Studies towards TGC limit extraction
-
Upload
jermaine-douglas -
Category
Documents
-
view
46 -
download
9
description
Transcript of Studies towards TGC limit extraction
Nick Edwards 1
Studies towards TGC limit extraction
Nick Edwards, University of Glasgow
Nick Edwards 2
Outline
Reminder on TGC binnings. Comparison of SM and TGC distributions at truth level in these bins. Check how CZZ looks in the four TGC samples compared to the SM. First results using a grid method to parameterise yields as a
function of the coupling.
Quick reminder on TGCs• We add them as extra terms in an
effective Lagrangian.• For on shell ZZ, we can parameterise
them in four independent constants:• f5Z, f4Z, f4γ,f5γ
• With the 1fb-1 analysis we set limits at the 0.1 level
Nick Edwards 3
Variables and Bins
Reminder: We have chosen to use the leading Z pT, four-lepton mass and leading lepton pT to calculate the TGC limits.
Chosen bins:– M(4l) : 0-240, 240-300, 300-400, 400-600, 600+– Pt(lead Z) : 0-60 , 60-100, 100-200, 200-300, 300+ – Pt(lead lepton) : 0-60, 60-100, 100-150, 150-250, 250+
Check how TGC distributions look in these binnings using the four Sherpa MC11c samples:– nTGC 0(f4γ=0.1, f5γ=f4Z=f5Z=0.0) – nTGC 1(f5γ=-0.1, f4γ=f4Z=f5Z=0.0)– nTGC 2(f4γ=f5γ=0.1, f4Z=f5Z=0.0) – nTGC 3 (f4γ=f5γ=0.0, f4Z=f5=0.1)
NB these samples are roughly at the boundary of what we have already excluded.
These distributions are at truth level.
Nick Edwards 4
Four Lepton Mass• Not really sensitive in first 2 bins
Nick Edwards 5
Leading Z pT• Not really sensitive in first 2 bins
Nick Edwards 6
Leading Lepton pT• Not really sensitive in first 2 bins
Nick Edwards 7
CZZ in TGC samples CZZ seems to be ~ 10% higher in the TGC samples in the four electron
channel and ~5% in the 2e2mu channel.
Nick Edwards 8
CZZ in TGC samples : mZZ
In the electron channels, CZZ
seems much higher in the TGC samples in the high bins than in the SM.
In four muon channel is a bit lower in the last bin – but could be stats.
Nick Edwards 9
CZZ in TGC samples : PTZ1
Nick Edwards 10
CZZ in TGC samples : PTLep1
Nick Edwards 11
Yield Parameterisation
Need to parameterise the predicted yield at as a function of the anomalous coupling, varying one and holding the others constant.
Doing this find a quadratic dependence: e.g. Yield = F00 + F44 (f5
Z)2
Previously we used afterburner, which takes the 4 fully reconstructed TGC samples (actually it can do it with one, but four are used as a cross check) and reweights to different values of the couplings to obtain the yield curves.
An alternative “brute force” approach is to generate a grid of samples:– Pick one constant to vary, holding the others constant.– Generate a load of samples at different values of the coupling.– Obtain the predicted yield at each value of the coupling, then can fit a
parabola to obtain the paramaterised yield.
From 1fb-1 analysis
Differential Cx including TGCs Set all except f5Z to 0
Nick Edwards 12
f5Z Parameterisation
Try this approach this for f5Z.
Generate 100k events with Sherpa 1.4.0 with f5Z = -0.4, -0.25, -0.1 0,
0.1, 0.25, 0.4. f5
Z = -0.4 sample not finished yet… Then run analysis code to obtain predicted yield at truth level and fit
to a parabola. These “predicted yields” are all at truth level so far. What we want is the predicted yield at reconstruction level – ie what
we observe. Unfortunately it’s not really practical to fully simulate and reconstruct a grid of samples, so instead predict the yield at truth level and use CZZ to come back to the reconstruction level– Relies on CZZ being the same in the TGC samples as the SM – have
shown it isnt’!
Also need to apply k-factor to get normalisation right.
Nick Edwards 13
f5Z Parameterisation -
Unbinned
Nick Edwards 14
f5Z Parameterisation - mZZ
• Same thing but in the 5 bins of mZZ that we’ve chosen.• Fit looks worse in the first 2 bins, but this is just because it’s very
zoomed in compared to the other 2. Can see by looking at the “b” parameter that sensitivity is very low in these bin.
Nick Edwards 15
f5Z Parameterisation – PT
Z1
• Same thing but in the 5 bins of PTZ1
that we’ve chosen.
Nick Edwards 16
f5Z Parameterisation – PT
lep1
• Same thing but in the 5 bins of PTlep1 that we’ve chosen.
Nick Edwards 17
Conclusions
CZZ seems to be 5-10% different in the TGC samples to the SM ones. Difference seems to be coming mainly from electron channels.– Need to understand where this comes from.
First attempts at parameterising “truth yield” as a function of f5
Z . If CZZ were the same would be simple to convert this to expected observed events.
Then can compare this to afterburner as a cross-check.