betaFeedback

WorksPhysics and Astronomy

Nuclear effects in the deuteron and global QCD analyses

Abstract

We report the results of a new global QCD analysis, which includes deep-inelastic e/μe/μ scattering data off proton and deuterium, as well as Drell-Yan lepton pair production in proton-proton and proton-deuterium collisions and W±/ZW^\pm/Z boson production data from pppp and ppˉp \bar p collisions at the LHC and Tevatron. Nuclear effects in the deuteron are treated in terms of a nuclear convolution approach with bound off-shell nucleons within a weak binding approximation. The off-shell correction is controlled by a universal function of the Bjorken variable xx describing the modification of parton distributions in bound nucleons, which is determined in our analysis along with the parton distribution functions of the proton. A number of systematic studies are performed to estimate the uncertainties arising from the use of various deuterium datasets, from the modeling of higher twist contributions to the structure functions, from the treatment of target mass corrections, as well as from the nuclear corrections in the deuteron. We obtain predictions for the ratios F2n/F2pF_2^n/F_2^p, and d/ud/u, focusing on the region of large xx. We also compare our results with the ones obtained by other QCD analyses, as well as with the recent data from the MARATHON experiment.

I Introduction

An accurate determination of the parton distribution functions (PDFs) in the proton and the neutron is of primary importance for modern high-energy physics, as PDFs determine the leading contribution to the cross sections of various high-energy processes. Since PDFs are not directly observable, they are usually extracted phenomenologically from global QCD analyses to experimental data at large momentum transfer, including lepton deep inelastic scattering (DIS), lepton-pair production (Drell-Yan process), jet production, and W±/ZW^{\pm}/Z boson production in hadron collisions (for a review see, e.g., Ref. [1]). While the abundant data available from a hydrogen target allow a reliable determination of the PDF content of the proton, data from various nuclei—most notably deuterium—are required as effective neutron targets to constrain the parton content of the neutron. Furthermore, a combination of hydrogen and deuterium data has been commonly used to separate the uu and dd quark PDFs, in particular, at large values of Bjorken xx.1

Precision studies require one to address the effects of nuclear environment at the parton level. While the nuclear effects in nuclear PDFs (nPDFs) analyses are usually treated empirically [4–7], a number of physics mechanisms are known to affect the PDFs and the structure functions (SFs) of the bound nucleons (for a review see, e.g., [8–10]). In the region of large xx, the relevant nuclear effects are related to the smearing of the cross sections with the nuclear momentum distribution [11] (Fermi motion), together with the nuclear binding correction [12]. In addition to these corrections, which have kinematical origin, nuclear effects related to the dynamical modification of the internal parton structure have to be addressed in bound nucleons. In Refs. [13, 14], such a modification is related to the off-mass-shell effect, i.e., the dependence of bound nucleon SFs on its virtual mass squared p2=p02−p2p^2=p_0^2-\boldsymbol{p}^2, where p0p_0 and p\boldsymbol{p} are the nucleon energy and momentum, respectively. This dependence is treated in Refs. [13, 14] as a perturbative correction in the nucleon virtuality v=(p2−M2)/M2v=(p^2-M^2)/M^2, relying on the fact that the typical nucleon momentum and energy are small compared to the nucleon mass MM in the nuclear ground state. Within this weak binding approximation, the corresponding nuclear correction is controlled by the SFs’ derivative with respect to p2p^2, which is described in Ref. [14] in terms of a dimensionless function δf(x)\delta f(x). Additional effects related to the meson-exchange currents and the nuclear shadowing are relevant at intermediate and small xx values. A model combining all of these effects has been successfully used to quantitatively explain the observed dependencies on xx, invariant momentum transfer squared Q2Q^2, nuclear mass number AA of the nuclear DIS data in a wide range of targets from 3He{}^{3}\mathrm{He} to 207Pb{}^{207}\mathrm{Pb} [14–16]. The same model also demonstrates an excellent agreement with the magnitude, the xx and mass dependence of the nuclear Drell-Yan (DY) data [17], as well as with the data on the differential cross sections for W±/ZW^{\pm}/Z boson production in proton-lead collisions at the LHC [18].

The off-shell effect is an important contribution to the full nuclear correction. The corresponding function δf\delta f was determined for the isoscalar nucleon from an analysis of nuclear DIS data on the cross-section ratios σA/σd\sigma^A/\sigma^d [14]. The function δf(x)\delta f(x) was also independently extracted together with the proton PDFs in global QCD analyses of proton and deuterium DIS data [19, 20]. The results of Ref. [19] on δf\delta f are consistent with the previous determination from nuclei with A≥4A\geq4 [14]. However, Refs. [20] and [19] strongly disagree on both the values of the function δf\delta f and on the ratio of the dd and uu quark PDFs at large xx. These observations motivate the present study, in which we perform a new global QCD analysis with updated sets of deuterium DIS data. We discuss a number of systematic studies aimed at understanding the uncertainties associated with a number of effects, including the consistency of various deuterium datasets, the treatment of target mass correction, and the modeling of higher-twist contributions and of the nuclear corrections in the deuterium. We also provide our predictions on the ratios F2n/F2pF_2^n/F_2^p and d/ud/u and compare them with the ones from the QCD analyses of Refs. [20, 21, 22, 23], as well as with the recent data from MARATHON experiment [24].

The paper is organized as follows. In § 0.2, we outline the theory framework used in our analysis of the proton and deuterium DIS. In § 0.3, we discuss the data samples and the details of our analysis. In § 0.4, we summarize our results, while in § 0.5, we discuss the uncertainties associated with the use of different deuterium datasets and with the modeling of the structure functions. In § 0.6, we compare our predictions on F2n/F2pF_2^n/F_2^p with MARATHON data and with the results of other QCD analyses, including the d/ud/u ratio. In Appendix § 0.9.2.1, we show the pulls obtained in our fit from different deuterium datasets. In Appendix § 0.9.2.2, we discuss in detail the phase space in the nuclear convolution equations employed in our analysis.

Theory framework

Nucleon structure functions

The inclusive spin-independent electron(muon)-nucleon inelastic cross section is described by two SFs, FT=2xF1F_T = 2xF_1 and F2F_2, which depend on two independent variables, the invariant momentum transfer squared Q2=−q2Q^2 = -q^2 and the dimensionless Bjorken x=Q2/(2p⋅q)x = Q^2/(2p \cdot q), where pp is the nucleon four-momentum and qq is the four-momentum transfer.

A common framework to describe the DIS is the operator product expansion (OPE), which introduces the power series in Q−2Q^{-2} (twist expansion). To the first order, i.e., in the leading twist (LT), the SFs are fully determined by the PDFs. Corrections from the higher-twist (HT) quark-gluon operators should also be supplemented by those arising from the finite nucleon mass (target mass correction, or TMC) [25]. We also note that for the sake of computing the nuclear SFs (see § 0.2 B), the nucleon SFs are required in the off-mass-shell region p2<M2p^2 < M^2, where MM is the nucleon mass. The unpolarized nucleon SFs in the DIS region can then be written as follows

Fi(x,Q2,p2)=FiTMC(x,Q2,p2)+HiQ2,(1)F_i(x,Q^2,p^2)=F_i^{\mathrm{TMC}}(x,Q^2,p^2)+\frac{H_i}{Q^2}, \tag*{(1)}

where i=T,2i = T, 2 and FiTMCF_i^{\mathrm{TMC}} are the corresponding LT SFs corrected for the target mass effect and HiH_i describe the dynamical twist-4 contribution (for brevity, we suppress explicit notation to the twists higher than four). In this study, we consider two different phenomenological HT models: (1) the additive HT model, in which we assume Hi=Hi(x)H_i = H_i(x) and (2) the multiplicative HT model [26], in which HiH_i is assumed to be proportional to the corresponding LT SF, Hi=FiLT(x,Q2)hi(x)H_i = F_i^{\mathrm{LT}}(x,Q^2)h_i(x). The HT terms from both models are addressed in this study.

The LT SFs are computed using the nucleon PDFs and coefficient functions, which are subject to a power series in the QCD coupling constant. The neutron LT SFs are computed in terms of the proton PDFs relying on the isospin symmetry of uu and dd quark PDFs. The isospin relations for the HT terms are not so obvious. By default, we assume Hip=HinH_i^p = H_i^n.2 We also consider the relation hip=hinh_i^p = h_i^n with the multiplicative HT model.

To account for the TMC, we follow the Georgi-Politzer OPE approach [25]. Since the calculation of the nuclear SFs requires the nucleon SFs in the off-shell mass region, we analytically continue the equations of Ref. [25] into the off-shell region by replacing the nucleon mass squared M2M^2 with p2p^2. We have

FTTMC(x,Q2,p2)=x2ξ2γFTLT(ξ,Q2,p2)+2x3p2Q2γ2∫ξ1duu2F2LT(u,Q2,p2),(2)F_T^{\mathrm{TMC}}(x,Q^2,p^2)=\frac{x^2}{\xi^2\gamma}F_T^{\mathrm{LT}}(\xi,Q^2,p^2)+\frac{2x^3p^2}{Q^2\gamma^2}\int_{\xi}^{1}\frac{du}{u^2}F_2^{\mathrm{LT}}(u,Q^2,p^2), \tag*{(2)}
F2TMC(x,Q2,p2)=x2ξ2γ3F2LT(ξ,Q2,p2)+6x3p2Q2γ4∫ξ1duu2F2LT(u,Q2,p2).(3)F_2^{\mathrm{TMC}}(x,Q^2,p^2)=\frac{x^2}{\xi^2\gamma^3}F_2^{\mathrm{LT}}(\xi,Q^2,p^2)+\frac{6x^3p^2}{Q^2\gamma^4}\int_{\xi}^{1}\frac{du}{u^2}F_2^{\mathrm{LT}}(u,Q^2,p^2). \tag*{(3)}

where ξ=2x/(1+γ)\xi= 2x/(1+\gamma) is the Nachtmann variable, and γ=(1+4x2p2/Q2)1/2\gamma= (1+4x^{2}p^{2}/Q^{2})^{1/2}. Note that in Eqs. (2) we drop the terms of order x4p4/Q4x^{4}p^{4}/Q^{4}, which produce numerically small contributions in the considered region. It should be commented that Eq. (2a) and (3) lead to a nonzero SFs at x→1x \to1. However, in practice this violation of the inelastic threshold behavior does not affect the DIS region, which is characterized by high values of the invariant mass WW of the produced hadronic states.

