More variables or more bins?

Impact on the EFT interpretation

of Drell-Yan measurements
arXiv:2404.10569v1 [hep-ph] 16 Apr 2024

Samuele Grossi1,2* and Riccardo Torre2*

1* Department of Physics, University of Genova, Via Dodecaneso 33,
Genova, 16146, Italy.
2* INFN, Sezione di Genova, Via Dodecaneso 33, Genova, 16146, Italy.

*Corresponding author(s). E-mail(s): [email protected];

[email protected];

Abstract
We generalize previous studies on constraining operators of the Standard Model Effec-
tive Field Theory using Drell-Yan (DY) measurements to include at the same time
all relevant operators and uncertainties. It has been shown that fully differential mea-
surements (triple differential for neutral and double differential for charged) are more
sensitive to EFT effects. Nevertheless, due to the finite statistics, the fully differen-
tial measurements sacrifice some statistical power on the shape (less invariant mass
or transverse momentum bins) in favour of more kinematic variables. We show that
when the observables are particularly sensitive to the shape of the distributions,
such as the invariant mass of the two leptons in neutral DY, the single differential
measurement with more bins, may be as sensitive as the fully differential one, at
least for specific EFT operators. This suggests to always supplement fully differential
analyses with projections into the relevant distributions evaluated with finer bins.

Keywords: SMEFT, Drell-Yan, LHC, Multi-differential measurements

Contents The so-called LHC high-pT precision pro-
gram of the LHC mainly relies on the presence
1 Introduction 2 of SMEFT operators whose effects on precision
observables grow with energy. A prototype exam-
2 Cross-sections parametrization 3 ple, within this program, is the Drell-Yan (DY)
process, which is experimentally clean and the-
3 Uncertainties 3
oretically well-understood, and is sensitive, in a
4 The likelihood 5 generic SMEFT basis, to a set of two-quark-two-
lepton four-fermion operators, and to a set of
5 Projections and results 5 operators that modify the properties of the SM
gauge bosons that contribute to the DY process
6 Conclusions 8 in the SM.
The DY process has originally been considered
A Cross-section parametrization 8 in this framework in Ref. [6], where it was shown
how, exploiting the growing-with-energy effects
B Likelihood 9 of some relevant EFT operators on measurable
observables, could lead to EW precision measure-
C Fine Binning 9
ments at the LHC that could compete, or even
surpass, the ones obtained at LEP. Studies tar-
1 Introduction geting DY have been improved [6–8], even with a
first attempt of interpretation of a NP search in
Our current best chance to investigate New
terms of EFT operators by the CMS Collabora-
Physics (NP) scenarios responsible for explaining
tion [9]. The same idea of targeting growing with
the Electroweak (EW) hierarchy hinges on par-
energy effects to indirectly explore heavy NP was
ticle colliders, which, with their high energy and
also applied to a variety of other channels, among
luminosity, are able to probe, both directly and
which di-jets [10–13], di-bosons [14–21], di-quarks
indirectly, the existence of new particles and inter-
[11], and di-tops [22–25], and even four-tops [26].
actions. In all scenarios where new particles are
In this work we focus again our attention
too heavy to be directly produced at the LHC, the
on the DY processes. Our aim is to generalize
only way to detect their presence is through the
previous constraints on the Wilson coefficient of
indirect effects they have on Standard Model (SM)
SMEFT operators, including at the same time
observables. The Standard Model Effective Field
all the relevant operators [8] and all the rele-
Theory (SMEFT) [1–5] is a fundamental tool in
vant uncertainties [7]. Moreover, we investigate
this context, since it helps to organize the possi-
the impact of different binning specifications and
ble NP effects of SM observables systematically. In
single vs multi-differential measurements to assess
this regard, the SMEFT can be seen as the mod-
the best strategy of presenting measurements to
ern version of the SM, of which the dimension-4
maximize the sensitivity to SMEFT operators.
SM Lagrangian is the long distance description.
We consider the dimension-6 two-quark-two-
Hadron colliders have long been recognized as
lepton four-fermion operators as written in the
challenging for precision physics due to the inher-
Warsaw basis [27], listed in Table 1. This repre-
ently noisy conditions of hadronic interactions,
sents, in the chosen basis, the full set of operators
which significantly complicate precise measure-
that lead to growing-with-energy effects in the DY
ments. Nevertheless, in the last decade, the LHC
process [8]. In our analysis we consider both the
ability to pursue a high-energy precision pro-
neutral DY, with a di-lepton final state, and the
gram, has emerged. This program complements
charged DY, with a lepton and a neutrino final
the Higgs precision program on one side, and the
state, resulting experimentally in a single lepton
“low-energy” flavor program on the other side,
plus missing energy final state. We exploit both
showcasing the versatility of hadron colliders in
the fully differential cross-section, triple differen-
advancing diverse research objectives, beyond the
tial for the neutral channel, and double differential
traditional search for new particles at the energy
for the charged one, and the single differential
frontier.
cross-section, in the di-lepton invariant mass mℓℓ
S 3

