betaFeedback

WorksPhysics and Astronomy

Weak (quasi-)PDFs from the lattice

Abstract

Future colliders will be able to probe the electroweak substructure of particles. An accurate description of this substructure may necessitate electroweak PDFs, similarly to QCD. We investigate therefore the Higgs and W/Z PDFs in a reduced Standard Model setup using lattice methods, determining the quasi-PDFs, and, to lowest order, the resulting PDFs. While uncertainties remain, the results show interesting structures in line with expectations from gauge invariance.

Preprint number: COMETA-2026-46

Copyright attribution to authors.

This work is a submission to SciPost Physics Core.

License information to appear upon publication.

Publication information to appear upon publication.

Introduction

Using parton distribution functions (PDFs) to describe the structure of bound states in collision processes of hadrons has a long and successful tradition [1, 2]. As higher energies become experimentally accessible at future colliders, such as the currently favored next large-scale collider FCC [3, 4], on-shell sea Higgs and weak gauge bosons become relevant, contributing substantially to cross sections [4, 5, 6]. Correspondingly, we need to understand contributions to PDFs originating from the electroweak sector [5]. These contributions are thought to be accessible through perturbative calculations, similar to QED PDFs [2], as the only contributions come from the sea, and splitting functions are suppressed by the small weak coupling. However, this holds only under the assumption that the structure of the asymptotic states is fully understood.

In the perturbative construction the electroweak asymptotic states do not need to be gauge invariant with respect to the weak interaction, since in the presence of the Brout-Englert-Higgs (BEH) effect, the gauge symmetry is broken by a gauge-variant vacuum. That the perturbatively constructed gauge-variant asymptotic states are the elementary ones is then responsible for the lack of valence contributions to the aforementioned PDFs. But this is not the complete picture.

In fact, the gauge symmetry is actually not literally broken, a consequence of Elitzur’s theorem [7], but merely hidden. The precise agreement of experimental measurements and perturbative phenomenological calculations is then not trivial, but can be explained by the Fröhlich-Morchio-Strocchi (FMS) mechanism [8, 9], and may not hold always. In fact, the FMS mechanism has been used to predict deviations [10, 11, 12] at sufficient precision. Hence, physical asymptotic states need to be weakly gauge-invariant. For particles like the proton or the electron, this implies the existence of an electroweak valence contribution of the Higgs [8, 9, 13], which is consistent with current data [14].

Future colliders may be sensitive to this effect. As manifestly invariant asymptotic states also yield a different composition of the PDFs [15] than in perturbation theory, it is worthwhile to take this possibility into account explicitly. Especially as data-driven extractions of the PDFs can hide new physics [16], and thus potentially also such effects. In fact, it may be that the full electroweak PDF structure can still be calculated by a suitable augmentation of perturbation theory [10, 11, 12, 17]. However, to avoid an unnecessary assumption at this stage, here a different route will be taken.

It has recently become possible to determine PDFs using first-principles lattice methods [18, 19]. The approach applies analogously to electroweak physics. However, a full Standard Model computation is hampered by the fact that parity violation in gauge theories is obstructed by the lattice regularization [20], and the large scale separations are computationally unfeasible. Thus, to test PDFs, we turn here to the use of a truncated Standard Model, which contains only the Higgs and the W/Z gauge bosons. This theory has been studied on the lattice in great detail [10], and the predicted deviations due to the compositeness of the asymptotic states from the perturbative results are readily available [11, 12, 17, 21, 22, 23]. We thus use this simplified theory here to determine the PDFs using lattice methods, providing insight into the structure of the composite gauge-invariant asymptotic states. Our results are compatible with the ones favored by general arguments [13] and constraints from data [14]. Future work may be concerned with computing PDFs from augmented perturbation theory (APT), incorporating the gauge invariance, and comparing them with the non-perturbative results in this work. This will ultimately pave the way to use APT to calculate the electroweak PDFs for the full standard model, rather than the present proxy.

We will also supplement the precision results with a first investigation which also include unquenched vectorial fermions, an improved proxy for the Standard Model [24, 25]. Some preliminary results can be found in [23]. The theoretical setup of our calculation will be given in § 2, in which we also discuss the details of the PDFs. Lattice simulation details can be found in § 3. While the lattice techniques are standard, the application to weak physics is new. The results for the PDFs, including the analysis of systematic lattice artifacts, are given in § 4. Our final results are presented in § 5 with a summary and conclusion in § 6.

Theoretical setup

Theory

The theory we study comprises an SU(2)w\mathrm{SU}(2)_{\mathrm{w}} Yang-Mills theory with gauge bosons WW coupled to a scalar field XX in the fundamental representation and two generations of degenerate vectorial fermions ψ\psi in the fundamental representation. The Lagrangian of this theory is given by

\mathcal{L}=-\frac{1}{4}W_{\mu\nu}^{a}W^{a\mu\nu}+\frac{1}{2}\operatorname{tr}\left[(D_{\mu}X)^{\dagger}(D^{\mu}X)\right]-\frac{\lambda}{4}\left(\operatorname{tr}(X^{\dagger}X)-v^{2}\right)^{2}+\bar{\psi}\left(i\slashed{D}-m_{\psi}\right)\psi. \tag*{(1)}

It is thus structurally similar to the Standard Model lepton sector, except that the vectorial nature of the fermions allow tree-level masses. This theory has several symmetries besides the SU(2)w\mathrm{SU}(2)_{\mathrm{w}} gauge symmetry: A global SU(2)c\mathrm{SU}(2)_{\mathrm{c}} symmetry of the scalar field XX and an SU(2)g\mathrm{SU}(2)_{\mathrm{g}} generational symmetry ψ\psi of the fermions for degenerate mass. As the gauge symmetry SU(2)w\mathrm{SU}(2)_{\mathrm{w}} yields a pseudoreal fundamental representation, the generation symmetry combines with the chiral symmetry to an enlarged SU(4)ψ\mathrm{SU}(4)_{\psi}, known as Pauli-Gürsey symmetry [26]. If mψ≠0m_{\psi}\ne0, this symmetry is explicitly broken to an Sp(4)ψ\mathrm{Sp}(4)_{\psi} symmetry. Besides this explicit breaking, the global symmetries remain intact, and can thus be used to classify states. Below, most of the time, the theory (1) will be used without the fermions.

As noted in the introduction, the physical spectrum of the theory needs to be manifestly invariant under the gauge symmetry. Thus, states can only be classified according to their global symmetries, as well as JPJ^{P}. A separate charge symmetry is, due to the pseudoreality, absent even for uncharged states. A detailed discussion of this can be found in [24].

Here, we consider two quantum number channels, the uncharged scalar 0+0^{+} and the custodial triplet 1−1^{-}. These correspond to the physical Higgs and the physical WW and ZZ bosons [8, 9, 10]. The simplest composite operators for these quantum numbers are

O0=det⁡X,(2)O_{0}=\det X , \tag*{(2)}
O1aμ=tr⁡(τaX†DμX),(3)O_{1}^{a\mu}=\operatorname{tr}\left(\tau^{a}X^{\dagger}D^{\mu}X\right), \tag*{(3)}

where τ\tau are the Pauli matrices acting in SU(2)c\mathrm{SU}(2)_{\mathrm{c}}. We note that the vector operator will generally have three polarization directions. We will consider here the cases, where the operator O1O_{1} is projected longitudinally and transversely with respect to the momentum separately, indicating the two possibilities as ∥\parallel and ⊥\perp respectively. With our momentum introduced later as (P0,0,0,P3)(P_{0},0,0,P_{3}), this implies that we either average over the μ=1,2\mu=1,2 polarizations or just consider the μ=3\mu=3 case. Since the vector operator is gauge invariant, all polarization directions are physical on-shell. Moreover, in our reduced Standard Model setups, there are stable states in both channels that produce on-shell asymptotic states, whose spin and spin components are well defined observable quantities. A brief exploratory discussion on how this relates to the gauge-dependent polarizations of the gauge bosons can be found in [21].

PDFs

In general, any PDFs will be derived from matrix elements of the structure [18, 19]

C3(P,z)=⟨Oi†(P)J(z,0)Oi(P)⟩ ,(4)C_{3}(P,z)=\langle O_i^\dagger(P)J(z,0)O_i(P)\rangle\,, \tag*{(4)}

where the OiO_i are any of (2-3), and represent the interacting state of the corresponding quantum number and momentum P=(P0,0,0,P3)P=(P_0,0,0,P_3). The quantity JJ is a gauge-invariant bi-local current operator, potentially carrying some charge, connecting the points 00 and zz along some path. A possibility are two terminating operators at the location of OiO_i and Oi†O_i^\dagger and a connecting Wilson line.

