How to generate events with NEUT and feed them to NUISANCE

This page aims to outline how to use NEUT to create a sample of interactions on your favourite target. We'll also outline how to feed this generated sample through the NUISANCE framework and get good results.

Contact: ​Clarence Wret

You can also try: ​Patrick Stowell ​Callum Wilkinson, ​Luke Pickering, or ​Hayato-san.

Compiling NEUT and running neutroot2

This guide assumes you've got your hand on a copy of NEUT and CERNLIB. To request a copy, please kindly inquire to Hayato-san at hayato@icrr.u-tokyo.ac.jp.

The main neutrino-nucleus interaction generator lives in $NEUT_ROOT/src/neutsmpl/neutroot2, which is the compiled program with source $NEUT_ROOT/src/neutsmpl/neutroot.F.

NEUT has two dependencies: ROOT and CERNLIB. You need to have $ROOTSYS, $CERN and $CERN_LEVEL environment variables set-up to compile NEUT. These should all be specified by the user in $NEUT_ROOT/src/neutsmpl/EnvMakeneutsmpl.csh.

Compilation is done by running ./Makeneutsmpl in $NEUT_ROOT/src/neutsmpl/: this sources the above EnvMakeneutsmpl.csh, runs through a host of makefiles, sets up symlinks and builds the project.

Running the main executable is done by ./neutroot2 NEUT_CARD_FILE.card OUTPUT_ROOT_FILE.root, where NEUT_CARD_FILE.card is a card file compliant with your chosen NEUT version, and OUTPUT_ROOT_FILE.root is the ROOT output file where the events are to be saved. This output file contains all the event information and is the input for NUISANCE analyses.

Examples of NEUT card files can be found in the $NEUT_ROOT/src/neutsmpl/Cards directory, or have a look at what team NUISANCE used for 2016 analyses at Winter2016 and at our ​box.

Hot tip: NEUT can be quite verbose and you might want to pipe the output to file (or /dev/null...) rather than dump it to screen. This can actually speed up NEUT considerably.

Structure of card files:

NEUT has a load of defaults set on run-time if the user does not supply them, which means you're running with "reference" NEUT.

Selecting different models is done in the cardfile, and all the inputs that a user can specify is found in src/neutcore/necard.F.

We can specify how many events we want, the input flux in TH1D format, the incoming neutrino type, the flux file, number of bound protons and neutrons and free protons, what interaction modes, and so on.

Here's a list for generating all modes on T2K H2O. The "C" at the start of each line is just a comment and will not be parsed by NEUT.

C Number of events ; EVCT-NEVT
EVCT-NEVT  100000
C
C Particle Code ; EVCT-IDPT
EVCT-IDPT 14
EVCT-MPOS 2
EVCT-RAD  100.
EVCT-MDIR 1
EVCT-DIR 0. 0. 1.
C
EVCT-MPV 3
EVCT-FILENM  'nuisance_fluxes/t2kflux_2016_plus250kA.root'
EVCT-HISTNM  'enu_nd280_numu'
EVCT-INMEV 0
C
C **** TARGET INFORMATION ****
NEUT-NUMBNDN 8
NEUT-NUMBNDP 8
NEUT-NUMFREP 2
NEUT-NUMATOM 16
C
C **** WHAT INTERACTION MODES DO WE WANT? (0 = ALL) ****
NEUT-MODE   0 
C **** WHAT CCQE MODEL ARE WE USING (2 = RFG WITH BBBA05) ***
NEUT-MDLQE 2
C
C **** RANDOM NUMBER AND IF WE WANT NEUT TO BE QUIET ****
NEUT-RAND 1
NEUT-QUIET 2

The EVCT-MPV 3 says the flux is coming from a file, EVCT-FILENM specifies the file location of the flux, EVCT-HISTNM is the name of the TH1D inside EVCT-FILENM, and EVCT-INMEV specifies if the input histogram is in MeV or not.

NEUT inputs for NUISANCE

The output of neutroot2, contains a ROOT file containing nevents neutvect objects which has all the information about each event in it. This is read in by NUISANCE to generate predictions. neutroot2 also saves the predicted flux (flux_numu) and event rate (evtrt_numu) predictions automatically into the output file so NUISANCE already has everything it needs.

To read NEUT inputs the NEUT type just needs to be specified in the NUISANCE input card file, e.g. for MiniBooNE CCQE

sample MiniBooNE_CCQE_XSec_1DQ2_nu  NEUT:/path/to/neut/input/file.root

Using NEUT ReWeight

NUISANCE has full support of running NEUT with its reweighting engine. We can generate error-bands, perform generator tunings to specific data-sets, and so on all through NUISANCE.

Compiling the reweight classes in NEUT

To compile NEUT's reweighting engine, make sure you've got the main executables compiled (see above), and

$ cd $NEUT_ROOT/src/reweight
$ make clean
$ make all

to compile the reweighting engine.

Running NUISANCE with NEUT ReWeight

We recommend talking to Hayato-san or us (or other NEUT experts) before fitting parameters to make sure that they are not deprecated.

The syntax for running NEUT with reweighting in NUISANCE is very similar to how it's done in GENIE and NuWro too.

The user specifies the parameter(s) and the range which should be varied as:

neut_parameter  MaCCQE  0.0  -4.0  4.0  1.0  FREE

where MaCCQE is the parameter to be varied, 0.0 is the starting values, -4.0 is the minimum value, 4.0 is the maximum value, 1.0 is the step-size to take, and FREE means the parameter is free to vary in the minimisation.

We can also specify FIX for the parameters if we want to reweight a systematic parameter to some value other than nominal and keep it there. This is useful for using best-fit parameter constraints from one experiment to predict another experiment (e.g. fit to MiniBooNE and predict MINERvA data, or fit to bubble-chamber data to predict T2K data).

The whole list of parameters can be found in $NEUT_ROOT/src/reweight/NSyst.h and the returned std::strings (e.g. MaCCQE and CA5RES).

The starting, minimum, maximum and step values are all in units of 1-sigma uncertainties. So -4.0 for the minimum means -4*1-sigma ranges, and 0.0 for the starting value means the nominal value nominal.

The hard-coded 1-sigma uncertainty defaults for NEUT can be found in $NEUT_ROOT/src/reweight/NSystUncertainity.cc and the NSystUncertainty::SetDefaults(void) function. These are in the physical units (e.g. MaCCQE has +/- 0.165 GeV uncertainty in NEUT for 1 sigma)

The hard-coded nominal defaults for NEUT can be found in $NEUT_ROOT/src/reweight/NFortFns.cc. These are also the physical units of the parameter (e.g. MaCCQE default is 1.21 GeV in NEUT).

The actual physical parameter value is found by

par_value = nominal * (1 + par_var * 1-sigma)

The output of the fit is in the units of par_var, not par_value. So an output for MaCCQE of 0.7 would equate to 1.21*(1+1*0.165) = 1.35 GeV.

Joint Flux Inputs

neutroot2 does not yet natively support combined beams or targets so if you want to use the measurement classes that require this (e.g. MINERvA CC0pi nue+nuebar) you will have to setup a JOINT input. See this? for more information on how to achieve this.