Options for producing OpenMC h5 cross sections

Jonathan Shimwell

IAEA Consultancy Meeting on Model Code Output

Application Nuclear Data Form Structure

Obtaining h5 cross section files

Download individual isotopes using the openmc-data-downloader Python package https://github.com/openmc-data-storage/openmc_data_downloader

Download compressed (tar) libraries from OpenMC.org

Process h5 files using the data repository https://github.com/openmc-dev/data
- Process ACE files using convert*.py scripts
- Process ENDF files using generate*.py scripts

Obtaining h5 cross section files

Library	Release	Processed by	Download from openmc.org	Download from openmc-data-downloader	Download ACE files and convert HDF5	Download ENDF files and generate HDF5	Convert local ACE files
CENDL	3.1 3.2					generate_cendl.py
ENDF/B	VII.0	LANL	✓				convert_mcnp70.py
ENDF/B	VII.1	LANL	✓				convert_mcnp71.py
ENDF/B	VII.1	NNDC	✓	✓	convert_nndc71.py	generate_endf.py
ENDF/B	VIII.0	LANL	✓				convert_lib80x.py
ENDF/B	VIII.0	NNDC	✓
FENDL	2.1, 3.0, 3.1a, 3.1d			✓ (3.1d)	convert_fendl.py
JENDL	4.0					generate_jendl.py
JEFF	3.2		✓		convert_jeff32.py
JEFF	3.3		✓		convert_jeff33.py
TENDL	2015, 2017, 2019			✓ (2019)	convert_tendl.py

There are several ways of obtaining the cross section data in the h5 file format required for neutronics simulations.

OpenMC data repository

Python scripts

Permissive open source license (MIT)

11 contributors

Storage and sharing opportunity to avoid reproducing the same scripts

Useful for developers

Generate scripts require NJOY install

PyPi package with pip install on the horizon

git clone https://github.com/openmc-dev/data.git

cd data

python convert_fendl.py --release 3.1d
>> downloading, extracting and processing ...

ls
>> fendl-3.1d-ace/
>> fendl-3.1d-endf/     
>> fendl-3.1d-download/
>> fendl-3.1d-hdf5/

ls fendl-3.1d-hdf5/
>> cross_sections.xml
>> neutron/
>> photon/

ls fendl-3.1d-hdf5/neutron
>> Ag107.h5
>> Ag109.h5
>> Al27.h5
>> Ar36.h5
>> ...

export OPENMC_CROSS_SECTIONS=fendl-3.1d-hdf5/cross_sections.xml

Example usage of one of the Python scripts

This particular script downloads 429MB of compressed ACE and ENDF files

Extracts the files

Processes the files into h5 file for use in OpenMC

Obtaining h5 cross sections - data repo

OpenMC.org

https://openmc.org includes a data section with downloads
- Official Libraries
- LANL-Based Data Libraries
- Other Libraries
- Depletion chains

OpenMC documentation pages includes information on the nuclear data interface https://docs.openmc.org

Obtaining h5 cross sections - OpenMC.org

The h5 files can be used in simulations once the OPENMC_CROSS_SECTION environmental variable has been set

wget -O nndc-b7.1.tar.xz https://anl.box.com/shared/static/9igk353zpy8fn9ttvtrqgzvw1vtejoz6.xz
>> dowloading compressed file ...

mkdir nndc-b7.1-hdf5

tar -xf nndc-b7.1.tar.xz -C nndc-b7.1-hdf5
>> extracting compressed file ...

ll nndc-b7.1-hdf5
>>cross_sections.xml
>>neutron/
>>photon/

ll nndc-b7.1-hdf5/neutron
>>Ac225.h5
>>Ac226.h5
>>Ac227.h5
>>Ag107.h5
>>Ag109.h5
>>...

export OPENMC_CROSS_SECTIONS=nndc-b7.1-hdf5/cross_sections.xml

Downloads can be scripted using wget

Obtaining h5 cross sections - OpenMC.org

OpenMC Docker image production

OpenMC automatically produces a new docker image with every code merge

Github Actions is used to build the docker image and upload it to Dockerhub

https://hub.docker.com/r/openmc/openmc

https://github.com/openmc-dev/openmc/actions

Data storage repo - processing

Contains repository for the openmc-data-downloader Python package.