Dimension-6 current-current operators to QCD radiative corrections. Such factorization

(3)
Olq = (lL σI γ µ lL )(q L σI γµ qL ) holds separately for each chirality channel, and
(1)
Olq = (lL γ µ lL )(q L γµ qL ) allows one to generate events for the full SM+EFT
Oqe = (q L γ µ qL )(ℓR γµ ℓR ) process, for any value of the Wilson coefficients,
Olu = (lL γ µ lL )(uR γµ uR )
and at NLO QCD accuracy, by just reweight-
Old = (lL γ µ lL )(dR γµ dR )
Oeu = (ℓR γ µ ℓR )(uR γµ uR ) ing events for the SM process at NLO QCD
Oed = (ℓR γ µ ℓR )(dR γµ dR ) accuracy. The reweighting coefficients depend on
the chirality of the quarks and leptons and on
Table 1 The seven dimension-6 contact operators
contributing to Drell-Yan processes written in the kinematic quantities such as the partonic center-
notation of ref. [27]. of-mass squared s. The parton shower at NLO
QCD accuracy is not modified by the EFT oper-
ators, therefore there is no need to account for
(neutral) and lepton transverse momentum pT their contribution. The SM cross-section and the
(charged), respectively. parton shower at NLO QCD accuracy have been
Since no new experimental measurements of obtained using respectively Powheg [28, 29] and
differential DY at high energy has appeared since pythia 8 [30]. To obtain results whose preci-
the last analysis of Ref. [8], we rely, in our anal- sion matches the experimental measurements, also
ysis, on simulated data. This allows us to explore the NNLO QCD corrections have to be taken
the impact of different binning strategies, and into account for the Standard Model contributions
to assess the importance of presenting indepen- (using FEWZ [31]). These contributions are com-
dent single and multi-differential measurements. pletely negligible for the new physics. Finally, the
We do so by comparing the sensitivity of analyses reweighting coefficients can be modified in order to
based on the fully-differential cross-section, on the include also the EW next-to-Leading-Log (NLL)
single differential cross-section obtained by inte- corrections which become important in the high
grating the fully-differential one over the angular energy regime we are interested in [7].
variables, and the “enhanced” single differential All relevant uncertainties have been taken into
cross-section obtained with a finer binning in the account following the prescription of Ref. [7].
dimensionful kinematic variables, that are the Exploiting the fact that the cross-section in each
invariant mass of the two leptons in the neutral bin, in both the neutral and the charged chan-
DY, and the lepton transverse momentum in the nels, is a quadratic polynomial in the Wilson
charged DY. coefficients, we can perform the Cholesky decom-
We find that, for parameters that are partic- position with coefficients parametrizing the SM
ularly sensitive to the shape of the distribution contribution, the interference between SM and
in the dimensionful kinematic variables, the single EFT operators, and the quadratic EFT contribu-
differential cross-section in which the full statis- tions. In this way we obtain a total of 36 weights
tics is used to optimize the binning compatibly for the neutral channel (1 for the SM, 7 for the
with the statistical uncertainty, can be as sensitive interference, and 28 for quadratic terms) and 3
as the fully differential one. This suggests that, weights (1 SM, 1 interference, and 1 quadratic)
in forthcoming experimental analyses, it would be for the charged one (see Appendix A for details).
optimal to have available both fully-differential All uncertainties are parametrized as fluctuations
and single differential (with optimized binning) of these weights, which we refer to as Cholesky
cross-section information. coefficients.

