Precise predictions for samesign Wboson scattering at the LHC
The European Physical Journal C
August 2018, 78:671  Cite as
Precise predictions for samesign Wboson scattering at the LHC
AuthorsAuthors and affiliations
Alessandro BallestreroBenedikt BiedermannSimon BrassAnsgar DennerStefan DittmaierRikkert FrederixPietro GovoniMichele GrossiBarbara JägerAlexander KarlbergEzio MainaMathieu PellenGiovanni PelliccioliSimon PlätzerMichael RauchDaniela RebuzziJürgen ReuterVincent RotheChristopher SchwanHuaSheng ShaoPascal StienemeierGiulia ZanderighiMarco ZaroDieter Zeppenfeld
Open Access
Regular Article  Theoretical Physics
First Online: 22 August 2018
Received: 04 April 2018
Accepted: 05 August 2018
76 Downloads
Abstract
Vectorboson scattering processes are of great importance for the current runII and future runs of the Large Hadron Collider. The presence of triple and quartic gauge couplings in the process gives access to the gauge sector of the Standard Model (SM) and possible newphysics contributions there. To test any newphysics hypothesis, sound knowledge of the SM contributions is necessary, with a precision which at least matches the experimental uncertainties of existing and forthcoming measurements. In this article we present a detailed study of the vectorboson scattering process with two positivelycharged leptons and missing transverse momentum in the final state. In particular, we first carry out a systematic comparison of the various approximations that are usually performed for this kind of process against the complete calculation, at LO and NLO QCD accuracy. Such a study is performed both in the usual fiducial region used by experimental collaborations and in a more inclusive phase space, where the differences among the various approximations lead to more sizeable effects. Afterwards, we turn to predictions matched to parton showers, at LO and NLO: we show that on the one hand, the inclusion of NLO QCD corrections leads to more stable predictions, but on the other hand the details of the matching and of the partonshower programs cause differences which are considerably larger than those observed at fixed order, even in the experimental fiducial region. We conclude with recommendations for experimental studies of vectorboson scattering processes.
1 Introduction
Vectorboson scattering (VBS) at a hadron collider usually refers to the interaction of massive vector bosons (\(\text {W} ^\pm ,\text {Z} \)), radiated by partons (quarks) of the incoming protons, which in turn are deflected from the beam direction and enter the volume of the particle detectors. As a consequence, the typical signature of VBS events is characterised by two energetic jets and four fermions, originating from the decay of the two vector bosons. Among the possible diagrams, the scattering process can be mediated by a Higgs boson. The interaction of longitudinally polarised bosons is of particular interest, because the corresponding matrix elements feature unitarity cancellations that strongly depend on the actual structure of the Higgs sector of the Standard Model (SM). A detailed study of this class of processes will therefore further constrain the Higgs couplings at a very different energy scale with respect to the Higgsboson mass, and hint at, or exclude, nonStandard Model behaviours.
Open image in new window
Fig. 1
Sample treelevel diagrams that contribute to the process \(\text {p} \text {p} \rightarrow \mu ^+\nu _\mu \text {e} ^+\nu _{\text {e}}\text {j} \text {j} \) at order \(\mathcal {O}{\left( \alpha ^{6}\right) }\). In addition to typical VBS contributions (left), this order also possesses schannel contributions such as decay chain (middle) and triboson contributions (right)
The VBS process involving two samesign \(\text {W} \) bosons has the largest signaltobackground ratio of all the VBS processes at the Large Hadron Collider (LHC): evidence for it was found at the centreofmass energy of \(8~\,\text {TeV} \) [1, 2, 3], and it has been recently measured at \(13~\,\text {TeV} \) as well [4]. Presently, the measurements of VBS processes are limited by statistics, but the situation will change in the near future. On the theoretical side, it is thus of prime importance to provide predictions with systematic uncertainties at least comparable to the current and envisaged experimental precision [5, 6].
\(\text {W} ^+\text {W} ^+\) scattering is the simplest VBS process to calculate, because the doublecharge structure of the leptonic final state limits the number of partonic processes and total number of Feynman diagrams for each process. Nonetheless, it possesses all features of VBS at the LHC and is thus representative of other VBS signatures. Therefore, it is the ideal candidate for a comparative study of the different simulation tools.
In the last few years, several nexttoleadingorder (NLO) computations have become available for both the VBS process [7, 8, 9, 10, 11, 12, 13] and its QCDinduced irreducible background process [13, 14, 15, 16, 17]. All these VBS computations rely on various approximations, typically neglecting contributions which are expected to be small in realistic experimental setups [12, 18]. Recently, the complete NLO corrections to \(\text {W} ^+\text {W} ^+\) have been evaluated in Ref. [19], making it possible for the first time to study in detail the quality of the VBS approximations at NLO QCD.1
This article starts with the definition of the VBS process in Sect. 2, before describing the approximations of the various computer codes in Sect. 3. In Sect. 4 a leadingorder (LO) study of the different contributions which lead to the production of two samesign \(\text {W} \) bosons and two jets is performed. In the same section predictions for VBS from different tools are compared at the level of the cross section and differential distributions. The comparison is extended to the NLO corrections to VBS in Sect. 5. The effect of the inclusion of matching LO and NLO computations to parton shower (PS) is discussed in Sect. 6. Finally, Sect. 7 contains a summary of the article and concluding remarks.
2 Definition of the process
The scattering of two positivelycharged \(\text {W} \) bosons with their subsequent decay into differentflavour leptons can proceed at the LHC through the partonic process:
$$\begin{aligned} \text {p} \text {p} \rightarrow \mu ^+\nu _\mu \mathrm{e}^+\nu _\mathrm{e}\,\text {j} \text {j} +\mathrm {X}. \end{aligned}$$
(1)
At LO, this process can proceed via three different couplingorder combinations: \(\mathcal {O}{\left( \alpha ^{6}\right) }\), \(\mathcal {O}{\left( \alpha _\text {s} ^{2}\alpha ^{4}\right) }\), and \(\mathcal {O}{\left( \alpha _\text {s} \alpha ^{5}\right) }\). The first, commonly referred to as EW contribution or VBS,2 receives contributions from Feynman diagrams such as those depicted in Fig. 1: in addition to genuine VBS contributions (left diagram), it also features schannel contributions with nonresonant vector bosons (centre diagram) or from tripleboson production (right diagram). Note that s, t, and uchannel contributions are defined according to the quark lines. The schannel denotes all Feynman diagrams where the two initialstate partons are connected by a continuous fermion line, while for the t and uchannel the fermion lines connect initial state quarks to final state quarks. The uchannel refers to contributions with crossed fermion lines with respect to tchannel, which appears for identical (anti) quarks in the final state. The schannel contributions play a particular role in the study of the various contributions in Sect. 4.1. When using approximations, care must be taken that only gaugeinvariant subsets are considered to obtain physically meaningful results. We discuss the commonlyused possible choices in detail in the next section.
The second coupling combination of order \(\mathcal {O}{\left( \alpha _\mathrm{s}^2\alpha ^{4}\right) }\) corresponds to diagrams with a gluon connecting the two quark lines, and with the \(\text {W} \) bosons radiated off the quark lines. Because of the different colour structure, this contribution features a different kinematic behaviour than VBS. Nonetheless it shares the same final state, and therefore constitutes an irreducible background to the EW process.
Finally, the third contribution of order \(\mathcal {O}{\left( \alpha _\mathrm{s}\alpha ^{5}\right) }\) is the interference of the two types of amplitudes described above. It is nonzero only for those partonic subprocesses which involve identical quarks or antiquarks. Such a contribution is usually small (\(3\%\)) within typical experimental cuts [19].
In the rest of this article, the notations LO or NLO(QCD) without any specification of coupling powers refer to the contributions at order \(\mathcal {O}{\left( \alpha ^{6}\right) }\) and \(\mathcal {O}{\left( \alpha _\mathrm{s}\alpha ^{6}\right) }\), respectively.
In experimental measurements, special cuts, called VBS cuts, are designed to enhance the EW contribution over the QCD one and to suppress the interference. These cuts are based on the different kinematical behaviour of the contributions. The EW contribution is characterised by two jets with large rapidities as well as a large dijet invariant mass. The two \(\text {W} \) bosons are mostly produced centrally. This is in contrast to the QCD contribution which favours jets in the central region. Therefore, the event selection usually involves rapiditydifference and invariantmass cuts for the jets. Note that, as pointed out in Ref. [19], when considering full amplitudes, the separation between EW and QCD production becomes illdefined. Hence, combined measurements which are theoretically better defined should be preferably performed by the experimental collaborations at the LHC.
3 Details of the calculations
3.1 Theoretical predictions for VBS
We now discuss the various approximations which are implemented in computer programs for the EW contribution at order \(\mathcal {O}{\left( \alpha ^{6}\right) }\). Since we are mostly interested in the scattering of two \(\text {W} \) bosons, which includes the quartic gaugeboson vertex, it may appear justified to approximate the full process by considering just those diagrams which contain the \(2\rightarrow 2\) scattering process as a subpart. However, this set of contributions is not gauge invariant. In order to ensure gauge invariance, an onshell projection of the incoming and outgoing W bosons should be performed. While this can be done in the usual way for the timelike outgoing W bosons, the treatment of the spacelike W bosons emitted from the incoming quarks requires some care. Following Refs. [22, 23] these Wboson lines can be split, the W bosons entering the scattering process can be projected onshell, and the emission of the W bosons from the quarks can be described by vectorboson luminosities. Such an approximation is usually called effective vectorboson approximation (EVBA) [24, 25, 26].
An improvement of such an approximation consists in considering all t and uchannel diagrams and squaring them separately, neglecting interference contributions between the two classes. These interferences are expected to be small in the VBS fiducial region, as they are both phasespace and colour suppressed [12, 18]. The schannel squared diagrams and any interferences between them and the t/uchannels are also discarded. This approximation is often called t/u approximation, VBF, or even VBS approximation. We adopt the latter denomination in the following. This approximation is gaugeinvariant, a fact that can be appreciated by considering the two incoming quarks as belonging to two different copies of the \(\mathrm {SU}(3)\) gauge group.
A further refinement is to add the squared matrix element of the schannel contributions to the VBS approximation.
The approximations performed at LO can be extended when NLO QCD corrections to the order \(\mathcal {O}{\left( \alpha ^{6}\right) }\) are computed. The VBS approximation can be extended at NLO in a straightforward manner for what concerns the virtual contributions. For the realemission contributions special care must be taken for the gluoninitiated processes. The initialstate gluon and initialstate quark must not couple together, otherwise infrared (IR) divergences proportional to schannels appear, which do not match with the ones found in the virtual contributions. The subset of diagrams where all couplings of the initial state gluon to initial state quark are neglected forms a gaugeinvariant subset, with the same argument presented above.
A further refinement is to consider the full real contributions, which include all interferences, and part of the virtual. In particular one can consider only oneloop amplitudes where there is no gluon exchange between the two quark lines and assuming a cancellation of the IR poles.
When considering the full NLO corrections of order \(\mathcal {O}{\left( \alpha _\mathrm{s}\alpha ^{6}\right) }\), besides real and virtual QCD corrections to the EW treelevel contribution of order \(\mathcal {O}{\left( \alpha ^{6}\right) }\) also real and virtual EW corrections to the LO interference of order \(\mathcal {O}{\left( \alpha _\mathrm{s}\alpha ^{5}\right) }\) have to be taken into account. Since some loop diagrams contribute to both types of corrections, QCD and EW corrections cannot be separated any more on the basis of Feynman diagrams, and the cancellation of IR singularities requires the inclusion of all of them [19].
3.2 Description of the programs used
In the following, the codes employed throughout this article and the approximations implemented in each of them are discussed:
Phantom [27] is a dedicated treelevel Monte Carlo for sixparton final states at \(\text {p} \text {p},\, \text {p} \bar{\text {p}}\), and \(\text {e} ^+\text {e} ^\) colliders at orders \(\mathcal O(\alpha ^6)\) and \(\mathcal O(\alpha _\text {s} ^2\alpha ^4)\) including interferences between the two sets of diagrams. It employs complete treelevel matrix elements in the complexmass scheme [28, 29, 30] computed via the modular helicity formalism [31, 32]. The integration uses a multichannel approach [33] and an adaptive strategy [34]. Phantom generates unweighted events at parton level for both the SM and a few instances of beyond the Standard Model (BSM) theories.
WHIZARD [35, 36] is a multipurpose event generator with LO matrixelement generator O’Mega. For QCD amplitudes it uses the colour flow formalism [37]. For NLO QCD calculations [38], where Whizard is in the final validation phase, it provides FKS subtraction terms [39, 40], while virtual matrix elements are provided externally by OpenLoops [41] or Recola [42, 43]. Furthermore, WHIZARD can automatically provide POWHEG matching to parton shower [44]. WHIZARD allows to simulate a huge number of BSM models as well, in particular in terms of higherdimensional operators for VBS processes including means to provide unitarity limits [45, 46].
The program Bonsay [47] consists of a generalpurpose Monte Carlo integrator and matrix elements taken from different sources: Born matrix elements are adapted from the program Lusifer [48], which have been generalised to calculate also real matrix elements. Virtual matrix elements have been calculated using an inhouse matrixelement generator. Oneloop integrals are evaluated using the Collier library [49, 50]. For the results presented here, it uses the VBS approximation at LO and NLO. The virtual corrections are additionally approximated using a doublepole approximation where the final state leptons are assumed to originate from the decay of two resonant Wbosons; see Ref. [51] for a detailed description and Ref. [52] for the onshell projection used. At LO the exact matrix elements can also be used.
