EW V + jets NLO

Asked by negin on 2018-06-26

Dear authors,

Is amC@NLO well suited for calculating EW V + jj with NLO corrections ?

with aMC v254 I ran
generate p p > z j j QCD=0 [QCD]

using
  1.0 = jetalgo ! FastJet jet algorithm (1=kT, 0=C/A, -1=anti-kT)
  0.4 = jetradius ! The radius parameter for the jet algorithm
20.0 = ptj ! Min jet transverse momentum
  5.0 = etaj ! Max jet abs(pseudo-rap) (a value .lt.0 means no cut)

and got
aMCatNLOError: Poles do not cancel, run cannot continue

Could you kindly let me know how I should proceed, if possible ?

Best Regards

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
marco zaro Edit question
Solved by:
marco zaro
Solved:
2019-01-14
Last query:
2019-01-14
Last reply:
2019-01-14

This question was reopened

marco zaro (marco-zaro) said : #1

Hi,
you can generate EW Z-production plus two jets at NLO QCD in two ways:
- with the standard (2.6.x) branch, you will not generate loop diagrams featuring EW bosons, and you have to turn off the check poles.
You can find the details (and how to turn them off) e.g. in this paper, where VBS has been studied
https://arxiv.org/pdf/1803.07943.pdf
- with the new, EW-capable branch (3.0.x), you can compute the full thing but only at fixed order (no event genration, no shower), because some corrections stem from EW corrections on top of the alpha^2 alpha_s interference. I fear you need to decay the Z in that case, I would advise you to start with the first option, which should normally be good enough.

Let us know
Cheers,

Marco

negin (neginsh) said : #2

Thanks marco zaro, that solved my question.

negin (neginsh) said : #3

Hi ,
thanks for your respond , I have tried that way and it worked but still one problem, I have large differences in cross-section of NLO and LO. I think 51.80 pb is large for NLO corrections as for LO is 37.42 pb! what do you think?

Cheers
Negin

marco zaro (marco-zaro) said : #4

Hi,
is 51.80 the NLO cross section or just the NLO corrections?
Here
https://arxiv.org/pdf/hep-ph/0310156.pdf
with the renormalization and factorization scales set to mz (and VBF-type cuts), one gets 10% corrections...
can you try to set the scale fixed to mz (you can do this in the run_card)?
Can you please also attach the process definition you used, just to cross-check?
thanks!

cheers,

Marco

negin (neginsh) said : #5

Hi,
 is 51.80 the NLO cross section or just the NLO corrections?
yes, it is NLO cross section( included NLO corrections)

I have attached The cards in following address for cross-check.
"https://github.com/nshafiei/VBF/"
Best regards
Negin

negin (neginsh) said : #6

Hi,

Sorry for sending you again the question because i have not received any respond. would you please check the cards and give me suggestion for this problem?!

Best
Negin

negin (neginsh) said : #7

Hi,

Sorry for sending you again the question because i have not received any respond. would you please check the cards and give me suggestion for this problem?!

Best
Negin

marco zaro (marco-zaro) said : #8

Hi Negin,
I was on hoidays in the past days.
Can you try setting the scale fixed to mz (you can do this directly from the run_card) and see how large the K-factor is?
In any case,the numbers you quoted does not seem pathological...
Let me know
cheers

Marco

negin (neginsh) said : #9

Hi Marco,

Thanks for your kind support.
I have run as you said with fixed scales and i have seen small differences between Fixed one and dynamical one

NLO/LO Fixed is 5.362e01/37.45
NLO/LO Dynamical is 5.184e01/37.42

So the NLO/LO difference is still a bit large.

I also have tried to change the drjj cut running LO (with fixed scales),,
but strangely I see no effect on the total cross-section, for example
LO Fixed is 37.4 with drjj=0.4 and 37.76 with drjj=0.0.

Do you know why the drjj cut does not change the results ?

