Seismological Research Letters, Volume 60, No.

2, April - June, 1989

A Student's Guide to and Review of Moment Tensors

M. L. Jost and R. B. Herrmann

Department of Earth and Atmospheric Sciences

Saint Louis University
P. 0. Box 8099
St. Louis, MO 63156

ABSTRACT
A review of a moment tensor for describing a general seismic point source is
presented to show a second order moment tensor can be related to simpler seismic
source descriptions such as centers of expansion and double couples. A review of
literature is followed by detailed algebraic expansions of the moment tensor into
isotropic and deviatoric components. Specific numerical examples are provided in
the appendices for use in testing algorithms for moment tensor decomposition.

INTRODUCTION was further extended by Backus and Mulcahy (1976), and

A major research interest in seismology is the Backus (1977a, b). Moment tensors can be determined
from free oscillations of the earth (e.g. Gilbert and
description of the physics of seismic sources. A common
approach is the approximation of seismic sources by a Dziewonski, 1975), long-period surface waves (e.g.
McGowan, 1976; Mendiguren, 1977; Patton and Aki,
model of equivalent forces that correspond to the linear
1979; Patton, 1980; Kanamori and Given, 1981, 1982;
wave equations neglecting non-linear effects in the near
source region (Geller, 1976; Aki and Richards, 1980; Ken- Romanowicz, 1981; Lay et al., 1982; Nakanishi and
Kanamori, 1982, 1984) or long-period body waves (e.g.
nett, 1983; Bullen and Bolt, 1985). Equivalent forces are
Stump and Johnson, 1977; Strelitz, 1978, 1980; Ward,
defined as producing displacements at the earth's surface
1980a, b; Fitch et al., 1980; Fitch, 1981; Langston, 1981;
that are identical to those from the actual forces of the
Dziewonski et al. , 1981; Dziewonski and Woodhouse,
physical process at the source. The equivalent forces are
1983a, b). Throughout this Student's Guide, we will
determined from observed seismograms that contain infor-
focus on second-rank, time independent moment tensors
mations about the source and path and distortions due to
(Appendix I). We refer to Dziewonski and Gilbert (1974),
the recording. Hence, the principle problem of source stu-
Gilbert and Dziewonski (1975), Backus and Mulcahy
dies is the isolation of the source effect by correcting for
(1976), Backus (1977a), Stump and Johnson (1977), Strel-
instrument and path.
itz (1980), Sipkin (1982), and Vasco and Johnson (1988)
The classical method of describing seismic sources, for a description of time dependent moment tensors.
having small dimensions compared to the wavelengths of Higher order moment tensors are discussed by Backus and
interest (point source approximation) is by their strength Mulcahy (1976), Backus (1977a, b), and Dziewonski and
(magnitudes, seismic moment) and their fault plane solu- Woodhouse (1983a).
tion (Honda, 1962; Hirasawa and Stauder, 1965;
The reason that moment tensors are important is
Herrmann, 1975). Recently, seismic moment tensors have
that they completely describe in a first order approxima-
been used routinely for describing seismic point sources
tion the equivalent forces of general seismic point sources.
(e.g. Kanamori and Given, 1982; Dziewonski and Wood-
The equivalent forces can be correlated to physical source
house, 1983b; Dziewonski et al. , 1983a-c, 1984a-c; Giar-
models such as sudden relative displacement at a fault
dini, 1984; EkstrZim and Dziewonski, 1985; Dziewonski
surface (elastic rebound model by H. F. Reid, 1910),
et al. , 1985a-d, 1986a-c, 1987a-f; EkstrOm et al , 1987;
rapidly propagating metastable phase transitions (Evison,
Sipkin, 1987; PDE monthly listings published by NEIS).
1963), sudden volume collapse due to phase transitions, or
Gilbert (1970) introduced moment tensors for calculating
sudden volume increase due to explosions (Kennett, 1983;
the displacement at the free surface which can be
Vasco and Johnson, 1988). The equivalent forces
expressed as a sum of moment tensorelements times the
representing a sudden displacement on a fault plane form
corresponding Green's function. An elastodynamic
the familiar double couple. The equivalent forces of a sud-
Green's function is a displacement field due to an uni-
den change in shear modulus in presence of axial strain
directional unit impulse, i.e. the Green's function is the
are represented by a linear vector dipole (Knopoff and
impulse response of the medium between source and
Randall, 1970). In conclusion, a seismic moment tensor is
receiver. The response of the medium to any other time
a general concept, describing a variety of seismic source
function is the convolution (Arfken, 1985) of that time
models, the shear dislocation (double couple source) being
function with the impulse response. The Green's function
just one of them.
depends on source and receiver coordinates, the earth
model, and is a tensor (Aki and Richards, 1980). The The equivalent forces can be determined from an
linearity between the moment tensor and Green's function analysis of the eigenvalues and eigenvectors of the
elements was first used by Gilbert (1973) for calculating moment tensor (Appendix I). The sum of the eigenvalues
moment tensor elements from observations (moment ten- of the moment tensor describes the volume change in the
sor inversion). The concept of seismic moment tensors source (isotropic component of the moment tensor). If the
37
 Jost and Herrmann
sum is positive, the isotropic component is due to an
explosion. The source has an implosive component if the North
sum is negative. If the sum of the eigenvalues vanishes,
then the moment tensor has only deviatoric components.
The deviatoric moment tensor represents a pure double
couple source if one eigenvalue equals zero. If none of the
eigenvalues vanishes and their sum still equals zero, the
moment tensor can be decomposed into a major and
minor double couple (Kanamori and Given, 1981), or a
double couple and a compensated linear vector dipole
(CLVD) (Knopoff and Randall, 1970). A CLVD is a
dipole that is corrected for the effect of volume change,
describing seismic sources which have no volume change,
net force, or net moment. In general, a complete moment
tensor can be the superposition of an isotropic component
and three vector dipoles (or three CLVD's, or three double
couples, Ben-Menahem and Singh, 1981).
This Student's Guide is an extension of 'A student's
guide to the use of P - and S- wave data for focal mechan-
ism determination" (Herrmann, 1975). Hence, emphasis is
given illustrating the relations between classical fault
plane solutions and seismic moment tensors. Addressing
general seismic point sources, we provide examples of
moment tensor decompositions into basic equivalent
source representations, as contributions of dipoles or dou-
ble couple sources. Clarification of terms such as major z
and minor double couple or compensated linear vector
dipole is provided. Moment tensor inversion schemes are Fig. 1. Definition of the Cartesian coordinates (x,y,z).
The origin is at the epicenter. Strike is measured
briefly summarized. In the appendices, examples of the
clockwise from north, dip from horizontal down,
use of notation by different authors are given along with
and slip counterclockwise from horizontal. u and v
some numerical results which are useful for testing com-
are the slip vector and fault normal, respectively
puter programs. Furthermore, the formulation of the
(modified after Aki and Richards, 1980).
basic Green's functions by Herrmann and Wang (1985) is
connected to a simple moment tensor inversion scheme.
arise due to differences between the model stress and the
GENERAL ELASTODYNAMIC SOURCE actual physical stress (stress glut, Backus and Mulcahy,
1976). Outside the source region, the stress glut vanishes
By using the representation theorem for seismic
as do the equivalent forces. The centroid of the stress
sources (Aki and Richards, 1980), the observed displace-
glut is then a weighted mean position of the physical
ment do at an arbitrary position x at the time t due to a
source region (Backus, 1977a; Aki and Richards, 1980;
distribution of equivalent body force densities, f k, in a
Dziewonski and Woodhouse, 1983a). It seems that the
source region is centroid of the stress glut gives a better position for the
co equivalent point source of an earthquake than the hypo-
dn (x,t) = f f G nk (x,t;r,t) f k (r,T) dV(r) dt , (1) center which describes just the position of rupture initiali-
-co v zation. The Taylor series expansion of the components of
where G n k are the components of the Green's function the Green's function around this new reference point is
containing the propagation effects, and V is the source (2)
G n k(x,t =
volume where f k are non-zero. We assume the summation
convention for repeated indices (Arfken, 1985). The sub- E - (7.1 ,—W • • • (rim —el. ) G„ k . . . (x,t 4T)
script n indicates the component of the displacement. m—o ml
Throughout, we will use the following coordinate system
The comma between indices in (2) describes partial deriva-
(Figure 1): The x-axis points towards north, the y-axis tives with respect to the coordinates after the comma.
towards east, and the z-axis down (this system is right
We define the components of the time dependent force
handed). Then, e z , e y , and e x are the unit vectors moment tensor as :
towards north, east, and vertically down, respectively.
By assuming that the Green's functions vary Mkji 2 (e,T) = f (rig—e) • • (7. 5. —e i )f k(r,t) dV. (3)
V
smoothly within the source volume in the range of
moderate frequencies, the Green's functions can be If conservation of linear momentum applies, such as for a
expanded into a Taylor series around a reference point to source in the interior of a body, then a term in Mk does
facilitate the spatial integration in (1) (Kennett, 1983; not exist in (3). With the Taylor expansion (2) and the
Arfken, 1985). The expansion is usually done around the definition of the time dependent moment tensor (3), the
centroid r = e. The physical source region is character- displacement (1) can be written as a sum of terms which
ized by the existence of the equivalent forces. These forces resolve additional details of the source (multipole expan-
38
 A Student's Guide to and Review of Moment Tensors

sion, Backus and Mulcahy, 1976; Stump and Johnson, 1982). Neglecting higher order terms, we get (Stump and
1977; Aki and Richards, 1980; Kennett, 1983; Dziewonski Johnson, 1977)
and Woodhouse, 1983a; Vasco and Johnson, 1988):
d n (x,t)= Mk1 [ Gnk,i * s(T)] (5)
d n (x,t) = (4)
Mk, are constants representing the components of the
xe<>
-, 1 r„ second order seismic moment tensor M, usually termed
j„,(Xlt 7e9 0
- * n, (e7 T) the moment tensor. Note that the displacement do is a
1
linear function of the moment tensor elements and the
where * denotes the temporal convolution. By using a
terms in the square brackets. If the source time function
seismic signal that has much longer wavelengths than the s (t) is a delta function, the only term left in the square
dimensions of the source (point source approximation), we
brackets is G nk1 describing nine generalized couples. The
need to consider only the first term in (4) (Backus and
derivative of a Green's function component with respect
Mulcahy, 1976; Stump and Johnson, 1977). Note, that
to the source coordinate e i is equivalent to a single couple
single forces will not be present in (4) if there are no
with arm in the e i direction. For k = j, i.e. force in the
externally applied forces (indigenous source). The total
same direction as the arm, the generalized couples are vec-
force, linear and angular momentum must vanish for the
tor dipoles (Figure 2; Maruyama, 1964). Thus, the
equivalent forces of an indigenous source (Backus and
moment tensor component Mkt gives the excitation of the
Mulcahy, 1976). The conservation of angular momentum
generalized (k,j) couple.
for the equivalent forces leads to the symmetry of the
seismic moment tensor (Gilbert, 1970).
DOUBLE COUPLE SOURCES
We assume that all components of the time depen-
The moment tensor components in (5) in an isotropic
dent seismic moment tensor in (4) have the same time
medium for a double couple of equivalent forces are given
dependence s(t) (synchronous source, Silver and Jordan,
by

y
z z z

z z z

z z
Fig. 2. The nine generalized couples representing G n k j in (5) (modified
after Aki and Richards, 1980).

