Download ebook Computational Geo Electromagnetics Methods Models And Forecasts Volume 5 Pdf full chapter pdf

Computational Geo-electromagnetics:
Methods, Models, and Forecasts:

Volume 5 - eBook PDF
Visit to download the full and correct content document:
https://ebooksecure.com/download/computational-geo-electromagnetics-methods-mo
dels-and-forecasts-volume-5-ebook-pdf/
Computational Geophysics
Volume 5
COMPUTATIONAL GEO-
ELECTROMAGNETICS
Methods, Models, and
Forecasts
VIACHESLAV V. SPICHAK
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
Copyright Ó 2020 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopying, recording, or any information storage and retrieval system, without permission in writing
from the Publisher. Details on how to seek permission, further information about the Publisher’s permissions
policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright
Licensing Agency, can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the Publisher (other than
as may be noted herein).
Notices
Knowledge and best practice in this field are constantly changing. As new research and experience broaden our
understanding, changes in research methods, professional practices, or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any
information, methods, compounds, or experiments described herein. In using such information or methods they
should be mindful of their own safety and the safety of others, including parties for whom they have a professional
responsibility.
To the fullest extent of the law, neither the Publisher, nor the authors, contributors, or editors, assume any liability
for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or
from any use or operation of any methods, products, instructions, or ideas contained in the material herein.
Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress
British Library Cataloguing-in-Publication Data
A catalogue record for this book is available from the British Library
ISBN: 978-0-12-819631-1
ISSN: 2468-547X
For information on all Elsevier publications visit our website at

https://www.elsevier.com/books-and-journals
Publisher: Candice Janco

Acquisitions Editor: Amy Shapiro
Editorial Project Manager: Andrea Dulberger
Production Project Manager: Omer Mukthar
Cover Designer: Christian J. Bilbow
Typeset by TNQ Technologies
Preface
The mainstream approach in the present-day interpretation
of electromagnetic (EM) data is to solve theoretically nonunique
inverse problem using some regularization technique. Many
examples could be found in the monographs published in the
past 20 years (e.g., Zhdanov, 2002, 2009; Simpson and Bahr, 2005;
Berdichevsky and Dmitriev, 2009; Chave and Jones, 2012; Spi-
chak, 2015). The situation is especially complicated in the 3-D
and 4-D cases, when the number of the unknowns increases
dramatically (e.g., Spichak, 1999; Spies and Oristaglio, 1999;
Zhdanov and Wannamaker, 2002; Spichak, 2019). The huge vol-
ume and high complexity of the data to be processed neces-
sitates the development and application of efficient quantitative
methods and computational tools, in particular, parallelization
of appropriate codes, exploiting high performance computer
resources (which are not always at hand).
Meanwhile, our knowledge of the Earth’s interior as well as
the efficiency of solutions of numerous related problems strongly
depends on accuracy of EM data interpretation. The approaches
used to this end should take into account the purpose of the
study, volume, and quality of the available data, prior geological
and geophysical information, scaling, signal to noise ratio, etc. If
one does not have a regular array of high quality and evenly
distributed EM sounding data, reliable and user friendly software
for their 3-D inversion and multi-processor workstation for
parallel calculations, reading this volume may help to extract
maximum information from the available data.
In contrary to previously published monographs on geo-
electromagnetics, this volume offers the techniques and appro-
priate algorithms for building geoelectrical models under the
conditions of sparse or irregularly distributed EM data, and lack
of prior geological or geophysical information. Secondly, it pro-
vides methodological guidelines for interpretation of electro-
magnetic sounding data depending on goals of the study. Finally,
one may find computational algorithms for using electrical
resistivity as a proxy parameter for forecasting petrophysical
properties of rocks beyond boreholes.
Accordingly, the book has three parts. In Part I (Methodology
of EM data interpretation) two alternative approaches to EM data
inversion are considered: the Bayesian statistical inversion and
neural networkebased algorithms. The former one provides
flexible tools for taking into account prior information and expert
xii Preface
estimates during EM data inversion and quantifying its results in

terms of posterior parameters’ uncertainties. The latter one is
especially useful in the case of constructing 3-D models (in
particular, in terms of macroparameters) from sparse, irregularly
distributed or a single-profile electromagnetic data. Finally, a
review of the methods used for joint analysis and inversion of EM
and other geophysical data is presented.
In Part II (Models of geological medium) the approaches
considered in the Part I are applied to study methodological issues
of EM modeling volcanoes (by examples of Vesuvio, Kilauea,
Elbrus, Komagatake, Hengill), geothermal and hydrocarbon res-
ervoirs. Conceptual models of the Icelandic type crust, a lens in
the upper crust and copper-porphyry ore formation are suggested
based on joint analysis of EM and other geophysical data.
In Part III (Forecasting petrophysical properties of rocks) the
techniques for estimating temperature, seismic velocities, and
porosity from the electrical resistivity as proxy parameter are
considered. This part is supplemented by Appendix, which
includes useful empirical relations between electrical resistivity,
seismic velocities and porosity.
The studies addressed in this volume were supported by
grants from Russian Foundation for Basic Research, Russian
Academy of Sciences, OYO Corporation (Japan), BRGM (France),
Einstein Consortium (Italy), INTAS and VI Framework Program
of European Community. I am very thankful for the funding
support provided by these organizations. I am deeply grateful to
Prof. M. Zhdanov (Utah University) who initiated my interest in
geo-electromagnetics in 1979 and was my PhD thesis mentor, I
am indepted to Prof. T. Mogi (Sapporo University) for providing
friendly support of my studies in computational geo-electro-
magnetics during almost 30 years., It is my great pleasure to
acknowledge Dr. H. Shima (OYO Corporation, Tokyo), Dr. M.
Mareschale (l’École Polytechnique de Montréal), Dr. R. Lewis
(Hobart University), Prof. D. Patella (University Federico II di
Napoli), Prof. A. Siniscalchi (Bari University), Prof. M. Menvielle
(University Paris Sud), Prof. T. Harinarayana (Institute of Science
and Technology, Hyderabad), Dr. A. Manzella (Institute of Geo-
science and Earth Resources, Pisa), Dr. H. Eysteinsson (ISOR,
Reykjavik), Dr. G. Gugunava (M. Nodia institute of Geophysics,
Tbilisi) Dr. P. Calcagno (BRGM, Orleans), Dr. A. Genter (ES-
Geothermie, Electricité de Strasbourg) for fruitful scientific col-
laboration in the framework of joint projects. While developing
computer codes for EM data analysis and interpretation I
Preface xiii
received much assistance from former employees of my Labo-

ratory Drs E. Fomenko and I. Popova.
I acknowledge Marina Nazarenko for partial English trans-
lation of this volume and Alexandra Goidina for assistance with
computer graphics.
Viacheslav V. Spichak
Moscow, July, 2019
References
Berdichevsky, M.N., Dmitriev, V.I., 2008. Models and Methods of
Magnetotellurics. Springer-Verlag, Berlin.
Chave, A., Jones, A. (Eds.), 2012. The magnetotelluric method: theory and
practice. Cambridge University Press, New York.
Simpson, F., Bahr, K., 2005. Practical magnetotellurics. Cambridge University
Press, New York.
Spichak, V.V., 1999. Magnetotelluric fields in three-dimensional geoelectrical
models. Scientific World, Moscow.
Spichak, V.V. (Ed.), 2015. Electromagnetic Sounding of the Earth’s Interior:
Theory, Modeling, Practice. Elsevier B.V., Amsterdam.
Spichak, V.V., 2019. Electromagnetic tomography of the Earth’s interior. Scientific
World, Moscow.
Spies, B., Oristaglio, M. (Eds.), 3D Electromagnetics. SEG Publ., GD7, Tulsa, USA.
Zhdanov, M.S., 2002. Geophysical inverse theory and regularization problems.
Elsevier, Amsterdam.
Zhdanov, M.S., 2009. Geophysical electromagnetic theory and methods. Elsevier,
Amsterdam.
Zhdanov, M.S., Wannamaker, P. (Eds.), 2002. Three-Dimensional
Electromagnetics. Elsevier, Amsterdam.
3-D EM forward modeling
1
techniques
1.1 Introduction
To solve the forward 3-D electromagnetic (EM) problem (in a fre-
quency domain) implies to determine the electromagnetic field (E;
H) satisfying the Maxwell equations
VH ¼ e
sðr; uÞE þ jext ; V E ¼ iumH; (1.1)
at a given frequency u, from the distribution of generalized

complex electrical conductivity esðr; uÞ ¼ s iuε (m is a magnetic
permeability, ε is a permittivity, jext ðr; uÞ is the external current
density) specified in a certain spatial region that includes an
inhomogeneity. In this chapter we assume for simplicity that j ext ¼
0; m ¼ m0 ¼ 4p,107 H/m, where m0 is the free-space magnetic
permeability, and neglect the displacement currents, i.e., e sh s. In
other words, we will consider modeling of the electromagnetic
fields induced in 3-D earth by natural sourcedplane wave (so-
called “magnetotelluric” (MT) fields) though the issues discussed
below in most cases refer equally to controlled source excitation.
Various approaches to the solution of this problem exist, all of
them falling into two large groups comprising integral equations
and differential equations techniques. Currently, numerical
computer-aided modeling of electromagnetic fields becomes a
powerful and relatively easy-of-access tool for analyzing
complicated situations. Whereas formerly the scope of geophys-
ical consideration was confined to simplified model media (1-D
e 2-D), the advent in designing powerful computer facilities
made it possible to pose the problem of numerical modeling of
electromagnetic fields excited by 3-D sources in a two- or
three-dimensional medium or by 2-D sources in a three-
dimensional medium (see review papers by Avdeev (2005,
2015); Boerner (2010); Newman (2014) and references therein).
In this chapter, we will consider briefly the numerical (Sec-
tions 1.2e1.4) and analog (Section 1.5) approaches to the forward
problem solution. Numerical methods, in turn, are subdivided
Computational Geo-Electromagnetics. https://doi.org/10.1016/B978-0-12-819631-1.00001-8

Copyright © 2020 Elsevier Inc. All rights reserved. 3
4 Chapter 1 3-D EM forward modeling techniques
into two large groups: integral equation methods (Section 1.2) and
differential equation methods (Section 1.3). A comparative anal-
ysis of these two groups of methods and hybrid approaches (Sec-
tion 1.4) based on their combination is present.
Section 1.6 describes main characteristic features of a balance
technique for EM field’s calculation in media with arbitrary 3-D
distribution of the electrical conductivity. At the same time, if the
electrical conductivity model shows a certain symmetry type, the
above-mentioned purpose can be achieved using smaller com-
puter resources. In particular, for 3-D models with a vertical axial
symmetry, the vector problem reduces to a scalar one. Section
1.7 addresses an algorithm for numerical calculation of quasista-
tionary electromagnetic fields in 3-D axially symmetric media
based on a finite element modification of the balance method.
1.2 Methods of integral equations

The integral equation method for the numerical computation of
electromagnetic fields was pioneered by Dmitriev (1969). The
basic ideas of this approach were developed by Raiche (1974),
Hohmann (1975), Tabarovsky (1975), and Weidelt (1975). Subse-
quently, the integral equation method was successfully elaborated
by Dmitriev and Farzan (1980), Ting and Hohmann (1981), Hoh-
mann (1983), Wannamaker et al. (1984a,b), Hvozdara (1985),
Hvozdara et al. (1987), Wannamaker (1991), Xiong (1992), Pankra-
tov et al. (1995), Singer (1995), Singer and Fainberg (1995, 1997),
Xiong and Tripp (1995), Avdeev et al. (1997), Farquharson and
Oldenburg (1999), Lee et al. (1999), Portniaguine et al. (1999),
Xiong et al. (1999a,b), Hursan and Zhdanov (2002), Zhdanov
et al. (2000, 2006), Singer (2008), Endo et al. (2009), Avdeev
(2015), Kruglyakov et al. (2016).
Parallel to the method of volume integral equations (VIE) that
involves integration over the entire volume of the anomalous
domain, a method of surface integral equations (SIE) implying
integration only over the boundary of the domain has been devel-
oped (Oshiro and Mitzner, 1967; Dmitried and Zakharov, 1970;
Tabarovsky, 1971; Smagin, 1980; Liu and Lamontagne, 1999).
Let us consider these two approaches.
1.2.1 The method of volume integral equations

Volume integral equation (VIE) technique is based on the numer-
ical solution of the second-type Fredholm equation, which can be
derived from Maxwell equations using Green functions:
Chapter 1 3-D EM forward modeling techniques 5
ZZZ ^
EðrÞ ¼ Eb ðr0 Þ þ ium0 ðs sb ÞG e ðr = r0 ÞEðr0 ÞdV 0 ; (1.2)
where Eb is a background field calculated with the assumption

that s ¼ sb (background electrical conductivity), G ce ðr =r0 Þ is the
0
electrical-type Green tensor, r and r are coordinates of the obser-
vation and source point, respectively, and V is the region in ques-
tion. (Hereinafter, for the sake of definiteness only the equation
for the electrical field E will be considered. Once the numerical
solution to this equation is obtained, the magnetic field H can
be readily calculated from, e.g., the second Maxwell equation
(see in this connection Section 1.6.4)).
The pivotal idea of this approach is as follows. The anomalous
region is divided by a spatial grid into cells. Within each cell, the
field is assumed to be constant. Therefore, in the second term of
(1.2), it can be taken outside the integral sign. To obtain a system
of linear algebraic equations for the field, one has to only calculate
the respective tensor coefficients within each unit cell.
Let us discuss the main advantages and drawbacks of this
method. The advantages comprise its greater physical transpar-
ency compared, e.g., to the methods of differential equations.
In addition, as is apparent from (1.2), the second term on the
right-hand side is nonzero only if sssb ; i.e., integration must be
carried out only over the anomalous region alone. Finally, the nu-
merical approximation of (1.2) does not involve the unstable pro-
cedure of numerical differentiation, which is typical of most the
differential equation methods.
On the other hand, significant difficulties arise in the numer-
ical approximation of the second term in the right-hand side of
Eq. (1.2). In particular, determination of the Green tensor
^
G e ðr =r0 Þ components is a nontrivial computational problem,
solution of which involves the Hankel transforms (Hohmann,
1975; Weidelt, 1975; Farquharson and Oldenburg, 1999; etc.) or
the linear filtration technique (Das and Verma, 1981; Verma and
Das, 1982). Computation of the tensor coefficients can be as
time consuming as itself solving the system of linear algebraic
equations. Therefore, it is a “bottleneck” of the technique, and
the success of the whole approach thus depends on how effec-
tively this problem is solved.
Another difficulty arises while solving the system of linear
algebraic equations. Dense matrix of the system makes limited
computer resources a critical factor. In turn, the limitation on
the dimensionality of algebraic system ultimately entails limita-
tions to the mathematical model: we can only compute the fields
for bodies that are not too large. Xiong (1992) overcame this draw-
back by partitioning the scattering matrix into many block sub-
matrices and solving the whole system by a block iterative
method. This reduces the computer memory requirements and
time of computation. (It is worth mentioning in this connection
that much more drastic reduction of the computer resources
could be achieved by using the so-called “sparse matrix tech-
nique” (Poggio and Miller, 1973) that implies ignoring the interac-
tion among sufficiently remote areas of the modeling domain; the
latter corresponds to zeroing those elements of the matrix that are
small compared to the diagonal ones.) Another simplification
(which also reduces the class of objects being modeled) is to solve
the problem within the long-wave approximation. This enables
one, firstly, to simplify computation of the Green tensor (Hvoz-
dara, 1981) and, secondly, to use in the numerical solution of
(1.2) the Born approximation (Torres-Verdin, 1985; Torres-Verdin
and Bostick, 1992), localized nonlinear approximations (Habashy
et al., 1993; Torres-Verdin and Habashy, 1994), quasilinear
approximation and series (Zhdanov and Fang, 1996, 1997) and
quasianalytical series (Zhdanov et al., 2000). Portniaguine et al.
(1999) suggested a way of reducing the time of computations by
the use of “compression matrix” that converts the original dense
matrix into the sparse one by constructing an interpolation pyra-
mid in multiple dimensions. Finally, Endo et al. (2009) suggested a
multidomain approach, while Zhdanov et al. (2006) proposed an
algorithm based on using inhomogeneous background conduc-
tivity, which increase competitiveness of this approach in com-
parison with differential equation techniques (Zhdanov et al.,
2013).
In evaluating the advantages and drawbacks of the VIE
method on a whole, it should be noted that this method seems
giving the most accurate resultsdin a comparatively short timed
for anomalies that are small compared to the wavelength within
the anomalous region and have a simple shape.
1.2.2 The method of surface integral equations