As a PDF is build from the current JJ of a conserved charge, it can be interpreted as the momentum distribution of that charge in the given particle described by OiO_i. In case the charge is carried by a specific species of particle, this becomes equivalent to the momentum distribution of this type of particle inside the parent particle. This interpretation is especially applying to the usual context of QCD, where all elementary particles are distinguished by global quantum numbers. These map one-to-one on physical currents, the six quark flavor currents and an unflavored spin-1 current for the gluon. The corresponding currents are then dominated at lowest order by the elementary particles, allowing this one-to-one interpretation. Of course, a literal interpretation as a probability is not possible beyond tree-level [2].

In our context, the situation is drastically different. The physical quantum numbers of (2-3) are not the same as those of the elementary ones. This makes a similar interpretation difficult, and needs some explanation. In particular, this will substantially affects the choice of currents.

In its very definition, the expression (4) is build from gauge-invariant operators. In QCD, the connection to the elementary gauge-dependent constituents is then usually drawn in the following way: The inserted bilocal current JJ has for quarks, e. g., the form [19]

J=qˉ(x)W(Cx→y)q(y) ,(5)J=\bar{q}(x)W(C_{x\to y})q(y)\,, \tag*{(5)}

where WW is a Wilson line along a path Cx→yC_{x\to y} from xx to yy, and the terminating quark fields qq are here taken to be of the same flavor. At leading order, W∼1W\sim1, and thus (5) provides the correlation of a quark inside each of the asymptotic states with fixed flavor. Likewise, the usual field-strength tensor will yield at leading order a gluon. As (5) is gauge-invariant, it averages over all possible colors of the quarks.

In the present case, the physical states for the WW and ZZ are described by a custodial triplet (3). However, the elementary ones are a gauge triplet and carry no custodial charge. Thus, building the current on (3) will not yield an identification with the WW and ZZ, but rather with multiple Higgs to create the adjoint charge. This is not what is needed for a perturbative approach1. Thus, to adequately describe the WW and ZZ in a PDF language will require to treat them in the same way as gluons in QCD, i. e. by a JJ built from the field strength tensors. Note that in the present case WW and ZZ are the same, because we do not include hypercharge. Thus, summing over WW and ZZ equally in the perturbative treatment is for now adequate.

This leaves the Higgs, which splits perturbatively into the fluctuation mode and the pseudo-Goldstone bosons contained in the Higgs field. By the Goldstone-boson equivalence theorem, the latter are expected to be related to the longitudinal polarization of the vector bosons. Doing the calculation in a suitable gauge, e. g. unitary gauge, all of their effects would be contained in the cases listed so far. If the aim is to use the PDFs in a calculation where the pseudo-Goldstone bosons remain dynamical, it may be necessary to also build from the Higgs fields a current in the adjoint of the custodial group. However, this is beyond the current scope. Here, we will concentrate on the scalar singlet current, to describe the fluctuation Higgs field, making contact to previous phenomenological investigations [13, 14].

In practice, the operator JJ is not unique, as usual. However, it has been argued that the results are universal [27]. In practice, we build our operators from the following basic operators

F0+(τ,z3)=tr⁡(X†(z)W(z,z0)X(z0)),(6)F_{0}^{+}(\tau,z_{3}) = \operatorname{tr}\left(X^{\dagger}(z)W(z,z_{0})X(z_{0})\right), \tag*{(6)}
F11−=wzx+wzy,(7)F_{1}^{1-} = w_{zx}+w_{zy}, \tag*{(7)}
F12−=wzx+wzy+wzt,(8)F_{1}^{2-} = w_{zx}+w_{zy}+w_{zt}, \tag*{(8)}
F13−=wtx+wty−2wxy=F14−−2F15−,(9)F_{1}^{3-} = w_{tx}+w_{ty}-2w_{xy}=F_{1}^{4-}-2F_{1}^{5-}, \tag*{(9)}
F14−=wtx+wty,(10)F_{1}^{4-} = w_{tx}+w_{ty}, \tag*{(10)}
F15−=wxy,(11)F_{1}^{5-} = w_{xy}, \tag*{(11)}
wμν(τ,z3)=tr⁡(Wμν(z)W(z,z0)Wμν(z0)W(0,z0)),w_{\mu\nu}(\tau,z_{3}) = \operatorname{tr}\left(W_{\mu\nu}(z)W(z,z_{0})W_{\mu\nu}(z_{0})W(0,z_{0})\right),
z=(τ,0,0,z3),z = (\tau,0,0,z_{3}),
z0=(τ,0,0,0),z_{0} = (\tau,0,0,0),

where WW is a straight Wilson line connecting the two points zz and z0z_{0}, which are at the same time τ\tau. As we choose the straight line along the third axis the only explicit dependency is on the difference in this component, called z3z_{3}. We note that the operators F1i−F_{1}^{i-} are the same as the ones used for the gluon PDFs in [28, 29, 30, 31], as discussed above. Only F0+F_{0}^{+} is genuine to the present context. As we found that in our case in most setups F14−F_{1}^{4-} and F15−F_{1}^{5-} cancel within statistical errors in F13−F_{1}^{3-}, we investigated each of them separately. We found that they yield similar results to F14−F_{1}^{4-} and F15−F_{1}^{5-} individually.

Together with the possible polarizations, we studied in total 18 (=3 external states×\times6 currents) cases. We studied these in three distinct setups, two are pure gauge-scalar systems without fermions, which differ (slightly) in terms of the parameters. The third set of ensembles includes dynamical fermions, but with fewer and smaller volumes and reduced statistics due to larger computational cost. Note that probing the fermion as a state is more involved due to the symmetry structure [15]. This will be addressed in the future. In total, we determine 54 different (quasi-)PDFs.

The PDFs are obtained from the matrix element (4) through the intermediate step of quasi-PDFs. To this end, the normalized operator

ROF(t,τ,P3,z3)=C3OF(t,τ,P3,z3)C2O(P3,t),R^{OF}(t,\tau,P_{3},z_{3}) = \frac{C_{3}^{OF}(t,\tau,P_{3},z_{3})}{C_{2}^{O}(P_{3},t)},
C3OF=⟨O†(t,P3)F(z,0,τ)O(P,0)⟩,C_{3}^{OF} = \langle O^{\dagger}(t,P_{3})F(z,0,\tau)O(P,0)\rangle,
C2O=⟨O†(t,P3)O(0,P3)⟩C_{2}^{O} = \langle O^{\dagger}(t,P_{3})O(0,P_{3})\rangle

is build from any choice of (2-3) for OO, separately for each polarization, and (6-11) for FF. The momentum dependence for the operators OO is introduced by performing a Fourier transformation with respect to the spatial directions. The time difference, measured by tt, is used to pull apart the probed particles, such that initial state and final state are far away from the probe time τ\tau, and should therefore be dominated by the asymptotic states. Ideally, 0≪τ≪t0 \ll\tau\ll t.

Assuming 0≪τ≪t0 \ll\tau\ll t, this behaves as [19]

lim⁡t→∞ROF(t,τ,P3,z)=KO(P3)MOF(P3,z),\lim_{t\to\infty} R^{OF}(t,\tau,P_{3},z)=K^{O}(P_{3})M^{OF}(P_{3},z),

where KK is a known function, a kinematic normalization. For (6) it is 1/(2E)1/(2E), where EE is the energy associated with the probed particle, and for (7-11) it is the inverse gauge coupling 1/g1/g. Both will be absorbed in the following, as we will eventually normalize the resulting PDFs.

The function MM is thus the one of interest here, and can be obtained as

ddt∫ϵt−ϵdτ ROF(t,τ,P3,z)=MOF(P3,z)+O(t).(12)\frac{d}{dt}\int_{\epsilon}^{t-\epsilon}d\tau\,R^{\mathrm{OF}}(t,\tau,P_{3},z)=M^{\mathrm{OF}}(P_{3},z)+\mathcal{O}(t). \tag*{(12)}

In turn, the quasi-PDF QQ is then defined as [19]

QOF(x,P3)=NF(xP3)∫dz2πe−izxP3MOF(P3,z)(13)Q^{\mathrm{OF}}(x,P_{3})=N^{F}(xP_{3})\int\frac{dz}{2\pi}e^{-izxP_{3}}M^{\mathrm{OF}}(P_{3},z) \tag*{(13)}

with N0+=1/2N^{0^{+}}=1/2 [32] and N1−=1/(xP3)N^{1^{-}}=1/(xP_{3}) [29] once more normalization factors.