39
 Jost and Herrmann

A ( ukui
Mkj =µ A vk ) , ( 6 )
The double couple u k vj + u j IA can equivalently be
described by its eigenvectors (Gilbert, 1973).
where it is the shear modulus, A is the area of the fault
plane, u denotes the slip vector on the fault surface, and uk Vjuj vk = tkti — Pj Pk (10 )

v is the vector normal to the fault plane (Ald and

= 0.5 [(tk + pk)(t j — p j )+(tk — pk)(t jpj)]
Richards, 1980; Ben-Menahem and Singh, 1981). Note
that the contributions of the vector of the fault normal v Comparing the terms in (10), we find the relation between
and the slip vector u are symmetric in (6). From the sym- tension and pressure axes and slip vector and fault normal
metry of M, we note that the roles of the vectors u and v (Appendix I):
could be interchanged without affecting the displacement 1
u= ( t p) (11)
field; i.e. the fault normal could equivalently be the slip
vector and vice versa. This well known fault plane - auxi- 1
liary plane ambiguity cannot be resolved from the seismic t — p) . (12)
V2 (
radiation of a point source. Hence studies of locations of
aftershocks, surface faulting, rupture directivity, or static The other nodal plane is defined by
final displacements (Backus, 1977a) need to be done in 1
u= (t—p) (13)
order to resolve this ambiguity.
The term u k vj + u i vk in (6) forms a tensor, D,
v= 1 ( t +p ) . (14)
describing a double couple. This tensor is real and sym-
metric, giving real eigenvalues and orthogonal eigenvec-
tors (Appendix I). The eigenvalues are proportional to (1, If strike, ', dip, .5, and slip, X, of the faulting are
0, -1). Hence, the characteristic properties of a moment known, the slip vector u and the fault normal v are given
tensor representing a double couple are 1) one eigenvalue by (Aki and Richards, 1980)
of the moment tensor vanishes, and ii) the sum of the u= cos X cos 1 + cos 8 sin X sin 4> ) e x
eigenvalues vanishes, i.e. the trace of the moment tensor is
zero (the other two eigenvalues are constrained to equal ( cos X sin 4> — cos 8 sin X cos 1 ) e y(15)
magnitude but opposite sign).
— Ti sin S sin X e z ,
Let t, b, and p designate the orthogonal eigenvectors
to the above eigenvalues (Herrmann, 1975; Backus, 1977a; where u is the mean displacement on the fault plane. The
Dziewonski and Woodhouse, 1983a). fault normal v is
v = —sin5 sin(I) e xsinS cos4> e y — toss e z . (16)
t— (v+u) (7)
The scalar product of u and 1/ is zero. The strike of the
b=vXu (8) fault plane, f, is measured clockwise from north, with the
fault plane dipping to the right when looking along the
p= 1 ( v—u). (0) strike direction. Equivalently, the hanging wall is then to
the right (Figure 1). The dip, 8, is measured down from
The tensor D corresponding to the terms in the brackets
the horizontal. The slip, X, is the angle between the strike
in (6) can be diagonalized (principal axis transformation,
direction and the direction the hanging wall moved rela-
see Appendix I), where the eigenvectors give the directions
tive to the foot wall (the slip is positive when measured
of the principal axes. The eigenvector b corresponding to
counterclockwise as viewed from the hanging wall side).
the eigenvalue zero gives the null-axis, the eigenvector t
The range of the fault orientation parameters are
corresponding to the positive eigenvalue gives the tension
axis, T, and the eigenvector p corresponding to the nega- 0 < < 27r, 0 < 8 < -2, and —7r < X < ir (Herrmann,
tive eigenvalue gives the pressure axis, P, of the tensor. 1975; Aki and Richards, 1980). The scalar seismic
These axes can be related to the corresponding axes of the moment is
fault plane solution, since we are focusing on pure double
couple sources. The P-axis is in the direction of max- M, = µA Ti (17)
imum compressive motion on the fault surface; the T-axis
Equation (6) together with (15), (16), and (17) lead to
is the direction of maximum tensional motion. Note that
the Cartesian components of the symmetric moment ten-
the P- and T-axes inferred from the motion on the fault sor in terms of strike, dip, and slip angles.
surface are not necessarily identical to the axes of
maximum tectonic stress, since the motion can be on a Mzx = —M0 (sin S cos X sin 24> + sin 28 sin X sin e 4>)
preexisting plane of weakness rather than on a newly My y = (sin S cos X sin 24> — sin 28 sin X cos t 4))
formed fault plane that would correspond to the max-
imum tectonic stress (McKenzie, 1969). However, this M„ Mo (sin 28 sin X) (18)
ambiguity cannot be resolved from the seismic radiation. Mxy = M, (sin S cos X cos 24> + 0.5 sin 2.5 sin X sin 21)
In order to determine the direction of maximum tectonic
stress, additional geological data such as in situ stress Mzz = —M0 (cos S cos X cos 4 + cos 25 sin X sin 4))
measurements and frictional forces is necessary. Lacking
My , = —M0 (cos S cos X sin 4> — cos 25 sin X cos 4))
this kind of information, it is generally assumed that the
P- and T- axes found from the seismic wave radiation are Different notation of the moment tensor elements are dis-
somewhat indicative of the direction of tectonic stress. cussed in Appendix II. In Appendix III, several simple

40
 A Student's Guide to and Review of Moment Tensors
moment tensors are related to fault plane solutions. general moment tensor decomposition by rewriting m as
Body-wave and surface wave radiation patterns from a Ftr (M) 0 0
source represented by a moment tensor are discussed by 1
m=3 0 tr (M) 0 (23)
Kennett (1988).
0 0 tr (M)
Since the seismic moment tensor is real and sym-
metric, a principal axis transformation can be found,
diagonalizing M (Appendix I). The diagonal elements are m 0 0
the eigenvalues m, of M. Then, the scalar seismic moment { 0 m2 o
can be determined from a given moment tensor by
0 0 M3
Mo = —
1 ( Imi I + lm2 I
2 (19)
tr (M) 0 0
where m 1 and m 2 are the largest eigenvalues (in the abso- = { 0 tr(M) 0 + rri
3
lute sense). The seismic moment can equivalently be 0 0 tr(M) i -1
estimated by the relations (Silver and Jordan, 1982):
where tr (M) = m i + m 2 + ni 3 is the trace of the
v 2mi 2
EMk) moment tensor and Tri i is a set of diagonal matrices
mo (20)
2 whose sum yields the second term in (23). The purely
deviatoric eigenvalues m ;` of the moment tensor are
+ m2 + m3 1
GENERAL SEISMIC POINT SOURCES mi = mi — m, tr(M) . (24)
3 3
In this section, it is assumed that the seismic source
The first term on the right hand side (RHS) of (23)
cannot be described by a pure double couple mechanism.
describes the isotropic part of the moment tensor. The
The moment tensor is represented as sum of an isotropic
eigenvalues of the isotropic part of the moment tensor are
part, which is a scalar times the identity matrix, and a important for quantifying a volume change in the source.
deviatoric part. The second term describes the deviatoric part of the
In order to derive a general formulation of the moment tensor consisting of purely deviatoric eigenvalues,
moment tensor decomposition, let's consider the eigen- which are calculated by subtracting 1/3 tr (M) from each
values and orthonormal eigenvectors of the moment ten- eigenvalue of M. This deviatoric part of the moment ten-
sor. Let m ; be the eigenvalue corresponding to the ortho- sor can be further decomposed, where the number of
normal eigenvector a, --= ((L iz ,a 1y ,a,,) T . Using the ortho- terms or the specific form of the decomposition will be
normality of the eigenvectors (Appendix I, (A1.5)), we can discussed in the next sections. Obviously, a multitude of
write the principal axis transformation of M in reverse different decompositions are possible. In Appendix IV, we
order as: give some numerical examples illustrating several methods
[al of moment tensor decomposition.

M= a2 a3
J m a2
T
(21) Vector Dipoles
A moment tensor can be decomposed into an isotro-
pic part and three vector dipoles. In equation (23) let N =

=r
a lz a2z a3zm 0 0 a12 a ly a lz 3 and
= n-11 a2y a3 1 0 P'2 0 {a2, 122 1 a2z •
00 0 0 0 0 0
a12 . a 2z a 3, 0 0 m3 a 3, a3 1 a 3, 0 0 m2= 74 0, m3= 00 0. (25)
00 0 0 0 0 0 0 m;
From (21), we find relations between components of the
eigenvectors and moment tensor elements: Applying (21) to E l , we get for the first deviatoric term
Mzx m 1 a i2x + m 2 + m 3 a 12 (i=1) in the decomposition
„, 2 + „„„ „, 2 ± m2
".2 142y 3-.3y
Myy ml
2
„2 2 "'12
Mzz = m l alz + m 2 - 22 . — 3 `. 3z
a 21 + m 3 a3 2 a3 1(22)
m; alzaly 2
a ly a ly a iz = mi a l a i , (26)
Mxy = m l alz a ly a2x
a ly a lz 2
a 3z alz alz a lz
aMxz = m l lz a lz + M 2 a 2x a 2z + m3 a 3 ,

a2z
Myz ml aly alz a2y a3y a 3„ where we identified the matrix as the dyadic al a i (Appen-
The effect of the eigenvalue decomposition (21) is that a dix I). The dyadic a l a i describes a dipole in the direction
new orthogonal coordinate system, given by the eigenvec- of the eigenvector a l . By applying (21) to rri 2 and rri3 in
tors, has been defined. In this new coordinate system, the (25), we get similar expressions involving a 2 a 2 and a3 a3 ,
source excitation is completely described by a linear com- describing the second and third deviatoric terms in the
bination of these orthogonal dipole sources. decomposition. Finally, equation (21) can be written for
m in (21) is the diagonalized moment tensor. The ele- the decomposition into three linear vector dipoles along
ments of m are the eigenvalues of M. We now define the the directions of the eigenvectors of M as
41
 Jost and Herrmann
1
M=— (m 1+m2+m3) (27) which is identical to equation (4.56) in Ben-Menahem and
3
Singh (1981).
m i*a2a2-1-m; a 3 a3 ,
Major and Minor Couple
which is identical to (22) and equation (4.55) in Ben- Next, we will decompose a moment tensor into an
Menahem and Singh (1981). isotropic component, a major and minor double couple.
The major couple seems to be the best approximation of a
Double Couples general seismic source by a double couple (Appendix IV),
Next, we decompose a moment tensor into an isotro- since the directions of the principal axes of the moment
pic part and three double couples. For the deviatoric part tensor remain unchanged. The major double couple is con-
in (23) let N = 6 and structed in the following way (Kanamori and Given, 1981;
* Wallace, 1985): The eigenvector of the smallest eigenvalue
mi 0 0 m1 0 0 (in the absolute sense) is taken as the null-axis. Let's
1
—= — 0 —m i* 0 , m2 =
1
0 0 0 _
assume that I m 3* I> Im 2* I> Im 1 I in (23). In (23), let
3 3 N=2 and use the deviatoric condition m 1* ±m 2* -km; = 0
0 0 0 0 0
to obtain
[0
0 0 0 r m 2* 0 0
[0 0 0 m1 0 0
1 1
M3 = m2 0 m4 = 0 m 2* 0 (28) = 0 —m; 0 0 —in 1 0 (31)
3 3 , m2 =
0 0 — m2 0 0 0 m3 0 0 0
0 0
0 0 0 3 0 0 Applying (21) to Fri i , we get the first deviatoric term in
1 1
1115 = 0 — m3 0 m6 = 0 0 0 the decomposition which corresponds to a pure double
3 3 couple termed major couple.
0 0 m3 0 0 M3
0 0 0 aiT
where each m — , is equivalent to a pure double couple
M m" = [ a i a 2 a 3 0 —7T/ 3 2 (32)
source (Appendix III). Notice that each double couple
consists of two linear vector dipoles (c.f. (25), (26) and 0 0 M 3 a3
(28)), e.g. (m i* /3) (a l a i — a 2 a 2 ) for E l . Each dipole con-
sists of two forces of equal strength but opposite direction Instead of the major double couple, a best double couple
(c.f. Figure 2). Then, the double couple can be seen this can be constructed similarly by replacing m; in (32) by
way: The first couple is formed by one force of each the average of the largest two eigenvalues (in the absolute
dipole, one force pointing in the positive a 1 , the other in sense, Giardini, 1984). Applying (21) to n12 gives the
the negative a 2 direction. The corresponding other couple second deviatoric term in the decomposition which also
is constructed by the complementary force of each dipole, corresponds to a pure double couple termed minor couple.
pointing toward the negative a 1 and positive a 2 direction.
Fajl
Using (21) with (23) and (28), we get the result that a mMIN = [a i
a 2 a3 1E2 a 2(33)
moment tensor can be decomposed into an isotropic part
and three double couples. a 3T
1 1 The complete decomposition is then:
M= — (m +m2+m3)I+ -3-- em 1 — m2) (a1a1 a2a2) —
(29)
3
1
(34
+ — (7n 2— 3) (a2 a 2 —a3 aQ)-1,--(m M= —
3 (m i±m2±m3)I )

