Overview

HTTPS SSH
Sedonu is a time-independent implicit Monte Carlo neutrino transport
code. Please read the copyright statement in LICENSE.txt and at the
beginning of each file. If anyone would like to contribute to
development of Sedonu, please let me know or send me a pull request on 
BitBucket.

This version of Sedonu (used in Richers et al. 2017) is rather
different than the previous version
(2962e1aa226f28cd3f188ac3fda933a0ecefdf17 used in Richers et
al. 2015). The tests and parameter descriptions are a bit out of
date. If you have questions before I manage to clean it up, send an
email. The 2015 version is available on branch named "2015".

REREQUISITES
==============================
To clone the repository, execute

    git clone  https://srichers@bitbucket.org/srichers/sedonu.git

In the sedona directory, copy the make.inc.template file to make.inc
(make.inc is used by the make system).

    cp make.inc.template make.inc

Open make.inc in your favorite text editor and make sure the compiler
and compiler flags are set to the values you want to use. Your mpi
compiler scripts should use the same compiler used to build NuLib and
HDF5. To check the compiler a script uses and that hdf5 was compiled
with, execute

    mpicc -v   #same for mpif90 and mpicxx
    grep 'Compiler' <hdf5_directory>/lib/libhdf5.settings

Note that the only compilers we've tested are GNU, Intel, and (only on
Hopper) PGI. Intel 12 and below appear not to work. On Zwicky I have
used Intel 14 and gcc 4.8.1.


########
# ZLIB #
########
In Debian-based systems, this should be as easy as

   apt-get install zlib1g-dev


#######
# MPI #
#######
In Debian-based systems, this should be as easy as

   apt-get install mpich2

On Zwicky, I use the intel MPI implementation and the intel compilers.

#######
# GSL #
#######
GSL is used to generate random numbers in sedona. If you do not
already have GSL installed, you may be able to install it using the
following in Ubuntu. This is the recommended way to install GSL. In
ubuntu you can check whether these are installed using "dpkg -l | grep
'gsl'" (without the quotes).

    apt-get install gsl-bin libgsl0ldbl libgsl0-dev

If you can't install software with a package manager on your machine
(e.g. a cluster), then you will have to build it from scratch in your
chosen directory. This almost never has to be done, so we won't
include directions here. If directions would be helpful, let us know
and we can add them. So far, the latest tested version of gsl is 1.15.


#######
# LUA #
#######
Lua is a scripting language that is used to read the parameter
files. If you do not already have Lua installed, you may be able to
install it using the following in Ubuntu. However, you will have to
change a flag in make.inc, as for some reason the Ubuntu package
manager renames the libraries. Also, the library will be in the
default search path, so remove the information that says where
everything is. If it works, this is the recommended way to install
lua. In ubuntu you can see if lua is installed using "dpkg -l | grep
'lua'" (without the quotes).

    apt-get install lua5.2 lua5.2-dev
    (in make.inc) #LUA_BASE=external/lua-5.2.2/install #commented out b/c it's unused
    (in make.inc) -I/usr/include/lua5.2                #ubuntu puts lua in a non-standard directory
    (in make.inc) LUA_LIBS=-llua5.2                    #at least this is the name Ubuntu gives the library ("liblua5.2.so")

If you can't install software with a package manager on your machine
(e.g. a cluster), then you will have to build it from scratch in your
chosen directory. We will install it into the "external" directory
here. First, download and extract the source code (this version is the
latest version tested with sedona).

    wget http://www.lua.org/ftp/lua-5.3.2.tar.gz
    tar -xvf lua-5.3.2.tar.gz -C external
    rm lua-5.3.2.tar.gz

Change into the lua directory to build it. If there are linking
errors, you will need to add MYLIBS=-ltermcap to the make line. See
Lua's readme for more detailed options. These work for me, however.

    cd external/lua-5.3.2
    make linux            # or "make macosx" or "make posix". See the lua readme for more information 
    make test             # success if you see the version string
    make local            # to choose installation directory, do "make install INSTALL_TOP=/path/to/directory"

If you installed it locally ("make local") it will put everything in a
subdirectory called "install". Everything actually needed for
compiling, linking, and running the code is in the hdf5
subdirectory. The default LUA_BASE variable in make.inc should reflect
the location of the install subdirectory.


