AN-20-069: The azimuthal correlation in Z+jets at 13 TeV

Contact: Qun Wang

Abstract

The production of Z bosons associated with jets is measured at a centre-of-mass energy of 13 TeV with data recorded at the LHC corresponding to a luminosity of 35.9 pb−1. The multiplicity of jets with pT > 30 GeV is measured for different regions of the Z transverse momentum, ranging from pT < 10 GeV to pT > 100 GeV. The azimuthal correlation ∆φ between the Z boson and the leading jet as well as the correlation between the two leading jets is measured for different pT of the Z boson. The measurements are compared with predictions at leading and next-to-leading order supplemented with parton shower and hadronization. Higher order calculations are able to describe the measurements reasonably well.

Authors

DESY, Hamburg, Germany: Qun Wang, Armando Bermudez Martinez, Hannes Jung, Mikel Mendizabal Morentin, Heng Yang Peking University, CN: Qun Wang, Heng Yang

Presentations

* 22.01.2021

* 08.01.2021

* 04.12.2020

Comments on paper draft

comments 1 (Vieri Candelise)

*- I think the title should be changed! This is not a standard Z+jets measurement xIntroduction - pp to be corrected*

Ok, it is corrected.

- Z boson (p.2) but sometimes you write Z-boson (p.4)

Ok.

- high energetic -> high energy

Done

- through the Intro part I often have the impression that sentences lead the reader thinking that this is just a standard Z+jets differential measurement: I would focus always on the correlation, the Zpt splitting, and the things that characterise your measurement.

- I would mention that high precision understanding of QCD/EW aspects of Z+jets in defined Zpt regions is also important for benchmark future precision measurements of the W mass, Higgs, NewPhysics searches

*x Theoretical Predictions - the two bullets refer to the same name (MG5_aMC) - shouldn’t be better to change the second name?*

Use MG5_aMC NLO and MG5_aMC LO to distinguish

- second bulltet: “5” ?

Corrected.

*x Data Analysis - antiKT -> ref*

done

- I think it would be beneficial for the paper to have a couple of detector level distribution, such as the dilepton pair invariant mass and the lepton isolation

Yes, we add the dilepton invariant mass from both muon and electron channel.

- I would mention here the JEC/JER

included.

xxx I think a Section on the background composition is needed…

Yes, it is added.

*x Uncertainties - I would put a table with relative, magnitude-ordered, % of systematics*

done

- are PDF, alphaS and scale uncertainties missing?

Included

- do MC xsecs have a systematic uncertainty?

Yes

- is there an uncertainty on the response matrix statistics?

Yes, update it in paper

*x Results - as discussed, I hope some more predictions will be added to the final result*

GENEVA: we are working on it, but having troubles with the generated lhe file, so we need to generate them from scratch (having problems with the geneva installation). But we are working on it.

- is double parton scattering playing a role? did you checked its contribution?

- No, we do not check it.

- plots: labels often overlap with data points

corrected.

- did you understand the behaviour of the uncertainty in the 2nd bin of the 1j unfolded xsecs?

This was mentioned on the pre-approval talk (https://indico.cern.ch/event/977173/contributions/4243369/attachments/2200002/3720712/ZJetsAzimuthalMarch2_Qun.pdf). In backup (slide 38), the PU uncertainty from bin-to-bin is less than 5%, while values in the table come from the covariance matrix. If we look at the response matrix (slide 16, which is normalised to Gen distribution), only round 1% of events with zero gen jet which have one jet at reco level are 70k events. That results in 59% events with one jet at Reco level coming from events with 0 gen jet, the migration is very large. Recently, we did some study with a tight PUMVA cut, but it improves very little in uncertainty, from 21% to 15%.

*x Conclusions - I would personally expand a bit this section, showing the physics final content of your work in more detail!*

Ok

- when you say “weakly” -> would it be better maybe to quantify these statements?

done

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

comments 2 (Juan Pablo Fernandez Ramos)

-+)Let me repeat what Vieri has said. TMD predictions should be added for the paper. This is esssential because I just think that comparing the shapes (data/prediction) is what can shed some light on whether theory is doing the right thing (with their resummation when soft gluon radiation takes place) for low Zpt. As you showed already, PB/TMD helps in the description of ptZ itself ( DY pT spectrum in the presentation in slide 12 , slide 32) https://indico.cern.ch/event/866746/contributions/3719531/attachments/1988054/3313404/LowMassDY_TMD.pdf I mean including those predictions in the azimuthal Z,jet plots of the paper (well in all the plots in fact) would be the key ingredient to proceed further from my point of view. Presented on Friday, still on-going, (NLO two jets), need a few checks.

