guide_sofi3D.tex 137 KB
Newer Older
Tilman Steinweg's avatar
Tilman Steinweg committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
\documentclass[11pt,onecolumn,oneside]{article}
\usepackage{epsf,pstricks,pst-grad,color}
\usepackage{graphicx,amssymb,amsmath,makeidx,psfrag,subfigure}
\usepackage{a4wide,authda14,sws,wasysym,epsfig}
\usepackage{longtable,rotating,latexsym,fancyhdr,eso-pic}
\usepackage[colorlinks=true,citecolor=blue,linkcolor=blue,bookmarks=true,urlcolor=blue,hyperfootnotes=true,plainpages=false,pdfpagelabels=true,breaklinks=true]{hyperref} %create interal an click-on links
\usepackage{breakurl}
\usepackage{tikz}
\usepackage{tikz-3dplot}
\usepackage{listings}
\usepackage{multirow}\lstset{basicstyle=\bfseries}
\bibliographystyle{authda3}
\makeindex

\parindent0em \parskip1.5ex plus0.5ex minus 0.5ex
% -------------------------------------------------------------------

%\geometry{a4paper, left=20 mm, right= 25mm, top=2.0cm, bottom=2.0 cm}

\newcommand\BackgroundPicture{%
   \put(0,-50){
     \parbox[b][\paperheight]{\paperwidth}{
       \vfill
       \centering
%       \includegraphics[width=\paperwidth,height=\paperheight,
%                        keepaspectratio]{eps/title_page1.ps}

       \includegraphics[width=\paperwidth,height=30cm]{eps/title_page1.pdf}
       \vfill
     }}}
% The picture is centered on the page background
% -------------------------------------------------------------------

% add background picture for the titlepage
\AddToShipoutPicture{\BackgroundPicture}

% with this we ensure that the chapter and section 
% headings are in lowercase 
%\renewcommand{\chaptermark}[1]{markboth{#1}{}} 
\renewcommand{\sectionmark}[1]{\markright{\thesection\ #1}} 
\fancyhf{} %delete the current section for header and footer 
\fancyhead[L,RO]{\bfseries\thepage}
%\fancyhead[LE,RO]{\bfseries\thepage}  
\fancyhead[LO]{\bfseries\rightmark} 
%\fancyhead[RE]{\bfseries\leftmark}
\fancyhead[R]{\bfseries\leftmark}  
\renewcommand{\headrulewidth}{0.5pt}
 % make space for the rule 
\fancypagestyle{plain}{% 
 \fancyhead{} %get rid of the headers on plain pages 
 \renewcommand{\headrulewidth}{0} % and the line 
}

\begin{document}
\thispagestyle{empty}
\newcommand{\Rule}{\rule{\textwidth}{1mm}}
\newtheorem{theorem}{Hypothese}[section]
\begin{center}
\Rule \vspace{5mm}
\sffamily \bfseries \Huge
- SOFI3D - \\ 
seismic modeling with finite differences \\
3D - acoustic and viscoelastic version
\vspace{1mm}\Rule\\
\vspace{1 cm}
\Large\emph{Users Guide}\\
\vspace{2 cm}
\large Thomas Bohlen \par
\large Denise De Nil \par
\large Daniel K\"ohn \par
\large Stefan Jetschny\par
\vspace{2 cm}

\small Karlsruhe Institute of Technology\par
\small Department of Physics, Geophysical Institute \par
\small Hertzstrasse 9, 76187 Karlsruhe, Germany \par

\vfill
\includegraphics[height=30mm]{eps/kit_logo_en_4c_positiv.pdf}
\vfill
\large Karlsruhe, May 15$^{\rm{th}}$, 2015
\end{center}
\cleardoublepage
\thispagestyle{empty} 
\cleardoublepage
\thispagestyle{empty}
\ClearShipoutPicture

\tableofcontents
\newpage
\section{Requirements}
\label{requirements}

SOFI3D can be generally run under Windows and Linux. You just have to make sure that an MPI implementation (one of the three listed below) is running and a C-Compiler is available. However, the preferred platform is Linux, since most of the large scale cluster computer run under Linux. On our local workstation cluster we use Suse Linux and OpenMPI, most of the test are performed under this system. The following programs should be installed on you machine for a proper run and processing of modeling data. 

\begin{center}
% use packages: array
\begin{small}
\begin{tabular}{lll}
Program & Description & Weblink \\ 
OpenMPI & MPI Implementation & \tiny{\url{http://www.mcs.anl.gov/research/projects/mpich2}} \\
 & (for the parallelization) & \\
MPICH2 & MPI Implementation & \url{http://www.open-mpi.org} \\  
& (for the parallelization) & \\
LAM/MPI & MPI Implementation & \url{http://www.lam-mpi.org} \\
& (for the parallelization) & \\
C-Compiler & the whole code is written in C,& \\
& any C-Compiler should be able & \\
& to compile it & \\
Seismic Un*x & Seismic processing package, & \url{http://www.cwp.mines.edu/cwpcodes} \\
(SU)  & SOFI3D outputs seismic data & \\
& in the SU format & \\
Matlab & Preferred program package for & \url{http://www.mathworks.de} \\
(commercial)& the visualization of snapshots, also  & \\
& useful for the display and & \\
& processing of seismograms & \\
xmovie & Console based movie program, can & usually included in Linux distribution\\
& be used for the quick display & otherwise install from repository\\
& of snapshot data  &
\end{tabular}
\end{small}
\end{center}

\newpage 
\section{Quick guide}\label{qguide}
To compile the program do: 

 \lstinline{cd sofi3D/src}

then

 \lstinline{make all} or \lstinline{make sofi3D}

to compile the program. You may also use the shell script  \lstinline{compileSOFI3D.sh} located in  \lstinline{sofi3D/par}. For a successful compiler run you probably need to change the compiler options in  \lstinline{sofi3D/src/Makefile}. 

To run the program on 8 CPUs do

 \lstinline{cd sofi3D/par}

 \lstinline{mpirun -np 8 ../bin/sofi3D ./in_and_out/sofi3D.json}


You may also use the shell script  \lstinline{startSOFI3D.sh} located in  \lstinline{sofi3D/par} that also writes the screen output to the file  \lstinline{sofi3D/par/in_and_out/sofi3D.jout}.

The modeling parameters are specified in the sofi3D input file  \lstinline{./in_and_out/sofi3D.json}. The old input file format (*.inp) is not supported anymore in the latest release. The parameters should be more or less self-explanatory. The synthetic seismograms and snapshots are written to the files specified in the sofi3D input file.

In the current distribution, the model is generated on the fly by the function sofi3D/src/hh.c. This function generates a homogeneous medium with $v_p$=3500 m/s, $v_s$=2000 m/s, and $\rho$=2000 kg/$\mathrm{m}^3$. Instead, the function readmod.c can be used to read model info from external grid files. See readmod.c for the format of the external files. You can change the function which generates the model grid by switching the READMOD parameter in the sofi3D input file.

\section{License}\label{license}

SOFI3D 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, version 2.0 of the License only.
 
SOFI3D 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. You should have received a copy of the GNU General Public License along with SOFI3D. See file COPYING and/or  \lstinline{http://www.gnu.org/licenses/gpl-2.0.html}.

The authors of SOFI3D are listed in file  \lstinline{AUTHORS}.

\section{Introduction}\label{intro}
In order to extract information about the structure and composition of the crust from seismic observations, it is necessary to be able to predict how seismic wavefields are affected by complex structures.
Since exact analytical solutions to the wave equations do not exist for most subsurface configurations, the solutions can be obtained only by numerical methods. For iterative calculations of synthetic seismograms with limited computer resources fast and accurate modeling methods are needed. 

The FD modeling program SOFI3D, that is described in this guide, is based on the FD approach described by \cite{virieux:86} and \cite{levander:88}. The present program SOFI3D has the following extensions

\begin{itemize}
\item considers viscoelastic wave propagation effects like attenuation and dispersion \\
\cite{robertsson:94,blanch:95,bohlen:02},
\item employs higher order FD operators,
\item applies Perfectly Matched Layer boundary conditions at the edges of the numerical mesh \cite{komatitsch:07},
\item works in MPI parallel environment ONLY, i.e. SOFI3D is a implementation based on a domain decomposition
\cite{bohlen:02}.
\end{itemize}

In the following sections, we give a short description of the  theoretical background, the different modeling parameters and options and how the program is used in a parallel MPI environment.
                       
\newpage
\section{Theoretical Background}
\subsection{(Visco)elastic Wave Equations}
What happens in a rock, during the passage of a seismic wave? To answer this question, we consider an infinitesimal small cube, which is part of the rock. In this cube are atoms, ions or molecules which move in a not specified manner. If we could describe the microscopic motion of the particles we would be able to predict the particle displacements and therefore the propagation of the seismic wave in the rock. One of the fundamental principles in physics is the conservation of momentum. The total momentum $\MB{P}$ in the rock can be changed by three different ways:
\begin{enumerate}

\item Not a bit
\EQ{1:3}{\frac{\partial \MB{P}}{\partial t}=\frac{\partial}{\partial t}\biggl \{\int_V \rho(\MB{x},t)\; \MB{v}(\MB{x},t) dV \biggr \} = 0.
\notag} 

where $t$ denotes the time, $\MB{x}$ the position vector, $\rho$ the density- and $\MB{v}$ the particle velocity distributions in the rock. 

\item By adding particles to the cube or remove them, which leads to a momentum gain or loss in the cube:

\EQ{1:5}{\Phi=\int_A \rho \MB{v} (\MB{v} \cdot dA). \notag}

\item By forces acting on the surface of the cube, for example the hydrostatic pressure P, or  more generally the stress $p_{ij}$ (surface forces) and forces $f_i$ acting on all the particles in the volume (body forces). 
\end{enumerate}

After a short calculation we get the following system of linear 1st order differential equations: 

\EQ{1:15}{\rho \frac{dv_i}{dt} + v_i \rho \MB{\nabla} v_i = - \frac{\partial p_{ij}}{\partial x_j} + f_i.}

This equation is satisfied for any volume, independently if it is filled with a liquid, a gas or rock. Until now no proposition on how the medium deforms under an external stress was made. For an {\bf{elastic medium}} exists a linear relationship between stress and strain:

\EQ{2:18}{p_{ij}=\lambda \theta \delta_{ij} + 2 \mu \epsilon_{ij}.}

with

\EQ{2:18:1}{\begin{split}
	\lambda\; \text{and}\; \mu:\; & \text{Lam$\acute{\rm e}$ parameter}\notag
\end{split}}

\EQ{2:18:1:1}{\epsilon_{ij}=\biggl(\frac{\partial u_i}{\partial x_j}+\frac{\partial u_j}{\partial x_i}\biggr)\; \text{deformation tensor} \notag}  
\EQ{2:18:2}{\theta = \epsilon_{11} + \epsilon_{22} + \epsilon_{33}:\; \text{cubic dilatation} \notag}
\EQ{2:18:3}{\delta_{ij}=
\begin{cases}
	1 & \text{for } i=j\\
	0 & \text{else}
\end{cases} 
\biggr \}: \;\text{Kronecker symbol} \notag}

For a {\bf{viscoelastic medium}} the stress-strain relationship is more complicated:

\EQ{2:19}{\begin{split}
	\dot{p}_{ij}&=\frac{\partial v_k}{\partial x_k}\{\pi (1+\tau^s)-2 \mu(1+\tau^s)\}+2\frac{\partial v_i}{\partial x_j}
	\mu(1+\tau^s)+\sum\limits_{l=1}^L r_{ijl}\; \text{if $i$=$j$}\\
	\dot{p}_{ij}&=\biggl(\frac{\partial v_i}{\partial x_j}+\frac{\partial v_j}{\partial x_i}\biggr) \mu (1+\tau^s)+\sum\limits_{l=1}^L r_{ijl}\; \text{if $i$$\ne$$j$}
\end{split}}

 with the memory variables 

 \EQ{2:20}{\begin{split}
	 \dot{r}_{ijl}&=-\frac{1}{\tau_{\sigma l}}\biggl\{(\pi \tau^p - 2 \mu \tau^s)\frac{\partial v_k}{\partial x_k} + 2 \frac{\partial
 	v_i}{\partial x_j} \mu \tau^s + r_{ijl} \biggr\}\; \text{if $i$ = $j$} \notag\\
 	\dot{r}_{ijl}&=-\frac{1}{\tau_{\sigma l}}\biggl\{\mu \tau^s \biggl(\frac{\partial v_i}{\partial x_j}+ \frac{\partial v_j}{\partial x_i} 
 	\biggr) + r_{ijl} \biggr\}\; \text{if $i$ $\ne$ $j$} \notag\\
 \end{split}}

$\tau^s$ and $\tau^p$ are related to the seismic quality factor $Q_s$, $Q_p$ for the S- respectively the P-wave velocity, by 

\EQ{2:21}{\begin{split}
	\tau^s &= 2/Q_s\\
	\tau^p &= 2/Q_p \notag
\end{split}}  

\section{Solving the Elastic Wave Equation with Finite Differences}
To solve the equations \ER{1:15} and \ER{2:18} for an elastic medium, or \ER{1:15} and \ER{2:19} for a viscoelastic medium, the particle velocities $\vec{v}$, the stresses $p_{ij}$, the Lame parameters $\lambda$ and $\mu$ are calculated at  discrete cartesian coordinates  $x=i\; \times dx$, $y=j\; \times dy$, $z=k\; \times dz$ and at discrete times $t=n\; \times dt$ on a grid. $dx$, $dy$ and $dz$, denote  the spatial grid point distance in x-, y-, z-direction and $dt$ the difference between two succeeding time steps with $i \in N | [1,NX]$, $j \in N | [1,NY]$, $k \in N | [1,NZ]$ 
and $n \in N | [1,NT]$, where $NX$, $NY$, $NZ$ and $NT$ denote the number of spatial grid points and time steps. Finally the partial derivatives  are replaced by {\bf{(finite)-difference}} operators. The derivative of a function y after a variable x can be approximated by a forward $D^+$ or a backward operator $D^-$:

\EQ{disc:1}{\begin{split}
	D^+_x y&= \frac{y[i+1]-y[i]}{dh} \hspace{1 cm} \text{forward operator}\\
	D^-_x y&= \frac{y[i]-y[i-1]}{dh} \hspace{1 cm} \text{backward operator}\\
\end{split}}

To calculate with a larger grid point distance the variables are arranged on a staggered grid (\cite{virieux:86} and \cite{levander:88}) (Figure \ref{fig_cell}).  Please note, that the vertical axis is denoted by \textit{Y}, e.g. the indices of the stress components are labeled accordingly. To satisfy the stability of the {\bf{Standard Staggered Grid (SSG)}} code the density $\rho$, respectively. the Lame parameter $\mu$ are arithmetically and harmonically averaged (\cite{bohlen:06}):

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/ssg_grid.pdf, width=13 cm}
\caption{\label{fig_cell} Grid geometry for a Standard Staggered Grid (SSG).}
\end{center}
\end{figure}   

\newpage
\subsection{Accuracy of FD-operators}
In the previous section, the partial derivation was simply replaced by a finite difference quotient. In this section, a more systematical approach  is used. First the first derivation of the variable f at a grid point i is calculated using the following Taylor expansion:

\EQ{op_acc:1}{\begin{split}
	(2k-1)\frac{\partial f}{\partial x}\biggr|_i&=\frac{1}{dh}(f_{i+(k-1/2)}-f_{i-(k-1/2)})\\
	&+\frac{1}{dh}\sum_{l=2}^N \frac{(k-\frac{1}{2} dh)^{2l-1}}{(2l-1)!}\frac{\partial^{(2l-1)}f}{\partial
	x^{(2l-1)}}\biggr|_i+{\mathcal{O}}(dh)^{2N} \notag
\end{split}}

For an FD operator with length 2N N equations with a weighting factor $\beta_k$ are added:

\EQ{op_acc:2}{\begin{split}
	[\sum_{k=1}^{N} \beta_k (2k-1)]\frac{\partial f}{\partial x}\biggr|_i&=\frac{1}{dh} \sum_{k=1}^{N} \beta_k (f_{i+(k-1/2)}-f_{i-(k-1/2)})\\
	&+\frac{1}{dh}\sum_{k=1}^{N} \sum_{l=2}^N \beta_k \frac{(k-\frac{1}{2} dh)^{2l-1}}{(2l-1)!}\frac{\partial^{(2l-1)}f}{\partial
	x^{(2l-1)}}\biggr|_i+{\mathcal{O}}(dh)^{2N}
\end{split}}

For the case N=1 we get the FD operator from the previous section with length 2N=2. The Taylor series expansion will be aborted after the first  term (${\mathcal{O}}(dh)^{2}$). This operator is called {\bf{2nd order FD-operator}} which denotes the abortion error of the Taylor series and not the order of the desired approximated derivation. To understand equation \ER{op_acc:2} in more detail the coefficients for a {\bf{4th order FD-operator}} are calculated. The 4th order FD operator has the length 2N = 4, so N=2. By calculating the sums in \ER{op_acc:2} we get:

\EQ{op_acc:3}{\begin{split}
	(\beta_1+3 \beta_2) \frac{\partial f}{\partial x}\biggr|_i&=\frac{1}{dh}(\beta_1(f_{i-1/2}-f_{i+1/2})+\beta_2(f_{i-3/2}-f_{i+3/2}))\\
	&+\frac{dh^3}{dh}\biggl[\beta_1\frac{1}{8 \cdot 3!}+\beta_2\frac{27}{8 \cdot 3!}\biggr]\frac{\partial^3 f}{\partial x^3}\biggr|_i
\end{split}}

The weights $\beta_k$ are calculated in the following way:\\ 
The coefficients in front of the derivation on the LHS of Equation \ER{op_acc:3} should be equal to 1:

\EQ{op_acc:4}{(\beta_1+3\beta_2)=1 \notag.}

The coefficients before the derivation $\frac{\partial^3 f}{\partial x^3}\biggr|_i$ on the RHS of Equation \ER{op_acc:3} are vanishing:

\EQ{op_acc:5}{(\beta_1+27\beta_2)=0 \notag.}

The weighting coefficients $\beta_k$ can be calculated by inverting the following matrix equation:

\EQ{op_acc:6}{
\left(
\begin{array}{ll}
	1 & 3 \\
	1 & 27 \\
\end{array}
\right)\cdot \hspace{0.2 cm}
\left(
\begin{array}{l}
	\beta_1 \\
	\beta_2 \\
\end{array}
\right)
=
\left(
\begin{array}{l}
	1 \\
	0 \\
\end{array}
\right) \notag
}

The resulting coefficients are $\beta_1=9/8$ and $\beta_2=-1/24$, so the forward and backward 4th order FD operators look like:

\EQ{op_acc:7}{\begin{split}
	\frac{\partial f}{\partial x}\biggr|_{i+1/2}&=\frac{1}{dh}[\beta_1 (f_{i+1}-f_i)+\beta_2 (f_{i+2}-f_{i-1})] \hspace{1 cm} \text{forward operator}\\
	\frac{\partial f}{\partial x}\biggr|_{i-1/2}&=\frac{1}{dh}[\beta_1 (f_{i}-f_{i-1})+\beta_2 (f_{i+1}-f_{i-2})] \hspace{1 cm} \text{backward operator}\\
\end{split}}

The coefficients $\beta_i$ in the FD operator are called {\bf{Taylor coefficients}}. The accuracy of higher order FD operators can be improved significantly simply by slightly changes in the FD coefficients \cite{holberg:87}. These numerically optimized coefficients are called {\bf{Holberg coefficients}}. An accuracy comparison between Taylor and Holberg coefficients can be found in section \ref{comp_taylor_holberg}.


