Interfacing LHE files to built-in Pythia

Asked by oleh on 2018-04-17

Dear MadGraph experts,

I apologize but I have to write rather a long introduction to explain my issue.

I develop my own LO generator and I would like to use built-it Pythia 8 to produce showers out of my own LHE files.
I also use MadGraph + Pythia 8 to check that my generator works properly.
As a "hello world" process I consider gg -> gg at LO. The total cross sections and differential distributions produced by my code coincide with those produced by MadGraph, however if I run parton shower I get a small, but non-negligible disagreement between distributions produced out of my LHE files and distributions produced out of MadGraph LHE files.

Here is the way my generator works:
it uses analytical expressions for the total cross sections (in this particular case gg ->gg LO cross section from Peskin-Schröder) averaged over spins and colours and Hit-or-Miss algorithm to produce parton events.

After events being produced my code writes them to a LHE file according to Les-Houches notation.

Since I use spin- and colour-averaged cross sections I do not have any information about spins and colours of colliding particles so in my LHE files I always write down THE SAME combination of colour and spins (taken out of MadGraph LHE files)

I set factorization scale equal to renormalization scale being equal equal to s_hat. I take a value of alphaS(s_hat) out of the LHAPDF and a value of alphaEM out of MadGraph event file.

Here below I have an example of my LHE event
<event>
4 1 1.289329e+07 9.781690e+01 7.546771e-03 1.377373e-01
21 -1 0 0 504 501 0.000000e+00 0.000000e+00 2.590066e+01 2.590066e+01 0.0 0.0 1.0
21 -1 0 0 503 502 0.000000e+00 0.000000e+00 -9.235426e+01 9.235426e+01 0.0 0.0 -1.0
21 1 1 2 503 501 9.460534e-01 -3.594737e+01 -7.330332e+01 8.164855e+01 0.0 0.0 1.0
21 1 1 2 504 502 -9.460534e-01 3.594737e+01 6.849718e+00 3.660638e+01 0.0 0.0 -1.0
</event>

and an example of MadGraph event
<event>
 4 1 +1.2890244e+07 7.97197500e+01 7.54677100e-03 1.42644000e-01
21 -1 0 0 504 501 +0.0000000000e+00 +0.00000000000e+00 +6.3448518807e+01 6.3448518807e+01 0.0 0.0 1.0
21 -1 0 0 503 502 -0.0000000000e+00 -0.00000000000e+00 -2.5040922650e+01 2.5040922650e+01 0.0 0.0 1.0
21 1 1 2 503 501 +3.1995951672e+01 +2.3310158620e+01 +1.4032426072e+01 4.2000159520e+01 0.0 0.0 1.0
21 1 1 2 504 502 -3.1995951672e+01 -2.3310158620e+01 +2.4375170085e+01 4.6489281936e+01 0.0 0.0 1.0
</event>

Total cross sections My code : 1.289329e+07 pb
                                     MadGraph: +1.2890244e+07 pb

Since my LHE file contains fake information about colours I was trying to switch off all processes in Pythia that may influence
shower generation.
Here is the set of flags I use :

PartonLevel:Remnants = off
Check:event = off
ColourReconnection:reconnect = off
PartonLevel:mpi = off
HadronLevel:all = off

However I get distributions that go faster to l pt_min and slower to pt_max as distributions produced out of MadGraph events.

I suspect that this difference comes ether out of the difference in the running of the couplings or out of some colour (spin) dependent processes in Pythia that I forgot to switch off.

So my questions is how one has to "mimic" LHE files to get the same distributions as produced out of MadGraph events?

Also am I right that Pythia will use values of alphaS out of the LHE file to produce showers ?

Is it also correct for alpha_EM ?

Thank you very much.

With best regards,
Oleh

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Olivier Mattelaer
Solved:
2018-04-17
Last query:
2018-04-17
Last reply:
2018-04-17

Hi,

The spin information can actually be left as "summed" in the official lhef convention.
You should check that paper and adpat your events accordingly (not sure if it is 9 or 0)
Pythia is actually not using that information (but other codes can so better to set that number in a coherent way)

The information of the color is actually a non trivial one. Since this correspond to the Leading Color information of the events. Since Pythia heavily use that information this can lead to potentially huge effect if you always provide the same color information.

I would suggest that you start by checking how large is that effect.
(the easiest is probably to modify MG events to put the color to the one that you use in your case)

> I suspect that this difference comes ether out of the difference in the running of the couplings or out of some colour (spin) dependent processes in Pythia that I forgot to switch off.

For this you should contact pythia author.

> Also am I right that Pythia will use values of alphaS out of the LHE file to produce showers ?
>
> Is it also correct for alpha_EM ?

This depends on how you run pythia actually. The starting point for both coupling is given in the events
But pythia will make those to run obviously.

