Peak in leading jet pT distribution for ggF H+jets with FxFx merging

Dear Alena,

There are some pythia8 options that you use that are not needed for FxFx (and I'm not sure what they do exactly). Could you try removing them and re-shower the LHE event file? They are:

    "JetMatching:nQmatch = "+str(maxjetflavor),
    "JetMatching:jetAlgorithm = 2",
    "JetMatching:slowJetPower = 1",
    "JetMatching:clFact = 1.0",
    "JetMatching:eTjetMin = "+str(qCut),
    "JetMatching:etaJetMax = 4.5", <-- should be set to "infinity"
    "JetMatching:exclusive = 1",
    "JetMatching:nJet = 0",
    "JetMatching:jetAllow = 1",
    "JetMatching:doShowerKt = off",

So, for the FxFx merging, you only need (with the required UserHook for jet matching):

JetMatching:doFxFx = on
JetMatching:merge = on
JetMatching:qCut = 50
JetMatching:qCutME = 10
JetMatching:coneRadius = 1.0
JetMatching:etaJetMax = 1000.0
JetMatching:nJetMax = 2
JetMatching:scheme = 1
JetMatching:setMad = off

(and of course the usual pythia8 flags that are needed for MC@NLO matching, which can be found at the bottom of the page
http://amcatnlo.web.cern.ch/amcatnlo/list_detailed2.htm#showersettings )

Note that if you use the interface to pythia8 that is shipped with the MG5_aMC, all of this is taken care of automatically.

Best regards,
Rikkert

Dear Rikkert,

Thanks for your comments.
I re-showered the events without the settings you mentioned, but there's still a peak in the pT jet distribution. However, compared to my previous settings, the peak is smaller and shifted to lower pT values (~40 GeV now). Let me know if logfiles or plots are useful for you.
I have to say that I completely removed the JetMatching:etaJetMax setting and forgot to set it to 1000 as you suggested. So I ran with the default value, which seems to be 2.5 (<previous setting of 4.5). However, it's hard for me to understand why this should be the reason for the pTj peak. Do you think it's anyway useful to re-run with JetMatching:etaJetMax = 1000.0 or do you have other suggestions?

Thanks,
Alena

Dear Alena,

You have to use JetMatching:etaJetMax = 1000.0 otherwise jets outside this region aren't included in the matching.

best,
Rikkert

Dear Rikkert,

Ok, I see! However, with JetMatching:etaJetMax = 1000.0 the peak around 50 GeV is still there.

One question concerning the setting JetMatching:eTjetMin:
You suggested to remove it, so it is set to 20 GeV by default. However on the pythia matching page
http://home.thep.lu.se/~torbjorn/pythia81html/JetMatching.html
it says that it should match qCut (so 50 GeV in my case)?

Best regards,
Alena

Dear Alena,

I believe that's only for Alpgen or LO MLM matching. Not for FxFx. I don't really know about the peak. How large is it? Does it change when varying the merging scale? A 10-20% dependence on the merging scale is to be expected for this process.

best,
Rikkert

Dear Rikkert,

yes, the peak depends on the merging scale. I ran the shower with qCut=30,50,70 GeV and the peak it always roughly around the corresponding merging scale. It also seems as if it gets broader with increasing qCut. One possibility is of course to go with the merging scale below the ptj threshold we use for our analysis (20 GeV), but I'm not sure if this is a sufficient solution.

What do you mean with 10-20% dependence on the merging scale? The distribution can change by 10-20% when varying the merging scale?

Best regards,
Alena

Hi Alena,

> What do you mean with 10-20% dependence on the merging scale? The distribution can change by 10-20% when varying the merging scale?

Yes.

Rikkert

Dear Rikkert,

I put an example plot of pt leading jet here:
 /afs/cern.ch/user/a/aloesle/public/MC_ggH2Np/ggH2Nptautauleplep_FxFx_ptLeadJet.eps
I hope you can access it.
You can clearly see the peak around 50 GeV (this is for qCut=50GeV).

Best regards,
Alena

Dear Alena,

Ok. That's clearly too large of an effect. It almost looks that the merging didn't reject any of the events. How large is the merging efficiency as reported by pythia8? I.e., what is the fraction of the events reject?

Best,
Rikkert

Dear Rikkert,

I'm not exactly sure if this is the merging efficiency but after the shower pythia gives the number of events as

'tried': 169974 ,
'selected': 169974 ,
'accepted': 40530

I generated 250000 events, so 169974/ 250000=0.68 -> 32% of the events are rejected?

Best regards,
Alena

Dear Alena,

Ok. This is a bit suspicious. Is there a way to download your LHE event files so that I can try to shower them myself to see what's going on?

Best,
Rikkert

Dear Rikkert,

I'm using gridpack to run on the grid, therefore I only have the evnt files. I can run again and output the lhe file, but this will probably take a few hours. I'll let you know where you can download the file then.

Thanks,
Alena

Dear Rikkert,

you can find an example lhe file here:
/afs/cern.ch/user/a/aloesle/public/MC_ggH2Np/tmp_run_01._00001.events.tar.gz

Concerning the peak: We received a comment that the pt jet peak around the merging scale is actually expected and that we therefore need to make sure qCut is set to a value well below the pt threshold for jets we consider in our analysis.
Can you confirm this?
If we require jets to be above 30,40 GeV in our signal region, should we then set qCut=20 GeV?

Best regards,
Alena

Dear Alena,

Thanks for the event file. I've showered it using the default mg5-aMC interface to pythia8. The inputs, log file and resulting jet1 pT plot can be found here
http://users.ph.tum.de/ga63zot/alena/

As you can see, there is no peak at 50 GeV (the merging scale used). Can you try with this pythia8 input file? Are you using a recent version of pythia8? I turned off hadronisation, MPI and Higgs decays to have a very clear, perturbative description only.

B.t.w., you should NOT limit the merging scale to be below the jet cut. What you use at the analysis level is unrelated to what you use as merging scale.

Best,
Rikkert

Dear Rikkert,

thanks for this, I'll try to reproduce your results.

One question/comment on the merging scale choice:
Just to be clear I didn't mean to set ptj>qCut on generation level, but rather in the analysis. If we generate events with qCut=20 GeV, ptj=10 GeV and then require ptj>40 GeV in our analysis signal region, we assume that matrix element jets give a proper description for our observable. Something similar is also discussed in https://arxiv.org/pdf/1209.6215.pdf around figure 1, where there's also a kink visible in the pt lead jet distibution. Can you maybe explain a bit more why you would not recommend this?

Best regards,
Alena

Dear Alena,

Figure 1 (and also figure 2) of 1209.6215 is not what's actually used in FxFx as currently implemented. Instead, all predictions include the Sudakov reweighting as used in e.g. figure 3 and all other figures of that paper. As you can see in the latter figures, as well as the plot I created from your LHE event file, there is hardly any bump/peak/shoulder visible around the merging scale. In other words, there must still be something wrong with your setup for the showering of the LHE events.

One cannot assume that using the matrix elements is the best description for jets with a small pT (as compared to the typical hard scale of the process). Hence, it's a priori not clear if a jet with a transverse momentum around 30 GeV is better described by the matrix elements or by the parton shower in Higgs production. Hence, one should explore the dependence of the predictions on the merging scale (which effectively determines which region is described by matrix elements and which by parton showers) to assign an uncertainty to the predictions. Hence, typically one would take values of the merging scale above and below the analysis level cuts on the jets.

Best,
Rikkert

Dear Rikkert,

To follow up on this: Our problem with the peak is solved when we apply the correct mc event weights. You can find the resulting ptj0 distribution here: /afs/cern.ch/user/a/aloesle/public/MC_ggH2Np/ptj0_mcweights.eps, it look much more reasonable now.

Thanks again for your help and effort!

Best regards,
Alena