SEM2DPACK Geophysical Simulation Software User's Guide

SEM2DPACK
A Spectral Element Method tool for 2D wave propagation
and earthquake source dynamics
User’s Guide
Version 2.3.8
August 2012
Jean-Paul Ampuero
California Institute of Technology
Seismological Laboratory
1200 E. California Blvd., MC 252-21
Pasadena, CA 91125-2100, USA
E-mail: [email protected]
Web: http://web.gps.caltech.edu/~ampuero
Phone: (626) 395-6958
Fax : (626) 564-0715
c 2003-2012 Jean-Paul Ampuero
Contents
1 Introduction
1.1 Overview . . . . . . .
1.2 History and credits . .
1.3 Application examples .
1.4 Download and updates
1.5 Requirements . . . . .
1.6 Installation . . . . . .
1.7 Documentation . . . .
1.8 Support . . . . . . . .
1.9 Contributions . . . . .
1.10 License . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
5
6
6
7
7
8
8
8
9
9
2 Physical background
2.1 General assumptions and conventions . . . . .
2.2 Material rheologies . . . . . . . . . . . . . . .
2.2.1 Linear elasticity . . . . . . . . . . . .
2.2.2 Linear visco-elasticity . . . . . . . . .
2.2.3 Coulomb plasticity and visco-plasticity
2.2.4 Continuum damage . . . . . . . . . .
2.3 Boundary conditions . . . . . . . . . . . . . .
2.3.1 Absorbing boundaries . . . . . . . . .
2.4 Fault interface conditions . . . . . . . . . . .
2.4.1 Linear slip law . . . . . . . . . . . . .
2.4.2 Normal stress response . . . . . . . . .
2.4.3 Friction . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
11
12
12
13
13
14
15
15
16
16
16
17
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
. . . . . . . . . . . .
generation software
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19
19
20
20
21
21
22
22
25
25
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3 Mesh generation
3.1 General guidelines . . . . . . . . . . . . . . . . . . .
3.2 Meshing features included in the solver . . . . . . . .
3.3 Meshing with the MESH2D Matlab utilities . . . . .
3.4 Generating a mesh with EMC2 . . . . . . . . . . . .
3.4.1 The mesh generator EMC2 . . . . . . . . . .
3.4.2 Notations . . . . . . . . . . . . . . . . . . . .
3.4.3 Basic step-by-step . . . . . . . . . . . . . . .
3.4.4 Further tips . . . . . . . . . . . . . . . . . . .
3.5 Importing a mesh from CUBIT, EZ4U or other mesh
CONTENTS
4 The
4.1
4.2
4.3
4.4
4.5
4.6
solver SEM2D
About the method . . . . . . . . . . . . . . . .
Basic usage flow . . . . . . . . . . . . . . . . .
General format of the input file . . . . . . . . .
Input Blocks Reference Guide . . . . . . . . . .
Verifying the settings and running a simulation
Outputs, their visualization and manipulation .
4.6.1 Spectral element grid . . . . . . . . . .
4.6.2 Source time function . . . . . . . . . . .
4.6.3 Snapshots . . . . . . . . . . . . . . . . .
4.6.4 Seismograms . . . . . . . . . . . . . . .
4.6.5 Fault outputs . . . . . . . . . . . . . . .
4.6.6 Stress glut . . . . . . . . . . . . . . . . .
4.6.7 Energies . . . . . . . . . . . . . . . . . .
4.6.8 Matlab utilities . . . . . . . . . . . . . .
4
.
.
.
.
.
.
.
.
.
.
.
.
.
.
27
27
28
28
32
54
59
59
59
59
61
61
63
63
63
5 Adding features to SEM2D (notes for advanced users)
5.1 Overview of the code architecture . . . . . . . . . . . . . . . . . . . . . . . . .
5.2 Accessible areas of the code . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
65
65
6 Frequently Asked Questions
6.1 SEM2D . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2 EMC2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
67
67
67
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Chapter 1
Introduction
1.1
Overview
The SEM2DPACK package is a set of software tools for the simulation and analysis of 2D
wave propagation and dynamic fracture, with emphasis on computational seismology and
earthquake dynamics. The core of the package is SEM2D, a solver for the 2D elastic wave
equations and dynamic earthquake rupture based on the Spectral Element Method (SEM)
with explicit time stepping. Chapter 2 of this User’s Guide summarizes the range of problems
that can be solved with SEM2DPACK. Section 4.1 provides some background on the SEM.
The essential properties of the method are its high order accuracy, affordable at competitive
computational cost, its geometrical flexibility to treat realistic, complicated crustal structures,
and its natural treatment of mixed boundary conditions such as fault friction.
SEM2DPACK provides tools for each step of the general flow of a simulation project:
1. Mesh generation: partition the domain into (deformed) quadrilateral elements. Whereas
no general mesh generation code is included, SEM2DPACK contains basic meshing utilities for structured and semi-structured grids and can import unstructured quadrilateral
meshes generated externally. These features are described in Chapter 3.
2. Mesh quality verification: check the accuracy, stability and computational cost,
applying tools described in Section 4.5. Return to previous step if needed.
3. Numerical simulation: run the SEM2D solver. Chapter 4 explains its usage.
4. Post-processing: visualization and analysis of the output. A number of post-processing
and graphic tools are included, as decribed in Section 4.6. Outputs are in the form of
binary data files, ASCII data files and PostScript figure files. Scripts are provided for
graphic display and analysis on Seismic Unix, Gnuplot and Matlab. We recommend the
usage of the Matlab functions included (see Section 4.6.8).
This is a research code, constantly under development and provided “as is”, and therefore
it should not be considered by the user as a 100% bug-free software package. We welcome
1.2 History and credits
6
comments, suggestions, feature requests, bug reports (see Section 1.8) and contributions to
the code itself (see Section 1.9).
1.2
History and credits
The main part of the elastic-isotropic solver was written by Dimitri Komatitsch as part of
his Ph.D. thesis (Komatitsch, 1997) under the direction of Prof. Jean-Pierre Vilotte at the
Institut de Physique du Globe de Paris (IPGP). The elastic-anisotropic solver and several
significant improvements were added by D. Komatitsch later as part of a research contract
with DIA Consultants. Further functionalities were added by Jean-Paul Ampuero, as part
of a Ph.D. thesis (Ampuero, 2002) also directed by Prof. Vilotte at IPGP. Most of these
additional features were motivated by an ECOS-NORD/FONACYT research project for the
study of the seismic response of the valley of Caracas, Venezuela. That became the version
1.0 of the SEM2DPACK, released in April 2002.
For version 2.0, most of the solver was rewritten in preparation for the implementation of
higher level functionalities. Developments for the simulation of earthquake dynamics (Ampuero, 2002) were included in the main branch of SEM2DPACK in October 2003 (version
2.2). Spontaneous rupture along multiple non-planar faults can be currently modelled, with
a range of friction laws.
Non-linear, inelastic materials were introduced in March 2008 (version 2.3). Damage and
visco-plastic rheologies are included especially for the modeling of earthquake rupture with
off-fault dissipation.
1.3
Application examples
SEM2DPACK has been utilized in a variety of applications in Earth sciences and has contributed to more than 20 publications:
– dynamic rupture on non-planar faults and seismic wave radiation (Madariaga et al., 2006)
– wave propagation in anisotropic TTI media (Dewangan et al., 2006)
– fault reflections from fluid-infiltrated faults (Haney et al., 2007)
– benchmark for anisotropic wave propagation (De la Puente et al., 2007)
– dynamic earthquake rupture with rate-and-state friction (Kaneko et al., 2008)
– benchmark for dynamic earthquake rupture simulation (De la Puente et al., 2009)
– non-linear wave propagation in damaged rocks (Lyakhovsky et al., 2009)
– modeling marine seismic profiles (Roberts et al., 2009)
– surface wave propagation in applied geophysics (Vignoli and Cassiani, 2010; Vignoli et al.,
2011; Vignoli et al., 2012; Boaga et al., 2012)
1.4 Download and updates
7
– wave propagation around a prototype nuclear waste storage tunnel (Smith and Snieder,
2010)
– earthquake dynamic rupture with off-fault plasticity (Harris and et al, 2011; Gabriel
et al., 2012a; Xu et al., 2012a; Xu et al., 2012b)
– earthquake rupture in heterogeneous media (Huang and Ampuero, 2011; Huang and
Ampuero, 2012)
– benchmark for wave propagation in heterogeneus media (O’Brien and Bean, 2011)
– dynamic rupture model of the 2011 Tohoku earthquake (Huang et al., 2012)
– dynamic rupture model of the 2012 off-Sumatra earthquake (Meng and Ampuero, 2012)
– earthquake rupture with velocity-and-state-dependent friction (Gabriel et al., 2012b)
1.4
Download and updates
SEM2DPACK is hosted by SourceForge at
http://sourceforge.net/projects/sem2d/.
All versions of the code can be downloaded from the package repository at
http://sourceforge.net/projects/sem2d/files/sem2dpack/.
Taking full advantage of the convenient features offered by SourceForge (subscribe to new
release announcements, submit and track bug reports) requires a SourceForge.net account,
which can be created at
http://sourceforge.net/account/registration/.
SEM2DPACK is updated regularly, typically every few months. To receive email notifications
about new releases you must sign up for the “Update Notifications” in the project’s main page
(scroll down a bit):
http://sourceforge.net/projects/sem2d/.
1.5
Requirements
Compiling the solver code requires the make utility and a Fortran 95 compiler. The code is
being developed with the Intel compiler for Linux. It works properly with the Intel compiler
starting with version 8.0.046 pe047.1, so make sure you have a recent version of ifort. Other
compilers are not being tested on a regular basis, so please report any related problems.
The solver runs under the Linux operating system. In particular, input/output file name
conventions are specific to Linux. Other operating systems have not been tested.
Pre-processing and post-processing tools, including graphic visualization, are provided for
Seismic Unix, Gnuplot, GMT and Matlab. The included Matlab tools are by far the most
complete, so a Matlab license is highly recommended. Matlab “clone” softwares have not
been tested.
1.6 Installation
1.6
8
Installation
1. Uncompress and expand the SEM2DPACK package: tar xvfz sem2dpack.tgz
2. Go to the source directory: cd SEM2DPACK/SRC
3. Edit the Makefile according to your FORTRAN 95 compiler, following the instructions
therein.
4. Modify the optimization parameters declared and described in SRC/constant.f90.
5. Compile: make
6. Move to the SEM2DPACK/POST directory, edit the Makefile and compile.
On normal termination you should end up with a set of executable files, among which
sem2dsolve, in /home/yourhome/bin/.
1.7
Documentation
Documentation is available through the following resources:
– This User’s Manual
– The EXAMPLES directory contains several examples, some have a README file
– The pre-processing and post-processing tools for Matlab are documented through Matlab’s help. For instance help mesh2d provides an overview of the MESH2D utilities, and
help mesh2d wedge provides detailed documentation for the wedge meshing function
– The ToDo file contains a list of known issues
1.8
Support
Support for users of SEM2DPACK is available through a tracking system at
http://sourceforge.net/tracker/?group_id=182742,
Three separate tracker lists deal with the following aspects:
– Feature Requests: requests for implementation of new features
– Support Requests: questions related to the usage of SEM2DPACK
– Bugs: bug reports
Before submitting an issue make sure that:
1. you have read the documentation (see Section 1.7), including the Frequently Asked
Questions (Chapter 6). Suggestions on how to improve the documentation are also
welcome.
1.9 Contributions
9
2. you are running the most recent version of SEM2DPACK. Your issue might have been
already fixed in a more recent version.
3. you understand the changes listed in SEM2DPACK’s ChangeLog file, especially changes
in the format of the input files
4. your problem has not been treated in previous submissions. You can browse the tracker
message titles or search for keywords.
A new submission must include the input files needed to reproduce your problem (Par.inp,
*.ftq, *.mesh2d, etc). You will receive email notifications of any update of your submitted
item, until it is closed. If the item is declared “Pending” you are expected to reply to the
last message of the developer within two weeks, otherwise the item will be closed. For more
instructions see
http://sourceforge.net/support/getsupport.php?group_id=182742.
1.9
Contributions
Contributions to SEM2DPACK by experienced programmers are always welcome and encouraged. Although the code is stable for typical applications in computational seismology
and earthquake dynamics, there is still a number of missing features. Their implementation could make SEM2DPACK interesting for a broader audience in mechanical engineering,
geotechnical engineering, applied geophysics and beyond.
The solver code is written in FORTRAN 95. Resources available for programmers include:
– A ToDo file included with SEM2DPACK contains a wish list that ranges from basic
functionalities to complex code re-engineering.
– Chapter 5 gives some guidelines for programmers.
– A “Developers Forum” to discuss the implementation of new features is available at
http://sourceforge.net/forum/forum.php?forum_id=635737,
1.10
License
This software is freely available for academic research purposes. If you use this software in
writing scientific papers include proper attributions to its author, Jean-Paul Ampuero.
This program is free software; you can redistribute it and/or modify it under the terms of the
GNU General Public License as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A
PARTICULAR PURPOSE. See the GNU General Public License for more details.
1.10 License
10
You should have received a copy of the GNU General Public License along with this program;
if not, write to the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
02111-1307, USA.
Chapter 2
Physical background
This chapter summarizes the physical assumptions and notations in SEM2DPACK. Footnotes
provide reference to the input arguments described in Chapter 4.
2.1
General assumptions and conventions
The coordinate sytem is Cartesian (rectangular). SEM2DPACK works in the two-dimensional
(x, z) plane, where x is the horizontal coordinate, with positive direction pointing to the right,
and z is the vertical coordinate, with positive direction pointing upwards. The coordinates
(x, y, z) will be also denoted as (x1 , x2 , x3 ). This notations carry also for subscripts. For
instance, the k-th component of displacement is denoted as uk , with k = 1, 2, 3 or with
k = x, y, z.
The reference frame is Eulerian. Infinitesimal strain is assumed. The (symmetric) infinitesimal strain tensor ǫ is defined as
∂uj
1 ∂ui
+
(2.1)
ǫij =
2 ∂xj
∂xi
Material density is deonted ρ(x, z). The displacements and stresses relative to an initial
equilibrium configuration are denoted uk (x, z, t) and σij (x, z, t), respectively. External forces
(sources) are denoted fi (x, z, t). SEM2DPACK solves the following equations of motion to
obtain the relative displacements uk (x, z, t):
ρ
∂σij
∂2u
=
+ fi
2
∂t
∂xj
(2.2)
where summation over repeated indices is assumed. The initial conditions are uk = 0 and
∂uk /∂t = 0. Stresses are related to strain, and possibly to other internal variables, by
constitutive equations described in Section 2.2. The governing equations are supplemented by
boundary conditions, described in Section 2.3. SEM2DPACK actually solves the governing
2.2 Material rheologies
12
equations in variational (weak) form, as described in any textbook on the finite element
method.
Two types of 2D problems are solved1 :
– Plane strain: Also known in seismology as P-SV, and in fracture mechanics as inplane
mode or mode II. It is assumed that u3 = 0 and ∂/∂x3 = 0. Hence, ǫ13 = ǫ23 = ǫ33 = 0
and there are two degrees of freedom per node, ux and uz .
– Antiplane shear : Also known in seismology as SH, and in fracture mechanics as antiplane
mode or mode III. It is assumed that u1 = u2 = 0 and ∂/∂x3 = 0. Hence, only ǫ13 and
ǫ23 are non-zero and there is one degree of freedom per node, uy .
2.2
Material rheologies
We describe here the constitutive equations implemented in SEM2DPACK, relating stress
(σij ), strain (ǫij ) and internal variables.
2.2.1
Linear elasticity
Linear isotropic elasticity
Stress and strain are linearly related by Hooke’s law, σij = cijkl ǫij , where cijkl is the tensor
of elastic moduli. In particular, for isotropic elasticity:
σij = λǫkk δij + 2µǫij
(2.3)
where λ and µ are Lamé’s first and second parameters, respectively. In 2D plane strain the
only relevant stress components are σ11 , σ22 and σ12 . The intermediate stress σ33 , although
p
not null, does
not
enter
in
the
equations
of
motion.