Cheers,

Olivier

> On 17 Apr 2018, at 10:33, oleh <email address hidden> wrote:
>
> New question #668008 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/668008
>
> Dear MadGraph experts,
>
> I apologize but I have to write rather a long introduction to explain my issue.
>
> I develop my own LO generator and I would like to use built-it Pythia 8 to produce showers out of my own LHE files.
> I also use MadGraph + Pythia 8 to check that my generator works properly.
> As a "hello world" process I consider gg -> gg at LO. The total cross sections and differential distributions produced by my code coincide with those produced by MadGraph, however if I run parton shower I get a small, but non-negligible disagreement between distributions produced out of my LHE files and distributions produced out of MadGraph LHE files.
>
> Here is the way my generator works:
> it uses analytical expressions for the total cross sections (in this particular case gg ->gg LO cross section from Peskin-Schröder) averaged over spins and colours and Hit-or-Miss algorithm to produce parton events.
>
> After events being produced my code writes them to a LHE file according to Les-Houches notation.
>
> Since I use spin- and colour-averaged cross sections I do not have any information about spins and colours of colliding particles so in my LHE files I always write down THE SAME combination of colour and spins (taken out of MadGraph LHE files)
>
> I set factorization scale equal to renormalization scale being equal equal to s_hat. I take a value of alphaS(s_hat) out of the LHAPDF and a value of alphaEM out of MadGraph event file.
>
> Here below I have an example of my LHE event
> <event>
> 4 1 1.289329e+07 9.781690e+01 7.546771e-03 1.377373e-01
> 21 -1 0 0 504 501 0.000000e+00 0.000000e+00 2.590066e+01 2.590066e+01 0.0 0.0 1.0
> 21 -1 0 0 503 502 0.000000e+00 0.000000e+00 -9.235426e+01 9.235426e+01 0.0 0.0 -1.0
> 21 1 1 2 503 501 9.460534e-01 -3.594737e+01 -7.330332e+01 8.164855e+01 0.0 0.0 1.0
> 21 1 1 2 504 502 -9.460534e-01 3.594737e+01 6.849718e+00 3.660638e+01 0.0 0.0 -1.0
> </event>
>
> and an example of MadGraph event
> <event>
> 4 1 +1.2890244e+07 7.97197500e+01 7.54677100e-03 1.42644000e-01
> 21 -1 0 0 504 501 +0.0000000000e+00 +0.00000000000e+00 +6.3448518807e+01 6.3448518807e+01 0.0 0.0 1.0
> 21 -1 0 0 503 502 -0.0000000000e+00 -0.00000000000e+00 -2.5040922650e+01 2.5040922650e+01 0.0 0.0 1.0
> 21 1 1 2 503 501 +3.1995951672e+01 +2.3310158620e+01 +1.4032426072e+01 4.2000159520e+01 0.0 0.0 1.0
> 21 1 1 2 504 502 -3.1995951672e+01 -2.3310158620e+01 +2.4375170085e+01 4.6489281936e+01 0.0 0.0 1.0
> </event>
>
> Total cross sections My code : 1.289329e+07 pb
> MadGraph: +1.2890244e+07 pb
>
> Since my LHE file contains fake information about colours I was trying to switch off all processes in Pythia that may influence
> shower generation.
> Here is the set of flags I use :
>
> PartonLevel:Remnants = off
> Check:event = off
> ColourReconnection:reconnect = off
> PartonLevel:mpi = off
> HadronLevel:all = off
>
> However I get distributions that go faster to l pt_min and slower to pt_max as distributions produced out of MadGraph events.
>
> I suspect that this difference comes ether out of the difference in the running of the couplings or out of some colour (spin) dependent processes in Pythia that I forgot to switch off.
>
> So my questions is how one has to "mimic" LHE files to get the same distributions as produced out of MadGraph events?
>
> Also am I right that Pythia will use values of alphaS out of the LHE file to produce showers ?
>
> Is it also correct for alpha_EM ?
>
> Thank you very much.
>
> With best regards,
> Oleh
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

oleh (fedkevych) said : #2

Dear Oliver,

many thanks for finding time to answer my question.

Just one extra question does Pythia use values of couplting

> This depends on how you run pythia actually. The starting point for both coupling is given in the events
Do you mean values of couplings given in para_card or values of couplings given in the first line of each LHE event
(eg 4 1 1.289329e+07 9.781690e+01 7.546771e-03 1.377373e-01)

With best regards,
Oleh

They should use the one given in the event not the one given in top of the file.
Actually even MadGraph typically does always use the value of alpha_s form the param_card.
As soon as a PDF is defined, then the value of the alphas is taken from the PDF rather than from the card.

Cheers,

Olivier

> On 17 Apr 2018, at 12:26, oleh <email address hidden> wrote:
>
> Question #668008 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/668008
>
> Status: Answered => Open
>
> oleh is still having a problem:
> Dear Oliver,
>
> many thanks for finding time to answer my question.
>
> Just one extra question does Pythia use values of couplting
>
>> This depends on how you run pythia actually. The starting point for both coupling is given in the events
> Do you mean values of couplings given in para_card or values of couplings given in the first line of each LHE event
> (eg 4 1 1.289329e+07 9.781690e+01 7.546771e-03 1.377373e-01)
>
> With best regards,
> Oleh
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

oleh (fedkevych) said : #4

Thanks Olivier Mattelaer, that solved my question.