The PowhegBox [53, 54, 55] is a framework for matching NLOQCD calculations with parton showers. It relies on the user providing the matrix elements and Born phase space, but automatically constructs FKS [39] subtraction terms and the phase space corresponding to the real emission. For the VBS processes all matrix elements are being provided by a previous version of VBFNLO [17, 56, 57] and hence the approximations used in the PowhegBox are similar to those used in VBFNLO.
VBFNLO [17, 56, 57] is a flexible partonlevel Monte Carlo for processes with EW bosons. It allows the calculation of VBS processes at NLO QCD in the VBS approximation, with process IDs between 200 and 290. Samesign \(\text {W} ^+\text {W} ^+jj\) production is provided via the process ID 250. The corresponding schannel contributions are available separately as triboson processes with semileptonic decays, with process IDs between 401 and 492. For the final state studied in this article, only \(\text {W} ^+\text {W} ^+\text {W} ^\) production with a hadronically decaying \(\text {W} ^\) boson, process ID 432, can contribute. These can simply be added on top of the VBS contribution. Interferences between the two are therefore neglected. The usage of leptonic tensors in the calculation, pioneered in Ref. [7], thereby leads to a significant speed improvement over automatically generated code. Besides the SM, also a variety of newphysics models including anomalous couplings of the Higgs and gauge bosons can be simulated.
MadGraph5_aMC@NLO [58] (henceforth MG5_a MC) is an automatic metacode (a code that generates codes) which makes it possible to simulate any scattering process including NLO QCD corrections both at fixed order and including matching to parton showers, using the MC@NLO method [59]. It makes use of the subtraction method by Frixione, Kunszt and Signer (FKS) [39, 40] (automated in the module MadFKS [60, 61]) for regulating IR singularities. The computations of oneloop amplitudes are carried out by switching dynamically between two integralreduction techniques, OPP [62] or Laurentseries expansion [63], and tensorintegral reduction [64, 65, 66]. These have been automated in the module MadLoop [67], which in turn exploits CutTools [68], Ninja [69, 70], IREGI [71], or Collier [50], together with an inhouse implementation of the OpenLoops optimisation [41]. Finally, scale and PDF uncertainties can be obtained in an exact manner via reweighting at negligible additional CPU cost [72]. The simulation of VBS at NLOQCD accuracy can be performed by issuing the following commands in the program interface:
Open image in new window
Table 1
Summary of the different properties of the computer programs employed in the comparison
Code
\(\mathcal O(\alpha ^6)\) s, t, u
\(\mathcal O(\alpha ^6)\) interf.
Nonres.
NLO
NF QCD
EW corr. to order \(\mathcal O(\alpha _\text {s} \alpha ^5)\)
Phantom
s, t, u
Yes
Yes
No
–
–
Whizard
s, t, u
Yes
Yes
No
–
–
Bonsay
t, u
No
Yes, virt. no
Yes
No
No
Powheg
t, u
No
Yes
Yes
No
No
VBFNLO
s, t, u
No
Yes
Yes
No
No
MG5_aMC
s, t, u
Yes
Yes
Yes
Virt. no
No
MoCaNLO+Recola
s, t, u
Yes
Yes
Yes
Yes
Yes
With these commands the complexmass scheme is turned on, then the NLOcapable model is loaded,3 finally the process code is generated (note the QCD=0 syntax to select the purelyEW process) and written to disk. No approximation is performed for the Born and realemission matrix elements. Only stronglyinteracting particles circulating in the loops are generated for the virtual matrix element. The version capable of computing both QCD and EW corrections will be released in the future. Such an approximation is equivalent to the assumption that the finite part of those loops which feature EW bosons is zero. In practice, since a part of the contribution to the single pole is also missing, the internal polecancellation check at run time has to be turned off, by setting the value of the IRPoleCheckThreshold and PrecisionVirtualAtRunTime parameters in the Cards/FKS_params.dat file to \(1\).
The program MoCaNLO+Recola is made of a flexible Monte Carlo program dubbed MoCaNLO and of the matrixelement generator Recola [42, 43]. It can compute arbitrary processes at the LHC with both NLO QCD and EW accuracy in the SM. This is made possible by the fact that Recola can compute arbitrary processes at tree and oneloop level in the SM. To that end, it relies on the Collier library [49, 50] to numerically evaluate the oneloop scalar and tensor integrals. In addition, the subtraction of the IR divergences appearing in the real corrections has been automatised thanks to the Catani–Seymour dipole formalism for both QCD and QED [73, 74]. The code has demonstrated its ability to compute NLO corrections for highmultiplicity processes up to \(2 \rightarrow 7\) [75, 76]. In particular the full NLO corrections to VBS and its irreducible background [19, 77] have been obtained thanks to this tool. One key aspect for these highmultiplicity processes is the fast integration which is ensured by using similar phasespace mappings to those of Refs. [28, 48, 78]. In MoCaNLO+Recola no approximation is performed neither at LO nor at NLO. It implies that, also contributions stemming from EW corrections to the interference are computed.
We conclude this section by summarising the characteristics of the various codes in Table 1. In particular, it is specified whether
all s, t, uchannel diagrams are included;
interferences between diagrams of different types are included at LO;
diagrams which do not feature two resonant W bosons are included;
the socalled nonfactorisable (NF) QCD corrections, i.e. the corrections where (real or virtual) gluons are exchanged between different quark lines, are included;
EW corrections to the interference of order \(\mathcal O (\alpha ^5\alpha _\text {s})\) are included. These corrections are of the same order as the NLO QCD corrections to the contribution of order \(\mathcal O (\alpha ^6\)) term.
3.3 Input parameters
The hadronic scattering processes are simulated at the LHC with a centreofmass energy \(\sqrt{s} = 13~\,\text {TeV} \). The NNPDF 3.0 parton distribution functions (PDFs) [79] with five massless flavours,4 NLOQCD evolution, and a strong coupling constant \(\alpha _\text {s} \left( M_\text {Z} \right) = 0.118\)5 are employed.6 Initialstate collinear singularities are factorised according to the \({\overline{\mathrm{MS}}}\) scheme, consistently with what is done in NNPDF.
For the massive particles, the following masses and decay widths are used:
$$\begin{aligned} m_\text {t}&= 173.21\,\text {GeV},&\quad \quad \quad \varGamma _\text {t}&= 0 \,\text {GeV}, \nonumber \\ M_\text {Z} ^\text {OS}&= 91.1876\,\text {GeV},&\quad \quad \quad \varGamma _\text {Z} ^\text {OS}&= 2.4952\,\text {GeV}, \nonumber \\ M_\text {W} ^\text {OS}&= 80.385\,\text {GeV},&\varGamma _\text {W} ^\text {OS}&= 2.085\,\text {GeV}, \nonumber \\ M_\mathrm{H}&= 125.0\,\text {GeV},&\varGamma _\text {H}&= 4.07 \times 10^{3}\,\text {GeV}. \end{aligned}$$
(2)
The measured onshell (OS) values for the masses and widths of the W and Z bosons are converted into pole values for the gauge bosons (\(V=\text {W},\text {Z} \)) according to Ref. [81],
$$\begin{aligned} \begin{aligned} M_V&= M_{V}^\text {OS}/\sqrt{1+(\varGamma _{V}^\text {OS}/M_{V}^\text {OS})^2}, \\ \varGamma _V&= \varGamma _{V}^\text {OS}/\sqrt{1+(\varGamma _{V}^\text {OS}/M_{V}^\text {OS})^2}. \end{aligned} \end{aligned}$$
(3)
The EW coupling is renormalised in the \(G_\mu \) scheme [51] according to
$$\begin{aligned} \alpha = \frac{\sqrt{2}}{\pi } G_{\mu } M_\mathrm{W}^2 \left( 1\frac{M_\mathrm{W}^2}{M_\mathrm{Z}^2} \right) , \end{aligned}$$
(4)
with
$$\begin{aligned} G_{\mu } = 1.16637\times 10^{5}\,\text {GeV} ^{2}, \end{aligned}$$
(5)
and where \(M_V^2\) corresponds to the real part of the squared pole mass. The numerical value of \(\alpha \), corresponding to the choice of input parameters is
$$\begin{aligned} 1/\alpha = 132.3572\ldots . \end{aligned}$$
(6)
The Cabibbo–Kobayashi–Maskawa matrix is assumed to be diagonal, meaning that the mixing between different quark generations is neglected. The complexmass scheme [28, 29, 30] is used throughout to treat unstable intermediate particles in a gaugeinvariant manner.
The central value of the renormalisation and factorisation scales is set to
$$\begin{aligned} \mu _\mathrm{ren} = \mu _\mathrm{fac} = \sqrt{p_\mathrm{T, j_1}\, p_\mathrm{T, j_2}}, \end{aligned}$$
(7)
defined via the transverse momenta of the two hardest jets (identified with the procedure outlined in the following), event by event.7 This choice of scale has been shown to provide stable NLOQCD predictions [12].
Following experimental measurements [1, 2, 3, 82], the event selection used in the present study is:
The two samesign charged leptons are required to fulfil cuts on transverse momentum, rapidity, and separation in the rapidity–azimuthalangle separation,
$$\begin{aligned} p_{\text {T},\ell }> 20\,\text {GeV},\qquad y_{\ell } < 2.5, \qquad \varDelta R_{\ell \ell }> 0.3. \end{aligned}$$
(8)
The total missing transverse momentum, computed from the vectorial sum of the transverse momenta of the two neutrinos, is required to be
$$\begin{aligned} p_\mathrm{T, miss} > 40\,\text {GeV}. \end{aligned}$$
(9)
QCD partons (light quarks and gluons) are clustered together using the anti\(k_T\) algorithm [83], possibly using the FastJet implementation [84], with distance parameter \(R=0.4\). We impose cuts on the jets’ transverse momenta, rapidities, and their separation from leptons,
$$\begin{aligned} p_{\text {T},\text {j}}> 30\,\text {GeV}, \qquad y_\text {j}  < 4.5, \qquad \varDelta R_{\text {j} \ell } > 0.3. \end{aligned}$$
(10)
VBS cuts are applied on the two jets with largest transverse momentum, unless otherwise stated. In particular, we impose a cut on the invariant mass of the dijet system, as well as on the rapidity separation of the two jets,
$$\begin{aligned} m_{\text {j} \text {j}}> 500\,\text {GeV},\qquad \varDelta y_{\text {j} \text {j}} > 2.5, \end{aligned}$$
(11)
if not explicitly stated otherwise.
When EW corrections are computed, real photons and charged fermions are clustered together using the anti\(k_T\) algorithm with radius parameter \(R=0.1\). In this case, leptons and quarks are understood as dressed fermions.
4 Leadingorder study
4.1 Contributions
In the present section, the cross sections and distributions are obtained without applying the VBS cuts on the variables \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\), Eq. (11). In Table 2, the cross sections of the three LO contributions are reported. The EW, QCD, and interference contributions amount to \(57\%\), \(37\%\), and \(6\%\) of the total inclusive cross section, respectively. The QCD contribution does not possess external gluons due to charge conservation. Thus the diagrams of order \(\mathcal {O}(g_\mathrm{s}^2g^4)\) only involve gluon exchange between the quark lines. This results in a small contribution even if the VBS cuts have not been imposed. The interference between EW and QCD contributions is small, due to colour suppression, but not negligible.
Table 2
Cross sections at LO accuracy for the three contributions to the process \(\mathrm{p}\mathrm{p}\rightarrow \mu ^+\nu _\mu \mathrm{e}^+\nu _\mathrm{e}\mathrm{j}\mathrm{j}\), obtained with exact matrix elements. These results are for the setup described in Sect. 3.3 but no cuts on \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\) are applied. The uncertainties shown refer to the estimated statistical errors of the Monte Carlo integrations
Order
\(\mathcal {O}(\alpha ^6)\)
\(\mathcal {O}(\alpha _\text {s} ^2\alpha ^4)\)
\(\mathcal {O}(\alpha _\text {s} \alpha ^5)\)
\(\sigma \;(\mathrm {fb})\)
\( 2.292 \pm 0.002 \)
\( 1.477 \pm 0.001 \)
\( 0.223 \pm 0.003 \)
Open image in new window
Fig. 2
Differential distribution in the dijet invariant mass \(m_{\text {j} \text {j}}\) (left) and the difference of the jet rapidities \(\varDelta y_{\text {j} \text {j}}\) (right) for the three LO contributions to the process \(\mathrm{p}\mathrm{p}\rightarrow \mu ^+\nu _\mu \mathrm{e}^+\nu _\mathrm{e}\mathrm{j}\mathrm{j}\). The EW contribution is in red, the QCD one in orange, and the interference one in grey. The sum of all the contributions is in blue. The cuts applied are the ones of Sect. 3.3 but no cuts on \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\) are applied
In Fig. 2, these three contributions are shown separately and summed in the differential distributions in the dijet invariant mass \(m_{\text {j} \text {j}}\) and the rapidity difference \(\varDelta y_{\text {j} \text {j}}\). For the dijet invariantmass distribution (left), one can observe that the EW contribution peaks around an invariant mass of about \(80\,\text {GeV} \). This is due to diagrams where the two jets originate from the decay of a W boson (see middle and right diagrams in Fig. 1). Note that these contributions are not present in calculations relying on the VBS approximation as these are only schannel contributions. The EW contribution becomes dominant for dijet invariant mass larger than \(500 \,\text {GeV} \). The same holds true for jetrapidity difference larger than 2.5 (right). This justifies why cuts on these two observables are used in order to enhance the EW contribution over the QCD one. In particular, in order to have a large EW contribution, rather exclusive cuts are required.
