photon conversion to e+ e- pairs on fixed target nucleus

Hi,

Looks like you are hitting/probing a singularity, and even if it is finite (is it?) the phase space integrator fails to integrate it correctly.

I would start by first looking at
> generate nuc a > nuc a /nuc

and see if you can have reasonable result for that case (and which cut you have to put to have it work).
Before adding one additional divergency into the game by decaying the photon.

Cheers,

Olivier

On 16 Apr 2015, at 22:01, Iftah Galon <email address hidden> wrote:

> New question #265332 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/265332
>
> Dear mg5amcnlo team,
> I have recently tried to mimic the photon conversion to e+ e- pair process for a fixed target study.
> To do this I have used Feynrules and defined a new model which interfaces with the SM implementation.
> I have defined a complex scalar particle, NUC, e.g the nucleus, which interacts with the photon through QED.
> I have also defined an interaction vertex photon-NUC-NUC which I dressed with the proper form factor structure, following
>
> https://cp3.irmp.ucl.ac.be/projects/madgraph/wiki/FormFactors
>
> I have tried to:
> generate nuc a > nuc e+ e- /nuc
>
> The amplitude for this process is peaked for collimated e+ e- pairs, so I have had to soften some of the cuts in run_card.dat
> Namely, the most effective change is the max eta cut which needs to be changed to ~20 (or alternatively removed with -1)
>
> After making various attempts I find two problems:
> 1. with the standard cuts the x-section is about 200pb with a very small error and the result is stable upon minor tunings of the cuts and through multiple attempts.
> 2. with the modified cuts I can go as high as 10^9-10^10pb but the x-section is somewhat unstable and the error is very large.
>
> Even if believe the somewhat large error result, this is still roughly 3 orders of magnitude below the actual result which is of order of 1barn.
>
> I was wondering if there is a way to ask mg5 to change the required accuracy in the run (I notice a --accuracy option in the interactive mode, but could make it work).
>
> In my runs, I turn off the pdfs lpp1,2 = 0 and the nucleus has the energy of its mass (silicon = 28GeV).
> I scatter an incoming photon at an energy of 100 GeV.
>
> I have also recently encountered
> https://answers.launchpad.net/mg5amcnlo/+question/262196
>
> and was wondering whether a similar cut could be affecting my simulations (though I have no pdfs).
>
> Thank you
> Sincerely
> Iftah
>
>
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Dear Olivier,
Thank you for the quick reply.

I think that this cross section should be finite.
Both Geant4 and XCOM in NIST give finite results for it.

I cannot do as you suggested as this model has a structure function coded in, and the NUC NUC a a vertex turned off.
This is because the structure function takes into account only one photon exchanged with the nucleus, and the nucleus is not allowed to propagate off-shell.
Therefore, with /nuc there are no contributing diagrms.

I think that the problems is that of numerical accuracy and not of probing a singularity.

If I allow the electron mass to be 5GeV instead of 5*10^(-4) GeV then the cross-section converges very fast and is very stable.
Is there a way to change the accuracy goal/percision in madgraph ?

Thank you
Sincerely
Iftah

HI,

I’m really confuse here.
How is it possible that
nuc a > nuc a /nuc
does not have diagram
but
nuc a > nuc e+ e- /nuc
does?
So the e+ e- are not coming from a photon? Are they coming from another particle (like a Z) or via a t-channel structure?

> If I allow the electron mass to be 5GeV instead of 5*10^(-4) GeV then the cross-section converges very fast and is very stable.

This is equivalent to put a cut in the invariant mass of the e+ e-.
If I’m correct, you have a 1/m**4 sensitivity in the mass of the electron.

> Is there a way to change the accuracy goal/percision in madgraph ?

We can not use quadruple precision, On the other hand, you can increase the number of PS points used for the integration.
either by using the “survey” command
(./bin/madevent)
then type help survey to see options

or if you want to generate events, you can modify the code

        #check difficult PS case
        if float(self.run_card['mmjj']) > 0.01 * (float(self.run_card['ebeam1'])+float(self.run_card['ebeam2'])):
            self.pass_in_difficult_integration_mode()

to
        #check difficult PS case
        if True:
            self.pass_in_difficult_integration_mode()

in madevent_interface.py at around line 2637.
That file to modify is either in bin/internal or in madgraph/interface depending of your way to launch MG.

Cheers,

Olivier

On 18 Apr 2015, at 03:56, Iftah Galon <email address hidden> wrote:

> Question #265332 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/265332
>
> Status: Answered => Open
>
> Iftah Galon is still having a problem:
> Dear Olivier,
> Thank you for the quick reply.
>
> I think that this cross section should be finite.
> Both Geant4 and XCOM in NIST give finite results for it.
>
> I cannot do as you suggested as this model has a structure function coded in, and the NUC NUC a a vertex turned off.
> This is because the structure function takes into account only one photon exchanged with the nucleus, and the nucleus is not allowed to propagate off-shell.
> Therefore, with /nuc there are no contributing diagrms.
>
> I think that the problems is that of numerical accuracy and not of
> probing a singularity.
>
> If I allow the electron mass to be 5GeV instead of 5*10^(-4) GeV then the cross-section converges very fast and is very stable.
> Is there a way to change the accuracy goal/percision in madgraph ?
>
> Thank you
> Sincerely
> Iftah
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Dear Olivier,
Sorry for taking long to answer.
Regarding nuc a > nuc a,
the use of a structure function, assumes a single photon exchange with the nucleus
so I cannot have a scattering and use the same function (as far as I know).

Indeed there is a singularity which depends on the invariant mass of the e+e- pair.
I was able to get to a good enough result using a cut on the maximal invariant mass, and the modification you suggested.

I still undershoot by a factor of ~5 but this is a significant improvement.

I guess I could do better if I could better focus the cuts on relevant phase-space region.

I had recently heard a talk by Scott Thomas who mentioned a recent modification you guys made to MG5's run cards.
The additional pt cuts on leading lepton number 1,2,3,4 ....
This help focus the ps sampling for Z->4 leps.

Perhaps a similar modification can help here if one could apply P (or E) cuts to different leptons.
P_t does not play such a role here as the e+e- pairs are collimated with the photon direction.

Thank you very much for all your help
Sincerely
Iftah