Sensitivity Reader


Data files, like the one used in this example, are not included with the python distribution. They can be downloaded from the GitHub repository, and accessed after setting the SERPENT_TOOLS_DATA environment variable

>>> import os
>>> sfile = os.path.join(
...     os.environ["SERPENT_TOOLS_DATA"],
...     "flattop_sens.m")

As of SERPENT 2.1.29, the ability to compute sensitivities using Generalized Perturbation Theory [gpt]. An overview of the functionality, and how to enable these features is included on the SERPENT wiki - Sensitivity Calculations. Sensitivity calculations produce _sens.m or _sensN.m files, depending on the version of SERPENT, and contain a collection of arrays and indexes, denoting the sensitivity of a quantity to perturbations in isotopic parameters, such as cross sections or fission spectrum. These perturbations can be applied to specific materials and/or isotopes.

The SensitivityReader is capable of reading this file and storing all the arrays and perturbation parameters contained therein. A basic plot method is also contained on the reader.

>>> import serpentTools
>>> sens =

The arrays that are stored in sensitivities and energyIntegratedSens are stored under converted names. The original names from SERPENT are of the form ADJ_PERT_KEFF_SENS or ADJ_PERT_KEFF_SENS_E_INT, respectively. Since this reader stores the resulting arrays in unique locations, the names are converted to a succinct form. The two arrays listed above would be stored both as keff in sensitivities and energyIntegratedSens. All names are converted to mixedCaseNames to fit the style of the project.

These arrays are quite large, so only their shapes will be shown in this notebook.

>>> print(sens.sensitivities.keys(), sens.energyIntegratedSens.keys())
dict_keys(['keff']) dict_keys(['keff'])
>>> print(sens.sensitivities['keff'].shape)
(1, 2, 7, 175, 2)
>>> print(sens.energyIntegratedSens['keff'].shape)
(1, 2, 7, 2)

The energy grid structure and lethargy widths are stored on the reader, as energies and lethargyWidths.

>>> print(sens.energies.shape)
>>> print(sens.energies[:10])
[1.00001e-11 1.00001e-07 4.13994e-07 5.31579e-07 6.82560e-07 8.76425e-07
1.12300e-06 1.44000e-06 1.85539e-06 2.38237e-06]
>>> print(sens.lethargyWidths.shape)
>>> print(sens.lethargyWidths[:10])
[9.21034  1.42067  0.25     0.249999 0.250001 0.247908 0.248639 0.253452
0.250001 0.249999]

Ordered dictionaries materials, zais, and perts contain keys of the names of their respective data, and the corresponding index, iSENS_ZAI_zzaaai, in the sensitivity arrays. These arrays are zero-indexed, so the first item will have an index of zero. The data stored in the sensitivities and energyIntegratedSens dictionaries has the exact same structure as if the arrays were loaded into MATLAB/Octave, but with zero-indexing.

>>> print(sens.materials)
OrderedDict([('total', 0)])
>>> print(sens.zais)
OrderedDict([('total', 0), (922380, 1)])
>>> print(sens.perts)
OrderedDict([('total xs', 0), ('ela scatt xs', 1), ('sab scatt xs', 2), ('inl
scatt xs', 3), ('capture xs', 4), ('fission xs', 5), ('nxn xs', 6)])


The SensitivityReader has a plot() method for visualizing the sensitivities.


Without additional arguments, other than the name of the array, the plot() method will plot all permutations of materials, isotopes, and isotope perturbations present. This can lead to a very busy plot and legend, so it is recommended that additional arguments are passed.

>>> sens.plot('keff');

The following arguments can be used to filter the data present:




Isotopes(s) of interest


Perturbation(s) of interest


Material(s) of interest

The sigma argument can be used to adjust the confidence interval applied to the plot. The labelFmt argument can be used to modify the label used for each plot. The following replacements will be made:

1. {r} - name of the response being plotted 1. {m} - name of the material 1. {z} - isotope zai 1. {p} - specific perturbation

>>> ax = sens.plot('keff', 922380, mat='total', sigma=0,
...                labelFmt="{r}: {z} {p}")
>>> ax.set_xlim(1E4);  # set the lower limit to be closer to what we care about

The argument normalize is used to turn on/off normalization per unit lethargy, while legend can be used to turn off the legend, or set the legend outside the plot.

>>> ax = sens.plot('keff', 922380, mat='total', sigma=0,
...                labelFmt="{r}: {z} {p}", legend='right')
>>> ax.set_xlim(1E4);  # set the lower limit to be closer to what we care about
>>> sens.plot('keff', zai='total', pert=['total xs', 'fission xs'], labelFmt="{z} -  {p}",
...           legend='above', ncol=2, normalize=False)
>>> pyplot.xlim(1E4, 1E8);


The SensitivityReader can quickly read sensitivity files, and stores all data present in the file. A versatile plot() method can be used to quickly visualize sensitivities.


Aufiero, M. et al. “A collision history-based approach to sensitivity/perturbation calculations in the continuous energy Monte Carlo code SERPENT”, Ann. Nucl. Energy, 152 (2015) 245-258.