Distribution of transvere W boson mass not modelled correctly by Delphes 3.4.0

[Soeren Stamm]

Hi,

I have generated single-top s-channel events (pp > t b~) with MadGraph5 2.4.2 and stored them in LHE files. These files are provided to Delphes 3.4.0 and showered/decayed by Pythia 8.219 using the executable "DelphesPythia8".

I found that the distribution of the transverse mass of the W boson (mtW) from the top decay is not modelled as expected. I did some research and was able to get a good modelling of the mtW distribution with Delphes 3.3.3 (and lower).

I attached a plot of mtW for Delphes 3.4.0 (red line) compared to Delphes 3.3.3 (blue line). The blue line is in agreement with my particle level study using Rivet. I also attached distributions for lepton pt (e/mu) and missing transverse energy and jet pt (highest pt jet only).

The content of the attached pdf is: page 1 - lepton pt, p.2 - jet pt, p.3 - mtW, p.4 - missing transverse energy.

What was changed between version 3.3.3 and 3.4.0 that could cause this odd looking shape of the mtW distribution?

Cheers,
Soeren

[Soeren Stamm]

Distributions for lepton pt, jet pt, mtW and Et miss

[Soeren Stamm]

I am using the ATLAS card "delphes_card_ATLAS.tcl".

[Michele Selvaggi]

Hi Soren,

thanks for spotting this. Can you please attach a small (~2k evts) LHE event sample and the root macro that produces the Mtw plot?
I can then have a look at when things started to go wrong.

Thanks
Michele

[Soeren Stamm]

Script for creating mtW plot

[Soeren Stamm]

2000 events MG5 pp > tb~ + pp > t~b

[Soeren Stamm]

2000 events MG5 pp > tb~ + pp > t~b

[Soeren Stamm]

Hi Michele,

thanks for looking into this.

I attached a root script "DelphesFastAnalysis.C" that you can use to analyse Delphes root files. I usually compile it using the ROOT cint interpreter via ".L DelphesFastAnalysis.C+" and then run it via DelphesFastAnalysis("input.root", "output.root").
Please let me know if you have trouble to run it. I am using ROOT 5.34.

I also attached 2x 2000 evts for pp > tb~ + pp > t~b.

Cheers,
Soeren

Replying to mselvaggi:

Hi Soren,

thanks for spotting this. Can you please attach a small (~2k evts) LHE event sample and the root macro that produces the Mtw plot?
I can then have a look at when things started to go wrong.

Thanks
Michele

[Michele Selvaggi]

Hi Soeren,

Given your final state the MT distribution is going to be very sensitive to which MET and isolation definiton are adopted and both these things have changed from the card in 3.3.3 and 3.4.0. Unfortunately the signal selection is very small so I can't really see if the changes I propose are really effective.
You'll have to try the following with larger stats:

Replace the inputArray to MissingET by "set InputArray EFlowMerger/eflow"

This will put back leptons in the MissingET definition (with EFlowMerger/eflow they are excluded)

Another thing you can try is putting back leptons in the isolation definition. So in every module XXXIsolation (where XXX=Electron/Photon/Muon) you can try to replace the IsolationInputArray EFlowFilter/eflow by EFlowMerger/eflow.

These changes should get you close to the card in 3.3.3.

Cheers
Michele

[Soeren Stamm]

Hi Michele,

I tried two different setups:

1st setup) I replaced the input array Calorimeter/towers with EFlowMerger/eflow in

module Merger MissingET {
# add InputArray InputArray
# add InputArray Calorimeter/towers

set InputArray EFlowMerger/eflow
set MomentumOutputArray momentum

}

I left the isolation as it was (i.e. so still using "set IsolationInputArray EFlowFilter/eflow")

This gives the expected mtW distribution. Very good!

However, I don't understand why this works... (I just started a week ago to look into Delphes)

• The EFlowMerger/eflow are made from HCal/eflowTracks, ECal/eflowPhotons, HCal/eflowNeutralHadrons.

I don't see electrons here?

Are they included via HCal/eflowTracks, i.e. the four-momentum of all electrons are included through their tracking information contained in the tracks? HCal/eflowTracks would then also include all charged particle track momenta (i.e. pions, etc.)?

• The Calorimeter/towers are made from ECal/ecalTowers, HCal/hcalTowers.

Shouldn't this include contributions from electrons and photons to the ecalTowers?

Is there any documentation on how the modules work?
I would like to optimize the ATLAS card to get better results with my analysis. For this I would like to understand the code a bit better.

2nd setup) I replaced in all XXXIsolationModules the lines:

set IsolationInputArray EFlowFilter/eflow
by
set IsolationInputArray EFlowMerger/eflow

but kept the Calorimeter/towers in the MissingET computation. Only changing the isolation, does not fix the problem.

Thank you, cheers,
Soeren

Replying to mselvaggi:

Hi Soeren,

Given your final state the MT distribution is going to be very sensitive to which MET and isolation definiton are adopted and both these things have changed from the card in 3.3.3 and 3.4.0. Unfortunately the signal selection is very small so I can't really see if the changes I propose are really effective.
You'll have to try the following with larger stats:

Replace the inputArray to MissingET by "set InputArray EFlowMerger/eflow"

This will put back leptons in the MissingET definition (with EFlowMerger/eflow they are excluded)

Another thing you can try is putting back leptons in the isolation definition. So in every module XXXIsolation (where XXX=Electron/Photon/Muon) you can try to replace the IsolationInputArray EFlowFilter/eflow by EFlowMerger/eflow.

These changes should get you close to the card in 3.3.3.

Cheers
Michele