The quasi-PDF relates to the PDF qq as [18, 19]

QOF(x,P3)=∫−11dy∣y∣DOF(x,y,P3,μ)qOF(y,μ),Q^{\mathrm{OF}}(x,P_{3})=\int_{-1}^{1}\frac{dy}{\lvert y\rvert}D^{\mathrm{OF}}(x,y,P_{3},\mu)q^{\mathrm{OF}}(y,\mu),

where DOFD^{\mathrm{OF}} is a perturbative expression, which thus introduces a factorization scale μ\mu. This function DOFD^{\mathrm{OF}} takes the form

DOF(x,y,P3,μ)=δ(1−xy)+αdOF(x,y,P3,μ),D^{\mathrm{OF}}(x,y,P_{3},\mu)=\delta\left(1-\frac{x}{y}\right)+\alpha d^{\mathrm{OF}}(x,y,P_{3},\mu),

with a perturbatively calculable function dOFd^{\mathrm{OF}}, which becomes less relevant for P3→∞P_{3}\to\infty, and α\alpha is the relevant coupling constant. As in our present case α\alpha is much smaller than in QCD, it can be expected that the correction due to dOFd^{\mathrm{OF}} is much smaller as in the QCD case. We thus take here the approximation

qOF(x)=lim⁡P3→∞QOF(x,P3).(14)q^{\mathrm{OF}}(x)=\lim_{P_{3}\to\infty}Q^{\mathrm{OF}}(x,P_{3}). \tag*{(14)}

Note that this implies that the factorization scale dependence is lost, and the so determined PDF will be at most useful at tree-level.

In the context of APT calculations, this will likely suffice for some time to come. However, it should be noted that we refer here to an infinite momentum frame for the PDFs, neglecting masses. It is therefore likely that our PDFs would be appropriate for very high-energy collisions. At low energies, of order a few hundred GeVs, the PDFs will likely be much more similar to their tree-level counterparts discussed in [13], i. e. trivial

HH=δ(1−x),(15)H_{H}=\delta(1-x), \tag*{(15)}
HW=0,(16)H_{W}=0, \tag*{(16)}
WW=δ(1−x),(17)W_{W}=\delta(1-x), \tag*{(17)}
WH=0,(18)W_{H}=0, \tag*{(18)}

irrespective of polarization. This is in line with measurements of cross sections in this theory using lattice simulations at very low energies [22]: Up to additional threshold effects due to extension of the states, they can be described well with the tree-level PDFs (15-18). One should therefore expect a QQ-evolution from (15-18) to the ones presented here. This is necessarily left for future work.

Normalization

For a final result, it is necessary to normalize the results for P3→∞P_{3}\to\infty. Assuming that all other PDFs, which are not determined in this work, are negligible, there is always the energy sum rule

∫dx x(∑iqi(x))=1.\int dx\,x\left(\sum_i q^i(x)\right)=1.

However, this requires a second sum rule for normalization, which is less simple to identify. For the scalar, there is no conserved quantum number available, and thus no obvious sum rule. However, we will adopt a constituent counting rule as

∫dxHH(x)=1.(19)\int dx H_H(x) = 1 . \tag*{(19)}

The reasoning here is the FMS mechanism. When working in a suitable fixed gauge, the operator (2) can be rewritten as [8, 9, 10]

det⁡X=v2+vℜtr⁡η+tr⁡(η†η)(20)\det X = v^2 + v\Re\operatorname{tr}\eta+ \operatorname{tr}\left(\eta^\dagger\eta\right) \tag*{(20)}

with X=v1+ηX=v1+\eta, where vv is the Higgs vacuum expectation value and η\eta the fluctuation field. Hence, on-shell, the scalar is resembled by a single elementary Higgs [11, 12, 17]. As a result, when neglecting terms of order η/v\eta/v, the PDFs for the Higgs will be just the tree-level PDFs (15) and (16) [13], justifying (and satisfying) this sum rule.

For the vector, the situation is different, as there is a conserved quantum number, the custodial charge. Now, for the same reason as in (20), follows

tr⁡(τaX†DμX)=v2Wμa+v(tr⁡(τaDμη)+Wμbtr⁡(τaητb))+tr⁡(η†Dμη).(21)\operatorname{tr}\left(\tau^a X^\dagger D_\mu X\right)=v^2W_\mu^a+v\left(\operatorname{tr}\left(\tau^aD_\mu\eta\right)+W_\mu^b\operatorname{tr}\left(\tau^a\eta\tau^b\right)\right)+\operatorname{tr}\left(\eta^\dagger D_\mu\eta\right). \tag*{(21)}

Thus, to leading order, this suggests that the vector can be approximated by a gauge field. This would suggest a similar sum rule as (19),

∫dxWW(x)=1,(22)\int dx W_W(x) = 1 , \tag*{(22)}

for both transverse and longitudinal polarization separately. However, this is problematic. On the one hand, e. g., in Landau gauge the WW has no longitudinal degrees of freedom. Thus, the elementary field will not contribute to the longitudinally polarized vector – at least not at leading order. In addition, in (21), the custodial charge is actually carried by the vacuum condensate vv, as well as η\eta. Thus, it is more appropriate to require the sum rule

∫dxWH(x)=2(23)\int dx W_H(x) = 2 \tag*{(23)}

for the two units of fundamental charge contributed to the state by the Higgs field. This sum rule would not be satisfied by (15–18). However, setting

WH=12δ(x)(24)W_H=\frac{1}{2}\delta(x) \tag*{(24)}

would do so. Such a PDF does not contribute at leading order to (elastic) cross sections [13], and is thus still compatible with existing (lattice) data. It will therefore be adopted here. The argument holds doubly true at high energies, as there the Higgs partons are expected to be treated on equal footing, i. e. all terms in (21) should be equally important [13, 15]. Then the fact that the custodial charge is carried by the Higgs partons needs to be fully evident. However, the crossover between both regions may be fairly involved, and may affect the effective choice of PDFs [15].

This also avoids a technical problem. The normalization for the vector in (13) yields a trivial 1/x1/x behavior for the gauge boson quasi-PDFs. For a sum rule like (22) to hold, this would need to be suppressed by the matrix element. As will be seen, our results do not (yet) see a suppression at small xx. While this may be just a lattice artifact, this would require a regularization to employ a sum rule like (22).

Lattice setup

Parameter sets

The configurations for this project have been created using the HiRep code [33] with an extension for scalar fields [34] in the manner discussed in [35], to which we refer for full details.

We consider here three different lattice setups. The parameter sets for the standard Wilson action employed are given in Table 1. The properties and spectroscopy of these setups have been previously studied in detail in [22, 35, 37], for which we refer for details, including for the generation of configurations. As indicated by the tree-level parameters, all our sets are weakly coupled, but the coupling is stronger than in the Standard Model. We therefore expect that splitting in the Standard Model would be weaker, which may affect the PDFs and slow down the evolution from (15-18) towards higher energies.

Nameβ\betaκ\kappaγ\gammaκF\kappa_Fa−1a^{-1} [GeV]mhm_h [GeV]NgN_gmψm_\psi [GeV]
H2.79840.29841.3172871480
L8.00000.13100.00334001250
F8.00000.13100.00330.31034001252128
Namevv [GeV]ggαW\alpha_Wλ\lambda
H68.11.180.11115.0
L1490.540.0230.69
F1490.540.0230.69
SM2500.320.00810.25

Table 1. The upper table lists the simulation parameters and sets. The lattice spacing aa has been set by requiring that mW=80.375m_W = 80.375 GeV. The lattice spacing and the masses have been determined in [22, 36] for set H, in [37] for set L, and in [35] for set F. The lower table lists the tree-level parameters in the renormalization scheme defined in [35], for comparison, along with the values in the Standard Model.

The parameters of the first set H are taken from [22, 36], and describe a fairly heavy Higgs boson, close but below the elastic threshold. The additional measurement of the phase shift in the scalar channel determined in [22] supplies information about the scattering length and thus an effective size parameter of the Higgs. Its value is about 40 GeV−1^{-1}; comparable to the value found for the physical vector bosons in [21], providing additional information on the bound state structure. Note, that this study was conducted at a relatively coarse lattice spacing, with a heavy Higgs mass at 1.85mW1.85m_W close to the elastic decay threshold and relatively strong couplings.

The second set L from [37] has a significantly finer lattice spacing with a Standard-Model-like Higgs-W mass hierarchy with weaker couplings aimed at more Standard-Model-like behavior. However, the vacuum expectation value is significantly lower than in the Standard Model, corresponding to stronger coupling causing substantial renormalization effects. Comparing its result to H will allow us to understand the sensitivity to the parameters of the theory, as well as to the discretization.

