deciphering dominant contributions

Hi,

> i.e. for G1300 i have diagram 57 of matrix9.ps, 54 of matrix10.ps, … contributing...

It is correct.

Now two remarks:

1) this does not take into account the .ddd number information which correspond to a slicing of the phase-space where some of the propagator
are forbidden to be on shell. This code is is a ternary code stating which propagator is on shell/below resonances/above resonances.
Since you have off-shell contribution, using constraint putting particle on shell might indeed not catch this contribution

2) This correspond to a phase-space parametrization which dominates. This channel can be high not only because of that diagram but also because of some other diagram (some of them can be completely different structure). By using the number associated to those directory, you can not claim that this is the dominant diagram.

Cheers,

Olivier

> On Jun 20, 2016, at 18:57, Tania Robens <email address hidden> wrote:
>
> New question #295497 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/295497
>
> Dear authors,
>
> just briefly, to see I understand it correctly
>
> if I want to do a mapping of the Gxxx. numbers to matrixyy.ps diagrams, i do the following:
>
> - Look at the results output; lets say I have identified G1300.ddd as the dominant contribution
>
> - then go to SubProcesses/P0_.../config_subproc_map.inc; here i see sth like
>
> DATA (CONFSUB(I,1300),I=1,18)/0,0,0,0,0,0,0,0,57,54,0,0,0,0,0,0
> $ ,7,7/
>
>
> i.e. for G1300 i have diagram 57 of matrix9.ps, 54 of matrix10.ps, ... contributing...
>
> correct ??
>
> I am asking because it looks quite counterintuitive, and when I try to reproduce this by constraining intermediate state (only for particles which for the *full* process are onshell, tested w madanalysis), i get numbers which are quite different...
>
> ?? Is there any way to further test this, e.g. from info in the G1300.ddd subdirectory ? eg. the results.dat there looks like
>
> 0.21730E+00 0.00000E+00 0.58000E+03 2040000 16 8 2 0.920E+01 0.14951E-09 0.21730E+00
> 6 0.49698E-04 0.18947E-04 0.19290E+03 0.32122E+01 0.49698E-04
> 7 0.19941E+01 0.19903E+01 0.71417E+03 0.10190E+07 0.19941E+01
> 8 0.24496E-02 0.21123E-02 0.87261E+03 0.21552E+04 0.24496E-02
>
> ??
>
> I can send/ attach further files (note the total run for this was quite long, so reproducing things might take quite a bit...)
>
> Let me know Best Tania
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier

answers to 1) and 2) above

>> 1) this does not take into account the .ddd number information which correspond to a slicing of the phase-space where some of the propagator
are forbidden to be on shell. This code is is a ternary code stating which propagator is on shell/below resonances/above resonances.
Since you have off-shell contribution, using constraint putting particle on shell might indeed not catch this contribution

--------------------------

a priori i agree, but i tested this plotting the invariant mass distributions of the events from the *full* process. e.g. in my scenario I have (some) diagrams with h-> gg, and then you can plot m (gg) to see whether these stem from on shell h's... which they do in my case. I am *not* restricting things in my full sample, but then, in hindsight, I test whether the full sample dominantly gives my gg from an onshell h, via madanalysis... which it (at least seemingly) does. So i should get the same number using pp > ... h, h> gg. right ?

2) dont understand your answer. does it mean that from my sample above, even if i look at all diagrams (57 of matrix9.ps, etc), I might still miss additional diagrams which are not listed in config_subproc_map.inc ? where can I find this information ?

Please clarify...

Thanks Tania

Hi Tania,

> a priori i agree, but i tested this plotting the invariant mass
> distributions of the events from the *full* process. e.g. in my scenario
> I have (some) diagrams with h-> gg, and then you can plot m (gg) to see
> whether these stem from on shell h's... which they do in my case. I am
> *not* restricting things in my full sample, but then, in hindsight, I
> test whether the full sample dominantly gives my gg from an onshell h,
> via madanalysis... which it (at least seemingly) does. So i should get
> the same number using pp > … h, h> gg. right ?

In most of the case, I expect this to be True. If it is not please check if you
apply the same cut in both sample. In particular, check the value of cut_decays
since depending of that switch the cut on the gluon coming from h > g g will be applied or not.

> 2) dont understand your answer. does it mean that from my sample above,
> even if i look at all diagrams (57 of matrix9.ps, etc), I might still
> miss additional diagrams which are not listed in config_subproc_map.inc
> ? where can I find this information ?

The easiest is that you look at the slide 79 of the following lecture:
https://cp3.irmp.ucl.ac.be/projects/madgraph/attachment/wiki/Shanghai2015/Tutorial_shangai_basic_mg5.pdf
This defines the number computed around those GX number.
If you look at the slide before you will see more details why we use such definition.

Now in that slide, I make believe (for the simplicity) that the sum is over all diagram, which is actually not True.
The only crucial point is that we have the same sum over i and over j. So indeed some diagram might not be directly associated to a channel of integration.
(Like the interference is never associated to a channel of integration).
The total cross-section is still correct, and their contribution is taken into account but this contribution is spread over multiple channel of integration.

Cheers,

Olivier

> On Jun 20, 2016, at 20:52, Tania Robens <email address hidden> wrote:
>
> Question #295497 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/295497
>
> Status: Answered => Open
>
> Tania Robens is still having a problem:
> Hi Olivier
>
> answers to 1) and 2) above
>
>>> 1) this does not take into account the .ddd number information which correspond to a slicing of the phase-space where some of the propagator
> are forbidden to be on shell. This code is is a ternary code stating which propagator is on shell/below resonances/above resonances.
> Since you have off-shell contribution, using constraint putting particle on shell might indeed not catch this contribution
>
>
> --------------------------
>
> a priori i agree, but i tested this plotting the invariant mass
> distributions of the events from the *full* process. e.g. in my scenario
> I have (some) diagrams with h-> gg, and then you can plot m (gg) to see
> whether these stem from on shell h's... which they do in my case. I am
> *not* restricting things in my full sample, but then, in hindsight, I
> test whether the full sample dominantly gives my gg from an onshell h,
> via madanalysis... which it (at least seemingly) does. So i should get
> the same number using pp > ... h, h> gg. right ?
>
> 2) dont understand your answer. does it mean that from my sample above,
> even if i look at all diagrams (57 of matrix9.ps, etc), I might still
> miss additional diagrams which are not listed in config_subproc_map.inc
> ? where can I find this information ?
>
> Please clarify...
>
> Thanks Tania
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi

are you sure you sent the correct link for your tutorial ? If I open

https://cp3.irmp.ucl.ac.be/projects/madgraph/attachment/wiki/Shanghai2015/Tutorial_shangai_basic_mg5.pdf

there is no slide 79 (also taking onslidexxx into account)... mb you meant a different tutorial ?

And re the cut_decays: I use the same run_card.dat for both cases, and get values that differ by 4 orders of magnitude... can it be that the integration calls need to be increased/ where do I find info whether I can believe the output or not ?

Thanks Best Tania

Hi,

Indeed I wanted to link the lecture to that school not the tutorial. Here is the correct link:
https://cp3.irmp.ucl.ac.be/projects/madgraph/attachment/wiki/Shanghai2015/MG5_Shangai.pdf

> And re the cut_decays: I use the same run_card.dat for both cases,

if you do “g g > g g HIG<=2” or “g g > h > g g HIG<=2” then the gluon will ALWAYS be but (whatever the value of cut_decays)
but if you do
“g g > h HIG<=1, h > g g HIG<=1” then the gluon will not be cut if you set cut_decays=F
Having the same card is not enough to ensure the same cut on different process definition.

> and
> get values that differ by 4 orders of magnitude... can it be that the
> integration calls need to be increased/ where do I find info whether I
> can believe the output or not ?

Do not have enough information to comment on that.

Cheers,

Olivier

> On Jun 21, 2016, at 15:47, Tania Robens <email address hidden> wrote:
>
> Question #295497 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/295497
>
> Status: Answered => Open
>
> Tania Robens is still having a problem:
> Hi
>
> are you sure you sent the correct link for your tutorial ? If I open
>
> https://cp3.irmp.ucl.ac.be/projects/madgraph/attachment/wiki/Shanghai2015/Tutorial_shangai_basic_mg5.pdf
>
> there is no slide 79 (also taking onslidexxx into account)... mb you
> meant a different tutorial ?
>
> And re the cut_decays: I use the same run_card.dat for both cases, and
> get values that differ by 4 orders of magnitude... can it be that the
> integration calls need to be increased/ where do I find info whether I
> can believe the output or not ?
>
> Thanks Best Tania
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi

>> Do not have enough information to comment on that.

which info do you need for this ??

Best Tania

> which info do you need for this ??

So far I do not even know which one of your two number are the smaller.

The ideal would be to have all the information to re-create such process (i.e.
1) the process definition
2) the cuts applied on the particle
3) the benchmark point
4) the model

I do not have that much cpu time available for the moment. So if your process is quite complicated, I will not have time to run it for a while.
If you want to have this faster, I will need:
5) the total cross-section obtained in each scenario
6) the cross-section by channel of integration (only those that you want to discuss)
7) the feynman diagram associated to the channel of integration.

Cheers,

Olivier

> On Jun 21, 2016, at 16:22, Tania Robens <email address hidden> wrote:
>
> Question #295497 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/295497
>
> Status: Answered => Open
>
> Tania Robens is still having a problem:
> Hi
>
>>> Do not have enough information to comment on that.
>
> which info do you need for this ??
>
> Best Tania
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi Olivier -

i suggest to send you 5-7, mb in a gzipped version or so ?? whatever suits you best...

Best Tania

Hi,

I do not care about the format.
You can either change this question in a bug report such that you can attach file.
or you can send me those files on my email address.

Cheers,

Olivier

> On Jun 22, 2016, at 17:27, Tania Robens <email address hidden> wrote:
>
> Question #295497 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/295497
>
> Status: Answered => Open
>
> Tania Robens is still having a problem:
> Hi Olivier -
>
> i suggest to send you 5-7, mb in a gzipped version or so ?? whatever
> suits you best...
>
> Best Tania
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Hi

it seems I had some wrong entries in one of my model files - now things look more sensible. Thanks anyways for the help and feedback !!

Best Tania