next up previous [pdf]

Next: Inverting for the virtual Up: Theory Previous: Wave mode separation and

Forward and adjoint isotropic elastic wave propagation

The isotropic elastic wave equation relates displacements to stresses via two elastic constants - the Lam$ \acute{e}$ parameters $ \lambda$ and $ \mu$ :

$\displaystyle \nabla \left( \left( \lambda + 2\mu \right) \nabla \cdot \mathbf ...
...\left( \mu \nabla \times \mathbf u \right) + \mathbf f = \rho \ddot{\mathbf u},$ (9)

where $ \mathbf u$ are the particle displacements in each dimension, $ \mathbf f$ is the force function and $ \rho$ is medium density. An alternate formulation is:

$\displaystyle \nabla \left( \left( \lambda + \mu \right) \nabla \cdot \mathbf u...
... \cdot \left( \mu \nabla \mathbf u \right) + \mathbf f = \rho \ddot{\mathbf u}.$ (10)

From equation 10, the explicit form for a heterogeneous two dimensional medium can be expressed in a matrix-vector notation as:

$\displaystyle \begin{bmatrix}\frac{1}{\rho} \partial_x \left( \left( \lambda + ...
...d{bmatrix} = \begin{bmatrix}\partial^2_t u_x \\ \partial^2_t u_z \end{bmatrix}.$ (11)

The forward elastic propagation operator can then be expressed as:

$\displaystyle \mathbf F = \begin{bmatrix}\frac{1}{\rho} \partial_x \left( \lamb...
...rtial_z + \frac{1}{\rho} \partial_x \mu \partial_x - \partial^2_t \end{bmatrix}$ (12)

For a homogeneous medium, and using a Green's function to describe the energy propagation between any two locations $ \mathbf x = \left( x, y, z \right)$ and $ \mathbf y = \left( x, y, z \right)$ , the equation takes the form:

$\displaystyle \left( \left( \lambda + \mu \right) \nabla \nabla + \mu \nabla^2 ...
...bf x, \mathbf y, \omega \right) = - \delta \left( \mathbf x - \mathbf y \right)$ (13)

The forward elastic propagation operator injects sources from a model into some location in the medium, and records the resulting wavefield at some other location:

$\displaystyle d \left( \mathbf x, \omega; \mathbf x_s \right) = \sum_{\mathbf y...
... y, \omega; \mathbf x_s \right) G \left( \mathbf y, \mathbf x, \omega \right) ,$ (14)

where $ m$ is the model of injected sources at location $ \mathbf y$ in the medium, and $ d$ are the recorded displacement fields $ \mathbf u$ at location $ \mathbf x$ in the medium. $ \omega$ is angular frequency and $ \mathbf x_s$ is the shot gather. In vector notation, this is expressed as

$\displaystyle \mathbf d = \mathbf {Fm},$ (15)

where $ \mathbf F$ is the forward operator. The adjoint operator injects the data from the same recording locations, and records the resulting wavefield at the model injection points:

$\displaystyle \tilde m \left( \mathbf y, \omega; \mathbf x_s \right) = \sum_{\m...
..., \omega; \mathbf x_s \right) G^* \left( \mathbf x, \mathbf y, \omega \right) ,$ (16)

which in vector notation is

$\displaystyle \mathbf {\tilde m} = \mathbf {F^*d},$ (17)

where $ \mathbf F^*$ is the adjoint operator.


next up previous [pdf]

Next: Inverting for the virtual Up: Theory Previous: Wave mode separation and

2012-05-10