Cheers
Negin

marco zaro (marco-zaro) said : #10

Hello Negin,
the reason of the large (40%) K factor is that, unless you require large jet-pair invariant mass, then you are dominated by
p p > z z with a Z decaying to jj (also contributions with a virtual photon are there)
Both zz and az production have rather large k-factor, see e.g. table 2 in https://arxiv.org/pdf/1405.0301.pdf
For the drjj, one should have some divergence if drjj->0, because of the contribution due to the virtual photon.
Best,
Marco

negin (neginsh) said : #11

Hello Marco,
Thanks so much , yes the problem was the k factor .

But for the drjj, I don't have have any divergence if drjj->0, i have number 37.76pb. Maybe as result of the cut that i have put for ptj=20?

Best,
Negin

marco zaro (marco-zaro) said : #12

Hi Negin,
I do not know, it can also be that the divergence is not seen by the
integrator or that the divergent contribution is rather small...
Just to understand, when you say LO you generate with or without [QCD]?
best,

Marco

On Wed, Sep 5, 2018 at 2:53 PM negin <email address hidden>
wrote:

> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> negin gave more information on the question:
> Hello Marco,
> Thanks so much , yes the problem was the k factor .
>
> But for the drjj, I don't have have any divergence if drjj->0, i have
> number 37.76pb. Maybe as result of the cut that i have put for ptj=20?
>
> Best,
> Negin
>
> --
> You received this question notification because you are subscribed to
> the question.
>

negin (neginsh) said : #13

Hello Marco,
when i say LO my mean is without [QCD].
 generate p p > z jj QCD=0

Best
Negin

negin (neginsh) said : #14

Dear Marco,

As you said me couple of months ago , for adding a Mjj for NLO in cut.f it seems doesn't work. what do you think?

double precision sumdot

  ! MJJ > 200 GeV
      if (dsqrt(sumdot(pjet(0,1),pjet(0,2),1d0)).lt.200d0) then
        passcuts_user=.false.
        return
      endif

Mjj=10 , 51.84pb AND Mjj= 200 , 52.58pb.

Cheers
Negin

marco zaro (marco-zaro) said : #15

Dear Negin, at which point of cuts.f do you add this cut?
Let me know

best,

Marco

negin (neginsh) said : #16

Dear Marco,
 I have added it in" DeltaR and invariant mass cuts " .

Cheers
Negin

marco zaro (marco-zaro) said : #17

Hi Negin,
DeltaR and invariant mass are for the leptons.
You should add the cut after the fastjet clustering.
In particular right after (before 122 continue)

         call amcatnlo_fastjetppgenkt_etamax_timed(
     $ pQCD,nQCD,rfj,sycut,etaj,palg,pjet,njet,jet)
c
c******************************************************************************

c Apply the jet cuts
         if (njet .ne. nQCD .and. njet .ne. nQCD-1) then
            passcuts_user=.false.
            return
         endif
      endif

This way you will be sure to have at least two jets.
Let me know if this works.

Cheers,

Marco

> On 11 Sep 2018, at 09:42, negin <email address hidden> wrote:
>
> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> Status: Needs information => Open
>
> negin gave more information on the question:
> Dear Marco,
> I have added it in" DeltaR and invariant mass cuts " .
>
> Cheers
> Negin
>
> --
> You received this question notification because you are subscribed to
> the question.

negin (neginsh) said : #18

Dear Marco,
Thanks so much, that approach of mjj cut seems to work well. Still one problem: for mjj >120 with reject the di-bosons
the NLO cross-section is 23.54pb and LO one is 14.23 . (still the k-faktor is very big) What do you think ?