3 —m 1 ) (a3 a 3 —a 1 a 1 ) ,
3
+ m 3* (a3 a 3 —a 2 a 2 ) +m (a 1 a 1 a 2 a2 )
which is identical to equation (4.57) in Ben-Menahem and —

Singh (1981).
Double Couple - CLVD
CLVD
Following Knopoff and Randall (1970) and Fitch et
Alternatively, a moment tensor can be decomposed
al. (1980), we can decompose a moment tensor into an iso-
into an isotropic part and three compensated linear vector

r
tropic part, a double couple and a compensated linear vec-
dipoles. Adding terms like Fri 1 and if i2 in (28) gives a
tor dipole. Let's assume again that Im; I>Im2 l>1 774
CLVD, 2a 1 a 1 a 2 a2
— a3 a3 . This CLVD represents a
—
in (23). We can write the deviatoric part in (23) as (N =
dipole of strength 2 in the direction of the eigenvector a 1 ,
1)
and two dipoles of unit strength in the directions of the
eigenvectors a 2 and a 3 , respectively. The decomposition F 0 0
can then be expressed as: m i = m; 0 (F-1) 0 , (35)
0 0 1
M= 1---(m 1 Fm 2 Fm 3 )I+ 1 m i (2a i a l a 2 a 2 a3 a 3 )
- - — — (30)
3 3 where F = - m i.* / m; and (F-1) = m; / 94. Note that
1 m 2 (2a2 a 2 —a j a i —a 3 a3 )± 1 3 (2a 3 a3 —a l a i —a 2 a 2 ),
O<F<0.5. This constraint on F arises from the deviatoric
3 3 m condition m 1* -Fm;+m; = 0. We can decompose (35)

42
 A Student's Guide to and Review of Moment Tensors

into two parts representing a double couple and a CLVD moment tensors from surface waves and body waves was
done by Fitch et al. (1981). Dziewonski et al. (1981) sug-
0 0 0 —1 0 0
gested an iterative inversion method, solving for the
m; (1 — 2F) 0 —1 0 + m; F 0 —1 0 ,(36) moment tensor elements and the centroid location (Backus
0 0 1 0 0 2 and Mulcahy, 1976; Backus, 1977a; see Dziewonski and
where we assumed that the same principal stresses Woodhouse, 1983a for a review). The reason for that
approach is that moment tensor elements trade off with
produce the double couple as well as the CLVD radiation.
the location of the earthquake. The lateral heterogeneity
The complete decomposition (21) is then:
of the earth was considered in inversion methods by Pat-
1 ton (1980), Romanowicz (1981), Nakanishi and Kanamori
M= — (m i ±m 2 -Fm 3 )I + m 3 (1-2F) (a 3 a 3 —a 2 a2 ) (37)
3 (1982), and Dziewonski et al (1984c).
m3 F (2a 3 a3 —a 2 a 2 —a l a l ) . The moment tensor inversion in the time domain can
use the formulation in (5) (e.g. Gilbert, 1970; McCowan,
1976; Stump and Johnson, 1977; Strelitz, 1978; Fitch
To estimate the deviation of the seismic source from
et al. , 1980; Ward, 1980b; Langston, 1981). If the source
the model of a pure double couple, Dziewonski et al.
time function is not known or the assumption of a syn-
(1981) used the parameter
chronous source is dropped (Sipkin, 1986), the frequency
domain approach is chosen (e.g. Gilbert, 1973; Dziewonski
M min
E— (38) and Gilbert, 1974; Gilbert and Dziewonski, 1975; Gilbert
max and Buland, 1976; Mendiguren, 1977; Stump and John-
son, 1977; Patton and Aki, 1979; Patton, 1980; Ward,
where m in* , n is the smallest eigenvalue (in the absolute
1980a, Kanamori and Given, 1981; Romanowicz, 1981):
sense) and max is the largest (in the absolute sense),
given by (24). From (35), we see that 6 = F. For a pure (x,f ) =-* Mki(f ) Gnk ,j(f ) • (39)
double couple source, ra m % = 0 and e = 0; for a pure
Both approaches, (5) and (39) lead to linear inversions in
CLVD, e = 0.5. Alternatively, e can be expressed in per-
the time or frequency domain, respectively. The advan-
centages of CLVD (multiply c by 200. The percentage of
tage of linear inversions is that a large number of fast
double couple is (1-2e) * 100). Dziewonski and Wood-
computational algorithms are available (e.g. Lawson and
house (1983b, see also Giardini, 1984) investigated the
Hanson, 1974; Press et al. , 1987). We can write either (5)
variation of e versus seismic moment and earthquake spa-
or (39) in matrix form:
tial distribution on the surface of the earth.
. (40)
MOMENT TENSOR INVERSION
In the time domain, the vector d consists of n sampled
There are various methods of inversion for moment values of the observed ground displacement at various
tensor elements. The inversion can be done in the time or arrival times, stations, and azimuths. G is a n X 6
frequency domain. Different data (e.g. free oscillations, matrix containing the Green's functions calculated using
surface- and body waves; different seismogram com- an appropriate algorithm and earth model, and m is a
ponents) can be used separately or combined. In addition, vector containing the 6 moment tensor elements to be
certain a priori constraints such as tr (M) = 0, or Mx, = determined (Stump and Johnson, 1977). In the frequency
Mr = 0 can be imposed to stabilize the inversion, result- domain, (40) can be written separately for each frequency.
ing in a decrease in number of resolved moment tensor d consists of real and imaginary parts of the displacement
elements. In this Student's Guide, we briefly outline cer- spectra. Weighting can be introduced which actually
tain approaches and refW to the original papers for smoothes the observed spectra subjectively (Mendiguren,
further reference. 1977; see also Ward, 1980b for weighting of body-wave
Gilbert (1970) introduced the seismic moment tensor data in the time domain). In the same way, G and Fri
for calculating the excitation of normal modes (Saito, contain real and imaginary parts. rTi contains also the
1967) of free oscillations of the earth. Gilbert (1973) sug- transform of the source time function of each moment
gested an inversion scheme for moment tensor elements in tensor element. If constraints are applied to the inversion,
the frequency domain. Gilbert and Dziewonski (1975) then m can contain a smaller number of moment tensor
used free oscillation data for their moment tensor inver- elements. In such a case, G has to be changed accordingly.
sion. Gilbert and Buland (1976) investigated on the smal- We refer to Aki and Richards (1980) for the details of
lest number of stations necessary for a successful inversion solving (40) for Fri (Note that (40) is identical to their
(see also Stump and Johnson, 1977). McCowan (1976), (12.83)).
Mendiguren (1977), Patton and Aki (1979), Patton (1980), The following presents an outline of the processing
Romanowicz (1981), Kanamori and Given (1981, 1982), steps in a moment tensor inversion. The first step is the
Lay et al. (1982), Nakanishi and Kanamori (1982, 1984), data acquisition and the preprocessing. We need data
and Scott and Kanamori (1985) used long-period surface with good signal to noise ratio that are unclipped and
waves (typically low pass filtered at 135 sec). Stump and that have a good coverage of the focal sphere (Satake,
Johnson (1977), Strelitz (1978, 1980), Ward (1980a, b), 1985). Glitches (non-seismic high amplitude spikes due to
Fitch et al. (1980), Langston (1981), Dziewonski et al. non-linearity of instruments e.g. Dziewonski et al., 1981)
(1981), and Dziewonski and Woodhouse (1983a, b), used have to be identified and possibly removed. Analog data
moment tensor inversion for body wave data (typically have to be digitized. The effect of non-orthogonality of
low pass filtered at 45 sec). A comparison between the analog recorder must be corrected. The digitized
43
 Jost and Herrmann
record has to be interpolated and resampled with a con- modes, causing systematic errors in the inversion.
stant sampling rate. At this point, a comparison of the Systematic errors in the inversion are also due to
sampled waveform with the original one can help to iden- deviations of the earth-model from the actual properties
tify digitization errors. The horizontal components will of the earth, affecting the synthetic Green's functions.
be rotated into radial and transverse components. Linear This is a fundamental problem in the sense that we are
trends have to be identified and removed. The instrument able to separate the source effect from the observed
effect is considered next (for WWSSN data see Hagiwara, seismogram only to a limited accuracy (Mendiguren, 1977;
1958; for SRO data see McCowan and Lacoss, 1978; for Langston, 1981; Silver and Jordan, 1982; O'Connell and
IDA data see Agnew et al., 1976). We can use either one Johnson, 1988). A major problem is the effect of lateral
of the two approaches: i) we can remove the instrument heterogeneity of the earth (Engdahl and Kanamori, 1980;
effect from the observed data and compare with theory or Romanowicz, 1981; Gomberg and Masters, 1988; Snieder
ii) we can apply the instrument response to the synthetic and Romanowicz, 1988). For example, a relative change
Green's functions and compare with observed data. The of 0.5 % due to lateral heterogeneity can cause a misloca-
nominal instrument response can be used or the calibra-
tion in the order of of 50 km at epicentral distances of
tion of the instrument can be checked by using f.e. the about 90 degrees (Dziewonski and Woodhouse, 1983b).
calibration pulse on the record. In addition, the polarity Giardini (1984) and Ekstr.5m and Dziewonski (1985)
of the instruments should be verified, e.g. from records of
reported on regional shifts in centroid positions due to
known nuclear explosions. High frequency noise in the lateral heterogeneity. In the inversion, lateral hetero-
data is removed by low-pass filtering. Amplitudes are geneity is often neglected, i.e. the calculation of the
corrected for geometrical spreading and reflections at the Green's functions is usually based on parallel layers of
free surface of the earth (Bullen and Bolt, 1985). For sur- lateral homogeneity (Harkrider, 1964, 1970; Langston and
face waves, the moving window analysis (Landisman et Helmberger, 1975; Harkrider, 1976). Nakanishi and
al., 1969) is applied in order to determine the group velo- Kanamori (1982) included the effect of lateral hetero-
city dispersion. From this analysis, we can identify the geneity into the moment tensor inversion. Another
fundamental mode Rayleigh and Love waves which can approach was developed for earthquakes within a small
then be isolated. source area: a calibration event is declared (mechanism
Second, synthetic Green's functions are calculated. known); the spectral ratio of any earthquake in that
Notice that the Green's functions are dependent on the region and the calibration event will result in isolating the
earth-model, the location of the point source (centroid of difference in source effects - the influence of the path is
the stress glut, or epicenter and focal depth), and the eliminated (Patton, 1980). It seems that the errors due to
receiver position. lateral heterogeneity are usually large enough to make a
The third step is the proper inversion, i.e. the solu- statistical significant detection of an isotropic component
tion of (40) (Aki and Richards, 1980). Usually, the inver- of the moment tensor difficult (Okal and Geller, 1979;
sion is formulated as least squares problem (Gilbert, 1973; Silver and Jordan, 1982; Vasco and Johnson, 1988).
Gilbert and Buland, 1976; Mendiguren, 1977; Stump and Patton and Aki (1979) investigated the influence of
Johnson, 1977). However, using other norms can have noise on the inversion of long-period surface wave data.
advantages in situations where less sensitivity to gross They found that additive noise such as background
errors like polarity reversions is required (Claerbout and recording noise does not severely affect the results of a
Muir, 1973; Fitch et al., 1980; Patton, 1980). linear inversion. However, multiplicative noise (signal
The source time function in (5) is often assumed to generated noise) caused by focusing, defocusing, mul-
be a step function (Gilbert, 1970, 1973; McCowan, 1976; tipathing, higher mode or body wave interference, and
Stump and Johnson, 197.7; Patton and Aki, 1979; Patton, scattering distorts the inversion results significantly
1980; Ward, 1980b; Dziewonski et al., 1981; Kanamori (overestimation or underestimation of moment tensor ele-
and Given, 1981). Aiming at the recovery of source time ments, deviation from the source mechanism; Patton,
functions, Burdick and Mellman (1976) used a powerful 1980; Ward, 1980b). Finally, body waves of events with
iterative waveform inversion method based on optimizing moments larger than 10 27 dyne-cm are severely affected by
the cross-correlation between observed, long-period body- finiteness of the source and directivity. If not corrected
wave trains and synthetics. The same approach was used for, an inversion can lead to severe errors in the moment
by Wallace et al. , (1981) inorder to invert for fault plane tensor elements (Dziewonski et al., 1981; Kanamori and
solutions. Other methods were employed by Strelitz (1980) Given, 1981; Patton and Aki, 1979; Lay et al., 1982;
and Kikuchi and Kanamori (1982) for large earthquakes Giardini, 1984).
(see also Lundgren et al. , 1988). Christensen and Ruff The inversion has only a limited resolution of
(1985) reported on a trade-off between source time func- moment tensor elements for certain data. If we have spec-
tion and source depth for shallow events. tra of fundamental mode Rayleigh waves only, the con-
If the focal depth is not known, then a linear inver- straint that the trace of the moment tensor vanishes (no
sion can be done for each depth out of a number of trial volume change) must be applied (Mendiguren, 1977; Pat-
depths. The most probable depth will minimize the qua- ton and Aki, 1979). This constraint is linear. Another
dratic error between observed and theoretical waveforms constraint which is often applied in addition is that one
(Mendiguren, 1977; Patton and Aki, 1979; Patton, 1980; eigenvalue vanishes (approximating the source by a dou-
Romanowicz, 1981). The influence of source depth on the ble couple). This constraint is not linear (Strelitz, 1978;
results of the moment tensor inversion was investigated Ward, 1980b). In such a case, the inversion is iterative,
by Sipkin (1982; Dziewonski et al , 1987b). Differences in using a linearized version of the constraints (Ward,
source depth influence the relative excitation of normal 1980b). For earthquakes at shallow depths (less than
44
 A Student's Guide to and Review of Moment Tensors
about 30 km), the moment tensor elements Mx , and Myz ACKNOWLEDGMENTS
corresponding to vertical dip slip faulting are not well
constrained from long-period surface wave data since the We thank two anonymous reviewers for their helpful
related excitation functions assume very small values near criticism of the manuscript. Critical remarks by Oznur
the surface of the earth (Fitch et al., 1981; Dziewonski et Mindevalli are appreciated. Funds for this research were
al., 1981; Kanamori and Given, 1981, 1982; Dziewonski provided by the Defense Advanced Research Projects
and Woodhouse, 1983a). In order to overcome this prob- Agency under contract F49628-87-K-0047 monitored by
lem, additional independent data, such as fault strike the Air Force Geophysics Laboratory.
(observed surface breakage) can be introduced into the
inversion. Another approach is to constrain these APPENDIX I
moment tensor elements to be zero. Thus, possible fault In the following, we give some mathematical
mechanisms are restricted to vertical strike slip or 45 definitions of tensors, the eigenvalue problem and dyadics
degree dip slip (Kanamori and Given, 1981, 1982). following Arfken (1985).
In Appendix V, we relate the Green's functions in the Let M be a moment tensor of second rank (order).
formulation of Herrmann and Wang (1985) to a simple Then, M is represented as a 3X3 matrix in a given refer-
moment tensor inversion scheme. This inversion example ence frame. Let apk be the cosine of the angle between the
is aimed at testing computer programs. p axis of another coordinate system and the k axis. Then
the components of M, Mki , transform into the new refer-
CONCLUSION ence frame by the relation
A seismic moment tensor describes the equivalent = akp Mkj (A1.1)
forces of a seismic point source. The eigenvectors are the Mp g
principal axes of the seismic moment tensor. For pure where we need to sum over repeated indices (summation
double couple sources, the principal axis corresponding to convention).
the negative eigenvalue is the pressure axis, the principal Given a moment tensor M, let's assume that there is
axis corresponding to the positive eigenvalue is the tension a vector a and a scalar m such that
axis, and the principal axis corresponding to the eigen-
value zero gives the null axis. The pressure, tension, and Ma=m a . (A1.2)
null axes can be displayed in the familiar focal mechanism a is called eigenvector of M and m is the corresponding
plot (fault plane solution). For general seismic sources, we eigenvalue. For calculating the eigenvalues and eigenvec-
can decompose the seismic moment tensor. First, we can tors of a given moment tensor (solving the eigenvalue
separate out the isotropic component which describes the problem), we transform (A1.2)
volume change in the source. The leftover part of the
moment tensor has, in general, three nonvanishing (M — m I) a = 0 , (A1.3)
eigenvalues. This deviatoric part of the moment tensor where I is the identity matrix. (A1.3) is a system of 3
can be decomposed into a number of simple combinations simultaneous homogeneous linear equations in a k . Non-
of equivalent forces. Obviously, there is no unique trivial solutions are found by solving the secular equation
moment tensor decomposition, i.e. unique model of (characteristic polynomial)
equivalent forces. We outlined methods of determining
moment tensor elements from observations, indicating det(M — m I) = 0 , (A1.4)
that recording noise as well as systematic errors due to an where "det' means the determinant. (A1.4) is a polyno-
insufficient knowledge of the Green's functions can intro- mial of third degree. It has three real roots, i.e. eigen-
duce errors into the moment tensor elements. This sug- values, since the moment tensor is real and symmetric
gests caution when apparent non-double couple sources (Faddeeva, 1959). Substituting each eigenvalue m 1 into
result from the inversion. (A1.3) gives the corresponding eigenvector a,. The eigen-
Randall and Knopoff (1970), Gilbert and Dziewonski vectors are orthogonal. Multiplying each eigenvector by
(1975), Dziewonski et al. (1981), Kanamori and Given its inverse norm, we get the orthonormal eigenvectors,
(1981, 1982), Dziewonski and Woodhouse (1983b), Giar- renaming them as a,:
dini (1984), and Scott and Kanamori (1985) reported that
a, a j = 6, 1 . (A1.5)
some seismic sources cannot be described by a pure double
couple. One explanation is that some fault planes show a Knowing the eigenvectors, we can diagonalize M (princi-
complex geometry (Dziewonski and Woodhouse, 1983b). pal axis transformation). Let A be the matrix whose
Another explanation can be that some sources deviate columns are the orthonormal eigenvectors of M. From
from the model of a sudden shear dislocation; they can be (A1.5), we see that A is orthogonal : A T = A 1 . Then,
-