########
# HDF5 #
########
HDF5 is used by NuLib to create and read the opacity tables. IT MUST
BE COMPILED BY THE SAME FORTRAN COMPILER COMPILING NULIB. If you can
get away with a copy already installed by "apt-get install
libhdf5-serial-dev" (check on ubuntu using "dpkg -l | grep 'hdf5'"
without the quotes) or some other means that's fine, but chances are
that it was compiled with a different compiler and you will have to
recompile it yourself. You will also have to change the values of the
hdf5 path in make.inc to reflect the system's location for the hdf5
files. Below are instructions for installing the latest version tested
into the "external" directory.

If you need to compile your own copy of hdf5, download the tarball and
extract the table:

    wget http://www.hdfgroup.org/ftp/HDF5/current/src/hdf5-1.8.13.tar.gz
    tar -xvf hdf5-1.8.16.tar.gz -C external
    rm hdf5-1.8.16.tar.gz

Go into the directory

    cd external/hdf5-1.8.16

Configure the make process. Some versions of gcc have a bug that
causes hdf5 to fail tests if any optimization is enabled, so there are
two examples below. The only tested compilers are intel and GNU.

    FC=gfortran CC=gcc CXX=g++  CFLAGS=-O0 ./configure --enable-fortran --enable-cxx   #for GNU compilers
    FC=ifort    CC=icc CXX=icpc            ./configure --enable-fortran --enable-cxx   #for Intel compilers

Build and test the binaries. This takes quite a while.

    make
    make test
    make install

Everything actually needed for compiling, linking, and running the
code is in the external/hdf5-1.8.13/hdf5 subdirectory. The default
hdf5 path in make.inc.template points to this location. Make sure that
the mpi compiler wrapper scripts (mpif90, etc) use the same compiler
that you used to compile hdf5 (see the beginning of the prerequisites
section).



#########
# NuLib #
#########
NuLib calculates the neutrino opacities based on a tabulated equation
of state (EOS). It must be built to yeild the executable to generate
the table and the libraries used by sedona to read the table. It must
be run to generate the opacity table before running Sedona. This is
all described below.

--Building------------

NuLib is publicly available at <nulib.org>, but is also contained
within Sedona as a git submodule. This makes it easy to pair a commit
of sedona with a commit of nulib, so results are entirely
reproducible. That is, each sedona commit knows which commit of NuLib
to check out to make the both codes exactly as they were at the time
of sedona's commit. To download NuLib, execute the following in the
base directory. It will automatically download into the "external"
directory.

    git submodule init
    git submodule update

Check that your parallel compiler wrapper script uses the same
compiler that was used to build HDF5 (see the top of the prerequisites
section). Then build NuLib

    make nulib

If you want to clean the nulib directory, execute

    make nulibclean


--Generating the Opacity Table--------------

You can choose to generate your own tables as described below or
request the tables used in a publication from srichers@tapir.caltech.edu
(they are about 1GB each and difficult to host online).

Now NuLib is compiled and ready to be used by Sedona. However, if you
wish to use neutrinos, you will have to generate an opacity
table. This requires a tabulated equation of state (EOS). There are
multiple EOS available at <stellarcollapse.org/equationofstate>. Here
we will configure things with the LS220 EOS, though it should be
generalizable to any EOS table that NuLib understands.