This can also be seen in Fig. 3 where the three contributions are displayed as doubledifferential distributions in the dijet invariant mass and jet rapidity difference. Again, it is clear that the region with low dijet invariant mass should be avoided in VBS studies as it is dominated by triboson contributions. This motivates in particular the choice of the cut \(m_{\text {j} \text {j}} > 200 \,\text {GeV} \) for our LO inclusive study in Sect. 4.2.
Open image in new window
Fig. 3
Doubledifferential distributions in the variables \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\) for the three LO contributions of orders \(\mathcal {O}(\alpha ^6)\) (top left), \(\mathcal {O}(\alpha _\text {s} \alpha ^5)\) (top right), and \(\mathcal {O}(\alpha _\text {s} ^2 \alpha ^4)\) (bottom). The cuts applied are the ones of Sect. 3.3 but no cuts on \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\) are applied
4.2 Inclusive comparison
In Fig. 4, ratios for doubledifferential cross sections in the variables \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\) are shown.8 Two plots are displayed: the ratios of the \(t^2 + u^2\) and \(s^2 + t^2 + u^2\) approximations over the full calculation. In the first case, the approximation is good within \(\pm \,10\%\) over the whole range apart from the low invariantmass region at both low and large rapidity difference. The low rapiditydifference region possesses remnants of the tribosons contribution that have a dijet invariant mass around the \(\text {W} \)boson mass. It is therefore expected that the \(t^2 + u^2\) approximation fails in this region. The second plot, where the \(s^2 + t^2 + u^2\) approximation is considered, displays a better behaviour in the previously mentioned region. The full calculation is approximated at the level of \(\pm \,5\%\) everywhere apart from the region where \(\varDelta y_{\text {j} \text {j}} < 2\).
Open image in new window
Fig. 4
Ratios for doubledifferential distributions in the variables \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\) at LO i.e. order \(\mathcal {O}(\alpha ^6)\) of approximated squared amplitudes over the full matrix element. The approximated squared amplitudes are computed as \(\mathcal {A}^2 \sim t^2 + u^2\) (left) and \(\mathcal {A}^2 \sim s^2 + t^2 + u^2\) (right). The cuts applied are the ones of Sect. 3.3 and no cuts on \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\) are applied
4.3 Comparison in the fiducial region
In Table 3, we report the total rates at LO accuracy at order \(\mathcal O (\alpha ^6)\) obtained in the fiducial region described in Sect. 3.3. Two things should be highlighted here: first, despite the different underlying approximations, the two mostdistant predictions (Powhegbox and MG5_aMC) are only \(0.7\%\) apart. This simply means that the details of the various VBS approximations have an impact below \(1\%\) at the level of the fiducial cross section at LO for a typical phasespace volume used by experimental collaborations. This is in agreement with the findings of Refs. [12, 18]. Second, the four complete predictions (Whizard, Phantom, MG5_aMC, and MoCaNLO+Recola) are not in statistical agreement. While we have checked the pointwise agreement of the matrixelement, we cannot exclude other reasons for the disagreement, for example a nonrepresentative (i.e. tooaggressive) estimate of the Monte Carlo uncertainty or a nonperfect mapping of the sixbody phasespace. However, the level of ambiguity is at the \(0.5\%\) level, which we deem satisfactory compared to the larger differences observed at NLO or when including matching to parton shower.
Table 3
Cross sections at LO accuracy and order \(\mathcal {O}(\alpha ^6)\). The complete \(2\rightarrow 6\) matrixelement, without any approximation, is employed by Phantom, Whizard, MG5_aMC, and MoCaNLO+Recola. The predictions are obtained in the fiducial region described in Sect. 3.3. The uncertainties shown refer to the estimated statistical errors of the Monte Carlo integrations
Code
\(\sigma \;(\mathrm {fb})\)
Bonsay
\(1.43636 \pm 0.00002\)
PowhegBox
\(1.44092 \pm 0.00009\)
VBFNLO
\(1.43796 \pm 0.00005\)
Phantom
\(1.4374 \pm 0.0006 \)
Whizard
\(1.4381 \pm 0.0002 \)
MG5_aMC
\(1.4304 \pm 0.0007\)
MoCaNLO+Recola
\(1.43476 \pm 0.00009\)
Open image in new window
Fig. 5
Differential distributions in the invariant mass (left) and rapidity difference of the two tagging jets (right) at LO accuracy i.e. at order \(\mathcal {O}(\alpha ^6)\). The description of the different programs used can be found in Sect. 3.2. The upper plots provides the absolute value for each prediction while the lower plots presents all predictions normalised to MoCaNLO+Recola which is one of the programs that provide the full prediction. The band corresponds to a sevenpoint variation of the renormalisation and factorisation scales. The predictions are obtained in the fiducial region described in Sect. 3.3
In Fig. 5, we show the distributions in the invariant mass (left) and the rapidity difference (right) of the two tagging jets which are key observables for VBS measurements. In both cases we show the absolute distributions in the upper plot, while the lower plot displays the ratio over the predictions of MoCaNLO+Recola, for which we also display the scaleuncertainty band (sevenpoints variation as in Eq. (3.11) of Ref. [19]). For both observables we find a relatively good agreement among the various tools, which confirms the fact that contributions from schannel diagrams as well as interferences are suppressed in the fiducial region. In general, the agreement is at the level of \(1\%\) or below in each bin. We have checked that the same level of agreement holds for other standard differential distributions such as rapidity, invariant mass, or transverse momentum. This means that at LO, in the fiducial volume and for energies relevant to the LHC, the VBS approximation is good to a per cent. This is in agreement with the findings of Sect. 4.2 as the present comparison completely excludes the phasespace region where triboson contributions could have a noticeable impact.
5 Nexttoleading order QCD
5.1 Inclusive comparison
According to the results of Sects. 4.1 and 4.2, the VBS approximation at LO fails drastically in the region \(m_{\text {j} \text {j}} < 200\) GeV, \(\varDelta y_{\text {j} \text {j}} < 2\). Therefore, we present an inclusive study at NLO QCD for the EW component, namely the order \(\mathcal {O}(\alpha _\text {s} \alpha ^6)\) for the setup described in Sect. 3.3 but imposing the requirements \(m_{\text {j} \text {j}}>200 \,\text {GeV} \) and \(\varDelta y_{\text {j} \text {j}}>2\).
Table 4
Cross sections at NLO QCD i.e. at order \(\mathcal {O}(\alpha _\text {s} \alpha ^6)\) for the full computation and two approximations. In addition to the cuts of Sect. 3.3, the VBS cuts take the values \(m_{\text {j} \text {j}}>200 \,\text {GeV} \) and \(\varDelta y_{\text {j} \text {j}}>2\). The uncertainties shown refer to the estimated statistical errors of the Monte Carlo programs
Prediction
\(\sigma _{\text {tot}}\,(\text {fb})\)
\(\delta \;(\%)\)
full
\(1.733\phantom {0} \pm 0.002\phantom {0}\)

\(t^2 + u^2\)
\(1.6292 \pm 0.0001\)
\(6.0\)
\(s^2 + t^2 + u^2\)
\(1.7780 \pm 0.0001\)
\(+2.6\)
We compare three different predictions at NLO QCD: the VBS approximation implemented in Bonsay (dubbed \(t^2+u^2\)), the VBS approximation with the schannel contributions from VBFNLO (dubbed \(s^2+t^2+u^2\)), and the full computation. The full computation employs exact matrix elements meaning that t/u/s interferences, factorisable and nonfactorisable QCD corrections, as well as EW corrections to the order \(\mathcal {O}(\alpha _\text {s} \alpha ^5)\) are included.
The total cross sections within the abovementioned kinematic cuts are shown in Table 4. The \(t^2+u^2\) approximation for NLO QCD predictions is lower by about \(6\%\) than the full calculation. The inclusion of schannel diagrams improves the approximate prediction, leading to an excess at the \(3\%\) level.
These differences are more evident in differential distributions. In Fig. 6, we show the differential distributions in the dijet invariant mass \(m_{\text {j} \text {j}}\) and rapidity separation \(\varDelta y_{\text {j} \text {j}}\). For large \(m_{\text {j} \text {j}}\) and large \(\varDelta y_{\text {j} \text {j}}\), as expected, the VBS approximation is performing well and its schannel extension agrees with the full calculation within \(10\%\). This is in contrast with the regions \(200 \,\text {GeV}< m_{\text {j} \text {j}} < 500 \,\text {GeV} \) and \(2<\varDelta y_{\text {j} \text {j}}<2.5\), where the difference between the \(t^2+u^2\) approximation and the full computation can be above \(30\%\). The inclusion of schannel contributions cures partly this behaviour by improving the approximation to about \(10\%\). This tends to indicate that interference contributions and/or nonfactorisable QCD corrections play a nonnegligible role in this phasespace region.
Open image in new window
Fig. 6
Differential distributions in the dijet invariant mass (left) and the rapidity separation (right) of the two tagging jets at NLO QCD i.e. at order \(\mathcal {O}(\alpha _\text {s} \alpha ^6)\) for the full computation and two approximations. The upper plots provide the absolute value for each prediction while the lower plots present all predictions normalised to MoCaNLO+Recola which is one of the programs that provide the full prediction. In addition to the cuts of Sect. 3.3, the VBS cuts take the values \(m_{\text {j} \text {j}}>200 \,\text {GeV} \) and \(\varDelta y_{\text {j} \text {j}}>2\)
In order to investigate further the jetpair kinematics, we study the doubledifferential distribution in the variables \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\). In particular, in Fig. 7, we compute in each bin the ratios of the approximated cross sections over the full ones [\(\sigma (t^2+u^2)/\sigma (\text {full})\) and \(\sigma (s^2+t^2+u^2)/\sigma (\text {full})\)]. As expected, in the low invariantmass and low rapidityseparation region of the jet pair (\(200 \,\text {GeV}< m_{\text {j} \text {j}} < 500 \,\text {GeV} \), \(2<\varDelta y_{\text {j} \text {j}}<2.5\)) the VBS approximation fails significantly (by more than \(40\%\)). Including the schannel contributions leads to a difference of less than \(10\%\) in this very region. However, in the region of large dijet invariant mass and low rapidity separation of the jets, the \(s^2+t^2+u^2\) approximation overestimates the full computation by more than \(40\%\).9 Again, this seems to support the fact that interferences and nonfactorisable corrections can be nonnegligible in this region. On the other hand, in the typical VBS region, the VBS approximation shows a good agreement with the full computation as documented in detail in Sect. 5.2.
Open image in new window
Fig. 7
Ratios for doubledifferential distributions in the variables \(m_{\text {j} \text {j}}\) and \(\varDelta y_{\text {j} \text {j}}\) at NLO QCD i.e. at order \(\mathcal {O}(\alpha _\text {s} \alpha ^6)\) of the approximated squared amplitudes over the full matrix element. The approximated squared amplitudes are computed as \(\mathcal {A}^2 \sim t^2+u^2\) (left) and \(\mathcal {A}^2 \sim s^2+t^2+u^2\) (right). In addition to the cuts of Sect. 3.3, the VBS cuts take the values \(m_{\text {j} \text {j}}>200 \,\text {GeV} \) and \(\varDelta y_{\text {j} \text {j}}>2\)
In Fig. 8, the distributions in the transverse momentum of the hardest jet and its rapidity are shown. At low transverse momentum, \(t^2+u^2\) and \(s^2+t^2+u^2\) approximations are lower and higher than the full computation by about \(20\%\), respectively. At high transverse momentum, they have a similar behaviour. They both diverge from the full computation towards larger transverse momentum (about \(10\%\) at \(1000\,\text {GeV} \)). Regarding the rapidity of the hardest jet, the two approximations have opposite behaviours. In the central region, the \(t^2+u^2\) approximation differs by \(12\%\) with respect to the full computation, while the \(s^2+t^2+u^2\) one is good within \(5\%\). In the peripheral region, the \(t^2+u^2\) approximation is rather close to the full computation (\(5\%\)), while the \(s^2+t^2+u^2\) one differs by \(10\%\).
Open image in new window
Fig. 8
Differential distributions in the transverse momentum (left) and rapidity (right) of the hardest tagging jet at NLO QCD i.e. at order \(\mathcal {O}(\alpha _\text {s} \alpha ^6)\) for the full computation and two approximations. The upper plots provide the absolute value for each prediction while the lower plots present all predictions normalised to MoCaNLO+Recola which is one of the programs that provide the full prediction. In addition to the cuts of Sect. 3.3, the VBS cuts take the values \(m_{\text {j} \text {j}}>200 \,\text {GeV} \) and \(\varDelta y_{\text {j} \text {j}}>2\)
Concerning leptonic observables, we show in Fig. 9 the distributions in the dilepton invariant mass and in the Zeppenfeld variable of the electron, defined as
$$\begin{aligned} z_{\text {e} ^+} = \frac{y_{\text {e} ^+}\frac{y_{\text {j} _1}+y_{\text {j} _2}}{2}}{\varDelta y_{jj}} . \end{aligned}$$
(12)
Analogous definitions are later also used for the Zeppenfeld variable of the muon and of the third jet. The \(s^2+t^2+u^2\) predictions for \(m_\mathrm{e^+\mu ^+}\) agree rather well with the full curve, obtained from MoCaNLO+Recola. The prediction from Bonsay is about \(10\%\) lower around \(1000 \,\text {GeV} \). The Zeppenfeld variable of the positron \(z_e\) is more strongly affected by the exclusion of schannel contributions. For increasing \(z_\mathrm{e}\), the \(t^2+u^2\) approximation diverges from the full computation to reach a difference of about \(25\%\) at 1.5. On the other hand, including schannel contributions leads to a better approximation, staying within \(10\%\) difference over the whole range.
Open image in new window
Fig. 9
Differential distributions in the leptonlepton invariant mass (left) and the electron Zeppenfeld variable (right) at NLO QCD i.e. at order \(\mathcal {O}(\alpha _\text {s} \alpha ^6)\) for the full computation and two approximations. The upper plots provide the absolute value for each prediction while the lower plots presents all predictions normalised to MoCaNLO+Recola which is one of the programs that provide the full prediction. In addition to the cuts of Sect. 3.3, the VBS cuts take the values \(m_{\text {j} \text {j}}>200 \,\text {GeV} \) and \(\varDelta y_{\text {j} \text {j}}>2\)
In conclusion, both the loose minimum dijet invariantmass cut and the inclusion of QCD radiative corrections render the schannel contributions less suppressed than at LO, making their inclusion mandatory, in order to provide trustworthy predictions at NLO accuracy. In the inclusive region studied here, neglecting schannel contributions, nonfactorisable corrections, and EW corrections can lead to discrepancies of up to \(30\%\) with respect to the full computation. Nevertheless, the VBS approximation at NLO provides a good approximation of full calculations in the kinematic region where \(m_{\text {j} \text {j}} \gtrsim 500 \,\text {GeV} \) and \(\varDelta y_{\text {j} \text {j}} \gtrsim 2.5\)), for both total cross section and differential distributions. This more exclusive region is studied in more detail in the next section.
5.2 Comparison in the fiducial region
In Table 5, the cross sections of the various tools at NLOQCD accuracy are presented. The order considered is again the order \(\mathcal {O}(\alpha _\mathrm{s}\alpha ^6)\), and the fiducial volume is the one described in Sect. 3.3. In contrast with Table 3, the NLO predictions differ visibly according to the approximations used.
Table 5
Cross sections at NLO accuracy and order \(\mathcal {O}(\alpha _\text {s} \alpha ^6)\). The predictions are obtained in the fiducial region described in Sect. 3.3. The uncertainties shown refer to estimated statistical errors of the Monte Carlo integrations
Code
\(\sigma \;(\mathrm {fb})\)
Bonsay
\(1.35039 \pm 0.00006\)
PowhegBox
\(1.3605 \pm 0.0007\)
VBFNLO
\(1.3916 \pm 0.0001\)
MG5_aMC
\(1.363 \pm 0.004\)
MoCaNLO+Recola
\( 1.378 \pm 0.001\)
The first observation is that the predictions featuring two versions of the VBS approximation (Bonsay and the PowhegBox) are relatively close.10 Bonsay uses a doublepole approximation for the virtual matrix element, and it is worth noticing that this approximation seems to be accurate at \(1\%\) level as compared to the PowhegBox. This means that the doublepole approximation on the two W bosons used in Bonsay constitutes a good approximation of the VBSapproximated virtual corrections implemented in the PowhegBox. Both predictions differ by about \(2\%\) with respect to the full computation (MoCaNLO+Recola). The second observation is that the inclusion of schannel contributions seems to have a significant impact. Indeed, their inclusion (as done in VBFNLO) approximates the full computation by a per cent. The main contribution due to the schannel diagrams thereby consists of realemission contributions, where one of the two leading jets is formed by one quark, or possibly also both quarks, originating from the Wboson decay, and the second one by the extra radiation emitted from the initial state. In such configurations, the hadronicallydecaying W boson can become onshell and hence yield larger contributions than at LO, where the invariant mass cut on the two jets forces the boson into the far offshell region. However, the agreement between MoCaNLO+Recola and VBFNLO is mostly accidental, as the inclusion of interference effects and some nonfactorisable corrections (in the real corrections) in MG5_aMC brings the prediction down and closer to the VBS approximation. Not unexpectedly none of the approximations used here agrees perfectly with the full calculation of MoCaNLO+Recola at NLO. Nevertheless, the disagreement seems never to exceed \(2\%\) at the fiducial crosssection level.
In Figs. 10, 11 and 12, several differential distributions are shown. All these predictions are performed at NLO accuracy at the order \(\mathcal {O}(\alpha _\text {s} \alpha ^6)\). In the upper panel, the absolute predictions are shown while in the lower panel, the ratio with respect to the full predictions are displayed. The band corresponds to a sevenpoints variation of the factorisation and renormalisation scales (as defined in Eq. (3.11) of Ref. [19]).
We start with Fig. 10 which displays the invariant mass (left) and the rapidity separation (right) of the two tagging jets. For high invariant mass, all predictions agree rather well. On the other hand, for low invariant mass, the hierarchy present at the level of the cross section is reproduced. The VBSapproximated predictions (Bonsay and PowhegBox) are lower than the full calculation (MoCaNLO+Recola). The full calculation is rather well approximated by the hybrid VBS approximation implemented in MG5_aMC. Finally, VBFNLO which includes also schannel contributions provides larger predictions at low invariant mass. For the rapidity difference between the two tagging jets, the hierarchy between the predictions is rather similar. Therefore, depending on the approximation used, it can vary by \(\pm \,7\%\) and \(\pm \,4\%\) with respect to the full computation at low invariant mass and low rapidity difference for the tagging jets, respectively
Open image in new window
Fig. 10
Differential distributions in the invariant mass (left) and rapidity difference (right) of the two tagging jets at NLO accuracy i.e. at order \(\mathcal {O}(\alpha _\mathrm{s}\alpha ^6)\). The description of the different programs used can be found in Sect. 3.2. The upper plots provide the absolute value for each prediction while the lower plots present all predictions normalised to MoCaNLO+Recola which is the full prediction. The band corresponds to a sevenpoint variation of the renormalisation and factorisation scales. The predictions are obtained in the fiducial region described in Sect. 3.3
Concerning the transverse momentum (left) and rapidity (right) of the hardest jet shown in Fig. 11, the situation is rather different. While MG5_aMC is very close to the full prediction for low transverse momentum, it departs from it at larger transverse momentum by about \(10\%\). This is in contrast with the VBSapproximated predictions such as Bonsay, Powheg, and VBFNLO which are lower than the full computation at low transverse momentum and higher for larger transverse momentum. The difference at high transverse momentum between the latter predictions and the full computation can be attributed to EW Sudakov logarithms that become large in this phasespace region. While the predictions of Bonsay and Powheg are rather close over the whole range, the one of VBFNLO is very different at low transverse momentum where it is even higher than the full computation. We note that for the transverse momentum of the second hardest jet, the predictions from MG5_aMC are in good agreement with the other VBSapproximated predictions. Concerning the rapidity of the hardest jet, VBFNLO is in good agreement with MoCaNLO+Recola in the rapidity range \(y_{j_1} < 3\). For larger rapidity, the other codes constitute a better description of the full process at order \(\mathcal {O}(\alpha _\mathrm{s}\alpha ^6)\).
Open image in new window
Fig. 11
Differential distributions in the transverse momentum (left) and rapidity (right) of the hardest jet at NLO accuracy i.e. at order \(\mathcal {O}(\alpha _\mathrm{s}\alpha ^6)\). The description of the different programs used can be found in Sect. 3.2. The upper plots provide the absolute value for each prediction while the lower plots present all predictions normalised to MoCaNLO+Recola which is the full prediction. The band corresponds to a sevenpoint variation of the renormalisation and factorisation scales. The predictions are obtained in the fiducial region described in Sect. 3.3
The last set of differential distributions is the invariant mass of the two charged leptons (left) and the Zeppenfeld variable for the antimuon (right). Concerning the comparison of the predictions, both distributions display a rather similar behaviour. Indeed, the hierarchy mentioned previously is here respected and enhanced towards high invariant mass or high Zeppenfeld variable. The predictions of MoCaNLO+Recola and VBFNLO are in rather good agreement for both distributions for the kinematic range displayed here. The other three VBS approximations are close to each other within few per cent.
Open image in new window
Fig. 12
Differential distributions in the invariant mass of the two charged leptons (left) and Zeppenfeld variable for the muon (right) at NLO accuracy i.e. at order \(\mathcal {O}(\alpha _\mathrm{s}\alpha ^6)\). The description of the different programs used can be found in Sect. 3.2. The upper plots provide the absolute value for each prediction while the lower plots present all predictions normalised to MoCaNLO+Recola which is the full prediction. The band corresponds to a sevenpoint variation of the renormalisation and factorisation scales. The predictions are obtained in the fiducial region described in Sect. 3.3
In the end, the quality of the VBS approximations is good up to \(10\%\) in the fiducial region. These differences are larger than those at LO.
The contributions from the schannel amplitude can be sizeable especially at low invariant mass for the two tagging jets (comparing the predictions of VBFNLO against the ones of Bonsay and Powheg). This can be explained by the fact that schannel contributions are less suppressed at NLO. As real radiation, an extra gluonjet can be radiated from any of the stronglyinteracting particles while the two quarks originating from the Wboson decay can be recombined in a single jet. Therefore, the jet requirements (\( m_{\text {j} \text {j}} > 500\,\text {GeV} \) and \(\varDelta y_{\text {j} \text {j}} > 2.5\)) that were suppressing schannel contributions at LO are partially lifted with the inclusion of a third jet at NLO. Such an effect has also been observed for top–antitop production in the lepton+jet channel at NLO QCD [85].
In phasespace regions where the schannel contributions are sizeable their interference with the t/uchannel can be of similar size. This can be observed by comparing the predictions of VBFNLO against the ones of MG5_aMC.
Finally, the effect of EW corrections and nonfactorisable contributions in the virtual corrections are usually small. But they can be relatively large (about \(10\%\)) for large transverse momentum of the hardest jet. These highenergy region of the phase space are where EW Sudakov logarithms become large. Nonetheless these regions are rather suppressed and thus these effects are hardly visible at the level of the cross section.
6 Matching to parton shower
We now discuss how different predictions compare when the matching to parton shower is included, both at LO (i.e. at order \(\mathcal O (\alpha ^6)\)) and at NLOQCD (i.e. at order \(\mathcal O (\alpha ^6\alpha _\text {s})\)) accuracy. For such a comparison we expect larger discrepancies than what we found at fixed order, as a consequence of the different matching schemes, parton showers employed, and of other details of the matching (such as the choice of the parton shower initial scale). Among the codes capable of providing fixedorder results, presented before, MG5_aMC, the PowhegBox, and VBFNLO can also provide results at (N)LO+PS accuracy. For VBFNLO matched to Herwig and the PowhegBox, we restrict ourselves to show results only in the VBS approximation, i.e. the schannel contributions are neglected here. Besides, also Phantom and Whizard are used for LO+PS results.
MG5_aMC, which employs the MC@NLO [59] matching procedure, is used together with Pythia8 [86] (version 8.223) and Herwig7 [87, 88] (version 7.1.2). For the latter, the default angularordered shower is employed. The same parton showers are employed for the LO results of Phantom. Pythia8 is also employed for the LO results of Whizard. For the PowhegBox, the namesake matching procedure is employed [53, 54], together with Pythia8 (version 8.230). VBFNLO serves as a matrixelement and phasespace provider for the Matchbox module [89] of Herwig7 [87, 88], using an extended version of the Binoth Les Houches Accord interface [90, 91, 92]. The Matchbox module makes it possible to choose between MC@NLOlike and Powheglike matching. As parton shower, both the default angularordered shower as well as the dipole shower can be employed. We use here the subtractive (MC@NLOtype) matching to these partonshower algorithms. Whenever Pythia8 is used, the Monash tune [93] is selected. Multipleparton interactions are disabled.
Results are presented within the cuts described in Sect. 3.3, applied after shower and hadronisation (this implies that jets are obtained by clustering stable hadrons, and not QCD partons). It follows that at the eventgeneration level, looser cuts (or even no cuts at all) must be employed in order not to bias the results. This also implies that the tagging jets, whose momenta enter in the renormalisation and factorisation scales, Eq. (7), are now defined without imposing the \(\varDelta R_{\text {j} \ell }\) cut. The effect of this change is below one per cent at the level of the fiducial cross sections at NLO.
Compared to the fixedorder computations, a slightly different setup has been employed for MG5_aMC in order to simplify the calculation: instead of generating the full \(\mathrm{p}\mathrm{p}\rightarrow \mu ^+\nu _\mu \mathrm{e}^+\nu _\mathrm{e}\mathrm{j}\mathrm{j}\) process, since it is dominated by doublyresonant contribution, the events are produced for the process with two stable \(\mathrm{W}^+\)bosons (\(\mathrm{p}\mathrm{p}\rightarrow \mathrm{W^+}\mathrm{W^+}\mathrm{j}\mathrm{j}\)), and the decay of these \(\mathrm{W}^+\)bosons is simulated with MadSpin [94] (ensuring spin correlations) before the parton shower. Since MadSpin computes the partial and total decay widths of the W bosons at LO accuracy only, while in Sect. 3.3 the NLO width is employed, an effect (\(6\%\)) on the normalisation is induced.
