Using associated top quark production to probe for new physics within the framework of effective field theory
SSNSN-323-63January 27, 2021
Using associated top quark production to probe for newphysics within the framework of effective field theory
Brent R. Yates
Department of PhysicsThe Ohio State University191 West Woodruff AveColumbus, OH 43210, USA
Signs of new physics are probed in the context of an Effective FieldTheory using events containing one or more top quarks in associationwith additional leptons. Data consisting of proton-proton collisions at acenter-of-mass energy of √ s =13 TeV was collected at the LHC by theCMS experiment in 2017. We apply a novel technique to parameterize 16dimension-six EFT operators in terms of the respective Wilson coefficients(WCs). A simultaneous fit is performed to the data in order to extractthe two standard deviation confidence intervals (CIs) of the 16 WCs. TheStandard Model value of zero is completely contained in most CIs, and isnot excluded by a statistically significant amount in any interval.PRESENTED AT th International Workshop on Top Quark PhysicsDurham, UK (videoconference), 14–18 September, 2020 a r X i v : . [ h e p - e x ] J a n Introduction
The Standard Model (SM) of particle physics is one of the most complete and pre-cise models to date, but it only accounts for 5% of the known universe. The SMcurrently provides no correct explanation for dark matter and dark energy, the hier-archy problem, and baryon asymmetry, to name a few. The Large Hadron Collider(LHC) located at CERN can currently probe a center-of-mass energy of √ s =13 TeV.Therefore, the natural question arises: what if new physics beyond the SM occurs atan energy scale above what the LHC can probe directly? The formalism of EffectiveField Theory (EFT) allows us to approximate new physics above this scale purely interms of SM fields. The strength of each new physics operator ( O ) is controlled bythe so called Wilson coefficients (WCs), and are suppressed by powers of the energyscale Λ. The effective Lagrangian may be written as L EFT = L SM + (cid:88) d,i c ( d ) i Λ d − O ( d ) i , (1)where L SM is the SM Lagrangian, c i is the i th WC, and d is the dimension of theoperator. It is important to note that all odd numbered dimensions violate leptonand/or baryon number conservation. This analysis focuses on dimension six; higherdimensions are suppressed by additional powers of Λ, making them unimportant atthis level of precision.The analysis described in this proceeding uses a novel technique to examine datacollected by the CMS experiment in 2017, corresponding to an integrated luminos-ity of 41 . − . It performs a global fit across all processes—including signal andbackground. We specifically probe EFT effects using multilepton final states. Theprocedure used helps to constrain the systematic uncertainties, and any correlationsrely solely on the data—no assumptions are made. The production channels exam-ined are: ttl ν , ttll, tllq, and tHq, where H → bb is specifically removed. The completedetails of this analysis may be found in [1]. The EFT may be parameterized in simulations by splitting the matrix elements ( M )into SM and EFT terms M = M SM + (cid:88) j c j Λ M j . (2)The cross section is proportional to M , and each simulated event may be viewed asa differential piece of the cross section with an event weight w . Therefore, we may1arameterize these weights using w i (cid:32) (cid:126)c Λ (cid:33) = s i + (cid:88) j s ij c j Λ + (cid:88) j s ij c j Λ + (cid:88) j,k s ijk c j Λ c k Λ , (3)where the structure constants ( s ) correspond to: the SM term ( s ), interference be-tween the SM and EFT ( s ), pure EFT terms ( s ), and interference between EFTterms ( s ). These weights may be summed to produce the predicted event yields asa function of the WCs.Simulations are generated with non-zero WC values at leading order, and ex-tra partons are included when possible to improve our sensitivity. Initial values arechosen to include all relevant phase space and to optimize the statistical power— σ = (cid:80) w i ( (cid:126)c ). The weight of each event accounts for variations in the yield dueto EFT effects, and are used to solve for the structure constants in the quadraticparameterization. These quadratic functions are then used to fit to the data.The simulations are made using the dim6TopEFT model [2]. Due to limitationsin the model, only tree-level simulations are possible. The 16 operators which havethe largest impact on the signal processes, and relatively small impact on the tt back-ground, are considered. Only the real components are considered since the imaginarycoefficients lead to CP violation, and are well constrained by EDM experiments andB → X s γ decays. The analysis is split into 35 sub-categories, including: lepton ( (cid:96) ) multiplicity, sum ofthe lepton charges, jet multiplicity, and b-tagged jet multiplicity. A BDT is appliedto help separate the prompt leptons from the non-prompt leptons. All final-state ob-servables are an admixture of the processes—the method does not require we separatethe states. Each analysis sub-category stores the sum of the quadratic coefficients,and therefore the event yields are fully parameterized by the WCs. Table 1 lists allthe categories used.Each category listed in Table 1 is treated as a Poisson experiment with a proba-bility of obtaining the observed data. A profiled likelihood is used simultaneously fitall categories and is used to extract the 2 standard deviation ( σ ) confidence intervals(CIs). Two fitting procedures are used: one where a single WC is fit while the other15 are treated as unconstrained nuisance parameters, and another where a single WCis fit while the other 15 WCs are fixed to their SM value of zero. The first fittingprocedure is the more physical of the two, as there is no reason for new physics to2able 1: Requirements for the different event categories. Requirements separated bycommas indicate a division into subcategories. The b jet requirement on individualjets varies based on the lepton category, as described in the text. Selection 2 (cid:96) ss 3 (cid:96) ≥ (cid:96) Leptons Exactly 2 leptons Exactly 3 leptons ≥ (cid:80) (cid:96) q < , (cid:80) (cid:96) q > (cid:80) (cid:96) q < , (cid:80) (cid:96) q > ≥ ≥ ≥ ≥ ≥ ≥ ≥ ≥ | m (cid:96)(cid:96) − m Z | >
10 GeV | m (cid:96)(cid:96) − m Z | ≤
10 GeV - only favor one WC. The second procedure is an extreme scenario where nature hasa single WC. The ability to fit this single WC is limited by the lack of knowledge ofthe other 15.Systematic uncertainties are treated as nuisance parameters in the profiled fit. Themost important systematic uncertainties in this analysis are: the misidentified lep-ton rate estimate, and simulation modeling including matrix-element parton-showermatching, missing parton uncertainties, and scale uncertainties.
Misidentified lepton rate estimate
Contamination from non-prompt leptonsentering into the analysis region are to be expected. This is overcome by examininga multijet enriched background region and comparing this to a tt + γ enriched back-ground. The limited statistics of the tt + γ background is taken into account, and istreated as an additional source of uncertainty. Simulation modeling uncertainties
Uncertainties in the process of matchingmatrix element simulations to those produced via parton shower models must beaccounted for. The leading term in this uncertainty is from matching the extra partonsadded to the final-state jets. An additional missing parton uncertainty must beapplied to any samples which could not be generated with extra partons. This involvescomparing leading order EFT effects without extra partons to next-to-leading orderSM simulations, and assigning an uncertainty to cover any discrepancies. Finally, thescale uncertainties due to initial- and final-state radiation are taken into account.
The 1 σ and 2 σ CIs are visualized in Figure 1. When the other 15 WCs are fixed tozero c tW , c t ϕ , and c ϕ t obtain broader disjoint 1 σ CIs. This is due to the quadraticnature of the parameterization, which broadens the profiled likelihood curves. None3 [TeV Λ Wilson coefficient CI / − − − − c tW c tZ c t φ ÷ 5 c − φ Q ÷ 2 c tG × 2 c bW c φ Q c φ tb c φ t ÷ 2 c ℓ )Q ℓ c −( ℓ )Q ℓ c ( ℓ )Qe c ( ℓ )t ℓ c ( ℓ )te c S ( ℓ )t c T ( ℓ )t ) σ Others profiled (2 ) σ Others profiled (1 ) σ Others fixed to SM (2 ) σ Others fixed to SM (1
CMS (13 TeV) Figure 1: Observed WC 1 σ (thick line) and 2 σ (thin line) confidence intervals (CIs).Solid lines correspond to the other WCs profiled, while dashed lines correspond to theother WCs fixed to the SM value of zero. In order to make the figure more readable,the c ϕ t interval is scaled by 1 /
2, the c tG interval is scaled by 2, the c − ϕ Q interval isscaled by 1 /
2, and the c t ϕ interval is scaled by 1 / Charge misid. Misid. leptons Diboson Triboson Conv. Httlltt ν ltt qltl tHq Total unc. Obs. E v en t s Prefit (13 TeV) CMS ℓ ss ( + ) ℓ ss ( − ) ℓ ( + ) ℓ ( − ) ℓ ( + ) ℓ ( − ) S FZ S FZ ℓ O b s . / p r ed . E v en t s Postfit (13 TeV) CMS ℓ ss ( + ) ℓ ss ( − ) ℓ ( + ) ℓ ( − ) ℓ ( + ) ℓ ( − ) S FZ S FZ ℓ O b s . / p r ed . Figure 2: Expected yields prefit (left) and postfit (right). The postfit values of theWCs are obtained from performing the fit over all WCs simultaneously. “Conv.”refers to the photon conversion background, “Charge misid.” is the lepton chargemismeasurement background, and “Misid. leptons” is the background from misiden-tified leptons. The jet multiplicity bins have been combined here, however, the fit isperformed using all 35 event categories. The lower panel is the ratio of the observationover the prediction.
ACKNOWLEDGEMENTS
We would like to acknowledge the CMS Collaboration for their work in maintainingthe CMS experiment and collecting all relevant data for this analysis. We also thankAdam Martin and Jeong Han Kim for their theoretical guidance in configuring anddebugging the EFT model used to generate the signal samples in this analysis.
References [1] A. M. Sirunyan et al. , “Search for new physics in top quark production withadditional leptons in proton-proton collisions at √ s = 13 TeV using effective fieldtheory,” 12 2020.[2] D. Barducci et al.et al.