The final set F is exploratory. It adds two generations of dynamical fermions to set L. More details about this set are given in [35]. The fermions are still fairly heavy, a concession to the substantial amount of statistics necessary to evaluate PDFs. Still, their mass is comparable to the Higgs and the vector bosons. As they appear not as valence particles for the states considered here, and we do not yet have tree-level Yukawa couplings, we expected their presence to produce only a negligible effect, which is confirmed by the numerical results.

We list the number of decorrelated configurations in Table 2, where we used the plaquette to measure residual autocorrelations. The relevant observables are three-point functions of composite operators which are fairly noisy for this theory [21] but also in general, see QCD applications [19]. As a result, a substantial number of configurations was needed. This restricted the maximum lattice size and mass of the fermions for set F. Fortunately, as will be seen, the results are fairly independent of volume, in accordance with other investigations of this theory [24, 36], such that relatively small volumes are enough to suppress systematic volume effects sufficiently for the present purpose.

NNHHLLFF
8972947
12436890404193
1639940433256837771
2020792022710319900
24199550216280
28119353139495
3296839

Table 2. Number of configurations used for each parameter set and lattice size.

Matrix element determination

For the involved operators, (2-3) and (6-11), the Wilson line and field strength tensors in (6-11) are both expressed in terms of products of links UμU_\mu along the shortest straight connecting path between both points. The choice follows for the operators [36, 38] and for the currents [29, 30, 31, 39]. This yields

O0(x)=det⁡X(x),O_0(x) = \det X(x),
O1aμ(x)=tr⁡(τaX(x)†det⁡X(x)†Uμ(x)X(x+μ)†det⁡X(x+μ)†),O_1^{a\mu}(x) = \operatorname{tr}\left(\tau^a \frac{X(x)^\dagger}{\sqrt{\det X(x)^\dagger}} U_\mu(x) \frac{X(x+\mu)^\dagger}{\sqrt{\det X(x+\mu)^\dagger}}\right),
W(Cz→z0)=(U3(z0)×⋯×U3(z))†,W(C_{z\to z_0}) = \left(U_3(z_0) \times\cdots\times U_3(z)\right)^\dagger,
Wμν(x)=i8(Uμν+Uν−μ+U−μ−ν+U−νμ−Uνμ−U−μν−U−ν−μ−Uμ−ν)(25)W_{\mu\nu}(x) = \frac{i}{8}\left(U_{\mu\nu} + U_{\nu-\mu} + U_{-\mu-\nu} + U_{-\nu\mu} - U_{\nu\mu} - U_{-\mu\nu} - U_{-\nu-\mu} - U_{\mu-\nu}\right) \tag*{(25)}

for the necessary lattice versions of the operators, where μ\mu is a lattice unit vector in μ\mu direction, UμU_\mu a link, and UμνU_{\mu\nu} the plaquette. Taking only the SU(2) part of the Higgs field (this refers to the additional division by the determinant instead of a trivial implementation of equation (3)) for O1O_1 has been found to substantially reduce fluctuations without changing the overlap with the asymptotic state [38]. The Wilson line WW is now taken along a straight line connecting points differing only in their 3-direction, ranging from zz to z0z_0, and determined in the usual way [40] by a product of links variables.

To further enhance the signal, we smeared the bosonic operators using up to 4 level of APE smearing [36, 41] as detailed in [35]. Note that on our smallest lattice of 848^4 this implies that smeared sinks and sources will overlap with the probe current, which will be discussed as a finite volume effect corresponding to keeping a fixed smearing radius relative to the lattice size, which will become point-like in the thermodynamic limit. We transform to momentum space using a standard discrete Fourier transform. Finally, the integral in (12) is computed by summing over 0<τ<t0 < \tau< t followed by a numerical derivative. To obtain the matrix element, we fit the plateau of the resulting quantity. The quasi-PDF (13) is determined by the discrete sum over the possible values of zz from the lattice. We further average over all possible axes for the choice of the straight path from zz to z0z_0.

We found that, while the function M(P3,z)M(P_{3},z) should decay for large zz to zero, volume artifacts lead to a non-zero value causing an oscillatory quasi-PDF behavior which persists at P3→∞P_{3} \to\infty and is not restricted to the interval x∈[0,1]x \in[0,1]. This constant contribution decreases for increasing P3P_{3}, but the decrease is not sufficiently fast to compensate for the increase in statistical fluctuations at higher momenta. We thus explicitly subtracted this constant from the function M(P3,z)M(P_{3},z), eliminating this finite volume artifact. This created a fairly smooth behavior in volume. In practice, we determine the constant by averaging the two largest zz values to determine the constant to be subtracted, which reduces the impact of statistical fluctuations. As this is not an exact procedure, a small oscillatory contribution remains.

Statistical errors are computed on the matrix-element level. Our primary matrix elements have very small errors, but the signal deteriorates exponentially with both P3P_{3} and tt. They are propagated to the quasi-PDF by varying the central values by ±1σ\pm1\sigma before determining the quasi-PDF. The errors at different values of xx are thus highly correlated due to the various averaging and fitting steps in between, especially in (12) and (13). The final error band is therefore not representing a situation where the error for each xx value is an independent statistical fluctuation. Rather, the propagation of the errors in the numerical process smooths out the fluctuations, giving the continuous error band shown, which is highly correlated between different xx values. Eventually, we take the Bjorken limit to be the point where the quasi-PDF remains stable within errors.

Lattice results

A study along different lines of constant physics is not strictly necessary at the current level of precision, since due to the weak coupling, the scale dependence is expected to be weak. This notion is corroborated by our data: The spectrum as well as form factors showed no strong dependence on the lattice spacing [21, 36]. Given the lack of excited states in the spectrum below the relevant inelastic thresholds [22, 36, 37], combined with a light ground state spectrum, this is not entirely surprising.

This leaves two major sources of systematic uncertainties, the volume and the extrapolation of P3→∞P_{3} \to\infty for the final result. Both are not entirely disjoint. On small volumes only a smaller range of P3P_{3} is available, which, however, already corresponds to much larger momenta and thus has higher statistical errors.

In total, we investigated 252 quasi-PDFs as a function of the lattice momentum P3P_{3} on different volumes, obtained from several millions of individual matrix element measurements for the different values of zz ranging from −N/2-N/2 to N/2N/2, t=3,…,N/2t = 3,\ldots,N/2, τ=1,…,t\tau= 1,\ldots,t, and P3=0,π/N,…,πP_{3} = 0,\pi/N,\ldots,\pi. The results for the matrix elements are very stable, justifying the correctness of our estimation of uncertainties. The matrix elements are made available in the dedicated data release [42] for reproducibility. This paper will only show the results on the final quasi-PDFs.

Volume effects

We measure the impact of the final volume effects by studying the ratio of the extracted quasi-PDF on a given volume to the one on the maximal volume. For all sets and all currents, we find that this ratio quickly approaches a constant when increasing the volume. To illustrate this, examples are shown in Figure 1. Even for fairly small volumes the ratio is almost constant within a few percent. Furthermore, the differences are mainly constrained to the largest xx-values. This is true for all momenta. However, due to the quick increase of noise when decreasing the volume and increasing the lattice momentum, this statement becomes less statistical reliable. towards the Bjorken limit. Still, for the largest volumes the ratio remain fairly constant within the larger errors.

Figure 1. The ratio of the vector quasi-PDFs HW2H_{W_2} of the scalar (top panels) and the scalar quasi-PDFs WHLW_H^L of the transverse polarized vector (bottom panels) as ratios between fixed volume to the largest volumes, and as a function of xx. The topmost row and the third row are for the lowest non-vanishing momentum and the second-to-top row and the bottom row for the second-lowest non-vanishing momentum. The left column is from the H ensemble and the right column from the L ensemble.

Figure 1 shows the scalar quasi-PDF for the transversely polarized vector WH⊥W_H^\perp and the vector current for the scalar HW2H_{W_2}. In the latter case, we use the current (8). Nevertheless, all vector currents with a statistically significant signal give the same qualitative result. There, it is directly visible that for the second-largest volumes the ratio to the largest volume is nearly xx-independent within a few percent, though the ratio itself is not necessarily one. The latter effect will be absorbed in the normalization. In fact, very often already volumes as small as 16416^4 give a reasonably flat ratio.

This pattern repeats itself for every quasi-PDF, operator, and set. Thus, we conclude that volume effects are essentially negligible for our largest volumes. We will therefore in the following only consider this largest volume. The only one, where some finite volume effects are still relevant at large xx is the computationally most expensive set F. However, the final results for the ensemble F are very close to the results of the pure gauge sets H and L. But our primary observation, that dynamical heavy fermions have negligible impact on the studied quasi-PDFs, is not affected by this. Hence, also in this case we will use in the following only the largest volume.