*+) A reminder of the checks to be done : Please provide us with -The fraction of events with a matched gen-reco jet pair;*

Ongoing

-inv mass of Z in the first Z-pt bin,

it is attached.

-This is new. I would like to see the delta_z between PV and Z (or if you wish the zPV of the daughther tracks of the Z)

nanoAOD Mikel dZ muon

- understand why there is a change in the uncertainty between the 2nd bin ( 1 jet) at the adyacent ones ( 0 and 2 jets)

This was mentioned on the pre-approval talk (https://indico.cern.ch/event/977173/contributions/4243369/attachments/2200002/3720712/ZJetsAzimuthalMarch2_Qun.pdf). In backup (slide 38), the PU uncertainty from bin-to-bin is less than 5%, while values in the table come from the covariance matrix. If we look at the response matrix (slide 16, which is normalised to Gen distribution), only round 1% of events with zero gen jet which have one jet at reco level are 70k events. That results in 59% events with one jet at Reco level coming from events with 0 gen jet, the migration is very large. Recently, we did some study with a tight PUMVA cut, but it improves very little in uncertainty, from 21% to 15%.

- explain why not doing just a single unfolding instead of two (Z->ee and Z->mumu)

The detector response is different for electron and muon channel, and a single unfolding means to have a response matrix with different elements for the two channels. Attempt to perform combination and unfolding in one go using such a matrix was tried for SMP-14-013 with no success. We are doing unfolding and combination in two steps, because it is simpler.

-Prediction from GENEVA to be added as well ?

GENEVA: we are working on it, but having troubles with the generated lhe file, so we need to generate them from scratch (having problems with the geneva installation). But we are working on it.

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

comments 3 (Paolo Azzurri)

I would like to ask more specifically about the motivation of studying these azimuthal correlations
in the two ptZ bins (<10 and 50-100). It is quoted that resummation is important at low pT of course,
but looking into the quoted references and other papers I did not find anything related to azimuthal correlations
at low pT, and the expected effects of resummation/TMDs for those. Sorry I probably just did not find it,
in case can you point us to that ?

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

comments Pre-approval notes

The main points to follow up are - Check distributions consistency between the two main eras (B-F vs. G-H)

Here are the distributions between the two main eras (B-F vs. G-H). We could see that, after the PU reweighting,the number of vertices of Run G-H is better described than Run B-F. But if we look at the exclusive jet multiplicity,the difference between two eras (B-F vs. G-H) only comes from the normalization, the step between Njet=0 andNjet=1 remains the same.

- Decide if the analysis reports results in either Zpt:[50-100]GeV or Zpt>50GeV

We have bin Zpt>100 GeV in AN, there is enough statistics inZpt>100 GeV bin. We will update the paper draft with Zpt>100 GeV.

- Cross-check single top cross sections (considering the appropriate leptonic branching ratios, when applicable)

Now we have the correct cross sections, we will update the results.

- Make use of more precise VV (WZ/ZZ) NLO samples

The background contamination from VV is very small, in total less than 0.5%. We believe the VV LO samples are enough.

- Check the status of the statCom questionnaire (make sure the questionnaire is "published")

yes

- Try to include the jet merged PB TMDs predictions

Armando is working on that to get the publication ready, which describes the procedure.

comments from ARC members

comments from Simone Gennai

Dear Authors, In the twiki I see that some of the questions which were present also before the pre-approval presentation are still unanswered. Like: “

- understand why there is a change in the uncertainty between the 2nd bin ( 1 jet) at the adyacent ones ( 0 and 2 jets)

This was mentioned on the pre-approval talk (https://indico.cern.ch/event/977173/contributions/4243369/attachments/2200002/3720712/ZJetsAzimuthalMarch2_Qun.pdf). In backup (slide 38), the PU uncertainty from bin-to-bin is less than 5%, while values in the table come from the covariance matrix. If we look at the response matrix (slide 16, which is normalised to Gen distribution), only round 1% of events with zero gen jet which have one jet at reco level are 70k events. That results in 59% events with one jet at Reco level coming from events with 0 gen jet, the migration is very large. Recently, we did some study with a tight PUMVA cut, but it improves very little in uncertainty, from 21% to 15%.

- explain why not doing just a single unfolding instead of two (Z->ee and Z->mumu)

The detector response is different for electron and muon channel, and a single unfolding means to have a response matrix with different elements for the two channels. Attempt to perform combination and unfolding in one go using such a matrix was tried for SMP-14-013 with no success. We are doing unfolding and combination in two steps, because it is simpler.

-Prediction from GENEVA to be added as well ?

GENEVA: we are working on it, but having troubles with the generated lhe file, so we need to generate them from scratch (having problems with the geneva installation). But we are working on it.

Are these being discussed during the pre-approval (and now fixed) or you are still working on them?

yes, and we are still working on GENEVA

Best, S.

Again on the MC to data disagreement in Fig.1 of the paper. The disagreement seems to come mostly from the 0-jet category and low pT Z bin, at least looking at fig. 2. Do the various object SFs have been applied in bins of N jets and Z pT or they are applied inclusively?

they are applied inclusively.

comments from Fabio Cossutti

The issue of the fluctuating uncertainty among multiplicity bins for low pt jets was mentioned in the pre-approval meeting, and explained there with the strong correlation between bins in the unfolding matrix. While the argument sounds plausible, such an effect is not visible for higher Z pts. Do you have the multiplicity unfolding response matrix available? Could you please put it in the AN, as the matrix for any plot that you desire to publish? I believe this is a request nowadays, and in any case it is useful to motivate the explanation provided.

We have the response matrix for jet multiplicities, and we showed it in the pre-approval talk. I will update the AN with the response matrix.

When do you expect to have a revised version of the documentation, addressing the various points discussed so far? When are the updated comparisons using GENEVA planned (question by Juan Pablo still not addressed)?

GENEVA: we are working on it, but having troubles with the generated lhe file, so we need to generate them from scratch (having problems with the geneva installation). But we are working on it.

It would be useful to have the check about different periods in the AN for the record. Could you please clarify the reason for the change in normalization responsible for (part of) the difference in the two ratios? What about other observable?

Yes, the PU reweighting needs to be modified according to the run era. We have other observables for the comparison, and I will update the AN with these observables.

comments from Sun Wong Lee

In the low pT(Z) region, soft gluon radiation becomes important, as the author emphasizes. Can the author compare this data set with the currently available specific NLO (or NNLO) resummation calculations? BTW, I am not sure if such specific calculations are available on the market.

For the low ptZ region, we cannot use any of the standard resummation methods, since they apply only for inclusive Z production (but not for Z+jet). For Z+jet we can use standard parton shower calculations or the PB TMDs.

Is this the first time this PB-TMD method is compared to the data? Or has this method already been compared to the experimental data through other physics processes?

PB TMD predictions: we have in our CMS paper on inclusive Z production (SMP-17-010) we have a prediction from PB TMDs (Fig 8) which uses the same generator (CACADE3) and the same PB-TMD set2. There are in addition several publications which describe the predictions for DY production at LHC and at low center of mass energies (arXiv 1906.00919, 2001.06488, both also published in journals).

Is there a particular reason to use 2016 data only? Adding another dataset (20178/2018) can help you get more statistics from a high Z(pT) region (for example, > 100 GeV). BTW, is it meaningful from this analysis perspective when expanding Z(pT) beyond 100 GeV? Is there a reason for choosing Z(pT) < 10GeV? How about z(pT) < 15GeV or something similar?

- the idea was just to have a bin with low pt(Z), with pt(Z) << pt(jet). Of course pt(Z) < 15 could also work, but then we come closer to the jet ptcut (and we have bin 10 <pt(Z) < 30). To have pt(Z) smaller than 10, say pt(Z)<5 or so, would make the analysis very difficult, since the statistics becomes very low.... so we had to find a compromise.

You used a medium lepton ID for the electron and muon, respectively. Have you ever used a tight IDs?

