Reading Neutrino-Generator Files

Correctly interpreting the weighting information for neutrino-generator can be difficult for a number of reasons. The icetray modules for Neutrino-Generator have a lot of options for generating primary neutrinos using a number of different generation surfaces. The scope of SimWeights is limited to configuration which have been used by simulation production and not try to attempt compatibility with every single configuration option which is available in neutrino-generator. SimWeights has been tested to work with all simulation production datasets going back to at least datasets 10634 which was produced in 2014. Reading older dataset is possible but will require additional work.

The primary way to weight neutrino-generator data is with the function simweights.Weighter which will inspect the I3MCWeightDict of the given file and construct the correct GenerationSurface and return an instance of simweights.Weighter.

How SimWeights reads I3MCWeightDict

The probability that a given neutrino will interact in the simulation volume was originally called TotalInteractionProbabilityWeight but was renamed TotalWeight in version 6 of simulation. SimWeights will use whichever column is present.

The biggest change in neutrino-generator that affected weighting was the change in the zenith distribution. In version 3 and before of simulation, the cosine of the zenith angle was drawn from a uniform distribution and the position of the particle was drawn from a circle perpendicular to the primary’s momentum. In version 4 this was changed to generate events on the surface of a cylinder with an ad hoc zenith distribution to account for a deficit of vertical neutrinos in certain analyses. The since neutrino generator includes the zenith weight in TotalWeight and TotalInteractionProbabilityWeight simweights takes a shortcut an just uses simweights.UniformSolidAngleCylinder for the zenith distribution and lets the factor in TotalWeight handle the zenith weighting. When this change was first implemented the size of the cylinder the events were generated on was hard coded to have a height of 1900 m and a radius of 950 m, the height and radius of the cylinder were added to I3MCWeightsDict in V6.

All neutrino-generator datasets have been produced with equal number of neutrinos and antineutrinos. Newer versions of simulation have save the fraction of the type in TypeWeight but this is assumed to be 0.5 in files where it is not present.

Event though neutrino-generator saves the minimum and maximum azimuth angle, for simplicity sake, SimWeights just assumes events were generated on all \(2\pi\) azimuth.

Note

If you want to use SimWeights to weight custom produced neutrino-generator data which does not conform to the weighting scheme used here, it would probably be easier to call simweights.Weighter() directly rather than attempting to shoehorn it into simweights.NuGenWeighter().

The complete list of parameters saved by neutrino-generator in the I3MCWeightDict are documented in the icetray documentation. The columns used by SimWeights are listed in the table below:

Name

Notes

MinZenith

MaxZenith

InjectionSurfaceR

before V04-01-00,

CylinderHeight

after V06-00-00, if not present 1900 is assumed

CylinderRadius

after V06-00-00, if not present 950 is assumed

MinEnergyLog

MaxEnergyLog

PowerLawIndex

PrimaryNeutrinoType

TypeWeight

Assumed to be 0.5 if not found

PrimaryNeutrinoEnergy

PrimaryNeutrinoZenith

TotalWeight

After V06-00-00

TotalInteractionProbabilityWeight

Before V06-00-00, replaced by TotalWeight

Why SimWeights doesn’t use OneWeight

The way most people choose to weight neutrino-generator datasets is to simply multiply the OneWeight by the flux an divide by the number of files. This works fine when you are limited to a single dataset and is exactly what OneWeight was designed to do. The issue is that OneWeight includes all the information in a way which is difficult to combine multiple datasets with different generation surfaces. SimWeights takes a different approach and tries to construct the correct representation of the generation surface.

Note

If you are happy with your existing setup using OneWeight to weight neutrino-generator data for uniform datasets, then you are under no obligation to switch to SimWeights.