dφ (9. we employ series solutions in place of the integrals of the previous sections. However. namely. θ. except that because each spherical surface is finite in size. n.4): 1) Express the source in terms of spherical harmonics:  π  2π b˜ mn (R. ω) dθ dφ (9.69) R2 ⎭ µ R sin φ 2 2 · · · · −2µ This rather complicated looking system of differential equations can also be solved by means of transforms.10. φ. ω) = J−1 sin φ Lnm Tn b(R. ω) = J−1 sin φ Lmn Tn u(R.  2π Tn T j dθ = π δ(nj) (1 + δn0 δ j0 ) (9.

see Section 10.76) Defining the diagonal matrix ⎧ ⎫ π (1 + δn0 ) (n + m)! ⎨ ⎬ 1 0 0 J=   0 m(m + 1) 0 (9.3 Spherical coordinates 139  π (n + m)! " # Lnm Lnk sin φ dφ = δ(mk)   diag 1 m(m + 1) m(m + 1) (9. . For examples of application.9.4.75) 0 m+ 1 2 (m − n)! Hence  π  2π  m  ∞  π  2π  sin φ Lnk T j u dθ dφ = sin φ Lnk T j Tn dθ Lnm dφ u˜ mn 0 0 m=0 n=0 0 0 ∞   π = π(1 + δn0 ) sin φ Lnk Lnm dφ u˜ mn m=0 0 π(1 + δn0 ) (n + m)! =   diag{1 m(m + 1)m(m + 1)}u˜ mn m + 12 (m − n)! (9.77) m + 2 (m − n)! ⎩ 0 1 0 m(m + 1) ⎭ we obtain the result given above for u˜ mn .

In these situations. which has the advantages over the propagator matrix method that are listed below and. The dispersion of surface waves on multilayered media. Indeed. readers familiar with the propagator method are strongly encouraged to familiarize themselves with and switch to this superior method. r Stiffness matrices involve half as many degrees of freedom as propagator matrices.10 Stiffness matrix method for layered media Closed-form solutions for layered media. A. 1950. Transmission of elastic waves through a stratified soil medium. on the other hand. and a factor of more than 4 on account of bandwidth. contain terms of exponential growth that require special treatment. Thus. source problems. 140 . 17–34. N. 1964. r Stiffness matrices remain robust and stable for thick layers and/or high frequencies. Among the advantages of the stiffness matrix method are: r Stiffness matrices are symmetric. layer interfaces decouple naturally as a result of coupling terms tending to zero. do not exist. W. Propagator matrices. r Stiffness matrices lead naturally to the solutions for normal modes (eigenvalue prob- lems). Bulletin of the Seismological Society of America. so their bandwidth is only half as large. Among these.. 98–93. which gives a total reduction of more than 8. or for homogeneous plates and strata with arbi- trary boundary conditions. a widely used scheme is the propagator matrix or transfer matrix method of Haskell1 and Thomson. at least in our judgment. r On account of the two previous items. and wave amplification problems. T. while propagator matrices are not.. we choose instead to present herein the related stiffness or impedance matrix method. the stiffness matrix method is nearly an order of magnitude faster than the propagator matrix method: The computational effort is smaller by a factor 2 due to symmetry. Thus. 43. even the free–free plate – the so-called Mindlin plate – is ultimately intractable by purely analytical means. pp. all without the need for 1 Haskell. no disadvantages in comparison with the latter. Vol. Journal of Applied Physics. 21. Since the computations must be repeated for each frequency and each wavenumber. pp. whereas the state vector in propagator matrices contains both stresses and displacements. The former involve only displacements. 2 Thomson. the savings are considerable.2 Nonetheless. Vol. such problems must be solved with the aid of numerical tools.

¨ J. in which K is a symmetric stiffness matrix. and spherical coor- dinates. M. Vol. solid or hollow. 1981.1 Summary of method 141 special manipulations or treatment for each class of problem. In principle. r For each frequency and wavenumber. for various sources. That such computation is fraught with difficulties and may thus introduce substantial errors if not carried out properly goes without saying. pp.10. The reader is referred to the literature for further details on the thin-layer method. the resulting expressions are generally intractable by purely analytical means and must ultimately be evaluated numerically. by appropriate superposition. Journal of Mathematics and Mechanics. Stiffness matrices for layered soils. cylindrical. especially in the dynamics of laminated plates and layered soils. However.3 the stiffness matrix method has found wide applications in elastodynamics. Bulletin of the Seismological Society of America. M. r Layered spheres of finite or infinite radius. . which are modeled as external tractions. and with virtually no added effort. 4 Biot. pp. These discrete matrices allow finding the normal modes – both propagating and evanescent modes – from the solution of a standard eigenvalue problem.. r Stiffness matrices lend themselves to the application of substructuring techniques. we shall show that sources and displacements in a layered medium are related by an equation of the form p = Ku. and consists of the following steps: r Transform the source(s) (if any). The method can be applied to a large class of continua. usually in closed form. and Roesset. 1743–1761. 1963. normal mode problems. but we restrict our presentation to the following three problems involving isotropic media formulated in Cartesian. and thus avoid the use of search techniques. Continuum dynamics of elastic plates and multilayered solids. 3 Kausel. Indeed.. it is an exact method in the sense that one obtains mathematical expressions for displacements that are free from approximations or dis- cretization errors. E. 12. This produces a source vector p. r Infinitely long layered cylinders of finite or infinite radius. A closely related. but a detailed discussion of these is beyond the scope of this book. 793–810. one can readily solve simultaneously. r Discrete versions of the stiffness matrices based on the finite element method – the so-called thin-layer method – can readily be obtained. but not identical. determine the stiffness matrix of each layer and. the stiffness matrix K of the complete layered system. cylindrical. such as beams and plates.4 In the follow- ing. method was also proposed much earlier by M. Since its inception in 1981. solid or hollow. from the space–time domain into the frequency–wavenumber domain. we present the stiffness matrix method for Cartesian. and spherical coordinates: r Laterally infinite. horizontally layered strata and/or half-spaces (plane strain). 10. Biot.1 Summary of method The stiffness or impedance matrix method is a tool for the analysis of wave propagation problems in elastic media. with the advantage that the matrix formalism remains exactly the same. namely source problems. In a nutshell. Vol. The method is based on the use of integral transforms. and wave amplification problems. 71.

k is (see Section 9.3) 2π −∞ n=0 r Spherical coordinates. ω) = Tn k Cn u˜ n (k. and obtain the displace- ments in the frequency–wavenumber domain.e. the requisite inverse integral transforms needed to obtain the displacement vector u in the spatial domain are of the following form: r Cartesian coordinates. ω) = Tn u˜ n (r. ω) dk (10. Damping is incorporated via complex moduli. r Carry out an inverse transform into the spatial–temporal domain. and in general its elements are complex. ω) = ˜ u(k. b = 0). φ. cylindrical layers:  +∞ $ ∞ % 1 u(r.. we obtain  2  k Dxx − n2 Dzz + n kBxz − ρω2 I v = 0 (10. ω) = Tn Lnmu˜ mn (R. z. ω) e−i kx dk (10. Nontrivial solutions exist if the determinant of the 3 × 3 matrix in parenthesis vanishes. θ.1) ∂u ∂ 2u k2 Dxx u + k Bxz − Dzz 2 − ρω2 u = 0 (10. θ. Omitting the inverse Fourier transform over frequencies.2 Stiffness matrix method in Cartesian coordinates Consider a homogeneous medium subjected to waves in the x. horizontal layers:  ∞  ∞ u(r. the elastic wave equation in the transformed frequency– wavenumber space ω.5) ∂z ∂z where for simplicity we have written the horizontal wavenumber as kx ≡ k. which for an isotropic medium reduces to a quadratic and a biquadratic equation for SH and SV-P . z. which yields the desired response.e.6) This equation is an eigenvalue problem in n for plane waves with vertical wavenumber kz = ±i n and eigenvectors v. z) = v(k..e.4) m=0 n=0 Details and examples are given in the pages that follow. ω. z. z plane.1) 2π −∞ r Cylindrical coordinates. narrowly banded) and symmetric. 10. In the absence of external sources (i. θ. Making an ansatz u(k. ω) e−ikz z dkz (10. spherical layers:  ∞  m u(R.142 Stiffness matrix method for layered media This matrix is block-tridiagonal (i. z. kz. horizontal layers:  +∞ 1 u(x. r Solve the system of equations p = Ku by standard methods. a condition that leads to a sixth order equation. ω) (10. SV-P and SH waves in plane strain. z.2) n=0 0 r Cylindrical coordinates. ω) enz and substituting it into the above expression. i..

Observe that SH waves are uncoupled from SV-P waves. a  u= uy = R1 E−1 R2 E z 1 (10. α= = P-wave velocity kα ρ  2 ! ω µ s= 1− .7) ⎩ ⎭ z a2 −i uz with ⎧ ⎫ ⎧ ⎫ ⎨ 1 0 −s ⎬ ⎨1 0 s⎬ R1 = 0 1 0 .9) p= 1− . Depending on the values of the wavenumber k and frequency ω (i. and !  ω 2 λ + 2µ (10. The terms in a1 are plane waves that propagate or decay in the positive z direction (i.a  s= σ yz = kµ −Q1 E−1 Q2 E z 1 (10.. or guided waves that evanesce in the vertical z direction.10. elastic layer of arbitrary thickness h as a free body in space. these solutions represent either plane body waves.12) ⎩ ⎭ ⎩ ⎭ −(1 + s 2 ) 0 2s 1 + s2 0 2s Consider next a homogeneous. the resulting displacement field can be shown to be given by ⎧ ⎫ ⎨ ux ⎬ + . and the six columns of R1 R2 are the eigenvectors. and assume that it is subjected to the spatially and temporally harmonic wave field determined previously.10) kβ ρ Observe that n1.4 = ±kp. horizontal.. we apply appropriate . The stresses in horizontal planes associated with the solution found above – after we modify these by an imaginary unit factor applied to the vertical component – can be shown to be given by ⎧ ⎫ ⎨ σxz ⎬ + . upwards).6 = ±ks are the six roots of the eigenvalue problem in n. the second and fifth are SH waves. Q2 = 0 s 0 (10. downwards). a2 are vectors of arbitrary constants (wave amplitudes). β= = S-wave velocity (10. n#3. whereas those in a2 propagate or decay in the negative z direction (i.2 Stiffness matrix method in Cartesian coordinates 143 waves. n2. To maintain dynamic equilibrium.. respectively.8) ⎩ ⎭ ⎩ ⎭ −p 0 1 p 0 1 + .e. Solving these equations and finding the eigenvectors v. The first and fourth columns in the solution for u above are P waves.11) ⎩ ⎭ z a2 −i σzz with ⎧ ⎫ ⎧ ⎫ ⎨ 2p 0 −(1 + s 2 ) ⎬ ⎨ 2p 0 1 + s2 ⎬ Q1 = 0 s 0 .e. R2 = 0 1 0 (10. Ez = diag ekpz eksz eksz Here a1 . and the third and sixth are SV waves.5"= ±ks.e. on real or imaginary values for p. s).

it is possible to separate the above equation into two separate matrix equations: a 4 × 4 for SV-P waves. the combined traction vectors are     $ %  p1 s1 −Q1 E−1 h/2 Q2 Eh/2 a1 = = kµ p2 −s2 Q1 E−1 −h/2 −Q2 E−h/2 a2 $ %  −Q1 E−1 Q2 Eh/2 a1 = kµ h/2 (10. Some authors refer to it also as the spectral element matrix. u2 are the displacements at the two interfaces. This leads to the impedance matrices for .3. Although somewhat tedious.a  + . and a 2 × 2 for SH waves. by discarding the waves that propagate or decay upwards. which can be shown to be symmetric. The matrix   K11 K12 K= (10.6). the amplitude vector is   $ %−1   a1 R1 E−1 R2 Eh/2 u1 = h/2 −1 (10. that is. respectively.15) a2 R1 Eh/2 R2 Eh/2 u2 Also. We also choose the origin of coordinates at the center of the layer and temporarily assign the suffixes 1 and 2.19) K21 K22 is the stiffness or impedance matrix of the layer.. p2 = −s2 at the upper and lower external interfaces. Because the SH components (second and fifth degree of freedom) are uncoupled from the SV-P degrees of freedom (1.18) p2 K21 K22 u2 which relates the two interface tractions with the observed interface displacements. which completely balance the internal stresses at these two locations.4. h → ∞) can also be considered by setting a1 = 0.144 Stiffness matrix method for layered media tractions p1 = s1 .a  u1 = R1 E−1 h/2 R2 Eh/2 1 . we have + . if u1 . respectively.e.1 separately for SV-P and SH waves. Hence. We give these in Table 10.16) Q1 Eh/2 −Q2 E−1 h/2 a2 Eliminating the amplitude vector between these two equations. the previous matrix operations can be carried out by hand and closed-form expressions obtained for the elements of K.14) u2 −1 R1 E−h/2 R2 E−h/2 a2 R1 Eh/2 R2 E−1 h/2 a2 Thus. Elastic half-spaces (i. we obtain   $ %$ %−1   p1 −Q1 E−1 Q2 Eh/2 R1 E−1 R2 Eh/2 u1 = kµ h/2 h/2 (10. to the upper and lower external surfaces.17) p2 Q1 Eh/2 −Q2 E−1 h/2 R1 Eh/2 R2 E−1h/2 u2 or more compactly      p1 K11 K12 u1 = (10.13) a2 a2 which can be combined into   $ %  $ %  u1 R1 E−1 h/2 R2 Eh/2 a1 R1 E−1 h/2 R2 Eh/2 a1 = = (10. u2 = R1 E−1 −h/2 R2 E−h/2 1 (10.

Im(kp) ≥ 0.   K11 K12 K= K21 K22 (continued ) . the SV-P elements are checkerboard symmetric and antisymmetric with respect to ±k. plane strain   kp = k2 − (ω/α)2 . Given ω. This is based on the (very close) approximation ! !   ! µc µ 1 µ βc = = 1 + 2i ξS sgn(ω) ≈ ρ ρ 1 − i ξS sgn(ω) ρ and a similar expression for αc . If material (hysteretic) damping is present. Also. ξS = 1/(2QS ). These approximations greatly simplify exponential terms such as    iωx iωx ξS ωx exp − = exp − exp − . ks = k2 − (ω/β)2 Note 1: In the first Riemann sheet. s themselves are in the first quadrant if k > 0. the products kp. Im(kp) ≥ 0. ω > 0 −1/ sinh ksh coth ksh   coth kh −1/ sinh kh K = kµ K = kµ k > 0. Cartesian coordinates. the complex wavenumbers for P and S waves are of the form ω ω  ω ω  = 1 − i ξ P sgn(ω) . QS = quality factor βc β β Note 2: Given k. ks are complex numbers in the first quadrant. ω > 0 −1/ sin S cot S β   µ 1 −1 K= K=0 k = 0. 1) SH waves Layer Half-space   coth ksh −1/ sinh ksh K = ksµ K = ksµ k > 0.1.2 Stiffness matrix method in Cartesian coordinates 145 Table 10. and Re(ks) ≥ 0. Stiffness matrices.10. This is the result of a deliberate change here in the sign of the imaginary factor applied to the vertical components. we replace r in said reference by p. S = K = iωρ β k = 0. whether or not the medium has material damping (attenuation). βc are the complex wave velocities. the numbers p. to avoid confusion with the radius or range. that is. all SH and SV-P elements exhibit complex-conjugate symmetry with respect to ±ω. Thus. and in the third if k < 0. ω = 0 h −1 1 2) SV-P waves Note: The horizontal–vertical coupling terms in the matrices for SV-P waves given here have opposite sign from those in Kausel and Roesset ¨ (footnote 6). Im(ks) ≥ 0. ω = 0 −1/ sinh kh coth kh   cot S −1/ sin S ωh K=ρβω . = 1 − i ξ S sgn(ω) αc α βc β in which αc . no matter what the sign of k should be.

ω = 0 β κ = kh. K21 = K12 T 2D − (C p − Cs ) 1 p ( ps Ss − Sp )  $ % 1 2 1 ( ps Ss − Sp ) − (C p − Cs ) K−1 12 = p 2kµ 1 − s2 C p − Cs 1 s ( ps Sp − Ss ) Lower half -space z < 0 (for an upper half-space. a=  2  2 α D = 1 + a 2 S2 − κ 2 1 − a 2     1 (1 + a 2 ) SC − κ (1 − a 2 ) −(1 + a 2 ) S2 0 1 K11 = 2kµ + D −(1 + a 2 ) S2 (1 + a 2 ) SC + κ (1 − a 2 ) 1 0 K22 = same as K11 . reverse the sign of the off-diagonal ter ms) :     1 − s2 p −1 0 1 K = 2kµ + 2 (1 − ps) −1 s 1 0 $ % 1 s 12 (1 − s 2 ) ps − 12 (1 + s 2 ) K−1 = 2kµ  ps − 12 (1 + s 2 ) p 12 (1 − s 2 ) b) Zero frequency. non-zero wavenumber: k > 0.146 Stiffness matrix method for layered media Table 10. with off-diagonal signs reversed  $ % 1 − s2 1 ( ps Sp − Ss ) C p − Cs K12 = 2kµ s .1.  = ps − = Rayleigh function ps 2  $ %   1 − s2 1 s (C p Ss − ps Cs Sp ) 1 − C p Cs + ps Sp Ss 1 + s2 0 1 K11 = 2kµ + 2D 1 − C p Cs + ps Sp Ss 1 p (Cs Sp − ps C p Ss ) 2 1 0 K22 = same as K11 . Sp = sinh kph. non-zero wavenumber: k > 0. (continued) a) Non-zero frequency. C = cosh κ. with off-diagonal signs reversed   2kµ κ (1 − a 2 ) C − (1 + a 2 ) S κ (1 − a 2 ) S K12 =   . S = sinh κ. K21 = K12 T D −κ (1 − a 2 ) S − κ (1 − a ) C + (1 + a ) S 2 2     1 − κ (1 − a 2 ) C + (1 + a 2 ) S −κ (1 − a 2 ) S K−1 12 = 2kµ κ (1 − a ) S 2 κ (1 − a 2 ) C − (1 + a 2 ) S . ω > 0 C p = cosh kph. Cs = cosh ksh. Ss = sinh ksh   2 1 1 + s2 D = 2 (1 − C p Cs ) + + ps Sp Ss .

. respectively. When the frequency and/or layer thickness are large. . K−1 = 1 + a2 a2 1 2kµ (1 − a 2 ) −a 2 1 1 Note : =1−ν 2 (1 − a 2 ) c) Non-zero frequency.. ⎪⎪ .. reverse the sign of the off-diagonal ter ms) :     2kµ 1 a2 1 1 −a 2 K= . and the material properties and thickness of the layer.N−1 K NN N . each of which is characterized by a stiffness matrix that depends on the horizontal wavenumber. ⎬⎪⎨ ⎪ ⎬ p3 = 0 K32 K33 .. we can move on and consider next a layered system with N−1 layers and N interfaces. S = α β ⎧ ⎫ ⎪ β cot S 0 −β/ sin S 0 ⎪ ⎪ ⎨ ⎪ ⎬ 0 α cot P 0 −α/ sin P K = ρω ⎪ ⎪ −β/ sin S 0 β cot S 0 ⎪ ⎪ ⎩ ⎭ 0 −α/ sin P 0 α cot P   β 0 Half-space K = iωρ 0 α d) Zero frequency. zero wavenumber: k = 0. ⎪ ⎪ ⎪ ⎪ . we can assemble the system impedance matrix by overlapping appropriately the matrices for neighboring layers. K N−1. K21 tend to zero.. The result is a block-tridiagonal matrix equation of the form ⎧ ⎫ ⎧K ··· ⎫⎧ ⎫ ⎪ p1 ⎪ ⎪ K12 0 0 ⎪ u1 ⎪ ⎪ ⎪ ⎪⎪ ⎪ 11 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ p2 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ K21 K22 K23 ··· 0 ⎪ ⎪ ⎪ ⎪ ⎪ u2 ⎪⎪ ⎨ ⎬ ⎨ . Thus.. . . ⎪ ⎪ ⎩ ⎭ ⎪ ⎪ ⎪ . . ⎪ .N ⎪ ⎪ ⎪ ⎪ ⎪ . . . the frequency. zero wavenumber: k = 0. ⎪ ⎪ pN ⎩ ⎭⎩u ⎭ 0 0 ··· K N.. ω = 0 ⎧ ⎫ ⎪ 1 0 −1 0 ⎪ ⎪ ⎪ µ⎨ 0 a −2 0 −a −2 ⎬ K= h⎪ ⎪ −1 0 1 0 ⎪ ⎪ ⎩ ⎭ 0 −a −2 0 a −2 Note : µ a −2 = λ + 2µ   0 0 Half-space K = 0 0 half-spaces listed in the tables.2 Stiffness matrix method in Cartesian coordinates 147 Lower half -space z ≤ 0 (for upper half-space. Having found the stiffness matrix for a given layer. . while the main elements K11 . . ω > 0 ωh ωh P = .20) ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ .10. u3 (10. K22 converge to the impedances of a lower and an upper half-space.. the coupling elements K12 .

This does not constitute a limitation of the method. We shall demonstrate the use of these matrices later on in this chapter by means of various examples.. Thus.1). the source vector in the frequency–wavenumber domain. and that there are no sources within the layer. Using super-indices ζ . Consider an individual layer of finite thickness h. this subdivision technique allows us also to obtain displacements within the layer in terms of the displacements at the interfaces by means of analytic continuation. however. ω) is the vector of external tractions applied at the interfaces. η to identify the two sub-layers.21) Here. solving this equation for the displacements and then carrying out appropriate integral transforms back into the spatial–temporal domain.1 Analytic continuation in the layers The stiffness matrix formulation described in the foregoing is based on the assumption that the external sources are applied solely in the form of tractions at the layer interfaces.22) ⎩ ⎭ ⎪ ⎩ ⎪⎩ ⎭ p2 η η ⎭ u2 0 K21 K22 Condensation of the internal degree of freedom gives  −1   ζ η ζ η u(z) = − K22 + K11 K21 u1 + K12 u2 (10. the dynamic equilibrium equation for this system is ⎧ ⎫ ⎧ Kζ ζ ⎫⎧ ⎫ ⎨ p1 ⎬ ⎪ ⎨ 11 K12 0 ⎪⎬ ⎨ u1 ⎬ η η 0 = Kζ21 ζ K22 + K11 K12 u (10.1: Analytic continuation in layer. and K is the symmetric system stiffness matrix. and subdivide it at elevation z it into two sub-layers of thicknesses ζ and η. i. 10. 10.e. η z2 x which can be written compactly as p = Ku (10.2. since it is always possible to define an auxiliary interface at the location of an internal source and subdivide the physical layer into two sub-layers with identical material properties. On the other hand.23) . respectively (so that ζ + η = h) (see Fig. which is generally composed of complex elements. p = p (k. we obtain the requisite solution for the layered medium.148 Stiffness matrix method for layered media z1 ζ h z Figure 10. as described next.

(10. K22 = K22 − K21 K12 K12 (10.24)  −1  −1 η ζ η ζ η η ζ K21 = −K21 K22 + K11 K21 .27) In the special case of a half-space loaded at its surface. K12 = −K12 K22 + K11 K12 (10.2 Stiffness matrix method in Cartesian coordinates 149 z1 ζ z Figure 10. η η K11 ≡ Khalf (i. Table 10. Observe also that when either ζ or η becomes zero. For example. when z = z2 .26) which provides the displacements at interior points at elevation z in terms of the interface displacements. so K21 K21 = I and u = u2 . this implies  −1 ζ ζ ζ ζ Khalf = K11 − K12 K22 + Khalf K21 (10. since condensation of the auxiliary interface must return a half-space. we have  −1  h −1 K021 = {∞} ⇒ K021 = 0.1 gives explicit expression for the inverses needed above. Of course.2: Analytic continuation in half- space. after condensation. This implies the identities  −1 ζ ζ  η −1 ζ ζ η η K11 = K11 − K12 K21 K21 .10. we recover the stiffness matrix for the complete layer. the above expression returns the correct value at each interface. ζ = z1 − z. x On the other hand. so that  −1 ζ ζ u(z) = − K22 + Khalf K21 u1 (10.30) Hence  −1   ζ ζ u(z) = K12 Khalf − K11 u1 (half-space z ≤ z1 ) (10. K12 = 0.29) so  −1  −1   ζ ζ ζ ζ K22 + Khalf K21 = − K12 Khalf − K11 (10.28) However. the half-space impedance).25) It follows that  η −1  −1 ζ u(z) = K21 K21 u1 + K12 K12 u2 (layer z2 ≤ z ≤ z1 ) (10. the latter must have been obtained first by solving the complete system..e. we have z ≤ z1 . η = ∞.31) .

Cs = cosh ksh = cosh (c + id) = ec C2 . Sp = ea S1 .36) sinh ksζ + cosh ksζ τ yz1 τ yz(ζ ) = (10. − 12 h ≤ z ≤ 12 h cosh 12 ksh sinh 12 ksh   1 cosh ksz sinh ksz = τm + τ (10.38) = ea C1 def where we have assumed a to be a positive number (appropriate changes are needed if a < 0). that there are no sources within the half-space.2.e. the displacements and stresses in the frequency– wavenumber domain at an interior point at a distance z from the midplane of the layer can be written explicitly as   1 cosh ksz sinh ksz u y (z) = um + u .32) 2 cosh 12 ksh sinh 12 ksh $ % sinh ksz cosh ksz τ yz(z) = ksµ um + u . i.34) 2 2 τ yz1 + τ yz2 τ yz1 − τ yz2 τm = . they break down in the case of a half-space.1 have terms containing hyperbolic functions of complex argument. Similarly. To illustrate the type of strategy needed..2 Numerical computation of stiffness matrices The elements of the stiffness matrices listed in Table 10. of course. to avoid errors associated with severe cancel- lations of large numbers. For details. SH waves). consider the following decomposition: C p = cosh kph = cosh(a + ib) = cosh a cos b + i sinh a sin b      = ea 21 1 + e−2a cos b + i 1 − e−2a sin b (10. s= 1 − (ω/kβ)2 (10. see the previously cited work by Kausel and Roesset¨ (footnote 6). The factor C1 is now well behaved. In that case. u = . − 12 h ≤ z ≤ 12 h (10.39) . In the anti-plane case (i.e..33) 2 cosh 12 ksh sinh 12 ksh in which u y1 + u y2 u y1 − u y2  um = . Ss = ec S2 (10. Thus. Hence.37) sinh ksζ + cosh ksζ 10.150 Stiffness matrix method for layered media which assumes. ζ = z1 − z (10.35) 2 2 These expressions for SH waves are based on placing the origin of coordinates at the center of the layer. which can attain very large numerical values even if the stiffness elements themselves will not. The condensation strategy can be used also to evaluate the effect of sources that are distributed vertically in a continuous fashion. sources distributed across the layers. it is essential to address this problem explicitly in the program routine that evaluates these functions. τ = (10. u y1 u y (ζ ) = .