Fits for P3→∞P_3 \to\infty

The results from the lattice for the quasi-PDFs depend on the momentum P3P_3. The Bjorken limit is defined as the limit P3→∞P_3 \to\infty. To get an estimate of this limit, the quasi-PDFs were fitted for each P3P_3 independently, followed by an extrapolation.

For the fits of the xx-dependence we used the following ansätze

f(x)=a1a2+xκ1,(26)f(x) = \frac{a_1}{a_2 + x^{\kappa_1}}, \tag*{(26)}
f(x)=a1x(1−a2xκ1),(27)f(x) = \frac{a_1}{x}\left(1-a_2 x^{\kappa_1}\right), \tag*{(27)}
f(x)=1x(1−x)κ1(a1+a2x2),(28)f(x) = \frac{1}{x}(1-x)^{\kappa_1}\left(a_1+a_2 x^2\right), \tag*{(28)}
f(x)=1x1a2+xκ1a1,(29)f(x) = \frac{1}{x}\frac{1}{a_2+\frac{x^{\kappa_1}}{a_1}}, \tag*{(29)}

which where sufficient to describe all PDFs. In particular, for any given PDF, the same fit form was working for all P3P_3. However, the P3P_3 dependence of the parameters a1a_1, a2a_2, and κ1\kappa_1 in (26–29) was more variable. To extrapolate to the Bjorken limit, we fitted these parameters with

the following fit functions

p(P3)=a+bP32,(30)p(P_3) = a + \frac{b}{P_3^2}, \tag*{(30)}
p(P3)=a+bP3,(31)p(P_3) = a + \frac{b}{P_3}, \tag*{(31)}
p(P3)=a+bP3c,(32)p(P_3) = a + \frac{b}{P_3^c}, \tag*{(32)}
p(P3)=a2,(33)p(P_3) = a^2, \tag*{(33)}
p(P3)=θ(P3−3.5)(1−bP31/c),(34)p(P_3) = \theta(P_3 - 3.5)\left(1 - \frac{b}{P_3^{1/c}}\right), \tag*{(34)}
p(P3)=θ(P3−4.5)(1−bP31/c),(35)p(P_3) = \theta(P_3 - 4.5)\left(1 - \frac{b}{P_3^{1/c}}\right), \tag*{(35)}
p(P3)=θ(3.5−P3)(a+b3.5−P3)+θ(P3−3.5)(c+d3.5−P3),(36)p(P_3) = \theta(3.5 - P_3)\left(a + \frac{b}{3.5-P_3}\right) + \theta(P_3 - 3.5)\left(c + \frac{d}{3.5-P_3}\right), \tag*{(36)}
p(P3)=θ(3.5−P3)(a+b3.5−P3)+θ(P3−3.5)(c+d3.5−P32),(37)p(P_3) = \theta(3.5 - P_3)\left(a + \frac{b}{3.5-P_3}\right) + \theta(P_3 - 3.5)\left(c + \frac{d}{3.5-P_3^2}\right), \tag*{(37)}
p(P3)=θ(4.5−P3)(a+b4.5−P3)+θ(P3−4.5)(c+d4.5−P32),(38)p(P_3) = \theta(4.5 - P_3)\left(a + \frac{b}{4.5-P_3}\right) + \theta(P_3 - 4.5)\left(c + \frac{d}{4.5-P_3^2}\right), \tag*{(38)}
p(P3)=a2(1+tanh⁡(P3−bc)),(39)p(P_3) = \frac{a}{2}\left(1+\tanh\left(\frac{P_3-b}{c}\right)\right), \tag*{(39)}
p(P3)=d+a2(1+tanh⁡(P3−bc)),(40)p(P_3) = d + \frac{a}{2}\left(1+\tanh\left(\frac{P_3-b}{c}\right)\right), \tag*{(40)}
p(P3)=1+b(1+cP3)d,(41)p(P_3) = 1 + \frac{b}{(1+cP_3)^d}, \tag*{(41)}
p(P3)=ab+P3c.(42)p(P_3) = \frac{a}{b+P_3^c}. \tag*{(42)}

The appearance of the θ\theta-function in some of these has to do with the fact that some features only start to become statistically non-zero for sufficiently large values of P3P_3. However, using the fit forms at large P3P_3 also for small P3P_3 lead to no stable fits. Therefore, we use here this solution with step-wise fit forms. With this ansatz, the fit parameters a1a_1, a2a_2, and κ1\kappa_1 for each PDF become, within statistical uncertainties, P3P_3-independent for sufficiently large values of P3P_3. This strongly suggests that we indeed can extrapolate to the Bjorken limit in this way.

We list the results for the parameters in the Bjorken limit from the fit functions (26–29) in Table 3, Table 4, and Table 5, and indicate up to which value of P3P_3 the data was used to perform an extrapolation using the fit formulas (30–42). To estimate the errors on the fitted parameters, we repeated the fits with the PDFs evaluated at the boundaries of their error bands. Given that our errors between different values of xx are strongly correlated, this appears justified, to get an estimate of the variability of the resulting Bjorken limit.

SetPDFFitmax P3P_3a1a_1a2a_2κ1\kappa_1
HHHH_{H}(26)50.0014−6+80.0014^{+8}_{-6} (30)0.012−11+160.012^{+16}_{-11} (32)2.1−2+32.1^{+3}_{-2} (30)
HHW1H_{W_1}(27)60.213−11+100.213^{+10}_{-11} (33)1 (34)3.4−3+43.4^{+4}_{-3} (36)
HHW2H_{W_2}(27)60.261−13+120.261^{+12}_{-13} (33)1 (34)3.3−3+43.3^{+4}_{-3} (36)
HHW3H_{W_3}(28)613−11+34×10−513^{+34}_{-11} \times10^{-5} (39)0.0002−1+150.0002^{+15}_{-1} (39)2.8(1)2.8(1) (36)
HHW4H_{W_4}(27)70.248−16+150.248^{+15}_{-16} (33)1.0(7)1.0(7) (39)1 (41)
HHW5H_{W_5}(27)70.176(11)0.176(11) (33)1.1(8)1.1(8) (39)1 (41)
HWHIIW^{\mathrm{II}}_{H}(26)60.0012−2+30.0012^{+3}_{-2} (30)0.010−4+50.010^{+5}_{-4} (32)2.10−8+92.10^{+9}_{-8} (30)
HWW1IIW^{\mathrm{II}}_{W_1}(27)90.039(3)0.039(3) (30)1.1−3+61.1^{+6}_{-3} (39)1.8−1+21.8^{+2}_{-1} (37)
HWW2IIW^{\mathrm{II}}_{W_2}(27)90.058(6)0.058(6) (30)1.1−3+61.1^{+6}_{-3} (39)1.83−10+111.83^{+11}_{-10} (37)
HWW3IIW^{\mathrm{II}}_{W_3}-9---
HWW4IIW^{\mathrm{II}}_{W_4}(27)90.053(4)0.053(4) (32)1.2−2+41.2^{+4}_{-2} (39)1 (41)
HWW5IIW^{\mathrm{II}}_{W_5}(27)90.026(2)0.026(2) (32)1.2−2+41.2^{+4}_{-2} (39)0.90(1)0.90(1) (40)
HWH⊥W^{\perp}_{H}(26)60.00007−5+70.00007^{+7}_{-5} (30)0.010−7+80.010^{+8}_{-7} (32)2.1(2)2.1(2) (30)
HWW1⊥W^{\perp}_{W_1}(27)80.0023(7)0.0023(7) (31)1.0−3+91.0^{+9}_{-3} (39)2.09(8)2.09(8) (37)
HWW2⊥W^{\perp}_{W_2}(27)80.0035(11)0.0035(11) (31)1.0−3+81.0^{+8}_{-3} (39)2.10(8)2.10(8) (37)
HWW3⊥W^{\perp}_{W_3}-8---
HWW4⊥W^{\perp}_{W_4}(27)80.0032(8)0.0032(8) (31)1.1−3+61.1^{+6}_{-3} (39)1 (41)
HWW5⊥W^{\perp}_{W_5}(27)80.0032(8)0.0032(8) (31)1.1−3+61.1^{+6}_{-3} (39)1 (41)

Table 3. Unnormalized fits of the raw PDFs for set H. Fit refers to the fit form (26–29) used to fit the PDF itself, and - implies none was possible. The references in parentheses for the fit parameters refers to which extrapolation form (30–42) has been used for the determination of the P3→∞P_3 \to\infty of the parameters. Note that a positivity cut has been applied to all fitted parameters.

