Answering the review questions in view of the approval for the TP text

Matt Herndon, via email

page 2 line 44: You state that all scenarios use 140 pileup and then say 50 pile for the non aged scenario.

  • Fixed in the forthcoming version

page 2 line 56: It would be useful to state the eta coverage of the muon upgrade as well.

  • we will include it in the next version

page 2 table You have “x”s for all scenarios for both full sim and delfts. Is this correct?

  • This is true for the signal only, and as a matter of fact the Delphes aged is not existing. Fixed in the forthcoming version.

page 3 line 106 The references to the figure number here are later for the muon plots are incorrect.

  • fixed in the forthcoming version

page 4 figure 2: The eta plot is given twice.

  • Fixed in the forthcoming version (the same is true for figure 6 in fact).

In general could these plots be superimposed so they can be compared. Also can the aged scenario be added.

  • there's work in progress on this.

Is it possible to compare full reconstruction and Delphes to show that the lepton reconstruction performance is well modeled?

  • In terms of shapes, this is addressed in figures 18 and 19. Concerning the normalisation, we checked that the efficiency "ladder" is very comparable between fullsim and delphes as well. Plots will be added to the forthcoming version of the note.

page 4 table 2 Why does the isolation number have to be changed to 0.6?

  • This is to preserve the same efficiency of 80%.

page 5 line 116 Is it possible to add some plots that access pile up jet rejection?

page 6 figure 9 Why is the efficiency in eta non symmetric?

  • we looked into this plot in detail. The asymmetry was an effect of the binning choice. When using several different binning options the effect disappears. The new plot in the note has about the same granularity of the previous one, and shows the expected behaviour.

page 7 line 159 Why do you cite “color flow” here. Color flow would effect things like underlying event activity between the jets but not the HT. HT should depend simply VBS topology.

  • As the HT adds up both the tag jets and the additional activity in the event, it is sensitive to the radiation induced by the color flow between the tag jets.

page 10 figure 15,16 I’m surprised there is no real difference in MET performance between the scenarios. The improved tracker coverage should allow better pile up jet rejection which should effect/improve MET performance. Could you explain why we would expect similar performance in these plots.

  • When comparing the Phase I and Phase II distributions, for the Puppi MET the reduction in the response plot is visible: the RMS goes from 45 to 37 GeV. Unfortunately, for the Reco MET the impact is not as important and the width of the distributions is very well comparable. This is probably due to the different PU conditions of the two samples (50 vs 140).

page 11 line 176 I think you need to make a statement on the effect of neglecting interference. For instance you could give the magnitude of the interference effect in the phase space of interest. You could then treat the effect as a systematic uncertainty. Note that in other analysis we have actually find that scale variation can lead to a large variation in the interference contribution even doubling it or flipping it’s sign.

  • We did not study in detail the interference with dedicated simulations so far, for this studies. Since the main target of this study is to compare the performances obtained with various upgrade scenarios, we thing that it is fair to report that the interference has been neglected, and this does not invalidate the conclusions of the study. Adding an uncertainty related to the interference seems like an overkill in this case: by the HL time, it's very likely that NLO simulation will exist for all the final states, therefore theoretical uncertainties will be much smaller.

page 19 line 346 You list a cut of m_jj 650 in the table and 850 in the text.

  • Fixed in the next version of the text.

page 19 line 354 Isn’t it a bad idea to use different physics variable cuts as you have done in the HT case. It makes the results not as cleanly comparable.

  • On the one side we see your point. On the other side, the different HT cut reflects the different geometry of the detector: the enlarged eta coverage allows for a different tuning of the selections, and it would be unfair not to profit of this to assess the improvements due to the updates. In other words, we are comparing the two detectors after optimisations tuned to each upgrade configuration.

page 23 figure 27 The measurement of logitudinal vector boson scattering is the most obvious place where the phase II detector outperforms the phase one detector. It would be good if you investigated further to determine what variable give the additional sensativity and then related that variable to specific detector upgrades. This sort of justification for the upgrades should be one of your primary conclusions.