Surface integral equation (SIE) technique makes use of electro-
magnetic field representation as an integral over the surface of
a domain by means of StrattoneChu formulas (Dmitriev and
Zakharov, 1970), method of auxiliary (fictitious) sources (Tabarov-
sky, 1971), method of potentials (Smagin, 1980), or the method
employing the Helmholtz scalar equations (Liu and Lamontagne,
1999). After passing to the limit with the observation point tending
to the boundary of the domain from inside and from outside in
turn, one can, using the continuity of the tangential components

of electromagnetic field on the boundary, obtain the necessary
equations only over the boundary of domain V with respect to un-
known field densities. Numerical solution of these equations in-
volves the KryloveBogolyubov method, variational approaches,
BubnoveGalerkinetype methods, etc. Smagin and Tsetsokho
(1982) have obtained the system of linear algebraic equations us-
ing the collocation method and the densities sought for were
approximated by smooth finite functions, which form a finite
partition of unity over the surface of the domain.
This technique has an obvious advantage over the integral
equation method. Numerical solution to the equations written
over the surface of the domain rather than over its volume in-
volves a considerable reduction in the dimensionality of the sys-
tem of linear algebraic equations. This is especially important
when modeling large inhomogeneities because the ratio of the di-
mensions of respective matrices decreases with an increase in the
characteristic linear dimension L as 1/L.
Despite this advantage, however, the use of SIE technique is
limited to modeling homogeneous bodies with rotational sym-
metry, embedded in a homogeneous space (half space) (Oshiro
and Mitzner, 1967; Liu and Lamontagne, 1999). This is due to
the imperfection of the mathematical apparatus for modeling
inhomogeneities with an arbitrary distribution of anomalous elec-
trical conductivity sa ¼ s(x,y,z) e sb(z) embedded in a horizontally
layered section and to the difficulties arising at the step of numer-
ical approximation of singular integral equations and of solving
the resulting system of linear algebraic equations with an ill-
conditioned matrix.
Therefore, integral equation methods are most helpful when
modeling inhomogeneities of comparatively simple shape and
small size (compared to the wavelength within them) embedded
in a horizontally layered medium with a small number of layers.
1.3 Methods of differential equations

Advances in geoelectrics call forth computing 3-D electromagnetic
fields in complicated situations when the inhomogeneity exhibits
an arbitrary shape and dimensions (in particular, it may not be
local) and the anomalous electrical conductivity in the model
varies arbitrarily, as, e.g., in regional models. In this case, as
pointed out above, integral equation methods are of limited use.
The only viable approach to such problems involves either a
direct solution of differential equations with partial derivatives,
or the use of so-called differential equation methods, of which

two main ones being the finite difference (FD) technique and
the finite element (FE) technique. Let us list the main features
of their application to the problem of numerical modeling of
3-D electromagnetic fields.
1.3.1 The finite difference technique

The FD technique based on the finite difference approximation of
derivatives is used for the numerical solution of a differential
equation with respect to the electrical or magnetic field in a
certain spatial region that contains an inhomogeneity:
DE V,ðV , EÞ þ k 2 E ¼ 0 (1.3)
2
k ¼ ium0 s; Rek > 0 ; or

1
V V H þ ium0 H ¼ 0; (1.4)
s
which follow directly from (1.1).

The advantages of this technique are its greater versatility
compared to integral equation methods, the simplicity of its nu-
merical implementation, band structure of the matrix of resulting
system of linear algebraic equations. The latter circumstance is
vital because it significantly reduces the time and amount of
operational computer memory required for solving the problem
and, consequently, enables modeling the targets of large size or
with complicated conductivity distribution in it.
The main factors influencing the effectiveness of this method
are as follows:
- choice of the appropriate governing equation;
- selection of the discretization scheme;
- approximation of the second derivatives in (1.3) or (1.4);
- specifying boundary conditions that are valid at nonindefinite
distance from the anomaly;
- conservation of currents and magnetic field flux;
- accuracy when solving the problems at frequencies approach-
ing the static limit (similar problem is met by the FE
technique);
- accuracy when solving the problems with sharp conductivity
contrasts.
Since the first 3-D modeling results were obtained with this
method by F. Jones (Jones and Pascoe, 1972; Lines and Jones,
1973; Jones, 1974; Jones and Lokken, 1975; Hibbs and Jones,
1978), a number of improvements were made that significantly

increased the effectiveness of FD technique (Zhdanov and Spi-
chak, 1980; Zhdanov et al., 1982; Lam et al., 1982; Spichak,
1983a,b; Zhdanov and Spichak, 1989, 1992; Mackie and Madden,
1993; Mackie et al., 1993, 1994; Druskin and Knizhnerman, 1994;
Alumbaugh and Newman, 1996; Smith, 1996a,b; LaBreque, 1999;
Spichak, 1999; Weidelt, 1999; Weaver et al., 1999; Xiong et al.,
1999b; Sasaki, 2001; Wang and Fang, 2001; Fomenko and Mogi,
2002; Newman and Alumbaugh, 2002; Weiss and Newman, 2002;
Streich, 2009; Mittet, 2010; Yavich and Zhdanov, 2016; Wang
and Tan, 2017; Cao et al., 2018).
Zhdanov and Spichak (1980) suggested a balance technique for
approximation of the FD equation in electrical field (see Section
1.6 later), which enabled the accuracy of the results to increase
due to reduction from the unstable calculation of the second-
order derivatives to the first-order ones. Spichak (1983a) proposed
to use a current divergence-free condition in the balance tech-
nique, which raised the accuracy of the results due to disappear-
ance of the second term in Eq. (1.3) in regions with zero
conductivity gradient and also at the static limit. Zhdanov et al.
(1982) and Spichak (1985, 1999, 2006) introduced the asymptotic
boundary conditions that greatly diminished the dimensions of
the modeling domain and really increased the accuracy of the for-
ward modeling. Smith (1996a) proposed to use a staggered grid
method (pioneered by Yee (1966)) that guarantees automatic con-
servation of currents and magnetic flux on the grid (though, at the
cost of some inconvenience when computing some field compo-
nents at the earth surface or the field transformations involving
the components specified on different grids). Davydycheva and
Druskin (1999), Weidelt (1999), and Wang and Tan (2017)
extended this approach to the anisotropic media.
Druskin and Knizhnerman (1994), Druskin et al. (1999), Jin
et al. (1999), Zhou et al. (2013) use a spectral Lanczos decomposi-
tion method (SLDM) with Krylov subspaces generated from the
inverse of the Maxwell operator. SLDM enables acceleration of
the forward modeling owing to possibility of getting the solution
for the whole frequency range practically at the cost of solution
for a single frequency.
Smith (1996b) developed a static correction procedure that
explicitly enforces the electrical current and magnetic field
divergence-free conditions that, in turn, increases the accuracy of
the EM field calculation when the frequency tends to zero. In order
to accelerate the solution of the forward problem at low induction
numbers (LINs), Newman and Alumbaugh (2002) proposed an LIN
preconditioner. It is based on splitting the electrical field into curl-
free and divergence-free projections that removes the null space of

the discrete curl-curl operator in the solution process.
A number of efficient solvers and preconditioners are used
presently in order to achieve both a good accuracy and high
convergence rate of the iteration process, especially when high
conductivity contrasts increase the condition number of the ma-
trix of the system of linear equations (SLE). Fomenko (1999), Var-
entsov (1999), Siripunvaraporn et al. (2002), Grayver and Streich
(2012), Varilsuha and Candansayar (2018) carried out a compara-
tive analysis of different preconditioners and solvers.
Finally, Zhdanov et al. (1982) and Spichak (1999) constructed
internal criterions for controlling the accuracy of the forward
modeling results (see Section 1.6.5.1 later).
1.3.2 The finite element technique

The finite element technique is especially useful for EM modeling
in the regions with complicated geometry or relief earth surface.
Due to this finite element method became very popular in compu-
tational electromagnetics (Reddy et al., 1977; Pridmore et al., 1981;
Boyse et al., 1992; Livelybrooks, 1993; Mogi, 1996; Haber, 1999;
Sugeng et al., 1999; Zunoubi et al., 1999; Zyserman and Santos,
2000; Badea et al., 2001; Jin, 2002; Mitsuhata and Uchida, 2004;
Farquharson and Miensopust, 2011; Mukherjee and Everett,
2011; Schwarzbach et al., 2011; Silva et al., 2012; Um et al., 2012,
2013; Vieira da Silva et al., 2012; Kordy et al., 2013; Persova et al.,
2013; Puzyrev et al., 2013; Ren et al., 2013a,b; Rivera-Rios
et al., 2013; Ansari and Farquharson, 2014; Cai et al., 2014; Jahan-
dari and Farquharson, 2014, 2015; Cai et al., 2017a,b; Jahandari
et al., 2017; Liu et al., 2018a,b; Xiao et al., 2019).
In the framework of this approach, the solution of Eq. (1.1) in
the finite domain V is equivalent to finding the stationary point of
the energy functional:
ZZZ

F ðEÞ ¼ V E , V E k2 E , E dV : (1.5)
V
(Note that we are dealing with the determination of a stationary

point, not the minimum of the functional F because the differential
operator on the right-hand side of (1.5) is not positively defined and
hence does not satisfy the conditions of the Minimum theorem).
Haber (1999), Badea et al. (2001), Ansari and Farquharson
(2014), Jahandari and Farquharson (2015) use to this end a
potentialecurrent formulation of the initial problem based on
the Helmholtz decomposition. The resulting matrix has a property
of the diagonal dominance that, in turn, allows its efficient solu-

tion even for very high conductivity contrasts.
An alternative way to overcome the same problem is based on
using of a single vector shape function at each edge of the grid
instead of three scalar functions defined in corner nodes. Due
to this, one has to only solve for tangential components of the
electrical field along the edges of cells (“edge-based finite ele-
ments”). Thus, avoiding a solution for normal components of
the field allows modeling for high contrasts (Sugent et al., 1999;
Mukherjee and Everett, 2011; Kordy et al., 2013; Rivera-Rios
et al., 2013; Cai et al., 2014; Xiao et al., 2018).
Jahandari and Farquharson (2014, 2015) have introduced
mixed “finite volume” schemes based on both the EM field and
the potential formulations of Maxwell’s equations (Nedelec,
1986). Comparison between three staggered grid finite volume
and two edgeebased finite element modeling schemes (Jahandari
et al., 2017) shows that while they are comparable in terms of ac-
curacy and computation resources, the finite elements schemes
are slightly more accurate but also more expensive than the finite
volume ones.
It is also important to optimize the hexahedral irregular
meshes. Persova et al. (2013) eliminate “unnecessary” nodes,
while Cherevatova et al. (2013) represents a full grid as a vertical
stack of subgrids each of which is a standard staggered grid. It is
refined only in the horizontal direction, uniformly across the
vertical layers, allowing only a factor of two between vertically
adjacent subgrids. Schwarzbach et al. (2011), Ren et al. (2013a),
and Liu et al. (2018b) use to this end adaptive refinement proce-
dure, which iteratively refines the finite element grid guided by
global posterior error estimators. Basing on numerical experi-
ments Ren et al. (2013a) concluded that the error estimator using
face jumps of normal components of current density embedded
in the goal-oriented adaptive refinement procedure shows the
most efficient performance.
In order to reduce the total computation time, Zyserman and
Santos, (2000) proposed a “mixed hybrid domain decomposed
iterative nonconforming” method that is based on the iterative
decomposition of the model domain and solving the local linear
systems of equations. This approach allows reduction of memory
and time required for solution of the appropriate system of linear
equations. Note in this connection that more significant
reduction of the computation time (not only in the FE based algo-
rithms) could be achieved my means of using parallelization
schemes (Alumbaugh and Newman, 1996; Newman and
Alumbaugh, 1999; Wilson et al., 1999; Kordy et al., 2013; Puzyrev

et al., 2013; Newman, 2014; Cai et al., 2017a).
Despite the fact that, theoretically, finite elements are well
suited to modeling the geological sections with complex distri-
butions of electrical conductivity, the flexibility of the method
is attained through considerable computing efforts associated
with the use of complicated finite elements. To characterize dif-
ferential equation methods in general, it should be emphasized
that they are generally more versatile compared to integral
equation methods. However, to increase their computing
efficiency, a number of problems discussed above have to be
properly solved.
1.4 Hybrid schemes

One way to refine the techniques of numerical modeling of 3-D
electromagnetic fields is to apply hybrid (“mixed” or “nonclas-
sical”) approaches that blend the advantages of the differential
and integral equation methods. This implies the differential
equation method to be used inside the modeling domain, which
permits consideration of a model with an arbitrary distribution of
the electrical conductivity and integral relations between field
components to be employed at the boundary, which ensures
the possibility of limiting the modeling domain to within an
area that only slightly exceeds the inhomogeneity in dimensions.
The first step on this way was made by Weidelt (1975) who
used the Green functions apparatus to remove from a consider-
ation the regions of the normal section above and below the
inhomogeneities:
Z Z nh i o
Ea ðr0 Þ ¼ b e ðr0 jrÞ ½n Ea ðrÞ dS;
VG (1.6)
S
where S is the surface of the discontinuity separating the layers, Ea

is the anomalous (secondary or scattered) field (Ea hE Eb ), and
the points r0 are selected in the region to be removed.
The efficiency of hybrid schemes was further improved by
reducing the modeling domain not only vertically but horizon-
tally as well. Thus, in (Gupta et al., 1987; Lee et al., 1981;
Pridmore and Lee, 1980; Best et al., 1985), an area of limited di-
mensions is covered over a small distance by a finite element
grid. Applying the finite element technique to the inner part of
the domain and the integral equation technique to the outer
part gives rise to two systems of equations: (1) finite element
equations in the inner nodes, with the respective part of the ma-
trix showing a characteristic band structure and (2) equations
that relate the unknowns on the boundary and inner nodes
(with the respective part of the matrix being filled) in accordance
with the formula
ZZZ
Ea ðr0 Þ ¼ b e EðrÞdV ;
sa G (1.7)
where r0 lies on the boundary of the domain V.

