[cig-commits] r18844 - seismo/3D/SPECFEM3D/trunk/src/specfem3D

[dkomati1 at geodynamics.org]

Author: dkomati1
Date: 2011-08-23 08:55:21 -0700 (Tue, 23 Aug 2011)
New Revision: 18844

Modified:
   seismo/3D/SPECFEM3D/trunk/src/specfem3D/specfem3D.f90
Log:
added a comment about the fact that acoustic-acoustic interfaces (for instance for purely acoustic simulations for the oil industry) are handled fine by the code because the acoustic potential used in a potential of (density * displacement), as it should be


Modified: seismo/3D/SPECFEM3D/trunk/src/specfem3D/specfem3D.f90
===================================================================
--- seismo/3D/SPECFEM3D/trunk/src/specfem3D/specfem3D.f90	2011-08-23 01:23:54 UTC (rev 18843)
+++ seismo/3D/SPECFEM3D/trunk/src/specfem3D/specfem3D.f90	2011-08-23 15:55:21 UTC (rev 18844)
@@ -271,6 +271,30 @@
 !    parallelized on 128 processors using Connection Machine Fortran
 !
 
+! in case of an acoustic medium, a potential Chi of (density * displacement) is used as in Chaljub and Valette,
+! Geophysical Journal International, vol. 158, p. 131-141 (2004) and *NOT* a velocity potential
+! as in Komatitsch and Tromp, Geophysical Journal International, vol. 150, p. 303-318 (2002).
+!
+! This permits acoustic-elastic coupling based on a non-iterative time scheme.
+! Displacement is then:
+!     u = grad(Chi) / rho
+! Velocity is then:
+!     v = grad(Chi_dot) / rho
+! (Chi_dot being the time derivative of Chi)
+! and pressure is:
+!     p = - Chi_dot_dot
+! (Chi_dot_dot being the time second derivative of Chi).
+!
+! The source in an acoustic element is a pressure source.
+!
+! First-order acoustic-acoustic discontinuities are also handled automatically
+! because pressure is continuous at such an interface, therefore Chi_dot_dot
+! is continuous, therefore Chi is also continuous, which is consistent with
+! the spectral-element basis functions and with the assembling process.
+! This is the reason why a simple displacement potential u = grad(Chi) would
+! not work because it would be discontinuous at such an interface and would
+! therefore not be consistent with the basis functions.
+
 ! ************** PROGRAM STARTS HERE **************
 
   ! force Flush-To-Zero if available to avoid very slow Gradual Underflow trapping