  • The improvement in the performances for the LL analysis is indeed an important benchmark for the upgrade. In particular, for each of the two final states the pair of most sensitive variables is chosen with an optimisation procedure, which fits all the possible combinations of the relevant variables. The outcome of the optimisation is $(\Delta\varphi_{jj}, p_T(\ell_1))$ for the both the same-sign WW and WZ cases. This is more clearly described in the text now for the same-sign WW as well.

follow-up with Matt

On interference. By putting this result out in public we are both making a statement on relative differences in detector performance and a projection of our expected physics reach. It's important that the statement on expected physics reach be accurate. I do not think that saying you neglect interference is quite good enough. If the effect were large then the results of the physics study would be questionable. I think this is simply addressed by running a QCD, EWK, and EWK+QCD madgraph job and comparing, at generator level with a VBS selection, the differences in cross section. In WZ the effect is small, but that does not mean it is in SS WW where the diagrams are different.

  • we checked this by performing the three generations, for the same sign different flavour case, with phantom. After mild cuts (pt > 20 GeV for the physics objects, mjj > 300 GeV and Detajj > 2) we find this distribution of mjj, where the bin-by-bin fluctuations are due to the limited statistics used in the plot (error bars on the red hits). As can be seen from the plot, the sum of the two samples (blue filled area) is very close to the generation done with the sum of the diagram sets (red line).
  • WWInterf_mjj_vbf_EWK.png

On Isolation. The difference in the isolation cut is large enough you should make some comment. The effect could be from pileup, in which case the pileup rejection is not working very well. The effect could also be from how the new calorimeter works. It would also be nice to show isolation distributions as well.

  • The main reason of the difference is the 140 PU instead of 50 PU. From the TP text, one can extract the very rough number of 0.5 GeV additional energy in a jet per PU vertex (140 PU events deposit ~ 70 GeV in a jet). This means that from 50 PU to 140 PU events one gets ~30 GeV more in a jet cone, and hence something similar in the lepton isolation cone. For leptons of pT of about 70 GeV, this is 0.5 of the energy, which is in the same ballpark of the 0.6 cut we apply.

On the conclusions. If you could demonstrate in a easy to interpret way that the DeltaPhi(jj) variable is improved due to the upgrade then you could directly motivate the upgrade in clear terms. I would encourage you to look into this in more detail.

  • in the combination section of the new version of the note, comments are being added on comparison between upgraded and current scenarios, which we think match the need of understanding the reasons of improvements in performances.

Sascha Savin, via email

in 13-015 we show sensitivity of the measurement to couplings and show that it starts already at low mll and it can be pronounced already at the 19 fb-1 we have, in the TP plot we see that even at the full 3000 fb-1 we have to expect only very few events and the effect we are looking at , is at extremely high mll. I don't think I get the TP plot by increasing the SMP-13-015 by a factor of 150. This should be somehow clarified.

  • a detailed comparison to the 8 TeV performances for the EFT is shown in this talk, at page 12. When performing the analysis at 14 TeV, with 20/fb of integrated luminosity, the limits obtained at 14 TeV are about 1/2 of the ones at 8 TeV. The reason of the difference is due to the binning choice (we use 7 bins instead of 4) and the different parton luminosity at 14 TeV wrt 8 TeV.

in WZ it is nicely shown how our sensitivity depends on the luminosity I think we need to have a similar plot for WW

  • The plot is actually available in the note, when fixing the fake rate scale factor to 1.

Fabio Cossutti, via email

1) the sensitivity plot as a function of the luminosity for WW is the one in Fig. 25 right, am I correct? I agree it would be good to have it in the conclusions, as done for the WZ one;

  • fixed in the forthcoming version.

2) I think a plot is missing in Fig. 47 right;

  • This figure will show qualitatively the difference in shape between the various polarisation states (LL, LT, TT). We will have it for the approval, its production is just lowish now in the priority because of the pre approval talk.

3) I notice that in SMP-13-015 the mll plot shows the effect of the FT0 operator, while the mll sensitivity plot you show describes the effect of try FS0 operator, and the mll ranges are completely different, due to different mll cuts adopted in the analyses, so it is difficult to directly compare them. Do you perhaps have a plot showing the FS0 sensitivity in terms of mll similar to the one of SMP-13-015?