Ren et al. (2014) have proposed a finite element scheme with a
hybrid boundary element (BE), which enables complete removal
of the volume discretization of the air space, which, in turn, pro-
vides the capability of simulating large-scale complicated geoelec-
tromagnetic induction problems. In the framework of this
approach, the surface integral formula in terms of the reduced
electrical vector potential is approximated using the point collo-
cation boundary element method. Numerical experiments show
that at low frequencies, where the quasistatic approximation is
applicable, standard FE methods prove to be superior to the
hybrid BEeFE solutions in terms of computational time, because
the FE method requires only a coarse discretization of the air
domain and offers an advantageous sparsity of the system matrix.
At radiomagnetotelluric frequencies of a few hundred kHz, the
hybrid BEeFE scheme turns to be superior to the FE method
because it avoids explicit storage of the system matrices as well
as dense volume discretization of the air domain required by FE
methods at high frequencies.
Hybrid schemes are commonly grouped into two types: direct
and iterative. In direct hybrid techniques, the matrix of the sys-
tem is inverted directly, whereas in iterative techniques, the
values of the unknowns at the domain boundaries are at first
assumed to be known, and then the values in the inner nodes
are computed, which are recalculated by (1.7) into boundary
values, etc. As appears from the estimation performed by Lee
et al. (1981), the iterative hybrid schemes are generally more
sparing in terms of the operational memory resources and CPU
time used. Similarly to other numerical modeling techniques their
further reduction requires using parallel computing.
Therefore, despite certain advantages of hybrid systems, their
practical application entails a number of difficulties. In partic-
ular, the requirements on computer resources remain rather
high, convergence of the iterative procedure is not guaranteed
(if the most resource-saving iterative scheme is considered) and
no internal criteria for the accuracy of the results obtained are
used (The last observation, which also applies to the rest of the
numerical approaches just mentioned, is perhaps most critical
for hybrid schemes).
1.5 Analog (physical) modeling approaches

Analog, or physical, modeling implies a physical nature for the
model of the primary field source, the medium, and the inhomo-
geneity. Two basic approaches are distinguished here: the contin-
uous media method and the electrical circuit method (Tetelbaum
and Tetelbaum, 1979). In the former approach, the model is deter-
mined by the field of an electrical current in a continuous medium
(Dosso, 1966), while the latter involves the use of electrical circuits
with concentrated parameters (Brewitt-Taylor and Johns, 1980).
The continuous media method is based on the use of the elec-
trodynamic similitude criterion (Stratton, 1941):
musL2 ¼ inv; (1.8)
where L is the characteristic linear dimension of the model. Among

the advantages of this approach, one should recall its simplicity
and low cost, the uniqueness of the equipment, and the possibility
of modeling the media with sharp electrical conductivity contrasts
(Berdichevsky et al., 1987; Farquharson et al., 2006).
Some of these advantages, however, prove to be drawbacks if
the issue is addressed in a broader context, namely, from the
standpoint of electromagnetic fields modeling in real situations.
The extremely small number of physical modeling installations
all over the world virtually obviates their application as a tool for
analyzing the observed fields, let alone the repeated use of the
modeling results, which is only possible through coupling physical
installations to computing facilities. In addition, one should note a
number of technologic difficulties, such as the choice of materials
with a sufficiently wide range of electrical conductivity variations
but not subject to frequency dispersion, the painstaking techno-
logic implementation of models for multilayer media, the poorly
developed modeling technique for use with hard materials and
low-melting metals, and the difficulties arising in obtaining exper-
imental curves over a broad frequency range.
The other important approach in physical modeling is the elec-
trical circuit method, which in actual practice is realized in two mod-
ifications. The first is based on discretization of the modeling
domain followed by the representation of the elementary volumes
obtained by means of elements of an electrical circuit (replacement
schemes) and measurement of voltages and currents in the circuit,

which model the electrical and magnetic fields, respectively. The
other is based on obtaining replacement circuits directly from finite
difference equations that describe the field being modeled (Brewitt-
Taylor and Johns, 1980).
As was pointed out in (Tetelbaum and Tetelbaum, 1979, p.
211), “the prime advantage of electrical circuits is that they enable
one to model three-dimensional fields, which are described by
equations with a right-hand side”. Another important advantage
of this approach is that it can be used in hybrid analog/numerical
installations. It is also worthwhile to note that with this approach,
the use of the diacoptics concept of Kron (1972) may prove
noteworthy.
Despite the outward merits of the above approach, however, it
is not free of the characteristic flaws of both the other physical
modeling techniques (measurement errors and technical diffi-
culties in constructing the models) and mathematical modeling
approaches (e.g., errors ensuing from the discretization of the
modeling domain and field equations).
We have discussed the main techniques and approaches
currently in use for modeling 3-D electromagnetic fields.
Collating their advantages and drawbacks shows the differential
equation methods to be the most versatile and best suited for
modeling the broadest spectrum of practical situations. However,
their application involves a number of theoretical, methodolog-
ical, and computing problems that affect the modeling efficiency.
In the following sections, these crucial issues will be addressed in
the context of the balance technique.
1.6 Balance technique for EM field

computation
In publications (Zhdanov and Spichak, 1980, 1989; Spichak,
1983a, 1985, 1999), basic principles were formulated of construct-
ing an algorithm for this problem bearing on a moderate-speed
computer with a limited core memory (for instance, minicom-
puters). In what follows, we will describe this algorithm and
demonstrate its application to solution of the forward problem
in 3-D axial symmetric situations.
1.6.1 Governing equations

Let some domain U in the Earth’s crust or upper mantle be
isotropic, nonmagnetic (mhm0 is the permeability of free space),
and to have a three-dimensional distribution of electrical conduc-

tivity sðx; y; zÞ that can be represented in the form:
0 sðzÞ; P˛U 1
sðPÞ ¼ @ sðx; zÞ; P˛U2 ;

sðx; y; zÞ; P˛U3
where U ¼ U1 WU2 WU3 , P ¼ Pðx; y; zÞ; with U3 u 0. The electro-

magnetic field in the domain U is induced by a plane wave verti-
cally incident on the Earth’s surface. The time dependence of the
field is given by factor expðiutÞ.
Given the conductivity distribution sðx; y; zÞ, it is necessary to
determineP electrical and magnetic fields everywhere in the
domain ¼ UWU0 , where U0 is the lower atmosphere adjacent
to U.
Over periods of interest to geophysics, the field in the domain
S is quasistationary and satisfies the Maxwell equations:
V H ¼ sE; (1.9)
V E ¼ ium0 H (1.10)
Eq. (1.9) e (1.10) yield the electrical field equations:
DE VðV , EÞ þ k 2 E ¼ 0; (1.11)
where k ¼ ðium0 Þ1=2 ; Rek > 0:

Taking the divergence from the two sides of Eq. (1.8), we
derive
sðV , EÞ þ ðE; VsÞ ¼ 0 (1.12)
With due account taken of (1.12), Eq. (1.11) takes the form:
DE þ VðE; Vln sÞ þ k2 E ¼ 0: (1.13)
To determine the electromagnetic field in the domain S we
have to solve a boundary value problem for the field E satisfying
Eq. (1.13) within S and then calculate H.
1.6.2 Boundary conditions

Let us now consider the choice of boundary conditions at the
boundaries of the domain S: The values of electrical field or of
its normal derivative are not known beforehand. So, a common
approach is based on the assumption that anomalous
electromagnetic field vanishes at the boundaries located at finite
distance from the anomaly (Dirichlet type boundary conditions).
However, this causes two problems: first, it is difficult to deter-

mine this distance correctly and, second, in order to ensure its
correctness it is necessary to place the boundaries of the modeling
domain far enough from the anomaly (especially in the noncon-
ducting parts), which, in turn, leads to unnecessary increasing
of the computation time.
One of possible solutions make use of integral boundary con-
ditions (Weidelt, 1972), but it could be very difficult to use them in
practice because of their entailing considerable computing
difficulties. In this connection, it would be interesting to consider
another type of boundary conditions that are based on the im-
plicit account of the character of the anomalous field decaying
far from the medium heterogeneities. Spichak (2006) suggested
a general approach, which may be used to construct the differen-
tial boundary conditions for the vector fields, satisfying the
Helmholtz equation. It is based on the expansion of the fields in
a series of multipoles and use of “canceling” operators.
In particular, the algorithm in question makes use of the
asymptotic boundary conditions of the first order:

v
1 ikr þ r ðE Eb Þ ¼ 0; (1.14)
vr
where Eb is the background electrical

P field corresponding to the
case when shsðzÞ for all P˛ ; r is the distance P between the
points lying on the boundary of the domain : and the origin
of the coordinates.
Basing on the results of the numerical modeling Spichak
(2006) has concluded that it is possible to make the following rec-
ommendations useful for application of the boundary conditions
when solving the external boundary value problems of
geoelectromagnetics:
- the earth responses are not equally affected by too small dis-
tance between the target and the modeling domain bound-
aries: the apparent resistivity is more sensitive in the case of
the resistive target while the phases are more sensitive in the
case of the conductive one;
- the “safe”adistance for using the Dirichlet-type boundary con-
!
ditions E ¼ 0 should not be less than 0.5 d, where d is a skin
depth in the background structure.
1.6.3 Discretization scheme

To obtain a system of linear algebraic equations and its subsequent
computer-aided solution, a transition from continuous to discrete
values is required. Various mechanisms for such a transition are

mainly based on (1) the application of Taylor series, (2) the varia-
tional formulation, and (3) integration of the primary equation.
Each of these approaches has its advantages and respective fields
of applicability. Thus, the first technique applicable to differential
equations in general case is used most often in defining the order
of approximation in difference schemes. The variational formula-
tion is inviting in that, in its context, natural boundary conditions
are a direct consequence of the relevant functional being stationary
(Pridmore et al., 1981). Lastly, the third approach, often referred to
as the “balance technique,” is applicable in general case and yields
particularly simple discrete schemes with internal boundaries and
nonuniform grids (Spichak, 1983a). Let us refer to this last tech-
nique to obtain discrete analogs of Eq. (1.13).
Provided that the electrical conductivity in the region
varies stepwise, derivatives of the function being sought for
may undergo breaks. It thus is worthwhile to compute the
values of the function itself at the nodes of a certain grid while
specifying values of the electrical conductivity function at the
nodes of an intermediate grid (Brewitt-Taylor and Weaver,
1976) with no worry to satisfy internal boundary conditions.
Similar idea is realized in the staggered grids approach
mentioned earlier.
To derive discrete equations for space grid nodes, let us
proceed from the continuous vector function E to the discrete
one Ul;m;n , defined at the nodes of the main grid. Integrating
Eq. (1.13) with respect to the volume of an elementary cell in
the vicinity of the node (l; m; n) (Fig. 1.1), we obtain an equation
for electrical current balance:
ZZ ZZ ZZZ
VUl;m;n ds þ ðUl;m;n ; Vlnsl;m;n Þds þ 2
kl;m;n Ul;m;n dV ¼ 0;
Sl;m;n Sl;m;n Vl;m;n
(1.15)
where l ¼ 1,2, ., L; m ¼ 1,2, ., M; n ¼ 1,2, ., N.

Substituting the derivatives of U and s in Eq. (1.15) by finite
differences and approximating the integrals by the trapezium
rule, we arrive at a linear algebraic equation relating the values
of the vector function U only in seven adjacent nodes:
1
^ ð0Þ ^ ð1Þ ^ ð2Þ ^ ð3Þ
Ul;m;n ¼ Dl;m;n ðDl;m;n Ul;m;n1 þ Dl;m;n Ul;m1;n þ Dl;m;n Ul1;m;n þ
^ ð4Þ ^ ð5Þ ^ ð6Þ

þ Dl;m;n Ulþ1;m;n þ Dl;m;n Ul;mþ1;n þ Dl;m;n Ul;m;nþ1 Þ
Figure 1.1 Elementary cell of a spatial grid.

ð1 ¼ 1; 2; :::L; m ¼ 1; 2; :::M; n ¼ 1; 2; :::N (1.16)
ðiÞ
b
where D l;m;n (i ¼ 1,2,.6) are the matrix coefficients having a size
(3 3) and determined by the grid geometry, distribution of the
electrical conductivity s and EM field frequency.
Note that numerical approximation of Eq. (1.13) has a number
of advantages compared to the approximation of the correspond-
ing second-order differential Eq. (1.11):
- the balance of currents is hold within each cell of the spatial
grid;
- the use of Eq. (1.13) makes it possible to avoid the approxima-
tion of mixed second derivatives in (1.11) and explicitly en-
forces current divergence-free conditions at each grid node
(compare with staggered grid approximation (Smith, 1996a));
- the approximation of the third term in Eq. (1.13) is accurate
enough in the vicinity of electrical conductivity contrasts as
well;
Figure 1.2 Structure of matrix D: L, M, Ndnumber of nodes in X, Y, and Z

axes, respectively; Q ¼ LMN, D b ðiÞ (3 3)dsubmatrix.
- in the regions where s ¼ 0 or Vs ¼ 0 the second term of the

equation automatically vanishes and in approximation of
(1.15) the total error decreases.
The matrix of a corresponding SLE has a block-banded shape
and is very sparse (Fig. 1.2).
Such systems are solved more efficiently by iterative methods,
which enable the most economical use of the CPU core memory.
In particular, in (Spichak, 1999) SLE (1.16) was solved using con-
jugate gradient technique BiCGstab (Steijpen et al., 1994) after its
preconditioning by means of the diagonal (Jacobi) scaling.
1.6.4 Calculation of the magnetic field

The magnetic field H can be determined directly from Eq. (1.4) or,
as noted above, it can be simply recalculated from the determined
electrical field by its differentiation in terms of the finite difference
(or finite element) approximation of Eq. (1.10). In doing so, we
have to overcome the difficulty arising in regions with large con-
ductivity gradients, which may cause blunders in the calculation
of the relevant electrical field derivatives and, eventually, incor-
rect values of the magnetic field. Particularly, in the calculation
of the horizontal components Hx and Hy at the Earth’s surface
according to formulas (1.10), the approximation of derivatives
dEx =dz and dEy =dz is not stable.
A possible way to tackle this problem lies in constructing a
spline over a set of values of the grid function and then using
the analytically obtained derivatives to compute the magnetic
field components from Eq. (1.10). This technique gives satisfac-

tory results in two-dimensional case when the solution involves
only one component of the electrical field, but it proves to be un-
stable in three-dimensional case (Pridmore et al., 1981).
Another approach lies in computing the anomalous magnetic
field by means of numerical integration of excess currents circu-
lating in the anomalous region Va :
Z
Ha ¼ G b m DsEdv; (1.17)
Va
where Ds ¼ s sb .
Despite the fact that this procedure is stable, in order to deter-
mine the magnetic field the relevant Greens tensor must be
known. If the electrical field was calculated using one of the inte-
gral equation techniques (see Section 1.2), then computing the
magnetic field would not require any additional calculations of
the Greens tensor. However, if the computation is performed by
means of a differential equation technique (see Section 1.3),
determining the magnetic field by Formula (1.16) would require
computing the Green tensor, which involves considerable compu-
tation expenses.
From this point of view, the way proposed by Spichak (1999)
appears to be more preferable. In this approach, magnetic field
components are determined in two steps. First, in the centers of
the grid cells, the vertical component of the magnetic field Hz is
found, and then, using Hilbert transforms (see, for instance,
(Zhdanov, 1988)), tangential components Hx and Hy are
computed at the grid nodes:
ZZ
1 Hz ðx; y; 0Þðx x0 Þ
Hx ðx0 ; y0 ; 0Þ ¼ Hx þ ð2pÞ
b
dxdy
r3
S
(1.18)
ZZ
Hz ðx; y; 0Þðy y0 Þ
Hy ðx0 ; y0 ; 0Þ ¼ Hyb þ ð2pÞ1 dxdy;
r3
S
where r ¼ ½ðx x0 Þ2 þ ðy y0 Þ2 1=2 ; the value of Hz is precalcu-

lated according to Formula (1.10); Hxb and Hyb are the background
magnetic field components at the Earth’s surface.
The above algorithm for the forward problem solution was
realized as a software package FDM3D (Spichak, 1983b) and
widely used for modeling EM fields in 3-D media (Spichak, 1999,
2015).
1.6.5 Controlling the accuracy of the results

1.6.5.1 Criteria for accuracy
In most cases, the accuracy of modeling EM fields can only be
evaluated indirectly, because the existing control techniques
provide, as a rule, necessary but not sufficient conditions for
the accuracy of the results.
“External” and “internal” techniques are employed for accu-
racy control. External techniques include
(a) comparison with the results of other authors;
(b) comparison with the results obtained by other methods,
including analytical ones.
Thus far, these two criteria have been used mainly for
checking the correctness of solutions. The results of numerous
comparisons, however, clearly show the above accuracy criteria
to be insufficient. Control techniques that enable the accuracy
estimation by an internal means of one or another numerical
approach should be applied.
As pointed out by Poggio and Miller (1973), “the idea of
numerical evaluation of the accuracy of a technique at times ap-
pears to be internally inconsistent, because it seems that numer-
ical results cannot be used to confirm their own correctness. In
reality, there are several approaches that enable a manifest incor-
rectness of numerical results to be detected.” Such criteria include
testing the following:
- the reciprocity theorem;
- the energy conservation law;
- the accuracy to which the equations and boundary conditions
are satisfied;
- convergence of the solution with decreasing size of grid cells,
number of unknowns, etc.
Finally, one may consider as a mixed type (in employing dif-
ferential equation techniques) the criterion proposed by Zhdanov
et al. (1982). It is based on an estimation of the accuracy to which
the integral identity
ZZZ
Eðr0 Þ ¼ Eb ðr0 Þ þ ium0 b e ðr0 jrÞsa EðrÞdy;
G
Va
is satisfied by substituting in it the solution obtained by the finite

difference or finite element techniques. Unfortunately, numerical
implementation of this criterion requires a time-consuming
computation of Green tensor for a layered medium, which re-
duces its practical value.
A more efficient internal criterion can be based on satisfying

the integral relations between components of the anomalous
EM field at the Earth’s surface (Spichak, 1999):
Exa ¼ Gx Ez ium0 GHya ; Eya ¼ Gy Ez þ ium0 GHxa (1.19)
where G is the free-space Green function and Gx and Gy are its de-
rivatives over x and y.
The developed code FDM3D (Spichak, 1983b) based on the al-
gorithm described above was tested on models that allowed an
analytical solution and on models computed by other techniques
in the framework of the COMMEMI project (Zhdanov et al., 1997).
1.6.5.2 Comparison with analytical solution