Cheers
Negin

marco zaro (marco-zaro) said : #19

Hi Negin,
I guess what is happening there is that at LO you kill diboson, but at NLO you are dominated by real emission configurations where the two tagging jets are the extra parton and one of those coming from the decaying boson. Note that in this way at NLO the diboson can be on shell, while at LO it cannot…
Can you try with tighter cuts?
cheers,

Marco

> On 13 Sep 2018, at 12:27, negin <email address hidden> wrote:
>
> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> Status: Answered => Open
>
> negin is still having a problem:
> Dear Marco,
> Thanks so much, that approach of mjj cut seems to work well. Still one problem: for mjj >120 with reject the di-bosons
> the NLO cross-section is 23.54pb and LO one is 14.23 . (still the k-faktor is very big) What do you think ?
>
> Cheers
> Negin
>
> --
> You received this question notification because you are subscribed to
> the question.

negin (neginsh) said : #20

Hi Marco,
Thanks so much, I have tried tighter cut as 500 and 700 mjj but for 1000 it takes long time do you have any suggestion for improving that ? The results for 500 and 700 are
MJJ>500 LO=6.68 NLO=7.45pb
MJJ>700 LO=4.77 NLO=5.05pb
so indeed the NLO/LO factor decreases as you predicted.

Stil I have a question about your suggested reason for large NLO/LO corrections with low MJJ.
Since I produce the Z jj final state, with no Z decay, is it that the Z boson is decayed anyhow with MG5_amcnlo
so that hadronic Z modes can contribute to the MJJ condition with kt jets ?
Otherwise the Z decay products would not contribute to the MJJ calculation, correct ?

Thank you for clarifying

Cheers
Negin

marco zaro (marco-zaro) said : #21

Hi Negin,

> On 9 Oct 2018, at 18:12, negin <email address hidden> wrote:
>
> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> Status: Answered => Open
>
> negin is still having a problem:
> Hi Marco,
> Thanks so much, I have tried tighter cut as 500 and 700 mjj but for 1000 it takes long time do you have any suggestion for improving that ? The results for 500 and 700 are
> MJJ>500 LO=6.68 NLO=7.45pb
> MJJ>700 LO=4.77 NLO=5.05pb
> so indeed the NLO/LO factor decreases as you predicted.
the long time is due to the fact that the harder the mjj cut the more difficult is for the phase space to find the region that passes cuts.
>
> Stil I have a question about your suggested reason for large NLO/LO corrections with low MJJ.
> Since I produce the Z jj final state, with no Z decay, is it that the Z boson is decayed anyhow with MG5_amcnlo
> so that hadronic Z modes can contribute to the MJJ condition with kt jets ?
> Otherwise the Z decay products would not contribute to the MJJ calculation, correct ?
no, that z remain stable. However, the two jets that you have can come from another Z or W boson which can become resonant.

Cheers,

Marco

>
>
> Thank you for clarifying
>
> Cheers
> Negin
>
> --
> You received this question notification because you are subscribed to
> the question.

negin (neginsh) said : #22

Dear Marco,

Ok thank you. I understand with the NLO configuration more diboson events can pass because
one of the quarks from a Z or W decay can make a big mass with a ISR parton, right ?

I wanted still to ask you, when I implement the mjj cut as you suggested with

 if (dsqrt(sumdot(pjet(0,1),pjet(0,2),1d0)).lt.200d0) then
        passcuts_user=.false.

how are the pjets ordered ? Am I picking the two with the largest pt values ?
I mean are the pjet(0,1) and pjet(0,2) the two pt leading jets of the event ?

Thank you

marco zaro (marco-zaro) said : #23

Hi Negin,

> On 16 Oct 2018, at 18:07, negin <email address hidden> wrote:
>
> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> Status: Answered => Open
>
> negin is still having a problem:
> Dear Marco,
>
> Ok thank you. I understand with the NLO configuration more diboson events can pass because
> one of the quarks from a Z or W decay can make a big mass with a ISR parton, right ?

yes, this is correct
>
> I wanted still to ask you, when I implement the mjj cut as you suggested
> with
>
> if (dsqrt(sumdot(pjet(0,1),pjet(0,2),1d0)).lt.200d0) then
> passcuts_user=.false.
>
> how are the pjets ordered ? Am I picking the two with the largest pt values ?
> I mean are the pjet(0,1) and pjet(0,2) the two pt leading jets of the event ?

the jets are ordered in pt, so, those are the two with largest pt in the event, as you write.

cheers,

marco
>
> Thank you
>
> --
> You received this question notification because you are subscribed to
> the question.

negin (neginsh) said : #24

Hi Marco,
 sorry for delay, i have actually run VBFNLO package and the differences between NLO/LO is too small but for the previous running on Madgraph we have seen large differneces.
Would please give me suggestion?