Table 6
Cross sections at LO+PS accuracy. The MG5_aMC results with \(\varGamma _\mathrm{resc}\) are rescaled to account for the effect related to the Wboson width computed by MadSpin (see the text for details). The uncertainties shown refer to estimated statistical errors of the Monte Carlo integrations
Code
\(\sigma \;(\mathrm {fb})\)
MG5_aMC+Pythia8
1.352 ± 0.003
MG5_aMC+Herwig7
1.342 ± 0.003
MG5_aMC+Pythia8, \(\varGamma _\mathrm{resc}\)
1.275 ± 0.003
MG5_aMC+Herwig7, \(\varGamma _\mathrm{resc}\)
1.266 ± 0.003
Phantom+Pythia8
1.235 ± 0.001
Phantom+Herwig7
1.258 ± 0.001
VBFNLO+Herwig7Dipole
1.3001 ± 0.0002
Whizard+Pythia8
1.229 ± 0.001
Table 7
Cross sections at NLO+PS accuracy. The MG5_aMC results with \(\varGamma _\mathrm{resc}\) are rescaled to account for the effect related to the Wboson width computed by MadSpin (see the text for details). For VBFNLO+Herwig7Dipole, the threepoint scale uncertainties are shown, while for MG5_aMC+Pythia8 the two displayed uncertainties are respectively the ninepoint scale uncertainty and the PDF one. The uncertainties shown refer to estimated statistical errors of the Monte Carlo integrations
Code
\(\sigma \;(\mathrm {fb})\)
MG5_aMC+Pythia8
\(1.491 ^{+1\%}_{2\%} {}^{+2\%}_{2\%} \pm 0.004\)
MG5_aMC+Herwig7
\(1.427 \pm 0.003\)
MG5_aMC+Pythia8, \(\varGamma _\mathrm{resc}\)
\(1.407 \pm 0.003\)
MG5_aMC+Herwig7, \(\varGamma _\mathrm{resc}\)
\(1.346 \pm 0.002\)
PowhegBox+Pythia8
\(1.3642 \pm 0.0004\)
VBFNLO+Herwig7Dipole
\(1.3389 ^{+0\%}_{1\%} \pm 0.0006\)
VBFNLO+Herwig7
\(1.3067 \pm 0.0006\)
Open image in new window
Fig. 13
Differential distribution in the exclusive jet multiplicity from predictions matched to parton showers, at LO (left) or NLO (right) accuracy (upper plot), compared with the fixedNLO result computed with VBFNLO (lower plot). At NLO+PS accuracy, for VBFNLO+Herwig7Dipole, the threepoint scale uncertainties are shown, while for MG5_aMC+Pythia8 the darker and lighter bands correspond respectively to the ninepoint scale uncertainty and the scale and PDF uncertainties combined linearly. The predictions are obtained in the fiducial region described in Sect. 3.3
We now present the results of predictions matched to parton showers. The total rates within VBS cuts are displayed in Tables 6 and 7, at LO and NLO accuracy respectively. For MG5_aMC, the numbers with \(\varGamma _\mathrm{resc}\) are rescaled to take into account the width effects described in the above paragraph. At NLO accuracy, for MG5_aMC + Pythia8 and VBFNLO+Herwig7Dipole, we also quote theoretical uncertainties. For the former, we show both PDF and scale uncertainties,11 obtained via exact reweighting [72] by varying independently the renormalisation and factorisation scales by a factor of two around the central value, Eq. (7) (ninepoint variations). For the latter, we show the threepoint scale uncertainties, obtained by considering correlated variations of the renormalisation, factorisation, and shower starting scale. Theory uncertainties should have very little dependence on the tool employed. We observe that, once the width effect is taken into account, total rates from different tools agree within some per cents (\(\le 7\%\)), both at LO and NLO. Larger discrepancies, however, appear for differential observables, which we discuss in the following. Theory uncertainties on the total rates are very small, regardless of whether scale variations are estimated with independent or correlated variations of the renormalisation and factorisation scales. Concerning differential distributions, for each observable we display results in two plots, shown sidebyside. In the plot on the left (right), (N)LO+PS predictions are shown with different colours in the main frame. In the inset, these predictions are compared in both cases with a fixedorder prediction at NLO accuracy (obtained with VBFNLO, i.e. the VBS approximation with schannel). For the differential observables, the MG5_aMC predictions are not rescaled to compensate for the width effect mentioned above. As for the table, we show theoretical uncertainties for the NLO+PS samples obtained with VBFNLO and MG5_aMC: again, for the first the band corresponds to threepoint variations, while for the second the darker (lighter) band corresponds to ninepoint scale variations (plus PDF uncertainties, linearly added).
The first observable we investigate is the exclusive jet multiplicity, shown in Fig. 13. Looking at the LO+PS predictions, one can appreciate that the main effects are driven by the parton shower that is employed (Herwig7 or Pythia8), with the clear tendency of producing more radiation for the latter, leading to higher jet multiplicities. Differences among tools that employ the same parton shower are typically smaller, and can be traced back to different values of the initial scale of the parton shower (the scalup entry of the Les Houches Event (LHE) file [95, 96]). This eventbyevent number corresponds to the maximum hardness (translated into the showerevolution variable) of the radiation that can be generated by the shower.12 The main effect of NLO corrections for this (rather inclusive) observable is to stabilise the predictions for the twojet bin, where discrepancies among tools are reduced to about \(10\%\). For the threejet bin, which is described only at LO accuracy, differences among tools remain large, and are possibly related to the underlying approximation performed (MG5_aMC is the only tool to use the full matrix element for the real radiation), in particular the inclusion of the schannel contributions: the largest rate is predicted by MG5_aMC, while the smallest rate is predicted by the PowhegBox, both matched to Pythia8. Despite the fact that the same parton shower is employed, the way emissions are treated is different among the two tools. In particular, for the PowhegBox, the first emission is generated with an internal Sudakov form factor (the prediction dubbed PowhegLHE corresponds to stopping after the first emission), while for MG5_aMC there is an interplay between the realemission matrix element and the shower emission. For this observable we also show the prediction obtained with MG5_aMC+Pythia8 by reducing the shower starting scale by a factor 2 with respect to the default value, dubbed MG5_aMC+Py8, \(Q_{sh}/2\).13 The main effect of reducing the shower scale is that events migrate from the threejet bin into the twojet bin, i.e. less radiation is generated. The size of this effect on the jet rate is \(+\,4\%\) (\(\,8\%\)) on the two (three) jet bin, while the total rate within cuts is left unchanged.
Open image in new window
Fig. 14
Differential distribution in the invariant mass of the two tagging jets from predictions matched to parton showers, at LO (left) or NLO (right) accuracy (upper plot), compared with the fixedNLO result computed with VBFNLO (lower plot). At NLO+PS accuracy, for VBFNLO+Herwig7Dipole, the threepoint scale uncertainties are shown, while for MG5_aMC+Pythia8 the darker and lighter bands correspond respectively to the ninepoint scale uncertainty and the scale and PDF uncertainties combined linearly. The predictions are obtained in the fiducial region described in Sect. 3.3
The next observable that we study is the invariant mass of the two tagging jets, shown in Fig. 14. For this observable, both at LO+PS and NLO+PS, the spread of predictions matched with parton shower is rather small (\(\lesssim 10\%\), if one compensates for the \(6\%\) width effect for MG5_aMC). The LO+PS predictions tend to be significantly softer than the fixed NLO one, with an effect of about \(\,30\%\) at the end of the displayed range. At NLO+PS, this effect is mitigated, owing to the better description of the first QCD emission which is now driven by the realemission matrix element. For this observable (and all the others which are NLO accurate) the effect of reducing the shower scale is negligible, hence it is not shown.
The rapidity difference between the two tagging jets, shown in Fig. 15, has some similarities with the invariantmass distribution. At LO+PS all predictions show the tendency to deplete the largeseparation region with respect to the fixedorder prediction, in a quantitatively similar way, except for VBFNLO+Herwig7 where the effect is mitigated. At NLO+PS, when the extra radiation is described by the real matrix element, such an effect is greatly reduced. A notable exception is the PowhegBox prediction, which still shows a suppression at large separations. Since such a suppression is already there for the PowhegLHE sample, it is very likely that it is driven by the way the first emission is generated. A minor effect in the same direction is visible in the last two bins of the MG5_aMC+Herwig7 prediction (although with rather large statistical uncertainties).
The transverse momentum of the hardest and secondhardest jets are shown in Figs. 16 and 17, respectively. In general, for both observables, predictions from different tools agree rather well with each other, with a spread at most at the 10% level. At LO+PS, typically the transversemomentum spectra are softer than the fixedNLO ones, and this effect is more marked for the secondhardest jet which, as expected, is more sensitive to the description of the extra radiation. Again, this effect is mitigated by NLO corrections. The only feature that may be worth noticing among the NLO+PS predictions is the tendency of the PowhegBox to suppress the hardestjet spectrum at low transverse momentum (\(p_{\text {T},\text {j} _1} <100 \,\text {GeV} \)).
Open image in new window
Fig. 15
Differential distribution in the rapidity separation of the two tagging jets from predictions matched to parton showers, at LO (left) or NLO (right) accuracy (upper plot), compared with the fixedNLO result computed with VBFNLO (lower plot). At NLO+PS accuracy, for VBFNLO+Herwig7Dipole, the threepoint scale uncertainties are shown, while for MG5_aMC+Pythia8 the darker and lighter bands correspond respectively to the ninepoint scale uncertainty and the scale and PDF uncertainties combined linearly. The predictions are obtained in the fiducial region described in Sect. 3.3
Open image in new window
Fig. 16
Differential distribution in the transverse momentum of the hardest jet from predictions matched to parton showers, at LO (left) or NLO (right) accuracy (upper plot), compared with the fixedNLO result computed with VBFNLO (lower plot). At NLO+PS accuracy, for VBFNLO+Herwig7Dipole, the threepoint scale uncertainties are shown, while for MG5_aMC+Pythia8 the darker and lighter bands correspond respectively to the ninepoint scale uncertainty and the scale and PDF uncertainties combined linearly. The predictions are obtained in the fiducial region described in Sect. 3.3
Open image in new window
Fig. 17
Differential distribution in the transverse momentum of the secondhardest jet from predictions matched to parton showers, at LO (left) or NLO (right) accuracy (upper plot), compared with the fixedNLO result computed with VBFNLO (lower plot). At NLO+PS accuracy, for VBFNLO+Herwig7Dipole, the threepoint scale uncertainties are shown, while for MG5_aMC+Pythia8 the darker and lighter bands correspond respectively to the ninepoint scale uncertainty and the scale and PDF uncertainties combined linearly. The predictions are obtained in the fiducial region described in Sect. 3.3
Open image in new window
Fig. 18
Differential distribution in the rapidity of the secondhardest jet from predictions matched to parton showers, at LO (left) or NLO (right) accuracy (upper plot), compared with the fixedNLO result computed with VBFNLO (lower plot). At NLO+PS accuracy, for VBFNLO+Herwig7Dipole, the threepoint scale uncertainties are shown, while for MG5_aMC+Pythia8 the darker and lighter bands correspond respectively to the ninepoint scale uncertainty and the scale and PDF uncertainties combined linearly. The predictions are obtained in the fiducial region described in Sect. 3.3
Open image in new window
Fig. 19
Differential distribution in the Zeppenfeld variable of the thirdhardest jet from predictions matched to parton showers, at LO (left) or NLO (right) accuracy (upper plot), compared with the fixedNLO result computed with VBFNLO (lower plot). At NLO+PS accuracy, for VBFNLO+Herwig7Dipole, the threepoint scale uncertainties are shown, while for MG5_aMC+Pythia8 the darker and lighter bands correspond respectively to the ninepoint scale uncertainty and the scale and PDF uncertainties combined linearly. The predictions are obtained in the fiducial region described in Sect. 3.3
If we consider the rapidity of the second jet, Fig. 18, we observe again rather small differences among tools, with the tendency towards a general stabilisation at NLO+PS. However, some (small) differences in the shape remain at NLO+PS, which are worth to be briefly discussed: predictions obtained with MG5_aMC are very close to the fixedorder prediction; the PowhegBox displays an enhancement of the central region, and a consequent suppression in the peripheral region, while VBFNLO shows an opposite behaviour. However, the effect is rather small, with the largest departure from the fixedorder prediction being at most \(10\%\).14
Finally, focusing on the third jet, we conclude the list of differential observables by showing the Zeppenfeld variable defined in Eq. (12), Fig. 19. This variable is closely related to the third jet rapidity, and small (large) values of z correspond to central (peripheral) rapidities. In general, for observables which involve the third jet, one can clearly see a degradation of the agreement among the various tools, because of the poorer perturbative description of these observables. The Zeppenfeld variable is a striking example: both at LO and NLO, the tendency of Pythia8 to generate more hard and central radiation, corresponding to low values of z, is clearly visible. Such an effect, which is related to the way Pythia8 deals with the recoil of the radiation in VBF(VBS)type processes, can be mitigated by setting SpaceShower:dipoleRecoil = on in the Pythia8 input file.15 It is interesting to notice that the effect survives beyond the first emission, as it can be observed by comparing PowhegLHE with Powheg+ Pythia8, and that it is only marginally attenuated when the shower scale is reduced. A similar behaviour of Pythia8 has also been observed in the study of EW production of a \(\text {Z} \) boson in association with two jets (see the recent CMS measurement, Ref. [97] Figure 12), where the experimental data seem to prefer the description by Herwig++ [98, 99]. The central enhancement is a bit mitigated if NLO+PS tools are used (compare LO+PS and NLO+PS from MG5_aMC+Pythia8 with the fixedNLO prediction), however even at NLO+PS the central region (\(z_{j_3}<0.5\)) is cursed by huge differences between tools. Large differences, reaching a factor 2, persist also away from the central region. These findings are consistent with behaviour displayed in Refs. [11, 100, 101, 102, 103] where the behaviour of NLO matching in VBS processes has been reported.
In conclusion, the comparison of tools including matching to parton shower clearly shows the benefits of the inclusion of NLO corrections: for most observables described effectively at NLO accuracy differences between tools are at (or below) the \(10\%\) level. Some exceptions exist, e.g. the rapidity separation of the two tagging jets, which on the one hand clearly suggest not to rely on a single tool/parton shower, and on the other make it worth investigating more in detail the way QCD radiation is generated, e.g. when fullydifferential computations at NNLO will become available (for VBF Higgs production, see Refs. [104, 105]). It is a remarkable fact that, even for those observables that display small discrepancies, the theoretical uncertainty obtained via scale variations (renormalisation, factorisation, and shower scale) systematically underestimates the spread of predictions. We note that in the only VBF process where NNLO corrections are known, i.e. VBF Higgs production [104, 105], the NLO scaleuncertainty band does not include the NNLO prediction. This suggests that the NLO scale variation underestimates the size of the perturbative uncertainty. Again, this stresses the need to employ at least two different tools in order to obtain a more realistic estimate of theoretical uncertainties. Finally, the size of discrepancies for observables that are described at a lower perturbative accuracy, notably those related to the third jet, suggests that experimental analyses should rely as little as possible on those observables and, in any case, use conservative estimates of the theory uncertainties. On the one hand, in order to improve the description of these observables, a simulation of VBS+j at NLO accuracy, currently unavailable but within the reach of modern automated tools, is certainly desirable. On the other hand, measurements of processes with similar colour flow (EW production of a single vector boson plus jets, VBF, \(\ldots \)) can certainly help in order to discriminate which tools perform better in the comparison with data [97, 106].
7 Conclusions and recommendations
In the present article, a detailed study of the process \(\text {p} \text {p} \rightarrow \mu ^+\nu _\mu \mathrm{e}^+\nu _\mathrm{e}\,\text {j} \text {j} +\mathrm {X}\) at the LHC has been presented, mainly focused on the EW production mechanism which involves the scattering of massive vector bosons. Until very recently, when the complete calculation became available for the NLO QCD corrections (order \(\mathcal O (\alpha _\text {s} \alpha ^6)\)), the socalled VBS approximation was the standard for this kind of simulations. For this reason, various theoretical predictions have been compared to the full computation, both in a typical VBS fiducial region and also in more inclusive phase space. We have precisely quantified the differences that arise for several physical observables, in particular for the dijet invariant mass and the rapidity separation of the leading two jets. This is the first time that such an indepth study is performed. Besides the study of fixedorder predictions, we have also investigated the impact of parton showering. To that end, several LO and NLO event generators which are able to perform matching to parton showers have been employed, and various observables have been thoroughly compared. While in general observables which are described at NLO accuracy show reasonable agreement among the tools, larger differences can appear for those observables described at a lower accuracy, such as those that involve the third jet. In particular such differences are quite prominent in the centralrapidity region, and are the largest for those simulations which employ Pythia8. The effect has been understood, and it can be partially mitigated by changing the recoil scheme of Pythia8 to distribute momenta within initial–final colour connections. These findings make it worth to further investigate these issues not only in the theoretical community, but also by experimental collaborations, for example by measuring related observables for similar processes.
The last part of our work is devoted to remarks and recommendations concerning the usage of theoretical predictions by experimental collaborations.
As found in Ref. [19], the NLO EW corrections of order \(\mathcal {O}{\left( \alpha ^{7}\right) }\) are the dominant NLO contribution to the process \(\text {p} \text {p} \rightarrow \mu ^+\nu _\mu \mathrm{e}^+\nu _\mathrm{e}\,\text {j} \text {j} +\mathrm {X}\). It is thus highly desirable to combine them with NLOQCD predictions matched with parton shower, or at least to include them into experimental analyses. Since, as shown in Ref. [77], these large EW corrections originate from the Sudakov logarithms which factorise, we recommend to combine them with QCD corrections in a multiplicative way. The estimate of missing higherorder EW corrections can be obtained, in a first approximation, by considering \(\pm \delta ^2_\mathrm{NLO EW}\),16 while the missing higherorder mixed QCDEW corrections can be estimated by taking the difference between the multiplicative and additive prescriptions. For more detailed studies of the combination of QCD and EW higherorder corrections, see e.g. Ref. [107] in the context of toppair production, or Ref. [108] for SM backgrounds for dark matter searches at the LHC.
For the typical fiducial region used by experimental collaborations for their measurements, the agreement between the approximations and the full calculation is satisfactory given the current experimental precision, as well as the one foreseen for the near future [5, 6, 109]. Nonetheless, care has to be taken when using such approximations, in particular if more inclusive phasespace cuts are used.
In addition to the standard interpretation of EW signal versus QCD background, combined measurements should also be presented as they are better defined theoretically. In fact, while at LO the interference term can be included in the background component, at NLO the separation of EW and QCD components becomes more blurred, as, e.g. at the order \(\mathcal {O}{\left( \alpha _\text {s} \alpha ^{6}\right) }\) both types of amplitudes contribute. Therefore, a combined measurement including the EW, QCD, and interference contributions is desirable. Note that with such a measurement a comparison to the SM would be straightforward and still be sensitive to the EW component. In addition, the QCD component could be subtracted based on a welldefined Monte Carlo prediction.
Since the inclusion of NLO QCD corrections gives a better control of extra QCD radiation and reduces the ambiguities related to the matching details and/or the parton shower employed, we encourage the use of NLOaccurate event generators in experimental analyses. In doing so, special care should be employed in order to estimate the theoretical uncertainties, as the standard prescription based on renormalisation and factorisationscale variation is clearly inadequate. Rather, different combinations of generators and parton showers should be employed.
The present study has focused on the orders \(\mathcal {O}{\left( \alpha ^{6}\right) }\) at LO and \(\mathcal {O}{\left( \alpha _\text {s} \alpha ^{6}\right) }\) at NLO. NLO computations and publiclyavailable tools also exist for the QCDinduced process [13, 14, 15, 16, 17, 19, 58].
For practical reasons, we have focused on the \(\mathrm{W}^+\mathrm{W}^+\) signature. Nonetheless, the observed features (e.g. validity of the VBS approximation or comparison of theoretical predictions matched to parton shower) should be qualitatively similar for other VBS signatures with massive gauge bosons. For these other signatures, similar quantitative studies should be performed.
Footnotes
1.
Preliminary results of the present study have already been made public in Ref. [20]. A similar study has also appeared very recently for the electroweak (EW) production of a Higgs boson in association with 3 jets [21].
2.
The name VBS is used even though not all Feynman diagrams involve the scattering of vector bosons.
3.
Despite the loop_qcd_qed_sm_Gmu model also includes NLO counterterms for computing EW corrections, it is not yet possible to compute such corrections with the current public version of the code.
4.
For the process considered, no bottom (anti)quarks appear in the initial or final state at LO and NLO, as they would lead to top quarks, and not light jets, in the final state.
5.
Note that the PowhegBox uses its own implementation of the two loop running for \(\alpha _\mathrm{s}\).
6.
The corresponding identifier lhaid in the program LHAPDF6 [80] is 260000.
7.
By default, the renormalisation and factorisation scales employed in the PowhegBox slightly differ from the ones defined in Eq. (7), as the momenta of two finalstate quarks in the underlying Born event are employed instead of those of the two hardest jets.
8.
In Fig. 4, the level of the accuracy of the predictions in each bin is around a per mille.
9.
The bin in the topleft corner of the righthandside plot of Fig. 7 suffers from large uncertainty (\(30\%\)) while the other errors are at the percent level.
10.
The VBFNLO prediction omitting schannel contributions amounts to \(1.3703 \pm 0.0001\) fb. This differs from the PowhegBox prediction mainly due to the different choice of scales used in the PowhegBox (cf. footnote 7).
11.
A preliminary study on PDF uncertainties in VBS has appeared in Ref. [47].
12.
At LO, the choice of such a scale is arbitrary and usually driven by common sense, as it is the case for the factorisation and renormalisation scales. At NLO, one has the freedom to change the shower scale without losing formal NLO accuracy within the MC@NLO matching, provided the Monte Carlo counterterms are also consistently updated. In the Powheg matching, the shower scale of the socalled \(\tilde{B}\) events is fixed to the transverse momentum of the radiation generated according the Powheg Sudakov factor, while it can be changed in the remnant events.
13.
The reduction of the shower scale for MG5_aMC+Herwig7 gives no visible effect on any of the observables discussed in this work.
14.
If the setting SpaceShower:dipoleRecoil = on (discussed in the following) is used when Pythia8 is employed together with the PowhegBox, the enhancement at central rapidities and the depletion at small value of transverse momentum are partially compensated.
15.
This requires version \(\ge 8.230\). Note that such a setting is not compatible with the NLO matching in MG5_aMC (but it is compatible with the Powheg matching). Also, this setting has other effects, though smaller, on the rapidity spectra of the two hardest jets.
16.
The quantity \(\delta _\mathrm{NLO EW}\) is defined through the relation \(\sigma _\mathrm{NLO EW} = \sigma _\mathrm{LO} \left( 1+ \delta _\mathrm{NLO EW}\right) \).
Notes
Acknowledgements
The authors are grateful to the members of the VBSCan collaboration for several discussions and comments on this work. The author also thank the Pythia8 developers, in particular Stefan Prestel, Torbjörn Sjöstrand, and Peter Skands for discussions and clarifications about the thirdjet rapidity spectrum. MZ would like to thank Andreas Papaefstathiou, Christian Reuschle and David Grellscheid for their help with Herwig7. RF, HSS, and MZ thank the MadGraph5_aMC@NLO authors for discussions, and they are particularly grateful to Valentin Hirschi for comments on this manuscript. The authors would like to acknowledge the contribution of the COST Action CA16108 which initiated this work. Moreover, this work was supported by several STSM Grants from the COST Action CA16108. Many authors acknowledge hospitality from Nikhef, where part of this work has been performed. BB, AD, and MP acknowledge financial support by the German Federal Ministry for Education and Research (BMBF) under Contract no. 05H15WWCA1 and the German Research Foundation (DFG) under reference number DE 623/61. SD and CS acknowledge support by the state of BadenWürttemberg through bwHPC and the DFG through Grant no. INST 39/9631 FUGG and Grant DI 784/3. The work of BJ was supported in part by the Institutional Strategy of the University of Tübingen (DFG, ZUK 63) and in part by the BMBF under Contract number 05H2015. AK acknowledges financial support by the Swiss National Science Foundation (SNF) under Contract 200020175595. MR acknowledges funding from the European Union’s Horizon 2020 research and innovation programme as part of the Marie SkłodowskaCurie Innovative Training Network MCnetITN3 (Grant agreement no. 722104). HSS is supported by the ILP Labex (ANR11IDEX000402, ANR10LABX63). GZ is supported by ERC Consolidator Grant HICCUP (no. 614577). MZ is supported by the Netherlands National Organisation for Scientific Research (NWO).
References
1.
ATLAS Collaboration, G. Aad et al., Evidence for electroweak production of \(W^{\pm }W^{\pm }jj\) in \(pp\) collisions at \(\sqrt{s}=8\) TeV with the ATLAS detector. Phys. Rev. Lett. 113(14), 141803 (2014). arXiv:1405.6241 [hepex]
2.
CMS Collaboration, V. Khachatryan et al., Study of vector boson scattering and search for new physics in events with two samesign leptons and two jets. Phys. Rev. Lett. 114(5), 051801 (2015). arXiv:1410.6315 [hepex]
3.
ATLAS Collaboration, M. Aaboud et al., Measurement of \(W^{\pm }W^{\pm }\) vectorboson scattering and limits on anomalous quartic gauge couplings with the ATLAS detector. Phys. Rev. D 96, 012007 (2017). arXiv:1611.02428 [hepex]
4.
CMS Collaboration, A.M. Sirunyan et al., Observation of electroweak production of samesign W boson pairs in the two jet and two samesign lepton final state in protonproton collisions at \(\sqrt{s} = \) 13 TeV. Phys. Rev. Lett. 120(8), 081801 (2018). arXiv:1709.05822 [hepex]
5.
CMS Collaboration, Prospects for the study of vector boson scattering in same sign WW and WZ interactions at the HLLHC with the upgraded CMS detector. http://cds.cern.ch/record/2220831
6.
ATLAS Collaboration, Studies on the impact of an extended inner detector tracker and a forward muon tagger on \(W^{\pm }W^{\pm }\) scattering in \(pp\) collisions at the HighLuminosity LHC with the ATLAS experiment. https://cds.cern.ch/record/2298958
7.