SetPDFFitmax P3P_3a1a_1a2a_2κ1\kappa_1
LHHH_H(26)40.0005−5+130.0005^{+13}_{-5} (30)0.008−8+120.008^{+12}_{-8} (32)2.6−5+62.6^{+6}_{-5} (30)
LHW1H_{W_1}(27)70.085(5)0.085(5) (33)1 (34)0.9−5+70.9^{+7}_{-5} (38)
LHW2H_{W_2}(27)70.104(6)0.104(6) (33)1 (34)0.7−5+70.7^{+7}_{-5} (38)
LHW3H_{W_3}-7---
LHW4H_{W_4}(29)60.1130×10−60.1^{30}_{1} \times10^{-6} (30)9.1(2)9.1(2) (33)2.10−3+42.10^{+4}_{-3} (30)
LHW5H_{W_5}(29)610−10+8×10−610^{+8}_{-10} \times10^{-6} (30)12.9(2)12.9(2) (28)2.1−3+42.1^{+4}_{-3} (30)
LWHIIW_H^{\mathrm{II}}(26)40.0006−3+50.0006^{+5}_{-3} (30)0.007−6+90.007^{+9}_{-6} (32)2.6(3)2.6(3) (30)
LWW1IIW_{W_1}^{\mathrm{II}}(27)80.0066(5)0.0066(5) (30)1 (35)1.1(2)1.1(2) (38)
LWW2IIW_{W_2}^{\mathrm{II}}(27)80.0099(8)0.0099(8) (30)1 (35)1.1(2)1.1(2) (38)
LWW3IIW_{W_3}^{\mathrm{II}}-8---
LWW4IIW_{W_4}^{\mathrm{II}}(29)80.00003(3)0.00003(3) (32)85−13+1085^{+10}_{-13} (27)2.15−5+62.15^{+6}_{-5} (30)
LWW5IIW_{W_5}^{\mathrm{II}}(29)82−2+7×10−62^{+7}_{-2} \times10^{-6} (32)178−8+10178^{+10}_{-8} (30)2.12−5+62.12^{+6}_{-5} (30)
LWH⊥W_H^{\perp}(26)44−4+8×10−54^{+8}_{-4} \times10^{-5} (27)0.007−6+70.007^{+7}_{-6} (32)2.6(3)2.6(3) (30)
LWW1⊥W_{W_1}^{\perp}(27)80.00−0+90.00^{+9}_{-0} (32)1 (35)1.0(2)1.0(2) (38)
LWW2⊥W_{W_2}^{\perp}(27)80.00−0+90.00^{+9}_{-0} (32)1 (35)0.9(2)0.9(2) (38)
LWW3⊥W_{W_3}^{\perp}-9---
LWW4⊥W_{W_4}^{\perp}(29)72−2+10×10−52^{+10}_{-2} \times10^{-5} (32)118−6+8118^{+8}_{-6} (42)2.14(4)2.14(4) (30)
LWW5⊥W_{W_5}^{\perp}(29)72−2+11×10−52^{+11}_{-2} \times10^{-5} (32)118−6+8118^{+8}_{-6} (42)2.14(4)2.14(4) (30)

Table 4. Unnormalized fits of the raw PDFs for set L. Fit refers to the fit form (26-29) used to fit the PDF itself, and - implies none was possible. The references in parentheses for the fit parameters refers to which extrapolation form (30-42) has been used for the determination of the P3→∞P_3 \to\infty of the parameters. Note that a positivity cut has been applied to all fitted parameters.

SetPDFFitmax P3P_3a1a_1a2a_2κ1\kappa_1
FHHH_H(26)30.0004−4+870.0004^{+87}_{-4} (30)0.00−0+140.00^{+14}_{-0} (30)3−2+43^{+4}_{-2} (30)
FHW1H_{W_1}(27)40.084−5+40.084^{+4}_{-5} (33)1 (34)3.1(3)3.1(3) (39)
FHW2H_{W_2}(27)40.103−6+50.103^{+5}_{-6} (33)1 (34)3.0(3)3.0(3) (39)
FHW3H_{W_3}-4---
FHW4H_{W_4}(29)40.000−0+20.000^{+2}_{-0} (30)9.2−3+49.2^{+4}_{-3} (33)2.11−7+92.11^{+9}_{-7} (30)
FHW5H_{W_5}(29)40.000−0+70.000^{+7}_{-0} (30)12.9−4+512.9^{+5}_{-4} (33)2.12−7+92.12^{+9}_{-7} (30)
FWHIIW_H^{\mathrm{II}}(26)30.001−1+30.001^{+3}_{-1} (30)0.00−0+70.00^{+7}_{-0} (30)2.7−5+102.7^{+10}_{-5} (33)
FWW1IIW_{W_1}^{\mathrm{II}}(27)50.007(2)0.007(2) (30)1 (34)2.5(4)2.5(4) (36)
FWW2IIW_{W_2}^{\mathrm{II}}(27)50.010(2)0.010(2) (30)1 (34)2.4−4+42.4^{+4}_{-4} (36)
FWW3IIW_{W_3}^{\mathrm{II}}-5---
FWW4IIW_{W_4}^{\mathrm{II}}(29)50.00−0+40.00^{+4}_{-0} (32)88−10+1488^{+14}_{-10} (30)2.1−1+22.1^{+2}_{-1} (30)
FWW5IIW_{W_5}^{\mathrm{II}}(29)50.000−0+20.000^{+2}_{-0} (32)177−20+30177^{+30}_{-20} (30)2.2−2+32.2^{+3}_{-2} (30)
FWH⊥W_H^\perp(26)30.0000−0+40.0000^{+4}_{-0} (30)0.00−0+70.00^{+7}_{-0} (30)2.7−5+102.7^{+10}_{-5} (33)
FWW1⊥W_{W_1}^\perp(27)50.00007−7+430.00007^{+43}_{-7} (32)1 (35)0.9992−9+40.9992^{+4}_{-9} (33)
FWW2⊥W_{W_2}^\perp(27)50.0001−1+50.0001^{+5}_{-1} (32)1 (34)2.3−3+42.3^{+4}_{-3} (36)
FWW3⊥W_{W_3}^\perp(28)50.00020(6)0.00020(6) (33)0.00008−4+90.00008^{+9}_{-4} (33)80−80+5080^{+50}_{-80} (39)
FWW4⊥W_{W_4}^\perp(29)50.0000−0+20.0000^{+2}_{-0} (32)225−122+41225^{+41}_{-122} (42)2.1−1+22.1^{+2}_{-1} (33)
FWW5⊥W_{W_5}^\perp(29)50.0000−0+20.0000^{+2}_{-0} (33)224−122+41224^{+41}_{-122} (42)2.1−1+22.1^{+2}_{-1} (33)

Table 5. Unnormalized fits of the raw PDFs for set F. Fit refers to the fit form (26-29) used to fit the PDF itself, and - implies none was possible. The references in parentheses for the fit parameters refers to which extrapolation form (30-42) has been used for the determination of the P3→∞P_3 \to\infty of the parameters. Note that a positivity cut has been applied to all fitted parameters.

Some of the fit parameters in tables (3–5) are compatible with zero. This is an artifact of the missing renormalization paired with a large statistical noise. While the finite-momentum quasi-PDFs are always statistically significant non-zero, the uncertainties when extrapolating to infinite P3P_3 allows for a compatibility with zero. This was, however, irrelevant for the final results in section 5, as the normalization is entirely determined by the sum rules of section 2.3.

A detailed example of the fitting procedure is shown in Figure 2. There are two main obstacles for a fit with the simple functions (26–29). One is the aforementioned finite volume

Figure 2. Fits and ratio to the fits for the quasi-PDFs HW2H_{W_2} (two top rows) and WH⊥W_H^{\perp} (two bottom rows), for the range of momenta where a fit was statistically possible. The top row and the third row show a direct comparison to the fits and the second and last row a ratio to the fit. The left column is set H and the right column set L.

effect in form of a constant offset in quasi-PDFs obtained from the scalar current (6). The remnant of this constant leads to an oscillatory behavior in xx, which increases with momenta. By fitting the whole range of xx with non-oscillating fit ansätze, this averages out the oscillations at the expense of an enlarged error band. Thus, this covers the consequences of this lattice artifact, which would eventually vanish in the infinite volume limit.

The second is that quasi-PDFs are defined for all2^{2} xx [19]. As a consequence, they will usually not vanish at x=1x = 1, even if the true PDF does. Indeed, we observe that with increasing P3P_3 some quasi-PDFs tend to vanish at x=1x = 1, and all become more suppressed towards x=1x = 1. In fact, due to the increase of statistical errors with increasing P3P_3, eventually all PDFs become consistent with zero at x=1x = 1 within errors. This feature is, to some extent, built into the fit forms (27-28), which at finite P3P_3 force the fit away from the data.

The preferred fit forms of the different sources (7-8) and (10-11) indicate a qualitative difference between them. Correspondingly, we will provide both versions in the result § 5, for the reader to gauge the associated systematic error.

Finally, we note that, up to lattice artifacts, there is good qualitative agreement between all three sets and quantitatively small differences between the parametrically matched pure gauge set L and the set F featuring dynamical fermions.

PDFs

In this section, we provide our best estimates of the PDFs investigated. This should provide an overall idea of their shape and size, and thus a basis for the use in phenomenological calculations.

These final estimates for the PDFs are shown in figures 3, 4, and 5 for the sets H, L, and F, respectively. The overall normalization has been fixed by the sum rules of § 2.3.

We find that the qualitative features of the results do not depend on the set and that the Higgs PDF is never enhanced at small xx, in strong contrast to quark PDFs. A similar saturation was already anticipated in phenomenological fits [13, 14]. From a gauge-fixed perspective, this could be interpreted as that low-energy Higgs particles are absorbed in the condensate, and cannot be excited to participate in the hard interaction. Low-momentum modes will then rather be part of the condensate than dynamically available, even at very high energies. The gauge boson PDFs, on the other hand, are strikingly different and behave like 1/x1/x. In that they are closer to QCD-like PDFs. This also suggests that the suppression of the Higgs PDF is not (only) a mass effect, as the Higgs and the gauge bosons have a similar mass.

Further, the PDFs obtained from the different currents for the gauge boson, the currents (7) and (10) in figures 3 to 5, overlap within their error bands in almost all cases, with the results from F1F_1 usually with smaller errors. It should also be noted that the fit forms are different depending on the channel and current. However, we found that the fit form (27) pairs usually with the current (7) and fit form (29) with the current (10). This is consistent with the anticipated universality [27]. Residual differences are most strongly pronounced in set L.

While the question of normalization is subtle, as described in § 2.3, we observe a trend that the Higgs PDF obtained from the stronger coupled set H is more heavily suppressed at larger xx compared to the sets L and F with weaker couplings and the W boson PDF on all other sets.

We further observe that the difference between the transversely polarized vector boson and longitudinally polarized vector boson are minor at best and that the difference between the scalar and the vector boson are also only moderate at large x≳0.1x \gtrsim0.1. The presence of

Figure 3. Final estimates for the PDFs for set H. The left column shows the PDFs themselves, the right column shows the PDFs multiplied by xx. The top line are the PDFs for the scalar boson, the middle line for the longitudinally polarized vector boson, and the bottom for the transversely polarized vector boson.

Figure 4. Final estimates for the PDFs for set L. The left column shows the PDFs themselves, the right column shows the PDFs multiplied by xx. The top line are the PDFs for the scalar boson, the middle line for the longitudinally polarized vector boson, and the bottom for the transversely polarized vector boson.

Figure 5. Final estimates for the PDFs for set F. The left column shows the PDFs themselves, the right column shows the PDFs multiplied by xx. The top line are the PDFs for the scalar boson, the middle line for the longitudinally polarized vector boson, and the bottom for the transversely polarized vector boson.

dynamical, heavy fermions does not have a qualitative impact. Probably the most interesting observations are that the all PDFs are modified substantially from their low-energy form (15–18) (or including (24) instead of (18)), and that the Higgs PDF behaves qualitatively different from the gauge boson PDFs. In particular, at low $x$, the gauge bosons dominate massively. In this context, it is necessary to keep in mind that our approximations will only reasonably work at large energies compared to the masses. Thus, the latter statement is likely only accurate at energy scales $x E \gg m_H$. Already for a modest value of $x = 0.1$ this requires multi-TeV interactions, the region where non-linear effects of the electroweak interactions are relevant [4, 15]`.

Conclusion

We have provided a first lattice investigation of weak PDFs in a reduced Standard Model, which contains only the Higgs and the W/ZW/Z, and a single exploratory look at the consequences of including heavy, vector-like fermions. This constitutes the first ever non-perturbative lattice investigation of PDFs for this theory, and thus the first ever non-perturbative investigation of the weak PDFs considered to be highly relevant for future experiments [4, 5, 6, 13, 15].

Our results suggest that the PDFs are relatively similar between the physical scalar and vector particles of the theory for x>0.1x > 0.1. However, the Higgs PDF is substantially suppressed at low xx, compared to the gauge boson PDF.

We have thoroughly investigated systematic effects. In particular, we have found that we have good control over the finite volume effects and the extrapolation of P3→∞P_3 \to\infty, the Bjorken limit. We examined the impact of different currents for the matrix elements, and studied thoroughly multiple fit ansatz choices, and found no qualitative, and only small quantitative, effects. While we did not yet investigate systematically discretization effects, there are strong indications for them to be small [36]. Also, our two ensembles H and L, while having slightly different Higgs masses, do have substantially different lattice spacing, but this did not lead to any qualitative, and only mild quantitative, differences. Of the remaining uncertainties, the largest effects stem from the choice of current, and the infinite-momentum extrapolation. Moreover, we neglected effects due to factorization, and thus our results will likely only be leading-order accurate.

There is still a long way to go before these results reach a similar precision as those employed in QCD calculations. Introducing factorization and improving statistical uncertainties are likely the most relevant candidates for future research. But for now, with experiments only starting to probe the relevant domain, our results can certainly serve as a first input to exploratory calculations based on the physical asymptotic states.

An interesting possibility will also be to evolve these results towards lower QQ, to pinpoint the necessary valence contribution, which is usually assumed to be zero [5] or trivial [13, 15]. That this will be a significant effect becomes evident by considering results for vector boson scattering [22], which at low energies cannot be captured by the trivial PDFs of section 2.3. Conversely, it will be also very interesting to use the present results for predictions of high-energy inclusive inelastic cross sections, e.g. again in vector boson scattering. There, deviations are also expected [15], and their determination using lattice methods is within reach [23].

Acknowledgements

We are grateful to S. Zafeiropoulos for helpful discussions. This research was funded in whole or in part by the Austrian Science Fund (FWF) [10.55776/PAT6443923]. The authors acknowledge support from the European Union (EU) COMETA COST Action [CA22130]. The computational results have been achieved using the Austrian Scientific Computing (ASC5) infrastructure and the Graz Scientific Cluster (GSC1). All matrix elements and quasi-PDFs can be downloaded from [42].