2 Cross-sections 3 Uncertainties
parametrization
In order to match the high precision measure-
The SM+EFT cross-section is obtained using the ments with theoretical predictions one has to
reweighting strategy introduced in Ref. [7]. This take into account the most important theoretical
is based on the fact that in DY the new physics and experimental sources of uncertainties. This is
contributions factorize not only with respect to done through the introduction of nuisance param-
the tree-level cross-section, but also with respect eters as in Ref. [7]. We assume that the nuisance
parameters deriving from theoretical uncertain- of the perturbative QCD series are accounted
ties modify the Cholesky coefficients (in each bin), for by the introducion of a nuisance param-
while the ones linked to the experimental uncer- eter θITU for each bin. We consider different
tainties have a direct impact on the number of values of the factorization and renormaliza-
expected events in each bin. We discuss the imple- tion scales, µF and µR , respectively: √ their
mentation of both classes of uncertainties in the central values are set to µR = µF = ŝ and
following. To do so we indicate with ci,I the i- we let them vary independently by multiplica-
1
th Cholesky coefficient in the I-th bin, calculated tive factors 2±1 , 2± 2 , and 1, with the latter
with Standard Model Central values of αS , Parton value corresponding to the central value. This
Density Functions (PDFs), and factorization and gives a grid with 25 values.
renormalization scales that we specify below, and Again, the missing higher order uncertainty
with ci,I the corresponding Cholesky coefficient is not leading in the SM. In particular, while
as function of the nuisance parameters. For sim- the contribution deriving from the truncation
plicity, we do not separate the discussion between of the NLO EW perturbative series could be
neutral and charged channels. It should never- completely neglected, the one linked to the
theless be clear that all quantities differ in each truncation of the QCD NNLO perturbative
channel. series is relevant only for the c0 SM coef-
• Theory uncertainties ficient. Such contribution is parameterized
as:
– Monte Carlo statistic
θ TU θ TU
θITU θITU
The uncertainty deriving from Monte Carlo c0,I (θITU ) = c0,I ekI = ekI ,
statistics is negligible if the simulations pro-
cmax cmin
!
TU | 0,I − c0,I | | 0,I − c0,I |
vide accurate enough predictions for the SM kIθ = max , ,
terms, well below 1%. This is guaranteed by 10 10
the fact that the new physics contributions (2)
are accounted for using reweighting, so that where cmax
0,I and cmin
0,I are the maximum and
their accuracy aligns with that of the SM minimum value of c0,I within the 25 different
terms. replicas specified above.
– Strong coupling constant – Parton Distribution Functions
The uncertainty associated to the value of PDF uncertainty is the most important the-
αS is accounted for through a single nuisance oretical uncertainty in the SM DY process
parameter θαS , which is the same across all [32–34]. Therefore, we account for it in the
channels and bins. The effect of θαS is esti- Cholesky coefficients of both the SM and
mated using Powheg SM DY [28] Monte the new physics contributions. The PDF
Carlo samples reweighted for upper (αSu = uncertainties are parametrized by a vector of
0.1195), lower (αSl = 0.1165), and central nuisance parameter θiPDF , corresponding to
value (αS = 0.1180) of αS at the scale of the the eigenvalues of the PDFs within the Hes-
Z mass. Since this uncertainty is not the lead- sian representation, for each bin. As before,
ing one in the SM, we can ignore its effect we use Powheg to get the weights of the
on the new physics Cholesky coefficients and different Hessian components in the SM cal-
only retain the SM part, parameterized by the culation. In our case it is enough to con-
coefficient c0 : sider the 30 components in the PDF set
PDF4LHC15 nlo 30 pdfas (code 90400 in
αS αS αS αS
c0,I (θαS ) = c0,I ekI θ = ekI θ , the LHAPDF database [35]). The advantage
  (1) of the Hessian set is that it automatically
kIαS = max | cu0,I − c0,I |, | cl0,I − c0,I | . provides a definition of the relevant nuisance
parameters that can be used across different
with cl0,I = c0,I (αSl ) and cu0,I = c0,I (αSu ). processes, simplifying the combination of dif-
– Missing higher orders (QCD and EW) ferent channels. The parametrization of the
The uncertainty deriving from the truncation Cholesky coefficients as function of the PDFs
S 5