Indirect assessment of the computation accuracy of the FDM3D
code was carried out by Spichak (1983a) using a high frequency
asymptotic solution for a solid sphere with radius a ¼ 2 km in a
homogeneous space, excited by a plane wave (Berdichevsky and
Zhdanov, 1984). The electrical conductivity of the target is
sT ¼ 104 S/m, and that of the host space is sH ¼ 106 S/m, the
period T ¼ 0.01 s.
With u/N and sT >> sH , for a point located above the solid
sphere at a distance r from its center, we have

Exa ¼ Exb ða=rÞ3 ium0 Hyb a3 = 2r 2 ;
where Exb and Hyb are the background electrical and magnetic
fields.
The study involved numerical modeling for a cube of equiva-
lent volume and 3.2 km on a side. Fig. 1.3 presents the results for a
profile through the center of the cube (model of “a solid sphere”).
One can see that with increasing distance from the center r,
the compatibility of the results increases. At the same time, the
divergence of the curves (10%) at short distances from the center
(r/a ¼ 2) is evidently due to the dissimilar geometries of the
models.
1.6.5.3 Comparison with results obtained by other techniques

Fig. 1.4 depicts the results of computing the horizontal compo-
nent of the magnetic field for the model 3D-1 (T ¼ 0.1s; Hb ¼
ðHx ; 0; 0Þ) from the project COMMEMI (Zhdanov et al., 1997),
which were performed by Spichak (1999) using different
techniques. The magnetic field was determined from the electrical
field by (1) finite difference approximation of the Maxwell
Figure 1.3 Comparison of numerical modeling results for a cube (1) with an
analytical solution for an equivalent solid sphere (2) (Spichak, 1999).
equation; (2) spline interpolation of the electrical field values; (3)

Hilbert transforms (1.19); and (4) integration of the excess cur-
rents according to (1. 17).
A comparison of these curves with the “statistical average”
values, derived for this model from the results obtained by
different authors and using a variety of techniques, shows that
computing the tangential components of the magnetic field by
means of Hilbert transforms yields the best results. In the next
section, we will dwell on the particular case of three-
dimensional models of the medium, where allowing for symme-
try results in a sharp reduction of computations.
1.7 Method of the EM field computation in

axially symmetrical media
Analysis and interpretation of electromagnetic field anomalies in
the Earth, investigation of the resolution of soundings as well as a
number of other methodological problems encountered today in
electromagnetics require calculation of many different models.
These problems can be solved in principle by using the existing
methods of numerical modeling of EM fields in media that
include arbitrary three-dimensional inhomogeneities (see
Sections 1.2e1.4 earlier). However, in many cases, the objective
Figure 1.4 Graphs of the normalized horizontal component of anomalous

magnetic field computed from the electrical field by (1) finite difference
approximation of the Maxwell equation; (2) spline interpolation of the
electrical field values; (3) Hilbert transforms (1.19); (4) integration of the
excess currents according to (1.17) (Spichak, 1999).
can be achieved with reduced computer resources if we confine

ourselves to models of specific type of symmetry. In particular,
axially symmetric three-dimensional models reduce the vector
problem to a series of independent tasks in a plane for two scalar
functions. In this case, the reduction to a discrete system may be
accomplished either by the integral equation method or by one of
the differential equation methods.
Below an algorithm for the forward modeling of quasistation-
ary electromagnetic fields in axially symmetric three-dimensional
media is described following (Zhdanov et al., 1990).
1.7.1 Governing equations

Consider an electromagnetic field excited by a plane vertically
incident wave in a layered medium that includes a three-
dimensional axially symmetric inhomogeneity (Fig. 1.5). The
medium is assumed to be isotropic and nonmagnetic. The
magnetic permeability in the whole space is equal to free-
space permeability (m ¼ m0 ). Displacement currents, as earlier,
are neglected, i.e., the field is assumed to be quasistationary.
The time dependence of the fields is defined by the factor
exp(iut). Introduce a cylindrical coordinate system (r; f; z)
whose vertical axis coincides with the axis of symmetry of the in-
homogeneity and is positive vertically downwards.
Represent the components of the vectors E; H as Fourier
series:
X
þN
ðnÞ
X
þN
ðnÞ
Er;f;z ¼ Er;f;z expðinfÞ; Hr;f;z ¼ Hr;f;z expðinfÞ (1.20)
n ¼ N n ¼ N
Substituting these expansions into Maxwell Eqs. (1.9 and 1.10)

we derive the following equations for harmonics:
ðnÞ
in ðnÞ vHf
H ¼ sErðnÞ ; (1.21)
r z vz
Figure 1.5 Model of the geoelectrical section (in cylindrical coordinates); U is

the domain of modeling.
ðnÞ
vHr vH ðnÞ ðnÞ
z ¼ sEf ; (1.22)
vz vr
ðnÞ
1 ðnÞ vHf in
Hf þ HrðnÞ ¼ sEzðnÞ ; (1.23)
r vr r
ðnÞ
in ðnÞ vEf
E ¼ ium0 HrðnÞ ; (1.24)
r z vz
ðnÞ
vEr vE ðnÞ ðnÞ
z ¼ ium0 Hf ; (1.25)
vz vr
ðnÞ
1 ðnÞ vEf in
Ef þ ErðnÞ ¼ ium0 HzðnÞ : (1.26)
r vr r
ðnÞ ðnÞ
Let u ¼ Ef and v ¼ Hf . Using Eqs. (1.21), (1.23), (1.24), and
ðnÞ ðnÞ ðnÞ ðnÞ
(1.26) express the components Er ; Ez and Hr ; Hz in terms of
u and v:
b r u ium0 r D
ErðnÞ ¼ inD b z v; E ðnÞ ¼ inD
b z u ium0 r D
b r v;
z
b z u þ inD
HrðnÞ ¼ sr D b r v; H ðnÞ ¼ sr D
b r u þ inD
b z v; (1.27)
z
where
br ¼ v 1 bz ¼ r v
vr þ r
D r and D are the components of the
a a vz
b
b ¼ Dr and a ¼ ium0 sr 2 n2 .
vector differential operator D bz
D
Substituting these expressions into (1.22) and (1.25) gives
equations in u and v:

b þ su inrot2 Dv
div2 sr Du b ¼ 0;

b þ ium0 v inrot2 Du
div2 ium0r Dv b ¼ 0; (1.28)
where
v v v v
div2 ¼ er þ ez ; rot2 ¼ er ez ;
vr vz vz vr
er and ez are the unit vectors of the cylindrical coordinate

system.
It is evident that when the field is excited by a plane wave it is
sufficient to solve the problem for harmonics n ¼ 1. If the
normal field is polarized linearly (with the magnetic field in the

f ¼ p=2 azimuth), the harmonics n ¼ 1 and n ¼ 1 are related
by
ð1Þ ð1Þ ð1Þ ð1Þ
Ef ðr; zÞ ¼ Ef ðr; zÞ; Hf ðr; zÞ ¼ Hf ðr; zÞ: (1.29)
Hence, to determine the electromagnetic field components
within an axially symmetric three-dimensional model, it is suffi-
cient to define functions u and v corresponding to value n ¼ 1
and then to determine the required field by Formula (1.20) with
reference to relations (1.27) and (1.29).
1.7.2 Boundary conditions

Without loss of generality, the modeling domain is assumed to be
a rectangle U on the plane (r;z), whose left-hand side lies on the z
axis, the upper side is in the atmosphere, and the lower side is
located in the underlying basement (Fig. 1.5). On the boundaries
of the domain U the following boundary conditions are specified.
(a) On the upper boundary in the atmosphere (and on the lower
boundary of the domain if the underlying basement is highly
resistive), the first-order asymptotic boundary conditions for
an anomalous field are valid:

v v v v
1þr þz ðu u Þ ¼ 0; 1 þ r þ z
b
ðv vb Þ ¼ 0;
vr vz vr vz
where u and v are the azimuthal components of the total field har-
monics, while ub and vb are those of the background field. These
conditions are readily established from the asymptotic boundary
conditions (1.14).
(b) At the interface of a highly conducting underlying basement,
which can be roughly considered as a perfect conductor, the
horizontal components of the electrical field are zero. This
leads to the boundary conditions
vv
u ¼ 0; ¼ 0:
vz
(c) On the axis of symmetry, the exact relations

vu vv
¼ 0; ¼ 0
vr vr
are satisfied.
(d) On the right-hand boundary of the modeling domain, the

total fields are locally approximated by a plane vertically inci-
dent wave. In this case, the boundary conditions are:
vu vv
¼ 0; ¼ 0:
vr vr
1.7.3 Discrete equations and their numerical

solution
In order to derive discrete equations a direct finite element
method (Norrie and de Vries, 1978) is used. In this case, it enables
to set up a conservative scheme around a nine-point pattern.
1.7.3.1 Discrete equations

P
Introduce a grid S on a plane ðr;zÞ : ðri ;rj Þ˛ ;1 < i < I;1 < j < J
Unit cells are rectangles Skl ðk ¼ 1; 2; :::; I; l ¼ 1; 2; :::; JÞ with
their vertices being in the middles of cells of the grid S adjacent
to grid points ðri ; rj Þ (Fig. 1.6).
Integrating Eq. (1.28) over an area of the cells Skl and using the
OstrogradskyeGauss and Stokes two-dimensional formulas, we
obtain the balance equations
Figure 1.6 Unit cell Sk;l of the rectangular mesh.

Z ZZ Z ^
^
sr Du en dl þ sudS in Dv et dl ¼ 0;
vSkl Skl vSkl
Z
ZZ Z ^
^
ium0 r Dv en dl þ ium0 vds in Du et dl ¼ 0; (1.30)
vSkl Skll vSkl
where vSkl is the boundary of the cell Skl , en and et are the unit
vectors directed along an outward pointing normal and a tangent
to the boundary vSkl , respectively; the contour vSkl is transversed
counterclockwise.
We seek u and v as an expansion in terms of finite basis
functions:
I X
X J I X
X J
uðr; zÞ ¼ uij fij ðr; zÞ; vðr; zÞ ¼ vij fij ðr; zÞ; (1.31)
i¼1 j¼1 i¼1 j¼1
where fij ðr; zÞ ¼ 0; if ðr; zÞ;½ri1 ; riþ1 ½zj1 ; zjþ1

Substituting expansions (1.31) into Eq. (1.30) we write:
X
kþ1 lþ1
X ij ij
Akl uij þ Bkl vij ¼ 0; (1.32)
i ¼ k1 j ¼ l1
X
kþ1 lþ1
X ij ij
Bkl uij þ Ckl vij Þ ¼ 0 ðk ¼ 1; 2; :::; I; l ¼ 1; 2; :::; JÞ;
i ¼ k1 j ¼ l1
where
Z ^
ZZ
ij
Akl ¼ sðr; zÞrðDfij Þen Þdl þ sðr; zÞfij dS;
vSkl Skl
Z ^
ij
Bkl ¼ in Dfij Þet dl;
vSkl
2 3
Z Z Z
^
ij 6 7
Ckl ¼ ium0 4 r Dfij en dl þ fij dS5:
vSkl Skl
Definite integrals entering the formulas for the coefficients

ij ij ij
Akl ; Bkl ; and Ckl are evaluated by means of the “rectangle rule”.
1.7.3.2 Basis functions

As it is known, allowance for the field behavior contributes to the
accuracy of equation approximation. Assuming that in each grid
point neighborhood total fields vary linearly in the horizontal and
exponentially in the vertical, it is possible to introduce the
following basis functions:
fij ðr; zÞ ¼ xij ðrÞzij ðzÞ; (1.33)
where
8
>
> 0 r < ri1
>
>
>
> r ri1
>
> ri1 r ri
<r r
i i1
xi ðrÞ ¼ ;
>
> r iþ1 r
>
> ri r riþ1
>
> riþ1 ri
>
>
:
0 riþ1 r
8 0 z zj1
>
>
>
>
>
>
>
> sh kij ðz zj1 Þ
>
>
>
> zj1 z zj
>
>
< sh kij ðzj zj1 Þ
zij ð; cÞ ¼ :
>
>
>
> sh kijþ ðz zjþ1 Þ
>
>
>
> zj z zjþ1
>
> sh kijþ ðzj zjþ1 Þ
>
>
>
>
:
0 zjþ1 z
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
where kijþ ¼ ium0 sþ
ij ; kij ¼ ium0 s þ
ij ; sij and sij are the
average conductivities in the upper and lower halves of a cell,
respectively.
Note, that the derived basis functions show fairly good
approximation properties. In particular, the background fields
calculated for a one-dimensional conducting medium from a sys-
tem of Eq. (1.32) with due account of the relation (1.33) coincides
with that calculated analytically. It is worth mentioning, that in
nonconductive medium (kij /0) the basis function (1.33) is
reduced to bilinear one, which, in turn, corresponds to the
geometry-only dependence of the field.
1.7.3.3 Numerical solution of discrete equations

The system of linear algebraic equations resulting from discreti-
zation is solved by employing the Crout algorithm of expanding
a matrix into the product of the upper and lower triangular
matrices (Tewarson, 1973). Below is a brief outline of the
algorithm.
b N Þ of the system in the form:
Represent the matrix AðN
b¼ d uT
A ;
v G
b is an
where d is a scalar, v is a column vector, uT is a row vector, G
(N-1)th order square matrix. It is easy to see that for ds0, the
following representation
b¼ d 0T 1 0T d uT =d
A , ,
v bI N1 0 G0 0 bI N 1
holds, where the ðN 1Þ th order square matrix G b 0 is defined as

Gb0 ¼ G b n , bI N 1 is an ðN 1Þ the order identity matrix, 0 is a
d
zero column vector. At the next step, the matrix Gb0 is expanded
in the same way, etc. As a result, at N steps, the initial matrix is
expanded into the product of the upper and lower triangular
matrices. Upon expansion, the lower and upper triangular sys-
tems of equations are solved.
The system of linear algebraic equations resulting from
discretization has a banded structure (the bandwidth M ¼ 4 þ
2minðI;JÞ, where I and J are the numbers of grid points in the ver-
tical and in the horizontal, respectively). The application of the
Crout algorithm to this matrix is distinguished by the following
feature. At each step of the algorithm all operators are executed
over the matrix elements lying inside a square LðM MÞ, which
slides diagonally downwards (Fig. 1.7). In doing so, we obtain the
relevant column of the lower triangular matrix and a row of the
upper triangular matrix. This permits matrix expansion by parts,
utilizing a hard disk direct-access file and a small portion of the
core memory.
1.7.4 Code testing

The above algorithm has been used to develop a code for forward
modeling of magnetotelluric fields in the Earth containing an
axially symmetric three-dimensional inhomogeneity. The appro-
priate code FDMS-3D was tested by calculation of the model
response in the DC asymptotic approximation.
b NÞ: N is matrix size, M is a bandwidth.