due to a rapidly propagating phase transition (Knopoff A T M A = m, where m is diagonal, consisting of the
and Randall, 1970; Dziewonski and Woodhouse, 1983b). eigenvalues of M.
However, the simple inversion experiment in Appendix V
pointed out that the deviation from a pure double couple We represent a dyadic by writing two vectors a and
can also be due to the presence of noise in the data b together as ab (see Appendix A in Ben-Menahem and
(Stump and Johnson, 1977; Patton and Aki, 1979; Pat- Singh, 1981). These two vectors forming the dyadic are
ton, 1980; Ward, 1980b; Wallace, 1985; O'Connell and not operating on each other, but define a 3X3 matrix. Let
Johnson, 1988). j, and k be unit vectors along a right handed Cartesian
coordinate system. The dyadic ab is defined as

45
 Jost and Herrmann
ab = (a z i + a y j + a z k) (b z i + b y j + b z k) v =-- sin S sin (I) e e + sin 6 cos I + cos 8 e r . (A2.3)
= iia z b x + ija z b y + ika z b, These two equations are identical to (4.122) in Ben-
Menahem and Singh (1981). The differences in sign com-
+ jia y b z + ha y by + jka y b z
pared to (15) and (16) can be fully explained by noting
+ kia z b x + kj a, b y + kka z b z (A1.6) that e r = - e 0 , = e y , and e s = - e x ; in other words,
e s , e o , and e r are unit vectors towards south, east, and
a z 1), a z b y a z 6, up, respectively (defining a right handed system).
= a y br a y b y ay 6,
a, b x a z by a z 6, APPENDIX III
In order to gain some experience in the relationships
For a = b, we get (26). The multiplication of a vector c between a moment tensor and a fault plane solution, three
from the left is simple focal mechanisms are discussed in detail. These
c • ij = [(ic zjc ykc z ) i]j = c z j (A1.7) will be vertical strike slip, 45 degree dip slip, and vertical
dip slip faults. These three fault plane solutions form a
If the dyadic is symmetric, the multiplication of any vec- complete set : The seismic radiation from a dislocation on
tor with the dyadic is commutative, i.e. ab ba. In a plane dipping an arbitrary angle (but striking north-
general, we can understand a dyadic as a tensor of second south) can be expressed as a linear combination of these
rank. By a proper choice of the coordinate system, a sym- three solutions (Burridge et al., 1964; Ben-Menahem and
metric dyadic can always be transformed into diagonal Singh, 1968).
form (principal axis transformation). As an example, we
can rewrite (10) using dyadics (Gilbert, 1973): Vertical strike slip fault
uv vu = tt — pp (A1.8) The following focal mechanism is assumed: (strike) (I)
= 0.5 [(t+p)(t—p) + (t—p)(t+p)] • = 0°, (dip) S = 90°, and (slip) X = 0°. From (15) and
(16), the slip vector on the fault plane is u = (1,0,0) and
the vector normal to the fault plane is v = (0,1,0). The
APPENDIX II moment tensor can be determined from (18).
The PDE monthly listings (NEIS) routinely publish
centroid moment tensor solutions in the notation of the 0 Mo 0
normal mode theory following Dziewonski et al. (1981). M= Mo 0 0 (A3.1)
For reference, the spherical moment tensor elements, f „
0 0 0
in the notation of Gilbert and Dziewonski (1975),
Dziewonski et al. (1981), and Dziewonski and Woodhouse The eigenvalues and eigenvectors of this tensor
(1983a) are compared to the moment tensor elements as (Example 4.6.1 in Arfken, 1985, see also Appendix I) are
given in (18) following Aki and Richards (1980). shown in Table A.1 (The components of the eigenvectors
f 1= Mrr = mzz are north, east, and down).