VBFNLO with
MDIJ_MIN = 120d0 ! dijet min mass cut on tag jet
RJJ_MIN = 0.4d0 ! min jet-jet R separation
PT_JET_MIN = 20.0d0 ! min jet pT

 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
     VBFNLO 2.7.1

  PROCESS: 120 : p p --> Z jj --> e+ e- jj

  TOTAL result (LO): 488.61713036518131 +- 2.0724201034012082 fb 0.42413987856961310 %
  NLO virtual result: 468.86270841137110 +- 2.2555515943560613 fb 0.48106866976016438 %
  QCD real emission: 12.108097076698595 +- 4.9990375157526206 fb 41.286731383852292 %
  TOTAL result (NLO): 480.97080548806969 +- 5.4843312334964143 fb 1.1402628124031640 %
  K-Factor: 0.98435109126976994
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

and if we multiply the above cross-sections by 30 for p p --> Z jj it is
 sigma(LO)= 14.66pb, sigma(NLO)=14.43pb

Cheers
Negin

marco zaro (marco-zaro) said : #25

Hi Negin,
I don’t know about VBFNLO, but it may be that they only include t-channel diagrams and not the diboson one. You should check this.
best,

Marco
> On 17 Dec 2018, at 17:07, negin <email address hidden> wrote:
>
> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> Status: Answered => Open
>
> negin is still having a problem:
> Hi Marco,
> sorry for delay, i have actually run VBFNLO package and the differences between NLO/LO is too small but for the previous running on Madgraph we have seen large differneces.
> Would please give me suggestion?
>
>
> VBFNLO with
> MDIJ_MIN = 120d0 ! dijet min mass cut on tag jet
> RJJ_MIN = 0.4d0 ! min jet-jet R separation
> PT_JET_MIN = 20.0d0 ! min jet pT
>
>
> !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
> VBFNLO 2.7.1
>
> PROCESS: 120 : p p --> Z jj --> e+ e- jj
>
> TOTAL result (LO): 488.61713036518131 +- 2.0724201034012082 fb 0.42413987856961310 %
> NLO virtual result: 468.86270841137110 +- 2.2555515943560613 fb 0.48106866976016438 %
> QCD real emission: 12.108097076698595 +- 4.9990375157526206 fb 41.286731383852292 %
> TOTAL result (NLO): 480.97080548806969 +- 5.4843312334964143 fb 1.1402628124031640 %
> K-Factor: 0.98435109126976994
> !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
>
>
> and if we multiply the above cross-sections by 30 for p p --> Z jj it is
> sigma(LO)= 14.66pb, sigma(NLO)=14.43pb
>
> Cheers
> Negin
>
> --
> You received this question notification because you are subscribed to
> the question.

negin (neginsh) said : #26

Hi Marco,

Yes in the vbfnlo manual for single V VBF the quoted reference is
https://arxiv.org/abs/hep-ph/0310156
where s-channel diboson contributions are considered negligible (and not included).

We are a bit puzzled that NLO emissions in diboson events can make a large invariant mass (mjj>120)
with one of the partons from on-shell hadronic V decays so frequently: would have to happen in
in a good fraction of all diboson events to obtain the +9pb cross section increase I see, what do you think ?

And on the other hand, at LO the s-channel diboson contributions to the EW Zjj process is very small with mjj>120:
do you think it “makes sense” that this “VBF” signal receives such important NLO corrections from dibosons ?
It seems more like moving into a different signal definition.
Do you think maybe a different definition of mjj at NLO could be more appropriate for this ?