Figure 1.7 Structure of matrix AðN
The model (Fig. 1.8) consists of a thin layer with thickness h1

and a constant resistivity re1 (integral electrical conductivity of
the layer S1e ¼ h1 =re1 ), an intermediate nonconducting layer
with thickness h2 ðr2 ¼ NÞ, and an ideally conducting underlying
basement ðr3 ¼ 0Þ. The upper layer contains a cylindrical insert
of radius a and a constant resistivity ri1 that matches the upper
layer in thickness (integral electrical conductivity S1i ¼ h1 =ri1 ).
Figure 1.8 Model of a cylindrical insert with resistivity ri1 (Berdichevsky and
Dmitriev, 1976).
For DC asymptotic approximation, an analytical solution is

available for such a model (Berdichevsky and Dmitriev, 1976),
from which the next relations follow:
Ea ¼ Fa Eab ; a ¼ r; f; (1.34)
b are components of the total and background

where Er;f and Er;f
fields, respectively, and
8
>
> S1i S1e
>
> 1 0ra
< S1i þ S1e
Fr ¼ ;
>
> i
e 2
>
> S S a
: 1 þ 1i 1
ra
S1 þ S1e r 2
8
>
> Si S1e
> 1 1i
> 0ra
< S1 þ S1e
Ff ¼ :
>
> i
e 2
>
> S S a
:1 i 1 1
ra
S1 þ S1e r 2
The value of an anomalous magnetic field for the DC asymp-
totic approximation is determined by the formula (Berdichevsky
and Dmitriev, 1976)
1
Ha ez ¼ S1 E S1e Eb ; (1.35)
2
where
S1 ¼ S1i , if 0 r a and S1 ¼ S1e , if r a; ez is a unit vector of
the Cartesian coordinate system.
Calculations were performed for a model with h1 ¼ 0.5 km;
h2 ¼ 10 km; a ¼ 5 km; S1e ¼ 500 S; S1i ¼ 2500 S; s2 ¼ 105 S/m;
s3 ¼ 105 S/m for periods T ¼ 21, 84 and 360 s.
Fig. 1.9 indicates normalized values of the azimuthal compo-
nent of the electrical field for a number of periods, computed
numerically using FDMS-3D and obtained by the asymptotic
Formula (1.34) (T / N). The diagram shows that at T ¼ 84s,
the graph Ef departs from the asymptotic curve by no more
than 2%e3%, merely smoothing out the break at the boundary
of the inclusion. The curve that corresponds to the period
T ¼ 21s differs significantly from the asymptotic one above the
anomaly due to the influence of the induction effect. However,
beginning at r ¼ 5.5 km (r/a ¼ 1.1), the curves converge, and
the discrepancy becomes no greater than 2%e4%.
Figure 1.9 Normalized values of Ef at f ¼ p=2 on the Earths surface for

model of a cylindrical insert shown in Fig. 1.8 (Zhdanov et al., 1990); the solid
curves represent the results calculated by the FDMS-3D program, the dashed
curve designates the results obtained by the asymptotic formula.
Fig. 1.10 shows values of the normalized anomalous magnetic

field Hfa =Hf;anal
a ð0Þ determined by means of the code FDMS-3D
for periods 21, 85 and 360s and using the asymptotic Formula
(1.35) (T /N).
Owing to the inductive influence, the curve Hf at T ¼ 21s dif-
fers considerably from the asymptotic one (by as much as 30% in
the center of the anomaly). As the period increases, the inductive
interaction decreases. Thus, the curve that corresponds to T ¼ 84s
departs from the asymptotic one by as much as 12% and the curve
that corresponds to T ¼ 360s virtually replicates the asymptotic
one, with a difference of 0.2% in the center of the anomaly, of
3% at r ¼ 7 km (r/a ¼ 1.4), and of 8% at the boundary of the anom-
aly (at r ¼ 5 km).
The results of test calculations and practical experience with
the FDMS-3D code demonstrate that the direct finite element
method with special basis functions is an effective means of nu-
merical modeling of quasistationary electromagnetic fields in
three-dimensional media exhibiting an axial symmetry. The
FDMS-3D code does not require appreciable computer resources
and applies equally to purely methodological calculations and to
the solution of a fairly wide range of practical problems.
Figure 1.10 Normalized values of Hfa at f ¼ 0 on the Earths surface for the
model of a cylindrical insert shown in Fig. 1.8 (Zhdanov et al., 1990); the solid
curves represent the results calculated by the FDMS-3D code, the dashed
curve designates the results obtained by the asymptotic formula.
Thus, we considered the 3-D forward modeling algorithms

based on the balance technique. The corresponding computer
codes enable calculations of EM fields in the models with a relief
topography, mixed type of the conductivity structure (1-D/2-D/
3-D), different 1-D-layering at infinities as well as at arbitrary
level in the earth and atmosphere.
1.8 Conclusions
To summarize the advantages and drawbacks of numerical and
physical modeling considered in this chapter, it is worth noting
that possibilities of latter techniques are limited to comparatively
simple conductivity models (though enabling large conductivity
contrasts) of the medium, which, needless to say, reduces the
field of its applicability. On the other hand, numerical modeling
(either finite difference or finite element) allows a more compre-

hensive understanding of EM field morphology than is yielded by
physical modeling methods. In addition, the study of how sensi-
tive the various components of the field are to the variations in
the model parameters, which necessitates multivariant calcula-
tions, can be performed in actual practice only on the basis of nu-
merical modeling. Therefore, by virtue of the above reasons,
mathematical modeling seems to be the most suitable tool for
analyzing 3-D electromagnetic fields in complex geophysical
situations.
As follows from the above discussion of the various approaches,
numerical modeling of 3-D electromagnetic fields in complex me-
dia is a rather painstaking exercise that involves the problems of
constructing an effective algorithm and its numerical implementa-
tion. The latter circumstance is further aggravated by the fact that,
in full-scale numerical modeling, requirements to computer re-
sources exceed the capacities of computers used in actual practice.
Hence, where possible, it is advisable, when stating the problem,
to bear in mind its peculiarity in order to obtain the most effective
solution of the narrow spectrum of issues being considered.
References
Alumbaugh, D.L., Newman, G.A., 1996. Three-dimensional wideband
electromagnetic modeling on massively parallel computers. Radio Sci. 43,
1e23.
Ansari, S., Farquharson, C.G., 2014. 3D finite-element forward modeling of
electromagnetic data using vector and scalar potentials and unstructured
grids. Geophysics 79, E149eE165.
Avdeev, D.B., 2005. Three-dimensional electromagnetic modelling and inversion:
from theory to application. Surv. Geophys. 26, 767e799.
Avdeev, D.B., 2015. 3-D EM forward modeling using integral equations. In:
Spichak, V.V. (Ed.), Electromagnetic Sounding of the Earth’s Interior: Theory,
Modeling, Practice. Elsevier, Amsterdam, pp. 145e159.
Avdeev, D.B., Kuvshinov, A.V., Pankratov, O.V., Newman, G.A., 1997. High-
performance three-dimensional electromagnetic modelling using modified
Neumann series. Wide-band numerical solution and examples. J. Geomagn.
Geoelectr. 49, 1519e1539.
Badea, E.A., Everett, M.E., Newman, G.A., Biro, O., 2001. Finite-element analysis
of controlled-source electromagnetic induction using Coulomb-gauged
potentials. Geophysics 66, 786e799.
Berdichevsky, M.N., Dmitriev, V.I., 1976. Basic principles of interpretation of
magnetotelluric sounding curves in geoelectrical and geothermal studies. In:
Adam, A. (Ed.), Geoelectric and Geothermal Studies. Acad.Kiado, Budapest,
pp. 165e221.
Berdichevsky, M.N., Moroz, I.P., Kobzova, V.M., Bilinsky, A.I., 1987. Physical
Modeling in Geoelectrics. Naukova Dumka, Kiev (in Russian).
Berdichevsky, M.N., Zhdanov, M.S., 1984. Advanced Theory of Deep

Geomagnetic Sounding. Elsevier, Amsterdam-London-NewYork-Tokyo.
Best, M.E., Duncan, P., Jacobs, E.J., Scheen, W.L., 1985. Numerical modeling of
the electromagnetic response of three-dimensional conductors in layered
earth. Geophysics 50, 665e676.
Boerner, R.-U., 2010. Numerical modelling in geo-electromagnetics: advances
and challenges. Surv. Geophys. 31, 225e245.
Boyse, W., Lynch, D., Paulsen, K., Minerbot, G., 1992. Nodal based finite
element modeling Maxwell’s equations. IEEE Trans. Antennas Propag. 40,
642e651.
Brewitt-Taylor, C.R., Johns, P.B., 1980. On the construction and numerical
solution of transmission-line and lumped network models of Maxwell’s
equations. Int. J. Numer. Methods Eng. 15 (2), 13e30.
Brewitt-Teylor, C.R., Weaver, J.T., 1976. On the finite-difference solution of
two-dimensional induction problems. Geophys. J. R. Astron. Soc. 47 (2),
375e396.
Cai, H., Hu, X., Li, J., Endo, M., Xiong, B., 2017a. Parallelized 3D CSEM modeling
using edge-based finite element with total field formulation and
unstructured mesh. Comput. Geosci. 99, 125e134.
Cai, H., Hu, X., Xiong, B., Auken, E., Han, M., Li, J., 2017b. Finite element time
domain modeling of controlled-source electromagnetic data with a hybrid
boundary condition. J. Appl. Geophys. 145, 133e143.
Cai, H., Xiong, B., Han, M., Zhdanov, M., 2014. 3D controlled-source
electromagnetic modeling in anisotropic medium using edge-based finite
element method. Comput. Geosci. 73, 164e176.
Cao, H., Wang, K., Wang, T., Hua, B., 2018. Three-dimensional magnetotelluric
axial anisotropic forward modeling and inversion. J. Appl. Geophys. https://
doi.org/10.1016/j.jappgeo.2018.04.015.
Cherevatova, M., Egbert, G.D., Smirnov, M., Kelbert, A., May 7e9, 2013. 3D
electromagnetic modeling using multi-resolution approach. In: Expanded
Abstract, Proceedings 5th Intern. Symp. on 3-D Electromagnetics. Sapporo,
Japan.
Das, U.C., Verma, S.K., 1981. Numerical considerations on computing the EM
response of three-dimensional inhomogeneities inalayered earth. Geophys. J.
R. Astron. Soc. 66 (3), 733e740.
Davydycheva, S., Druskin, V., 1999. Staggered grid for Maxwell’s equations in 3-D
anisotropic media. In: Spies, B., Oristaglio, M. (Eds.), Three Dimensional
Electromagnetics, SEG Monograph. GD7, Tulsa, USA, pp. 138e145.
Dmitriev, V.I., 1969. Elektromagnitnye Polya v Neodnorodnyh Sredah
(Electromagneic Fields in Inhomogeneous Media). Moscow State University,
Moscow (in Russian).
Dmitriev, V.I., Farzan, R.H., 1980. Method for calculating an anomalous
electromagnetic field from a local inhomogeneity. In: Adam, A. (Ed.),
Mathematical Models of Electromagnetic Sounding and Mapping in
Geophysics. Lorand Eotvos Univ, Budapest, pp. 95e111.
Dmitriev, V.I., Zakharov, E.V., 1970. Method of the solution of the problems of
the electrodynamics of inhomogeneous media. J. Comput. Math. Math. Phys.
10 (6), 1458e1564.
Dosso, H.W., 1966. A plane-wave analogue model for studying electromagnetic
variations. Can. J. Phys. 44 (1), 67e80.
Druskin, V.L., Knizhnerman, L.A., 1994. Spectral approach to solving three-
dimensional Maxwell’s equations in the time and frequency domain. Radio
Sci. 29, 937e953.
Druskin, V.L., Knizhnerman, L.A., Lee, P., 1999. New spectral Lanczos
decomposition method for induction modeling in arbitrary 3D geometry.
Geophysics 64, 701e706.

Endo, M., Cuma, M., Zhdanov, M.S., 2009. Multiple domain integral equation
method for 3D electromagnetic modeling in complex geoelectrical
structures. In: Proceedings SEG 79th Annual Meeting, Expanded Abstracts,
pp. 679e683.
Farquharson, C.G., Duckworth, K., Oldenburg, D.W., 2006. Comparison of
integral equation and physical scale modeling of the electromagnetic
responses of models with large conductivity contrasts. Geophysics 71 (4),
G169eG177.
Farquharson, C.G., Miensopust, M.P., 2011. Three-dimensional finite-element
modelling of magnetotelluric data with a divergence correction. J. Appl.
Geophys. 75, 699e710.
Farquharson, C., Oldenburg, D., 1999. An integral equation solution to the
geophysical electromagnetic forward-modeling problem. Salt Lake City, USA.
In: Wannamaker, P., Zhdanov, M. (Eds.), Proceedings 2nd Int. Symp. on
Three Dimensional Electromagnetics, pp. 8e11.
Fomenko, E., 1999. MT and control source modeling algorithms for 3-D media
-with topography and large resistivity contrasts. USA,. In: Wannamaker, P.,
Zhdanov, M. (Eds.), Proceedings 2nd Int. Symp. on Three Dimensional
Electromagnetics Salt Lake City, pp. 21e24.
Fomenko, E.Y., Mogi, T., 2002. A new computation method for a staggered grid
of 3D EM field conservative modeling. Earth Planets Space 54, 499e509.
Grayver, A., Streich, R., 2012. Comparison of iterative and direct solvers for 3D
CSEM modeling. In: Expanded Abstr. 82th Annual International Meeting.
SEG, pp. 1e6.
Gupta, P.K., Bennett, L.A., Raiche, A.P., 1987. Hybrid calculations of the three-
dimensional electromagnetic response of buried conductors. Geophysics 52,
301e306.
Habashy, T.M., Groom, R.W., Spies, B.R., 1993. Beyond the Born and Rytov
approximations. A nonlinear approach to electromagnetic scattering.
Geophys. Res. 98 (B2), 1759e1775.
Haber, E., 1999. Modeling of 3D EM using potentials and mixed finite elements.
USA. In: Wannamaker, P., Zhdanov, M. (Eds.), Expanded Abstr. 2nd Int.
Symp. on Three Dimensional Electromagnetics Salt Lake City, pp. 12e15.
Hibbs, R.D., Jones, F.W., 1978. Electromagnetic induction in three-dimensional
structures for various source fields. Geomagn. and Geoel 30 (1), 1e18.
Hohmann, G.W., 1975. Three-dimensional induced polarization and
electromagnetic modeling. Geophysics 40 (2), 309e324.
Hohmann, G.W., 1983. Three-dimensional EM modeling. Geophys. Surv. 6, 27e53.
Hursan, G., Zhdanov, M.S., 2002. Contraction integral equation method in three-
dimensional electromagnetic modeling. Radio Sci. 37, 1089e2002.
Hvozdara, M., 1981. Electromagnetic induction of a three-dimensional
conductivity inhomogeneity in the two-layered Earth. Pt 1. Theory. Stud.
Geophys. Geod. 25 (2), 167e180.
Hvozdara, M., 1985. Solution of the stationary approximation for MT fields in a
layered Earth with 3D and 2D inhomogeneities. Geophysics 55, 214e221.
Hvozdara, M., Kaikkonen, P., Varentsov, I.M., 1987. Algorithm for solving 3D
problems of EM induction by means of a vector integral equation. Stud.
Geophys. Geod. 31, 369e385.
Jahandari, H., Ansari, S.M., Farquharson, C.G., 2017. Comparison between
staggered grid finiteevolume and edgeebased finiteeelement modelling of
geophysical electromagnetic data on unstructured grids. J. Appl. Geophys.

