This is a modified version of lanl/CGMF for use in uncertainty quantification. The new capabilities include the added options listed below. Includes the following dependencies which are automatically downloaded and integrated by the build system:

• nlohmann/json for parsing json formatted parameter files
• beykyle/osiris a library of optical model potentials and solvers

The goal of this modification is to provide an easy to use platform for running CGMF with modified internal parameters (e.g. optical model parameter, and more), and analyzing the results. To do this Python bindings are provided to run CGMF, directly populating event-by-event information into a CGMFtk.Histories object, without writing to disk or copying large amounts of data in memory.

To build the pyCGMF module:

py setup.py build -j{nproc}
pip install -e .

This will:

• build CGMF as a library
• build the pyCGMF module
• build and install CGMFtk, if it's not already installed
• install the pyCGMF module

Now we can import pyCGMF within Python, to run CGMF, and populate a CGMFtk.Histories object without ever writing to disk or copying data, like so:

from pyCGMF import Input, run
from CGMFtk.histories import Histories
import numpy as np

inp = Input(
    nevents = 100,
    zaid    = 98252,
    einc    = 0.0
)

histories = run(inp)
print("nubar: {}".format( histories.nubartot() ) )

# we can save the histories to disk 
histories.save("histories.npy")

# later, we can load them back into memory
hist2 = Histories.load("histories.npy")

When we do do disk I/O by calling Histories.load/save, we make use of numpy.load/save under the hood for speed and minimal disk usage. In particular, storing CGMF histories using this method rather than the standard CGMF output cuts down the time to parse the histories by about half!

We can even run in parallel, for example, by using mpi4py. First, create a script called run_cgmf.py:

from pyCGMF import Input, run
from CGMFtk import Histories

from mpi4py import MPI

comm = MPI.COMM_WORLD
size = comm.Get_size()
rank = comm.Get_rank()

# 1000 events per MPI rank
inp = Input(
    nevents  = 1000,
    zaid     = 98252,
    einc     = 0.0,
    MPI_rank = rank
)

# run worker on each MPI rank
# note: we don't have to write to disk!
hists = run(inp)

# gather histories from all MPI ranks
result = comm.gather(hists.histories, root=0)

# concatenate all results into single Histories object and write the result
if rank == 0:
    all_histories = Histories(  from_arr=np.concatenate( result, axis=0  ) )
    all_histories.save("all_histories.npy")

This can be executed with mpirun, e.g.:

mpirun -n 4 python run_cgmf.py

More extensive examples are contained in notebooks in pyCGMF/examples/.

The original documentation to build CGMF as an executable is preserved below, with the addition of the extra command line arguments.

Patrick Talou, XCP-5, Los Alamos National Laboratory, [email protected]

Ionel Stetcu, T-2, Los Alamos National Laboratory, [email protected]

Patrick Jaffke, Institute for Defense Analyses, [email protected]

Michael E. Rising, XCP-3, Los Alamos National Laboratory, [email protected]

Amy E. Lovell, T-2, Los Alamos National Laboratory, [email protected]

Toshihiko Kawano, T-2, Los Alamos National Laboratory, [email protected]

CGMF is a code that simulates the emission of prompt fission neutrons and gamma rays from excited fission fragments right after scission. It implements a Monte Carlo version of the Hauser-Feshbach statistical theory of nuclear reactions to follow the decay of the fission fragments on an event-by-event basis. Probabilities for emitting neutrons and gamma rays are computed at each stage of the decay. Each fission event history records characteristics of the parent fragment (mass, charge, kinetic energy, momentum vector, excitation energy, spin, parity) and the number (multiplicity) and characteristics (energy, direction) of the prompt neutrons and gamma rays emitted in this event.