nuisance parameters is: with the test-statistic tµ defined by

 
30 c(i) − c
k,I PDF
 
PDF k,I
X
ck,I (θi ) = ck,I exp  θi , tµ = −2 supδ log L (µ, δ)−sup(µ,δ) log L (µ, δ) .
ck,I
i=1
(5)
for k = 0, 1, ..., 35 (Neutral channel), Here L is the likelihood as function of the param-
for k = 0, 1, 2 (Charged channels). eters, µ represents the parameters of interest, that
(3) are the seven Wilson coefficients corresponding to
the EFT operators in Table 1, and δ are the nui-
• Experimental uncertainties sance parameters.
Uncertainties associated to the experimental Assuming that tµ follows asymptotically a χ2 dis-
setup could only be accounted for within an tribution with a number of degrees of freedom
analysis or with available information from the equal to the number of parameters of interest [36],
experiments. Despite our code gives us full flexi- we can set confidence level boundaries on the Wil-
bility in accounting for experimental systematic son coefficients. Bin by bin the likelihood is a
uncertainties with full correlation information, Poisson distribution of the number of observed
possibly different in the electron and muon events nI with mean µI , multiplied by the like-
channels, in this work we considered a relatively lihood that constrains each nuisance parameter
simple pattern of uncertainties, consistent with (with auxiliary data), which can be parametrized
past analyses. We assume a 2% uncorrelated as a standard Normal distribution, since the rele-
uncertainty, parametrized by a single nuisance vant scales have already been taken into account
parameters θL , across all bins and channels, in defining the dependence of the Cholesky coeffi-
from the determination of the integrated lumi- cients on the nuisance parameters. The combined
nosity. For all the other experimental systematic likelihood is then simply written as:
uncertainties, we consider 2% and 5% uncer-
tainties in each bin, uncorrelated among bins, Lcomb (µ, δ) = Ln (µ, δ n ) × Lc,+ (µ, δ c )
for the neutral and charged channels, respec- (6)
tively. These are parametrized by a set of nui- × Lc,− (µ, δ c ) × Laux (δ) ,
sance parameters θIexp for each bin and each
channel. where the explicit definition of each term is given
We assume that the experimental uncertainties in Appendix B.
do not affect directly the Cholesky coefficients,
and only modify the number of expected events 5 Projections and results
in each bin µI from its theoretical prediction µth
I
as follows: We present here the 95% confidence level (CL)
! projected bounds, at the LHC at 13 TeV, on each
X h√ iJ of the Wilson coefficients of the operators in Table
th exp exp L
µI = µI exp Σ θJ + 0.02θ . 1. We assume that the SM central value for the
I
J number of observed events in each bin is given by
(4) nI = L · σ I , where L is the luminosity and σ I is
Here Σexp is the covariance matrix of the exper- the cross-section in each bin calculated assuming
imental systematic uncertainties in the space of the central values for αS , PDFs, and factoriza-
bins, that we take proportional to the identity tion and renormalization scale, and setting to zero
matrix, and µth th
I = L · σI , where L is the inte- all the experimental uncertainties and the EFT
grated luminosity, and σIth is the cross-section effects. We consider luminosity values of 100, 300,
in each bin including all sources of theoretical and 3000 fb−1 as in Ref. [7]. The binning specifica-
uncertainties specified before. tion for the multi-differential analysis is analogous
to the one of Ref. [8]. In particular, given the
4 The likelihood dilepton invariant mass mℓℓ , the scattering angle
c∗ = cos θ∗ , and the absolute value of the ratio
The constraints on the Wilson coefficients are between the dilepton rapidity and its maximum
obtained using the profiled likelihood ratio test, value |y/ymax |, the binning for the neutral channel
 95% CL L = 100 fb−1 L = 300 fb−1 L = 3000 fb−1
−9 −2
[10 GeV ] Fully-Dif Single integrated Single fine bins Fully-Dif Single integrated Single fine bins Fully-Dif Single integrated Single fine bins
(3)
Glq [−2.04, 2.10] [−2.23, 2.31] [−2.07, 2.14] [−1.43, 1.47] [−1.65, 1.71] [−1.50, 1.56] [−0.73, 0.75] [−1.02, 1.08] [−0.91, 0.96]
(1)
Glq [−8.50, 14.7] [−9.48, 20.2] [−8.85, 18.0] [−5.75, 9.01] [−6.72, 15.5] [−6.07, 12.1] [−2.66, 3.36] [−3.68, 9.22] [−3.04, 4.54]
Gqe [−8.72, 15.2] [−11.1, 18.0] [−10.4, 16.8] [−5.98, 10.8] [−8.10, 13.9] [−7.38, 12.5] [−2.88, 5.26] [−4.68, 9.10] [−3.95, 7.19]
Glu [−8.21, 13.5] [−11.3, 21.1] [−10.5, 18.5] [−5.40, 8.02] [−8.03, 16.0] [−7.22, 12.7] [−2.41, 3.03] [−4.43, 10.0] [−3.65, 5.93]
Gld [−27.1, 18.1] [−30.5, 21.6] [−29.3, 20.5] [−20.3, 12.8] [−23.6, 16.1] [−22.3, 14.9] [−11.7, 6.70] [−15.2, 9.72] [−13.6, 8.41]
Geu [−6.29, 7.37] [−6.91, 8.54] [−6.27, 7.42] [−4.16, 4.64] [−4.77, 5.57] [−4.16, 4.68] [−1.88, 1.99] [−2.52, 2.75] [−2.00, 2.11]
Ged [−27.7, 15.5] [−29.9, 16.7] [−27.1, 15.4] [−20.5, 10.9] [−23.2, 12.2] [−19.8, 10.9] [−11.4, 5.56] [−15.4, 7.09] [−11.1, 5.84]

Table 2 One dimensional single parameter 95% confidence intervals for the seven EFT Wilson coefficients in units of
10−9 GeV−2 , for integrated luminosity values of 100 fb−1 , 300 fb−1 and 3000 fb−1 . The results are obtained considering a
fully-differential cross-section, a single differential cross-section obtained from the fully-differential one integrating over
angular and rapidity variables and a single differential cross-section with a fine binning.

Fig. 1 Comparison between the one dimensional single parameter (switch on one coefficient at a time and set the other
six to 0) 95% confidence intervals for the seven EFT Wilson coefficients. These results are obtained considering a multi-
differential cross-section (green), a single differential cross-section obtained from the multi-differential one integrating over
angular and rapidity variables (red) and a single differential cross-section with a fine binning (blue). Collider energy is set
to 13 TeV and integrated luminosity of 100 fb−1 , 300 fb−1 and 3000 fb−1 are considered.

is given by For the single differential analysis we con-

 sidered two different scenarios. In the first one

 mℓℓ : {300, 360, 450, 600, 800, 1100, the single differential cross-sections are obtained
1500, 2000, 2600, 13000} GeV , integrating the fully differential ones over angu-




c∗ : {−1, −0.6, −0.2, 0.2, 0.6, 1} , (7) lar and rapidity variables (neutral) and over


 y pseudo-rapidity variable (charged) keeping the
: {0, 1/3, 2/3, 1} . aforementioned binning for the dimensionful



ymax
kinematic variables. In the second one the large
Analogously, for the charged channel, given the number of expected events is exploited to consider
lepton transverse momentum pT and the absolute a finer binning in the mℓℓ /pT (neutral/charged)
value of the ratio between the lepton pseudo- variables. Such binning is obtained requiring that
rapidity and its maximum value |η/ηmax |, the the statistical uncertainty of our Monte Carlo
binning is given by simulation remains negligible (below 0.2%) in all
bins. With this constraint we get 113 bins in mℓℓ

p : {150, 180, 275, 300, 400, 550, for the neutral channel and 26 bins in pT for the
 T

charged one (the exact binning is reported in
 750, 1000, 1300, 7500} GeV ,
(8) Appendix C).
 η

