Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Geophysical Prospecting, 2009, 57, 169–185 doi: 10.1111/j.1365-2478.2008.00775.

Vertically fractured transversely isotropic media: dimensionality


and deconstruction
Michael A. Schoenberg†
Earth Sciences Division, Lawrence Berkeley National Lab, 1 Cyclotron Road, Berkeley, CA 74720, USA

Received February 2008, revision accepted August 2008

ABSTRACT
A vertically fractured transversely isotropic (VFTI) elastic medium is one in which
any number of sets of vertical aligned fractures (each set has its normal lying in the
horizontal x 1 , x 2 -plane) pervade the medium and the sets of aligned fractures are the
only features of the medium disturbing the axi-symmetry about the x 3 -axis implying
that in the absence of fractures, the background medium is transversely isotropic (TI).
Under the assumptions of long wavelength equivalent medium theory, the compliance
matrix of a fractured medium is the sum of the background medium’s compliance
matrix and a fracture compliance matrix. For sets of parallel rotationally symmetric
fractures (on average), the fracture compliance matrix is dependent on 3 parameters
− its normal and tangential compliance and its strike direction. When one fracture set
is present, the medium is orthorhombic and the analysis is straightforward. When two
(non-orthogonal) or more sets are present, the overall medium is in general elastically
monoclinic; its compliance tensor components are subject to two equalities yielding an
11 parameter monoclinic medium. Constructing a monoclinic VFTI medium with n
embedded vertical fracture sets, requires 5 TI parameters plus 3×n fracture set param-
eters. A deconstruction of such an 11 parameter monoclinic medium involves using
its compliance tensor to find a background transversely isotropic medium and several
sets of vertical fractures which, in the long wavelength limit, will behave exactly as
the original 11 parameter monoclinic medium. A minimal deconstruction, would be
to determine, from the 11 independent components, the transversely isotropic back-
ground (5 parameters) and two fracture sets (specified by 2 × 3 = 6 parameters). Two
of the background TI medium’s compliance matrix components are known immedi-
ately by inspection, leaving nine monoclinic components to be used in the minimal
deconstruction of the VFTI medium. The use of the properties of a TI medium, which
are linear relations on its compliance components, allows the deconstruction to be
reduced to solving a pair of non-linear equations on the orientations of two fracture
sets. A single root yielding a physically meaningful minimum deconstruction yields a
unique minimal representation of the monoclinic medium as a VFTI medium. When
no such root exists, deconstruction requires an additional fracture set and uniqueness
is lost. The boundary between those monoclinic media that have a unique minimal
representation and those that do not is yet to be determined.

INTRODUCTION

† Michael Fractures in subsurface rock are the pathways for fluid flow
A. Schoenberg passed away on 29 August 2008. This nearly
finished contribution to this special issue of Geophysical Prospecting underground and thus intimately linked to the permeability,
is published in his honour. which is a tensor and generally an anisotropic tensor. Thus


C 2009 European Association of Geoscientists & Engineers 169
170 M.A. Schoenberg

the information about fracture networks that can be obtained sound in a fractured rock is reduced, phenomenologically, to
from seismic data is essential for predicting how fluids will be finding the values of a 3 × 3 FCM. This formulation is based
displaced under conditions of fluid extraction or fluid injec- on the assumption that in the elastic compliance matrix, the
tion. It has been observed from outcrops that most fractures excess fracture compliance is at least an order of magnitude
occur as members of a set of aligned, or roughly parallel, frac- larger than any diffraction and multiple scattering effects and
tures. Sets of aligned fractures affect the propagation of long effects due to scattering from intersections that contribute to
wavelength seismic waves in very particular ways because the the fractured medium’s overall compliance but that cannot be
seismic anisotropy of the fractured elastic medium is sensitive included in this parameterization.
to the orientation, distribution and fracture compliance of In subsurface rock, fractures generally occur as a few sets
each set of fractures in a very specific way. Thus, this paper is of aligned fractures, the orientation of which seem to be as-
concerned with first the elastic properties of formations with sociated with past and present effective stress acting on the
multiple sets of aligned fractures and second, with attempting rock. If all orientations of fractures are allowed, it is very dif-
to work backward, that is, to deconstruct the long wavelength ficult, if not impossible, to disentangle the anisotropy due to
elasto-dynamic properties of a rock to find the elastic proper- the fractures, from the elastic compliance of the unfractured
ties of the background rock (in its unfractured state) and the background rock (some results might be obtainable when that
orientation and compliances of the several fracture sets, which unfractured background rock is isotropic, this being a subject
contribute to the overall seismic anisotropy of the medium. for further study). However, an important case occurs when
The fundamental principle on which this paper is based is the fracture sets have the following assumed properties: a)
that the compliance tensor of a homogeneous elastic medium all sets of aligned fractures are vertical, so each set’s orienta-
that is equivalent, in the long wavelength limit, to a back- tion is fixed by an angle θ between the horizontal 1-axis and
ground medium with one or more sets of aligned fractures the normal to the fractures (equivalently between the 2-axis
(which below will be merely called the compliance tensor of and the strike direction) and b) all sets are, on average, rota-
the fractured medium) is, to 0th order, composed of the com- tionally symmetric. These assumed properties of the fracture
pliance tensor of the background medium, denoted by sub- sets, verticality and rotational symmetry will be understood
script b, plus an excess fracture compliance tensor, denoted by in the analysis in the rest of this paper. To the extent that
subscript f that is itself a sum of fracture compliance tensors, the fracture behaviour is linearly elastic, each such fracture
each of which being associated with a single one of the sets set is specifiable by a pair of real positive numbers, a tan-
of aligned fractures in the medium. The operative words here gential compliance, denoted by B T and a normal compliance,
are ‘plus’ and ‘sum’; this property of additivity in the com- denoted by B N . [Aside: previously I have used η T,N for an
pliance domain (Hood 1989) is the reason that this paper is individual fracture’s compliance (Schoenberg 1980) and I and
written from the point of view of compliance rather than stiff- others have used Z T,N for fracture system compliances, i.e.,
ness. Further, the excess fracture compliance tensor associated total of individual fracture compliances per unit width normal
with any particular set of aligned fractures has a very partic- to the fractures. Recently, however, Michael Worthington, in
ular form that depends on the orientation of the fractures presentations at the 2007 EAGE meeting and the Septem-
and a 3 × 3 symmetric, positive non-negative definite, frac- ber 2007 EAGE summer research workshop on ‘Integrating
ture compliance matrix (FCM) that relates the excess strain Geosciences for Fractured Reservoir Description’, used B T,N
(i.e., the strain in the fractured medium minus the strain that and as B sounds softer and thus more suggestive of compli-
would occur if the set of aligned fractures were absent) with ance than Z, which is sharp and generally associated with
the stress traction acting on the fracture plane, see Schoen- impedance. I have also decided to switch to the use of B T,N
berg and Sayers (1995). This is not to say that the properties for fracture system compliance].
of the background medium and the presence of other fracture The analysis is begun by looking at a single fracture set
sets may not influence the actual values of a particular frac- with its strike direction in the 2-direction (so that the normal
ture set’s FCM; in general it will. However, the assumption to the fractures is in the 1-direction) that pervades a large
made here is that the influence is manifest only in the val- (compared to wavelength) region of a transversely isotropic
ues of the components of the fracture compliance matrix but medium with its axis of symmetry taken to be the vertical
not in its form. Thus the question of how the presence of a x 3 -axis (also called a vertically transversely isotropic (VTI)
set of aligned fractures, with its indescribable complexity and medium that will be unstated but understood below). The
unknowable microscopic properties, affects the propagation excess compliance matrix associated with this set of fractures


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
Vertically fractured transversely isotropic media 171

given by Schoenberg and Sayers (1995) has the form, monoclinic stiffness tensor can conceivably be determined by
⎡ ⎤ laboratory measurements on large rock samples with small
BN 0 0 0 0 0 fractures, admittedly a difficult experiment to set up, or, hope-
⎢ ⎥
⎢ 0 ⎥ fully, in future, from in situ field measurements. It would be
⎢ 0 0 0 0 0 ⎥
⎢ ⎥
⎢ ⎥ necessary to determine a large variety of slowness vectors and
⎢ 0 0 0 0 0 0 ⎥
Sf = ⎢