326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
\subsection{Adams-Bashforth higher-order accurate time integrators}
To improve the precision of the temporal discretization the Adams-Bashforth third-order and fourth-order accurate time integrator can be used. 
The staggered Adams-Bashforth method (ABS) is a multi-step method that uses previously calculated wavefields to increase the accuracy order in time.\\
In \cite{bohlen2015higher2} and \cite{bohlen2015higher} we give a detailed study of the performance of the ABS method. The analysis shows that the numerical dispersion is much lower than that of the widely used second-order leapfrog method. Numerical dissipation is introduced by the ABS method which is significantly smaller for the ABS method of fourth-order accuracy. In 1-D and 3-D simulation experiments the convincing improvements of simulation accuracy of the fourth-order ABS method is verified. In a realistic elastic 3-D scenario the computing time reduces by a factor of approximately 2.2, whereas  the memory requirements increase by approximately the same factor. \\
The ABS method thus provides an alternative strategy to increase the time accuracy by investing computer memory instead of computing time.\\

In the following we provide an extract of \cite{bohlen2015higher2} to demonstrate the underlying theory. Additionally seismograms of different temporal orders of accuracy and an analysis of the numerical simulation error are shown.\\ 
\subsubsection{Theory}
For the sake of simplicity we illustrate the staggered Adams-Bashforth method (ABS-method) using the 1-D acoustic wave equation in velocity-stress formulation
\begin{eqnarray}
\frac{\partial p(x,t)}{\partial t}&=& - \pi (x) \frac{\partial v(x,t)}{\partial x}  \notag \\
\frac{\partial v(x,t)}{\partial t}&=& -\rho^{-1}(x) \frac{\partial p(x,t)}{\partial x}
\label{waveequation}
\end{eqnarray}
The wavefield variables are the pressure $p(x,t)$ and the particle velocity $v(x,t)$. The material is described by the P-wave modulus $\pi (x)$ and the mass density $\rho(x)$. For simplicity we omit the temporal ($t$) and spatial dependencies ($x$)  in the following.

Using the conventional second order staggered grid approximation to the first order time derivative we obtain
\begin{eqnarray}
\frac{p |^{n+1/2}-p|^{n-1/2}}{\Delta t}&=& - \pi \frac{\partial v}{\partial x}\bigg |^n  + \mathcal{O}(\Delta t^2) \nonumber \\ 
\frac{v|^n-v|^{n-1}}{\Delta t}&=& -\rho^{-1} \frac{\partial p}{\partial x}\bigg |^{(n-1/2)} + \mathcal{O}(\Delta t^2)
\label{2ndorder}
\end{eqnarray}

This results in the conventional explicit second order accurate time-stepping (leapfrog) scheme
\begin{eqnarray}
p|^{n+1/2}&=&p|^{n-1/2} - \Delta t \pi \frac{\partial v}{\partial x}\bigg |^n  + \mathcal{O}(\Delta t^2) \nonumber \\
v|^n&=&v|^{n-1} -\Delta t \rho^{-1} \frac{\partial p}{\partial x}\bigg |^{(n-1/2)}  + \mathcal{O}(\Delta t^2)
\label{2ndorderexplicit}
\end{eqnarray}
which requires no additional storage of the wavefield variables $p$ and $v$. 

With the ABS-method the order of the temporal integration  can be increased by using previous time levels of the right hand sides of equation \ref{2ndorder} \cite{ghrist2000staggered}.
The ABS-method thus requires the storage of previous time levels of spatial derivatives of the pressure $p$ and particle velocity $v$. The time update for the ABS-method thus reads
\begin{eqnarray}
p|^{n+1/2}&=&p|^{n-1/2} - \Delta t \pi \sum_{k=0}^{M-1} a_k \frac{\partial v}{\partial x}\bigg |^{n-k}   + \mathcal{O}(\Delta t^M) \nonumber \\
v|^n&=&v|^{n-1} -\Delta t \rho^{-1}  \sum_{k=0}^{M-1} a_k  \frac{\partial p}{\partial x}\bigg |^{(n-1/2-k)}  + \mathcal{O}(\Delta t^M)
\label{absmethod}
\end{eqnarray}
The weights for the time accuracy orders $M=2,3,4$ are given in Table \ref{absweights}. 

\begin{table}[ht]
\begin{center}
\begin{tabular}{|c|c|c|c|c|}\hline
$M$ & $a_0$ & $a_1$ & $a_2$ & $a_3$  \\ \hline
2 & 1             &  0           &  0           &     0   \\
3 &  25/24     &  -1/12         & 1/24   & 0   \\
4 & 13/12      &  -5/24      &  1/6       &  -1/24     \\ \hline
\end{tabular}
\end{center}
\caption{\label{absweights}Weights used for the summation of previous time levels in the Adams-Bashforth method \protect\cite{ghrist2000staggered}.}
\end{table}
\newpage
\subsubsection{Accuracy in 1-D}
We first compare seismograms calculated with the 1-D ABS-method using equations \ref{absmethod} for different orders of accuracy in time ($M$). We use a 1-D homogeneous acoustic 
medium with a wave velocity of $c=3500$\,m/s and constant density of $\rho=2000$\,kg/m$^3$ . The source signal is a Ricker signal with a center frequency of 600\,Hz. We choose a large source-receiver distance of 120 dominant wavelength to emphasize the effects of numerical dispersion and numerical dissipation in the synthetic seismograms. The spatial derivatives are
computed with high accuracy using a centered staggered FD stencil of 8th order accuracy.  The spatial grid spacing is held constant at $\Delta x=0.4$\,m corresponding to approximately 14 grid points per dominant wavelength. Discrepancies to the analytical solution (time-shifted Ricker signal) are thus mainly caused by the chosen time step interval $\Delta t$ and the order of the temporal discretization ($M=2,3,4$). The numerical results for Courant numbers $r=c \Delta t/\Delta x=0.4, 0.2, 0.1$ and temporal orders of accuracy $M=2, 3, 4$ are compared with the analytical solution in Figure \ref{fig_1dfdseis}.  For a large Courant number of $r=0.4$ (corresponding to a large time step interval) and second order approximation of the time derivative ($M=2$) we can observe a large time dispersion error causing a leading phase with high amplitude (Figure \ref{fig_1dfdseis}, top-left). Figure  \ref{fig_1dfdseis} compares two ways to reduce this time discretization error. The conventional way is to  stay with the given temporal accuracy order (typically $M=2$) and then reduce the time step interval, i.e. Courant number. The resulting waveforms are shown in the rows of Figure  \ref{fig_1dfdseis}.
Reducing the Courant number (time step interval) will increase the computation time proportional to $1/r$.  The alternative  way proposed in this study is to increase the order of the temporal discretization with the ABS-method which requires to store $M-1$ previously calculated spatial derivates of wavefields. The resulting waveforms are shown in the columns of  Figure  \ref{fig_1dfdseis}. \\ \\
%
%In Figure \ref{fig_1dfdseis} we see that the simulation accuracy increases with decreasing Courant number and increasing accuracy order.  In order to quantify the corresponding numerical simulation error we calculated the normalized L2-misfit between the numerical and analytical solution for different Courant numbers and accuracy orders 
%$M = 2, 3, 4$. 
%In Figure \ref{fig_L2} we plot the normalized L2-misfit over the number of time steps $NT=T/\Delta t=T c/(r \Delta x)$ where the propagation time of waves is held fixed at $T=0.24$\,s. We can observe the expected higher order reduction of the simulation accuracy which reduces proportional to $\Delta t^{M}$ or $NT^{-M}$. For a given accuracy level of $E=0.1\%$ (horizontal line in Figure \ref{fig_L2}) the required numbers of time steps are 39233, 13938, and 8704 for $M=2,M=3$ and $M=4$, respectively. The number of time steps thus reduces  to $36\%$  and $22\%$  when we increase the temporal order from $M=2$ to  $M=3$ and $M=4$, respectively (Figure  \ref{fig_L2} ). The ABS methods thus allows to significantly decrease the number of time steps and thus to save computing time. \\ \\
\begin{figure*}[!h]
\begin{center}
\includegraphics[width=\textwidth]{eps/seismo.pdf}
\caption{Seismograms (red lines) calculated for the 1-D case for different Courant numbers $r$ and accuracy orders $M$. $M=2$ corresponds to the classical second order leapfrog scheme (equations \ref{2ndorder}). $M=3,4$ correspond to the multi-step ABS method (equations \ref{absmethod}). The analytical solution is plotted as a black line. }
\label{fig_1dfdseis}
\end{center}
\end{figure*}
%\begin{figure}[h]
%\begin{center}
%\includegraphics[scale=0.7]{eps/error_1D_courant}
%\caption{Relative error of 1-D simulations (Figure \ref{fig_1dfdseis}). 
%The normalized L2 norm is plotted over the required number of time steps ($NT$). The temporal accuracy orders are $M=2,3,4$. The spatial accuracy is fixed at order $N=8$. 
%The number of time steps to achieve a given accuracy of $E=0.01\%$ (horizontal line) reduce to $36\%$ and $22\%$ for the temporal orders $M=3$ and $M=4$, respectively. 
%For large $NT$ (small time step intervals) the error converges to the fixed error of the spatial discretization. }
%\label{fig_L2}
%\end{center}
%\end{figure}

\newpage
Florian Wittkamp's avatar
Florian Wittkamp committed
406
\subsubsection{Accuracy in 3-D}
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
In Figure \ref{fig_1dfdseis} we see that the simulation accuracy in 1-D increases with decreasing Courant number and increasing accuracy order.  In order to quantify the corresponding numerical simulation error of the 3-D elastic update scheme we calculated the normalized L2-misfit between the numerical and analytical solution for different Courant numbers and accuracy orders 
$M = 2, 3, 4$. 
We compared the synthetic seismograms with an analytical solution for an explosive point source in a homogeneous full space. The elastic model parameters are $v_p=3500$\,m/s 
and $v_s=2000$\,m/s for the P- and S-velocities, respectively,  and $\varrho=2000$\,kg/m$^3$ for the density. The explosive point source generates  a Ricker signal with a dominant frequency of $600$\,Hz. We analyze the accuracy of the direct P-wave in a distance of $180$\,m corresponding to 30 dominant wavelength. The size of the model grid is 800x400x400 grid points. The direct P-wave is spatially sampled with approximately 14  grid points per dominant wavelength. The influence of numerical dispersion due to the discretization in space  is small because of the
chosen high spatial accuracy  order of $N=8$. The simulations are performed on 80 cores on a small-scale shared memory cluster.

As a measure of accuracy we use the normalized L2-norm between the numerical and analytical seismograms which only contain the direct P-wave. The results are shown
in Figure \ref{fig_L23D} (left). The L2-error reduces with increasing number of time steps $NT=T/\Delta t$ because of the decreasing time step interval $\Delta t$. (The wave propagation time $T=0.07$\,s is held constant).
At higher orders of the time accuracy ($M$) the error reduces more rapidly  proportional to $\Delta t^M$. 
%The observed overall behavior of the convergence of accuracy in the  3-D elastic case (Figure \ref{fig_L23D}, left) is quite similar to the convergence obtained for the 1-D scheme shown in Figure \ref{fig_L2}. 
%This indicates that the numerical dispersion and dissipation properties derived in chapters
%\ref{chap1Ddisp} and \ref{chap1Ddiss} for the 1-D case are also applicable to P-waves in 3-D media.

In Figure \ref{fig_L23D} (left)  we see see that we can reduce the number of time steps that are required to achieve a certain level of accuracy by increasing the temporal accuracy order $M$. For a given error level of $E=0.1\%$ the required number of  time steps reduce to approximately $43\%$ and $29\%$ when increasing the temporal order from $M=2$ to  $M=3$ and $M=4$, respectively. The corresponding run times decrease less significant because the number of floating point operations also increase with $M$. The observed relation between the number of time steps and the total run time is plotted in Figure  \ref{fig_L23D} (right). The computational requirements are also summarized in Table \ref{tab:3D_times}. We see that the corresponding run times reduce only to $58\%$ and $42\%$ for $M=3$ and $M=4$, respectively, which is still a substantial improvement. The downside of these significant run time savings is the boost of the memory requirements which increase to $174\%$ and $220\%$ (Table \ref{tab:3D_times}). Interestingly, the factor for the run time reduction and the increase factor for memory are quite similar for the same accuracy order $M$. 

\begin{table}[hb]
  \begin{center}
    \begin{tabular}{l|l|l|ll}
          & $M=2$  & $M=3$   & $M=4$      &   \\\hline
    Run time & 4349\,s & 2505\,s & 1833\,s & \\
          & \textbf{100\%}   &\textbf{58\%}  & \textbf{42\%}   & \\\hline
    Time steps & 8133 & 3505 & 2350 & \\
          & \textbf{100\%}   &\textbf{43\%}  & \textbf{29\%}   & \\\hline
    Total memory & 19.9\,GB  &  34.7\,GB& 43.8\,GB  &  \\
          & \textbf{100\%}   & \textbf{174\%} &\textbf{220\%} & \\\hline
    Error & 0.10013\% &0.10014\% & 0.10013\% & \\
    \end{tabular}
   \end{center}
   \caption{\label{tab:3D_times} Comparison of computational demands for 3-D elastic FDTD simulations with the ABS time integrator of accuracy $M$ to achieve the same level of accuracy. The required run time and number of time steps
   decrease with $M$. At the same time the total requirements of memory increase by approximately the same factor due to the required storage of previously calculated wavefields. }
\end{table}
\begin{figure}[ph]
\centerline{\includegraphics[width=\textwidth]{eps/error_3D_courant}}
\caption{Left: Relative error of 3-D elastic simulations. 
The normalized L2 norm is plotted over the required number of time steps ($NT$) and the courant number. The temporal accuracy orders are $M=2,3,4$. The spatial accuracy is fixed at order $N=8$. 
The number of time steps to achieve a given accuracy of $E=0.1\%$ (horizontal line) reduce to $43\%$ and $29\%$ for the temporal orders $M=3$ and $M=4$, respectively. 
For large $NT$ (small time step intervals) the error converges to the fixed error of the spatial discretization. Right: The required run time  of the program as a function of the number of
time steps. The total run time to achieve an accuracy level of $E=0.1\%$ reduces to $58\%$ and $42\%$ for the temporal orders of $M=3$ and $M=4$, respectively.}
\label{fig_L23D}
\end{figure}
Tilman Steinweg's avatar
Tilman Steinweg committed
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913


\newpage

\subsection{Boundary Conditions}\label{bound_cond}
To find to a unique solution of the problem, boundary conditions have to be defined. We can roughly distinguish two types of boundary conditions:

\begin{enumerate}

\item \underline{Dirichlet Boundary Conditions:}

The Dirichlet boundary condition specifies the values a solution needs to take on the boundary of a domain.
The Dirichlet boundary condition could take the form

\EQ{FD:5}{\begin{split}
	v_i(0)&=v_i(NX)=0 \\
	p_{ij}(0)&=p_{ij}(NX)=0
	\end{split}}

The Dirichlet boundary conditions have the disadvantage that it is not possible to model an infinite full space, because the wavefield is reflected at the boundaries of the model grid.

\item \underline{Absorbing Boundary Conditions:}

To overcome this disadvantage an absorbing boundary condition can be applied. A commonly used implementation is described in \cite{cerjan:85}: The numerical grid is enlarged by a few grid points ''$\rm{FW}$'' (typically $\rm{FW=30 gridpoints}$) in each direction. The values of the stress and particle velocity in this boundary frame are multiplied by a factor ''damp'':

\EQ{FD:5:1}{\rm{damp=exp(-a^2x^2)}}

with $\rm{a=\sqrt{-log(amp)/(FW)}}$ and $\rm{amp=0.92}$. The seismic waves are damped inside the boundary frame and cannot be reflected back into the model. This type of absorbing boundary
is not able to damp the reflected waves from the boundary completely.

A more effective way to damp the waves near the boundaries are the {\bf{Perfectly Matched Layers (PMLs)}}. This can be achieved by a coordinate stretch of the wave equations in the frequency domain \cite{komatitsch:07}. The coordinate stretch creates exponentially decaying plane wave solutions in the absorbing boundary frame. The PMLs are only reflectionless if the exact wave equation is solved. As soon as the problem is discretized (for example using finite differences) you are solving an approximate wave equation and the analytical perfection of the PML is no longer valid. To overcome this shortcoming the wavefield is damped by the damping function 

\EQ{FD:6:1}{\rm{c=-V_{pml} \cdot \frac{log(\alpha)}{L}}}

where $V_{pml}$ denotes the typical P-wave velocity of the medium in the absorbing boundary frame, $\alpha=1 \cdot 10^{-4}$ and L is the thickness of the absorbing boundary layer.
A comparison between the exponential damping and the PML boundary is shown in section \ref{comp_EXP_PML}.   
\end{enumerate} 
\clearpage

\section{Numerical Artefacts and Instabilities}\label{num_instab}
To avoid numerical artefacts and instabilities during a FD modeling run, a spatial and temporal sampling condition for the wavefield has to be satisfied. These will be discussed in the following two sections. 

\subsection{Grid Dispersion}\label{grid-dispersion}
The first question when building a FD model is: What is the maximum spatial grid point distance dh, so that the wavefield is correctly sampled? To answer this question we take a look at this simple example: The particle velocity in x-direction is defined by a sine function:

\EQ{grid_disp:1}{v_x=\sin\biggl(2 \pi \frac{x}{\lambda}\biggr),}

where $\lambda$ denotes the wavelength. When calculating the derivation of this function analytically at x=0 and setting $\lambda=1\;m$ we get:

\EQ{grid_disp:2}{\frac{d v_x}{d x}\biggl|_{x=0}=\frac{2 \pi}{\lambda} \cos\biggl(2 \pi \frac{x}{\lambda}\biggr)\biggl|_{x=0}=2 \pi.}

In the next step, the derivation is approximated numerically by a 2nd order finite-difference operator:

\EQ{grid_disp:3}{\frac{d v_x}{d x}\biggl|_{x=0} \approx \frac{v_x(x+\Delta x)-v_x(x)}{\Delta x}\biggl|_{x=0}=\frac{\sin \biggl(\frac{2 \pi \Delta
x}{\lambda} \biggr)}{\Delta x}.}

Using the Nyquist-criteria it should be sufficient to sample the wavefield with $\Delta x = \lambda/2$. In table \ref{grid_disp.1}, the numerical solutions of Equation \ER{grid_disp:3} and the analytical solution \ER{grid_disp:2} are compared for different sample intervals  $\Delta x = \lambda /n$, where n is the number of gridpoints per wavelength. For the case n=2, which corresponds to the Nyquist criteria, the numerical solution is $\frac{d v_x}{d x}|_{x=0}=0.0$, which is not equal with the analytical solution $2 \pi$. A refinement of the spatial sampling of the wavefield results in an improvement of the finite difference solution. For n=16 the numerical solution is accurate to one decimal place. The effect of a sparsly sampled pressure field is illustrated in \FIG{grid_disp_pics}. The dimensions of the FD grid are fixed and the central frequency of the source signal is increased systematically.  When using a spatial sampling of 16 grid points per minimum wavelength (\FIG{grid_disp_pics}, top) the wavefronts are sharply defined. For n=4  grid points a slight numerical dispersion of the wave occurs (\FIG{grid_disp_pics}, center). This effect is obvious when using the Nyquist criteria (n=2)  (\FIG{grid_disp_pics}, bottom). Since the numerical calculated wavefield seem to be dispersive this numerical artefact is called {\bf{grid dispersion}}. To avoid the occurrence of grid dispersion the following criteria for the spatial grid spacing dh has to be satisfied:

\EQ{grid_disp:4}{dh \le \frac{\lambda_{min}}{n} = \frac{V_{s,min}}{n\; f_{max}}.}

Here $\lambda_{min}$ denotes the minimum wavelength, $V_{s,min}$ the minimum S-wave velocity in the model and $f_{max}$ is the maximum frequency of the source signal.Depending on the accuracy of the used FD operator the parameter n is different.  In table \ref{grid_disp.2}, n is listed for different FD operator lengths  and types (Taylor and Holberg operators). For short operators n should be chosen relatively large, so the spatial grid spacing is small, while for longer FD operators n is smaller and the grid spacing can be larger.            

\begin{table}[hbt]
\begin{center}
\begin{tabular}{ccc}\hline \hline
n &  $\Delta x\; [m]$ & $\frac{d v_x}{d x}|_{x=0}\; []$ \\ \hline 
analytical & - & $2\pi \approx 6.283$ \\ 
2 & $\lambda/2$ & 0.0 \\ 
4 & $\lambda/4$ & 4.0 \\ 
8 & $\lambda/8$ & 5.657 \\ 
16 & $\lambda/16$ & 6.123 \\ \hline \hline
\end{tabular}
\caption{\label{grid_disp.1} Comparison of the analytical solution of Equation \ER{grid_disp:2} with the numerical solution (Equation \ER{grid_disp:3}) for different grid spacings $\Delta x = \lambda /n$}
\end{center}
\end{table}
 
\begin{table}[hbt]
\begin{center}
\begin{tabular}{ccc}\hline \hline
FDORDER & n (Taylor) & n (Holberg) \\ \hline 
2nd   &   12       &  12         \\
4th   &   8        &  8.32       \\
6th   &   7        &  4.77       \\
8th   &   6        &  3.69       \\
10th  &   5        &  3.19       \\
12th  &   4        &  2.91       \\   
\hline \hline
\end{tabular}
\caption{\label{grid_disp.2} The number of grid points per minimum wavelength n for different orders (2nd-12th) and types (Taylor and Holberg) of FD operators.}
\end{center}
\end{table} 

\clearpage

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/homogenous_grid_n_16_5.pdf, width=7 cm}\epsfig{file=eps/homogenous_grid_n_16_10.pdf, width=7 cm}
\epsfig{file=eps/homogenous_grid_n_4_5.pdf, width=7 cm}\epsfig{file=eps/homogenous_grid_n_4_10.pdf, width=7 cm}
\epsfig{file=eps/homogenous_grid_n_2_5.pdf, width=7 cm}\epsfig{file=eps/homogenous_grid_n_2_10.pdf, width=7 cm}
\caption{\label{grid_disp_pics} The influence of grid dispersion in FD modeling: Spatial sampling of the wavefield using  16 (top), 4 (center) and 2 gridpoints (bottom).}
\end{center}
\end{figure}
\clearpage

\subsection{The Courant Instability}\label{courandt}
In analogy to the spatial, the temporal discretization has to satisfy a sampling criterion to ensure the stability of the FD code. If a  wave is crossing a discrete grid, then the timestep $dt$ must be less than the time for the wave to travel between two adjacent grid  points with grid spacing $dh$. For a 3-D grid this means mathematically: 

\EQ{courandt:1}{dt \le \frac{dh}{r_{\text{max}} \cdot v_{p,max}}.}

Where $v_{p,max}$ is the maximum P-wave velocity in the model. The factor $r_{\text{max}}$ represents the Courant-Friedrichs-Lewy (CFL) number. In table  \ref{courant.1}, r is listed for different FD operator lengths and types (Taylor and Holberg operators). Criterion \ER{courandt:1} is called {\bf{Courant-Friedrichs-Lewy criterion}}. \FIG{courandt_pics} shows the evolution of the pressure field when the Courant 
criterion is violated. After a few time steps the amplitudes are growing to infinity and the calculation becomes unstable.\\
At the beginning of the simulation the CFL criterium will be checked automatically. 

\begin{table}[hbt]
\begin{center}
\begin{tabular}{c|ccc|ccc}\hline \hline
&  \multicolumn{3}{c|}{Taylor} &  \multicolumn{3}{c}{Holberg}\\
FDORDER & M=2   & M=3 & M=4 & M=2   & M=3 & M=4\\ \hline 
2nd   &   0.577  & 0.494     & 0.384 &0.577&0.494&0.384  \\
4th   &   0.494   &  0.424    &  0.329&0.487&0.418& 0.325 \\
6th   &   0.464  &   0.398   & 0.309   &0.45&0.386&0.3
\\
8th   &   0.448   &  0.384    &  0.299  &0.429&0.368&0.286\\
10th  &   0.438  &  0.375    &  0.292   &0.416&0.357&0.277\\
12th  &   0.431 &  0.369    &  0.287   &0.407&0.349&0.272\\
\hline \hline
\end{tabular}
\caption{\label{courant.1} The factor r is the Courant-Friedrichs-Lewy number for different orders FDORDER (2nd-12th) and types (Taylor and Holberg) of FD operators. M represents the temporal order (FDORDER\_TIME).}
\end{center}
\end{table} 

\clearpage

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/courandt_1.pdf, width=7 cm}\epsfig{file=eps/courandt_2.pdf, width=7 cm}
\epsfig{file=eps/courandt_3.pdf, width=7 cm}\epsfig{file=eps/courandt_4.pdf, width=7 cm}
\caption{\label{courandt_pics} Temporal evolution of the Courant instability.}
\end{center}
\end{figure}
\clearpage

\section{Seismic Modeling with SOFI3D}
The viscoelastic FD algorithm discussed in the previous section is implemented in the C program SOFI3D \cite{bohlen:02}. In the following sections, we give a short description of the different modeling  parameters, options and how the program is used in a parallel MPI environment.

\subsection{Parallel Environment}
The parallelization employs functions of the Message Passing Interface (MPI). MPI has to be installed when compiling and running the SOFI3D software. At least two implementations exist for Unix-based networks: OpenMPI and MPICH2. The LAM-MPI implementation is no longer supported by the developers. Currently all three implementation work with SOFI3D. OpenMPI and MPICH2 are MPI programming environments and development systems for heterogeneous computers on a network. With OpenMPI or MPICH2, a dedicated cluster or an existing network computing  infrastructure can act as one parallel computer solving one problem. The latest version of OpenMPI can be obtained from \url{http://www.open-mpi.org}. MPICH2 is available at \url{http://www-unix.mcs.anl.gov/mpi/mpich}. LAM-MPI can be still downloaded here: \url{http://www.lam-mpi.org}.

\subsection{LAM}
Even though outdated, LAM-MPI will be used to illustrate the setting up of the MPI implementation. In order to reproduce the following instructions, you must first install the LAM-MPI software. 
On SUSE LINUX systems the package can be installed with yast. The software can also be downloaded from \url{http://www.lam-mpi.org}. A good documentation of LAM/MPI is available at this site. Before you can run the SOFI3D software on more than one node you must boot LAM. This requires that you can connect to all of your nodes with ssh without a password. To enable ssh connection without a password you must generate authentication keys on each node using the command ssh-keygen -t rsa. These are saved in the file  \lstinline{$\sim$/.ssh/id_rsa.pub}. Copy all authentication keys into one file named  \lstinline{authorized_keys} and copy this file on all nodes into the directory  \lstinline{$\sim$/.ssh}. 

Before you can boot LAM you must specify the IP addresses of the different processing elements in an ASCII file. An example file is par/lamhosts. Each line must contain one IP address. The first IP number corresponds to node 0, the second line to node 1 and so forth. Note that in older LAM-MPI implementations the mpirun command to run the FD programs must always be executed on node 0 !,
i.e. you must log on node 0 (first line in the file par/lamhosts) and run the software on this node. In the last stable release of LAM-MPI, the node 0 just has to be listed in the lamhosts file (the order does not matter). To boot lam just do  \lstinline{lamboot -v par/lamhosts}. To run the software in serial on a single PC just type  \lstinline{lamboot} without specifying a lamhosts file. You can still specify more PEs in the FD software but all processes will be executed on the same CPU. Thus this only makes sense if you run the software on a multicore system. In this case, you should boot it without
a lamhosts file and specify a total number of 2 processing elements (PEs). To shut down LAM again (before logout) use the command  \lstinline{lamclean -v}. However, it is not necessary to shut down/restart LAM after each logout/login. 

\section{Installation}
\label{installation}
After unpacking the software package (e.g. by  \lstinline{tar -zxvf sofi3D.tgz}) and changing to the directory sofi3D ( \lstinline{cd sofi3D})  you will find different subdirectories:

\textbf{.svn}\\
The software is updated using the version control system subversion (SVN). This directory contains internal information on recent software updates etc. You can ignore this directory, a clean release should not have any .svn directories at all. Each subdirectory described below also contains a subdirectory SVN which can be ignored as well.

\textbf{bin}\\
This directory contains all executable programs, generally sofi3D and snapmerge. These executables are generated using the command  \lstinline{make $<$program$>$} (see below).

\textbf{doc}\\
This directory contains documentation on the software (this users guide) as well as some important papers in PDF format on which the software is based on. It also contains the full documentation of the SOFI3D overnight built testing suite (see \textbf{overnightbuilt}).

\textbf{genmod}\\
Contains the model and benchmark files for SOFI3D.

\textbf{mfiles}\\
Here some Matlab routines (m-files) are stored. These Matlab programs can be used to find optimal relaxation frequencies to approximate a constant Q (qapprox.m) or to plot Q as a function of frequency for
certain relaxation frequencies and value of tau (qplot.m). For further details on the theory behind these algorithms we refer to the Phd thesis of T. Bohlen \cite{bohlen:98} and to the paper in which the so-called tau-method is described \cite{blanch:95}. The parameter tau is used in this software to define the level of attenuation.

\textbf{overnightbuilt}\\
This directory contains a complete bechmarking suite designed to test whether your installed release of SOFI3D works as intended. During the test run, several model scenarios will be simulated and ,afterwards, the created seismogram and snapshot files will be compared with reference values created by the SOFI3D developer team. This benchmarking test can be used as an overnight built check while developing the code, i.e. after modifications to the code, the test evaluates if standard modeling scenarios still result in the same simulation output. For further information we refer to the separate documentation  \lstinline{sofi3D_overnight_documentation} located in the folder \textbf{doc}.

\textbf{par}\\
Parameter files for SOFI3D modeling.

\textbf{scripts}\\
Here, you will find examples of script-files used to submit modeling jobs on cluster-computers. This directory also contains the shell script  \lstinline{convert_silo_new.sh} to convert 3-D snapshot data into the \textit{silo data format} for further visualization using VisIT (see section \ref{visual}). With the c++ code  \lstinline{cut_plane.cxx} a x-y-plane in z-direction can be extrated and written into a separate file for further visualization.   

\textbf{src}\\
This directory contains the complete source codes.  The following programs are available and my be compiled using make $<$program$>$.

\begin{table}[hbt]
\begin{tabular}{ll}
program & purpose \\ \hline
snapmerge & program to merge snapshot files \\
sofi3D & 3-D-Viscoelastic FD code, 2nd-12th order spatial FD operators\\
sofi3D\_acoustic & 3-D-Acoustic FD code, 2nd-12th order spatial FD operators\\
\end{tabular}
\caption{Names of executables located in the folder \textbf{bin} and purpose}
\label{tab_programs}
\end{table}

\cleardoublepage
\section{Example problem: Homogenous block model}\label{simple test}
\subsection{Preliminary Considerations}
In order to illustrate the usage of SOFI3D, we use a homogenous block model adapted from \cite{kang:04} with the dimensions 5400.0 x 5400.0 x 5400.0 m. The material parameters of the block are:  P-wave velocity $v_p=3500.0\; \frac{\mathrm{m}}{\mathrm{s}}$, S-wave velocity $v_s=2000.0\; \frac{\mathrm{m}}{\mathrm{s}}$ and density $\rho=2000.0\; \frac{\mathrm{kg}}{\mathrm{m}^3}$. An explosive source is located inside the model at point  $x_s=2592.0\;$m, $y_s=2106.0\;$m and $z_s=2592.0\;$ km (y denotes the vertical direction). The source function is a Ricker wavelet (Equation \ER{eq_ricker}) with a center frequency $f_c$ of 5 Hz. To calculate the dimensions of the grid the spatial grid spacing is estimated using the grid dispersion criterion \ER{grid_disp:4}. For an 8th order Holberg FD operator using $v_{s,min}=2000\;$m/s, $f_{max}=2 \cdot f_c=10\;$ Hz and $n=3.69$ (Table \ref{grid_disp.2}) we get:

\EQ{prob1:1}{dh \le \frac{V_{s,min}}{n\; f_{max}} = 54.2\;\mathrm{m}.}

The whole model grid has the dimensions 100 x 100 x 100 gridpoints. To avoid a violation of the Courant criterion the time step size DT is calculated according to \ER{courandt:1} using $v_{p-max}=3500\; m/s$, $h=1.345$ (see Table \ref{courant.1}) and $dh=54.0\; m$:

\EQ{prob1:2}{DT \le \frac{dh}{h \sqrt{3} v_{p-max}} = 6.62\times 10^{-3} s \approx 6.6\; \mathrm{ms}.}

The modeling covers a time span of 5.0 s, so about NT=758 time steps are needed. A line of 100 Geophones is placed between $x_{start}=54.0\;$ m and $x_{end}=5400.0\;$ m and $y=2106.0\;$ m, $z=2592.0\;$m. Please note, that due to the small size of the model, more than half of the grid is covered by the absorbing boundary (if ABS=2, see Chapter \ref{abs}). 

\subsection{Coordinate system}
\label{coord_system}
The original 3-D modeling code SOFI3D bases on the 2-D version called SOFI2D (formerly FDVEPS\_MPI). There, a Cartesian coordinate system denoted by the horizontal \textit{X} and the vertical \textit{Y} axis has been introduced. While considering the third dimension and expanding the 2-D code to SOFI3D, this coordinate system had been re-used with an additional axis \textit{Z} which was not the vertical but the second horizontal axis. Consequently, SOFI3D still uses \textit{Y} as the vertical coordinate. All cooordinate related parameters read from the input file (such as DY and DZ, IDY and IDZ, REFREC,...) follow the logic that \textit{Y} denotes the depth (vertical axis). A previous attempt to switch \textit{Y} and \textit{Z} within the input file in order to make \textit{Z} the vertical direction has been reversed. Even though the coordinate system with \textit{Z} as the depth axis was more intuitive, maintaining and developing the code has been more complicated. For this reason, \textit{Y} is still the vertical depth axis!

Please note that if you choose the Seismic Unix output format ( \textbf{SEIS\_FORMAT = 1}) the coordinate orientation written into the header will not follow the SOFI3D coordinate convention as described above. Instead, the header word \textit{gy} denotes the second horizontal axis (equivalent to the \textit{Z} coordinate in SOFI3D) while the header word \textit{gelev} denotes the vertical axis (equivalent to the \textit{Y} coordinate in SOFI3D).

\subsection{Definition of the Wave Modeling Parameters}
\label{modelgeom}
The geometry of the FD grid and all parameters for the wave field simulation have to be defined in a parameter file (which we name in this case { \lstinline{sofi3D/par/in_and_out/sofi3D.json}}). In the following we will first list the full input file and later explain every input parameter section in detail:
\input{sofi3D.json}
All lines in the parameter file are formated according to the JSON standard (\url{http://www.json.org}) and organized as follows: 

\begin{verbatim}
"VARNAME" = "Parameter value",
\end{verbatim}

a comment line can look like this

\begin{verbatim}
"Comment" = "This is a useful comment",
"3D Grid information" = "comment",
\end{verbatim}

where VARNAME denotes the name of the global variable in which the value is saved in all functions of the program. The possible values are described in comments, feel free to add own comments. Basically all non JSON conform line will be ignored. The order of parameters can be arbitrary organized. The built-in JSON parser will search for the need parameters and displays found values. If critical parameters are missing the code will stop and an error message will appear. The meaning of the different parameters is described in the following.

\subsubsection{Domain decomposition}

\begin{verbatim}
"Domain Decomposition" : "comment",
			"NPROCX" : "2",
			"NPROCY" : "2",
			"NPROCZ" : "2",
\end{verbatim}

with

NPROCX : number of processors in x-direction\\
NPROCY : number of processors in y-direction\\
NPROCZ : number of processors in z-direction\\



Parallelization is based on domain decomposition (see Figure \ref{fig_grid}), i.e each processing element (PE) updates the wavefield within his portion of the grid. The model is  decomposed
by the program into sub grids. After decomposition each processing elements (PE) saves only his sub-volume of the grid. NPROCX, NPROCY and NPROCZ specify the number of
processors in x-, y- and z-direction, respectively (Figure  \ref{fig_grid}). The total number of processors thus is NP=NPROCX*NPROCY*NPROCZ. This value must be specified when starting the program with the mpirun command:  \lstinline{mpirun -np $<$NP$>$ ../bin/sofi3D ./in_and_out/sofi3D.json} (see section \ref{compexec1}). If the total number of processors in sofi3D.json
and the command line differ, the program will terminate immediately with a corresponding error message (Try it !). Obviously, the total number of PEs (NPROCX*NPROCY*NPROCZ) used to decompose the model  should be less equal than the total number of CPUs which are available on your parallel machine. If you use LAM and decompose your model in more domains than CPUs are available two or more  domains will be updated on the same CPU (the program will not terminate and will produce the correct results). However, this is only efficient if more than one processor is available on each node. In order to reduce the amount of data that needs to be  exchanged between PEs, you should decompose the model into more or less cubic sub grids. In our example, we use 2 PEs in each direction: NPROCX=NPROCY=NPROCZ=2. The total number of PEs used by the program is NPROC=NPROCX*NPROCY*NPROCZ=8. 
\begin{figure}
\begin{center}
\includegraphics[width=\textwidth,angle=0]{eps/grid.pdf}
\end{center}
\caption{Geometry of the numerical FD grid using 4 processors in x-direction (NPROCX=4) and 2 processors in y-direction (NPROCY=2). Each processing element (PE) is updating the wavefield in his domain.
At the top of the numerical mesh the PEs apply a free surface boundary condition if FREE\_SURF=1, otherwise an absorbing boundary condition (ABS or PML) is applied. The width of the absorbing frame is FW grid points.  The size of the total grid is NX grid points in x-direction, NY gridpoints in y-direction and NZ gridpoints in z-direction. The size of each sub-grid  thus is NX/NPROCX x NY/NPROCY x NZ/NPROCZ gridpoints. The source is located at (XSRC,YSRC,ZSRC). The first receiver is at (XREC1,YREC1,ZREC1) and the  last receiver at (XREC2,YREC2,ZREC2). The origin of the Cartesian coordinate system (x,y,z) is at the top left corner of the grid. }
\label{fig_grid}
\end{figure}
\subsubsection{Discretization}
\begin{verbatim}
"3-D Grid" : "comment",
			"NX" : "100",
			"NY" : "100",
			"NZ" : "100",
			"DX" : "54.0",
			"DY" : "54.0",
			"DZ" : "54.0",
\end{verbatim}

with

NX : number of gridpoints in x-direction\\
NY : number of gridpoints in y-direction\\
NZ : number of gridpoints in z-direction\\
DX : distance between gridpoints(in m) in x-direction\\
DY : distance between gridpoints(in m) in y-direction\\
DZ : distance between gridpoints(in m) in z-direction\\
Caution: first gridpoint is not at {0.0,0.0,0.0} but at {dx,dy,dz}!\\
Note that Y denotes the vertical direction !\\


These lines specify the size of the total numerical grid (Figure  \ref{fig_grid}). NX, NY and NZ give the number of grid points in the x-, y- and z-direction, respectively, and DX, DY and DZ
specify the grid spacing in x-, y- and z-direction, respectively. The grid spacing may thus vary in the three directions to allow for a detailed discretization of small borehole tool features.
The size of the total internal grid in meters in x-direction is NX*DX, in y-direction NY*DY  and in z-direction NZ*DZ. To allow for a consistent domain decomposition NX/NPROCX, NY/NPROCY and NZ/NPROCZ must be integer values.

To avoid numerical dispersion the wavefield must be discretized with a certain number of gridpoints per wavelength. The number of gridpoints per wavelength required, depends on the order of the spatial
FD operators used in the simulation (see section \ref{grid-dispersion}). In the current FD software, 2nd, 4th, 6th, 8th, 10th and 12th order operators are implemented. The criterion to avoid numerical dispersion reads:
\begin{equation}
max(DX,DY,DZ)\le\frac{v_{s,min}}{2 f_c n} \label{eq_dispersion}
\end{equation}
where $\frac{v_{s,min}}{2 f_c}$ is the smallest wavelength propagating through the model. $v_{s,min}$ denotes the minimum shear wave velocity in the model, and $f_c=1/TS$ is the center frequency of the source wavelet. The program assumes that the maximum frequency of the source signal is approximately two times the center frequency. The center frequency is approximately one over the duration time TS. The value of n for different FD operators is tabulated in table \ref{grid_disp.2}. The criterion \ref{eq_dispersion} is checked by the FD software. If the criterion is violated a warning message will be displayed in the SOFI3D output (default :  \lstinline{./in_and_out/sofi3D.jout}), section ``--- CHECK FOR GRID DISPERSION ---``. Please note, that the FD-code will NOT terminate due to grid dispersion, only a warning is given in the output file.

\subsubsection{Order of the FD operator}

\begin{verbatim}
"FD order" : "comment",
			"FDORDER" : "8",
			"FDORDER_TIME" : "2",
			"FDCOEFF" : "2",
\end{verbatim}

The order of the used spatial FD operator is defined by the option FDORDER. The possible values are 2, 4, 6, 8, 10, 12. 
The variable FDORDER\_TIME represents the used temporal FD order. FDORDER\_TIME=2 correspondents to the classical leapfrog scheme. Available higher order temporal FD operators are 3 and 4. Higher temporal orders will increase memory usage significant at least by a factor of 2.2. With the option FDCOEFF the user can switch between Taylor (FDCOEFF=1) and Holberg (FDCOEFF=2) FD coefficients. The chosen FD operator and FD coefficients have an influence on the numerical stability and grid dispersion (see section \ref{grid-dispersion}).


\subsubsection{Time stepping}
\begin{verbatim}
"Time Stepping" : "comment",
			"TIME" : "5.0",
			"DT" : "6.6e-3",
\end{verbatim}

with

TIME : time of wave propagation\\
DT : timestep (in seconds)\\


The propagation time of seismic waves in the entire model is TIME. The time stepping interval (DT) has to fulfill the stability criterion \ER{courandt:1} in section \ref{courandt}. 
The program checks these criteria for the entire model, outputs a warning message if these are violated , stops the program and will output the time step interval for a stable model run. 


\subsubsection{Sources}
\begin{verbatim}
			"SOURCE_SHAPE" : "1",
			"source shape values: ricker=1;fumue=2;from_SIGNAL_FILE=3;SIN**3=4" : "comment",
			"SIGNAL_FILE" : "signal_mseis.tz",

			"SOURCE_TYPE" : "1",
			"source_type values (point_source):" : "comment",
			"explosive=1;force_in_x=2;in_y=3;in_z=4;custom=5;double_couple=6" : "comment",
			"SOURCE_ALPHA, SOURCE_BETA" : "0.0 , 0.0",
			"AMON, STR, DIP, RAKE" : "1.0e2 , 45.0 , 90.0 , 45.0",
			"SRCREC" : "1",
			"srcrec values :  read from SOURCE_FILE=1, PLANE_WAVE=2 (internal)" : "comment"
			
			"SOURCE_FILE" : "./sources/sources.dat", 
			"RUN_MULTIPLE_SHOTS" : "0", 
			
			"PLANE_WAVE_DEPTH" : "2106.0",
			"PLANE_WAVE_ANGLE" : "0.0",
			"TS" : "0.2",
\end{verbatim}

with

SOURCE\_SHAPE : Shape of source-signal (ricker=1, fumue=2, from SIGNAL\_FILE=3, SIN$^3$=4) \\
SOURCE\_TYPE : point source (explosive=1, force in x=2, in y=3, in z=4, custom=5, double couple=6) \\
If SOURCE\_TYPE $<$ 5 the SOURCE\_ALPHA and SOURCE\_BETA are ignored \\
SOURCE\_ALPHA : force angle between X and Z coordinate (in degree)\\
SOURCE\_BETA : force angle between:: Y and Z coordinate (in degree)\\
if SOURCE\_TYPE $\ne$ 6 the AMON, STR, DIP and RAKE are ignored\\
AMON: Seismic Moment of the earthquake (in newton meter)\\
STR: strike angle of the earthquake fault plane (in degree)\\
DIP: dip angle of the earthquake fault plane (in degree)\\
RAKE: rake angle of the slip vector (in degree)\\
PLANE\_WAVE\_DEPTH : depth of plane wave excitation (no$<$=0) (in meter)\\
if PLANE\_WAVE\_DEPTH$>$0, SRCREC is treated as 0\\
if PLANE\_WAVE\_DEPTH = 0 the PLANE\_WAVE\_ANGLE and TS  are ignored\\
PLANE\_WAVE\_ANGLE : dip of plane wave from vertical (in degrees)\\
TS : duration of source-signal (in seconds)\\
SIGNAL\_FILE : external signal file name \\
SRCREC : read source positions from SOURCE\_FILE (yes=1)\\
SOURCE\_FILE : external source file name \\
RUN\_MULTIPLE\_SHOTS : run multiple shots defined in SOURCE\_FILE (yes=1)\\
Note that Y denotes the vertical direction!\\


The following source types are availabe: explosive sources that excite compressional waves
only (SOURCE\_TYPE=1), point forces in x-, y-, and z-direction (SOURCE\_TYPE=2,3,4), point forces in an arbitrary direction (SOURCE\_TYPE=5) and a double couple source (SOURCE\_TYPE=6).

When using the earthquake double couple source, the orientation of the slip vector on the fault plane is defined by the angles STR, DIP and RAKE. For their definition see Figure \ref{Fig:EQS}. The earthquake source is implemented via a moment-tensor description \cite{graves:96}, while the moment-tensor components are derived from the strike, dip and rake angles \cite{aki:80}. The radiated energy of the earthquake is defined by the seismic moment, which is denoted by AMON. When using the earthquake double couple source output seismograms are in units of meters per second.

\tdplotsetmaincoords{0}{0}

\begin{figure}[ht]
\centering
\begin{tikzpicture}[scale=4]
\draw[thick,-stealth] (0,0,0) -- (1,0,0) node[anchor=west]{\textsf{\textbf{x East}}};
\draw[thick,-stealth] (0,0,0) -- (0,-1,0) node[anchor=west]{\textsf{\textbf{y Depth}}};
\draw[thick,-stealth] (0,0,0) -- (0,0,-1) node[anchor=south west]{\textsf{\textbf{z North}}};

\draw[-stealth,color=red] (0.35,-0.7,-0.75) -- (0.55,-0.35,-0.75);
\draw[dashed, color=red] (0,0,0) -- (0.1,-0.7,-0.5);
\draw[dashed, color=red] (0,0,0) -- (0.5,0,-0.5);
\draw[dashed, color=red] (0.5,0,-0.5) -- (0.6,-0.7,-1.0);
\draw[dashed, color=red] (0.1,-0.7,-0.5) -- (0.6,-0.7,-1.0);

\tdplotdrawarc{(0,0,0)}{0.4}{15}{45}{anchor=west}{\textsf{\textbf{STR}}};%$\phi$}
\tdplotdrawarc{(0.35,-0.7,-0.75)}{0.65}{33.5}{50.5}{anchor=west}{\textsf{\textbf{RAKE}}};%$\Lambda$}

\draw[dashed, color=black] (0.8,0,0.8) -- (-0.2,0,-0.2);
\draw[bend left=90] (0.1,-0.7,-0.5) -- (0.5,0,0.5) node[anchor=north west]{\textsf{\textbf{DIP}}};%$\delta$};
% \draw[thick,-stealth] (0.1,-0.69,-0.5) -- (0.1,-0.7,-0.5);

\end{tikzpicture}
\caption{The orientation of a simplified earthquake rupture plane is described by the strike angle STR, which is measured clockwise from north to the fault line, and the angle DIP, which denotes the angle between fault plane and the horizontal plane. The slip direction on the fault plane is denoted by the angle RAKE, which is measured counterclockwise from the horizontal fault line.}
\label{Fig:EQS}
\end{figure}

Three built-in wavelets of the seismic source are available. The corresponding time functions are defined in src/wavelet.c. You may modify the time functions in this file and recompile to include your
own analytical wavelet or to modify the shape of the built-in wavelets.

Ricker wavelet:
\begin{equation}
r(\tau)=\left(1-2\tau^2\right)\exp(-\tau^2) \quad \mbox{with} \quad \tau=\frac{\pi(t-1.5/f_c-t_d)}{1.0/f_c}
\label{eq_ricker}
\end{equation}

Fuchs-M\"uller wavelet:
\begin{equation}
f_m(t)=\sin(2\pi(t-t_d)f_c)-0.5\sin(4\pi(t-t_d)f_c) \quad \mbox{if} \quad t\in[t_d,t_d+1/fc] \quad \mbox{else} \quad fm(t)=0
\label{eq_fm}
\end{equation}

$sin^3$ wavelet:
\begin{equation}
s3(t)=0.75 \pi f_c \sin(\pi(t+t_d)f_c)^3\quad \mbox{if} \quad t \in[t_d,t_d+1/fc] \quad \mbox{else} \quad s3(t)=0
\label{eq_s3}
\end{equation}

In these equations, t denotes time and $f_c=1/TS$ is the center frequency. $t_d$ is a time delay which can be defined for each source position in SOURCE\_FILE. Variable time delays for different sources locations allow the simulation of passive acoustic emission. Note that the symmetric (zero phase) Ricker signal is always delayed by $1.0/f_c$, which means that after
one period the maximum amplitude is excited at the source location.  These three source wavelets and the corresponding amplitude spectra for a center frequency of $f_c=50$ Hz and
a delay of $t_d=0$ are plotted in Figure \ref{fig_source_wavelets}. Note the delay of the Ricker signal described above. The Fuchs-M\"uller wavelet has a slightly higher center frequency and covers a broader frequency range.

\begin{figure}
\begin{center}
\includegraphics[width=8cm,angle=0]{eps/signals.pdf}
\end{center}
\caption{Plot of built-in source wavelets (equations \ref{eq_ricker}, \ref{eq_fm}, \ref{eq_s3}) for a center frequency of $f_c=50$ Hz 
($TS=1/f_c=0.02$s): Ricker signal (solid), Fuchs-M\"uller signal (dashed), $sin^3$-signal (dotted). a) Time function, b) amplitude
spectrum, c) phase spectrum.  }
\label{fig_source_wavelets}
\end{figure}

You may also use your own digitized time function as the source wavelet (for instance the signal of the first arrival recorded by a geophone at near offsets). Specify SOURCE\_SHAPE=3 and save the samples of
your source wavelet in ASCII-format in SIGNAL\_FILE. SIGNAL\_FILE should contain one sample per line. It should thus look like:

\begin{verbatim}
0.0
0.01
0.03
...
\end{verbatim}

The time interval between the samples must equal the time step interval (DT) of the FD simulation (see above) ! It it therefore mostly necessary to resample/interpolate a given source time function with a smaller sample rate. You may use the matlab script mfiles/resamp.m to resample your external source signal to the required sampling interval.