138, 185e197.
Jahandari, H., Farquharson, C.G., 2014. A finite-volume solution to the
geophysical electromagnetic forward problem using unstructured grids.
Geophysics 79, E287eE302.
Jahandari, H., Farquharson, C.G., 2015. Finite-volume modelling of geophysical
electromagnetic data on unstructured grids using potentials. Geophys. J. Int.
202, 1859e1876.
Jin, J., 2002. Finite Element Method in Electromagnetics, second ed. Wiley-IEEE
Press, New York.
Jin, J., Zunoubi, M., Donepudi, K.C., Chew, W.C., 1999. Frequency-domain and
time domain finite-element solution of Maxwell’s equations using spectral
Lanczos decomposition method. Comput. Methods Appl. Mech. Eng. 169,
279e296.
Jones, F.W., 1974. The perturbation of geomagnetic fields by two-dimensional
and three-dimensional conductivity inhomogeneities. Pure Appl. Geophys.
112 (5), 793e800.
Jones, F.W., Lokken, J.E., 1975. Irregular coast line and channeling effects in
three-dimensional geomagnetic perturbation models. Phys. Earth Planet.
Inter. 10 (2), 140e150.
Jones, F.W., Pascoe, L.J., 1972. The perturbation of alternating geomagnetic fields
by three-dimensional conductivity inhomogeneities. Geophys. J. R. Astron.
Soc. 27 (2), 479e485.
Kordy, M., Maris, V., Wannamaker, P., Cherkaev, E., May 7e9, 2013. 3D edge
finite element solution for scattered electric field using a direct solver
parallelized on an SMP workstation. In: Expanded Abstr. 5th Intern. Symp.
on 3-D Electromagnetics. Sapporo, Japan.
Kron, G., 1972. Study of Complex Systems by Parts-Diacoptics (in Russian).
Nauka, Moscow.
Kruglyakov, M., Geraskin, A., Kuvshinov, A., 2016. Novel accurate and scalable
3-D MT forward solver based on a contracting integral equation method.
Computers and Geosciences 96, 208e217.
La Brecque, D.J., 1999. Finite-difference modeling of 3-D EM fields with scalar
and vector potentials. In: Spies, B., Oristaglio, M. (Eds.), Three Dimensional
Electromagnetics, SEG Monograph. GD7, Tulsa, USA, pp. 146e160.
Lam, Y.L., Jones, F.W., Hibbs, R.D., 1982. The response of perturbation and
induction arrows to a three-dimensional buried anomaly. Geophysics 47, 51e59.
Lee, K.H., Pridmore, D.F., Morrison, Y.F., 1981. A hybrid three-dimensional
electromagnetic modelling scheme. Geophysics 46 (5), 769e805.
Lee, S., Kim, H., Song, Y., Morrison, H., Lee, K., 1999. A new formulation of
magnetic field integral equation for 3-D EM modeling. Salt Lake City, USA.
In: Wannamaker, P., Zhdanov, M. (Eds.), Expanded Abstr. 2nd Int. Symp. on
Three Dimensional Electromagnetics, pp. 41e44.
Lines, L.R., Jones, F.W., 1973. The perturbation of alternating geomagnetic fields
by three-dimensional island structures. Geophys. J. R. Astron. Soc. 32 (2),
133e154.
Liu, E.H., Lamontagne, Y., 1999. Electromagnetic modeling with surface integral
equations. In: Spies, B., Oristaglio, M. (Eds.), Three Dimensional
Electromagnetics. SEG Publ., GD7, Tulsa, USA, pp. 76e89.
Liu, S., Lang, C., Yang, H., Wang, W., 2018a. A developed nearly analytic discrete
method for forward modeling in the frequency domain. J. Appl. Geophys.
149, 25e34.
Liu, Y., Xu, Z., Li, Y., 2018b. Adaptive finite element modelling of three-
dimensional magnetotelluric fields in general anisotropic media. J. Appl.
Geophys. https://doi.org/10.1016/j.jappgeo.2018.01.012.
Livelybrooks, D., 1993. Program 3D FEEM: a multidimensional electromagnetic
finite element model. Geophys. J. Int. 114, 443e458.
Mackie, R.L., Madden, T.R., 1993. Conjugate gradient relaxation solutions for
three-dimensional magnetotelluric modeling. Geophysics 58, 1052e1057.
Mackie, R.L., Madden, T.R., Wannamaker, P.E., 1993. Three-dimensional
magnetotelluric modeling using difference equations. Theory and
comparisons to integral equation solutions. Geophysics 58, 215e226.
Mackie, R.L., Smith, J.T., Madden, T.R., 1994. Three-dimensional electromagnetic
modeling using finite difference equations: the magnetotelluric example.
Radio Sci. 29, 923e935.
Mitsuhata, Y., Uchida, T., 2004. 3D magnetotelluric modeling using the TeU
finite element method. Geophysics 69, 108e119.
Mittet, R., 2010. High-order finite-difference simulations of marine CSEM
surveys using a correspondence principle for wave and diffusion fields.
Geophysics 75, F33eF50.
Mogi, T., 1996. Three-dimensional modeling of magnetotelluric data using finite
element method. J. Appl. Geophys. 35, 185e189.
Mukherjee, S., Everett, M.E., 2011. 3D controlled-source electromagnetic edge-
based finite element modeling of conductive and permeable heterogeneities.
Geophysics 76, F215eF226.
Nedelec, J.-C., 1986. A new family of mixed finite elements in R3. Numer. Math.
50, 57e81.
Newman, G.A., 2014. A review of high-performance computational strategies for
modeling and imaging of electromagnetic induction data. Surv. Geophys. 35,
85e100.
Newman, G.A., Alumbaugh, D.L., 1999. 3-D electromagnetic modeling and
inversion on massively parallel computers. In: Spies, B., Oristaglio, M. (Eds.),
Three Dimensional Electromagnetics, SEG Monograph. GD7, Tulsa, USA,
pp. 299e321.
Newman, G.A., Alumbaugh, D.L., 2002. Three-dimensional induction logging
problems, Part 2: a finite-difference solution. Geophysics 67 (2), 484e491.
Norrie, D.H., de Vries, G., 1978. An Introduction to Finite-Element Analysis.
Academic Press, New York.
Oshiro, F.K., Mitzner, K.M., 1967. Digital computer solution of three-dimensional
scattering problems. In: Mittra, R. (Ed.), Proceedings IEEE Intern. Antennas
and Propagation Symp. Mich., Ann Arbor, pp. 257e258.
Pankratov, P.V., Avdeev, D.B., Kuvshinov, A.V., 1995. Scattering of
electromagnetic field in inhomogeneous earth: forward problem solution.
Izvestya, Fizika Zemli 3, 17e25.
Persova, M.G., Soloveichik, Y.G., Tokareva, M.G., Domnikov, P.A., 2013. Finite
element 3D modeling of geoelectromagnetic fields for controlled and natural
sources. In: Expanded Abstr. 5th Intern. Symp. on 3-D Electromagnetics.
Sapporo, Japan.
Poggio, A.J., Miller, F.L., 1973. Integral Equation Solutions of Three-Dimensional
Scattering Problems. In: Mittra, R. (Ed.), Computer Techniques for
Electromagnetics. Pergamon Press, Inc., New York, pp. 159e265.
Portniaguine, O., Hursan, G., Zhdanov, M., 1999. Compression in 3-D EM
integral equation modeling. Salt Lake City, USA. In: Wannamaker, P.,
Zhdanov, M. (Eds.), Proceedings of the 2nd Int. Symp. on Three Dimensional
Electromagnetics, pp. 209e212.
Another random document with
no related content on Scribd:
A question now arises of greater moment both for our present
purpose and for the wider interests of Comparative Religion. Was the
purport and significance of the sacrificial act the same in the Western
society as that which is revealed in the sacred literature of
Babylonia? No part of the ancient religious system has been the
theme of so much study and speculation in recent years as the
ancient sacrifice. Robertson Smith in his epoch-making book, The
Religion of the Semites, was the pioneer of a new theory; which has
since been developed or modified by certain English and a few
Continental scholars following on his track. The result of these
labours has been to formulate and define various forms of sacrifice
that prevailed in the Mediterranean area. Three main types appear to
emerge: (a) the gift sacrifice, where an oblation is given over entirely
to the deity, whether generally to win his favour, or in special
circumstances—for instance, after sin has been committed—to
appease his wrath, or as a thank-offering for favour received; (b) the
communion sacrifice, where the community or the individual eat with
the deity, strengthening their feeling of fellowship by a common
meal; (c) the sacramental type, where the community or the
individual may be said to “eat the god,” that is, to partake of food or
drink made sacred by the infusion of the divine spirit or personality,
which is thus communicated to the partaker. It is best for the present
to regard these three as separate and independent, without trying to
determine which is prior and which posterior.235.1
The first type, which is almost ubiquitous in the human societies
that have arrived at the belief in personal deities, is sufficiently
attested by Homer of the early Greeks, who promise and perform the
sacrifice partly as an offering to please or to appease the deity. What
is more important is the evidence, which I have dealt with
elsewhere,236.1 that Homeric society was familiar also with the more
genial conception of the sacrifice as a communion-meal where the
worshipper and the deity meet around the altar; this emerges clearly
in the accounts that Homer has given us of an Achaean sacrificial
feast.236.2 Even the germs are already visible of the idea from which
the third or more mystic type of sacrifice, what I have called the
sacramental type, might be evolved; for special significance attaches
to the acts described in the phrases οὐλοχύτας προβάλοντο and
σπλάγχν᾽ ἐπάσαντο, “they threw forward the barley-shreddings” and
“they tasted the entrails”: the first phrase is not wholly clear, but it
may signify that stalks of barley are first placed on the altar, and
thereby consecrated or filled with the divine virtue that is inherent
there, and then the beast is touched with these on the forehead and
thereby becomes himself filled with the spirit of godhead;236.3 the
second is also a mystic act, for the σπλάγχνα specially contain the
life, which is now infused with divinity, and by tasting them the
worshippers partake of the divine life. All this arises solely from the
extraordinary degree of supernatural force or “Mana” which the altar
itself possesses, a force which may have been an inheritance from
long ages of pillar-worship, if we believe the altar to have been
evolved from the sacred pillar.237.1 It explains other details in the old
Hellenic sacrificial act; such as the casting the hairs of the victim into
the altar-flame,237.2 which established a communion between the
animal and the deity, the practice of solemnly consecrating the lustral
water by ceremoniously carrying it round the altar,237.3 and charging
it with a still more potent infusion of divinity by plunging into it a
lighted brand from the altar-fire.237.4
The communion sacrifice must then have been in vogue in pre-
Homeric times; and the idea that gave it its meaning never wholly
faded from the State-ritual; for the rule, expressed by such formulae
as οὐκ ἀποφορά, δαινύσθων αὐτοῦ, bidding the worshipper
conclude the feast round the altar and take none of the flesh home,
seems to arise partly from the feeling that the ceremony was
meaningless unless he feasted wholly with the deity.237.5 But it was
most vividly realised in the religious services of the θίασοι, the later
religious orders or fraternities devoted each to the cult of its special
divinity; for these a common religious meal formed the chief binding-
tie.237.6
Apart from the Homeric evidence, we have the record of the Attic
Bouphonia as attestation of the great antiquity of this type of sacrifice
in Greece. To the actual statement of the details given us by
Theophrastos and Pausanias, much is added by the curious
aetiological legends that grew up around it.238.1 We see the ox
marking himself for sacrifice by voluntarily going up to the altar and
eating the corn upon it, being thus called, as it were, by the god into
communion with himself. As he is thus full of the spirit of the god, it is
regarded a sacrilegious act to slay him; but all the citizens partake of
his flesh, and even the stranger who eats becomes himself a citizen,
as through this feast he enters into kinship with Zeus Polieus. All this
can be explained by the belief that Zeus Polieus is in the altar; and
we need not resort to such theories as that the ox is a totem-animal
or the spirit of vegetation.
We must, however, beware of concluding that because the victim
was thus temporarily possessed with godhead and in this holy state
devoured, he was therefore literally regarded as the full incarnation
of the god, or that the worshipper consciously believed he was
eating his own deity who died in the sacrifice. For religious
consciousness by no means always draws the full logical corollaries
of a religious act. The more mystic idea, that has played a great part
in the religion of Europe, can only be detected or suspected, apart
from written direct ancient testimony, where the animal is treated
with reverence apart from and before his association with the altar,
or is regarded as the habitual incarnation of the deity. The
immolation and devouring of such a victim would be of the true
sacramental type, which Robertson Smith believed was the
aboriginal form of all sacrifice. But we have no clear example of it
from the earliest period of genuine Hellenic religion, unless we force
the evidence or exaggerate its meaning.239.1 We have only certain
myths that we may doubtfully venture to interpret by means of this
hypothesis. And these are no myths about animals, but about human
victims devoured in sacrifice: the most significant is the story of the
cannibalistic feast held by King Lykaon, who cooked his own son
and offered the flesh to Zeus, a ritual of which some survival,
whether mimetic or half-real, was witnessed by Pausanias himself on
Mount Lykaon.239.2 This would point to sacramental cannibalism, if
we assume that King Lykaon was the human incarnation of Zeus
Lykaios and that his son was therefore a divine infant. But it is
possible that the story enshrines the remembrance of the more
ordinary clan-sacrifice of the life of a clansman to procure them
communion with the clan-god by the common partaking of his flesh:
the kinsman is offered rather than the animal, not so much that the
sacrificers may eat their god, but that the god by consuming their
most valued life may be more closely incorporate with them.
Again, in one of the darkest and most perplexing of Greek
legends, the story of Klumenos of Argos and his incestuous love of
his daughter Harpalyke, who revenges herself by slaying her own
child and offering it to the father in a sacrificial meal, we may discern
the glimmer of a remembrance of a cannibal sacrament. The
associations of the story dimly indicate a Thracian origin.239.3 And it
is in the range of the Greek Dionysiac cult, which according to the
most probable view was adopted from Thrace, that we find imprinted
on legend and ritual the tradition of a savage type of sacrament, in
which the human or animal incarnation of the god was devoured.
Such is the significance that we may fairly attach to the Titan-story of
the murder of the infant Dionysos, to the σπαραγμός of the goat or
bull or snake periodically practised by the Bacchoi or Bacchae, and
to the death of Pentheus.240.1 And in later Greek ritual the
consciousness here and there survived that the victim offered to
Dionysos incarnated the very deity, even before it acquired the
temporary mystic afflatus from contact with the altar. The record of
the ritual of Tenedos, in which a sacred pregnant cow was tended
reverently and the calf that she bore was dressed in the buskins and
then sacrificed to Dionysos, “the render of men,” is the most piquant
example.240.2
This Dionysiac tradition reaches back undoubtedly to the second
millennium in Greece; the evidence of a similar sacramental ritual in
purely Hellenic worship is shadowy and slight, for the critical
examination of the Eleusinian mysteries does not clearly reveal it;
and the growth and diffusion of the idea in later Paganism does not
concern us now.
But the consideration of the early Hellenic sacrifice, of which the
salient features have been slightly sketched above, is of signal value
for our present purpose. For it reveals at once a marked contrast to
Babylonian ideas so far as these at present are revealed to us. The
Babylonian-Assyrian liturgies, epics, and chronicles have failed to
disclose any other theory of the sacrifice than that which is called the
gift-theory. A general term for the Babylonian sacrifice is “kishtu” or
“present.”241.1 The deities are supposed to eat what is given them; in
the Epic of the Deluge the naïve phrase occurs, “The gods smell the
savour, the delightful savour, the gods swarm like flies around the
sacrificer.” No evidence is as yet forthcoming that the sacrificer was
supposed or was allowed to eat with the deity, as in the Hellenic
communion-sacrifice. On the contrary, certain texts can be quoted
which seem expressly to forbid such a thing.241.2 The document
already cited that was found in Assurbanipal’s library, containing the
Job-like lament of the good man who had found no profit in
goodness, contains a verse in which he compares himself with the
sinner who neglects all religious ordinances, and who has even
“eaten the food of God.” And so we find that among the various evil
or impure or unlucky deeds that could bring a man under the ban of
the gods, the “devouring of sacrificial flesh” is expressly
mentioned.241.3
Unless, then, documents yet to be revealed contradict this positive
and negative evidence, we have here a fact of great weight to set
against the theory that we are discussing. Whencesoever the
Hellenes derived their genial conception of the sacrifice as a
communion-meal, they did not derive it from Babylon. And all
Robertson Smith’s speculations concerning the inner significance of
the Semitic sacrifice cannot yet be applied to Mesopotamia, whence
he was not able to glean any evidence. In Babylonia the sacrificer
got no share of the victim. He might eat with the spirits of the dead in
certain ritual, but he was not, it appears, privileged to eat with the
god or goddess. The deity took the victim, or the sacrifice of cereals
and fruits, as a present, and the priest got his share. But we are not
told that the priest ate with the god, or where he ate; nor can we say
that the priest represented the worshipper.
If a true sacrament is yet to be discovered in Babylonian religion, it
will probably be found in some document of the Tammuz ritual. For it
is probable that Tammuz was identified with the corn as with other
parts of vegetation, and that the mourning for him was accompanied
with abstinence from bread. His resurrection ended the fast, and if in
their joy the worshippers ceremoniously broke bread, they may have
supposed they were eating the body of their risen lord. But such a
reconstruction of the old Tammuz ritual rests at present only on the
indirect evidence of the later records of Attis-Adonis cult and of the
Tammuz-worship among the heathen Syrians of Harran in the tenth
century of our era.242.1
It belongs to the Babylonian conception of the sacrifice as a gift,
that the animal was often regarded as a ransom for the man’s own
life; that is to say, when sin had been committed, the deity might be
placated by the gift of an agreeable victim, and be persuaded to
accept it in place of the sinner whose life was properly forfeit. For
instance, a sick man is always supposed to have sinned; and the
priest who is performing an animal sacrifice in his behalf uses the
prayer, “Take his present, take his ransom”;242.2 and the formula of a
sacrifice offered by way of exorcism is very explicit: “A male sheep, a
female sheep, a living sheep, a dead sheep shall die, but I shall
live.”243.1 Another inscription243.2 throws further light on the ritual of
the vicarious sacrifice: “To the high-priest may one cry out: the kid of
substitution for the man, the kid for his soul he hath brought: place
the head of the kid to the head of the man, place the neck of the kid
to the neck of the man, place the breast of the kid to the breast of the
man.” Whether this solemn manipulation was performed after or
before the sacrifice, its object must have been to establish by contact
a communion between the man and the victim, so that the kid might
be his most efficacious representative. It thus became part of the
higher ethical teaching of Babylon that “sacrifice brings life,” just as
“prayer takes away sin,” a doctrine expressed in a fragmentary tablet
that contains a text of striking spiritual import.243.3
The type of sacrifice that may be called vicarious must have been
an ancient as well as a later tradition in Greece; for the legends
associated with many sacrifices clearly attest it, explaining certain
animal victims as substitutes for a human life that was formerly
demanded by the offended deity, the vicarious sacrifice usually
carrying with it the ideas of sin and atonement.243.4 The substitution
might occasionally be apologised for by a legal fiction, as, for
instance, when in the ritual of the Brauronian Artemis the angry
goddess demanded the life of a maiden, and the Athenian parent
sacrificed a goat, “calling it his daughter.”244.1
But though this idea is common to the Mesopotamian and Hellenic
communities, they differ widely in respect of the evidence they afford,
of the prevalence of human sacrifice. As regards ancient Greece, the
evidence is indubitable, though much that has been brought by
modern scholars is due to false interpretation of ritual, such as the
scourging of the Spartan boys; later, the human sacrifice became
repugnant to the advancing ethical thought of the nation, but
according to one authority did not wholly die out till the age of
Hadrian. On the other hand, no literary text nor any monument has
yet been found that proves the existence of such a ritual in
Babylonia. In one of his biographical inscriptions Assurbanipal
proclaims that he “sacrificed” prisoners of war to avenge his
murdered father on the spot where his father was slain.244.2 He
boasts of worse things than this; and we can well believe that he
murdered them in cold blood. But the words by no means suggest a
ceremonious tomb-ritual with human sacrifice.
Slightly more important is a passage in a legal document, to which
Mr. Johns245.1 calls attention; whence it appears that a forfeit for
reopening an already adjudicated cause was the consecration of
one’s eldest child by fire to a god or goddess; and, as incense and
cedar-wood are mentioned in the same context as concomitants of
the threatened ceremony, the conclusion seems natural that this was
once at least a real threat of human sacrifice inflicted as a legal
punishment. This legal clause gives us the right to conclude that in
the earliest period the Semites and Sumerians of Mesopotamia
occasionally resorted to this rite. They would be indeed a peculiar
people and a favoured nation if they had always been innocent of it.
It is sufficiently attested by direct evidence, either of record or
excavation, or by the suggestion of legend, of the Arabs, Syrians, the
early Canaanites,245.2 the Israelites, the Phoenicians; also of the
Phrygians and other non-Semitic peoples of Anatolia. Yet it must be
put to the credit of the Babylonian culture of the second millennium,
that the Mesopotamians had either completely or almost abandoned
it. At this time it was doubtless in full vogue in Greece; and certainly
Babylon could not have been their evil teacher in this matter. But
they needed no teachers in what was an ancient tradition of their
northern ancestors, and of the people with whom they mingled. Yet
only twenty years ago a distinguished writer on Greek ritual could
say, “It is certain that the Hellenes borrowed the practice of human
sacrifice from the Orientals.”246.1
As we discover no trace of the idea of communion in the ordinary
Babylonian sacrifice, we are the less surprised that scarcely any hint
is given by the sacred literature or monuments of any mystic
application of the blood of the victim, which was used for so many
purposes of communion-ritual by the early Hellenes and Hebrews. I
can find no other evidence for this in Mesopotamia except one
passage quoted by Zimmern,246.2 in which the sacrificer is ordered to
sprinkle some part of the door with the blood of the lamb. It is not
probable that the Babylonians were incapable of the notion that by
physical contact with certain sacred objects a temporary communion
could be established between the mortal and the divinity: it appears,
for instance, in one of the formulae of the purification-ritual—“May
the torch of the Fire-god cleanse me”246.3—in the yearly practice of
the king grasping the hands of the idol of the god, perhaps in the
custom of attaching the worshipper to the deity with a cord,246.4 and
in the diviner’s habit of grasping the cedar-staff, which is called “the
beloved of the gods.”246.5 But it may be that they never applied this
notion to the sacrifice, so as to evolve the institution of the
communion-meal; or they may have evolved it in early times, and
through long ages of power the priests may have become strong
enough to suppress it and to substitute for it the gift-ritual, which
would be more profitable for themselves. It accords with this
absence of any mystic significance in the sacrifice that we do not find
in the Babylonian service any mystic use of the sacrificial skins, of
which some evidence can be gathered from various details of Greek
ritual.247.1
Again, the Babylonian records have so far failed to reveal any
evidence for any such public ceremony as the sending forth of the
scapegoat, whether human or animal, charged with the sins of the
community. This ritual was common to the Hellenes, Egyptians, and
Hebrews, and probably to other Semitic communities.247.2 The idea
of sin-transference on which it rests was familiar enough to the
ancient Babylonian; but he seems only to have built upon this a
private system of exorcism and purgation of sin and disease for the
individual. As far as we know, it did not occur to him to effect by this
method a solemn annual expulsion of all the sins of the nation.
On the other hand, there is another type of sacrifice common to
Babylonia and Greece, by which an oath or an engagement might be
cemented: the animal is slaughtered with an imprecation that the
same fate may befall him who breaks his oath or violates the
compact. Zimmern quotes a good example of this, relating to the
compact made between the Assyrian king Assurnirari and Mati’ilu,
prince of Arpad:248.1 a sheep is sacrificed and the formula
pronounced: “This head is not the head of a sheep, it is the head of
Mati’ilu, of his sons, of his great ones, of the people of his land. If
Mati’ilu breaks this oath, the head of Mati’ilu will be cut off, like the
head of this sheep.” The same idea underlies the oath-sacrifice in
the Iliad,248.2 though it is not expressed with such naïve make-
believe or such logic in the detail; but as the beast is slaughtered or
wine poured, a curse is uttered invoking on the perjured a similar
fate, or with a prayer that “his brains may be poured out like this
wine.” The original idea is magical: the symbolic explanation is later.
Another parallel is the Latin oath over the stone.248.3
Such resemblance in special forms by no means weakens the
impression that we receive from the striking differences discernible in
the Babylonian and Hellenic significance of sacrifice. To those
already noted we may add yet another, which concerns the
association of sacrifice with divination. It is Professor Jastrow’s