The
S
and
P
wave
speeds
are
c
=
µ/ρ
S
p
and cP = (λ + 2µ)/ρ, respectively. In 2D antiplane shear only the stress components σ13
and σ23 are relevant, and only S waves are generated.
Linear anisotropic elasticity
Transverse anisotropy with vertical symmetry axis (VTI) is implemented for 2D P-SV (Komatitsch et al., 2000). The stress-strain constitutive relation for P-SV in Voigt notation
is:

 
 

σxx
c11 c13 0
ǫxx
 σzz  =  c13 c33 0   ǫzz 
(2.4)
σxz
0
0 c55
2ǫxz
where the cij are elastic moduli.
1
In the &GENERAL input block, plane strain is selected by ndof=2 and antiplane shear by ndof=1.
2.2 Material rheologies
2.2.2
13
Linear visco-elasticity
Generalized Maxwell material
Not implemented yet.
Kelvin-Voigt material
Kelvin-Voigt damping can be combined with any of the other constitutive equations by replacing the elastic strain ǫ by ǫ∗ = ǫ + η ∂ǫ/∂t, where η is a viscosity timescale.
The resulting quality factor Q is frequency-dependent, Q−1 (f ) = 2πηf . This rheology is
not approriate to model crustal attenuation with constant Q, unless the source has a narrow
frequency band and η is selected to achieve a given Q value at the dominant frequency of the
source.
The main application of Kelvin-Voigt viscosity is the artificial damping of high-frequency
numerical artifacts generated by dynamic faults. Dynamic source simulations using methods that discretize the bulk, such as finite difference, finite element and spectral element
methods, are prone to high frequency numerical noise when the size of the process zone
is not well resolved. Efficient damping is typically achieved by a thin layer of KelvinVoigt elements surrounding the fault, with thickness of 1 or 2 elements on each side of
the fault and η/∆tf ault = 0.1 to 0.3, where ∆tf ault is the critical time step size based on
the size of the spectral elements along the fault (not necessarily equal to the critical time
step over the whole mesh). The value of ∆tf ault can be obtained with the Matlab function
PRE/critical timestep.m.
2.2.3
Coulomb plasticity and visco-plasticity
Perfect plasticity
Perfect plasticity with a Coulomb yield function is implemented for 2D plane strain, as in
Andrews (2005).
The total strain is the sum of an elastic and a plastic contribution, ǫ = ǫe + ǫp . The plastic
strain is assumed to be purely deviatoric (ǫpkk = 0). Plastic yield occurs when the maximum
shear stress over all orientations,
p
2 + (σ
2
(2.5)
τmax = σxz
xx − σzz ) /4,
reaches the yield strength,
Y = c cos(φ) − (σxx + σzz ) sin(φ)/2,
where c is the cohesion and φ is the internal friction angle.
(2.6)
2.2 Material rheologies
14
Visco-plasticity
In classical Duvaut-Lions visco-plasticity the (visco-)plastic strain rate is proportional to the
excess of stress over the yield strength:
τij
1
hτmax − Y i
(2.7)
2µ Tv
τmax
.
where Tv is the visco-plastic relaxation time, hxi = (x + |x|)/2 is the ramp function and
τij = σij − 13 σkk δij is the deviatoric part of the stress tensor.
ǫ̇pkl =
Visco-plasticity is often employed as a regularization of plasticity to avoid or delay the occurrence of strain localization features, such as shear bands, that are mesh-dependent. For that
particular application, Tv is typically set to the average P wave traveltime across a few grid
points, i.e. a few times the average spacing between GLL nodes divided by the P wave speed.
2.2.4
Continuum damage
The continuum damage formulation by Lyakhovsky et al. (1997), including damage-related
plasticity as introduced by Hamiel et al. (2004), is implemented with modifications for 2D
plane strain.
The first and second invariants of the 2D elastic strain tensor are defined
as I1 = ǫekk and
√
I2 = ǫeij ǫeij , respectively. A strain invariant ratio is defined as ξ = I1 / I2 . The following
non-linear stress-strain relation is assumed (Lyakhovsky et al., 1997, eq. 12):
σij = (λ − γ/ξ) I1 δij + (2µ − γξ) ǫeij
(2.8)
where γ is an additional elastic modulus. The elastic moduli depend on a scalar damage
variable, 0 ≤ α ≤ 1, through (Lyakhovsky et al., 1997, eq. 19):
λ = λ0
(2.9)
µ = µ0 + γr ξ0 α
(2.10)
γ = γr α
(2.11)
where λ0 and µ0 are Lamé’s parameters for the intact material (α = 0). The parameter ξ0
is the threshold value of the strain invariant ratio ξ at the onset of damage. It is related to
the internal friction angle φ in a cohensionless Mohr-Coulomb yield criterion by the 2D plane
strain version of Lyakhovsky et al. (1997, eq. 37):
√
− 2
(2.12)
ξ0 = p
1 + (λ0 /µ0 + 1)2 sin2 φ
The scaling factor γr is chosen such that convexity is lost at α = 1 when ξ = ξ0 . It is derived
from the 2D plane strain version of Lyakhovsky et al. (1997, eq. 15):
p
γr = p + p2 + 2µ0 q
(2.13)
2.3 Boundary conditions
15
where
q = (2µ0 + 2λ0 )/(2 − ξ02 )
p = ξ0 (q + λ0 )/2
The evolution equation for the damage variable is (Lyakhovsky et al., 1997, eq. 20)
Cd I2 (ξ − ξ0 ) if ξ > ξ0
α̇ =
0
otherwise
(2.14)
(2.15)
(2.16)
No healing is assumed below ξ0 . The evolution of the plastic strain ǫpij is driven by the damage
variable α (Hamiel et al., 2004, eq. 9):
τij Cv α̇ if α̇ ≥ 0
p
(2.17)
ǫ̇ij =
0
otherwise
where τij = σij − 31 σkk δij is the deviatoric part of the stress tensor. The parameter Cv is
of order 1/µ0 and2 is related to the seismic coupling coefficient 0 < χ < 1 by (Ben-Zion and
Lyakhovsky, 2006)
1−χ 1
Cv =
(2.18)
χ µ0
2.3
2.3.1
Boundary conditions
Absorbing boundaries
Two approximate absorbing boundary conditions (ABC) are implemented to model the outwards propagation of waves at the boundaries of the computational domain. Both conditions
are of paraxial type. Their performance is appropriate at normal incidence but degrades at
grazing incidence.
Clayton-Engquist ABC
In the local coordinate frame (t, n) related to the tangential (t) and outgoing normal (n)
directions to the boundary, the first-order accurate ABC proposed by Clayton and Engquist
(1977) reads:
∂ut
(2.19)
∂xn
∂un
(2.20)
u̇n = −cP
∂xn
The implementation is based on an equivalent formulation as a mixed boundary condition
that relates tractions T to displacements u:
u̇t = −cS
Tt = −ρcS u̇t
Tn = −ρcP u̇n
The formulation above is for P-SV mode. In SH mode the ABC is Ty = −ρcS u̇y .
2
In &MAT DMG, the input argument R is defined as R = µ0 Cv .
(2.21)
(2.22)
2.4 Fault interface conditions
16
Stacey ABC
The second-order accurate ABC introduced by Stacey (1988) under the name “P3” is:
∂ut
∂un
− (cP − cS )
∂xn
∂xt
∂ut
∂un
− (cP − cS )
= −cP
∂xn
∂xt
u̇t = −cS
(2.23)
u̇n
(2.24)
Its formulation as a mixed boundary condition is:
∂un
∂xt
∂ut
= −ρcP u̇n − ρcS (2cS − cP )
∂xt
Tt = −ρcS u̇t + ρcS (2cS − cP )
(2.25)
Tn
(2.26)
This ABC is only implemented in P-SV mode.
2.4
Fault interface conditions
2.4.1
Linear slip law
Represents a compliant fault zone with elastic contact. See Haney et al. (2007). [...]
2.4.2
Normal stress response
Unilateral contact
No interpenetration during contact, free stress during opening. [...]
Modified Prakash-Clifton regularization
Regularization of the normal stress response, as required for bimaterial rupture problems, is
implemented following Rubin and Ampuero (2007). The frictional strength is proportional
to a modified normal stress σ ∗ , related to the real fault normal stress, σ, by either of the
following evolution laws:
– Version with a regularization time scale:
1
(σ − σ ∗ )
σ˙∗ =
Tσ
where Tσ is a constitutive parameter3 .
3
In &BC DYNFLT NOR, this law is set by kind=2, and the relevant parameter is T.
(2.27)
2.4 Fault interface conditions
17
– Version with a regularization slip scale:
|V | + V ∗
(σ − σ ∗ )
σ˙∗ =
Lσ
(2.28)
where V is slip rate, and V ∗ and Lσ are constitutive parameters4 .
2.4.3
Friction
Slip-weakening friction
Slip occurs when the fault shear stress reaches the shear strength τ = µσ (or τ = µσ ∗ if the
Prakash-Clifton law is assumed). [...] The friction coefficient µ is a function of the cumulated
slip D, given by one of the following laws:
– Linear slip-weakening law:
µs − µd
D
µ = max µd , µs −
Dc
(2.29)
– Exponential slip-weakening law:
µ = µs − (µs − µd ) exp(−D/Dc )
(2.30)
“Fast” rate-and-state-dependent friction
Friction with fast (power law) velocity weakening at fast slip speed is a first order proxy
for physical weakening processes that operate on natural fault zones at coseismic slip velocities. A rate-and-state dependent friction law with fast velocity-weakening is implemented in
SEM2DPACK, similar to that adopted e.g. by Ampuero and Ben-Zion (2008). The friction
coefficient depends on slip velocity (V ) and a state variable (θ):
µf = µs + a
θ
V
−b
.
V + Vc
θ + Dc
(2.31)
The state variable has units of slip and is governed by the evolution equation
θ̇ = V − θVc /Dc .
(2.32)
The friction law is defined by the following constitutive parameters: µs is the static friction coefficient, a and b are positive coefficients of a direct effect and an evolution effect,
respectively, Vc is a characteristic velocity scale5 , and Dc is a characteristic slip scale.
The steady-state (θ̇ = 0) friction coefficient
µf = µs + (a − b)
4
5
V
V + Vc
In &BC DYNFLT NOR, this law is set by kind=3, and the two relevant parameters are V and L.
vstar in &BC DYNFLT RSF
(2.33)
2.4 Fault interface conditions
18
weakens asymptotically as 1/V when V ≫ Vc , if a < b, approaching its dynamic value (µd =
µs + a − b) over a relaxation timescale Dc /Vc . The value of the relaxation time Dc /Vc tunes
the weakening mechanism between two limit cases: slip-weakening and velocity-weakening. If
Dc /Vc is much longer than the typical time scale of fluctuation of the state variable (≈ θ/θ̇),
Equation 2.32 becomes θ̇ ≈ V , implying that θ is proportional to slip and that the evolution
term of the friction coefficient is effectively slip-weakening, with characteristic slip-weakening
distance Dc . Conversely, if Dc /Vc is short the relaxation to steady state is fast, θ/Dc ≈ V /Vc
and the friction is effectively velocity-weakening, with characteristic velocity scale Vc .
Logarithmic rate-and-state friction
Not implemented yet.
Dieterich and Ruina classical rate-and-state laws, with aging or slip state evolution law.
Chapter 3
Mesh generation
3.1
General guidelines
The Spectral Element Method (SEM) requires an initial decomposition of the space domain
into quadrilateral elements (a quad mesh). Obtaining the best performance (accuracy/cost)
out of the SEM imposes constraints on the mesh design:
– The interfaces between different materials, at which sharp contrasts of material properties
occur, should preferably coincide with faces of the elements. This is sometimes called an
adapted mesh and is the only way to preserve spectral accuracy at material interfaces.
– Fault planes, across which displacement discontinuities occur, must coincide with element
faces. Faults are implemented with a split node formulation.
– Elements can be deformed, but extremely small and extremely large angles between faces
of a same element must be avoided. This would penalize both accuracy and stability.
– The linear size of the elements must be small enough, so that each element contains
enough computational nodes per minimum wavelength, and each fault boundary element
contains enough nodes per rupture process zone.
– Unnecessarily small elements should be avoided, they penalize the stability of the method.
Generating high quality quad meshes for complicated geological models is not yet a fully
automated process, and can be very time-consuming. Iterations between mesh generation and
mesh quality check are sometimes required. The last two constraints above are addressed more
quantitatively in Section 4.5. Mesh quality assessment tools are also presented in Section 4.5.
The remainder of this chapter describes three possible ways to generate quad meshes, by
order of complexity:
1. If the geometrical structure of the model is simple or if the user prefers to sacrifice
accuracy by using a non-adapted structured mesh, i.e. a logically cartesian mesh where
the element faces do not necessarily follow the material interfaces, the basic built-in
meshing capabilities of the solver SEM2D, described in Section 3.2, are sufficient.
3.2 Meshing features included in the solver
20
2. Moderately complicated meshes can be generated with the included Matlab tools, described in Section 3.3.
3. Adapted meshes for more complicated geological models must be generated with some
external software. As an illustration, the usage of the freely available 2D mesh generation software EMC2 is described in Section 3.4.
SEM2DPACK provides only basic meshing capabilities and does not include an unstructured
mesh generator for complicated, realistic geological models. This chapter describes how to
achieve that task with an external software, EMC2.
Generating a high quality unstructured quad mesh can be a time-consuming task. Let’s note
that, for wave propagation problems without dynamic faults, if the acceptable accuracy is low
(or large computational resources are available to work with a very fine mesh) a structured
mesh in which the element faces do not necessarily follow the material interfaces can be
generated with the basic built-in meshing capabilities of SEM2D.
3.2
Meshing features included in the solver
The solver itself has very limited meshing capabilities. It can only generate a structured mesh
for a single quadrilateral domain, possibly with curved sub-horizontal boundaries and curved
sub-horizontal layer interfaces. The domain can be cut in the horizontal direction by a single
fault, possibly curved or kinked.
For further details see the Reference Guide for the input blocks MESH CART and MESH LAYERED
in Section 4.4.
3.3
Meshing with the MESH2D Matlab utilities
A number of Matlab functions for 2D meshing are provided in POST/mesh2d. These can
generate structured meshes for quadrilateral domains with curved boundaries, and merge
several such meshes to generate a more complicated, globally unstructured mesh. Functions
for manipulating, visualizing and exporting these meshes are included. Here is an overview
of available tools:
SEM2DPACK/PRE/mesh2d provides Matlab utilities for the generation, manipulation
and visualization of structured 2D quadrilateral meshes, and unstructured
compositions of structured meshes.
The mesh database is stored in a structure described in MESH2D_TFI’s help.
Mesh generation:
MESH2D_TFI
generates a structured mesh by transfinite interpolation
3.4 Generating a mesh with EMC2
21
MESH2D_QUAD
generates a structured mesh for a quadrilateral domain
MESH2D_CIRC_HOLE generates a mesh for a square domain with a circular hole
MESH2D_WEDGE generates a mesh for a triangular wedge domain
MESH2D_EX0 mesh for a vertical fault
MESH2D_EX1 mesh for a shallow layer over half-space with dipping fault
Mesh manipulation:
MESH2D_ROTATE
rotates the node coordinates
MESH2D_TRANSLATE translates the node coordinates
MESH2D_MERGE
merges several meshes into a single mesh
MESH2D_WRITE
writes a 2D mesh database file (*.mesh2d)
MESH2D_READ reads a 2D mesh database from a *.mesh2d file
Reading mesh files from other mesh generation software:
READ_DCM reads a 2D mesh in the DCM format of EZ4U (http://www-lacan.upc.es/ez4u.htm)
READ_INP reads a 2D mesh in the INP format of ABAQUS exported by CUBIT
Mesh visualization:
MESH2D_PLOT
plots a 2D mesh
Miscellaneous tools:
SAMPLE_SEGMENTS generates points that regularly sample multiple segments of a line
The functions MESH2D TFI and MESH2D MERGE are the core tools. The script MESH2D EXAMPLE1
is a good starting point. The syntax of the mesh database file, *.mesh2d, is described in
Section 4.4.