−i uz. Example 10. use of the imaginary factor is essential to preserve symmetry. +i uz. 3) Form the system matrix K and solve for u : u = K −1 p. Thus. they would not be Hermitian either.41) and from Table 10. ˜ 5) Find the actual (displacements at each interface with the inverse Fourier transform +∞ −i kx u(x. z. after solving for the transformed displacements. ω) = 2π 1 −∞ u(k. ω) ei kx dk. √ 4) Multiply all axial (vertical) components of u by +i = −1 : u → u. z. the solution is then obtained along the following lines: 1) Form the (physical) load( vector p˜ (the tilde refers to the frequency–wavenumber +∞ domain).. the dynamic equilibrium equation is Ku = 0 ⇒ det K = 0 (10. we chose a negative sign instead so as to achieve matrices in cylindrical coordinates that are identical to those in Cartesian coordinates.42) .. Thus.3 Summary of computation Without the imaginary unit factor applied to each of the vertical components of displace- ments and stresses (i. z. +i σzz) would also have led to symmetric matrices. √ 2) Multiply all axial components p˜ of by −i = − −1 : p˜ → p. though a positive sign for this factor (i. p(k.2. On the other hand. ω) e ˜ dk. −i σzz).40) 2 e−(a+c) − C1 C2 + ps + ps S1 S2 which is now numerically stable and well behaved. ω) = −∞ p(x.1.e. 10.˜ z. introducing this into the first term in the first element of K11 .e.10.1: Rayleigh waves in a half-space In the absence of external tractions at the surface of a half-space. More- over. Formally. the vertical components must be multiplied by +i to recover the physical components. *  *  2 * 1 − s2 p −1 0 1 * 1 + s2 *2kµ + *=0 ⇒  = ps − =0 * 2 (1 − ps) −1 s 1 0 * 2 (10. if the soil had damping.2 Stiffness matrix method in Cartesian coordinates 151 For example. the stiffness matrices would not be symmetric. we obtain C p Ss ea+c C1 S2 =   D 2 (1 − e C1 C2 ) + ps a+c 1 + ps ea+c S1 S2 C1 S2 =   1  (10.

45) −iw a−1 w a−1 2 (1 − ps) Thus.0276v3 (10.43) α β 2 β k kR This equation can be changed into a bicubic equation with three roots. The system equation is then ⎧ ⎫ ⎨ coth ks1 h −1/ sinh ks1 h ⎬ K = ks1 µ1 s2 µ2 (10. so Rayleigh waves are non-dispersive. j = 1.50) µ1 k2 − k2 1 . which are       $ % u 1 u 1 1 − s2 = ap ⇒ = i ap with a = (10. tanh h k2 − k12 +  =0 (10.48) s1 µ1 sinh2 ks1 h and after brief algebra   µ2 k2 − k22 s2 µ2 tanh ks1 h + = 0. Example 10. The above characteristic equation can also be written as   µ2 k − k2 2 2 tan h k1 − k = 2 2  =0 (10.e. A very tight approx- imation to the speed of Rayleigh waves as a function of Poisson’s ratio is CR = 0. of which only one is physically meaningful and the other two are spurious roots that result from the rationalization process of squaring the two radicals.47) ⎩ −1/ sinh ks1 h coth ks1 h + ⎭ s1 µ1 Free waves require the determinant of the system stiffness matrix to vanish.197v − 0..056v2 − 0. the stiffness matrices for the layer (1) and the half-space (2) are.44) β The relative amplitudes of Rayleigh waves at the surface follow from the eigenvectors. From Table 10.49) s1 µ1 µ1 k2 − k12 with kj = ω/β j . 2. the vertical displacements are 90◦ out of phase with the horizontal. i. respectively.2: Love waves in a layer over an elastic half-space Consider an elastic layer of thickness h underlain by an elastic half-space.874 + 0. The single physical root is found to be independent of frequency. s2 µ2 1 coth2 ks1 h + coth ks1 h − =0 (10. or in full. CR = ≡ (10.152 Stiffness matrix method for layered media which when written in full can be recognized as the well-known Rayleigh function     2 CR 2 CR 2 1 CR 2 ω ω 1− 1− − 1− = 0.1.46) sinh ks1 h −1 cosh ks1 h  with s j = 1 − (ω/kβ j )2 .   ks1 µ1 cosh ks1 h −1 K1 = . K2 = ks2 µ2 (10.

While the above systems could be solved analytically and the result simplified with the aid of basic trigonometric identities. that is. and they will exhibit sharp peaks and undulations that . i. u˜ y = u˜ y (k. over frequencies. Example 10.3: SH line source at some depth in a layer over an elastic half-space The geometry is as in Example 10. z.10. if needed. we obtain the displacements in space–time by carrying out numerically an inverse Fourier transform over wavenumbers and.51) Carrying out a Fourier transform in x and t. as it is also in the propagator matrix method. we can obtain the displacement amplitudes at the interfaces as eigenvectors of the equation Ku = 0. we obtain the source in the frequency– wavenumber domain as b˜ y (k. displacements caused by Love waves within the layer or the half-space can be obtained by means of the analytic continuation technique given previously.2 Stiffness matrix method in Cartesian coordinates 153 which is the classical equation for Love waves. the kernels for layered media are likely to be wavy.52) which is equivalent to a surface traction py = 1 in a horizontal plane at elevation z = z . Thereafter. which corresponds to a body load by (x. After thus finding the response functions. If the source is in the upper layer at a depth ζ = z1 − z with complement η = h − ζ = z − z2 and defining κ = s2 µ2 /s1 µ1 then ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎨ coth ks1 ζ −1/ sinh ks1 ζ 0 ⎬ ⎨ u˜ y1 ⎬ ⎨ 0 ⎬ ks1 µ1 −1/ sinh ks1 ζ coth ks1 ζ + coth ks1 η −1/ sinh ks1 η u˜ y = 1 ⎩ ⎭⎩ ⎭ ⎩ ⎭ 0 −1/ sinh ks1 η coth ks1 η + κ u˜ y2 0 (10. ω) = δ(z − z ) (10.e. However. then the system is ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎨ coth ks1 h −1/ sinh ks1 h 0 ⎬ ⎨ u˜ y1 ⎬ ⎨ 0 ⎬ ks1 µ1 −1/ sinh ks1 h coth ks1 h + κ coth ks2 ζ −κ/ sinh ks2 ζ u˜ = 0 ⎩ ⎭ ⎩ y2 ⎭ ⎩ ⎭ 0 −κ/ sinh ks2 ζ κ (coth ks2 ζ + 1) u˜ y 1 (10. but now the medium is subjected to an impulsive line source at some elevation z and in direction y. t) = u˜ y ei(ωt−kx) dk dω (10.. t) = δ(t) δ(x) δ(z − z ) (10. Once we have found the roots of this equation.53)  whereas if the source is in the half-space at ζ = z2 − z and defining again κ = s2 µ2 /s1 µ1 .54) We added tildes to the components to remind us that that the displacements in these expressions are cast in the frequency–wavenumber domain.2.55) 2π −∞ −∞ and similar expressions for the other components. If the medium has attenuation. so the numerical integrals are feasible. the sys- tem has no singularities.  2  +∞  +∞ 1 u y (x. z. in the normal use of the stiffness matrix method – especially for a system with several layers – this is done numerically for each frequency and wavenumber. ω).

and Roesset. 118. our problem would be of the general form Ku= p. Vol. multiply these by p˜ y (k). and the two independent line loads require replacing the load vector on the right- hand side by a two-column matrix. 5 Phinney.4: Mindlin plate subjected to horizontal and vertical line loads at upper surface A homogeneous plate of arbitrary thickness h is subjected to either a horizontal or a vertical line load at the upper surface. this problem is similar to the previous one. which is a well-established technique in computational seismology. such as the Cholesky decomposition of K. In principle. and finally carry out inverse Fourier transformations back into the spatial domain. Thus. say. More generally. this succinct description glosses over many computational details that cannot be addressed herein. M.57) −∞ Alternatively. we could continue using a unit load on the right-hand side and. This would allow us to solve simultaneously for various load types. except that the Mindlin plate calls for the use of the SV-P stiffness matrix (Tables. only after solving for the displacements. we strongly recommend the use of complex frequencies5 (the so-called complex exponential window method 6 ). pp. a strip load of various widths.56) −a ik ka ka Thus. Vol. Of course. or a line load and. a single couple (whose Fourier transform is −i k). and to minimize this undesirable effect. No. Example 10. then multiply the vertical components of these vectors by +i (to remove the implied imaginary factors from these). the wavenumber step. Theoretical calculation of the spectrum of first arrivals in a layered medium. 6 Kausel. 721–734. we first solve simultaneously for the two displacement vectors. the wavenum- ber representation of the distributed load would have been  +a  q0  i ka  sin ka sin ka p˜ y = q0 eikx dx = e − e−i ka = 2a q0 = (10. These may affect the accuracy with which the transforms can be evaluated. If instead of the line source we had applied in this problem a strip load q0 distributed uniformly over a horizontal area of width 2a with total intensity 2a q0 = 1. the only difference from our previous problem is that the number 1 on the right-hand side is replaced by (sin ka)/ka. any arbitrary external traction py (x) dis- tributed horizontally over an interface at elevation z would lead to a load term of the form  +∞ p˜ y (k) = py (x) eikx dx (10. Journal of Geophysics Research. Frequency domain analysis of undamped systems. ¨ J. .2). the tails of integrals. R. pp. 70.. Journal of Engineering Mechanics. 1992. or in full ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎪ ⎪ K11 K12 K13 K14 ⎪ ⎪ u˜ 1x ⎪ u˜ 1z ⎪ ⎪ 1 0 ⎪ ⎨ ⎬⎪⎨ ⎪ ⎬ ⎪ ⎨ ⎪ ⎬ K21 K22 K23 K24 −i w˜ 1x −i w˜ 1z 0 −i = (10. 10. In principle. 1965.1.58) ⎪ K ⎪ 31 K K K ⎪ ⎪ 34 ⎪ ⎪ u ˜ u˜ ⎪ 2z ⎪ ⎪ ⎪ 0 0 ⎪ ⎪ ⎩ 32 33 ⎭ ⎩ 2x ⎭ ⎩ ⎭ K41 K42 K43 K44 −i w˜ 2x −i w˜ 2z 0 0 Observe the extra factor −i in the second column on the right-hand side applied to the vertical load in the frequency–wavenumber domain ( p˜ 1z = 1). for example.. and so on. A. 5107–5123. 10.154 Stiffness matrix method for layered media increase in number with increasing frequency. 4. E.

U.65)   s = 1 − (ω/kβ)2 = 1 − (c/β)2 (10. ϕ2 = ±J ϕ1 . D. Hence. Asbury Park. These are deceptively simple-looking transcendental eigenvalue equations. Proceedings 11th Annual Symposium on Frequency Control. R.64) 2   p = 1 − (ω/kα)2 = 1 − (c/α)2 (10. J = (10.. this eigenvalue prob- lem admits two type of modes: symmetric and antisymmetric. Pergamon Press.S. the real modes of propagation can be found using search techniques. evanescent modes) is exceedingly difficult7. it is possible to work out the two determinants in closed form. R. . These are obtained from the eigenvalue problem Kϕ = 0 (10.66) in which c is the phase velocity (or celerity) of the modes. 8 Mindlin. Proceedings 1st Symposium on Naval Structural Mechanics. Mathematical theory of elastic plates. They are ps tanh 12 kph − q2 tanh 12 ksh = 0 (10. 10. the 4 × 4 eigenvalue problem can be written as           K11 K12 ϕ1 0 K11 ±K12 J ϕ1 0 = ⇒ = (10. although only after considerable and tedious algebra. one for each sign. Waves and vibrations in isotropic. The displacement compo- nents for the upper and lower interfaces in these two types of modes satisfy the kinematic constraint     ϕ1 1 0 ϕ= . 1960. For given values of the frequency ω.63b) with 1 + s2 q= (10. CA (Aug.3).2 Stiffness matrix method in Cartesian coordinates 155 An alternative solution method could consist in carrying out a superposition of the so-called normal modes (or Lamb’s modes) of the plate.62) The eigenvalues for the symmetric and antisymmetric modes are obtained by setting equal to zero the two determinants of this equation. the second row is identical to the first. yet their full solution for complex wavenumbers k (i.e.. the results should agree with those of a direct numerical search of the determinant of the stiffness matrix. Since this problem involves a single homogeneous layer.59) Because of the symmetry of the plate with respect to the midplane.8 (see Fig. NJ. Of course. Army Signal Engineering Laboratory. Stanford University. elastic plates. 7 Mindlin.61) K21 K22 ±J ϕ1 0 ±J K21 J K22 J ϕ1 0 Because of the structure of the stiffness matrices.63a) and ps coth 12 kph − q2 coth 12 ksh = 0 (10. the reduced eigenvalue problem is simply [K11 ± K12 J] ϕ1 = 0 (10. D.60) ϕ2 0 −1 Hence. 1958).10. 1957.

Bottom: antisym- metric modes.3: Wave spectrum for Mindlin plate. the phase velocity of some of these modes converges to the Rayleigh wave velocity. such superposition must include both the prop- agating modes and the evanescent modes (real vs. which is useful for showing that at high frequencies. Top: symmetric modes.31. ωh / π 15 10 5 0 0 5 -5 10 0 5 15 10 15 Re kh / π Im kh / π By the way. very much along the lines of what was done in the example in Section 9. ν = 0. complex eigenvalues or characteristic wavenumbers k).1. . in which case the propagating modes may suffice. observe that  = ps − q2 is the Rayleigh function.156 Stiffness matrix method for layered media ωh / π 15 10 5 0 0 5 -5 10 0 5 15 10 15 Re kh / π Im kh / π Figure 10. After finding the modes. The great advantage of modal solutions is that the inverse transforms back into the space domain can be carried out analytically. In principle. the displacements for the line load could be obtained by modal superposition. The exception is when we are interested only in the response at some distance from the source.

2 Stiffness matrix method in Cartesian coordinates 157 A substantial simplification to this problem can be achieved by means of the thin-layer method9 (TLM). the propagation modes – propagating as well as evanescent – can readily be found. Hence. . see references in that paper to earlier work on the subject. kz = cos θ. the plate is subdivided into an adequate number of sub-layers that are thin in the finite element sense. In the absence of the upper layers. kx p j = sin2 θ − for P waves (10. so that together they form a homogeneous full space.4c. these waves have horizontal and vertical wavenumbers  2 ω ω ω α kx = sin θ. Hence. This medium is subjected to plane P or SV waves that originate in the half-space with prescribed incli- nation θ with respect to the vertical. where we consider a reference problem with known solution that consists either of a lower half-space by itself alone (i. it suffices to apply a fictitious source at the half-space interface equal to Khalf u∗outcrop = Kfull u∗full (the second of which is much simpler). pp. Either way. and the method of weighted residuals is applied to obtain the discrete equations of motion.67) α α α αj  2 ω ω ω β kx = sin θ. see Fig. a quadratic eigen- value problem in narrowly banded matrices that can be cast as a linear eigenvalue problem of double size.e. International Journal for Numerical Methods in Engineering. Compare this against Fig. We refer to this response as the outcropping motion. Focusing attention on the lower substructure in each fig- ure. We shall show that to obtain the solution to the wave amplification problem with the stiffness matrix method.10. Consider a stratified medium consisting of N−1 layers (N interfaces) underlain by an elastic half-space. 10. kx s j = sin2 θ − for SV waves (10. Thin layer method: formulation in the time domain.4b. and formulas can be found in many standard references. Also. 927–941. Vol. we see that the motions at the surface of the two lower half-spaces differ from one 9 Kausel. 10. 37.4c). and in either case the seismic source is placed in the lower medium. Consider Fig. In this alternative.. 10. Example 10. then the motion is simply that of a plane wave. we distinguish the outcropping motion or full space with a superscript star (i.5: Amplification of plane SV-P waves in a layered medium We demonstrate next the application of stiffness matrices to the wave amplification prob- lem by recourse to substructuring techniques.. which shows the layered system separated into two substruc- tures or free bodies. 1994. u∗ ) and assume it to be known (see Fig. kz = cos θ. if we replace the upper layers by an upper half-space with the same properties as the half-space underneath.4. a rock outcrop) or of a full space divided into a lower and an upper half-space. Even simpler. the displacements are then discretized in the direction of layering via interpolation polynomials. The TLM then changes the eigenvalue problem from transcendental to conventional – in essence. 10.68) β β β βj in which the non-subscripted wave velocities are those of the half-space.e. the response caused by these plane waves at the surface of the half- space is well known. namely the layers and a lower half-space. E.

s*. a) Actual layered system with seismic source. Since the stress imbalance and deviations in displacements are observed on the same horizon. This is ostensibly the result of the interface stresses acting between the layers and the half-space. u* Outcrop another. and u∗ is the (presumably known) outcropping motion elicited by the source.4: Amplification of plane SV-P waves using stiffness sN . c) Upper half-space . c) Half-space alone (either a rock outcrop or a divided full space). despite the fact that the lower substructures are identical and the source is exactly the same. s = 0 is the stress-free condition at the surface of the half- space. so that − s N = −s∗ + Khalf u∗ − Khalf u N (10. they must satisfy the dynamic equilibrium equation s N − s∗ = Khalf (u N − u∗ ) .158 Stiffness matrix method for layered media a) 1 N b) 1 N −sN . u* s*. For .69) ∗ In the case of a rock outcrop. b) Lay- ers and half-space as free bodies. so the difference in motions is solely the result of the differences in stresses. u N Figure 10. u N matrices. which act as secondary “sources” applied at the half-space interface.

Finally. ⎪ ⎪ ⎬ ⎪ ⎪ ⎨ . ⎪ ⎪ ⎬ 21 . and it must be assembled using the appropriate horizontal wavenumber k = kx given previously for either of the two wave types being considered. ⎪⎪ ⎪ ⎪ . .71) Clearly. In the ensuing. ⎪ ⎪ ⎬ ⎪ ⎪ ⎨ . ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨K K22 . on the other hand. and Kfull is the impedance of the full space. without worrying about critical angles and the like.3 Stiffness matrix method in cylindrical coordinates 159 the full space. For example. Finally.. and boundary conditions – but not necessarily the loads – exhibit cylindrical symmetry with respect to the vertical axis z...4b. ⎪ ⎪ . we distinguish between the cases of horizontal. = .. ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ . we conclude that the external stresses −s∗ applied to it must satisfy the dynamic equilibrium equation   −s∗ = Kupper half u∗ . 10. it suffices for us to multiply the motions u∗ caused by a plane wave in the interior of a full space by the impedance of that full space.10. ⎪ ⎪ ⎪⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ ⎪ 0 K N−1. 10. The latter is simply a diagonal matrix whose elements equal twice the diagonal elements of the stiffness matrix of the half-space.72) −i cos θ z=0 Observe that – as in all previous application of the stiffness matrix method in plane strain – the vertical components of u∗ and s∗ must carry a factor −i.73) which constitutes an easy way to compute the reflection of plane waves at the surface of an elastic half-space.. Taking into account the fact that there are no sources in the upper half of the full space. so Khalf u∗ − s∗ ≡ Klower half + Kupper half u∗ = Kfull u∗ (10.70) in which Kupper half is the impedance of the upper half-space (identical to Klower half ≡ Khalf . . critical angles. but with signs changed in checkerboard fashion). applying the stresses −s N as external tractions at the bottom of the upper layers in Fig. = . u∗ and s∗ are the displacements and stresses observed at the elevation that in the layered system constitutes the half-space interface. observe that Khalf u∗outcrop = Kfull u∗full implies u∗outcrop = K−1 ∗ half Kfull ufull (10. geometric.N−1 K NN + Khalf Khalf u − s ∗ ∗ Khalf u ∗ outcrop Kfull u ∗ full (10. in the case of a P wave with amplitude A propagating at some angle θ with the vertical.. .. we obtain after brief algebra ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎪ ⎪ K11 K12 ··· 0 ⎪ ⎪ ⎪ u1 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ 0 ⎪ ⎪ 0 ⎪ ⎪ ⎪ . This second alternative has the advantage that there is no need for us to be concerned with reflections. and apply these as external loads to the layered medium.. and the like. ⎬⎪ ⎪ ⎨ u2 ⎪⎬ ⎪ ⎪ ⎪ ⎨ . the matrix on the left-hand side is that of the system of layers together with the half-space. u∗ is simply   * sin θ * u∗full = A e−i (kx x+kz z) * (10.3 Stiffness matrix method in cylindrical coordinates Consider an elastic system subjected to dynamic sources whose material. Thus.N ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎭ ⎩ uN ⎭ ⎪ ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ ⎩ 0 ··· K N. . ⎪ = ⎪ . .

10. it suffices to summarize the equations in Table 10. . d(kr) ⎪ ⎩ kr ⎪ ⎭ 0 0 Jn The stiffness matrix K is assembled with the elements of the plane strain matrix: ⎧ SVP SVP ⎫ ⎪ ⎨ K11 0 K12 ⎪ ⎬ K= 0 KSH 0 from Table 10. t) e−iω t dt dθ dr 2π 0 0 −∞ p˜ = K u. the equations of motion in cylindrical coordinates for radially homogeneous media.1 ⎪ ⎩ SVP ⎪ ⎭ SVP K21 0 K22 Form system matrix. the stiffness matrix method for flat layers in cylindrical coordinates is virtually identical to that in plane strain. z.3.160 Stiffness matrix method for layered media Table 10. for they are uncoupled. ω) = u˜ θ . z. and are handled in exactly the same way as in plane strain. and b) a Hankel transform is used in place of the Fourier transform to obtain the field quantities in the spatial domain. n. are identical to those in plane strain. Hence. parallel layers that are perpendicular to the symmetry axis z. ˜ Solve separately for the SV-P and SH components. ω) = r Cn Tn p(r. Thus. θ. z.2. ⎩ ⎭ u˜ z  +∞ ∞  ∞ 1 = eiω t Tn ˜ k Cn u(k. The only two differences are that a) the vertical components of stresses and displacements in cylindrical coordinates do not carry the implied imaginary unit factor −i. The stiffness matrices for single or multiple layers are otherwise identical. z.1.2 and illustrate the method by means of examples. when expressed in the frequency–wavenumber domain. n.1 Horizontally layered system As we saw in Section 9.2. and of cylindrical layers that are concentric with it. Flat layers in cylindrical coordinates ⎧ ⎫ ⎨ u˜ r ⎬ ˜ u(k. ω) dk dω 2π −∞ n=0 0    cos nθ − sin nθ cos nθ Tn = diag (upper or lower elements used) sin nθ cos nθ sin nθ ⎧ n ⎫ ⎪ ⎪  0⎪ ⎪ ⎨ Jn kr Jn ⎬ dJ n (kr) Cn = n Jn = ⎪ Jn Jn 0 ⎪.  ∞   +∞ 2 − δ0n 2π ˜ p(k.

give a unit load in direction x:  R  2π   δ(r )  R cos2 θ + sin2 θ r dr dθ = δ(r ) dr = 1 (10. Formally. added together.3 Stiffness matrix method in cylindrical coordinates 161 Example 10. . horizontal point load bx = δ(x) δ(y) δ(z). an = 1 − δn0 (10. In cylindrical coordinates. the two should be assembled and solved separately – and this is especially true for layered media. and integrated over a small circular area of radius R enclosing the origin.77) This implies that of all azimuthal components.10. the stiffness matrix will be of the form ⎧ SVP SVP ⎫ ⎪ ⎨ K11 0 K12 ⎪ ⎬ K= 0 KSH 0 (10. only the term n = 1 survives. when projected in the x direction.79) ⎪ ⎩ SVP ⎪ SVP ⎭ K21 0 K22 but because the SV-P components are uncoupled from the SH components. this corresponds to a surface traction with azimuthal index n = 1 and components ⎧ ⎫ ⎧ ⎫ ⎧ ⎫⎧ ⎫ ⎨ pr ⎬ cos θ ⎬ cos θ 1 ⎨ δ(r ) ⎨ 0 0 ⎬ ⎨1⎬ δ(r ) p1 = pθ = − sin θ = 0 − sin θ 0 1 ⎩ ⎭ 2π ⎩ ⎭ r ⎩ ⎭ ⎩ ⎭ 2π r pz (1) 0 0 0 cos θ 0 ⎧ ⎫ ⎨1⎬ δ(r ) = T1 1 (10.6: Elastic half-space subjected to a tangential point load at its surface (Chao’s problem) A lower elastic half-space z < 0 is subjected at its surface to a harmonic.74) ⎩ ⎭ 2π r 0 This is so because the radial and tangential components of the load.78) in which K is the stiffness matrix of the half-space. only u˜ 1 will exist. we begin by casting the load in the frequency–wavenumber domain: ⎧ ⎫  ∞  2π ⎨ 1 ⎬ δ(r )  1 1 p˜ n = an r Cn Tn T1 1 dθ dr . The displacement com- ponents in the frequency–wavenumber domain are then u˜ 1 = K−1 p˜ 1 (10. which is assembled with the elements given in Table 10.1 for SV-P waves and for SH waves. which is equivalent to a surface traction in the horizontal plane z = 0 in direction x of the form px = δ(x) δ(y).75) 0 0 2π r 0 To solve this problem using stiffness matrices.76) 0 0 ⎩ ⎭ 2π r π 2 0 Because of orthogonality conditions satisfied by the integral in the azimuth θ. so ⎧ ⎫ ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ p˜ r ⎬  ∞ ⎨ J1 1 0 ⎬ ⎨1 ⎬  ∞ ⎨ J0 ⎬ 1 ⎨ ⎬ J kr 1 1  δ(r ) δ(r ) p˜ 1 = p˜ θ = 1 J1 J1 0 1 dr = J0 dr = 1 ⎩ ⎭ 0 ⎩ kr ⎭ ⎩ ⎭ 2π 0 ⎩ ⎭ 2π 2π ⎩ ⎭ p˜ z (1) 0 0 J1 0 0 0 (10.

81) 1 1 1 = = ks µ 2π 2π k s µ Observe that.7: Same as Example 10.6.80) 4π kµ  ps − 12 (1 + s 2 ) and 1 u˜ θ (1) = p˜ SH θ (1) K  (10.162 Stiffness matrix method for layered media Therefore. Finally. Example 10.6. the vertical components do not carry an imag- inary factor. unlike the plane-strain case.82) ⎪ ⎩ ⎭ ⎪ 0 ⎩ ⎪ kr 1 ⎪ ⎭⎩ ⎪ ⎪ ⎭ uz 0 0 J1 u˜ z (1) that is      ∞ s 1 (1 − s2) ∞ 1 J1 (kr ) 1 1 J1 (kr ) ur = (cos θ) 2 J0 (kr ) − dk + dk 4π µ 0  kr 2π µ 0 s kr (10.85) 4π µ 0  which can be evaluated by numerical integration or.1.83)      ∞ s 1 (1 − s 2 ) J1 (kr ) ∞ 1 1 1 J1 (kr ) uθ = (− sin θ) 2 dk + J0 (kr ) − dk 4π µ 0  kr 2π µ 0 s kr (10. the displacements in the spatial domain are ⎧ ⎫ ⎧  1 ⎫⎧ ⎫ ⎪ ⎨ ur ⎪ ⎬  ∞⎪ ⎨ J1 J kr 1 0⎪⎬⎪⎨ u˜ r ⎪ ⎬  u = uθ = T1 1 J1 J 0 u˜ θ k dk (10.84)    ∞ ps − 1 (1 +s ) 2 1 uz = (cos θ) 2 J1 (kr ) dk (10. by analytical means. as Chao does. The only difference now is that in the frequency–wavenumber domain. but for a layer over an elastic half-space The procedure is identical to that in Example 10. we have $ % $ % $ % SVP SVP −1 u˜ r K11 K12 p˜ r = SVP SVP u˜ z (1) K21 K22 p˜ z (1) $ % $ % 1 s 12 (1 − s 2 ) ps − 12 (1 + s 2 ) 1 1 = 2kµ  ps − 1 2 (1 +s ) 2 p 1 2 (1 −s ) 2 2π 0 $ % 1 s 12 (1 − s 2 ) = (10. we must first solve the two systems ⎧ ⎫ ⎧ ⎫ ⎪ ⎪ u˜ r 1 ⎪ ⎪ ⎪ ⎪ ⎪ 1 ⎪     ⎨ ⎬  SVP −1 1 ⎨ 0 ⎬  −1 1 1 u˜ z1 u˜ θ1 = K and = KSH (10. from Table 10.86) ⎪ ⎪ u˜ ⎪ ⎪ 2π ⎪ ⎪ 0 ⎪ u˜ 2π 0 ⎩ r 2 ⎭ ⎩ ⎪ ⎭ θ2 (1) u˜ z2 (1) 0 .