opinion248.4 that the chief motive of the Babylonian sacrifice was the
inspection of the liver of the victim, from the markings of which the
skilled expert could interpret the future by a conventional system
revealed to us in certain ancient Babylonian documents. This
superstition is so elaborate and artificial that if we find it in adjacent
countries, it is more reasonable to suppose that one borrowed it from
the other, than that it was developed independently in each. We find
it in later Greece, Etruria, and Rome; but the evidence of the
Homeric poems suggests that it was unknown to the Hellenes of the
earlier period. They are very likely to have borrowed it from
Babylonian sources in post-Homeric times; and we note here, as in
other cases where the influence of Babylon upon Greece can be
reasonably posited, it reaches the western shores of the Aegean at a
post-Homeric rather than a pre-Homeric epoch.
The comparative study of Mesopotamian and Hellenic sacrifice
confronts us finally with another problem belonging to the history of
religions, and one of the greatest, the dogma of the death of the
divinity and the origin and significance of that belief. For where the
mystic type of communion-sacrifice is found, where the animal that is
slain for the sacrament is regarded as the incarnation of the deity,
the divinity may be supposed himself to die temporarily, doubtless to
rise again to life, either immediately or at some subsequent festival.
This momentous conclusion need not always have been drawn, for
religious logic is not often persistently thorough, nor does the
evolution of the idea belong necessarily to the sphere of totemism,
as Robertson Smith supposed, and M. Reinach is still inclined to
maintain. It is not my concern here to discuss the totemistic
hypothesis, but I may point out that in the rare examples where the
totem-animal is slain, it is not clear that it is slain as a divinity.
Again, the belief in the periodic death of a deity might arise
independently of the sacrifice, namely, from the essential idea of the
godhead itself, when the divine life is identified with the annual life of
vegetation: the phenomena of nature in autumn and spring may
suggest to the worshipper the annual death and resurrection of the
god or goddess. It is important to note that in this, as in the other
source of the belief, the conclusion need not always have been
drawn, for the vegetation-deity might be supposed not to die in
autumn or winter, but merely to disappear, and the story of his or her
disappearance need not carry the same religious consequences as
the story of the divine death.
The immolation of the divine victim in a communion-service,
wherein the worshippers partake symbolically or realistically of the
divine flesh and blood, though suggested by a thought that we must
call savage, may be pregnant with consequences momentous for
higher religion, as the history of Christian dogma attests. And even
the annual death of the nature-god may be raised to a higher
significance than its mere nature-meaning, and may be linked with
the promise of human immortality.
We may note, finally, that a religion which expresses in its ritual
the idea of the deity’s death and resurrection is likely to be charged
with a stronger emotional force than one that lacks it; for the two
events will excite an ecstasy of sorrow and of joy in the believer.
As the phenomenon, then, is of such importance, it is necessary to
be critical and unbiased in the collection of statistics. Our present
field of inquiry is the Eastern and Western Mediterranean area; and
here our conspicuous example is the ancient Sumerian-Babylonian
ritual of Tammuz,250.1 a folk-service of lamentation and rapture,
psychologically akin in many respects to Christianity, and of most
powerful appeal.
The Tammuz hymns preserved to us are of the highest Babylonian
poetry, and though they are chiefly litanies of lamentation, sorrowing
over the death of the young god, yet one or two echoes are heard at
their close of the rapturous rejoicing over his resurrection.251.1 With
them is associated the story of the descent of his consort Ishtar, or of
his goddess-sister; another great motive of the religious imagination
which neighbouring peoples and faiths were quick to capture and
adapt to their own religious use. We have seen251.2 that the
evidence is clear that the life of Tammuz is the life of the crops and
fruits; and we discern a pure nature-religion unmoralised and without
dogma, but evoking a mood and a sentiment that might supply the
motive force to more complex and more spiritual creeds. It was not
suited to the religious atmosphere of the Assyrian and Babylonian
courts; but its influence spread far through Asia Minor. It captivated
the other polytheistic Semites, and at times, as Ezekiel shows us,
the women of Israel, revealing to these latter, no doubt, a vein of
religious sentiment unknown in the austere Mosaic monotheism. The
ritual of Adonis is mainly borrowed from the Tammuz service. For
instance, the rite of planting the short-lived “garden of Adonis,” of
which possibly the earliest record is in Isaiah xvii. 10, appears to be
alluded to in a verse of a Tammuz hymn.251.3 The figure of Tammuz
is primevally Sumerian; therefore the diffusion of his cult among the
various Semitic communities does not enable us to conclude that the
death and resurrection of a divinity is an aboriginal Semitic tradition.
As regards other evidence on the strength of which this dogma has
been attributed to them by some scholars, it is of late authority and
of doubtful validity. Josephus251.4 tells us that at Tyre the resurrection
of Herakles was once celebrated by Hiram; but this might well be a
derivative of the non-Semitic Sandon cult of Tarsos, which will be
considered below. And the legend of the death of Dido at Carthage,
even if there is no doubt that the queen was originally the great
goddess of Tyre, is no sufficient proof of a Phoenician ritual in which
the divinity died annually.
But as regards the non-Semitic peoples of anterior Asia, the
question of borrowing is more difficult to answer with certainty.
No Hittite monument nor any Hittite text has as yet revealed to us
any figure that we can identify with Tammuz. But certain indications
incline us to believe that the idea of the death of the god was not
unfamiliar to the Hittite religion or to some of the communities under
Hittite influence. On the Boghaz-Keui relief we have noted the
presence in the religious procession of those mysterious animals,
calves, or bulls, wearing caps of peculiar Hittite fashion.252.1 Are not
these “theanthropic animals” to be sacrificed as a communion-link
between man and God? We know that the bull was worshipped as
an incarnation of a Hittite deity; and therefore from the sacrifice of
the bull might emerge the dogma that the deity ceremoniously died
at certain periods. From the sanctity of the bull in ancient Hittite cult-
centres may have descended the mystic communion rite of the
Taurobolion or Tauropolion, which Cumont has shown good reason
for supposing to have arisen in the worship of the Persian Anahita,
and to have been adopted into the service of Kybele.253.1
More direct evidence is to be gleaned from the cult of Sandon or
Sandes of Tarsos, a city which was once within the area of Hittite
culture. The god of Tarsos comes later to be identified with the Tyrian
Baal and the Hellenic Herakles; and the legend of the death of the
latter hero may be an echo of a ἱερὸς λόγος of Tarsos, inspired by an
annual rite in which the god of the city was consumed on a funeral
pyre, and was supposed to rise again from the flames in the form of
an eagle.253.2 The later Tarsian coins display the image of the god,
the pyre, the eagle, the double-headed axe, and the lion;253.3 and
the last three of these symbols belong to the oldest religious art of
the Hittites. The proof would be complete if it could be shown that
the name Sandon or Sandes belongs to the Hittite language. All we
know at present is that it is not a Babylonian or Sumerian word, or
found in the vocabulary of any Semitic people. Prof. Sayce believes
himself to have found it in a cuneiform inscription of Boghaz-Keui.
This would be the direct proof that we require; but the word that he
transliterates as Sandes is said to be the ideogram of Hadad, the
Syrian Semitic god, and that Hadad is used as the Semitic
equivalent of Sandes is merely a conjuncture.253.4
A still clearer and more striking example of the phenomenon with
which we are dealing is the Phrygian and Lydian cult and legend of
Attis. The various and often conflicting details in the story of his birth,
life, and death, the various elements in his cult, are known to us from
late sources; the consideration of the whole question would not be
relevant here; but it is necessary for our purpose to determine, if
possible, what are the aboriginal motives of the myth and cult. It
seems likely that the earliest form of the Phrygian religion was the
worship of the great mother-goddess, coupled with a son or
lover,254.1 a young and beautiful god who dies prematurely, and
whose death was bewailed in an annual ritual, whose resurrection
was presented in a subsequent or accompanying service. Of the
death and the lamentation we have older evidence than for the
resurrection and the rejoicing, but the one seems to be a necessary
complement of the other. The family likeness of Attis to Tammuz
strikes us at first sight. As Tammuz appears as a young vegetation
deity, identified partly with the life of trees, partly with the corn, so
Attis in the Phrygian legend and ritual is presented as a tree-divinity,
and in the verse of a late hymn, which is inspired by an ancient
tradition, is invoked as “the corn cut by the reaper.”
And these two personalities of the Sumerian and Phrygian
religions evoked the same psychologic sentiment, sorrowful,
romantic, and yearning. The hypothesis naturally suggests itself that
the more Western people borrowed the cult from Mesopotamia, and
that this had happened as early as B.C. 1500.254.2 All scholars are
agreed at all events that the figure of Attis belongs to the older pre-
Aryan stratum of the population of Phrygia; modern speculation is
sometimes inclined to regard this as Hittite, and we know that the
Hittites adopted some part of the Babylonian religion. But the name
Attis itself is a stumbling-block to the hypothesis of borrowing from
Mesopotamia. Believing Adonis to be a Western-Semitic form of
Tammuz, we can explain the name as meaning merely “the Lord,” a
natural appellative of the Sumerian god. But we cannot so explain
“Attis.” It is non-Semitic, and must be regarded as belonging to an
Anatolian language-group, nor can we yet discover its root-meaning.
Again, there are many features of the Attis-worship and legend
that are not found in the corresponding Sumerian, and one at least
that seems essentially alien to it. The death of the vegetation-god,
originally suggested by the annual phenomenon of nature, may be
explained by various myths, when the personal deity has so far
emerged from his nature-shell that he is capable of personal drama.
The death of Tammuz does not appear to have been mythologically
explained at all. We may suppose that the killing of Adonis by the
boar was borrowed from the Attis legend, for in Phrygia, and also in
Lydia—as the Herodotean Ates story proves—this animal was
sometimes regarded as the enemy that slew the god. It is a
reasonable belief that the boar came to play this part in the story
through a misunderstanding of certain ritual, in which this victim was
annually offered as incarnating the deity, or was reverentially spared
through a sacrificial law of tabu. If this is an original fact of Attis-cult,
it counts somewhat against the hypothesis of derivation from
Mesopotamia, for the pig does not appear to have played any such
part, positive or negative, in Mesopotamian, as in the ritual of the
Western Semites and on the shores of Asia Minor; nor can any
connection at present be discovered between Tammuz and this
animal.
But another version of the death of Attis, current at some time
among his worshippers, was that he died from the effects of self-
mutilation, a motive suggested by the emasculation of the Phrygian
Galloi. We have here a phenomenon in the cult and myth that was
alien to the religious habits of the Mesopotamian communities. The
eunuch as a secular functionary is a figure belonging to an
immemorial social tradition of the East; but the eunuch-priest is the
morbid product of a very few religions, and there is no trace of such
in Mesopotamia. The Babylonian church-law demanded an
unblemished priesthood of strong virility, agreeing in this respect with
the Judaic and the Hellenic, and according with an ancient sentiment
that the vigour of the priest was the pledge of that continued flow of
divine power which supported the vigour of the community. Self-
emasculation was penalised in the religious rule of Jahwé, and the
Gallos was excluded from temple-worship by the ritual code of
Lesbos. The records of modern savagery and the history of
asceticism, whether in modern and mediaeval India or in early
Christianity, afford us varied illustration of the wildest excesses of
self-inflicted cruelty against the human body, but not—so far as I am
aware—of this particular form of self-destruction.256.1
As a religious practice it is a special characteristic of Phrygia, a
land always fascinating to the student on account of its strange
freaks of religious psychology; and from Phrygia the practice spread
into some adjacent communities, such as Bambyke. One may be
allowed to pause a little to consider the original motive that prompted
it. At first sight one is tempted to explain it as due to a morbid
exaggeration of the craving for purity. But elsewhere, where this
impulse was most powerful, for instance, in the later Orphic and Isis
cults, and in early and mediaeval Christianity, it produced many
mental aberrations but not this particular one. Nor, again, have we
any reason for supposing this craving to have been strong in the
devotees of Phrygia; the Galloi of Bambyke, according to Lucian,
were possessed by strong though impotent sexual desires and were
allowed full license with women. The form of communion most
ardently sought with the Phrygian goddess and with the later
Sabazios was a marriage symbolised by a sexual act; and Greek
and Latin writers, both pagan and Christian, agree in reprobating the
obscenities of Kybele-Attis worship; we may note also that Phrygian
sacred mythology is somewhat grosser than the Hellenic.
We are compelled to seek another explanation, and I can suggest
none other than that which I have hazarded elsewhere; namely, that
Phrygian religious emasculation was an act performed in a frenzy of
exaltation by the priest or mystes desirous of assimilating himself as
far as possible to the female nature of the great divinity.257.1 The
worship was under male ministration for the highest part; but for the
full exercise of divine power the male priest must become quasi-
female and wear a female dress, the latter part of the rôle being
common enough in primitive “theurgy.” The priest is himself at times
the incarnation of the young god, and is called Attis. Therefore Attis
was himself supposed to have performed the same act, even at the
cost of his life. How early was this institution of a eunuch priesthood
in Phrygia we have no direct evidence to prove. It may be a “Hittite”
tradition; for figures that Perrot reasonably interprets as eunuch-
priests are seen on the reliefs of Boghaz-Keui.
Returning to the topic of the death of the divinity, we may assume
that in Phrygia this was a very ancient tradition, enacted yearly by a
ceremonious laying out of the vegetation puppet on a bier, or the
suspension of it on a pine-tree. We have no direct or otherwise trusty
evidence for the immolation of the priest who incarnates the god;
doubtless the stories about the death of Marsyas and the harvest-
sacrifice of Lityerses point to a ritual of human-sacrifice; whom
Marsyas stands for is doubtful, but in the Lityerses legend it is
merely the passing stranger who is slain, and neither of these
traditions is explicitly linked with Attis-cult.
Finally, we may pronounce the hypothesis of the derivation of the
Phrygian cult from Mesopotamia to be unproved and unnecessary.
Pursuing this phenomenon further afield, we come to the area of
Minoan-Mycenaean culture. If the legend of the death of the
Minotaur could be safely interpreted as arising from the periodic
immolation of a bull-god, the idea that we are in quest of would be
proved to belong to the Minoan Cretans; but the frescoes of Knossos
present that event with such a gay and sporting entourage that one
feels shy of forcing a solemn religious significance into it. More
important for our purpose is the traditional Zeus-legend of Crete. It is
generally felt to be alien to the genuine Hellenic tradition concerning
their high god, as something adopted by the immigrant Hellenes
from an earlier Eteocretan ritual and creed.259.1 We have a glimpse
of a ritual in which the deity is born, is worshipped as an infant, and
then as a boy—κοῦρος, as he is invoked in the new fragment of the
hymn of the Kouretes—and especially as the son of a great mother,
not as a mature independent personality. Again, there appears an
orgiastic emotion and passion in the ritual that strikes a note in
harmony with the Phrygian Kybele-Attis worship. The very early
associations of Crete and the countries adjacent to the Troad are
now being revealed by accumulating evidence, and may point to an
affinity of stock. It may well have been, then, that the Minoan
Cretans had their counterpart to Attis, a young god who was born
and died periodically, whom they may have named Velchanos, the
name of the young deity seen sitting under a tree on a Cretan coin of
the fifth century. Though in age and character so unlike the Hellenic
Zeus, we may suppose that the incoming Hellenes named him so
because they found him the chief god of the island. We can also
understand why the later Bacchic mystery flourished so fruitfully in
Crete, if it found here already the ritual of a young god who died and
rose again, and why in later times the inhabitants celebrated with
such enthusiasm the Hilaria,259.2 the Easter festival of the
resurrection of the Phrygian divinity.
This attempt to reconstruct a portion of old Cretan religion on the
lines of the early Phrygian has only a precarious value, until some
more positive evidence is forthcoming from the Minoan art-record,
which hitherto has revealed to us nothing concerning an annual
divine birth and death. The ritual-legend is incomplete: we hear
sufficiently of the birth, and we may argue a priori that a periodic
ritual of the god’s birth implies a periodic death. Unfortunately all that
we glean from ancient literature is that there was a grave of Zeus,
perhaps in the Idaean cave, on which Pythagoras is said to have
written an epitaph.260.1 But a sceptical doubt arises here from the
fact which was pointed out by Rohde, that the grave of a divine
personage was often a misnomer of the underground sanctuary of a
chthonian deity; and either the Idaean cave or the great cavern on
Mount Dikte, whence the interesting relics of an immemorial cult
have recently been gathered,260.2 might at a later period have come
to be regarded as a grave. Still, we may ask, could the phrase “the
grave of Zeus” have become prevalent among a people with whom
the worship of this god was still a living creed, unless the faith also
prevailed that the god who died rose again to power? In that case
the “grave of Zeus” could be a name for a sanctuary where the ritual
of the death was enacted preliminary to the ritual of the birth.
This reconstruction then, and the a priori deduction emphasised
above, may claim to be at least legitimate.
Finally, some evidence may be added from Cyprian cult for the
view that the Minoan civilisation was cognisant of the dogma of the
death of the divinity. We hear of the grave of Ariadne-Aphrodite
which was shown in later times in Cyprus,261.1 and the Cretan and
Cypriote legends of the maidens called Gorgo, Parakuptousa, and
Galatea261.2 reveal to us an Aphrodite who died periodically and was
laid out on a bier and revived. The Aphrodite of Cyprus is most
probably of “Minoan” origin; and, being a goddess of vegetation, the
idea of periodic death and resurrection might naturally attach to her,
and might be associated with another type of ritual also, the annual
casting of her puppet into the sea, which probably gave rise to such
stories as the leap of Britomartis or Derketo into the waves.261.3
We can now deal with the purely Hellenic evidence. Confining our
view first to the cults and legends of the higher divinities of Hellas,
we cannot affirm that the death and resurrection of the deity is a
primitive tradition of the Hellenes. We may suspect it to have been a
leading motive in some of the local Arcadian cults of Artemis, if, for
instance, we interpret the Arcadian Kallisto as a form of the great
goddess herself; but it is very probable that Artemis in Arcadia and
many other of her cult-centres represents the pre-Hellenic divinity of
birth and fruitfulness. What we may dare to call the Hellenic spirit
seems to speak in the answer given by Xenophanes to the people of
Elea when they asked him whether they ought to sacrifice to
Leukothea and bewail her: “If you regard her as a deity, do not
bewail her; if as human, do not sacrifice.”261.4
But when we descend from the higher religion to the old Hellenic
agrarian cults associated with the heroes or daimones of the soil and
field, we find evidence of sorrowful rites, ceremonies of bewailing,
which belong to the same type as that of Tammuz; and in the Greek,
as in the Babylonian, the personage to whom they are attached is a
youthful hero or heroine of vegetation: such are Linos, perhaps the
earliest theme of a melancholy harvest-song of pre-Homeric days;
Hyakinthos, the “youth” of the Laconian land who may or may not
have been Hellenic, to whom the greater part of the Hyakinthia were
consecrated; Eunostos of Tanagra; Erigone, “the early-born,” of
Ikaria. The life of all these passes away as the verdure passes, or as
the crops are gathered in; and to one of them at least, Hyakinthos,
and perhaps to the others, the idea of an annual resurrection was
attached. But none of these came like Tammuz to play a world-part;
they remain the naïve, half-realised forms of poetic folk-religion. Like
to them is Bormos of Bithynia,262.1 whose death was bewailed at the
harvest-time with melancholy songs, accompanied by sad flute-
music, and Lityerses of Phrygia, whom the reapers lamented around
the threshing-floor.
Shall we say that all these are merely reflections cast afar by the
great cult-figure of Babylon?
Then we must say the same of the peasant-hero “German” whom
the modern Bulgarians adore and bewail, of the Russian Yarilo,262.2
and our own John Barleycorn. And at this point we shall probably fall
back on the theory of independent similar developments, and shall
believe that peasant religion in different parts of the world is capable
of evolving strikingly parallel figures in obedience to the stimulus of
similar circumstances and needs.
We have no surety, then, for a belief that Tammuz, or any shadow
of Tammuz, was borne to the western shores of the Aegean in the
days before Homer. And we know that Adonis, his nearest Anatolian
representative, only arrived late in the post-Homeric period.
Meantime, whatever view we may hold concerning prehistoric
religious commerce between East and West, this vital difference
between Mesopotamian and Hellenic religion must be strongly
emphasised: Babylonian liturgy is mainly a service of sorrow, and
part of that sorrow is for Tammuz; Hellenic worship was mainly
cheerful and social, and only in a few chthonian cults is a gloomier
tone discernible, nor can we anywhere hear the outbursts of violent
and ecstatic grief. In this respect, and in its remoteness from any
idea of the death and resurrection of the deity, Hellenic religion was
further removed from that of Catholic Europe than was the old
Phrygian or the Sumerian.
The Babylonian temple-service was complex and varied, and
offers many problems of interest to the comparative student. We
gather that a Holy Marriage was part of a religious drama perhaps
performed annually; for instance, we find reference to the solemn
nuptials of Ninib and Bau, and to the marriage presents given to
Bau.263.1 In every anthropomorphic polytheism, especially when
idolatry provides images that could be used for religious drama, this
ritual act is likely to occur. It is recorded of the southern Arabians in
the days before Islam, an ancient inscription speaking of the
marriage ceremony of Athtar.263.2 It is a marriage of the great God
and Goddess that according to the most reasonable interpretation is
represented on the Hittite reliefs of Boghaz-Keui. We may conjecture
that it was a ceremony of Minoan worship; a Mycenaean signet-ring
shows us a seated goddess with a young god standing before her
and joining his forearm to hers, while both make a peculiar gesture
with their fingers that may indicate troth-plighting;264.1 also, the later
legends and the later ritual commemorating the marriages of
Aphrodite and Ariadne may descend from the pre-Hellenic religious
tradition. Finally, we have fairly full evidence of the same religious
act in purely Hellenic cult. The ἱερὸς γάμος of Zeus and Hera was
enacted in many communities with certain traits of primitive
custom;264.2 the nuptials of Kore and the lower-world god might be
found in the ceremonies of certain temples;264.3 while the central
scene of the Eleusinia, the greater if not the lesser, included a Holy
Marriage.264.4 The Roman religion, in the original form of which there
may have been no marrying or giving in marriage, no family ties or
genealogies of divinities, no doubt borrowed its “Orci nuptiae” from
the Greek. But for the other cases, there is no need to resort to any
theory of borrowing to explain a phenomenon so natural at a certain
stage of religion.
Nor is it an important phenomenon, so long as the ceremony was
enacted merely by puppets or idols, as in the Boeotian Daidala.264.5
It only begins to be of higher significance for the history of religious
practice and thought, when the part of one of the divinities in this
drama is played by a human representative. For not only does this
offer indefinite possibilities of exaltation for the mortal, but it may
engender the mystic ideal and practice of communion with the
divinity through sexual intercourse, which played a great part in
Phrygian religion, and left a deep impress on early Christian
symbolism. The question whether the Mesopotamian religion
presents us with evidence of a “holy marriage” solemnised between
a mortal and the divinity must finally involve the more difficult

