sPHENIX discussion of software

[Aaron Angerami <angerami AT cern.ch>]

Hi Dennis,

Regarding the self-energy bias. 

-I agree that we should re-run with a lower threshold. Reporting a JES/JER 
for a response distribution that is double-peaked is kind of pointless and a 
bit misleading, since those numbers don’t really quantify the shape. 

- 15 GeV is definitely too high. This is especially bad because we aren’t 
applying a JES calibration, thus the jets that pass the 15 GeV cut probably 
correspond to jets with pT 25 or 30 GeV on average, which is too high even 
for LHC-like conditions. 

-This lack of calibration probably explains why you see a significant 
contribution of biased-jets out to > 30 GeV. It also points to the fact that 
if we want to use an iterative procedure, we really probably need a JES 
calibration to make it work properly.

-The optimal choice of this number is going to be driven both by the size of 
the UE fluctuations, the calorimetric response to soft particles at the “EM 
scale” and a tradeoff between the hard scattering rate for jets at low pT and 
how low you want to measure. 

-The way to do this right is to study the fluctuation spectrum. We will need 
to do this eventually to understand the JER. In the meanwhile I would say 
that if we want to be able to measure jets down to 10-15 GeV, and assuming 
the EM-scale response is ~0.5 down there, then we want to pick this threshold 
between 5-8 GeV. 

-One thing we can do right away is make the pT spectrum for jets in central 
Au+Au. Instead of matching to truth jets, you could make the spectrum that 
includes fake jets (normalized per event), and then try to gain some 
intuition for the width of this distribution. I would just do this for one of 
your jet pTmin slices since applying weights to the fake jets will screw up 
their rates. To understand this it would also be good to know whether the 
HIJING has been restricted in how high in pT it can produce hard scatterings. 
This study gives us a feel for the likelihood that a fake jet will get picked 
up as a seed, which is the reason we don’t take the threshold down to ~zero.

Aaron






> On Apr 18, 2018, at 4:18 PM, Perepelitsa, Dennis <dvp AT bnl.gov> wrote:
> 
> Hi all,
> 
> As promised in the Jet Structure meeting on Monday, please find some slides 
> with a mid-week update.
> 
> Key points: 
> 
> 1. The strange JER in pp collisions arises from an aspect of the HI-style 
> jet reconstruction. The minimum seed jet pT, which essentially sets how 
> permissive is the identification of jet-like regions to exclude from the 
> background determination, was chosen to be quite large. This can result in 
> a “self-energy” bias for low-pT jets, in which their energies are included 
> in the background determination, causing a systematically low response. 
> This resulted in a double-peak structure in the response distributions, 
> first noticed by Jeff, arising from jets which do / do not have a seed jet 
> nearby. 
> 
> While in Au+Au events, it makes sense to have a somewhat high minimum seed 
> jet pT (so that one does not accidentally exclude background fluctuations — 
> although just where to set it is still an open question), this is clearly 
> too high for p+p collisions. I will rerun some p+p MC with a lower seed jet 
> pT threshold, although that will mean that we SHOULD expect some difference 
> in the mean pT response between pp and Au+Au at low-jet-pT.
> 
> I’d be particularly happy for input on this point.
> 
> 2. I have updated the dijet xJ plot with full statistics, and I have made 
> two versions: one with small cone size and high-pT where there is no 
> appreciable centrality-dependence in the reco-level distribution, and one 
> with large cone size / small-pT where there is. This followed from a nice 
> suggestion by Aaron (who also had other suggestions I have not yet gotten 
> to).
> 
> 3. The photon+jet simulations have finished and Kurt Hill, a student at 
> Colorado, has made some xJg distributions in the same spirit as the dijet 
> xJ distributions, again showing a case where one can / cannot observe a 
> centrality-dependence in the reco-level distribution.
> 
> Happy for any feedback — will keep working towards the presentation of 
> these at the General Meeting,
> 
> Dennis 
> 
> Dennis V. Perepelitsa
> Assistant Professor, Physics Department
> University of Colorado Boulder
> 
> 
> <dvp-sPHENIX-update-4-18-18.pdf>_______________________________________________
> sPHENIX-jet-structure-l mailing list
> sPHENIX-jet-structure-l AT lists.bnl.gov
> https://lists.bnl.gov/mailman/listinfo/sphenix-jet-structure-l