Delphes behaviour with displaced vertices

[Nishita Desai]

We’re trying to understand the limits from a standard ATLAS 0lepton+jets+MET analysis for a particular model that predicts displaced vertices. The issue is that we find a very reduced efficiency even though there is no explicit veto on displaced vertices in the analysis.

To understand what is going on, we wrote a simple Pythia code with jet reconstruction and smeared the jet pT. We can reproduce the cutflow for the ATLAS benchmark both with Delphes (as a part of CheckMATE) and with our simple Pythia code.

The trouble arises when we try to run the events for our model with displaced vertices. Here, our simple code gives a much higher efficiency as compared to Delphes. We don’t understand what may be causing this discrepancy. We also assume a cylindrical box to not count anything that decays outside the detector.

Does anyone know of any DELPHES feature that is responsible for us losing so many events?

Thanks for you time.

[Pavel Demin]

Could you please send us some events that we could use to reproduce this problem?

[Nishita Desai]

HepMC file with 100 events (@ 8 TeV)

[Nishita Desai]

Replying to pavel:

Could you please send us some events that we could use to reproduce this problem?

Thanks for your quick reply.
I've attached 100 events in HepMC format.

[Pavel Demin]

Thanks for the events. I've just tried to run them through the default ATLAS card and noticed that the output ROOT file contains 26 isolated electrons and 36 isolated muons. Is it expected? What is your 0lepton selection?

[Nishita Desai]

Replying to pavel:

Thanks for the events. I've just tried to run them through the default ATLAS card and noticed that the output ROOT file contains 26 isolated electrons and 36 isolated muons. Is it expected? What is your 0lepton selection?

This is not unexpected. The selection cuts on events are applied later, this sample contains all events (since I did not want to do any detector simulation for isolation before passing the events to you.)

I am attaching a second file here for events that pass the 6jt signal region for our simplified simulation, but most do not pass the same with Delphes.

[Nishita Desai]

Events passing 6jt signal region of ATLAS analysis ​http://arxiv.org/abs/1405.7875

[Pavel Demin]

I've just checked the new file and the article. Can't see anything suspicious like this. Probably, the effects that you observe can't be seen without redoing the analysis.

About 1% of particles are rejected because they are outside of the cylindrical volume defined in the particle propagation module.

Is there any particular event selection criteria that is more efficient with your code than with Delphes?

[Nishita Desai]

I checked the cut flow and it seems we have a large mismatch in the first (trigger) cut itself: MET > 160 GeV; (jet1.pt. jet2.pt) > (130 , 60) GeV.

Here are the comparisons:

Cut        Py8       Delphes
MET+J1,2  :0.64400   0.179784
J3        :0.63140   0.173097
J4        :0.60060   0.163517
J5        :0.55200   0.14705
J6        :0.46760   0.123299
dPhi      :0.34080   0.0903995
dPhi2     :0.21120   0.0557827
MET/ht    :0.16840   0.00792746
Final     :0.05920   0.00419143