Matching parameter choice in model without clear transition between multiplicites

Asked by Denis Rathjens on 2018-09-25

Dear experts,
in a model with a BSM Z' (in this case with a mass of 500 GeV) that only couples to bottom quarks, we are trying to establish the appropriate xqcut and qcut parameters. Unlike in examples like https://www.ipmu.jp/sites/default/files/webfm/pdfs/FocusWeek_10 page 11, in our case, there is no clean transition between the multiplicities, even if the sum of them all appears smooth. Is this an issue or is the matching still valid, when
1) the sum of multiplicites is smooth and
2) qcut variations do not further increase the cross section dramatically?

The process is:
define j = j b b~
define p = p b b~
generate p p > zp > j j
add process p p > zp > j j j
add process p p > zp > j j j j

with zp being the particle coupling only to b b~
We use lhapdf 306000
The matching parameters are:
xqcut = 30
maxjetflavor = 5
use_syst=True

JetMatching:qCut = 60 (the variations it automatically added were 90 and 120)
JetMatching:doShowerKt=off
JetMatching:nJetMax=4
SysCalc:FullCutVariation=on

The result is in https://drive.google.com/file/d/1LdZihIQGSrYD9lgydCd5tZMADycgJW1c/view?usp=sharing and we are looking at the transitions 2->3 and 3->4
The lower edge of the green band is the qcut 90 and the upper edge the qcut 120 version, so by that logic, is xqcut30, qcut90 what we ought to choose?

With kind regards,
Denis

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
2018-10-03
Last reply:
2018-10-04

Hi,

In your case nJetMax should be put on 2 and not on 4. The reason for that is that you have two jet not participating to the matching since they (always?) come from the Zp decay.

Now my first question is why do want to allow the zp to be very offshell?
Would not make it more sense to use the following syntax:

> generate p p > zp, zp > j j
> add process p p > zp j, zp > j j
> add process p p > zp j j, zp > j j

Obviously, in that case, radiation from the zp are not included in the matching procedure and should be taken care at 100% by the parton-shower.

The second question is why do you use the lines:
> define j = j b b~
> define p = p b b~

If your model is 5 flavor, you do not need those lines (i.e. those are done automatically).
If your model is 4 flavor then the computation will be ok but you will loose in efficiency.
(Note that converting a 4 flavor model to a 5 flavor model is trivial)

Third question if zp only couples to b b~, Why not using
> generate p p > zp, zp > b b~
> add process p p > zp j, zp > b b~
> add process p p > zp j j, zp > b b~
or
> generate p p > zp > b b~
> add process p p > zp > b b~ j
> add process p p > zp > b b~ j j

Are you also sure that it is better to use a 5flavor matching in that case?

Finally in order to validate the matching/merging, I prefer to first start with the pythia-pgs package.
With that package (+MadAnalysis4+root) the matching plot are done in a very clean way (that PY8 does not allow us to do) that simplify the validation. Can you try?

Cheers,

Olivier

> On 26 Sep 2018, at 01:52, Denis Rathjens <email address hidden> wrote:
>
> New question #674422 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/674422
>
> Dear experts,
> in a model with a BSM Z' (in this case with a mass of 500 GeV) that only couples to bottom quarks, we are trying to establish the appropriate xqcut and qcut parameters. Unlike in examples like https://www.ipmu.jp/sites/default/files/webfm/pdfs/FocusWeek_10 page 11, in our case, there is no clean transition between the multiplicities, even if the sum of them all appears smooth. Is this an issue or is the matching still valid, when
> 1) the sum of multiplicites is smooth and
> 2) qcut variations do not further increase the cross section dramatically?
>
> The process is:
> define j = j b b~
> define p = p b b~
> generate p p > zp > j j
> add process p p > zp > j j j
> add process p p > zp > j j j j
>
> with zp being the particle coupling only to b b~
> We use lhapdf 306000
> The matching parameters are:
> xqcut = 30
> maxjetflavor = 5
> use_syst=True
>
> JetMatching:qCut = 60 (the variations it automatically added were 90 and 120)
> JetMatching:doShowerKt=off
> JetMatching:nJetMax=4
> SysCalc:FullCutVariation=on
>
> The result is in https://drive.google.com/file/d/1LdZihIQGSrYD9lgydCd5tZMADycgJW1c/view?usp=sharing and we are looking at the transitions 2->3 and 3->4
> The lower edge of the green band is the qcut 90 and the upper edge the qcut 120 version, so by that logic, is xqcut30, qcut90 what we ought to choose?
>
> With kind regards,
> Denis
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Denis Rathjens (rathjd) said : #2

Hi!