 : {0, 1/3, 2/3, 1} .
ηmax
S 7

95% CL L = 100 fb−1 L = 300 fb−1 L = 3000 fb−1

−9 −2
[10 GeV ] Fully-Dif Single integrated Single fine bins Fully-Dif Single integrated Single fine bins Fully-Dif Single integrated Single fine bins
(3)
Glq [−2.17, 2.17] [−2.35, 2.39] [−2.16, 2.18] [−1.53, 1.54] [−1.73, 1.77] [−1.57, 1.59] [−0.82, 0.82] [−1.05, 1.10] [−0.93, 0.98]
(1)
Glq [−14.1, 17.2] [−15.7, 20.8] [−14.9, 19.5] [−10.5, 12.5] [−12.1, 16.3] [−11.2, 14.8] [−6.11, 6.78] [−7.77, 11.2] [−6.52, 9.22]
Gqe [−12.3, 15.6] [−16.9, 18.3] [−16.1, 17.1] [−9.01, 11.5] [−13.2, 14.2] [−12.2, 12.8] [−5.12, 6.50] [−8.77, 9.36] [−7.42, 7.52]
Glu [−10.9, 17.4] [−18.3, 22.1] [−17.4, 20.5] [−7.56, 12.4] [−14.1, 17.0] [−13.0, 15.2] [−3.84, 6.55] [−9.11, 11.2] [−7.59, 8.81]
Gld [−27.9, 24.6] [−32.7, 32.0] [−30.9, 30.8] [−20.9, 18.4] [−25.7, 25.5] [−23.6, 24.1] [−12.2, 10.7] [−17.4, 17.8] [−14.6, 15.8]
Geu [−10.4, 17.3] [−11.8, 19.2] [−11.3, 17.6] [−7.47, 12.9] [−8.85, 14.9] [−8.30, 13.1] [−4.05, 7.75] [−5.45, 9.88] [−4.74, 7.65]
Ged [−28.7, 25.7] [−30.2, 28.7] [−28.1, 27.7] [−21.7, 19.5] [−23.5, 22.9] [−21.1, 21.7] [−13.1, 11.5] [−15.6, 16.0] [−12.5, 14.3]

Table 3 One dimensional profiled (constraining one by one each coefficient while treating the other six as nuisance
parameters) 95% confidence intervals for the seven EFT Wilson coefficients in units of 10−9 GeV−2 , for integrated
luminosity values of 100 fb−1 , 300 fb−1 and 3000 fb−1 . The results are obtained considering a fully-differential
cross-section, a single differential cross-section obtained from the fully-differential one integrating over angular and
rapidity variables and a single differential cross-section with a fine binning.

Fig. 2 Comparison between the one dimensional profiled 95% confidence intervals for the seven EFT Wilson coefficients.
These results are obtained considering a multi-differential cross-section (green), a single differential cross-section obtained
from the multi-differential one integrating over angular and rapidity variables (red) and a single differential cross-section
with a fine binning (blue). Collider energy is set to 13 TeV and integrated luminosity of 100 fb−1 , 300 fb−1 and 3000 fb−1
are considered.

The constraints we obtained in the multi- cross-section, a single differential cross-section