12 = M ee = Mzx TABLE A.1

f 3 = M OO = MYY
EIGENVALUE EIGENVECTOR
(A2.1) 0 ( 0.0000, 0.0000,-1.0000)
f 4 = Mr e = Mxz
Mo ( 0.7071, 0.7071, 0.0000)
/ 5 = Mr(k = Myz -M, (-0.7071, 0.7071, 0.0000)
f 6 = Me0 = Mxy
The eigenvector b corresponding to the eigenvalue
where (r,19,0) are the geographical coordinates at the zero is the null-axis, the eigenvector t corresponding to
source. 8 is the colatitude (0 = 0 at the north pole) and the positive eigenvalue gives the tension axis, T, and the
ck is the longitude of the point source. The sign of the off- eigenvector p corresponding to the negative eigenvalue
diagonal moment tensor eleinents depend on the orienta- gives the pressure axis, P, of a focal mechanism.
tion of the coordinate system (Fitch et al., 1981). But the The focal mechanism is obtained by using (7)-(14)
eigenvalues and the eigenvectors of the moment tensor in (Herrmann, 1975). For the trend and plunge (in degrees)
the formulation of (18) or (A2.1) are identical, which can of the X-, Y-, null-, T-, and P-axes, we get (90, 0), (180,
be shown by comparing the solutions to the secular equa- 0), (270, 90), (45, 0), and (135, 0), respectively. The trend
tion (Appendix I). This result is expected since physical of both the P and T axes can be shifted by 180° (Figure
laws should not depend on the choice of the reference A.1a); i.e. the P-axis can also be described by (315°, 0°)
frame. The slip vector u and fault normal v are and the T-axis by (225°, 0°). This ambiguity can be fol-
(Dziewonski and Woodhouse, 1983a) lowed through to the moment tensor: The sign of an
u = ( — cos X cos 4> — cos 6 sin X sin (I) ) e s
. eigenvector is not constrained by the solution of the eigen-
value problem (Arfken, 1985). However, any choice of
▪ u ( cos X sin (I) — cos S sin X cos (I) ) e o(A2.2)
—
sign leads to the same focal mechanism.
▪ u sin X sin (5 e,. ,
—

and
46
 A Student's Guide to and Review of Moment Tensors

N b

Fig. A.1. Focal mechanisms of a vertical strike slip fault (strike = 0°, dip
= 90°, slip = 0° ), (a), a 45 degree dip slip fault (strike = 0°, dip
= 45°, slip = 90°), (b), and a vertical dip slip fault (strike = 0 ° ,
dip = 90°, slip = 90° ), (c) (Appendix III).

45 degree dip slip fault TABLE A.3

The following focal mechanism is assumed: (strike) 43 EIGENVALUE EIGENVECTOR
= 0°, (dip) 6 = 45°, and (slip) X = 90°. From (15) and
0 (-1.0000, 0.0000, 0.0000)
(16), u = (0,-0.7071,-0.7071) and v = (0,0.7071,-0.7071).
Mo ( 0.0000, 0.7071,-0.7071)
The moment tensor is calculated from (18).
( 0.0000, 0.7071, 0.7071)
0 0 0
M = 0 -M 0 0 (A3.2) The fault plane solution is obtained from (7)-(14)
(Herrmann, 1975). For the trend and plunge (in degrees)
00 Mo of the X-, Y-, null-, T-, and P-axes, we get (0, 90), (90, 0),
The corresponding eigenvalues and eigenvectors are (180, 0), (270, 45), and (90, 45), respectively. The trend
of the null axis can be shifted by 180° (Figure A.1c) to
shown in Table A.2.
(360 ° , 0° ).
TABLE A.2
APPENDIX IV
EIGENVALUE EIGENVECTOR In the following, examples of the five methods of
0 (-1, 0, 0) moment tensor decomposition are presented.
Mo ( 0, 0,-1) In order to construct a moment tensor that does not
- Mn ( 0, 1, 0) lead to a simple double couple mechanism, let
The fault plane solution is obtained from (7)-(14) 1 00
(Herrmann, 1975). For the trend and plunge (in degrees) 1\4 1 = 1 0 1 0 (A4.1)
0 01
of the X-, Y-, null-, T-, and P-axes, we get (00, 45), (270, 0 10 I
45), (360, 0), (180, 90), and (270, 0), respectively. The
M2 = 6 1 0 0 (A4.2)
trend of the P and null axes can be shifted by 180° (Fig- 0 00
ure A.1b) to (90 ° , 0 ° ) and (180 ° , 0°), respectively. 0 0 0
Vertical dip slip fault M 3 = 3 0 -1 0 (A4.3)
0 0 1
The following focal mechanism is assumed: (strike) (I)
= 0°, (dip) 6 = 90°, and (slip) X = 90°. From (15) and 0 0 0
(16), u = (0,0,-1) and v = (0,1,0). The moment tensor is M4 =1 0 0 -1 (A4.4)
calculated from (18). 0 -1 0
{0 0 0 The first moment tensor represents an explosion, the
M= 0 0 -Mo (A3.3) others are the familiar ones from Appendix III, represent-
0 -M0 0 ing a vertical strike-slip, a 45 degree dip-slip, and a verti-
cal dip-slip fault, respectively. All four moment tensors
The corresponding eigenvalues and eigenvectors are are superimposed in order to describe a complex source
shown in Table A.3. that is dominated by a vertical strike slip mechanism.
47
 Jost and Herrmann

Fig. A.2. Focal mechanisms of the double couples from the moment tensor
decomposition (Appendix IV). (a) major couple of the moment ten-
sor in (A4.5), elementary moment tensor EMT3 in (A4.6), and
second term on the RHS of (A4.9) (strike = 355°, dip = 80°, slip
= 16° ), (b) elementary moment tensor EMT2 in (A4.6) (strike =
125°, dip = 63°, slip = -95° ), (c) elementary moment tensor
EMT4 in (A4.6) (strike = 199°, dip = 44°, slip = 63°).
The result is the eigenvector (-0.2938, -0.1397, -0.9456) with eigenvalue
1 6 0 2.9 as the null-axis. The fault plane solution of the major
M= 6 -2 -1 (A4.5) double couple gives for the X-, Y-, null-, T-, and P-axes
0 -1 4 (in degrees): (172, 16), (265, 10), (25, 71), (219, 18), and
(128, 4), respectively (Figure A.2a). The major double
Table A.4 shows the eigenvalues of (A4.5) and the
couple gives a good estimate of the major contribution to
corresponding eigenvectors, which are the principal axes of
the faulting which is predominantly strike slip (compare
M. Figures A.la and A.2a).
TABLE A.4 Next, the moment tensor in (A4.5) is decomposed
into an isotropic part and three double couples following
EIGENVALUE EIGENVECTOR (29) which is evaluated by using (26) together with the
3.8523 (-0.2938, -0.1397, -0.9456) data in Table A.4. The numbering of the eigenvalues and
5.8904 ( 0.7352, 0.5992, -0.3170) eigenvectors in (29) follows the columns of Table A.4, but
-6.7427 ( 0.6109, -0.7883, -0.0734) that is not relevant to the solution. The calculation gives
{1 0 0 0.4542 0.3995 -0.51091
The sum of the eigenvalues is equal to 3, which is the M= 1 0 1 0+ 0.6794 0.3995 0.3396-0.3220
expected value for the sum of the eigenvalues of (A4.1),
001 -0.5109 -0.3220 -0.7936
describing an explosion.
[ 0 1673 0.9221 -0.1882
In order to calculate the deviatoric part of the given
+ 4.2110 0 9221 -0.2623 -0.2477 (A4.6)
moment tensor, the isotropic part is removed by subtract-
-0.1882 -0.2477 0.0951
ing one third of the trace of (A4.5) from each diagonal ele-
ment. The solution to the corresponding eigenvalue prob- -0.2869 0.5226 0.3227
lem leads to the same eigenvectors as above. This + 3.5316 0.5226 -0.6019 0.0743 .
indicates that the principal axes of the complete moment 0.3227 0.0743 0.8887
tensor are the same as the principal axes of the
corresponding deviatoric tensor. The deviatoric eigen- This equation is identical to (A4.5). The first ele-
values are 2.8523, 4.8904, and -7.7427 in the order of mentary moment tensor (EMT1) on the RHS of (A4.6)
Table A.4 (see (24)). From (38), e = 0.37, i.e. the given describes the explosion (isotropic component of (A4.5))
moment tensor has a double couple component of 26 % and is identical to (A4.1). The last three elementary
and a CLVD component of 74 %. moment tensors on the RHS (EMT2, EMT3, EMT4,
respectively) represent pure double couple sources since
For the determination of the major couple from (32),
the eigenvalues of each tensor is 0 and ± 1. The three ele-
we identify the eigenvector (0.6109, -0.7883, -0.0734)
mentary moment tensors have identical eigenvectors
corresponding to the deviatoric eigenvalue of -7.7 as the
which are the same vectors as shown in Table A.4.
P-axis, the eigenvector (0.7352, 0.5992, -0.3170)
However, the correlation between eigenvector and eigen-
corresponding to the eigenvalue of 4.9 as the T-axis, and
value (i.e. null-, P-, and T-axes) varies. Note that replac-
48
 A Student's Guide to and Review of Moment Tensors
ing Mk) by -Mkt switches the sign of the eigenvalues moments of the elementary moment tensors are given by
(leaving the eigenvectors untouched), which is identical to the coefficients in (A4.7) which are identical to the devia-
interchanging the P- and T-axes. toric eigenvalues of (A4.5). This exercise demonstrated
From the eigenvalues and the eigenvectors, the fault that vector dipoles are related to the eigenvectors scaled
plane solution for each elementary moment tensor is by the corresponding eigenvalue of a given moment ten-
determined and shown in Table A.S. sor, which makes an evaluation of (A4.7) obsolete.
Alternatively, the moment tensor in equation (A4.5)
TABLE A.5 can be decomposed into an isotropic part and three com-
pensated linear vector dipoles using (30).
EMT2 EMT2 EMT3 EMT3 EMT4 EMT4
AXIS TRD PLG TRD PLG TRD PLG I. 0 0 -0.7411 0.1231 0.83361
(deg.) (deg.) (deg.) (deg.) (deg.) (deg.) M = 1 0 1 0 + 1.2841 0.12310.8336 -0.9415 0.3963
X 36 26 172 16 324 38 [ 0 0 1 0.3963 1.6823
Y 226 63 265 10 109 46 [ 0 6215 1.3216 -0.69911
NULL 128 4 25 71 219 18 + 1.9635 1 3216 0.0773 -0.5697 (A4.8)
T 219 18 219 18 25 71 - 0.6991 -0.5697 -0.6985
P 25 71 128 4 128 4 [ 0.1196 -1.4447 -0.1345
The focal mechanisms corresponding to EMT2 - - 2.2476 -1.4447 0.8642 0.1734 .
EMT4 are shown in Figures A.2b, A.2a, and A.2c, respec- - 0.1345 0.1734 -0.9838
tively. Note that the positions of the axes remain fixed in
these figures, where only the correlation to the eigenvalues This equation is identical to (A4.5). The seismic
changes. The fault plane solution representing the third moments of the elementary moment tensors are given by
elementary moment tensor EMT3 in (A4.6) is identical to the product of the respective coefficient and . The
the fault plane solution of the major couple (Figure A.2a). eigenvalues and eigenvectors for (A4.8) are shown in
Notice that this solution has also the largest coefficient in Table A.6, using the same notation as above. Note that
(A4.6). This solution is an approximation to the major the eigenvectors are identical to those in Table A.4.
contributor of the moment tensor (Figure A.la and
(A4.2)). However, the other fault plane solutions (Figure TABLE A.6
A.2b and A.2c) do not show similarities to the input fault
mechanisms (Figure A.lb and A.1c). EMT EIGENVALUE EIGENVECTOR
The seismic moments of the elementary moment ten- 2 -1 ( 0.6109,-0.7883,-0.0734)
sors are given by the coefficients in (A4.6). The sum of the -1 ( 0.7352, 0.5992,-0.3170)
seismic moments of the elementary moment tensors is 1.4 2 (-0.2938,-0 1397,-0.9456)
times larger than the seismic moment of the composite 3 -1 (-0.2938,-0.1397,-0.9456)
moment tensor in (A4.5). -1 ( 0.6109,-0.7883,-0.0734)
Next, the moment tensor in equation (A4.5) is 2 ( 0.7352, 0.5992,-0.3170)
decomposed into an isotropic part and three vector dipoles 4 -1 ( 0.7352, 0.5992,-0.3170)
following (27) which is evaluated by using (26) together -1 (-0.2938,-0.1397,-0.9456)
with Table A.4. 2 ( 0.6109,-0.7883,-0.0734)
10 0 0.0863 0.0410 0.2779 Next, the moment tensor in (A4.5) is decomposed
M=1 0 1.-0 + 2.8523 D.0410 0.0195 0.1321 into an isotropic part, a double couple and CLVD follow-
00 1 0.2779 0.1321 0.8941 ing (37), where c = F = 0.3684.
100 0.1673 0.9221 0.1882
-

