Monte-Carlo description of VBS - VBSCAN COST action
---------------------------------------------------
16-17 November - Nikhef Amsterdam - the Netherlands
https://indico.cern.ch/event/673764/
-> Thursday (16 November)
# Bruni
- Measuring all contribution together (EW, QCD, int): this is possible experimentally.
This is much better to measure all contributions in a given fiducial volume.
- Try to use NLO results in experimental collaborations as much as possible, when code are available or ask theorists.
For NLO QCD they can be computed with (good) approximations in VBSNLO and POWHEG.
For now, for NLO EW this can be only be done on demand by Pellen et al.
- For semi-leptonic channels, is it possible to isolate W+W+, W+W-, WZ?
For example looking at some observables (angles?) might give some hint.
But more work is needed in this direction (possible project in the action?).
In particular one could use MadGraph_aMC@NLO and Phantom at LO for some preliminary study.
- What could be the next VBS process measured at the LHC?
Maybe p p -> Z \gamma 2 j or p p -> Z Z 2j.
- For ZW, one could use b-tagging to remove the tZj contributions that are doimnated by resonant top.
# Schwan
- Limit of Higgs mass -> infinity: maximum difference with respect to SM.
Is it true?
If you have two Higgs doublet model, can you have larger differences?
It seems to be correct by one should check this in details.
- One probes high-x of the pdfs, so the uncertinty gets larger (up to 7% at large mjj, while incusively it is at the order of 2%).
Is there a preferred choice of PDF?
At least of the order of the calculation.
But no problem (in principle) with using NLO PDF with LO matrix element.
- Can we learn about PDF from VBS?
Probably one learn the same things than from VBF.
In di-jet very sensitive but maybe in VBS/F, you get only some flavour combinations.
This would need a dedicated study.
- No Higgs (Higgs mass -> infinity): the cross-section grows by 9% @ LO, with the default selections of the study.
Large differences in the spectra are probably due to the different polarisations in the interaction.
For the noH case (always TT) wrt the Higgs case (there's also LL).
- Infinite Higgs, how does it look taking into account the 125 GeV constraints?
Is it possible to translate it in terms of couplings?
Plots from Pietro (SM-14-008 in CMS) and Michael.
You can have sensitvity.
Possibility to use the kappa framework?
# Karlberg
- Technically easy to implement BSM in the process, since the EWK component of the diagrams is easily decoupled from the QCD structure, which does not change.
This is only true for Electroweak-sector BSM models.
- Scale variation?
1-2% for NLO observables and ~20% for LO obs. as the third jet ones.
More to be done on third jet -> VVjjj @ NLO seems to favour Pythia results.
- POWHEG, VBS@NLOPS: can be long.
But with the special settings for the grid of Alexander, this can speed up.
- Third jet not yet known at NLO in VBS.
One could check VBF Hjjj, where it seems to favour the pythia shower wrt other showers:
the NLO tends to let the third jet to populate the central region wrt the NLO VBS calculation.
# Rauch
- The PS after VBS cuts reduces the XS, it seems due to the VBS cuts.
- Larger scale variation for LO observables (expected) such as j3 observables.
To know which one is the best Powheg+Phythia vs. VBFNLO+Herwig, can we look at LO?
Maybe look at H+3j? Because the QCD structure is the same.
In the same spirit, (VBF H NLO + the 2 showers) vs. (VBF Hjjj (VBF + 1j NLO).
- Suggestion: put LO+PS on top for the comparison.
# Zaro
- Are differences due to Powheg or/and Pythia?
Could it be the Powheg Sudakov?
# Pellen
- Is there a good way to estimate the EW?
[Denner, Pozzorini; hep-ph/0010201] for LL.
- In the 2D ratio, adding s-channel contribution to the t-u approximation is expected to remove the negative peak in bottom left in figure 2D ratio.
This will be done for the report.
- p.16, the NLO band is asymmetric, why? Interplay between the scales.
# Pigard
- Seems challenging to compute the gg background very fast.
The loop induced process are part of the NNLO corrections.
###################################################
-> Friday (17 November)
Review of the objectives of WG1.
Original document:
https://docs.google.com/spreadsheets/d/17ysuUKidqKLU3Y88qPhIiSjnXyPMZ49R_MhHG2_iRrs/edit#gid=0
- Common possible study WG1/WG2:
-> study of the different contributions (WW, WZ) for semi-leptonic process at LO
-> polarisation for ZZ adn WZ using Phantom / MadGraph_aMC@NLO at LO
### Definition of the signal and its simulation
- Theory would like to have all the components of the VBS (QCD, EWK, and all interferences) simulated and considered as signal in analyses, to have a proper measurement.
Do not subtract the QCD, always measure the sum in the VBS region, even if purity is high (unless we get in situation where QCD becomes negligible).
At LO generate the 3 contributions alpha^6, alphas alpha^5, alphas^2 alpha^4.
At NLO do the same and generate the 4 contributions.
This is much better to measure all contributions in a given fiducial volume instead of subtracting some contributions based on MC.
-> Signal definition: use of experimental cuts is OK
- First do the SM measurement, then think of BSM.
-> open issues:
- how to apply corrections?
One could argue that QCD corrections play mostly in the low energy part, while the EWK corrections work at high energy tails, but this is handwaving only and not necessairily true.
For now the additive prescription should be sufficient.
- what about EM showers? do they bring in double counting?
Yes, there is double counting when one do a QED shower when applying EW corrections.
Nonetheless this should be a small effect as the QED shower should have a very small impact.
Thus given the precision now, this should be OK.
This will be checked in the MC report.
- move to NLO + PS generators for SM.
For BSM it's not bad to make it LO for the time being but on the long run when codes are available this should also move to NLO.
To have the two simulations (SM and BSM) at different order is not a big issue, according to the audience, and in general the limits would probably be larger than QCD corrections.
One should be pragmatic though.
For BSM EWK corrections things are more complicated, it could be done with RECOLA2, which is has just been released this week.
Is there a corresponding effort going on in Madgraph?
- If results are presented unfolded, people can to any detailed future analyses by themselves.
- What about approximations, like the VBS one? Is it applicable all the time? -> the fact that this could be different should go in the report.
- Strategy to isolate S from B usable for theoreticians
what is the impact of MVA selections?
For the time being, the calculations should be given to experimentalists for the analysis.
- Higgs XSWG is doing simplified cross-sections, can we apply that in our case as well?
-- Discussion on BDT (boosted decision tree)
1. build a BDT
2. give a number in a fiducial volume which is similar to the BDT
- how can one translate a MVA-fitted cross section in a result obtained in a fiducial cross-section?
- when there's enough events, theoreticians want to have a definition of the signal region which is reproducible, in which to count events.
This region should have simple cuts.
These are unfolded numbers basically (corrected for acceptance of course).
- Polarisation of vector bosons
- sensitivity of specific variables needs to be proved, look at the paper from Phantom authors.
Going to NLO QCD in this framework is straight-fwd for the fully-leptonic case, as QCD corrections factorise.
For EW corrections, this is tricky since there's non-factorisable corrections.
--> LO comparisons between phantom and madgraph (in particular MadSpin) could be useful (possible project of WG1/WG2).
In the fully leptonic cases for WZ and ZZ, it would be feasible to study polarisation as it was done for the WW case.
There is an interest from both theorists and experimentalists.
Two scenarios for polarisation studies, extension with HL-LHC and current detector acceptance should be studied.
Approximations needed for polarisation predictions (on-shell projection), see recent paper from Torino group, but the error of this approximaiton should also be studied.
Suggestion to add calculation for ZZ channel that has the best potential for experimental separation of polarisations, should be straightforward.
Long discussion on producing calculation for WV in semi leptonic, very complicated, would require months of work.