⎥,
⎥ (1) perhaps transmission and reflection coefficients at an inter-
⎢ 0 0 0 0 0 0 ⎥
⎢ ⎥ face between an unknown fractured medium and a known
⎢ ⎥
⎢ 0 0 0 0 BT 0 ⎥ isotropic medium. A VFTI medium with two fracture sets de-
⎣ ⎦
0 0 0 0 0 BT pends on five TI constants of the background medium and
2 × 3 = 6 parameters specifying the two fracture sets and it is
with B T , B N ≥ 0 and all the results of this paper follow convenient then that the deconstruction of any VFTI medium
from the assumption that the total compliance of the frac- should begin with the search for a transversely isotropic back-
tured medium is the sum of the compliance matrix of the ground and two sets of aligned fractures that yield the same
background, whatever it may be, plus an excess compliance monoclinic compliance tensor as the original medium. If such
matrix of this form. Multiple fracture sets are accommo- a deconstruction is possible, it may be called a ‘minimal’ de-
dated by ‘adding’ fracture sets at different orientations (see, construction and the medium is called a VFTI 2 medium.
for example, Nichols, Muir and Schoenberg 1989; Hood and A word about the representation of the compliance tensor
Schoenberg 1992). Such media are called vertically fractured used here. The elastic stiffness and compliance tensors are of
transversely isotropic (VFTI) media and the implication is that 4th rank and are customarily written as c ijkl and s ijkl , respec-
the only features disturbing rotational symmetry about the tively. Due to the required symmetries of the elasticity tensors,
vertical axis are these sets of vertical fractures. i.e.,
For a single fracture set, the VFTI medium is orthorhombic,
skli j = si jkl , si jlk = si jkl , s jikl = si jkl ,
i.e., it has three mutually orthogonal planes of mirror symme-
try but no further symmetry. The introduction of additional which also hold for the stiffness tensor, there are 21 inde-
fracture sets, in general, renders the medium monoclinic with a pendent elastic constants that can be, and customarily are,
horizontal plane of mirror symmetry, although there are situa- represented in Voigt notation by a 6 × 6 symmetric matrix
tions when multiple fracture sets are present but the equivalent that relates the six independent stresses to the six independent
medium remains orthorhombic. These are considered briefly strains. The relationship between the 4th rank elastic com-
in a section on orthorhombic VFTI media. A continuous dis- pliance tensor and its conventional representation as a 6 × 6
tribution of fracture system orientations can also be accom- symmetric matrix S = C−1 in Voigt notation is,
modated by specifying a pair of functions BT (θ ) and BN (θ).
s(i)(i)( j)( j) = Si j ,
The only adaptation needed for this continuous case would Si p+3
be the replacement of the summation of compliance matrices s(i)(i) jk = s jk(i)(i) = , p = j = k,
2
with the corresponding integration over angle θ . The mono- S p+3q+3
si jkl = , p = i = j, q = k = l,
clinic behaviour of a VFTI medium is not completely general 4
(a general monoclinic medium has 13 independent elastic con- where () around an index denotes ‘no summation’.
stants); instead there are two constraints on the compliance Rotation of coordinates for elasticity matrices in Voigt no-
components that are always satisfied leaving an 11 parame- tion is often done by pre- and post-multiplying by a Bond
ter monoclinic medium. The constraints come about because transformation matrix and its transpose, see for example,
vertical fracturing cannot affect certain compliance compo- Auld (1990). Coordinate rotation about the 3-axis, whether
nents and thus those compliance components are the same as in 4th rank tensor notation or using Bond transformation ma-
they are in the underlying background transversely isotropic trices, is essentially the multiplication of a vector composed of
medium. compliance components by a block diagonal rotation matrix.
The deconstruction of a VFTI medium does not depend By definition, the TI background’s tensor components are un-
at all on the number of fracture sets or whether there is a affected by rotation about the 3-axis. Rotation of coordinates
continuous or discrete distribution of orientation angles. The only affects the components of the excess fracture compliance
components of the compliance matrix could be obtained by matrix. The rotation matrix needed for rotation of this mono-
the above construction but practically, the components of a clinic excess compliance matrix is a block diagonal consisting


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
172 M.A. Schoenberg

in a 3 × 3 and a 6 × 6 block; these two blocks are derived in S = Sb + S f


Appendix A. ⎡ ⎤
Sb11 + BN Sb12 Sb13 0 0 0
⎢ ⎥
⎢ S ⎥
⎢ b12 Sb11 Sb13 0 0 0 ⎥
⎢ ⎥
⎢ ⎥
CONSTRUCTION OF VFTI MEDIA ⎢ Sb13 Sb13 Sb33 0 0 0 ⎥
=⎢⎢
⎥.

⎢ 0 0 0 Sb55 0 0 ⎥
Let the compliance matrix of an unfractured TI background ⎢ ⎥
⎢ ⎥
medium, in Voigt 6 × 6 matrix notation, be denoted by S b , ⎢ 0 0 0 0 Sb55 + BT 0 ⎥
⎣ ⎦
which must have the form, 0 0 0 0 0 Sb66 + BT
⎡ ⎤ (4)
Sb11 Sb12 Sb13 0 0 0
⎢ ⎥
⎢S ⎥
⎢ b12 Sb11 Sb13 0 0 0 ⎥
⎢ ⎥
⎢ ⎥ The non-negative parameters B T and B N are the excess tan-
⎢ Sb13 Sb13 Sb33 0 0 0 ⎥
Sb = ⎢

⎥,
⎥ (2) gential and normal fracture compliances, respectively. Equa-
⎢ 0 0 0 Sb55 0 0 ⎥ tion (4) is the orthorhombic compliance matrix of the frac-
⎢ ⎥
⎢ ⎥
⎢ 0 0 0 0 Sb55 0 ⎥ tured medium with a single set of fractures. Note that
⎣ ⎦
S 55 ≥ S 44 and S 11 ≥ S 22 . The deconstruction of such a medium
0 0 0 0 0 Sb66
can be done by inspection: B T = S 55 − S 44 , B N = S 11 − S 22
and from the requirement that the fracture set be rotationally
subject to
symmetric and from equation (3),
Sb66
Sb12 + − Sb11 = 0. (3)
2 S66 = 2(S22 − S12 ) + (S55 − S44 ). (5)

This relation is the analogue, in the compliance domain, to If this constraint is not satisfied, this is an indication that the
the well-known relation on the components of the stiffness fracture set is not rotationally symmetric and that the hori-
matrix, Cb12 + 2Cb66 − Cb11 = 0. This matrix is dependent on zontal tangential compliance (which adds in the 6,6 position)
5 independent parameters and the values of the components is not the same as the vertical tangential compliance (which
of this compliance matrix of the TI background medium, S b , adds in the 5,5 position). If S 55 − S 44 and S 11 − S 22 are not of
is unchanged by any coordinate system rotation about the the same sign, then two fracture sets would be required for a
3-axis. valid deconstruction.
Let the contribution of the fractures to the compliance ma- However, for the simple case of a single set of rotation-
trix of the fractured medium, known as the excess fracture ally symmetric fractures, the coordinate system can be rotated
compliance matrix, be denoted by S f . For a single set of frac- about the x 3 -axis to a primed ( ) system an angle −θ, so in the
tures, let θ be the angle between a line normal to the fractures, new coordinate system, the fracture set is oriented so that the
called the fracture normal and the +x 1 -axis and as the frac- fracture normal is at angle +θ from the x 1 -axis. The compli-
ture normal is a line, its sense can always be chosen such that ance matrix of the TI background medium, S b , is unchanged
−π /2 < θ ≤ π /2. For θ = 0, i.e., with the plane of the by this rotation. The rotation of the coordinate system by any
aligned fractures being the 2,3-plane, the fractured medium’s angle about the x 3 -axis is discussed in Appendix A. The frac-
compliance matrix is the sum of S b and the excess fracture tured medium’s compliance matrix in the primed system is
compliance matrix of equation (1), or, given by,

⎡ ⎤
Sb11 + S f11 Sb12 + S f12 Sb13 0 0 S f16
⎢ ⎥
⎢ Sb + S f Sb11 + S f22 Sb13 0 0 S f26 ⎥
⎢ 12 12 ⎥
⎢ ⎥
⎢ Sb Sb13 Sb33 0 0 0 ⎥
⎢ 13 ⎥
S = Sb + S f = ⎢ ⎥,
⎢ 0 0 0 Sb55 + S f44 S f45 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ 0 0 0 S f45 Sb55 + S f55 0 ⎥
⎣ ⎦
S f16 S f26 0 0 0 Sb66 + S f66


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
Vertically fractured transversely isotropic media 173

⎡ ⎡ ⎤ ⎤
S f55 1 + C2
⎢ ⎥ ⎢ ⎥
⎢ S f ⎥ = ⎢ S2 ⎥ BT ,
⎣ 45 ⎦ ⎣ ⎦ 2
S f44 1 − C2
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
s f11 (1 + C2 )2 BN + S22 BT /4 (1 + C2 ) [BT + BN − C2 (BT − BN)] /4
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢s ⎥ ⎢[(1 + C )B − C B ] S /2 ⎥ ⎢ S2 [BN − C2 (BT − BN)] /2 ⎥
⎢ f16 ⎥ ⎢ 2 N 2 T 2 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢s f12 ⎥ ⎢ S2 (BN − BT )/4
2
⎥ ⎢ −S2 (BT − BN)/4
2

⎢ ⎥=⎢ ⎥≡⎢ ⎥, (6)
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢s f66 ⎥ ⎢ S2 BN + C2 BT
2 2
⎥ ⎢ BT + BN + (C2 − S2 )(BT − BN) /2 ⎥
2 2
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢s f26 ⎥ ⎢[(1 − C2 )BN + C2 BT ] S2 /2 ⎥ ⎢ S2 [BN + C2 (BT − BN)] /2 ⎥
⎣ ⎦ ⎣ ⎦ ⎣ ⎦

s f22 (1 − C2 ) BN + S2 BT /4
2 2
(1 − C2 ) [BT + BN + C2 (BT − BN)] /4

where,
Note that due to the assumed rotational symmetry of all
C2 ≡ cos 2θ, S2 ≡ sin 2θ. the fracture sets, a constraint exists among the components
The medium is still orthorhombic but because it is no longer of S f :
in its natural coordinate system (when the coordinate planes S f55 + S f44 = S f66 − 4S f12 . (8)
coincide with the three planes of mirror symmetry), the com-
pliance matrix appears to be that of a monoclinic medium The compliance matrix S = S b + S f now, in general, is that of
with the 1,2-plane as its mirror plane of symmetry. With no a monoclinic medium.
loss of generality, the primes can be dropped and this is merely If fractures exist at all, orientations of the compliances may
the compliance matrix for a VFTI medium with a single frac- be specified by a pair of functions BT (θ) and BN (θ). Then,
ture set whose fracture normal is at angle θ , −π /2 < θ ≤ π /2, the summation over the n sets in equation (7) is replaced by
from the +x 1 -axis. integrals over θ from −π /2 to π /2. As an example consider
In addition, to 0th order, it is assumed that the inclusion S f55 , which for a functional representation, is given by,
of another fracture set at another orientation, merely adds π/2
1
S f55 = (1 + cos 2θ)BT (θ)dθ.
another fracture compliance matrix, although rotated to its 2 −π 2
new orientation. Thus, when there are multiple fracture sets, Letting compliance be a distribution of the form BT (θ ) =
say n such sets, the form of the compliance matrix is the i=1
n
BTi δ(θ − θi )returns the discrete representation.
same as equation (6) but with the fracture compliance matrix
components given by the sums over all the sets of aligned
fractures, PROPERTIES OF VFTI MEDIA
⎡ ⎤ ⎡ n ⎤
S f55 i=1 (1 + C2i )BTi VFTI media are those with compliance matrices that can be
⎢ ⎥ 1⎢ ⎥
⎢ ⎥ expressed as the sums of a) an excess fracture compliance
⎢ S f45 ⎥ = ⎢ i=1
n
S2i BTi ⎥,
⎣ ⎦ 2⎣ ⎦ matrix constructed as above, for any distribution of tangential
Sf i=1 (1 − C2i )BTi
n
and normal compliances over angle and b) any transversely
44

