Meeting in Karlsruhe 30/08-31/08 2017.
The speakers can be found on the website of the meeting:
https://indico.cern.ch/event/652320/
Attendance in the room is on the spread sheet.
Attendance on Vidyo: Joany, Michal Szleper, Giulia Gonella.
EFT:
----
One can try to set limits on EFT operator without unitarisation but it is unlikely that the LHC 13 TeV (even at high luminosity) will ever be sensitive (see Matthias's presentation).
Alternative to unitarisation could be to cut (clipping) on s hat below the unitarisation bound.
Depending on the cut, you would probe weaker or stronger couplings.
It would also be interesting to see if there are observables that are (strongly) correlated with the s hat.
Then one could use these observables to set constraints.
But this cut on s hat is only possible for some clean VBS channels like for example ZZ (very clean) or WZ semi-leptonic.
Also cutting on s hat might decrease the integration performances in the Monte Carlo programs.
Concerning the unitarisation procedure, there are several implementations available in various code.
It would then be useful to review them and compare them.
Also, it would be interesting to make comparisons of sensitivity between the unitarisation procedures and the cut based procedures (like on s hat).
In principle, it would be nice to be able to fit all relevant parameters for VBS at the same time.
Unfortunately, this is not possible due to computing resources.
It could be useful to fit them one after the other.
It is not optimal as all possible combinations are not covered but it could still give some information.
Also, there are mathematicians (statistics) in Milano involved in the COST action.
It might useful to interact with them concerning these kind of issues.
Another possibility is to fit one, two or three operators (as many as possible).
This is not ideal but it is better than nothing.
Also, it is possible to select a number of operators for each measurement.
In particular one should determine what operators are the most relevant for each measurements.
For the case of VBS, one could then select only few operators which are thought to be testable in these measurements.
This would imply to set the other parameters to zero assuming that they are constrained by other measurements.
Some basis might be more appropriate than others for specific measurements.
Having then constrained some operators in a given basis, one should then translate these constraints into a common basis (using Rosetta for example) to be able to make global fits.
Nonetheless, one should be careful regarding gauge invariance as it could be sometimes fairly counter intuitive (see Ilaria's example).
For this, it is important to infer what operators are relevant for each measurement.
In particular, they could modify the SM processes or add new physics contributions.
The implementation of these EFT operators is also important.
In particular, shall one take only the EFT insertion squared with SM contribution or the EFT contribution squared?
This could impact the final result and gauge invariance has to be conserved.
It would thus be worth to review the different implementation of EFT and compare them in the context of VBS.
In that respect studying the 4-fermion operator seems to be important.
What are their contributions in VBS? Can they be constrained by di-jet measurements for example?
Can their effect be removed by experimental cuts?
Also, some operators might be more or less relevant depending on the experimental cuts and this is important to infer.
Finally, one should keep in mind that some operators have been constrained at LEP.
But this is a very different energy scale than the one at the LHC.
Therefore, all the constraints should be taken in the TeV regime.
In that respect, Drell-Yan or other simple process could be useful to apply these ideas before moving to a more complicated process such as VBS.
It is useful to have exp./th. wish lists.
What is always useful for theorists is:
95% CL upper bound on the fiducial cross section.
This is not so much work from the experimental side.
The choice of the fiducial volume can always be discussed.
Ideally, one would also have the same limits for distributions.
The issue is that for now, the experimental uncertainty is dominated by statistics and in the tails of the distributions, there are only few events.
Experiments can always provide the data points but it is not sure that is very useful due to the low statistics.
One thing that would be useful for theorists is double differential distributions, for example:
transverse momentum of the bosons and the invariant mass of them.
But again the problem is how to reconstruct the invariant mass for WW for example.
In principle, any distributions are fine but maybe some are more sensitive to new physics (? -> job of the theorists).
From the experimental side, it would be useful to provide a matching between UV complete models such as supersymmetry and EFT.
In that way, experimental collaborations could also easily constraint full models.
In particular, could it be useful to have benchmarks of EFT coefficients to be checked against data (in the same way as for SUSY searches sometimes)? (not discussed during the meeting)
At NLO, things are even more complicated because operators mixes and thus one should fit even more operators.
This seems to be too complicated for now to tackle for now.
But at least for the NLO QCD correction to VBS, one might be able to reweight the LO EFT predictions with NLO QCD in the SM.
In particular the QCD corrections should factorise.
But for VBS this seems to be the case and this should also be studied in details.
In particular, refs. like Maltoni et al. on tth might be worth to study (see reference in Raquel presentation).
Note that for EW corrections, this would be much more difficult and it should be tackle later.
Another issue is that: is the experimental accuracy good enough to be sensitive to NLO corrections?
This is a priori not the case yet.
For dim-8 operators, the same discussion as above applies.
But it is even more complicated now as one cannot rely on others measurements to constraint some operators.
Nonetheless, it is worth to study them as only VBS (and tri-boson production) have to these.
In VBFNLO and WHIZARD (some of them) have been implemented.
In order to constraint the operators, it might be useful to constraint combinations of them across different measurements.
Does VBS has some sensitivity to CP violation?
This could be studied with angular distributions for example in ZZ.
But it seems that VBS is probably not the most sensitive process for this kind of studies in particular regarding the low statistics which is available.
In this regards, it seems OK to remove CP violating operators for VBS studies.
Also, in a first approximation removing operators following flavour symmetries seems OK
Concerning the use of MVA (multi-variate analysis) and BDT (boosted decision tree) techniques, it should not worry the theorists.
This is not used to measure the observables.
As long as a fiducial volume is defined, theres is no problem.
It might be useful to study HEFT basis as it provides a different perspective.
In particular, it might be more appropriate to some measurements.
Nonetheless, it seems to introduce more parameters than the Warsaw basis for example.
Therefore, it is maybe be not useful for experimental analysis in a first approach but should be kept in mind for theorists.
Plans for future:
- Review of various EFT and unitarisation procedure for VBS.
This could lead to comparison and agreement on some procedures.
- Review of EFT analysis in experiments for VBS.
- Procedure to constraint EFT parameters.
Eventually, we should aim at some recommendations for both the theoretical and experimental analysis (this would be extremely valuable).
MC SM:
------
Concerning SM predictions, using Rivet analysis could be useful.
But this should rather be used after parton shower.
For the comparison, this should be rather done with one of the predictions that have all possible corrections (Recola or WHIZARD).
Also, we should have plots LO vs. NLO (for a same code) and not only LO and NLO comparisons between different codes.
One could also study the variations of the predictions under several PDF sets (at both LO and NLO as at NLO there is the gq channels opening up).
We could have extra distributions such as pT and eta of the third jet.
One should also provide a set of uncertainties:
- the differences between the various approximations.
- the scale variation uncertainty but this will not give an estimate of the NNLO computation as it has been shown for VBF.
- The estimate NNLO EW ~ (delta NLO ew)^2
Maybe the effect of merging could be interesting? (not discussed during the meeting)
Also, it would be useful to have recommendations for the matching of Parton Shower for both the fully leptonic and semi-leptonic processes.
Finally, we reached an agreement on the implementation of the delta R j l in W+W+ measurements:
At NLO, one can have up to three jets.
For the event to be accepted at least two jets should fulfil the pT and rapidity requirements as well as the delta R j l cut.
This means that in the definition of a jet enters the pT, the rapidity and the delta R j l cut.
To accept an event, two of such jet should be present in the final state.