Contains separate repositories for each library release

Repositories can be templated

Open source, permissively licensed (MIT) https://github.com/openmc-data-storage

Data storage repo - enviroment

Use of the openmc/openmc-dev image this contains NJOY as well

Addition of the openmc-data repo to the base image

FROM openmc/openmc:develop

RUN git clone https://github.com/openmc-dev/data.git

COPY tests tests/

ENV OPENMC_CROSS_SECTIONS=/share/cross_sections.xml

Copying test suite

See environmental variable

4 Lines

+ + =

Data storage repo - processing

Github action (CI) then commits the new h5 files from the share folder to the repository

Github actions (CI) build and run the new docker image

name: process_test_h5_files
on:  
  push:    
    branches:
    - main

env:
  PROCESS_CMD: 'python data/convert_*****.py -d /share/h5_files -r *****'
jobs:
    build:
      runs-on: ubuntu-latest
      steps:
      - uses: actions/checkout@master
        with:
          persist-credentials: false # otherwise, the token used is the GITHUB_TOKEN, instead of your personal token
          fetch-depth: 0 # otherwise, you will failed to push refs to dest repo
      - name: Create custom readme
        run: |
            python readme_writer.py
            cat README.md
      - name: Build and test with Docker
        run: |
            docker build --tag nuc_data_maker .
            rm -rf h5_files
            docker run -v /home/runner/work/${{ github.event.repository.name }}/${{ github.event.repository.name }}:/share nuc_data_maker /bin/bash -c "${{ env.PROCESS_CMD }} && pytest -rP tests/test_xs_in_simulations.py"
      - name: Commit files
        run: |
          git config --local user.email "action@github.com"
          git config --local user.name "GitHub Action"
          git add h5_files/*.h5
          git add h5_files/cross_sections.xml
          git add README.md
          git commit -m "Added readme with repo name (bot commit)" -a
      - name: Push changes
        uses: ad-m/github-push-action@master
        with:
          github_token: ${{ secrets.GITHUB_TOKEN }}
          branch: ${{ github.ref }}

The new docker image is then run externally with a shared volume mount

40 Lines

4 Lines

+ + =

Data storage repo - testing

Each h5 cross section file is then used in an OpenMC simulation

Testing the h5 file produced has converted successfully and does not cause errors during usage

import os
import unittest
import xml.etree.ElementTree as ET

import openmc

os.environ["OPENMC_CROSS_SECTIONS"] = '/share/h5_files/cross_sections.xml'


class TestH5FilesInSimulations(unittest.TestCase):

    def test_simulation_runs_with_isotopes(self):
        tree = ET.parse('/share/h5_files/cross_sections.xml')
        root = tree.getroot()   

        for elem in root:
            if elem.attrib['type'] == 'neutron':
                isotope_name = elem.attrib['materials']

                sett = openmc.Settings()
                sett.batches = 100
                sett.inactive = 0
                sett.particles = 500
                sett.run_mode = 'fixed source'

                mat1 = openmc.Material(1, "pure_isotope_material")
                mat1.add_nuclide(isotope_name, 1)
                mat1.set_density('g/cm3', 1)
                mats = openmc.Materials([mat1])

                outer_surface = openmc.Sphere(r=10, boundary_type='vacuum')
                cell1 = openmc.Cell(region=-outer_surface)
                cell1.fill = mat1
                universe = openmc.Universe(cells=[cell1])
                geom = openmc.Geometry(universe)

                model = openmc.model.Model(geom, mats, sett)
                model.run()

40 Lines

4 Lines

39 Lines

+ + =

Data storage repo - storage

Gigabytes of h5 files each with their own URL endpoint

https://github.com/openmc-data-storage/FENDL-3.1d/blob/main/h5_files/neutron/Ag107.h5?raw=true

+ + =

40 Lines

4 Lines

39 Lines

Data storage repo - openmc-data-downloader

Easy to install Python package

pip install openmc_data_downloader

https://github.com/openmc-data-storage/openmc_data_downloader

Knowledge of all the URLs of all the h5 files available.

