This section describes the things we must do to handle model boundaries.
Handling a free surface is probably the most complex of the various problems that arise in seismic modeling exercises. The literature on this aspect of the synthesis is quite vast and outside the scope of what we wish to discuss here. We leave detailed investigation of this to you, if you are interested.
However, one of the more appealing methods is discussed by Lavendar in his 1988 paper on P-SV modeling, and illustrated in Figure 25. The essential difference lies in how each layer is handled. Turning on (or off) free surface reverberations controls whether or not synthetic data contains multiples and ghosts.
(a)
Free
Surface
Boundary
Layers
|
The free surface at the top of the model is padded above with a fictitious set of nodes. Since a free surface implies that no normal or shear stress are active there, we can set τ 2 ; 2 = 0 and τ 1 ; 2 = 0 at the top. The shear stress boundary condition is handled by setting it to zero at z = 0 as well. The normal stress is not defined at the top boundary but is forced to zero by making the normal stress antisymmetric for the first two rows above the free surface, as shown in Equation 101.
There are a variety of approaches for handling the other boundaries in a typical seismic Earth model. The three most popular are what are called sponge boundary conditions, absorbing, boundary conditions, and the so called perfectly matched layers.
The idea behind sponge boundary conditions is to modify the propagating equation by adding viscosity to equation along the boundary. This is normally accomplished by writing Equation 102, where 0 γ is an absorbing parameter chosen to produce a wave that decreases in amplitude with distance.
![]() | (102) |
Gamma is usually chosen to have exponential decay within the defined boundary zone and is zero within the model dimensions. Note that when γ = 0, the solution to the equation is, in fact, p.
For the finite element method, sponge boundaries can be implemented by changing the definition of φ ( U;V ) to Equation 103, where α and β are the damping factors in each of the boundary layers.
![]() | (103) |
Perfectly matched layers represent a more modern treatment of the sponge boundary conditions. In this setting, the spatial derivatives are modified so that we have Equation 104.
![]() | (104) |
Paraxial boundaries are based on the one-way wave equation and within the boundary layers take the form in
Equation
105, where
j
α
j
j
<
for all
j (Higdon 1991).
![]() | (105) |
This equation works because each factor cos (