gmapy is a Python package to produce nuclear data evaluations of cross sections including uncertainties based on the combined data from various experiments.

This package can be regarded as a modernized version of the GMAP code developed by Wolfgang P. Poenitz, which was employed for the evaluation of the neutron data standards. A number of test cases have been created to prove that the latest neutron standards evaluation can be reproduced with gmapy. To learn more about improvements of gmapy over GMAP, please see the respective section below.

This package is still evolving and not well documented yet. To get a first impression of the usage of gmapy, you can inspect the Jupyter notebook examples/example-000-basic-usage.ipynb. You can directly open this notebook in the browser by clicking here.

gmapy depends on the SuiteSparse library. With appropriate build tools, it can probably be installed on Windows but this has not been tested. On Linux it can be installed by

sudo apt install libsuitesparse-dev

and on MacOs via

brew install suite-sparse

It is also important that the source and library files of SuiteSparse can be found by the compiler and linker. If the installation instructions for gmapy in the next section fail with the error message that cholmod.h cannot be found, execute the following instructions with appropriate paths:

export SUITESPARSE_INCLUDE_DIR=/opt/local/include
export SUITESPARSE_LIBRARY_DIR=/opt/local/lib

The given paths are examples of commonly used locations and may need to be adjusted on your system.

We recommend to create a virtual environment with Python version 3.9. With conda you can create a new environment by

conda create -y -n gmapy python=3.9 pip

Then activate the environment (conda activate gmapy) and install the gmapy package:

pip install git+https://github.com/iaea-nds/gmapy.git

Here are the capabilities of gmapy at present that go beyond those of the Fortran GMAP code:

• Exact maximum likelihood optimization based on a custom version of the Levenberg-Marquardt algorithm with convergence checking.
• Adaptive numerical integration by Romberg's method with convergence checking for a precise computation of spectrum averaged cross sections.
• Indidvidual energy meshes for different reactions are possible to account for reaction-specific structures.
• Experimental data need not be aligned perfectly to the the computational mesh thanks to the use of linear interpolation.
• A new data type ratios of spectrum averaged cross sections is available.
• Data can be read in the database format used by GMAP as well as in a new JSON format.
• Unknown uncertainties can be introduced and estimated during the estimation of the cross sections and their uncertainties via maximum likelihood estimation.

This code is distributed under the MIT license, see the accompanying license file for more information.

Copyright (c) International Atomic Energy Agency (IAEA)