References

  1. [1]G. Dissertori, I. G. Knowles and M. Schmelling, Quantum Chromodynamics: High energy experiments and theory, Oxford University Press, Oxford, Oxford, UK: Clarendon (2003) 538 p (2003).
  2. [2]F. J. Yndurain, The Theory of Quark and Gluon Interactions, Theoretical and Mathematical Physics. Springer, Berlin, Germany, ISBN 9783540332091, 9783540332107, doi:10.1007/3-540-33210-3 (2006).DOI
  3. [3]M. Benedikt et al., Future Circular Collider Feasibility Study Report: Volume 1. Physics, Experiments, Detectors, Eur. Phys. J. C 85(12), 1468 (2025), doi:10.1140/epjc/s10052-025-15077-x, 2505.00272.
  4. [4]J. de Blas et al., Physics Briefing Book: Input for the 2026 update of the European Strategy for Particle Physics (2025), doi:10.23731/CYRM-2025-008, 2511.03883.
  5. [5]C. W. Bauer, N. Ferland and B. R. Webber, Standard Model Parton Distributions at Very High Energies, JHEP 08, 036 (2017), doi:10.1007/JHEP08(2017)036, 1703.08562.DOI
  6. [6]K. Moral Figueroa, E. Gallo, H. Jung and S. Taheri Monfared, Determination of parton densities for QCD partons and electroweak bosons, Eur. Phys. J. C 86(5), 449 (2026), doi:10.1140/epjc/s10052-026-15577-4, 2511.08321.
  7. [7]S. Elitzur, Impossibility of Spontaneously Breaking Local Symmetries, Phys. Rev. D12, 3978 (1975), doi:10.1103/PhysRevD.12.3978.DOI
  8. [8]J. Fröhlich, G. Morchio and F. Strocchi, Higgs phenomenon without a symmetry breaking order parameter, Phys.Lett. B97, 249 (1980), doi:10.1016/0370-2693(80)90594-8.DOI
  9. [9]J. Fröhlich, G. Morchio and F. Strocchi, Higgs phenomenon without a symmetry breaking order parameter, Nucl.Phys. B190, 553 (1981), doi:10.1016/0550-3213(81)90448-X.DOI
  10. [10]A. Maas, Brout-Englert-Higgs physics: From foundations to phenomenology, Progress in Particle and Nuclear Physics 106, 132 (2019), doi:10.1016/j.ppnp.2019.02.003, 1712.04721.DOI
  11. [11]D. Dudal, D. van Egmond, M. Guimaraes, L. Palhares, G. Peruzzo and S. Sorella, Spectral properties of local BRST invariant composite operators in the SU(2) Yang–Mills–Higgs model, Eur. Phys. J. C 81(3), 222 (2020), doi:10.1140/epjc/s10052-021-09008-9, 2008.07813.DOI
  12. [12]A. Maas and R. Sondenheimer, Gauge-invariant description of the Higgs resonance and its phenomenological implications, Phys. Rev. D 102, 113001 (2020), doi:10.1103/PhysRevD.102.113001, 2009.06671.DOI
  13. [13]L. Egger, A. Maas and R. Sondenheimer, Pair production processes and flavor in gauge-invariant perturbation theory, Mod. Phys. Lett. A32, 1750212 (2017), doi:10.1142/S0217732317502121, 1701.02881.DOI
  14. [14]S. Fernbach, L. Lechner, A. Maas, S. Plätzer and R. Schöfbeck, Constraining the higgs valence contribution in the proton, Phys. Rev. D 101(11), 114018 (2020), doi:10.1103/PhysRevD.101.114018, 2002.01688.DOI
  15. [15]A. Maas and F. Reiner, Restoring the Bloch-Nordsieck theorem in the electroweak sector of the standard model, Phys. Rev. D 108(1), 013001 (2023), doi:10.1103/PhysRevD.108.013001, 2212.08470.DOI
  16. [16]J. Gao, L. Harland-Lang and J. Rojo, The Structure of the Proton in the LHC Precision Era, Phys. Rept. 742, 1 (2018), doi:10.1016/j.physrep.2018.03.002, PDF, 1709.04922.DOI
  17. [17]A. Maas, D. M. van Egmond and S. Plätzer, Subleading Higgs effects at lepton colliders, In 42nd International Conference on High Energy Physics (2024), 2409.20131.arxiv.org/abs/2409.20131
  18. [18]X. Ji, Parton Physics on a Euclidean Lattice, Phys. Rev. Lett. 110, 262002 (2013), doi:10.1103/PhysRevLett.110.262002, 1305.1539.DOI
  19. [19]K. Cichy and M. Constantinou, A guide to light-cone PDFs from Lattice QCD: an overview of approaches, techniques and results, Adv. High Energy Phys. 2019, 3036904 (2019), doi:10.1155/2019/3036904, 1811.07248.DOI
  20. [20]P. Hasenfratz and R. von Allmen, Towards Weyl fermions on the lattice without artefacts, JHEP 02, 079 (2008), doi:10.1088/1126-6708/2008/02/079, Parity violation lattice, 0710.5346.DOI
  21. [21]A. Maas, S. Raubitzek and P. Törek, Exploratory study of the off-shell properties of the weak vector bosons, Phys. Rev. D99(7), 074509 (2019), doi:10.1103/PhysRevD.99.074509, 1811.03395.DOI
  22. [22]P. Jenny, A. Maas and B. Riederer, Vector boson scattering from the lattice, Phys. Rev. D 105, 114513 (2022), doi:10.1103/PhysRevD.105.114513, 2204.02756.DOI
  23. [23]S. Martins, P. Jenny, A. Maas and G. Wieland, Weak and Higgs physics from the lattice, In 42th International Symposium on Lattice Field Theory (2026), 2603.12882.
  24. [24]V. Afferrante, A. Maas, R. Sondenheimer and P. Törek, Testing the mechanism of lepton compositeness, SciPost Phys. 10, 062 (2021), doi:10.21468/SciPostPhys.10.3.062, 2011.02301.DOI
  25. [25]G. Wieland and A. Maas, Spectrum of the SU(2) scalar-fermion-gauge system under the influence of the Brout-Englert-Higgs effect, PoS EPS-HEP2025, 296 (2026), doi:10.22323/1.485.0296, 2510.27457.DOI
  26. [26]J. Kogut, M. A. Stephanov, D. Toublan, J. Verbaarschot and A. Zhitnitsky, QCD - like theories at finite baryon density, Nucl.Phys. B582, 477 (2000), doi:10.1016/S0550-3213(00)00242-X, hep-ph/0001171.DOI
  27. [27]J.-H. Zhang, X. Ji, A. Schäfer, W. Wang and S. Zhao, Accessing Gluon Parton Distributions in Large Momentum Effective Theory, Phys. Rev. Lett. 122(14), 142001 (2019), doi:10.1103/PhysRevLett.122.142001, 1808.10824.DOI
  28. [28]Z. Fan, R. Zhang and H.-W. Lin, Nucleon gluon distribution function from 2 + 1 + 1-flavor lattice QCD, Int. J. Mod. Phys. A 36(13), 2150080 (2021), doi:10.1142/S0217751X21500809, 2007.16113.DOI
  29. [29]T. Khan et al., Unpolarized gluon distribution in the nucleon from lattice quantum chromodynamics, Phys. Rev. D 104(9), 094516 (2021), doi:10.1103/PhysRevD.104.094516, 2107.08960.DOI
  30. [30]I. Balitsky, W. Morris and A. Radyushkin, Gluon Pseudo-Distributions at Short Distances: Forward Case, Phys. Lett. B 808, 135621 (2020), doi:10.1016/j.physletb.2020.135621, 1910.13963.DOI
  31. [31]W. Good, K. Hasan and H.-W. Lin, Toward the first gluon parton distribution from the LaMET, J. Phys. G 52(3), 035105 (2025), doi:10.1088/1361-6471/ada815, 2409.02750.
  32. [32]W. Wang, S. Zhao and R. Zhu, Gluon quasidistribution function at one loop, Eur. Phys. J. C 78(2), 147 (2018), doi:10.1140/epjc/s10052-018-5617-3, 1708.02458.DOI
  33. [33]L. Del Debbio, A. Patella and C. Pica, Higher representations on the lattice: numerical simulations. SU(2) with adjoint fermions, Phys. Rev. D81, 094503 (2010), doi:10.1103/PhysRevD.81.094503, 0805.2058.DOI
  34. [34]M. Hansen, T. Janowski, C. Pica and A. Toniato, SU(2) with fundamental fermions and scalars, EPJ Web Conf. 175, 08010 (2018), doi:10.1051/epjconf/201817508010, Partial compositeness, 1710.10831.DOI
  35. [35]A. Maas, S. Martins and G. Wieland, Spectroscopy of the SU(2) scalar-fermion-gauge system under the influence of the Brout-Englert-Higgs effect (2026), 2610.33611.arxiv.org/abs/2610.33611
  36. [36]A. Maas and T. Mufti, A spectroscopical analysis of the phase diagram of Yang-Mills-Higgs theory, Phys. Rev. D91(11), 113011 (2015), doi:10.1103/PhysRevD.91.113011, 1412.6440.DOI
  37. [37]M. Wurtz and R. Lewis, Higgs and W boson spectrum from lattice simulations, Phys.Rev. D88, 054510 (2013), doi:10.1103/PhysRevD.88.054510, 1307.1492.DOI
  38. [38]H. G. Evertz, J. Jersak, C. B. Lang and T. Neuhaus, SU(2) HIGGS BOSON AND VECTOR BOSON MASSES ON THE LATTICE, Phys. Lett. B171, 271 (1986), doi:10.1016/0370-2693(86)91547-9.DOI
  39. [39]M. Constantinou et al., Parton distributions and lattice-QCD calculations: Toward 3D structure, Prog. Part. Nucl. Phys. 121, 103908 (2021), doi:10.1016/j.ppnp.2021.103908, 2006.08636.DOI
  40. [40]I. Montvay and G. Münster, Quantum fields on a lattice, Cambridge University Press, Cambridge, Cambridge, UK: Univ. Pr. (1994) 491 p. (Cambridge monographs on mathematical physics) (1994).DOI
  41. [41]O. Philipsen, M. Teper and H. Wittig, On the mass spectrum of the SU(2) Higgs model in (2+1)- dimensions, Nucl.Phys. B469, 445 (1996), doi:10.1016/0550-3213(96)00156-3, hep-lat/9602006.DOI
  42. [42]P. Jenny, A. Maas, S. Martins and G. Wieland, Weak (quasi-)PDFs from the lattice – data release (2026), doi:10.5281/zenodo.23683441.

Paper details

vlnc:W_s4mzfzgx
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