3.4
3.4.1
Generating a mesh with EMC2
The mesh generator EMC2
EMC2 is one of the few public domain 2D mesh generation softwares that includes quadrilateral elements and a Graphical User Interface. Its C code sources and executables can be
freely downloaded from
http://www.ann.jussieu.fr/~hecht/ftp/emc2/.
We show here an example featuring the most useful functionalities of EMC2. For further
details you should refer to the complete documentation of the EMC2 package,
Before starting you must prepare files containing in 2-column data the coordinates (X,Z) of
all the points needed to define the geometry of the model (topography, sediment bottom).
3.4 Generating a mesh with EMC2
22
Once installed, you can run EMC2 by typing emc2.
3.4.2
Notations
The following notations are assumed in the next section:
• (XXX) = click XXX on top menu bar
• (xxx) = click xxx on bottom menu bar
• <XXX< = click XXX on left menu bar
• >XXX> = click XXX on right menu bar
• $xxx$ = enter xxx from keyboard or from the calculator in the right panel
• ”xxx” = type xxx in bottom prompt
• {xxx} = perform action xxx
• *xxx = do xxx as many times as needed
• n*xxx = do xxx n times
3.4.3
Basic step-by-step
A typical EMC2 session has three steps:
STEP I: CONSTRUCT, defines the geometry of the model
1. Switch to the construction tool:
<CONSTRUCTION<
2. Load the points:
(POINT) (xy file) ”palosgrandes.dat”
You must give the full path to your points-file, the root directory being the
one where you launched emc2.
3. Reset the figure window to fit all points:
>SHOW ALL>
The original data has some geometrical features that are too complex to be
meshed by quadrilaterals, for instance the corners at the N and S ends of the
basin, you may want to smooth out these features. You also need to define the
extreme boundaries of the region to be modelled (N,S and bottom absorbing
boundaries) and some additional points on the free surface outside the basin.
You must modify the data set (add and delete points):
4. Add new points:
3.4 Generating a mesh with EMC2
23
a. with the mouse:
(POINT) (mouse) *{click in figure window}
b. by coordinates:
(POINT) (xy pt) *{ $x=y=$ }
This is the safest way to get really vertical and horizontal boundaries needed
for the absorbing conditions in SPECFEM90. You probably need to get the
coordinates of an existing reference point:
(POINT) <QUERY< (point) *{click on point}
c. you can also reload another point-file (I2)
5. Delete points,
(POINT) <DESTRUCT< (point) *{click on point}
Now you must define the geometry of the domains. These macro-blocks are
intended to be internally meshed by deformed quadrilaterals. Their geometry
follows the geometry of the geological model (one domain per material). Each
domain must be bounded by segments or splines:
6. Segments:
(SEGMENT) (point) 2*{click extreme point}
7. Splines:
(SPLINE) (point) *{click point}
You will see the spline evolve as you click points.
STEP II: PREPARE, defines the properties of the discrete spectral element mesh
1. Switch to the preparing mesh tool:
<PREP MESH<
2. Define domains with rock n:
(DOMAIN REF) $n=$ (any) *{click inside domain}
You will see the domains edges get colored and the domains get numbered
with n.
3. At any moment you can decide to show or not the domain decomposition:
To hide the domain decomposition:
>REFRESH>
Show the domain decomposition:
(SHOW) (ALL)
4. Remove a domain definition:
(REMOVE) (DOMAINE) (any) {click inside domain}
WARNING: corrections to the domain decomposition are sometimes displayed
only after refreshing the figure window.
5. Now you must define the subdivision of each domain in quadrilateral finite
elements. Define the number n of elements on each edge:
(NB INTERVAL) $n=$ (any) {click edge}
You will see the intermediate points appear. The number of intervals n is
mainly dictated by the resolution criterion: elements should be smaller than
the smallest wavelength you want to propagate. Moreover, a domain can
3.4 Generating a mesh with EMC2
24
be quadrangulated only if the total number of intervals along its
perimeter is even (the sum of all n along its boundaries). However, a
quality mesh is not always guaranteed and you need to proceed by trial and
error (emc2 allows you to jump back and forth between the different steps of
the meshing procedure).
6. Finally you must define the external boundaries of the modelled region which
will have a special treatment. You must associate a tag (a number) to each
absorbing boundary. No convention is assumed but you should remember those
tags later when setting the boundary conditions in SEM2D. It is also useful
to assign a tag to the free surface boundary, that will be eventually used by
SEM2D to locate the receivers or sources.
Define a boundary with index n:
(LINE REF) $n=$ (any) *{click edge}
Of course each boundary can be composed of many domain edges. Refresh
the display to better see the boundaries. The same procedure applies to define
split-node interfaces such as faults and cracks: you must assign a different tag
to each side of the fault.
7. Save your work in EMC2 format:
<SAVE< ”name”
The resulting file is name.emc2 bd
STEP III: EDIT, generates the mesh
1. Switch to the edit mesh tool:
<EDIT MESH<
Press ENTER 4 times.
A triangles mesh appears. You must convert it to a quad mesh:
2. Convert the triangle mesh to a quad mesh:
<QUADRANGULATE> <ALL>
You can smooth the mesh with: <REGULARIZE> *<ALL>
The final mesh is displayed. If there remain some triangles come back to the
previous step and figure out how to modify the points per edge to help the
mesher. Some experience is needed here.
3. Renumber the mesh, in order to optimize computations:
*<RENUMBER>
4. Define the boundary condition for the 4 corner nodes of the model: (these
nodes belong to 2 external boundaries so they were given a reference number
=0)
(MODIF REF) $n=$ (corner) {click close to corner, inside element}
Where n is the reference number of one of the 2 boundaries containing the
corner node. Zooming can be useful. The same operation must be performed
for the corner nodes of the subdomains belonging to an external boundary,
and for the the crack tip nodes. However, as a special case, crack tip nodes
must be assigned the -1 tag.
3.5 Importing a mesh from CUBIT, EZ4U or other mesh generation software 25
5. Export the mesh:
<SAVE<
Two questions are asked in the bottom prompt:
• Format of the file, you must select:
”ftq”
• Prefix name for the file
”name”
The resulting file name will be name.ftq
3.4.4
Further tips
• Whenever possible it is better to mesh a domain with a structured mesh (a deformed
cartesian grid). This can be done with (QUADRANGULATE), during the PREPARE
step. See our FAQ for further details.
• To load an existing project, in the construction tool or in the preparation mesh tool:
<RESTORE< ”name”
EMC2 will look for the file name.emc2 bd. Beware: the project loaded will replace the
actual project if any, there is no superposition.
• BUG WARNING (13/07/01): the Sun release of EMC2 has a bug with the reference
indices in the ftq format This bug is fixed in the 2.12c version. If you work on a Sun
station, download the most recent version of the sources, rather than the executable,
and compile it yourself.
• To densify (h-refinement) an existing mesh use the script SEM2DPACK/POST/href.csh.
It edits the *.emc2 bd file. You can then restore it in EMC2 and save it in *.ftq
format.
• To create a fault, in EDIT MESH mode:
a. Crack an existing edge:
(CRACK) (segment)
b. Give a reference number to each side of the fault :
(MODIF REF) $n=$ (segment)
c. Give the tag ”-1” to crack tip nodes:
(MODIF REF) $-1=$ (corner) *{click close to crack tip node, inside element}
• Note that only Q4 elements (4 control nodes) are supported. For a smoother description
of boundaries Q9 would be desirable.
3.5 Importing a mesh from CUBIT, EZ4U or other mesh
generation software
The most convenient way of doing this is by writing a Matlab function that reads the
external mesh file, performs mesh manipulations if required and exports a mesh file with
3.5 Importing a mesh from CUBIT, EZ4U or other mesh generation software 26
the MESH2D format Section 3.3.
Two examples of this are provided in POST/mesh2d. The functions READ DCM and
READ INP read a mesh in, respectively, the DCM format of EZ4U (http://www-lacan.upc.es/ez4u.htm)
and in the INP format of ABAQUS exported by CUBIT.
Chapter 4
The solver SEM2D
4.1
About the method
Given a crustal model meshed with quadrilateral elements and a set of material properties,
sources, receivers and boundary conditions, SEM2D solves the elastic wave equation applying
the Spectral Element Method (SEM) for the space discretization and a second-order explicit
scheme for the time discretization. The range of physical problems solved by SEM2D (material
constitutive equations and boundary conditions) is described in more detail in Chapter 2.
The SEM, introduced by Patera (1984) in Computational Fluid Dynamics, can be seen as a
domain decomposition version of the Spectral Method or as a high order version of the Finite
Element Method. It inherits from its parent methods the accuracy (spectral convergence),
the geometrical flexibility and the natural implementation of mixed boundary conditions.
Introductory texts to the SEM can be found at www.math.lsa.umich.edu/~karni/m501/
boyd.pdf (chapter draft, by J.P. Boyd), at www.mate.tue.nl/people/vosse/docs/vosse96b.
pdf (a tutorial exposition of the SEM and its connection to other methods, by F.N. van de
Vosse and P.D. Minev) and at www.siam.org/siamnews/01-04/spectral.pdf (a perspective
paper). Details about the elastodynamic algorithm and study of some of its properties are
presented by Komatitsch (1997), Komatitsch and Vilotte (1998), Komatitsch et al. (1999),
Komatitsch and Tromp (1999) and Vai et al. (1998).
The implementation of fault dynamics is similar to that in FEM with the “traction at split
nodes” method explained by Andrews (1999). More details can be found in the author’s
Ph.D. dissertation (Ampuero, 2002)1 , in Gaetano Festa’s Ph.D. dissertation2 and in Kaneko
et al. (2008).
A more accesible tutorial code, SBIEMLAB written in Matlab, can be downloaded from the
author’s website, at web.gps.caltech.edu/~ampuero/software.html.
1
2
web.gps.caltech.edu/~ampuero/publications.html
people.na.infn.it/~festa/
4.2 Basic usage flow
4.2
28
Basic usage flow
In general, a simulation requires the following steps:
1. Prepare the input file Par.inp (Section 4.3 and Section 4.4).
2. Run the solver in “check mode” (iexec=0 in the GENERAL input block of Par.inp):
sem2dsolve > info &.
3. Verify the resolution, stability, estimated CPU cost and memory cost (Section 4.5).
4. If needed go back to step 1 and modify Par.inp (Section 4.5), else proceed to next step.
5. Run the solver in “production mode” (iexec=1): sem2dsolve.
6. Plot and manipulate the solver results (Section 4.6).
Full details are given in the following sections.
4.3
General format of the input file
The input file must be called Par.inp. Its typical structure is illustrated by two examples in
Figure 4.1 and Figure 4.2. Most of the file is made of standard FORTRAN 90 NAMELIST input blocks. Each block gives input for a specific aspect of the simulation: material properties,
sources, receivers, boundary conditions, etc.
The general syntax of a NAMELIST block can be found in any FORTRAN 90 textbook. In
summary, a block named STUFF, with possible input arguments a, b and c, must be given as
&STUFF a=..., b=..., c=... /
where ... are user input values. Line breaks and comments preceded by ! are allowed within
an input block.
The complete Reference Guide of the input blocks is presented in Section 4.4. For each block
the documentation includes its name, possibly the name of a group of blocks to which it
belongs, its purpose, its syntax, the list of its arguments with their description, and some
important notes. In the syntax description, a vertical bar | between two arguments means
“one or the other”. In the argument list, each item is followed by two informations within
brackets []. The first bracketed information is the type of the argument: double precision
(dble), integer (int), logical (log), single character (char), fixed length word (e.g. char*6 is
a 6 characters word), arbitrary length word (name) or vectors (e.g. int(2) is a two element
integer vector). The second bracketed information is the default value of the argument. Some
arguments are optional, or when absent they are automatically assigned the default values.
Some arguments have a second version with a suffix H that allows to set values that are
spatially non uniform. The H-version of the argument must be set to the name of any of the
input blocks of the DISTRIBUTIONS group. The appropriate &DIST xxxx block must follow
immediately. For example, to set the argument eta to a Gaussian distribution:
4.3 General format of the input file
29
&MAT_KV etaH=’GAUSSIAN’ /
&DIST_GAUSSIAN length=1d6,100d0, ampli=0.1d0 /
Arguments that accept an H-version are indicated in Section 4.4. When more than one Hversion argument is present, the &DIST xxxx blocks must appear in the same order as in the
argument list of Section 4.4.
In the next section, Input Block Reference Guide, you should get acquainted with the syntax
of the blocks you are most likely to use. The mandatory or more important input blocks are:
– &GENERAL
– &MESH DEF, followed by a &MESH Method block
– &MATERIAL, followed by a &MAT Material block
– &BC DEF, one for each boundary condition, each followed by a &BC Kind block
– &TIME
– &SRC DEF, followed by &STF SourceTimeFunction and &SRC Mechanism blocks
– &REC LINE
4.3 General format of the input file
30
Printed by Jean Paul Ampuero
Mar 06, 08 10:48
# Parameter file for SEM2DPACK 2.0
Par.inp
Page 1/1
#−−−−− Some general parameters −−−−−−−−−−−−−−−−
&GENERAL iexec=1, ngll= 6, fmax=1.25d0 , ndof=1 ,
title = ’Test SH’, verbose=’1111’, ItInfo = 1000 /
#−−−−− Build the mesh −−−−−−−−−−−−−−−−−−−−−−−−−−−
&MESH_DEF method = ’CARTESIAN’ /
&MESH_CART xlim=0.d0,30.d0 ,zlim=0.d0,30.d0 , nelem=60,60/
#−−−− Elastic material parameters −−−−−−−−−−−−−−
&MATERIAL tag=1, kind=’ELAST’ /
&MAT_ELASTIC rho=1.d0, cp=1.7321d0, cs=1.d0 /
#−−−−− Boundary conditions −−−−−−−−−−−−−−−−−−−−−
&BC_DEF tag = 2 , kind = ’ABSORB’ /
&BC_ABSORB stacey=F/
&BC_DEF tag = 3 , kind = ’ABSORB’ /
&BC_ABSORB stacey=F/
#−−−− Time scheme settings −−−−−−−−−−−−−−−−−−−−−−
&TIME TotalTime=35.d0, courant = 0.3d0 /
#−−−− Sources −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
&SRC_DEF stf= ’RICKER’ ,coord= 0.d0,0.d0 , mechanism= ’FORCE’ /
&STF_RICKER f0= 0.5d0, onset = 3.d0, ampli = 0.25d0 /
&SRC_FORCE angle = 0d0/
#−−−−− Receivers −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
&REC_LINE number = 7 , field=’D’, first = 0.d0,0.d0, last = 30d0,0.d0, isamp=1 /
#−−−−−−−−− Plots settings −−−−−−−−−−−−−−−−−−−−−−
&SNAP_DEF itd=100000, ps=F , bin=F /
Thursday March 06, 2008
1/1
Figure 4.1: Input file Par.inp for an elementary example in EXAMPLES/TestSH/ : a boxed
region with a structured mesh.