⎡ ⎡⎤  ⎤ isotropic compliance matrix. Any medium that is equivalent


S f11 i=1
n
(1 + C2i ) BTi + BNi − C2i (BTi − BNi ) /2
⎢  ⎥ to at most two fracture sets in a TI background is called
⎢ ⎥ ⎢ ⎥
⎢S f ⎥ ⎢ i=1
n
S2i BNi − C2i (BTi − BNi ) ⎥ VFTI 2 , so clearly VFTI 2 is a subset of vertically fractured
⎢ 16 ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢S f ⎥ −i=1 S2i (BTi − BNi )/2
n 2 transversely isotropic. S f for a VFTI 2 medium has 2 × 3 =
⎢ 12 ⎥ 1⎢ ⎥
⎢ ⎥= ⎢

⎥.
⎥ 6 independent parameters. S b depends on the 5 transversely
⎢S f ⎥ 2⎢  n BT + BN + (C 2 − S 2 )(BT − BN ) ⎥
⎢ 66 ⎥ ⎢ i=1 i i 2 2 i i
⎥ isotropic parameters. Thus any VFTI 2 media could be thought
⎢ ⎥ ⎢ 
i i

⎢S f ⎥ ⎢  n
+ − ⎥
⎣ 26 ⎦ ⎣ i=1 iS2 B Ni C 2i (BTi BN i ) ⎦ of as a point in 11 dimensional space, although this is reduced

S f22 i=1
n
(1 − C2i ) BTi + BNi + C2i (BTi − BNi ) /2 by one if all media that can be rotated into one another are
considered to be the same medium.
(7) The components in equation (7) have been expressed in a
form that isolates the contribution of the values of B T − B N .


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
174 M.A. Schoenberg

This was done because these quantities show up directly in the and two on just the BTi − BNi ,
equations and are usually quantities of interest for reservoir
2S12 + S66 − (S11 + S22 ) = 2S f12 + S f66 − (S f11 + S f22 )
description. Also this formulation makes clear how to confine
the analysis to scalar fracture systems for which B T − B N =
n
 2  
= C2i − S22i BTi − BNi ,
0 (this special case to be discussed below). Since the nine i=1
  
cos 4θi ≡C4i
components of equation (7) are the only possible non-zero
components of S f , the presence of vertical fractures cannot 2(S26 − S16 ) = 2(S f26 − S f16 )
n
affect the 1,3, 2,3, 3,3 and 3,6 components of the monoclinic
= 2S2i C2i (BTi − BNi ). (12)
compliance matrix. Consequently, these components are those   
i=1
of the background TI medium. Any VFTI medium then is sin 4θi ≡S4i

subject to the constraints on these components that exist in In addition, from Appendix A, S 55 + S 44 is invariant under
the background transversely isotropic medium, i.e., rotation about the x 3 -axis, as are S 66 − 4S 12 and S 11 + S 22 +
2S 12 . These invariant linear combinations of the elements of
S13 =S23 , S36 = 0. (9)
S b for a VFTI medium are given by,
A monoclinic medium satisfying these constraints is a po- n
S55 + S44 = 2Sb55 + S f55 + S f44 = 2Sb55 + BTi , (13)
tential VFTI (p-VFTI) medium. Clearly, VFTI is a subset of
i=1
p-VFTI, which is also an 11 dimensional space (2 constraints
on 13 monoclinic parameters), which can also be reduced by
S66 − 4S12 = (Sb66 − 4Sb12 ) +S f 66 − 4S f12
one as all media that can be rotated into one another are the   
3Sb66 −4Sb11
same and the medium can always be rotated to its natural
n
coordinate system in which the 4,5-component vanishes. (14)
= 3Sb66− 4Sb11 + BTi ,
In addition to the 1st of equation (9), there are three other i=1
relations among the non-vanishing components of the trans-
versely isotropic background’s compliance matrix S b : S11 + S22 + 2S12 = 2(Sb11 + Sb12 ) +S f11 + S f22 + 2S f12
  
4Sb11 −Sb66
Sb55 − Sb44 = 0, Sb11 − Sb22 = 0,
n
2Sb12 + Sb66 − (Sb11 + Sb22 ) = 0; = 4Sb11− Sb66 + BNi . (15)
i=1
and further, there are three non-vanishing components of the
Note that from equations (13) and (14),
monoclinic compliance matrix whose corresponding compo-
nents for the TI background vanish: S55 + S44 − 2Sb55 = S66 − 4S12 − 3Sb66 + 4Sb11 . (16)

Sb45 = 0, Sb16 = 0, Sb26 = 0. If the background were actually isotropic, then S33 = Sb33 =
Sb11 = 1/Eb and S13 = Sb13 = Sb12 = −νb /Eb , where E b , ν b
From these six transversely isotropic relations and equation are the isotropic background’s Young’s modulus and Pois-
(7), there are two linear equations on just the BTi , son’s ratio, respectively. In addition, Sb55 = Sb66 = 1/μb =
n
2(1 + νb )/Eb = 2(S33 − S13 ) where μ b is the isotropic back-
S55 − S44 = S f55 − S f44 = C2i BTi , ground’s shear modulus, so that every component of S b can be
i=1 written in terms of S 33 and S 13 . Then equation (16) becomes,
n
2S45 = 2S f45 = S2i BTi , (10) S55 + S44 − S66 + 4S12 = 2(S33 +S13 ). (17)
i=1
The satisfaction of this relation is a strong indicator that the
two linear equations on just the BNi , background is isotropic.
n Note that the compliance matrix of the form of equation (6)
S11 − S22 = S f11 − S f22 = C2i BNi , with components for multiple fracture sets from equation (7)
i=1 is not in its natural coordinate system since in general S 45 =
n
(11) 0. Yet measurements will often be made in the natural coordi-
S16 + S26 = S f16 + S f26 = S2i BNi ,
i=1 nate system, for now designated as the starred (∗ ) system, since


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
Vertically fractured transversely isotropic media 175

this is defined by the polarization directions of vertically prop- 1 Solve linear equation (10) when n = 2 and S 45 = 0 for
agating shear waves and by the axes of the slowness ellipse BT1 , BT2 in terms of θ 1 and θ 2 giving,
for horizontally propagating shear waves with displacement ⎡ ⎤ ⎡ ⎤−1 ⎡ ⎤
vectors in the vertical direction, so called cross-plane waves BT1 C21 C22 S55 − S44
⎣ ⎦=⎣ ⎦ ⎣ ⎦
(Helbig 1994). From Appendix A, equation (A1), it is seen BT2 S21 S22 0
that setting S∗45 = 0 implies the starred system must be at   
D−1
angle θ from the unstarred system, where 2
⎡ ⎤
S22
2S45 =⎣ ⎦ S55 − S44 , (18)
tan 2θ = ,
S55 − S44 −S21 sin 2(θ2 − θ1 )

or, in particular, θ is such that, recalling that in this rotated system, S 55 − S 44 > 0. Since all
fracture compliances must be positive, this solution implies
S55 − S44 that θ 1 and θ 2 are of opposite sign, so that it may be assumed
cos 2θ ≡ C2 =  ,
(S55 − S44 )2 + 4S45
2 that −π /2 < θ 1 < 0 < θ 2 < π /2, thus defining R, a square
2S45 region of the θ 1 , θ 2 -plane. Further, the positivity of the BTi
sin 2θ ≡ S2 =  , implies,
(S55 − S44 )2 + 4S45
2

|D2 | = sin 2(θ2 − θ1 ) > 0, or θ2 − θ1 < π/2, (19)


and
reducing the solution space to the lower right triangular half

S55 − ∗
S44 = C2 (S55 − S44 ) + S2 (2S45 ) of R.
 2 Similarly, solve linear equation (11) with n = 2 for BN1 , BN2

= (S55 − S44 )2 + 4S45 2
> 0, S45 = 0, in terms of θ 1 and θ 2 giving,
⎡ ⎤ ⎡ ⎤
ensuring that S∗55 >S∗44 whatever the sign of S 55 − S 44 . In this BN1 S 11 − S22
⎣ ⎦ = D−12
⎣ ⎦. (20)
natural coordinate system, shear waves propagating in the BN2 S16 + S26
vertical 3-direction are polarized in the slow 1∗ -direction and
 ∗  ∗ 3 Substitution of the results of steps 1 and 2 into equation
the fast 2∗ -directions with slownesses ρ S55 and ρ S44 , re-
spectively, where ρ is density. These are also the lengths of (12) with n = 2 yields,
⎡ ⎤ ⎡ ⎤⎡ ⎤
the semi-axes of the slowness ellipse for horizontally propa- 2S12 + S66 − (S11 + S22 ) C41 C42 BT1 − BN1
gating cross-plane waves. Below, unless otherwise specified, ⎣ ⎦=⎣ ⎦⎣ ⎦
2(S26 − S16 ) S41 S42 BT2 − BN2
it will be assumed that the compliance tensor has been ob-      
tained/given in the natural coordinate system and the stars q D4

will be dropped. ⎡ ⎤
(S55 − S44 ) − (S11 + S22 )
= D4 D−1
2
⎣ ⎦,
−(S16 + S26 )
  