In the off-mass-shell region, the SFs explicitly depend on the nucleon invariant mass squared p2p^{2}. This dependence has two different sources: (i) the terms p2/Q2p^{2}/Q^{2} in Eqs. (2), which lead to power terms at large values of Q2Q^{2} and (ii) nonpower terms from the off-shell dependence of the LT SFs. Following Refs. [13, 14], we note that for computing the nuclear SFs, it would be sufficient to know the proton and the neutron SFs in the vicinity of the mass shell p2=M2p^{2}=M^{2}. We then treat the nucleon virtuality v=(p2−M2)/M2v=(p^{2}-M^{2})/M^{2} as a small parameter and expand SFs in series in vv. To the leading order in vv, we have

FiLT(x,Q2,p2)=FiLT(x,Q2,M2)[1+δfi(x,Q2)v],(4)F_i^{\mathrm{LT}}(x,Q^{2},p^{2})=F_i^{\mathrm{LT}}(x,Q^{2},M^{2})\left[1+\delta f_i(x,Q^{2})v\right], \tag*{(4)}
δfi(x,Q2)=M2∂p2ln⁡FiLT(x,Q2,p2),(5)\delta f_i(x,Q^{2})=M^{2}\partial_{p^{2}}\ln F_i^{\mathrm{LT}}(x,Q^{2},p^{2}), \tag*{(5)}

where FiLTF_i^{\mathrm{LT}} on the right-hand side in Eq. (4) are the structure functions i=T,2i=T,2 of the on-mass-shell nucleon, and ∂p2\partial_{p^{2}} in Eq. (5) denotes the partial derivative with respect to p2p^{2} taken on the mass shell p2=M2p^{2}=M^{2}. According to Eq. (5), the function δfi\delta f_i describes the relative modification of the nucleon LT FiF_i in the vicinity of the mass shell, which is related to the corresponding PDF modification.

In this study, we assume the function δf\delta f to be the same for FTF_T and F2F_2 motivated by the fact that FT≈F2F_T \approx F_2 in the region of large xx. The function δf\delta f drives the nuclear correction associated with the modification of the bound nucleon in the nuclear environment [14]. Detailed studies of nuclear DIS, DY production of the lepton pair and W/ZW/Z boson in Refs. [19] [14, 16, 17, 18] are consistent with no significant scale and nucleon isospin dependencies of δf\delta f. We thus assume the same δf=δf(x)\delta f=\delta f(x) function for the proton and the neutron.

Note that Eq. (4) holds in the vicinity of the mass shell where ∣v∣≪1|v|\ll1. In computing the nuclear SFs, we integrate over the bound nucleon momentum as discussed in § 0.2 B. For kinematics reason, p2<M2p^{2}<M^{2} and v<0v<0 for bound nucleons. Using the results of Ref. [14], we have δf∼1\delta f\sim1 at large x>0.6x>0.6. Then the off-shell correction in Eq. (4) is large and negative for v∼−1v\sim-1, and the off-shell SFs may be negative in this region. Since the values ∣v∣≳1|v|\gtrsim1 are outside of the region of applicability of the linear approximation in vv, Eq. (4), we consider the following model in the full region of vv:

FiLT(x,Q2,p2)=FiLT(x,Q2,M2)exp⁡[δf(x)v].(6)F_i^{\mathrm{LT}}(x,Q^{2},p^{2})=F_i^{\mathrm{LT}}(x,Q^{2},M^{2})\exp[\delta f(x)v]. \tag*{(6)}

This equation ensures the positivity of SFs in the off-shell region, and for a small off-shell correction Eq. (6) is identical to Eq. (4). In the study of the deuteron SFs, we consider both Eq. (4) and Eq. (6).

Deuteron structure functions

We assume that the nuclear DIS in the region x>0.1x>0.1 is dominated by the incoherent scattering off the bound protons and neutrons and consider the process in the target rest frame. The deuteron structure functions can be written as follows [27] [14]:

Fid(x,Q2)=∫d3p ∣Ψd(p)∣2Kij[Fjp(x′,Q2,p2)+Fjn(x′,Q2,p2)],(7)F_i^d(x,Q^{2})=\int d^{3}p\,|\Psi_d(\boldsymbol{p})|^{2}K_{ij}\left[F_j^p(x',Q^{2},p^{2})+F_j^n(x',Q^{2},p^{2})\right], \tag*{(7)}

where i,j=T,2i,j=T,2, and we assume a summation over the repeated subscript jj. The integration is performed over the bound nucleon momentum p\mathbf{p}, and Ψd(p)\Psi_d(\mathbf{p}) is the deuteron wave function in the momentum space, which is normalized as

∫d3p ∣Ψd(p)∣2=1.(8)\int d^3\mathbf{p}\,|\Psi_d(\mathbf{p})|^2=1. \tag*{(8)}

Because of the energy-momentum conservation, the four-momentum of the struck proton (neutron) is p=(Md−M2+p2,p)\mathbf{p}=(M_d-\sqrt{M^2+\mathbf{p}^2},\mathbf{p}), where MdM_d is the deuteron mass, and MM is the mass of residual nucleon [M=MnM=M_n for the proton contribution and M=MpM=M_p for the neutron contribution in Eq. (7)]. We use a coordinate system in which the momentum transfer q\mathbf{q} is antiparallel to the zz axis, and pzp_z and p⊥p_\perp are the longitudinal and transverse component of the nucleon momentum, p2=p02−p2\mathbf{p}^2=p_0^2-\mathbf{p}^2 and x′=Q2/(2p⋅q)x'=Q^2/(2p\cdot q) are the invariant mass and the Bjorken variable of the off-shell nucleon, respectively. The kinematic factors KijK_{ij} are [14]

KTT=(1+γpzM),KT2=2x′2p⊥2Q2,(9)\begin{aligned} K_{TT}&=\left(1+\frac{\gamma p_z}{M}\right),& K_{T2}&=2\frac{x'^2p_\perp^2}{Q^2}, \tag*{(9)} \end{aligned}
K2T=0,K22=(1+γpzM)(1+x′2(4p2+6p⊥2)Q2)1γ2,(10)\begin{aligned} K_{2T}&=0,& K_{22}&=\left(1+\frac{\gamma p_z}{M}\right)\left(1+\frac{x'^2(4p^2+6p_\perp^2)}{Q^2}\right)\frac{1}{\gamma^2}, \tag*{(10)} \end{aligned}

where γ=(1+4x2M2/Q2)1/2\gamma=(1+4x^2M^2/Q^2)^{1/2}. Note that Eq. (7) and Eq. (8) are the result of a series expansion of relativistically covariant operators in the parameters p/Mp/M and (p0−M)/M(p_0-M)/M to order p2/M2p^2/M^2 (for more detail, see [13, 14] and Appendices B and C of Ref. [28]). The factor 1+γpz/M1+\gamma p_z/M in Eq. (8) describes the change in the virtual photon flux for a bound nucleon with the momentum pp compared to the corresponding flux for the nucleus at rest. Note also the term KT2∼x′2p⊥2/Q2K_{T2}\sim x'^2p_\perp^2/Q^2 resulting from a mixing effect between the longitudinal and transverse structure functions at finite values of Q2Q^2, which is due to the transverse motion of the bound nucleon.

Assuming no p2p^2 dependence of the nucleon structure functions, in the limit Q≫MQ\gg M, Eq. (7) reduces to the standard convolution of the nucleon SFs with the nucleon distribution over the light-cone momentum y=(p0+pz)/My=(p_0+p_z)/M in the deuteron. In the presence of an off-shell p2p^2 dependence we have a generalized convolution, which involves the integration over the light-cone momentum yy and the nucleon virtuality p2p^2 [13]. The phase space at finite Q2Q^2 used in Eq. (7) is discussed in more detail in Appendix B.

In the region x<0.1x<0.1, the corrections due to the meson-exchange currents and the nuclear shadowing, at even smaller values of x≪0.1x\ll0.1, are relevant. In this study, while focusing on x>0.1x>0.1, we treat these effects following Refs. [14, 17].

Off-shell Function within Global QCD Analysis

Data samples

The present study is an update of our former analysis [19] based on the data on the DIS of charged leptons off hydrogen and deuterium combined with the ones on W±/ZW^\pm/Z boson production at hadron colliders. The latter samples allow the separation of the uu and dd quark distributions in a wide range of xx that, in turn, provides a basis for studying nuclear effects in the deuteron for the DIS structure functions. The deuterium datasets employed for this purpose are listed in Table I. They comprise the ones used in the analysis of [19] supplemented by the most recent results on σd/σp\sigma^d/\sigma^p by the MARATHON experiment at JLab [24]. Due to the increased energy of the upgraded JLab beam, the MARATHON data cover a much wider kinematics as compared to the earlier JLab

FacilityExperimentReferenceBeamBeam energy (GeV)ObservableNormalization factorNormalization error(s) (%)χ2/NDP\chi^2/\mathrm{NDP}
SLACE49a[29, 30]ee11 ÷\div 19.5d2σddE′dΩ\frac{d^2\sigma^d}{dE'd\Omega}0.988(10)2.1325/59
”E49b””4.5 ÷\div 18”0.996(10)”187/145
”E87””8.7 ÷\div 20”1.000(9)”114/109
”E89b[30, 32]”10.4 ÷\div 19.5”0.987(9)”52/72
”E139[30, 33]”8 ÷\div 24.5”1.002(9)”8/17
”E140[30, 34]”3.7 ÷\div 19.5”11.725/26
CERNBCDMS[35]μ\mu100 ÷\div 280d2σddxdQ2\frac{d^2\sigma^d}{dxdQ^2}0.989(7)3273/254
”NMC[36]”90 ÷\div 280F2d/F2pF_2^d/F_2^p1<0.15<0.15155/165
DESYHERMES[37]ee27.6σd/σp\sigma^d/\sigma^p11.421/30
JLabE00-116[38]”5.5d2σddE′dΩ\frac{d^2\sigma^d}{dE'd\Omega}0.981(10)1.75208/136
”BoNuS[39]”4.2, 5.2F2n/F2dF_2^n/F_2^d0.97(9)7 ÷\div 1090/63
”MARATHON[24]”10.6σd/σp\sigma^d/\sigma^p10.558/7
Total1166/1083

Table 1. The list of DIS data on the deuterium target employed in the present analysis alongside with the values of χ2/NDP\chi^2/\mathrm{NDP} and normalization factors obtained in the fit in comparison with the experimentally determined normalization errors.

BoNuS experiment [39].4 Besides, a dedicated study performed by MARATHON allowed one to reduce the normalization uncertainty in its measurements to unprecedented level of 0.55%. This guarantees a superior statistical significance of the MARATHON data over both the original BoNuS sample [39] and the results of the study [40] based on the BoNuS measurements. The results on F2d/F2NF_2^d/F_2^N derived in Ref. [40] from the BoNuS data on F2n/F2dF_2^n/F_2^d using a parametrization of F2pF_2^p were employed in our earlier study [19]. However, since they are sensitive to model assumptions about the F2pF_2^p shape, in the present study, we select the original BoNuS data in order to reduce the model dependence of the analysis. To provide a complete representation of the relevant data, we also add to the fit the DIS data collected in the Jlab-E00-116 [38] and DESY-HERMES [37] experiments. Finally, we replace the deuteron NMC data [41] used in Ref. [19] by more recent measurements [36] performed with a higher luminosity. In Figure 1, we illustrate the accuracy of various datasets by plotting the ratio F2d/F2pF_2^d/F_2^p in the range 0.1<x<0.50.1 < x < 0.5 and for Q2=14x (GeV2)±10%Q^2 = 14x\ (\mathrm{GeV}^2) \pm10\% from various measurements [24, 31, 36, 37].5 This selection of the Q2Q^2 band is motivated by kinematics of the MARATHON experiment [24].