  • The different range of the plot is due to the enhanced kinematic because of the increase in parton luminosity when going to 14 TeV.
  • the plot you asked for is here, and is now also in the note at the place of the previous one.
  • mll_T0.png

Markus Klute during the pre approval

* Why is the aged EWK WW XS that worse with respect to the phase II case?*

  • the precision is systematically dominated, and the main player is the uncertainty on the b-tagging scale factor, which we doubled with respect to the phase 2 case. Quoting an email from Petra Van Mulders:
    The systematic uncertainty on the b-tagging and mistagging scale factor for the aged configuration 
    is something that is very hard to predict as we have not (yet) assessed these.
    The dominant source of systematic uncertainty on the mistag scale factor for 8TeV 
    are the event kinematics (in particular the jet pt), secondary interactions in the material 
    and decays of long-lived K short and Lambda. 
    However, it is very likely that other sources will start dominating for the aged phase 1 scenario, 
    like pile up, fake or mismeasured tracks and so on. The same is true for the b-tagging efficiency.
    I assume that the guess of doubling the uncertainty is probably the best that can be done 
    without dedicated studies.

Pre-aprroval reprise (2016)

Senka Duric (17th Jan 2016)

suggestion: L104 and L110 both say that CTEQ6L PDF was used,  so perhaps 
mention it only in L110 saying it applys for both MG and Phantom.

L312: missing "m" in brackets: ("jj")->("mjj")

L317 and L318 are the same?
Indeed: L318 kept only

L347 it is not clear to me what is the value of the uncertainty on the lepton fake rate. 
From Sec 5 I concluded that it is 30% for all scenarios. But here in L347 it says it is 
increased "by 30%" for Phase 2 wrt Phase 1..
Beside the normalisation uncertainty of 30%, we also have a shape uncertainty on fakes and that is the one which is mentioned in L347.

Determination of EWK cross section. The fit of templates in 2D distribution was done. 
Is the bkg template a sum of all backgrounds or do all processes from Fig 3 have their 
own templates? Bkg template/templates are allowed to float within their uncertainties? 
Are you using the combine for the fit?

Yes each background has its own template and are allowed to float within their uncertainties. The fit is indeed performed with the Higgs combine tool. L.354-355 was modified to say that we use combine for all results in this PAS.

L422 usually for anomalous coupling what is done in MC models is variation of c/Lambda 
(=Wilson coefficient / scale), not c alone. From the text and table 2 looks like you are setting 
limits on c, and from Fig 10 like you are setting limits on c/Lambda. Can you please check in 
the MG model which quantity is varied?

  • It is indeed c/Lambda that we are setting limits on, as we are not setting the scale for the new physics. I've seen the name Wilson coefficient been used both for the dimensionless coefficient as for coefficient / scale (perhaps wrongly). Anyway we should clarify this in a few places:
    l423: Wilson coefficients -> coefficients with dimension 1/TeV^4
    l424: ones -> operators
    table 2: add units! ( TeV^-4)

Table 2: Describe in the caption what are the values in the brackets [] in last column.
These are the references from which these numbers are coming from but seems some of the reference where missing, now it is fixed

Josh Kunkle (18th Jan 2016)

205-218 : The description seems consistent with a phase-1 detector, but in the phase-2 case 
we'll be using the HGCal. I think it would be good to give some description of how electron 
identification works in the HGcal. The HGCal enters in other parts of the event reconstruction, 
so you could have some separate description on HGcal reconstruction.
DONE: We added a few lines pointing at the specificity of the HGCAL electron reconstruction for Phase-II detector.

332-333 : Here you mention the sign flip estimate, but I don't seen any description of how 
you make the estimate besides this sentence. I think a bit more detailed description is warranted.
DONE: Well, we took it from ttH 8 TeV analysis. This is mow given as reference and made clearer in the text.

359 : Here you mention the fake rate scale factor which is used to show many of the results. 
But I didn't understand from Section 5 that this was part of the background estimate. I think 
it would be good to clearly state in Section 5 that this scale factor will be used in displaying the results.
This is stated in L.292-294 at the end of section 5.

Minor comments :

Introduction : First you say that the discovery of the Higgs completes the SM, then you explain 
why the SM isn't complete. The statement on line 3 is a bit too strong.
DONE: changed "completed" by "confirmed". Removed last part of sentence as it is repeated later in more details.

19 : sing -> sign

48-50 : Its difficult to claim that we know what the HL-LHC will do, its better to say that its planned 
or expected to meet those benchmarks.
DONE: changed to "is expected to ..."

102 : extra alpha_s, should go with the previous line somehow?
DONE: Forced carriage return to have the whole formula on next line, but perfect but ...

136 : missing \times

272+9 : form -> from

311 : remove 'units'

317, 318 : repeated
DONE: L317 removed

Qiang Li (5th Feb 2016)

L35-41 Is it possible/better to move these sentences ahead to 2nd paragraph in Sec.1?

  • Moved whole paragraph up

L55-56: Please give reference to the Zeppenfeld variable and the following statement "Usually, the two jets with the largest p T in the event correspond to the deflected quarks".

  • the reference is an email from Dieter Z. to Sasha Nikitenko, back in 200*something (like 2006?), so this reference is not trivial to identify

L64-66: A bit confusing to readers. Suggest to mention WZ+2Jets first (and add e.g. "with one lepton out of detection"), and then TTbar with jet fake lepton, in order to make sense for the subsequent "The W well in a similar manner".

  • DONE

L108: CTEQ6L or CTEQ6L1?

L111: Please provide a reference to "the studies performed for the CMS Upgrade Technical Proposal"

  • DONE

L123: Which version of MADGRAPH here?

  • Same as above, we only give the full version for the first occurence of a MC and then use the short name.

L244-246, L255-256: Lepton efficiency numbers etc. Would it be better to provide refs?

  • These numbers are for folding our MC simulations with the tabulated efficiencies from the TP and choosing isolation worjing point within the analysis. So not reallt a reference but a "measurement" in the analysis.

L256: "an efficiency of ~70% for b-quark jets": Doesn't it depend on jet pt?

  • Ok added "average" in front of efficiency

L280-281: "Jets associated to the tag lepton", what is the association criteria?

  • This is within a cone of 0.3 in DeltaR, added to the text

Sec. 5, it seems only samples with PU20 has been exploited? How large is the fake rate dependence on PU, considering Phase II will have typical PU of 140?

  • Well this is part of the check done a l.297 using the 3 different upgrade scenario. To make it explicit, we added ", including the high pile-up ones," after scenario.

L323-324: "the distance between the di-jet and the di-lepton systems DeltaR ( ll, jj ) has to be smaller than 6 units" What is the definition and motivation? Is this DeltaR related to the last subplot on R in Fig.3?

  • The selection kills some fake lepton background with almost 100% efficiency on the signal.
  • Fig. 3 shows another variable (defined as the ratio of the products of lepton and jet pT).

L327-330: Is the scalar sum over all tracks or just only over the ones excluding the 2 tagged jets?

  • Only tracks between the 2 tag jets in pseudo-rapidity are used, so excluding the jets indeed. Also the leptons are removed. Text modified accordingly.

Fig.3: Is it better to plot mjj in log scale, as the high tail region is the most interesting part for VBS process? * here is the link to the log scale plot: Here the plot.

Fig.4: left plot has lower resolution and should be improved.

  • Changed

L463 jj->mjj

  • Fixed

Fig.14, please make the y title consistent with Fig.7.

  • Done

Fig.9 and Table.2: The total excess with e.g. S0~1.10 (T0~0.07) over SM is only around 1 event, but how can you set 95%CL limit for that S0 (T0)?

  • Well as usual we quote the median of the 95% CL, indeed the error band around it can be large.

-- PietroGovoni - 2015-04-28

Topic attachments
I Attachment History Action Size Date Who Comment
PDFpdf Eff_LeadvsEta.pdf r1 manage 15.7 K 2015-05-03 - 19:15 PietroGovoni  
PDFpdf Eff_LeadvsEta_P2.pdf r1 manage 15.8 K 2015-05-03 - 19:17 PietroGovoni  
PDFpdf Eff_TrailvsEta.pdf r1 manage 15.9 K 2015-05-03 - 19:15 PietroGovoni  
PDFpdf Eff_TrailvsEta_P2.pdf r1 manage 16.1 K 2015-05-03 - 19:17 PietroGovoni  
PNGpng WWInterf_mjj_vbf_EWK.png r1 manage 9.0 K 2015-04-30 - 09:35 PietroGovoni  
PNGpng mll_T0.png r1 manage 14.2 K 2015-04-30 - 12:07 PietroGovoni  
Edit | Attach | Watch | Print version | History: r17 < r16 < r15 < r14 < r13 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r17 - 2016-05-25 - XavierJanssen
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Main All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright &© 2008-2021 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback