Madspin with LO production

Asked by Lailin Xu

Hi,

I'm running VBF Higgs ->ZZ->4l with MG5, with off-shell Higgs effects included. For example, I use the syntax:
p p > h > j j e+ e- mu+ mu- QCD=0 QED=6

To speed up the event generation, I want to produce Higgs ->ZZ decay, but no further Z boson decay, and use
MadSpin for Z boson decay. I use the syntax:
generate p p > h > j j z z QCD=0 QED=4
and then with madspin:
decay z > e+ e-; decay z > mu+ mu-

Since I'm more interested in the off-shell region, so on-shell Z boson here is fine with me.

But when I compare the kinematics of the generated LHE events with the two above methods, I find there
is some obvious difference in kinematics, like lepton pt, eta and even the 4lepton invariant mass is different (in
high mass region, > 220 GeV for example).
I'm wondering if the difference is caused by MadSpin, due to the spin correlation effect.

Thanks for your help in advance! Btw, I used sm model, not heft.

Lailin

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Lailin Xu
Solved:
Last query:
Last reply:

This question was reopened

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#1

Hi,

If you want to check that MadSpin is doing the correct result for LO, you need to compare your result with the following syntax:
> generate p p > h > j j z z QCD=0 QED=4 , z > e+ e-, z > mu+ mu-

I expect that the above syntax should be in agreement with MadSpin.

The difference with that syntax and this one:
> p p > h > j j e+ e- mu+ mu- QCD=0 QED=6

is exactly the difference of forcing of having only Z diagram (close to be on shell) which is only valid in the Narrow-width-approximation

Cheers,

Olivier

> On Dec 6, 2016, at 17:23, Lailin Xu <email address hidden> wrote:
>
> New question #406515 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/406515
>
> Hi,
>
> I'm running VBF Higgs ->ZZ->4l with MG5, with off-shell Higgs effects included. For example, I use the syntax:
> p p > h > j j e+ e- mu+ mu- QCD=0 QED=6
>
> To speed up the event generation, I want to produce Higgs ->ZZ decay, but no further Z boson decay, and use
> MadSpin for Z boson decay. I use the syntax:
> generate p p > h > j j z z QCD=0 QED=4
> and then with madspin:
> decay z > e+ e-; decay z > mu+ mu-
>
> Since I'm more interested in the off-shell region, so on-shell Z boson here is fine with me.
>
> But when I compare the kinematics of the generated LHE events with the two above methods, I find there
> is some obvious difference in kinematics, like lepton pt, eta and even the 4lepton invariant mass is different (in
> high mass region, > 220 GeV for example).
> I'm wondering if the difference is caused by MadSpin, due to the spin correlation effect.
>
> Thanks for your help in advance! Btw, I used sm model, not heft.
>
> Lailin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#2

Just want to add one point.

MadSpin does not have any cut on the final state particle.
So in order to have a fair comparison, you have to not have any cut on the other syntax as well.

Cheers,

Olivier

> On Dec 6, 2016, at 21:47, Olivier Mattelaer <email address hidden> wrote:
>
> Question #406515 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/406515
>
> Status: Open => Answered
>
> Olivier Mattelaer proposed the following answer:
> Hi,
>
> If you want to check that MadSpin is doing the correct result for LO, you need to compare your result with the following syntax:
>> generate p p > h > j j z z QCD=0 QED=4 , z > e+ e-, z > mu+ mu-
>
>
> I expect that the above syntax should be in agreement with MadSpin.
>
> The difference with that syntax and this one:
>> p p > h > j j e+ e- mu+ mu- QCD=0 QED=6
>
> is exactly the difference of forcing of having only Z diagram (close to
> be on shell) which is only valid in the Narrow-width-approximation
>
> Cheers,
>
> Olivier
>
>> On Dec 6, 2016, at 17:23, Lailin Xu <email address hidden> wrote:
>>
>> New question #406515 on MadGraph5_aMC@NLO:
>> https://answers.launchpad.net/mg5amcnlo/+question/406515
>>
>> Hi,
>>
>> I'm running VBF Higgs ->ZZ->4l with MG5, with off-shell Higgs effects included. For example, I use the syntax:
>> p p > h > j j e+ e- mu+ mu- QCD=0 QED=6
>>
>> To speed up the event generation, I want to produce Higgs ->ZZ decay, but no further Z boson decay, and use
>> MadSpin for Z boson decay. I use the syntax:
>> generate p p > h > j j z z QCD=0 QED=4
>> and then with madspin:
>> decay z > e+ e-; decay z > mu+ mu-
>>
>> Since I'm more interested in the off-shell region, so on-shell Z boson here is fine with me.
>>
>> But when I compare the kinematics of the generated LHE events with the two above methods, I find there
>> is some obvious difference in kinematics, like lepton pt, eta and even the 4lepton invariant mass is different (in
>> high mass region, > 220 GeV for example).
>> I'm wondering if the difference is caused by MadSpin, due to the spin correlation effect.
>>
>> Thanks for your help in advance! Btw, I used sm model, not heft.
>>
>> Lailin
>>
>> --
>> You received this question notification because you are an answer
>> contact for MadGraph5_aMC@NLO.
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Lailin Xu (xlltoade) said :
#3

Thanks Olivier Mattelaer, that solved my question.

Revision history for this message
Lailin Xu (xlltoade) said :
#4

Hi Olivier,

Many thanks for your reply.
I'm aware of that MadSpin does not apply any filters that are defined in my proc card. I already applied some cuts
when I did the LHE analysis to compare the two.
But as you mentioned, it's very likely that the difference is due to the other diagrams without two intermediate Z
bosons.
I'll try "generate p p > h > j j z z QCD=0 QED=4 , z > e+ e-, z > mu+ mu- " and compare the results with MadSpin.

Thanks,
Lailin

Revision history for this message
Lailin Xu (xlltoade) said :
#5

Hi Olivier,

So I did some comparison with the following 3:
1) full decay with MadGraph
generate p p > h > j j mu+ mu- mu+ mu- QCD=0 QED=6
add process p p > h > j j e+ e- e+ e- QCD=0 QED=6
add process p p > h > j j e+ e- mu+ mu- QCD=0 QED=6

2) decay with MadSpin
generate p p > h > j j z z QCD=0 QED=4
decay z > l+ l-; decay z > l+ l-

3) decay with MadGraph
generate p p > j j z z QCD=0 QED=4, z > e+ e-, z > mu+ mu-
add process p p > j j z z QCD=0 QED=4, z > e+ e-, z > e+ e-
add process p p > j j z z QCD=0 QED=4, z > mu+ mu-, z > mu+ mu-

Here are the cross section numbers I got:
1) 0.0004933 +- 8.639e-07 pb
2) 0.000129936411292 pb
3) 0.0001091 +- 3.424e-07 pb

I understand that 1) has more diagrams, but 2) gives higher cross section than 3), which is a little strange to me.

I also compared the four lepton invariant mass of the 3, after requiring some kinematics cuts ( lepton pt, eta, dilepton mass, etc)
to make sure they are in the same phase space.
The plot can be found below and you can see the difference.
https://dl.dropboxusercontent.com/u/6608183/HZZ/ZZ4l/mass_ZZ_VBF_MZ2_madspin_VBF_2e2mu_H4l_baseline_log.png
[The 3 colors are: black for 1), red for 2) and blue for 3).]

I don't understand the difference and I'm not sure which is the best one to use. I guess option 1) is always the best, but it's really
CPU consuming.
Do you have any suggestions?

Thanks a lot!
Lailin

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#6

Hi,

I guess that your problem are related to cut present for 3 which are not present for 2.
From what you say, you apply the same cut for doing the plot but it is not clear if you check the cut at production time.

I have performed the computation for 2 and 3)
and I get
for 2: 0.0005811
for 3: 0.0005527
So we have a 5% difference which sounds reasonable (mainly due to the error related to the narrow-width approximation of MadSpin to get the cross-section)

> I'm not sure which is the best one to use.

I can not answer that question since this depends of the analysis that you are performing.
and of the cut that you are putting on your analysis.
For example, syntax 1 will have a huge contribution at 125 GeV when the Higgs is on-shell. Contribution which is missed by 2/3.
So it really depends of your analysis cut and analysis strategy.

Cheers,

Olivier

Revision history for this message
Lailin Xu (xlltoade) said :
#7

Hi Olivier,

Sorry to bother you again. But I'm still confused from what I see for 1) and 3).
I tried to set bwcutoff as a very large number (10000000).
For 1) the cross section number is unchanged, 0.0004933 +- 8.639e-07 pb.
For 3), the cross section changed from 0.0001325 +- 3.965e-07 pb (bwcutoff=15) to 0.000679 +- 2.909e-06 pb.