Figure 1. Data on F2d/F2pF_2^d/F_2^p (or σd/σp\sigma^d/\sigma^p) for 0.1<x<0.50.1 < x < 0.5 from various experiments. Data legend is shown in the plot. The data points were selected within interval Q2=14xQ^2 = 14x (GeV2^2) ±10%\pm10\% to facilitate the comparison with Ref. [24]. The inner error bar shows the statistical and systematic error taken in quadrature, while the outer error bar in addition includes the normalization error listed in Table I.

Whenever possible, we select the data on the cross sections and their ratios rather than on the structure functions F2F_2. This makes the modeling more involved, however, allows for a consistent account of the contribution from the structure function FLF_L, since the various experiments do not follow a common convention on the shape of FLF_L when extracting F2F_2 from the cross-section measurements. As an exception, the BoNuS [39] and the NMC [36] data are taken in the form of the ratios F2n/F2dF_2^n/F_2^d and F2d/F2pF_2^d/F_2^p, respectively, as the cross-section results have not been released by these experiments. Such an inconsistency can be, however, justified since the FLF_L contribution to great extent cancels out in the ratios. To ensure a perturbative QCD description of the leading-twist terms in the DIS structure functions, we impose a general cut of Q2>2.5 GeV2Q^2 > 2.5\ \mathrm{GeV}^2 and W2>3 GeV2W^2 > 3\ \mathrm{GeV}^2. For the BoNuS data [40] used in our earlier study [19], a relaxed cut of Q2>1.5Q^2 > 1.5 GeV2^2 was selected in order to increase the statistical significance of this sample. In the present analysis, which includes the precision MARATHON data, this exemption is not applied and the BoNuS data [39] are considered within a common framework.

Information about the point-to-point correlation of systematic errors in the data is taken into account in the fit whenever available. In particular, a detailed breakdown of the systematic uncertainties over independent sources is provided for the SLAC, CERN-BCDMS, CERN-NMC, and JLab-BoNuS experiments. For the Jlab-E00-116 and MARATHON datasets, only the overall systematic uncertainty is published, and in the present fit, it is combined in quadrature with the statistical (uncorrelated) uncertainties. The systematic uncertainties of DESY-HERMES measurements are separated by sources; however, no information about their point-to-point correlation was provided. For this reason, we select for our fit the DESY-HERMES data on the ratio σd/σp\sigma^d/\sigma^p, where the correlated uncertainties partially cancel. The remaining systematic uncertainty, except of the normalization one, are combined with the statistical uncertainty, in line with the approximation adopted in the DESY-HERMES analysis of their own data [37].

The normalization uncertainty, a peculiar case of systematic errors, often dominates the uncertainty of the datasets considered. Furthermore, the normalization factors for the available SLAC-E49a, E49b, E87, E89b, E139, and CERN-NMC datasets were estimated by comparing them to the measurements of the SLAC-E140 experiment with the normalization uncertainty of 1.7%. Following a similar approach, we release the normalization factors of those data and determine such factors from a fit simultaneously with other parameters. Furthermore, this procedure is also applied to the CERN-BCDMS and Jlab-E00-116 data allowing for improvement of their instrumentally determined normalizations. For the CERN-NMC and DESY-HERMES data, which were taken in the form of ratios, the impact of the normalization uncertainty is greatly reduced; therefore, their normalizations were kept fixed. A similar treatment was applied to the MARATHON data, which have a very accurate luminosity monitoring and for which we also avoid normalization tuning. This allows us to use those data for the calibration of the other datasets, in addition to the SLAC-E140 data set.

Analysis setup

The leading-twist PDFs, which are necessary for computing the nucleon SFs Eq. (1), are parametrized using the shape employed in the ABMP16 fit [43] and in our earlier analysis [19]. The DIS SFs are treated as outlined in § 0.2 A. The functions HT(x)H_T(x) and H2(x)H_2(x), which describe the HT contributions, are treated independently and are parametrized in a model-independent form of spline polynomials interpolating between the points x=(0,0.1,0.3,0.5,0.7,0.9,1)x = (0, 0.1, 0.3, 0.5, 0.7, 0.9, 1) with the values (HT coefficients) determined on this grid. We assume the HT terms to be independent of the nucleon isospin state. The Q2Q^2 dependence of the LT component of the nucleon SFs was computed taking into account NNLO perturbative QCD (pQCD) corrections, while for the HT coefficients, possible pQCD effects have been neglected.

The nuclear effects in the deuteron are accounted by Eq. (7) with the off-shell correction governed by Eq. (5). The deuteron AV18 wave function is used [44, 45]. The function δf(x)\delta f(x) is determined along with the proton PDFs and HTs in a fit to the deuterium data listed in Table I and the proton data from Table II in Ref. [19]. This function is parametrized as a polynomial:

δf(x)=c0+c1x+c2x2.(11)\delta f(x) = c_0 + c_1 x + c_2 x^2. \tag*{(11)}

Note that Eq. (7) describes the nuclear corrections driven by the momentum distribution, the nuclear binding, and the off-shell effect, which dominate in the present analysis. We also verified [19] that the other nuclear effects, such as the meson-exchange currents and the nuclear shadowing, are within experimental uncertainties, and for this reason, they are not considered in the present analysis.

Results

We simultaneously fit the parameters of the off-shell function δf\delta f with those of the PDFs and HTs in order to provide a consistent separation of the various contributions to the SFs by exploiting the broad xx and Q2Q^2 coverage of available data. The resulting data normalization factors and χ2\chi^2 values corresponding to various deuterium datasets are listed in Table I, and the parameters of the δf(x)\delta f(x) function are c0=−0.16±0.11c_0 = -0.16 \pm0.11, c1=−2.04±0.73c_1 = -2.04 \pm0.73, and c2=4.86±1.13c_2 = 4.86 \pm1.13. This function is shown in Figure 2, together with the results of other determinations of this quantity from Refs. [19, 20] [14]. For all data points included in our fit we have χ2/d.o.f.=4842/4044\chi^2/\mathrm{d.o.f.} = 4842/4044. The present results are in a good agreement with both our former global QCD analysis [19] and the analysis of Ref. [14], in which the function δf(x)\delta f(x) was determined from a global fit to the data on the ratios σd/σp\sigma^d/\sigma^p for the DIS cross section off nuclear targets with mass number 4≤A≤2084 \leq A \leq208 using the proton and neutron SFs of Ref. [46]. However, our results are in a strong disagreement with those of Ref. [20]. Below we trace possible reasons of the discrepancy with Ref. [20] by verifying the differences in the experimental datasets and in the underlying model.

Figure 2. The 1σ1\sigma uncertainty band for the off-shell function δf\delta f as a function of xx obtained in the present analysis (shaded cyan area) in comparison with the results of the earlier AKP [19] (right-tilted hash area), KP [14] (left-tilted hash area), and CJ15 [20] (dash-dotted curve) fits.

In Figure 3 we compare our results with the recent measurement of σd/σp\sigma^d/\sigma^p by the MARATHON experiment [24]. Also shown are the predictions from the CT18 [21], MSHT20 [22], NNPDF4.0 [23], and CJ15 [20] QCD analyses. Note that the CJ15 analysis was performed to the NLO approximation, while all others were done to the NNLO one. The CT18, MSHT20, and NNPDF4.0

Data on $\sigma^{d}/\sigma^{p}$ from the MARATHON measurement compared with the present analysis and global QCD analyses

Figure 3. Data on σd/σp\sigma^{d}/\sigma^{p} from MARATHON measurement [24] compared with corresponding 1σ1\sigma band of present analysis (left-tilted area). Also shown are the predictions of various global QCD analyses: CT18 [21] (dots), MSHT20 [22] (connected dots), NNPDF4.0 [23] (short dashes), and CJ15 [20] (long dashes).

structure functions are computed in the 3-flavour scheme using the code OPENQCDRAD (version 2.1) [47] combined with the LHAPDF (version 6) PDF grids [48, 49] CT18NNLO, MSHT20nnlo_nf3, and NNPDF40_nnlo_pch_as_01180_nf_3, respectively. We use F2d=F2p+F2nF_{2}^{d}=F_{2}^{p}+F_{2}^{n} for CT18 and NNPDF4.0 in Figure 3, as those analyses do not account for the deuteron correction. Both MSHT20 and CJ15 account for the deuteron effect in their PDF fits. For MSHT20, we take F2d=Rd(F2p+F2n)F_{2}^{d}=R_{d}(F_{2}^{p}+F_{2}^{n}) with the correction factor RdR_{d} obtained in the NNLO global QCD fit of Ref. [22]. For CJ15 we use their results obtained from Ref. [24]. All the predictions are in agreement with MARATHON σd/σp\sigma^{d}/\sigma^{p} data within uncertainties in the region about x=0.3x=0.3 and somewhat overshoot the data for xx values about 0.2 and 0.4.6

The quality of our fit for the newly added BoNuS data on F2d/F2dF_{2}^{d}/F_{2}^{d} [39], the cross-section measurements from JLab-E00-116 experiment [38], NMC data on F2d/F2pF_{2}^{d}/F_{2}^{p} [36], and the HERMES data on σd/σp\sigma^{d}/\sigma^{p} [37] are illustrated in detail in Figure 8 to Figure 11 in Appendix § 0.9.2.1. In general, we observe a good agreement of the fit and the data with no regular and/or statistically significant trend in the pulls.

Figure 8. The pulls of the present analysis for the BoNuS data on the ratio F2n/F2dF_2^n/F_2^d [39] vs W2W^2 displayed in the panels of Q2Q^2 bins. The inner error bars reflect the statistical and uncorrelated systematic uncertainties of data, while the outer bars are the total experimental error. The shaded region shows ±1σ\pm1\sigma uncertainty of the fit.

Image not extracted

Figure 11. The pulls of the DESY-HERMES data [37] on σd/σp\sigma^{d}/\sigma^{p} vs Q2Q^{2}. The panels correspond to the bins of xx. Notations are similar to those in Figure 10.

Figure 10. The pulls of the NMC data [36] on F2d/F2pF_2^d/F_2^p vs Q2Q^2. The panels correspond to the bins of xx. Notations are similar to those in Figure 8.

Systematic Studies

We verify the stability of our results on δf\delta f by performing a number of fits with a modified ansatz. Some of these modifications, like the parametrization of higher-twist terms (additive vs multiplicative form) and the off-shell correction [a linear dependence on vv in Eq. (4) vs exponentiated form of Eq. (6)], reflect uncertainties in the theory framework of the fit. The other modifications are motivated by other studies in the field and aimed to facilitate a comparison with those studies. The impact of the modifications considered is summarized in Figure 4, and their detailed description is given below in this section. Note that for all considered modifications of the fit, the data normalization factors are kept at the values of Table I, thus allowing us to avoid the interplay with the data shift.

Figure 4. The 1σ1\sigma uncertainty band on δf\delta f from our nominal fit same as in Figure 2 (shaded cyan area) compared with the results of a modified fit framework. Left panel: higher twists (HTs) parametrized in a multiplicative ansatz by Eq. (12) (dashed curve); using approximate TMC scheme by Eq. (13) (see Eq. (61) of Ref. [50]) and the data on F2F_{2} (dash-dotted curve); exponential form of the off-shell correction by Eq. (6) (dotted curve). Right panel: using NMC cross-section data [41] instead of NMC F2d/F2pF_{2}^{d}/F_{2}^{p} data [36] in the nominal fit (right-tilted hash area); released normalization of MARATHON σd/σp\sigma^{d}/\sigma^{p} data (left-tilted hash area).