Regarding the interspersed questions:
1) We used j instead of b b~ in this case, because there's a second coupling b s~ > zp > b s~ that's an FCNC which we want to be able to use with the same syntax for consistency. For the purpose of the initial test, it was set to 0 though.
2) Allowing the Z' to be offshell was based on a misunderstanding on my part what the p p > zp > j j syntax actually does.
3) Regarding 5 flavour-matching, the general advice we received from asking various people was that bottoms are effectively massless for our energy scale of interest, so the 5 flavor scheme would be preferable. So we didn't get a clear answer to this, anywhere, either.

Using the proposed syntax
generate p p > zp, zp > j j
add process p p > zp j, zp > j j
add process p p > zp j j, zp > j j

Merging scale Cross-section [pb] MC uncertainty [pb]
45 1.722538e+01 7.12e-02
67.5 1.844202e+01 7.36e-02
90 1.882250e+01 7.44e-02

works with nJetMax=2. The old syntax crashes with that setting in pythia8.
The resulting matching plots are https://drive.google.com/file/d/1M-gLiB5wUQooV8202d1LNRl8ED-0y47e/view?usp=sharing and there's still no clear non-overlap, although the curves are again, smooth.

There's still the question though why it's valid to leave the FSR from the resonance's decay products to the parton shower.
The associated bottom jets from the initial state are very soft while the resonance's decay products are comparatively very hard. Using the following syntax instead increases the cross section by a factor of 2:

generate p p > zp, zp > j j
add process p p > zp j, zp > j j
add process p p > zp, zp > j j j
add process p p > zp j, zp > j j j
add process p p > zp j j, zp > j j
add process p p > zp, zp > j j j j

Merging scale Cross-section [pb] MC uncertainty [pb]
45 3.164708e+01 1.29e-01
67.5 3.592874e+01 1.38e-01
90 3.756187e+01 1.41e-01

And the matching plots: https://drive.google.com/file/d/1GOU_HlprmmmPtJJy45YNs65QnCIgrx_j/view?usp=sharing and cross section match the original plots in the original request rather accurately.
So which way is now correct and why?

Cheers,
Denis

Hi,

FSR from onshell resonance are NOT include in the merging procedure.
Therefore your second syntax suffer from double counting.
If you want to use such generated sample of events, you have to apply yourself some kind of merging.
(I would actually use in that case a second merging scale for those FSR) .
Adding both the Sudakov form factor and the merging cut.
Note that since the Parton-shower has the constraint to keep the mass of the resonance on-shell.
You have to be smart in the way you prevent the double-counting (I do not know how to do that and if that was ever done in a consistent way actually)

Cheers,

Olivier

> On 4 Oct 2018, at 00:52, Denis Rathjens <email address hidden> wrote:
>
> Question #674422 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/674422
>
> Status: Answered => Open
>
> Denis Rathjens is still having a problem:
> Hi!
>
> Regarding the interspersed questions:
> 1) We used j instead of b b~ in this case, because there's a second coupling b s~ > zp > b s~ that's an FCNC which we want to be able to use with the same syntax for consistency. For the purpose of the initial test, it was set to 0 though.
> 2) Allowing the Z' to be offshell was based on a misunderstanding on my part what the p p > zp > j j syntax actually does.
> 3) Regarding 5 flavour-matching, the general advice we received from asking various people was that bottoms are effectively massless for our energy scale of interest, so the 5 flavor scheme would be preferable. So we didn't get a clear answer to this, anywhere, either.
>
> Using the proposed syntax
> generate p p > zp, zp > j j
> add process p p > zp j, zp > j j
> add process p p > zp j j, zp > j j
>
> Merging scale Cross-section [pb] MC uncertainty [pb]
> 45 1.722538e+01 7.12e-02
> 67.5 1.844202e+01 7.36e-02
> 90 1.882250e+01 7.44e-02
>
> works with nJetMax=2. The old syntax crashes with that setting in pythia8.
> The resulting matching plots are https://drive.google.com/file/d/1M-gLiB5wUQooV8202d1LNRl8ED-0y47e/view?usp=sharing and there's still no clear non-overlap, although the curves are again, smooth.
>
> There's still the question though why it's valid to leave the FSR from the resonance's decay products to the parton shower.
> The associated bottom jets from the initial state are very soft while the resonance's decay products are comparatively very hard. Using the following syntax instead increases the cross section by a factor of 2:
>
> generate p p > zp, zp > j j
> add process p p > zp j, zp > j j
> add process p p > zp, zp > j j j
> add process p p > zp j, zp > j j j
> add process p p > zp j j, zp > j j
> add process p p > zp, zp > j j j j
>
> Merging scale Cross-section [pb] MC uncertainty [pb]
> 45 3.164708e+01 1.29e-01
> 67.5 3.592874e+01 1.38e-01
> 90 3.756187e+01 1.41e-01
>
> And the matching plots: https://drive.google.com/file/d/1GOU_HlprmmmPtJJy45YNs65QnCIgrx_j/view?usp=sharing and cross section match the original plots in the original request rather accurately.
> So which way is now correct and why?
>
> Cheers,
> Denis
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi,