Cheers
Negin

marco zaro (marco-zaro) said : #27

Dear Negin,
well, mjj> 120 is not at all a strong cut.
If you look here
https://arxiv.org/pdf/1803.07943.pdf <https://arxiv.org/pdf/1803.07943.pdf>
figs 4 and 7 show, in their left panel, the ratio t+u channel / complete (s+t+u with interferences), at the LO (fig 4) and NLO (fig 7). This is WW VBS, but it is very similar to EW production of a Z boson
You see that in fig 4 (LO) effects are mild (-10 / 15%) in the low M(jj) and low Dy(jj). At NLO effects can instead be large (-50%). Note that the NLO panel has the Dy(jj) that starts at 2, while at LO it starts at 1. Note also that, for larger Mjj, the VBF approximation works better.
As I said, this is due to the fact that at NLO the two jets do not have to come from the decay products of the other vector boson, and one can be an ISR jet.
So to go back to your question, mjj> 120 is not enough to kill the diboson+jet background at NLO, and a large fraction of the Z+jj events come from this production channel.
Best,

Marco

> On 18 Dec 2018, at 15:52, negin <email address hidden> wrote:
>
> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> Status: Answered => Open
>
> negin is still having a problem:
> Hi Marco,
>
> Yes in the vbfnlo manual for single V VBF the quoted reference is
> https://arxiv.org/abs/hep-ph/0310156
> where s-channel diboson contributions are considered negligible (and not included).
>
> We are a bit puzzled that NLO emissions in diboson events can make a large invariant mass (mjj>120)
> with one of the partons from on-shell hadronic V decays so frequently: would have to happen in
> in a good fraction of all diboson events to obtain the +9pb cross section increase I see, what do you think ?
>
> And on the other hand, at LO the s-channel diboson contributions to the EW Zjj process is very small with mjj>120:
> do you think it “makes sense” that this “VBF” signal receives such important NLO corrections from dibosons ?
> It seems more like moving into a different signal definition.
> Do you think maybe a different definition of mjj at NLO could be more appropriate for this ?
>
> Cheers
> Negin
>
> --
> You received this question notification because you are subscribed to
> the question.

negin (neginsh) said : #28

Hi Marco,
Thanks so much. Things are more clear to me.
I still have one more question. If I would like to try excluding the s-channel contributions
to study the effects what would be the syntax ? what do I have to change ?
Thank you Cheers
Negin

marco zaro (marco-zaro) said : #29

Dear Negin,
to exclude s-channels the syntax is $$ (see below the example for single top)
 generate p p > h t b~ j $$ w+ w- [QCD]

best,

Marco
> On 14 Jan 2019, at 11:33, negin <email address hidden> wrote:
>
> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> Status: Answered => Open
>
> negin is still having a problem:
> Hi Marco,
> Thanks so much. Things are more clear to me.
> I still have one more question. If I would like to try excluding the s-channel contributions
> to study the effects what would be the syntax ? what do I have to change ?
> Thank you Cheers
> Negin
>
> --
> You received this question notification because you are subscribed to
> the question.

negin (neginsh) said : #30

Dear Marco,
thanks so much, but i have to exclude 3 s-channel zw+ zw- and zz , if i want to add them in one line what should i do ?
cheers
Negin

Best marco zaro (marco-zaro) said : #31

The syntax $$ a b c … vetoes diagrams with any of the particle a,b,c appearing as s-channel
Î

> On 14 Jan 2019, at 17:02, negin <email address hidden> wrote:
>
> Question #670477 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/670477
>
> Status: Answered => Open
>
> negin is still having a problem:
> Dear Marco,
> thanks so much, but i have to exclude 3 s-channel zw+ zw- and zz , if i want to add them in one line what should i do ?
> cheers
> Negin
>
> --
> You received this question notification because you are subscribed to
> the question.

negin (neginsh) said : #32

Thanks so much i have done it with comma.

negin (neginsh) said : #33

Thanks marco zaro, that solved my question.