Les Houches

Photon topics:

1. Measurements: gg, gg + jet, gg + 2jets

–All processes:Try for some calculation including direct and fragmentation contributions in as high orders as possible (something equivalent to joining of 2gammaNNLO and DIPHOX/Gamma2MC) In what areas in phase space is 'missing' (meaning wrt 2gammaNNLO) fragmentation important? What is the relationship with asymmetric pt cuts? Pertinence to ATLAS recent results and CMS results in progress, theory-data agreement in low DeltaPhi region for example

–gamma gamma + jet: Try GoSam gamma gamma + 1 jet + MG compared to SHERPA/aMC@NLO/POWHEG/ALPGEN/MG/PYTHIA (codes producing hadronised events) Is there a way to 'map' kinematic effects from fragmentation to experimental level (as we do today running for example PYTHIA with and without UE to correct results from the partonic codes for UE today), so as to be able to correct/extrapolate from purely partonic codes?

==⇒Discussions/conclusions from June 5 meeting:

–Opinion from theory is that a calculation at NNLO with both direct and fragmentation contributions is ~10 years away. Not feasible for this LH…

–Differential diphoton cross-section as a function of delta_pt: 'Turnover' at low delta_pt expected to be more pronounced for NNLO than for NLO

–GoSam does not include the box diagram, but will in the future!

–After checking, it seems that amc@nlo (http://amcatnlo.web.cern.ch/amcatnlo/) does not yet have validated direct photon processes except for diphoton production in the ADD model, not directly relevant for us since relatively high m_gg (>100 GeV). Other yet-to-be-validated processes are: p p > a a a [QCD], p p > a j j [QCD], p p > a j [QCD]. Should ask Rikert F. anyway what the timetable is. For all these processes, it is recommended to use Frixione isolation for the photon at generation level and analysis level with parameters: They use d0 = 0.38, epsilon = 1 and n = 2

—POWHEG does not include direct photon processes yet either. There had been some work in HERWIG++, see 1106.3939.

— pp → AAA at NLO is done by VBFNLO: http://www.itp.kit.edu/~vbfnloweb/wiki/doku.php? id=documentation:proclist

==⇒Slides shown by (June 5): Daniel (talk at Paris photon workshop, delta_pt study)

==⇒Discussions/conclusions from June 10 meeting:

–Where does the missing data/MC agreement in the LHC diphoton measurements come from? Fragmentation, missing higher orders, parton shower? Daniel thinks missing orders… Frank K thinks that Sherpa seems to work, so the statement is not completely correct. In Sherpa it is the interplay of merging with fragewmtnation.

–Discussion of eventual possibility of GoSam interfacing to either SHERPA or POWHEG including parton shower

–BTW: NLO corrections and box in SHERPA + OpenLoops available.

==⇒Studies to be undertaken (People)

–Redo study of differential cross section vs. delta_pt with 2gammaNNLO, see how 'turnover' at low delta_pt compares with NLO case and what is minimum 'safe' delta_pt (Daniel, Leandro et al)

–Study of k-factor NNLO/NLO as function of delta_pt for (pt_gamma1, pt_gamma2)={(40, 25), (28, 25), (25, 25) GeV} (Nicolas et al)

–Comparisons between GoSam gamma gamma + 1 jet/MG (parton-level) compared to SHERPA/MG/PYTHIA (Gudrun, Nicolas, Nicolas Greiner)

Observables : Mgg, pT1, pT2, pTj, dR(j,g1), dR(j,g2)

Comparisons: Diphoton+up to 2jets Madgraph with Pythia parton shower Pythia Box Diphoton+up to 3jets Sherpa with shower

Acceptance cuts proposed for the above studies: pT_gamma1>40, pT_gamma2>25 |eta_gamma|<1.44 || 1.56<|eta_gamma|<2.5 ! to be applied at analysis level dR(gamma1,gamma2)>0.45 pT_jet>30 GeV |eta_jet|<4.7 Standard isolation: ETiso<0.1*ET in dR<0.4 cone (necessary if comparing with the showered samples)

sqrt(s)=8 TeV CT10, Scales={1.0,1.0,1.0} anti-kT jets with R=0.5

Add a second set more tailored for Higgs searches.

