Trouble with overlap in matching of 0/1 jet multiplicites

Hi,

> I have been varying xqcut and qcut using the rule qcut=xqcut+5,

This sounds a too small gap. I typically use at least 10 and even more when qcut is larger than 60GeV.

> It works fine in the sense that the 1-jet case starts at xqcut, but the 0-jet only very slowly dies off, and is of the same order as the 1-jet multiplicity, creating ~ a factor of two spike in my DJR1->0 graph around xqcut, qcut.

Did you use pythia6? If yes, do you have use_syst=T in your run_card?
If you do then the plot need to be done by taking account of the weight of the events (which is not done with the automatic plot) and therefore can lead to what you observe.

> 2) It appears that ATLAS and CMS use ptj1>0 when they are trying this matching process, but setting it non-zero only seems to increase my current problems with the DJR1 graph, lowering the 0-jet sample everywhere, and making it extremely not smooth. Does ptj1>0 remove all 0-jet events, since there are no leading jets, and thus none make the cut?

If you include such ptj1>0 you need to be careful on how you implement the matching/merging, but the detailed should also depend of the comparison between xqcut and ptj1.
Depending of this value, it does not make any sense (or it does) to include the 0j sample). In the case that you did not include it, then obviously the DJR1 will not be smooth at all.

> 3) More of a notation question. The legend lists 0-jet sample, 1-jet sample, etc. Do these events correspond directly to the jet number of the hard processes I generated in Madgraph, or does it mean, given the xqcut, this event has n-jets after showering, where n could be more than 1?

This refer to the number of jet participating to the matching/merging. So if you have events coming from Higgs/W/…, those jet will not be include in the merging algorithm and therefore those sample of events will appear with a lower multiplicity in the plot.

Cheers,

Olivier

> On 7 Feb 2017, at 23:52, Zach <email address hidden> wrote:
>
> New question #452493 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/452493
>
> So there are three parts to my question, but overall I just can't quite get matching to work the way I think it should. I am trying my own dark matter, dm/dm~ model right now, but my issue seems to hold for any process I try(if it is even an issue!)
>
> 1) I ( generate p p > dm dm~, add process p p > dm dm~ j ) and then set ickkw=1 for matching. I have been varying xqcut and qcut using the rule qcut=xqcut+5, or similar variations of it, with xqcut typically around 25.
>
> My problem lies in the differential jet rate 1->0 graph. My understanding is that there should be a smooth transition from the 0-1 jet runs, with the 1-jet starting at xqcut, and the 0-jet ending at qcut. It works fine in the sense that the 1-jet case starts at xqcut, but the 0-jet only very slowly dies off, and is of the same order as the 1-jet multiplicity, creating ~ a factor of two spike in my DJR1->0 graph around xqcut, qcut.
>
> I have many pictures if you would like some. It looks to me as if qcut is not matching the 0-jet/pythia showered jet to the harder 1-jet/hard process case, and instead double counts the cases.
>
> 2) It appears that ATLAS and CMS use ptj1>0 when they are trying this matching process, but setting it non-zero only seems to increase my current problems with the DJR1 graph, lowering the 0-jet sample everywhere, and making it extremely not smooth. Does ptj1>0 remove all 0-jet events, since there are no leading jets, and thus none make the cut?
>
> 3) More of a notation question. The legend lists 0-jet sample, 1-jet sample, etc. Do these events correspond directly to the jet number of the hard processes I generated in Madgraph, or does it mean, given the xqcut, this event has n-jets after showering, where n could be more than 1?
>
> Thanks a lot!
> Zachary
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Great! That use_syst parameter escaped me, but correctly fixes the automatic DJR plots coming out of pythia. Just a quick couple follow up questions.

1. Could you elaborate on your comment to number 2? Or put me on the right path to implementing ptj1>0? Generally I would like to be able to simulate up to very high MET regions, past 700 GeV. Without the ability to set ptj1 or etmiss>0 for the 0-jet case, it would seem I would need to simulate an enormous number of events to get decent statistics in high met regions.
I wouldn't expect the 0-jet case to change much in such a high limit anyway, so I could run 1-jet, maybe matching with 2-jet, but generally it would be nice to generate all of my phase space with one consistent run.

2. Is there any limit to a reasonable xqcut? I saw you, or someone else, mention that qxcut should be ~1/3 or 1/6 of the hard process in your simulation. If I am looking in the MET>700 GeV region, does this mean xqcut should be ~>120 ?

Regards,
Zachary

Thanks Olivier Mattelaer, that solved my question.