ggF meeting on Higgs pT.

Europe/Zurich
40/R-D10 (CERN)

40/R-D10

CERN

20
Show room on map
Minutes of the meeting.
Participants beyond the speakers:
Chiara Mariotti, Sara Bolognesi, Paolo Nason, Giulia Zanderighi, Elisabetta Pianori, Roberto Di Nardo, ...

1) In YR2 a difference between MC@NLO and HRES was observed above all at high pT, that was tracked back to the use of a fixed scale set at mH in HRES and a dynamic scale,
set at mT = sqrt(pT^2+mH^2) in MC@NLO;

2) also a difference  between Powheg and HRES was observed, Powheg running with the scale set at mH. The difference between HRES and Powheg was fixed by adding a damping factor to the 
contribution of the sudakov form factor to the collinear divergent amplitude. This damping factor puts to 0 its contribution in the pT ->inf limit, and enhance it in the pT -> 0 limit.
The damping factor (hfact) is a tuning parameter of Powheg, it avoids extra emission on top of the ME one in the high pT limit. Powheg was tuned with hfact = Mh/1.2 in order to match the HRES 
contribution (see talk from Vicini summarising the approach);

3) Heavy Quark mass effect contribution and electro-weak correction were included in Powheg and used in both ATLAS and CMS as default generator. ATLAS showering with Pythia8 and CMS with Pythia6.
A difference between MC@NLO and Powheg on the effect of the HQ mass effect in the Higgs pT was observed, in particular for the bottom quark contribution. The meeting is mainly focused to
follow up on this issue.

Massimiliano Grazzini reviewed the status of HNNLO and HRES2.0, now including both the top and the bottom contribution to the gg triangle. While the top contribtion doesn't pose problems to the resummation
scheme, the bottom contribution introduce a new scale mb. For pTH >> mb the collinear factorisation approximation that is at the basis of the resummation techniques doesn't hold. In particular when using
a resummation scale for the bottom contribution of order mH/2 the pT spectrum doesn't match the NLO calculation at moderate pT values (~40 GeV). It does match MC@NLO spectrum,
that Frixione confirms to have an implementation of the ME and matching scheme, called "additive",  very similar to HRES. By the way, when a scale of order mb is chosen for the resummation of the b quark contribution, the spectrum is qualitatively closer to Powheg
default with hfact. This matching seems accidental (as explained by Vicini in his talk). In resummation calculation is possible to resum only one set of logarithms, and the Higgs pT with HQ mass effect is a multi-scale problem:
mH, mT, mb and pT Higgs. mH and mT looks to be close enough to not need the resummation of their logs, therefore the top implementation is anumbiguous, while when including the bottom quark,
logarithm of the type: pTHiggs/mb, pTHiggs/mH appears in the calculation. In order to peform it, Massimiliano introduce a second scale, called Q2 and for pTHiggs > Q2 the resummation is switched off.
The preferred values of Q2 is set at mb, because for Q2=4mb a relevant discrepancy at high pT between resummed calculation and NLO one is visible.
To asses the systematics on the spectrum, he illustrates  the effect of the usual recipe: mb/2, 2mb. Giulia Zanderighi commented that using this variation as systematics could be aggressive, giving that we are anyway missing all
the resummed calculation for pT > mb, whose contribution we cannot compute at the moment, as stressed many times from Massimiliano. One of the core point of the discussion is how to assess the systematic
due to the HQ mass effect (CMS using the error on the b and t quark mass, ATLAS planning to use the difference between MC@NLO and Powheg).

A. Vicini.
Alessandro summarised the implementation of the HQ mass effect in Powheg, given that hfact practically splits the collinear and the non collinear contribution in Powheg, it is possible to use hfact to reproduce the resummed calculation used in HRES according the prescription from Massimiliano. The idea is to try to run Powheg with different hfact values for the b and top contribution. We can write:

|M(t+b)|^2 = |M(t)|^2+[|M(t+b)|^2-|M(t)|^2]
therefore
sigma(t+b) = sigma(t,hfact = mH/1.2) + [sigma(t+b, hfact=mb)-sigma(t,hfact=mb)]
So the idea is to generate 3 samples. One with only top quark effect, that run at hfact=mH/1.2 (the default value to match HRES), the second with t and b quark contribution, running at mb and the third with only t,
to subtract the top contribution from the second sample, running at hfact=mb. With this approach an agreement at 2% level except the first bin pT < 5GeV is observed between HRES (Grazzini prescription) and  Powheg.
The best matching is obtained for mb=4.75 GeV. Plan to include the full treatmenet in Powheg to avoid the generation of 3 samples.

Frixione.
A new verision of MC@NLO 4.10 was released. This version contains mt and mb finite mass effect in loop using the possibility to use two different scales in top and bottom loops.
I particular there are 2 run options. One producing only the top contribution at scale = mH,, the other producing the top-bottom interference and the bottom at scale equal mb.
One needs to run the two samples and sum them at analysis level. Also in this case the agreement is at per cent level except in the first bin pT < 5 GeV. The need for the 2 sample generation can be avoided upon requests from experiments.


Conclusion on Higgs pT.

Proposal from Massimiliano, use the generator you prefer tuned to reproduce the HRES central value with HQ mass effect, use dynamic scale at high pT where it makes more sense,
take the uncertainty band from HRES and find the range of tuning parameters able to cover that uncertainty, use these samples as systematic samples.


ATLAS/CMS talks.

There were report from both ATLAS and CMS.
There are few points on which we should follow up.

1) ATLAS, using Pythia8, doesn't find appreciable effects of Hadronisation and Underlying Event while CMS finds order 10% effect;
2) The jets in ATLAS look more abundant than in CMS

need to follow up on this issue. test CMS configurations and vice-versa. Question from Frixione if ATLAS was using the "interleaving" Pythia8 optiion,
Biagio followed up on this issue and it looks that we are using it, but ATLAS is now testing new matching schemes without this option.

A bit off topic we discussed how to treat the BR(H->ZZ) when the m4l is far off-shell. Need to follow up with Nikolas Kauer, the immediate answer is to correct the spectrum using the
mass depndent Higgs Br.

ATLAS and CMS showed results of scale variation using HRES and PDF. ATLAS bands look slightly larger than CMS.

Sherpa.

Frank Kraus showed first stuies with Sherpa. The attempt is to have an multi-jet NLO matched sample, using the same technique of aMC@NLO.
Frixione made notice that the inclusion of HQ mass effect in aMC@NLO is not trivial giving that it is a 2 loop process, so it cannot be automatically computed. 
The same problems is valid for Sherpa. Anyway, NLO mathcing in 2 jet could be of interest for the 2 jet topologies.







There are minutes attached to this event. Show them.