Download ebook Computational Geo Electromagnetics Methods Models And Forecasts Volume 5 Pdf full chapter pdf

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Download ebook Computational Geo Electromagnetics Methods Models And Forecasts Volume 5 Pdf full chapter pdf

Uploaded by

Copyright:

Available Formats

Computational Geo-electromagnetics:

Methods, Models, and Forecasts:

Library of Congress Cataloging-in-Publication Data

For information on all Elsevier publications visit our website at

Publisher: Candice Janco

estimates during EM data inversion and quantifying its results in

received much assistance from former employees of my Labo-

at a given frequency u, from the distribution of generalized

Computational Geo-Electromagnetics. https://doi.org/10.1016/B978-0-12-819631-1.00001-8

1.2 Methods of integral equations

1.2.1 The method of volume integral equations

where Eb is a background ﬁeld calculated with the assumption

1.2.2 The method of surface integral equations

turn, one can, using the continuity of the tangential components

1.3 Methods of differential equations

or the use of so-called differential equation methods, of which

1.3.1 The finite difference technique

which follow directly from (1.1).

1978), a number of improvements were made that signiﬁcantly

free and divergence-free projections that removes the null space of

1.3.2 The finite element technique

(Note that we are dealing with the determination of a stationary

of the diagonal dominance that, in turn, allows its efﬁcient solu-

Alumbaugh, 1999; Wilson et al., 1999; Kordy et al., 2013; Puzyrev

1.4 Hybrid schemes

where S is the surface of the discontinuity separating the layers, Ea

where r0 lies on the boundary of the domain V.

1.5 Analog (physical) modeling approaches

where L is the characteristic linear dimension of the model. Among

schemes) and measurement of voltages and currents in the circuit,

1.6 Balance technique for EM field

1.6.1 Governing equations

and to have a three-dimensional distribution of electrical conduc-

sðPÞ ¼ @ sðx; zÞ; P˛U2 ;

where U ¼ U1 WU2 WU3 , P ¼ Pðx; y; zÞ; with U3 u 0. The electro-

where k ¼ ðium0 Þ1=2 ; Rek > 0:

1.6.2 Boundary conditions

However, this causes two problems: ﬁrst, it is difﬁcult to deter-

where Eb is the background electrical

1.6.3 Discretization scheme

values is required. Various mechanisms for such a transition are

where l ¼ 1,2, ., L; m ¼ 1,2, ., M; n ¼ 1,2, ., N.

^ ð4Þ ^ ð5Þ ^ ð6Þ

Figure 1.1 Elementary cell of a spatial grid.

Figure 1.2 Structure of matrix D: L, M, Ndnumber of nodes in X, Y, and Z

- in the regions where s ¼ 0 or Vs ¼ 0 the second term of the

1.6.4 Calculation of the magnetic field

ﬁeld components from Eq. (1.10). This technique gives satisfac-

where r ¼ ½ðx x0 Þ2 þ ðy y0 Þ2 1=2 ; the value of Hz is precalcu-

1.6.5 Controlling the accuracy of the results

is satisﬁed by substituting in it the solution obtained by the ﬁnite

A more efﬁcient internal criterion can be based on satisfying

1.6.5.2 Comparison with analytical solution

1.6.5.3 Comparison with results obtained by other techniques

equation; (2) spline interpolation of the electrical ﬁeld values; (3)

1.7 Method of the EM field computation in

Figure 1.4 Graphs of the normalized horizontal component of anomalous

can be achieved with reduced computer resources if we conﬁne

1.7.1 Governing equations

Substituting these expansions into Maxwell Eqs. (1.9 and 1.10)

Figure 1.5 Model of the geoelectrical section (in cylindrical coordinates); U is

er and ez are the unit vectors of the cylindrical coordinate

normal ﬁeld is polarized linearly (with the magnetic ﬁeld in the

1.7.2 Boundary conditions