For 3), with bwcutoff at 10000000, I can now see both on-shell and off-shell Higgs. But I don't understand why
the cross section becomes even larger than 1). I compared the kinematic shapes and they look consistent, only
the cross sections are different.

Do you have any idea?

Thanks,
Lailin

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#8

Hi,

This is likely because you have cut on the lepton in the case of 1)
that you do not have in the case of 3)
if you change cut_decays in the run_card from False to True, then you should get your agreement.
(assuming that you have all the order input the same in the run_card)

Cheers,

Olivier

Revision history for this message
Lailin Xu (xlltoade) said :
#9

Hi Olivier,

Thanks for your reply.
Indeed it's due to the lepton cuts. Though I already set cut_decays to False in the run_card, somehow the cuts
were still applied. So I manually changed the cuts.

Now I set bwcutoff to be 10000000 for 1) and 3) to be consistent and set cut_decays to False for both of them.
The cross section numbers are now closer
1) 0.0007293 +- 1.193e-06 pb
3) 0.000679 +- 2.909e-06 pb

But the event kinematics are still a little different.
You can find the LHE files I generated and also some distribution comparison below:
https://www.dropbox.com/sh/vev0pjxdpx9brts/AACWrh2RfIw9vpEYVOhaNbSfa?dl=0
The plots are made with the same cuts (lepton pt, eta, m4l> 220 GeV).

Could you take a look at my plots and LHE files to see what could be the reason for the difference?
Really sorry to keep asking you those questions, but I really would like to understand the problem.

Thanks,
Lailin

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#10

Hi,

You do not have the same Feynman Rules for both diagram definitions.
So this can clearly include difference both in shape and cross-section.
Additionally if you use the default dynamical scale of Madgraph, then the scale is defined via a CKKW merging algorithm which is Feynman diagram dependent. So having additional Feynam diagram can lead to different scale choice. (therefore the difference of
cross-section should be compare to the scale uncertainty)

Cheers,

Olivier

Revision history for this message
Lailin Xu (xlltoade) said :
#11

Hi Olivier,

Thanks for your reply.
Yes, I understand that the different Feynman rules could lead to difference in cross sections and shapes.
But the puzzling part to me is that, I checked the diagrams generated for the two syntaxes, and they are
exactly the same. So I'd expect the cross sections and shapes should be consistent in the same phase space.

Thanks,
Lailin

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#12

Hi,

I checked as well and they are not the same.
You have those diagram for syntax 1: q q > z, (z > h Z, h > Z Z)
that does not appear with syntax 3

Cheers,

Olivier

> On Dec 12, 2016, at 20:15, Lailin Xu <email address hidden> wrote:
>
> Question #406515 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/406515
>
> Lailin Xu posted a new comment:
> Hi Olivier,
>
> Thanks for your reply.
> Yes, I understand that the different Feynman rules could lead to difference in cross sections and shapes.
> But the puzzling part to me is that, I checked the diagrams generated for the two syntaxes, and they are
> exactly the same. So I'd expect the cross sections and shapes should be consistent in the same phase space.
>
> Thanks,
> Lailin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Lailin Xu (xlltoade) said :
#13

Hi Olivier,

Sorry, but I do see both WH and ZH processes for syntax 3, as you can see below:
https://www.dropbox.com/s/xljmdb64pq7bmr7/VBF_ZZ4l_sigBW_125.0_W0.004097_gridpack_Run_4_P1_qq_qqzz_z_ll_z_ll.pdf?dl=0

Thanks,
Lailin

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#14

Ok my bad.

But the dynamical scale choice will still differ in both computation.