No, we have not used tight IDs. We use the same IDs as Z+jets 2016 analysis (SMP-19-009) and Drell-Yan pT over a wide dilepton mass range at 13 TeV SMP-20-003.

For jets, loose WP was applied to the puMVA. Is this WP optimized for this analysis? Have you checked other WP points?

We did not optimize the WP, we keep the same cuts as other Z+jets analysis SMP-19-009 and SMP-20-003. Yes, we checked the tight working point: it reduces the migrations, but it also reduces statistics.

The final Z-mass distributions (after all corrections are applied) generally indicates that the data are lower than expected. What is the reason?

The ratio is simulation/data, and the data is slightly higher than expected. In our measured Z mass region (76-106 GeV), the simulations agree quite well with data. (Z mass distribution for muon and electron channel are attached.)

In jet multiplicity (pT(Z) < 10GeV), there appears to be a significant difference between N(j=1) and N(j=2) in the total uncertainty. Do you understand this?

The jets pT are required to be larger than 30 GeV in the analysis. If Z boson has pT < 10 GeV, there is 20 GeV imbalance for the Z+1 jet system. For this case, some soft jets should be produced at the same time. We checked the response matrix for jet multiplicity (pT(Z) < 10GeV), there is large migration from Gen N(j=0).

While for N(j=2) events (pT(Z) < 10GeV), the two jets are produced back to back. This hard process could be described by Madgraph NLO.

Often in Z+1 jet events, measurements of the rapidity distributions (e.g. the rapidity of a particle,y, and the invariant rapidity differences (y(diff) = |y(Z-y)j|/2 and y(sum) = |y(Z+y)j|/2) can also provide an important check of QCD and theoretical predictions. Have you checked the rapidity distributions?

These two rapidity distributions have been measured by SMP-19-009.

Have you compared your data with SHERPA and MCFM?

Sherpa we could perhaps do, MCFM seems to be difficult, since it is a fixed order calculation, and we see, that we are sensitive to resummation effects and also in certain regions need a merging of different jet multiplicities, which cannot be done in MCFM.

We do not intend to have a full set of MC comparisons, but rather focus on a consistent set of comparisons, so that we can learn the underlying physics. We do not aim for the best description with the most super tune, but rather a basic physics comparsion. Therefore, at the moment we did not consider SHERPA or all the other generators. We do NOT want to conclude at the end: generator A is describing the measuremnt in some region better than generator B, which is differnt from generator C.

comments from Guillelmo Gomez Ceballos Retuerto

I think for clarity, it's maybe better if you include all the comments in the Q&A twiki, and reply over there. Otherwise, it's going to be hard to track them down, no?

yes, sure, we'll include the comments also in the Twiki....

comments from Philip Chang

Overall question - Why was only pt(Z) < 10 and pt(Z) > 50 - 100 are presented in the paper? In AN I see other pt regions, but it seems paper draft completely ignores this.

We just took as examples these two extreme bins. Of course we would like to have all the other distributions also released as supplementary material.

Figure 1: - Generally the data is lower even after supposedly all the corrections applied. Do we understand this? AN says SMP-17-010 SFs are used, but SMP-17-010’s AN2017-203_v9 Figure 3 shows very good agreement for both electrons and muons while your Figure 1, electrons Mee is worse by few percent. (dimuon seems OK however.) Is there any missing trigger/lepton SF? - To some degree, I can accept that the modeling of off-shell Z regions are not as good as the peak for the DY, but I am not sure I understand the difference between ee and µµ for the low mll regions. The differences is of order of 10% between the channels. (For high mll, they seem to show similar trend, however, which is less concerning.) Is this an effect of missing smearing? From AN I did not find whether the smearing was applied. At the end of the day, given the selection includes 76 - 106 GeV selection, I do not think that this is a major concern, but I would like to know what was done exactly.

In the plots, the data ia higher than MC after all corrections applied. We use the same SFs as SMP-17-010, and the smearing is considered in the analysis. SMP-17-010 has more backgrounds with Z boson, and they slit backgrounds in two components: the resonant background and the nonresonant background resonant backgrounds. While we use TTbar, Single Top, Wjets, diboson for our study. The way we treats background is similar to analysis SMP-20-003 (Drell-Yan pT over a wide dilepton mass range at 13 TeV) CMS AN-2020/026, v13 Figure 19, one could find nearly the same results for Z mass.

