Matching and efficiency

Hi,

> 1) For the case of pp> tt~ + 0,1,2 jets I get smooth DJR-plots for many xqcut / qcut values (20-40GeV) (See here : https://drive.google.com/drive/folders/1A0ZU-XSNya-3z2qBT8D7cfLzR1-YmDcb?usp=sharing)
> Only at xqcut=50GeV / qcut=75GeV a kink appears. Does that mean I'm "allowed" to choose any value in between? If so i would chose the highest passable xqcut because that results in only 50% of ME-events being tossed by pythia instead of ~80%.

Yes you can choose in that range.

Once you have found a good value for Qcut, you should try to increase xqcut which is also going to reduce the number of event tossed away (again you can do that up to the point that you have a kink, and probably take a security marging obviously).

>> 2) For the process pp> a a + 0,1,2 jets with lhaid=315000 I have a strong dependence of the cross-section on the matching. My guess is that this comes from the soft-divergence of this process, so i set pta = 18GeV. I found on here that you once suggested xqcut=10GeV to someone but for this the generation fails entirely for me. Is there something I'm missing?
>> Also at xqcut=12.5 (the lowest that is reliably possible) and qcut=22.5 I get a weird plateau in the DJR-plot (see the link above). Is that something to worry about? Even at xqcut=15 the plot gets worse.

I'm actually not an expert for that process. Here they are two point to check
First, for this process, the isolation cut between the photon and the quark are quite important as well.
Which might be something that you have to check what the impact is on such distribution.

Independently, I'm worry that your Qcut is larger than your pta, this implies that you do not have a scale separation between your parton-shower and your matrix-element. This is likely the main issue why your DJR plot are problematic.

> 3) If the analysis after simulating the detector response calls for eta<2.4 is it okay to set etaj to, say 2.6 to allow for some smearing or would such a small margin of error furnish artifacts?

I would say that it should be fine, but I'm not a Parton-Shower expert. They should know better than me which variable are deeply changed by the parton-shower. In any case, the best is to do two small sample to check that your are not sensitive for such type of cuts.

> 4) An unrelated question: Are the cards used for arXiv:1405.0301 still somewhere accessible? The link in the publication seems to be broken.

I do not have them and I do not see them on the amcatnlo website. Did we put a link to such card in the paper?
If yes those path are likely moved now.

Cheers,

Olivier

> On 2 Mar 2023, at 08:30, Bickendorf <email address hidden> wrote:
>
> New question #705679 on MadGraph5_aMC@NLO:
> https://answers.launchpad.net/mg5amcnlo/+question/705679
>
> Hi,
> i have a few questions that came up while trying to understand matching better.
>
> 1) For the case of pp> tt~ + 0,1,2 jets I get smooth DJR-plots for many xqcut / qcut values (20-40GeV) (See here : https://drive.google.com/drive/folders/1A0ZU-XSNya-3z2qBT8D7cfLzR1-YmDcb?usp=sharing)
> Only at xqcut=50GeV / qcut=75GeV a kink appears. Does that mean I'm "allowed" to choose any value in between? If so i would chose the highest passable xqcut because that results in only 50% of ME-events being tossed by pythia instead of ~80%.
>
> 2) For the process pp> a a + 0,1,2 jets with lhaid=315000 I have a strong dependence of the cross-section on the matching. My guess is that this comes from the soft-divergence of this process, so i set pta = 18GeV. I found on here that you once suggested xqcut=10GeV to someone but for this the generation fails entirely for me. Is there something I'm missing?
> Also at xqcut=12.5 (the lowest that is reliably possible) and qcut=22.5 I get a weird plateau in the DJR-plot (see the link above). Is that something to worry about? Even at xqcut=15 the plot gets worse.
>
> 3) If the analysis after simulating the detector response calls for eta<2.4 is it okay to set etaj to, say 2.6 to allow for some smearing or would such a small margin of error furnish artifacts?
>
> 4) An unrelated question: Are the cards used for arXiv:1405.0301 still somewhere accessible? The link in the publication seems to be broken.
>
> Thank you!
>
>
> --
> You received this question notification because you are an answer
> contact for MadGraph5_aMC@NLO.