artefacts in (b) jet eta distribution w/ high statistics and binning

[Fabian Bach]

Running Delphes over single top s channel samples (i.e. evbb parton level final states showered with Pythia 6.4) with high statistics (2M events), I observe systematic fluctuations in the eta distributions of the jets when choosing high enough binning (e.g. 0.05/bin), cf. histograms inside the attached ROOT files. I checked that parton and truth level (jets reconstructed from truth level showered particles, Delphes v2.0.0) distributions are OK, which indicates that the jet algorithms are not the problem. Nonetheless, I tried SIScone and anti-kt algorithms with cone radii of 0.4 resp. 0.7 in Delphes v2.0.0 and v3.0.5, always getting these artefacts at detector level. The histograms in the ROOT files are named "eta_" + "j" (Btag==0) / "b" (Btag==1) + "1" (hardest jet) / "2" (2nd jet) / "" (sum). I used the default ATLAS detector cards, and only edited the jet algorithm/cone radius as well as b-tagging efficiency.

[Pavel Demin]

We're trying to address a similar problem by smearing the calorimeter tower position in eta and phi.

Could you, please check if the following prerelease is working better for you?

​http://cp3.irmp.ucl.ac.be/downloads/Delphes-3.0.6.pre1.tar.gz

[Fabian Bach]

Thanks for the prerelease, I now get smooth eta distributions in the central region < 2.5, but with a new additional broad dip around 1.5 and a sharp peak at 0, while the "zigzag" remains for > 2.5 (cf. ROOT file attached). Please excuse my lack of profound knowledge in jet calorimetry, but is this what's supposed to come out?

[Pavel Demin]

Normally, you should obtain something that more or less resembles the MC truth jets with some effects due to detector resolution. So, I would say that the eta distributions that you obtained with Delphes 3.0.5 are OK. I don't think that the irregularities in the eta and phi distributions should have any effect on the final results.

I've just noticed that I introduced a bug while adding the smearing of the calorimeter tower position. Could you, please try another prerelease:

​http://cp3.irmp.ucl.ac.be/downloads/Delphes-3.0.6.pre2.tar.gz

[Pavel Demin]

We've just discovered another problem with the 3.0.6.pre2 prerelease and fixed it in the new prerelease:

​​http://cp3.irmp.ucl.ac.be/downloads/Delphes-3.0.6.pre3.tar.gz

[Fabian Bach]

Thanks again for this new version, results are looking fine now, at least to my eye. Are there any further bugfixes imminent? Can you already tell when v3.0.6 will be released officially?

[Pavel Demin]

Thanks for all the tests. The eta distributions look indeed better now.

We'll try to release this new version at the beginning of next week.

[Pavel Demin]

Delphes 3.0.6 is released.