0.5405 0.4405 -0.23301 M= 1 0 1 0+ 2.0379 0.9221 -0.2623 -0.2477

+ 4.8904 0.4405 0.3591 -0.1899 (A4.7) 001 -0.1882 -0.2477 0.0951
-0.2330 -0.1899 0.1005 0 1196 -1.4447 -0.1345
0.3732 -0.4816 -0.0448 - 2.8523 -1.4447
1.4447 0.8642 0.1734 . (A4.9)
- 7.7427 -0.4816 0.6214 0.0578 . [
-0.1345 0.1734 -0.9838
-0.0448 0.0578 0.0054 This equation is identical to (A4.5). Notice that the
This equation is identical to (A4.5). In the notation second term on the RHS corresponds to EMT3 in (A4.6)
used above, each of the elementary moment tensors and to the major double couple. These three tensors all
EMT2, EMT3, and EMT4 have two eigenvalues equal to have the same fault plane solution (Figure A.2a). The
zero, the third one equals one. EMT2 is represented by the third term in (A4.9) corresponds to EMT4 in (A4.8),
eigenvector (0.2938, 0.1397, 0.9456), EMT3 by (-0.7352, representing a CLVD (see Table A.6).
-0.5992, 0.3170), and EMT4 by (0.6109, -0.7883, -0.0734). As final remark, let's consider the decomposition
These vector dipoles are mutually orthonormal. Notice equations (27), (29), and (30) for a simple double couple
that these vector dipoles are identical to the eigenvectors source (e = 0 ), e.g. let m 1 = - m 2 = 1, and m 3 = 0.
of equation (A4.5), which are the principal axes of the ten- Then, M = a l a i a 2 a 2 for all three equations. That is,
-

sor (Table A.4). EMT2 represents the null-, EMT3 the we get one pure double couple out of the decomposition.
tension-, and EMT4 the pressure axis. The seismic For a CLVD (e = 0.5), let's assume that m 1 = m 2 = 1, -

49
 ▪
Jost and Herrmann
and m 3 = 2. Then all three formulas give M = 2 a 3a3 - regroup and assume the presence of an isotropic com-
a l a i - a 2 a 2 , representing one CLVD. ponent (ZEP *0, REP *0):

APPENDIX V ZSS ZDD + ZEP

dz(r = Mzxcos(2 az) —
2 3
In this section, we relate the Green's functions in the
formulation of Herrmann and Wang (1985) to a moment Alvy {—ZSS ZDD ZEP
tensor inversion scheme. Following the theory given by cos(2 az) ±
2 3
Herrmann and Wang (1985), the Fourier transformed dis-
placements at the free surface at the distance r from the {ZEP
origin due to an arbitrarily oriented double couple ▪ --zz 3
without moment is
dz (r ,z;:),cu) = ZSS A 1 ZDS A 2 ZDD A 3 ▪ Mxy [ZSS sin(2az )] (A5.4)
dr (r = RSS A 1 RDS A 2 RDD A 3 (A5.1)
d o (r ,z),c4)) = TSS A4 TDS A5 , ▪ M„ [ZDS cos(az )]
where dz is the vertical displacement (positive upward), dr
is the radial displacement, and d 15 is the tangential dis- ▪ Myz [ZDS sin(az))]
placement (positive in a direction clockwise from north).
The functions ZSS, ZDS, ZDD, RSS, RDS, RDD, TSS,
and TDS together with ZEP and REP are the ten Green's
functions required to calculate a wave field due to an arbi- [RSS RDD + REP I
dr (r,z,w) = Mxx cos(2az)
(2az)
trary point dislocation source or point explosion buried in 2 2 3
a plane layered medium (Wang and Herrmann, 1980;
Herrmann and Wang, 1985). As before, let u=(u x ,uy ,u z ) 4_ Ain RSS
[ — RDD REP
cos(2az) +
and v=(vz ,vy ,v,) be the dislocation vector and vector 2 3
normal to the fault plane, respectively. Note that (15) and ii „. [REP I
(16) are identical to the formulation used by Herrmann + '-. zz 3
and Wang (1985), where our u equals their f and our v
equals their n (1 = x-axis, 2 = y-axis, 3 = z-axis). Then
▪ Mxy [RSS sin(2az) (A5.5) ]

A i =(u x vz —u y vy ) cos(2az )+(u x vy d-u y vz ) sin(2 az)

A 2 =(u x vz 1-u z vx ) cos( az )+(u y vz +u z vy ) sin(az)
-I- Mx, [ RDS cos(az)
A 3 =u, v z (A5.2)
A 4 4u, vz —u y vy ) sin(2az )—(u x vy -Fu y vz ) cos(2az ) ▪ Myz [RDS sin(az))]
A 5 =(u x vz ±u z vz ) sin(az )—(u y +u z vy ) cos(az ) ,
where az is the azimuth of observation. Equivalently,
1 [T2S sin(2 az ) I
A 1 --= --2-(Mrx Myy) cos(2az )+Mxy sin(2az )
• d ,k(r ,z = Mxx

A 2= Mx, cos( az )+Myz sin( az ) [—T2SS

MYY sin(2az )1
A 3= --12— (Mxx +Myy) (A5.3)

A 4= l(Mxx—Myy) sin(2 az )—Mxy cos(2 az) ▪ Mxy [—TSS cos(2az)] (A5.6)

A 5= -Myz cos(az)+Mx , sin(az)-.
▪ Mxz [TDS sin(az))]
These equations are identical to (A5.2) which can be
proven by using (18) together with (15) and (16). Note
that the coefficients given in (A5.3) agree with the + Myz [—TDS cos(az)] .
moment tensor elements as defined by Aki and Richards
(1980; (A5.3) differs in sign with the coefficients of Langs-
ton (1981) due to conventions on displacements and Equations (A5.4), (A5.5), and (A5.6) each set up a
Green's functions). moment tensor inversion scheme. Equations (A5.4) and
(A5.5) are formulated for the general case where the inver-
Note that either definition of the coefficients of the sion expects a moment tensor that is a composition of an
Green's functions can be used for the calculation of the isotropic part and a deviatoric part. An inversion based
displacement at the free surface, depending on whether on transverse data, (A5.6), cannot resolve Mzz . In such a
the focal mechanism or the moment tensor is given. Here, case, we assume that the moment tensor is purely devia-
equations (A5.3) and (A5.1) are used in order to develop toric and constrain Mzz - (M.= Myy). The same con-
an inversion scheme for the moment tensor elements. We straint can be applied to (A5.4) and (A5.5) in the case of
50
 A Student's Guide to and Review of Moment Tensors
an inversion that looks for a pure deviatoric moment ten-
sor (set formally ZEP = REP 0 in (A5.4) and (A5.5), ZSS
Dziewonski et al. , 1981).
From the last three equations we see that the
observed displacement at the free surface is a linear com-
bination of the station specific Green's functions, within
the square brackets, with the moment tensor elements as
scalar multipliers. We also note that if the source time 2.388E 09
-

function is known and a point source approximation is

acceptable, the moment tensor elements are independent
of frequency (linear inversion) and similar equations arise ZDD
relating observed time histories to temporal Green's func-
tions within the square brackets.
Next, we performed a simple moment tensor inver-
sion using the vertical component of synthetic teleseismic
P-wave first motion peak amplitudes as suggested by
7.323E 09
-

Stump and Johnson (1977). We assumed a pure devia-

toric source (ZEP = 0 in (A5.4)).
Let az i , azn be azimuths of n different stations. ZDS
Then the expressions in the square brackets of (A5.4)
define components of a matrix as a, i (azi ) , a, 5 (az,) for
the i-th azimuth. A system of linear equation arises:

dz (az i ) a ii(azi) • • a 15 ( az i )
Mxx
6. 402E-09
Myy
Mxy (A5.7)
Fig. A.3. Synthetic Green's functions ZSS, ZDD, and ZDS
Mxz (Herrmann and Wang, 1985) for a half-space model
dz (az„) a ni (azn ) . . a n5 (azn ) Myz ( V = 8 km/sec, V8 = 4.6 km/sec, p = 3.3
g/cm 3 , h = 30 km, t * = 0.7) calculated by using
For observations at more than 5 distinct azimuths, the Haskell formalism. The time window ranges
the system (A5.7) is overdetermined. The solution can be from 4.0 to 55.1 sec (dt = 0.05 sec). Maximum
reached by the classical least squares approach. The five amplitudes are in cm (Appendix V).
moment tensor elements can be determined by using the
numerical stable singular value decomposition. We
imposed the deviatoric constraint M20 = (Mxx M yy)*
-

Hence the inversion gives a purely deviatoric source.

However, we were not constraining one eigenvalue as zero
(double couple), letting the inversion tell us about double
couple and CLVD components. The eigenvalues and
eigenvectors can be calculated using the Householder
transformation with further QL decomposition. The
implementation of these numerical concepts was done
using code by Press et al. (1987).
In the following, some results of inverting synthetic
data are presented. First, Green's functions were calcu-
lated using a Haskell formalism for a simple half-space
model ( V p 8 km/sec , = 4.6 km/sec and p = 3.3
g/cm 3 , h = 30 km ). Figure A.3 shows the three basic
Green's functions ZSS, ZDD, and ZDS. The assumed
focal mechanism (Figure A.4: strike = 180°, dip = 40°,
slip = 110°) is the same as used by Herrmann (1975, Fig-
ure 2). Teleseismic P-wave first motions were synthesized
at 12 equidistant azimuths (epicentral distance = 50° ).
Note that an instrument response was not included in the
synthetics. Due to the simple model and the fact that all
stations are equidistant from the source, a correction for
anelastic attenuation ( t * = 0.7 ) or geometrical spread-
ing is not required. A correction for an extended source is Fig. A.4. Assumed focal mechanism for the synthetic
not necessary since the moment used is 10 20 dyne-cm and seismograms: strike = 180°, dip = 40°, slip =
the duration of the source time function is 0.2 sec. We 110° (Appendix V).
used (A5.4) for time domain measurements.
51
 Jost and Herrmann

TABLE A.7: RESULTS OF THE MOMENT

TENSOR INVERSION (MAJOR COUPLE)

Case 0 Case I Case II Case III Case IV Case V

M 0 -0.037 -0.050 -0.109 -0.202 0.301

Myy -0.925 -0.902 -0.951 -0.966 -1.023 -1.091

M„ 0.925 0.939 1.002 1.075 1.225 0.791

Msy -0.220 -0.199 -0.200 -0.194 -0.176 0.257

M„ -0.262 -0.262 -0.260 -0.264 -0.257 -0.172

M5, -0.163 -0.168 -0.162 -0.168 -0.156 -0.324

EV(NULL) 0.00 -0.04 -0.05 -0.11 -0.20 0.26

EV(T) 1.00 1.01 1.07 1.14 1.28 0.92
EV(P) -1.00 -0.97 -1.02 -1.03 -1.08 -1.18

% of DC 100 92 90 81 69 56

% of CLVD 0 8 10 19 31 44

Me 1.00 0.99 1.04 1.09 1.19 1.07

STRIKE 180.0 179.5 179.2 177.8 176.4 211.8

DIP 40.0 39.7 40.1 39.9 40.5 40.8

SLIP 110.0 109.0 107.8 106.1 103.2 123.1

STRIKE 334.6 335.4 336.5 337.2 339.3 351.1

DIP 52.8 52.9 52.2 51.9 50.8 56.8

SLIP 74.0 74.9 75.6 77.0 79.0 64.8

T (TRD) 192.7 194.9 194.7 196.8 198.2 209.7

T (PLG) 75.6 76.2 77.1 78.1 80.0 67.3

P (TRD) 75.9 76.1 76.7 76.4 77.1 98.8

P (PLG) 6.6 6.7 6.2 6.1 5.2 8.5

For Case 0, moment tensor elements are calculated from (18) assuming a double couple source
(strike 180 ° , dip = 40°, slip = 110°). The eigenvalues of the moment tensor corresponding to
the null-, T-, and P-axes are shown as EV(NULL), EV(T), and EV(P), respectively. Equation (38)
is used to determine the percentage of double couple or CLVD from the eigenvalues of the moment
tensor. The seismic moment is calculated using (20). The orientation of the fault plane and auxili-
ary plane is given together with the trend and plunge of the T- and P-axes (Herrmann, 1975).
Cases I - IV are for additive pseudo-random noise ( 0 %, 14 %, 28 %, and 56 %, respectively) in
the synthetic seismograms at 12 different azimuths. Case V assumes that one of the 12 seismo-
grams has a reversed polarity (0 % pseudo random noise).