1. except that the vertical load in cylindrical coordinates corresponds to a surface traction with azimuthal index n = 0.88) ⎩ ⎭ 2π ⎩ ⎭ r ⎩ ⎭ 2π r pz (0) 1 1 Hence.90) 4π kµ  p 1 2 (1 −s ) 2  1 1 u˜ θ (0) = p˜ θ (0) = (0) = 0 (10.87) ⎩ ⎭ 0 ⎩ kr ⎭ ⎩ θj ⎭ uzj 0 0 J1 u˜ zj (1) Example 10. 2 (10. and its compo- nents are ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ pr ⎬ 1 ⎨ ⎬ δ(r ) 0 ⎨ 0 ⎬ δ(r ) " # p0 = pθ = 0 ≡ T0 0 . In this case. it will be found that the transforms back into space for each elevation are intractable by purely analytical means.10.91) K SH ks µ Finally.89) ⎩ ⎭ 0 ⎩ ⎭ ⎩ ⎭ 2π 2π ⎩ ⎭ p˜ z (0) 0 0 J0 1 1 The solution for the displacement vector in ω. T0 = diag 1 0 1 (10. using for each elevation the inversion formula ⎧ ⎫ ⎧ ⎫⎧ ⎫ ⎨ ur j ⎬  ∞ ⎨ J1 1 J kr 1 0 ⎬ ⎨ u˜ r j ⎬  u j = uθ j = T1 1 J J1 0 u˜ k dk.8: Elastic half-space subjected to a vertical point load at its surface (Pekeris’s problem) This is again very similar to Example 10. the solution in the space domain is ⎧ ⎫ ⎧ ⎫⎧ ⎫ ⎨ ur ⎬  ∞ ⎨ J0 0 0 ⎬ ⎨ u˜ r ⎬ u = uθ = T0 0 J0 0 u˜ k dk (10. so they must be done numerically.92) ⎩ ⎭ 0 ⎩ ⎭⎩ θ ⎭ uz 0 0 J0 u˜ z (0) or in full  ∞ 1 ps − 12 (1 + s 2 ) ur = − J1 (kr ) dk (10.3 Stiffness matrix method in cylindrical coordinates 163 in which the matrices KSVP and KSH for the layered system are assembled with the elements in Table 10. the load in the frequency–wavenumber domain is ⎧ ⎫ ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎨ p˜ r ⎬  ∞ ⎨ J0 0 ⎬⎨0⎬ 1 ⎨ ⎬ 0 0 δ(r ) p˜ 0 = p˜ θ = 0 J0 0 0 dr = 0 (10. j = 1.93) 4π µ 0  . k is then $ % $ % $ % SVP SVP −1 u˜ r K11 K12 p˜ r = SVP SVP u˜ z (0) K21 K22 p˜ z (0) $ %   1 s 12 (1 − s 2 ) ps − 12 (1 + s 2 ) 1 0 = 2kµ  ps − 12 (1 + s 2 ) p 12 (1 − s 2 ) 2π 1 $ % 1 ps − 12 (1 + s 2 ) = (10.6.

radius) at which the matrices are being evaluated.e. = = a2 u2 p2 −r2 s2 −r2 Fn2 ⎩ ⎭ (1) (2) (1) (2) Hn2 Hn2 a2 −r2 Fn2 (10. a) Single layer Let’s consider first the outermost layer as an isolated.98)  T s = s(r ) = σ˜ r σ˜ r θ −iσ˜ r z = F(1) n a1 + Fn a2 (2) (10. The overbars are a reminder that this is a solution in the frequency–wavenumber domain with an added imaginary factor. As shown in Section 9. z. n a1 + Hn a2 e (2) ≡ Tn u e−ikz z (10.95) 4π µ 0  10.96)  T u = u(r. kz. Hn are as listed in Table 10. and superscripts indicate the kind of Hankel functions used. Fn are given in Table 10.2 Radially layered system Consider a system of N − 1 concentric cylindrical layers. ω) = u¯ r u¯ θ −i u¯ z = Tn H(1) u(r.e.94)  ∞ s 1 (1 1 − s2) uz = 2 J0 (kr ) dk (10. defining the tractions per radian as the products of the stresses and the interface radius. and are constructed with first and second Hankel functions.. we obtain $ % $ (1) %$ % $ % $ % $ %⎧a ⎫ r1 Fn1 ⎨ 1 ⎬ (2) (1) (2) u1 Hn1 Hn1 a1 p1 r1 s1 r1 Fn1 = . for given axial (i.99) (1) (2) in which Fn . respectively. the tildes indicate that the variations with z and θ have been separated. respectively.97) (1) (2) in which a1 . a particular solution for the displacement vector at a point in the interior of an individual cylindrical layer is of the form  T   −ikz z ¯ θ. vertical) wavenumber kz and frequency ω. Evaluating the displacements and stresses at these two surfaces..2. r2 . and the matrices Tn . The negative sign . and writing both together in matrix form.4. each of arbitrary thickness. the stresses in cylindrical surfaces are  T   −ikz z σ¯ r σ¯ r θ −iσ¯ r z = Tn F(1) n c1 + Fn c2 e (2) = Tn s e−ikz z (10.2. We assume the material properties in each layer to be independent of the azimuth θ.100) in which subscripts indicate the location (i.164 Stiffness matrix method for layered media uθ = 0 (10.3.3. free layer whose bounding outer and inner surfaces have radii r1 . Hn . Also. These layers could optionally be surrounded by an unboundedly large exterior region. as indicated by the superscript. a2 are arbitrary constants. whose N interfaces we number from the outside to the inside. and to be homogeneous within each cylindrical layer. ω) = u˜ r u˜ θ −i u˜ z = H(1) n a1 + Hn a2 (2) (10.

101) K21 K22 u2 . Matrix for displacements in cylindrical layers Obtain u by solving u = K −1 p for the system of layers. Hn (kα r ) layer Hαn = (2) Hn (kα r ) exterior ⎪ ⎪ ⎩ Jn (kα r ) core ⎧ (1) ⎪ (2) ⎪ Hn (kβ r ). Hnj = Hn Hn (kr j ) ⎪ ⎪ kα r kβ kβ r ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ kz ⎪ ⎪ ⎪ ⎩ − Hαn 0 Hβn ⎭ kα in which ⎧ (1) ⎪ ⎪ (2) ⎨ Hn (kα r ).3 Stiffness matrix method in cylindrical coordinates 165 Table 10. Hn (kβ r ) layer ⎨ Hβn = (2) Hn (kβ r ) exterior ⎪ ⎪ ⎩ Jn (kβ r ) core  dHαn  dHβn Hαn = . Hβn = d(kα r ) d(kβ r ) in the second row comes from the fact that external tractions are opposite in direction to the internal stresses at the inner surface. ω) = uθ = Tn u˜ n e−ikz z dkz ⎩ ⎭ 2π −∞ n=0 uz Tn = diag[ cos nθ − sin nθ cos nθ ] or Tn = diag[ sin nθ cos nθ sin nθ ] ⎧ Hβn kz  ⎫ ⎪ ⎪  ⎪ ⎪ ⎪ Hαn ⎪ n kβ r H kβ βn ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ H kz Hβn ⎬   αn  (i) (i) Hn = n Hβn n . ω) = u˜ θ → u˜ = u˜ θ ⎩ ⎭ ⎩ ⎭ −i u˜ z u˜ z ⎧ ⎫ $ % ⎨ ur ⎬  +∞ ∞ 1 u(r.10. kz. Eliminating the constants a1 . z. θ.3. we obtain $ % $ (1) (2) %$ (1) (2) %−1 $ % p1 r1 Fn1 r1 Fn1 Hn1 Hn1 u1 = p2 −r2 Fn2 (1) −r2 Fn2 (2) (1) Hn2 (2) Hn2 u2 $ %$ % K11 K12 u1 = (10. n. then for each interface: ⎧ ⎫ ⎧ ⎫ ⎨ u˜ r ⎬ ⎨ u˜ r ⎬ u(r. a2 between these two matrices.

kz. Fn = f21 f22 f23 ⎩ ⎭ ⎩ ⎭ ⎩ ⎭ −iσ˜ r z σ˜ r z f31 f32 f33 " #   (i) (i) Fn = fi j (Hαn . Elements of matrix for stresses in cylindrical surfaces ⎧ ⎫ ⎧ ⎫ ⎧ ⎫ ⎨ σ˜ r ⎬ ⎨ σ˜ r ⎬ ⎨ f11 f12 f13 ⎬ s(r. We list below the elements for a generic version of Fn : $   2      %  kz Hαn n 2 f11 = −kα λ 1 + Hαn + 2µ + 1− Hαn kα kα r kα r  2nµ  Hβn f12 = Hβn − r kβ r $   2  %  Hβn n f13 = −2kzµ + 1− Hβn kβ r kβ r 2nµ  Hαn f21 = Hαn − r kα r       Hβn n 2 f22 = −kβ µ 2 + 1−2 Hβn kβ r kβ r 2n µ  Hβn f23 = kz Hβn − kβ r kβ r  f31 = −2 kz µ Hαn Hβn f32 = −kzn µ kβ r   2  kz  f33 = kβ µ 1 − Hβn kβ in which $ % $ (1) (2) %$ (1) (2) %−1 K11 K12 r1 Fn1 r1 Fn1 Hn1 Hn1 K= = (1) (2) (1) (2) (10.166 Stiffness matrix method for layered media Table 10. etc.3. the external region can be regarded as a layer whose external radius is infinitely large. we must set a1 = 0 (to satisfy the radiation and boundedness conditions at infinity) and .. b) Unbounded external region In the case of a cylindrical cavity within an unbounded homogeneous space. Hβn ) . In this case.102) K21 K22 −r2 Fn2 −r2 Fn2 Hn2 Hn2 is the symmetric stiffness (or impedance) matrix of the cylindrical layer. n.4. Fnj = Fn Hn (kr j ) in which Hαn . are the same as in Table 10. ω) = σ˜ r θ . s˜ = σ˜ r θ .

.. ω) = 2−δ p(r. . K N−1. the 3 × 3 impedance matrix of the exterior region is  −1 Kext = −r F(2) n H(2) n (10. u → u. u3 (10. the solution is then obtained along the following lines: 1) Form the (physical) load vector  2π  +∞ ˜ n. ⎪ ⎪ pN ⎩ ⎭⎩u ⎭ 0 0 ··· K N. θ.N−1 K NN N or more compactly p = Ku (10..N ⎪ ⎪ ⎪ ⎪ ⎪ . √ 2) Multiply all axial components of p˜ by −i = − −1. ⎬⎪⎨ ⎪ ⎬ p3 = 0 K32 K33 . z. both of which must be constructed using second Hankel functions evaluated at the radius r of the external region. ⎪ ⎪ ⎪ ⎪ . p˜ → p. . 2π 0n Tn p(r.3 Stiffness matrix method in cylindrical coordinates 167 use thus only second Hankel functions.. 3) Form system matrix K and solve for u = K −1 p. Formally. . . θ...3 and 10. Hence. it suffices to overlap the layer matrices for each layer as appropriate. To construct the global (system) stiffness matrix. ⎪ .. consider a layered system. . z. The result is a system of equations of the block-tridiagonal form ⎧ ⎫ ⎧K ··· ⎫⎧ ⎫ ⎪ p1 ⎪ ⎪ K12 0 0 ⎪ u1 ⎪ ⎪ ⎪ ⎪ ⎪ 11 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ p2 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ K21 K22 K23 ··· 0 ⎪ ⎪ ⎪ ⎪ ⎪ u2 ⎪⎪ ⎨ ⎬ ⎨ ..10. ω) e−ikz zdz dθ (10.106) 0 −∞ (usually requires multiplying external pressures by r to obtain tractions per radian). beginning with the impedance matrix of the exterior region. ⎪⎪ .107) 2π −∞ n=0 . ω) = Tn u˜ n e−ikz z dkz (10.103) which is assembled with the matrices Hn and Fn in Tables 10.. √ 4) Multiply all axial (vertical) components of u by +i = −1. kz.104) ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ . ˜ 5) Find the actual displacements at each interface from the inverse transform  +∞ $ ∞ % 1 u(r. c) Layered system Finally. which act as external tractions at the layer interfaces. .105) in which p is the vector of external sources (tractions per radian) in the frequency– wavenumber domain. . ⎪ ⎪ ⎩ ⎭ ⎪ ⎪ ⎪ .4. which may optionally be surrounded by an unbounded exterior region.

10.108) K21 K22 u2 p2 . unbounded exterior region. H2 = H22 − H21 H−1 11 H12 This form is computationally more robust.   ⎩ F(2) = F H(2) (kr ) n n ni i  Hn = Hn (Jn (kr )) Kcore = r Fn H−1 n . the single layer matrix can be shown to be given by ⎧     ⎫   ⎨ r1 F11 − F12 H−1 H21 H1 ⎬ K11 K12 r1 F12 − F11 H−1 11 H12 H2 K= =  22    K21 K22 ⎩ −r2 F21 − F22 H−1 H21 H1 −r2 F22 − F21 H−1 H 12 H2 ⎭ 22 11 in which  −1  −1 H1 = H11 − H12 H−1 22 H21 . Fn = Fn (Jn (kr )) System of layers: p = Ku ( j) ( j) Note: Expressing the matrices above as Fni = Fi j . the force–displacement equation for the thick tube can be written as      K11 K12 u1 p1 = (10. we need to carry out a condensation of the degrees of freedom on the axis prior to taking the limit r1 → 0.3.   ⎩−r F(1) −r2 Fn2 ⎭ ⎩ Hn2 (2) (1) Hn2 ⎭ (2) ⎩ F( j) = F H( j) (kr ) 2 n2 ni n n i ⎧ (2)     ⎨ Hn = Hn Hni(2) (kri ) (2) (2) −1 Kext = −r Fn Hn . single layer.5.168 Stiffness matrix method for layered media Table 10. kβ = kS2 − kz2 ⎧ ⎫⎧ ⎫ ⎧ ( j)   ⎨ r1 F(1) (2) −1 r1 Fn1 ⎬ ⎨ Hn1 Hn1 ⎬ ⎨ Hni = Hn Hn( j) (kri ) (2) (1) n1 K= . To avoid the presence of sources on the axis that effectively vanish from the exterior. the layer matrix is symmetric. Stiffness matrix for cylindrical layers See Tables 10. 2 → inner surface kP = ω/α. kS = ω/β. Though it is not obvious from the above. Hni = Hi j . Fn used herein 1 → outer surface. Using the simplified notation in Table 10.5. d) Solid core As the inner radius of a thick tube approaches zero and forms a solid cylinder (or perhaps the core of a layered system). solid core. a singularity develops at the axis that decouples it from the exterior.4 for matrices Hn . kz= axial wavenumber   kα = kP2 − kz2 .

114)     K11 = r1 F11 − F12 H−1 22 H21 H1 → lim r1 F11 H−1 11 (10. and H−1 11 → O (the null matrix). H11 (Yn (kr1 )) (10. K11 is not particularly interesting to us. a 3 × 3 matrix that is constructed with conventional Bessel functions.113) Hence.119) so   −1 K22 u2 = p2 + r2 F21 − F22 H−1 22 H21 (r1 F11 ) p1 (10. On the other hand. we shall construct the stiffness matrices in Table 10.110) F12 (Jn (kr1 )) . We explore next the limit of the matrices when the inner radius goes to zero. For this purpose. H11 (but only these) will be singularly large when r1 → 0. but because the source on the axis is also infinitely large. the elements of K11 become either zero. the product of stiffness and displacement at this location is indeterminate. when r1 → 0.10. On the other hand.118) When r1 → 0.117)   K22 = −r2 F22 − F21 H−1 −1 11 H12 H2 → r 2 F22 H22 (10.111) F21 (Yn (kr2 )) . the coupling forces are well behaved. so that their product will remain finite.112) F22 (Jn (kr2 )) . we obtain    −1 K21 K−1 −1 11 = −r2 F21 − F22 H22 H21 H1 H1 −1 r1 F11 − r1 F12 H−1 22 H21   −1 → −r2 F21 − F22 H−1 22 H21 (r1 F11 ) (10. H2 = H22 − H21 H−1 11 H12 → H−1 22 (10. namely F11 (Yn (kr1 )) . as will be seen next. we observe also that the submatrix K22 converges to what we refer as the stiffness of the solid core.115) r1 →0   K12 = r1 F12 − F11 H−1 11 H12 H2 → O (10.116)   K21 = −r2 F21 − F22 H−1 22 H21 H1 → O (10. or infinite.3 Stiffness matrix method in cylindrical coordinates 169 Condensation of the inner degrees of freedom yields   K22 − K21 K−1 −1 11 K12 u2 = p2 − K21 K11 p1 (10. Thus.120) . H22 (Jn (kr2 )) (10. This implies the following trends:  −1  −1 H1 = H11 − H12 H−1 22 H21 → H−111 . From Table 10. finite. the product K21 K−111 on the right-hand side will tend to zero at the same rate as the singularly large forces p1 go to infinity.5 using Bessel and Neumann functions.109) As will be seen. all elements of F11 . H12 (Jn (kr1 )) (10. H21 (Yn (kr2 )) (10.5.

While all terms are zero in the limit. We proceed next to examine the limit limr1 →0 F−1 11 when the inner radius converges to the axis. and a lateral point load acting on the axis of a solid cylinder. we obtain the following: n = 0:  ⎧  2 ⎫ ⎪ ⎪ − kα − kz −2 kkβα kz ⎪ ⎪ ⎪ ⎪ kβ 1 zβ 0 ⎪ ⎪ ⎨ kβ  kβ ⎬  2 lim (r1 F11 )−1 →  π  2  − 1+ kz r1 →0 kz ⎪ ⎪ 0 zβ 0 ⎪ ⎪ 4µ 1+ ⎪ ⎪ kβ ⎪ ⎪ kβ ⎩ −2 kkβz zβ 0 2 ⎭ (10. an axial point load. for that matter.170 Stiffness matrix method for layered media −1 in which K22 = r2 F22 H−1 22 is the stiffness of the cylindrical core.121) At this point.124) To illustrate matters.  & ' p2eq = p2 + r2 F21 − F22 H−1 22 H21 (r1 F11 )−1 p1 (10. k β = k 2S − kz2 .122) n = 1: ⎧ ⎫  2  2  2 ⎪ ⎪ kα kα −3 kkβα kkβz zβ ⎪ ⎪ ⎪ ⎪ kβ kβ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪  2  2  2 ⎪ ⎪ ⎨ ⎬ 1+ kz 1+ kz − 1 + kkβz kz (10. When converted to the axial wavenumber–frequency domain. Thus. .123) lim (r1 F11 )−1 →  π  2  kβ kb z kβ β r1 →0 k ⎪ ⎪ ⎪ ⎪ 8µ 1+ z ⎪ ⎪ ⎪ ⎪ kβ ⎪ ⎪ ⎪  2 ⎪ ⎪ ⎪ ⎪ ⎩ kz kz 2 − kkβz zβ ⎪ ⎭ kβ kβ n ≥ 2: ⎧ ⎫  n+1  n+1  n+1 ⎪ ⎪ n kkβα kα − kα kz ⎪ ⎪ ⎪ ⎪ n z kβ β ⎪ ⎪ ⎪ ⎪ kβ kβ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨  2  2  2 ⎪ ⎬ π zβn−1 lim (r1 F11 )−1 →    n 1 + kkβz n 1 + kkβz − 1 + kkβz a 2 kkβz zβ r1 →0 k 2 ⎪ ⎪ ⎪ ⎪ 2 n!µ 1− α n ⎪ ⎪ ⎪ ⎪ kβ ⎪ ⎪   2 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ n kkβz n kkβz 1 − a2 1 + kz zβ ⎪ ⎭ kβ (10. This shows that the source on the axis can be expressed in terms of a fictitious source applied at the exterior. let’s consider the three particular cases of a torsional point source. on any internal cylindrical surface. so we shall omit them in the ensuing. as will be seen. these terms give unity. which have a spatial–temporal variation δ(z) δ(t). we may generalize the above by allowing K22 to receive stiffness (impedance) contributions from external cylindrical layers. zβ = kβ r1 . kα = k 2P − k 2z . Defining a = β/α. if any. because the source itself will be singular to someorder. we shall give the expressions for the leading terms in zβ = kβ r1 . zα = kα r1 . If the product (r 1 F11 ) p1 remains finite as F−1 11 → O and p 1 → ∞. k P = ω/α. k S = ω/β and using MATLAB’s symbolic toolbox. provided the core is appropriately subdivided into two sub-layers). this system of equations has a finite solution. a point source on the axis can be expressed in terms of fictitious loads on the exterior surface of the core (or.

⎧ ⎫ kβ ⎨ ⎬ 0 −1 lim (r1 F11 ) p1 = 1 (use to form p2eq ) (10.126) that is.128) ⎩ θ ⎭ 2π ⎩ ⎭ −i p˜ z 1  −i π lim (r1 F11 )−1 p1 =   2 r1 →0 2π 4µ 1 + kz kβ ⎧   2 ⎫ ⎪ ⎪ kz ⎪ ⎪⎧ ⎫ ⎪− kβ 1 − kβ −2 kkβα kα kz ⎪ zβ 0 kβ ⎪ ⎪ 0 ⎨  ⎬⎨ ⎬  2 × − 1+ kz 0 ⎪ ⎪ 0 zβ 0 ⎪ ⎪ ⎩ ⎭ ⎪ ⎪ kβ ⎪ 1 ⎪ ⎩ −2 kkβz zβ 0 2 ⎭ (10. which elicits a torsional moment Mt = 2π p˜ θ r1 = 1 (observe that p˜ θ is the traction per radian.10.3 Stiffness matrix method in cylindrical coordinates 171 Torsional point source (n = 0) A torsional point source of unit intensity can be obtained as the limit of a tangential ring load pθ acting on a ring of radius r1 → 0. ⎧   2 ⎫ ⎪ ⎪ ⎧ ⎫ ⎪ ⎪ − kα 1 − kkβz zβ 0 −2 kkβα kz ⎪ ⎪  ⎪ kβ ⎨  kβ ⎪⎪ ⎬ ⎪0⎪ ⎨ ⎪ ⎬ 1 π  2 lim (r1 F11 )−1 p1 =   2 ⎪ 0 − 1 + kkβz zβ 0 1 r1 →0 2πr1 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ 4µ 1 + kkβz ⎪ ⎪ ⎪ ⎪ ⎩ ⎪ ⎭ ⎩ ⎭ 0 −2 kkβz zβ 0 2 (10.125) ⎩ ⎭ 2πr1 ⎩ ⎭ −i p˜ z 0 Hence. which implies an axial load Pz = 2π p˜ z = 1. and multiplication by the radius gives the moment).127) r1 →0 8µ ⎩ ⎭ 0 Axial point load (n = 0) This load can be visualized as the limit of an axial ring traction per radian. Hence. so the total ring load is 2π times larger. Hence. the load vector in the frequency– axial-wavenumber domain for a torsional point source of unit intensity is ⎧ ⎫ ⎧ ⎫ ⎨ p˜ r ⎬ 1 ⎨ ⎬ 0 p1 = p˜ θ = 1 (10. the load vector in the frequency–axial-wavenumber domain for a unit axial load is ⎧ ⎫ ⎧ ⎫ ⎨ p˜ r ⎬ −i ⎨ ⎬ 0 p1 = p˜ = 0 (10.129) . acting on a ring of radius r1 → 0. pz.

ω)dθ = 1 (10.135) ⎪ ⎪ ⎪ ⎪ ⎩ ⎭ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪  2 ⎪ ⎪ ⎪ ⎪ ⎩ kz kz 2 − kkβz zβ ⎪ ⎭ kβ kβ that is.136) r1 →0 kz ⎪ ⎪ ⎪ 16µ 1+ kβ ⎪ ⎩ kz ⎪ ⎭ kβ .132) 0 0 The load vector in the azimuthal wavenumber domain follows from the Fourier transform ⎧ ⎫  2π 1 ⎨ ⎬ 1 p˜ = π1 T1 p(r1 .133) 0 4π ⎩ ⎭ 0 ( 2π (the additional factor 1/2 arises from the integral π1 0 T1 T1 dθ = 12 I). Hence.131) ⎩ ⎭ 2π ⎩ ⎭ ⎩ ⎭ 2π ⎩ ⎭ pz 0 0 cos θ 0 0 Projecting this onto the direction of the load and integrating over the ring. we obtain a unit load in the x direction:  2π  2π   ( pr cos θ − pθ sin θ) r1 dθ = 2πr 1 1 cos2 θ + sin2 θ r1 dθ = 1 (10. kz. ⎧ ⎫ ⎧ ⎫ ⎨ p˜ r ⎬ 1 ⎨ ⎬ 1 p1 = p˜ θ = 1 (10. ⎧  2 ⎫ ⎪ ⎪ kα ⎪ ⎪ ⎪ ⎨ kβ ⎪ ⎬  2 lim (r1 F11 )−1 p1 =  1  2  1 + kkβz ⎪ (use to form p2eq ) (10.172 Stiffness matrix method for layered media that is. ⎧  2 ⎫ ⎪ ⎨ kkβα kz ⎪ ⎬ kβ lim (r1 F11 )−1 p1 =  i  2  0 (use to form p2eq ) (10. θ.130) r1 →0 4µ 1+ kz ⎪ ⎩ ⎪ ⎭ kβ −1 Lateral point load (n = 1) A lateral load of unit intensity can be visualized as the combination of radial and tangential ring loads per radian of equal strength acting on a ring of radius r1 → 0: ⎧ ⎫ ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎨ pr ⎬ cos θ 1 ⎨ 0 0 ⎬⎨1⎬ ⎨1⎬ 1 p = pθ = 0 − sin θ 0 1 = T1 1 (10.134) ⎩ ⎭ 4π ⎩ ⎭ −i p˜ z 0  1 π lim (r1 F11 )−1 p1 =  2 r1 →0 4π 8µ 1 + kkβz ⎧  2  2  2 ⎫ ⎪ ⎪ kα kα −3 kα kz z⎪ ⎪ ⎪ ⎪ kβ kβ kβ kβ β⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎧ ⎫ ⎪ ⎨  2  2  2 ⎪⎨1⎬ ⎬ × 1+ kz kβ 1+ kz kβ − 1 + kβ kz kz z kβ β 1 (10.

