MadGraph5_aMC@NLO

Large uncertainties on some cross sections when doing a parameter scan

Asked by Kevin Urquía

Hi all.

I am using MadGraph 3.3.2. I was doing a parameter scan on the model mlrsm-nu-loop (you can find the project here: http://feynrules.irmp.ucl.ac.be/wiki/LRSM_NLO), specifically MN and MWR, the mass of a fermion and a gauge boson.

I was running the process
p p > n1 e /w+
there should be an intermediate WR in the s-channel for this process. I did everything at LO.

For some specific runs, the uncertainty of the cross section becomes of the order of the cross section itself, for example, for:

MN = 20 GeV and MWR = 12 TeV, cross_section = 7.161e-6, error = 1.3e-6
MN = 30 GeV and MWR = 12 TeV, cross_section = 6.732e-6, error = 1.3e-6
MN = 50 GeV and MWR = 12 TeV, cross_section = 6.735e-6, error = 1.2e-6

This is the second parameter scan I do. The first time happened for different values of MWR and MN. Is this a common thing? Is there a way to solve it?

I also performed the same scan at NLO, and I did not get anything crazy like this.

The kinematics for all runs look fairly reasonable. I was comparing them with the NLO, and there is very little difference between them.

Question information

Language:
English Edit question
Status:
Answered
For:
MadGraph5_aMC@NLO Edit question
Assignee:
No assignee Edit question
Last query:
Last reply:
Revision history for this message
waqas khan (waqaskh12) said : 		#1

It sounds like you're seeing significant uncertainties in your cross-section calculations at LO for certain parameter values of `MN` and `MWR` in MadGraph. This can sometimes occur with high-mass scenarios or where intermediate particles, like `WR` in your s-channel, are very heavy. At LO, the statistical sampling can produce higher variances, especially in complex models.

Running the same process at NLO with stable results suggests that the uncertainties at LO may be due to limitations in sampling precision or loop effects that NLO handles more accurately. To reduce uncertainty at LO, you could try increasing the event sample size or adjusting integration parameters. Additionally, if computationally feasible, NLO might be more suitable for scans with high-mass configurations like yours.

Revision history for this message
Kevin Urquía (kurquia97) said : 		#2

@waqas khan, thanks for your reply! I ran some things during the weekend at LO for the places where I was getting problems. Now I am getting consistently bad uncertainties:

For 75000 events:
MWR = 14 TeV
MN = 10 GeV, cross_section = 7.147e-6, error = 1.4e-6
MN = 20 GeV, cross_section = 7.139e-6, error = 1.6e-6
MN = 30 GeV, cross_section = 6.875e-6, error = 1.3e-6
MN = 40 GeV, cross_section = 6.719e-6, error = 1.4e-6
MN = 50 GeV, cross_section = 6.623e-6, error = 1.5e-6

Out of curiosity, I ran the same events with 1000 events
MWR = 14 TeV
MN = 10 GeV, cross_section = 6.673e-6, error = 9.1e-7 (events = 62)
MN = 20 GeV, cross_section = 7.158e-6, error = 8.6e-8
MN = 30 GeV, cross_section = 6.792e-6, error = 1.2e-6 (events = 439)
MN = 40 GeV, cross_section = 6.792e-6, error = 7e-8
MN = 50 GeV, cross_section = 6.677e-6, error = 8.9e-8

I have no cuts on either of the processes. The only difference between both runs was the nevents. I seriously do not understand what is going on

Revision history for this message
Sihyun Jeon (shjeon) said : 		#3

what is your beam energy?
if it's ~14tev i think it's natural to get very unstable cross section due to pdf uncertainty

Can you help with this problem?

Provide an answer of your own, or ask Kevin Urquía for more information if necessary.

To post a message you must log in.