The main publication (and documentation) to cite for CGMF is:

Patrick Talou, Ionel Stetcu, Patrick Jaffke, Michael E. Rising, Amy E. Lovell, and Toshihiko Kawano, “Fission Fragment Decay Simulations with the CGMF Code,” Comp. Phys. Comm., 269 (2021), Article 108087. DOI: 10.1016/j.cpc.2021.108087

• Patrick Talou, Ionel Stetcu, Patrick Jaffke, Michael E. Rising, Amy E. Lovell, and Toshihiko Kawano, “Fission Fragment Decay Simulations with the CGMF Code,” Comp. Phys. Comm., 269 (2021), Article 108087. Los Alamos Technical Report LA-UR-20-21264 (2020).

• Open source, BSD-3
• Copyright: Triad National Security, LLC. All rights reserved.
• Programming language: C++ (and Python for post-processing)
• Fission reactions handled: spontaneous fission of Pu-238,240,242,244 and Cf-252,254; neutron-induced fission reactions from thermal up to 20 MeV for n+U-233,234,235,238, n+Np-237, and n+Pu-239,241.

0. Optional, set CGMFDATA environment variable to point to data/ directory
1. Create a build directory
2. Change to build directory
3. Configuration: type cmake .. and then make for default build
   • The .. needs to point to the top-level CGMF directory
   • Options:
     • CMAKE_BUILD_TYPE [Debug, RelWithDebInfo, Release]
     • CMAKE_INSTALL_PREFIX [path/to/install/directory]
     • cgmf.shared_library [ON/OFF (default)]
     • cgmf.x.MPI [ON/OFF (default)]
     • cgmf.tests [(default) ON/OFF]
4. Building: type make for default build
   • This creates the static library libcgmf.a in the build/libcgmf directory
   • This creates the executable cgmf.x in the build/utils/cgmf directory
5. Testing: type make test or ctest to run the default tests
6. Installing: type make install to install CGMF
   • This creates the following directory structure in the CMAKE_INSTALL_PREFIX directory:
     • bin/ [contains cgmf executable]
     • include/cgmf-1.1.0 [contains cgmf header files]
     • lib/cgmf-1.1.0 [contains libcgmf library]
     • share/data/cgmf-1.1.0 [contains cgmf data files]

Example configuration in release mode with build, test and install:

• cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=../install -Dcgmf.x.MPI=ON ..
• make && ctest && make install
• If the build and tests pass:
  • This creates the static library libcgmf.a in the build directory
  • This creates the executable cgmf.mpi.x in the build/utils/cgmf directory
  • This installs bin, lib, cgmf/data, and cgmf/include directories into install directory

./cgmf.x [options]

Options:

-i $ZAIDt   		[required]	1000*Z+A of target nucleus, or fissioning nucleus if spontaneous fission
-e $Einc    		[required]	incident neutron energy in MeV (0.0 for spontaneous fission)
-n $nevents 		[optional]	number of Monte Carlo fission events to run or to be read. If $nevents is negative, produces initial fission fragments yields Y(A,Z,KE,U,J,p)
-s $startingEvent	[optional]	skip ahead to particular Monte Carlo event (1 is default)
-f $filename		[optional]	fission histories or yields result file (default: "histories.cgmf" or "yields.cgmf")
-t $timeCoinc		[optional]	time coincidence window for long-lived isomer gamma-ray emission cutoff (in sec)
-d $datapath		[optional]	overrides the environment variable CGMFDATA and default datapath
-o $omppath		[optional]	use custom global OM params, formatted in json at given path
-r $seed                [optional]      use custom RNG seed (Note: RANDOM_SEED_BY_ARG must be set to true in config.h)
-g $ZAIDsf              [optional]      Only de-excites fragments correspinding to ZAIDsf, with excitation energy, spin, parity, and TKE
                                        all sampled as if that fragment was produced by the fissioning nucleus

The -o option currently accepts the following global optical models: Koning-Delaroche, WLH, and Chapel-Hill 89. Example parameter files are contained in data/optical

A concise summary of average results is displayed on the standard output, and event-by-event results or initial fragment distributions are saved in "histories.cgmf" or "yields.cgmf" (default names).