MadGraph5_aMC@NLO

Matching & large HT cut

Asked by Martin Stoll

Hi,

I want to produce events with HT > 1000 GeV (ihtmin), e.g. p p > b b b~ b~ + 1 j matched in the MLM scheme.

If I choose a small xqcut around 40, I can get quite smooth djr plots, however such a small scale is probably not what I want when I look for hard jets (also the cross-section for 4b+j is much larger than for 4b).
So I chose the matching scale around 200 GeV which makes nice djr plots again. Now I am wondering about the following things:

- Somehow I feel 200 GeV is too large a matching scale; it will force the event to be very symmetric.
- Should I set maxjetflavor to 5 instead of 4? Why does the cross-section decrease by a factor of ~200?
- The (matched) cross-section varies between 0.05 and 0.08 for xqcut in [190,210], which makes me feel uncomfortable.

Bottom line: What could be a reasonable setup and why?
I ask myself similar questions about t pair production + jets, too (also with large HT cut).

Thanks in advance~ v(^^)

Question information

Language:
English Edit question
Status:
Solved
For:
MadGraph5_aMC@NLO Edit question
Assignee:
de Visscher Edit question
Solved by:
Martin Stoll
Solved:
Last query:
Last reply:
Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#1

Hi Martin,

I'm in fact not sure that a cut on HT makes any sense for a Matched/merged sample.
Indeed this cut will not be applied on the same way for the 0 and 1 jet sample generated by MG which is probably the reason why this introduce dependencies in the matching cut.

I assign Simon to this bug since it knows better that me.
But I think that the following should works:
1) introduce a very loose cut on HT at the MG level (should be lower than the final value+ xqcut* number of matched jet).
2) run the mlm normally
3) re-introduce the (real) cut after pythia (with the use of a jet algorithm)

Cheers,

Olivier

Revision history for this message
Martin Stoll (stoll) said : 		#2

Thanks for your answer, Olivier, I'm trying your method right now.
Looks promising so far!

Revision history for this message
Martin Stoll (stoll) said : 		#3

Hi,

I ran some tests and now the distributions look quite reasonable, at least to me...
I do still have some questions, though:

- Is there some argument about how to choose the xqcut? (e.g. 10% of the final HT cut for 4/5 outgoing partons or something like this)
- What is a good way to determine the value of the cross-section?

And on an unrelated subject:
- Why is the cross-section so much lower when I chose maxjetflavor=5?

Cheers, Martin

Revision history for this message
Olivier Mattelaer (olivier-mattelaer) said : 		#4

Hi Martin,

The real important parameter is qcut and not xqcut. xqcut is just there to try to not throw away too many events.

- Why is the cross-section so much lower when I chose maxjetflavor=5?

Did you change the definition of the jets/proton at the diagram generation level, this change the way to do matching since this control if you do a 4 flavor or a 5 flavor matching (this also change the cuts associated to the b)

Cheers,

Olivier

Revision history for this message
Martin Stoll (stoll) said : 		#5

Thanks Olivier!