4.3 General format of the input file
31
Printed by Jean Paul Ampuero
Mar 06, 08 10:49
Par.inp
#−−−−− Some general parameters −−−−−−−−−−−−−−−−
&GENERAL Iexec =0 , Ngll = 6 , fmax = 1.5 , ndof=1,
Title = ’Palos Grandes NS meshed with EMC2’ ,
Verbose=’1111’, ItInfo = 1000/
Page 1/1
#−−−−− Build the mesh −−−−−−−−−−−−−−−−−−−−−−−−−−−
&MESH_DEF Method = ’EMC2’ /
&MESH_EMC2 File= ’NS03qb.ftq’ /
#−−−− Elastic material parameters −−−−−−−−−−−−−−
&MATERIAL tag=1, kind=’ELAST’ /
&MAT_ELASTIC rho=1800.d0, cp=850.d0, cs=450.d0/
&MATERIAL tag=2, kind=’ELAST’ /
&MAT_ELASTIC rho=2100.d0, cp=1800.d0, cs=650.d0/
&MATERIAL tag=3, kind=’ELAST’ /
&MAT_ELASTIC rho=2400.d0, cp=2300.d0, cs=850.d0/
&MATERIAL tag=4, kind=’ELAST’ /
&MAT_ELASTIC rho=2600.d0, cp=3800.d0, cs=2200.d0/
#&MAT_ELASTIC rho=2500.d0, cp=5000.d0, cs=2900.d0/
#−−−−− Boundary conditions −−−−−−−−−−−−−−−−−−−−−
&BC_DEF Tag = 2, Kind = ’ABSORB’ /
&BC_ABSORB Stacey=F /
&BC_DEF Tag = 3, Kind = ’ABSORB’ /
&BC_ABSORB Stacey=F, let_wave=T /
&BC_DEF Tag = 4, Kind = ’ABSORB’ /
&BC_ABSORB Stacey=F /
#−−−− Time scheme settings −−−−−−−−−−−−−−−−−−−−−−
&TIME TotalTime=25.d0, Courant = 0.55d0 /
&TIME_NEWMARK alpha=1.d0, beta=0.d0, gamma=0.5d0 /
#−−−− Sources −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
&SRC_DEF stf=’RICKER’, Mechanism=’WAVE’ Coord= −1160000.d0,−2000.d0 /
&STF_RICKER f0 = 1.d0 , Onset = 1.5d0 ,Ampli = 1.d0 /
&SRC_FORCE Angle = 90. /
&SRC_WAVE Angle = 30. , phase=’S’ /
#−−−−− Receivers −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
# receivers located at the surface by giving a very large vertical position
# locating them at the nearest computational node (AtNode=.true. is the default)
&REC_LINE Number = 31 , First = −1163068.0d0,1.d3, Last = −1159697.36d0,1.d3,
Isamp=10 /
#−−−−−−−−− Plots settings −−−−−−−−−−−−−−−−−−−−−−
&SNAP_DEF itd=100000, fields=’V’, components=’x’ / # itd = 3500
&SNAP_PS Mesh=T,Vectors=F,Color=T, Interpol = T, DisplayPts=7,
ScaleField=0.2d0 /
Thursday March 06, 2008
1/1
Figure 4.2: Input file Par.inp for a more realistic example: a sedimentary basin with an
unstructured mesh generated by EMC2. Available in EXAMPLES/UsingEMC2/.
4.4 Input Blocks Reference Guide
4.4
32
Input Blocks Reference Guide
=============================================================
= Self-documentation for the INPUT BLOCKS of the SEM2D code =
=============================================================
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
BC_ABSORB
BOUNDARY_CONDITION
Absorbing boundary
&BC_ABSORB stacey, let_wave /
stacey
[log] [F] Apply Stacey absorbing conditions for P-SV.
Higher order than Clayton-Engquist (the default).
let_wave [log] [T] Allow incident waves across this boundary
if mechanism=’WAVE’ in &SRC_DEF
NOTE
: Only implemented for vertical and horizontal boundaries.
---------------------------------------------------------------------------NAME
: BC_DIRNEU
GROUP : BOUNDARY_CONDITION
PURPOSE: Dirichlet (null displacement)
and/or Neumann (null or time-dependent traction)
boundary conditions on vertical or horizontal boundaries
SYNTAX : &BC_DIRNEU h, v, hsrc, vsrc /
possibly followed by one or two STF_XXXX blocks
h
v
hsrc
vsrc
[char][’N’] Boundary condition on the horizontal component
[char][’N’] Boundary condition on the vertical component :
’N’ : Neumann
’D’ : Dirichlet
[name][’none’] Name of the source time function for a
time-dependent horizontal traction:
’RICKER’, ’TAB’, ’USER’, etc (see STF_XXXX input blocks)
[name][’none’] Same for the vertical component
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
BC_DYNFLT
BOUNDARY_CONDITION, DYNAMIC_FAULT
Dynamic fault with friction
&BC_DYNFLT friction, cohesion|cohesionH, opening, Tn|TnH, Tt|TtH,
4.4 Input Blocks Reference Guide
33
Sxx|SxxH, Sxy|SxyH, Sxz|SxzH, Syz|SyzH, Szz|SzzH
ot1, otd, oxi, osides /
followed, in order, by:
1. &DIST_XXX blocks (from the DISTRIBUTIONS group) for arguments
with suffix H, if present, in the order listed above.
2. &BC_DYNFLT_SWF, &BC_DYNFLT_TWF or &BC_DYNFLT_RSF block(s)
(if absent, default values are used)
3. &BC_DYNFLT_NOR block (if absent, default values are used)
friction [name(2)] [’SWF’,’’] Friction law type:
SWF = slip weakening friction
TWF = time weakening friction
RSF = rate and state dependent friction
Some friction types can be combined. E.g. to set the
friction coefficient to the minimum of SWF and TWF, set
friction=’SWF’,’TWF’
cohesion [dble] [0d0] part of the strength not proportional to normal stress
opening [log] [T] Allow fault opening instead of tensile normal stress
Tn
[dble] [0d0] Initial normal traction (positive = tensile)
Tt
[dble] [0d0] Initial tangent traction (positive antiplane: y>0)
Sxx
[dble] [0d0] Initial stress sigma_xx
Sxy
[dble] [0d0] Initial stress sigma_xy
Sxz
[dble] [0d0] Initial stress sigma_xz
Syz
[dble] [0d0] Initial stress sigma_yz
Szz
[dble] [0d0] Initial stress sigma_zz
otd
[dble] [0.d0] Time lag between outputs (in seconds)
Internally adjusted to the nearest multiple of the timestep
Its value can be found in the output file FltXX_sem2d.hdr
The default internally resets otd = timestep
ot1
[dble] [0.d0] Time of first output (in seconds)
Internally adjusted to the nearest multiple of the timestep
Its value can be found in the output file FltXX_sem2d.hdr
oxi
[int(3)] [(1,huge,1)] First, last node and stride for output
The default resets oxi(2) = last fault node
osides
[log] [F] Export displacement and velocities on each side
of the fault
NOTE: the initial stress can be set as a stress tensor (Sxx,etc), as
initial tractions on the fault plane (Tn and Tt) or as the sum of both.
NOTE: we recommend to use dynamic faults with the leapfrog time scheme
and a layer of Kelvin-Voigt damping material near the fault.
---------------------------------------------------------------------------NAME
GROUP
: BC_DYNFLT_NOR
: DYNAMIC_FAULT
4.4 Input Blocks Reference Guide
34
PURPOSE: Normal stress response for dynamic faults.
SYNTAX : &BC_DYNFLT_NOR kind, V, L, T /
kind
T
V
L
[int] [1] Type of normal stress response:
0 = shear strength is independent of normal stress
(the cohesive strength is set as the product of
friction coefficient and initial normal stress)
1 = Coulomb
2 = Prakash-Clifton with regularizing time scale
3 = Prakash-Clifton with regularizing length scale
[dble] [1d0] Regularization time scale if kind=2
[dble] [1d0] Characteristic velocity if kind=3
[dble] [1d0] Regularization length scale if kind=3
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
kind
Dc
MuS
a
b
Vstar
BC_DYNFLT_RSF
DYNAMIC_FAULT
Velocity and state dependent friction
&BC_DYNFLT_RSF kind, Dc | DcH, Mus | MusH ,
a | aH, b | bH, Vstar | VstarH /
followed by &DIST_XXX blocks (from the DISTRIBUTIONS group) for
arguments with suffix H, if present, in the order listed above.
[int] [1] Type of rate-and-state friction law:
1 = strong velocity-weakening at high speed
as in Ampuero and Ben-Zion (2008)
[dble] [0.5d0] Critical slip
[dble] [0.6d0] Static friction coefficient
[dble] [0.01d0] Direct effect coefficient
[dble] [0.02d0] Evolution effect coefficient
[dble] [1d0] Characteristic or reference slip velocity
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
kind
Dc
MuS
BC_DYNFLT_SWF
DYNAMIC_FAULT
Slip-weakening friction
&BC_DYNFLT_SWF Dc | DcH, MuS | MuSH , MuD | MuDH, healing /
followed by &DIST_XXX blocks (from the DISTRIBUTIONS group) for
arguments with suffix H, if present, in the order listed above.
[int] [1] Type of slip weakening function:
1 = linear
2 = exponential
[dble] [0.5d0] Critical slip
[dble] [0.6d0] Static friction coefficient
4.4 Input Blocks Reference Guide
MuD
healing
35
[dble] [0.5d0] Dynamic friction coefficient
[log] [F] Instantaneous healing upon slip arrest
Healing is currently valid only with the leapfrog time scheme
---------------------------------------------------------------------------NAME
: BC_DYNFLT_TWF
GROUP : DYNAMIC_FAULT
PURPOSE: Time weakening friction for dynamic faults
with prescribed rupture speed.
SYNTAX : &BC_DYNFLT_TWF kind, MuS, MuD, Mu0, X, Z, V, L, T /
kind
MuS
MuD
Mu0
X,Z
V
L
T
NOTE
[int] [1] Type of time-weakening history:
1 = expansion at constant speed V up to time T
2 = expansion at decreasing speed then contraction
as in Andrews and Ben-Zion (JGR 1997, eqs 2 and 3)
[dble] [0.6d0] Static friction coefficient
[dble] [0.5d0] Dynamic friction coefficient
[dble] [0.6d0] Friction coefficient at the hypocenter at time=0
[dble] [0d0] Position of hypocenter
[dble] [1d3] Rupture propagation speed (initial speed if kind=2)
[dble] [1d0] Size of weakening zone
[dble] [huge] Total duration
: Time-weakening is usually applied as an artificial nucleation procedure.
The maximum size of the nucleation region is 2*V*T if kind=1, V*T/2 if kind=2
---------------------------------------------------------------------------NAME
: BC_DEF
PURPOSE: Define a boundary condition
SYNTAX : &BC_DEF tag, tags, kind /
possibly followed by &BC_kind blocks
tag
tags
kind
[int] [none] A number assigned to the boundary. If you are
using SEM2D built-in structured mesher the conventions are:
1
bottom
2
right
3
up
4
left
If you are importing a mesh, you must use the tags assigned
to the boundaries during the mesh construction.
[int(2)] [none] Two tags are needed for split-node interfaces (faults)
and for periodic boundaries.
[char*6] [none] Type of boundary condition. The following are
implemented:
’DIRNEU’, ’ABSORB’, ’PERIOD’, ’LISFLT’, ’DYNFLT’
4.4 Input Blocks Reference Guide
NOTE
36
: Most of the boundary conditions need additional data, given
in a BC_kind input block of the BOUNDARY_CONDITIONS group
immediately following the BC_DEF block.
---------------------------------------------------------------------------NAME
: BC_LSF
GROUP : BOUNDARY_CONDITION
PURPOSE: Linear slip fault, a displacement discontinuity interface
where stress and slip are linearly related
SYNTAX : &BC_LSF Ktang | Ctang, Knorm | Cnorm /
Ktang
Ctang
Knorm
Cnorm
[dble]
[dble]
[dble]
[dble]
[Inf]
[0d0]
[Inf]
[0d0]
Tangential stiffness
Tangential compliance
Normal stiffness
Normal compliance
NOTE: For each component:
You can set K _or_ C, but _not_both_
If C=0d0 or K=Inf then no discontinuity is allowed (transparent)
If K=0d0 the fault is free stress boundary
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
DIST_GAUSSIAN
DISTRIBUTIONS_2D
Bell shaped Gaussian 2D distribution
&DIST_GAUSSIAN centered_at, length, offset, ampli, order /
centered_at
length
offset
ampli
order
[dble(2)] [0,0]
[dble(2)] [1]
[dble] [0]
[dble] [1]
[int] [1]
Coordinates of the center point.
Characteristic lengths on each axis.
Background level.
Amplitude from background.
Exponent
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
file
DIST_GRADIENT
DISTRIBUTIONS_2D
Constant gradient 2D distribution.
&DIST_GRADIENT file,valref ,grad,angle/
[name] [none]
Name of the file containing the coordinates
of the points defining the reference line.
It is an ASCII file with 2 columns per line:
(1) X position (in m) and
4.4 Input Blocks Reference Guide
valref
grad
angle
NOTE
37
(2) Z position (in m)
[dble] [none]
Value along the reference line
[dble >0] [none] Positive gradient (valref_units/meter)
[dble] [none]
Angle (degrees) between the vertical down
and the grad+ direction. Anticlockwise convention (grad+
points down if 0, right if 90)
: Make sure the angle and ref-line are compatible. The code will
abort if the ref-line is too short: some points of the domain
cannot be projected to ref-line in the angle direction.
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
file
DIST_HETE1
DISTRIBUTIONS_2D
Linear interpolation of values from a regular 2D grid.
&DIST_HETE1 file, col /
[name] [none] Name of the file containing the definition
of the regular grid and values at grid points.
The format of this ASCII file is:
Line 1 : ncol nx nz x0 z0 dx dz
ncol = [int] number of data columns
nx,nz = [2*int] number of nodes along x and z
x0,z0 = [2*dble] bottom-left corner
dx,dz = [2*dble] spacing along x and z
Line 2 to nx*nz+1 : [ncol*dble] values at grid points
listed from left to right (x0 to x0+nx*dx),
then from bottom to top (z0 to z0+nz*dx)
[int] [1] Column of the file to be read
col
NOTE
: The same file can contain values for (ncol) different properties,
(e.g. rho, vp, vs) but each DIST_HETE1 block will read only one.
NOTE
: Even if the original model domain has an irregular shape,
the regular grid where input values are defined must be rectangular
and large enough to contain the whole model domain.
The regular grid possibly contains buffer areas with dummy values.
These dummy values should be assigned carefully (not random nor zero)
because SEM2D might use them during nearest-neighbor interpolation.
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
DIST_LINEAR
DISTRIBUTIONS_1D
Piecewise linear 1D distribution along X or Z.
&DIST_LINEAR n,dim,length /
4.4 Input Blocks Reference Guide
38
followed immediately by the interpolation data,
one line per point, two columns:
position (X or Z), value
or
&DIST_LINEAR file,dim,length /
and the interpolation data is read from a two-column file
n
dim
file
length
[int] [0] Number of points to be interpolated
[int] [1] Interpolate along X (dim=1) or along Z (dim=2)
[name] [none] Name of the ASCII file containing the data
[dble] [0] Smoothing length for sliding average window
No smoothing if length=0
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
xn
zn
x
z
v
DIST_ORDER0
DISTRIBUTIONS_2D
Blockwise constant 2D distribution.
&DIST_ORDER0 xn, zn /
x(1) ... x(xn-1)
z(1) ... z(zn-1)
v(1,1) ... v(xn,1)
...
...
...
v(1,zn) ... v(xn,zn)
[int] [none] Number of zones along X
[int] [none] Number of zones along Z
[dble(xn-1)] [none] Boundaries of X-zones: first zone X < x(1),
second zone x(1) < X < x(2), ... , last zone x(xn-1) < X
[dble(zn-1)] [none] Boundaries of Z-zones
[dble(xn,zn)] [none] Values inside each zone
---------------------------------------------------------------------------NAME
: DIST_PWCONR
GROUP : DISTRIBUTIONS_2D
PURPOSE: Piecewise constant radial (2D) distribution.
This distribution defines a set of annular zones, centered
at an arbitrary reference point, and assigns constant values
within each zone.
SYNTAX : &DIST_PWCONR num, ref /
r(1) ... ... r(num-1)
v(1) v(2) ... v(num-1) v(num)
num
ref
r
[int] [none] Number of annular zones (including inner and exterior)
[dble(2)] [(0d0,0d0)] Reference point: center of radial zones
[dble(num-1)] [none] External radius of zones:
4.4 Input Blocks Reference Guide
v
39
first zone R <= r(1),
second r(1) < R <= r(2), ...
last r(num-1) < R
[dble(num)] [none] Value inside each zone
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
file
dim
DIST_SPLINE
DISTRIBUTIONS_1D
Spline interpolated 1D distribution along X or Z.