(1) -- Download and extract the LS220 EOS table

    wget http://stellarcollapse.org/EOS/LS220_234r_136t_50y_analmu_20091212_SVNr26.h5.bz2
    bunzip2 -c LS220_234r_136t_50y_analmu_20091212_SVNr26.h5.bz2 > external/tables/LS220.h5
    rm LS220_234r_136t_50y_analmu_20091212_SVNr26.h5.bz2

(2) -- First, we want to set a few of the processes to be included. Using your favorite text editor, open the file shown below.

    emacs external/NuLib/src/requested_interactions.inc

Inelastic electron scattering is not yet programmed in, so turn on
(very approximate) elastic electron scattering, and turn off the
inelastic scattering. Also, we need to turn off the emission processes
from pair annihilation and from NN bremsstrahlung, since the inverse
rates involve two neutrinos as input, and NuLib does not deal with the
inverse processes correctly. Change the following parameters.

    (in requested_interactions.inc)
    logical :: add_nue_scattering_electrons = .true.
    logical :: add_anue_scattering_electrons = .true.
    logical :: add_numu_scattering_electrons = .true.
    logical :: add_anumu_scattering_electrons = .true.
    logical :: add_nutau_scattering_electrons = .true.
    logical :: add_anutau_scattering_electrons = .true.

    logical :: add_nue_Iscattering_electrons = .false.
    logical :: add_anue_Iscattering_electrons = .false.
    logical :: add_numu_Iscattering_electrons = .false.
    logical :: add_anumu_Iscattering_electrons = .false.
    logical :: add_nutau_Iscattering_electrons = .false.
    logical :: add_anutau_Iscattering_electrons = .false.

    logical :: add_nue_emission_epannihil = .false.
    logical :: add_anue_emission_epannihil = .false.
    logical :: add_numu_emission_epannihil = .false.
    logical :: add_anumu_emission_epannihil = .false.
    logical :: add_nutau_emission_epannihil = .false.
    logical :: add_anutau_emission_epannihil = .false.

    logical :: add_nue_emission_NNBrems = .false.
    logical :: add_anue_emission_NNBrems = .false.
    logical :: add_numu_emission_NNBrems = .false.
    logical :: add_anumu_emission_NNBrems = .false.
    logical :: add_nutau_emission_NNBrems = .false.
    logical :: add_anutau_emission_NNBrems = .false.

(3) -- NuLib has an example program that generates a rather unresolved
opacity table. This program serves as an example that is intended to
be modified by the end user. You can feel free to modify it, but if
you check the file out again it will be overwritten, so back it up or
rename it if you want to prevent accidents. Here we will just change
the file shipped with nulib so we don't have to mess with the
makefiles, but if you want to create multiple custom table-generating
programs you will have to do that. Anyway, we have to modify the
fortran code to say how big we want our table to be and where the EOS
table is. Open the source file in your favorite text editor.

    emacs external/NuLib/src/make_table_example.F90

A more detailed description of what everything means is in the NuLib
README file (external/NuLib/README). First, notice the two variables
"mytable_neutrino_scheme" and "mytable_number_species". Do NOT change
mytable_number_species. If you want to change how neutrinos are
represented in the table change "mytable_neutrino_scheme". Also notice
that there is a variable "mytable_number_groups" that is set to 24. If
you want finer resolution in neutrino energy, increase this
number. However, for this example we will leave all three of these
variables as they are. We must tell the program where to find the EOS
table.

    (in make_table_example.F90:L28) character*200 :: eos_filename = "<path_to_sedona>/external/tables/LS220.h5"

Each EOS table uses a 'reference mass' that is used to convert
densities to number densities (e.g. the SL EOS's require a number
density input in terms of fm^-3). The default is set for any of the LS
tables. I don't know what reference mass would be required for the
Shen EOS. You would have to email Evan O'Connor and ask. If you use a
non-LS table you will have to change the line below to something else,
but if you use a LS table, just leave it alone.

