[cig-commits] [commit] devel: added more details about PML to the manual (about regular elements inside the PMLs for better time stability) (783b2cd)
cig_noreply at geodynamics.org
cig_noreply at geodynamics.org
Mon Jan 26 16:49:26 PST 2015
Repository : https://github.com/geodynamics/specfem3d
On branch : devel
Link : https://github.com/geodynamics/specfem3d/compare/4720296dd98ef99c70e9ad3d9fd6fc550475f60b...367727d9b25a3dc40b9c10fee9dc3275eff7b717
>---------------------------------------------------------------
commit 783b2cdc0a7434927200c1b08b1ed160ffa291d4
Author: Dimitri Komatitsch <komatitsch at lma.cnrs-mrs.fr>
Date: Tue Jan 27 01:43:03 2015 +0100
added more details about PML to the manual (about regular elements inside the PMLs for better time stability)
>---------------------------------------------------------------
783b2cdc0a7434927200c1b08b1ed160ffa291d4
doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.pdf | Bin 14049638 -> 14050378 bytes
doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.tex | 7 +++++++
2 files changed, 7 insertions(+)
diff --git a/doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.pdf b/doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.pdf
index 7c40a59..ab2f74d 100644
Binary files a/doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.pdf and b/doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.pdf differ
diff --git a/doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.tex b/doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.tex
index a391639..701c7ca 100644
--- a/doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.tex
+++ b/doc/USER_MANUAL/manual_SPECFEM3D_Cartesian.tex
@@ -1837,6 +1837,13 @@ model will change from viscoelastic or poroelastic to elastic at the
entrance of the PML, but in practice this is sufficient and produces
only tiny / negligible spurious reflections.
+If you use PML and an external mesh (created using an external meshing tool
+such as CUBIT/TRELIS or similar), try to have elements inside the PML as regular as possible,
+i.e. ideally non-deformed cubes obtained by `extrusion' of regular surface mesh elements meshing the
+outer edges of the computational domain without PML; by doing so, the PMLs obtained will be far more stable
+in time (PML being weakly unstable from a mathematical point of view, very deformed mesh elements
+inside the PMLs can trigger instabilities much more quickly).
+
If you use PML and an external tomographic velocity and density model,
you should be careful because mathematically a PML cannot handle heterogeneities
along the normal to the PML edge inside the PML layer. This comes
More information about the CIG-COMMITS
mailing list