Figure 2: - Several points in the bottom panels are off the range. - Is there a plan to add systematic uncertainties for LO calculation? - Uncertainty diff of Njet = 1 v. 2. So is there a fix to this?

We do not plan to add the systematic for LO calculation, since the scale uncertainties are not well defined at LO. We have all the uncertainties for the NLO calcualtions. Of course the scale of the ratio plot will be adjusted. At present we are still investiagting the pileup contribution and we saw already that using the tight puMVA cut, the uncertainties of Njet=1 will be smaller.

L234 and L240-241: - The authors claim that the high Njet + low pT(Z) events are “QCD + Z radiation from higher order EW correction” events and notes that MC overestimates. But the difference is ~order of magnitude with fairly low error. (something like 20 +- small (Prediction) v. 3 +- 3 (measurement)). It is a bit hard pill to swallow for me to say missing EW correction is the reason.

Maybe the formulation is not clear: of course normal EW corrections should be small. However here we mean EW corrections (that is emmisson of real Z) from a dijet process. Since dijet production is orders of magnitude larger than pure Z production, the emmission of a Z from a dijet process makes a significant correction which can be large in a particular region of phase space. We shoudl try to make this clearer in the text.

Are there other generators that can be used to compare here? What about powheg? - From the AN, Figure 20 (top right) I see that the prediction agrees fairly well for pt(Z) ~10-30 GeV. It seems like a drastic change.

We still will try to use GENEVA, although at the moment we face technical issues.

Figure 3: - I don’t think I can reconcile Figure 2 (left) against Figure 3 (left). If I look at Figure 2 (left) for Njet >= 1, I see that for LO, the 1 bin has 30% deficit, and =2 bin has 5% deficit, and much larger for =3 and =4. Yet, Figure 3 (left) this is gone. Do the inclusive cross section measurement for Figure 2 (for njet > 1) and Figure 3 differ by a lot? It is as if the normalization is ignored.

In the low Z pt bin, the delta phi (Z,j) is mainly driven by Z+2jet events, since the Z cannot take the recoil of the jet. Therefore the normalisation of the N=2 bin is the one which is relevant for the delta phi (Z,j) distribution. We will check in detail that the normalization in both distribution is consistent, thanks for pointing out a potential inconsistency.

Topic attachments
I Attachment History Action Size Date Who Comment
PDFpdf NVtx_Zinc0jet_runBF.pdf r1 manage 23.9 K 2021-03-10 - 13:39 QunWang distributions between the two main eras (B-F vs. G-H)
PDFpdf NVtx_Zinc0jet_runGH.pdf r1 manage 24.8 K 2021-03-10 - 13:39 QunWang distributions between the two main eras (B-F vs. G-H)
PDFpdf ZMass_Zinc0jet(1).pdf r1 manage 30.1 K 2021-03-17 - 00:47 QunWang Zmass inside mass window (76-106 GeV)
PDFpdf ZMass_Zinc0jet.pdf r1 manage 30.0 K 2021-02-09 - 15:25 QunWang Zmass for different Z pT
PDFpdf ZMass_Zinc0jet_Ele(1).pdf r1 manage 30.4 K 2021-03-17 - 00:47 QunWang Zmass inside mass window (76-106 GeV)
PDFpdf ZMass_Zinc0jet_Zpt10.pdf r1 manage 29.8 K 2021-02-09 - 15:25 QunWang Zmass for different Z pT
PDFpdf ZMass_Zinc0jet_Zpt100.pdf r1 manage 32.0 K 2021-02-09 - 15:25 QunWang Zmass for different Z pT
PDFpdf ZNGoodJets_Zexc_Zpt10_runBF.pdf r1 manage 17.2 K 2021-03-10 - 13:39 QunWang distributions between the two main eras (B-F vs. G-H)
PDFpdf ZNGoodJets_Zexc_Zpt10_runGH.pdf r1 manage 17.1 K 2021-03-10 - 13:39 QunWang distributions between the two main eras (B-F vs. G-H)
Edit | Attach | Watch | Print version | History: r15 < r14 < r13 < r12 < r11 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r15 - 2021-03-25 - QunWang
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    CMSPublic 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