Testing
Testing
MSc. Thesis
May 2020
A thesis submitted to Khalifa University of Science and Technology in accordance with the
requirements of the degree of MSc. in Petroleum Geosciences in the Department of Earth
Sciences.
i
Computations of the complex phase-, group-,
ray-velocity and attenuation for seismic ray
tracing in viscoelastic anisotropic media
by
at
Khalifa University
Thesis Committee
May 2020
ii
Abstract
Tayammum Mohamed Ali Abdulqader Alkatheeri, Computations of the complex phase-,
group-, ray-velocity and attenuation for seismic ray tracing in viscoelastic anisotropic
media, M.Sc. Thesis, MSc in Petroleum Geosciences, Department of (Earth Sciences), Khalifa
University of Science and Technology, United Arab Emirates, May 2020.
Seismic imaging is still the sole reconnaissance means to provide subsurface information with
no direct tangible contact. The resulting outcome of the seismic exercise greatly depends on
the understanding of the sent energy, as well as the medium that the energy waves are travelling
through. Too many assumptions and approximations would render the seismic depreciating in
value in direct proportion to the degree of violation of reality by those assumptions. The earth
is often assumed elastic and isotropic to simplify calculations while it is known that the earth
is in fact attenuating and does exhibit directional trends. To accurately model Earth (through
seismic surveying), it should be regarded as more complex than is being done currently. The
preliminary stages of ray tracing include the calculation of complex-valued wave quantities,
such as the complex phase slowness vector, the phase-, group- and ray-velocity and the ray
attenuation. These parameters become complex-valued because of the viscoelasticity of the
rock media which has to be taken into account because it causes amplitude decay of the wave,
hence the imaginary part, which is directly linked with attenuation of the propagating wave
energy. This complicates the problem. We present here three different methods which enable
the computation of the wave quantities: the iterative solver method, the perturbation method,
and the RSD method. The first proved satisfactory but computationally expensive due to
iterations with fine steps. The second proved successful in most of the cases but failed with
some highly complex media. The third and last proved efficient and fitting even at the
challenging triplication of wave-fronts in vertical shear waves. Further development of such
methods is needed especially for more complex media with strong anisotropy and strong
attenuation, because they mimic the earth more truthfully therefore they need to be
investigated. Subsequently, after obtaining such parameters, seismic ray tracing for accurate
complex travel-times can be conducted.
Indexing Terms: viscoelastic anisotropic media (VEAM); amplitude attenuation; ray and
phase velocities; transverse isotropy; seismic ray tracing
iii
Acknowledgement
I would like to acknowledge and thank all those who contributed greatly to this study. First and
foremost, I would like to express my sincerest appreciation and gratitude to my advisor, Dr.
Bing Zhou for his unrelenting support and constant observation throughout this study. I would
also like to extend my appreciation to my co-advisor Dr. Mohammad AlKhaleel for his
acknowledge the support of the committee members, Dr. Youcef Bouzidi and Dr. Mohamed
Riahi, and the pleasure it was interacting with them. Special thanks to the team members, Dr.
Jianlu Wu, Dr. Qijnjie Yang, Mr. Shangbei Yang, Mr. Moosoo Won and Mr. Weining Liu for
their helpful contributions. Many thanks go to my family and close friends for their
unconditional support. Completing this would not have been possible without those incredible
people.
iv
Declaration and Copyright
Declaration
I declare that the work in this thesis was carried out in accordance with the regulations of
Khalifa University of Science and Technology. The work is entirely my own except where
indicated by special reference in the text. Any views expressed in the thesis are those of the
author and in no way represent those of Khalifa University of Science and Technology. No part
of the thesis has been presented to any other university for any degree.
07/05/2020
Date: _________________________________
Copyright ©
No part of this thesis may be reproduced, stored in a retrieval system, or transmitted, in any
form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise,
without prior written permission of the author. The thesis may be made available for
consultation in Khalifa University of Science and Technology Library and for inter-library
lending for use in another library and may be copied in full or in part for any bona fide library
or research worker, on the understanding that users are made aware of their obligations under
copyright, i.e. that no quotation and no information derived from it may be published without
the author's prior consent.
v
Contents
Abstract .................................................................................................................................................. iii
Acknowledgement ................................................................................................................................. iv
Declaration and Copyright ...................................................................................................................... v
List of Figures ....................................................................................................................................... vii
List of Tables .......................................................................................................................................... x
1. Introduction ......................................................................................................................................... 1
1.1 Viscoelastic anisotropic rocks ....................................................................................................... 1
1.2 Seismic ray tracing in viscoelastic anisotropic media................................................................... 4
1.3 Goals of this study ........................................................................................................................ 5
2. Plane-wave theory in VEAM .............................................................................................................. 7
2.1 Inhomogeneous complex phase slowness vector .......................................................................... 7
2.2 Complex travel-time in the phase velocity form ......................................................................... 11
2.3 Phase attenuation ........................................................................................................................ 13
2.4 The Christoffel Equation ............................................................................................................. 16
3. Group and ray velocity in VEAM ..................................................................................................... 23
3.1 Group velocity vector ................................................................................................................. 23
3.2 Ray velocity vector ..................................................................................................................... 27
3.3 Complex travel-time in the ray velocity form ............................................................................. 29
3.4 Ray attenuation ........................................................................................................................... 31
4. Computations of complex phase and ray velocity ............................................................................ 33
4.1 Iterative algorithm ....................................................................................................................... 33
4.2 Perturbation method .................................................................................................................... 40
4.3 Real Slowness Direction approximation ..................................................................................... 48
5. Numerical Results ............................................................................................................................. 52
5.1 VTI models ................................................................................................................................. 52
5.2 TTI model ................................................................................................................................... 60
5.3 ORT models ................................................................................................................................ 62
6. Conclusions ....................................................................................................................................... 65
References ............................................................................................................................................. 69
Appendices............................................................................................................................................ 73
Appendix A ....................................................................................................................................... 73
Appendix B ....................................................................................................................................... 75
Appendix C ....................................................................................................................................... 76
vi
List of Figures
Figure 1: A very simplistic illustration showing the absorption/attenuation effect on seismic
waves as they travel through Earth and how wave energy decays with time. Wave energy and
shape are altered (vertical axis is energy, horizontal axis is travel-time) .................................. 1
Figure 2: Simplistic models of different types of transversely isotropic media: (a) vertical
symmetry axis (vertical transverse isotropy VTI), (b) horizontal symmetry axis (horizontal
transverse isotropy HTI), and (c) tiled symmetry axis (tilted transverse isotropy TTI) (Zhang,
2017) .......................................................................................................................................... 2
Figure 3: The stress and strain (cause & effect) impacts relationship on different materials at
different points in time. Note the different recovery modes which are associated with the
Figure 4: Illustrating the wave-field models which arise from the nature of the source wave as
Figure 5: Illustration behind the concept of plane waves. The source is far away enough from
the receiver that until reaching there, the wave arrives in the form of its asymptote (planar).
ȓ, with two distinct slowness directions associated with the complex slowness vectors p, and τ
is the travel-time; the real parts indicated in red and imaginary parts in blue ......................... 10
Figure 7: Multiple models of VTI media which are derived from geological features ........... 19
Figure 8: Multiple models of TTI media which are derived from geological features and are a
Figure 9: Simple illustration of the polarization direction in 3D of the different wave modes
qP, qSV, qSH. The q prefix added to any mode denotes quasi- which is a more
vii
realistic/representative visualization of particle motion with respect to the overall direction of
the wave travel; where the angle between wave propagation and particle vibration is not exactly
900 ............................................................................................................................................ 23
Figure 10: A simplistic illustration of a ray-path going from a source S to a receiver R through
time τ with slowness vector p (implicitly incorporating phase velocity c) and group velocity U
Figure 11: A flowchart summarizing the first branch of the iterative solver algorithm steps,
focusing on the iterative element, to obtain the ray velocity, attenuation and quality factor .. 38
Figure 12: A flowchart summarizing the second branch of the iterative solver algorithm steps,
focusing on the iterative element, to obtain the ray velocity, attenuation and quality factor .. 39
Figure 13: Simplistic illustration of the concept of perturbation. Properties of the background
are summed with properties of a distinct disturbance deviating from the background, resulting
Figure 14: A flowchart summarizing the perturbation method steps to obtain the ray velocity,
Figure 15: A flowchart summarizing the RSD approach steps to obtain the ray velocity,
Figure 16: Ray and phase velocities (first row), ray and phase attenuations (second row), and
ray and phase quality factors for the three wave modes: qP (first column), qSV (second column),
qSH (third column) for a shale model of a viscoelastic VTI medium ...................................... 53
Figure 17: The computed ray velocity, attenuation, and quality factor of the qP-wave in Model
A1. The absolute relative error is also plotted with respect to each method in comparison with
Figure 18: The computed ray velocity, attenuation, and quality factor of the qP-wave in Model
B1. The absolute relative error is also plotted with respect to each method ........................... 56
viii
Figure 19: The computed ray velocity, attenuation, and quality factor of the qSV-wave in Model
A1. The absolute relative error is also plotted with respect to each method ........................... 57
Figure 20: The computed ray velocity, attenuation, and quality factor of the qSV-wave in Model
B1. The absolute relative error is also plotted with respect to each method ........................... 57
Figure 21: The computed ray velocity, attenuation, and quality factor of the qSH-wave in Model
A1. The absolute relative error is also plotted with respect to each method ........................... 58
Figure 22: The computed ray velocity, attenuation, and quality factor of the qSH-wave in Model
B1. The absolute relative error is also plotted with respect to each method ........................... 59
Figure 23: Ray and phase velocities (first row), ray and phase attenuations (second row), and
ray and phase quality factors for the three wave modes: qP (first column), qSV (second column),
qSH (third column) for a tilted shale model of a viscoelastic TTI medium based on the Bond
transformation .......................................................................................................................... 61
Figure 24: The absolute relative errors of the ray velocity, ray attenuation, and ray quality factor
Figure 25: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality
Figure 26: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality
ix
List of Tables
Table 1: Elastic moduli and quality matrix of a viscoelastic VTI shale .................................. 52
Table 3: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality
Table 4: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality
Table 5: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality
Table 6: Elastic moduli and quality matrix of 3 viscoelastic ORT media ............................... 62
Table 7: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality
x
1. Introduction
Through seismic wave analysis, it is possible to visualize the subsurface and image it.
This is highly dependent on how the waves travel through the ground and are received back on
the surface. The better the understanding gets for how these waves propagate through the
multiple rock layers, the better and the closer to reality the imaging is going to be. This is
especially needed for complex geological features such as heavy thrusting and cumulative
processing of the seismic data; and it is that the rock media which the seismic waves travel
through are elastic and isotropic. This is a violation of the real situation because rock media
are almost always not elastic nor isotropic; they facilitate energy decay, otherwise known as
attenuation/absorption (Figure 1) as waves travel through them, this is also associated with the
medium properties like pore fluids and the presence of fractures or joints (Carcione, 1995). Not
only that, but these changes vary in magnitude along different directions, which is otherwise
known as anisotropy (Figure 2). Waves’ velocities vary as they cross different rock types, and
Figure 1: A very simplistic illustration showing the absorption/attenuation effect on seismic waves as they travel through Earth
and how wave energy decays with time. Wave energy and shape are altered (vertical axis is energy, horizontal axis is travel-
time)
1
Figure 2: Simplistic models of different types of transversely isotropic media: (a) vertical symmetry axis (vertical transverse
isotropy VTI), (b) horizontal symmetry axis (horizontal transverse isotropy HTI), and (c) tiled symmetry axis (tilted transverse
isotropy TTI) (Zhang, 2017)
It is known throughout the literature that rock layers are typically anisotropic and
attenuating (Tao and King, 1990; Burton 2007; Behura et al., 2012; Shekar and Tsvankin, 2012;
Zhubayev et al., 2016; Vavryčuk et al., 2016). Anisotropy means that the medium is not
homogeneous throughout; its properties change from sample to sample across the medium
along certain axes (directional dependency). In the real world, there are several examples of
naturally anisotropic geologic formations. Many factors result in forming such properties in
rocks; it might come down to the mineralization of the rock’s constituents where they have
preferred orientations reflected in the grain deposition like in the example of shales or olivine-
rich rocks, or in other cases where the depositional environment plays a role like in the cases
of mud cracks in clays or thin beds in limestones (Thomsen, 1986). This intrinsic or inherent
anisotropy can be divided into four broad categories: crystalline anisotropy, constraint induced
Attenuation practically represents the gradual loss of energy as that energy travels
through a medium due to the various interactions with the various components of the
surroundings. A medium that is perfectly elastic does not exhibit any attenuation, and that is
why attenuation is associated with inelastic or viscoelastic media; media which get
2
permanently deformed when subjected to stress (Figure 3). These types of media get
permanently deformed because they absorb portions of the energy travelling across them.
Figure 3: The stress and strain (cause & effect) impacts relationship on different materials at different points in time. Note the
different recovery modes which are associated with the different medium responses to stress (inflicted energy)
In this particular context, the energy is the seismic waves and the medium is Earth’s
subsurface. The combined effects of both of the mentioned phenomena result in the directional
variation of the seismic wave velocities and amplitudes as they travel underground. Velocity is
the main parameter for seismic studies since only through it can the subsurface be modelled
and visualized. There have been some theoretical and practical approaches to investigate the
anisotropic attenuation of rocks but they are considered rough estimates. The theory has
succeeded in the modelling but as for experimental results, they have been based on somewhat
invalid/limiting assumptions. The difficulty arises from the association of viscoelasticity with
complex numbers, wherein the real part of the quantities describes the propagation of wave
energy while the imaginary part represents the attenuation of the wave energy (Corcione,
2007). The challenges are significantly less apparent when the medium is assumed to be elastic;
no attenuation of energy is exhibited and hence all parameters are real-valued. This approach
3
have already been studied and explored (Thomsen, 1986). Reality also has another layer of
complexity; anisotropy, where a medium’s attributes vary across it with different directions.
One of the numerical modelling techniques of seismic waves is seismic ray tracing. It
involves simulating the receiver response and predicting the ray-path which the waves follow
as they cross Earth’s media from the starting point, the source, to the end point, the receiver.
In seismic wave modelling, most studies assume that the wave-field is a 2D model, i.e. a line
source (planar waves) and a 2D geological structures. However, the actual seismic sources are
always point sources (spherical/curving waves). Thus, the assumption of 2D wave-field will
undoubtedly cause some discrepancies in data processing and interpretation and ultimately
yield a slightly different image than what the subsurface actually looks like. In addition, most
studies typically regard acoustic media or elastic isotropic media in seismic wave modelling,
but most geological structures in reality are more likely to be anisotropic due to the usually
naturally thin horizontally deposited layers, vertical fractures or parallel cracks. To simplify
the mathematical computations, as abovementioned, most previous research assumes the wave-
field to be 2D, i.e. both the seismic source and the geological model are 2D; however, in
practice, the seismic sources generated are always 3D (called point sources), even though the
geological model is considered 2D, the resultant wave-field caused by the point source become
2.5D (Figure 4). Attention rose to such conditions in order to more realistically approach the
modelling.
Reconstructing the subsurface structures is the main value of seismic data acquisition
and processing. If the seismic data failed to do so, then it becomes irrelevant. Additionally, the
more accurate the subsurface representation by the seismic is, the higher its value becomes and
the closer we are to studying and understanding what we cannot physically make direct contact
4
with. Therefore, it is also critical to study reality as it. The current ray tracing methods mainly
focus on elastic media, far fewer ones deal with viscoelastic anisotropic media (VEAM). On
top of that, most of the seismic ray tomography procedures use the assumption that equates the
response of ray tracing in VEAM and in elastic media, but the fact remains that this is not a
wholly valid assumption in reality. Hence, there is a need to address VEAM differently
Figure 4: Illustrating the wave-field models which arise from the nature of the source wave as well as the geological model
We can consider this study a pioneer in its field as nobody has undertaken the seismic
ray tracing in viscoelastic 2D and 3D heterogeneous anisotropic media for its complexity and
involvement in computation. However, doing this will tremendously increase the accuracy in
seismic data processing and interpretation, as well as the overall quality and perhaps resolution,
because the image that is going to be examined is essentially closer to the reality of Earth’s
natural layering. And this practical application is advantageous over most of the research
ignoring the viscoelastic anisotropic rocks in carbonate reservoirs for example, which make up
roughly half of the world’s hydrocarbon reserves. Moreover, this study aids in eventually
carrying out seismic anisotropy ray tomography to improve the imaging of the aforementioned
and reconstruct the heterogeneity of viscoelastic anisotropic layers using the complex travel-
times of the multi-arrivals. The stress is on carbonates here because they are well known for
5
their complexity and inherent heterogeneity. On top of that, studying and developing in this
domain could prove to be useful and applicable for research involving micro-seismicity, e.g.
technology.
computing the ray-paths and travel-times in both isotopic and anisotropic media, in other
words, seismic ray tracing in arbitrary 2D & 3D VEAM. What is applicable for anisotropic
media is surely applicable for isotropic media. Since VEAM properties are closer to the actual
Earth properties than the traditional assumptions discussed above, understanding and defining
them allow for much more accurate reconstruction of Earth’s layers with a higher resolution of
subsurface interfaces. Finding the complex ray velocity gives the seismic ray speed and
attenuation in arbitrary VEAM, then integrating several factors aids in determining the ray-
path and the calculation of the real and imaginary travel-time (wave energy attenuation), and
anisotropic media. Having a method to accurately trace ray-paths would surely create
opportunities for a more meaningful interpretation of seismic images and deduce more realistic
subsurface structures. Having a robust ray tracing technique, is a critical part of seismic ray
tomography or seismic depth migration for reconstructing the subsurface image of the
viscoelastic anisotropic Earth. This research is to find a more effective and efficient algorithm
to compute the phase and ray velocities, as well as the ray attenuation for the final goal of ray
tracing in any given arbitrary VEAM, which is the natural next step and continuation of the
study. Computing these quantities is fundamental for seismic ray tracing as it cannot be
6
2. Plane-wave theory in VEAM
According to the complex ray theory, the elastic moduli which describe the properties
of a medium; the phase velocity and ray velocity, the ray travel-time and the ray polarization
vector all become complex-valued with real parts to represent the actual energy propagation
aspect of the waves, and imaginary parts to represent the attenuation the wave energy goes
through as it travels an inelastic medium (Hearn and Krebes, 1990a, b; Chapman et al., 1999;
Hanyga and Seredyńska, 2000). Moreover, all of these viscoelastic quantities in our case are,
obviously, complex in value but they are also proved to be frequency dependent (Auld, 1973;
Carcione, 2014). Involving the aforementioned theory in order to address the problems in ray
tracing in viscoelastic media is a solution; but it is often overlooked due to the unavoidable
incorporation of imaginary numbers in the calculations, which renders dealing with the
This is a study involving seismic waves in VEAM, the basis principle is understandably
governed by the equation of motion, or the wave equation, the derived subset. For the sake of
simplifying the problem, the waves are considered planar, meaning that the source is assumed
to be far enough that the wave-front’s bend is negligible and the wave is abstracted into its
tangent resulting in a plane wave (Figure 5). In addition, according to the Fourier Transform,
any seismic signal can be decomposed into its constituents of different combinations of plane
waves.
7
Figure 5: Illustration behind the concept of plane waves. The source is far away enough from the receiver that until reaching
there, the wave arrives in the form of its asymptote (planar). Where τ is the travel-time of the wave and d is distance
where u is the displacement vector, A is the amplitude vector of the wave, i is the imaginary
unit, is the angular frequency of the wave, is the travel-time of the wave at a certain point
in elapsed time, p is the slowness vector and its magnitude can be thought of as an inverse to
the phase velocity with a certain direction, x is the position vector denoting the wave travelling
location in space. This representation of the wave can be further extended to this special case
namely changing the travel-time and the phase slowness to be complex-valued. Equation (2-1)
becomes:
(2-2)
=Ae ( p x ) i ( p x )
(I ) (I ) (R) (R)
e
where any superscript R denotes the real component of the complex parameter, and any
superscript I denotes the imaginary component of the complex parameter. The final form of (2-
2) is rearranging the variables in the equation aiming at separating the real components form
the imaginary components. This is with the intent to distinguish between the portion describing
8
the propagation of the wave (the real part) and the portion describing the attenuation of the
wave (the imaginary part). Carefully examining equation (2-2), it is possible to deduce these
physical meanings from the mathematical forms. First, by setting up two obvious conditions
( I ) p ( I ) x C1 0
(2-3)
( R ) p ( R ) x C2 0
The imaginary portion in the exponential function should and must represent amplitude decay.
And since there is a negative sign in the exponent term, the article within must yield a positive
and be a value above zero, otherwise the attenuation of energy and the amplitude decay will
not happen. It is worth to note that the same conditions are also prevalent with the real term of
the wave equation. Although it is true that the real portion could be represented in a similar
satisfy this condition and be a value above zero, but looking carefully in the wave equation’s
real part in the exponential function there is not a negative sign; meaning that inversely, the
article within must satisfy a positive value too as to reflect the propagation of the energy of the
wave.
The travel-time and the slowness are complex-valued while the wave amplitude,
frequency and position are real-valued. Subsequently, taking the gradient of equation (2-3)
yields
( I )
p( I ) ( I ) ,
x
(2-4)
( R )
p( R ) ( R ) .
x
As aforementioned, the slowness p is thought of as the inverse of the phase velocity vector;
time over distance, and at the same time it can be interpreted as the gradient of the travel-time,
9
as seen above in (2-4). This complex slowness vector is in general considered inhomogeneous
in nature. Vector inhomogeneity in this context is defined as any complex vector where the
directions of its real and imaginary components are not the same (Vavryčuk, 2008a).
Accordingly, and with having these two components, the slowness could be represented in
another form,
whereas established previously, p is the complex slowness vector with the two parts indicated
by the superscripts, p is the magnitude of p, n ( R ) and n ( I ) are the directional unit vectors shown
in Figure 6. A homogeneous complex vector has only one directional vector to represent it, as
the real and imaginary parts’ directions coincide, but generally this is not the case of the
Figure 6: A very concise illustration of a source generating a wave-front tracked by a ray-path ȓ, with two distinct slowness
directions associated with the complex slowness vectors p, and τ is the travel-time; the real parts indicated in red and imaginary
parts in blue
Finally, iterating on the definition of the slowness as an inverse of velocity, it is worth to note
that that velocity is the phase velocity symbolized by c. Throughout, there is a distinction
between phase quantities and ray quantities. Many of the wave quantities have those two sides
or versions; where the first describes the propagation of a plane wave while the latter describes
the propagation of wave energy following a ray-path and is more often associated with curved
10
wave-fronts related to point sources (Vavryčuk, 2016). Concerning the slowness vector, its
form with respect to the phase velocity is simply p n / c , and decomposing that into its
n( R ) n( I )
p p ( R ) ip ( I ) i (2-6)
cR cI
Therefore, to determine the slowness vector, one must find the unit vectors n ( R ) and n ( I ) , and
1 1
p( R) , p( I ) (2-7)
cR cI
Kinematic ray tracing mainly involves travel-times and ray-path calculations (Zhang,
2017). Essentially, to trace the rays as they are transmitted through the earth is the main
objective of this field of seismic modelling. Hence, the quantity representing the travel-time of
a wave is such a critical element as it is involved in multiple aspects of wave modelling and is
a characteristic by which a wave could be defined. Simply put, a travelling wave can be
comprehended as energy propagating through either of the domains; the spatial domain where
the position of the wave and its associated properties at a certain point is characterized by its
location in the 3D space coordinates, or the temporal domain where the position of the wave
and its associated properties at a certain point is characterized by the time elapsed since the
wave is triggered. The link between those two domains is the speed or velocity at which the
wave travels across a certain distance in the medium in a period of time. Great care should be
taken as in this case of a VEAM, the problem differs on two levels; there is an additional
imaginary part which must be considered, linked to the attenuating aspects of the media, as
11
well as the anisotropic aspects which results in the splitting and the heterogeneity of parameters
at different positions in the media (real and imaginary components of the slowness vector and
travel-times, etc., and different velocities of phase and group quantities, etc.).
Referring back to the definition of the phase slowness in terms of the travel-time, i.e.
dx p (n rˆ ) ds (2-8)
where the gradient of the travel-time is equal to the slowness vector p pn , and dx is an
the ray-path. Furthermore, the real and imaginary travel-times could be obtained by
( R ) ( R ) dx p ( R ) (n( R ) rˆ )ds
(2-9)
( I ) ( I ) dx p ( I ) (n( I ) rˆ )ds
p ( I ) (n ( I ) rˆ ) ( I )
= . (2-10)
p ( R ) (n ( R ) rˆ ) ( R )
Having proven equation (2-10), which translates to show that a minute change in the travel-
time of a wave or a very fine step in time is dependent on the ray-path direction of where the
wave is going and its phase slowness vector, or conversely the inverse of the velocity by which
the wave is propagating. This suggests that in order to obtain a function for the travel-time of
a wave, an integration along the ray is needed. And it could simply read
(R)
Ray
p ( R ) (n( R ) rˆ )ds,
(2-11)
p ( I ) (n ( I ) rˆ )ds.
(I )
Ray
Similarly, this integral equation could be manipulated and presented in the form of the phase
velocity. Recalling equation (2-6), which prompts that the phase velocity is the inverse of the
slowness with along its direction. The combination of equations (2-6) and (2-11) results in,
12
(n rˆ )
Ray
c
ds . (2-12)
When the travel-time is complex ( ( R ) i ( I ) ), and the slowness direction vector n is taken
c* c ic
( R ) i ( I )
Ray
(n rˆ )
cc *
ds (n rˆ ) R * I ds
Ray
cc
(2-13)
c c
(n rˆ ) R 2 ds i (n rˆ ) I2 ds
Ray
|c| Ray
|c|
where the asterisks represent complex conjugate quantities, and the modulus signifies the
magnitude of a vector. Comparing equation (2-11) and equation (2-13) to split the travel-time
cR
(R) (n rˆ ) | c |
Ray
2
ds,
(2-14)
c
(I )
(n rˆ ) I2 ds.
Ra y
|c|
And as such, obtaining the complex phase velocity grants the arrival at the function of travel-
be a property of the medium. A strongly attenuating medium has its Q factor lower than 5, and
weakly attenuating media’s Q factor is higher than 30 (Vavryčuk, 2008b). This quality factor
is unitless as it is a ratio, more on that below. Affinity to attenuating energy through it, the
quality factor of the medium is inversely proportional to that: the higher the attenuation
inflicted on waves travelling through the medium, the lower the quality factor of said medium
is. Attenuation is only an issue if the medium is inelastic, elastic media do not exhibit
13
attenuation. Waves’ energy gradually decay holding other factors constant as they travel across
viscoelastic media; this prompts a relationship between the travel-time of the wave, and the
media.
Like the phase slowness, there are other wave quantities. Additionally, many of these
quantities could either be defined as phase quantities or ray quantities as mentioned previously.
Being one of the wave quantities, attenuation could be obtained through calculations using
phase quantities where it then becomes phase attenuation, while on the other hand, if it is
obtained through calculations using ray quantities, it becomes ray attenuation. However,
measuring phase quantities and obtaining them is theoretically more desirable as then doing
the inversion with them is much easier than with the ray quantities (Vavryčuk, 2016). Phase
velocity and phase attenuation both depend on the complex phase velocity c , and as just
illustrated, the travel-time of the wave could be formulated in terms of this complex phase
velocity through substitution of the phase slowness vector (equation 2-12 and assuming
c* ( R ) c ic ( R )
( R ) i ( I ) Ray cc* (n dx) Ray Rcc* I (n dx)
cR c
Ray
c cI
2
R
2
( n ( R ) dx ) i 2 I 2 ( n ( R ) d x )
c cI
Ray R
(2-15)
(n ( R ) dx)
phase
i A phase (n ( R ) dx)
Ray
V Ray
where i is the imaginary unit, and c* is the complex conjugate of c . The real portion of the
travel-time is the inverse of one of the phase quantities mentioned earlier, the phase velocity:
14
cR2 cI2
V phase . (2-16)
cR
All the while the imaginary portion is equal to another phase quantity and it is the subject of
cI
A phase . (2-17)
c cI2
2
R
To honor the theory, one must highlight that an approximation/assumption have been made
here. The unit vector n used in equation (2-12) is in reality complex-valued and is generally
unknown, therefore it is common to make the supposition where the vector is real
n n ( R ) n ( I ) for equations (2-13) and (2-14) for constraining the problem. Following behind
this idea, it is frequently considered better to resort to the opposite side of the coin and choose
to utilize the laterally adjacent workflow of ray quantities over phase quantities in determining
the velocities and attenuations. Lastly, the third interesting phase quantity which is oftentimes
incorporated in such studies is the phase quality factor Q phase . It is a scalar which directly
represents a medium’s attenuating magnitude: the higher the Q factor of a medium, the less
attenuating it is. It mathematically reads (Carcione, 2000; Chichinina et al., 2006; Vavryčuk,
2007),
(c 2 ) ( R )
Q phase . (2-18)
(c 2 ) ( I )
Moreover, Vavryčuk (2016) argues that ray quantities are preferable for another reason
involving practical laboratory experiments complications. Plane waves, which are associated
with phase quantities, are generally difficult to generate and control in lab settings, the
alternative is generating point source waves which naturally have a curved wave-front in 3D
15
2.4 The Christoffel Equation
All VEAM may be described by a density-normalized (in order to eliminate the effects
2008b),
aijkl aijkl
( R)
iaijkl
(I )
(2-19)
where the superscripts R and I respectively represent the real component of the elastic moduli
and the imaginary component of the elastic moduli. The combination of the two effectively
describe the viscoelastic medium since each part denotes properties pertaining to a certain
behavior; the real part is the elastic behavior and the imaginary part is the viscous behavior
mirroring the attenuation of energy. Taking a step back from this, and to find where this elastic
moduli came from, it is inevitable to examine the plane wave (equation 2-1) and recognize it
presents a trial solution for plane waves in a homogeneous VEAM (Zhang, 2017) which satisfy
the differential equation based on Hook’s Law and the motion equation (Aki and Richards,
2u j i (cijkl k ul ) f j
(2-20)
or 2u j i (aijkl k ul ) f j
where is density of the medium, is the frequency of the travelling wave, u j is the jth
component of the displacement vector u , i is the partial derivative of u with respect to the ith
coordinate, cijkl is the four-ordered elastic moduli tensor of the medium, and finally, f j is the
jth component of the body force. It is worth to note here that most commonly, the density-
normalized moduli aijkl is the one used in formulation instead of cijkl in order to eliminate the
density out of the equation and disregard its individual effects, and have it incorporated from
within the elastic moduli. Moreover, this moduli is not dealt with the way it is in its higher
order tensor form, but it is rather simplified in order to be able to manipulate it and its
16
calculations more conveniently. Hence, this four-ordered tensor aijkl is reduced into a two-
(Vavryčuk, 2008b):
p i ij (9 i j )(1 ij ),
(2-21)
q k kl (9 k l )(1 kl ).
Kronecker delta jk is the operator used here for carrying out the transformation. Applying
equation (2-21) to the four order moduli aijkl , we obtain the two order moduli a pq , which
replaces its counterpart in equation (2-20). For example, to describe the parameter a2312 , it is
dictates. Additionally, the tensor also exhibits symmetry when considered in its matrix form;
the subscripts can be addressed interchangeably as the two resulting moduli would have the
same value; a12 equals a21 ( a1122 a2211 ). Originally, and since there are four subscripts in the
elastic tensor (i, j, k, l = 1, 2, 3), and each of these subscripts can take up to three values (3D
space), the total number of independent moduli resulting out of that will be
34 3 3 3 3 81 . After the Voigt recipe however, that number is reduced as there is now
two variables only (p, q = 1, 2, 3, 4, 5, 6), and both of them can take up to six values this time
62 6 6 36 . On the other hand, the assumption of the matrix symmetry equates all the
moduli values diagonally, and this leads to having the amount of independent moduli reach 21
independent parameters to completely describe any arbitrary VEAM. Equation (2-22) shows
the matrix form of the density-normalized elastic moduli reduced in two dimensions by the
Voigt recipe exhibiting symmetry, hence showing only the 21 independent variables needed to
17
a11 a12 a13 a14 a15 a16
a22 a23 a24 a25 a26
a33 a34 a35 a36
apq . (2-22)
a44 a45 a46
a55 a56
a66
This is considered to be as the most general case for defining a VEAM, or any arbitrary media.
However, in specifically oriented media, the number of describing variables decreases as the
medium becomes simpler and simpler and more homogeneous in its properties. Ten cases have
been showed by Crampin (1984) as he applied the symmetric properties of rocks and deduced
the elastic moduli of different anisotropy cases. For example, to define an isotropic medium, it
is only necessary to define a11 and a44 as they are the sole independent variables out of the 21
(equation 2-23). Similarly, to define a transversely isotropic medium (VTI, Figure 7), the
moduli a11 , a44 , a13 , a33 , a66 are the ones needed (equation 2-24). In the case of a tilted
transversely isotropic media (TTI, Figure 8), the number of parameters increases to seven as
the additional two represent the tilt angles (inclination and azimuth ) of the symmetry axis
as given by equation (2-25). And as the medium becomes more and more complex, more
moduli are needed to fully describe media with this orientation (equation 2-26).
18
a11 a11 2a44 a11 2a44 0 0 0
a11 a11 2a44 0 0 0
a44 0 0 0
ISO
a pq . (2-23)
a44 0 0
a44 0
a44
Figure 7: Multiple models of VTI media which are derived from geological features
Figure 8: Multiple models of TTI media which are derived from geological features and are a result of tiltation of VTI media
19
TTI
aijkl ai ' j ' k 'l 'ei 'i e j ' j ek ' k el 'l (2-25a)
where ai ' j ' k 'l ' are given by equation (2-24) and
e ( cos 0 , 0, sin 0 )
x' (2-25b)
e ( 0 , 1, 0 )
y'
e ( sin 0 , 0, cos 0 )
z'
An additional property that is often also associated with viscoelasticity and the study of
viscoelastic media is the matrix of quality factor qijkl . This quality factor is the quotient of the
two previously mentioned parts of the elastic moduli and it represents the attenuation
quantitatively. The higher the quality factor value, the less absorption of energy the medium
qijkl aijkl
( R) (I )
/ aijkl (2-27)
obtain some of the aforementioned viscoelastic parameters like the phase slowness and the
phase velocity. The famous Christoffel Equations is the integral relationship which is
cornerstone to almost all discussions about seismic waves in anisotropic media (Vavryčuk,
2007; 2008a; 2008b; Vavryčuk et al., 2016; Zhang, 2017). The Christoffel equation could be
obtained by the motion equation (2-20) in a VEAM, which leads to the following form,
20
2u j aijkl ,i uk ,l aijkl uk ,li (2-28)
where i u u,i and so on and so forth with respect to any partial derivative of any variable.
The trial solution is u j Aj (x)ei ( x ) (see equation 2-1), and then we calculate the other
quantities:
2 Aj (x)ei ( x ) cijkl ,i [ Ak ,l (x) i ,l Ak ]ei ( x ) cijkl {[ Ak ,l (x) i ,l Ak ],i [ Ak ,l ( x) i ,l Ak ]i ,i }ei ( x )
2 Aj (x) cijkl ,i [ Ak ,l (x) i ,l Ak ] cijkl {[ Ak ,li (x) i ( ,li Ak ,l Ak ,i )] [ Ak ,l (x) i ,l Ak ]i ,i }
2 Aj (x) cijkl ,i Ak ,l (x) cijkl Ak ,li (x) i[cijkl ,i ,l Ak cijkl ( ,li Ak ,l Ak ,i ) cijkl Ak ,l (x) ,i ] 2cijkl ,l ,i Ak
(2-30)
thus, there are two parts to the last equation, a real one and an imaginary one. The latter is
concerned with the amplitude and is not the focus. The real part is the main concern and it
gives,
and with applying the density normalization to the elastic moduli, as well as making the
Aj aijkl ,i ,l Ak
(2-32)
or
jk Ak aijkl ,i ,l Ak
(2-33)
and
21
(aijkl ,i ,l jk ) Ak 0
, (2-34)
where the matrix aijkl ,i ,l is often written as aijkl pi pl recalling equation (2-4) relating the travel-
time to the slowness vector. This matrix defines the so called Christoffel tensor:
which is a 3 3 symmetric matrix dependent on the density-normalized elastic moduli and the
We express the amplitude vector by A Agˆ , where A is the amplitude of the wave, and
gˆ ( gˆ1 , gˆ 2 , gˆ 3 ) is the polarization vector of the wave. Plugging this form into equation (2-34)
( jk c 2 jk ) gˆ k 0 , (2-37)
where jk is the Christoffel tensor given by equation (2-36), c is the phase velocity, and jk
is the Kronecker delta. Subsequently, as gˆ k is a unit vector which represents the polarization
direction of the wave, and cannot be equal to zero, hence for the equation to be true, the
following is implicit,
det( jk c 2 jk ) 0 , (2-38)
where in this determinant, the squared phase velocity is sometimes denoted as G because it
represents the three eigenvalues of the Christoffel equation, which in general has three
G n aijkl ni nl g j g k c 2 (2-39)
the eigenvalues’ matrix in terms of the direction vector. It can also be reformed in terms of the
G p aijkl pi pl g j g k 1 . (2-40)
22
The three eigenvalues give three phase velocities of wave modes, and three eigenvectors ĝ
describing the three different polarization directions of the three wave modes. When addressing
vector ĝ , it is defined as the vector denoting the directions of the eigenvector. And in other
words, the polarization direction of the particle motion, which is characteristic of each wave
All of these values are dependent either directly or indirectly on the complex elastic
moduli, which in turn makes them complex in nature as well. And this is what is distinct about
VEAM; a layer of complexity is there taking the form of anisotropy, but another layer is also
added from the viscoelastic aspect of the medium reflecting the attenuating features.
Figure 9: Simple illustration of the polarization direction in 3D of the different wave modes qP, qSV, qSH. The q prefix added
to any mode denotes quasi- which is a more realistic/representative visualization of particle motion with respect to the overall
direction of the wave travel; where the angle between wave propagation and particle vibration is not exactly 90 0
The group velocity is defined by the travelling of the wave amplitude. As opposed to
all the previous discussion about the phase velocity, the physical meaning of the group velocity
is the wave speed of the seismic energy along a ray-path. However, in the isotropic case, the
distinction between the phase velocity and the group velocity becomes meaningless as they
23
both are equivalent, but that is not the case in other types of media, namely, elastic anisotropic
media (Zhou and Greenhalgh, 2004) (Figure 10). In addition, and as emphasized previously, in
VEAM, all of the quantities become complex-valued; therefore illustrating them would be
complicated as the different values have different vectors, and those vectors have different
directions, and on top of that, real and imaginary quantities also must be considered and their
Figure 10: A simplistic illustration of a ray-path going from a source S to a receiver R through time τ with slowness vector p
(implicitly incorporating phase velocity c) and group velocity U in an elastic anisotropic medium
Mathematically, and since the group velocity denotes the energy flux alongside the
U (3-1)
κ
where U is the group velocity, and κ is the wave number (spatial frequency) vector
κ (1 , 1 , 1 ) n . The wave number is also related to the frequency in another form; the
12 22 32 (3-2)
c
where c is the phase velocity. Having established the above relationships, we can further
24
( c) c
Ui c
i i i i
(3-3)
c c c
cni cni ( )
i i i
where c cm ( , ) the phase velocity is a function of two angles reflecting the inclination ( )
and the azimuth ( ) of the phase slowness direction. Moreover, the two derivatives of the
angles with respect to the wave number read sin 1 ( 12 22 / ) , tan 1 ( 2 1 ).
i i
Then, with this, the derivatives can be replaced by their expansions in the equation of the group
velocity,
where the superscripts or subscripts ‘m’ takes the integers 1, 2 ad 3 and stand for the wave
modes (qP, qSV, qSH). Thus, with obtaining the group velocity in the three directions in the 3D
cm 2 1 c
U ( m) cm2 ( ) 2 ( m )2 . (3-5)
sin
Like so, the group velocity could be obtained, and viewing it from equation (3-5) suggests that
it is dependent on the phase velocity and the inclination and azimuth angles. This confirms the
concept and supports the abovementioned idea about the association between the phase
velocity and the group velocity in isotropic media; in isotropic media where the two angles are
zero (thus eliminating the derivative terms in (3-5)), the velocities end up being equivalent.
Whereas, in anisotropic media, the two velocities are indeed proven to be distinct. As for
25
calculating these derivatives of the phase velocity with respect to the angles, analytic
calculations could be carried out by several formulations (Zhou and Greenhalgh, 2005; Zhang,
2017).
In another way, and recalling equation (2-39) describing the eigenvalues, the phase
velocity could be defined differently by also employing the Christoffel tensor equation (2-39),
In addition to that, the definition of the Christoffel tensor above could be put to use further by
i
recognizing a property of the wave number vector with its directional descriptors where ni
l
and nl (due to n ), so we have:
i l ( m ) ( m )
cm2 aijkl ( ) gˆ gˆ . (3-7)
2 j k
Afterwards, to manipulate the above equation aiming to come up with an alternative expression
of the group velocity, we proceed with deriving equation (3-7) with respect to the wave number
y and calculate,
cm 1 i l
[aijkl gˆ (jm ) gˆ k ( 2 ) aijkl pi pl ( gˆ (jm ) gˆ k( m ) )]
2cm
1 il 2 i l gˆ (jm ) gˆ k( m )
[ 2 ( l i i l ) il ( ˆ
g (m)
ˆ
g ( m)
)]
2cm 3
jk k j jk
(3-8)
26
gˆ (jm) gˆ k( m)
where il aijkl gˆ j gˆ k , and the identity jk gˆ k( m) gˆ (jm) jk 0 have been applied in
these calculations. Substituting equation (3-8) into equation (3-3) for the derivative in the group
cm aijkl nl ( m ) ( m )
U i( m ) cm ni gˆ j gˆ k
i cm (3-9)
= aijkl p (m)
l gˆ ( m)
j gˆ ( m)
k
Thus far, presented here are two methods by which the group velocity could be obtained, which
in turn demonstrated why the phase velocity and the group velocity generally have significant
difference when it comes to anisotropic media as opposed to isotropic media in which they are
both interchangeable. The first way of finding the group velocity utilized defining it in terms
of the phase velocity and additional terms associating with the wave normal inclination and
azimuth angles, while the second approach required finding the eigenvectors along with the
slowness (or exchangeably the phase velocity). Any parameter based on a complex parameter
(since the main subject is the viscoelastic realm) is rendered by default to be complex-valued
as well.
The ray velocity vector is a member of the ray quantities, the category of parameters
briefly cited earlier opposing phase quantities. And in contrast to the latter, the former as its
name suggests is directly linked to the ray-path of a wave. It can be explained as the propagation
speed of a wave along its ray-path. Therefore, it can be defined simply in a mathematical form
27
dx ds
v rˆ , (3-10)
d d
where x is the coordinate vector of the ray-path, and it is exactly the same if defined by a
segment ds in the direction of the ray-path r̂ . Based on the definition of the slowness as the
identity by applying the a dot product of the phase slowness vector and the ray velocity (Helbig,
1994; Vavryčuk, 2007; 2008a; 2008b; Vavryčuk et al., 2016; Zhang, 2017):
ds d ds
p v rˆ 1, (3-11)
d ds d
c
n r . (3-12)
v
where v is the -scalar- ray velocity coming from v rˆ . By this, a direct link is found between
the slowness vector and the ray velocity vector. The dot product could suggest the nature of
the relationship between the two vector variables, as it encloses a trigonometric function of the
angle between the two directions of the two vectors. If a dot product of two vectors is zero then
they are perpendicular to each other, if it is equal to 1 then they are parallel to each other (given
they are both normalized to length equal to 1), any other number indicates any other angle.
In an attempt to further develop the expression of the ray velocity, the step of
manipulating equation (3-6) by dividing both sides of the equation by cm2 could be taken, and
also by accompanying that with the definitions of the Christoffel tensor and the slowness
vector, we get:
and with modifying this equation to isolate pi( m) , and recalling the famous identity equation (3-
28
vi( m ) aijkl pl( m ) gˆ (jm ) gˆ k( m ) . (3-14)
This equation will look familiar as it is exactly identical to equation (3-9). This quickly prompts
to deduce that the group velocity is in fact equivalent to the ray velocity in anisotropic media
as it is in isotropic media. But note that both are not equal to the phase velocity in anisotropic
media as they are in isotropic media. One will also note that both the group and ray velocities
have similar physical meanings where both describe the propagation of the wave energy along
a ray-path. This might prove important for some specific applications or particular
methodologies for when for example measuring and calculating ray velocities is not reasonable,
hence one would resort to measuring the easier quantity and then that would be equivalent to
the actual quantity sought after. Again, it is worthy to reiterate on the fact that most of these
parameters are complex-valued in the subject case of viscoelasticity and attenuating media.
Therefore, all equations shown here are applicable to VEAM so long as the parameters are
Having defined the ray velocity (or the group velocity for that matter, it makes no
difference in isotropic as well as anisotropic media), which corresponds to the phase velocity
when mirroring ray and phase quantities, subsequent parameters can easily be derived and
calculated. Most of the relationships between these viscoelastic quantities have been shown
above where linking the phase with ray calculations is unavoidable and measuring or handling
either one or the other purely depends on the nature of the anticipated application. Therefore,
some view both domains as being quite alike where they only differ in physical meanings or
concept of study; “the only difference is that the ray quantities are defined along a ray, whereas
the phase quantities are defined along a wave normal” (Vavryčuk, 2007).
29
The complex ray velocity relates to the complex travel-time through the ray coordinate
vector as equation (3-11) indicates. But additionally, and to make things clearer, the equation
could be further developed to show the two components of the complex ray velocity; the real
dx
v vrˆ
d
ds
v rˆ v rˆ ( v ( R ) iv ( I ) ) (3-15)
d
vR rˆ v ( R ) , vI rˆ v ( I )
where bold characters indicate vectors while others indicate components or magnitudes of
vectors. The ray velocity is simple and straightforward to deduce, it is intuitive when the term
encompasses the basic definition of speed which is the distance over the time. Consequently,
in order to obtain the travel-time it need be isolated and have the equation rewritten in terms of
ds
Ray
v
(3-16)
If we were to continue to make analogy to the phase realm, equation (3-16) should be
equivalent to equation (2-12) in phase terms. This form of the travel-time is then inserted to
v* v iv
(R)
i (I )
* ds R * I ds
Ray
vv Ray
vv
(3-17)
v vI
R 2 ds i ds
Ray
|v| Ray
| v |2
Correspondingly, equation (3-17) correlates to equation (2-13) of the phase travel-times. The
formulation is identical, however note that the dot product between the slowness direction
30
vector and the ray direction vector is not explicitly required in the ray realm, which is the reason
why oftentimes calculating the travel-time by equation (3-17) rather than by equation (2-13) is
more preferable and reliable as it does not necessitate making any approximation which could
very well impose inaccuracies and errors on the results of the calculations.
The exact same correlation applies when realizing the ray attenuation and ray velocity
parameters as with their counterparts as phase quantities. In an almost identical fashion to the
travel-times in phase form, the real and imaginary components of equation (3-17) demonstrate
the two sought after quantities. We rewrite (3-17) to highlight that as such,
v* v iv
( R ) i ( I ) Ray vv* ds Ray Rvv* I ds
vR v
Ray
v vI
2
R
2
ds i 2 I 2 ds
v vI
Ray R
(3-18)
ds
V
Ray
Ray
i
Ray
ARay ds
The real portion is the inverse of one of the ray quantities mentioned earlier, the ray velocity:
vR2 vI2
V ray . (3-19)
vR
All the while, the imaginary portion is equal to another ray quantity and it is the subject of
discussion; the ray attenuation, which guides the amplitude decay of a wave along a ray-path:
vI
Aray . (3-20)
v vI2
2
R
Lastly, the third interesting ray quantity which is always incorporated in such studies is the ray
quality factor Q ray . It is a scalar which directly represents a medium’s attenuating capacity; the
31
higher the Q factor of a medium, the less attenuating it is. It mathematically reads (Carcione,
(v 2 ) ( R )
Q ray (3-21)
(v 2 ) ( I )
All of these quantities are associated with v , the complex ray velocity parameter. Similarly,
and as have been demonstrated previously, it is simple to obtain the same quantities but for the
counterparts; the real phase velocity V phase , the real phase attenuation A phase , and the real phase
quality factor Q phase shown in equations (2-16) ~ (2-18), simply by replacing the complex ray
An obvious deduction can be made clearer here. If it is assumed that the imaginary part
parameter, the abovementioned six quantities (expressed in equations (2-16) ~ (2-18) and (3-
19) ~ (3-21)) change drastically to accommodate the absence of any imaginary component,
leading to:
vI 0
Aray 2 0
v vI vR 0
2
R
2
cI 0
A phase 2 0 (3-22)
c cI c R 0
2
R
2
(v 2 ) ( R ) (v 2 ) ( R )
Q ray
(v 2 ) ( I ) 0
32
(c 2 ) ( R ) (c 2 ) ( R )
Q phase
(c 2 ) ( I ) 0
This perfectly represents elastic media; the ray velocity and phase velocity are exactly equal to
their real parts, the attenuation is brought to zero because there is zero energy loss in perfect
elastic media, which is also translated in an infinitely rising Q factor for both of the ray and
phase quantities.
equations (2-16) ~ (2-18) (phase quantities) for computing the travel-time and attenuation
because the former does not incorporate any conditions, but the latter implies the assumption
n n( R ) .
For the ultimate goal of conducting ray tracing in VEAM, the determination of the
phase slowness vector and hence the ray velocity vector is inevitable. Having stated that, the
dimensions of the problem differ as the conditions of the settings vary. It is an undeniable fact
that dealing with elastic media is much more convenient in terms of calculations/computations
and the limitations which have be to be taken into account. This is all due to the simplicity of
the medium’s properties when compared with the more tortuous viscous medium or even the
viscoelastic medium. The complexity sprouts from the absorbent properties of such media
which theoretically translate into complex-valued numbers with imaginary parts which must
be considered in the calculations. Conversely, in elastic media, there are only real numbers to
be involved in calculations; the elastic moduli aijkl and the slowness vector p only have to be
dealt with as their real numbers values. However, in the subject case (the viscoelastic case),
33
there is a distinction between p ( R ) and p ( I ) , and one way to calculate this imaginary part is by
iteratively varying it through a range of values (for a constant p ( R ) ) with the purpose of
minimizing the ray direction vector’s imaginary part (Vavryčuk, 2008b). Commonly, the
imaginary part of the slowness pales in value with respect to the real part, and this makes for
solving a system of polynomial equations. The latter though exhibits some hindering
limitations seeing that it was initially developed for dealing with elastic media, however this is
not an alarming issue as it is possible to adaptively convert the solution into the viscoelastic
domain by incorporating the imaginary parts of the complex quantities. The real issue with the
polynomial equations approach, even after adopting complex numbers, is that it might still
produce highly invalid results when the assumption that the complex ray velocity vector for
the given slowness vector being homogenous is violated (Vavryčuk, 2006). This leaves the
iterative procedure. On top of that, this procedure also has limitations where it does not provide
accurate results; “the iterative procedure is fast and works well, provided that a wave-front is
free of triplications” (Vavryčuk, 2007). The wave-front triplication only occurs with the qSV
homogeneous complex ray velocity v , whose real and imaginary parts have a real ray direction
N . The initialization is by calculating the Christoffel tensor jk (n) aijkl ni nl with the complex-
34
sin( R i I ) cos( R i I )
n sin( R i I )sin( R i I ) . (4-1)
cos( R i I )
Subsequently, the Christoffel tensor has three eigenvalues and three eigenvectors each to be
The eigenvalues ( cm ) and equation (2-6) may then be used to calculate the complex slowness
equation (3-14). The ray velocity vector calculated by equation (3-14) is generally
inverse approach of the above stated procedure. There are two very similarly-working branches
which can be followed to inversely iterate to find the directional vector n , and the misfit
function explained by Vavryčuk (2008b). One, is when a ray direction vector N is fixed and
then the actual search is for two parameters: the two angles ( ( I ) , ( I ) ) describing the
imaginary part of the slowness direction n , with given angles ( ( R ) , ( R ) ) the other two
describing the real part’s. Consequently, the slowness direction n with its complex
components ( n ( R ) , n ( I ) ) is obtained. Henceforth, the rest of the parameters are orderly obtained
(the Christoffel tensor, the stationary slowness and the ray velocity vectors, etc.), and then
through the several iterations, the to-be optimized misfit function could be constructed by
setting a condition relating the initially fixed ray vector N and the derived ray vector and the
35
The other branch is to define the slowness direction as,
where n ( R , ) is the unit vector perpendicular to n ( R ) , and ẑ1 , ẑ2 are complex values defined
by:
1 cos 2 1 cos 2
,
2 2
sin 2 (4-5)
.
1 cos 2 1 cos 2
2
2 2
Accordingly, given any angle , one is capable of obtaining ẑ1 and ẑ2 , hence knowing n ( R )
and n ( R , ) . Therefore, one only needs to search a single angle for the homogeneous ray
At this point, the two branches analogously converge as the coming step is to calculate
the remaining parameters (the Christoffel tensor, the stationary slowness and the velocity
vectors, etc.). This time around however for the second approach, the aim of the to-be
optimized, through the several iterations, misfit function is to minimize the imaginary part of
the ray direction vector N in order to end up with the real-valued component alone –after
searching and varying the angle–. And when that condition is satisfied, the iteration halts and
one last run through the calculations is done with the optimum values: starting from the
36
searched angles up to calculating the real ray velocity V ray , the real ray attenuation Aray , and
the real ray quality factor Q ray using equations (3-19) ~ (3-21).
These methods are thought to be more accurate and more dependable when compared
with one of the predecessors, the aforementioned method utilizing the polynomial equations,
which is considered to have a high possibility of producing some unreliable results (Vavryčuk,
2006; Vavryčuk, 2008b). So, this was the alternative concept; to deal with wave energy in its
ray-path carried by higher frequency ranges of a stationary slowness direction knowing the ray
vector direction with an iterative inversion. Figures 11 and 12 show and summarize the general
workflows of these two branches, and where the iteration element actually is.
Utilizing this method for the ultimate objective of ray tracing in viscoelastic anisotropic
media is reasonable. The iterative method incorporates less assumptions; there is only one
assumption, which dictates that the complex ray velocity vector is homogeneous i.e. there is
only the real ray direction to be taken into account as both the real and imaginary directions
are parallel ( n n ( R ) ). This is highly advantageous for this method. In fact, this is the particular
reason why this certain method is used to generate the solution by which we benchmark
modelling results in order to compare any other methods’ against. Additionally, the results
proved to show high accuracy for the qP-wave even in cases of strong attenuation, as well as
at the interfaces of two media (Vavryčuk, 2008a real ray; Vavryčuk, 2010). Having stated that,
the iterative process is considered to be a time-consuming process due to the fine step required
for searching for the solution of the complex slowness direction for VTI media for example at
the cusps of qSV, and fails for the qSV- and qSH-waves in a highly complex VEAM, for instance
the ORT (orthorhombically anisotropic) medium. For this reason among others, other methods
37
Figure 11: A flowchart summarizing the first branch of the iterative solver algorithm steps, focusing on the iterative element,
to obtain the ray velocity, attenuation and quality factor
38
Figure 12: A flowchart summarizing the second branch of the iterative solver algorithm steps, focusing on the iterative element,
to obtain the ray velocity, attenuation and quality factor
39
4.2 Perturbation method
This method is, as the name suggests, based on the concept of an anomalous
perturbation happening in the vicinity of normal conditions (Figure 13). The assumption here
is basing on a hugely complex ray vector based on considering the normal/default conditions
being the elasticity of the media and the disturbance (hence, perturbation) from that is a
component that is behaving in a viscoelastic manner. There is a key assumption here that is
made and is essential in order for this concept to yield valid results; the assumption is that the
complex ray velocity vector is homogenous, in other words, for this complex vector quantity,
the real part’s vector’s direction is identical to the imaginary part’s vector’s direction. This
might not always be the case but it makes for a decent approximation.
Figure 13: Simplistic illustration of the concept of perturbation. Properties of the background are summed with properties of
a distinct disturbance deviating from the background, resulting in an aggregate of the two properties
which the thorough work is a work-in-progress lead by Dr. Jianlu Wu. Concisely, the
Hamiltonian function is employed alongside the phase slowness vector –in conjunction with a
few approximations and conditions– to derive the eigenvalues representing the phase velocity
vector, and eventually the final form of the ray velocity vector. Moving on with that, the
magnitudes of the ray velocity vector, ray attenuation, and ray quality factor can be calculated.
The Hamiltonian function is a function of the position vector x , the slowness vector p , and
the elastic moduli aijkl , and it is equal to zero in a viscoelastic medium (Červený, 2001):
40
H x, p, aijkl 0 (4-6)
The point here then, as mentioned above, is to split the response of the viscoelastic medium
into the background elastic behavior and the perturbing viscoelastic behavior as the figure
above demonstrated; the total elastic moduli becomes: aijkl aijkl iaijkl , and the total
(0)
slowness becomes: p p(0) ip . Note that both the background and the perturbation are dealt
with as anisotropic media. Hereafter, the Hamiltonian function and its components in turn are
H x, p, aijkl H ( 0) x, p ( 0) , aijkl
(0 )
iH x, p, aijkl (4-7)
(0)
Accordingly, the first part denotes the elastic parameters ( aijkl , p (0) ) and the latter part refers
to the complex parameters ( aijkl , p ). The elastic slowness is clearly a real number which
can be calculated simply because of all the real quantities of the H function (Zhou and
Greenhalgh, 2004). Another note to add relating to the slowness, is that the resulting
The same paper, and others (Červený, 2001) also proves how the Hamilton equation is directly
linked to the ray velocity vector which is sought after. The link is found by partially deriving
the Hamiltonian function with respect to the slowness from the Eikonal equation. Applying
41
that will yield a ray velocity vector of two parts, the real part arising from the background
response ( vi ) and an imaginary part arising from the perturbation response ( vi ),
(0)
Both parts of the ray velocity are considered to have the same direction as mentioned earlier to
preserve the assumption of the homogeneity of the complex vector. By this, it is possible to
calculate all the parameters and finally obtain the ray velocity vector ( vi aijkl pl g j gk ).
Subsequently, and as long as that the two components, the real and the imaginary, of the ray
velocity are known, the vector magnitude V ray can be calculated, as well as the attenuation
magnitude Aray , along with the associated quality factor Q ray as demonstrated previously
(equations (3-19) ~ (3-21)) but not before calculating from the here mentioned two velocity
N v/v
vR N R iv I N I
v (4-11)
v vR ivI vi vi v v 2ivR vI ( N R N I )
2
R
2
I
The main advantage of this method is that it enables the direct calculation of the
imaginary parts of the slowness vector, ray velocity, and quality factor of all of the three wave
modes, the qP-, qSV- and qSH- waves. Up to here, is the broad demonstration of the method
and its procedure for any general case. This could be further developed and extended to be
applied for specific media, like VTI and other types of transversely isotropic/anisotropic media
such as HTI, TTI, and ORT, all of which is in the viscoelastic domain. Here, VTI and ORT are
discussed.
42
For the modes qP and qSV in the VTI medium, they are dealt with separately from qSH
because the former two depend on the elastic moduli a11 , a33 , and a44 , while the latter only
depends on a44 and a66 , hence the equations differ. Accordingly, the Hamiltonian function for
H x, p a44 a11 px2 a33 a44 pz2 a11a44 px4 a33a44 pz4 Dpz2 px2 1 0
(4-12)
where D is the coefficient of the Hamiltonian function which is entirely dependent on the
elastic moduli alone. Referring back to the totals of the elastic moduli and slowness vector, the
two Hamiltonian function components, the background and the perturbation of equation (4-9)
become
where,
p a a a a p D p p
(0) 4
z
(0)
44 11
(0)
11 44
(0) 4
x
(0) 2
x
(0) 2
z ,
W1 2 a a p 4a a p 2 D p p ,
(0)
11
(0)
44
(0)
x
(0) (0)
11 44
(0) 3
x
(0) (0)
x
(0) 2
z
W2 2 a a p 4a a p 2 D p p .
(0)
33
(0)
44
(0)
z
(0) ( 0)
33 44
(0) 3
z
(0) (0) 2
x
(0)
z
(4-15)
In the same way, referring to the ray velocity in terms of the Hamiltonian function derivative,
43
vx H x , p px vx(0) ivx vx(0) i A1 A2 px A3pz
(4-16)
vz H x , p pz vz(0) ivz vz(0) i B1 B2 px B3pz
where,
vx(0) 2 a11(0) a44 px(0) 4a11(0) a44(0) px(0) 2D(0) px(0) pz(0) ,
(0) 3 2
A1 2 px(0) a11 a44 4 px(0) a44 a11 a11(0) a44 2 p x(0) p z(0) D;
(0)3 2
(4-17)
and
(4-18)
Employing the Hamiltonian function perturbation and the ray velocity definitions above, along
with the identity vx vz vx vz , which arises from the condition of the complex ray
(0) (0)
velocity vector homogeneity (the real and the imaginary parts of the direction vectors are
parallel), the perturbation slowness vector p (px , pz ) , and the ray velocity vector
v (vx , vz ) are finally obtained. This covers the details associated with the qP and qSV wave
modes. For qSH, the case is similar, however with different parameters. The Hamiltonian
function perturbation and the complex ray velocity components in terms of the Hamiltonian
function are similarly governed by the equations (4-13) and (4-14), where this time, the
44
W0 px(0) a66 pz(0) a44 ,
2 2
W1 2a66
(0) (0)
px , W2 2a44
(0) (0)
pz ,
vx(0) 2a66
(0) (0)
px , A1 2 px(0) a66 , (4-19)
A2 2a66
(0)
, vz(0) 2a44
(0) (0)
pz ,
B1 2 pz(0) a44 , B2 2a44
(0)
.
And once again, similarly like it was for the other two wave modes’ case, and using the same
identity from the homogeneity condition of the ray velocity, it is a matter of plugging in the
equations to obtain the perturbation slowness vector, and the ray velocity vector, associated
As for the ORT media, the subject is a bit more involved and complicated as the elastic
moduli needed for describing this category of anisotropic media are nine independent variables
( a11 , a12 , a13 , a22 , a23 , a33 , a44 , a55 , a66 ) out of the total of twenty one independent variables
(equation 2-26) which could describe any media in general. The Hamiltonian function in this
case reads,
(4-20)
where the vector of the coefficients of the Hamiltonian function Di is constituted of the nine
elastic moduli and the can be found in details in Appendix A. Henceforward, the perturbation
45
where the vector of the coefficients of the perturbation of the Hamiltonian Wi is dependent on
the slowness and the coefficients of the Hamiltonian Di , and can be found in details in
Appendix B. Finally, the complex ray velocity and in its components form show as,
y iv y v y i F1 F2 p x F3 p y F4 p z ,
v y H p y v (0) (0)
(4-22)
where these ray velocity coefficients are dependent on the slowness and the coefficients Di ,
and can be found in details in Appendix C. Figure 14 below summarizes the procedure above
broadly. In short, it starts from the elastic moduli inserted in to the Hamiltonian function. Then,
the derivative of that function with respect to the slowness gives the ray velocity as the literature
have proven. Finally, the total slowness (real and imaginary parts) is found along with the ray
velocity, which is employed to eventually obtain the real ray velocity V ray , the real ray
attenuation Aray , and the real ray quality factor Q ray as equations (3-19) ~ (3-21) demonstrate.
To ensure the validity of this method’s results, two essential assumptions must be
considered. The first assumption is that the complex ray velocity vector is homogenous. The second
is that the Hamiltonian function perturbation is linear, which limits the applicability of the method
to weak anisotropy (as the aijkl is relatively very small), and that will be investigated and verified.
The main advantage of this method is that, on top of having significantly reduced computational
expenses due to the lack of an iterative search for the slowness angle, it promotes the direct
calculation of the imaginary parts of the slowness vector, ray velocity, and quality factor of all of
the three wave modes, the qP-, qSV- and qSH- waves through the Hamiltonian perturbation function.
In addition, this method is expected to more accurately solve for the real ray velocity and also better
tackle less complicated media and wave-fronts and shape the optimum solution for such settings.
46
Figure 14: A flowchart summarizing the perturbation method steps to obtain the ray velocity, attenuation and quality factor
47
4.3 Real Slowness Direction approximation
This approach could be thought of as similar to its previous counterpart when examined
from a certain point of view. A comparable assumption is put in place in order to simplify the
problem, so long as the results are valid and the approximation does not lead to the violations
of base principles. As stated previously, and generally but specially in viscoelastic anisotropic
media, all the complex vector parameters could be thought of as being inhomogeneous as both
of the two components –the real and the imaginary– could very well have different directions.
However, and for the purposes of this method, the slowness vector is chosen to abide by the
homogeneous complex vector condition. This facilitates neglecting dealing with the
inhomogeneous state of the slowness vector, and on top of that, enables the direct
approximation of the complex ray velocity which is similarly going to be bound under the
condition of the homogeneous complex ray velocity vector. The team dubbed this approach the
Real Slowness Direction (RSD) approximation, in reference to the fact that the direction of the
slowness vector is only one instead of two, and it is by default denoted by the real direction,
the imaginary direction is substituted by the real one (but it is established that the real direction
component, a direction for each component, and a complex phase velocity. Decomposing it
from equations (2-5) ~ (2-6) and forcing homogeneity on the slowness vector ( n ( R ) n ( I ) )
yields:
n( R )
p p ( R )n ( R ) ip ( I )n ( I ) , (4-23)
c
1
p p ( R ) ip ( I ) , (4-24)
c
and
48
cR
p( R) n( R ) ,
c cI
2
R
2
(4-15)
c
p (I )
2 I 2 n( R) .
cR cI
vector are defined by the same direction (which is taken to be the real one), where generally
the two direction could very well be unequal. Thus, by means of substitution, a simple
cI ( R )
p( I ) p (4-16)
cR
The slowness vector could also be defined in another way; associated with the wave travel-
time ( p ). And the travel-time itself is in turn constitutional in the definition of ray
velocity by simple physics by means of a partial derivative involving the ray coordinates (recall
equations (3-10) and (3-15)). The ray velocity vector is a homogeneous complex vector too,
forcibly. The above leads to the deduction that the slowness vector and the velocity vector are
parallel following the famous identity of equation (3-11) ( p v 1 ) (Vavryčuk, 2008a). This is
the main distinction between this method and the perturbation method. This identity is not used
as a condition constraining the solution. Having established this identity formula, expanding in
terms of direction vectors results in equation (3-12) but only in the real domain as we forced
cR
(n ( R ) rˆ ) . (4-27)
c
vR [1 ( I ) 2 ]
cR
This suggests that the direction of the slowness n and the direction of velocity/ray-path r̂ are
not aligned, in proportion to the difference of the phase velocity cR and the ray velocity vR ,
which is normally the case in VEAM. Through the above identity as well, it is possible to
49
p v 1 (p ( R ) ip ( I ) ) (vR ivI )rˆ 1
cI ( R )
(p ( R ) i p ) (vR ivI )rˆ 1
cR
cI
(1 i )(vR ivI )(p ( R ) rˆ ) 1
cR
(4-28)
c c n( R)
(vR I vI ) i (vI I vR )( rˆ ) 1
cR cR cR
cI cI n ( R ) rˆ
( vR vR )( ) 1
cR cR cR
cI 2 c
[1 ( ) ](n ( R ) rˆ ) R ,
cR vR
which enables the discovery of yet another relationship used amidst expanding the terms:
cI
vI vR
cR (4-29)
relating the real and imaginary parts of both the phase and the ray velocities. The advantage of
this equation is that it imposes and satisfies the homogeneity condition when calculating the
real and imaginary ray velocities. Having known these quantities of the ray velocity
components, the velocity vector magnitude V ray , the attenuation magnitude Aray , along with
the associated quality factor Q ray can straight away be calculated as well, as priorly
demonstrated through equations (3-19) ~ (3-21). The procedure is summarized in the following
As the name of the method indicates, we regard only the real slowness direction and
enforce homogeneity on this vector. This limits the overall reach of the method however,
enables the direct approximation of the complex ray velocity, and also enforces its
particularly, and it ensures the parallel relationship between the two vectors. This method
optimizes computation time as well as it is suitable for targeting more complex media with
stronger anisotropy and higher attenuation, even for the problematic tripling of q-SV wave-
front having cusps with several spatial values for a single ray direction.
50
Figure 15: A flowchart summarizing the RSD approach steps to obtain the ray velocity, attenuation and quality factor
51
5. Numerical Results
The easiest natural VTI media that one would think of as fitting in a geological sense
complying with vertical symmetry is the shale lithology. Shale is naturally layered with
horizontal planes overlying each other (principle of original horizontality), so long as no further
deformation caused any inclination after sedimentation. There are many natural geological
phenomena which could result in encountering VTI media in the environment because there
are many sources of transverse isotropy. We assume that we have a viscoelastic perfectly VTI
Using the iterative method workflow (Figure 11), the above information is used to construct,
for the three wave modes (qP, qSV, qSH) at different ray directions, their real ray velocity, real
ray attenuation, and ray quality factor based on the given elastic moduli. The phase quantities
are also obtainable easily as long as the phase velocity is found. Since the medium is symmetric,
the results are displayed in half a 2D plane due to the reflection of the results on the other half
(Figure 16).
52
Figure 16: Ray and phase velocities (first row), ray and phase attenuations (second row), and ray and phase quality factors
for the three wave modes: qP (first column), qSV (second column), qSH (third column) for a shale model of a viscoelastic VTI
medium
The three wave modes slightly behave similarly. The qP-wave’s behavior in terms of each pair
of quantities in both realms is quite regular, as it is the case for the qSH-wave whose phase and
ray quantities are almost inseparable. Both of the velocities certainly do change with the
changing directions, however, they both converge the more we move away from the center and
get closer towards the horizontal and the vertical axes where they meet and become equal,
which is the similar case reported in elastic VTI media by Thomsen (1986). The same behavior
is also noticed in the case of the attenuation and the quality factor. As for the qSV-wave, it
53
always exhibits bizarre shapes as that is due to the aforementioned triplication of wave-front
the wave undergoes from its natural structure. In this instance and concluding from the
supporting plot, for single ray directions there is more than one value for these wave quantities
and that explain the cusps and twists of the shape that is a signature of the qSV-wave alone.
With this, there is proof that the iterative algorithm provides satisfactory results when it comes
to finding the complex ray velocity and the slowness vectors to deduce the three wave modes’
phase and ray quantities in a viscoelastic purely VTI medium such as shale.
In addition, eight other models are used to verify the efficiency of the perturbation
approach and the RSD method. The Thomsen models’ parameters’ can be found below (Table
Upon examining the values, models belonging to the same category (either A or B) have the
same elastic moduli but different quality factors, with Model A1 being the most complex with
the highest anisotropy and attenuation. Vavryčuk (2008b) derived the analytic solutions which
our numerical solutions will compare against. The deviation or the difference between the two
will be measured by an absolute value of the relative error between the numerical and the
54
analytical results ( V ray , Aray , Qray ). Figure 17 below shows the results of the 3 ray quantities
using the two methods against the exact solution with the error plotted as well, and that is for
the qP-wave in Model A1. These two methods achieved solutions that are in excellent
agreement with the exact solutions, even though this model is the most complex out of the 8.
The maximum errors of the ray velocity and the quality factor are 0.01% and 0.22%
respectively for the perturbation approach. As for the attenuation, the errors are comparable to
the quality factor (0.21%). The RSD slightly falters behind for the velocity but is on par for the
other two quantities in terms of maximum errors. In Figure 18, however, the results are even
brighter with much reduced maximum errors due to the weaker anisotropy of Model B1,
although the same level of attenuation. The error reduction is more than the double for the
perturbation method, while it is minimal for RSD. Thus far, the perturbation method yielded
slightly better results than the RSD approximation, and by examining Table 3, one could
observe that RSD only performs better than its peer in the four A Models (having stronger
anisotropy) only with respect to the quality factor. Overall, the perturbation method is superior
Figure 17: The computed ray velocity, attenuation, and quality factor of the qP-wave in Model A1. The absolute relative error
is also plotted with respect to each method in comparison with the exact solution form the literature
55
Figure 18: The computed ray velocity, attenuation, and quality factor of the qP-wave in Model B1. The absolute relative error
is also plotted with respect to each method
Table 3: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality factor of the two methods for the
qP-wave
As for the other wave modes, Figures 19 and 20 show the same as Figures 17 and 18 (Models
A1 and B1) but with respect to the qSV-wave. A remarkable observation needs to be made here
as the RSD method succeeds in better estimating the complicated tripling wave-front of the
qSV-wave, where other methods fall behind. Table 4 explicitly shows how the maximum errors
of RSD are significantly less, by orders of x10, than those of the perturbation method when it
comes to the qSV-wave. We continue to observe the insensitivity of RSD to the degree of
anisotropy.
56
Figure 19: The computed ray velocity, attenuation, and quality factor of the qSV-wave in Model A1. The absolute relative error
is also plotted with respect to each method
Figure 20: The computed ray velocity, attenuation, and quality factor of the qSV-wave in Model B1. The absolute relative error
is also plotted with respect to each method
57
Table 4: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality factor of the two methods for the
qSV-wave
The remaining wave mode is the horizontal shear wave and its results for Model A1 are given
in Figure 21, and for Model B1 given in Figure 22. The accuracy of both methods proved to be
generally higher and better for the qSH-waves when compared with the other two modes. All
the while, the RSD performed generally much better than the perturbation approach with
respect to the ray attenuation and ray quality factor, however, using the perturbation approach
is far more accurate for estimating the ray velocity. The maximum absolute errors of the two
methods for all models particularly for qSH are reported in Table 5.
Figure 21: The computed ray velocity, attenuation, and quality factor of the qSH-wave in Model A1. The absolute relative
error is also plotted with respect to each method
58
Figure 22: The computed ray velocity, attenuation, and quality factor of the qSH-wave in Model B1. The absolute relative
error is also plotted with respect to each method
Table 5: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality factor of the two methods for the
qSH-wave.
59
5.2 TTI model
A tiltation of rock units is very common and can be observed widely, obviously in
tectonically active areas, but even in passive margins and other inactive regions. For example,
clinoforms highly mimic tilted transverse isotropy conditions, assuming that each single layer
is homogeneous; that yields a TTI model which simplifies reality but is at least the best
approximation. Other than that, having a TTI medium could be as simple as applying an
inclination, and/or azimuthal, dip to a VTI medium. Which is the case here as we attempt
employing the iterative solver on a more complicated model, we use the same VTI medium
model in Table 1 to estimate a TTI medium by applying a dip in 3D space through the Bond
transformation (Bond, 1943). TTI is considered a more anisotropic case than VTI which
naturally makes the problem more complicated as more variables enter the equation and
therefore have to be taken into account, on top of the trigonometric relations which have to be
made to work with the two reference coordinates in space. But for simplicity sake, the
inclination angle 0 is arbitrarily chosen to be 30, while the azimuthal shift is considered to
subsequently shows the results of the TTI model in the same fashion as Figure 16 does with
the VTI model, for the 3 wave modes and wave quantities plots. There are no major differences
in the graphs when comparing them to the VTI case, similar trends can be readily observed but
with a tilt corresponding to the axis tilt; the cusps in the qSV-wave-front, and the convergence
at the vertical and horizontal axes of most of the quantity pairs. However, slight changes have
to be noted too where the pairs of quantities of different modes coincide in less space and the
contrast between them becomes more apparent. This is probably due to the higher level of
complexity of the medium which unsurprisingly causes irregularities and highlights the
accuracy of ray quantities over phase quantities, this is not the main aim but worth to not. The
qSH-wave though seems to always have the most regular behavior in all cases with all
60
quantities. Once again, obtaining such results validates the iterative solver as a satisfactory
method which works reasonably well with somewhat complex conditions of anisotropy.
Figure 23: Ray and phase velocities (first row), ray and phase attenuations (second row), and ray and phase quality factors
for the three wave modes: qP (first column), qSV (second column), qSH (third column) for a tilted shale model of a viscoelastic
TTI medium based on the Bond transformation
61
5.3 ORT models
complex, and is in need of 9 independent elastic moduli to fully characterize it. On top of that,
adding the viscoelasticity complicates the subject further. However, as new techniques arise
and computing power enhances, the more complicated a model is, the more probable it is closer
to reality than what we simplify it as. Hence, it would be a good challenge to try and tackle this
problem with a new approach. From Hao and Alkhalifah (2017), the model parameters are
given in the Table 6. There are three models which will be used to validate the results of the
perturbation and the RSD methods as the analytic solution is available and proved by Vavryčuk
(2008b). As previously, the aim is to calculate the ray quantities and then compute the error
Figure 24 shows the errors of the computed ray velocity, attenuation, and quality factor using
the two proposed methods for the qP-wave mode for Model C1, which is anisotropic with the
strongest attenuation. Table 7 summarizes the maximum errors for these ray quantities of the
qP- wave mode for the all of three models. In light of these results, one may conclude that the
results obtained by the RSD method have higher accuracy where the error is less than half a
percent than those obtained by the perturbation approach in almost all cases, particularly for
62
ray attenuation and quality factor for all of the three models. For the ray velocity however the
two methods’ results are quite similar with the perturbation method performing slightly better.
Figure 24: The absolute relative errors of the ray velocity, ray attenuation, and ray quality factor of the two methods for the
qP-wave in Model C1
Table 7: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality factor of the two methods for the
qP-wave
For the two shear waves, the qSV- and qSH-wave modes' results for Models C1 and C2, which
exhibit stronger attenuation, it is found that the iteration algorithm proposed by Vavryčuk
(2008b) is time consuming because of the choice of a fine step of the complex slowness angle
for searching for an acceptable slowness vector for a given ray direction. And on the other
hand, the perturbation method results in unacceptable large errors of more than ten percent for
63
the two shear modes for all the models, in contrast with the tolerable order of errors. Having
made that statement, the displayed results are only of the RSD approach which offered much
more appropriate outcomes at least for the weakest attenuation model C3. From these results
one can see that highest errors are commonly around 0.1% except for some which reach about
1~5% (especially for the complex qSV), and that is unavoidable and occurs as a consequence
of the triplication of the qSV-wave-fronts (Figures 25 and 26). Overall, these results confirm
that the RSD approach is an effective and efficient technique to follow for dealing with
Figure 25: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality factor of the two methods for the
qSV-wave in Model C3
Figure 26: The maximum absolute errors of the ray velocity, ray attenuation, and ray quality factor of the two methods for the
qSH-wave in Model C3
64
6. Conclusions
In the grand-scheme of things, the focus of the study is but a pathway for the next step.
The main topic is seismic ray tracing, and the ultimate objective in this case is seismic ray
tracing in viscoelastic anisotropic media. Many studies have delved into studying elastic
isotropic media, which might be fine for certain purposes but not all. It superimposes
errors. Thus, there is a need to develop more and more efficient ways to achieve and deal with
Here, the problem is the accuracy of the ray tracing through the subsurface, which
reflects back as the image of the underlying structures. If the directed waves are not actually
behaving as expected, then the resulting modelling and estimation could vary and might very
well be misleading. When we make the assumption that the earth is elastic and does not
attenuate wave energy as it is travelling through it, then the received energy is going to tell a
whole different story than anticipated. So, in order to do accurate and correct ray tracing,
several parameters need to be studied and obtained; the ray velocity, ray attenuation and ray
quality factor. Not to mention that the calculations become held in the complex domain to
account for the attenuating wave energy. Naturally, these quantities are complex-valued due to
viscoelasticity and the associated attenuation. The ray velocity describes the speed of a wave’s
amplitude propagation, while the ray attenuation describes that amplitude’s decay as the wave
propagates through a viscoelastic medium. The ray quality factor is essentially a measure of
the strength of the attenuation; the lower the quality factor is, the more the attenuation to be
expected is.
What I presented are three methods by which a step closer to more accurate ray tracing
takes place. In order to conduct ray tracing, several of the waves’ as well as the mediums’
properties should be known, but most importantly, three ray quantities which are addressed
65
extensively above: the real ray velocity, the real ray attenuation, and the real ray quality factor.
Phase quantities are also important and could be as important but due to some aspects related
To validate the results of the presented methods, which are numerical approximation
attempts at being comparable with the existing analytic solutions, we compare the outcomes
and measure them against the exact solution and examine the relative errors. The methods are
the iterative algorithm, the perturbation method, and the real slowness direction approach.
Some methods work for some media better than others, and some methods work for some wave
We demonstrate the success in the application of the iterative method to compute the
complex ray velocity which gives the real ray velocity and ray attenuation in a homogeneous
VEAM, namely VTI and TTI. The comparison of the computational expenses for different
media shows that the computing time consumption is increased due to the process of iteratively
running to determine the homogeneous ray velocity vector in a specific direction. It would be
more efficient for a direct solver or a parallel computing scheme to be implemented which will
naturally, the more complex the medium is, the more time it takes to arrive at a convergent
solution. However, and overall, the iterative method is quite accurate in producing reliable
results –for simple and more complex anisotropic media– as it does not incorporate too many
constraining assumptions. This makes it the scale which we can compare the results of other
methods against to assess their validity; as emphasized, the main pitfall of iteration is that it is
time consuming.
As for the perturbation method and the RSD method, they seem to provide highly
satisfactory results, which approach the exact solution with minimal errors in most of the cases.
Even for the different wave modes, and even for moderate and strong viscoelastic VTI media.
66
There are two dimensions which each method should be assessed by; the performance in
handling the degree of complexness of the media (denoted by the strength of anisotropy and
attenuation), and the performance in handling the different wave modes (qP, qSV, qSH) with
the vertical shear wave being the most problematic as it always exhibits wave-front triplication.
The perturbation method excelled in resolving the 3 ray quantities, especially the ray velocity,
for qP and qSH wave modes in the most complex VTI medium model, yet it disappoints for
qSV as the wave-front triplication kicks in and challenges this method. Additionally, for the
more complex ORT media, the results suggest that using the perturbation method is only
satisfactory for the qP-wave, and that is also only for the ray velocity, while it is notably less
effective for highly complex media for both shear waves. Therefore, this method is only
optimal for simpler media and specifically qP-waves, and also particularly for the ray velocity
The RSD approach is quite peculiar, it performs better for more complex cases than it
does for simpler cases. For example, the results indicate that for VTI’s qP wave mode, RSD
yields just reasonable solutions for ray quality and sometimes ray attenuation. On the other
hand, specifically for qSV, RSD always proves to be the best approach to follow to obtain more
trustworthy results. However, it is worthy to note that these most of the results strongly suggest
that this method is insensitive to anisotropy, as several models with varying anisotropy degrees
share very similar accuracies. What is more promising is that although the qSV-wave is
generally problematic due to its wave-front triplications, that was handled well especially by
the RSD approach. However, for the most complicated media tackled here, the viscoelastic
orthorhombically anisotropic media, it was discovered that the perturbation method fails in
computations of both of the shear waves –as stressed above– due to the wave-front triplication.
Moreover, the iterations make for extensive computational time, which is disadvantageous.
Fortunately enough, in these cases, the RSD approach becomes available to efficiently obtain
67
highly acceptable results for all three wave modes (qP, qSV, qSH). Mainly, this method is the
optimal choice when aiming to obtain accurate solutions for higher complexity media, for qSV
for all anisotropy and attenuation cases, and specifically for ray attenuation and ray quality for
less complex media for qP and qSH. However, it still requires further development to be able
Overall, the perturbation approach and the RSD method both show significant results
and exhibit notable success, each for certain settings and specific targets . They surely work
purposefully and yield better results at lower expense (as they cost less computing efforts to
run when compared with the iterative algorithm for example). Both can be applied to seismic
ray tracing in VEAM and yield reliable results depending on the user’s desired application and
goals.
68
References
Aki, K., Richards, P.G., 1980. Quantitative Seismology: Theory and Methods. W.H. Freeman
Auld, B. A., 1973, Acoustic fields and waves in solids. JohnWiley & Sons, Inc.
Behura, J., I. Tsvankin, E. Jenner, and A. Calvert, 2012, Estimation of interval velocity and
attenuation anisotropy from reflection data at Coronation Field, The Leading Edge, 31, 580–587.
Bond, W. L., 1943, The mathematics of the physical properties of crystals, The Bell System
Burton, N., 2007, Rock quality, seismic velocity, attenuation and anisotropy: Taylor & Francis.
Carcione, J.M. & Cavallini, F., 1995. Forbidden directions for inhomogeneous pure shear
Carcione, J. M., 2000, A model for seismic velocity and attenuation in petroleum source rocks,
Carcione, J.M., 2007. Wave Fields in Real Media: Wave Propagation in Anisotropic, Anelastic,
Carcione, J.M., 2014. Wave Fields in Real Media: Theory and Numerical Simulation of Wave
Červený, V., 1972. Seismic rays and ray intensities in inhomogenous anisotropic media
Červený, V., 2001. Seismic Ray Theory, Cambridge University Press, Cambridge, U.K.
Červený, V., 2004, Inhomogeneous harmonic plane waves in viscoelastic anisotropic media:
69
Stud. Geophys. Geod., 48, 167–186.
Chapman, S.J., Lawry, J.H., Ockendon, J.R. & Tew, R.H., 1999. On the theory of complex
Chichinina, T., Sabinin, V., & Ronquillo-Jarillo, G. 2006, QVOA analysis: P-wave attenuation
Crampin, S., E. M. Chesnokov, and R. G. Hipkin, 1984, Seismic anisotropy—the state of the
Hanyga, A. & Seredyńska, M., 2000. Ray tracing in elastic and viscoelastic media, Pure Appl.
Hao, Q., and T. Alkhalifah, 2017, An acoustic eikonal equation for attenuating orthorhombic media,
Hearn, D.J. & Krebes, E.S., 1990a. On computing ray‐ synthetic seismograms for anelastic
Hearn, D.J. & Krebes, E.S., 1990b. Complex rays applied to wave propagation in a viscoelastic
Helbig, K., 1994, Foundations of Anisotropy for Exploration Seismics, Elsevier, NY.
Musgrave, M. J. P., 1970, Crystal Acoustics: Introduction to the Study of Elastic Waves and
Pereyra, V., W. Lee, and H. Keller, 1980, Solving two-point seismic-ray tracing problems in a
heterogeneous medium Part 1. A general adaptive finite difference method: Bulletin of the
70
Shekar, B., and I. Tsvankin, 2012, Anisotropic attenuation analysis of crosshole data generated
Tao, G., and M. S. King, 1990, Shear-wave velocity and Q anisotropy in rocks, A laboratory
study: International Journal of Rock Mechanics and Mining Sciences and Geomechanics Abstracts,
27, 353–361.
Thomsen, L., 1986, Weak Elastic Anisotropy, Geophysics, 51, (10): 1954–1966.
Vavryčuk, V, 2008b, Velocity, attenuation and quality factor in anisotropic viscoelastic media:
Vavryčuk, V. 2008b. Real ray tracing in anisotropic viscoelastic media, Geophys. J. Int., 175,
617–626.
Vavryčuk, V., Svitek, T., & Lokajíček, T., 2016. Anisotropic attenuation in rocks: theory,
Vavryčuk, V., 2007b. Ray velocity and ray attenuation in homogeneous anisotropic
Vavryčuk, V, 2006, Calculation of the slowness vector from the ray vector in anisotropic
Zhou, B., & Greenhalgh, S., 2004. On the computation of elastic wave group velocities for a
Zhou, B., & Greenhalgh, S. A., 2005, ‘Shortest path’ ray tracing for most general 2D/3D
71
and attenuation anisotropy of shales, Whitby, United Kingdom, Geophysics, 81(1), D45–D56.
Zhang, L., 2017. Seismic Ray Tracing and Traveltime Inversion Using Multi-arrivals in
72
Appendices
Appendix A
D1 a11a55 a66 ,
D3 a33a44 a55 ,
D6 a11a22 a66 a66 a44 a22 a55 a66 a12 a66 a44 ,
2
D7 a22 a55 a44 a66 a44 a22 a33a66 a23 a44 a66 ,
2
D8 a55 a66 a11a44 a33 a44 a55 a55 a13 a55 a44 ,
2
D9 a22 a55 a44 a66 a33 a44 a44 a55 a23 a44 a55 ,
2
D10 a22 a55 a44 a66 a55 a55 a66 a11a44 a44 a11a22 a66a66 a33
a13 a55 a22 2 a12 a66 a23 a44 a13 a55 a23 a44 a11
2 2
73
D14 a12 a66 a66 a22 a55 a11 a66 a44 a11a22 a66 a66 ,
2
D15 a13 a55 a55 a44 a55 a11 a66 a33 a55 a66 a11a44 ,
2
D16 a23 a44 a55 a44 a44 a66 a22 a33 a22 a55 a44 a66 ,
2
74
Appendix B
D p p D p p D p p
5
(0) 4
x
(0) 2
z 6
(0) 2
x
(0) 4
y 7
(0) 4
y
(0) 2
z
D p p D p p D p p p
8
(0) 2
x
(0) 4
z 9
(0) 2
y
(0) 4
z 10
(0)
x
(0)
y
(0) 2
z
D p D p D p D p p
11
(0) 4
x 12
(0) 4
y 13
(0) 4
z 14
(0)
x
(0) 2
y
D p p D p p D p D p
15
(0)
x
(0) 2
z 16
(0)
y
(0) 2
z 17
(0) 2
x 18
(0) 2
y
D p ,
19
(0) 2
z
W1 2[ D6 p (0)
y D8 p z D10 p y p D p
4
(0) (0) 4 2 (0) 2 (0) 2
z 14 y
W3 2[ D5 px(0) D7 p (0)
y D10 p x p D p
4 (0) 4 2 (0) 2 (0) 2
y 15 x
+D16 p (0)
y D19 ] p z 6 D3 p z 4[ D p
2 (0) (0) 5 (0) 2
8 x
D9 p (0)
y +D13 ] p z .
2 (0) 3
where the parameters Di i 1 ~ 19 are the imaginary part of the coefficients Di i 1 ~ 19 .
75
Appendix C
2 D06 p 0y D08 pz0 D10 p 0y pz0 D14 p 0y D15 pz0 D17 p x0 ,
4 4 2 2 2
2 D08 pz0 D10 px0 pz0 p y2 D14 p y0 D15 pz0 D17 px0 ,
4 2 2 2
2 D08 pz0 2 D10 p 0y pz0 2 D14 p 0y 2 D15 pz0 2 D17 ,
4 2 2 2
2 D09 pz0 D10 px0 pz0 D14 px0 D16 pz0 D18 p 0y ,
4 2 2 2
2 D09 pz0 D10 px0 p 0y D14 px0 D16 pz0 D18 p y0 ,
4 2 2 2
2 D09 pz0 2 D10 px0 pz0 2 D14 px0 2 D16 p z0 2 D18 ,
4 2 2 2
76
Vz0 6 D03 pz0 4 D08 px0 D09 p 0y D13 pz0 2 D05 px0 pz0
5 2 2 3 4
2 D07 p 0y D10 px0 px0 D15 px0 D16 p 0y C33 pz0 ,
4 2 2 2
2 D07 p 0y D10 px0 p 0y D15 px0 D16 p 0y D19 pz0 ,
4 2 2 2
G2 8 D05 px0 pz0 8D08 px0 pz0 4 D10 px0 pz0 p 0y 4 D15 px0 pz0 ,
3 3 2
G4 30 D03 pz0 2 D05 px0 2 D07 p 0y 12 D08 px0 pz0 12 D09 p 0y pz0
4 4 4 2 2
77