–GoSAM with parton shower (SHERPA, POWHEG)[maybe not for the proceedings] (Gudrun et al)

2. Understanding finely the background to the H125 or X125 (or 126!) –Make 2d reweighting of LO/LO+ codes used for prompt diphoton irreducible backgrounds to 2gammaNNLO(+gamma2MC) –Make 2d reweightings of LO/LO+ codes used for photon + jet and jet-jet (pure QCD) backgrounds to DIPHOX in gamma-h and h-h modes (above are for training of multivariate methods in Hgg analysis)

–An obstacle to the use of physics motivated truth models from partonic/hadron-level calculations for background modelling in Hgg is the 'sculpting' of distributions at the experimental level by acceptance/other selection criteria not possible to model at parton- or hadron-level, can we work on techniques for these extrapolations?

–Precise measurements of the different background components (gamma-gamma, gamma-jet, jet-jet) *in the nbd of the H125/6*

==⇒Studies to be undertaken (People)

–Make 2d reweighting (mass X qT or qT X deltaPhi) of LO/LO+ codes used for prompt diphoton irreducible backgrounds (MG) to 2gammaNNLO (+gamma2MC), should have m_gg distributions in slices of qt (Nicolas,Leandro et al)

–Make 2d reweightings of LO/LO+ codes used for photon + jet [and jet-jet (pure QCD)] reducible backgrounds to DIPHOX in gamma-h and h-h modes [see arxiv.org/abs/hep-ph/0203064](Susan et al)

–Observe shape differences, use the above for training and test of MVA discriminators at particle level (Nicolas, Leandro et al…). For both of the above, need unbiased generator-level samples

3. Isolation –For or against trying to use or at least evaluating Frixione isolation on experimental level? Other cones? –Use in Vgamma –Prior LH studies, which Frixione parameters are the best –Same as usual question of surviving collinear (to photon) radiation

==⇒Discussions/conclusions from June 5 meeting:

–It was the general conclusion that Frixione isolation is a technique developed to address a theoretical problem (treatment of collinear divergences) and its use should be limited to theoretical predictions. If not for a criterion like Fx, a large number of additional subprocesses (dijet production, gamma + jet) would have to be calculated at NNLO. But what about Eric's generalized Fx (see Eq. (19.2.1) of 2011 LH proceedings http://arxiv.org/pdf/1203.6803v2.pdf)? Decided it was useful only if to be used by experiment, which we are not recommending

–Studies at NLO by Daniel, Leandro et al (LPNHE Photon workshop Paris 2012) showed that inclusive diphoton cross-section changed by between<1-3% (corresponding to fragmentation component of between 6-16% [but of course not physical to consider it alone]) for 'standard' (solid DIPHOX dir/no box + 1frag + 2frag) and 'smooth' (continuous Frixione DIPHOX dir/no box only) isolations, except for quite loose isolation cuts (==Ehad_Tmax=0.5Pt for Pt>30, 40 GeV) Fx params used: R0=0.4, n=1 epsilon=0.05 and 0.5 tried)

–Results by Gudrun et al. (1303.0824) gamma gamma + jet NLO including fragmentation with GoSam (slides to be shown at next meeting) for inclusive and exclusive cases, compared scale dependence stability and some differential k-factors between standard cone and Frixione isolation.

–Question I had after the meeting: Does the difference wrt Fx decrease even more when using other cones available in DIPHOX (such as crown), which is close to what exp. use? This seems not to have been checked at partonic level in 2011.

–Other question after the meeting: How do these results change for lower-Pt photons

–It was suggested to define a set of observables in order to identify the kinematical regimes/zones where fragmentation contribution really is significant (cos theta*, mgg, dr(gamma,jet), deltaPhi…)

–Uniform 'LH accord' on recommended Fx parameters (n, epsilon, R0) for TH to use?

–Eric P. since the meeting has proposed a 'hadron machine-friendly' recast of the Fx Et profile chi(r [sometimes noted as 'delta']), to be discussed in the next (Monday) meeting:

“the function giving the E_T profile vs. the cone radius R currently used in the Frixione criterion is… [( 1 - \cos r)/(1-\cos R_max)]^{n} This form is inherited from Stefano's original PRL paper, which illustrated the criterion in the e+e- case. Indeed in this case the motivation is, the collinear pole in the final state read 1/(1-\cos \theta) where \theta was the orthorial angle in spherical coordinates in the e+e- c.m.s.frame, and the diff element of phase space read dtheta \sin \theta = d \cos \theta, so Stefano F. found it shrewd to use a function that led to easily handled analytic formulae.