B. Jäger, C. Oleari, D. Zeppenfeld, Nexttoleading order QCD corrections to W\(^{+}\)W\(^{}\) production via vectorboson fusion. JHEP 07, 015 (2006). arXiv:hepph/0603177 [hepph]Google Scholar
8.
B. Jäger, C. Oleari, D. Zeppenfeld, Nexttoleading order QCD corrections to Z boson pair production via vectorboson fusion. Phys. Rev. D 73, 113006 (2006). arXiv:hepph/0604200 [hepph]ADSGoogle Scholar
9.
G. Bozzi, B. Jäger, C. Oleari, D. Zeppenfeld, Nexttoleading order QCD corrections to W\(^+\)Z and W\(^\)Z production via vectorboson fusion. Phys. Rev. D 75, 073004 (2007). arXiv:hepph/0701105 [hepph]ADSGoogle Scholar
10.
B. Jäger, C. Oleari, D. Zeppenfeld, Nexttoleading order QCD corrections to \(W^+ W^+ jj\) and \(W^ W^ jj\) production via weakboson fusion. Phys. Rev. D 80, 034022 (2009). arXiv:0907.0580 [hepph]ADSGoogle Scholar
11.
B. Jäger, G. Zanderighi, NLO corrections to electroweak and QCD production of \(W^+W^+\) plus two jets in the POWHEGBOX. JHEP 11, 055 (2011). arXiv:1108.0864 [hepph]Google Scholar
12.
A. Denner, L. Hošeková, S. Kallweit, NLO QCD corrections to \(W^+ W^+ jj\) production in vectorboson fusion at the LHC. Phys. Rev. D 86, 114014 (2012). arXiv:1209.2389 [hepph]ADSGoogle Scholar
13.
M. Rauch, Vectorboson fusion and vectorboson scattering. arXiv:1610.08420 [hepph]
14.
T. Melia, K. Melnikov, R. Röntsch, G. Zanderighi, Nexttoleading order QCD predictions for \(W^+W^+jj\) production at the LHC. JHEP 12, 053 (2010). arXiv:1007.5313 [hepph]ADSGoogle Scholar
15.
T. Melia, P. Nason, R. Röntsch, G. Zanderighi, \(W^+W^+\) plus dijet production in the POWHEGBOX. Eur. Phys. J. C 71, 1670 (2011). arXiv:1102.4846 [hepph]ADSGoogle Scholar
16.
F. Campanario, M. Kerner, L.D. Ninh, D. Zeppenfeld, Nexttoleading order QCD corrections to \(W^+W^+\) and \(W^W^\) production in association with two jets. Phys. Rev. D 89(5), 054009 (2014). arXiv:1311.6738 [hepph]ADSGoogle Scholar
17.
J. Baglio et al., Release note—VBFNLO 2.7.0. arXiv:1404.3940 [hepph]
18.
C. Oleari, D. Zeppenfeld, QCD corrections to electroweak \(\ell \nu _\ell j j\) and \(\ell ^+ \ell ^ j j\) production. Phys. Rev. D 69, 093004 (2004). arXiv:hepph/0310156 [hepph]ADSGoogle Scholar
19.
B. Biedermann, A. Denner, M. Pellen, Complete NLO corrections to W\(^{+}\)W\(^{+}\) scattering and its irreducible background at the LHC. JHEP 10, 124 (2017). arXiv:1708.00268 [hepph]ADSGoogle Scholar
20.
C.F. Anders et al., VBSCan split 2017 workshop summary (2018). arXiv:1801.04203 [hepph]
21.
F. Campanario, T.M. Figy, S. Plätzer, M. Rauch, P. Schichtel, M. Sjödahl, Stresstesting the VBF approximation in multijet final states. arXiv:1802.09955 [hepph]
22.
I. Kuss, H. Spiesberger, Luminosities for vector boson  vector boson scattering at highenergy colliders. Phys. Rev. D 53, 6078–6093 (1996). arXiv:hepph/9507204 [hepph]ADSGoogle Scholar
23.
E. Accomando, A. Denner, S. Pozzorini, Logarithmic electroweak corrections to \(e^+ e^\rightarrow \nu _e \bar{\nu }_e W^+ W^\). JHEP 03, 078 (2007). arXiv:hepph/0611289 [hepph]ADSGoogle Scholar
24.
S. Dawson, The effective \(W\) approximation. Nucl. Phys. B 249, 42–60 (1985)ADSMathSciNetGoogle Scholar
25.
M.J. Duncan, G.L. Kane, W.W. Repko, \(WW\) physics at future colliders. Nucl. Phys. B 272, 517–559 (1986)ADSGoogle Scholar
26.
R.N. Cahn, S. Dawson, Production of very massive Higgs bosons. Phys. Lett. 136B, 196 (1984). [Erratum: Phys. Lett. 138B, 464 (1984)]ADSGoogle Scholar
27.
A. Ballestrero, A. Belhouari, G. Bevilacqua, V. Kashkan, E. Maina, PHANTOM: a Monte Carlo event generator for six parton final states at high energy colliders. Comput. Phys. Commun. 180, 401–417 (2009). arXiv:0801.3359 [hepph]ADSGoogle Scholar
28.
A. Denner et al., Predictions for all processes \(e^+ e^\rightarrow 4\,\)fermions \({}+ \gamma \). Nucl. Phys. B 560, 33–65 (1999). arXiv:hepph/9904472 ADSGoogle Scholar
29.
A. Denner et al., Electroweak corrections to chargedcurrent \(e^+ e^\rightarrow 4\,\) fermion processes: technical details and further results. Nucl. Phys. B 724, 247–294 (2005). arXiv:hepph/0505042 ADSMathSciNetGoogle Scholar
30.
A. Denner, S. Dittmaier, The complexmass scheme for perturbative calculations with unstable particles. Nucl. Phys. Proc. Suppl. 160, 22–26 (2006). arXiv:hepph/0605312 [hepph]. [,22(2006)]ADSGoogle Scholar
31.
A. Ballestrero, PHACT: helicity amplitudes for present and future colliders. In: High Energy Physics and Quantum Field Theory. Proceedings, 14th International Workshop, QFTHEP’99, Moscow, Russia, May 27–June 2, 1999, pp. 303–309. (1999). arXiv:hepph/9911318 [hepph]
32.
A. Ballestrero, E. Maina, A New method for helicity calculations. Phys. Lett. B 350, 225–233 (1995). arXiv:hepph/9403244 [hepph]ADSGoogle Scholar
33.
F.A. Berends, P.H. Daverveldt, R. Kleiss, Complete lowest order calculations for four lepton final states in electronpositron collisions. Nucl. Phys. B 253, 441–463 (1985)ADSGoogle Scholar
34.
G.P. Lepage, A new algorithm for adaptive multidimensional integration. J. Comput. Phys. 27, 192 (1978)ADSzbMATHGoogle Scholar
35.
W. Kilian, T. Ohl, J. Reuter, WHIZARD: simulating multiparticle processes at LHC and ILC. Eur. Phys. J. C 71, 1742 (2011). arXiv:0708.4233 [hepph]ADSGoogle Scholar
36.
M. Moretti, T. Ohl, J. Reuter, O’Mega: an optimizing matrix element generator. arXiv:hepph/0102195 [hepph]
37.
W. Kilian, T. Ohl, J. Reuter, C. Speckner, QCD in the colorflow representation. JHEP 10, 022 (2012). arXiv:1206.3700 [hepph]ADSGoogle Scholar
38.
B. Chokoufé Nejad, W. Kilian, J.M. Lindert, S. Pozzorini, J. Reuter, C. Weiss, NLO QCD predictions for offshell \(t \bar{t}\) and \(t \bar{t} H\) production and decay at a linear collider. JHEP 12, 75 (2016). arXiv:1609.03390 [hepph]Google Scholar
39.
S. Frixione, Z. Kunszt, A. Signer, Three jet crosssections to nexttoleading order. Nucl. Phys. B 467, 399–442 (1996). arXiv:hepph/9512328 [hepph]ADSGoogle Scholar
40.
S. Frixione, A general approach to jet crosssections in QCD. Nucl. Phys. B 507, 295–314 (1997). arXiv:hepph/9706545 [hepph]ADSGoogle Scholar
41.
F. Cascioli, P. Maierhöfer, S. Pozzorini, Scattering amplitudes with open loops. Phys. Rev. Lett. 108, 111601 (2012). arXiv:1111.5206 [hepph]ADSGoogle Scholar
42.
S. Actis et al., Recursive generation of oneloop amplitudes in the standard model. JHEP 04, 037 (2013). arXiv:1211.6316 [hepph]ADSGoogle Scholar
43.
S. Actis et al., RECOLA: REcursive Computation of OneLoop Amplitudes. Comput. Phys. Commun. 214, 140–173 (2017). arXiv:1605.01090 [hepph]ADSzbMATHGoogle Scholar
44.
J. Reuter, B. Chokoufé Nejad, A. Hoang, W. Kilian, M. Stahlhofen, T. Teubner, C. Weiss, Automation of NLO processes and decays and POWHEG matching in WHIZARD. J. Phys. Conf. Ser. 762(1), 012059 (2016). arXiv:1602.06270 [hepph]Google Scholar
45.
A. Alboteanu, W. Kilian, J. Reuter, Resonances and unitarity in weak boson scattering at the LHC. JHEP 11, 010 (2008). arXiv:0806.4145 [hepph]ADSGoogle Scholar
46.
W. Kilian, T. Ohl, J. Reuter, M. Sekulla, Highenergy vector boson scattering after the Higgs discovery. Phys. Rev. D 91, 096007 (2015). arXiv:1408.6207 [hepph]ADSGoogle Scholar
47.
Schwan C, Vectorboson scattering at the LHC. In: Proceedings, 13th International Symposium on Radiative Corrections: Application of Quantum Field Theory to Phenomenology (RADCOR2017): St. Gilgen, Austria, September 24–29, 2017, vol. RADCOR2017 (2018). https://doi.org/10.22323/1.290.0081
48.
S. Dittmaier, M. Roth, LUSIFER: a LUcid approach to six FERmion production. Nucl. Phys. B 642, 307–343 (2002). arXiv:hepph/0206070 [hepph]ADSGoogle Scholar
49.
A. Denner, S. Dittmaier, L. Hofer, COLLIER—a fortranlibrary for oneloop integrals. PoS LL2014, 071 (2014). arXiv:1407.0087 [hepph]zbMATHGoogle Scholar
50.
A. Denner, S. Dittmaier, L. Hofer, Collier: a fortranbased complex oneloop library in extended regularizations. Comput. Phys. Commun. 212, 220–238 (2017). arXiv:1604.06792 [hepph]ADSzbMATHGoogle Scholar
51.
A. Denner, S. Dittmaier, M. Roth, D. Wackeroth, Electroweak radiative corrections to \({e}^+ {e}^ \rightarrow W W \rightarrow \) 4 fermions in doublepole approximation: The RACOONWW approach. Nucl. Phys. B 587, 67–117 (2000). arXiv:hepph/0006307 [hepph]ADSGoogle Scholar
52.
S. Dittmaier, C. Schwan, Nonfactorizable photonic corrections to resonant production and decay of many unstable particles. Eur. Phys. J. C 76(3), 144 (2016). arXiv:1511.01698 [hepph]ADSGoogle Scholar
53.
P. Nason, A new method for combining NLO QCD with shower Monte Carlo algorithms. JHEP 11, 040 (2004). arXiv:hepph/0409146 [hepph]ADSGoogle Scholar
54.
S. Frixione, P. Nason, C. Oleari, Matching NLO QCD computations with parton shower simulations: the POWHEG method. JHEP 11, 070 (2007). arXiv:0709.2092 [hepph]ADSGoogle Scholar
55.
S. Alioli, P. Nason, C. Oleari, E. Re, A general framework for implementing NLO calculations in shower Monte Carlo programs: the POWHEG BOX. JHEP 06, 043 (2010). arXiv:1002.2581 [hepph]ADSzbMATHGoogle Scholar
56.
K. Arnold et al., VBFNLO: a parton level Monte Carlo for processes with electroweak bosons. Comput. Phys. Commun. 180, 1661–1670 (2009). arXiv:0811.4559 [hepph]ADSGoogle Scholar
57.
K. Arnold et al., VBFNLO: a parton level Monte Carlo for processes with electroweak bosons—manual for Version 2.5.0. arXiv:1107.4038 [hepph]
58.
J. Alwall et al., The automated computation of treelevel and nexttoleading order differential cross sections, and their matching to parton shower simulations. JHEP 07, 079 (2014). arXiv:1405.0301 [hepph]ADSGoogle Scholar
59.
S. Frixione, B.R. Webber, Matching NLO QCD computations and parton shower simulations. JHEP 06, 029 (2002). arXiv:hepph/0204244 [hepph]ADSGoogle Scholar
60.
R. Frederix, S. Frixione, F. Maltoni, T. Stelzer, Automation of nexttoleading order computations in QCD: the FKS subtraction. JHEP 10, 003 (2009). arXiv:0908.4272 [hepph]ADSGoogle Scholar
61.
R. Frederix, S. Frixione, A.S. Papanastasiou, S. Prestel, P. Torrielli, Offshell singletop production at NLO matched to parton showers. JHEP 06, 027 (2016). arXiv:1603.01178 [hepph]ADSGoogle Scholar
62.
G. Ossola, C.G. Papadopoulos, R. Pittau, Reducing full oneloop amplitudes to scalar integrals at the integrand level. Nucl. Phys. B 763, 147–169 (2007). arXiv:hepph/0609007 [hepph]ADSMathSciNetzbMATHGoogle Scholar
63.
P. Mastrolia, E. Mirabella, T. Peraro, Integrand reduction of oneloop scattering amplitudes through Laurent series expansion. JHEP 06, 095 (2012). arXiv:1203.0291 [hepph]. [Erratum: JHEP 11, 128 (2012)]ADSMathSciNetGoogle Scholar
64.
G. Passarino, M.J.G. Veltman, Oneloop corrections for \(e^+ e^\) annihilation into \(\mu ^+ \mu ^\) in the Weinberg model. Nucl. Phys. B 160, 151–207 (1979)ADSGoogle Scholar
65.
A.I. Davydychev, A Simple formula for reducing Feynman diagrams to scalar integrals. Phys. Lett. B 263, 107–111 (1991)ADSMathSciNetGoogle Scholar
66.
A. Denner, S. Dittmaier, Reduction schemes for oneloop tensor integrals. Nucl. Phys. B 734, 62–115 (2006). hepph/0509141ADSzbMATHGoogle Scholar
67.
V. Hirschi, R. Frederix, S. Frixione, M.V. Garzelli, F. Maltoni, R. Pittau, Automation of oneloop QCD corrections. JHEP 05, 044 (2011). arXiv:1103.0621 [hepph]ADSzbMATHGoogle Scholar
68.
G. Ossola, C.G. Papadopoulos, R. Pittau, CutTools: a program implementing the OPP reduction method to compute oneloop amplitudes. JHEP 03, 042 (2008). arXiv:0711.3596 [hepph]ADSGoogle Scholar
69.
T. Peraro, Ninja: automated integrand reduction via laurent expansion for oneloop amplitudes. Comput. Phys. Commun. 185, 2771–2797 (2014). arXiv:1403.1229 [hepph]ADSzbMATHGoogle Scholar
70.
V. Hirschi, T. Peraro, Tensor integrand reduction via Laurent expansion. JHEP 06, 060 (2016). arXiv:1604.01363 [hepph]ADSzbMATHGoogle Scholar
71.
H.S. Shao, IREGI user manual (unpublished) Google Scholar
72.
R. Frederix et al., Fourlepton production at hadron colliders: aMC@NLO predictions with theoretical uncertainties. JHEP 02, 099 (2012). arXiv:1110.4738 [hepph]ADSGoogle Scholar
73.
S. Catani, M.H. Seymour, A general algorithm for calculating jet crosssections in NLO QCD. Nucl. Phys. B 485, 291–419 (1997). arXiv:hepph/9605323 [hepph]. [Erratum: Nucl. Phys. B 510, 503 (1998)]ADSGoogle Scholar
74.
S. Dittmaier, A general approach to photon radiation off fermions. Nucl. Phys. B 565, 69–122 (2000). arXiv:hepph/9904440 ADSGoogle Scholar
75.
A. Denner, R. Feger, NLO QCD corrections to offshell topantitop production with leptonic decays in association with a Higgs boson at the LHC. JHEP 11, 209 (2015). arXiv:1506.07448 [hepph]ADSGoogle Scholar
76.
A. Denner, J.N. Lang, M. Pellen, S. Uccirati, Higgs production in association with offshell topantitop pairs at NLO EW and QCD at the LHC. JHEP 02, 053 (2017). arXiv:1612.07138 [hepph]ADSGoogle Scholar
77.
B. Biedermann, A. Denner, M. Pellen, Large electroweak corrections to vectorboson scattering at the Large Hadron Collider. Phys. Rev. Lett. 118(26), 261801 (2017). arXiv:1611.02951 [hepph]ADSGoogle Scholar
78.
F.A. Berends, R. Pittau, R. Kleiss, All electroweak four fermion processes in electron–positron collisions. Nucl. Phys. B 424, 308–342 (1994). arXiv:hepph/9404313 [hepph]ADSGoogle Scholar
79.
NNPDF Collaboration, R.D. Ball et al., Parton distributions for the LHC Run II. JHEP 04, 040 (2015). arXiv:1410.8849 [hepph]
80.
A. Buckley, J. Ferrando, S. Lloyd, K. Nordström, B. Page, M. Rüfenacht, M. Schönherr, G. Watt, LHAPDF6: parton density access in the LHC precision era. Eur. Phys. J. C 75, 132 (2015). arXiv:1412.7420 [hepph]ADSGoogle Scholar
81.
D.Yu. Bardin, A. Leike, T. Riemann, M. Sachwitz, Energydependent width effects in \({e}^+ {e}^\)annihilation near the Zboson pole. Phys. Lett. B 206, 539–542 (1988)Google Scholar
82.
CMS Collaboration, Observation of electroweak production of samesign W boson pairs in the two jet and two samesign lepton final state in proton–proton collisions at 13 TeV. CMSPASSMP17004Google Scholar
83.
M. Cacciari, G.P. Salam, G. Soyez, The anti\(k_t\) jet clustering algorithm. JHEP 04, 063 (2008). arXiv:0802.1189 [hepph]ADSzbMATHGoogle Scholar
84.
M. Cacciari, G.P. Salam, G. Soyez, FastJet user manual. Eur. Phys. J. C72, 1896 (2012). arXiv:1111.6097 [hepph]ADSGoogle Scholar
85.
A. Denner, M. Pellen, Offshell production of topantitop pairs in the lepton\(+\)jets channel at NLO QCD. JHEP 02, 013 (2018). arXiv:1711.10359 [hepph]ADSGoogle Scholar
86.
T. Sjöstrand, S. Ask, J.R. Christiansen, R. Corke, N. Desai, P. Ilten, S. Mrenna, S. Prestel, C.O. Rasmussen, P.Z. Skands, An introduction to PYTHIA 8.2. Comput. Phys. Commun. 191, 159–177 (2015). arXiv:1410.3012 [hepph]ADSzbMATHGoogle Scholar
87.
J. Bellm et al., Herwig 7.0/Herwig++ 3.0 release note. Eur. Phys. J. C76(4), 196 (2016). arXiv:1512.01178 [hepph]ADSGoogle Scholar
88.
J. Bellm et al., Herwig 7.1 release note. arXiv:1705.06919 [hepph]
89.
S. Plätzer, S. Gieseke, Dipole Showers and automated NLO matching in Herwig++. Eur. Phys. J. C 72, 2187 (2012). arXiv:1109.6256 [hepph]ADSGoogle Scholar
90.
T. Binoth, A proposal for a standard interface between Monte Carlo tools and oneloop programs. Comput. Phys. Commun. 181, 1612–1622 (2010). arXiv:1001.1307 [hepph]. [1 (2010)]ADSzbMATHGoogle Scholar
91.
S. Alioli et al., Update of the Binoth Les Houches Accord for a standard interface between Monte Carlo tools and oneloop programs. Comput. Phys. Commun. 185, 560–571 (2014). arXiv:1308.3462 [hepph]ADSzbMATHGoogle Scholar
92.
J.R. Andersen et al., Les Houches 2013: physics at TeV colliders: standard model working group report. arXiv:1405.1067 [hepph]
93.
P. Skands, S. Carrazza, J. Rojo, Tuning PYTHIA 8.1: the Monash 2013 tune. Eur. Phys. J C74(8), 3024 (2014). arXiv:1404.5630 [hepph]ADSGoogle Scholar
94.
P. Artoisenet, R. Frederix, O. Mattelaer, R. Rietkerk, Automatic spinentangled decays of heavy resonances in Monte Carlo simulations. JHEP 03, 015 (2013). arXiv:1212.3460 [hepph]ADSGoogle Scholar
95.
J. Alwall, A Standard format for Les Houches event files. Comput. Phys. Commun. 176, 300–304 (2007). arXiv:hepph/0609017 [hepph]ADSGoogle Scholar
96.
J.M. Butterworth et al., The Tools and Monte Carlo Working Group summary report from the Les Houches 2009 workshop on TeV colliders. In: Physics at TeV colliders. Proceedings, 6th Workshop, dedicated to Thomas Binoth, Les Houches, France, June 8–26, 2009. 2010. arXiv:1003.1643 [hepph]. http://inspirehep.net/record/848006/files/arXiv:1003.1643.pdf
97.
Sirunyan, AM and others. Electroweak production of two jets in association with a Z boson in protonproton collisions at \(\sqrt{s}= \) 13 TeV. Eur. Phys. J. C78(7), 589 (2018). https://doi.org/10.1140/epjc/s1005201860499
98.
M. Bahr et al., Herwig++ physics and manual. Eur. Phys. J. C 58, 639–707 (2008). arXiv:0803.0883 [hepph]ADSGoogle Scholar
99.
J. Bellm et al., Herwig++ 2.7 release note. arXiv:1310.6877 [hepph]
100.
B. Jäger, S. Schneider, G. Zanderighi, Nexttoleading order QCD corrections to electroweak \(Zjj\) production in the POWHEG BOX. JHEP 09, 083 (2012). arXiv:1207.2626 [hepph]Google Scholar
101.
B. Jäger, G. Zanderighi, Electroweak W\(^+\)W\(^jj\) prodution at NLO in QCD matched with parton shower in the POWHEGBOX. JHEP 04, 024 (2013). arXiv:1301.1695 [hepph]Google Scholar
102.
B. Jäger, A. Karlberg, G. Zanderighi, Electroweak \(ZZjj\) production in the Standard Model and beyond in the POWHEGBOX V2. JHEP 03, 141 (2014). arXiv:1312.3252 [hepph]ADSGoogle Scholar
103.
F. Schissler, D. Zeppenfeld, Parton shower effects on \(W\) and \(Z\) production via vector boson fusion at NLO QCD. JHEP 04, 057 (2013). arXiv:1302.2884 [hepph]ADSGoogle Scholar
104.
M. Cacciari, F.A. Dreyer, A. Karlberg, G.P. Salam, G. Zanderighi, Fully differential vectorbosonfusion Higgs production at nexttonexttoleading order. Phys. Rev. Lett. 115(8), 082002 (2015). arXiv:1506.02660 [hepph]ADSGoogle Scholar
105.
J. CruzMartinez, T. Gehrmann, E.W.N. Glover, A. Huss, Secondorder QCD effects in Higgs boson production through vector boson fusion. Phys. Lett. B 781, 672–677 (2018). https://doi.org/10.1016/j.physletb.2018.04.046 ADSGoogle Scholar
106.
ATLAS Collaboration, M. Aaboud et al., Measurement of the crosssection for electroweak production of dijets in association with a Z boson in pp collisions at \(\sqrt{s}\) = 13 TeV with the ATLAS detector. Phys. Lett. B 775, 206–228 (2017). arXiv:1709.10264 [hepex]
107.
M. Czakon, D. Heymes, A. Mitov, D. Pagani, I. Tsinikos, M. Zaro, Toppair production at the LHC through NNLO QCD and NLO EW. JHEP 10, 186 (2017). arXiv:1705.04105 [hepph]ADSGoogle Scholar
108.
J.M. Lindert et al., Precise predictions for \(V+\) jets dark matter backgrounds. Eur. Phys. J. C77(12), 829 (2017). arXiv:1705.04664 [hepph]ADSGoogle Scholar
109.
D. Contardo, M. Klute, J. Mans, L. Silvestris, J. Butler, Technical proposal for the phaseII upgrade of the CMS detectorGoogle Scholar
Copyright information
© The Author(s) 2018
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Funded by SCOAP3
Authors and Affiliations
Alessandro Ballestrero1Benedikt Biedermann2Simon Brass3Ansgar Denner2View author's OrcID profileStefan Dittmaier4Rikkert Frederix5Pietro Govoni6Michele Grossi78Barbara Jäger9Alexander Karlberg10Ezio Maina111Mathieu Pellen2Email authorGiovanni Pelliccioli111Simon Plätzer12Michael Rauch13Daniela Rebuzzi7Jürgen Reuter14Vincent Rothe14Christopher Schwan4HuaSheng Shao15Pascal Stienemeier14Giulia Zanderighi16Marco Zaro17Dieter Zeppenfeld131.INFN, Sezione di TorinoTurinItaly2.Institut für Theoretische Physik und AstrophysikUniversität WürzburgWürzburgGermany3.Department PhysikUniversität SiegenSiegenGermany4.Physikalisches InstitutAlbertLudwigsUniversität FreiburgFreiburgGermany5.Technische Universität MünchenGarchingGermany6.University and INFN, MilanoBicoccaMilanItaly7.Dipartimento di Fisica and INFN, Sezione di PaviaUniversitá di PaviaPaviaItaly8.IBM Italia s.p.a. Circonvallazione IdroscaloSegrateItaly9.Institute for Theoretical PhysicsUniversity of TübingenTübingenGermany10.PhysikInstitutUniversität ZürichZurichSwitzerland11.Dipartimento di FisicaUniversità di TorinoTurinItaly12.Particle Physics, Faculty of PhysicsUniversity of ViennaViennaAustria13.Institute for Theoretical PhysicsKarlsruhe Institute of Technology (KIT)KarlsruheGermany14.DESY Theory GroupHamburgGermany15.Laboratoire de Physique Théorique et Hautes Énergies (LPTHE), UMR 7589Sorbonne Université et CNRSParis Cedex 05France16.Theoretical Physics DepartmentCERNGeneva 23Switzerland17.NikhefAmsterdamThe Netherlands