138) which must be assembled with Bessel functions of the first kind.4) (10. solid rod composed of N material layers numbered from the outside to the core. Example 10.3) F21 = Fn (Yn (kr2 )) . solid rod subjected to load on axis Consider a layered.e. n = 1 k 2 ⎪ ⎪ ⎪ 16µ 1+ z kβ ⎪ ⎩ kz ⎪ ⎭ kβ H21 = Hn (Yn (kr2 )) .   det K = det r Fn (Hn )−1 = 0 ⇒ det Fn = 0 (10.137) Example 10. inasmuch as r2 = 0 at the axis. we replace the Hankel functions by Bessel functions. we begin by discarding the degrees of freedom associated .9. H22 = Hn (Jn (kr2 )) (Table 10.3 Stiffness matrix method in cylindrical coordinates 173 In summary..10. axial wavenumber kz . As in Example 10. Jn . n = 0 8µ ⎩ ⎭ 0 ⎧  2 ⎫ ⎪ ⎪ kz ⎪ ⎨ kβ kβ ⎪ kα ⎬ qaxis =  i 2  0 Axial point source. n = 0 kz ⎪ ⎪ ⎪ ⎪ 4µ 1+ ⎩ ⎭ kβ −1 ⎧  2 ⎫ ⎪ ⎪ kα ⎪ ⎪ ⎪ ⎨ ⎪  2 ⎬ kβ qaxis = 1    1 + kβ ⎪ kz Lateral point source. a solid core of radius r is characterized by the following properties: Kcore = r Fn H−1 n Stiffness matrix (using Jn )   pequiv = pcore + r F21 − F22 H−1 22 H21 qaxis Total load per radian at external surface qaxis = lim (r1 F11 )−1 p1 Fictitious load due to source on axis r1 →0 ⎧ ⎫ kβ ⎨ ⎬ 0 qaxis = 1 Torsional point source. Hence. i. which is satisfied only if the determinant of the stiffness matrix vanishes.10: Layered. to avoid a singularity on the axis. F22 = Fn (Jn (kr2 )) (Table 10.9: Normal modes of solid rod The dynamic stiffness matrix of a solid rod of radius r1 = r is obtained by disregarding in the formulation all of the elements of the layer matrix associated with the inner radius. and azimuthal index n. Also. the symmetric stiffness matrix is K = r Fn (Hn )−1 (10.139) the roots of which can be obtained by numerical search for any combination of frequency ω. A free vibration problem is characterized by K u = 0.

homogeneous solid with the same material properties as the core. u s*. The strategy is to define an arbitrary cylindrical reference surface within the core (which could be the external surface itself). the details are as follows: Separate the inner and outer systems at the reference surface within the core for both the actual and the infinite solid. see Example 10. For convenience. However. but we could deal with this situation by means of the rigorous method that we presented in Section 10. homogeneous solid (right). and apply the substructuring technique used in Example 10. we shall now demonstrate instead an alternative way of solving this problem.5.10. with the axis by arguing that the inner radius of the core’s cylindrical “layer” is zero.9.5: Layered solid rod subjected to source on axis (left). u* Source Source Figure 10. 10.3.174 Stiffness matrix method for layered media Source Source Exterior Exterior −s − s* Core Core s. so the core of the layered system has the same stiffness matrix as Example 10. However.2d. and modeling via reference solution for infinite. we now have a difficulty in that the source acts directly on the extirpated axis. evaluate at this location the displacements caused by the source in an infinite. choose the reference surface to . With reference to Fig.5.

.. s N .4. These are avoided by adding some damping to the system. K N−1. 10. it follows that the difference in displacements at the reference surface is solely the result of the difference in internal tractions per radian. r N (s N − s∗ ) = Kcore (u N − u∗) with Kcore = r N F N (H N )−1 (Use Bessel functions Jn . etc. we effectively discard the source at the axis. s∗ ..) Hence −r N s N = −Kcore u N + K∗full u∗ with K∗full = Kcore + K∗ext (10. . ⎪ ⎪ ⎪ ⎪ ⎪ .) By taking the difference.140) (Use 2nd Hankel functions. has infinite elements at the resonant frequencies of the core with fixed exterior. which adds to both sides of the equations (i.e.) can be obtained explicitly from the equations presented in Section 4. The source in the actual layered medium elicits at rN displacements and internal stresses u N . u3 = 0 (10. which is bounded by external and internal spherical surfaces of radii Re . As we have seen in Section 1.e. infinite medium elicits displacement and internal stresses u∗ .. .10.. which we define as p¯ e = Re2 s¯ R| R=Re . kz. ⎪⎪ ⎪ ⎪ . the solution to the actual problem follows from ⎧ ⎫⎧ ⎫ ⎧ ⎫ ⎪ K K12 0 ··· 0 ⎪ ⎪ ⎪ 11 ⎪ ⎪ ⎪⎪⎪ u1 ⎪ ⎪ ⎪ ⎪ 0 ⎪ ⎪ ⎪ ⎪ K21 K22 K23 ··· 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ ..4 Stiffness matrix method for layered spheres 175 coincide with the core’s external interface (i.. it follows that the stresses for the homogeneous reference problem satisfy the equilibrium condition   (2) −1 −r N s∗ = Kext u∗ with K∗ext = −r N F N H N (2) (10. ⎪ ⎪ ⎪ ⎪ . The only difficulty here is that Kcore . Also.142) ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ .3.9. Ri . across any arbitrary spherical surface in the layer there exist internal stresses of the form s¯ R = [σ R σ R φ σ R θ ]T = LTR u¯ ∂ 1 ∂ 1 ∂ 1 cot φ = DR R + DRφ + DRθ + DR1 + DR2 Tn Lnm Hm a ∂R R∂ φ R sin φ ∂ θ R R (10. K NN Kfull ). .143) Also.. ω) for any type of axial load (a spatially harmonic line load. on each of the external surfaces act external tractions per unit solid angle.. ⎪ ⎩ ⎭ uN ⎭ ⎩ ∗ ∗⎪ ⎭ 0 0 · · · K N. rN )..N−1 K NN K full u in which K NN receives contribution from Kcore as well as the adjoining external layer. respectively. The reference displacements u∗ = u∗ (r N . Because the core of the actual and the homogeneous problem are identical and so also is the source. . p¯ i = −Ri2 s¯ R| R=Ri (10... . dipoles.N ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ .e. ⎪ ⎪ .144) .. ⎪ ⎪⎪⎪ . ⎪⎪ ⎬ ⎨ u2 ⎬ ⎪ ⎪ ⎨ 0 ⎪ ⎬ 0 K32 K33 . whereas the same source in the homogeneous. a line of pressure. i. since the exterior region is free from sources.4 Stiffness matrix method for layered spheres Consider an elastic spherical layer of arbitrary thickness.141) Finally.

145) a2 . Ri . m) (not a function of n) (i) (i) (i) (i) Fmj ≡ Fm (Rj ) Hmj ≡ Hm (Rj ) $ % Kee Kei Spherical layer: (6 × 6) K = Kie Kii $ (1) (2) %$ (1) (2) %−1 Re2 Fme Re2 Fme Hme Hme = (1) (2) (1) (2) −Ri2 Fmi −Ri2 Fmi Hmi Hmi Solid sphere: (3 × 3) K = R2 Fm (Hm)−1 Core (use jm in Fm. ω) dθ dφ traction per steradian u˜ = K−1 p˜ ) ∞ ) m ) ∞ ) m u(R. ω) = ¯ u(R.2 on how to make them symmetric. n as   ¯ u(R. and with reference to Tables 10. p = pφ = tractions per steradian ⎩ ⎭ ⎩ ⎭ uθ pθ ⎧ ⎫ π(1 + δn0 ) (n + m)! ⎨ ⎬ 1 0 0 J= 0 m(m + 1) 0 (m + 12 ) (m − n)! ⎩ ⎭ 0 0 m(m + 1) The negative sign in the second term comes from the fact that external tractions are opposite in direction to the internal stresses at the inner surface.8.7 and 10.6. θ. φ. we write the displacement and traction vectors for specific indices m.7 and the results in Section 9.4. Hm)   (2) (2) −1 (2) Cavity: (3 × 3) K = −R2 Fm Hm Infinite external space (use hm for matrices) Wavenumber–space transform: (π +( . m. for greater computational efficiency. ˜ m. ω) = J−1 0 sin φ Lnm 0 Tn p(R. ω.a  1 = Tn Lnm H(1) m H(2) m (10. n. Considering that the wave field in the layer is composed of both outgoing and ingoing waves. Re = external radius Ri = internal radius R = generic radius p˜ = Ku˜ K = K(Re . 10.176 Stiffness matrix method for layered media Table 10. n. see Tables 10. θ. θ.6. µ. ω) = Tn Lnmu(R. φ. ω) = Tn Lnm H(1)m a1 + Hm a2 (2) + . See Section 10. 2π ˜ p(R. solve separately for the SH and SVP degrees of freedom in K when solving for u. Stiffness matrices in spherical coordinates Note: Stiffness matrices given below are not symmetric. For definition of the submatrices below. Also. m. ˜ since they are uncoupled. ω) m=0 n=0 m=0 n=0 with ⎧ ⎫ ⎧ ⎫ ⎨ uR ⎬ ⎨ pR ⎬ u= uφ = displacements.3. n. φ. λ.

146) a2 with subscripted matrices T.7. n. Spheroidal. F defined in Table 10. Fm → hPm = hm (zP ).m+1 − (m − 1) zS Hm. H.m+1 − (m + 1) (m − 1) zS zS  2µ h Sm f12 = −kS m (m + 1) hS. the numbers in parenthesis refer (1) to the kind of spherical Bessel functions used to construct the matrices: (1) refers to hm . hSm = hm (zS ) (use 2nd spherical Hankel functions) ω ω zP = kP R ≡ P . kS = α β   ¯ p(R.7.m+1 + m (m − 1) zP zP  2µ hSm f22 = −kS µ hSm − hS. azimuthal and spherical Bessel matrices ⎧ n ⎫ ⎪ ⎪ Pm 0 0 ⎪ ⎪ ⎪ ⎪ d Pmn n Pmn ⎪ ⎪ ⎨ 0 ⎬ Lnm = dφ sin φ . hSm = hm (zS ) (use 1st spherical Hankel functions) (2) (2) (2) (2) Hm . φ. (2) and (2) to hm . Fm = f21 f22 0 ⎪ ⎪ 0 ⎪ ⎪ ⎩ ⎭ ⎪ ⎪ zP zS dzS ⎪ ⎪ 0 0 f33 ⎪ ⎩ ⎪ ⎭ 0 0 h Sm  2µ hPm f11 = −kP (λ + 2µ) hPm − 2hP. Pmn = Pmn (φ) = associated Legendre function ⎪ ⎪ ⎪ ⎪ ⎪ ⎩ 0 n Pmn d Pmn ⎪ ⎪ ⎪ ⎭ sin φ dφ Tn = diag[ cos nθ cos nθ − sin nθ ] or Tn = diag[ sin nθ sin nθ cos nθ ] ⎧ dh hSm ⎫ ⎪ ⎪ Pm m(m + 1) 0 ⎪ ⎪ ⎪ ⎪ dz z ⎪ ⎪ ⎧ ⎫ ⎪ ⎨ P S ⎪ ⎬ ⎨ f11 f12 0 ⎬ Hm = hPm 1 d(zS h Sm) .10. Also. kP = . hSm = jm(zS ) (use spherical Bessel functions) (1) (1) (1) (1) Hm . ω) = Tn Lnm F(1) m a1 + Fm a2 (2) + .m+1 − (m − 1) zS zS  2µ hPm f21 = −kP hP. Fm → h Pm = hm (zP ). θ. and the overbar being a reminder that this is a particular solution for given m. spherical Bessel functions of the first and second kind can be . Fm → hPm = jm(zP ).m+1 − (m − 1) zP zP  hSm f33 = −kS µ hS. Alternatively.a  1 = Tn Lnm F(1) m F(2) m (10. zS = kS R ≡ S .4 Stiffness matrix method for layered spheres 177 Table 10.

8. Evaluating these expressions at the inner and outer surfaces. L11 = − 0 cos φ 1 ⎩ ⎭ ⎩ ⎭ 0 0 − sin φ 0 1 cos φ ⎧ ⎫ 3 cos 2φ + 1 1⎨ ⎬ 0 0 L02 = 0 −6 sin 2φ 0 4⎩ ⎭ 0 0 −6 sin 2φ ⎧ ⎫ 3⎨ ⎬ sin 2φ 0 0 L12 = − 0 2 cos 2φ 2 cos φ 2⎩ ⎭ 0 2 cos φ 2 cos 2φ ⎧ ⎫ 1 − cos 2φ 3⎨ 0 0 ⎬ L22 = 0 2 sin 2φ 4 sin φ 2⎩ ⎭ 0 4 sin φ 2 sin 2φ ⎧ ⎫ 5 cos 3φ + 3 cos φ 1⎨ ⎬ 0 0 L03 = 0 −3 (5 sin 3φ + sin φ) 0 8⎩ ⎭ 0 0 −3 (5 sin 3φ + sin φ) ⎧ ⎫ sin φ + 5 sin 3φ 3⎨ ⎬ 0 0 L13 = − 0 cos φ + 15 cos 3φ 6 + 10 cos 2φ 8⎩ ⎭ 0 6 + 10 cos 2φ cos φ + 15 cos 3φ ⎧ ⎫ cos φ − cos 3φ 15 ⎨ ⎬ 0 0 L23 = 0 3 sin 3φ − sin φ 4 sin 2φ 4 ⎩ ⎭ 0 4 sin 2φ 3 sin 3φ − sin φ ⎧ ⎫ 3 sin φ − sin 3φ 15 ⎨ ⎬ 0 0 L33 = − 0 3 (cos φ − cos 3φ) 6 (1 − cos 2φ) 4 ⎩ ⎭ 0 6 (1 − cos 2φ) 3 (cos φ − cos 3φ) used. i. and adding appropriate sub-indices e. we can write $ % $ % $ (1) (2) %$ % u¯ e Tn Lnm 0 Hme Hme a1 = (1) (2) (10. to identify these surfaces. The spheroidal or co-latitude matrices up to order and rank 3 ⎧ ⎫ ⎨1 ⎬ L00 = 0 ⎩ ⎭ 0 ⎧ ⎫ ⎧ ⎫ ⎨ cos φ 0 0 ⎬ ⎨ sin φ 0 0 ⎬ L01 = 0 − sin φ 0 .178 Stiffness matrix method for layered media Table 10.147) u¯ i 0 Tn Lnm Hmi Hmi a2 and $ % $ %$ (1) (2) %$ % p¯ e Tn Lnm 0 Re2 Fme Re2 Fme a1 = (1) (2) (10.148) p¯ i 0 Tn Lnm −Ri2 Fmi −Ri2 Fmi a2 .

151) Kie Kii u˜ i or more compactly p˜ = Ku˜ (10. .152) K is the dynamic stiffness or impedance matrix. . namely the quantities with tilde. and indeed.. ˜ and from here the actual displacements by appropriate multiplication by the azimuthal and co-latitude matrices. . . ⎪ . and then applying the azimuthal and co-latitude transfor- mations to obtain the actual displacements. More generally. . in the case of a layered sphere composed of N−1 layers whose N interfaces have radii R1 .n ⎪ ⎪ ⎪ ⎪ ⎪ ..4 Stiffness matrix method for layered spheres 179 Next. we can construct the system impedance matrix by appropriately overlapping the layer impedance matrices. to all concentric surfaces in the case of a layered sphere. = (1) (2) (10. so they can be factored out for the system as a whole. ⎪ ⎪ ⎪ . we obtain $ % $ (1) (2) %$ (1) (2) %−1 $ % p˜ e Re2 Fme Re2 Fme Hme Hme u˜ e = (1) (2) (1) (2) p˜ i −Ri2 Fmi −Ri2 Fmi Hmi Hmi u˜ i $ %$ % Kee Kei u˜ e = (10. ⎪ ⎪ ⎪ ⎪ ⎪ ⎪⎪⎪ ⎪ . ⎪ ⎪ ⎪ ⎩ ⎭ ⎪ ⎪ ⎪ .10. for given tractions p.150) p¯ i 0 Tn Lnm p˜ i p˜ i −Ri2 Fmi −Ri2 Fmi a2 which express the displacement and load vectors at the two interfaces in terms of local amplitudes.. The end result is of the block-tridiagonal form ⎧ ⎫ ⎧K ··· ⎫⎧ ⎫ ⎪ p˜ 1 ⎪ ⎪ K12 0 0⎪ u˜ 1 ⎪ ⎪ ⎪ ⎪⎪ ⎪ 11 ⎪ ⎪ ⎪K ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ p˜ 2 ⎪ ⎪ 21 K22 K23 ··· 0⎪ ⎪ ⎪ ⎪ u˜ 2 ⎪ ⎪ ⎨ ⎪ ⎬ ⎪ ⎨ .. .R2 . Kn−1. Hence. ⎬⎪ ⎪ . Observe that the azimuthal and co-latitude matrices are common to both surfaces. . ⎨ ⎪ ⎬ p˜ 3 = 0 K32 K33 .. ⎪ ⎪ p˜ n ⎩ ⎭ ⎩ u˜ ⎭ 0 0 ··· Kn. u˜ 3 (10..n−1 Knn n which again allows obtaining the displacement amplitudes from the load amplitudes by solving the system of equations. . . Eliminating the integration constants between the traction and displacement amplitudes. ˜ we can obtain the displacements u. ⎪⎪ ⎪ ⎪ .153) ⎪ ⎪ . . we define $ % $ %$ % $ % $ (1) (2) %$ % u¯ e Tn Lnm 0 u˜ 1 u˜ e Hme Hme a1 = . . = (1) (2) (10.149) u¯ i 0 Tn Lnm u˜ 2 u˜ i Hmi Hmi a2 and $ % $ %$ % $ % $ (1) (2) %$ % p¯ e Tn Lnm 0 p˜ e p˜ e Re2 Fme Re2 Fme a1 = . which relates the amplitudes of the external tractions applied at both surfaces to the amplitudes of displacements observed at these locations. RN and the interfaces are numbered from the outside to the inside. ..

the 3 × 3 matrix of a solid sphere of radius R (or of the core of a layered system) must be constructed with spherical Bessel functions of the first kind jm so as to avoid a singularity on the axis: u˜ = Hmc. depends solely on the index m. but they are not recommended here for numerical computation. 10.1 Properties and use of impedance matrices The impedance matrix for any layer is a function of the inner and outer radii. because they may become quasi-linearly dependent for complex.2 Asymmetry As written. the frequency of excitation.180 Stiffness matrix method for layered media 10. is given by a 3 × 3 matrix:  −1 u˜ = H(2) m c.4. As we have seen. which is the radial component. and especially negative imaginary. all elements of H except for h11 can be set to zero. Here. and in fact. m = 0 (10. K = R2 Fm (Hm)−1 (core) (10. The impedance for the outer region. one defines the scaling factor 1 ξ= . it is independent of the azimuthal index n. The case m = 0 is also special. the off-diagonal element h12 of H is zero. hm to construct this matrix. K = −R2 F(2) m H(2) m (infinite exterior) (10. the impedance matrices for each layer (and therefore. Also. the system matrix) are not symmetric. the co-latitude matrix L00 is zero except for the first diagonal element.156) The imaginary part in the resulting impedance matrix represents the radiation damping in the outer region. so as to satisfy the radiation condition at infinity. all other degrees of freedom should be removed before computation. they can be brought into a symmetric form by a similarity transformation that. fortunately. Fm must be constructed using second spherical Hankel functions hm (kR). Hence. the material properties of the layer. however. as seen from the surface of the cavity. the above outer region matrix must be added to the element K11 to complete the global system matrix. arguments. p˜ = −R2 F(2) m c. ym could be used in place of the spherical (1) (2) Hankel functions hm .157) m(m + 1) .154) Kie Kii −Ri2 Fmi −Ri2 Fmi Hmi Hmi In principle. in the case of a spherical cavity of radius R in an infinite space. and the radial index m. spherical Bessel functions jm. On the other hand.155) By contrast.4. the 6 × 6 stiffness matrix for a spherical layer is of the form $ % $ (1) (2) %$ (1) (2) %−1 Kee Kei Re2 Fme Re2 Fme Hme Hme K= = (1) (2) (1) (2) (layer) (10. the component (2) matrices Hm. If desired. However. For a system of layers surrounded by a homogeneous infinite space. It follows that each interface is characterized by a single degree of freedom. For each component matrix. p˜ = R2 Fmc.

which we assume to be defined over a spherical surface of constant radius R. Thus. Expansion into spheroidal harmonics yields  ∞  m p= Tn Lnmp˜ mn (10. The advantage of using a symmetric stiffness matrix lies in the saving in computational effort. and write the system equation simply as p˜ = K ˜ u˜ ⇒ ˜ −1 p˜ u˜ = K (10.10. if any.160) = Tn Ln Q−1 H m ˜ mc m and p¯ = Tn Lnm Q−1 m Qm Fm c p¯ = Tn Lnm Q−1 m p˜ ⇒ (10. to avoid proliferation of symbols.163) In this equation. p˜ ≡ p˜ mn (R. θ. we obtain a symmetric global system matrix.4 Stiffness matrix method for layered spheres 181 such that ⎧ ⎫⎧ ⎫ ⎨ξ ⎬ ⎨ f11 f12 0 ⎬ F˜ m = Qm Fm = 1 f f22 0 (10. the scaled. must be expressed in terms of spheroidal harmonics of order m.159) ⎩ ⎭ ⎩ 21 ⎭ 1 0 0 h33 which implies u¯ = Tn Lnm Q−1 m Qm H m c ⇒ u¯ = Tn Lnm Q−1 m u˜ (10. we continue labeling the scaled vectors for the system as we did for any individual layer. in the case of a system of layers.3 Expansion of source and displacements into spherical harmonics The stiffness matrix method described herein is based on a formulation in the frequency– spheroidal-wavenumber domain..161) = Tn Ln Q−1 F˜ m c m m = Tn Ln Q−1 K m ˜ u˜ m For example. 10. every third (SH) degree of freedom is uncoupled from the preceding two (SV-P) degrees of freedom. More generally. non-symmetric form. φ.162) m which relates the scaled force and displacement amplitudes vectors at the interface of the outer region. the sources p = p(R.n. i. ω) Let p be the 3 × 1 vector of external tractions per solid angle at a given frequency. However. symmetric impedance matrix would be  −1 K ˜m ˜ = −R2 F˜ m H ⇒ ˜ = Qm K Q−1 K (10.164) m=0 n=0 .4. we present the ensuing examples in the standard. for an exterior region. For notational transparency. so they should be solved separately.e.158) ⎩ ⎭ ⎩ 21 ⎭ 1 0 0 f33 ⎧ ⎫⎧ ⎫ ⎨ξ ⎬ ⎨ h11 h12 0 ⎬ ˜ m = Qm Hm = H 1 h h22 0 (10. however. which is a factor of two. ω).

(10. namely three translations and three rotations. the displacements in the physical domain are obtained for each interface as  ∞  m u= Tn Lnmu˜ mn (10.173) .4 Rigid body spheroidal modes An unconstrained.170) Rotations (2) (1) (2) x = c T1 L11 eˆ 3 .165) 0 0 with ⎧ ⎫ π(1 + δn0 ) (n + m)! ⎨ ⎬ 1 0 0 J= 0 m(m + 1) 0 (10. and     n = diag cos nθ T(1) cos nθ − sin nθ .182 Stiffness matrix method for layered media whose inversion is  π  2π  p˜ mn = J−1 sin φ Lnm Tn p dθ dφ (10.172)  T  T eˆ 12 = 1 1 0 .  y = c T1 L11 eˆ 3 .169) m=0 n=0 Application of these equations is demonstrated in Example 10. n = diag sin nθ T(2) sin nθ cos nθ (10.  π Lnm Lnm sin φ dφ = J (10. spherical system of finite size admits six rigid body modes that produce no net stresses within the system.171) in which c is an arbitrary constant. Uy = c T1 L11 eˆ 12 . eˆ 3 = 0 0 1 (10. x = c T0 L01 eˆ 3 (10. 10.168) 0 Having found the displacements u˜ ≡ u˜ mn (ω) in the spheroidal domain.167) 0  2π Tn T j dθ = π δ(nj) (1 + δn0 δ j0 ).13. Uz = c T0 L01 eˆ 12 (10.166) (m + 2 ) (m − n)! ⎩ 0 1 0 m(m + 1) ⎭ This is based on the orthogonality conditions of the spheroidal and azimuthal matrices. They are: Translations (1) (2) (1) Ux = c T1 L11 eˆ 12 .4.

10. m>0 ⇒ torsional modes (10. the spheroidal (SV-P) degrees of freedom. Nontrivial solutions of this problem exist if det K(ω) = 0 ⇒ det Fm = 0 (10. m=0 f33 = 0. they should be solved separately.177) the roots of which can be obtained by numerical search for any given parameter m. we obtain the free vibration problem Ku˜ = 0. inasmuch as Ri = 0. we consider the special case of a unit point load in the radial direction at location R0 . u˜ are the displacements on the surface of the sphere. one can then proceed to find the vibration modes in transformed space by solving the equation Ku˜ = 0 while assigning an arbitrary value to one component ˜ say the first. The result is of the form p˜ = Ku. Having found the frequencies. The inner core. Example 10. Observe that the natural frequencies do not depend on the azimuthal index n. at the center (i. the inner surface shrinks to a point).174) in which the 3 × 3 matrices used to form K are assembled with spherical functions of the first kind. the core matrix must be left out. m>0 ⇒ spheroidal modes (10.4 Stiffness matrix method for layered spheres 183 Example 10.e. the actual modes in physical space are obtained by applying the co-latitude and azimuthal matrices to the vectors thus found. since the torsional (SH) degree of freedom is uncoupled from of u. is characterized by the same 3 × 3 stiffness matrix of the previous exam- ple.11: Free vibration modes of a solid sphere The dynamic stiffness matrix of the solid sphere is obtained from the preceding by dis- regarding the terms associated with the inner surface. because K does not depend on n.175) or in full  f11 f22 − f12 f21 = 0. θ 0 .13: Response to applied load We examine next the case of a layered sphere being acted upon by a harmonic load with some known spatial distribution. Again. To keep the presentation simple.. Having found the modes in transformed space. and p˜ are the exter- nal tractions per solid angle. When these tractions vanish.12: Free vibration modes of a layered sphere This is similar to the preceding example. whereas the outer layers have matrices of size 6 × 6. Example 10.176) f11 = 0. if any. In the case of a hollow layered sphere. φ 0 . and then search for the roots of det K = 0. This can be expressed . ˜ K = Re2 Fm (Hm)−1 (10. It suffices to set up the system matrix by appro- priately overlapping the layer matrices.

. we augment the vector e with sufficient number of zeros to match the size of the problem. Also. Thereafter.179) 4π 2 (1 + δn0 ) (n + m)! sin nθ0 Observe that the load has both symmetric and antisymmetric spheroidal components as indicated by the two alternative variations with the azimuth. This is the result of using herein only spheroidal harmonics of the first kind. for if this were not the case. we can always assume that the interface at radial distance R (where the load is applied) coincides with one of the interfaces of the layers of the problem at hand.184 Stiffness matrix method for layered media as a radial traction per solid angle of the form ⎧ ⎫ δ(φ − φ0 ) δ(θ − θ0 ) ⎨1⎬ pR = .180) m=0 n=0 Observe that K does not depend on n. and proceed to solve for the displacements as  ∞  m u˜ = K−1 p˜ ⇒ u= Tn Lnmu˜ (10. it would suffice to subdivide the physical layer in which the load resides into two sub-layers with identical material properties. e= 0 (10.178) 4π ⎩ ⎭ 0 The load vector in transformed space is then  π  2π p˜ = J−1 sin φ Lnm Tn p dθ dφ = 4π 1 sin φ0 J−1 Lnm(φ0 ) Tn (θ0 ) e 0 0  (m + 12 ) (m − n)! cos nθ0 = sin φ0 Pmn (φ0 ) e (10. Now. p = pRe. notice that the radial load cannot be located at either the north or the south pole. because in such a case sin φ0 = 0.

. 11. These functions behave asymptotically as complex exponentials exp(±ikr ). or away from it (second kind). We refer in the ensuing to any of the Bessel functions by means of the symbol Cn = Cn (z). 1980. and Jn (kr ). Academic Press. Hankel functions can be interpreted as waves that propagate toward the origin (first kind). 1970.1) dx 2 r dx r2 This equation has solutions of the form y = c1 Jn (kr ) + c2 Yn (kr ).4) dz 2 1 Extensive tables and properties for mathematical functions can be found in the following two references: a) Abramovitz.1. 185 .1. with c1 . in combination with a harmonic term of the form exp(+iωt). Table of integrals.1 Bessel functions 11. A. Alternatively. Yn (kr ) the Bessel functions of order n and of the first and second kind. I.3) z 2 dCn 1 = (Cn−1 − Cn+1 ) (11. Handbook of mathematical functions.1 Differential equation  d2 y 1 dy n2 + + k2 − y=0 (11. c2 arbitrary con- stants. so they are often used in wave propagation problems. SECTION VI: APPENDICES 11 Basic properties of mathematical functions1 11. and Stegun. I.2) (1) (2) in which Hn = Jn + i Yn and = Jn − i Yn are the Hankel functions of the first and Hn second kind (or Bessel functions of the third kind). which is shorthand for cylindrical functions with a (generally complex) argument z = kr . b) Gradshteyn. solutions can be written in the form y = c1 Hn(1) (kr ) + c2 Hn(2) (kr ) (11. series and products.. respectively (some books write Nn (kr ) instead of Yn (kr ) and refer to it as the Neumann function). National Bureau of Standards.2 Recurrence relations n 1 Cn = (Cn−1 + Cn+1 ) (11. and Ryzhik. Thus. I. M.

To a first approximation. 3. 1. (11. and znj = knj R.1: Bessel functions Jn (x).3 Derivatives  dkCn 1  k = k (−1) j kj Cn−k+2 j .4 Wronskians 2 Jn+1 (z) Yn (z) − Jn (z) Yn+1 (z) = (11. the zeros are given by znj = knj R = these π j + 12 n − 14 . Y1 Y2 -0.10) 0 2 dz z=knj R When n = 0. 2.5 Orthogonality conditions First condition If Jn (znj ) = 0 are the zeros of Jn . . .5) dzk 2 j=0  k 1 d (z n Cn ) = z n−kCn−k (11. j = 1.5 0 Figure 11. . k = 0. Y0 Yn (x). . (11. . 2. .186 Basic properties of mathematical functions 1. there exists a zero zn0 = 0 that leads to the trivial condition 0 = 0.6) z dz  k 1 d   z−n Cn = (−1)k z−n−kCn+k (11.9) πz 11.1.8) πz (2) (1) 4i Hn(1) (z) Hn+1 − Hn+1 Hn(2) (z) = (11.5 -1. then   * 2 R 1 2 d * Jn (kni r ) Jn (knj r ) r dr = R Jn (z)** δi j .1. .7) z dz 11.0 0 5 x 10 15 11. so  must be excluded. i.1.0 J0 J1 J2 0.

14) 0 2 α dr * αR r =R  R R [Jn (α R) Yn (β R) − Jn (α R) Yn (β R)] − 2 π (α/β)n Jn (αr ) Yn (βr ) r dr = .6 Useful integrals  R R [Jn (α R) Jn (β R) − Jn (α R) Jn (β R)] Jn (αr ) Jn (βr ) r dr = . α = β (11.2 Spherical Bessel functions 187 Second condition * d If dz Jn (z)*z = 0 are the zeros of the derivative of Jn and znj   = knj R.1 Differential equation ∂2 y 2 ∂y m(m + 1) + + k2 − y=0 (11.13) 0 α2 − β 2  R $ *  n 2 % R2 1 d Jn (αr ) 2 ** Jn2 (αr ) r dr = * + 1− Jn2 (α R) (11.11.2 Spherical Bessel functions 11. α = β 0 α2 − β 2 (11.1. there exists a zero z00 = 0 that satisfies  R J02 (0) r dr = 12 R2 (11. 2.16) αr r =R α π  * b r [Yn (αr ) Yn (βr ) − Yn (αr ) Yn (βr )] **b Yn (αr ) Yn (βr ) r dr = * .17) a α −β 2 2 a $ 2 %*b  b r2  n 2 * 1 dYn (αr ) * Yn2 (αr ) r dr = + 1− Yn2 (αr ) * . j = 1.2. α = β. i. znj = 0.15)  R 2   r 1 d Jn (αr ) dYn (αr ) Jn (αr ) Yn (αr ) r dr = 0 2 α2 dr dr  n 2 * #* n + 1− Jn (αr ) Yn (αr ) ** − 2 (11. (11.11) 0 2 kni R   The zeros znj d of dz Jn (z) interlace the zeros znj of Jn (z).12) 0 11. a > 0 (11.19) ∂R2 R∂R R2 .18) 11. . α = β. . In addition. When n = 0. . a>0 a 2 α dr αr * a (11.  for n = 0. then nj     R   1 2 2  n 2 Jn (kni r ) Jn (knj r ) r dr = R Jn (kni R) 1 −  δi j .

5 j1 j2 0 Figure 11.2. Alternatively. j1 (z) = 2 − . For integer m. second. c2 arbitrary constants. h2 (z) = i 3 + 3 i z − z2 (11.0 j0 0.2: Spherical Bessel functions jn (x). yn (x). respectively. and jm(kr ). ym(z) = Ym+1/2 (z) (11.23) z z2 z z z z (2) e−iz (2) e−iz (2) e−iz   h0 (z) = i . y0 y 1 y2 -0. The first three spherical Bessel functions of the first.22) z z z z z z  cos z cos z sin z 1 3 3 y0 (z) = − . y1 (z) = − − . these functions relate to the Bessel functions of half-integer order n = m + 12 as ! ! π π jm(z) = Jm+1/2 (z). with c1 .2 Trigonometric representations The spherical Bessel functions can be represented in terms of elementary functions.188 Basic properties of mathematical functions 1. ym(kr ) the spherical Bessel functions of order m and of the first and second kind. and generally complex argument z = kr . h1 (z) = i (1 + iz) .24) z z2 z3 . with complex argument z are  sin z sin z cos z 3 1 3 j0 (z) = .21) 2z 2z 11.20) (1) (2) in which hm = jm + i ym and hm = jm − i ym are the spherical Hankel functions of the first and second kind (or spherical Bessel functions of the third kind). j2 (z) = 3 − sin z − 2 cos z (11. solutions can be written in the form m (kr ) + c2 hm (kr ) y = c1 h(1) (2) (11. and third kinds.0 0 5 x 10 15 This equation has solutions of the form y = c1 jm(kr ) + c2 ym(kr ).5 -1. y2 (z) = − 3 cos z − 2 sin z (11.

. |x| ≤ 1 (11. 2.11.2. and Pm(x) the Legendre polynomial of order m. . The trigonometric form of the differential equation is d2 Pm d Pm + cot φ + m(m + 1) Pm = 0 (11. . . m = 0.8 1.0 P2 P3 0.30) dφ 2 dφ . -0. 1. c2 being arbitrary constants. 1.29) dx 2 dx For non-negative integer m. .     1 d m sin z 1 d 1 d 1 d sin z jm(z) = (−z)m = (−z)m ··· . x = cos φ. with c1 .3 Legendre polynomials 11. .1 Differential equation d2 y dy (1 − x 2 ) − 2x + m(m + 1)y = 0. the solution to this equation is y = c1 Pm + c2 Qm. they vanish at the origin except for j0 .5 P2 -1.3: Legendre polynomials.28) dz 11.0 πφ 1 The functions jm satisfy jm(0) = δm0 . 2.27) z dsm(z) (2m + 1) = msm−1 (z) − (m + 1) sm+1 (z) (11. that is. More generally.26) z dz z 11.3 Legendre polynomials 189 1. the ym functions are singular at z = 0. z dz z z dz z dz z dz z (11. Also.3 Recurrence relations sm(z) = any of the spherical Bessel functions 2m + 1 sm(z) = sm−1 (z) + sm+1 (z) (11. m = 0.3.6 0.5 P4 P5 0 Figure 11. In most applications. The second solution Qm(x) to the Legendre differential equation is used much less often.0 0 0. .25)  m 1 d cos z ym(z) = −(−z)m (11.2 0.4 0.

the orthogonality condition is  π 2 Pm(cos φ) Pn (cos φ) sin φ dφ = δmn (11. P3 = (5 cos 3φ + 3 cos φ) (11. 2mm! dx m 2 (11.3 Trigonometric expansion (x = cos φ) (2m − 1)!! 1 m Pm(φ) = 2 cos mφ + cos(m − 2)φ (2m)!! 1 (2m − 1) 1×3 m (m − 1) + cos(m − 4)φ + · · · (11.3. P1 (x) = x.5 Orthogonality condition  +1 2 Pm(x) Pn (x) dx = δmn (11.31) Observe that Pm(1) = 1 and Pm(−1) = (−1)m.3.34) 11. P2 (x) = (3x 2 − 1). (2m − 1)!! = 1 × 3 × 5 × · · · × (2m − 1) and (2m)!! = 2 × 4 × 6 × · · · × 2m. P5 = (63 cos 5φ + 35 cos 3φ + 30 cos φ) 64 128 (11. with x = cos φ. P2 = (3 cos 2φ + 1) .4 Recurrence relations (m + 1)Pm+1 = (2m + 1)x Pm − m Pm−1 (simplest way to find Pm) (11.32) 1 × 2 (2m − 1) (2m − 3) Divide the last term by 2 if m is even! Here. P1 = cos φ.38) −1 2m + 1 In its trigonometric form. P0 (x) = 1.37) dφ 2m + 1 11. The first six Legendre polynomials in trigonometric form are 1 1 P0 = 1.36) dx 2m + 1 d Pm m (m + 1) sin φ = m (cos φ Pm − Pm−1 ) = (Pm+1 − Pm−1 ) (11. 11.3. .190 Basic properties of mathematical functions 11.2 Rodrigues’s formula (−1)m dm(1 − x 2 )m 1 Pm(x) = .35) d Pm m (m + 1) (1 − x 2 ) = m (Pm−1 − x Pm) = (Pm−1 − Pm+1 ) (11. .39) 0 2m + 1 .3.33) 4 8 1 1 P4 = (35 cos 4φ + 20 cos 2φ + 9) . .

. . dm = m + 12 g(φ) Pm(φ) sin φ dφ (11.4: Associated Legendre functions. Qmn the associated Legendre functions or spheroidal harmonics of the first and second kind.1 Differential equation d2 y dy n2 (1 − x 2 ) 2 − 2x + m(m + 1) − y=0 dx dx 1 − x2 m = 0. . respectively. -0. The first of these functions is related to the Legendre polynomials as dn Pm(x) Pmn (x) = (−1)n (1 − x 2 )n/2 (11. Also.0 4 9 P31 1 15 P32 P30 0. so it has no derivatives higher than the mth.42) The solution to this differential equation is y = + with c1 . 2. 1.4 Associated Legendre functions (spheroidal harmonics) 191 1. and Pmn .6 Expansion in Legendre series For any function f (x) defined in the interval −1 ≤ x ≤ 1 that satisfies the so-called Dirich- let conditions (most practical functions do). . .5 1 15 P33 -1. 1. n = 0.41) m=0 0 When the series is truncated at some fixed m. one can express it in a Legendre series of the form  ∞    +1 f (x) = cm Pm(x).0 0 0.43) dx n Recall that Pm is a polynomial of order m. 11. . constants. Pm0 = Pm.8 1.6 0. .3.5 0 Figure 11. cm = m + 12 f (x) Pm(x) dx (11.0 πφ 1 11.11. which implies n ≤ m. . 2. |x| ≤ 1 (11. .40) m=0 −1  ∞  π   g(φ) = dm Pm(φ).2 0.4.4 0. c2 arbitrary c1 Pmn (x) c2 Qmn (x). the resulting mth degree polynomial provides the best fit to f(x) in the least squares sense. n ≤ m.4 Associated Legendre functions (spheroidal harmonics) 11.

tesseral harmonics for m > n.192 Basic properties of mathematical functions The related functions cos nθ Pmn (cos φ) and sin nθ Pmn (cos φ) are called surface harmonics of the first kind – in particular.48) −1 (m + 12 ) (m − n)! ⎧  +1 ⎨ (n + m)! 2 −1 n 0<n=k≤m (1 − x ) Pm (x) Pm (x) dx = n (m − n)! k (varying order) −1 ⎩ 0 n = k (11. we obtain the trigonometric form of the differential equation. these orthogonality conditions are  π (n + m)! Pmn (φ) Pkn (φ) sin φ dφ =   δmk (varying degree) (11. and sectoral harmonics for m = n.44) dφ 2 dφ sin φ 2 11.2 Recurrence relations (m − n + 1)Pm+1 n = (2m + 1) x Pmn − (m + n) Pm−1 n (11.49) In their trigonometric form.51) ⎩ 0 0 n = k .50) 0 m + 12 (m − n)! ⎧  π ⎨ (n + m)! 0<n=k≤m Pmn (φ) Pmk(φ) dφ = n (m − n)! (varying order) (11.4.3 Orthogonality conditions  +1 (n + m)! Pmn (x) Pkn (x) dx = δmk (varying degree) (11.46) dx d Pmn m+ n n = m cot φ Pmn − P .4.45) dPn (1 − x 2 ) m = (m + n) Pm−1 n − m x Pmn (11.47) dφ sin φ m−1 d Pmn m− n+ 1 n = Pm+1 − (m + 1) cot φ Pmn dφ sin φ List of spheroidal harmonics Pmn up to third degree and order n\m 0 1 2 3 0 P00 =1 P10 = cos φ P20 = 1 4 (1 + 3 cos 2φ) P30 = 1 8 (5 cos 3φ + 3 cos φ) 1 — P11 = −sin φ P21 = − 32 sin 2φ P31 = − 38 (sin φ + 5 sin 3φ) 2 — — P22 = 3 2 (1 − cos 2φ) P32 = 15 4 (cos φ − cos 3φ) 3 −/+ — — P33 = − 15 4 (3 sin φ − sin 3φ) 11. Setting x = cos φ. which is d2 Pmn d Pn n2 + cot φ m + m (m + 1) − Pmn = 0 (11. (11.

58) 1×2×3 This rule is often useful in the context of Legendre functions.55) 0 dφ dφ 11.4. Observe that the coefficients are the same as those of the binomial expansion.  π (n + m)! " # Lnm Lnk sin φ dφ = δmk   diag 1 m(m + 1) m(m + 1) (11.4 Orthogonality of co-latitude matrix ⎧ ⎫ ⎪ ⎪Pmn 0 0 ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ ⎨ d Pmn n Pmn ⎪ ⎪ ⎬ 0 Lm = n dφ sin φ (11.5 Expansion of arbitrary function in spheroidal harmonics  ∞  +1   f (x) = cm Pmn (x). dm = m + 12 (m−n)! (m+n)! g(φ) Pkn (φ) sin φ dφ (11. Also.53) 0 m+ 1 2 (m − n)! This is based on the integral  π d Pmn d Pkn n2 m (m + 1) (n + m)! + Pmn Pkn sin φ dφ =   δmk (11.4 Associated Legendre functions (spheroidal harmonics) 193 11. . for example when f (x) = (1 − x 2 )k.4. and applying the orthogonality condition.54) 0 dφ dφ sin φ 2 m + 12 (m − n)! which can be proved integrating the first term by parts.52) ⎪ ⎪ ⎪ ⎪ ⎪ ⎪ n Pmn d Pmn ⎪ ⎪ ⎪0 ⎩ ⎪ ⎭ sin φ dφ (used in problems formulated in spherical coordinates). cm = m + 12 (m−n)! (m+n)! f (x) Pkn (x) dx (11.56) m=0 −1  ∞  π   g(φ) = dm Pmn (cos φ).57) m=0 0 11.  π n d Pm n d Pn *π n Pk + Pmn k dφ = n Pmn Pkn *0 = 0 (11.11.6 Leibniz rule for the derivative of a product of two functions dn [ f (x) g(x)] n (n − 1)  (n−1) = f g (n) + n f  g (n−1) + f g dx n 1 1×2 n (n − 1) (n − 3)  (n−2) + f g + · · · + f (n) g (11. using the differential equation for the associated Legendre functions.4.

12 Brief listing of integral transforms 12. that ( +∞they * have * a finite number of discontinuities and satisfy the so-called Dirichlet condition −∞ * f (t)* dt = ∞.1 Fourier transforms For any sufficiently well-behaved function f(t) there exists a Fourier transform F(ω) such that  +∞  +∞ 1 f (t) = F(ω) eiωt dω ↔ F(ω) = f (t) e−iωt dt 2π −∞ −∞ ∂ n f (t) ↔ (i ω)n F(ω) ∂t n ∂ n F(ω) t n f (t) ↔ in ∂ωn provided that f(t) and/or F(ω) decay sufficiently fast with t or ω. Although the derivatives of such functions do not necessarily satisfy these conditions – and hence the integrals do not converge – the above rules can still be used to obtain formal transforms to the derivatives of f or F in the sense of distributions (or singularity functions). a) Transforms in time–frequency  +∞  +∞ 1 f (t) = F(ω) eiωt dω ↔ F(ω) = f (t) e−iωt dt 2π −∞ −∞ δ(t) ↔ 1 ∗ H (t − tS ) π i (2)  ↔ − H (ωtS ) t 2 − tS2 2 0  π i tS (2) t 2 − t S2 H (t − tS ) ↔ H (ωtS ) ∗ 2 ω 1 H (t − tS ) π i ω (2)  3/2 ↔ H (ωtS ) t 2 − tS2 ∗ 2 tS 1 * t 1 + iωt **tS [H (t − tP ) − H (t − tS )] ↔ e−iωt tS2 ωtS2 *tP 194 . Such transforms are identified in the listings below with an asterisk. that is.

1 Fourier transforms 195 b) Wavenumber–space integrals From Gradshteyn and Ryzhik (Chapter 11. eq. eq. footnote 1). Integrals that violate the Dirichlet conditions are valid only in the sense of distributions. 1151. 17. k > 0 0 (x + a ) 2 2 2a (ν + 12 ) We use this expression to derive some of the results given below. 6. we obtain  +∞  +∞ 1 f (x) = F(k) e∓ikx dk ↔ F(k) = f (x) e±ikx dx 2π −∞ −∞  ν 1 √ |k| Kν (|k| a) ν+1/2 ↔ 2 π Re a > 0. Results for Im(k0 ) = 0 are principal value (PV).12) 2n 2 n! Starting from the above. using symmetry and antisymmetry considerations. p. together with the relations ⎧ ⎨12 π iν+1 Hν(1) (i z) −π < arg z ≤ 12 π Kν (z) = (Abramovitz & Stegun.4. ⎧  ∞ ±i k x ⎪ ⎪ i −i k 0 |x| 1 e ⎨− e Im k0 < 0 2k0 dk = 2π −∞ k 2 − k02 ⎪ ⎪ 1 ⎩− sin k0 |x| Im k0 = 0 2k0 ⎧  ∞ ⎪ ⎪ i 1 k e±i k x ⎨ ± e−i k 0 |x| sgn(x) Im k0 < 0 dk = 2 2π −∞ k 2 − k02 ⎪ ⎪ i ⎩ ± cos k0 x sgn(x) Im k0 = 0 2 .1. Integrals for Im(k0 ) > 0 can be obtained by conjugation. Re ν > − 12 . Re ν > − 12 (x 2 + a 2 ) 2a (ν + 12 )  1 2n+1 n! |k| n n+1/2 ↔ Kn (|k| a) Re a > 0 (x 2 + a 2 ) (2n)! a 1 √ ↔ 2K0 (|k| a) Re a > 0 x2 + a 2  1 ( 12 ) − |x| ν (2) 1 Hν (k0 |x|) ↔  ν+1/2 Im k0 < 0. Re ν > − 12 2i (ν + 12 ) 2k0 k 2 − k02  1 2n n! − |x| n (2) 1 Hn (k0 |x|) ↔  n+1/2 Im k0 < 0 2i (2n)! k0 k 2 − k02 1 (2) 1 H (k0 |x|) ↔  Im k0 < 0 2i 0 k 2 − k02 Additional formulas may be obtained by contour integration and derivation under the integral sign with respect to parameter. We assume throughout that Im(k0 ) ≤ 0.12.6.34. 9. we have  +∞  ν cos kx dx √ k Kν (ka) ν+1/2 = π .4b) ⎩1 (2) 2 π (−i)ν+1 Hν (−i z) − 12 π < arg z ≤ π and 1 × 3 × 5 × · · · × (2n − 1) 1 (2n)! √ (n + 12) = ( 2 ) = 2n π (Abramovitz & Stegun. Re a > 0.eq.

. Since the Bessel function behaves asymptotically as a harmonic function ( ∞ *while its amplitude * decays as r −1/2 .2 Hankel transforms The Hankel transform is defined by the integral  ∞ Fn (k) = r ν f (r ) Jn (kr ) dr 0 in which ν is defined in various references as either 1/2 or 1. if the Dirichlet . the Hankel transform exists only if the integral 0 *r ν−1/2 f (r )* dr exists (i.196 Brief listing of integral transforms ⎧ ⎪ i  −i k 0 |x|   ∞ ±i k x ⎪ ⎨± 2 e − 2 sgn(x) Im k0 < 0 1 e 2 k0   dk = 2π −∞ k k 2 − k02 ⎪ ⎪ i ⎩ ± 2 ( cos k0 x − 2) sgn(x) Im k0 = 0 2 k0 ⎧ ⎪ 1 −i k 0 |x|  ∞ ⎪ ⎨ − 3 (k0 |x| − i ) e Im k0 < 0 1 e±i k x 4k0  2 dk = ⎪ 1 2π −∞ k 2 − k02 ⎪ ⎩ − 2 |x| cos k0 x Im k0 = 0 4k0 ⎧ x −i k 0 |x| ⎪ 1  ∞ ke ±i k x ⎨ ± 4k e ⎪ Im k0 < 0 0  2 dk = ⎪ i 2π k 2 − k02 ⎪ −∞ ⎩ ± 2 (cos k0 x − k0 x sin k0 x) sgn x Im k0 = 0 4k0 ⎧  ⎪ ⎪ 1 1 ∞ k 2 e±i k x ⎨ (1 − i k0 |x|) e−i k 0 |x| Im k0 < 0 = 4i k0  2 dk ⎪ 2π −∞ k 2 − k0 2 ⎪ |x| ⎩ cos k0 x Im k0 = 0 4 ⎧  ⎪ ⎪ 1 (2) 1 ∞ e±i k x ⎨ H0 (k0 |x|) Im k0 < 0 2i  1 2 dk = ⎪ 1 2π −∞ k 2 − k02 / ⎪ ⎩ J0 (k0 |x|) Im k0 = 0 i ⎧  ∞ ⎨ ± 1 k0 H(2) (k0 |x|) sgn(x) Im k0 < 0 1 k e±i k x 2 1   dk = 2π −∞ k 2 − k02 / 1 2 ⎩ ±k J (k |x|) sgn(x) Im k0 = 0 0 1 0 ⎧ ⎪ i |x|  ∞ ⎪ ⎨ (2) H (k0 |x|) Im k0 < 0 1 e±i k x 2 k0 1   dk = 2π −∞ k 2 − k02 / 32 ⎪ ⎪ i |x| ⎩ J1 (k0 |x|) Im k0 = 0 k0 ⎧  ∞ ⎨ ± 1 x H0(2) (k0 |x|) Im k0 < 0 1 k e±i k x 2   dk = 2π −∞ k2 − k2 / 32 ⎩ ±x J (k |x|) Im k0 = 0 0 0 0 12.e.

r ≤ R k 2i J (kR)Jν (kr ) dk = π 2 ν Im(k0 ) < 0 0 k − k0 2 ⎪ ⎩ Jν (k0 R) Hν(2) (k0 r ). namely  ∞ f (r ) = kν Fn (k) Jn (kr ) dk 0 We list below some Hankel transforms using ν = 1. R≤r 2i k0 ⎧ π r ⎪ (2)  ∞ 1 ⎨2i k 2 J1 (k0 r ) H1 (k0 R) − 2R k 2 . (∞ (∞ f (r ) = 0 Fn (k) Jn (kr ) k dk Fn (k) = 0 f (r ) Jn (kr ) r dr δ(r ) ←→ n=0 1 r δ(r − a) ←→ n Jn (ka) r ←→ 1 −k 2 /4α e−αr 2 n=0 e 2α ⎧π  ∞ ⎪ ⎨ Jν (k0 r ) Hν(2) (k0 R).3 Spherical Hankel transforms 197 condition is satisfied). in which case  ∞  ∞ f (r ) = k Fn (k) jn (kr ) dk ⇔ Fn (k) = r f (r ) jn (kr ) dr 0 0 . in which case the transform is self-reciprocating. r ≤ R ⎪ 0 0 J1 (kR)J1 (kr ) dk = π R Im(k0 ) < 0 0 k(k − k0 ) 2 2 ⎪ ⎪ (2) J (k R) H1 (k0 r ) − . we conclude that an appropriate spherical Hankel transform is  ∞ 1  ∞ 1 f (r ) = kν+ 2 Fn (k) jn (kr ) dk ⇔ Fn (k) = r ν+ 2 f (r ) jn (kr ) dr 0 0 ( ∞* * ( ∞* * provided that the integrals 0 *r ν f (r )* dr and/or 0 *kν F(k)* dk exist. R ≤ r 2i ⎧ ⎪ ⎪ π 1  ∞ ⎨ (2) J0 (k0 r ) H1 (k0 R) − . A convenient value is ν = 1/2.3 Spherical Hankel transforms √ Starting from the expression jn (kr ) = π/2kr Jn+1/2 (kr ).12. which relates the spherical Bessel function of the first kind to the half-integer order conventional Bessel function. r≤R 1 2i k0 Rk02 J (kR)J (kr ) dk = Im(k0 ) < 0 0 k − k0 2 2 1 0 ⎪ ⎪ π (2) ⎩ J1 (k0 R) H0 (k0 r ). R≤r ⎩ 2 1 0 2 2i k0 2r k0 12.

198 . the following functions are called by some of the programs EASYPLOT Writes text files for plotting (most programs) CISIB Sine and cosine integrals (cavity 2-D and 3-D) ELLIPINT3 Elliptic integrals (Lamb 3-D) Note: Some programs have same-named functions embedded within. but they are not identical to each other.13 MATLAB programs The following Matlab programs are listed herein: SH2D FULL SH line load in full space SVP2D FULL SVP line load in full space BLAST2D Blast line source in full space CAVITY2D 2-D cavity in full space POINT FULL Point source in full 3-D space TORSION FULL Torsional point source in full space CAVITY3D 3-D cavity in full space SH2D HALF SH line source in half-space GARVIN Blast line source in half-space LAMB2D SVP line source in half-space LAMB3D Point source in half-space (horizontal and vertical) TORSION HALF Torsional point source in half-space SH PLATE SH line load in homogeneous plate SH STRATUM SH line source in homogeneous stratum SVP PLATE SVP line source in plate with mixed boundary conditions SPHEROIDAL Spheroidal and torsional modes of homogeneous sphere In addition. even if functionally related. they are not interchangeable. Thus.

plot (W. function [ ] = SH2D Full( ) 199 function [] = SH2D Full() % SH line source in a full. titx = ‘Dimensionless frequency w*r/Cs = kr’. homogeneous 2D space % Default data cs = 1. grid on. G = -i*0. grid on. xlabel(titx). titx. W.5. Uyy(3) = Uyy(4). pause. plot (W.25*besselh(0.real(G)).*W2]. G. % Time domain T1 = [0 1 1].ˆ0. W = [0. % maximum dimensionless frequency for plotting (rad/s) mu = rho*csˆ2.imag(G).W)/mu./Uyy].real(G)). tit = ‘Full-space displacement due to SH line load * sqrt(kr)’. plot (W. xlabel(titx). title(tit). % Frequency domain W2 = W. hold on. % avoid singularity and discontinuity in plot T = [T1. hold off. W]. T2].ezp’. % Maximum dimensionless time for plotting wmax = 10. EasyPlot(‘SH1-FD. pause.imag(G). ‘c’). Uyy = 2*pi*mu*sqrt(T2. title(tit). plot (W. hold off. W = [dw:dw:wmax].‘r’). hold on. Uyy = [0 0 inf 1. dt = tmax/np. T2 = [1+dt:dt:tmax]. plot (T. % Number of time and/or frequency intervals tmax = 10. G = [0. Uyy). % Shear modulus dw = wmax/np. G. tit = ‘Full-space displacement due to SH line load’.‘r’). .2.ˆ2-1). tit. % Shear wave velocity rho = 1 % Mass density np = 200.

200 MATLAB programs

tit = ‘Full-space displacement due to SH line load’;
title(tit);
titx = ‘Dimensionless time, t*Cs/r’;
xlabel(titx);
grid on;
pause;
EasyPlot(‘SH1-TD.ezp’, tit, titx, T, Uyy, ‘r’);
close;
return

function [ ] = SVP2D Full(x, z, poi) 201

function [] = SVP2D Full(x, z, poi)

% SVP line source in a full 2D space
% Arguments:
% x, z = coordinates of receiver
% pois = Poisson’s ratio
% Note: Uxz = Uzx, Gxz = Gzx
% To obtain the components in cylindrical coordinates,
% simply run with x=1,z=0, and then x=0,z=1

% Default data
rho = 1; % mass density
cs = 1; % shear wave velocity
nt = 500; % number of time points
tmax = 5; % maximum time
nf = 500; % number of frequency points
wmax = 20; % max. frequency

mu = rho*csˆ2; % Shear modulus
alfa = sqrt((1-2*poi)/(2-2*poi)); % cs/cp
r = sqrt(xˆ2+zˆ2);
theta = atan2(z, x);

% Time domain solution
dt = tmax/nt;
T = [0:nt]*dt;
T2 = T.ˆ2;
U = impulse response(r, theta, cs, mu, alfa, T2);
T = T*cs/r; % dimensionless time
plot (T, U(1,:));
tit = sprintf(...
‘Uxx at x=%5.3f, z=%5.3f due to SVP line load in full space’, x, z);
titx = ‘Dimensionless time, t*Cs/r’;
grid on;
title(tit);
xlabel(titx);
EasyPlot(‘SVP TDxx.ezp’, tit, titx, T, U(1,:), ‘r’);
pause;
plot (T, U(2,:));
tit = sprintf(...
‘Uxz at x=%5.3f, z=%5.3f due to SVP line load in full space’, x, z);
grid on;
title(tit);
xlabel(titx);
EasyPlot(‘SVP TDxz.ezp’, tit, titx, T, U(2,:), ‘r’);
pause;
plot (T, U(3,:));
tit = sprintf(...
‘Uzz at x=%5.3f, z=%5.3f due to SVP line load in full space’, x, z);
grid on;

202 MATLAB programs

title(tit);
xlabel(titx);
EasyPlot(‘SVP TDzz.ezp’, tit, titx, T, U(3,:), ‘r’);
pause;

% Frequency domain solution
dw = wmax/nf;
W = [dw:dw:wmax];
H = Green(r, theta, cs, mu, alfa, W);
W = W*r/cs; % dimensionless frequency
plot (W, real(H(1,:)));
hold on;
plot (W, imag(H(1,:)),‘r’);
tit = sprintf(...
‘Gxx at x=%5.3f, z=%5.3f due to SVP line load in full space’, x, z);
titx = ‘Dimensionless frequency w*r/Cs = k*r’;
grid on;
title(tit);
xlabel(titx);
EasyPlot(‘SVP FDxx.ezp’, tit, titx, W, H(1,:), ‘c’);
pause;
hold off;
plot (W, real(H(2,:)));
hold on;
plot (W, imag(H(2,:)),‘r’);
tit = sprintf(...
‘Gxz at x=%5.3f, z=%5.3f due to SVP line load in full space’, x, z);
grid on;
title(tit);
xlabel(titx);
EasyPlot(‘SVP FDxz.ezp’, tit, titx, W, H(2,:), ‘c’);
pause;
hold off;
plot (W, real(H(3,:)));
hold on;
plot (W, imag(H(3,:)),‘r’);
tit = sprintf(...
‘Gzz at x=%5.3f, z=%5.3f due to SVP line load in full space’, x, z);
grid on;
title(tit);
xlabel(titx);
EasyPlot(‘SVP FDzz.ezp’, tit, titx, W, H(3,:), ‘c’);
pause;
hold off;
close all;
return

function [G] = Green(r, theta, cs, mu, alfa, W)
% Green’s functions for SVP line load in full space
Ws = W*r/cs;

function [ ] = SVP2D Full(x, z, poi) 203

Wp = alfa*Ws;
psi = (besselh(1,2,Ws)-alfa*besselh(1,2,Wp))./Ws - besselh(0,2,Ws);
chi = alfaˆ2*besselh(2,2,Wp)- besselh(2,2,Ws);
c = cos(theta);
s = sin(theta);
f = 0.25*i/mu;
G = f*[psi+chi*cˆ2; chi*c*s; psi+chi*sˆ2]; % [gxx; gxz=gzx; gzz]
% G = f*[psi+chi; psi]; % radial & tangential components
% Radial changes as cos(theta), tangential as -sin(theta)
return

function [U] = impulse response(r, theta, cs, mu, alfa, T2)
% Impulse response functions for SVP line load in full space
ts = r/cs;
tp = alfa*ts;
ts2 = tsˆ2;
tp2 = tpˆ2;
k = find(tp2<T2);
if isempty(k)
PSI = zeros(size(T2));
CHI = PSI;
else
k = k(1);
k1 = k-1;
n = length(T2);
S = sqrt((T2(k:n)-tp2));
PSI = [zeros(1,k1),-S/ts2]; % yes, we divide by ts2, not tp2
CHI = [zeros(1,k1),alfaˆ2./S]-2*PSI;
if tp2==T2(k1), CHI(k1)=inf; end
k = find(ts2<T2);
if ∼isempty(k)
k = k(1);
k1 = k-1;
n = length(T2);
S = sqrt((T2(k:n)-ts2));
PSI = PSI + [zeros(1,k1),S/ts2+1./S];
CHI = CHI − [zeros(1,k1),2*S/ts2+1./S];
if tp2==T2(k1), PSI(k1)=inf; CHI(k1)=inf; end
end
end
c = cos(theta);
s = sin(theta);
f = 0.5/pi/mu;
U = f*[PSI+CHI*cˆ2; CHI*c*s; PSI+CHI*sˆ2]; % [uxx; uxz=uzx; uzz]
% U = f*[PSI+CHI; PSI]; % radial & tangential components
% Radial changes as cos(theta), tangential as −sin(theta)
return

204 MATLAB programs

function [] = Blast2D(pois)

% Computes the plane-strain response elicited by an
% SV-P line blast source in a full, homogeneous 2D space
% Arguments:
% pois = Poisson’s ratio

% Default data
nt = 300; % Number of time intervals
nf = 200; % Number of frequency intervals
cs = 1; % Shear wave velocity
rho = 1; % Mass density
r = 1; % epicentral distance
tmax = 5; % Maximum time for plotting
wmax = 20; % maximum frequency for plotting (rad/s)

mu = rho*csˆ2; % Shear modulus
cp = cs*sqrt((2-2*pois)/(1-2*pois)); % P-wave velocity
a = cs/cp;
a2 = aˆ2;

% Frequency domain
f = 0.25/mu/cp;
dw = wmax/nf;
Ws = [dw:dw:wmax]*r/cs; % dimensionless frequency for S waves
Wp = a*Ws; % dimensionless frequency for P waves
H1p = besselh(1,2,Wp);
Gr = f*Wp.*H1p; % Green’s function for radial displacement
tit = ‘Radial displacement in full space due to line blast load’;
plot (Ws,real(Gr));
hold on;
plot (Ws,imag(Gr),‘r’);
grid on;
title(tit);
titx = ‘Dimensionless frequency w*r/Cs = k*r’;
xlabel(titx);
pause;
hold off;
EasyPlot(‘B2D FDr.ezp’, tit, titx, Ws, Gr, ‘c’);
clear G* H*

% Time domain
f = 1/(2*pi*mu*r);
tp = r/cp; % arrival time of P waves
ts = r/cs; % arrival time of SV waves
ts2 = tsˆ2;
tp2 = tpˆ2;
dt = tmax/nt;
T = [0:dt:tmax]+eps;
P = sqrt(T.ˆ2-tp2);

T = T*cs/r. % dimensionless time plot (T. T. tit = ‘Radial displacement in full space due to line blast load’. Ur. titx. titx = ‘Dimensionless time. close. Ur).*real(1. EasyPlot(‘B2D TDr. grid on. tit. pause. t*Cs/r’.function [ ] = Blast2D(pois) 205 Ur = f*T.ezp’. xlabel(titx). return . title(tit)./P). ‘r’).

tit = sprintf( . % Cp/Cs Cp = a*Cs. hold on. %plot(w1.25/i. rr. h1 = besselh(1. wmax = 2*pi*Cp/r0. % transfer function for line of pressure w1 = [0.*h0. homogeneous space % subjected to a harmonic/impulsive pressure % The pressure is scaled so that p*pi*r0ˆ2=1 % % Arguments (in consistent units) % r = distance to receiver % r0 = radius of cavity % pois = Poisson’s ratio % % Returned transfer functions are g*mu*r vs. step % Frequency domain solution f1 = r/r0/pi. for pressure in cylindrical cavity. w = [dw:dw:wmax]. % factor for cavity f2 = 0.. r0=%5. y = 2*h1-a2*w0. % P-wave velocity rr = r0/r.4f. % dimensionless frequency wmax = wmax*r/Cp. pois) % Cylindrical (2-D) cavity in a full. f2*w1. ‘T.5/pi. w*r/Cp % Returned time histories are u*rho*rˆ2 vs.w0).‘r’). \nu=%5. % factor for line of pressure h0 = besselh(0. nf = 1024. H = [0. rho is the mass density.2. plot(w1. h1 = besselh(1.w1). % frequency vector plot(w1./y]. F. dw = dw*r/Cp.real(G).206 MATLAB programs function [] = Cavity2D(r.2. % Compare with TF for line of pressure %plot(w1. t*Cp/r % in which mu is the shear modulus. % true frequency w0 = w*r0/Cp. dimensionless freq.real(H)).w0). % and Cp is the P-wave velocity % Basic data Cs = 1.imag(H). f1*h1. dw = wmax/nf.2. % dimensionless freq. % max. w1].‘:’).*h1]..2f’. pois). w1 = w*r/Cp. % S-wave velocity (the results do not depend on this value) a2 = (2-2*pois)/(1-2*pois).5/pi. .imag(G). a = sqrt(a2). r0. % transfer function for cylindrical cavity G = [0.‘r:’).

EasyPlot(‘Cav2D-FD. u = fac*(1-f*cisib(wmax*T)). close.. xlabel(titx).. % delay TF by arrival time tmax = 2.2f’. %axis([0 wmax -1 1]). plot(T.. end U = U*dw*a/pi.*exp(i*w1*t0). tit.G].5/a/pi/rrˆ1.. hold off. r0. response at r/r0=%5. ‘Cylindrical cavity under pressure. \nu=%5. % maximum dimensionless (delayed) time nt = 100. % tail U = U+u. ‘r’). pois) 207 titx = ‘Frequency (rad/s)’. return . % dimensionless.U). fac = 0. of time steps U(j) = sum(real(H. xlabel(titx).function [ ] = Cavity2D(r. tit. ‘c’). f = 2/pi. T. for j=1:length(T) % Fourier transform by direct integration (no FFT here) t = T(j). % dimensionless arrival time H = H. tit = sprintf(. T = [0 t0 T+t0]. % time step T = [0:nt]*dt. pois). title(tit).ezp’.*exp(i*w1*t))). % Advantage: can use arbitrary size & No. titx. close.. title (tit). pause. % No. 1/rr. U. pause. [H. EasyPlot(‘Cav2D-TD. % Response at receiver in the time domain t0 = 1-rr. w1.ezp’. % shift delayed time axis by wave arrival time U = [0 0 U]. delayed time axis U = zeros(size(T)).2f. % plot response grid on. titx. titx = ‘Time (s)’. grid on.5. of time steps dt = tmax/nt.

Fp = Ep.ezp’. titx. Ws = [dw:dw:wmax]. % Shear wave velocity rho = 1.pois). hold on. pause. % Mass density mu = rho*csˆ2. % epicentral distance tmax = 2. xlabel(titx). Ws.*(i+1.pois). % Displacements in spherical coordinates due to point load in x direction PSI = Es-Fs+Fp. ‘c’). cp = cs*sqrt((2-2*pois)/(1-2*pois)). % Maximum dimensionless time for plotting wmax = 20. plot (Ws.real(PSI)). % Rigid-body component CHI = Ep-Es-3*(Fp-Fs)./Ws. EasyPlot(‘PSI-3D-FD. \nu=%5./Wp). % Dimensionless frequency for P waves Es = exp(-i*Ws).imag(PSI). PSI.real(CHI)). title(tit).*(i+1. a2 = aˆ2. % Dimensionless frequency for S waves Wp = a*Ws. grid on. % maximum dimensionless frequency for plotting (rad/s) % Frequency domain dw = wmax/nf. Fs = Es. tit = sprintf(‘\chi in full 3D space. titx = ‘Dimensionless frequency w*r/Cs’./Wp.2f’. % P-wave velocity a = cs/cp. dt = tmax/nt. \nu=%5. tit. 3D homogeneous space % Input arguments: % pois = Poisson’s ratio % Basic data nt = 400./Ws). . plot (Ws.208 MATLAB programs function [] = Point Full(pois) % Computes the response elicited by a point load % in a full. % Shear modulus r = 1. ylabel(‘\psi’).2f’. plot (Ws. % Number of time intervals nf = 200. hold off. Ep = a2*exp(-i*Wp).‘r’). % Distortional component clear E* F* tit = sprintf(‘\psi in full 3D space. % Number of frequency intervals cs = 1.

return . CHI = [0. hold off.tp1.0. axis([0 tmax -0.5 0. titx.PSI(jp).ezp’. tit = sprintf(‘\chi in full 3D space.5]). plot (Ws. T = [0. \nu=%5.pois).ts1. PSI = Ts-Tp. Ts = T. EasyPlot(‘CHI-3D-TD.ts1. title(tit). ylabel(‘\chi’). titx = ‘Dimensionless time.imag(CHI). xlabel(titx).0.PSI(jp:js). title(tit). tit = sprintf(‘\psi in full 3D space.*(T>=ts)/ts2. T. titx. grid on. pause. % Time domain clear G* tp = r/cp.amax. CHI. tp1 = 0. grid on. titx = ‘Dimensionless time.tp1. PSI.0]. tit. close. T. PSI). titx. t*Cs/r’. % arrival time of P waves ts = r/cs.function [ ] = Point Full(pois) 209 hold on.2f’.ezp’.ezp’.5*(T(jp-1)+T(jp)).*(T>=tp)/ts2. xlabel(titx). CHI. ‘c’).2f’.0. axis([0 tmax -1. EasyPlot(‘PSI-3D-TD.amax..0. amax = 10*max(abs(PSI)). CHI = -3*PSI.amax. T = [0:dt:tmax]. CHI). grid on. ‘r’). tit.CHI(js). Ws.0.CHI(jp).5*(T(js)+T(js+1)).tp1.pois).ts1.5 5]). pause.T(jp:js). jp = ceil(tp/dt)+1. titx = ‘Dimensionless frequency w*r/Cs’. js = floor(ts/dt). Tp = T.CHI(jp:js). EasyPlot(‘CHI-3D-FD. plot (T. PSI = [0. % arrival time of SV waves ts2 = tsˆ2.PSI(js). t*Cs/r’. title(tit). xlabel(titx). pause. plot (T. ‘r’).0]. tit.tmax]. % to simulate a Dirac delta mm = length(PSI).‘r’). ts1 = 0. \nu=%5.

.. % cos of vertical angle ap = pi/2/td. tit = sprintf(. T = [0:dt:td]. y.rho. R = sqrt (xˆ2 + yˆ2 + zˆ2). x/R.%5. % shear modulus nt = 200.. y/R. tau = ang*T.z)=(%5. % Scaling factor for response ta = R/cs. b = ang*A/cs/R.z/R)=(%5.%5.2f. % plot dimensionless response & time tau = T*cs/R.210 MATLAB programs function Torsion Full(x. A = ap*phi/8/pi/mu.y.2f. phi = r/R. v = (mu*Rˆ2)*v.v) tit = sprintf(. % amplitude of sine pulse ap = 1. xlabel(‘Time’).y. a = A/Rˆ2. x.. z = coordinates of receiver % td = duration of pulse % rho = mass density % cs = shear wave velocity mu = rho*csˆ2.v.. title(tit) grid on.e. .z). z/R).2f) due to torsional sine pulse’. pause. actual time plot(T. T = [0. or unit impulse) % The torsional moment is applied at the origin and has vertical axis % Arguments % x. 0<=t<=td % (i.ta+2*td].y.0].2f) due to torsional sine pulse’. v = a*sin(tau)+b*cos(tau). title(tit). ‘Response at (x.%5...2f..0. plot(tau.v). % Number of points dt = td/nt.ta+td. % plot actual response vs.%5.z.cs) % Computes the response of a full 3D space to a torsional % point source with time variation (pi/2/td)*sin(pi*t/td).. a sine pulse with unit area.. ‘Response at (x/R. ylabel(‘v*mu*Rˆ2’). v = [0.T+ta. xlabel(‘Dimensionless time t*Cs/R’).2f. % arrival time ang = pi/td. grid on.y/R.ta. % time step r = sqrt(xˆ2 + yˆ2).td.0.

fclose (fout). fprintf (fout.5e\n’. ‘/sd off\n’). fprintf (fout. y. ‘/og on\n’). ‘/sm off\n’). tit). fprintf (fout. ‘%15. fprintf (fout. return . fprintf (fout.’w’). close % Make file for plotting with EasyPlot fout = fopen (‘tors. [tau. cs) 211 pause. z.5e%15. td. ‘/et g “%s”\n’. ‘/et x “t*Cs/R”\n’).v]). rho. fprintf (fout.function Torsion Full(x.ezp’.

% dimensionless. plot(w... r. w1 = w0*r. % Receiver rho = 1. H = G. % No. % shear modulus % Frequency and time vectors nf = 1024. % Radius of cavity R = r*R0. % P-wave velocity mu = rho*Csˆ2. G0 = 1+i*w0.ˆ2)). G = fac*(G0. tit = sprintf(. homogeneous space % subjected to a harmonic/impulsive pressure p % Arguments: % r = R/R0 = (distance to receiver)/(radius of cavity) % pois = Poisson’s ratio % % This program assumes p0=1 (unit pressure). % maximum dimensionless time nt = 200. a = sqrt(a2)../(G0-0. grid on. % Cp/Cs R0 = 1..*G1.25*a2*w0. % time step T = [0:nt]*dt.imag(G).‘:’). . %Receiver at R plot(w.2f’. % assuming p=1 here w0 = w*R0/Cp. xlabel(titx). \nu=%5. titx = ‘Frequency (rad/s)’.F. pois). dw = wmax/nf. title (tit). tmax = 6. G1 = 1+i*w1. and r0=1 (radius of cavity) % Basic parameters a2 = (2-2*pois)/(1-2*pois).25*R0/mu.*exp(-i*(r-1)*w0).‘r:’).‘r’). % Cavity wall at R0 plot(w. delayed time (= t*Cp/R) % Response in frequency domain fac = 0. w = [dw:dw:wmax].imag(H). for pressure in spherical cavity.4f. plot(w.real(H). hold on. pois) % Spherical (3-D) cavity in a full. wmax = 16*pi.212 MATLAB programs function [] = Cavity3D(r. R/R0=%5. of time steps dt = tmax/nt. % mass density Cs = 1.real(G)). % S-wave velocity Cp = a*Cs. ‘T./G0/rˆ2..

‘/og on\n’).5f %15. titx). fprintf (fout.‘r:’). tit = sprintf(. ‘%15. % Create EasyPlot file fout = fopen(‘Cav3D-TD0. U = fac*exp(-b*T). ‘/et x “%s”\n’. of time steps U(j) = u0+sum(real(G. % damping wd = wn*sqrt(1-xiˆ2).*exp(i*w*t))). titx).5f %15.. % factor for unit strength pressure f1 = f0*rˆ2. ‘/og on\n’). xlabel(titx). % factor for unit strength pressure . fprintf (fout. \nu=%5. fprintf (fout.. phi = asin(pois/(1-pois)). plot(T. ‘/et x “%s”\n’. pois). [w0. response at wall by numerical Fourier transform U = zeros(size(T)). fprintf (fout. titx = ‘Time (s)’.ezp’. tit). % a) Time domain. plot(T.5f %15.U. % natural frequency xi = Cs/Cp.function [ ] = Cavity3D(r.5f %15.75*mu/R0/pi.. fout = fopen(‘Cav3D-FD. ‘Cylindrical cavity under pressure.f0*imag(G)]). f0 = 0. pause. ‘/et g “%s”\n’. ‘%s \n’.ezp’.5f\n’.*cos(wd*T-phi). title(tit).2f’. pause. fprintf (fout.5f\n’. response at wall. fclose (fout).f1*real(H). end U = U*dw/pi. [w1. tail = (1-2/pi*cisib(wmax*T))*2*fac*Csˆ2/Cp/R0. pois) 213 grid on. fprintf (fout.5*fac. % Advantage: can use arbitrary size & No. % b) Time domain. response at wall by exact formula fac = sqrt((1-pois)/2)/rho/Cs. f0 = 0.75*rho*Cs*R0ˆ3/pi. % Exact response hold on. u0 = 0. % plot response by FT grid on. wn = 2*Cs/R0. for j=1:length(T) % Fourier transform by direct integration (no FFT here) t = T(j). hold off. ‘%15... ‘//nc’).U). %tail of Fourier integral U = U+tail. ‘w’). % damped frequency b = xi*wn. ‘w’). fprintf (fout. fprintf (fout. tit).f0*real(G). ‘/et g “%s”\n’.f1*imag(H)]). fprintf (fout.

U*f0]). pause. fprintf (fout. ‘/og on\n’).5f\n’. % delay TF by arrival time V = zeros(size(T)). plot(T1. ‘%15. close. % c) Response at receiver by numerical Fourier transform hold off. hold on.. end V = V*dw/pi. fclose (fout). ‘w’). titx).*(cos(wd*T-phi)-(1-1/r)*sin(wd*T))/r. V = fac*exp(-b*T). tit). title(tit).5*fac/rˆ2. \nu=%5. fclose (fout). return . % plot response by FT grid on.. for j=1:length(T) % Fourier transform by direct integration (no FFT here) t = T(j).214 MATLAB programs f1 = f0*rˆ3. ‘%15. ‘/et g “%s”\n’. of time steps V(j) = u0+sum(real(H.2f’. % Advantage: can use arbitrary size & No.‘r:’). T1 = [0 t0 T+t0]. fprintf (fout. [T1. ‘/et x “%s”\n’.. % Create EasyPlot file fout = fopen(‘Cav3D-TD1. u0 = 0. V = V+tail/r. ‘Cylindrical cavity under pressure. fprintf (fout.. % shift delayed time by wave arrival time V = [0 0 V].ezp’.*exp(i*w0*t0). titx = ‘Time (s)’. % dimensionless arrival time H = H.. fprintf (fout. % Exact response V = [0 0 V].*exp(i*w*t))). response at r/r0=%5.2f. fprintf (fout. tit = sprintf(.5f %15. t0 = r-1.5f %15.5f\n’.V*f1]). xlabel(titx). % d) Response at receiver by exact formula fac = sqrt((1-pois)/2)/rho/Cs. pois). plot(T1. [T.V).V. r.

Hence. r1 = sqrt(dxˆ2+dz1ˆ2).. zs. tit = sprintf(. Source at (%5. zr) % Computes displacements due to SH line source in a homogeneous half-space % The source and the receiver can be placed anywhere % % Input arguments: % xs. xs.‘r’)..W*r))/mu. dz1 = zr-zs. dz2 = zr+zs. grid on. if r<1. r=1/r... % The vertical axis may be taken either up or down. dx = xr-xs.W)+besselh(0. zs = Coordinates of source % xr. % maximum dimensionless frequency for plotting (rad/s) np = 500.receiver distance r = r2/r1. hold on.2f)’.25*(besselh(0. xr. zr) 215 function [] = SH2D Half(xs. function [ ] = SH2D Half(xs. hold off. G = -i*0. plot (W.real(G)). xr.2. end W = [dw:dw:wmax].2. % image source . zs.%5. xr. pause.2f. zs. % source-receiver distance r2 = sqrt(dxˆ2+dz2ˆ2). titx = ‘Dimensionless frequency w*r/Cs’. % Create EasyPlot file . zr = Coordinates of receiver % % The origin of coordinates is taken at the free surface. % Shear modulus tmax = 10. Receiver at (%5.%5. % Number of time and/or frequency intervals dw = wmax/np. plot (W. zr). title(tit). % Shear wave velocity rho = 1.2f). % Mass density mu = rho*csˆ2. dt = tmax/np. ‘Half-space SH displacement..imag(G). xlabel(titx).2f. zs and zr in call to SH2 must have the same sign!’ return end % Basic data cs = 1. % Maximum dimensionless time for plotting wmax = 10. zs and zr % must have the same sign (either both positive or both negative) if (zs>0 & zr<0) | (zs<0 & zr>0) ‘Error.

imag(G)]). ‘w’). ‘/sm off\n’). Uyy)./sqrt((T2. ‘/sd off\n’). fprintf (fout. fprintf (fout. fprintf (fout. ‘/sd off\n’). titx = ‘Dimensionless time.ˆ2-rˆ2))). f = 1/(2*pi*mu).ezp’. titx). ‘/et x “%s”\n’. fclose (fout).ˆ2-1)+real(f. ‘w’). ‘/et g “%s”\n’. ‘/et x “%s”\n’. fprintf (fout. grid on.5f %15. tit). fprintf (fout. t*Cs/r’.5f %15. [T. % Create EasyPlot file fout = fopen(‘SH2-TD. Uyy(3) = Uyy(4).5f %15. ‘/et g “%s”\n’. fclose (fout). ‘/sm off\n’). close. titx). ‘/og on\n’). .real(G). T1 = [0 1 1]. ‘%15. T2 = [1+dt:dt:tmax]. Uyy = [0 0 inf Uyy]. xlabel(titx). plot (T. fprintf (fout. [W.Uyy]).216 MATLAB programs fout = fopen(‘SH2-FD. title(tit).5f\n’. fprintf (fout. fprintf (fout. pause. Uyy = f.ezp’. fprintf (fout. fprintf (fout. fprintf (fout./sqrt(T2. ‘/og on\n’). tit). fprintf (fout. T2]. % avoid singularity & discontinuity in plot T = [T1.5f\n’. ‘%15.