The following source types are availabe: explosive sources that excite compressional waves only (SOURCE\_TYPE=1), and point forces in the x-, y-, and z-direction (SOURCE\_TYPE=2,3,4).
The force sources excite both P- and S-waves. The explosive source is located at the same position as the diagonal elements of the stress tensor, i.e. at (i+1/2, j+1/2, k+1/2) (Figure \ref{fig_cell}).
The forces are located at the same position as the corresponding components of particle velocity (Figure \ref{fig_cell}). If (x,y,z) denotes the position at which the source location is defined in source.dat, then the actual force in x-direction is located at (x+DX/2,y,z), the actual force in y-direction is located at (x,y+DY/2,z) and the actual force in z-direction is at (x,y,z+DZ/2). 
By choosing SOURCE\_TYPE=5 a custom directive force can be defined by the force angle between horizontal x- and z-coordinate (SOURCE\_ALPHA) and the force angle between horizontal x- and vertical y-coordinate (SOURCE\_BETA). The orientation of the custom force source angles is plotted in Figure \ref{coordinate_system_customforcesource}. The unit of both SOURCE\_ALPHA and SOURCE\_BETA is in degree from 0 to 360. Please note, that due to numerical imprecision of the calculation of the values of the sine or cosine function, there is a small difference in the seismogram output when considering, e.g., the vertical y-force source (SOURCE\_TYPE=3) and the equivalent custom force source (SOURCE\_TYPE=5, SOURCE\_ALPHA=0, SOURCE\_BETA=90). Usually, this difference is orders of magnitude smaller than the actual seismogram.

\begin{figure}
\begin{center}
\includegraphics[width=9cm,angle=0]{eps/coordinate_system_customforcesource.pdf}
\end{center}
\caption{Scheme for SOURCE\_TYPE=5. The parameter SOURCE\_ALPHA denotes the angle between the horizontal x- and z-coordinate, while SOURCE\_BETA denotes the angle between horizontal x- and vertical y-coordinate. As an example, parameters for SOURCE\_ALPHA and SOURCE\_BETA are given in order to gain force sources in single coordinate direction (as specified by SOURCE\_TYPE=2-4)}
\label{coordinate_system_customforcesource}
\end{figure}


The locations of multiple sources must be defined in an external ASCII file (SOURCE\_FILE) that has the following format:
\begin{verbatim}
 % 	XSRC		YSRC		ZSRC		TD		FC		AMP  
\end{verbatim}

More than one source position may be specified for example to simulate acoustic emission. In the following lines, you can define certain parameters for each source point:
XSRC is the x-coordinate of a source point (in meter), YSRC is the vertical y-coordinate of a source point (in meter), ZSRC is the horizontal z-coordinate of a source point (in meter).
TD is the excitation time (time-delay) for the source point [in seconds], FC is the center frequency of the source signal [in Hz], and AMP is the maximum amplitude of the source signal.

Please note that SOFI3D considers real physical units as input and output. AMP specifies a seismic moment of 1 Nm if an explosive source is chosen, in case of force source AMP specifies a force of 1 N. The seismogram output scales to this seismic moment or force and is given in particle velocities or pressure. Only the component \textit{div} and \textit{curl} cannot be treated as physical measures. For more information on the seismogram output see section \subsubsection{Receivers}.

The  SOURCE\_FILE =  \lstinline{./sources/source.dat} that defines an explosive source  at $x_s=2610.0\;$ m, $y_s=2610.0\;$ m and $z_s=2100.0\;$ m (depth) with
a center frequency of 5 Hz (no time delay) reads, for example
\begin{verbatim}
	2592.0        2106.0        2106.0        0.0           5.0           1.0
\end{verbatim}

If the option RUN\_MULTIPLE\_SHOTS=0 in the parameter file all the shot points defined in the SOURCE\_FILE are used in one simulation. Setting RUN\_MULTIPLE\_SHOTS=1 will start model runs 
from i=1 to i=NSRC with source locations and properties defined at line i+1 of the SOURCE\_FILE. 

Instead of a single source or multiple sources specified in the SOURCE\_FILE, you can also specify to excite a plane wave parallel (or tilted by an angle PLANE\_WAVE\_ANGLE) to the top of the model. This plane wave is approximated by a plane of single sources at every grid point at a depth of PLANE\_WAVE\_DEPTH below. The center source frequency $f_c$ is specified by the inverse of the duration of the source signal TS. SOURCE\_SHAPE and SOURCE\_TYPE are taken from the parameters as described above. If you choose the plane wave option by specifying a PLANE\_WAVE\_DEPTH$>$0, the parameters SRCREC and SOURCE\_FILE will be ignored.

\subsubsection{Generation of models}
\label{gen_of_mod}
\begin{verbatim}
"Model" : "comment",
			"READMOD" : "0",
			"MFILE" : "model/test",
			"WRITE_MODELFILES" : "0",

\end{verbatim}

READMOD : read model parameters from MFILE (yes=1)\\
MFILE : String for the model file names, used to read from file and/or to write model for validation \\
WRITE\_MODELFILES : switch to decide whether all models RHO, U, Pi (and Taup, Taus) models (=1) or only the density model (=2) are written to file. In case of WRITE\_MODELFILES=0, no model file is written to file.

If READMOD=1, the P-wave, S-wave, and density model grids are read from external binary files. MFILE defines the basic file name that is expanded by the following extensions: P-wave model: ''.vp'', S-wave model: ''.vs'', density model: ''.rho''.  In the example above, the model files thus are: ''model/test.vp'' (P-wave velocity model),''model/test.vs'' (S-wave velocity model), and ''model/test.rho'' (density model). 

In these files, each material parameter value must be saved as 32 bit (4 byte) native float. Velocities must be in meter/second, density values in $kg/m^3$. The fast dimension is the y direction. See src/readmod.c. The number of samples for the entire model in the x-direction is NX, the number of values in the y-direction is always NY  and the number of values in the z-direction is always NZ. The file size of each model file thus must be NX*NY*NZ*4 bytes. You may check the model structure using the SU command ximage:

ximage n1=$<$NX$>$ n2=$<$NY$>$ $<$ model/test.vp .

If the variable TAU = 0.0 (see the next section \textit{Q-approximation}), it is also possible to read Qp, and Qs grid files to allow for spatial variable attenuation. In case of TAU > 0, constant damping is assumed with $Q_p=Q_s = \frac{2}{\mbox{TAU}}$. Please not that in order to ensure proper visualization using xmovie or ximage, the binary models are structured in terms of coordinate priority as follows : [Y,X,Z]. That means a 3-D model consits of NZ planes of the size NY $\times$ NX. The fast dimension of each NY-NX plane is the vertical Y-direction, i.e., first, columns of NY material parameters are written to file, in total NX columns of the length NY are written. To give an example of how to create a 3-D model using Matlab, the M-File  \lstinline{create_simple_tunnel_mod3D.m} is located in the folder  \lstinline{mfiles} (see Chapter \ref{installation}).

If READMOD=0 the model is generated ''on the fly'' by SOFI3D, i.e. it is generated internally before the time loop starts. We would prefer this way of generating the model grids because this way the size and the discretization intervals (DX, DY and DZ) can be controlled by the parameter file sofi3D.json. Furthermore, it is not necessary to generate (large) files by an external program. See section \ref{model_def_func} for an example function that generates the simple block model ''on the fly''. If READMOD=0 this function is called in sofi3D.c and therefore must be specified in src/Makefile (at the top of src/Makefile, see section \ref{compexec}). If you change this file, for example to change the model structure, you need to re-compile SOFI3D by changing to the src directory and ''make sofi3D''.

Due to the fact that the variable MFILE is also used to write the model - expanded by the extension-string ''SOFI3D'' to file for verification (also see the output file, section ''MODEL CREATION AND OUTPUT''), you can specify, whether only the density model is written (WRITE\_MODELFILES=0) or the Vp-velocity, Vs-velocity, density as well as Qp-damping and Qs-damping models (WRITE\_MODELFILES=1). BE AWARE that the output of additional models besides density cause extra but temporal memory allocation of the size of the local subgrid times the number of models! To give an example, the file name of the density models written for verification looks like this  \lstinline{model/test.SOFI3D.rho}.

\subsubsection{Q-approximation}
\begin{verbatim}
"Q-approximation" : "comment",
			"L" : "0",
			"FL1" : "5.0", 
			"FREF" : "5.0",
			"TAU" : "0.00001",
\end{verbatim}

with

L : Number of relaxation mechanisms\\
FL1 : First relaxation frequency \\
FREF : Reference frequency where no velocity dispersion occurs \\
TAU : TAU value for entire model\\

These lines may be used to define an overall level of intrinsic (viscoelastic) attenuation of seismic waves, i.e. a constant (spatially not varying) TAU = 2/Q for the complete model. The level of attenuation is defined by Q. Note that small Q values ($Q<50$) may lead to significant amplitude decay and velocity dispersion. The relaxation frequency (FL1) can be specified to approximate the constant Q for all frequencies. The frequency dependence on attenuation, i.e., Q and phase velocity as a function  of frequency, may be calculated using the Matlab functions in the directory mfiles.
In the current version, only one (so-called) relaxation mechanism (L=1) is implemented \cite{bohlen:98,blanch:95,bohlen:02}. The possible values for L thus are L=1 and L=0. In case of L=0, a purely elastic simulation is performed (no absorption). In the elastic case (L=0), the memory requirements and run time of the program are reduced by approximately 30 per cent. The parameter FREF determines the frequency where no velocity dispersion occurs. Generally, FREF equals the largest center frequency FC of sources defined in SOURCE\_FILE. In fact, if FREF is not specified or set to 0.0, FC is used instead.

If you set TAU=0.0, it is further possible to simulate any spatial distribution of absorption by assigning the gridpoints with different Q-values by reading external grid files for Qp (P-waves) and Qs (S-waves) (see src/readmod.c) or by generating these files ''on the fly'' (see section \ref{model_def_func}). If TAU$>$0.0 no external Qp and Qs-file will be read in, instead TAU will be used for a constant Q-model.

Please note, that due to dispersive media properties the viscoelastic velocity model is only defined for the reference frequency only. In sofi3D, this reference frequency is specified as the center source frequency. At the exact reference frequency, elastic and viscoelastic models are equivalent. As a consequence, slightly smaller and larger minimum and maximum velocity values occur in the viscoelastic model.


\subsubsection{Boundary conditions}
\label{abs}

\begin{verbatim}
"Boundary Conditions" : "comment",
			"FREE_SURF" : "1", 
			"ABS_TYPE" : "2",
			"FW" : "30.0",
			"DAMPING" : "8.0",
			"FPML" : "5.0",
			"VPPML" : "3500.0",
			"NPOWER" : "2.0",
			"K_MAX_CPML" : "10.0",
			"BOUNDARY" : "0",
\end{verbatim}

with

FREE\_SURF : free surface (yes=1)\\
ABS\_TYPE : type of boundary condition (PML=1/ABS=2)\\
FW : width of boundary frame (in grid points) (No$<$=0)\\
DAMPING : parameter for absorbing boundary (exponential damping), percentage of amplitude decay at outer edge\\
FPML : parameter for PML, Dominant frequency (in Hz)\\
VPPML : p wave velocity near the grid boundary (in m/s)\\
NPOWER :  power to compute d0 profile, i.e. exponent used in CMPL frame update, the larger the more damping \\
K\_MAX\_CPML :  Coefficient used for the CMPL frame update \\
BOUNDARY : apply periodic boundary condition at edges. Periodic boundary (no=0,left/right/ front/back=1) \\
%A : thickness of the layer between aquisition geometry and PML \\

The boundary conditions are applied on each side face and the bottom face of the model grid. A plane stress free surface is applied at the top of the global grid if FREE\_SURF $\neq$ 0 using the imaging method proposed by Levander \cite{levander:88}. If FREE\_SURF = 0 the boundary conditions are also applied at the top face of the model grid. Note that the absorbing frames are always located INSIDE the model space, i.e. parts of the model structure are covered by the absorbing frame, in which no physically meaningful wavefield propagates. You should therefore consider the frame width when you design the model structure and the acquisition geometry (shot and receivers should certainly be placed outside).

Two different boundary conditions are available: an unsplit convolutional perfectly matched layer (cPML) boundary condition (in the elastic case) and split perfectly matched layer (PML) boundary condition (in the acoustic case), respectively, and a conventional exponential wavefield damping condition (ABS) (see section \ref{bound_cond} and section \ref{comp_EXP_PML}). The PML implementation is based on the following papers \cite{collino:01}, \cite{komatitsch:07} and \cite{martin:09}. It should be much more effective than the ABS condition. If ABS\_TYPE=1 (PML) a width of the absorbing frame of FW=10-20 grid points should be sufficient. In the case of ABS\_TYPE=2 (ABS) exponential wavefield damping according to \cite{cerjan:85} is applied. For this case, you should increase the width of the frame to 30-50 grid points. DAMPING specifies attenuation in percent at the edges of the numerical grid, i.e. amplitudes are multiplied by a factor of 1-DAMPING at the outer edges. A good choice for DAMPING is 8.0 per cent. For larger values, reflections at the onset of the absorbing frame might occur. In order to avoid such an impedance contrast between the model domain and the absorbing boundary layer, you may as well increase FW and decrease DAMPING.

For the optimal realization of the CPML boundary condition you have to specify the dominant signal frequency FMPL occurring during the wave simulation. This is usually the center source frequency FC specified in the source file. VPPML should be approximately the P-wave velocity near the model boundaries. For the definition of the option VPPML, have a look at $V_p$ in equation \ER{FD:6:1}. If the wave incident angle is not perpendicular to CPML frame, K\_MAX\_CPML can be increased in order to stabilize the CPML update, otherwise instabilities might occure.

 In some cases, it is usefull to apply periodic boundary conditions (see section \ref{bound_cond}). IF BOUNDARY=1 no absorbing boundaries are installed at the left/right and front/back sides of the grid. Instead, wavefield information is copied from left to right and vice versa, as well as from front to back and vice versa. The effect is, for example, that a wave which leaves the model at the left side enters the model again at the right side.

\subsubsection{Wavefield snapshots}
\begin{verbatim}
"Snapshots" : "comment",
			"SNAP" : "4",
			"TSNAP1" : "6.6e-3",
			"TSNAP2" : "4.8",
			"TSNAPINC" : "0.2",
			"IDX" : "1",
			"IDY" : "1",
			"IDZ" : "1",
			"SNAP_FORMAT" : "3",
			"SNAP_FILE" : "./snap/test",
			"SNAP_PLANE" : "1",
\end{verbatim}

with

SNAP : output of snapshots (yes$>$0) (particle-velocities=1, pressure field=2, curl and divergence energy=3, particle velocities and energy=4)\\
TSNAP1 : first snapshot (in sec)\\
TSNAP2 : last snapshot (in sec)\\
TSNAPINC : increment (in sec)\\
Note that Y denotes the vertical direction!\\
IDX : increment x-direction (in grid points)\\
IDY : increment y-direction (in grid points)\\
IDZ : increment z-direction (in grid points)\\
SNAP\_FORMAT : data-format (ASCII(2);BINARY(3))\\
SNAP\_FILE : basic filename, the output will look like SNAP\_FILE.bin.z.000, if SNAP = 1,2 SNAP\_PLANE is ignored\\
SNAP\_PLANE : output of snapshots as energy (without sign=1, with sign true for x-z-plane=2, with sign true for x-y-plane=3, with sign true for y-z-plane=4)\\


If SNAP$>0$, wavefield information (particle velocities, pressure, or curl and divergence of particle velocities) for the entire model is saved on the hard disk (assure that enough free space is on disk!). Each PE is writing his sub-volume to disk. The filenames have the basic filename SNAP\_FILE plus an extension that indicates the PE number in the logical processor array (see Figure \ref{fig_grid}), i.e. the PE with number PEno writes his wavefield to SNAPFILE.PEno. The first snapshot is written at TSNAP1 seconds of seismic wave traveltime to the output files, the second at TSNAP1+TSNAPINC seconds etc. The last snapshots contains wavefield at TSNAP2 seconds. Note that the file sizes increase during the simulation. The snapshot files might become quite LARGE. It may therefore be necessary to reduce the amount of snapshot data by increasing IDX, IDY and IDZ and/or TSNAPINC. A detailed description how to visualize 3-D wavefields is given in section \ref{visual}. In order to merge the separate snapshot of each PE after the comletion of the wave modeling, you can use the program snapmerge (see Chapter \ref{installation}, section \textbf{src}). The bash command line to merge the snapshot files can look like this:  \lstinline{../bin/snapmerge ./in_and_out/sofi3D.json}.

\subsubsection{Receivers}
\begin{verbatim}
"Receiver" : "comment",
			"SEISMO" : "4",
			"READREC" : "0",
			"REC_FILE" : "./receiver/receiver.dat",
			"REFRECX, REFRECY, REFRECZ" : "0.0 , 0.0 , 0.0",
			"XREC1,YREC1, ZREC1" : "54.0 , 2106.0, 2592.0",
			"XREC2,YREC2, ZREC2" : "5400.0 , 2106.0, 2592.0",
			"NGEOPH" : "1",
\end{verbatim}

with 

SEISMO : output of seismograms (no seismograms=0, particle-velocities=1, pressure (hydrophones)=2, curl and div=3, everything=4)\\
READREC : read receiver positions from file (yes=1)\\
REC\_FILE : external receiver file name\\
REFREC : reference point for receiver coordinate system\\
if READREC=1 the REC1,YREC1,ZREC1, XREC2,YREC2,ZREC2 and NGEOPH are ignored\\ 
REC1,YREC1,ZREC1 : position of first receiver (in m)\\
XREC2,YREC2,ZREC2 : position of last receiver (in m)\\
NGEOPH : distance between two adjacent receivers (in gridpoints)\\


If SEISMO$>$0, seismograms are saved on hard disk. If SEISMO equals 1 x-, y-, and z-component of particle velocity will be written according to parameters specified in Chapter \ref{seismograms}.
If SEISMO==2 pressure (sum of the diagonal components of the stress tensor) recorded at the receiver locations (receivers are hydrophones!) is written. if SEISMO=3 the curl and divergence are saved. As described in section \subsubsection{Sources} the seismogram output is physically correct for a seismic moment of 1 Nm (explosive source) and a force of 1 N (force source), respectively. A validation of SOFI3D and analytical data can be found in section \ref{analytic}.

The curl and divergence of the particle velocities are useful to separate between P- and S-waves in the snapshots of the wavefield. SOFI3D calculates the divergence and the magnitude of the curl of the particle velocity field \cite{dougherty:88}. The motivation for this is as follows. According to Morse and Feshbach \shortcite{morse:53} the energy of P- and S-wave particle velocities is, respectively,
\begin{equation}
E_p=\left(\lambda + 2 \mu\right) (div(\vec{v}))^2 \quad \mbox{and} \quad E_s=\mu \left|rot(\vec{v})\right|^2 \quad\mbox{.}
\label{eq_E}
\end{equation}
$\lambda$ and $\mu$ are the Lam\`{e} parameters, and $\vec{v}$ is the particle velocity vector. In order to preserve the divergence and curl sign information  while showing relative compressional
and shear particle velocity amplitudes, we plot the following quantities:
\begin{equation}
\bar{E}_p=sign(div \vec{v}) E_p^{1/2} \quad \mbox{and} \quad \bar{E}_s= sign(rot\vec{v}) E_s^{1/2} \quad\mbox{,}
\label{eq_e}
\end{equation}
The magnitudes of $\bar{E}_p$ and $\bar{E}_s$ are proportional to the magnitudes of the P- and S- particle velocities, respectively. Note that interface waves like Rayleigh waves contain both a P- and S-wave component and therefore show up on both quantities of equation \ref{eq_e}.

The locations of the receivers may either be specified in a separate file REC\_FILE or in this parameter file. If READREC=1 receiver locations are read from the ASCII-file REC\_FILE. Each line contains the coordinates of one receiver, the first two number in each line indicate the horizontal x-coordinate, the vertical y-coordinate and the third number denotes the horizontal z-coordinate of each receiver position. To give an example of a receiver file, the following 3 lines specify 3 receivers located at constant depth (2106.0 m) and horizontal y-coordinate (2592.0 m). However, the receiver coordinates change in x-direction (starting at 540 m) and therefore lining up along the x-axis.  
\begin{verbatim}
540.0   2106.0   2592.0   
1080.0  2106.0   2592.0
1620.0  2106.0   2592.0
\end{verbatim}
These receiver coordinates in REC\_FILE are shifted by REFREC[1], REFREC[2] and REFREC[3] into the  x-, y- and z-direction, respectively. This allows for completely moving the receiver spread without modifying REC\_FILE. This may be useful for the simulation of moving profiles in reflection seismics.

If READREC=0 the receiver locations must be specified in the parameter file. In this case, it is assumed that the receivers are located along a straight line. The first receiver position is defined by (XREC1, YREC1, ZREC1), and the last receiver position by (XREC1, YREC1, ZREC1) (see Figure \ref{fig_grid}). The spacing between receivers is NGEOPH grid points.  A Vertical Seismic Profile (VSP)
is realized when XREC1=XREC2, YREC1$<$YREC2.  

Receivers are always located on full grid indices, i.e. a receiver that is located between two grid points will be shifted by the FD program to the closest next grid point. It is not yet possible
to output seismograms for arbitrary receiver locations since this would require a certain wavefield interpolation.

\textbf{It is important to note that the actual receiver positions defined in REC\_FILE or in sofi3D.json may vary by DX/2 and/or DY/2 and/or DZ/2 due to the staggered positions of the particle velocities and stress tensor components. }

In our example, we specify 100 receiver location. Due to the small size of the model, most of the specified receiver positions will be located inside this absorbing boundary (if ABS=2, see Chapter \ref{abs}). A corresponding warning message will appear. If you choose to read the receiver location from REC\_FILE  \lstinline{./receiver/receiver.dat} (READREC=1), only 10 receivers locations are considered. The list of receivers specified in file  \lstinline{./receiver/receiver.dat} is equivalent to the parameters in the input file with only a larger distance between adjacent receivers (NGEOPH = 10.)

\subsubsection{Receiver array}
\begin{verbatim}
"Receiver array" : "comment",

			"REC_ARRAY" : "0",
			"REC_ARRAY_DEPTH" : "1350.0",
			"REC_ARRAY_DIST" : "640.0", 
			"DRX" : "2",
			"DRZ" : "2",
\end{verbatim}

A horizontal 2-D array of receivers is simulated if REC\_ARRAY$>$0. This option specifies the number of receiver planes horizontal to the x-y-plane (surface of the model). The first plane is located in depth REC\_ARRAY\_DEPTH meter. The second plane in REC\_ARRAY\_DEPTH + REC\_ARRAY\_DIST, the last in REC\_ARRAY\_DEPTH + (REC\_ARRAY-1) $\times$ REC\_ARRAY\_DIST. The distance between receivers within each plane is DRX and DRY grid points (DRX*DX and DRY*DY meters).

\subsubsection{Seismograms}
\label{seismograms}
\begin{verbatim}
"Seismograms" : "comment",
			"NDT, NDTSHIFT" : "1, 0",
			"SEIS_FORMAT" : "1",
			"SEIS_FILE" : "./su/test",
\end{verbatim}

with

NDT, NDTSHIFT : samplingrate and timelag (in timesteps!). Write every (NDT)th sample, leaving NDTSHIFT samples at the beginning out\\
default: NDT=1 and NDTSHIFT=0. (NDT=0 is set to NDT=1, NDT$<$0 are set to -NDT.)\\
caution: previous versions used NDTSHIFT=NDT-1\\
SEIS\_FORMAT : data-format, values:
\par
\begingroup
\leftskip=0.5cm
0 = SEG-Y (ASCII-text/native 4-byte-floats (IEEE on PC)/little endian on PC)\\
1 = SU (native 4-byte-floats (IEEE on PC)/little endian on PC)\\
2 = TEXTUAL (native ASCII)\\
3 = BINARY (IEEE-4-byte-floats on PC/little endian on PC)\\
4 = SEG-Y (ASCII-text/native 4-byte-floats (IEEE on PC)/little endian on PC)\\
5 = SEG-Y (ASCII-text/IBM-4-byte-floats on PC/big endian on PC) \\
\par
\endgroup
SEIS\_FILE : basic filename according to output\_ of\_seismograms (SEISMO), separate seismogram files are outputted and will look like  \lstinline{SEIS_FILE_vz.su} or  \lstinline{SEIS_FILE_div.segy}.

If SEISMO$>$0 seismograms recorded at the receiver positions are written to the corresponding output files. The sampling rate of the seismograms is NDT*DT seconds. In case of a small
time step interval and a high number of time steps, it might be useful to choose a high NDT in order to avoid a unnecessary detailed sampling of the seismograms and consequently large files of seismogram data. Possible output formats of the seismograms are SU, ASCII, BINARY, SEGYa or SEGYe. It is recommended to use SU format as seismogram output. The main advantage of this format is that the
time step interval (NDT*DT) and the acquisition geometry (shot and receiver locations) are stored in the corresponding SU header words. Also, additional header words like offset are set by SOFI3D. This format thus facilitates a further visualization and processing of the synthetic seismograms.

Note, however, that SU cannot handle sampling rates smaller than 1.0e-6 seconds and the number of samples is limited to about 32.000. In such cases, you should increase the sampling rate by increasing NDT. If this is impossible (for example because the Nyquist criterion is violated) you must choose a different output format (ASCII or binary).

Each PE internally stores seismograms which are recorded in his portion of the grid. After finishing the time step loop, the seismogram parts are internally exchanged and the merged seismograms are collectively saved to file. The filenames are expanded according to basic file name SEIS\_FILE and the SEISMO option (pressure, particle velocity, curl, divergence). To give an example, PE0 writes the merged seismograms of the x-component of particle velocity to  \lstinline{SEIS_FILE_vx.su} if the output is chosen to be according to the SU file format. 

%If SU-files are output these can be merged together by using the Unix command cat. For example to merge seismograms of the vx-component of particle velocity into one single SU-file use:
%\emph{cat SEIS\_FILE\_VR.* $>$ SEIS\_FILE\_VR} 
%The shell script  \lstinline{sucat.sh} merges all seismograms at the same time.

\subsubsection{Monitoring the simulation}
\begin{verbatim}
"Monitoring the simulation" : "comment",
			"LOG_FILE" : "log/test.log",
			"LOG" : "1",
			"OUT_SOURCE_WAVELET" : "0",
			"OUT_TIMESTEP_INFO" : "10",

\end{verbatim}

with

LOG: output of logging information of node (no output=0; to stdout=1; to file LOG\_FILE=2)\\
LOG\_FILE : log-file for information about progress of program (each PE is printing log-information to LOG\_FILE.MYID) \\
OUT\_SOURCE\_WAVELET : output of used source wavelet (yes=1/no=0) \\
OUT\_TIMESTEP\_INFO : every OUTNTIMESTEPINFOth time step information on update and exchange times are written in LOG\_FILE 


SOFI3D can output a lot of useful and possibly not useful information about the modeling parameters and the status of the modeling process etc. The major part of this information is output by PE 0.
If LOG=1, PE 0 writes this info to stdout, i.e. on the screen of your shell. This is generally recommended  to monitor the modeling process. You may want to save this screen info to a output file by adding  \lstinline{$>$ sofi3D.jout} or  \lstinline{|tee sofi3D.jout}. to your starting command. If LOG=1 all other processes with PE number (PEno) greater than zero will write their information to LOG\_FILE.PEno. If you specify LOG=2 also PE 0 will output his information to LOG\_FILE.0. As a consequence only little information is written directly to the screen of your shell. On supercomputers where you submit modeling jobs to a queuing system as batch jobs LOG=2 may be advantageous. In case of LOG=2, you may still watch the simulation by checking the content of LOG\_FILE.0 for example by using the Unix commands more or tail. After finishing the program the timing information is written to the ASCII file log/test.log.0.timings. This feature is useful to benchmark your local PC cluster or supercomputer (see section \ref{benchmark_sofi3D}). If LOG=0 no output from node 0 will be written, neither to stdout nor to an LOG file. There will be also no output of timing information to the ASCII file log/test.log.0.timings. 

By using the switch OUT\_SOURCE\_WAVELET you can decide to output the used source wavelet (yes=1/no=0). Every PE that includes a source locatation will than outputs the source signal time series as SU output. Due to the fact that the output of information on the update and exchange via fprint can be both slowing down the computation for very small models and producing large output files, you can choose by OUT\_TIMESTEP\_INFO after how many time steps such intermediate information are given.

\subsubsection{Checkpointing}
\begin{verbatim}
"Checkpoints" : "comment",
			"CHECKPTREAD" : "0",
			"CHECKPTWRITE" : "0",
			"CHECKPT_FILE" : "tmp/checkpoint_sofi3D",
\end{verbatim}

with 

CHECKPTREAD : read wavefield from checkpoint file (yes=1/no=0)\\
CHECKPTWRITE : save wavefield to checkpoint file (yes=1/no=0)\\
CHECKPTFILE : checkpoint file name\\

On most supercomputers with a queuing system the run time of job is limited. Sometimes the allowed run time is not sufficient to finish a FD simulation. In such a case, check-pointing can be performed: the first job saves the complete elastic wavefield (CHECKPTWRITE=1) but does not! read the wavefield from a checkpoint file (CHECKPTREAD=0). The subsequent jobs read and write the wavefield to the CHECKPTFILE, i.e. CHECKPTREAD=1 and CHECKPTWRITE=1. In this manner, one simulation can be divided on different batch jobs. The resulting seismograms may be catenated using the SU-command sucat. But be aware that the checkpointing option saves the COMPLETE wavefield information of the last time step, which can produce a huge amount of data.

\subsection{''On the fly'' definition of material parameters}
\label{model_def_func}
If you choose to create the model ``on the fly'', the distribution of the material parameters P-wave velocity $v_p$, S-wave velocity $v-s$ and density $\rho$ for the simple block model has to be defined in the function hh.c that can be found in the src folder. While ``programming'' the model, please note the orientation of the coordinate system (see Chapter \ref{coord_system} for more information). An example C-file ( \lstinline{hh_elastic.c}) for the generation of an elastic model is given in the following:

\input{hh_elastic.c}


As you can see from lines
\begin{verbatim}
	sprintf(modfile,"%s.mod",MFILE);
	writemod(modfile,rho,3);
\end{verbatim}
the density model is written to file to be displayed as described in Chapter \ref{gen_of_mod}. As a base name for the model file, the variable MFILE from the input file is used that actually specifies the external velocity and density files. Note, that the thickness of the first layer \textit{h} is set to 10000 m (10km) which is larger than the vertical extension (NY*DH=5400 m). This way, we effectively created a homogeneous model while demonstrating a two layer case. Simple decrease the variable \textit{h} to 2700 and a real two layer model is created. Generally, you only have to modify the material property definitions in the upper part of the model file and code some if statements to construct simple structures such as planes, cubes, tubes, etc.

\subsection{Compilation of SOFI3D}\label{compexec}
The source codes are located in the directory src. To compile SOFI3D the name of the model function has to be entered in the MAKEFILE. Depending on your MPI environment (MPI distribution) you may need to modify the compiler options in src/Makefile. For a few typical platforms the compiler options are available in src/Makefile. It is often usefull to enable the highest level of optimization (typically -03 or -04). For example the optimization option -04 of the hcc LAM compiler leads to a speedup of SOFI3D of approximately 30 per cent. No other changes in the Makefile should be necessary. 
\begin{verbatim}
# Makefile for SOFI3D 

#--------------------------------------------------------
# edit here

# source code for model generation (viscoelastic)
MODEL_SRC_V = hh_visco.c

# source code for model generation (elastic)
MODEL_SRC_E = hh_elastic.c

# source code for model generation (acoustic)
MODEL_SRC_A = hh_acoustic.c

# source code for benchmark model
MODEL_SRC_BENCH = benchmod.c

# Compiler (LAM: CC=hcc, CRAY T3E: CC=cc, SHARCNet: mpicc)

# ON Linux cluster running LAM
CC=hcc
LFLAGS=-lm -lmpi
CFLAGS=-g -Wall -O4

# On CRAY T3E
#CC=cc

# On Linux Cluster running OpenMPI
#CC=mpicc
#LFLAGS=-lm
#CFLAGS=-Wall -O4 

# On JUROPA cluster
#CC=mpicc
#LFLAGS=-lm
#CFLAGS=-O3 -ipo 


# On HLRN system
#CC=mpcc
#LFLAGS=-lm  

# ON Linux cluster ALTIX.RZ.UNI-KIEL:DE
#CC=icc
#CFLAGS=-mp -O3 -ip0q
#LFLAGS=-lmpi -lm

# ON Linux cluster ALTIX.HRZ.TU-FREIBERG.DE
#CC=icc
#CFLAGS=-mp -O3 -ip0
#LFLAGS=-lmpi -lm -i-static


# after this line, no further editing should be necessary
# --------------------------------------------------------

\end{verbatim} 
To compile the program SOFI3D you must change to the  \lstinline{src} directory and execute:


 \lstinline{make all}
or
 \lstinline{make sofi3D}


The following (or a similar) output should occur:
\begin{verbatim}
icc -c  -mp -O3 -ip0 sofi3D.c
icc -c  -mp -O3 -ip0 comm_ini_s.c
icc -c  -mp -O3 -ip0 checkfd_ssg.c
icc -c  -mp -O3 -ip0 seismo_ssg.c
icc -c  -mp -O3 -ip0 matcopy.c
icc -c  -mp -O3 -ip0 surface_ssg.c
icc -c  -mp -O3 -ip0 update_s_ssg.c
icc -c  -mp -O3 -ip0 update_s_ssg_PML.c
icc -c  -mp -O3 -ip0 update_v_ssg.c
icc -c  -mp -O3 -ip0 update_v_ssg_PML.c
icc -c  -mp -O3 -ip0 snap_ssg.c
[...]
icc -c  -mp -O3 -ip0 sources.c
icc -c  -mp -O3 -ip0 splitrec.c
icc -c  -mp -O3 -ip0 splitsrc.c
icc -c  -mp -O3 -ip0 timing.c
icc -c  -mp -O3 -ip0 util.c
icc -c  -mp -O3 -ip0 wavelet.c
icc -c  -mp -O3 -ip0 writedsk.c
icc -c  -mp -O3 -ip0 writemod.c
icc -c  -mp -O3 -ip0 writepar.c
icc -c  -mp -O3 -ip0 zero.c
icc -c  -mp -O3 -ip0 zero_sofi3D.c
icc -lmpi -lm -i-static  sofi3D.o comm_ini_s.o checkfd_ssg.o seismo_ssg.o 
matcopy.o surface_ssg.o update_s_ssg.o update_s_ssg_PML.o update_v_ssg.o 
update_v_ssg_PML.o snap_ssg.o exchange_v.o exchange_s.o psource.o 
hh.o absorb.o absorb_PML.o av_mat.o comm_ini.o info.o 
initproc.o initsour.o merge.o mergemod.o note.o outseis.o PML_ini.o 
rd_sour.o read_checkpoint.o readdsk.o read_par.o exchange_par.o receiver.o 
save_checkpoint.o saveseis.o sources.o splitrec.o splitsrc.o timing.o 
util.o wavelet.o writedsk.o writemod.o writepar.o zero.o zero_sofi3D.o 
-o ../bin/sofi3D
\end{verbatim} 
The program snapmerge that is used to merge the snapshots (see below) can be compiled with ''make snapmerge''. Since this is not a MPI program (no MPI functions are called) the MPI libraries are not required and any standard compiler (like gcc and cc) can be used to compile this program. The executables sofi3D and snapmerge are located the directory bin.

\subsection{Running the program}\label{compexec1} 
Each SOFI3D run reads the required parameters from a parameter file. An example parameter file was defined in section \ref{modelgeom}  \lstinline{par/in_and_out/sofi3D.json}. 
The command to start a simulation on 8 processor with the lowest priority of -19 (in order to allow working on the a workstation while running a simulation) is as follows. Please note, that we assume you have navigated to the folder  \lstinline{sofi3D/par}.

 \lstinline{mpirun -np 8 nice -19 ../bin/sofi3D ./in_and_out/sofi3D.json }

If you use LAMMPI the command  \lstinline{lamboot -v lamhost} must be executed on node 0 which is the PE where ./par/lamhosts is the file containing IP addresses of all computing nodes. It if often useful to save the standard output of the program for later reference. The screen output may be saved to sofi3D.jout using 

 \lstinline{mpirun -np 8 nice -19 ../bin/sofi3D ./in_and_out/sofi3D.json > ./in_and_out/sofi3D.jout}

Combining both the screen and the file output you can execute the following line (this is used in the shell script  \lstinline{startSOFI3D.sh})

 \lstinline{mpirun -np 8 nice -19 ../bin/sofi3D ./in_and_out/sofi3D.json | tee ./in_and_out/sofi3D.jout}

After the output of geometry and model parameters the code starts the time stepping and displaying timing information:

\begin{verbatim} 
**************************************************
 *********** STARTING TIME STEPPING ***************
 **************************************************
 real time before starting time loop: 0.26 s.

 MYID=0 *****  Starting simulation for shot 1 of 1  **********

 **Message from splitsrc:
 Splitting of source positions from global to local grids finished.
 MYID= 0                 no. of sources= 1

 **Message from splitsrc:
 Table of local source positions (in m), time-shift, centre frequency, 
 amplitude and source type:

 MYID  x         y         z         tshift  fc    amp   stype
 0     2592.000  2592.000  2106.000  0.00    5.00  1.00  1


 Message from function wavelet written by PE 0
 1 source positions located in subdomain of PE 0
 have been assigned with a source signal.

 Computing timestep 1 of 758
 Real time for particle velocity update:         0.04 s.
 Real time for particle velocity exchange:       0.03 s.
 Real time for stress tensor update:             0.04 s.
 Real time for stress tensor exchange:           0.01 s.


 PE 0 is writing snapshot-data at T=0.006600s to
        ./snap/test.bin.vx.000
        ./snap/test.bin.vy.000
        ./snap/test.bin.vz.000

        ./snap/test.bin.p.000

        ./snap/test.bin.div.000
        ./snap/test.bin.curl.000

 total real time for timestep 1 :                0.17 s.

 Computing timestep 11 of 758
 Real time for particle velocity update:         0.08 s.
 Real time for particle velocity exchange:       0.01 s.
 Real time for stress tensor update:             0.26 s.
 Real time for stress tensor exchange:           0.03 s.
 total real time for timestep 11 :               0.37 s.
\end{verbatim}
After some waiting time - depending on the computer performance - we finally get the message:
\begin{verbatim}
 Computing timestep 751 of 758
 Real time for particle velocity update:         0.05 s.
 Real time for particle velocity exchange:       0.02 s.
 Real time for stress tensor update:             0.07 s.
 Real time for stress tensor exchange:           0.01 s.
 total real time for timestep 751 :              0.15 s.


 *********** Finish TIME STEPPING ****************
 **************************************************


 PE 0 is writing 100 merged seismogram traces (vx)   to  ./su/test_vx.su
 PE 0 is writing 100 merged seismogram traces (vy)   to  ./su/test_vy.su
 PE 0 is writing 100 merged seismogram traces (vz)   to  ./su/test_vz.su
 PE 0 is writing 100 merged seismogram traces (p)    to  ./su/test_p.su
 PE 0 is writing 100 merged seismogram traces (dif)  to  ./su/test_div.su
 PE 0 is writing 100 merged seismogram traces (curl) to  ./su/test_curl.su

**Info from function timing (written by PE 0)
 Average times and standard deviations (in percent) for
   velocity update:       0.003 seconds +/- 325.760 per cent
   stress update:         0.005 seconds +/- 483.790 per cent
   velocity exchange:     0.003 seconds +/- 371.565 per cent
   stress exchange:       0.004 seconds +/- 479.052 per cent
   timestep:              0.026 seconds +/- 415.614 per cent
 PE 0 is writing timing information in ASCII format to log/test.log.0.timings

 **Info from main (written by PE 0):
 Total real time of program: 128.22 seconds.

 ******************************************************
 **************** SOFI3D has finished *****************
 ******************************************************
\end{verbatim}  

\subsection{Postprocessing}

The wavefield snapshots can be merged using the program  \lstinline{snapmerge}. The program snapmerge is not a MPI program. Therefore, it can be executed without MPI and the mpirun command. You can run snapmerge on any PC since a MPI environment (e.g. LAM) is not required. You may therefore copy the snapshot outputs of the different nodes to another non-MPI computer to merge the files together.  \lstinline{snapmerge} reads the required information from the SOFI3D parameter file. Simply type

 \lstinline{../bin/snapmerge ./in_and_out/sofi3D.json}

Depending on the model size the merge process may take a few seconds or hours. For the simple block model it only takes a few seconds. The output should read like this:

\begin{verbatim}
***********************************************************
 This is program SNAPMERGE.
 Merge of snapshot files from the parallel
 3-D Viscoelastic Finite Difference Modelling

 written by  T. Bohlen
 Geophysical Institute, Department of Physics,
 Institute of Technology, Karlsruhe, Germany
 http://www.gpi.kit.edu
 ***********************************************************

 Syntax example if excecuted from ./par directory: 
    ../bin/snapmerge in_and_out/sofi3D.json
 Input file for the snapmerge process from command line : 
    in_and_out/sofi3D.json
 Opening input file was successful.


From input file in_and_out/sofi3D.json, 75 objects have been read in.

[...]

 ==============================================================
 P-wave energyfield (files: ./snap/test.bin.div.??? ).

 writing merged snapshot file to  ./snap/test.bin.div
 Opening snapshot files: ./snap/test.bin.div.???  ... finished.
 Copying... ... finished.
 Use
 xmovie n1=100 n2=100 < ./snap/test.bin.div loop=1 label1=Y label2=X title=%g
 to play movie.
 ==============================================================
 S-wave energyfield (files: ./snap/test.bin.curl.??? ).

 writing merged snapshot file to  ./snap/test.bin.curl
 Opening snapshot files: ./snap/test.bin.curl.???  ... finished.
 Copying... ... finished.
 Use
 xmovie n1=100 n2=100 < ./snap/test.bin.curl loop=1 label1=Y label2=X title=%g
 to play movie.
 ==============================================================
\end{verbatim} 

By using the visualization tutorial in section \ref{visual} the 3-D snapshots of the wavefield shown in \FIG{test_1_hom_snap.pdf} can be produced.
\newpage
\subsection{Evolution of the Wavefield}
It is easy to see from \FIG{test_1_hom_snap.pdf} how the P-wave is traveling from the source to the boundaries of the model grid. Here the wave is 
reflected back into the model. After some time a very complex wavefield is generated due to constructive and destructive interference of the different
reflected phases. The resulting seismic section  \lstinline{test_p.su} is shown in \FIG{seis_hom_test_2.ps}. Note the direct P wave and the strong 
reflections at the boundaries of the model grid.
\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/Khang_problem_snap.pdf, width=14 cm}
\caption{\label{test_1_hom_snap.pdf} Evolution of the Pressure wavefield for the simple block model.}
\end{center}
\end{figure}
\clearpage
\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/test_Khang_no_abs_p.pdf, width=14 cm}
\caption{\label{seis_hom_test_2.ps} Pressure seismogram for the simple block model.}
\end{center}
\end{figure} 
\clearpage

\section{Comparison with analytical solutions}
\label{analytic}

In order to gain an absolute validation of Finite Difference data gained from SOFI3D, we perform comparisons of SOFI3D seimogram traces and data calculated by analytic solutions. The analytical solution of a point explosive source within a homogeneous full space and a point vertical force source at the top of a homogenous halfspace are provided by the Matlab Script  \lstinline{analytic_solution_main.m} located in the folder  \lstinline{mfiles/analytical_solution_fullspace} and  \lstinline{mfiles/analytical_solution_halfspace}. In each folder you also find the Matlab script  \lstinline{comparison_analytic_vs_sofi3D.m} to directly compare the SOFI3D and analytical data and gain the same figures which are presented in the following.

We first consider a point explosive source within a homogeneous full space. We find this validation so useful that we already included the SOFI3D modeling into the overnight built test. For further information on the input parameters and elastic properties we refer to the Documentation of the SOFI3D overnight built (see  \lstinline{doc/overnightbuilt_sofi3D/sofi3D_overnight_documentation.pdf}, section \textit{Homogeneous Fullspace}). For several offsets the SOFI3D and analytical traces are displayed in Figure \ref{analytic_sofi3D_p} (pressure component) and Figure \ref{analytic_sofi3D_vx} ($v_x$ component). Each Figure contains the analytical data (red line), SOFI3D data (dashed blue line) and the difference (dotted green line). Both the FD and the analytical data are tracewise normalized to the maximum of the analytical data. The amplitudes are thus truly scaled and the difference is absolute. For each trace the RMS (L2 Norm) is calculated and the RMS error given by

\begin{equation}
\mbox{RMS}_{\mbox{error}} = \frac{\mbox{RMS}_{\mbox{analytical}} -\mbox{RMS}_{\mbox{SOFI3D}} }{\mbox{RMS}_{\mbox{analytical}}} = \frac{\mbox{RMS}_{\mbox{difference}} }{\mbox{RMS}_{\mbox{analytical}}}\mbox{.}
\end{equation}

Considering the center source frequency of $f_c = 500 $ Hz and the P-wave velocity of 3500 $\frac{\mbox{m}}{\mbox{s}}$ we have a typical wavelength of $\lambda_p = 7$ m. At this offset we have an RMS error of less than 0.2 \%. Even after almost three wavelength (offset = 20 m), we observe a RMS error of about half a percent which is well below the tolerated error due to the spatial discretization of about 3 \%. Note that both the SOFI3D and the analytical solution calculate the response of explosive source with a seismic moment of 1 Nm and give the output in particle velocities (in case of Figure \ref{analytic_sofi3D_vx}).


\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/SOFI3D_vs_analytic_p.pdf, width=10 cm}
\caption{\label{analytic_sofi3D_p} Comparison of SOFI3D data and analytical data for various offsets with respect to the pressure component. Note that analytical traces are tracewise normalized and the SOFI3D traces are normalized to the maximum of each analytical trace. True amplitude are thus displayed.}
\end{center}
\end{figure} 


\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/SOFI3D_vs_analytic_vx.pdf, width=10 cm}
\caption{\label{analytic_sofi3D_vx} Comparison of SOFI3D data and analytical data for various offsets with respect to the $v_x$ component (horizontal particle velocity). Note that analytical traces are tracewise normalized and the SOFI3D traces are normalized to the maximum of each analytical trace. True amplitude are thus displayed.}
\end{center}
\end{figure} 

\clearpage

\section{Visualization of 3-D wavefield data}\label{visual}


\subsection{2-D visualization of time slices}
After the data is merged, snapshots can be visualized directly using the SU program xmovie (the exact command is output by snapmerge). You can also cut x-y-slices in 
z-direction out of the 3-D wavefield model with the C++ program  \lstinline{cut_plane.cxx} located in the directory visual/:
\begin{verbatim}
/*------------------------------------------------------------------------
 *   Data Conversion Program from Binary to VTK-Format
 *   Daniel Koehn
 *
 *   Raisdorf, the 22nd of february 2004
 *   update: 29.3.2004 - included VTK Binary IO
 *           01.4.2004 - added Decomposition of Data-Set in smaller 
 *                       blocks
 *           28.10.2007 - using this code to cut a plane in x-y-direction out
 *                        of the 3-D block
 *
 *  ----------------------------------------------------------------------*/
#include <iostream.h>
#include <fstream.h>
#include <math.h>
#include <stdlib.h>
#include <stddef.h>
#include <string.h>
#include <time.h>

int main( int argc, char *argv[] )
{

/* define input and output files here */
ifstream input("~/sofi3D/par/snap/test.bin.p",ios::in | ios::binary);
ofstream output("~/sofi3D/par/snap/test_cut_plane.bin.p",ios::out | ios::binary);
int i,j,k,l,h,NX,NY,NZ;
int IDX, IDY, IDZ, NSNAP;
int CUTPOINT;
float vp, DH;
char test[50];

if(input.fail()){
    cout << "Sorry, can't read that file" << endl;
    exit(1);
    }

/* define geometry (like in *.json)*/
NX=100;
NY=100;
NZ=100;

IDX=1;
IDY=1;
IDZ=1;

/* number of time steps */
NSNAP=20;

/* define gridpoint in z-direction where the cut plane is located */
CUTPOINT=48;

for(k=1;k<=NSNAP;k++){
cout << "Cutting wavefield time step ... " << k << endl;

    for(l=1;l<=NZ;l+=IDZ){
      for (i=1;i<=NX;i+=IDX){
        for (j=1;j<=NY;j+=IDY){
	
	  input.read((char*)&vp,sizeof(float));
	
	  if(l==CUTPOINT){
	    output.write((char*)&vp,sizeof(float));
	  }
	  
        }
     }
    
   }
}
input.close();
output.close();
return 0;
}
\end{verbatim} 

Like in sofi3D.json you have to define the number of grid points NX, NY, NZ and increments IDX, IDY, IDZ in x-, y- and z-direction. Additionally the complete directory path and 
name of the 3-D snapshot file SNAP\_FILE, an output file for the slices, the number of snapshots NSNAP and the gridpoint in z-direction where the cut plane should be located 
(CUTPOINT) have to be defined.
For the compilation of  \lstinline{cut_plane.cxx} a C++ compiler is needed (for example g++):   

 \lstinline{g++ cut_plane.cxx -o cutplane }

After typing  \lstinline{./cutplane} the program cuts for every time step x-y-planes at the gridpoint CUTPOINT in z-direction:
\begin{verbatim}
Cutting wavefield time step ... 1
Cutting wavefield time step ... 2
Cutting wavefield time step ... 3
Cutting wavefield time step ... 4
Cutting wavefield time step ... 5
\end{verbatim}

The output of  \lstinline{cutplane.cxx} can either be visualized using xmovie or you can create stand-alone movies of seismic wave propagation that can be integrated into (PowerPoint) presentations we recommend to create ''nice'' images of each snapshot using MATLAB. An example Matlab-script is  \lstinline{mfiles/snap3D_allplanes.m}. Currently, the output format of each snapshot is a jpeg image and an eps file (other output format is possible, check for ``print'' in the MATLAB manual). The jpg output can serve as input for the program mpeg\_encode to create mpeg movies. mpeg\_encode is a free mpeg encoder that can be downloaded from the following web-site: \url{http://bmrc.berkeley.edu/frame/research/mpeg/mpeg_encode.html} . A short users guide is availabe here: \url{http://www-vis.lbl.gov/NERSC/HowTos/mpeg/help/tools/mpeg_encode.html}.

If the input snapshots have ppm format the mpeg-movies build with mpeg\_encode are quite small (file size is generally less than 1 MB) and quite low resolution. The resolution is, however, mostly
sufficient for (PowerPoint) presentations. Furthermore, mpeg movies created from ppm files are compatible with mpeg players that work under Microsoft Windows (like MS Media Player or Quick Time). This
is often not the case for mpeg movies that are build from jpeg files. The Linux media player xanim can play both ppm based and jpeg based mpeg movies. xanim has the advantage that is allows
to play the movie step by step, i.e. snapshot by snapshot.

Some time slices for the example problem are shown in \FIG{slices_snap2.pdf}.
\begin{sidewaysfigure}[ht]
\begin{center}
\epsfig{file=eps/slices_snap2.pdf, width=18 cm}
\caption{\label{slices_snap2.pdf} Visualization of the propagation of P waves for the simple block model by using time slices.}
\end{center}
\end{sidewaysfigure}
\clearpage

\subsection{3-D visualization using Matlab}

For the convenient visualization of both 2-D slices of the wavefield or 3-D data visualization (2-D slices within a 3-D volume) you can use the Matlab script file  \lstinline{snap3D_allplanes.m} located in the folder  \lstinline{mfiles}. A few input parameters have to be taken from the SOFI3D input file, e.g. grid size or snapshot increments. By the help of various switches you can decide to plot only the 3-D model or 2-D slices of both the div and curl component of the wavefield. Additional contour lines of the model, clipping of the wavefield and the definition of where to cut the 3-D snapshot volume in order to generate 2-D slices will help to evaluate the wavefield in a custom manner. In Figure \ref{matlab_visu_model} (Pseudo 3-D model) and Figure \ref{matlab_visu_snap} (2-D wavefield snapshots) there are examples of the Matlab visualization. Each Matlab figure can be written as eps or png output.


\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/matlab_model.pdf, width=10 cm}
\caption{\label{matlab_visu_model} Pseudo 3-D visualization using Matlab: 2-D slices of the 3-D density model (2 layer case) used for the simulation.}
\end{center}
\end{figure} 

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/matlab_snapshots.pdf, width=10 cm}
\caption{\label{matlab_visu_snap} 2-D visualization using Matlab: 2-D slices of the 3-D div and curl component of the synthetic wavefield at different time steps.}
\end{center}
\end{figure} 


\clearpage



\subsection{3-D visualization using VisIt}
The 3-D wavefield visualization can be done with the visualization toolkit VisIt. The source code or binaries, including a very detailed documentation, can be downloaded from \\ \url{http://www.llnl.gov/visit}. Before the snapshot output of SOFI3D can be used in VisIt the binary data has to be converted into the SILO data format using the C-binary  \lstinline{silo_convert_snap_new} and the shell script  \lstinline{convert_silo_new.sh} in the visual/ directory. Each snapshot will be written to a separate file. The shell script  \lstinline{convert_silo_new.sh} needs
some modifications according the parameters in the SOFI3D-input file. You need to specify:

\begin{enumerate}
\item Name of the input file.
\item Basic name of the output files.
\item Increments in x-,y- and z-Richtung: IDX, IDY, IDZ
\item Number of gridpoints in x-,y- and z-direction: NX, NY, NZ
\item Number of time steps: NT
\item Distance between gridpoints in x-, y- and z-direction: DX, DY, DZ
\item First snapshot: TSNAP1.
\item Time step: DT. 
\end{enumerate}  

To run the shell script navigate to the  \lstinline{scripts} folder and type 

 \lstinline{./convert_silo_new.sh}

We can visualize the pressure wavefield of the simple block model in 7 steps:
\begin{enumerate}

\item Open model files ({\bf{File $\rightarrow$ Select File}}, a list of snapshot files should be visible as shown in \FIG{open_file.pdf},{
{\bf{$\rightarrow$ Select $\rightarrow$ OK}}}).

\item Deactivate the option {\bf{Apply operators and selection to all plots}} and choose {\bf{Plots $\rightarrow$ Pseudocolor $\rightarrow$ div}}. A cube representing the 3-D model domain appears on the screen. You can rotate the model with the left mouse button and zoom in and out using the right mouse button. Rotate the model until the coordinate axes are orientated as shown in \FIG{fig_grid} (y-axes downwards and x-axes to the right).  

\item Choose {\bf{Controls $\rightarrow$ Color table}} and define your own color table optimized to visualize seismic wavefields. Select the colortable \textit{hot} from the Manager menu, type a name, for example  \textit{seismic} and click the \textit{New} button.  Right click on the arrows above the color bar shown in the Editor and select the color white for every arrow, expect the blue and the red on the left and right of the color bar. The color table should now look like \FIG{seismic_color_table.pdf}.    

\item By clicking on the white arrow in the Active plots menu (on the left side of \textit{Pseudocolor}) a pulldown menu will pop up with all the operators that are applied on the Plot. A double click with the left mouse button on \textit{Pseudocolor} opens the {\bf{Pseudocolor plot attributes}} menu (see \FIG{pseudo_color_attrib.pdf}). Activate the limits options \textit{min} and \textit{max} to define the clipping of the wavefield. The correct clipping is found when the wavefronts in the snapshots are neither too pale nor too strong. Set {\bf{Min=-0.001 and Max=0.001}}. Choose the just created {\bf{color table seismic}} and deactivate the {\bf{Legend}}. Now you see a white cube in  the background. 

\item Now it's time to activate some wanted and deactivate some unwanted annotations. Choose {\bf{Controls $\rightarrow$ Annotation}}. Deactivate the options \textit{User information}, \textit{Database} and \textit{Legend}. Go to the \textit{Objects} slide and create a \textit{time slider}. Move the slider to the top left of the screen by pressing the left mouse button on the arrow right to the option Lower left. Add \textit{ms} after time in the option \textit{Text label}. Your VisIt window looks now like \FIG{slider_annotation.pdf}.  

\item Now simply press the {\bf{Play}} button and you can see an animation of the pressure wave field propagating over the surface of the cube (\FIG{first_movie_1.pdf}).      

\item In the last step, we want to see what is going on inside the model domain. We simply choose {\bf{Operators $\rightarrow$ ThreeSlice}} and locate the intersection point of the three planes at the source position at x = 2610 m, y=2100 m and z=2610 m. The result after pressing \textit{Apply} should look like \FIG{three_cut_planes.pdf}. When pressing the \textit{Play} button you can see how the P-wave propagates in the block model and is reflected at the model boundaries (see \FIG{test_1_hom_snap.pdf}). 

\end{enumerate}

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/open_file.pdf, width=\textwidth}
\caption{\label{open_file.pdf} Snapshot of VisIt showing how to open snapshot files in SILO format with VisIt.}
\end{center}
\end{figure}

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/seismic_color_table.pdf, width=\textwidth}
\caption{\label{seismic_color_table.pdf} Snapshot of VisIt showing how to modify the color tables.}
\end{center}
\end{figure}

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/pseudo_color_attrib.pdf, width=\textwidth}
\caption{\label{pseudo_color_attrib.pdf} Snapshot of VisIt showing how to access the pseudocolor plot attributes menu.}
\end{center}
\end{figure}

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/slider_annotation.pdf, width=\textwidth}
\caption{\label{slider_annotation.pdf} Snapshot of VisIt showing how to modifiy the annotating of the plot.}
\end{center}
\end{figure}

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/first_movie_1.pdf, width=0.5\textwidth}\epsfig{file=eps/first_movie_2.pdf, width=0.5\textwidth}\\
\epsfig{file=eps/first_movie_3.pdf, width=10 cm}
\caption{\label{first_movie_1.pdf} Snapshot of VisIt showing the evolution of the pressure field for the block model.}
\end{center}
\end{figure}

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/three_cut_planes.pdf, width=\textwidth}
\caption{\label{three_cut_planes.pdf} Snapshot of VisIt showing three cut planes within the 3-D model volume.}
\end{center}
\end{figure}
\clearpage

\section{Benchmarking with SOFI3D}
\label{benchmark_sofi3D}

A slightly modified version of the example problem mentioned above can be used for benchmarking. Generally, we can on one hand test the scaling behavior of the SOFI3D modeling code and on the other hand - if the SOFI3D scaling behavior is known - the cluster computer (hardware + software installation) itself. Benchmarking can be performed assuming two different strategies: \textit{speed-up} or \textit{scale-up}. For the speed-up scenario we keep the modeling domain constant (NX=NY=NZ=512 gridpoints) and increase the number of cores (PEs) for the simulation. This way we assume that we have a fixed modeling problem and just want to get the results faster. In theory, if we double the number of cores (PEs) we reduce the total runtime by half. By plotting the runtime 

\begin{equation}
t = \frac{t_8}{t_N} \label{eq_speedup}
\end{equation}


with the total runtimes using 8 cores $t_8$, the total runtime $t_N$ using N cores (N = NPROCX $\times $ NPROCY $\times$ NPROCZ) versus the number of cores N, we can theoretically gain a linear speed-up. This means, every time we double the number of cores (e.g. 16, 32, 64, 128, ...) we speed-up the calculation by factor 2. If the speed-up is not linear this can be caused by two different reasons. If the number of grid points per PE is decreasing due to increasing number of PEs, caching effects (storing data in the core cache instead of the memory) can speed-up the update of the wave field (over-linear behavior). If the amount of data that has to be exchanged is increasing due to the increasing number of cores and the communication interface (network interconnect) can not handle that increased amount of exchange, the total runtime can be overpowered by the exchange times and thus the speed-up is below linear. The latter case will most likely occur at some point. 

The speed-up benchmarks performed on the several super computers are displayed in Figure \ref{speedup_bench}. For more information on the hardware architecture and further basic specifications of each cluster computer see Table \ref{tab:cluster}. From Figure \ref{speedup_bench} we see that SOFI3D scales both at the Juropa and the Hermit cluster excellent. The speed-up is close to the linear curve up to 4096 cores (see Figure \ref{speedup_bench}a). In contrast the IC2 speed-up decreases significantly with increasing number of cores. At some point above 1000 cores, the speed-up is even almost 1, i.e. further increasing the  number of cores does not significantly speed-up the simulation. The pure speed-up benchmark does not tell the absolute performance of SOFI3D, this is displayed in Figure \ref{speedup_bench}b where only the total computation times are plotted. It can be seen that these total times can be quite different, especially at large number of cores which is usually due to the different node interconnect (network connection). Depending on the hardware architecture the massive simultaneous exchange of data and, thus, the time needed for exchange of data, can significantly dominate the total time resulting in a good or bad speed-up behavior.

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/bench_speedup.pdf, width=\textwidth}
\caption{\label{speedup_bench} Speed-up benchmarking results performed on the several super computers: a) speed-up factors, b) total computation time.}
\end{center}
\end{figure}

In contrast to the speed-up benchmark, the scale-up benchmarks assume that we have a simulation that reaches the hardware limitations by a certain number of PEs. By increasing the modeling domain we might exceed the hardware resources (e.g. memory consumption). If we increase the number of PEs and increase the modeling domain such that the number of gridpoints per PE is constant (NX=NY=NZ=128 gridpoints per PE!), we can expect to have a constant runtime. Usually the runtime will increase with increasing number of PEs due to an increase in communication overhead, i.e., the update times will be constant but the exchange times will increase. The SOFI3D scale-up behavior performed on several super computers (see Table \ref{tab:cluster}) is displayed in Figure \ref{scaleup_bench}. As can be seen from the subfigures, the interpretation can be quite complicated. The general scale-up behavior of SOFI3D on both Juropa and Hermit is reasonably stable. Only at larger numbers of cores, the total computation at Hermit increases. However, each simulation at constant input parameters takes more than twice the time at the Hermit cluster which already have been shown in Figure \ref{speedup_bench}b. In contrast, the IC2 behavior can be seen as a bad example. Instead of being constant, the total computation time steadily increases with the increasing number of cores. From Figure \ref{scaleup_bench}c we see that this is not due to the update times but due to a steady increase in the exchange times (see Figure \ref{scaleup_bench}c). Obviously the interconnect between each node is too slow or cannot handle massive simultaneous exchange of data very well. In the end these exchange times overpower the good scaling behavior of the update times and result in a bad scale-up behavior.

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/bench_scaleup.pdf, width=\textwidth}
\caption{\label{scaleup_bench} Scale-up benchmarking results performed on several super computers: a) total computation times, b) communication times, c) update times.}
\end{center}
\end{figure}

Alltogether, we believe that the scale-up test is more realistic and can be directly used to also compare the absolute performace of each cluster but it is generally more consuming. For both the scale-up and speed-up benchmark test, input files are prepared and located in the folder  \lstinline{par/in_and_out/benchmark}. The corresponding source-location files are located in the folder  \lstinline{par/sources/benchmark}. In order to ensure comparable benchmark scenarios especially for the scale-up test, each PE has been assigned a source. This way, every PE has to update the wavefield right from the start and not just update with zero values until the wavefield propagates in each PE. As a result, the source file for the scale-up considering 4096 cores, 4096 source locations are defined, too. 

For convenience, there is a shell-script located in the folder  \lstinline{par/} that can be used to perform the benchmarks:

 \lstinline{./startBENCH.sh 8}

The script  \lstinline{startBENCH.sh} will have to be modified according to the job-submission syntax on each cluster computer. Table \ref{tab:speedup_benchmark} can be used to estimate the range of total runtime that has to be specified for the job-submission. The total runtime of the scale-up benchmark can range from 1200 s (Juropa) to 3000 s (IC2).

\begin{table}[hbt]
\begin{center}
% use packages: array
\begin{tabular}{|l|l|} \hline 
Name &  web address\\ \hline
Juropa &  \href{http://www.fz-juelich.de/ias/jsc/EN/Expertise/Supercomputers/JUROPA/JUROPA_node.html}{www.juelich.de}   \\
Hermit &  \href{http://www.hlrs.de/systems/platforms/cray-xe6-hermit/}{www.hlrs.de} \\
IC2    &  \href{http://www.scc.kit.edu/dienste/ic2.php}{www.kit.edu}  \\ 
GRID KIT & \href{http://www.scc.kit.edu/dienste/7349.php}{www.kit.edu} \\
GRID Esslingen & \href{http://www.hs-esslingen.de/de/hochschule/fakultaeten/informationstechnik/it-forschung/bwgrid.html}{www.hs-esslingen.de} \\
HC3 & \href{http://www.scc.kit.edu/dienste/hc3.php}{www.kit.edu} \\ \hline
\end{tabular} 
\caption{Web addresses for information on the architecture and hardware information of each super computer considered for the speed-up and scale-up benchmark.}
\label{tab:cluster}
\end{center}
\end{table}


\begin{table}[hbt]
\begin{center}
% use packages: array
\begin{tabular}{|c|c|c|l|} \hline 
Number of PEs &  Allow simulation time  & Memory consumption \\
Number of PEs & in h (Juropa / Hermit) & per PE in MB\\  \hline
8    &  16:00:00 / 32:00:00 & 1500 \\
16   &  09:00:00 / 15:00:00 & 750  \\
32   &  05:00:00 / 08:00:00 & 400  \\
64   &  03:00:00 / 03:30:00 & 210 \\
128  &  01:00:00 / 01:45:00 & 120 \\
256  &  00:45:00 / 01:00:00 & 70  \\
512  &  00:30:00 / 00:45:00 & 35  \\
1024 &  00:15:00 / 00:30:00 & 20  \\
1536 &  00:10:00 / 00:20:00 & 15  \\
2048 &  00:05:00 / 00:00:00 & 12  \\
2560 &  00:05:00 / 00:10:00 & 10  \\
3072 &  00:05:00 / 00:10:00 & 9   \\
3584 &  00:05:00 / 00:05:00 & 8   \\
4096 &  00:05:00 / 00:05:00 & 7   \\ \hline
\end{tabular} 
\caption{Compilation of number of PEs, minimum expected total simulation times (gained from Juropa Cluster computer at Juelich and estimated for job-script file) and memory consumption (rounded up) for the speed-up benchmarking.}
\label{tab:speedup_benchmark}
\end{center}
\end{table}


\clearpage
\section{Appendix A: Comparison between Taylor and Holberg coefficients}
\label{comp_taylor_holberg}
First, we change some parameters in the input file: Set TIME = 0.8 s, $DT=6.0 \times 10^{-3}\;$s and TSNAP$=0.8\;$s. In the source file, the central frequency $f_c$ should be increased
from 5 to 12 Hz. Using the switches FDORDER and FDCOEFF,  we can use FD operators of different order (2nd-12th) and different FD coefficients (Taylor or Holberg). In figure 
\ref{Taylor_Holberg_2nd_comp.pdf} - \ref{Taylor_Holberg_12th_comp.pdf}, the resulting seismic sections are compared. For 2nd order FD operators the distinct grid dispersion after
the inset of the direct wave dominates the wavefield (\FIG{Taylor_Holberg_2nd_comp.pdf}). Using 4th order FD operators strongly reduce the amount of grid dispersion, but the
accuracy of the Taylor and Holberg coefficients seem to be equal (\FIG{Taylor_Holberg_4th_comp.pdf}). By using 6th order FD operators the operators using Holberg coefficients seem
to be slightly more accurate (\FIG{Taylor_Holberg_4th_comp.pdf}). In the case of the 8th order FD operator, we can barely see any grid dispersion when using Holberg coefficients,
while the Taylor coefficients still show some grid dispersion (\FIG{Taylor_Holberg_8th_comp.pdf}). This is much more noticeable when using 10th or 12th order FD operators (figure
\ref{Taylor_Holberg_10th_comp.pdf} and \ref{Taylor_Holberg_12th_comp.pdf}). This demonstrates the higher accuracy of the Holberg compared to the Taylor coefficients, when using
higher order FD operators.      

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/Taylor_Holberg_2nd_comp.pdf, width=14 cm}
\caption{\label{Taylor_Holberg_2nd_comp.pdf} Pressure seismograms for the block model using a 2nd order FD operator with Taylor (top) and Holberg coefficients (bottom) .}
\end{center}
\end{figure}
\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/Taylor_Holberg_4th_comp.pdf, width=14 cm}
\caption{\label{Taylor_Holberg_4th_comp.pdf} Pressure seismograms for the block model using a 4th order FD operator with Taylor (top) and Holberg coefficients (bottom) .}
\end{center}
\end{figure}
\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/Taylor_Holberg_6th_comp.pdf, width=14 cm}
\caption{\label{Taylor_Holberg_6th_comp.pdf} Pressure seismograms for the block model using a 6th order FD operator with Taylor (top) and Holberg coefficients (bottom) .}
\end{center}
\end{figure}
\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/Taylor_Holberg_8th_comp.pdf, width=14 cm}
\caption{\label{Taylor_Holberg_8th_comp.pdf} Pressure seismograms for the block model using a 8th order FD operator with Taylor (top) and Holberg coefficients (bottom) .}
\end{center}
\end{figure}
\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/Taylor_Holberg_10th_comp.pdf, width=14 cm}
\caption{\label{Taylor_Holberg_10th_comp.pdf} Pressure seismograms for the block model using a 10th order FD operator with Taylor (top) and Holberg coefficients (bottom) .}
\end{center}
\end{figure}
\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/Taylor_Holberg_12th_comp.pdf, width=14 cm}
\caption{\label{Taylor_Holberg_12th_comp.pdf} Pressure seismograms for the block model using a 12th order FD operator with Taylor (top) and Holberg coefficients (bottom) .}
\end{center}
\end{figure}
\clearpage

\section{Appendix B: Exponential damping vs. PML boundaries}\label{comp_EXP_PML} 
To demonstrate the improvement of the absorbing boundary condition by introducing PMLs accoring to \cite{komatitsch:07} and \cite{martin:09}, we first apply the exponential damping by setting the parameter ABS\_TYPE=2 and the width of
the absorbing frame to FW=30 gridpoints. The resulting seismic section is shown in \FIG{ABS_cPML_comp.pdf} (top). You can clearly see the direct P-wave. The amount of reflected waves is strongly reduced, compared to \FIG{seis_hom_test_2.ps} where no absorbing boundary was applied, but there are still some noticeable boundary reflections visible. 

Next we activate the PML boundaries by setting ABS\_TYPE=1 and use a thinner absorbing boundary layer FW=10. The resulting seismic section calculated using 4th order FD operators is shown in \FIG{ABS_cPML_comp.pdf}(bottom). When using PMLs no dominant boundary reflections are visible. In \FIG{ABS_PML_comp_shots.pdf}, the evolution of the pressure for the block model with exponential damping
(left) and PMLs (right) are compared. While the absorbing frame with exponential damping produces strong boundary reflections, only few reflections can be seen when using PMLs, despite
the fact that the PML frame is thinner than the exponential damping frame.            

\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/ABS_cPML_comp.pdf, width=14 cm}
\caption{\label{ABS_cPML_comp.pdf} Pressure seismograms for the block model with exponential damping (top) and PMLs (bottom).}
\end{center}
\end{figure}
\begin{figure}[ht]
\begin{center}
\epsfig{file=eps/ABS_PML_comp_shots.pdf, width=14 cm}
\caption{\label{ABS_PML_comp_shots.pdf} Evolution of the pressure field for the block model with exponential damping (left) and PMLs (right).}
\end{center}
\end{figure}

\bibliography{thesis}
\end{document}