    (in make_table_example.F90:L84) m_ref = m_n   !m_n is defined in constants.inc

Finally, we have to change how big our table is going to be. As
described in the NuLib README file, Evan recommends at least 10 points
per decade in rho, 20 points per decade in temperature, and a point
for every 0.01 or 0.02 in Ye. The LS220 table has the following lower
and upper limits: T:{1e-2,1e2.4}MeV rho:{1e3,1e16}g/cm^3
Ye:{0.035,0.55}, and if calculating beta-equilibrium Lattimer and
Swesty recommend staying at temperatures above 0.5MeV. Find the chunk
of code commented "!set up table" and notice that the table boundaries
(min_ye, etc) are within these recommendations, but the number of
points (final_table_size_ye) is far too small. (final_Itable_* deals
with the inelastic scattering table, which is not yet implemented so
we won't worry about it). For this example we will leave the ranges as
they are and adjust the number of points to get a decently-resolved
table, but you are free to set whatever you want. Ye has a range of
0.515, so we need 26-52 points.

    (in make_table_example.F90:L90) final_table_size_ye = 40

rho goes over 9.8 decades, so we need about 100 points in rho

    (in make_table_example.F90:L91) final_table_size_rho = 100

Temp goes over 3 decades, so we need about 60 points

    (in make_table_example.F90:L92) final_table_size_temp = 60

You can change where the energy bins lie, but we will leave them as
they are. If you want to make a custom table that deviates from this
program, feel free. However, if you change the structure or what
variables represent, you will also have to modify
nulibtable_reader.F90 to read everything in correctly. Assuming you
don't make any further changes, this should be ready to create a
table.

(4) -- Save the file, exit the editor, rebuild the nulib directory,
and execute the file.

    make nulibclean
    make nulib
    external/NuLib/make_table_example

This table took about 25 minutes to generate on my 2.4 GHz
processor. Finally, move the final product into the external/tables
directory (the filename will depend on the options you give it).

    mv NuLib_LS220_rho100_temp60_ye40_ng24_ns3_Itemp10_Ieta10_version1.0_20130915.h5 external/tables/NuLib.h5

NuLib is now ready to use!



###########
# DOXYGEN # (optional)
###########

Doxygen will generate useful documentation to show the structure of files and classes in Sedonu. You can in stall it on Ubuntu using

    apt-get install doxygen doxygen-gui graphviz

Then, to build the documentation, from the Sedonu base directory, execute

    cd doc
    doxygen Doxyfile

To see the documentation, in a browser open sedonu/doc/html/index.html


COMPILING
=====================================
If all of the prerequisites are ready, simply execute

    make

This will use the settings in make.inc. To clean the src directory execute

    make clean







TESTING
=====================================
There are a couple of tests in the 'tests' directory.
A brief list of the tests is included below. More information can be
found in the tests's individual directory in its README file. To run
each test, simply go into the test directory and execute "make".

 - equilibrium_ye - sets up a spherical system where no neutrinos are
   emitted and only Ye is solved for. Tests that the Ye solver works
   properly with the nulib table.
 - nulib_plots - generates lots of plots for easily examining the
   values of the nulib tables.
 - nut_blackbody - tests to see that a blackbody distribution is
   actually built up in an optically thick medium
 - nut_neutron_star - optically thick problem to test that the
   neutrinos propagate as expected and lead to a reasonable result
 - nut_spherical_emis - emits neutrinos isotropically from a central
   source and checks that the output spectrum is as expected.
 - nut_therm_shell - check that fluid relaxes to the proper {T,Ye}
   when irradiated by neutrinos of a certain {T,chempot}
 - nut_visc_ball - see what spectrum from a ball of gas, sourced by
   "viscosity" (really just an arbitrary source term) looks like

IF IT IS NOT IN THE TESTS FOLDER, IT IS PROBABLY NOT TESTED!




RUNNING
===========================
When sedonu is built, it produces and executable called "sedonu" in the
base directory. Move this executable to the run directry (where all
the data will be put). See test problems for example parameter
files. Go to that directory and execute

    ./sedonu [parameter_file]

The argument is optional. By default it will look for a parameter file
called "param.lua" in the same directory. However, it is good practice
to copy the parameter file into the run directory to have a record of
the settings. The output will be spectrum files and "ray" files. The
spectrum files contain data about the radiation that escapes the
system. The ray files contain the system properties (e.g. temperature,
Ye) along a ray from the center.