Use LSF

[ktht]

Instead of running our own NanoAOD production to create lepton-subtracted AK8 jet collections, we should instead try what's already available in NanoAOD: lepton subjet fractions (LSF). The relevant branches are:

• FatJet_lsf3 -- floating point number between 0 and 1, corresponds to the ratio of lepton pT to 3rd subjet pT (or -1 if there are no leptons matched to the subjet) that results from reclustering the constituents of existing AK8 jet with three subjets instead of two;
• FatJet_electronIdx3SJ -- index to the Electron collection (or -1 if there are no matches or if the match is a muon);
• FatJet_muonIdx3SJ -- index to the Muon collection (or -1 if there are no matches or if the match is an electron).

Possible action items:

• modify RecoJetAK8 and its reader class accordingly;
• introduce index variable for leptons, which tells the position of the lepton collection in the input NanoAOD Ntuple, so that the *Idx3SJ branches can be used to identify the LSF matches;
• add isBoosted flags to leptons and/or AK8 jets if the selected fakeable lepton is LSF-matched to a selected AK8 jet (and if LSF is greater than, say, 0.7) when writing them to the output Ntuple?

Resources:

• Paper on LSF: https://arxiv.org/pdf/1410.0362.pdf
• Implementation in CMS analyses: https://indico.cern.ch/event/871916/contributions/3677160/attachments/1969064/3274986/LeptonIsolation_B2G_14Jan.pdf

[ktht]

Hmn looks like the PR link does not show up here even though it's linked in the comments. I'll back-reference the PR then and close the issue: #25
Another task is to actually test that it works, but it can be done at the analysis level, not here.

[saswatinandan]

@ktht , since you just mentioned about file size in other chat, I was thinking that we can drop one branch here, we can get this information form lsf3 value which is -1 by default i.e not boosted. We can do the changes later as it is not important, though it will not reduce the file size a lot.