As we know, isotropic elastic models are described by three familiar parameters—density, compressional velocity, and shear velocity. To model elastic wavefields in such three-parameter media, we need to derive a suitable equation or set of equations that describe the wavefield propagation at any given instant. Unfortunately, these three parameters are not the most useful for this purpose. On the other hand, the required parameters are much more directly related to actual rock properties. Moreover, relating the required parameters to more useful quantities is fairly straightforward.
We begin (Equation 8 and Equation 7) by observing that Equation 20 is true, so that Equation 21 is also true. In continuous terms, Equation 21 can be restated as Equation 22.
![]() | (21) |
![]() | (22) |
Equation 22 is a first order partial-differential equation relating a second order change in time to a first order change in force per unit area. Force per unit area is generally referred to as stress, so our equation relates particle acceleration to stress. In this setting, the stress is one-dimensional and acts along or parallel to the layout of the string. There is also only one compressional wavefield described by this equation. Stepping up to the simple isotropic elastic models described by the three familiar parameters above, means that it is necessary to include one additional wavefield in the mix, namely shear. While including just two wavefields is certainly an option, there isn't any reason not to move up to full anisotropic elasticity by incorporating two shear waves for a total of three wavefields. In three dimensions, Equation 22 takes the form of Equation 23, where i = 1 ; 2 ; 3 and we have arbitrarily chosen to set x = x 1, y = x 2, and z = x 3.
![]() | (23) |
Clearly, this is a three-dimensional equation with nine stress factors, τ ij, one for each of the three dimensions and wavefields. To make this into a system of equations governing the three wavefields, we must find a way to relate the stresses, τ ij to the u i. As before, Hooke's law comes to the rescue. What it says in this anisotropic case is:
Each component of stress is linearly proportional to every component of strain:
Strain, which measures the deformation (compression, extension, ...) of a solid, is defined in the notation of the previous equation as Equation 24.
![]() | (24) |
The mathematical expression of Hooke's law then takes the form of Equation 25.
![]() | (25) |
Inserting Equation 25 into Equation 23, we finally get the complex system ( i = 1 ; 2 ; 3) of fully anisotropic equations of motion, Equation 26.
![]() | (26) |
Because equation Equation 26 is three-dimensional, each of the c ijmn coefficients is actually a three-dimensional volume. Even today's massive supercomputers may not have sufficient memory to handle a problem of this size.
It would not be hard to throw up our hands at this point and give up, but, before we panic too much, we might want to analyze the situation a bit. As it turns out there are at least two things we can do to simplify the situation considerably. First, we can simplify the mathematical notation to put us into a setting where we can make some sense of the parameters, and second, we can reformulate the c ijmn coefficients in a way that will make a great deal more physical sense.
We are not really interested as much in the math as we are in understanding the kinds of Earth models these c ijmn coefficients define for us. We need to know how the various velocities of the wavefields that propagate in the medium are defined. We want to see if we can understand how direction changes the speed of propagation, and then see if we can find ways to estimate parameters that can be converted into c ijmn volumes to both synthesize data and image data we have recorded over fully elastic models.
The first simplification to the complexity of Equation 26 is based on the symmetry of stress and strain. Here, the ij indices representing stress can be switched so that c ijmn = c jimn. Similarly, the strain based indices can also be switched so that c ijmn = c ijnm. Finally, it is also true that c ijmn = c mnij. This triple symmetry means that the total number of c ijmn volumes has been reduced from only 21! Thus, defining the most general model we can imagine requires only 21 independent parameters (volumes), instead of the 81 parameters we would need without symmetry.
By applying the indexing scheme (known as the Voigt scheme) in Equation 27, we arrive at the 6x6 matrix shown in Equation 28.
![]() | (27) |
![]() | (28) |
This matrix completely describes the unique set of 21 coefficients fully defining anisotropic elasticity. In this case, the C matrix is the most complicated form of anisotropy we can encounter. For the interested reader this case is called "triclinic" symmetry and is probably something we will not be able to investigate computationally until computers have advance significantly beyond where they are today. Moreover, we may never be able to measure sufficient data to be able to estimate all of these parameters. Consequently, we will focus on what we consider reasonable today.
In what perhaps is overkill, the C matrix, for a purely acoustic medium, takes the form shown in Equation 29.
![]() | (29) |
Here,
λ is the first of the two so-called "Lame" parameters, which are material properties
(proportionality constants) that relate stress to strain. In this case,
λ is directly related to the bulk
modulus
K and the compressional velocity is
v
p
=
=
. If we define the actual
c
ijmn from the elements of
C, plug them back into the fully anisotropic Equation
26, we see that
u
1
=
u
2
=
u
3
=
u, and consequently that
Equation
26 reduces to Equation
30, which is, of course, the normal three-dimensional scalar wave
equation.
![]() | (30) |