22 january 2009 david1 look at dead material and fake met in jx samples mc08 10 tev simulations,...
DESCRIPTION
22 January 2009 David3 J1 Top left: no correlation between dead material ‘all’ and the true eta of the lead jet Top right: just 5 events out of many tens of thousands has a Delta MET exceeding 5 sigma Bottom left: nice positive correlation between dead material ‘all’ and the error on the missing ET Bottom right: somewhat weaker, but still positive correlation between the crack1 dead material and the error on METTRANSCRIPT
22 January 2009 David 1
Look at dead material and fake MET in Jx samples
• mc08 10 TeV simulations, release 14.2.20.3• J0 to J6 are tag s479_r586, ‘ideal geometry’ samples• The available J7/J8 samples are s479_r563, hence not the
ideal geometry they’re supposed to be, but they suffer from using offline conditions for misaligned ID, by mistake– Are being redone right now, first look at jet resolution in J7 and J8
shows large tails and unhealthy distributions, hence take these two samples in the following with a grain of salt
• Looked at the dead material variables in the AOD, the ones labeled ‘All’, which is Crack1 + Crack2 + Cryo, and ‘Crack1’, which is 1.1 < fabs(eta) < 1.7– ‘Crack2’ would be 2.9 < fabs(eta) < 3.5, ‘Cryo’ is the barrel cryostat
22 January 2009 David 2
J0 • Top left: no correlation between dead material ‘all’ and the true eta of the lead jet
• Top right: just 1 event out of many tens of thousands has a Delta MET exceeding 5 sigma for the given event Sigma ET
• Bottom left: nice positive correlation between dead material ‘all’ and the error on the missing ET
• Bottom right: somewhat weaker, but still positive correlation between the crack1 dead material and the error on MET
Green bullets are always profile plots of the colored 2d histogram!
22 January 2009 David 3
J1 • Top left: no correlation between dead material ‘all’ and the true eta of the lead jet
• Top right: just 5 events out of many tens of thousands has a Delta MET exceeding 5 sigma
• Bottom left: nice positive correlation between dead material ‘all’ and the error on the missing ET
• Bottom right: somewhat weaker, but still positive correlation between the crack1 dead material and the error on MET
22 January 2009 David 4
J2 • Top left: correlation starts to appear, profile plot peaks at eta 1.5 and 3.2
• Top right: still not many events with large fake MET
• Bottom left: nice positive correlation between dead material ‘all’ and the error on the missing ET
• Bottom right: somewhat weaker, but still positive correlation between the crack1 dead material and the error on MET
22 January 2009 David 5
J3 • Top left: profile plot shows clear correlation between dead material and eta lead jet – good
• Top right: we start to see the detector gaps in the eta of the poorer measured of the 2 lead jets when fake MET is large
• Bottom left: nice positive correlation Bottom right: somewhat weaker, but still positive correlation
22 January 2009 David 6
J4 • Top left: eta 1.5 and 3.2 are clearly seen in the profile plot
• Top right: sharp peak at eta 3.2 for events with large fake MET, 1.5 also suggested
• Bottom left: nice positive correlation Bottom right: somewhat weaker, but still positive correlation
22 January 2009 David 7
J5 • Top left: jets start to get more central, only the first crack really appears
• Top right: prominent peaks at eta 0 and 1.5 show how jets in this area cause large fake MET
• Bottom left: nice positive correlation Bottom right: nice positive correlation
22 January 2009 David 8
J6 • Top left: again jets are rather central, crack 1 nicely visible
• Top right: jets hitting eta 0 and 1.5 cause large fake MET – as seen before, eta=0 has a problem (in the jet reco?)