For the validation of the first syntax, I would first try with PY6.
I found the PY8 validation plot very confusing and not very informative.
While the merging scale is validated with PY6 then you should be able to use the same with PY6 in a quite confident way.

Cheers,

Olivier

> On 4 Oct 2018, at 09:16, Olivier Mattelaer <email address hidden> wrote:
>
> Hi,
>
> FSR from onshell resonance are NOT include in the merging procedure.
> Therefore your second syntax suffer from double counting.
> If you want to use such generated sample of events, you have to apply yourself some kind of merging.
> (I would actually use in that case a second merging scale for those FSR) .
> Adding both the Sudakov form factor and the merging cut.
> Note that since the Parton-shower has the constraint to keep the mass of the resonance on-shell.
> You have to be smart in the way you prevent the double-counting (I do not know how to do that and if that was ever done in a consistent way actually)
>
> Cheers,
>
> Olivier
>
>
>> On 4 Oct 2018, at 00:52, Denis Rathjens <email address hidden> wrote:
>>
>> Question #674422 on MadGraph5_aMC@NLO changed:
>> https://answers.launchpad.net/mg5amcnlo/+question/674422
>>
>> Status: Answered => Open
>>
>> Denis Rathjens is still having a problem:
>> Hi!
>>
>> Regarding the interspersed questions:
>> 1) We used j instead of b b~ in this case, because there's a second coupling b s~ > zp > b s~ that's an FCNC which we want to be able to use with the same syntax for consistency. For the purpose of the initial test, it was set to 0 though.
>> 2) Allowing the Z' to be offshell was based on a misunderstanding on my part what the p p > zp > j j syntax actually does.
>> 3) Regarding 5 flavour-matching, the general advice we received from asking various people was that bottoms are effectively massless for our energy scale of interest, so the 5 flavor scheme would be preferable. So we didn't get a clear answer to this, anywhere, either.
>>
>> Using the proposed syntax
>> generate p p > zp, zp > j j
>> add process p p > zp j, zp > j j
>> add process p p > zp j j, zp > j j
>>
>> Merging scale Cross-section [pb] MC uncertainty [pb]
>> 45 1.722538e+01 7.12e-02
>> 67.5 1.844202e+01 7.36e-02
>> 90 1.882250e+01 7.44e-02
>>
>> works with nJetMax=2. The old syntax crashes with that setting in pythia8.
>> The resulting matching plots are https://drive.google.com/file/d/1M-gLiB5wUQooV8202d1LNRl8ED-0y47e/view?usp=sharing and there's still no clear non-overlap, although the curves are again, smooth.
>>
>> There's still the question though why it's valid to leave the FSR from the resonance's decay products to the parton shower.
>> The associated bottom jets from the initial state are very soft while the resonance's decay products are comparatively very hard. Using the following syntax instead increases the cross section by a factor of 2:
>>
>> generate p p > zp, zp > j j
>> add process p p > zp j, zp > j j
>> add process p p > zp, zp > j j j
>> add process p p > zp j, zp > j j j
>> add process p p > zp j j, zp > j j
>> add process p p > zp, zp > j j j j
>>
>> Merging scale Cross-section [pb] MC uncertainty [pb]
>> 45 3.164708e+01 1.29e-01
>> 67.5 3.592874e+01 1.38e-01
>> 90 3.756187e+01 1.41e-01
>>
>> And the matching plots: https://drive.google.com/file/d/1GOU_HlprmmmPtJJy45YNs65QnCIgrx_j/view?usp=sharing and cross section match the original plots in the original request rather accurately.
>> So which way is now correct and why?
>>
>> Cheers,
>> Denis
>>
>> --
>> You received this question notification because you are an answer
>> contact for MadGraph5_aMC@NLO.
>

Can you help with this problem?

Provide an answer of your own, or ask Denis Rathjens for more information if necessary.

To post a message you must log in.