differential analysis and in the single differential obtained integrating the fully-differential one over
one after the integration over the rapidity and the angular and rapidity variables, and a single differ-
angular variables are compatible with the results ential cross-section with a finer binning. The three
reported in Ref. [8]. Our bounds are slightly more aforementioned integrated luminosity values are
conservative because we also included the uncer- considered. The numerical values in the table are
tainties deriving from the value of the strong obtained considering only one Wilson coefficient
coupling constant αS , and the truncation of the at a time, with all others set to zero, building the
perturbative series. Furthermore, we considered a test-statistic tµ as function of a single parameter,
2% experimental uncertainty in the neutral chan- and assuming it is asymptotically distributed as
nel and a 5% in the charged one, while in Ref. [8] a χ2 with one degree of freedom. The results
a 2% uncertainty was considered for both chan- are also shown in Figure 1, which highlights the
nels. The single differential analysis with a finer differences between the three binning strategies.
binning was not considered in previous works.
From the figures and the table we can draw
Table 2 reports the 95% CL projected bounds some important conclusions.
on the Wilson coefficients in the three scenarios
discussed above: considering a fully-differential
• As expected, the fully-differential analysis of sensitivity for a subset of parameters. This
always provides the tightest constraints on the highlights the importance of presenting results in
Wilson coefficients. different ways, distributing the largest possible
• The single-differential analysis with fine bins amount of information contained in the analysis.
often reaches the same level of sensitivity as
the fully-differential one. This is particularly
(3)
true for the Wilson coefficients Glq , Geu , and Acknowledgments
Ged , which are therefore much less sensitive
We thank A. Wulzer and L. Ricci for useful
to angular distributions. The sensitivity is still
discussions. We also thank L. Ricci for help in
weaker than the fully-differential analysis for
comparing our results with those of Ref. [8], and
the other Wilson coefficients, but the differ-
E. Rizvi and G. Poddar for clarifications about the
ence is, especially in the negative end of the
experimental analyses in Ref.s [37, 38].
bound, generally less than 10%, which can be
considered within the precision of our determi-
nation, based on asymptotic formulae for the A Cross-section
likelihood-ratio test-statistic distribution. parametrization
• The single differential analysis obtained by inte-
grating the fully-differential cross-section over The cross-section as a function of the Wilson coef-
the angular and rapidity variables is gener- ficients is a non-negative quadratic polynomial in
(3) (1)
ally less sensitive than the fully-differential one. Glq , Glq , Gqe , Glu , Gld , Geu , Ged . The cross-
(1)
This is particularly true for the operators Glq , section in each bin I can therefore be parametrized
Gqe , Glu , and Gld , whose upper bounds reach using the Cholesky decomposition. In the neutral
differences from the fully-differential analysis channel, this can be written as:
of more than a factor of two. This implies
2
that the bound on these parameters is par- 8
X
ticularly sensitive to the angular and rapidity σIth,n = σ SM,n
I c20,I CIj Gj , (9)
distributions. j=1

The same results, but obtained marginalizing with the Cholesky matrix
the likelihood over the other Wilson coefficients,
instead of setting them to zero, are reported in
 
1 c1,I c2,I c3,I c4,I c5,I
c7,I c6,I
Table 3 and shown in Figure 2. The conclusions 0 c8,I c15,I c16,I c17,I c18,I
c20,I  c19,I
 
are unchanged. 0 0 c9,I c21,I c22,I c23,I
c25,I  c24,I
 
0 0 0 c10,I c26,I c27,I
c29,I  c28,I
CIj= 
6 Conclusions 0
 0 0 0 c11,I c30,I
c32,I 
 c31,I
0 0 0 0 0 c12,I
c34,I  c33,I
 
In this work we generalized previous work on 0 0 0 0 0 0
c35,I  c13,I
the determination of the bounds on the Wilson 0 0 0 0 0 c14,I0 0
coefficients of the EFT operators entering in the (10)
DY processes at the LHC. We have put together and the vector of Wilson coefficients Gj with j =
the “all-operators” approach of Ref. [8] with the 1, ..., 8
“all-uncertainties” approach of Ref. [7]. We found  
consistent results with the previous works in the (1) (3)
Gj = 1, Glq , Glq , Gqe , Glu , Gld , Geu , Ged ,
multi-differential and integrated single differential
(11)
analysis [8] and we added a single differential
where the 1 accounts for the SM contribution,
“fine binning” analysis. In turn, we have discussed
and the Cholesky coefficients ck,I are functions of
the sensitivity of different Wilson coefficients to
the nuisance parameters θαS , θiPDF , and θITU , as
angular and rapidity distributions, and we have
described in the main text. The same parametriza-
shown that while the fully-differential analysis is
tion holds for the charged channel, where it is more
always the most sensitive, the single differential
analysis with fine bins can reach a similar level
S 9

simply written as • charged

pT : {150, 163, 177, 191, 207, 225, 244, 264, 288,
! 2

1 c1,I

1 313, 342, 373, 407, 445, 488, 537, 591, 652, 723,
σIth,c = σ SM,c
I c20,I (3) . (12) 802, 896, 1003, 1130, 1292, 1493, 1770, 6500}
0 c2,I Glq
GeV.

B Likelihood References
Each factor of the combined likelihood is:
[1] I. Brivio and M. Trott, “The Standard Model
Nn
Y as an Effective Field Theory”, Phys. Rept.
Ln = Poisson [nIn | µIn (G, θ n )] , 793 (2019) 1, arXiv:1706.08945 [Semantic
In =1 Scholar].
Nc
Y [2] T. Cohen, N. Craig, X. Lu and D. Sutherland,
Lc,± = Poisson n± ±
 
Ic | µIc (G, θ c ) ,
“Is SMEFT Enough?”, JHEP 03 (2021) 237,
Ic =1
arXiv:2008.08597 [Semantic Scholar].

Y Nc Y
Nn Y 30 [3] A. Falkowski and R. Rattazzi, “Which EFT”,
Laux = fαS (θαS )fPDF (θiPDF )f (θITU
n
) JHEP 10 (2019) 255, arXiv:1902.05936
In =1 Ic =1 i=1 [Semantic Scholar].
×f (θITU
c
)fexp (θIexp )fexp (θIexp )fL (θL ) ,
n c
[4] J. de Blas, J. C. Criado, M. Perez-Victoria
where G is the vector defined in eq. (11), and and J. Santiago, “Effective description of
general extensions of the Standard Model:
θ n = θαS , θiPDF , θITU , θIexp , θL ,


n n the complete tree-level dictionary”, JHEP 03
, θIexp (2018) 109, arXiv:1711.10391 [Semantic


θ c = θαS , θiPDF , θITU
c c
, θL .
Scholar].
C Fine Binning [5] M. McCullough, L. Ricci and M. Riem-
The “fine binning” have been obtained imposing bau, “Boundaries of Universal Theories”,
to have a negligible MC statistical error in each arXiv:2312.03834 [Semantic Scholar].
bin (below 1%). We achieved this imposing to have
[6] M. Farina, G. Panico, D. Pappadopulo, J. T.
∼ 2 · 105 MC events for each bin in our simulation,
Ruderman, R. Torre and A. Wulzer, “Energy
that corresponds to a statistical error of ∼ 0.22%.
helps accuracy: electroweak precision tests
The resulting binning, for the neutral and charged
at hadron colliders”, Phys. Lett. B 772
channels, is:
(2017) 210, arXiv:1609.08157 [Semantic
• neutral Scholar].
mℓℓ : {300, 305, 309, 315, 320, 326, 332, 337, 342,
348, 355, 362, 368, 375, 381, 389, 397, 405, 413, [7] R. Torre, L. Ricci and A. Wulzer, “On the
420, 427, 435, 444, 453, 462, 470, 478, 487, 497, W&Y interpretation of high-energy Drell-
507, 518, 527, 536, 546, 556, 567, 580, 592, 602, Yan measurements”, JHEP 02 (2021) 144,
614, 626, 639, 653, 669, 684, 696, 709, 723, 739, arXiv:2008.12978 [Semantic Scholar].
756, 774, 792, 805, 819, 834, 850, 868, 888, 908,
924, 941, 959, 979, 1001, 1026, 1053, 1077, 1095, [8] G. Panico, L. Ricci and A. Wulzer, “High-
1115, 1137, 1161, 1188, 1218, 1251, 1275, 1297, energy EFT probes with fully differen-
1322, 1349, 1379, 1413, 1451, 1494, 1521, 1548, tial Drell-Yan measurements”, JHEP 07
1577, 1610, 1647, 1689, 1737, 1792, 1823, 1854, (2021) 086, arXiv:2103.10532 [Semantic
1888, 1926, 1969, 2018, 2076, 2142, 2184, 2220, Scholar].
2259, 2305, 2357, 2417, 2489, 2575, 2680, 2910,
[9] CMS Collaboration, A. Tumasyan et al.,
3105, 3365, 3752, 4126, 4802, 13000} GeV.
“Search for new physics in the lepton plus
 missing transverse momentum√ final state in from gluon fusion”, JHEP 11 (2023) 132,
proton-proton collisions at s = 13 TeV”, arXiv:2306.09963 [Semantic Scholar].
JHEP 07 (2022) 067, arXiv:2202.06075
[Semantic Scholar]. [19] C. Degrande and H.-L. Li, “Impact of
dimension-8 SMEFT operators on dibo-
[10] S. Alioli, M. Farina, D. Pappadopulo and son productions”, JHEP 06 (2023) 149,
J. T. Ruderman, “Catching a New Force arXiv:2303.10493 [Semantic Scholar].
by the Tail”, Phys. Rev. Lett. 120 101801,
arXiv:1712.02347 [Semantic Scholar]. [20] ATLAS Collaboration, G. Aad et al., “Inter-
pretations of the ATLAS measurements of
[11] S. Alioli, M. Farina, D. Pappadopulo and Higgs boson production and decay rates and
J. T. Ruderman, “Precision Probes of QCD differential
√ cross-sections in pp collisions at
at High Energies”, JHEP 07 (2017) 097, s = 13 TeV”, arXiv:2402.05742 [Seman-
arXiv:1706.03068 [Semantic Scholar]. tic Scholar].

[12] T. Han, S. C. I. Leung and M. Low, “Higgs [21] P. Englert, “New Physics Probes at Present/-
to bb̄ from Vector Boson Fusion for High- Future Hadron Colliders via Vh Production”,
Scale Physics”, arXiv:2305.01010 [Seman- . PhD thesis, Humboldt U., Berlin, 2023.
tic Scholar]. [InSpireHEP].

[13] SMEFiT Collaboration, J. J. Ethier, [22] J. Ellis, M. Madigan, K. Mimasu, V. Sanz

G. Magni, F. Maltoni, L. Mantani, E. R. and T. You, “Top, Higgs, Diboson and Elec-
Nocera, J. Rojo et al., “Combined SMEFT troweak Fit to the Standard Model Effec-
interpretation of Higgs, diboson, and top tive Field Theory”, JHEP 04 (2021) 279,
quark data from the LHC”, JHEP 11 arXiv:2012.02779 [Semantic Scholar].
(2021) 089, arXiv:2105.00006 [Semantic
Scholar]. [23] CMS Collaboration, T. C. Collaboration
et al., “Search for new physics in top quark
[14] G. F. Giudice, C. Grojean, A. Pomarol production with additional
√ leptons in proton-
and R. Rattazzi, “The Strongly-Interacting proton collisions at s = 13 TeV using
Light Higgs”, JHEP 06 (2007) 045, effective field theory”, JHEP 03 (2021) 095,
hep-ph/0703164 [Semantic Scholar]. arXiv:2012.04120 [Semantic Scholar].

[15] D. Liu, A. Pomarol, R. Rattazzi and [24] S. Tong, J. Corcoran, M. Fieg, M. Fenton
F. Riva, “Patterns of Strong Coupling for and D. Whiteson, “New Physics in Single
LHC Searches”, JHEP 11 (2016) 141, Resonant Top Quarks”, arXiv:2311.00121
arXiv:1603.03064 [Semantic Scholar]. [Semantic Scholar].

[16] LHC Higgs Cross Section Working [25] N. Elmer, M. Madigan, T. Plehn and
Group Collaboration, D. de Florian et al., N. Schmal, “Staying on Top of SMEFT-
“Handbook of LHC Higgs Cross Sections: 4. Likelihood Analyses”, arXiv:2312.12502
Deciphering the Nature of the Higgs Sector”, [Semantic Scholar].
arXiv:1610.07922 [Semantic Scholar].
[26] C. Englert, G. F. Giudice, A. Greljo and
[17] CMS Collaboration, A. M. Sirunyan et al., M. Mccullough, “The Ĥ-Parameter: An
“Combined measurements of Higgs boson √ Oblique Higgs View”, JHEP 09 (2019) 041,
couplings in proton–proton collisions at s = arXiv:1903.07725 [Semantic Scholar].
13 TeV”, Eur. Phys. J. C 79 (2019) 421,
arXiv:1809.10733 [Semantic Scholar]. [27] B. Grzadkowski, M. Iskrzynski, M. Misiak
and J. Rosiek, “Dimension-Six Terms in
[18] A. N. Rossia, M. O. A. Thomas and E. Vry- the Standard Model Lagrangian”, JHEP 10
onidou, “Diboson production in the SMEFT (2010) 085, arXiv:1008.4884 [Semantic
11

Scholar]. [37] ATLAS Collaboration, M. Aaboud et al.,

“Measurement of the Drell-Yan triple-
[28] S. Alioli, P. Nason, C. Oleari and E. Re, differential cross section in pp collisions
√
“NLO vector-boson production matched with at s = 8 TeV”, JHEP 12 (2017) 059,
shower in POWHEG”, JHEP 07 (2008) 060, arXiv:1710.05167 [Semantic Scholar].
arXiv:0805.4802 [Semantic Scholar].
[38] ATLAS Collaboration, G. Aad et al., “Mea-
[29] S. Frixione, P. Nason and C. Oleari, “Match- surement of the double-differential high-mass
ing NLO QCD computations with Parton Drell-Yan cross section in pp collisions at
√
Shower simulations: the POWHEG method”, s = 8 TeV with the ATLAS detector”,
JHEP 11 (2007) 070, arXiv:0709.2092 JHEP 08 (2016) 009, arXiv:1606.01736
[Semantic Scholar]. [Semantic Scholar].
[30] T. Sjöstrand, S. Ask, J. R. Christiansen,
R. Corke, N. Desai, P. Ilten et al., “An intro-
duction to PYTHIA 8.2”, Comput. Phys.
Commun. 191 (2015) 159, arXiv:1410.3012
[Semantic Scholar].

[31] Y. Li and F. Petriello, “Combining QCD and

electroweak corrections to dilepton production
in FEWZ”, Phys. Rev. D 86 (2012) 094034,
arXiv:1208.5967 [Semantic Scholar].

[32] J. Gao, L. Harland-Lang and J. Rojo, “The

Structure of the Proton in the LHC Pre-
cision Era”, Phys. Rept. 742 (2018) 1,
arXiv:1709.04922 [Semantic Scholar].

[33] E. Hammou, Z. Kassabov, M. Madigan,

M. L. Mangano, L. Mantani, J. Moore
et al., “Hide and seek: how PDFs can con-
ceal new physics”, JHEP 11 (2023) 090,
arXiv:2307.10370 [Semantic Scholar].

[34] S. Amoroso et al., “Snowmass 2021 Whitepa-

per: Proton Structure at the Precision Fron-
tier”, Acta Phys. Polon. B 53 (2022) 12,
arXiv:2203.13923 [Semantic Scholar].

[35] A. Buckley, J. Ferrando, S. Lloyd, K. Nord-

ström, B. Page, M. Rüfenacht et al.,
“LHAPDF6: parton density access in the
LHC precision era”, Eur. Phys. J. C 75 ,
arXiv:1412.7420 [Semantic Scholar].

[36] G. Cowan, K. Cranmer, E. Gross and

O. Vitells, “Asymptotic formulae for
likelihood-based tests of new physics”, Eur.
Phys. J. C 71 (2011) 1554, arXiv:1007.1727
[Erratum: Eur.Phys.J.C 73, 2501 (2013)]
[Semantic Scholar].