NMC data choice

Our fit includes the NMC data on the ratio F2d/F2pF_{2}^{d}/F_{2}^{p} [36]. These data are derived from the cross-section data assuming the same ratio R=σL/σTR = \sigma_{L}/\sigma_{T} for the proton and deuteron, which was verified experimentally by the NMC with a good accuracy. Alternatively, in our former analysis [19], the cross-section data for the deuterium target [41] have been employed instead. In order to verify the impact of the particular NMC data choice, we perform a variant of our nominal fit with the NMC data on F2d/F2pF_{2}^{d}/F_{2}^{p} replaced with the cross-section data on the deuterium target [41]. This fit results in χ2/d.o.f.=4693/3988\chi^{2}/\mathrm{d.o.f.} = 4693/3988. As can be seen in Figure 4, the difference in the function δf\delta f obtained in these two versions of the fit is significant only for x≲0.4x \lesssim0.4. In this region, the uncertainty in the value of δf\delta f extracted from the deuterium cross-section data is somewhat larger, due to less statistical significance of this sample. Meanwhile, the error bands for the two determinations almost overlap with each other and with the determination based on the heavy-nuclear data [14] (see also Figure 2). Note also that δf(x)\delta f(x) obtained in the fit with the NMC deuterium cross-section data is almost identical to our earlier result [19].

MARATHON data normalization

The normalization of the ratio σd/σp\sigma^{d}/\sigma^{p} in the MARATHON experiment is determined experimentally with a very high accuracy of 0.55%. Nonetheless, the MARATHON data go somewhat lower than the other samples used in our analysis, cf. Figure 1. To quantify this tension, we perform a variant of fit with the normalization of the MARATHON σd/σp\sigma^{d}/\sigma^{p} data released and adjusted simultaneously with other fit parameters. The normalization factor of 1.014(4) obtained in this way is at about 2σ2\sigma off the nominal value 1.0000(55). However, this renormalization of the MARATHON σd/σp\sigma^{d}/\sigma^{p} ratio has a negligible impact on the value of δf\delta f extracted from the data, and the corresponding change in its value is well within 1σ1\sigma uncertainty band. For this fit, we have χ2/d.o.f.=4834/4044\chi^{2}/\mathrm{d.o.f.} = 4834/4044.

Higher-twist correction

In our analysis we compute the structure functions following Eq. (1) with an additive model of the higher-twist (HT) terms motivated by the OPE. However, a multiplicative HT model is often used in the literature (see, e.g., Refs. [20, 26, 51]):

Fi(x,Q2,p2)=FiTMC(x,Q2,p2)+FiLT(x,Q2)hi(x)/Q2,(12)F_i(x,Q^2,p^2) = F_i^{\mathrm{TMC}}(x,Q^2,p^2) + F_i^{\mathrm{LT}}(x,Q^2)h_i(x)/Q^2, \tag*{(12)}

where i=2,Ti = 2,T. To compare the additive and multiplicative HT models, one should confront the coefficients HiH_i in Eq. (1) with the corresponding product FiLThiF_i^{\mathrm{LT}}h_i in Eq. (12). These terms have a different Q2Q^2 dependence driven by assumptions about anomalous dimensions of the HT operators: For the additive form, they are neglected, and for the multiplicative one, they are similar to the leading-twist case. This difference is important at large xx, where the leading-twist evolution and the TMC are most significant (for illustration, see Figure 5). The same trend appears in the determination of the off-shell function, which is sensitive to the assumed HT model at x≳0.5x \gtrsim0.5, although the shape of δf(x)\delta f(x) (left panel of Figure 5) does not change essentially under the HT model variation. For the fit with multiplicative HT model χ2/d.o.f.=4798/4044\chi^{2}/\mathrm{d.o.f.} = 4798/4044.

Twist-4 corrections and target mass corrections for the proton structure function

Figure 5. Left panel: The 1σ1\sigma bands of twist-4 correction to the proton F2F_2 using the additive HT model [Eq. (1), gray band] and multiplicative HT model [Eq. (12)]. The multiplicative HT is shown for Q2=5Q^2 = 5 (right-tilted band) and 20 GeV220\ \mathrm{GeV}^2 (left-tilted band) in order to illustrate their residual Q2Q^2 dependence. Also shown are the corresponding CJ best fit results [20] (dashed and dashed-dotted curves). Right panel: Target mass correction for the proton F2F_2 in terms of Q2[F2TMC(x,Q2)−F2LT(x,Q2)]Q^2[F_2^{\mathrm{TMC}}(x,Q^2)-F_2^{\mathrm{LT}}(x,Q^2)] suitable for comparison with HT in the left panel. Georgi-Politzer TMC [25] is shown for Q2=5Q^2 = 5 and 20 GeV220\ \mathrm{GeV}^2 (solid and long-dashed curves). The short-dashed and dotted lines show an empirical TMC by Eq. (13), which was used in analysis [20].

Target mass effect

The target mass effects are taken into account in our analysis using the Georgi-Politzer formalism in the off-shell region [see Eqs. (2)]. In other global QCD fits, the TMC is either neglected or treated differently, assuming p2=M2p^2 = M^2. For example, in the CJ15 analysis [20, 52] the TMC is accounted using an approximation to Eq. (2) (see Eq.(61) in Ref. [50]):

F2TMC(x,Q2)≈(1+γ)24γ3F2LT(ξ,Q2)[1+3(γ−1)γ(1−ξ)2].(13)F_2^{\mathrm{TMC}}(x,Q^2) \approx\frac{(1+\gamma)^2}{4\gamma^3}F_2^{\mathrm{LT}}(\xi,Q^2)\left[1+\frac{3(\gamma-1)}{\gamma}(1-\xi)^2\right]. \tag*{(13)}

To verify the sensitivity of our results to a particular TMC treatment, we perform a fit using Eq. (13) instead of Eq. (2b). A similar approximation for FTTMCF_T^{\mathrm{TMC}} is not available in the formalism of Ref. [50]. For this reason, in our modified fit, we employ data on F2F_2, instead of cross-section data, for the SLAC, CERN-BCDMS, CERN-NMC, and JLab-E00-116 experiments. The DESY-HERMES and MARATHON data on the cross-section ratio are treated using the relation σd/σp≈F2d/F2p\sigma^d/\sigma^p \approx F_2^d/F_2^p since R=σL/σTR = \sigma_L/\sigma_T is similar for proton and deuteron at moderate xx [36]. Finally, for the HERA cross-section data, we take FTF_T according to Eq. (2a). Such an approach does not cause a serious inconsistency due to the HERA data being localized at small xx, where the TMC is generally small. For this variant of the fit, we have χ2/d.o.f.=4831/4050\chi^2/\mathrm{d.o.f.} = 4831/4050. The best fit result on δf(x)\delta f(x) is shown by the dashed-dotted curve in Figure 4, which is within the 1σ1\sigma band of our nominal fit for almost all xx values but a region around x=0.4x = 0.4. In this region Eq. (13) gives a rather poor approximation on Eq. (2b), as illustrated in Figure 5 (right panel). The maximal difference between the two implementations is observed in the range 0.2≤x≤0.50.2 \le x \le0.5. Note that in the same region we observe opposite deviations between the HT contributions (left panel of Figure 5) obtained by the CJ15 analysis and by our fit using the same multiplicative HT form and the TMC from Eq. (2) отк.

Exponential model of off-shell correction

In our present study, we compute the off-shell correction using Eq. (4). As noted in § 0.2 A, at high nucleon momentum ∣p∣≳M|p| \gtrsim M in the nuclear convolution Eq. (7) the off-shell structure function from Eq. (4) may be negative, thus signaling a violation of the linear approximation in vv. To verify the relevance of Eq. (4), we performed a fit using the exponential model of off-shell correction from Eq. (6), which gives a positive SF at any value of vv. The resulting function δf(x)\delta f(x), shown in Figure 4, is identical to our nominal fit result within the fit uncertainties and χ2/d.o.f.=4847/4044\chi^2/\mathrm{d.o.f.} = 4847/4044 for this variant of the fit.

Discussion

In § 0.4 and V, we discussed determination of the quantity δf(x)\delta f(x) from a global QCD analysis of the most recent DIS data off hydrogen and deuterium combined with the ones on WW- and ZZ-boson production at hadron colliders. We reiterate that δf(x)\delta f(x) describes the modification of the nucleon PDFs in the off-shell region for bound nucleons and, as such, it is expected to be a universal quantity independent of the nucleus considered. This quantity has a considerable impact on the nuclear corrections obtained within the nuclear convolution approach and is required to describe available nuclear DIS [14, 16] and Drell-Yan data [17, 18]. The results from our current analysis are in good agreement with our previous analysis [19], as well as with the study of nuclear ratios of DIS cross sections for nuclear targets with nuclear number A≥3A \ge3 [14, 16] (Figure 2). However, our results disagree with the ones of Ref. [20].

In our previous analysis [19], we evaluated the uncertainties associated with the modeling of the deuteron wave function and with the use of different datasets. In § 0.5, we performed additional systematic studies on both the input ansatz and the datasets in order to verify our results and to further investigate the observed discrepancies with the analysis of Ref. [20]. In all cases, our results on δf\delta f are stable against the modifications of the fit considered, and the corresponding variations are generally consistent with the quoted uncertainties. We find that while δf\delta f has some sensitivity to the implementation of the HT corrections (i.e., additive vs multiplicative) at large Bjorken xx, its shape is essentially unchanged (Figure 4). In general, the systematic uncertainties related to the use of different deuterium datasets, in particular NMC cross-section data vs NMC F2d/F2pF_2^d/F_2^p ones, are comparable to the ones related to the input model assumptions.

We further verify our results by comparing our predictions on the ratio Rnp=F2n/F2pR_{np}=F_2^n/F_2^p with the recent MARATHON data [24], which were not included in our fits. The calculations are performed for the MARATHON kinematics, which is roughly consistent with Q2=14xQ^2=14x (GeV2^2), and are shown in Figure 6. Our independent predictions for RnpR_{np} are in excellent agreement with the MARATHON measurement over the entire xx range available.

Figure 6. Data from the MARATHON measurement of F2n/F2pF_2^n/F_2^p [24] compared with the predictions of the present analysis. In the left panel, shown are the predictions from recent PDF analyses: CT18 [21] (dash-dotted line), MSHT20 [22] (dotted line), NNPDF4.0 [23] (short dashes), CJ15 [20] (long dashes), and JAM21 [51] (dash-double-dot). In the right panel, shown are our predictions based on modified fit results discussed in § 0.5.