Cheers,

Olivier
> On Dec 12, 2016, at 22:43, Lailin Xu <email address hidden> wrote:
>
> Question #406515 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/406515
>
> Lailin Xu posted a new comment:
> Hi Olivier,
>
> Sorry, but I do see both WH and ZH processes for syntax 3, as you can see below:
> https://www.dropbox.com/s/xljmdb64pq7bmr7/VBF_ZZ4l_sigBW_125.0_W0.004097_gridpack_Run_4_P1_qq_qqzz_z_ll_z_ll.pdf?dl=0
>
> Thanks,
> Lailin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Lailin Xu (xlltoade) said :
#15

Hi Olivier,

I think finally I figured out the problem, i.e. why I saw different event kinematics between the two syntax.
1) generate p p > h > j j mu+ mu- mu+ mu- QCD=0 QED=6
3) generate p p > j j z z QCD=0 QED=4, z > mu+ mu-, z > mu+ mu-

It was simply due to that I missed the event weights. After including event weights, then they agree very well.
Now the question is, how should I get weighted events? I run the gridpack and used the gridpack to generate LHE events.
So I assume those should be unweighted events. But when I check the events, I do see they have different weights.
You can check the LHE files I sent last time.
https://www.dropbox.com/sh/vev0pjxdpx9brts/AACWrh2RfIw9vpEYVOhaNbSfa?dl=0

Do you know the reason for that?

Thanks,
Lailin

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#16

Hi,

I noticed that you are using a very old version of the code (2.3.3)
For your events file you are likely using a file that we use to generate the unweighted sample.

Cheers,

Olivier

> On Dec 15, 2016, at 08:47, Lailin Xu <email address hidden> wrote:
>
> Question #406515 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/406515
>
> Lailin Xu posted a new comment:
> Hi Olivier,
>
> I think finally I figured out the problem, i.e. why I saw different event kinematics between the two syntax.
> 1) generate p p > h > j j mu+ mu- mu+ mu- QCD=0 QED=6
> 3) generate p p > j j z z QCD=0 QED=4, z > mu+ mu-, z > mu+ mu-
>
> It was simply due to that I missed the event weights. After including event weights, then they agree very well.
> Now the question is, how should I get weighted events? I run the gridpack and used the gridpack to generate LHE events.
> So I assume those should be unweighted events. But when I check the events, I do see they have different weights.
> You can check the LHE files I sent last time.
> https://www.dropbox.com/sh/vev0pjxdpx9brts/AACWrh2RfIw9vpEYVOhaNbSfa?dl=0
>
> Do you know the reason for that?
>
> Thanks,
> Lailin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Lailin Xu (xlltoade) said :
#17

Hi Olivier,

I think my problem is that I used the gridpack to generate events, and from the run.sh file:

if [[ -e ./Events/GridRun_${seed}/unweighted_events.lhe.gz ]]; then
        gunzip ./Events/GridRun_${seed}/unweighted_events.lhe.gz
fi

if [[ ! -e ./Events/GridRun_${seed}/unweighted_events.lhe ]]; then
    echo "Error: event file not found !"
    exit
else
    echo "Moving events from events.lhe"
    mv ./Events/GridRun_${seed}/unweighted_events.lhe ../events.lhe
    cd ..
fi

The event I got should be unweighted_events.lhe.gz, which is assumed to be unweighted.
But I can try with the latest code to see if I still have the problem.

Thanks,
Lailin

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said :
#18

Hi,

From the gridpack, you should not use the events file in the Events directory.
Actually you should not touch to any of the file created inside the gridpack.
The unweighted events file is actually written outside of the grid pac directory at the same level of the executable.

Cheers,

Olivier

> On Dec 15, 2016, at 17:33, Lailin Xu <email address hidden> wrote:
>
> Question #406515 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/406515
>
> Lailin Xu posted a new comment:
> Hi Olivier,
>
> I think my problem is that I used the gridpack to generate events, and
> from the run.sh file:
>
> if [[ -e ./Events/GridRun_${seed}/unweighted_events.lhe.gz ]]; then
> gunzip ./Events/GridRun_${seed}/unweighted_events.lhe.gz
> fi
>
> if [[ ! -e ./Events/GridRun_${seed}/unweighted_events.lhe ]]; then
> echo "Error: event file not found !"
> exit
> else
> echo "Moving events from events.lhe"
> mv ./Events/GridRun_${seed}/unweighted_events.lhe ../events.lhe
> cd ..
> fi
>
>
> The event I got should be unweighted_events.lhe.gz, which is assumed to be unweighted.
> But I can try with the latest code to see if I still have the problem.
>
> Thanks,
> Lailin
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Revision history for this message
Lailin Xu (xlltoade) said :
#19

Hi Olivier,

Thanks for pointing this out! It looks I made some stupid mistake...

Thanks,
Lailin