Table A.7 displays the inversion results for the major graphs. Hence it was assumed that one of the 12 seismo-
couple. The moment tensor elements, the percentage of grams of Case I had a wrong polarity.
double couple and CLVD, the seismic moment, and the The theoretical focal parameters (Case 0) agree
focal mechanism parameters-are shown. For Case 0, the within the measurement errors with the observed ones
moment tensor elements were calculated from the given (Case I). This justifies the technique. The effect of noise is
fault plane solution and (18). Next, three experiments to severely distort the moment tensor elements. The iso-
were performed: 1.) synthetic seismograms were calculated tropic moment tensor components seem to be more sensi-
using the Haskell method (Case I). Figure A.5a shows the tive to noise than the deviatoric ones. Notice that the
vertical component of a synthetic seismogram at an moment tensor gains a contribution of a CLVD due to the
azimuth of 0 degrees. 2.) Different amounts of pseudo- noise. The percentage of CLVD versus double couple
random noise were added to the synthetic seismograms increases with increasing noise. The effect of random
calculated in Case I with amplitudes of ± 0.25 X 10 -9 cm noise on the fault plane solution that is derived from the
(Case II, Figure A.5b), ± 0.5 X 10 -9 cm (Case III, Figure moment tensor elements is minor; i.e. the fault plane
A.5c), and ± 1.0 X 10 -9 cm (Case IV, Figure A.5d). solution for the major double couple is very close to the
Averaged over the 12 azimuths, these noise levels original focal mechanism. However, with increasing noise,
correspond to 14 %, 28 %, and 56 % pseudo-random the fault plane solution deteriorates. 8 % polarization
additive noise, respectively. 3.) The final experiment errors in otherwise perfect data lead to worse results than
(Case V) relates to possible polarity errors of seismo- 56 % additive random noise (Case IV). A doubling of the
52
 A Student's Guide to and Review of Moment Tensors
REFERENCES
2. 267E -09 a
Agnew, D., J. Berger, R. Buland, W. Farrell, and F. Gil-
bert (1976). International deployment of
accelerometers: a network for very long-period
seismology, EOS, Trans. Am. Geophys. Union 57,
180-188.

Aki, K. and P. G. Richards (1980). Quantitative Seismol-

ogy: Theory and Methods, W. H. Freeman and Co.,
New York, San Francisco, 932 pp.

Arfken, G. (1985). Mathematical Methods for Physicists,

2. 413E-09 b 3rd ed., Academic Press Inc., Orlando, Florida, 985
pp.

Backus, G. E. (1977a). Interpreting the seismic glut

moments of total degree two or less, Geophys. J. R.
Astr. Soc. 51, 1-25.

Backus, G. E. (1977b). Seismic sources with observable

glut moments of spatial degree two, Geophys. J. R.
Astr. Soc. 51, 27-45.

Backus, G. and M. Mulcahy (1976). Moment tensors and

2.604E-09 C other phenomenological descriptions of seismic
sources - I. continuous displacements, Geophys. J.
R. Astr. Soc. 46, 341-361.

Ben-Menahem, A. and S. J. Singh (1968). Eigenvector

expansions of Green's dyads with applications to
geophysical theory, Geophys. J. R. Astr. Soc. 16,
417-452.

Ben-Menahem, A. and S. J. Singh (1981). Seismic Waves

and Sources, Springer Verlag, New York, 1108 pp.
2.984E-09 d Bullen, K. E. and B. A. Bolt (1985). An Introduction to
the Theory of Seismology, 4th ed., Cambridge
University Press, Cambridge, 499 pp.

Burdick, L. J. and G. R. Mellman (1976). Inversion of the

body waves from the Borrego Mountain earthquake
to the source mechanism, Bull. Seism. Soc. Am.
66, 1485-1499.

Burridge, R., E. R. Lapwood, and L. Knopoff (1964). First

motions from seismic sources near a free surface,
Bull. Seism. Soc. Am. 54, 1889-1913.
Fig. A.S. Vertical components of synthetic teleseismic
Christensen, D. H. and L. J. Ruff (1985). Analysis of the
seismograms at 50 degrees and azimuth of 0 ° . The
trade-off between hypocentral depth and source
time window ranges from 4.0 to 29.6 sec (dt =
time function, Bull. Seism. Soc. Am. 75, 1637-
0.05 sec). Maximum amplitudes are in cm. (a) No
1656.
pseudo-random noise added; (b) - (d) pseudo-
random noise is added with amplitudes of
Claerbout, J. F. and F. Muir (1973). Robust modeling
±0.25 X 10 -9 cm, ±0.50 X 10 -9 cm, ±1.0 X 10 -9 with erratic data, Geophysics 38, 826-844.
cm, respectively (Appendix V).
Dziewonski, A. M. and F. Gilbert (1974). Temporal varia-
tion of the seismic moment tensor and the evidence
polarization errors gives meaningless results. Due to the of precursive compression for two deep earth-
setup of the experiment, a minor couple would be a pure quakes, Nature 247, 185-188.
artifact of the noise.
53
 Jost and Herrmann

Dziewonski, A. M. and J. H. Woodhouse (1983a). Studies Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse
of the seismic source using normal-mode theory, in (1985d). Centroid-moment tensor solutions for
Earthquakes: Observation, Theory and Interpreta- January March 1985, Phys. Earth Planet. Interiors
-

tion, H. Kanamori and E. Boschi, Editors, North- 40, 249-258.

Holland, Amsterdam, New York, Oxford, 608 pp.
Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse
Dziewonski, A. M. and J. H. Woodhouse (1983b). An (1986a). Centroid-moment tensor solutions for
experiment in systematic study of global seismicity: April June 1985, Phys. Earth Planet. Interiors 41,
-

centroid-moment tensor solutions for 201 moderate 215-224.

and large earthquakes of 1981, J. Geophys. Res.
88, 3247-3271. Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse
(1986b). Centroid-moment tensor solutions for
Dziewonski, A. M., T.-A. Chou, and J. H. Woodhouse July September 1985, Phys. Earth Planet. Interiors
-

(1981). Determination of earthquake source param- 42, 205-214.

eters from waveform data for studies of global and
regional seismicity, J. Geophys. Res. 86, 2825- Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse
2852. (1986c). Centroid-moment tensor solutions for
October December 1985, Phys. Earth Planet. Interi-
-

Dziewonski, A. M., A. Friedman, D. Giardini, and J. H. ors 43, 185-195.

Woodhouse (1983a). Global seismicity of 1982:
centroid-moment tensor solutions for 308 earth- Dziewonski, A. M., G. EkstrOm, J. E. Franzen, and J. H.
quakes, Phys. Earth Planet. Interiors 33, 76 90.
- Woodhouse (1987a). Centroid-moment tensor solu-
tions for January-March 1986, Phys. Earth Planet.
Dziewonski, A. M., A. Friedman, and J. H. Woodhouse Interiors 45, 1 10.
-

(1983b). Centroid-moment tensor solutions for

January March 1983, Phys. Earth Planet. Interiors
- Dziewonski, A. M., G. EkstrOm, J. E. Franzen, and J. H.
33, 71-75. Woodhouse (1987b). Global seismicity of 1977:
centroid-moment tensor solutions for 471 earth-
quakes, Phys. Earth Planet. Interiors 45, 11 36.
-
Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse
(1983c). Centroid-moment tensor solutions for
Dziewonski, A. M., G. EkstrOm, J. E. Franzen, and J. H.
April June 1983, Phys. Earth Planet. Interiors 33,
-
Woodhouse (1987c). Centroid-moment tensor solu-
243-249.
tions for July-September 1986, Phys. Earth Planet .

Interiors 46, 305-315.

Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse
(1984a). Centroid-moment tensor solutions for
Dziewonski, A. M., G. EkstrOm, J. E. Franzen, and J. H.
July September 1983, Phys. Earth Planet. Interiors
-
Woodhouse (1987d). Global seismicity of 1978:
34, 1-8. centroid-moment tensor solutions for 512 earth-
quakes Phys. Earth Planet. Interiors 46, 316 342.
-
Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse
(1984b). Centroid-moment tensor solutions for Dziewonski, A. M., G. EkstrOm, J. H. Woodhouse and G.
October December 19$3, Phys. Earth Planet. Interi-
-
Zwart (1987e). Centroid-moment tensor solutions
ors 34, 129-136. for October-December 1986, Phys. Earth Planet .

Interiors 48, 5 17.

-

Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse

(1984c). Centroid-moment tensor solutions for Dziewonski, A. M., G. EkstrOm, J. E. Franzen, and J. H.
January March 1984, Phys. Earth Planet. Interiors
-
Woodhouse (1987f). Global seismicity of 1979:
34, 209-219.
centroid-moment tensor solutions for 524 earth-
quakes Phys. Earth Planet. Interiors 48, 18 46.
-

Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse

(1985a). Centroid-moment tensor solutions for
EkstrOm, G. and A. M. Dziewonski (1985). Centroid-
April June 1984, Phys. Earth Planet. Interiors 37,
-
moment tensor solutions for 35 earthquakes in
87-96. Western North America (1977-1983), Bull. Seism.
Soc. Am. 75, 23-39.
Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse
(1985b). Centroid-moment tensor solutions for EkstrOm, G., A. M. Dziewonski, and J. H. Woodhouse
July September 1984, Phys. Earth Planet. Interiors
-
(1987). Centroid-moment tensor solutions for the
38, 203-213. 51 IASPEI selected earthquakes, 1980-1984, Phys.
Earth Planet. Interiors 47, 62 66.
-

Dziewonski, A. M., J. E. Franzen, and J. H. Woodhouse

(1985c). Centroid-moment tensor solutions for Engdahl, E. R. and H. Kanamori (1980). Determination of
October December 1984, Phys. Earth Planet. Interi-
-
earthquake parameters, EOS, Trans. Am. Geophys.
ors 39, 147-156. Union 61, 60-65.
54
 A Student's Guide to and Review of Moment Tensors

Evison, F. F. (1963). Earthquakes and faults, Bull. Seism. Harkrider, D. G. (1964). Surface waves in multilayered
Soc. Am. 53, 873-891. elastic media, 1. Rayleigh and Love waves from
buried sources in a multilayered elastic half-space,
Faddeeva (1959). Computational Methods of Linear Alge- Bull. Seism. Soc. Am. 54, 627-679.
bra, Dover, New York, 252 pp.
Harkrider, D. G. (1970). Surface waves in multilayered
Fitch, T. J. (1981). Correction and addition to 'Estima- elastic media, 2. Higher mode spectra and spectral
tion of the seismic moment tensor from teleseismic ratios from point sources in plane layered earth
body wave data with applications to intraplate and models, Bull. Seism. Soc. Am. 60, 1937-1987.
mantle earthquakes' by T. J. Fitch, D. W.
McCowan, and M. W. Shields, J. Geophys. Res. Harkrider, D. G. (1976). Potentials and displacements for
86, 9375-9376. two theoretical seismic sources, Geophys. J. R.
Astr. Soc. 47, 97-133.
Fitch, T. J., D. W. McCowan, and M. W. Shields (1980).
Estimation of the seismic moment tensor from Herrmann, R. B. (1975). A student's guide to the use of P
teleseismic body wave data with applications to and S wave data for focal mechanism determina-
intraplate and mantle earthquakes, J. Geophys. tion, Earthquake Notes 46, 29-39.
Res. 85, 3817-3828.
Herrmann, R. B. and C. Y. Wang (1985). A comparison of
Fitch, T. J., R. G. North, and M. W. Shields (1981). synthetic seismograms, Bull. Seism. Soc. Am. 75,
Focal depths and moment tensor representations of 41-56.
shallow earthquakes associated with the great
Sumba earthquake, J. Geophys. Res. 86, 9357- Hirasawa, T. and W. Stauder, S.J. (1965). On the seismic
9374. body waves from a finite moving source, Bull.
Seism. Soc. Am. 55, 237-262.
Geller, R. J. (1976). Body force equivalents for stress-drop
seismic sources, Bull. Seism. Soc. Am. 66, 1801- Honda, H. (1962). Earthquake mechanism and seismic
1804. waves, Geophysical Notes, Geophys. Inst., Fac. of
Science, Tokyo Univ. 15, 1-97.
Giardini, D. (1984). Systematic analysis of deep seismicity:
200 centroid-moment tensor solutions for earth- Kanamori, H. and J. W. Given (1981). Use of long-period
quakes between 1977 and 1980, Geophys. J. R. surface waves for rapid determination of
Astr. Soc. 77, 883-914. earthquake-source parameters, Phys. Earth Planet .

Interiors 27, 8-31.

Gilbert, F. (1970). Excitation of the normal modes of the
earth by earthquake sources, Geophys. J. R. Astr. Kanamori, H. and J. W. Given (1982). Use of long-period
Soc. 22, 223-226. surface waves for rapid determination of earth-
quake source parameters: 2. Preliminary determina-
tion of source mechanisms of large earthquakes (Ms
Gilbert, F. (1973). Derivation of source parameters from > 6.5) in 1980, Phys. Earth Planet. Interiors 30,
low-frequency spectra, Phil. Trans. R. Soc. A 274, 260-268.
369-371.
Kennett, B. L. N. (1983). Seismic Wave Propagation in
Gilbert, F. and R. Buland (1976). An enhanced deconvolu- Stratified Media, Cambridge University Press,
tion procedure for retrieving the seismic moment Cambridge, 342 pp.
tensor from a sparse network, Geophys. J. R. Astr.
Soc. 47, 251-255. Kennett, B. L. N. (1988). Radiation from a moment-
tensor source, in Seismological Algorithms - Com-
Gilbert, F. and A. M. Dziewonski (1975). An application putational Methods and Computer Programs, D. J.
of normal mode theory to the retrieval of struc- Doornbos, Editor, Academic Press, San Diego, Cal-
tural parameters and source mechanisms from
ifornia, 427-441.
seismic spectra, Phil. Trans. R. Soc. A 278, 187-
269. Kikuchi, M. and H. Kanamori (1982). Inversion of com-
plex body waves, Bull. Seism. Soc. Am. 72, 491-
Gomberg, J. S. and T. G. Masters (1988). Waveform
modeling using locked-mode synthetic and 506.
differential seismograms: application to determina-
tion of the structure of Mexico, Geophys. J. 94, Knopoff, L. and M. J. Randall (1970). The compensated
193-218. linear-vector dipole: a possible mechanism for deep
earthquakes, J. Geophys. Res. 75, 4957-4963.
Hagiwara, T. (1958). A note on the theory of the elec-
tromagnetic seismograph, Bull. Earthquake Res. Landisman, M., A. Dziewonski, and Y. Sato (1969).
Inst. Tokyo Univ. 36, 139-164. Recent improvements in the analysis of surface
wave observations, Geophys. J. R. Astr. Soc. 17,
55 369-403.
 Jost and Herrmann

Langston, C. A. (1981). Source inversion of seismic Patton, H. (1980). Reference point equalization method
waveforms: The Koyna, India, earthquakes of 13 for determining the source and path effects of sur-
September 1967, Bull. Seism. Soc. Am. 71, 1-24. face waves, J. Geophys. Res. 85, 821-848.

Langston, C. A. and D. V. Helmberger (1975). A pro- Patton, H. and K. Aki (1979). Bias in the estimate of
cedure for modeling shallow dislocation sources, seismic moment tensor by the linear inversion
Geophys. J. R. Astr. Soc. 42, 117-130. method, Geophys. J. R. Astr. Soc. 59, 479-495.

Lawson, C. H. and R. J. Hanson (1974). Solving Least Press, W. H., B. P. Flannery, S. A. Teukolsky, and W. T.
Squares Problems, Prentice-Hall, Englewood Cliffs, Vetterling (1987). Numerical Recipes: The Art of
New Jersey. Scientific Computing, Cambridge University Press,
Cambridge, 818 pp.
Lay, T., J. W. Given, and H. Kanamori (1982). Long-
period mechanism of the 8 November 1980 Eureka, Randall, M. J. and L. Knopoff (1970). The mechanism at
California, earthquake, Bull. Seism. Soc. Am. 72, the focus of deep earthquakes, J. Geophys. Res.
439-456. 75, 4965-4976.

Lundgren, P. R., E. A. Okal, and S. Stein (1988). Body- Reid, H. F. (1910). Elastic rebound theory, Univ. Calif.
wave deconvolution for variable source parameters; Publ., Bull. Dept. Geol. Sci. 6, 413-433.
application to the 1978 December 6 Kuriles earth-
quake, Geophys. J. 94, 171-180. Romanowicz, B. (1981). Depth resolution of earthquakes
in central Asia by moment tensor inversion of
Maruyama, T. (1964). Statical elastic dislocations in an long-period Rayleigh waves: Effects of phase velo-
infinite and semi-infinite medium, Bull. Earthquake city variations across Eurasia and their calibration,
Res. Inst. 42, 289-368. J. Geophys. Res. 86, 5963-5984.

McCowan, D. W. (1976). Moment tensor representation of Saito, M. (1967). Excitation of free oscillations and surface
surface wave sources, Geophys. J. R. Astr. Soc. waves by a point source in a vertically heterogene-
44, 595-599. ous earth, J. Geophys. Res. 72, 3689-3699.

McCowan, D. W. and R. T. Lacoss (1978). Transfer func- Satake, K. (1985). Effects of station coverage on moment
tions for the seismic research observatory seismo- tensor inversion, Bull. Seism. Soc. Am. 75, 1657-
graph system, Bull. Seism. Soc. Am. 68, 501-512. 1667.
McKenzie, D. P. (1969). The relation between fault plane Scott, D. R. and H. Kanamori (1985). On the consistency
solutions for earthquakes and the directions of the of moment tensor source mechanisms with first-
principal stresses, Bull. Seism. Soc. Am. 59, 591- motion data, Phys. Earth Planet. Interiors 37, 97-
601. 107.
Mendiguren, J. A. (1977). Inversion of surface wave data Silver, P. G. and T. H. Jordan (1982). Optimal estimation
in source mechanism studies, J. Geophys. Res. 82, of scalar seismic moment, Geophys. J. R. Astr. Soc.
889-894. 70, 755-787.

Nakanishi, I. and H. Kanamori (1982). Effects of lateral Sipkin, S. A. (1982). Estimation of earthquake source
heterogeneity and source process time on the linear parameters by the inversion of waveform data:
moment tensor inversion of long-period Rayleigh synthetic waveforms, Phys. Earth Planet. Interiors
waves, Bull. Seism. Soc. Am. 72, 2063-2080. 30, 242-255.

Nakanishi, I. and H. Kanamori (1984). Source mechanisms Sipkin, S. A. (1986). Interpretation of non-double-couple
of twenty-six large, shallow earthquakes (MS > earthquake mechanisms derived from moment ten-
6.5) during 1980 from P-wave first motion and sor inversion, J. Geophys. Res. 91, 531-547.
long-period Rayleigh wave data, Bull. Seism. Soc.
Am. 74, 805-818. Sipkin, S. A. (1987). Moment tensor solutions estimated
using optimal filter theory for 51 selected earth-
O'Connell, D. R. H. and L. R. Johnson (1988). Second quakes, 1980-1984, Phys. Earth Planet. Interiors
order moment tensors of microearthquakes at The 47, 67-79.
Geysers geothermal field, California, Bull. Seism.
Soc. Am. 78, 1674-1692. Snieder, R. and B. Romanowicz (1988). A new formalism
for the effect of lateral heterogeneity on normal
Okal, E. A. and R. J. Geller (1979). On the observability modes and surface waves - I: isotropic perturba-
of isotropic seismic sources: the July 31, 1970 tions, perturbations of interfaces and gravitational
Colombian earthquake, Phys. Earth Planet. Interi- perturbations, Geophys. J. 92, 207-222.
ors 18, 176-196.
56
 A Student's Guide to and Review of Moment Tensors

Strelitz, R. A. (1978). Moment tensor inversions and Wallace, T. C., D. V. Helmberger, and G. R. Mellman
source models, Geophys. J. R. Astr. Soc. 52, 359- (1981). A technique for the inversion of regional
364. data in source parameter studies, J. Geophys. Res.
86, 1679-1685.
Strelitz, R. A. (1980). The fate of the downgoing slab: A
study of the moment tensors from body waves of Wang C. Y. and R. B. Herrmann (1980). A numerical
complex deep-focus earthquakes, Phys. Earth study of P-, SV-, and SH-wave generation in a
Planet. Interiors 21, 83-96. plane layered medium, Bull. Seism. Soc. Am. 70,
1015-1036.
Stump, B. W. and L. R. Johnson (1977). The determina-
tion of source properties by the linear inversion of Ward, S. N. (1980a). Body wave calculations using
seismograms, Bull. Seism. Soc. Am. 67, 1489- moment tensor sources in spherically symmetric,
1502. inhomogeneous media, Geophys. J. R. Astr. Soc.
60, 53-66.
Vasco, D. W. and L. R. Johnson (1988). Inversion of
waveforms for extreme source models with an Ward, S. N. (1980b). A technique for the recovery of the
application to the isotropic moment tensor com- seismic moment tensor applied to the Oaxaca, Mex-
ponent, in Regional Studies with Broadband Data, ico earthquake of November 1978, Bull. Seism. Soc.
T. V. McEvilly and L. R. Johnson, Editors, Report Am. 70, 717-734.
No. 1, Air Force Geophysics Laboratory AFGL-
TR-88-0131.
Received June 23, 1988
Wallace, T. C. (1985). A reexamination of the moment Revised October 4, 1988
tensor solutions of the 1980 Mammoth Lakes Accepted November 1, 1988
earthquakes, J. Geophys. Res. 90, 11,171-11,176.

57