Translation from user instructions to a list of h5 files to download

Internal data based that gets filtered by library / isotopes etc

{
  'element': 'Zn',
  'isotope': 'Zn66',
  'library': 'FENDL-3.1d',
  'local_file': 'FENDL-3.1d_Zn66.h5',
  'remote_file': 'Zn66.h5',
  'url': 'https://github.com/openmc-data-storage/FENDL-3.1d/raw/main/h5_files/neutron/Zn66.h5'
 },
 ...

Data storage repo - openmc-data-downloader

Command line usage

openmc_data_downloader --help

Data storage repo - openmc-data-downloader

openmc_data_downloader -l FENDL-3.1d -i Li6

Downloads one isotope

openmc_data_downloader -l TENDL-2019 -i Li6 Li7

Downloads two isotopes

openmc_data_downloader -l TENDL-2019 -e Li

Downloads one element

openmc_data_downloader -l TENDL-2019 -e Li Si Na

Downloads two elements

openmc_data_downloader -l ENDFB-7.1-NNDC -i Be9 -d my_h5_files

Downloads to a directory

openmc_data_downloader -l TENDL-2019 -e Li Si Na -i Fe56 U235

Downloads a combination of elements and isotopes

openmc_data_downloader -l TENDL-2019 -m materials.xml

Downloads all isotopes in a materials.xml file

openmc_data_downloader -l FENDL-3.1d TENDL-2019 -i Li6 U235

Downloads isotopes with library preference

Just in time library generator

import openmc
import openmc_data_downloader as odd

mat1 = openmc.Material(1, "breeder_mat")
mat1.add_element('Pb', 84.2, percent_type='ao')
mat1.add_element('Li', 15.8, percent_type='ao')
mat1.set_density('g/cm3', 11.3)
mats = openmc.Materials([mat1])

odd.just_in_time_library_generator(
    libraries='FENDL-3.1d',
    materials=mat1
)

Extra import

Material

Download h5 cross section files for the materials, accepts lists or individual items

Resulting download of h5 files and setting of OPENMC_CROSS_SECTIONS environmental varible

Re-running the same Python script and h5 downloading can be skipped as they already exist

Just in time library generator - in simulation

import openmc
import openmc_data_downloader as odd

mat1 = openmc.Material(1, "breeder_mat")
mat1.add_element('Pb', 84.2, percent_type='ao')
mat1.add_element('Li', 15.8, percent_type='ao')
mat1.set_density('g/cm3', 11.3)
mats = openmc.Materials([mat1])

outer_surface = openmc.Sphere(r=500, boundary_type='vacuum')
cell1 = openmc.Cell(region=-outer_surface)
cell1.fill = mat1
universe = openmc.Universe(cells=[cell1])
geom = openmc.Geometry(universe)

sett = openmc.Settings()
sett.batches = 100
sett.particles = 500
sett.run_mode = 'fixed source'

source = openmc.Source()
source.space = openmc.stats.Point((0, 0, 0))
source.angle = openmc.stats.Isotropic()
source.energy = openmc.stats.Discrete([14e6], [1])
sett.source = source

odd.just_in_time_library_generator(
    libraries='FENDL-3.1d',
    materials=mat1
)

model = openmc.model.Model(geom, mats, sett)
model.run()

Extra import

Materials

CSG

Settings

Point source

Download h5 cross section files for the materials

Run the simulation

Scale using repository templates

Less time spent making a repo for new nuclear data

Architecture + customization = new nuclear data repo

Live demo

~ 1 min human time

Building blocks

OpenMC Python API

Openmc-dev data repository

Automated processing and testing via CI

Downloading of reproducible collections of isotopes

+URL endpoints for ACE and ENDF libraries

+ Run commands and containerized enviroment

+URL end points for h5 files and libraries

IncidentNeutron, from_endf and from_ace

convert*.py and generate*.py

docker build, docker run and pytest

Summary

There are several methods of obtaining the h5 cross section files needed to run OpenMC simulations.

Each method has a specific use case.

OpenMC.org - GUI, script, entire libraries, h5 data
OpenMC-dev data repo - scripts for processing libraries
openmc-data-downloader - reproducible mixed micro libraries

Future work

Ability to specify particle type to allow photons
Ability to specify temperature