DECONSTRUCTION OF MONOCLINIC a (21)
VFTI MEDIA
a pair of non-linear equations on orientations θ 1 , θ 2 . As shown
Given a p-VFTI compliance matrix, the assumption of two in Appendix B, this pair of non-linear equations can be trans-
fracture sets implies that there are six fracture parameters, formed to a pair of non-linear equations on ≡ θ2 + θ1
BTi , BNi , θi , i = 1, 2 that need to be determined. These pa- and υ/ ≡ θ2 − θ1 . These can be reduced to equation (B7) on
rameters specify S f . Equations (10), (11) and (12) consist in cos υ/ in terms of sines and cosines of and equation (B9) on
six equations on these parameters. This is at least suggestive sin which, when squared, gives a cubic equation (B10) on
that it should be possible to find S f and subsequently a TI sin2 . Three of the six roots are extraneous, as they do not
background with compliance matrix S b . If this is possible, the satisfy equation (B9). From equation (B7), each of the non-
original medium is VFTI 2 . extraneous three roots gives a corresponding value of cos υ/
The deconstruction ultimately requires solving a pair of that must be positive from equation (19) and which must sat-
non-linear equations on two orientation angles θ 1 and θ 2 . isfy constraint equation (B3). Other roots must be discarded
The method has the following steps: to guarantee the positivity of theBTi . Each of the remaining


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
176 M.A. Schoenberg

roots then yield a pair of orientation angles θ 1 , θ 2 that lie in compliance matrix of the form,
the acceptable region of the θ 1 , θ 2 -plane. ⎡ ⎤
−B 0 0 0 0 0
4 Substitute these acceptable pairs of orientations into equa- ⎢ ⎥
⎢ 0 0 0 0 0 0 ⎥
tions (18) and (20). Positive values for fracture compliances ⎢ ⎥
⎢ ⎥
BT1 , BT2 are guaranteed. Pick the pairs that yield positive val- ⎢ ⎥
⎢ 0 0 0 0 0 0 ⎥
ues ofBN1 , BN2 . If there is one such pair, there may be a unique Snon− f = ⎢

⎥.

⎢ 0 0 0 0 0 0 ⎥
minimal representation consisting in two fracture sets; if there ⎢ ⎥
⎢ ⎥
⎢ 0 0 0 0 BT 0 ⎥
are two or three solutions then there may be non-unique min- ⎣ ⎦
imal representations; if there are no such pairs, then there is 0 0 0 0 0 BT
no minimal representation. These results depend on six linear
This is not that of a fracture set, since energy considerations
combinations of the components of the original monoclinic
require that a fracture set have a non-negative definite excess
compliance tensor S, namely,
fracture compliance matrix. However, this compliance matrix
can be deconstructed since S non-f can be written as the sum of
q1 ≡ 2S12 + S66 − (S11 + S22 ), q2 = 2(S26 − S16 ), three matrices,
⎡ ⎤
−B 0 0 0 0 0
and S 55 − S 44 > 0, 2S 45 = 0, S 11 − S 22 and S 16 + S 26 . Note ⎢ ⎥
⎢ 0 −B 0 0 0 ⎥
that these six linear combinations would vanish if no fracture ⎢ 0 ⎥
⎢ ⎥
sets were present. ⎢ ⎥
⎢ 0 0 0 0 0 0 ⎥
⎢ ⎥
5 From the values of θ 1 , θ 2 and the corresponding fracture ⎢ ⎥
⎢ 0 0 0 −B 0 0 ⎥
compliances, the S fi j are found from equation (7) with n = 2. ⎢ ⎥
⎢ ⎥
Subtracting these values from the corresponding S ij yields the ⎢ 0 0 0 0 −B 0 ⎥
⎣ ⎦
TI background, as, 0 0 0 0 0 −2B
  
ST I
S44 + S55 1
2 ⎡ ⎤
Sb55 = − BTi , 0 0 0 0 0 0
2 2 ⎢ ⎥
i=1 ⎢0 ⎥
⎢ 0 0 0 0 0 ⎥
1
2
 ⎢ ⎥
Sb66 = S66 − BTi + BNi + C4i (BTi − BNi ) , ⎢ ⎥
⎢0 0 0 0 0 0 ⎥
2 i=1 +⎢



⎢0 0 0 0 0 0 ⎥
S11 + S22 1
2
 ⎢ ⎥
Sb11 = − BTi + BNi − C22i (BTi − BNi ) , ⎢ ⎥
⎢0 0 0 0 B + BT 0 ⎥
2 4 i=1
⎣ ⎦
2 0 0 0 0 0 B + BT
1 Sb66   
Sb12 = S12 + S22i (BTi − BNi ) = Sb11 − . S0
4 i−1
2 ⎡ ⎤
0 0 0 0 0 0
⎢ ⎥
Note that the final conditions for a valid representation as ⎢0 0⎥
⎢ B 0 0 0 ⎥
⎢ ⎥
a VFTI 2 medium are thatSb55 > 0, Sb66 > 0, Sb11 − Sb66 /4 > ⎢ ⎥
⎢0 0 0 0 0 0⎥ (22)
Sb213 /Sb33 , guaranteeing a positive definite TI background. +⎢

⎥.

In all examples that have been considered, the cubic equa- ⎢0 0 0 B 0 0⎥
⎢ ⎥
⎢ ⎥
tion yielded a unique solution for the pair of angles, θ 1 and ⎢0 0 0 0 0 0⎥
⎣ ⎦
θ 2 , which satisfied constraint equation (B3). These constraints 0 0 0 0 0 B
then guarantee that the B T will be positive but not the B N .   
Sπ/2
Cases have been found for vertically fractured transversely
isotropic media where one or both of the B N were negative. S TI when added to a TI background compliance matrix is also
In this case, a minimal representation in terms of two fracture TI provided the sum is positive definite. The second, S 0 , is
sets does not exist and the medium is not VFTI 2 . However, the excess compliance matrix of a fracture set oriented nor-
deconstruction can still be achieved. This is based on the fact mal to the x 1 -axis with zero normal compliance. The third,
that a ‘fracture set’ with B T > 0 and B N = −B < 0, if it S π /2 , is the excess fracture compliance matrix of a scalar frac-
were normal to the 1-direction, would have an excess fracture ture set (i.e., normal and tangential compliances are equal)


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
Vertically fractured transversely isotropic media 177

oriented normal to the x 2 -axis. In general, ifBNi < 0, the or- Table 1 Fracture parameters
thogonal fracture sets that replace this non-fracture set ma-
Set 1 Set 2 Set 3
trix would be aligned with fracture normals at angles θ i and
θ i + π /2. Then it is clear that one negative BNi implies a
θ −60◦ 0◦ +45◦
deconstruction into a representation of 3 fracture sets; two δT 0.45 0.35 0.16
negative BNi would result in a deconstruction into 4 fracture BT 0.07719 0.05080 0.01797
sets. δN 0.37 0.49 0.27
BN 0.01712 0.02801 0.01078

A NUMERICAL EXAMPLE For this example, let the fracture parameters for constructing
the medium be given by Table 1.
Consider three sets of aligned fractures in a background TI
The contribution of the fractures to the compliance matrix
medium which, for this exercise, will be chosen to be Green-
of the fractured medium is
horn shale with the measured elastic stiffnesses (Jones and ⎡ ⎤
Wang 1981), 50.74 −13.06 0 0 0 −15.03
⎢ ⎥
⎢−13.06 31.29 0 0 0 10.98 ⎥
⎢ ⎥
⎢ ⎥
Cb11 = 34.3, Cb33 = 22.7, Cb55 = 5.4, ⎢ 0 0 0 0 0 0 ⎥
Sf = ⎢⎢


Cb66 = 10.6, Cb13 = 10.7 GPa. ⎢ 0 0 0 66.87 −24.44 0 ⎥
⎢ ⎥
⎢ 0 0 0 −24.44 79.08 0 ⎥
⎣ ⎦
Note that these moduli yield the background compliance −15.03 10.98 0 0 0 93.72
matrix, ×10−3 ,

⎡ ⎤
36.99 −10.18 −12.63 0 0 0
⎢ ⎥
⎢−10.18 36.99 −12.63 0 ⎥
⎢ 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢−12.63 −12.63 55.96 0 0 0 ⎥
Sb = ⎢

⎥ × 10−3 ,
⎥ (23)
⎢ 0 0 0 185.19 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ 0 0 0 0 185.19 0 ⎥
⎣ ⎦
0 0 0 0 0 94.34

in units of (GPa)−1 . To choose meaningful values for the B T ,


which, when rotated by −37.990◦ to put the fracture system
B N , the dimensionless parameters δ T , δ N will be employed, as
compliance matrix in its natural coordinates, yields,
described by Schoenberg and Sayers (1995), i.e., ⎡ ⎤
45.00 −12.72 0 0 0 22.12
B × stiffness ⎢ ⎥
δ≡ . ⎢ ⎥
1 + B × stiffness ⎢−12.72 36.36 0 0 0 −4.23 ⎥
⎢ ⎥
⎢ ⎥
⎢ ⎥
These δ’s lie between 0 and 1 and specify what part of a ⎢ 0 0 0 0 0 0 ⎥
⎢ ⎥
Sf = ⎢ ⎥
material’s total displacement over a given width normal to ⎢ ⎥
⎢ 0 0 0 49.79 0 0 ⎥
a set of fractures is attributable to slip across the fractures. ⎢ ⎥
⎢ ⎥
⎢ 0 0 ⎥
Normal compliances are derived from values of δ N relative ⎢ 0 0 0 98.17 ⎥
⎣ ⎦
to Cb11 ; tangential compliances from values of δ T relative to
22.12 −4.23 0 0 0 95.08
Cb66 . Thus,
× 10−3 . (24)
1 δT 1 δN
BT = , BN = .
Cb66 1 − δT Cb11 1 − δ N


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
178 M.A. Schoenberg

Adding this result to S b gives,


⎡ ⎤
81.98 −22.90 −12.63 0 0 22.12
⎢ ⎥
⎢ −22.90 73.34 −12.63 −4.23 ⎥
⎢ 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ −12.63 −12.63 55.96 0 0 0 ⎥
S = Sb + S f = ⎢

⎥ × 10−3 ,
⎥ (25)
⎢ 0 0 0 232.97 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ 0 0 0 0 283.35 0 ⎥
⎣ ⎦
22.12 −4.23 0 0 0 189.42

and the required constants for the deconstruction are, in since, from equations (18) and (19), they will necessarily give
(GPa)−1 , negative values for theBTi . However, the third root, which is
⎡ ⎤ ⎡ ⎤ equivalent to,
2S12 + S66 − (S11 + S22 ) −11.72
q≡⎣ ⎦=⎣ ⎦ × 10−3 ,
2(S26 − S16 ) −52.70
= θ2 + θ1 = arcsin(0.3906) = 22.99◦
⎡ ⎤ ⎡ ⎤
and υ/ = θ2 − θ1 = arccos(0.4302) = 64.52◦ ,
S55 − S44 50.38
⎣ ⎦=⎣ ⎦ × 10−3 ,
2S45 0
or
⎡ ⎤ ⎡ ⎤
S11 − S22 8.64
⎣ ⎦=⎣ ⎦ × 10−3 .
S16 + S26 17.89 θ1 = −20.77 and θ2 = 43.76, (26)

Subsequently,
⎡ ⎤ ⎡ ⎤ ⎡ ⎤ gives,
S55 − S44 S11 − S22 41.74
a≡⎣ ⎦−⎣ ⎦=⎣ ⎦ × 10−3 . ⎡ ⎤ ⎡ ⎤
2S45 S16 + S26 −17.89 C21 C22 0.7486 0.0434
D2 = ⎣ ⎦=⎣ ⎦
From q and a, equation (B9) gives the set of constants, S21 S22 −0.6630 0.9991
   
c3 = 2a1 q1 q2 + a2 q22 − q12 − 3a12 − a22 = 92.1 × 10−6 ,
   
c2 = 2a2 q1 q2 − a1 q22 − q12 − 3a22 − a12 = −164.9 × 10−6 , and positive compliances,
 
c1 = 3a1 q1 q2 + a2 q22 − 2q12 − 2(2a12 − a22 ) = 145.8 × 10−6 , ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
−6 BT1 S55 − S44 0.0648
c0 = q2 (q2 a1 − q1 a2 ) − 2a1 a22 = 100.2 × 10 . ⎣ ⎦ = D−1 ⎣ ⎦=⎣ ⎦,
2
BT2 2S45 0.0430
from which are found the coefficients in cubic equation (B10)
⎡ ⎤ ⎡ ⎡ ⎤⎤
of the various powers of
BN1 S11 − S22
0.0101 (27)
⎣ ⎦ = D−1 ⎣ ⎦=⎣ ⎦.
sin2 ≡ sin2 (θ2 + θ1 ). BN2
2
S16 + S26 0.0246
For this example, after multiplying the coefficients by 109 , the
cubic equation is, The parameters of the original 3 fracture sets, rotated 37.990◦
142.7(sin2 )3 − 228.6(sin2 )2 + 97.4 sin2 − 10.0 = 0, so the medium is in its natural coordinate system, from
Table 1, are given in Table 2. These can be compared with
with non-extraneous roots from equation (B9), the parameters of the derived 2 fracture sets that are given in
sin ≡ sin(θ2 + θ1 ) = 0.9884, −0.6873, 039.06. Table 3. Derived Set 1 can be seen to be close to original Set 1,
somewhat less compliant, while derived Set 2 can be seen to
For each of these roots, corresponding values of be close to derived Set 2, also somewhat less compliant but ro-
cos υ/ ≡ cos(θ2 − θ1 ) = −0.6115, −0.4651, 0.4302, tated about 6◦ towards the original Set 3, the least compliant
of the original 3 fracture sets.
are found from equation (B6). The first two values ofcos υ/, From the parameters given in Table 3 for the derived 2
being negative, implies the first two roots must be rejected fracture sets, equation (7) yields the minimal representation


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
Vertically fractured transversely isotropic media 179

Table 2 Parameters of the original three sets any TI background medium, S f − S fmin must be non-negative
definite, which, for this example, it is.
Set 1 Set 2 Set 3

θ −22.010◦ 37.990◦ +82.990◦


BT 0.0772 0.0508 0.0180
BN 0.0171 0.0280 0.0108 ORTHORHOMBIC VFTI MEDIA

An orthorhombic medium has three orthogonal planes of mir-


ror symmetry. AVFTI medium can be orthorhombic in one of
Table 3 Parameters of the derived two sets
two ways: in a TI background, a) it could be due to a sym-
metrical arrangement of fracture sets about a given vertical
Set 1 Set 2
plane, call it the 1,3-plane (necessarily there is also a symmet-
θ −20.766◦ +43.755◦ rical arrangement about the 2,3-plane), or b) it could be due
BT 0.0648 0.0430 to any number of arbitrarily oriented vertical fracture sets,
BN 0.0101 0.0246 all of which are scalar, meaning that BNi = BTi ≡ Bi for all
i. Case b) can be approximately valid when all the fracture
sets are composed of dry fractures. The basic difference is that
fracture compliance matrix, for case b), from equation (7), S f12 ≡ 0, although whether
⎡ ⎤ or not this is true cannot be known before attempting the
32.28 −10.60 0 0 0 22.12
⎢ ⎥ deconstruction.
⎢−10.60 −4.23 ⎥
⎢ 23.64 0 0 0 ⎥ In either case, to attempt a minimal deconstruction of an
⎢ ⎥
⎢ ⎥ orthorhombic p-VFTI medium, one can start with the sim-
⎢ 0 0 0 0 0 0 ⎥
S fmin =⎢


⎥ plest assumption that the non-axisymmetric part of the com-
⎢ 0 0 0 28.72 0 0 ⎥
⎢ ⎥ pliance tensor is equivalent to a pair of congruent fracture
⎢ ⎥
⎢ 0 0 0 0 79.09 0 ⎥ sets, each with fracture compliances B T /2, B N /2, where one
⎣ ⎦
22.12 −4.23 0 0 0 65.41 member of the pair has its fracture normal at angle θ , the
−3 other has its fracture normal at angle − θ, from the x 1 -axis.
× 10 .
(28) The fracture compliance matrix for the pair is the same as
that from equation (6) except that all terms that are antisym-
which is significantly less compliant than the original frac-
metric with respect to θ (terms proportional to S 2 (θ ) which
ture compliance matrix given in equation (24) in its natu-
are the 16, 26 and 45 components) will vanish. The result
ral coordinate system; consequently, the background medium
is,
needed to give the total compliance matrix for the minimal
representation,
⎡ ⎤
49.70 −12.30 −12.63 0 0 0
⎢ ⎥
⎢−12.30 49.70 −12.63 0 ⎥
⎢ 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢−12.63 −12.63 55.96 0 0 0 ⎥
Sbmin = S − S fmin = ⎢⎢ ⎥ × 10−3 . (29)

⎢ 0 0 0 204.26 0 0 ⎥
⎢ ⎥
⎢ ⎥
⎢ 0 0 0 0 204.26 0 ⎥
⎣ ⎦
0 0 0 0 0 124.01
⎡ ⎤
S f11 S f12 0 0 0 0
is significantly more compliant than than the original back- ⎢ ⎥
⎢S 0 ⎥
ground matrix given in equation (23). ⎢ f12 S f22 0 0 0 ⎥
⎢ ⎥
From Sb + S f = Sbmin + S fmin , if the original background ⎢ ⎥
⎢ 0 0 0 0 0 0 ⎥
medium were very stiff so that S b were O(ε) (almost all Sf = ⎢

⎥,

⎢ 0 0 0 S f44 0 0 ⎥
the fractured medium’s compliance due to the fractures), ⎢ ⎥
⎢ ⎥
deconstruction would give the new background medium as ⎢ 0 0 0 0 S f55 0 ⎥
⎣ ⎦
Sbmin = S f − S fmin + O(ε). Thus, for S b + S f to be VFTI 2 for 0 0 0 0 0 S f66


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
180 M.A. Schoenberg

⎡ ⎤ ⎡ ⎤
compliance fracture sets, which violate energy conservation.

S f55
⎦ 1 ⎣1 + C2 ⎦
= BT , It was seen from equation (22) that a ‘fracture set’ orientation
S f44 2 1 − C2
with a negative value for B T can be deconstructed into physi-
⎡ ⎤ ⎡ ⎤
S f11 (1 + C2 ) [BT + BN − C2 (BT − BN)] cally realizable sets of fractures plus a set of components that
⎢ ⎥ ⎢ ⎥
⎢S f ⎥ 1⎢ −S22 (BT − BN) ⎥ appear to be TI, so that often deconstruction would lead to a
⎢ 12 ⎥ ⎢ ⎥
⎢ ⎥ = ⎢  ⎥. TI background whether or not it was included from the start.
⎢S f ⎥ 4 ⎢2 BT + BN + (C 2 − S 2 )(BT − BN) ⎥
⎣ 66 ⎦ ⎣ 2 2 ⎦ Thus, it seemed reasonable to assume that the background
S f22 (1 − C2 ) [BT + BN + C2 (BT − BN)] was TI from the outset, giving the problem a lot of flexibility
(30)
and using the six independent relations to try to determine
This leads directly to three equations on B T , B N and θ analo- uniquely the six parameters of two fracture sets. Any possible
gous to equations (18), (20) and (21), equally spaced congruent fracture sets are absorbed into the
TI background.
S55 − S44 = S f55 − S f44 = C2 BT > 0,
The centrepiece of this paper is the method for finding two
S11 − S22 = S f11 − S f22 = C2 BN, vertical fracture sets that yield a fracture compliance matrix
2S + S66 − (S11 + S22 ) = 2S f12 + S f66 − (S f11 + S f22 ) whose properties mimic those of a class of monoclinic media,
 12  
q which has been called potentially VFTI (p-VFTI), i.e., mon-
= C4 (BT − BN), (31) oclinic media with the 1,2-plane as its plane of mirror sym-
metry, for which constraint equations (9) on the components
and if these equations have a solution with B T , B N ≥ 0, this of the compliance tensor are satisfied. This leaves 6 TI con-
is a minimal decomposition of the medium. The requirement straints which, when applied to the monoclinic compliances,
that S f55 − S f44 > 0 forces C2 > 0 or θ < π /4. Note that the yield six equations on the six parameters of a minimal repre-
special case θ = 0 is merely a single set of fractures with normal sentation consisting in two fracture sets,BTi , BNi and θ i , i = 1,
in the 1-direction. If there is no apparent minimal solution, 2. This is called a minimal representation, since in reality that
one can look for a pair of congruent fracture sets at ±θ with non-axisymmetric behaviour could be due to more than two
additional sets at θ = 0 or θ = π /2 or both. fracture sets, or possibly to something else entirely. Elimina-
tion of the unknown compliances leads to a deceptively simple
pair of non-linear equations on the two orientation angles of
DISCUSSION
the two fracture sets which, after a change of variables, is re-
A key choice was made to consider a TI background medium duced to a single bi-cubic equation on the sine of the sum of
instead of an isotropic background. The basic reason for this the angles, as shown in Appendix B. If one or more pairs of
choice is the following. If the background is assumed isotropic, acceptable angles can be found from the three non-extraneous
then the isotropic background’s Young’s modulus, E b = 1/S 33 roots of the bi-cubic equation, then there exists one or more (2
and Poisson’s ratio, ν b = −S 13 /S 33 , are known by inspection, or 3) minimal representations consisting in two sets of frac-
immediately giving the isotropic background compliance S b tures and the medium is VFTI 2 . However, from the many
and the 9 possible non-zero components of S f = S − S b from cases that have been looked at numerically, non-unique min-
equation (7) subject to constraint (8). If that constraint is not imal representations have not been found and it remains to
satisfied, then clearly the choice of an isotropic background show whether or not non-unique minimal representations are
was wrong. If constraint (8) is satisfied, then the assumption possible.
of a minimal representation of 2 fracture sets (6 unknowns) Alternatively, there have been examples for which no min-
yields 8 independent equations on 6 unknowns. Three fracture imal representation exists, leading to non-unique solutions
sets with 9 unknowns are needed for a minimal representation consisting in three or four fracture sets in which case the
leaving a one parameter family of solutions. medium is VFTI but not VFTI 2 . It is not clear that this is
In addition, it can be shown that three or more equally always possible, i.e., that there is always a non-extraneous
spaced (in angle) congruent fracture sets, or an orthogonal root of bi-cubic equation (B10) that satisfies constraint (B3).
pair of congruent scalar fracture sets or a uniform (in angle) If there is, this would imply that all p-VFTI media are VFTI.
distribution of fractures give rise to an TI fracture compliance Further work is needed to find the boundary between subsets
matrix, see, for example, Schoenberg (2002). Thus there is consisting in VFTI 2 media and its complement in the set of all
a fundamental uncertainty that can and does allow negative VFTI media.


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
Vertically fractured transversely isotropic media 181

CONCLUSIONS the data with the required precision necessary to conclude that
a given formation is monoclinic and in addition, a p-VFTI
When a medium is shown to be monoclinic with the horizon-
formation, will require much new technology and most likely
tal plane a plane of mirror symmetry, subject to constraint
new experimental well configurations and improved estimates
equations (9), i.e., p-VFTI, a reasonable starting assumption
of reflection and transmission coefficients at interfaces. How
is that the azimuthal anisotropy is caused by multiple sets of
best to do this is a question of optimizing seismic data acqui-
vertical (or near vertical) fractures. Long wavelength elastic
sition and to answer such questions will require, in future, a
wave data can yield, at most, the elasticity tensor from which
combination of theoretical research and a series of field ex-
deconstruction yields the orientation and associated compli-
periments. Thus the purpose of this paper is not to provide
ances of two or more vertical planes of weakness. Here the
an immediately useful tool for exploration and reservoir de-
starting point was not data but the compliance tensors of
scription. It is to extend the boundaries of what could be done
numerically constructed VFTI media with two or more ver-
in regard to monoclinic media if such data were available, to
tical fracture sets. The deconstruction when the medium is
provide the motivation for making the appropriate measure-
constructed with two fracture sets, returns those fracture sets
ments on core samples under a variety of stress conditions and
and the original TI background. Of the media constructed nu-
to finally lead to experiments in fractured reservoirs. To the
merically with three or more fracture sets, most have yielded
extent that permeability anisotropy is connected to fracture
a physically meaningful unique minimal deconstruction into
networks, the analysis of fractures as a cause of azimuthal
two fracture sets and a TI background. Occasionally, no min-
anisotropy will be a problem of growing interest.
imal deconstruction was possible. A surprise was that the ex-
cess fracture compliance tensor of the two fracture sets was
not the same as the excess fracture compliance tensor used to ACKNOWLEDGEMENTS
construct the VFTI medium. Thus the TI backgrounds associ-
The author would like to thank Enru Liu for discussions mo-
ated with each were also different. However, if the procedure
tivating this study that took place at and just after the 12th
is started with a measured compliance tensor, the original
International Workshop on Seismic Anisotropy held in Bei-
VFTI medium is unknowable from long wavelength data and
jing in the autumn of 2006. In addition, many thanks are
we are left with just the unique minimal representation, if it
due to Valeri Korneev of Lawrence Berkeley National Lab for
exists. In any event, the minimal representation reveals much
helping with some of the numerical calculations in the early
about the behaviour of the medium and the effective planes
stages of this investigation. Finally, much of the work was
of weakness. This can have secondary implications for frac-
performed while the author was a guest scientist in the Earth
ture density, whether they are wet or dry and perhaps, when
Sciences Department of Lawrence Berkeley National Lab in
looked at with multiple frequencies (i.e., beyond pure elastic
the spring of 2007.
analysis), possible information on properties of the saturating
fluids.
There are many possibilities for further work in this di- REFERENCES
rection. The restriction that the fracture sets be rotationally Auld B.A. 1990. Acoustic Fields and Waves in Solids, Vol. 1, 2nd edn.
symmetric could be dropped in future investigations; allowing Krieger Publishing Company.
different values for horizontal and vertical tangential compli- Helbig K. 1994. Foundations of Anisotropy for Exploration Seismics,
ance is one way to do that (not the most general way though) 1st edn. Pergamon.
Hood J.A. 1991. A simple method for decomposing fracture induced
while maintaining the monoclinicity of the VFTI. Also, look-
anisotropy. Geophysics 56, 1275–1279.
ing into sets of dipping fractures as well as vertical sets could
Hood J.A. and Schoenberg M.A. 1992. NDE of fracture induced
be very informative. An interesting question is what might be anisotropy. Review of Progress in Quantitative Nondestructive
a minimal representation for fractured media when the sets of Evaluation 11, 2101–2108.
fractures are distributed in 3D. A related problem is to find Jones L.E.A. and Wang H.F. 1981. Ultrasonic velocities in Cretaceous
the set of all anisotropic media that can be represented by an shales from the Williston basin. Geophysics 46, 288–297.
Nichols D., Muir F. and Schoenberg M.A. 1989. Elastic properties
isotropic background with multiple sets of fractures.
of multiply fractured Rock. 59th SEG meeting, Dallas, Texas, USA,
Within the confines of linear elasticity theory, seismic data Expanded Abstracts, P471.
(in its broadest sense) can at most give us the full anisotropic Sayers C.M. and Kachanov M. 1991. A simple technique for find-
description of a particular subsurface formation. To acquire ing effective elastic constants of cracked solids for arbitrary crack


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
182 M.A. Schoenberg

orientation statistics. International Journal of Solids and Structures 4 ‘3’s with one equation on s 3333 = S 33 that vanishes in S f and
12, 81–97. in general is unchanged by rotation about the 3-axis. This
Schoenberg M.A. 1980. Elastic wave behavior across linear slip inter-
equation is not needed for coordinate rotation.
faces. Journal of the Acoustical Society of America 68, 1516–1521.
3 ‘3’s with 2 equations on:
Schoenberg M.A. 2002. Time-dependent anisotropy induced by pore
pressure vaiation in fractured rock. Journal of Seismic Exploration s3331 = S35 /2 and s3332 = S34 /2,
11, 83–105.
Schoenberg M.A. and Douma, J. 1988. Elastic wave propa- both of which vanish in S f and in any monoclinic matrix with
gation in media with parallel fractures and aligned cracks. the 1,2-plane as its mirror plane of symmetry. Thus these
Geophysical Prospecting 36, 571–590. doi:10.1111/j.1365-
equations are not needed for coordinate rotation.
2478.1988.tb02181.x
Schoenberg M.A. and Sayers C.M. 1995. Seismic anisotropy of frac-
2 ‘3’s with six equations on six components which itself is
tured rock. Geophysics 60, 204–211. block diagonal yielding two sets of three equations, called set
2 a on components:

APPENDIX A s1313 = S55 /4, s1323 = S45 /4 and s2323 = S44 /4

Coordinate rotation about the 3-axis and set 2 b , the identical equations, on components:

When rotating about the x 3 -axis an angle θ from an unprimed s1133 = S13 , s2233 = S23 and s1233 = S36 /2.
to a primed coordinate system, the transformed components The 2 b components vanish in S f (although not in a general
of a 4th rank tensor are given in terms of the original unprimed monoclinic compliance tensor) but S f in its simplest form
components according to the following linear transformation when the fractures are normal to the 1-direction has, from
rule: equation (4), S f55 = BT > 0 so the 2 a equations on S 55 , S 45
si jkl = ai p a jq akr als s pqr s , and S 44 are transformed using a 3 × 3 rotation matrix called
⎡ ⎤ A 2 (θ).
cos θ sin θ 0
⎢ ⎥ 1 ‘3’ with 6 equations on 6 components that all vanish in S f
where, a(θ) = ⎣ −sin θ ⎢ 0⎥
cos θ ⎦. and in any monoclinic matrix.
0 0 1 0 ‘3’s with 6 equations on 6 components
It is not necessary to rotate S b as shown in equation (4), since s1111 = S11 , s1122 = S12 , s2222 = S22 , s1112 = S16 /2,
TI matrix components are unaffected by rotation about its
s2212 = S26 /2 and s1212 = S66 /4.
symmetry axis, here the x 3 -axis. Due to the form of a(θ), this
transformation rule, which is essentially a 21 × 21 matrix, is S f from equation (4) has S f11 = BN > 0 and S f66 = BT > 0,
block diagonal with the blocks set according to the number of so that the 0 ‘3’ components must be transformed using a
‘3’s there are out of the four tensor indices. The following 6 × 6 × 6 rotation matrix called A 0 (θ).
6 matrix shows the number of ‘3’ indices for the corresponding Thus only the 3-‘vector’ associated with 2 a and the 6-‘vector’
4th rank tensor: associated with 0 ‘3’ components need be considered.
⎡ ⎤ For the 2 a components of the compliance tensor si jkl =
0 0 2b 1 1 0
⎢ ⎥ aip ajq akr als spqrs yields,
⎢ 0 0 2b 1 1 0 ⎥
⎢ ⎥ 
⎢ ⎥ S55 
⎢ 2b 2b 4 3 3 2b ⎥ = s1313 = a112
s1313 + 2a11 a12 s1323 + a12
2
s2323
⎢ ⎥ 4
⎢ ⎥,
⎢ 1 1 3 2a 2a 1 ⎥ S55 S45 S44
⎢ ⎥ = cos2 θ + 2 sin θ cos θ + sin2 θ ,
⎢ ⎥ 4 4 4
⎢ 1 1 3 2a 2a 1 ⎥
⎣ ⎦ 
S45 
0 0 2b 1 1 0 = s1323 = a11 a21 s1313 + (a11 a22
4
where the meaning of subscripts a and b are described be- + a12 a21 )s1323 + a12 a22 s2323
low. The components of the tensor with a certain number of S55 S45 S44
= − cos θ sin θ +(cos2 θ − sin2 θ) +cos θ sin θ ,
‘3’s in a primed coordinate system can depend only on the 4 4 4

S44 
components with the same number of ‘3’s in the unprimed = s2323 = a212
s1313 + 2a21 a22 s1323 + a22
2
s2323
4
coordinate system. Thus each block consists in components S55 S45 S44
with: = sin2 θ = 2 sin θ cos θ + cos2 θ ,
4 4 4


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
Vertically fractured transversely isotropic media 183

or, after noting a11 a22 + a21 a12 = cos 2θ ≡ C2 and 2a11 a12 = 
s1122 = a11
2 2
a21 s1111 + 2a11 a21 C2 s1112 + (a11
2 2
a22 + a12
2 2
a21 )s1122
sin 2θ ≡ S2 ,
+ 4a11 a12 a21 a22 s1212 + 2a12 a22 C2 s2221 + a12
2 2
a22 s2222
⎡ ⎤
1 + C2 1 − C2 = cos2 θ sin2 θs1111 − 2 cos θ sin θC2 s1112
⎡ ⎤ S2 ⎡ ⎤

S55 ⎢ 2 2 ⎥ S55
⎢ ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥ + (1 − 2 cos2 θ sin2 θ)s1122
⎢ S  ⎥ = ⎢ − S2 C2
S2 ⎥ ⎢ S45 ⎥ . (A1)
⎣ 45 ⎦ ⎢ ⎥⎣ ⎦
⎢ 2 2 ⎥ − 4 cos2 θ sin2 θs1212 + 2 cos θ sin θC2 s2221
S ⎣ 1 − C2 1 + C2 ⎦ S44
44
−S2 + cos2 θ sin2 θs2222
 2  2   
A2 (θ) S22 S2
= s1111 − S2 C2 s1112 + 1 − 2 s1122 − S22 s1212
4 2
Note that |A 2 (θ )| ≡ 1 and that S22
+ S2 C2 s2221 s2222 ,
    4
      
= a11
2 2
a21 s1111 + 2a11 a21 C2 s1112 + 2a11 a21 a12 a22 s1122
 S44 S45   S44 S45  s1212
  ≡ S  S  − (S  )2 = S55 S44 − S 2 ≡  
   55 44 45 45  
 S45 
S55   S45 S55  + C22 s1212 + 2a12 a22 C2 s2221 + a12
2 2
a22 s2222
= cos2 θ sin2 θs1111 − 2 cos θ sin θC2 s1112
so this determinant is invariant under rotation about the 3-
− 2 cos2 θ sin2 θs1122 + C22 s1212
axis. Also the trace of the matrix of 2 a elements is invariant,

i.e., S55 
+ S44 = S55 + S44 . + 2 cos θ sin θC2 s2221 + cos2 θ sin2 θs2222
Consider rotation about the 3-axis of components with 0 S22 S2
= s1111 − S2 C2 s1112 − 2 s1122 + C22 s1212
‘three’s. After noting that 2 cos2 θ = 1 + C2 and 2 sin2 θ = 1 − 4 2
C2 so that S22
+ S2 C2 s2221 + s2222 ,
4

a11 a22 + 3a12 a21 = 2C2 − 1, s2221 = a21
3
a11 s1111 + a21
2
(2C2 + 1)s1112

3a11 a22 + a12 a21 = 2C2 + 1, + a22 a21 C2 (s1122 + 2s1212 )


3
4a11 a12 = 2a11
2
S2 = S2 (1 + C2 ), + a22
2
(2C2 − 1)s2221 + a12 a22
3
s2222
S2 (1 − C2 ) 1 − C2
3
4a11 a12 = 2a12
2
S2 = S2 (1 − C2 ), =− s1111 + (2C2 + 1)s1112
4 2
S2 C2
the transformation equation si jkl = aip ajq akr als spqrs applied to − (s1122 + 2s1212 )
2
the 0 ‘3’ components of the compliance tensor yields, 1 + C2 S2 (1 + C2 )
+ (2C2 − 1)s2221 + s2222 ,
2 4


s1111 = a11
4
s1111 + 4a11
3
a12 s1112 + 2a11
2 2
a12 (s1122 + 2s1212 ) s2222 = a21 s1111 + 4a21 a22 s1112 + 2a22 a21 (s1122 + 2s1212 )
4 3 2 2

+ 4a11 a12
3
s2221 + a12
4
s2222 + 4a21 a22
3
s2221 + a22
4
s2222

(1 + C2 )2 S2 (1 − C2 )2 S2
= s1111 + S2 (1 + C2 )s1112 + 2 (s1122 + 2s1212 ) = s1111 − S2 (1 − C2 )s1112 + 2 (s1122 + 2s1212 )
4 2 4 2
(1 − C2 )2 (1 + C2 )2
+ S2 (1 − C2 )s2221 + s2222 , − S2 (1 + C2 )s2221 + s2222 .
4 4

s1112 = a11
3
a21 s1111 + a11
2
(2C2 − 1)s1112 Recalling that
+ a11 a12 C2 (s1122 + 2s1212 ) ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
 
S11 s1111 S11 s1111
+ a12
2
(2C2 + 1)s2221 + 3
a12 a22 s2222 ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ S  ⎥ ⎢2s  ⎥ ⎢ S ⎥ ⎢2s ⎥
⎢ 16 ⎥ ⎢ 1112 ⎥ ⎢ 16 ⎥ ⎢ 1112 ⎥
S2 (1 + C2 ) (1 + C2 ) ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
=− s1111 + (2C2 − 1)s1112 ⎢  ⎥ ⎢  ⎥ ⎢ ⎥ ⎢ ⎥
4 2 ⎢ S12 ⎥ ⎢ s1122 ⎥ ⎢ S12 ⎥ ⎢ s1122 ⎥
⎢ ⎥=⎢ ⎥ and ⎢ ⎥=⎢ ⎥,
S2 C2 ⎢  ⎥ ⎢  ⎥ ⎢ ⎥ ⎢ ⎥
+ (s1122 + 2s1212 ) ⎢ S66 ⎥ ⎢4s1212 ⎥ ⎢ S66 ⎥ ⎢4s1212 ⎥
2 ⎢ ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢  ⎥ ⎢  ⎥ ⎢ ⎥ ⎢ ⎥
1 − C2 S2 (1 − C2 ) ⎢ S26 ⎥ ⎢2s2221 ⎥ ⎢ S26 ⎥ ⎢2s2221 ⎥
⎣ ⎦ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦
+ (2C2 + 1)s2221 + s2222  
2 4 S22 s2222 S22 s2222


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
184 M.A. Schoenberg

then the previous set of equations gives, Note that since −π /4 < (θ 2 + θ 1 )/2 < π /4, the constraint
⎡ ⎤ ⎡ ⎤ on is − π /2 < < π /2; similarly, the constraint on υ/ is

S11 S11
⎢ ⎥ ⎢ ⎥ 0 < θ 2 − θ 1 = υ/ < π /2. However, from inequality (B2), the
⎢S ⎥ ⎢S ⎥
⎢ 16 ⎥ ⎢ 16 ⎥ constraints on , υ/ can be written,
⎢ ⎥ ⎢ ⎥
⎢  ⎥ ⎢ ⎥
⎢ S12 ⎥ ⎢ S12 ⎥
⎢ ⎥ = Azero (θ ) ⎢ ⎥ , | | < υ/ < π/2, (B3)
⎢  ⎥ ⎢ ⎥
⎢ S66 ⎥ ⎢ S66 ⎥
⎢ ⎥ ⎢ ⎥ again defining a triangular region, now in the , υ/-plane where
⎢  ⎥ ⎢ ⎥
⎢ S26 ⎥ ⎢ S26 ⎥
⎣ ⎦ ⎣ ⎦ a solution must lie. Then using the identities

S22 S22

⎡ ⎤
(1 + C2 )2 /4 S2 (1 + C2 )/2 S22 /2 S22 /4 S2 (1 − C2 )/2 (1 − C2 )2 /4
⎢ ⎥
⎢−S (1 + C )/2 (1 + C )(2C − 1)/2 S2 C2 /2 (1 − C2 )(2C2 + 1)/2 S2 (1 − C2 )/2 ⎥
⎢ 2 2 2 2 S2 C2 ⎥
⎢ ⎥
⎢ ⎥
⎢ S2 /4
2
−S2 C2 /2 1 − S22 /2 −S22 /4 S2 C2 /2 S2 /4
2

Azero (θ ) = ⎢

⎥.
⎥ (A2)
⎢ S22 2S2 C2 −2S22 C22 2S2 C2 S22

⎢ ⎥
⎢ ⎥
⎢−S2 (1 − C2 )/2 (1 − C2 )(2C2 + 1)/2 −S2 C2 −S2 C2 /2 (1 + C2 )(2C2 − 1)/2 S2 (1 + C2 )/2 ⎥
⎣ ⎦
(1 − C2 )2 /4 −S2 (1 − C2 )/2 S22 /2 S22 /4 −S2 (1 + C2 )/2 (1 + C2 )2 /4

    
Note that S66 − 4S12 = S 66 − 4S 12 and S11 + S22 + 2S12 = S21 S22 ≡ sin2 cos2 υ/ − cos2 sin2 υ/ = sin2 cos2 υ/
S 11 + S 22 + 2S 12 , so these linear combinations are always
− cos2 (1 − cos2 υ/) = cos2 υ/ − cos2 ,
invariant under rotation about the 3-axis. Also note that
|Azero (θ )| ≡ 1 and that C21 C22 ≡ cos 2
sin υ/ − sin
2 2
sin υ/ = cos
2 2
cos2 υ/
     − sin2 (1 − cos2 υ/) = cos2 υ/ − sin2 ,
 S11 S12   
S16  
   S11 S12 S16 
   
S      and
 12 S22 S26  =  S12 S22 S26  ,
   
S 
S26  
S66 S S26 S66  S22 − S21 ≡ 2 cos sin υ/, C21 − C22 ≡ 2 sin sin υ/,
16 16

so this determinant is also invariant under rotation about the we can write, after noting the common factor 2sinυ/ in all
3-axis. matrix elements,

APPENDIX B

Solution of non-linear equation (21)

Explicitly, D 4 D−1
2 from equation (21) can be expanded in
terms of the C21 , S21 as follows:
⎡ ⎤
C41 S22 − C42 S21 −C41 C22 + C42 C21
sin (θ2 − θ1 )D4 D2 = ⎣
2 −1 ⎦
S41 S22 − S42 S21 −S41 C22 + S42 C21
⎡ ⎤
(1 − 2S221 )S22 − (1 − S222 )S21 −(2C221 − 1)C22 + (2C222 − 1)C21
=⎣ ⎦
2S21 C21 S22 − 2S22 C22 S21 −2S21 C21 C22 + 2S22 C22 C21
⎡ ⎤
(1 + 2S22 S21 )(S22 − S21 ) −(1 + 2C22 C21 )(C21 − C22 )
=⎣ ⎦. (B1)
2S21 S22 (C21 − C22 ) 2C21 C22 (S22 − S21 )

Let ≡ θ 2 + θ 1 and υ/ ≡ θ2 − θ1 where


− υ/ + υ/
= θ1 < 0 < θ2 = . (B2)
2 2


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185
Vertically fractured transversely isotropic media 185

⎡  ⎤
1 ⎣ 1 + 2(cos υ/ − cos ) cos − 1 + 2(cos2 υ/ − sin2 ) sin
2 2
D4 D−1 = ⎦
2
cos υ/ 2(cos2 υ/ − cos2 ) sin 2(cos2 υ/ − sin2 ) cos
⎡  ⎤
1 ⎣ 2 cos υ/ − cos − 2 cos2 υ/ + sin 2 sin
2 2
cos
= ⎦
cos υ/ 2(cos2 υ/ − cos2 ) sin 2(cos2 υ/ −sin2 ) cos
⎡ ⎤ ⎡ ⎤
cos −sin 1 cos 2 cos cos 2 sin
= 2 cos υ/ ⎣ ⎦− ⎣ ⎦
sin cos cos υ/ sin 2 cos sin 2 sin
⎡ ⎤ ⎡ ⎤
cos − sin 1 cos 2
= 2 cos υ/ ⎣ ⎦− ⎣ ⎦ [ cos sin ]. (B4)
sin cos cos υ/ sin 2

With this formulation of D 4 D−1


2 in terms of and υ/, equation
(21) becomes,
⎧ ⎡ ⎤ ⎡ ⎤ ⎫
⎨ cos − sin 1 cos 2 ⎬
q= 2 cos υ/ ⎣ ⎦− ⎣ ⎦ cos sin a. (B5)
⎩ sin cos cos υ/ sin 2 ⎭

 
1
Pre-multiplying these equations by cos 2 sin 2 q= 2 cos υ/ − cos sin a
   cos υ/   
  
⎡ ⎤ Q2 Q1 2A A2
sin 2 − cos 2 A1 − Q1
1
⎣ ⎦
cos 2 sin 2 Q21 − 2A21 (B8)
= A2 ,
Q1 A1

yields, where Q 2 and A 2 are two additional scalar trigonometric


functions of . Multiplication by Q 1 A 1 yields A 1 Q 1 Q 2 =
⎡ ⎤
sin 2 − cos 2 A 2 Q21 − 2A21 A 2 , which when expanded, gives,
⎣ ⎦q 
cos 2 sin 2 c0 + 2c2 sin2
sin = 1 − sin2 , (B9)
⎡ ⎤⎡ ⎤ c1 − 2c3 sin2
sin 2 − cos 2 cos − sin
= 2 cos υ/ ⎣ ⎦⎣ ⎦a  2   
cos 2 sin 2 sin cos c3 = 2a1 q1 q2 + a2 q2 − q12 − 3a12 − a22 ,
      
⎡ ⎤ c2 = 2a2 q1 q2 − a1 q22 − q12 − 3a22 − a12 ,
⎢ sin − cos ⎥    
⎢ ⎥
⎣ ⎦ c1 = 3a1 q1 q2 + a2 q22 − 2q12 − 2 2a12 − a22 ,
cos sin
⎡ ⎤ c0 = q2 (q2 a1 − q1 a2 ) − 2a1 a22 .
1 ⎣ 0 0
⎦ a,
− (B6) a single equation in sin . Squaring both sides of equa-
cos υ/ cos sin
tion (B9) gives, after multiplying by (c 1 − 2c 3 sin2 )2 and re-
arranging terms, a cubic equation on sin2 in standard form,
still a pair of scalar equations on and υ/. The first of these   3  
gives cos υ/ in terms of the, as yet unknown, : 4 c32 + c22 sin2 − 4 c3 c1 − c2 c0 + c22 (sin2 )2
  (B10)
+ c12 + c02 − 4c2 c0 sin2 − c02 = 0,
sin 2 − cos 2 q Q1 which has three roots, (sin2 ) i , i = 1, 2, 3, at least one of
cos υ/ = ≡ , (B7)
2 sin − cos a 2A1 which is real between 0 and 1. The actual roots of equation
(B9) then are
where Q 1 and A 1 are scalar trigonometric functions of ; the c0 + 2c2 (sin2 )i 
sin i 1 − (sin2 )i , i= 1, 2, 3. (B11)
second (making use of equation (B7) for cos υ/) yields, c1 − 2c3 (sin2 )i


C 2009 European Association of Geoscientists & Engineers, Geophysical Prospecting, 57, 169–185

You might also like