Minimum ptj for p p > j j?

Asked by Jae Hyeok Chang on 2018-03-15

Hi,

I just want to see all p p > j j events without any cuts. So I put ptj=0, and etaj=-1. However, it seems that something is going wrong. I plotted cross sections for this event in terms of ptj (with etaj=-1), and it is flat between ptj=0 and ptj=2. Also, I compared p p > j j and p p > j j j cross sections, and p p > j j j has larger cross section for ptj=0, and similar cross section for ptj=2 while I think it should be smaller by at least an order of magnitude. What is the minimum ptj that gives reliable results?

Thank you so much in advance!

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Solved by:
Jae Hyeok Chang
Solved:
2018-03-16
Last query:
2018-03-16
Last reply:
2018-03-16

Hi,

Since the cross-section is infinite if you do not apply any pt cut.
I'm not surprise that you face plenty of issue here.

For the 3 jet case, having just a pt cut is not enough to cut all the divergency and having just a pt cut will
also lead to an infinite cross-section. (the third jet can still be soft collinear to one of the other jet).

> What is the minimum ptj that gives reliable results?

If you only apply such cut, then no value should return reliable results.
If put in addition to other cuts (like DR cuts) then 10 GeV should work for sure.
Now if you consider the typical matching/merging method, you will realise that the merging scale is typically taken higher than that.

Cheers,

Olivier

> On 15 Mar 2018, at 19:57, Jae Hyeok Chang <email address hidden> wrote:
>
> New question #665683 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/665683
>
> Hi,
>
> I just want to see all p p > j j events without any cuts. So I put ptj=0, and etaj=-1. However, it seems that something is going wrong. I plotted cross sections for this event in terms of ptj (with etaj=-1), and it is flat between ptj=0 and ptj=2. Also, I compared p p > j j and p p > j j j cross sections, and p p > j j j has larger cross section for ptj=0, and similar cross section for ptj=2 while I think it should be smaller by at least an order of magnitude. What is the minimum ptj that gives reliable results?
>
> Thank you so much in advance!
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Jae Hyeok Chang (jjlophy) said : #2

Thanks Oliver!

Can I ask a few more questions?

For p p > j j, is it ok to give a eta cut with ptj=0? If so, what's the maximum etaj that gives reliable results? (I think I don't need DR cuts here since it is 2->2 process, right?)

For p p > j j j, can I use the default DR cuts with some etaj and ptj=0?

When I use pythia8 for hadronic showers, do I need to change some settings for such low cuts?

Thank you a lot agian!

Hi,

> For p p > j j, is it ok to give a eta cut with ptj=0?

In principle such eta cut should be equivalent to an effective pt cut. So yes it should be enough.
The only thing that this does not cut is for process at very low energy but those should be cut
by our hardcoded cut that the factorization scale should be larger than 4 GeV^2.

> If so, what's the
> maximum etaj that gives reliable results?

I do not know.

They are multiple limit here:
1) the limit where LO computation does not make sense anymore
2) the limit where our code starts to be numerically unstable.
3) the limit where neglecting some physical effects does not make sense anymore.

For 1) I have in mind perturbation theory. For soft radiation, you should not use the full LO matrix-element
but resommation method and/or parton-shower which are all order method.
(Remember that madgraph generates inclusive events and that additional soft radiation can be generated by Parton-Shower)

2) This is quite clear and one sign of this is that the requested number of events is not the one that you request (and/or the error on the cross-section is quite large)

3) Here I have in mind the mass of the quark, if you want radiation with a pt of 0.5 GeV, then you should not perform a 4 flavor computation and you should restore the mass of (at least some) quark.

Looks like from your email, you are only focussing on the second limit, where you can probably but very low cut and be safe on that one. But even if MG5aMC will work as expected the two other points will make the computation meaningless anyway.

> For p p > j j j, can I use the default DR cuts with some etaj and
> ptj=0?

Not sure that this is enough to kill all singularity. But you can obviously try.

> When I use pythia8 for hadronic showers, do I need to change some
> settings for such low cuts?

Clearly you might have some issue where Pythia fail to shower events in this context since
your hardscale will be so close (or even lower) of the hadronization scale that Pythia will fail to either shower the event
and potentially even to hadronize such event. (which will be a clear indication that you have generate event in a non perturbative part of the phase-space)

Cheers,

Olivier

> On 15 Mar 2018, at 21:13, Jae Hyeok Chang <email address hidden> wrote:
>
> Question #665683 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/665683
>
> Status: Answered => Open
>
> Jae Hyeok Chang is still having a problem:
> Thanks Oliver!
>
> Can I ask a few more questions?
>
> For p p > j j, is it ok to give a eta cut with ptj=0? If so, what's the
> maximum etaj that gives reliable results? (I think I don't need DR cuts
> here since it is 2->2 process, right?)
>
> For p p > j j j, can I use the default DR cuts with some etaj and
> ptj=0?
>
> When I use pythia8 for hadronic showers, do I need to change some
> settings for such low cuts?
>
> Thank you a lot agian!
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Jae Hyeok Chang (jjlophy) said : #4

Thank you Olivier, this is really helpful! I'll try with a eta cut and a small pt cut.

Can I ask one more question? I tried "p p > j j" and "p p > j j + p p > j j j (with the 'add process' command)" in the default setting, but the cross section for the latter is smaller than the former. I read a note that the latter processes can be doubly counting, but then the cross section should be overestimated, not underestimated. Why is the cross section for two processes are smaller? I don't understand this.

Thank you in advance again!

By default we setup the run_card for MLM setup such everything is set to remove such double counting.
This also means that the default run_card has
1) ickkw =1 (i.e. we use special scale)
2) xqcut=20 (i.e we define additional cut)
3) auto_ptj_mjj (i.e. we automatically set the value of ptj and mjj accordeing to xqcut)

If you did not change any of the two last parameter then It is normal that you have a smaller cross-section.

Cheers,

Olivier

> On 16 Mar 2018, at 13:27, Jae Hyeok Chang <email address hidden> wrote:
>
> Question #665683 on MadGraph5_aMC@NLO changed:
> https://answers.launchpad.net/mg5amcnlo/+question/665683
>
> Status: Answered => Open
>
> Jae Hyeok Chang is still having a problem:
> Thank you Olivier, this is really helpful! I'll try with a eta cut and a
> small pt cut.
>
> Can I ask one more question? I tried "p p > j j" and "p p > j j + p p >
> j j j (with the 'add process' command)" in the default setting, but the
> cross section for the latter is smaller than the former. I read a note
> that the latter processes can be doubly counting, but then the cross
> section should be overestimated, not underestimated. Why is the cross
> section for two processes are smaller? I don't understand this.
>
> Thank you in advance again!
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.

Jae Hyeok Chang (jjlophy) said : #6

Hi Olivier,

Thank you so much for you kind explanations. It was very helpful!