In Figure 6 (left panel), we show the predictions on RnpR_{np} obtained from the QCD analyses CJ15, MSHT20, NNPDF4.0, CT18, and JAM21. The structure functions are computed as described in § 0.4. The calculations include the target mass corrections of Ref. [25]. The CJ15 prediction is obtained from Figure 3 of Ref. [24]. At large Bjorken xx (x>0.6x > 0.6), significant differences are observed. While the predictions from CJ15, MSHT20, and NNPDF4.0 are consistent with each other, they differ substantially from CT18 and our results. Note that different assumptions on the HT contributions are used for the various calculations: additive HT for our results, multiplicative HT for CJ15, and no HT (only LT term) for the others. To verify the impact of the dd-quark PDF on such differences, we compare the corresponding predictions for the Rdu=d/uR_{du}=d/u ratio for the MARATHON kinematics using the PDFs from the LHAPDF library [49].7 Figure 7 shows our 1σ1\sigma

x≈0.03x \approx0.03, because of fermion number conservation. While this effect is relevant for the electroweak studies at LHC, we leave its analysis for future, as in the present work we focus on the region relevant for the MARATHON measurement. uncertainty band together with the central values obtained for the other analyses. The uncertainties on these latter are relatively large for x>0.6x > 0.6 due to the use of tight WW cuts, which effectively exclude high-xx DIS data from the QCD analysis.8 Figure 6 indicates that the CJ15-, MSHT20-, and NNPDF4.0-based predictions are in clear disagreement with the MARATHON RnpR_{np} data.

The effect of the variations of the model assumptions discussed in § 0.5 on our predictions for RnpR_{np} is also illustrated in Figure 6 (right panel). Although in most cases the results are consistent with the nominal one within the quoted uncertainties, a significant deviation is apparent for the variant with multiplicative HT contributions from Eq. (12). In this latter case our predictions for RnpR_{np} appear to be closer to the CJ15-, MSHT20- and NNPDF4.0-based calculations. However, the corresponding RduR_{du} ratio shown in Figure 7 (right panel) is consistent with our nominal fit. These observations indicate that the MARATHON data are sensitive to the HT contributions in the region x>0.6x > 0.6.

Figure 7. The 1σ1\sigma band of the PDF ratio d/ud/u vs. xx computed at variable scale μ2=14x (GeV2)\mu^2=14x\ (\mathrm{GeV}^2) (left-tilted area). Also shown are the results from PDF analyses, same as in Figure 6 (left panel). The right-tilted (yellow) area shows the nucleon resonance region of W≲2 GeVW \lesssim2\ \mathrm{GeV}. All shown analyses but Ref. [20] are to NNLO pQCD order. Right panel shows our results from modified fits, same notations as in Figure 6.

The comparison with the CT18-based predictions is instructive, as they provide a good LT description of the MARATHON data on RnpR_{np} (Figure 6) without any HT contributions. This agreement is explained by a much larger value of the RduR_{du} ratio at x>0.6x > 0.6 compared to the other QCD analyses, as illustrated in Figure 7 (left panel).9 By contrast, we obtain a good description of MARATHON data on RnpR_{np} with Rdu→0R_{du} \to0 as x→1x \to1 and a sizable HT contribution, which is maximal at x∼0.7x \sim0.7 (Figure 5). In general, W±W^{\pm} boson production and the corresponding lepton asymmetries from D0 and LHCb data at high rapidity could help to clarify the differences observed on the dd-quark distribution at large Bjorken xx [53, 54]. However, the calculation of the cross-section for WW-boson production in the NNLO pQCD approximation suffers from uncertainties in the available numerical codes [55].

As discussed in § 0.2 A, we assume isoscalar HT terms in the additive HT model (Hp=HnH^p = H^n). We also assume hp=hnh^p = h^n for the multiplicative HT model. However, in the latter case the overall HT correction is different for protons and neutrons since hh is multiplied by the corresponding LT structure functions. Note that for the multiplicative HT model the contribution from hh cancels out in the ratio RnpR_{np}, which has a similar behavior as the corresponding LT approximation. Conversely, in the case of the additive HT model used in our nominal fit the ratio RnpR_{np} receives a finite HT correction. The MARATHON RnpR_{np} data seem to prefer a common additive HT contribution (Figure 6) for both the neutron and the proton. Although the disagreement observed for x>0.6x > 0.6 with the multiplicative HT form may be mitigated by the introduction of an explicit isospin dependence in the hh terms, such an effect could result in observable deviations for other DIS data sensitive to isospin effects.

A recent paper [51] reports the results of a global QCD analysis (JAM21) including MARATHON data on the cross-section ratios σd/σp\sigma^d/\sigma^p and σ3H/σ3He\sigma^{3\mathrm{H}}/\sigma^{3\mathrm{He}} for the three-body nuclei, as well as the previous measurement of σ3He/σd\sigma^{3\mathrm{He}}/\sigma^d from E03-103 at JLab [56]. The study includes multiplicative HT corrections and a calculation of nuclear effects based on the convolution approach supplemented by off-shell corrections. However, the treatment of these latter corrections is rather different with respect to our implementation. In the JAM21 analysis, the off-shell correction depends on both the specific nucleus considered and on the isospin of the target nucleon (different for protons and neutrons), resulting in multiple functions that are extracted from data. In particular, the need of an explicit isovector contribution in the off-shell functions is advocated to describe the MARATHON data. Our analysis indicates that this result may be affected by the assumption of multiplicative HT terms. The result of the JAM21 fit on RnpR_{np} is shown in the left panel of Figure 6 and appears to be in disagreement with the MARATHON F2n/F2pF_2^n/F_2^p data [24].

The determination of δf\delta f described in this paper is based on deuterium DIS data and is therefore only sensitive to the isoscalar combination F2p+F2nF_2^p + F_2^n. Our results are in a good agreement with the study of the ratios of nuclear DIS cross sections σA/σd\sigma^{A}/\sigma^{d} with nuclear number A≥3A \ge3, in which the nuclear EMC effect was successfully described in terms of a nuclear convolution approach with a universal off-shell function δf(x)\delta f(x) independent of the nucleus. Although the model [14] could naturally incorporate an isospin dependence into the off-shell correction, the good agreement with data on nonisoscalar nuclei obtained using the same off-shell function δf(x)\delta f(x) for the proton and neutron [14, 16] seems to indicate that potential isospin dependence of δf\delta f is small. Dedicated studies of nuclear effects using DIS data from mirror nuclei 3H{}^{3}\mathrm{H} and 3He{}^{3}\mathrm{He} [24] and upcoming DIS data on proton and deuterium from JLab12 [57] could provide new insights on the origin of modification of parton structure in bound nucleons, as well as improved constraints on nucleon dd-quark distribution at large xx and on the isospin dependence of HT corrections. A more comprehensive study of these effects would require future data from high-energy processes, which can provide a flavor selection like the hadronic Drell-Yan reaction or DIS using both the electron and (anti)neutrino charged-current (CC) process. To this end, the availability of precision measurements at future electron-ion collider [58] and of both neutrino and antineutrino CC interactions off hydrogen and various nuclear targets [59, 60] at the long-baseline neutrino facility could provide valuable insights.

Figure 9. The pulls of the JLab-E00-116 data [38] on d2σdE′dΩ\frac{\mathrm{d}^{2}\sigma}{\mathrm{d}E'\mathrm{d}\Omega} for the proton (left panels) and the deuteron (right panels) targets vs Q2Q^{2}. The panels correspond to the bins of the electron scattering angle θ\theta. Notations are similar to those in Figure 8.

Acknowledgments

We would like to thank A. Accardi for useful discussions and communications on the CJ15 results, G. G. Petratos for useful communications on the MARATHON results, M. V. Garzelli and S.-O. Moch for useful discussions and a careful reading of the manuscript, and Fr. Kok for the help with manuscript preparation. The work of S.A. is supported by the DFG grants MO 1801/5-1, KN 365/14-1. R.P. is supported by Grant No. DE-SC0010073 from the Department of Energy, USA.

Comparisons of the fit results with data

Phase space integration in the nuclear convolution

Here, we discuss in more detail the integration in Eq. (7) for general kinematics of xx and Q2Q^2. Recall that integration region is constrained by the condition,

W2≥Wth2(14)W^2 \ge W_{\mathrm{th}}^2 \tag*{(14)}

where W2=(p+q)2W^2=(p+q)^2 and pp is the four-momentum of the bound nucleon, and WthW_{\mathrm{th}} is the threshold mass. The pion production threshold corresponds to Wth=M+mπW_{\mathrm{th}}=M+m_\pi, and by setting Wth=MW_{\mathrm{th}}=M, we also include the elastic channel. Because of the energy-momentum conservation, p=pd−pSp=p_d-p_S, where pdp_d is the deuteron four-momentum, and pS=(mS2+p2,−p)p_S=(\sqrt{m_S^2+p^2},-\boldsymbol{p}), the four-momentum of the spectator system with the mass mSm_S.10 In Appendix A of Ref. [14], the nuclear convolution integral with the constraint (B1) was considered for nonrelativistic spectator assuming mS2+p2=mS+p2/(2mS)\sqrt{m_S^2+p^2}=m_S+p^2/(2m_S). This approximation makes sense as the deuteron is a weakly bound system and most of the momentum distribution is in the nonrelativistic region. However, the high-momentum part with ∣p∣|\boldsymbol{p}| of order of a few hundred MeV requires a relativistic analysis. Here, we discuss the fully relativistic case of spectator kinematics that would allow us to better describe the contribution from the high-momentum region of the spectator.

In terms of the four-vectors pdp_d, qq, and pSp_S, we can write Eq. (14) as follows:

(pd+q)2+mS2−2(pd+q)⋅pS≥Wth2.(15)(p_d+q)^2+m_S^2-2(p_d+q)\cdot p_S \ge W_{\mathrm{th}}^2. \tag*{(15)}

In order to facilitate the discussion of Eq. (15), we use the following notations

S=(pd+q)2=Md2+Q2(1/xd−1),(16)S=(p_d+q)^2=M_d^2+Q^2\left(1/x_d-1\right), \tag*{(16)}
E=Md+q0=S+q2,(17)E=M_d+q_0=\sqrt{S+\boldsymbol{q}^2}, \tag*{(17)}
α=(S+mS2−Wth2)/(2EmS),(18)\alpha=(S+m_S^2-W_{\mathrm{th}}^2)/(2Em_S), \tag*{(18)}
β=∣q∣/E,(19)\beta=|\boldsymbol{q}|/E, \tag*{(19)}

where SS and EE are, respectively, the invariant mass squared and the energy of the virtual photon–deuteron system, and xd=Q2/(2pd⋅q)x_d=Q^2/(2p_d\cdot q) is the natural Bjorken variable for the deuteron. As it follows from the definitions in Eqs. (B3), α>0\alpha>0, and 0<β<10<\beta<1 at any finite Q2Q^2 value. In the limit Q2→∞Q^2\to\infty, we have β=1\beta=1 and α=(1−xd)Md/mS\alpha=(1-x_d)M_d/m_S. Using Eqs. (B3), we can write Eq. (15) as follows:

αmS−mS2+p2+βpz≥0.(20)\alpha m_S-\sqrt{m_S^2+p^2}+\beta p_z\ge0. \tag*{(20)}

For completeness, we also give here Eq. (20) for nonrelativistic spectator kinematics

2(α−1)mS2−p2+2βmSpz≥0.(21)2(\alpha-1)m_S^2-p^2+2\beta m_Sp_z\ge0. \tag*{(21)}

Below we discuss the solution to Eq. (20) and Eq. (21) in terms of both the spherical coordinates and the (pz,p⊥)(p_z,p_\perp) basis for the momentum pp.