• Bottom left: nice positive correlation
• Bottom right: nice positive correlation
22 January 2009 David 9
J7 • Top left: jets are more and more central of course, crack1 is still there, less pronounced
• Top right: large fake MET events show still that eta=0 is particularly bad, although it’s less clear here. Note, J7 and J8 have this misaligned ID problem
• Bottom left: correlation starts to disappear, these events are in general very busy
• Bottom right: also here, less and less correlation
22 January 2009 David 10
J8 • Top left: not much left to say given the really restricted eta range
• Top right: No special detector region visible in this plot
• Bottom left: no more correlation of dead material depot and Delta MET
• Bottom right: same here, no more correlation
22 January 2009 David 11
Conclusions about dead material and poor MET
• The correlation between the dead material depot and the error on MET shows nicely that at least for the 2 cracks the dead material hits are correctly collected
• The peak at 0 in the distribution of poorly measured lead jets, when fake MET is large, is still mysterious– There seemed to be some theory about it, probably need to check with
Ana
• Next we should probably investigate how to get rid of such events without truth variables, see also next slide
22 January 2009 David 12
Next to do
• Verify that correlation between crack1 / crack2 dead material and DeltaMET increases further if poorly measured lead jet hits crack1 or crack2, i.e. if true eta is around eta = 1.5 or eta = 3.2
• Plot reconstructed MET, true MET, reconstructed MET corrected for ET in dead material in the crack for lead jets hitting the cracks
– If all is consistent the true and corrected reconstructed MET should be almost identical– One could also see above which MET this fake MET from lost energy doesn’t play a role
any more• Compare true MET to reco MET without jets near eta=0 or near cracks, and to full
reco MET– This would tell us what kind of fake MET we are to expect from dead material, and what
the true irreducible MET level is in such events which we cannot get rid of– Scale with cross section to get a sense of the lumi, too
• Investigate eta=0 defficiency– Talking to Ana, the theory is that the LAr gap of a few millimetres at exactly eta=0 is the
cause– Could be longitudinal leakage due to missing 1.5 lambda from LAr– Could be big variation of Tile sampling fraction due to shower starting ‘too’ late
• Seen in testbeam, but only if no LAr in front– Plan to look at fraction of LAr / tile energy deposits for such events to verify
22 January 2009 David 13
J0
• Now the next plots show always on the left the absolute value of Delta METx, fabs(rec.METx - trueMETx) versus the true eta of the lead jet; overlaid is a profile plot of the same distribution
• On the right is always plotted the Delta METx with sign versus the true lead jet eta, again with a profile overlaid; the error bars ar the real spread in the bin rather than the spread/sqrt(N), which is the default plotting option of root
• Not much to see here…
22 January 2009 David 14
J1
• We start to see a variation of the bin spread beyond eta 3.2, where there is a dip in the mean DeltaMET on the left, which comes from a reduced spread in DeltaMET, as seen on the right-hand side
22 January 2009 David 15
J2
• Same as before, there is a dip in DeltaMET, around crack2 the error distribution tends to be narrower, hence better (!) than everywhere else
• This is a small effect though
22 January 2009 David 16
J3
• Now the detector structures around crack2 start to show up as expected• The mean error on the left peaks at the crack position, and this is due to
increased spread, in other words worsened resolution, as seen on the right• Note the absence of any bias, as expected, as we’re looking at MET in the
lab frame, not projected onto the jet axis
22 January 2009 David 17
J4
• Also here, very pronounced peak around 3.2• Crack1 starts to show up, too• Maybe also a glitch around eta 0.7
22 January 2009 David 18
J5
• Again, clearly worse resolution in the crack regions• Note that around eta 0, where we have seen a peak when
cutting on really bad MET, there isn’t so much to see; maybe a slight degradation, not clear
22 January 2009 David 19
J6
• Jets are now realle central, peaks are visisble at 0.7 and 1.5• Again nicely apparent from the plot on the right what this is,
the spread increase in these areas
22 January 2009 David 20
J7
• Low stats now in the crack1 region, still the same picture
22 January 2009 David 21
J8
• All jets are central, we hardly reach crack1
22 January 2009 David 22
Conclusions about DeltaMET vs. jet eta plots
• Detector structures are nicely visible in these plots• Interesting to see how the MET immediately broadens if the
lead jet hits the cracks; just as expected• Again, time to look at how we can get rid of such events