&DIST_SPLINE file,dim /
[name] [none] Name of the ASCII file containing
the interpolation data, one line per point, two columns:
one line per point, two columns:
position (X or Z), value
[int] [1] Interpolate along X (dim=1) or along Z (dim=2)
---------------------------------------------------------------------------NAME
: GENERAL
PURPOSE: General parameters
SYNTAX : &GENERAL iexec, ngll, fmax, title, verbose, itInfo /
iexec
ngll
fmax
ndof
title
verbose
itInfo
[int] [0] Run level:
0 = just check
1 = solve
[int] [9] Number of GLL nodes per edge on each spectral element
( polynomial order +1 ). Usually 5 to 9.
[dble] [1.d0] The code checks if this maximum frequency is
well resolved by the mesh and issues a warning if not.
This parameter is not used in computations, only for checking.
To improve the resolution for a given fmax you must increase ngll
(but you will have to use shorter timesteps) or refine the mesh.
[int] [2] Number of degrees of freedom per node
1 = SH waves, anti-plane
2 = P-SV waves, in-plane
[word] [none] Title of the simulation
[char(4)] [’1101’] Print progress information during each phase:
verbose(1) = input phase
verbose(2) = initialization phase
verbose(3) = check phase
verbose(4) = solver phase
Example: ’0001’ is verbose only during solver.
[int] [100] Frequency (in number of timesteps) for printing
progress information during the solver phase.
4.4 Input Blocks Reference Guide
40
---------------------------------------------------------------------------NAME
: MAT_DAMAGE
GROUP : MATERIALS
PURPOSE: Set material properties for the damage rheology of
Lyakhovsky, Ben-Zion and Agnon (J. Geophys. Res. 1997)
and Hamiel et al (Geophys. J. Int. 2004)
SYNTAX : &MAT_DAMAGE cp,cs,rho,phi,alpha,Cd,R,e0,ep /
cp
cs
rho
phi
alpha
Cd
R
e0
ep
[dble][0d0] P wave velocity
[dble][0d0] S wave velocity
[dble][0d0] density
[dble][0d0] internal friction angle
[dble][0d0] initial value of damage variable
[dble][0d0] damage evolution coefficient
[dble][0d0] damage-related plasticity coefficient Cv
normalized by the inverse of the intact shear modulus
[dble(3)][0d0] initial total strain (11, 22 and 12)
[dble(3)][0d0] initial plastic strain (11, 22 and 12)
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
MAT_ELASTIC
MATERIALS
Set material properties for a linear elastic medium
For isotropic material:
&MAT_ELASTIC rho|rhoH, cp|cpH, cs|csH /
followed by DIST_XXXX blocks, for arguments with suffix H,
if present, in the same order as listed above.
For transverse anisotropy with vertical symmetry axis:
&MAT_ELASTIC rho, c11,c13,c33,c55 /
cp
[dble][0d0] P wave velocity
cs
[dble][0d0] S wave velocity
rho
[dble][0d0] density
c11,c13,c33,c55 [dble][0d0] anisotropic elastic moduli
---------------------------------------------------------------------------NAME
: MATERIAL
PURPOSE: Define the material type of a tagged domain
SYNTAX : &MATERIAL tag, kind /
followed by one or two MAT_XXXX input blocks.
tag
kind
[int] [none]
Number identifying a mesh domain
[name(2)] [’ELAST’,’’] Material types:
’ELAST’, ’DMG’,’PLAST’, ’KV’
4.4 Input Blocks Reference Guide
NOTE
41
: Some combinations of material kinds can be assigned to the same domain.
Any material type can be combined with ’KV’, for instance:
&MATERIAL tag=1, kind=’ELAST’,’KV’ /
followed by a &MAT_ELAST block and a &MAT_KV block
sets an elastic material with Kelvin-Voigt damping.
---------------------------------------------------------------------------NAME
: MAT_KV
GROUP : MATERIALS
PURPOSE: Sets material properties for a Kelvin-Voigt viscous material.
Adds a damping term C*v = K*eta*v, where eta is a viscous time.
This produces attenuation with frequency-dependent quality factor
Q(f) = 1/(eta*2*pi*f)
Its main usage is for artificial damping of high-frequency
numerical artifacts generated by dynamic faults, which requires a
thin layer of Kelvin-Voigt elements surrounding the fault
with eta/dt = 0.1 to 0.3 and a layer thickness of 1 to 2 elements
on each side of the fault.
SYNTAX : &MAT_KV eta, ETAxDT /
&MAT_KV etaH, ETAxDT / followed by a DIST_XXX input block
eta
ETAxDT
NOTE
[dble][0d0] Viscosity coefficient
[log][T] If eta is given in units of dt (timestep)
: Kelvin-Voigt viscosity modifies the stability of time integration.
The timestep (or the Courant number) must be set to a value
smaller than usual. The critical timestep for a Kelvin-Voigt material
integrated with the leapfrog time scheme is
dtc_kv = eta*( sqrt(1+dtc^2/eta^2)-1 )
where dtc is the critical timestep for a purely elastic medium (eta=0).
In terms of the normalized viscosity (if ETAxDT=T):
dtc_kv = dtc / sqrt( 1+ 2*eta)
---------------------------------------------------------------------------NAME
: MAT_PLASTIC
GROUP : MATERIALS
PURPOSE: Set material properties for elasto-plastic material
with Mohr-Coulomb yield criterion
and non-dilatant (null volumetric plastic strain)
SYNTAX : &MAT_PLASTIC cp,cs,rho,phi,coh,Tv,e0 /
cp
cs
rho
[dble][0d0] P wave velocity
[dble][0d0] S wave velocity
[dble][0d0] density
4.4 Input Blocks Reference Guide
phi
coh
Tv
e0
42
[dble][0d0] internal friction angle
[dble][0d0] cohesion
[dble][0d0] visco-plastic relaxation time
[dble(3)][0d0] initial total strain (11, 22 and 12)
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
xlim
zlim
nelem
ezflt
fztag
fznz
FaultX
MESH_CART
MESH_DEF
Rectangular box with structured mesh.
&MESH_CART xlim, zlim, nelem, ezflt,fztag, FaultX /
[dble(2)] [none] X limits of the box (min and max)
[dble(2)] [none] Z limits of the box (min and max)
[int(2)] [none] Number of elements along each direction
[int][0] introduce a horizontal fault between the ezflt-th
and the (ezflt+1)-th element rows. Rows are numbered from
bottom to top, starting at ezflt=1.
If ezflt=0, (default) no fault is introduced inside the box
(for symmetric problems a fault can still be set at an external boundary)
If ezflt=-1, a fault is introduced at/near the middle of the box
(ezflt is reset to int[nelem(2)/2])
[int][0] fault zone tag for elements close to the fault
Useful to set a damping layer near the fault.
If ezflt=0, a fault is assumed at the bottom boundary
[int][1] vertical size (number of elements) of near-fault layer
[log] [F] Same as ezflt=-1. Obsolete (will be deprecated)
NOTE: the following tags are automatically assigned to the boundaries:
1
Bottom
2
Right
3
Top
4
Left
5
Fault, bottom side
6
Fault, top side
---------------------------------------------------------------------------NAME
: MESH_CART_DOMAIN
PURPOSE: Define a subdomain within a structured meshed box.
SYNTAX : &MESH_CART_DOMAIN tag,ex,ez /
tag
ex
ez
[int] [none] Tag number assigned to this domain.
[int(2)] [none] Horizontal index of the first and last elements.
The leftmost element column has horizontal index 1.
[int(2)] [none] Vertical index of the first and last elements.
The bottom element row has vertical index 1.
4.4 Input Blocks Reference Guide
NOTE
43
: If you ignore this input block a single domain (tag=1) will span
the whole box
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
file
MESH_EMC2
MESH_DEF
Imports a mesh from INRIA’s EMC2 mesh generator in FTQ format
&MESH_EMC2 file /
[name] [none] Name of the FTQ file, including suffix
---------------------------------------------------------------------------NAME
: MESH_DEF
PURPOSE: Selects a method to import/generate a mesh.
SYNTAX : &MESH_DEF method /
followed by a &MESH_method input block
method
[name] [none] Meshing method name:
’CARTESIAN’, ’LAYERED’, ’EMC2’, ’MESH2D’
---------------------------------------------------------------------------NAME
: MESH_LAYERED
GROUP : MESH_DEF
PURPOSE: Structured mesh for layered medium
with surface and interface topography.
SYNTAX : &MESH_LAYERED xlim,zmin,nx,file,nlayer,ezflt,fztag /
xlim
zmin
nx
file
nlayer
[dble(2)] [none] X limits of the box (min and max)
[dble] [none] bottom Z limit of the box
[int] [1] Number of elements along the X direction.
Not needed if ztopH=’QSPLINE’ in a &MESH_LAYER block.
[string] [’’] Only for flat layers,
name of ASCII file containing layer parameters,
one line per layer, listed from top to bottom,
3 columns per line:
(1) vertical position of top boundary,
(2) number of elements along Z direction
(3) material tag
[int] [none] Number of layers
If a file name is not given the layer parameters
must be given immediately after the &MESH_LAYERED block
by nlayer &MESH_LAYER input blocks,
one for each layer, listed from top to bottom.
4.4 Input Blocks Reference Guide
ezflt
fztag
fznz
44
[int][0] introduce a fault between the ezflt-th and the
(ezflt+1)-th element rows, numbered from bottom to top.
If ezflt=0 (default), no fault is introduced.
If ezflt=-1, a horizontal fault is introduced at/near the
middle of the box: ezflt is reset to int[nelem(2)/2]
[int][0] tag for elements near the fault
Useful to set a damping layer near the fault.
[int][1] vertical size of near-fault layer
(half thickness in number of elements)
NOTE: the following tags are automatically assigned to the boundaries:
1
Bottom
2
Right
3
Top
4
Left
5
Fault, lower side
6
Fault, upper side
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
nz
ztop
ztopH
tag
MESH_LAYER
MESH_DEF
Define mesh parameters for one layer
&MESH_LAYER nz, ztop|ztopH, tag /
followed by a DIST_XXXX block if ztopH is set
[int] [none] Number of elements in layer along Z direction
[dble] [none] Only for layers with flat top surface:
vertical position of top boundary
[string] [’none’] Name of the type of spatial distribution to
generate an irregular (non flat) top boundary. In general it is
one of the 1D distribution available through a DIST_XXXX block:
ztopH = ’LINEAR’, or
ztopH = ’SPLINE’, etc.
There are two methods to generate a curve with a smooth normal,
typically to guarantee smooth boundary conditions on curved faults.
The first method is based on quadratic splines and sometimes
produces degenerated elements:
ztopH=’QSPLINE’, followed by a &QC_SPLINE block
The second method is based on cubic splines and is more robust:
ztopH=’CSPLINE’, followed by a &QC_SPLINE block
[int] [none] Material tag
If not given, a tag is automatically assigned to the layer,
sequentially numbered from top to bottom (top layer tag =1)
NOTE: If ztopH=’LINEAR’ the mesh uses linearly deformed (Q4) elements,
otherwise it uses quadratically deformed (Q9) elements
4.4 Input Blocks Reference Guide
45
---------------------------------------------------------------------------NAME
: QC_SPLINE
GROUP : MESH_LAYER
PURPOSE: Define the boundary of a layer using quadratic or cubic splines and
enforcing smooth (continuous) normal between elements, for instance
to guarantee smooth boundary conditions on curved faults.
SYNTAX : &QC_SPLINE file /
file
[string] [’’] Name of ASCII file containing information of
all the element vertex nodes lying on the boundary curve.
One line per node, ordered by increasing x, 3 columns per line:
(1) x position
(2) z position
(3) derivative dz/dx of the curve at the node
All QC_SPLINE curves in a mesh must have the same number of nodes.
The parameter nx in &MESH_LAYERED is automatically reset
(nx = number of nodes in QC_SPLINE - 1)
---------------------------------------------------------------------------NAME
: MESH_MESH2D
GROUP : MESH_DEF
PURPOSE: Imports a mesh in mesh2d format
as defined by the PRE/mesh2d mesh generator tools for Matlab
SYNTAX : &MESH_MESH2D file /
file
[name] [none] Name of the MESH2D file, including suffix.
The format of this file is:
"NEL NPEL NNOD NBC"
1 line with 4 integers:
nb of elements, nodes per element, total nb of nodes, nb of boundaries
"NID X Y"
NNOD lines, one per node, with 1 integer and 2 reals:
node id, x, y
"EID NODES TAG"
NEL lines, one per element, with NPEL+2 integers:
element id, NPEL node ids, tag.
"BCTAG NBEL"
|
2 integers: boundary tag, nb of boundary elements
|
"BID EID EDGE"
| repeat for each of
NBEL lines, one per boundary element, 3 integers:
| the NBC boundaries
boundary element id, bulk element id, edge id
|
4.4 Input Blocks Reference Guide
46
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
SNAP_DEF
SNAPSHOT_OUTPUTS
Set preferences for exporting snapshots
&SNAP_DEF it1, itd, fields, components, bin, visual3, avs, ps, gmt /
Followed by a &SNAP_PS block if ps=T.
it1
itd
fields
[int] [0]
Time step of first snapshot output
[int] [100] Number of timesteps between snapshots
[char*] [’V’] fields to export in snapshots (the prefix of the
output file names is given in parenthesis):
’D’
displacements (dx,dy,dz,da)
’V’
velocity (vx,vy,vz,va)
’A’
acceleration (ax,ay,az,aa)
’E’
strain (e11,e22,e12,e23,e13)
’S’
stress (s11,s22,s12,s33,e13,e23)
’d’
divergence rate (dvx/dx + dvz/dz)
’c’
curl rate (dvx/dz - dvz/dx)
components [char*] [’ya’] components for PostScript outputs:
in P-SV: ’x’,’z’ and/or ’a’ (amplitude). ’y’ is ignored
in SH:
’y’ only. Other values are ignored.
ps
[log] [T] PostScript (see &SNAP_PS input block)
gmt
[log] [F] output triangulation file grid_sem2d.gmt
to be used in "pscontour -T" of the General Mapping Tool (GMT)
avs
[log] [F] AVS (only for D,V and A fields)
visual3 [log] [F] Visual3 (only for D,V and A fields)
bin
[log] [T] binary
NOTE
: E and S fields are exported only as binary.
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
SNAP_PS
SNAPSHOT_OUTPUTS
Preferences for PostScript snapshots
&SNAP_PS vectors, mesh, background, color,
isubsamp, boundaries, symbols, numbers, legend,
ScaleField, Interpol, DisplayPts /
vectors
mesh
background
[log] [F] Plots a vectorial field with arrows
[log] [F] Plots the mesh on background
[char] [’’] Filled background, only for vector plots:
’’
none
’P’ P-velocity model
’S’ S-velocity model
’T’ domains
4.4 Input Blocks Reference Guide
isubsamp
boundaries
symbols
numbers
legend
color
ScaleField
Interpol
DisplayPts
47
[int] [3] Subsampling of the GLL nodes for the
output of velocity model.
The default samples every 3 GLL points.
[log] [T] Colors every tagged boundary
[log] [T] Plots symbols for sources and receivers
[log] [F] Plots the element numbers
[log] [T] Writes legends
[log] [T] Color output
[dble] [0d0] Fixed amplitude scale (saturation),
convenient for comparing snapshots and making movies.
The default scales each snapshot by its maximum amplitude
[log] [F] Interpolate field on a regular subgrid
inside each element
[log] [3] Size of interpolation subgrid inside each
element is DisplayPts*DisplayPts. The default plots at
vertices, mid-edges and element center.
---------------------------------------------------------------------------NAME
: REC_LINE
PURPOSE: Defines a line of receivers
SYNTAX : If single receiver line:
&REC_LINE number,first,last,AtNode,isamp,field,irepr /
If receiver locations from file:
&REC_LINE file,AtNode,isamp,field,irepr /
number
first
last
file
AtNode
isamp
field
irepr
NOTE
[int] [0] Number of stations in the line
[dble(2)] Receivers can be located along a line,
this is the position (x,z) of the first receiver
[dble(2)] Position (x,z) of the last receiver,
other receivers will be located with regular spacing
between First and Last.
[name] [’none’] Station positions can instead be read
from an ASCII file, with 2 columns: X and Z (in meters)
[log] [T] Relocate the stations at the nearest GLL node
[int] [1] Sampling stride (in number of timesteps). Note that
for stability reasons the timestep can be very small.
[char] [’V’] The field in the seismogram:
’D’
displacement
’V’
velocity
’A’
acceleration
[char] [’D’] Abscissa for the seismic multitrace plot:
’X’ Horizontal position
’Z’ Depth
’D’ Distance to the first station
: to locate receivers at the free surface set their vertical position
4.4 Input Blocks Reference Guide
48
above the free surface and AtNode=T
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
angle
SRC_FORCE
SOURCE MECHANISM
Point force source
&SRC_FORCE angle /
[dble] [0d0] For P-SV, the angle of the applied force,
in degrees, counterclockwise from Z-UP, e.g.:
90 points left, 180 points down
For SH, angle is ignored and the SRC_FORCE block is not required.
---------------------------------------------------------------------------NAME
: SRC_DEF
PURPOSE: Define the sources.
SYNTAX : &SRC_DEF stf, mechanism, coord /
&SRC_DEF stf, mechanism, file /
followed by one SOURCE TIME FUNCTION block (STF_XXXX)
and one SOURCE MECHANISM block (SRC_XXXX)
stf
mechanism
coord
file
[name] [none] Name of the source time function:
’RICKER’, ’TAB’, ’HARMONIC’, ’BRUNE’ or ’USER’
[name] [none] Name of the source mechanism:
’FORCE’, ’EXPLOSION’, ’DOUBLE_COUPLE’, ’MOMENT’ or ’WAVE’
[dble(2)] [huge] Location (x,z) of the source (m).
[name] [’none’] Name of file containing source parameters.
The file format is ASCII with one line per source and
2, 3 or 4 columns per line:
(1) X position (in m)
(2) Z position (in m)
(3) time delay (in seconds)
(4) relative amplitude
If column 4 is absent, amplitude = 1.
If columns 3 and 4 are absent, delay = 0 and amplitude = 1.
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
dip
SRC_DOUBLE_COUPLE
SOURCE MECHANISM
Define a double-couple source
&SRC_DOUBLE_COUPLE dip /
[dble] [90] Dip angle, in degrees, clockwise
from the positive X direction
4.4 Input Blocks Reference Guide
49
NOTE
: Sign convention: if the source amplitude is positive the right block
moves up (positive Z direction) in PSV and forward (positive Y
direction) in SH.
NOTE
: The source time function gives the cumulative seismic moment Mo(t),
NOT the seismic moment rate.
NOTE
: The seismic moment Mo must be rescaled because a 2D point source is
equivalent to a 3D line source. A proper scaling is obtained by
dividing the original 3D moment by the characteristic size of the
rupture area in the off-plane dimension. An approximate scaling for
a fault area with aspect ratio close to unity is
Mo_2D = (Mo_3D/dtau)^2/3 * dtau
where dtau is the stress drop (typically a few MPa).
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
SRC_MOMENT
SOURCE MECHANISM
Define a moment tensor source
&SRC_MOMENT Mxx,Mxz,Mzx,Mzz /
&SRC_MOMENT Myx,Myz /
Mxx,Mxz,Mzx,Mzz [dble] [0] Tensor components for PSV
Myx,Myz
[dble] [0] Tensor components for SH
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
angle
phase
NOTE
SRC_WAVE
SOURCE MECHANISM
Incident plane wave through the absorbing boundaries
&SRC_WAVE angle, phase /
[dble] [0d0]
Incidence angle in degrees within [-180,180]
counterclockwise from the positive Z (up) direction
to the wave vector direction:
Exs: incidence from below if angle in ]-90,90[
normal incidence from below if angle=0
from bottom right if angle=+45
from bottom left if angle=-45
[char] [’S’]
’S’ or ’P’ (only needed in PSV, ignored in SH)
: Incident waves enter through the absorbing boundaries.
An incident wave is applied on every absorbing boundary
unless "let_wave = F" in the respective BC_ABSO block.
Incident waves are not implemented for "Stacey" absorbing boundaries.
4.4 Input Blocks Reference Guide
50
---------------------------------------------------------------------------NAME
: STF_BRUNE
GROUP : SOURCE TIME FUNCTIONS
PURPOSE: Brune (1970)’s model with omega-squared spectral fall-off:
stf(t) = ampli*( 1 - (1+2*pi*fc*t)*exp(-2*pi*fc*t) )
SYNTAX : &STF_BRUNE ampli, fc /
ampli
fc
[dble] [1d0] Amplitude (usually the seismic moment)
[dble] [1d0] Corner frequency (Hz)
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
ampli
f0
STF_HARMONIC
SOURCE TIME FUNCTIONS
Harmonic source time function f(t) = ampli*sin(2*pi*t*f0)
&STF_HARMONIC ampli, f0 /
[dble] [0d0] Amplitude
[dble] [0d0] Frequency
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
ampli
f0
onset
NOTE
NOTE
STF_RICKER
SOURCE TIME FUNCTIONS
The Ricker wavelet is the second derivative of a gaussian.
&STF_RICKER ampli, f0, onset /
[real] [1.] Signed amplitude of the central peak
[real >0] [0] Fundamental frequency (Hz).
distribution: it has a peak at f0 and an exponential
decay at high frequency. The cut-off high frequency is usually
taken as fmax = 2.5 x f0.
[real >1/f0] [0] Delay time (secs) with respect to the peak value.
: The spectrum has a peak at f0 and decays exponentially at high
frequencies. Beyond 2.5*f0 there is little energy, this is a
recommended value for fmax.
: onset>1/f0 is needed to avoid a strong jump at t=0, which can cause
numerical oscillations. Ignore if using incident waves.
---------------------------------------------------------------------------NAME
: STF_TAB
GROUP : SOURCE TIME FUNCTIONS
PURPOSE: Source time function spline-interpolated from values in a file
4.4 Input Blocks Reference Guide
51
SYNTAX : &STF_TAB file /
file
NOTE
[string] [’stf.tab’] ASCII file containing the source time function,
two columns: time and value. Time can be irregularly sampled and
must increase monotonically.
: assumes value(t<min(time))=value(min(time))
and value(t>max(time))=value(max(time))
---------------------------------------------------------------------------NAME
: STF_USER
GROUP : SOURCE TIME FUNCTIONS
PURPOSE: A template for user-supplied source time function.
File stf_user.f90 must be modified by the user to fit
special needs.
SYNTAX : &STF_USER ampli, onset, par1, par2, ipar1, ipar2 /
ampli
onset
par1
par1
par1
par1
[dble] [1.] Amplitude
[dble] [0] Delay time (secs)
[dble] [0] Example parameter
[dble] [0] Example parameter
[int] [0] Example parameter
[int] [0] Example parameter
---------------------------------------------------------------------------NAME
: TIME
PURPOSE: Defines time integration scheme
SYNTAX : &TIME kind, {Dt or Courant}, {NbSteps or TotalTime} /
Possibly followed by a TIME_XXXX block.
kind
[char*10] [’leapfrog’] Type of scheme:
’newmark’
Explicit Newmark
’HHT-alpha’
Explicit HHT-alpha
’leapfrog’
Central difference
’symp_PV’
Position Verlet
’symp_PFR’
Position Forest-Ruth (4th order)
’symp_PEFRL’
Extended PFR (4th order)
Dt
[dble] [none] Timestep (in seconds)
Courant
[dble] [0.5d0] the maximum value of the Courant-Friedrichs-Lewy
stability number (CFL), defined as
CFL = Dt*wave_velocity/dx
where dx is the distance between GLL nodes. Tipically CFL<= 0.5
NbSteps
[int] [none] Total number of timesteps
TotalTime [int] [none] Total duration (in seconds)
4.4 Input Blocks Reference Guide
NOTE
52
: The leap-frog scheme is recommended for dynamic faults. It is equivalent
to the default Newmark scheme (beta=0, gamma=1/2). However it is
faster and requires less memory.
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
beta
gamma
TIME_NEWMARK
TIME SCHEMES
Explicit Newmark time integration scheme
&TIME_NEWMARK gamma, beta /
[dble] [0d0] First Newmark parameter.
If beta=0 the scheme is fully explicit (the update of
displacement depends only on the last value of acceleration),
otherwise it is a single-predictor-corrector scheme
[dble] [0.5d0] Second Newmark parameter.
Second order requires gamma=1/2.
---------------------------------------------------------------------------NAME
:
GROUP :
PURPOSE:
SYNTAX :
alpha
rho
TIME_HHTA
TIME SCHEMES
Explicit HHT-alpha time integration scheme, second order
&TIME_HHTA alpha, rho /
[dble] [0.5d0] Parameter in the HHT-alpha method. Values in [0,1].
Defined here as 1 + HHT’s original definition of alpha.
When alpha=1 it reduces to second order explicit Newmark
(beta=0, gamma=0.5).
[dble] [0.5d0] Minimum damping factor for high frequencies.
Values in [0.5,1]. Rho=1 is non-dissipative.
NOTE: We consider only second order schemes, for which alpha+gamma=3/2
If alpha<1, Newmark’s beta is related to the HHT parameters by
beta = 1 -alpha -rho^2*(rho-1)/[(1-alpha)*(1+rho)^3]
If alpha=1, we set rho=1 (beta=0, gamma=0.5)
NOTE: Dissipative schemes (rho<1) require slightly smaller Courant number
(0.56 for rho=0.5, compared to 0.6 for rho=1)
NOTE: This is an explicit version of the HHT-alpha scheme of
H.M. Hilber, T.J.R. Hughes and R.L. Taylor (1977) "Improved numerical
dissipation for time integration algorithms in structural dynamics"
Earthquake Engineering and Structural Dynamics, 5, 283-292
implemented with a slightly different definition of alpha (1+original).
Its properties can be derived from the EG-alpha scheme of
4.4 Input Blocks Reference Guide
53
G.M. Hulbert and J. Chung (1996) "Explicit time integration
algorithms for structural dynamics with optimal numerical dissipation"
Comp. Methods Appl. Mech. Engrg. 137, 175-188
by setting alpha_m=0 and alpha=1-alpha_f.
4.5 Verifying the settings and running a simulation
4.5
54
Verifying the settings and running a simulation
Once the code has been successfully compiled, the simulation can be started by typing
sem2dsolve from your working directory, which contains the file Par.inp. The computations can be run in background and the screen output saved in a file (e.g. info) by typing
sem2dsolve > info &.
A typical screen output of SEM2D, corresponding to the first example, is shown on the
following pages. The parameters of the simulation and some verification information are
reported there in a self-explanatory form. You are advised to do a first run with iexec=0 in
the GENERAL input block and check all these informations prior to the real simulation. You
should always verify the following:
• Stability: the CFL stability number should be smaller than 0.55 ∼ 0.60 for second
order time schemes, and much smaller for highly deformed meshes (see Section 6.1 on
“Instabilities in very distorted elements”). This number is defined at each computational
node as
CFL = cP ∆t/∆x
where ∆t is the timestep, cP the P-wave velocity and ∆x the local grid spacing. Note
that ∆x is usually much smaller than the element size h (≈ Ngll2 times smaller) because
SEM internally subdivides each element onto a non-regular grid of Ngll×Ngll nodes
clustered near the element edges (Gauss-Lobatto-Legendre nodes). If the computation
is unstable, the maximum displacement, printed every ItInfo time steps, increases
exponentially with time. Stability can be controlled by decreasing Dt or Courant in
Par.inp.
• Resolution: the number of nodes per shortest wavelength λmin should be larger than
4.5 ∼ 5. The minimum wavelength is defined as
λmin = min(cS )/fmax
where cS is the S-wave velocity and fmax the highest frequency you would like to resolve,
e.g. the maximum frequency at which the source spectrum has significant power (for
a Ricker wavelet fmax = 2.5 × f0 ). For an element of size h and polynomial order
p = Ngll − 1, the number of nodes per wavelength G is
p λmin
.
h
Typical symptoms of poor resolution are ringing and dispersion of the higher frequencies.
However, in heterogeneous media these spurious effects might be hard to distinguish
from a physically complex wavefield, so mesh resolution must be checked beforehand.
If resolution is too low the mesh might be refined by increasing Ngll in Par.inp (prefinement) or by generating a denser mesh (h-refinement). If you were using EMC2 as
a mesh generator, the script PRE/href.csh can be useful for h-refinement.
G=
• Cost: the total CPU time an memory required for the simulation are as much as you
can afford. Estimates of total CPU time are printed at the end of check mode. Details
about memory usage can be found in MemoryInfo sem2d.txt.
4.5 Verifying the settings and running a simulation
55
Stability
350
3000
300
2000
1000
250
0
200
−1000
150
−2000
−3000
100
−4000
−1.164
−1.162
−1.16
−1.158
−1.156
50
−1.154
6
x 10
−1
−0.5
0
0.5
0
−2
−1
0
1
0
0.5
1
1
Resolution
400
3000
350
2000
300
1000
250
0
−1000
200
−2000
150
−3000
100
−4000
50
−1.164
−1.162
−1.16
−1.158
−1.156
−1.154
6
x 10
−0.6
−0.4
−0.2
0
0.2
0.4
0
−0.5
0.6
Figure 4.3: Checking the quality of a mesh with PRE/ViewMeshQuality.m for the example
in EXAMPLES/UsingEMC2/. The balance of the stability and resolution properties of the mesh
can be analyzed: logarithmic stability index (top) and logarithmic resolution index (bottom).
Histograms of these indices (in number of elements) are shown on the right.
4.5 Verifying the settings and running a simulation
56
The quality of the mesh can be inspected with the Matlab script PRE/ViewMeshQuality.m
which produces plots like Figure 4.3. The proper balance of the mesh with respect to the
following two criteria can be analyzed:
• Stability criterion, related to the largest stable timestep. The stability of each element
is quantified by
S = min(∆x/cP ).
We also define a stability index as
SI = log[S/median(S)].
where the median value is taken over the whole mesh. Red elements (small SI) are
relatively unstable and require small timesteps ∆t. Because ∆t is constant over the
whole mesh and the computational cost is inversely proportional to ∆t, these red elements penalize the computational efficiency. The mesh should be redesigned to increase
their size, as much as possible, while keeping them small enough to resolve the shortest
wavelength (see next).
• Resolution criterion, related to the number of nodes per shortest wavelength. The
resolution of each element is quantified by
R = min(cS /h).
We also define a resolution index as
RI = log[R/median(R)].
where the median value is taken over the whole mesh. Red elements (small RI) have
relatively poor resolution, in their vicinity the maximum frequency resolvable by the
mesh is limited. The mesh should be redesigned to decrease their size, as much as
possible. Conversely, elements with very high RI (blue) are smaller than required and
might increase the computational cost.
To minimize the CPU and memory cost of a simulation an ideal mesh design should minimize
the spread of the two indices above, by aiming at a ratio of element size to wave velocity,
h/c, as uniform as possible across the whole mesh. However, in some cases a poorly balanced
mesh is inevitable: in the example of Figure 4.3 the worst elements are near the edges of
the sedimentary basin, at a sharp velocity contrast. Small element sizes on the rock side are
inherited from the sediment mesh.3
Similar information is plotted by gv Stability sem2d.ps and gv Resolution sem2d.ps.
The indices in these files are however not logarithmic and are not normalized by the median.
3
In future releases of SEM2DPACK this penalty on computational efficiency will be reduced by nonconformal meshing with mortar elements, by timestep subcycling or by implicit/explicit timestep partitioning.
info
Mar 06, 08 18:04
Page 1/4
info
Mar 06, 08 18:04
Poisson’s ratio . . . . . .
First Lame parameter Lambda
Second Lame parameter Mu. .
Bulk modulus K. . . . . . .
Young’s modulus E . . . . .
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
Program S P E C F E M : start
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Page 2/4
.
.
.
.
.
= 250.021E−03
=
1.000E+00
=
1.000E+00
=
1.667E+00
=
2.500E+00
B o u n d a r y
C o n d i t i o n s
=====================================
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
D a t e : 06 − 03 − 2008
T i m e : 18:03:23
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
***********************************************
*
I n p u t
p h a s e
*
***********************************************
Boundary tag. . . . . . . .
Boundary condition. . . . .
Type of absorbing boundary.
Allow incident wave . . . .
.
.
.
.
. . . (tag)
. . .(kind)
. .(stacey)
.(let_wave)
=
=
=
=
2
ABSORB
Clayton−Engquist
T
Boundary tag. . . . . . . .
Boundary condition. . . . .
Type of absorbing boundary.
Allow incident wave . . . .
.
.
.
.
. . . (tag)
. . .(kind)
. .(stacey)
.(let_wave)
=
=
=
=
3
ABSORB
Clayton−Engquist
T
G e n e r a l
P a r a m e t e r s
===================================
Execution mode . . . . . . . . . . . . . . . (iexec)
Number of nodes per edge . . . . . . . . . . .(ngll)
Number of d.o.f per node . . . . . . . . . . .(ndof)
Highest frequency to be resolved . . . . . . .(fmax)
Print progress information during
input phase . . . . . . . .(verbose(1))
initialization phase . . . .(verbose(2))
checking phase . . . . . . .(verbose(3))
solver phase . . . . . . . .(verbose(4))
Frequency for solver progress information .(itInfo)
S o u r c e s
=============
= solve
= 6
= 1
=
1.250E+00
=
=
=
=
=
X−position (meters). . . . .(coord(1))
Y−position (meters). . . . .(coord(2))
Source time function . . . . . . . . .
Fundamental frequency (Hz) . . . .(f0)
Time delay (s) . . . . . . . . (onset)
Multiplying factor . . . . . . (ampli)
Source Type. . . . . . . . . . . . . .
If P−SV: counterclockwise angle / up .
T
T
T
T
1000
Method. . . . . . . . . . .
Minimum X . . . . . . . . .
Maximum X . . . . . . . . .
Minimum Z . . . . . . . . .
Maximum Z . . . . . . . . .
Number of elements along X.
Number of elements along Z.
Cut by horizontal fault . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .(method)
. (xlim(1))
. (xlim(2))
. (zlim(1))
. (zlim(2))
.(nelem(1))
.(nelem(2))
. .(faultx)
=
=
=
=
=
=
=
=
Number of receivers . . . .
Subsampling for seismograms
Field recorded. . . . . . .
Axis of the seismogram plot
CARTESIAN
0.000E+00
3.000E+01
0.000E+00
3.000E+01
60
60
F
.
.
.
.
.
. . .(kind)
. (NbSteps)
. . . .(Dt)
. (Courant)
(TotalTime)
=
=
=
=
=
leapfrog
will be set later
will be set later
0.30
35.000E+00
M a t e r i a l
P r o p e r t i e s
=====================================
Number of materials . . . . . . . . . . . = 1
Thursday March 06, 2008
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. (tag)
.(kind)
. .(cp)
. .(cs)
. (rho)
.
.
.
.
.
.
.
.
.
.
.
.
(number)
.(isamp)
.(field)
.(irepr)
=
=
=
=
7
1
D
D
Timestep of first snapshot output . . . . . . (it1)
Number of timesteps between snapshots. . . . . (itd)
Save results in PS file or not . . . . . . . . .(ps)
Save grid triangulation for GMT. . . . . . . . (gmt)
Save results in AVS file or not. . . . . . . . (avs)
Save results in Visual3 file or not. . . . (visual3)
Save results in binary file or not . . . . . . (bin)
Selected fields :
Displacement . . . . . . . . . . . . . . . . . . .
Velocity . . . . . . . . . . . . . . . . . . . . .
Acceleration . . . . . . . . . . . . . . . . . . .
Strain . . . . . . . . . . . . . . . . . . . . . .
Stress . . . . . . . . . . . . . . . . . . . . . .
Selected components for PostScript snapshots :
X . . . . . . . . . . . . . . . . . . . . . . . .
Y . . . . . . . . . . . . . . . . . . . . . . . .
Z . . . . . . . . . . . . . . . . . . . . . . . .
Amplitude . . . . . . . . . . . . . . . . . . . .
=
=
=
=
=
=
=
0
100000
F
F
F
F
F
=
=
=
=
=
F
T
F
F
F
=
=
=
=
F
T
F
F
= 1
= Elastic
=
1.732E+00
=
1.000E+00
=
1.000E+00
***********************************************
*
I n i t i a l i z a t i o n
p h a s e
*
***********************************************
info
1/2
57
.
.
.
.
.
. . . . .
recording
. . . . .
. . . . .
S n a p s h o t
O u t p u t s
===============================
T i m e
i n t e g r a t i o n
===============================
Material index.
Material type .
P−wave velocity
S−wave velocity
Mass density. .
0.000E+00
0.000E+00
Ricker
500.000E−03
3.000E+00
250.000E−03
Collocated Force
0.00
R e c e i v e r s
=================
M e s h
G e n e r a t i o n
=============================
Scheme. . . . . . . . . .
Number of time steps. . .
Time step increment . . .
Courant number. . . . . .
Total simulation duration
=
=
=
=
=
=
=
=
4.5 Verifying the settings and running a simulation
Printed by Jean Paul Ampuero
info
Mar 06, 08 18:04
Page 3/4
info
Mar 06, 08 18:04
5
6
7
Defining the FEM mesh ...... [OK]
Saving node coordinates in file MeshNodesCoord_sem2d.tab ...... [OK]
Saving element connectivity in file ElmtNodes_sem2d.tab ...... [OK]
20.000E+00
25.000E+00
30.000E+00
0.000E+00
0.000E+00
0.000E+00
Maximum distance between asked and real =
Sampling rate (Hz)
Sampling timestep (secs)
Total number of samples
Number of receivers
S p e c t r a l
e l e m e n t s
g r i d
===========================================
Numbering GLL points ...... [OK]
Total number of GLL points. . . . . . . = 90601
Page 4/4
20.000E+00
25.000E+00
30.000E+00
0.000E+00
0.000E+00
0.000E+00
0.000E+00
0.000E+00
0.000E+00
0.000E+00
=
56.751E+00
=
17.621E−03
= 1988
= 7
... [OK]
Initializing sources ...
Saving element/node table in binary file ibool_sem2d.dat ...... [OK]
Defining nodes coordinates ...... [OK]
S o u r c e s
=============
Saving the grid coordinates (coord) in a text file ...... [OK]
Saving the grid coordinates (coord) in a binary file ...... [OK]
Sources have been relocated to the nearest GLL node
M a t e r i a l
p r o p e r t i e s
=====================================
Translating input model ...... [OK]
Exporting model ...... [OK]
z−requested
x−obtained
z−obtained
distance
0.000E+00
0.000E+00
0.000E+00
0.000E+00
0.000E+00
RESOLUTION: nodes per min wavelength =
for maximum frequency
=
minimum wavelength =
−−− CPU TIME ESTIMATES (in seconds) :
CPU time for initialization . . 806.877E−03
CPU time per timestep . . . . . 18.997E−03
Total solver CPU time . . . . . 37.747E+00
(mins) . . . . 629.118E−03
(hours). . . . 10.485E−03
8.000E+00
1.250E+00 Hz
1.600E+00 m
Dump PostScript Resolution_sem2d.ps ...... [OK]
Dump PostScript Stability_sem2d.ps ...... [OK]
Timestep #
T i m e
s o l v e r
=====================
Time step (secs)
Number of time steps
Total duration (secs)
Courant number
=
1000
t =
17.621E+00
−−− CPU TIME INFORMATION (in
CPU time for initialization
CPU time per timestep . . .
Total solver CPU time . . .
(mins) . .
(hours). .
=
17.621E−03
= 1987
=
35.013E+00
= 300.000E−03
CFL number
0.000E+00
***********************************************
*
S o l v e r
p h a s e
*
***********************************************
Checking mesh ...... [OK]
Max mesh size = 142.616E−03
Min mesh size =
58.736E−03
Ratio max/min =
2.428E+00
300.000E−03
vmax =
91.661E−03
dmax =
28.653E−03
seconds) :
. . 806.877E−03
. . 20.395E−03
. . 40.524E+00
. . 675.397E−03
. . 11.257E−03
Storing sismos data (SEP format) ...
Defining work arrays for elasticity ...... [OK]
Initializing kinematic fields ...... [OK]
Max displ =
0.000E+00
Max veloc =
0.000E+00
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
Program S P E C F E M : end
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
Test SH
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
D a t e : 06 − 03 − 2008
T i m e : 18:04:11
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−
Building the mass matrix ...... [OK]
Defining boundary conditions ...... [OK]
Initializing receivers ...
R e c e i v e r s
=================
Receivers have been relocated to the nearest GLL node
x−requested
z−requested
x−obtained
z−obtained
1
2
3
4
0.000E+00
5.000E+00
10.000E+00
15.000E+00
0.000E+00
0.000E+00
0.000E+00
0.000E+00
0.000E+00
5.000E+00
10.000E+00
15.000E+00
0.000E+00
0.000E+00
0.000E+00
0.000E+00
distance
0.000E+00
0.000E+00
0.000E+00
0.000E+00
info
2/2
58
Receiver
Thursday March 06, 2008
x−requested
1
Maximum distance between requested and real =
0.000E+00
... [OK]
Timestep #
0 t =
0.000E+00 vmax =
0.000E+00 dmax =
M e s h
p r o p e r t i e s
=============================
STABILITY:
Source
4.5 Verifying the settings and running a simulation
Printed by Jean Paul Ampuero
4.6 Outputs, their visualization and manipulation
4.6
59
Outputs, their visualization and manipulation
In addition to the screen output described above, sem2dsolve generates different files and
scripts that allow the user to control the parameters of the simulation and to display the
results. All the outputs files follow the naming convention SomeName sem2d.xxx, where xxx
is one of the following extensions: tab for ASCII data files, txt for other text files, dat for
binary data files, etc. This makes it easy to clean a working directory with a single command
like rm -f * sem2d*.
4.6.1
Spectral element grid
As explained in the previous section, sem2dsolve generates two PostScript files for mesh
quality checking purposes: Stability sem2d.ps and Resolution sem2d.ps. The relevant
information is contained in the files Stability sem2d.tab and Resolution sem2d.tab and
can also be inspected with the Matlab script PRE/ViewMeshQuality.m.
4.6.2
Source time function
sem2dsolve generates a file called SourcesTime sem2d.tab containing the source time function sampled at the same rate as the receivers. It is important to verify that the spectrum of
the source has little power at those high frequencies that are not well resolved by the mesh
(those that correspond to less than 5 nodes per wavelength). If this is not the case you must
be very cautious in the interpretation of the seismograms in the high frequency range, or
low-pass filter the results.
4.6.3
Snapshots
sem2dsolve generates snapshots at a constant interval defined, in number of solver timesteps,
by the input parameter itd of the SNAP DEF input block. An example is shown in Figure 4.4. Requested fields are exported in binary data files called xx XXX sem2d.dat, where
xx is the field code defined in the documentation of the PLOTS input block and XXX is
the 3-digit snapshot number. The user is encouraged to inspect the Matlab s function
POST/sem2d snapshot read.m to find more about the data formats and their manipulation.
Snapshots can also be exported as PostScript files xx XXX sem2d.ps. These can be merged
into an animated GIF (movie) file movie.gif by the script POST/movie.csh and displayed
by xanim movie.gif or animate movie.gif. An animated GIF can also be created by the
Matlab function POST/sem2d snapshot movie.m.
4.6 Outputs, their visualization and manipulation
60
Z
SEM2DPACK 2.11 - Spectral Element Method
Palos Grandes NS meshed with EMC2
Max = 2.252E+00
X
Time = 7.912E+00 s
Time step = 5000
Figure 4.4: Sample snapshot from EXAMPLES/UsingEMC2/: an obliquely incident SH plane
wave impinging on a sedimentary basin. The unstructured mesh of spectral elements is
plotted on background.
4.6 Outputs, their visualization and manipulation
4.6.4
61
Seismograms
The seismograms are stored using the SEP format, a simple binary block of single precision
floats. The components of the vector field (velocity by default) are stored in separate files
U* sem2d.dat, where * is x or z in P-SV and y in SH. The seismograms header is in the file
SeisHeader sem2d.hdr. Its second line contains the sampling timestep DT, the number of
samples NSAMP and the number of stations NSTA. The stations coordinates, XSTA and ZSTA,
are listed from the third line to the end of file. With this notations, U* sem2d.dat contains
a NSAMP×NSTA single precision matrix.
You can view the seismograms using any tool that is able to read the SEP format, which is
the case of almost all the softwares able to deal with seismic data. sem2dsolve generates
scripts for the XSU-Seismic Unix visualization tool4 :
– Xline sem2d.csh displays all seismograms together on screen
– PSline sem2d.csh plots all seismograms on PostScript files U*Poly sem2d.ps
– Xtrace sem2d.csh prompts the user for a trace number (between 1 and NSTA) and then
displays this particular trace on screen
– PStrace sem2d.csh does the same as Xtrace, but exports the traces as PostScript files
U*TraceXXX sem2d.ps where XXX is the number of that particular trace
The program post seis.exe performs similar basic manipulation and plotting (through
gnuplot) of the seismograms. Its interactive menu is self-explanatory. It is usually called inside a script, as in POST/seis b2a.csh (converts all seismograms to ASCII) or POST/seis plot.csh
(plots all seismograms together, an example is shown in Figure 4.5).
The script POST/sample seis.m shows how to manipulate and plot seismogram data in Matlab. It uses the functions POST/sem2d read seis.m and POST/plot seis.m.
4.6.5
Fault outputs
Fault data from dynamic rupture simulations is stored in three files (where XX is the boundary
tag of the first side of the fault, tags(1) of the BC SWFFLT input block):
– FltXX sem2d.hdr contains the information needed to read the other fault data files. Its
format, line by line, is:
1. NPTS NDAT NSAMP DELT (name of parameters)
2. Value of parameters above
3. Name of fields exported in FltXX sem2d.dat, separated by “:”
4. XPTS ZPTS (name of coordinate axis)
5. from here to the end of file: a two-column table of coordinates of the output fault
nodes
4
Seismic Unix is freely available from the Colorado School of Mines at http://timna.mines.edu/cwpcodes
4.6 Outputs, their visualization and manipulation
62
3500
3000
2500
D (m)
2000
1500
1000
500
0
-500
0
5
10
15
20
25
Time (secs)
Figure 4.5:
Sample
POST/seis plot.csh.
seismograms
from
EXAMPLES/UsingEMC2/
generated
with
– FltXX sem2d.dat contains the space-time distribution of fault data such as slip, slip rate,
stress and strength. Every DELT seconds a block of fault data values is written. The total
number of blocks is NSAMP. Each block has NDAT lines (one per fault data field) and NPTS
columns (one per fault node) 5 . Stresses are relative to their initial values.
– FltXX init sem2d.tab contains the spatial distribution of initial shear stress, initial
normal stress and initial friction (3 columns).
– FltXX potency sem2d.tab contains time-series of seismic potency and potency rate. The
seismic potency tensor pij is defined by the following integral along the fault:
Z
1
(ni ∆uj + nj ∆ui ) dx
(4.1)
pij =
2 f ault
where ∆u is slip and n is the local unit vector normal to the fault. The file contains one
line per timestep. In SH (ndof=1) each line has 4 columns: 2 components of potency
(p13 and p23 ) and 2 components of potency rate (ṗ13 and ṗ23 ). In P-SV (ndof=2) each
line has 3 components of potency (p11 , p22 and p12 ) and 3 components of potency rate
(ṗ11 , ṗ22 and ṗ12 ).
Some tools are available to manipulate the data in FltXX sem2d.dat:
– The script FltXX sem2d.csh shows how to extract ASCII time series of different fields
at given locations on the fault, using Seismic Unix tools.
5
The actual number of columns is NPTS +2: Fortran adds a one-word tag at the front and end of each
record.
4.6 Outputs, their visualization and manipulation
63
– The program post fault.exe performs basic manipulations of the fault data, including
conversion to an ASCII file readable by gnuplot. Its interactive menu is self-explanatory.
– The script POST/sample fault.m and function POST/sem2d read fault.m show how to
manipulate and plot fault data in Matlab.
4.6.6
Stress glut
For damage and plastic materials, the solver can export the plastic and damage components
of the cumulative stress glut tensors defined, respectively, as
ZZ
p
sij (t) = −
2µ ǫpij (t) dx dz
(4.2)
ZZ
(4.3)
sdij (t) =
[σij (t) − c0ijkl ǫekl (t)] dx dz
where ǫp is the plastic strain, ǫe the elastic strain, σ the absolute stress and c0 the tensor of
elastic moduli of the undamaged medium.
To enable this feature: set COMPUTE STRESS GLUT = .true. in file SRC/constants.f90, then
re-compile the code. The stress glut output is exported in the file stress glut sem2d.tab
in 7 columns: time, sp11 , sp22 , sp12 , sd11 , sd22 , sd12 .
4.6.7
Energies
The solver can export the cumulative plastic energy, the kinetic energy and the total change
of elastic energy, defined respectively by
ZZZ t
p
σij (t′ )ǫ̇pij (t′ ) dx dz dt′
(4.4)
E (t) =
0Z Z
E k (t) = 1/2
ρvi2 (t) dx dz
(4.5)
ZZ
E e (t) =
U [ǫe (t)] − U [ǫe (0)] dx dz
(4.6)
where U is the elastic potential and summation over subindices is implied.
To enable this feature: set COMPUTE ENERGIES = .true. in file SRC/constants.f90, then recompile the code. The energy output is exported in the file energy sem2d.tab in 4 columns:
time, E p , E k and Ee .
4.6.8
Matlab utilities
A range of functions and sample scripts for Matlab are available to read, manipulate and plot
output data. Add the directory POST/ to your Matlab path (addpath). For an overview of
existing utilities, type help POST:
4.6 Outputs, their visualization and manipulation
64
SEM2DPACK/POST provides Matlab utilities for the manipulation
and visualization of SEM2DPACK simulations results.
Reading simulation data:
SEM2D_READ_SPECGRID
SEM2D_SNAPSHOT_READ
SEM2D_READ_SEIS
SEM2D_READ_FAULT
reads
reads
reads
reads
a spectral element grid
snapshot data
seismogram data
fault data
Data manipulation:
SEM2D_EXTRACT_POINT
SEM2D_EXTRACT_LINE
ARIAS_INTENSITY
RESPONSE_SPECTRUM
extracts field values at an arbitrary point
extracts field values along a vertical or horizontal line
computes Arias Intensity and Significant Duration
computes response spectra (peak dynamic response
of single-degree-of-freedom systems)
Data visualization:
SEM2D_PLOT_GRID
plots a spectral element grid
SEM2D_SNAPSHOT_PLOT plots snapshot data
SEM2D_SNAPSHOT_GUI interactively plots snapshot data
SEM2D_SNAPSHOT_MOVIE makes an animation of snapshot data
PLOT_MODEL
plots velocity and density model
PLOT_SEIS
plots multiple seismograms
PLOT_FRONTS
space-time plot of rupture front and process zone tail
SAMPLE_FAULT
example of visualization of fault data
SAMPLE_SEIS
example of visualization of seismogram data
Miscellaneous tools:
XCORRSHIFT
SPECSHIFT
SPECFILTER
cross-correlation time-delay measurement
signal time shift by non-integer lag via spectral domain
zero-phase Butterworth filter via spectral domain
Chapter 5
Adding features to SEM2D (notes
for advanced users)
Sometimes you will need to add new capabilities to the SEM2DPACK solver, by modifying
the program. The following notes are intended to guide you through this process. We will
not give here a comprehensive description of the code architecture, only enough details to get
you started in performing safely the most usual and evident modifications.
5.1
Overview of the code architecture
[ ... in progresss ...]
This code uses a mixture of procedural (imperative) and object-oriented paradigms. Historically, it evolved from a purely procedural code.
Extensive use of modularity.
Object Oriented Programming (OOP) features (principles) applied in this code: encapsulation, classes, static polymorphism. These are not applied everywhere in the code, for different
reasons: reusage of legacy code, performance, difficulty related to the limits of Fortra 90, or
sections of code yet to be updated.
Added cost of structures containing pointer components: the possibility of pointer aliasing
prevents more agressive compiler optimizations and adds overhead for safety checks.
5.2
Accessible areas of the code
Some areas of the code have been written in such a way that a moderately experienced Fortran
95 programmer, with a limited understanding of the code architecture, can introduce new fea-
5.2 Accessible areas of the code
66
tures without breaking the whole system. This is achieved through modularity, encapsulation
and templates. The modifications that are currently accessible are:
• boundary conditions, see bc gen.f90
• material rheology, see mat gen.f90
• source time functions, see stf gen.f90
• spatial distributions, see distribution general.f90
The source files listed above contain step-by-step instructions, just follow the comments starting by !!.
Chapter 6
Frequently Asked Questions
6.1
SEM2D
Segmentation fault
This problem is often related to a small stack size in your computer settings. In your
Linux shell do: ulimit -s unlimited under bash or limit stacksize unlimited under
csh. Place this command in your startup files (.login, .bashrc or .cshrc).
Instabilities on very distorted elements
Very distorted elements (with very small or very large angles) are usual close to wedges of
sedimentary basins, fault branching points, etc. In general, distorted elements are less stable
than square elements: spurious motions with exponentially increasing amplitude might appear
in their vicinity. In most cases these instabilities can be suppressed by reducing the Courant
input parameter. There is currently no simple recipe to determine the maximum value of this
parameter, so trial and error is required.
6.2
EMC2
I can’t get rid of a few triangles
Obtaining a quality quad mesh is not always a trivial task. Trial and error and experience is
needed. This can be by far the most time consuming stage of modeling.
First make sure that the total number of element edges along the perimeter of each mesh
domain is even. This is a necessary topological condition to generate a quad-only mesh.
6.2 EMC2
68
When the geometry seems too complicated for quad meshing you should consider simplifying
the geometry, especially those details that are much smaller than the dominant wavelength.
If the above fails or does not apply, you have to help the mesher. The recommended procedure
in EMC2 is:
1. Divide your original mesh into simple domains, in such a way that most domains have
exactly four sides (possibly curved) and the remaining non-four-sided domains are as
small as possible.
2. Generate a structured quad-mesh (a regular grid) inside each four-sided domain with
the (QUADRANGULATE) tool of the PREP MESH mode, as described in section 5.2.13
of EMC2’s manual (note that this is not the same as the <QUADRANGULATE> button
in the EDIT MESH mode).
3. Proceed as usual (triangulation followed by quadrangulation) inside the remaining nonfour-sided domains. If these are small enough EMC2 should not have problems doing a
correct tri-to-quad meshing.
Bibliography
Ampuero, J.-P. (2002). Etude physique et numérique de la nucléation des séismes. Ph. D.
thesis, Université Paris 7, Denis Diderot, Paris.
Ampuero, J.-P. and Y. Ben-Zion (2008). Cracks, pulses and macroscopic asymmetry of
dynamic rupture on a bimaterial interface with velocity-weakening friction. Geophys. J.
Int. 173 (2), 674–692, doi:10.1111/j.1365–246X.2008.03736.x.
Andrews, D. (1999). Test of two methods for faulting in finite difference calculations. Bull.
Seis. Soc. Am. 89, 931–937.
Andrews, D. J. (2005). Rupture dynamics with energy loss outside the slip zone. J. Geophys.
Res. 110 (B1), B01307, doi:10.1029/2004JB003191.
Ben-Zion, Y. and V. Lyakhovsky (2006). Analysis of aftershocks in a lithospheric model
with seismogenic zone governed by damage rheology. Geophys. J. Int. 165 (1), 197–210.
Boaga, J., S. Renzi, G. Vignoli, R. Deiana, and G. Cassiani (2012). From surface wave
inversion to seismic site response prediction: Beyond the 1d approach. Soil Dyn. Earthq.
Engng. 36, 38–51.
Clayton, R. and B. Engquist (1977). Absorbing boundary conditions for acoustic and elastic
wave equations. Bull. Seis. Soc. Am. 67 (6), 1529–1540.
De la Puente, J., J. P. Ampuero, and M. Käser (2009). Dynamic rupture modeling on
unstructured meshes using a discontinuous Galerkin method. J. Geophys. Res. 114,
B10302, doi: 10.1029/2008JB006271.
De la Puente, J., M. Käser, M. Dumbser, and H. Igel (2007). An arbitrary high order
discontinuous Galerkin method for elastic waves on unstructured meshes IV: Anisotropy.
Geophys. J. Int. 169 (3), 1210–1228, doi:10.1111/j.1365–246X.2007.03381.x.
Dewangan, P., I. Tsvankin, M. Batzle, K. Van Wijk, and M. Haney (2006). PS-wave moveout inversion for tilted TI media: A physical-modeling study. Geophys. 71 (4), D135–
D143, doi: 10.1190/1.2212274.
Gabriel, A. A., J. P. Ampuero, L. A. Dalguer, and P. M. Mai (2012a). Source properties of
dynamic rupture pulses with off-fault plasticity. Geophys. Res. Let., submitted.
Gabriel, A. A., J. P. Ampuero, L. A. Dalguer, and P. M. Mai (2012b). The transition of
dynamic rupture styles in elastic media under velocity-weakening friction. J. Geophys.
Res., submitted.
Hamiel, Y., Y. F. Liu, V. Lyakhovsky, Y. Ben-Zion, and D. Lockner (2004). A viscoelastic damage model with applications to stable and unstable fracturing. Geophys. J.
Int. 159 (3), 1155–1165.
BIBLIOGRAPHY
70
Haney, M., R. Snieder, J.-P. Ampuero, and R. Hofmann (2007). Spectral element modelling of fault-plane reflections arising from fluid pressure distributions. Geophys. J.
Int. 170 (2), 933–951.
Harris, R. A. and et al (2011). Verifying a computational method for predicting extreme
ground motion. Seis. Res. Let. 82 (5), 638–644, doi: 10.1785/gssrl.82.5.638.
Huang, Y. and J.-P. Ampuero (2011). Pulse-like ruptures induced by low-velocity fault
zones. J. Geophys. Res. 116, B12307, doi: 10.1029/2011JB008684.
Huang, Y. and J.-P. Ampuero (2012). The competitive effects on slip pulse properties
of low-velocity fault zones and strong velocity-weakening friction. Geophys. Res. Let.,
submitted.
Huang, Y., L. Meng, and J.-P. Ampuero (2012). A dynamic model of the frequencydependent rupture process of the 2011 Tohoku-Oki earthquake. EPS in press,
doi: 10.5047/eps.2012.05.011.
Kaneko, Y., N. Lapusta, and J.-P. Ampuero (2008). Spectral element modeling of earthquake rupture on rate-and-state faults: Effects of velocity-strengthening friction at
shallow depths. J. Geophys. Res. 113 (B9), B09317, doi: 10.1029/2007JB005553.
Komatitsch, D. (1997). Méthodes spectrales et éléments spectraux pour l’équation de
l’élastodynamique 2D et 3D en milieu hétérogène. Ph. D. thesis, Institut de Physique
du Globe de Paris, Paris.
Komatitsch, D., C. Barnes, and J. Tromp (2000). Simulation of anisotropic wave
propagation based upon a spectral element method. Geophys. 65 (4), 1251–1260,
doi:10.1190/1.1444816.
Komatitsch, D. and J. Tromp (1999). Introduction to the spectral-element method for 3-D
seismic wave propagation. Geophys. J. Int. 139, 806–822.
Komatitsch, D. and J. P. Vilotte (1998). The spectral element method: an efficient tool
to simulate the seismic response of 2D and 3D geological structures. Bull. Seis. Soc.
Am. 88, 368–392.
Komatitsch, D., J. P. Vilotte, R. Vai, and F. J. Sánchez-Sesma (1999). The Spectral Element
method for elastic wave equations: application to 2D and 3D seismic problems. Int. J.
Num. Meth. Engng. 45 (9), 1139–1164.
Lyakhovsky, V., Y. Ben-Zion, and A. Agnon (1997). Distributed damage, faulting, and
friction. J. Geophys. Res. 102 (B12), 27635–27649.
Lyakhovsky, V., Y. Hamiel, J.-P. Ampuero, and Y. Ben-Zion (2009). Non-linear damage rheology and wave resonance in rocks. Geophys. J. Int. 178 (2), 910–920,
doi: 10.1111/j.1365–246X.2009.04205.x.
Madariaga, R., J.-P. Ampuero, and M. Adda-Bedia (2006). Seismic radiation from simple
models of earthquakes. In A. McGarr, R. Abercrombie, H. Kanamori, and G. di Toro
(Eds.), Earthquakes: Radiated Energy and the Physics of Earthquake Faulting, Volume
170 of Geophysical Monograph, pp. 223–236. Am. Geophys. Union.
Meng, L. and J.-P. Ampuero (2012). Slow rupture and weakly pressure-sensitive strength
enables compressional branching: Dynamic rupture simulations of the 2012 off-Sumatra
earthquake. Geophys. Res. Lett., submitted.
BIBLIOGRAPHY
71
O’Brien, G. and C. Bean (2011). An irregular lattice method for elastic wave propagation.
Geophys. J. Int. 187 (3), 1699–1707, doi: 10.1111/j.1365–246X.2011.05229.x.
Patera, A. (1984). A spectral element method for fluid dynamics: laminar flow in a channel
expansion. J. Comp. Phys. 54, 468–488.
Roberts, A. W., R. S. White, and P. A. F. Christie (2009). Imaging igneous rocks on
the North Atlantic rifted continental margin. Geophys. J. Int. 179 (2), 1024–1038,
doi: 10.1111/j.1365–246X.2009.04306.x.
Rubin, A. M. and J. P. Ampuero (2007). Aftershock asymmetry on a bimaterial interface.
J. Geophys. Res. 112, B05307, doi: 10.1029/2006JB004337.
Smith, S. and R. Snieder (2010). Seismic modeling and analysis of a prototype heated
nuclear waste storage tunnel, Yucca Mountain, Nevada. Geophys. 75 (1), T1–T8,
doi:10.1190/1.3273868.
Stacey, R. (1988). Improved transparent boundary formulations for the elastic-wave equation. Bull. Seism. Soc. Am. 78 (6), 2089–2097.
Vai, R., J. M. Castillo-Covarrubias, F. J. Sánchez-Sesma, D. Komatitsch, and J. P. Vilotte
(1998). Elastic wave propagation in an irregularly layered medium. Soil Dyn. Earthquake
Eng. 18 (1), 11–18.
Vignoli, G. and G. Cassiani (2010). Identification of lateral discontinuities via multi-offset
phase analysis of surface wave data. Geoph. Prosp. 58 (3), 389–413, doi: 10.1111/j.1365–
2478.2009.00838.x.
Vignoli, G., G. Cassiani, M. Rossi, R. Deiana, J. Boaga, and P. Fabbri (2012). Geophysical
characterization of a small pre-Alpine catchment. J. App. Geophys. 80, 32–42.
Vignoli, G., C. Strobbia, G. Cassiani, and P. Vermeer (2011). Statistical multioffset phase
analysis for surface-wave processing in laterally varying media. Geophys. 76 (2), U1–U11,
doi:10.1190/1.3542076.
Xu, S., Y. Ben-Zion, and J. P. Ampuero (2012a). Properties of inelastic yielding zones generated by in-plane dynamic ruptures: I. Model description and basic results. Geophys.
J. Int., submitted.
Xu, S., Y. Ben-Zion, and J. P. Ampuero (2012b). Properties of inelastic yielding zones
generated by in-plane dynamic ruptures: II. Detailed parameter-space study. Geophys.
J. Int., submitted.
">