Convolution integral using spherical coordinates

We consider Eq. (20) and Eq. (21) in spherical coordinates, in which pz=∣p∣cos⁡θp_z=|p|\cos\theta with θ\theta the zenith angle. Both Eq. (20) and Eq. (21) have two nodes, for which we will use the notation p±(cos⁡θ)p_\pm(\cos\theta). For Eq. (B4), we have

p±(cos⁡θ)=mSαβcos⁡θ±β2cos⁡2θ+α2−11−β2cos⁡2θ(22)p_{\pm}(\cos\theta)=m_S\frac{\alpha\beta\cos\theta\pm\sqrt{\beta^2\cos^2\theta+\alpha^2-1}}{1-\beta^2\cos^2\theta} \tag*{(22)}

while for Eq. (B5), we have

p±(cos⁡θ)=mS(βcos⁡θ±β2cos⁡2θ+2(α−1))(23)p_{\pm}(\cos\theta)=m_S\left(\beta\cos\theta\pm\sqrt{\beta^2\cos^2\theta+2(\alpha-1)}\right) \tag*{(23)}

In solving the inequalities (B4) and (B5), it is convenient to consider the cases α≤1\alpha\leq1 and α>1\alpha>1. As a result, the solution involves two different regions:

−1≤cos⁡θ≤10≤∣p∣≤p+(cos⁡θ),for α>1,(24)\begin{aligned} -1&\leq\cos\theta\leq1\\ 0&\leq|p|\leq p_{+}(\cos\theta),\qquad\text{for }\alpha>1, \tag*{(24)} \end{aligned}

and

c≤cos⁡θ≤1p−(cos⁡θ)≤∣p∣≤p+(cos⁡θ),for α0≤α≤1.(25)\begin{aligned} c&\leq\cos\theta\leq1\\ p_{-}(\cos\theta)&\leq|p|\leq p_{+}(\cos\theta),\qquad\text{for }\alpha_{0}\leq\alpha\leq1. \tag*{(25)} \end{aligned}

The parameters α0\alpha_{0} and cc are different for Eq. (B4) and (B5). For relativistic kinematics c=1−α2/βc=\sqrt{1-\alpha^{2}}/\beta and p±(cos⁡θ)p_{\pm}(\cos\theta) are given by Eq. (B6), while for the nonrelativistic spectator, c=2(1−α)/βc=\sqrt{2(1-\alpha)}/\beta and p±(cos⁡θ)p_{\pm}(\cos\theta) are given by Eq. (B7). The minimum value of α\alpha in Eq. (B9) is derived from the condition c=1c=1. We have α0=1−β2\alpha_{0}=\sqrt{1-\beta^{2}} and α0=1−12β2\alpha_{0}=1-\frac{1}{2}\beta^{2} for Eq. (B4) and (B5), respectively.

Also, the condition c=1c=1 determines the maximum allowed value of xdx_{d} for given Q2Q^{2} which is consistent with Eq. (B1). For Q2→∞Q^{2}\to\infty, we have xdmax⁡=1x_{d}^{\max}=1 and xdmax⁡=1−M/(2Md)≈3/4x_{d}^{\max}=1-M/(2M_{d})\approx3/4 for the relativistic and nonrelativistic spectator kinematics, respectively. For finite values of Q2Q^{2} we have in case of Eq. (B4),

xdmax⁡=(1+(M+Wth)2−Md2Q2)−1≈(1+4mπMQ2)−1.(26)x_{d}^{\max}=\left(1+\frac{(M+W_{\mathrm{th}})^{2}-M_{d}^{2}}{Q^{2}}\right)^{-1}\approx\left(1+\frac{4m_{\pi}M}{Q^{2}}\right)^{-1}. \tag*{(26)}

We use Eq. (B8) and (B9) to cast the momentum integral in Eq. (7) as follows:

∫d3p ∣Ψd(p)∣2 θ(W2−Wth2)={12∫−11dcos⁡θ∫0p+(cos⁡θ)dp p2(ψ02(p)+ψ22(p)),for α>1,12∫c1dcos⁡θ∫p−(cos⁡θ)p+(cos⁡θ)dp p2(ψ02(p)+ψ22(p)),for α≤1.(27)\int\mathrm{d}^{3}p\,|\Psi_{d}(p)|^{2}\,\theta\left(W^{2}-W_{\mathrm{th}}^{2}\right) = \begin{cases} \frac{1}{2}\int_{-1}^{1}\mathrm{d}\cos\theta\int_{0}^{p_{+}(\cos\theta)}\mathrm{d}p\,p^{2}\left(\psi_{0}^{2}(p)+\psi_{2}^{2}(p)\right), & \text{for }\alpha>1,\\[1.2ex] \frac{1}{2}\int_{c}^{1}\mathrm{d}\cos\theta\int_{p_{-}(\cos\theta)}^{p_{+}(\cos\theta)}\mathrm{d}p\,p^{2}\left(\psi_{0}^{2}(p)+\psi_{2}^{2}(p)\right), & \text{for }\alpha\leq1. \end{cases} \tag*{(27)}

where p=∣p∣p=|\mathbf{p}|, and ψ0\psi_{0} and ψ2\psi_{2} are the deuteron orbital momentum wave functions for l=0l=0 and l=2l=2, respectively:

∣Ψd(p)∣2=(ψ02(p)+ψ22(p))/(4π).(28)|\Psi_{d}(p)|^{2}=\left(\psi_{0}^{2}(p)+\psi_{2}^{2}(p)\right)/(4\pi). \tag*{(28)}

Following Eq. (8), the functions ψ0\psi_{0} and ψ2\psi_{2} are normalized as follows:

∫0∞dp p2(ψ02+ψ22)=1.(29)\int_{0}^{\infty}\mathrm{d}p\,p^{2}\left(\psi_{0}^{2}+\psi_{2}^{2}\right)=1. \tag*{(29)}

Note that in numerical applications, we apply a cut on the bound nucleon momentum pcut∼1 GeVp_{\mathrm{cut}}\sim1\ \mathrm{GeV} in the nuclear convolution. For this reason, we replace the upper limit on the momentum in Eq. (B11) with min⁡(pcut,p+(cos⁡θ))\min(p_{\mathrm{cut}},p_{+}(\cos\theta)). The integration region in Eq. (B11) is illustrated in Figure 12 for both the relativistic and the nonrelativistic spectator and for a few fixed values of xx and Q2Q^{2}. The integration region systematically shrinks with rising xx, and the allowed kinematical region is somewhat larger for the relativistic case, although the difference is only visible for high nucleon momenta p>0.5 GeVp>0.5\ \mathrm{GeV}. As a somewhat extreme example of the deuteron kinematics, in the last panel of Figure 12, we show the integration region for x=1.3x=1.3, which is limited to high values of cos⁡θ\cos\theta and momentum p>300 MeVp>300\ \mathrm{MeV}.

Image not extracted

Figure 12. The shaded area shows the region constrained by Eq. (24) and (B9) together with the cut p<1p < 1 GeV for a few fixed values of xx and Q2Q^2 shown in the panels. The light-gray region corresponds to a relativistic spectator, while the dark-gray is for the nonrelativistic one.

Convolution integral using (y,p⊥)(y, p_\perp) basis

Consider Eq. (7) and note that x′=x/yx' = x/y, where the dimensionless variable y=p⋅q/(Mq0)=(p0+γpz)/My = p \cdot q/(M q_0) = (p_0 + \gamma p_z)/M is usually referred to as the nucleon light-cone momentum. The integral over the nucleon momentum in Eq. (7) can be cast in terms of integration over yy and p⊥p_\perp:

Fid(x,Q2)=∫dydp⊥2dij(y,p⊥2,γ)FjN(x/y,Q2,μ2),(30)F_i^d(x,Q^2) = \int\mathrm{d}y\mathrm{d}p_\perp^2 d_{ij}(y,p_\perp^2,\gamma)F_j^N(x/y,Q^2,\mu^2), \tag*{(30)}

where i=T,2i = T,2 and we assume the sum over the repeated subscript j=T,2j = T,2 and

dij(y,p⊥2,γ)=π∫dpz∣Ψd(p)∣2Kijδ(y−p0+γpzM),(31)d_{ij}(y,p_\perp^2,\gamma) = \pi\int\mathrm{d}p_z \lvert\Psi_d(p)\rvert^2 K_{ij}\delta\left(y-\frac{p_0+\gamma p_z}{M}\right), \tag*{(31)}

and μ2=p02−p2\mu^2 = p_0^2 - p^2, and p0p_0 is the energy of the active nucleon, and the kinematical factors KijK_{ij} in Eq. (30) are given by Eq. (8). Note that the kernel dijd_{ij} in the convolution integral Eq. (30) depends on the kinematic variables xx and Q2Q^2 through a dimensionless parameter γ=(1+4x2M2/Q2)1/2\gamma= (1 + 4x^2M^2/Q^2)^{1/2}. We now briefly consider the γ=1\gamma= 1 case corresponding to the light-cone kinematics of Q2→∞Q^2 \to\infty. In this limit, the matrix KijK_{ij} has only the diagonal components, and KTT=K22=1+pz/MK_{TT} = K_{22} = 1 + p_z/M. Then the nuclear convolution for FTF_T and F2F_2 have the same form with the kernel dTT=d22=D(y,p⊥2)d_{TT} = d_{22} = D(y,p_\perp^2), which has the meaning of distribution over the corresponding variables:

D(y,p⊥2)=π∫dpz∣Ψd(p)∣2(1+pzM)δ(y−p0+pzM).(32)D(y,p_{\perp}^{2})=\pi\int dp_{z}\lvert\Psi_{d}(p)\rvert^{2}\left(1+\frac{p_{z}}{M}\right)\delta\left(y-\frac{p_{0}+p_{z}}{M}\right). \tag*{(32)}

The distribution by Eq. (32) is normalized to 1:

∫dydp⊥2D(y,p⊥2)=∫d3p∣Ψd(p)∣2(1+pzM)=1.(33)\int dydp_{\perp}^{2}D(y,p_{\perp}^{2})=\int d^{3}p\lvert\Psi_{d}(p)\rvert^{2}\left(1+\frac{p_{z}}{M}\right)=1. \tag*{(33)}

The term proportional to pzp_{z} vanishes after angular integration.

Note that in the off-shell region, the nucleon SF in Eq. (30) depends on the virtual nucleon mass square μ2\mu^{2}. We use Eq. (4) in order to separate the off-shell dependence of the bound nucleon structure function and integrate over p⊥2p_{\perp}^{2}. Then Eq. (30) can be cast in terms of a one-dimensional convolution integral as follows:

F2d(x,Q2)=∫xymax⁡dy[S0(y)F2N(x/y,Q2)+S1(y)δf(x/y)F2N(x/y,Q2)].(34)F_{2}^{d}(x,Q^{2})=\int_{x}^{y_{\max}}dy\left[S_{0}(y)F_{2}^{N}(x/y,Q^{2})+S_{1}(y)\delta f(x/y)F_{2}^{N}(x/y,Q^{2})\right]. \tag*{(34)}

where the light-cone smearing functions S0S_{0} and S1S_{1} are as follows:

S0(y)=∫dp⊥2D(y,p⊥2),(35)S_{0}(y)=\int dp_{\perp}^{2}D(y,p_{\perp}^{2}), \tag*{(35)}
S1(y)=∫dp⊥2D(y,p⊥2)v,(36)S_{1}(y)=\int dp_{\perp}^{2}D(y,p_{\perp}^{2})v, \tag*{(36)}

where v=(μ2−M2)/M2v=(\mu^{2}-M^{2})/M^{2} is the nucleon virtuality. The function S0(y)S_{0}(y) makes sense of the nucleon light-cone distribution in the deuteron and normalized to unity according to Eq. (33). Note that Eq. (34) was derived for γ=1\gamma=1, i.e., light-cone kinematics. In this limit, the constraint by Eq. (14) reduces to y>xy>x and ymax⁡=Md/My_{\max}=M_{d}/M.

Below we discuss in more detail the nuclear convolution by Eq. (30) and (B15) for both the relativistic and nonrelativistic kinematics of the nucleon spectator while keeping finite Q2Q^{2} effects.

Nonrelativistic spectator

We first consider Eq. (31) assuming the nonrelativistic nucleon with energy p0=M+εd−p2/(2M)p_{0}=M+\varepsilon_{d}-p^{2}/(2M), where εd=Md−2M\varepsilon_{d}=M_{d}-2M is the deuteron binding energy. Taking the integral in Eq. (31) we have

∫dpzδ(y−p0+γpzM)=M2t2−p⊥2,(37)\int dp_{z}\delta\left(y-\frac{p_{0}+\gamma p_{z}}{M}\right)=\frac{M^{2}}{\sqrt{t^{2}-p_{\perp}^{2}}}, \tag*{(37)}

where

t2=2M2(ymax⁡−y),(38)t^{2}=2M^{2}(y_{\max}-y), \tag*{(38)}
ymax⁡=1+γ22+εdM.(39)y_{\max}=1+\frac{\gamma^{2}}{2}+\frac{\varepsilon_{d}}{M}. \tag*{(39)}

Note also that by integrating the δ\delta function in Eq. (37), we have pzp_{z} as a function of yy and p⊥2p_{\perp}^{2}:

pz=γM−t2−p⊥2.(40)p_{z}=\gamma M-\sqrt{t^{2}-p_{\perp}^{2}}. \tag*{(40)}

Note that tt makes sense of the maximum p⊥p_{\perp} for the given yy. The condition t2=0t^2=0 determines the maximum value y=ymax⁡y=y_{\max}; see Eq. (39). Note that for γ=1\gamma=1 and neglecting a small correction due to the deuteron binding energy in Eq. (39), we have ymax⁡=3/2y_{\max}=3/2. This is different from the kinematical maximum ymax⁡=Md/My_{\max}=M_d/M in the relativistic case which is discussed in Sec. B 2 b and illustrated in Figure 13.

Image not extracted

Figure 13. The integration region in Eq. (41) for different values of xx and Q2Q^2. The region x<y<ymax⁡x<y<y_{\max} and p⊥2<t2(y)p_{\perp}^2<t^2(y) is shown by dashed lines. The shaded area is a region restricted to W2>(M+mπ)2W^2>(M+m_{\pi})^2 and ∣p∣<1|\mathbf{p}|<1 GeV computed for both the relativistic and the nonrelativistic spectator, and for fixed values of xx and Q2Q^2 indicated in the panels. The light-gray region corresponds to the relativistic spectator, while the dark-gray region is for the nonrelativistic one.

Using Eq. (37) to (B24), we cast Eq. (30) as follows:

Fid(x,Q2)=M24∫ymin⁡ymax⁡dy∫0t2dp⊥2ψ02(p)+ψ22(p)t2−p⊥2KijFjN(xy,Q2,μ2)(41)F_i^d(x,Q^2)=\frac{M^2}{4}\int_{y_{\min}}^{y_{\max}} \mathrm{d}y\int_0^{t^2}\mathrm{d}p_{\perp}^2\frac{\psi_0^2(p)+\psi_2^2(p)}{\sqrt{t^2-p_{\perp}^2}}K_{ij}F_j^N\left(\frac{x}{y},Q^2,\mu^2\right) \tag*{(41)}

where p=pz2+p⊥2p=\sqrt{p_z^2+p_{\perp}^2} and pzp_z is given by Eq. (40). Note the p⊥2p_{\perp}^2 integration in Eq. (41) has a singularity at p⊥2=t2p_{\perp}^2=t^2. Although this is an integrable singularity, it may cause an instability in numerical applications. For this reason, it is convenient to change the integration variable in Eq. (41) from p⊥2p_{\perp}^2 to u=t2−p⊥2u=\sqrt{t^2-p_{\perp}^2}. Then we have

Fid(x,Q2)=M22∫ymin⁡ymax⁡dy∫0tdu(ψ02(p)+ψ22(p))KijFjN(xy,Q2,μ2)(42)F_i^d(x,Q^2)=\frac{M^2}{2}\int_{y_{\min}}^{y_{\max}}\mathrm{d}y\int_0^t\mathrm{d}u\left(\psi_0^2(p)+\psi_2^2(p)\right)K_{ij}F_j^N\left(\frac{x}{y},Q^2,\mu^2\right) \tag*{(42)}

where p⊥2=t2−u2p_{\perp}^{2}=t^{2}-u^{2}, pz=γM−up_z=\gamma M-u, p=p⊥2+pz2\mathbf{p}=\sqrt{p_{\perp}^{2}+p_z^{2}}, and μ2=M2+2Mεd−2p2\mu^{2}=M^{2}+2M\varepsilon_d-2p^{2}. The lower limit of integration over the light-cone variable is ymin⁡=xy_{\min}=x. Note, however, that this integration region in the nuclear convolution is modified for finite Q2Q^{2}. The corresponding region can be inferred from Eq. (14). Unlike the case of spherical coordinates discussed in Sec. B1, the analytic solution to the inequality (B1) in terms of (y,p⊥)(y,p_{\perp}) is somewhat cumbersome and not shown here. The resulting integration region in the nuclear convolution Eq. (41) is illustrated in Figure 13.

In conclusion of this section we present the explicit expressions for the light-cone smearing functions by Eq. (35) and (B20):

S0(y)=M22∫0tdu (ψ02(p)+ψ22(p))(2−uM),(43)S_0(y)=\frac{M^2}{2}\int_0^t du\,(\psi_0^2(p)+\psi_2^2(p))\left(2-\frac{u}{M}\right), \tag*{(43)}
S1(y)=M22∫0tdu (ψ02(p)+ψ22(p))(2−uM)v,(44)S_1(y)=\frac{M^2}{2}\int_0^t du\,(\psi_0^2(p)+\psi_2^2(p))\left(2-\frac{u}{M}\right)v, \tag*{(44)}

where v=μ2/M2−1v=\mu^2/M^2-1 and the other notations are similar to those in Eq. (42). The function S0(y)S_0(y) and S1(y)S_1(y) computed for the AV18 deuteron wave function are plotted in Figure 14 (left panel).

Image not extracted

Figure 14. Left panel shows the smearing functions S0(y)S_0(y) (solid line) and −S1(y)-S_1(y) (dashed line) computed using Eq. (43) and (B28) for the AV18 deuteron wave function. The right panel illustrates the relativistic effects in the smearing functions (see also text).

Relativistic spectator

For the relativistic kinematics, we have p0=Md−M2+p2p_0=M_d-\sqrt{M^2+p^2}. Integrating the δ\delta function in Eq. (31), we have

∫dpzδ(y−p0+γpzM)=γMEaM+(γ2−1)E,(45)\int dp_z\delta\left(y-\frac{p_0+\gamma p_z}{M}\right)=\frac{\gamma M E}{aM+(\gamma^2-1)E}, \tag*{(45)}

where a=Md/M−ya=M_d/M-y, E=pz2+p⊥2+M2E=\sqrt{p_z^2+p_\perp^2+M^2}, and we should replace pzp_z with the solution of the following equation:

γpz=pz2+p⊥2+M2−aM.(46)\gamma p_z=\sqrt{p_z^2+p_\perp^2+M^2}-aM. \tag*{(46)}

For γ>1\gamma>1, this equation has a solution for any value of the parameter aa:

pz=−γaM+a2M2+(γ2−1)(M2+p⊥2)γ2−1.(47)p_z=\frac{-\gamma aM+\sqrt{a^2M^2+(\gamma^2-1)(M^2+p_\perp^2)}}{\gamma^2-1}. \tag*{(47)}

Note, that for γ→1\gamma\to1 (or Q2→∞Q^2\to\infty), only the region a>0a>0 is allowed, and we have

pz=p⊥2+(1−a2)M22aM.(48)p_z=\frac{p_\perp^2+(1-a^2)M^2}{2aM}. \tag*{(48)}

In this case, the condition a=0a=0 determines the upper limit on yy, ymax⁡=Md/My_{\max}=M_d/M. For finite values of Q2Q^2, the integration region extends to y>Md/My>M_d/M.

For the deuteron structure functions, we have

Fid(x,Q2)=γ4∫ymin⁡ymax⁡dy∫dp⊥2MEaM+(γ2−1)E(ψ02(p)+ψ22(p))KijFjN(xy,Q2,μ2).(49)F_i^d(x,Q^2)=\frac{\gamma}{4}\int_{y_{\min}}^{y_{\max}}\mathrm{d}y\int\mathrm{d}p_\perp^2\frac{ME}{aM+(\gamma^2-1)E}\left(\psi_0^2(p)+\psi_2^2(p)\right)K_{ij}F_j^N\left(\frac{x}{y},Q^2,\mu^2\right). \tag*{(49)}

The integration region in Eq. (49) is limited by Eq. (14). The resulting region is shown in Figure 13, in which we also illustrate the impact of the momentum cut on the integration region.

The light-cone distributions by Eq. (35) and (B20) can be written as

S0(y)=14a∫dp⊥2(ψ02(p)+ψ22(p))E(1+pzM),(50)S_0(y)=\frac{1}{4a}\int\mathrm{d}p_\perp^2\left(\psi_0^2(p)+\psi_2^2(p)\right)E\left(1+\frac{p_z}{M}\right), \tag*{(50)}
S1(y)=14a∫dp⊥2(ψ02(p)+ψ22(p))E(1+pzM)Md2−2MdEM2.(51)S_1(y)=\frac{1}{4a}\int\mathrm{d}p_\perp^2\left(\psi_0^2(p)+\psi_2^2(p)\right)E\left(1+\frac{p_z}{M}\right)\frac{M_d^2-2M_dE}{M^2}. \tag*{(51)}

Note that these functions have a pole at a=0a=0 that corresponds to y=Md/My=M_d/M. However, this value of yy requires an infinite nucleon momentum as pz→∞p_z\to\infty at y=Md/My=M_d/M. Such configurations should be suppressed by the deuteron wave function. In practice, the region of large y∼Md/My\sim M_d/M, and therefore the singularity, can be avoided by applying a reasonable cut on the nucleon momentum in the convolution integral.