Uz] = garvin(x. 528-541 % Basic data cs = 1. rho=1) % % Sign convention: % x from left to right. % Garvin’s line blast problem. W. % Source-receiver angle w. Series A % Vol. Uz] = Garvin(x. % Source-receiver distance c = abs(z)/r. % Shear modulus a2 = (1-2*pois)/(2-2*pois).r. z=0 is the upper boundary % Reference: % W. pois) 217 function [T. % Max. March 1956.function [T. T = [a+dt:dt:tmax]. Ux. tau = t*Cs/r % Ux = Horizontal displacement at surface % Uz = Vertical displacement at surface % Unit soil properties are assumed (Cs=1. Ux. % scaling factor for displacements T0 = [0 a].. z. Room 1-271. z=0 is the lower boundary % If z < 0 ==> a lower half-space is assumed. Cambridge.pois). z=0 at the surface. pp. No.ˆ2. MIT. vertical fac = 1/(pi*mu*r). % direction cosine w. . % Mass density np = 200. a = sqrt(a2). % Time vector T2 = T. % If z > 0 ==> an upper half-space is assumed. z axis s = abs(x)/r.t. % Proceedings of the Royal Society of London. 0].r. Exact transient solution of the buried line source problem.z. % Cs/Cp dt = (tmax-a)/np. time for plotting mu = rho*csˆ2. % Number of time intervals tmax = 2. Uz] = Garvin(x. % ” ” ” x axis theta = asin(s).z. 234. z points up. Uz0 = [0. 1199. % Shear wave velocity rho = 1. plane strain (2-D) % Step blast line load applied at depth z below surface of elastic half-space % % Written by Eduardo Kausel. MA % % [T. Garvin. % Displacements are positive up and to the right. % time step r = sqrt(xˆ2+zˆ2). 0]. % time interval before arrival of P waves Ux0 = [0.t.pois) % Input arguments: % x = range of receiver on surface (z = 0) % z = depth of source at x = 0 % pois = Poisson’s ratio % Output arguments % T = Dimensionless time vector. Ux.

ˆ2 . ‘/sd off\n’).r*Ux]).4*Q1.*p1. tit). close. . fprintf (fout. title (tit). fprintf (fout. \nu=%5. plot (T. fprintf (fout. fprintf (fout..5f\n’. % ” vertical ” ” ” Q1 = q1.Uz1]. ‘Horizontal displacement due to line blast load. grid on. ‘/et g “%s”\n’. Uz = [Uz0. [T.*s1.*D1). fclose (fout).T].pois). % Plot results plot (T. ‘w’). P waves p1 = c*T+i*s*T1. ‘/sd off\n’). ‘%15. fprintf (fout.218 MATLAB programs T1 = conj(sqrt(T2-a2)). pause. %axis([0 tmax -2 2]). T = [T0.r*Uz).5f\n’./T1.ezp’. % Rayleigh function. fprintf (fout. title (tit).r*Ux).*s1. % Derivative of q1 divided by R1 Ux1 = 2*fac*sign(x)*imag(q1. ‘/sm off\n’). ‘w’). % Create EasyPlot file fout = fopen(‘garvin-z. \nu=%5. ‘/og on\n’). ‘/et g “%s”\n’.5f %15. grid on.ˆ2. tit = sprintf(. % make the imaginary part negative q1 = c*T1+i*s*T. % Create EasyPlot file fout = fopen(‘garvin-x. fprintf (fout.. P waves D1 = p1.2f’. Uz1 = -fac*sign(z)*real(S1.r*Uz]).5f %15.2f’. fclose (fout). s1 = sqrt(Q1+1). ‘/sm off\n’). R1 = S1.Ux1]. S1 = 2*Q1+1. tmax = max(T). fprintf (fout.*D1).pois). ‘/og on\n’). ‘Vertical displacement due to line blast load..ezp’. pause. fprintf (fout. % Complex horizontal slowness. tit = sprintf(. %axis([0 tmax -2 2]). fprintf (fout. ‘%15.. [T./R1. Ux = [Ux0. tit).

% Max. MIT. % Shear modulus a2 = (1-2*pois)/(2-2*pois). 2. % Pages 606/607.5.pois) % Input arguments: % x.z. Uxx. pois) 219 function [T. Uxz. eqs. 7. z. 7.z. 7. Uxz.16. Cambridge.8-10 % Basic data cs = 1. i. up) % If z = 0 ==> a lower half-space is assumed. % Max. tau = t*Cs/r % Uxx = Horizontal displacement caused by a horizontal load % Uxz = Horizontal displacement caused by a vertical load % Uzx = Vertical displacement caused by a horizontal load % Uzz = Vertical displacement caused by a vertical load % % Unit soil properties are assumed (Cs=1. and reverse their signs. z points up.15.0) % pois = Poisson’s ratio % Output arguments % T = Dimensionless time vector. 1979 % Eringen & Suhubi. Uxz. eq.pois) % Lamb’s problem in plane strain (2-D) % Impulsive line load applied onto the surface of an elastic % half-space and receiver at depth z % % To obtain the solution for an interior source and receiver at the % surface.4999*pi.e. Elsevier. . eq. z=0 is the upper boundary % Vertical impulse at z=0 is tensile. % Displacements are positive up and to the right. Uxx. Pilant.22 and 7. time for plotting mang = 0. z=0 is the lower boundary % Vertical impulse at z=0 is compressive.16. z=0 at the surface.9. z = coordinates of receiver relative to source at (0. rho=1) % % Sign convention: % x from left to right. exchange the coupling terms. up) % If z < 0 ==> a lower half-space is assumed. Uzz] = lamb2D(x. page 615. Elastodynamics. Uzz] = lamb2(x.14. 7. Room 1-271. i. eqs.27 % Page 612. Uzz] = lamb2D(x.e. % If z > 0 ==> an upper half-space is assumed. % % Written by Eduardo Kausel. % Mass density np = 200. MA % % [T.function [ T. page 617. angle for interior solution mu = rho*csˆ2. % Number of time intervals tmax = 2. % Shear wave velocity rho = 1.. Elastic Waves in the Earth. Vol. Uxx. z=0 is the upper boundary % both the load and the displacements are at the surface % % References: % Walter L.14.

tr = cs/cr. q = 2*t2-1. Uzz0 = [0. Uxz0 = [0. p = sqrt(p). vertical fac = cs/(pi*mu*r). % Time vector T2 = T. % dimensionless arrival time of R waves xr = (cr/cs)ˆ2. 0].*p. Uxx1 = 4*fac*t2. % t > ts=1 t2 = T2. Uxz2 = zeros(size(Uxx2)).r.*q. p = t2-a2.*p.r. % ” ” ” x axis theta = asin(s).*s.r. s = sqrt(s).220 MATLAB programs a = sqrt(a2). vertical crang = asin(a). Uzx0 = [0.0276*pois)*pois)*pois). T = [a+dt:dt:tmax]. .ˆ2. Uxz1 = 2*fac*T1. % t > ts=1 % a=tp <= t <= ts=1 t2 = T1./d.25*fac*pi*(1-0.*p. p = sqrt(t2-a2). % Source-receiver distance c = abs(z)/r.874+(0. z axis s = abs(x)/r.t.*s. % Source-receiver angle w. s = 1-t2. % interval from tp to ts T2 = [1+dt:dt:tmax].056+0. d = q.ˆ4 + 16*p./d. s = sqrt(t2-1).t.ˆ2.*s. % Cs/Cp dt = (tmax-a)/np. d = q. jl = length(T).ˆ2. 0].t. % time step r = sqrt(xˆ2+zˆ2).125*xrˆ3-a2). wr = 0. % scaling factor for displacements % t < tp=a (two points suffice) T0 = [0 a].197-(0./d. 0].ˆ2. Uxx2 = -fac*s.*p.5*xr)ˆ3/(1-0. 0]. % direction cosine w./d. q = 2*t2-1. if theta > mang % Displacements on surface T1 = [a:dt:1].*t2.ˆ2-4*t2. Uzz2 = -fac*p. jr = floor((tr-a)/dt+1). cr = cs*(0. Uxx0 = [0. Uzz1 = -fac*q. % Critical angle w.ˆ2.5*xrˆ2+0.*s./d.

ˆ2 . S waves D1 = p1.ˆ2 ./sqrt(T2-1). T = [T1 T2]./R2. Uxz = [Uxz1. Q1 = q1. Uxx(js:jl) = Uxx(js:jl) + T2./D2. z. else % Displacements in the interior T1 = conj(sqrt(T2-a2)). P waves (Cagniard-De Hoop path) q2 = c*T2+i*s*T. % ” q2 ” ” R2 % Check critical angle if (theta > crang) tcrit = cos(theta-crang). end % Apply Heaviside step function k = floor((tcrit-a)/dt+1). S2 = 2*Q2+1. elseif theta < 1e-3 % Displacements on epicentral line D1 = (T2-a2+0.ˆ2. S1 = 2*Q1+1.*s1./T1. else tcrit = 1.*sqrt(T2+a2-1). Uzz(js:jl) = Uzz(js:jl) .*(T2-1).*sqrt(T2-1+a2).ˆ2 . % first element after arrival of S waves T2 = T2(js:jl). Uxz2]. Uxx./D1.T1.*sqrt(T2-a2+1). s2 = sqrt(Q2+a2). Uzz] = lamb2D(x.4*Q1./D2.*s2. Uxx = -T. Q2 = q2. f = 0. T1 = T(js:jl). % Rayleigh function.ˆ2. Uzz = f*Uzz.5). Uzz2]. % make the imaginary part negative T2 = conj(sqrt(T2-1)). Uxz = zeros(size(T)). p2 = c*T+i*s*T2. s1 = sqrt(Q1+1). Uzx = Uxz. Uxx2].*p1. Uzx = -Uxz. q1 = c*T1+i*s*T.ˆ2 .5*fac.T. js = 1+floor((1-a)/dt+1).5).*sqrt(T2-a2+1).5). clear T1. % Rayleigh function.*(T2-a2). .*sqrt(T2-a2). Uzz = [Uzz1. % Complex slowness.5). R1 = S1. pois) 221 % Combine solutions Uxx = [Uxx1./sqrt(T2-a2). P waves R2 = S2. Uxx = f*Uxx.*sqrt(T2-1)./D1.*p2. % “ “ S ” p1 = c*T+i*s*T1./R1. Uxz. % Derivative of q1 divided by R1 D2 = p2. Uzz = T2.function [ T.4*Q2. D2 = (T2-0.*(T2-0.*(T2-a2+0.T1./T2.

*p1...5 0.pois). end hold off.*p2.5]). %axis([0 tmax -2 2]).Uzz). plot (T. Uxz = [Uxz0. % Displacements due to impulsive line load sgn = fac*sign(x)*sign(z). Uzx = sgn*imag(q1. pause.*S2-Q1.*S1-q2. pause.Uxx]. Uzz = fac*real(p1. %axis([0 tmax -0.Uxz].pois). tit = sprintf(. title (tit).. ‘Vertical displacement due to horizontal line load.Uzx].Uzz]. Uzz = [Uzz0.. ‘Horizontal displacement due to horizontal line load. ‘Vertical displacement due to vertical line load.*D1..*D1-q2.2f’. grid on.*D2.2f’. plot (T. Uxx = [Uxx0. title (tit). Uxz = sgn*imag(q1..*S2). . grid on. [0 -wr]).*s2.*S1-Q2. title (tit).*D2). \nu=%5. D1 = 2*D1. grid on. S2 = S2.Uzx).*s1.Uxx). %axis([0 tmax -2 2]). titx = ‘Time’. end S1 = S1. end T = [T0.2f’. tmax = max(T). tit = sprintf((.*D2).. pause. if theta > mang hold on plot ([tr tr].T].222 MATLAB programs if k >=1. Uxx = fac*real(p2. pois). Uzx = [Uzx0. D2(1:k)=0.. tit = sprintf((. % Plot results plot (T. \nu=%5. D2 = 2*D2.. \nu=%5.*D1).

. titx. T. T. tit. pois) 223 plot (T. z. tit. ‘r’). end close. pfac*Uzx. -pfac*Uxz. Uxz. pfac*Uxz. pfac*Uxx. titx. tit. titx. end pause.ezp’. T. grid on.pois). T.2f’. pfac = rho*cs*r. EasyPlot(‘Lamb2Dzz.ezp’. if z∼=0 % reverse sign & coupling term for source at depth. ‘r’). ‘r’). if theta > mang hold on plot ([tr tr]. receiver at surface EasyPlot(‘Lamb2Dxz.Uxz). titx.5 0. title (tit). EasyPlot(‘Lamb2Dzx. ‘r’).. ‘r’). tit. tit. [0 wr]). T. else % source+ receiver are at the surface of a lower half-space EasyPlot(‘Lamb2Dxz.ezp’. hold off. Uxx.5]).ezp’.ezp’. Uzz] = lamb2D(x. tit = sprintf(. .function [ T. T. -pfac*Uzx. %axis([0 tmax -0. ‘r’). EasyPlot(‘Lamb2Dxx. titx. pfac*Uzz. EasyPlot(‘Lamb2Dzx. ‘Horizontal displacement due to vertical line load. titx.ezp’. \nu=%5. tit.

473-491. September 1960. z=0). Uxx. MIT. No. but % vertical component only % b) Horizontal loads: (pois = 0. pois) % Point loads suddenly applied onto the surface of a lower elastic half-space.2631. pois > 0. rho=1) % % Written by Eduardo Kausel.26 % Mooney. Uxx.0) % pois = Poisson’s ratio % Output arguments (cylindrical coordinates) % T = Dimensionless time vector. except that % Uzx varies as cos(theta) (reciprocity) % % Sign convention: % Vertical load and vertical displacements at the surface point up % Response given in terms of dimensionless time tau=t*Cs/r % % Note: % Solution for vertical loads for any Poisson’s ratio. % Maximum time for plotting (t=1 => arrival of S waves) fac = 1/16/pi/r.25 only) % Chao. because Uzx = -Urz. Dynamical response of an elastic halfspace to % tangential surface loading. but coupling term % Urz available only up to Poisson’s ratio 0. Urz] = Lamb3D(r. pp 559-567 mu = 1. Vol. Journal of Applied Mechanics. % Both horizontal and vertical loads are considered. tau = t*Cs/r % Urx = Radial displacement caused by a horizontal load % Varies as cos(theta) with azimuth % Utx = Tangential displacement caused by a horizontal load % Varies as -sin(theta) with azimuth % Uzz = Vertical displacement caused by a vertical load % (no variation with azimuth) % Urz = Radial displacement caused by a vertical load % (no variation with azimuth) % Uzx = Vertical displacement caused by a horizontal load % Not returned. tmax = 2. II. pois) % Input arguments: % r = range of receiver relative to source at (0. Uzz. . Utx. 64. % Time variation of load is a unit step function (Heaviside). Cambridge.2631. 1974. C. % Unit soil properties are assumed (Cs=1. Elastodynamics.224 MATLAB programs function [T. This is the value % for which the two false roots of the Rayleigh equation turn complex. pp. BSSA. % References: % a) Vertical loads: % Eringen and Suhubi. Uzz. Urz] = point3D(r.2. pois<0. Utx. MA % % [T. Room 1-271. 748-750. % Vol 27. % Shear modulus Nt = 400. C. V. which is % applied at the origin on the surface (x=0.

*Z). U2 = 2*f*(4 . if pois==0 x2 = a2. x = sort(roots(p)). x3 = x(3). T0 = [0 tp]./D./sqrt(x3-t2)). % ” ” ” S ” tr = sqrt(x3). a2 = (1-2*pois). end end Q = (1. Uzz./Q). tp = sqrt(a2). p = [-16*(1-a2). U1 = f*(4-A1. A1 = (1-2*x1)ˆ2*(a2-x1)/((x1-x2)*(x1-x3)). U1 = f*(4 . % save last point Uzz = 8*f. % ” ” ” R ” dt = (ts-tp)/Nt. 8*(3-2*a2). end T2 = [ts+dt:dt:tr-dt].*D.function [ T. else D = t2-a2. T3 = [tr]. t2 = T1. T1 = [tp+dt:dt:ts]./sqrt(t2-x1)-A2. Utx. U0 = [0 0].A3. pois) 225 % Vertical displacements due to vertical step load f = (1-pois)*fac./(2 .A3. Q = 1+2*Z+2*sqrt((Z+1). Uxx. for k=1:length(Q) if abs(Q(k))>1. Urz] = Lamb3D(r.Q). 1]. b2 = 1-a2. Z = (a2-x1)./sqrt(x3-t2))./Q . else n2 = b2/(a2-x2).ˆ2. n2 = inf. x2 = x(2).8*Q .ˆ2. if (imag(x1)==0) A1 = (1-2*x1)ˆ2*sqrt(a2-x1)/((x1-x2)*(x1-x3)). A3 = (1-2*x3)ˆ2*sqrt(x3-a2)/real((x3-x1)*(x3-x2)).2*pois)./sqrt(t2-x2)-A3. end n3 = b2/(a2-x3). y = U2(length(U2)). . % Arrival time of P waves ts = 1. A2 = (1-2*x2)ˆ2*sqrt(a2-x2)/((x2-x1)*(x2-x3)). Q = real(A1. Q(k)=1/Q(k). x1 = x(1). A3 = (1-2*x3)ˆ2*sqrt(x3-a2)/((x3-x1)*(x3-x2)). t2 = T2. -8./sqrt(x3-t2)). n1 = b2/(a2-x1).

axis ([0 tmax -0. T4 = [tr+dt:dt:tmax]. titx = ‘Dimensionless time’. U1.n3.n2. U3 = U2(length(U2))..0. . B2 = ellipint3(90. T = [T0. U2 = f*U2. ‘r’). T4].. grid on. tit. U3. pois).B3. U4].3f’. t2 = T2. k2 = (t2-a2)/b2. title (tit).ezp’..2 0.k2)*(1-2*x2)*(1-x2)/(x2-x1)/(x2-x3). % inverse of kˆ2 B1 = ellipint3(90. k2 = b2.n3*k2..226 MATLAB programs U3 = [-sign(A3)*inf]. B2 = ellipint3(90.B1 .2]).B2 . U4]. U2 = 2*ellipint3(90. pause./(t2-a2).*T2. % = kˆ2 B1 = ellipint3(90.n2*k2. Uzz.k2)*(1-2*x1)*(1-x1)/(x1-x2)/(x1-x3).. U1 = f*U1. ‘Radial displacement due to vertical point (step) load. titx.ˆ2. Uzz).. U4 = Uzz*ones(size(T4)). %axis ([0 tmax -0.ˆ2.k2)*(1-2*x3)*(1-x3)/(x3-x1)/(x3-x2). B3 = ellipint3(90. tit = sprintf(.. U1 = 2*ellipint3(90. xlabel(titx). Urz = [U0.1 0. T2. f = fac/sqrt(b2ˆ3). hold off. U4(1) = y.k2) .6]). C = (2*x3-1)ˆ3/(1-4*x3+8*b2*x3ˆ3). T3. \nu=%5.0. title (tit).*sqrt(k2).ˆ2. if (imag(x1)==0) t2 = T1. grid on. T..B2 .*T1.k2)*(1-2*x2)*(1-x2)/(x2-x1)/(x2-x3). ‘Vertical displacement due to vertical point (step) load.Urz).k2) . U4 = 2*pi*fac*C*T4. plot (T..k2)*(1-2*x1)*(1-x1)/(x1-x2)/(x1-x3). U2.n1*k2.n1./sqrt(t2-x3).B3. \nu=%5. T1. U3.k2)*(1-2*x3)*(1-x3)/(x3-x1)/(x3-x2). plot(T. tit = sprintf(. U2. U1. Uzz = [U0. pois)..3f’. EasyPlot(‘Lamb3Dzz.B1 . % Radial displacements due to vertical load fac = 1/(8*pi*pi*r). B3 = ellipint3(90. % save last point t2 = T4.

titx. titx. c2 = 0.Urx). U2 = fac*(1 -2*c3*t2. and multiplied by pi (but not here) % Radial displacements fac = 0. Urx = [U0.ˆ2. Urz. T4]./sqrt(t2-x2)-c3. xlabel(titx). x3 = 0.function [ T.25*(3-sqrt(3)). x2 = 0. T. tit. pois) 227 xlabel(titx). ‘r’). U4]. T = [T0. ‘r’). tit = ‘Radial displacement due to horizontal point (step) load. U2 = fac*(1 .f*c3*sqrt(x3-t2)).25’.25’. plot (T. \nu=0. T1. . U2../sqrt(x3-t2)).375/pi/mu/r. ‘Sorry: solution for Urz not available for Poisson’ ‘ratio > 0. title (tit). t2 = T2. xlabel(titx). % Tangential displacements fac = -0. U4]. EasyPlot(‘Lamb3Drz.ˆ2. t2 = T1.263’ break end % Horizontal step load (Chao’s solution) % Note: Chao’s figures are upside down. Uzz. U1 = fac*t2. Uxx. tit = ‘Tangential displacement due to horizontal point (step) load. title (tit). t2 = T1. t2 = T2./sqrt(x3-t2)).ˆ2.Utx). tit.f*(c1*sqrt(t2-x1)-c2*sqrt(t2-x2)+c3*sqrt(x3-t2))). grid on. pause.25.125*sqrt(6*sqrt(3)-10). pause. c3 = 0.5/pi/mu/r. Urz] = Lamb3D(r.75*sqrt(3).125*sqrt(6*sqrt(3)+10). Utx. \nu=0. x1 = 0. U1 = 0. EasyPlot(‘Chao3Drx.ˆ2. U1..25*(3+sqrt(3)). c1 = 0./sqrt(t2-x1)-c2. U2. T2. U4 = fac*ones(size(T4)). Urx.ezp’. plot (T.5*fac*(1.ezp’. else % THIS PART NOT YET FUNCTIONAL FOR Urz Warning =. f = 8/3. T. grid on. U1.*(c1. U4 = fac*ones(size(T4)). Utx = [U0.

U = [U0. U1 = -fac*f. f = t. ‘r’).ˆ2.5/pi/mu/r.*sqrt(t2+2/3). under load. t = [T2. f = (2*t2+1/3). g = t2.ezp’. % Chao’s functions on the axis (point load. T.*(t2-1/3). f = (2*t2+1/3).*sqrt(t2+2/3).*(2*t. grid on.U1.*sqrt(t2+2/3). tit.T1. U2 = fac*(0. ‘r’). tit.*(t2-1/3).t].5*(t2+1)-f+g).*sqrt(t2-2/3)-2*t2+1).*(t2-1). f = t. t2 = t./f. \nu=0.228 MATLAB programs EasyPlot(‘Chao3Dtx.U2]. plot (T.*sqrt(t2-2/3).*(t2-1/3).*(t2-1).T4]. EasyPlot(‘Chao3Daxis. . t2 = t.ezp’.ˆ2-4*t. title(tit)./f.U).) fac = 0. t = T1.*sqrt(t2+2/3). pause.ˆ2. close all. titx. U. T. tit = ‘Displacement at axis under horizontal point (step) load. T = [T0. titx.*(t2-1/3). pause.ˆ2-4*t./g.ˆ2-4*t. xlabel(titx).25’. displ. Utx. g = (2*t2-1).

3f due to torsional point load at z=%5. r. tit = sprintf(.. z. z = coordinates of receiver (horizontal range and depth) % z0 = depth of source if (z>0&z0<0) | (z<0&z0<0) disp(‘Error: z and z0 must have the same sign’).0.3f’. % Time domain solution m = 200.. Tp = pi/m*[0:m]. nt=2*nt. z. 0<=t<=td % Origin is at surface % Arguments: % r. R2 = sqrt(rˆ2+(z+z0)ˆ2). U). % mass density cs = 1. The source has a time % variation of the form f(t) = sin(pi*t/td).5*dM(m+1). U(k2:k2+m) = U(k2:k2+m)+S2*(M+R2*dM). dM(1) = 0. % latest arrival while nt*dt<tarr. z0). % duration of pulse fac = 1/8/pi/mu. z0) % Torsional point source with vertical axis acting within an % elastic half-space at depth zo. return end mu = 1. M = sin(Tp). end T = [0:nt]*dt. plot (T. . % fac*sin(phi-1)/R1ˆ2 S2 = fac*r/R2ˆ3.t2). k1 = floor(t1/dt)+1. z0) 229 function [] = Torsion Half(r.5*dM(1). S1 = fac*r/R1ˆ3. dM = pi/td/cs*cos(Tp). tarr = 2*max(t1.3f. R1 = sqrt(rˆ2+(z-z0)ˆ2). t2 = R2/cs. z. % shear wave velocity td = 1. % smooth out discontinuity for FFT dM(m+1) = 0.. z=%5..function [ ] = Torsion Half (r.. titx = ‘Time’. U(k1:k1+m) = S1*(M+R1*dM). t1 = R1/cs. ‘Response at r=%5. % Shear modulus rho = 1. U = zeros(size(T)). % Number of intervals in pulse dt = td/m. nt = m. k2 = floor(t2/dt)+1.

df = fmax/nf. T = T(k1:k2+m). EasyPlot(‘TorHalf TD. tit. A1 = abs(H). U. hold on. plot(f. tit = ‘Transfer function by FT ‘. % Transfer function by frequency domain solution W1 = i*2*pi*R1/cs*f. hold off. transfer function by FT ‘). xlabel(titx). real(H)). xlabel(titx). % Transfer function by brute force Fourier Transform % Reason: Only a dense set of low frequencies is of interest % (difficult and computationally expensive to accomplish with FFT) nf = 200. Tp = dt*[0:m]. grid on.‘r’). plot (f. A1). for j=1:nf+1 w = i*2*pi*f(j).nf+1). T. title(tit). .230 MATLAB programs grid on.*exp(-w*T)). title (‘Angle transfer function by FT ‘). Val. plot(f. f = [0:df:fmax]. H = zeros(1. end plot (f.ezp’. grid on. Um = sum(M. titx = ‘Frequency’. xlabel(titx). Uf = sum(U. pause. pause. pause.*exp(-w*Tp)). title (‘Abs. titx. H(j) = Uf/Um. title (tit). ‘r’). xlabel(titx). fmax = 2. hold off. ang1 = unwrap(angle(H)). U = U(k1:k2+m). ang1). imag(H). pause. grid on.

pause. title (‘A1/A2’). grid on. A1. z0) 231 W2 = i*2*pi*R2/cs*f. z=%5. z0). H. title (‘Diff... title (‘Angle transfer function.. A2). direct’.. real(H)). plot(f.*exp(-W1)+S2*(1+W2). pause. grid on. %H = S1*(1+W1). z.‘r’). transfer function. title (tit). Angle / pi’). f. tit. plot (f. ang2 = unwrap(angle(H)).*exp(-W2). title (tit)..function [ ] = Torsion Half (r. ang2). pause./A2). pause.imag(H). xlabel(titx). ang = (ang2-ang1)/pi. titx. ‘c’).*exp(-W2). return . z. ang). xlabel(titx). xlabel(titx). xlabel(titx). grid on. grid on. pause. plot(f. tit = ‘Abs. grid on. xlabel(titx). plot(f. ‘Transfer function at r=%5. close all.*exp(-W1)+S2*(1+W2).3f due to torsional load at z=%5. plot(f.3f’.3f. A2 = abs(H). tit = sprintf(. plot(f. hold on.ezp’. Val. xlabel(titx). hold off. H = S1*(1+W1). direct’). r. EasyPlot(‘TorHalf FD.

. % Plate has unit thickness and material properties % Origin is at the bottom surface. for j=-N:N t1 = (x2+(dz-j*h2)ˆ2)/cs2. z.. frequency if z>h | z0>h disp(‘Error. end U = fac*U. U). grid on. dz = z-z0. h2 = 2*h. ‘Impulse response at x=%5.0001. cs2 = csˆ2. z0). plot (T. t2). T = [0:nt]*dt. z = coordinates of receiver % z0 = elevation of source % Default data mu = 1. % parameter for max. % Shear modulus rho = 1. return end % Time domain solution via method of images % Choose number of images ni=N such that other images arrive after tmax tfund = 2*h/cs. U = zeros(size(T)).3f. t1) + response(T2.. % shear wave velocity h = 1. 232 MATLAB programs function [] = SH Plate(x. z is positive up % Arguments: % x. .3f’. x2 = xˆ2. x. % fundamental period of plate ffund = 1. % plate thickness xi = 0. z=%5. T2 = T.ˆ2. % material damping N = 10. % fundamental frequency (Hz) fac = 1/2/pi/mu. source/receiver z must be <= than thickness of plate’). tmax = N*tfund. dt = tmax/nt. % mass density cs = 1. nt = 400.. time and max. U = U + response(T2.3f due to SH line load in plate at z=%5. tit = sprintf(.. z0) % SH line source at elevation zo above the origin in an elastic plate./tfund. z. sz = z+z0. t2 = (x2+(sz-j*h2)ˆ2)/cs2.

EasyPlot(‘SH plate TD. EasyPlot(‘SH plate FD1. z=%5. z. imag(H). z0). xlabel(titx).. W = W*h*(1-i*xi)/cs. hold off. ‘r’).. z = z/h.3f. pause. H = zeros(size(F)). T. H. % Frequency domain solution via method of images ni = 20.2. ‘Green function at x=%5. z. real(H)).. z=%5. % Number of images fac = 0.3f due to SH line load in plate at z=%5. titx = ‘Frequency (rad/s)’.3f’. titx = ‘Time’. grid on. . tit.ezp’. plot (F. W2 = W.3f due to SH line load in plate at z=%5. ‘Green function at x=%5.3f’..2. tit = sprintf(.. hold on. H = 0.t1*W) + besselh(0. F = [df:df:fmax].‘r’). x = abs(x/h). W = 2*pi*F. xlabel(titx). fmax = N*ffund. x. % number of normal modes fac = 1/mu/i. titx... z0). nf = 600. ‘c’). df = fmax/nf. x.ˆ2. function [ ] = SH Plate(x. z0) 233 title(tit). z./W.25/mu/i. pause. z0 = z0/h. H = H + besselh(0. F. plot (F. % Frequency domain solution via normal modes nm = 20.. for j=-ni:ni t1 = sqrt(x2+(dz-j*h2)ˆ2)/cs.t2*W). end H = fac*H. U. for j=1:nm j1 = pi*j. titx..3f. title(tit).. tit.ezp’. t2 = sqrt(x2+(sz-j*h2)ˆ2)/cs. tit = sprintf(.5*exp(-i*x*W).

end H = fac*H. plot (F. if t2==T2(k1). F.ezp’. t2) k = find(t2<T2). H. xlabel(titx). title(tit). return function [U] = response(T2. plot (F./real(sqrt((T2(k:n)-t2)))]. hold on. hold off.‘r’). real(H)). end end return . titx. tit. k1 = k-1. 1. grid on./K. EasyPlot(‘SH plate FD2.k1). U = [zeros(1. U(k1)=inf. pause. imag(H).234 MATLAB programs K = sqrt(W2-j1ˆ2). else k = k(1). if isempty(k) U = zeros(size(T2)). H = H + cos(j1*z)*cos(j1*z0)*exp(-i*x*K). n = length(T2). ‘c’). close all.

dz = z-z0. frequency if z>h | z0>h disp(‘Error. x2 = xˆ2. tmax = N*tfund. t2))*(-1)ˆj. cs2 = csˆ2.response(T2. for j=-N:N t1 = (x2+(dz-j*h2)ˆ2)/cs2.3f due to SH line load in stratum at z=%5.ˆ2. end U = fac*U. ’Impulse response at x=%5. z is positive up % Arguments: % x. x. plot (T.. % fundamental period of plate ffund = 1. titx = ‘Time’.. .0001. z. return end % Time domain solution via method of images % Choose number of images ni=N such that other images arrive after tmax tfund = 2*h/cs. source/receiver z must be <= than thickness of plate’). the frequency domain solution is determined again with % the normal modes method (gives more accurate answers). % mass density cs = 1.3f’. nt = 400. sz = z+z0. z./tfund. U = zeros(size(T)).. % shear wave velocity h = 1. % Shear modulus rho = 1. function [ ] = SH Stratum(x. z=%5. T2 = T. h2 = 2*h. time and max. z. z0) % SH line source at elevation zo above the origin in an elastic stratum % Stratum has unit thickness and material properties % Origin is at the bottom surface.3f. U). % parameter for max. z0). dt = tmax/nt. tit = sprintf(. % fundamental frequency (Hz) fac = 1/2/pi/mu.. t2 = (x2+(sz-j*h2)ˆ2)/cs2. z0) 235 function [] = SH Stratum(x. z = coordinates of receiver % z0 = elevation of source % Method of images used for both the time domain and the frequency domain % In addition.. U = U + (response(T2. t1) . T = [0:nt]*dt. % material damping N = 10. % plate thickness xi = 0. mu = 1.

title(tit). for j=1:nm j1 = pi*(j-0.ezp’.3f’.. ‘r’). x = abs(x/h). titx. titx.3f’. % Frequency domain solution via method of images ni = 100. z = z/h. tit. pause.3f due to SH load in stratum at z=%5. for j=-ni:ni t1 = sqrt(x2+(dz-j*h2)ˆ2)/cs. hold on. z0).. % Frequency domain solution via normal modes nm = 20.‘r’). z.3f. end H = fac*H. F = [df:df:fmax]. tit = sprintf(.2.. H = H + (besselh(0. H = zeros(size(W)). W = 2*pi*F.. F. t2 = sqrt(x2+(sz-j*h2)ˆ2)/cs.5). plot (F. imag(H). tit = sprintf(.. EasyPlot(‘SH stratum FD1. W2 = W. ‘Green function at x=%5. df = fmax/nf. EasyPlot(‘SH stratum TD. W = W*h*(1-i*xi)/cs.. z.ˆ2. 236 MATLAB programs xlabel(titx). x. plot (F. hold off. z=%5.ezp’. U. T. . xlabel(titx). ‘c’).. grid on.. z=%5. tit. z0). real(H)). z0 = z0/h. % number of normal modes fac = 1/mu/i. H. grid on. ‘Green function at x=%5.25/mu/i. nf = 600.t2*W))*(-1)ˆj.2. title(tit)..3f. % Number of images fac = 0. pause. fmax = N*ffund. titx = ‘Frequency (rad/s)’.besselh(0.3f due to SH load in stratum at z=%5. x. H = zeros(size(F))..t1*W) .

grid on.ezp’. xlabel(titx). t2) k = find(t2<T2).function [ ] = SH Stratum(x. ‘c’). 1. F. hold off./K. else k = k(1). plot (F. if isempty(k) U = zeros(size(T2)). U(k1)=inf. if t2==T2(k1). end H = fac*H. tit. return function [U] = response(T2. real(H)). U = [zeros(1. hold on. n = length(T2). plot (F. close all. title(tit). k1 = k-1. H = H + sin(j1*z)*sin(j1*z0)*exp(-i*x*K)./real(sqrt((T2(k:n)-t2)))]. z. EasyPlot(‘SH stratum FD2.‘r’). pause. H. z0) 237 K = sqrt(W2-j1ˆ2). titx. end end return . imag(H).k1).

BC2). % cs/cp tfund = 2*h*alfa/cs. z = coordinates of receiver % z0 = elevation of source % pois = Poisson’s ratio % BC1 = Boundary condition at bottom (‘FC’ or ‘CF’) % BC2 = Boundary condition at top (‘FC’ or ‘CF’) % Method of images used for both the time domain and the frequency domain % In addition. 238 MATLAB programs function [] = SVP Plate(x. and it is MUCH faster). % Default data mu = 1.0001. % fundamental period of plate (dilatation) ffund = 1. poi. T2 = T. z. T = [0:nt]*dt. return end alfa = sqrt((1-2*poi)/(2-2*poi)). % choose ni=N such that other images arrive after tmax tmax = 10. n = length(T). z0. BC2) % SVP line source at elevation zo in an elastic plate % with mixed boundary conditions % Plate has unit thickness and material properties % Origin is at the bottom surface. z is positive up % Arguments: % x. BC = strcat(BC1. % parameter for max. % plate thickness xi = 0. % shear wave velocity h = 1. cs2 = csˆ2. dt = tmax/nt. %%%%%%%%%%%% nt = 400./tfund. % material damping N = 10. sz = z+z0. % mass density cs = 1. % boundary conditions % Method of images % Parameters for time domain %tmax = N*tfund. BC2 = upper(BC2). % Shear modulus rho = 1. time and max. BC1. frequency if z>h | z0>h disp(‘Error. % fundamental frequency (Hz) x2 = xˆ2. . source/receiver z must be <= than thickness of plate’). the frequency domain solution is determined again with % the normal modes method (Gives more accurate answers. dz = z-z0. BC1 = upper(BC1).ˆ2. h2 = 2*h.

:).ezp’. z0.function [ ] = SVP Plate(x. xlabel(titx).3f due to SVP line load in plate at z=%5. direc = [‘xx’.:)-U2(2:3.:)-U2(2:3. tit(2:3) = direc(j1:j2). z1. titx = ‘Time’. cs.:)-U2(1:2.U1(2:3. grid on. alfa.U1(2:3. EasyPlot(fname. fname(2:3) = direc(j1:j2). ‘Uxx at x=%5.:)+U2(1:2.3f’.:). x.:). j1 = j2-1. ‘zx’.U1(2:3. U2 = impulse response(r2. BC1. nf = 500. df = fmax/nf. title(tit).:)+U2(2:3. plot (T. % Number of images frequency domain fmax = 5. .3f. U(j. mu. z.:)).:)+U2(1:2. z=%5.. U(j. for j=-N:N z1 = dz-j*h2. W = 2*pi*F. elseif BC==‘CFCF’ U = U+[U1(1:2. else disp(‘Not a valid boundary condition’).:).. if BC==‘FCFC’ U = U+[U1(1:2. T. x.:)].n)..:)-U2(1:2. elseif BC==‘FCCF’ U = U+(-1)ˆj*[U1(1:2.:)].. ‘xz’. ‘zz’]. n = length(W). alfa.U1(2:3.:)]. z. z2. r2 = sqrt(x2+z2ˆ2). for j=1:4 j2 = 2*j. T2). tit. T2)..:)+U2(2:3. x. z0). F = [df:df:fmax]. ‘r’). poi. pause. U1 = impulse response(r1. end % Frequency domain solution via method of images % Parameters frequency domain ni = 50. z2 = sz-j*h2. elseif BC==‘CFFC’ U = U+(-1)ˆj*[U1(1:2. cs. return end end tit = sprintf(.:). r1 = sqrt(x2+z1ˆ2). fname = ‘Uxx SVP plate TD. BC2) 239 U = zeros(4.:)]. titx. mu.

H2 = Green(r2.:)]. tit(2:3) = direc(j1:j2). z1.:).:).:)-H2(2:3. W). x. title(tit). for j=-ni:ni z1 = dz-j*h2.:)-H2(1:2. EasyPlot(fname.:).. . elseif BC==‘FCCF’ H = H+(-1)ˆj*[H1(1:2. ‘Hxx at x=%5.. z0). x. tit. pause. z. z0). ‘Green function at x=%5. z2. real(H(j.H1(2:3. ‘c’). plot (F. H(j.:)))..:).:)+H2(2:3. ‘xz’.:)-H2(1:2.H1(2:3. hold on. j1 = j2-1.3f due to SVP load in plate at z=%5. z2 = sz-j*h2.. % number of normal modes tit = sprintf(. cs. direc = [‘xx’.H1(2:3. 240 MATLAB programs H = zeros(4. titx = ‘Frequency (rad/s)’. mu.3f’. x.:)+H2(1:2. imag(H(j. fname = ‘Hxx SVP plate FD1. z=%5.:)]. elseif BC==‘CFFC’ H = H+(-1)ˆj*[H1(1:2. else disp(‘Not a valid boundary condition’). z. H1 = Green(r1..:)-H2(2:3. elseif BC==‘CFCF’ H = H+[H1(1:2. alfa.3f.3f. plot (F. cs.:)+H2(1:2.3f due to SVP line load in plate at z=%5. grid on. mu.. x. end clear U H T % Frequency domain solution via normal modes nm = 20.ezp’.:)]. alfa. ‘zx’. titx.3f’..H1(2:3.. z=%5. for j=1:4 j2 = 2*j.. ‘r’). hold off. W). r1 = sqrt(x2+z1ˆ2).. ‘zz’].:)).:). return end end tit = sprintf(.n). fname(2:3) = direc(j1:j2). if BC==‘FCFC’ H = H+[H1(1:2.:)]. xlabel(titx). r2 = sqrt(x2+z2ˆ2). F.:)+H2(2:3.

H(3..5). ‘zx’.ˆ2. z=%5. if BC1==‘FC’ H(1. H(4.:) = H(1. H(4./Ks. H4 = Ks. z. Ws = W*h*(1-i*xi)/cs.cs*sr*H2. end end clear H1 H2 H3 tit = sprintf(. cr = cos(j1*z).3f due to SVP line load in plate at z=%5. else j1 = pi*(j-0.:) = H(2. ‘xz’. H = zeros(4. H(3. z0 = z0/h.ˆ2. cs = cos(j1*z0). z0).:) + cs*cr*H4.:) = H(1.5*Ws. if BC==‘FCFC’ H(1.:) = 0. elseif BC1==‘CF’ H(1. H2 = j1*(Ep-Es). Wp2 = Wp. x = abs(x/h).:) = H(4. Wp = alfa*Ws. Kp = sqrt(Wp2-j2).:) = H(3.:) + ss*cr*H2. end for j=1:nm if BC1==BC2 j1 = pi*j. H(2..:) . poi.:) + ss*sr*H1.*Ep+j2*Es.3f. x.:) .:) + ss*sr*H4.:) + ss*cr*H2.cs*sr*H2. ‘Hxx at x=%5. z.function [ ] = SVP Plate(x. BC1.5*Wp. Ep = exp(-i*x*Kp). ‘zz’]. H1 = Kp.*exp(-i*x*Wp). . H(2. direc = [‘xx’.:) + cs*cr*H1.*exp(-i*x*Ws). z0..3f’.*Es+j2*Ep. end j2 = j1ˆ2. Ks = sqrt(Ws2-j2)... Es = exp(-i*x*Ks). Ws2 = Ws. elseif BC==‘CFCF’ H(4. z = z/h.:) = H(2.:) = H(4.length(Ws)).:) = 0. BC2) 241 sgn = sign(x).:) = H(3. ss = sin(j1*z0). sr = sin(j1*z)./Kp.

c = x/r.besselh(0. Ws = ts*W. F. hold off. Wp = alfa*Ws.:)). title(tit).Ws). s=sintheta).2. tp = alfa*ts. plot (F. j1 = j2-1. mu. for j=1:4 j2 = 2*j. cs. psi = (besselh(1. return else ts2 = tsˆ2.2. . f = 0. ‘c’).:). chi = alfaˆ2*besselh(2. gzx. z. imag(H(j. ‘r’).Ws).besselh(2. xlabel(titx). x. W) % Returns the Green’s functions for an SVP line load in a full space % r=source receiver distance.2. return function [G] = Green(r. gzz] return function [U] = impulse response(r.25*i/mu. chi*c*s. x. end close all.ezp’. H(j. T2) % Returns the impulse response functions for an SVP line load in a full space ts = r/cs. grid on.2. tit(2:3) = direc(j1:j2). % sin(theta) G = f*[psi+chi*cˆ2. k = find(tp2<T2). EasyPlot(fname. real(H(j./Ws2. hold on. H(j. tp2 = tpˆ2.:). sgn. % cos(theta) s = z/r. fname(2:3) = direc(j1:j2). z. sgn. % [gxx. k1 = k-1./Ws .Wp). pause.Wp)).:) = f(j)*H(j. plot (F.242 MATLAB programs fname = ‘Hxx SVP plate FD2.2. -i]/mu. alfa. cs. k = k(1).Ws)-alfa*besselh(1. psi+chi*sˆ2]. alfa=cs/cp ts = r/cs. tit. if isempty(k) U = zeros(3. alfa.length(T2)). f = [-i.:))). c=cos(theta). mu. titx.

if tp2==T2(k1). S = sqrt((T2(k:n)-ts2)).k1). CHI(k1)=inf.function [ ] = SVP Plate(x. c = x/r. k1 = k-1. % sin(theta) U = f*[PSI+CHI*cˆ2. CHI = CHI . uzz] return .-S/ts2]. end k = find(ts2<T2).[zeros(1. CHI(k1)=inf. % yes.k1).k1). z.S/ts2+1. BC2) 243 n = length(T2).alfaˆ2. % cos(theta) s = z/r. PSI+CHI*sˆ2]. % [uxx.2*S/ts2+1. if tp2==T2(k1).k1). PSI = [zeros(1. end end end f = 1/2/pi/mu./S]-2*PSI. poi. PSI = PSI + [zeros(1. not tp2 CHI = [zeros(1. if ∼isempty(k) k = k(1). n = length(T2). uzx. we must divide by ts2. CHI*c*s./S]./S]. S = sqrt((T2(k:n)-tp2)). z0. PSI(k1)=inf. BC1.

spheroidal modes % Actual frequencies (rad/s) are obtained multiplying by Cs/R % where Cs is the shear wave velocity. it uses the Bessel functions of half order % instead of the actual spherical Bessel functions. zspher] = spheroidal (pois. end tit = sprintf(‘Torsional modes for m =%d’. % initial search step if wmin<dz zs=[dz:dz:wmax]. else zs=[wmin:dz:wmax]. if m>0 f = dettors(m. z2.zs(1)). pause. z2 = ztors(k)+dz. Reason: % The factors sqrt(pi/2z) cancel out in the determinant equations % % Makes first a rough search in the search interval to determine approximate % values of the roots. nz = length(ztors). m.dz.e-5. and R is the radius % % For enhanced accuracy. then refines the results using the bisection method % % Results agree with Eringen & Suhubi. m. wmax) % Computes the torsional and spheroidal modes of a solid sphere. titx = ‘Mode number’. II. 1.zs). z1. .pois).244 MATLAB programs function [ztors.f). torsional modes % zspher = dimensionless frequencies. and % plots the two determinant functions whose zero crossings define the roots % Assumes a sphere of radius R=1 and shear wave velocity Cs=1 % Note: the modes are independent of the azimuthal index n % % Input arguments: % pois = Poisson’s ratio % m = radial index % wmin = minimum frequency for search % wmax = maximum frequency for search % Output arguments: % ztors = dimensionless frequencies.m). 814. Vol.1. ztors(k) = bisect2(‘dettors’. m). ztors = zerox(f. grid on. pp.ezp’. wmin. axis([0 wmax -1 1]).818 dz = 0. fname = sprintf(‘Tors%d. Elastodynamics. for k=1:nz z1 = ztors(k)-dz. end plot(zs.

f = jsm1-(m+2)*jsm.ezp’./zs). end tit = sprintf(‘Spheroidal modes for m =%d’. % Cs/Cp zp = a*zs.pois). f = detspher(m. titx. tit./zs./zp. [1:nz].2*(2*jpm1-(m+1)*(m+2)*jpm. jsm1 = besselj(m-0. z = bisect2(‘detspher’.pois)./zp. end end EasyPlot(fname. % Constrained modulus a=sqrt(1/cm). axis([0 wmax -1 1]).function [ztors. jpm1 = besselj(m-0.pois) jsm = besselj(m+0. ‘r’) else disp(‘No torsional modes exist for m=0’).zp). pause. 1. nz = length(zspher). z2 = zspher(k)+dz. grid on./zs. f = f11.zs).f). titx. zspher.5. EasyPlot(fname.4*jpm1. close. [1:nz].5. plot(zs./zp. jsm1 = besselj(m-0. f12 = 2*m*(m+1)*(jsm1-(m+2)*jsm.dz./zp). f11 = -cm*jpm.*f21. if m>0 jsm = besselj(m+0.zs. else f = -cm*jpm.5./zs). wmax) 245 close. m.m).e-5.zs).zs). f22 = -jsm-2*(jsm1-m*(m+2)*jsm. zspher(k) = z./zs. ztors = []. tit.5. z1./zp). if z∼=[]. z2.zs. ‘r’) return function [f] = dettors(m. zspher] = spheroidal (pois. ztors. fname = sprintf(‘Spher%d. f21 = 2*(jpm1-(m+2)*jpm. .pois) cm = 2*(1-pois)/(1-2*pois). m.5. wmin. for k=1:nz z1 = zspher(k)-dz. zspher= zerox(f.5.zp).zs).*f22-f12. jpm = besselj(m+0.zs. m). return function [f] = detspher(m.zs(1)).

dz. fright = feval(f. while (sg1 == 0) nz = nz+1. return elseif fright == 0 center = right.m.left.m. z = []. z1) % Finds the zero crossings of a vector n = length(a). end end return function z = zerox(a. pois) % Finds the root of a function f in interval [left. left. m. if (n < 2) return.pois) <= 0 right = center. z = [z. else left = center.246 MATLAB programs end return function center = bisect2(f. sg1 = sign(a(j)). while j<=n & sign(a(j))==sg1 j = j+1. tolerance. while j<n sg1 = sign(a(j)). end j = 1. right. j = j + 1. if (fleft == 0) center = left. nz = 0. end j = j + 1.pois). right] % where a root is known to exist. end if j<=n .m. if feval(f.pois)* feval(f. return elseif fleft* fright > 0 center = [].right.m. return end while (right-left) > tolerance center = (left + right)/ 2.pois).left.j].center. Uses the bisection method fleft = feval(f.

j-dj].function [ztors. wmin. dj = abs(a(j))/(abs(a(j-1))+abs(a(j))). z = [z. zspher] = spheroidal (pois. wmax) 247 % found a root nz = nz+1. return . end end z = (z-1)*dz+z1. m.

5f\n’. return . end end fclose (fout). titx. ‘‘%15. ‘/sm off\n’).real(U(n. [T. ‘%15.nc] = size(U).248 MATLAB programs function EasyPlot(fname.5f %15. if flag==‘r’ fprintf (fout. ‘/et x “%s”\n’.:)). elseif flag==‘c’ fprintf (fout. ‘/sd off\n’).:)]). tit). U. else ‘Not a valid option in EasyPlot’. flag) % Creates files for display with the EasyPlot program fout = fopen (fname. ‘//nc\n’). fprintf (fout.imag(U(n. tit.5f %15. [T. ‘/og on\n’). fprintf (fout. [nr. T. fprintf (fout. titx). fprintf (fout. ‘/et g “%s”\n’.5f %15.:))]). for n=1:nr fprintf (fout. fprintf (fout.U(n. ‘w’).5f\n’.

1d-7*x2-2../xx.555556d-002).570796327d0-fx.function [si] = cisib(x) 249 function [si] = cisib(x) % ============================================= % Purpose: Compute cosine and sine integrals % Si(x) and Ci(x) % Input: x --..67732d0). while (n3<n & x(n3)<=1) n3 = n3+1./((((x2 +40. 482.834d-5)./xx-gx.*x2-5..ˆ2.23628d0).*x2+.. si(n3:n) = 1.018498d0). fx = ((((x2+38. x2 = xx. end if (n3>n1) n2 = n3-1.690326d0).Si(x) % ============================================= %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% added stuff n = length(x)..*x2+. gx = ((((x2+42.*x2+265. 335.*x2+157. end ..10d-6).*x2+.021433d0).102495d0). end if (n3<=n) xx = x(n3:n).0d-8*x2+3.*cos(xx). 1.66667d-003).*x2+1.187033d0). *x2+352. ci(n3:n) = fx.0).Ci(x) % SI --.*x2+302.ˆ2./xx.196927d0).577215665d0+log(xx).Argument of Ci(x) and Si(x) % Output: CI --..624911d0).*x2+449.*x2+38.3148d-4). n1 = n1+1.*x2-0. end n3 = n1. 1.485984d0)..*sin(xx).*x2+570. ci(n1:n2) = ((((-3.*x2+1114. 21.*x2-2.*x2+.757865d0).041667d-2). end %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% if (n1>1) n0 = n1-1.242855d0).105423d0)..978885d0)./((((x2 +48.*sin(xx). x2 = xx./xx.027264d0).*x2+.*x2+0. ci(1:n0) = -1.*x2+.. si(1:n0) = 0.821899d0).*xx.25). n1 = 1./xx-gx..*cos(xx).0d+300. si(n1:n2) = ((((3.0d0. 322.. while (n1<n & x(n1)<=0) x(n1) = 0. xx = x(n1:n2).

. . t = [..1316886384491766..ˆ2. . ..’ ‘Should be size(N)=1 or size(N)=size(M)’ el3 = 0.. % Arguments: phi .9122344282513259.N... .M) % [EL3] = ELLIPINT3 (phi...8391169718222188. psi = ang/2. el3 = zeros(size(M)). first argument in ellipint3 cannot be negative’ el3 = 0. .1761400713915212d-1.parameter (some authors use c=-n) % N can be also be scalar or vector... it must agree in size with M % Definition: If n. w = [..7463319064601508... elseif length(N) ∼= length(M) ‘Error.... s1 = sin(t1).. m are elements of N..3737060887154195.1019301198172404. M.250 MATLAB programs function [el3] = ellipint3(phi.1527533871307258].. wrong size of second argument in ellipint3. s2 = sin(t2). but if % the latter.9931285991850949..1181945319615184..5108670019508271. evaluated for each value of N. . ang = phi*pi/180.4060142980038694d-1.M) returns the elliptic integral of % the third kind.7652652113349734d-1]. then % % phi % el3 = integ | [dt/((1+n*sinˆ2(t))*sqrt(1-m*sinˆ2(t))) % 0 % % Observe that m = kˆ2 is the square of the argument % used by some authors for the elliptic integrals % Method: 10-point Gauss-Legendre quadrature if (phi < 0) ‘Error. return end tol = 1-1d-8.... t2 = psi*(1-t).1420961093183820..1491729864726037.Upper limit of integration (in degrees) % M ..N. return end if length(N) ==1 N = N*ones(size(M)).2277858511416451. .Modulus (some authors use k=sqrt(m)) % M can be a scalar or vector % N --. M % Can also be used to obtain the elliptic integral % of the first kind by setting N=0.9639719272779138. t1 = psi*(1+t).6360536807265150.ˆ2.6267204833410907d-1.8327674157670475d-1.

if phi<=90 % assuming phi is in degrees here if k2*s>=tol | -n*s >= tol flag = 0.*sqrt(1-k2*s1)). n = N(j). end elseif k2 >= tol | -n >=tol flag = 0. flag = 1. else el3(j) = inf. end end el3 = psi*el3. end if flag f1 = 1. for j=1:length(M) k2 = M(j).function [el3] = ellipint3(phi. return . f2 = 1. el3(j) = (f1+f2)*w’./((1+n*s2)./((1+n*s1). M) 251 s = sin(ang)ˆ2. N.*sqrt(1-k2*s2)).