In the hadron collider framework the collinear poles instead look conveniently like \propto 2(\cosh \Delta y) - \cos (\Delta \phi)} in terms of rapidity differences and azim angle differences: nothing to do with the e+e- inspired formula. On the other hand in the nearly colln limit 2 (\cosh( \Delta y) - \cos (\Delta \phi)} ~ [( \Delta y)^2 + (\Delta \phi)^2] = r^2: therefore a E_T profile E_T = E_Tmax * (r / R_max)^{2n) would do the job equally well. Since we seek a simple form, why not adopting this minimalist choice?”

==⇒Slides/plots shown by (June5): Gudrun (plots from 1303.0824), Daniel (LPNHE slide 12),Leandro (https://indico.cern.ch/conferenceTimeTable.py?confId=215483#20130423 slides 12-16), Suzanne(Daniel's presentation at CMS QCD photons group 14dec11, slide 4 but also LH proceedings 11: http://arxiv.org/pdf/1203.6803v2.pdf pp 165-179 )

==⇒Discussions/conclusions from June 10 meeting:

–Daniel presented results from he and Leandro showing differential diphoton cross-section behaviour for kinematical observables m_gg, deltaPhi and costheta* with a) DIPHOX full direct + 1-frag + 2-frag (NLO frags) with standard solid cone b) DIPHOX with direct and LO frag and c) DIPHOX with direct only and smooth Fx. Results in the slides but notably the largest difference was between a) and b) (~5%). These were just preliminary first studies and need more statistics for almost all observables. Need to compare with GoSam results from Gudrun. A fixed Et limit at the outer cone of 2 GeV (maybe a bit too tight) was taken. They tried several Et profiles (chi®) including Eric's 'hadron-friendly' one, see above. The results for different profiles can be used to determine the uncertainty on the theoretical isolation. Leandro thinks the sensitivity to outer cone size (from 0,4 to 0,3) is not too significant.

–Gudrun presented results from 1303.0824: in slide 4, stability of inclusive diphoton + jet cross-section as a function of epsilon_c and Fx_epsilon for scale variations for inclusive and exclusive jet cuts, conclusion is that the scale variation bands are thicker for lower epsilon values for Fx (but Fx epsilon does not mean the same thing as cone epsilon. In standard cone, 'upturn' at low values of epsilon may be due to missing NLO frag. On slide 6, m_gg distribution, the black curve (NLO, z_c=0,1) is the one to compare with Daniel's/Leandro's results.

–Decision taken to push adoption of Eric's 'hadron-collider-friendly' Et profile for use by TH pending completion of studies being undertaken by the group

==⇒Slides/plots shown by (June10): Daniel (slides), Gudrun (slides linked from main LH program under June 10 photons meeting)

==⇒Studies to be undertaken (People)

–Identify zones in differential phase space where fragmentation component is important (treated also in 1. )

–Check Standard/discretized Fx/annulus cones for total cross-section and differential distributions, frag. component (DIPHOX/Gamma2MC, GoSam) (Daniel/Leandro et al.) continuation of above studies, try a few more distributions including qTand z and try Eric's 'hadron-friendly' Et profile. Also evaluate change in size of outer cone from 0.4 to 0.3 [now used more by some experiments like CMS] (Gudrun et al): Consider cases where renormalisation and factorisation scales are not equal to each other, and check pt_jet dependence as well)

–Discuss and decide on 'LH Fx Accord' for TH, a good name is probably 'LH Tight Photon Isolation Accord' (all)

4. Hgg signal/background –Propagate new knowledge on the gg-ggf background-signal interference (Martin,Dixon et al) to the exps. signal modelling (mass shifts..) –Investigate new possible observables for greater S/B sensitivity (using knowledge from 2. above)

==⇒Studies to be undertaken (People)

–Try to implement Dixon recent code (see 1305.3854v1) and evaluate with experimentally realistic signal functions (not just a simple Gaussian)…(Philippe)

This has been done in Sherpa and the process will be included in version 2.0.0, as advertised here

Synthesis of Studies to be undertaken/finished for the proceedings (June 11 2013)