The effect of relativistic kinematics is illustrated in Figure 14 (right panel), in which we show the ratio of the function S0S_0 computed with Eq. (50) and (B27) and a similar ratio for S1S_1. For the most important region ∣y−1∣<0.2|y-1|<0.2, which drives the nuclear convolution, this relativistic effect is negligible for S0S_0. For this reason the relativistic effect has only a small impact on the deuteron structure function for x<1x<1. The relativistic correction is somewhat larger for S1S_1 but does not exceed 3% in this region. The region of large ∣y−1∣|y-1| is driven by a high-momentum component of the deuteron wave function. For this reason, the effect of relativistic kinematics on the smearing functions is more important in this region, as illustrated by Figure 14. As the region y>1y>1 drives the deuteron structure functions for x≳1x\gtrsim1, one cannot ignore the effect of relativistic kinematics in this region of xx.

Q2Q^{2}000.20.20.40.40.60.60.80.80.90.91.01.0
1.01.01.0001.0005.032E−015.032\mathrm{E}{-01}2.086E−012.086\mathrm{E}{-01}6.188E−026.188\mathrm{E}{-02}9.986E−039.986\mathrm{E}{-03}2.977E−032.977\mathrm{E}{-03}8.851E−048.851\mathrm{E}{-04}
10.010.01.0001.0005.034E−015.034\mathrm{E}{-01}2.095E−012.095\mathrm{E}{-01}6.198E−026.198\mathrm{E}{-02}8.711E−038.711\mathrm{E}{-03}1.785E−031.785\mathrm{E}{-03}2.359E−042.359\mathrm{E}{-04}
100.0100.01.0001.0005.035E−015.035\mathrm{E}{-01}2.096E−012.096\mathrm{E}{-01}6.198E−026.198\mathrm{E}{-02}8.550E−038.550\mathrm{E}{-03}1.631E−031.631\mathrm{E}{-03}1.649E−041.649\mathrm{E}{-04}

Table 2. The values of the deuteron structure function F2dF_2^d computed by Eq. (7) using the AV18 deuteron wave function [45] and the test functions F2N=(1−x)3F_2^N=(1-x)^3 and δf=x\delta f=x.

Benchmarks of the convolution integral

In order to facilitate the comparison with the present approach, in Table II we list our results for F2dF_2^d computed for the test functions F2N=(1−x)3F_2^N=(1-x)^3 and δf=x\delta f=x.

References

References

  1. [19]S. I. Alekhin, S. A. Kulagin, and R. Petti, “Nuclear effects in the deuteron and constraints on the d/u ratio,” Phys. Rev. D 96, 054005 (2017), arXiv:1704.00204 [nucl-th].DOI
  2. [20]A. Accardi, L. T. Brady, W. Melnitchouk, J. F. Owens, and N. Sato (CJ), “Constraints on large-x parton distributions from new weak boson production and deep-inelastic scattering data,” Phys. Rev. D 93, 114017 (2016), arXiv:1602.03154 [hep-ph].DOI
  3. [21]T.-J. Hou et al., “New CTEQ global analysis of quantum chromodynamics with high-precision data from the LHC,” Phys. Rev. D 103, 014013 (2021), arXiv:1912.10053 [hep-ph].DOI
  4. [22]T. Cridge, L. A. Harland-Lang, A. D. Martin, and R. S. Thorne (MSHT), “An investigation of the αₛ and heavy quark mass dependence in the MSHT20 global PDF analysis,” Eur. Phys. J. C 81, 744 (2021), arXiv:2106.10289 [hep-ph].DOI
  5. [23]R. D. Ball et al. (NNPDF), “The path to proton structure at 1% accuracy,” Eur. Phys. J. C 82, 428 (2022), arXiv:2109.02653 [hep-ph].
  6. [24]D. Abrams et al. (MARATHON), “Measurement of the nucleon F₂ⁿ/F₂ᵖ structure function ratio by the Jefferson Lab MARATHON Tritium/Helium-3 deep inelastic scattering experiment,” Phys. Rev. Lett. 128, 132003 (2022), arXiv:2104.05850 [hep-ex].arxiv.org/abs/2104.05850
  7. [25]H. Georgi and H. D. Politzer, “Freedom at moderate energies: masses in color dynamics,” Phys. Rev. D 14, 1829 (1976).DOI
  8. [26]M. Virchaux and A. Milstzajn, “A measurement of αₛ and of higher twists from a QCD analysis of high statistics F₂ data on hydrogen and deuterium targets,” Phys. Lett. B 274, 221 (1992).DOI
  9. [27]S. I. Alekhin, S. A. Kulagin, and S. Liuti, “Isospin dependence of power corrections in deep inelastic scattering,” Phys. Rev. D 69, 114009 (2004), arXiv:hep-ph/0304210 [hep-ph].DOI
  10. [29]A. Bodek et al., “Experimental studies of the neutron and proton electromagnetic structure functions,” Phys. Rev. D 20, 1471 (1979).DOI
  11. [30]L. W. Whitlow, S. Rock, A. Bodek, E. M. Riordan, and S. Dasu, “A precise extraction of R = σL/σT from a global analysis of the SLAC deep inelastic ep and ed scattering cross-sections,” Phys. Lett. B 250, 193 (1990).
  12. [31]L. Whitlow, Deep inelastic structure functions from electron scattering on hydrogen, deuterium, and iron at 0.6 ≤ Q² ≤ 30 GeV², Ph.D. thesis, SLAC (1990).
  13. [32]M. D. D. M. e. a. Masters et al., “The Ratio σL/σT From Deep Inelastic Electron Scattering,” Phys. Rev. D 27, 285 (1983).
  14. [33]J. Gomez et al. (SLAC E-139), “Measurement of the A-dependence of deep inelastic electron scattering,” Phys. Rev. D 49, 4348 (1994).
  15. [34]S. Dasu et al., “Measurement of kinematic and nuclear dependence of R = σL/σT in deep inelastic electron scattering,” Phys. Rev. D 49, 5641 (1994).
  16. [35]A. C. Benvenuti et al. (BCDMS), “A high statistics measurement of the deuteron structure functions F₂(x, Q²) and R from deep inelastic muon scattering at high Q²,” Phys. Lett. B 237, 592 (1990).DOI
  17. [36]M. Arneodo et al. (NMC), “Accurate measurement of F₂ᵈ/F₂ᵖ and Rᵈ − Rᵖ,” Nucl. Phys. B 487, 3 (1997), arXiv:hep-ex/9611022.
  18. [37]A. Airapetian et al. (HERMES), “Inclusive measurements of inelastic electron and positron scattering from unpolarized hydrogen and deuterium targets,” JHEP 05, 126 (2011), arXiv:1103.5704 [hep-ex].DOI
  19. [38]S. Malace et al. (Jefferson Lab E00-115), “Applications of quark-hadron duality in F₂ structure function,” Phys. Rev. C 80, 035207 (2009), arXiv:0905.2374 [nucl-ex].DOI
  20. [39]S. Tkachenko et al. (CLAS), “Measurement of the structure function of the nearly free neutron using spectator tagging in inelastic ²H(e,e′p)X scattering with CLAS,” Phys. Rev. C 89, 045206 (2014), [Addendum: Phys. Rev. C 90, 059901 (2014)], arXiv:1402.2477 [nucl-ex].DOI
  21. [40]K. Griffioen et al., “Measurement of the EMC Effect in the deuteron,” Phys. Rev. C 92, 015211 (2015), arXiv:1506.00871 [hep-ph].DOI
  22. [41]M. Arneodo et al. (NMC), “Measurement of the proton and deuteron structure functions, F₂ᵖ and F₂ᵈ, and of the ratio σL/σT,” Nucl. Phys. B 483, 3 (1997), arXiv:hep-ph/9610231 [hep-ph].
  23. [42]G. G. Petratos, private communication.
  24. [43]S. Alekhin, J. Blümlein, S. Moch, and R. Plačakytė, “Parton distribution functions, αₛ, and heavy-quark masses for LHC Run II,” Phys. Rev. D 96, 014011 (2017), arXiv:1701.05838 [hep-ph].DOI
  25. [44]R. B. Wiringa, V. G. J. Stoks, and R. Schiavilla, “An accurate nucleon-nucleon potential with charge independence breaking,” Phys. Rev. C 51, 38 (1995), arXiv:nucl-th/9408016 [nucl-th].
  26. [45]S. Veeramasy and W. N. Polyzou, “A momentum-space Argonne V18 interaction,” Phys. Rev. C 84, 034003 (2011), arXiv:1106.1934 [nucl-th].
  27. [47]https://www-zeuthen.desy.de/~alekhin/OPENQCDRAD/.
  28. [48]A. Buckley et al., “LHAPDF6: parton density access in the LHC precision era,” Eur. Phys. J. C 75, 132 (2015), arXiv:1412.7420 [hep-ph].DOI
  29. [49]https://hepdata.info.org/.
  30. [50]I. Schienbein et al., “A review of target mass corrections,” J. Phys. G 35, 053101 (2008), arXiv:0709.1775 [hep-ph].DOI
  31. [51]C. Cocuzza et al. (JAM), “Isovector EMC effect from global QCD analysis with MARATHON data,” Phys. Rev. Lett. 127, 242001 (2021), arXiv:2104.06946 [hep-ph].DOI
  32. [52]A. Accardi, private communication.
  33. [53]S. Alekhin, J. Blümlein, S. Moch, and R. Plačakytė, “Isospin asymmetry of quark distributions and implications for single top-quark production at the LHC,” Phys. Rev. D 94, 114038 (2016), arXiv:1508.07923 [hep-ph].DOI
  34. [54]S. Alekhin, J. Blümlein, S. Kulagin, S.-O. Moch, and R. Petti, “Strange and non-strange distributions from the collider data,” PoS 316, 008 (DIS2018), arXiv:1808.06871 [hep-ph].arxiv.org/abs/1808.06871
  35. [55]S. Alekhin, A. Kardos, S. Moch, and Z. Trócsányi, “Precision studies for Drell–Yan processes at NNLO,” Eur. Phys. J. C 81, 573 (2021), arXiv:2104.02400 [hep-ph].DOI
  36. [56]J. Seely et al., “New measurements of the EMC effect in very light nuclei,” Phys. Rev. Lett. 103, 202301 (2009), arXiv:0904.4448 [nucl-ex].DOI
  37. [57]S. P. Malace et al. (E12-10-002), “Precision measurements of the F₂ structure function at large x in the resonance region and beyond,” https://misportal.jlab.org/mis/physics/experiments/viewProposal.cfm?paperId=622.
  38. [58]R. Abdul Khalek et al., “Science requirements and detector concepts for the electron-ion collider: EIC yellow report,” (2021), arXiv:2103.05419 [physics.ins-det].DOI
  39. [59]R. Petti, “Precision measurements of fundamental interactions with (anti)neutrinos,” in Proceedings, 27th International Workshop on Deep Inelastic Scattering and Related Subjects (DIS 2019): Torino, Italy, April 8-12, 2019 (2019) arXiv:1910.05995 [hep-ex].arxiv.org/abs/1910.05995
  40. [60]H. Duyang, B. Guo, S. R. Mishra, and R. Petti, “A novel approach to neutrino-hydrogen measurements,” (2018), arXiv:1809.08752 [hep-ph].DOI

Paper details

vlnc:W_kk7mzrfe
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents