betaFeedback

WorksPhysics and Astronomy

Generalized transverse-momentum-dependent parton distributions of a photon and orbital angular momentum

Abstract

The coefficients of generalized transverse-momentum-dependent distributions (GTMDs) that encode the canonical fermion orbital angular momentum and spin–orbit correlation do not contribute to the direct transverse-momentum-dependent (TMD) and generalized-parton-distribution (GPD) projections. We present, to our knowledge, the first explicit calculation of the leading-twist, time-even, chiral-even fermion GTMDs of an on-shell real photon at zero skewness and order αα. The coefficients FA\mathcal{F}_A and GA\mathcal{G}_A, which multiply the k⊥×Δ⊥k_\perp\timesΔ_\perp structure, yield the fermion orbital angular momentum ℓzf\ell_z^f and spin–orbit correlation Czf\mathcal{C}_z^f. For photon helicity +1+1, both are obtained in closed form in the leading fermion–antifermion sector and obey Czf=−ℓzf/2\mathcal{C}_z^f=-\ell_z^f/2 for a common transverse regulator independent of xx, because the channel with the active fermion's spin anti-aligned with its orbital motion carries three times the integrated weight of the aligned channel. The projections reproduce the known photon TMDs and leading-logarithmic photon-to-fermion kernels, and energy-momentum-tensor overlaps reproduce the known one-loop photon gravitational form factors F1F_1, F2F_2, and F3F_3. Our result provides an analytic benchmark for fermionic orbital and spin–orbit structure in the real photon.

Introduction

The real photon’s fermion–antifermion component is perturbatively calculable [1]. Its structure functions and polarized parton distribution functions (PDFs) [2, 3, 4, 5], generalized parton distributions (GPDs) [6, 7, 8, 9, 10], and transverse-momentum-dependent distributions (TMDs) [11, 12, 13, 14] resolve its partonic structure. Gravitational form factors (GFFs) describe its momentum, angular-momentum, and stress distributions [15], with related studies for a dressed quark [16, 17].

Generalized transverse-momentum-dependent distributions (GTMDs) depend jointly on the parton transverse momentum k⊥\boldsymbol{k}_{\perp} and momentum transfer Δ⊥\boldsymbol{\Delta}_{\perp} [18, 19]. Setting Δ⊥=0\boldsymbol{\Delta}_{\perp}=0 gives TMDs, integrating over k⊥\boldsymbol{k}_{\perp} gives GPDs, and Fourier transformation in Δ⊥\boldsymbol{\Delta}_{\perp} gives Wigner distributions [20, 21, 22]. Parton orbital angular momentum (OAM) depends on its operator definition [23, 24, 25]. For spin-12\frac{1}{2} targets, k⊥2k_{\perp}^{2} moments of F1,4F_{1,4} and G1,1G_{1,1}, respectively, give OAM and spin–orbit correlation [26, 27, 28, 29, 30, 31]. Kinetic OAM and spin–orbit correlation also follow from GPD moments [24, 30].

Proposed exclusive measurements probe nucleon GTMDs [32, 33, 34], and lattice calculations compute the relevant moments [35]. Spin–orbit correlations have been extracted from lattice-QCD GPD moments [36], and their measurement has been proposed for the Electron-Ion Collider [37]. They have also been analyzed in the twist expansion [38], and computed for spin-0 and spin-1 hadrons and light-front two-body systems [39, 40, 41, 42]. Light-front wave-function overlaps relate these moments to the orbital and helicity content of individual amplitudes [26, 43, 44, 45].

A spin-1 target admits tensor-polarized structures [46, 47, 48], including a tensor-polarized spin–orbit correlation [49]. The real photon’s two transverse polarizations allow helicity conservation and helicity flip by two units.

In this Letter we derive the leading-twist, time-even, chiral-even fermion GTMDs of a real photon at zero skewness and order α\alpha from its leading fermion–antifermion light-front wave function. We identify the GTMDs that carry the canonical OAM and spin–orbit correlation and resolve their contributions by fermion helicity. We then obtain the TMD, GPD, and PDF projections. GTMD sum rules determine two photon GFFs, while the transverse energy-momentum tensor gives the third.

Kinematics and light-front wave function

We use the symmetric zero-skewness frame

P=p′+p2,Δ=p′−p,t=Δ2=−Δ⊥2,P=\frac{p'+p}{2},\qquad\Delta=p'-p,\qquad t=\Delta^{2}=-\Delta_{\perp}^{2},
p2=p′2=Mγ2=0,Δ+=0,P⊥=0.(1)p^{2}=p'^{2}=M_{\gamma}^{2}=0,\qquad\Delta^{+}=0,\qquad\boldsymbol{P}_{\perp}=0. \tag*{(1)}

Here pp and p′p' are the initial and final photon momenta. Our light-front conventions are v±=v0±v3v^{\pm}=v^{0}\pm v^{3} and v⋅w=(v+w−+v−w+)/2−v⊥⋅w⊥v\cdot w=(v^{+}w^{-}+v^{-}w^{+})/2-\boldsymbol{v}_{\perp}\cdot\boldsymbol{w}_{\perp}. We retain the following two-sector Fock expansion of the photon:

∣γ,h⟩=Zγ∣γ,h⟩1+∑s,sˉ∫dx d2q⊥ Ψssˉh(x,q⊥)∣ffˉ;x,q⊥,s,sˉ⟩2.(2)|\gamma,h\rangle=\sqrt{Z_{\gamma}}|\gamma,h\rangle_{1}+\sum_{s,\bar{s}}\int dx\,d^{2}\boldsymbol{q}_{\perp}\,\Psi_{s\bar{s}}^{h}(x,\boldsymbol{q}_{\perp})\left|f\bar{f};x,\boldsymbol{q}_{\perp},s,\bar{s}\right\rangle_{2}. \tag*{(2)}

The first sector is a bare photon and the second a fermion–antifermion pair; ZγZ_{\gamma} fixes the state normalization. Here h=±1h=\pm1 is the photon helicity, and s,sˉs,\bar{s} are the fermion and antifermion light-front helicities, with s,sˉ=±1s,\bar{s}=\pm1. The variables x=k+/P+∈(0,1)x=k^{+}/P^{+}\in(0,1) and q⊥\boldsymbol{q}_{\perp} are the fermion momentum fraction and relative transverse momentum. The wave function [7, 9, 50, 51, 52] is

Ψssˉh(x,q⊥)=−eef2(2π)3χs†Kh(q⊥)χ−sˉq⊥2+m2,(3)\Psi_{s\bar{s}}^{h}(x,\boldsymbol{q}_{\perp})=-\frac{ee_{f}}{\sqrt{2(2\pi)^{3}}}\frac{\chi_{s}^{\dagger}\mathcal{K}_{h}(\boldsymbol{q}_{\perp})\chi_{-\bar{s}}}{\boldsymbol{q}_{\perp}^{2}+m^{2}}, \tag*{(3)}
Kh(q⊥)=[(β σ⋅q⊥−im)σi−xσi(σ⋅q⊥)]ϵ⊥ih.\mathcal{K}_{h}(\boldsymbol{q}_{\perp})=\left[(\beta\,\boldsymbol{\sigma}\cdot\boldsymbol{q}_{\perp}-im)\sigma^{i}-x\sigma^{i}(\boldsymbol{\sigma}\cdot\boldsymbol{q}_{\perp})\right]\epsilon_{\perp i}^{h}.

Here m>0m>0 and efe_{f} are the fermion mass and charge, β=1−x\beta=1-x, ϵ⊥h=(−h,−i)/2\epsilon_{\perp}^{h}=(-h,-i)/\sqrt{2}, and χ+=(1,0)T\chi_{+}=(1,0)^{T}, $\chi_{-}= (0,1)T(0,1)^T are two-component spinors. The σi\sigma^i are Pauli matrices, with the transverse index i=1,2i=1,2 summed. Each helicity amplitude has relative orbital angular-momentum projection ℓz=h−(s+sˉ)/2\ell_z=h-(s+\bar{s})/2 [53].

Fermion GTMDs of the photon. The correlator and its leading ffˉf\bar{f} overlap are

Wh′h[Γ]=14∫dz− d2z⊥(2π)3eik⋅z⟨γ(p′),h′∣ψˉ(−z/2)UΓψ(z/2)∣γ(p),h⟩∣z+=0=Nc∑s′,s,sˉ′,sˉΨs′sˉ′h′∗(x,q+) Ss′s[Γ] Ψssˉh(x,q−).(4)\begin{aligned} W_{h'h}^{[\Gamma]}=\frac{1}{4}\int\frac{dz^-\,d^2z_\perp}{(2\pi)^3}e^{ik\cdot z}\langle\gamma(p'),h'|\bar{\psi}(-z/2)\mathcal{U}\Gamma\psi(z/2)|\gamma(p),h\rangle\big|_{z^+=0} \\ &=N_c\sum_{s',s,\bar{s}',\bar{s}}\Psi_{s'\bar{s}'}^{h'*}(x,q_+)\,S_{s's}^{[\Gamma]}\,\Psi_{s\bar{s}}^h(x,q_-). \tag*{(4)} \end{aligned}

Here ψ\psi is the fermion field, zz the field separation, and Γ=γ+\Gamma=\gamma^+ or γ+γ5\gamma^+\gamma_5 the Dirac projection. The Wilson line U\mathcal{U} is taken as unity in light-cone gauge A+=0A^+=0 with A⊥(+∞)=0A_\perp(+\infty)=0. The factor NcN_c equals 1 for a lepton and 3 for a quark flavor. For light quarks, this describes the pointlike component of the photon; its nonperturbative hadronic component is not included. The primed helicities label the final state; the spectator helicity sˉ/2\bar{s}/2 is unchanged. The relative transverse momenta are q±=k⊥±βΔ⊥/2q_\pm=k_\perp\pm\beta\Delta_\perp/2 in the final (+)(+) and initial (−)(-) wave functions. We evaluate the bare two-body overlap through O(α)O(\alpha), with α=e2/(4π)\alpha=e^2/(4\pi). The two-component reduction of the currents [54] gives the active-fermion spin matrices S[γ+]=1S^{[\gamma^+]}=1 and S[γ+γ5]=σ3S^{[\gamma^+\gamma_5]}=\sigma^3. The Supplemental Material (SM), appended after the references, gives the helicity amplitudes and overlap details.

The trace, antisymmetric part, and symmetric traceless part of the photon’s 2×22\times2 polarization matrix give scalar, vector, and tensor sectors [46]. We use the polarization structures

Ih′h=Bh′hii,Vh′h=−iϵ⊥ijBh′hij,I_{h'h}=B_{h'h}^{ii},\qquad V_{h'h}=-i\epsilon_\perp^{ij}B_{h'h}^{ij},
Th′hij=2Bh′h(ij)−δijIh′h,T~h′hij=iϵ⊥ikTh′hkj.(5)T_{h'h}^{ij}=2B_{h'h}^{(ij)}-\delta^{ij}I_{h'h},\qquad\widetilde{T}_{h'h}^{ij}=i\epsilon_\perp^{ik}T_{h'h}^{kj}. \tag*{(5)}

Here Bh′hij=ϵ⊥′∗iϵ⊥jB_{h'h}^{ij}=\epsilon_\perp^{\prime*i}\epsilon_\perp^j, with final helicity h′h' and initial helicity hh. Parentheses denote symmetrization, 2Bh′h(ij)=Bh′hij+Bh′hji2B_{h'h}^{(ij)}=B_{h'h}^{ij}+B_{h'h}^{ji}; transverse indices run over 1, 2 and repeated indices are summed, δij\delta^{ij} is the unit tensor, and ϵ⊥12=1\epsilon_\perp^{12}=1. We write k⊥×Δ⊥=kxΔy−kyΔxk_\perp\times\Delta_\perp=k_x\Delta_y-k_y\Delta_x. The structures II, VV, and TT describe scalar, circular, and linear polarization, respectively. The dual T~\widetilde{T} supplies the axial tensor structure with the same polarization components. The SM gives the decomposition and tensor identities. For h′=hh'=h, Ihh=1I_{hh}=1 and Vhh=hV_{hh}=h, while Thhij=T~hhij=0T_{hh}^{ij}=\widetilde{T}_{hh}^{ij}=0; the tensors select the helicity-flip channel with ∣h′−h∣=2|h'-h|=2. We parametrize the calculated time-even overlap with eight GTMDs. Their common arguments (x,k⊥2,k⊥⋅Δ⊥,Δ⊥2)(x,k_\perp^2,k_\perp\cdot\Delta_\perp,\Delta_\perp^2) are suppressed:

Wh′h[γ+]=Ih′hFS+iVh′hk⊥×Δ⊥m2FA+Th′hijk⊥ik⊥jm2FDk+Th′hijΔ⊥iΔ⊥jm2FDΔ,(6)W_{h'h}^{[\gamma^+]}=I_{h'h}F_S+iV_{h'h}\frac{k_\perp\times\Delta_\perp}{m^2}F_A+\frac{T_{h'h}^{ij}k_\perp^ik_\perp^j}{m^2}F_{Dk}+\frac{T_{h'h}^{ij}\Delta_\perp^i\Delta_\perp^j}{m^2}F_{D\Delta}, \tag*{(6)}
Wh′h[γ+γ5]=Vh′hGS+iIh′hk⊥×Δ⊥m2GA+T~h′hijk⊥ik⊥jm2GDk+T~h′hijΔ⊥iΔ⊥jm2GDΔ.(7)W_{h'h}^{[\gamma^+\gamma_5]}=V_{h'h}G_S+iI_{h'h}\frac{k_\perp\times\Delta_\perp}{m^2}G_A+\frac{\widetilde{T}_{h'h}^{ij}k_\perp^ik_\perp^j}{m^2}G_{Dk}+\frac{\widetilde{T}_{h'h}^{ij}\Delta_\perp^i\Delta_\perp^j}{m^2}G_{D\Delta}. \tag*{(7)}

Here FXF_X and GXG_X denote the vector (γ+)(\gamma^+) and axial (γ+γ5)(\gamma^+\gamma_5) GTMDs, respectively, with X∈{S,A,Dk,DΔ}X\in\{S,A,Dk,D\Delta\}. For Mγ=0M_\gamma=0, we normalize tensor prefactors with the fermion mass mm. Using a reference scale μ\mu in these prefactors rescales the corresponding GTMDs by μ2/m2\mu^2/m^2 and requires 1/μ21/\mu^2 in their moment definitions, preserving the correlator and observables.

Matching the photon-helicity overlaps gives

FS=Cfm2+c q+⋅q−D+D−,FA=−Cfcβm2D+D−,F_S=C_f\frac{m^2+c\,q_+\cdot q_-}{D_+D_-},\qquad F_A=-C_f\frac{c\beta m^2}{D_+D_-},
FDk=−2Cfxβm2D+D−,FDΔ=Cfxβ3m22D+D−,F_{Dk}=-2C_f\frac{x\beta m^2}{D_+D_-},\qquad F_{D\Delta}=C_f\frac{x\beta^3m^2}{2D_+D_-},
GS=Cfm2+d q+⋅q−D+D−,GA=−Cfdβm2D+D−,G_S=C_f\frac{m^2+d\,q_+\cdot q_-}{D_+D_-},\qquad G_A=-C_f\frac{d\beta m^2}{D_+D_-},
GDk=GDΔ=0.(8)G_{Dk}=G_{D\Delta}=0. \tag*{(8)}
D±=q±2+m2,Cf=Nce22π2,af=πCf.(9)D_\pm=q_\pm^2+m^2,\qquad C_f=\frac{N_c e^2}{2\pi^2},\qquad a_f=\pi C_f. \tag*{(9)}

Here c=x2+β2c=x^2+\beta^2 and d=2x−1d=2x-1; the denominators and normalization factors are

The factor afa_f normalizes the distributions after transverse integration. The axial photon-helicity-flip GTMDs GDkG_{Dk} and GDΔG_{D\Delta} vanish in the leading two-particle calculation. Thus six of the eight GTMDs are nonzero. Figure 1 shows their xx dependence at k⊥2/m2=1k_\perp^2/m^2=1. The transfer enters through βΔ⊥\beta\Delta_\perp, so its effect vanishes as x→1x\to1. At −t≲4m2-t\lesssim4m^2, FAF_A and GAG_A have peak magnitudes comparable to FSF_S and GSG_S; their relative size decreases at larger transfer. The peak of FDΔF_{D\Delta} is more than an order of magnitude smaller than those of FSF_S and GSG_S. The factor d=2x−1d=2x-1 fixes the zero of GAG_A.

Projection limits. In the forward limit, Δ⊥=0\Delta_\perp=0, the GTMDs FSF_S and GSG_S reduce to the unpolarized TMD f1γ(x,k⊥2)f_1^\gamma(x,k_\perp^2) and helicity TMD g1Lγ(x,k⊥2)g_{1L}^\gamma(x,k_\perp^2), respectively:

f1γ=Cfm2+c k⊥2(k⊥2+m2)2,g1Lγ=Cfm2+d k⊥2(k⊥2+m2)2.(10)f_1^\gamma=C_f\frac{m^2+c\,k_\perp^2}{(k_\perp^2+m^2)^2},\qquad g_{1L}^\gamma=C_f\frac{m^2+d\,k_\perp^2}{(k_\perp^2+m^2)^2}. \tag*{(10)}

These reproduce the basis light-front quantization benchmarks [11, 12] and the real-photon results of Ref. [13]. At Δ⊥=0\Delta_\perp=0, the helicity-flip amplitude is W+−[γ+]=−kL2FDk/m2W_{+-}^{[\gamma^+]}=-k_L^2F_{Dk}/m^2, with kL=kx−ikyk_L=k_x-ik_y. In the fermion-mass normalization of Eq. (6), we define f1TTγ(x,k⊥2)≡FDk∣Δ⊥=0f_{1TT}^\gamma(x,k_\perp^2)\equiv F_{Dk}|_{\Delta_\perp=0}, the real-photon counterpart of the tensor-polarized coefficient f1TTf_{1TT} [46]. Reference [13] treats real and virtual photons and normalizes f1TTf_{1TT} with Mγ2M_\gamma^2. Our definition fixes the coefficient directly from the physical-helicity overlap of a real photon.

Integrating FSF_S and GSG_S over k⊥k_\perp at fixed momentum transfer gives the helicity-nonflip vector and axial GPDs, respectively, at zero skewness:

Figure 1. All six nonzero photon GTMDs in units of Cf/m2C_f/m^2 at k⊥2/m2=1k_\perp^2/m^2=1: (a) FS\mathcal{F}_S, (b) FA\mathcal{F}_A, (c) FDk\mathcal{F}_{D_k}, (d) FDΔ\mathcal{F}_{D_\Delta}, (e) GS\mathcal{G}_S, and (f) GA\mathcal{G}_A. Colors identify −t/m2=1-t/m^2=1, 4, and 16; for each transfer the solid and dashed curves are θ=∠(k⊥,Δ⊥)=0\theta=\angle(\boldsymbol{k}_\perp,\boldsymbol{\Delta}_\perp)=0 and π/2\pi/2, and the band contains all intermediate angles.

Ff(x,t)=af[clog⁡ΛUV2m2+2β(x−cβτ)S(βτ)],F^f(x,t)=a_f\left[c\log\frac{\Lambda_{\mathrm{UV}}^2}{m^2}+2\beta(x-c\beta\tau)S(\beta\sqrt{\tau})\right],
F~f(x,t)=af[dlog⁡ΛUV2m2+2β(1−dβτ)S(βτ)].(11)\widetilde{F}^f(x,t)=a_f\left[d\log\frac{\Lambda_{\mathrm{UV}}^2}{m^2}+2\beta(1-d\beta\tau)S(\beta\sqrt{\tau})\right]. \tag*{(11)}

The GPDs are Ff=Hhh[γ+]F^f=H^{[\gamma^+]}_{hh} and F~f=hHhh[γ+γ5]\widetilde{F}^f=hH^{[\gamma^+\gamma_5]}_{hh}, where Hhh′[γ+](x,0,t)=∫d2k⊥Whh′[γ+]H^{[\gamma^+]}_{hh'}(x,0,t)=\int d^2k_\perp W^{[\gamma^+]}_{hh'}. The integration uses a shifted circular cutoff ΛUV\Lambda_{\mathrm{UV}} (SM). Here τ=−t/(4m2)\tau=-t/(4m^2) and S(z)=arsinh⁡z/[z1+z2]S(z)=\operatorname{arsinh}z/[z\sqrt{1+z^2}], with S(0)=1S(0)=1. The terms multiplying FA\mathcal{F}_A and GA\mathcal{G}_A integrate to zero by reflection of the momentum component perpendicular to Δ⊥\boldsymbol{\Delta}_\perp. The logarithms reproduce the nonflip photon-GPD kernels [7]. The finite terms use the on-shell denominator and stated cutoff; O((m2−t)/ΛUV2)O((m^2-t)/\Lambda_{\mathrm{UV}}^2) terms are omitted. For Δ⊥=Δ⊥x^\boldsymbol{\Delta}_\perp=\Delta_\perp\hat{x}, transverse integration of the FDk\mathcal{F}_{D_k} and

FDΔ\mathcal{F}_{D_\Delta} structures gives the double-helicity-flip GPD

E1(x,t)=−12m2∫d2k⊥[(kx2−ky2)FDk+Δ⊥2FDΔ]=−afxβ[1−S(βτ)].(12)\begin{aligned} E_1(x,t)=-\frac{1}{2m^2}\int d^2k_\perp\left[(k_x^2-k_y^2)\mathcal{F}_{D_k}+\Delta_\perp^2\mathcal{F}_{D_\Delta}\right] \\ &=-a_fx\beta\left[1-S(\beta\sqrt{\tau})\right]. \tag*{(12)} \end{aligned}

with H+−[γ+]=2E1H^{[\gamma^+]}_{+-}=2E_1 in our helicity convention. Its quadrupole structure agrees with Ref. [9]. At t=0t=0, these GPDs give the unpolarized and helicity PDFs; their leading logarithms (LL) are

qLLγ=afclog⁡ΛUV2m2,ΔqLLγ=afdlog⁡ΛUV2m2,(13)q^\gamma_{\mathrm{LL}}=a_fc\log\frac{\Lambda_{\mathrm{UV}}^2}{m^2},\qquad\Delta q^\gamma_{\mathrm{LL}}=a_fd\log\frac{\Lambda_{\mathrm{UV}}^2}{m^2}, \tag*{(13)}

where qγq^\gamma and Δqγ\Delta q^\gamma are the unpolarized and helicity distributions. The coefficients are the leading-order photon-to-fermion splitting kernels [2, 4]. Table I collects the projections. The FDΔ\mathcal{F}_{D_\Delta} contribution vanishes at Δ⊥=0\boldsymbol{\Delta}_\perp=0; FDk\mathcal{F}_{D_k} gives f1T⊥f_{1T}^{\perp} but has zero angular integral in a PDF. The orbital information resides in the weighted FA\mathcal{F}_A and GA\mathcal{G}_A moments.

TABLE I. TMD, GPD, PDF, and GFF projections of the six nonzero GTMDs in Eqs. (8)–(9), together with their orbital and spin–orbit moments, at leading two-particle order. A dash denotes no contribution to a projection. Daggers mark GTMDs that contribute jointly to the same GPD or GFF. The last column gives k⊥2k_\perp^2 moments; F1−1F_1-1 uses the forward subtraction in Eq. (21).

GTMDTMDGPDPDFGFFk⊥2k_\perp^2 mom.
FS\mathcal{F}_Sf1γf_1^\gammaFγF^\gammaqγq^\gammaF1−1F_1-1–
FA\mathcal{F}_A––––ℓzf\ell_z^f
FDk\mathcal{F}_{Dk}f1TTγf_{1TT}^\gammaE1γ†E_1^{\gamma\dagger}–(2tF3)†(2tF_3)^\dagger–
FDΔ\mathcal{F}_{D\Delta}–E1γ†E_1^{\gamma\dagger}–(2tF3)†(2tF_3)^\dagger–
GS\mathcal{G}_Sg1Lγg_{1L}^\gammaF~γ\widetilde{F}^\gammaΔqγ\Delta q^\gamma––
GA\mathcal{G}_A––––CzfC_z^f

Table 1.

Orbital motion and spin–orbit correlation

We define the fermion OAM ℓzf\ell_z^f and spin–orbit correlation CzfC_z^f using light-front canonical (Jaffe–Manohar) operators with ordinary transverse derivatives [23, 30]. The GTMDs FA\mathcal{F}_A and GA\mathcal{G}_A determine these moments, respectively. For equal initial and final photon helicities, h′=hh'=h, FA\mathcal{F}_A multiplies ih k⊥×Δ⊥/m2ih\,\boldsymbol{k}_\perp\times\boldsymbol{\Delta}_\perp/m^2 in the vector correlator, while GA\mathcal{G}_A multiplies ik⊥×Δ⊥/m2i\boldsymbol{k}_\perp\times\boldsymbol{\Delta}_\perp/m^2 in the axial correlator. Their moments follow the construction for the nucleon GTMDs F1,4F_{1,4} and G1,1G_{1,1} [26, 27, 30]. With our signs, the corresponding coefficients are FA\mathcal{F}_A and −GA-\mathcal{G}_A. For photon helicity h=+1h=+1,

(ℓzf,Czf)=−∫01dx∫d2k⊥ k⊥2m2(FA,GA),(14)(\ell_z^f,C_z^f)=-\int_0^1 dx\int d^2k_\perp\,\frac{k_\perp^2}{m^2}(\mathcal{F}_A,\mathcal{G}_A), \tag*{(14)}

both evaluated at Δ⊥=0⊥\boldsymbol{\Delta}_\perp=\boldsymbol{0}_\perp. The OAM adds the orbital contributions from the two fermion helicities s=±1s=\pm1. The spin–orbit correlation measures the alignment of the fermion spin with its orbital motion. The axial insertion weights each orbital contribution by ss, giving the s=+1s=+1 contribution minus the s=−1s=-1 contribution.

Both moments depend logarithmically on the ultraviolet regulator and its scheme. We use a common, xx-independent disk k⊥2<ΛUV2k_\perp^2<\Lambda_{\mathrm{UV}}^2 and write

L=log⁡(1+ΛUV2m2),R=ΛUV2ΛUV2+m2.(15)\mathcal{L}=\log\left(1+\frac{\Lambda_{\mathrm{UV}}^2}{m^2}\right),\qquad\mathcal{R}=\frac{\Lambda_{\mathrm{UV}}^2}{\Lambda_{\mathrm{UV}}^2+m^2}. \tag*{(15)}

At Δ⊥=0\boldsymbol{\Delta}_\perp=0, D+D−=(k⊥2+m2)2D_+D_-=(k_\perp^2+m^2)^2, giving

(ℓzf,Czf)=af(L−R)∫01dx β(c,d).(16)(\ell_z^f,C_z^f)=a_f(\mathcal{L}-\mathcal{R})\int_0^1 dx\,\beta(c,d). \tag*{(16)}

To understand the sign and relative size of the two moments, it is instructive to separate the helicity-channel contributions already summed in cc and dd. The factor β=1−x\beta=1-x is the active fermion’s share of the pair’s relative OAM about the transverse center of momentum. For h=+1h=+1, the two OAM-carrying channels are (s,sˉ)=(+,−)(s,\bar{s})=(+,-) and (−,+)(-,+), both with ℓz=+1\ell_z=+1 and with the active-fermion spin aligned and anti-aligned, respectively. Their wave-function numerators are proportional to x(qx+iqy)x(q_x+iq_y) and −β(qx+iqy)-\beta(q_x+iq_y), respectively (SM). Both channels have the same denominator and normalization, so taking their absolute squares gives longitudinal weights x2x^2 and β2\beta^2. Multiplying these weights by the fermion’s orbital share β\beta gives the channel contributions x2βx^2\beta and β3\beta^3:

∫01x2β dx=112,∫01β3 dx=14.\int_0^1 x^2\beta\,dx=\frac{1}{12},\qquad\int_0^1\beta^3\,dx=\frac{1}{4}.

The anti-aligned channel therefore has three times the integrated weight of the aligned channel. Their sum and signed difference give ∫01cβ dx=1/3\int_0^1 c\beta\,dx=1/3 and ∫01dβ dx=−1/6\int_0^1 d\beta\,dx=-1/6, respectively, so that

ℓzf=af3(L−R),Czf=−ℓzf2.(17)\ell_z^f=\frac{a_f}{3}(\mathcal{L}-\mathcal{R}),\qquad C_z^f=-\frac{\ell_z^f}{2}. \tag*{(17)}

The common transverse integral cancels in the ratio for any xx-independent transverse regulator. Reversing h,s,sˉh,s,\bar{s} reverses the channel weights. Since ℓz=h−(s+sˉ)/2\ell_z=h-(s+\bar{s})/2, this gives ℓzf(−h)=−ℓzf(h)\ell_z^f(-h)=-\ell_z^f(h), while the spin–orbit weight sℓzs\ell_z is unchanged. Thus Czf(h)=−hℓzf(h)/2C_z^f(h)=-h\ell_z^f(h)/2 remains negative for both photon helicities. For an unpolarized photon, the OAM averages to zero while the spin–orbit correlation remains negative.

In the perturbative dressed-quark model, all OAM-carrying amplitudes have the same active-quark helicity, so Czq=ℓzqC_z^q=\ell_z^q [43]; the photon has two OAM-carrying channels with opposite active-fermion helicity.

Figure 2(b) shows the orbital and spin–orbit densities ℓzf(x)\ell_z^f(x) and Czf(x)C_z^f(x), defined by the integrands of the xx integral in Eq. (14), divided by af(L−R)a_f(\mathcal{L}-\mathcal{R}). At x→0x\to0, the anti-aligned channel dominates, giving normalized densities +1+1 and −1-1. Equal channel weights make Czf(x)C_z^f(x) vanish at x=1/2x=1/2, while the OAM remains positive. The aligned channel makes the spin–orbit density positive for x>1/2x > 1/2. Both densities vanish as the fermion’s orbital share β→0\beta\to0. Their ratio is Czf(x)/ℓzf(x)=d/cC_z^f(x)/\ell_z^f(x) = d/c for a common spin-independent transverse weight. Charge conjugation gives Czfˉ=CzfC_z^{\bar f} = C_z^f.

Figure 2. For photon helicity +1+1: (a) total constituent-spin and orbital fractions of JzJ_z in the leading fixed-flavor ffˉf\bar f sector, Eq. (20). (b) Single-fermion orbital and spin–orbit densities divided by af(L−R)a_f(\mathcal{L}-\mathcal{R}), giving cβc\beta and dβd\beta, respectively; LL and RR are defined in Eq. (15).

We next examine how constituent spin and orbital motion share the angular momentum of the photon’s two-particle sector, as illustrated in Figure 2(a). For h=+1h = +1, the bare pair probability and total constituent spin are

Pffˉ=Nc∑s,sˉ∫01dx∫d2q⊥∣Ψssˉ+(x,q⊥)∣2,P_{f\bar f} = N_c \sum_{s,\bar{s}} \int_0^1 dx \int d^2q_\perp\left|\Psi^{+}_{s\bar{s}}(x,\boldsymbol{q}_\perp)\right|^2,
Szffˉ=Nc∑s,sˉ∫01dx∫d2q⊥s+sˉ2∣Ψssˉ+(x,q⊥)∣2.S_z^{f\bar f} = N_c \sum_{s,\bar{s}} \int_0^1 dx \int d^2q_\perp\frac{s+\bar{s}}{2}\left|\Psi^{+}_{s\bar{s}}(x,\boldsymbol{q}_\perp)\right|^2.

Here (s+sˉ)/2(s+\bar{s})/2 is the total constituent helicity. Inserting Eq. (3) and integrating with the same transverse cutoff gives

Szffˉ=afR,Pffˉ=af(23L+13R).(18)S_z^{f\bar f} = a_f\mathcal{R}, \qquad P_{f\bar f} = a_f\left(\frac{2}{3}\mathcal{L}+\frac{1}{3}\mathcal{R}\right). \tag*{(18)}

The state normalization gives Zγ−1=1+PffˉZ_\gamma^{-1} = 1 + P_{f\bar f} and Zγ=1−Pffˉ+O(α2)Z_\gamma= 1 - P_{f\bar f} + O(\alpha^2). The perturbative truncation requires Pffˉ≪1P_{f\bar f} \ll1. The spin SzffˉS_z^{f\bar f} includes both constituents. Charge conjugation gives equal integrated fermion and antifermion OAM, so Lzffˉ=2ℓzfL_z^{f\bar f}=2\ell_z^{f} and

Zγ+Szffˉ+Lzffˉ=1+O(α2).(19)Z_\gamma+ S_z^{f\bar f} + L_z^{f\bar f} = 1 + O(\alpha^2). \tag*{(19)}

Since Jz=hJ_z = h holds channel by channel, this checks that the FA\mathcal{F}_A moment reproduces the wave-function OAM. The leading logarithm of the pair OAM and this helicity sum rule have analogs for a polarized gluon [55]; here the GTMDs also resolve the fermion share and spin–orbit correlation.

Figure 2(a) plots, within the regulated bare two-body sector, the fractions of angular momentum carried by spin and orbital motion,

SzffˉPffˉ=3R2L+R,LzffˉPffˉ=2(L−R)2L+R.(20)\frac{S_z^{f\bar f}}{P_{f\bar f}} = \frac{3\mathcal{R}}{2\mathcal{L}+\mathcal{R}}, \qquad\frac{L_z^{f\bar f}}{P_{f\bar f}} = \frac{2(\mathcal{L}-\mathcal{R})}{2\mathcal{L}+\mathcal{R}}. \tag*{(20)}

so the two curves sum to unity at every cutoff. At large cutoff, the spin approaches afa_f while the OAM grows logarithmically, making the orbital fraction approach unity.

GTMD sum rules and gravitational form factors

The xx-weighted (second Mellin) moment of the vector bilocal gives the fermion energy-momentum tensor (EMT) matrix element [24, 56]. We normalize the full T++T^{++} matrix element by 2(P+)22(P^+)^2. In Milton’s photon GFF basis [57], its diagonal entry is F1(t)F_1(t) and helicity flip determines F3(t)F_3(t). Adding the equal antifermion contribution and the one-photon normalization term gives the GTMD sum rules for F1F_1 and F3F_3, up to O(α2)O(\alpha^2) corrections:

F1(t)−1=2∫01dx x∫d2k⊥[FS(x,k⊥,Δ⊥)−FS(x,k⊥,0⊥)],(21)F_1(t)-1 = 2\int_0^1 dx\,x\int d^2k_\perp\left[\mathcal{F}_S(x,\boldsymbol{k}_\perp,\boldsymbol{\Delta}_\perp)-\mathcal{F}_S(x,\boldsymbol{k}_\perp,\boldsymbol{0}_\perp)\right], \tag*{(21)}
2tF3(t)=2m2∫01dx x∫d2k⊥[(kx2−ky2)FDk+Δ⊥2FDΔ],(22)2tF_3(t) = \frac{2}{m^2}\int_0^1 dx\,x\int d^2k_\perp\left[(k_x^2-k_y^2)F_{Dk}+\Delta_\perp^2F_{D\Delta}\right], \tag*{(22)}

for Δ⊥=Δ⊥x^\boldsymbol{\Delta}_\perp=\Delta_\perp\hat{\boldsymbol{x}}. The one-photon normalization Zγ=1−Pffˉ+O(α2)Z_\gamma=1-P_{f\bar f}+O(\alpha^2) supplies the forward subtraction in Eq. (21), removing the ultraviolet logarithm and enforcing F1(0)=1F_1(0)=1. Comparison with Eq. (12) gives 2tF3=−4∫01dx x E1(x,t)2tF_3=-4\int_0^1dx\,x\,E_1(x,t), so the quadrupole that defines E1E_1 also fixes F3F_3. The resulting closed forms are

F1(t)−1af=3518−1367−(4τ−5)η3τ−(2τ−1)η22τ2,(23)\frac{F_1(t)-1}{a_f}=\frac{35}{18}-\frac{13}{67}-\frac{(4\tau-5)\eta}{3\tau}-\frac{(2\tau-1)\eta^2}{2\tau^2}, \tag*{(23)}
2tF3(t)af=13−7τ+6ητ−(2τ−1)η2τ2.(24)\frac{2tF_3(t)}{a_f}=\frac{1}{3}-\frac{7}{\tau}+\frac{6\eta}{\tau}-\frac{(2\tau-1)\eta^2}{\tau^2}. \tag*{(24)}

Here η=arsinh⁡τ\eta=\operatorname{arsinh}\sqrt{\tau} and r=1+1/τr=\sqrt{1+1/\tau}. For a unit-charge lepton, af=α/(2π)a_f=\alpha/(2\pi), and these expressions agree term by term with the one-loop QED results of Ref. [15], which reproduce Refs. [57, 58]. Figure 3 compares them with direct numerical integration of Eqs. (21)–(22) (SM).

Figure 3. Photon GFFs (a) (F1−1)/af(F_1-1)/a_f and (b) 2tF3/af2tF_3/a_f. Solid curves: analytic results, Eqs. (23)–(24); open symbols: numerical GTMD sum rules, Eqs. (21)–(22).

The remaining form factor F2F_2 probes the transverse EMT. For Δ⊥=Qx^\boldsymbol{\Delta}_\perp=Q\hat{\boldsymbol{x}}, Q=−tQ=\sqrt{-t}, the full matrix element in Milton’s basis [57], with our helicity conventions, obeys ⟨p′,+∣Tyy∣p,−⟩=−t2F2(t)\langle p',+|T^{yy}|p,-\rangle=-t^2F_2(t) and ⟨p′,+∣Txx∣p,−⟩=0\langle p',+|T^{xx}|p,-\rangle=0. Subtracting the zero-transfer pair overlap at the same transverse cutoff before removing the cutoff gives (SM)

2tF2(t)af=13+3τ−2rητ−η2τ2.(25)\frac{2tF_2(t)}{a_f}=\frac{1}{3}+\frac{3}{\tau}-\frac{2r\eta}{\tau}-\frac{\eta^2}{\tau^2}. \tag*{(25)}

which matches the one-loop QED result in Eq. (35c) of Ref. [15].

Conclusion. We have calculated the leading-twist, time-even, chiral-even fermion GTMDs of a real photon at zero skewness and order α\alpha. Six of the eight GTMDs are nonzero. The FAF_A and GAG_A moments trace canonical OAM and spin–orbit correlation to two orbital channels with opposite fermion helicities. Their 1:3 aligned-to-anti-aligned weights explain Czf=−ℓzf/2C_z^f=-\ell_z^f/2 for photon helicity +1. The negative sign reflects the greater weight of the anti-aligned channel. In the dressed-quark model, a single active-quark helicity instead gives Czq=ℓzqC_z^q=\ell_z^q. The ratio Czf/ℓzf=−1/2C_z^f/\ell_z^f=-1/2 is exact for a common xx-independent transverse regulator, although each moment is cutoff dependent. The pair’s OAM contains the ultraviolet logarithm, while its spin remains finite; adding the bare photon satisfies the angular-momentum sum rule. Agreement with photon TMDs, splitting kernels, and all three one-loop GFFs tests the same wave function through parton projections and distinct EMT components: F1F_1 and F3F_3 follow from GTMD sum rules, while F2F_2 requires the transverse EMT. These closed forms provide a perturbative reference for orbital and spin–orbit structure. Higher Fock sectors and radiative corrections can test how additional helicity channels modify the ratio.

Acknowledgments. S.N. acknowledges support from the Fundamental Research Funds for the Central Universities under Grant Nos. lzujbky-2023-stlt01, lzujbky2024-oy02, and lzujbky-2025-eyt01, and from the National Natural Science Foundation of China under Grant No. 12335003. AI use: OpenAI’s GPT-6 Astra and Anthropic’s Claude Opus 5.5 were used for proofreading and assistance with two rounds of simulated peer review.

References

Supplemental Material for

“Generalized transverse-momentum-dependent parton distributions of a photon and orbital angular momentum”

Helicity overlaps and polarization structure

Helicity overlaps

The two-particle wave function in Letter Eq. (3) has the helicity numerators

N+=2(imxqR−βqR0)(26)N^{+}=\sqrt{2}\begin{pmatrix} im & xq_{R} \\ -\beta q_{R} & 0 \end{pmatrix} \tag*{(26)}
N−=2(0βqL−xqL−im).(27)N^{-}=\sqrt{2}\begin{pmatrix} 0 & \beta q_{L} \\ -xq_{L} & -im \end{pmatrix}. \tag*{(27)}

Here Nssˉh=χs†Khχ−sˉN^{h}_{s\bar{s}}=\chi_{s}^{\dagger}\mathcal{K}_{h}\chi_{-\bar{s}}, β=1−x\beta=1-x, and qR,L=qx±iqyq_{R,L}=q_{x}\pm iq_{y}. The fermion and antifermion light-front helicities s/2s/2 and sˉ/2\bar{s}/2 label the rows and columns, respectively, with s,sˉ=±1s,\bar{s}=\pm1; h=±1h=\pm1 is the photon helicity. The mass terms have ℓz=0\ell_{z}=0 and the momentum terms ℓz=h\ell_{z}=h. The ffˉf\bar{f} component is generated by annihilating the photon in the initial ket, so the vertex selects the term ϵ⊥hah\epsilon_{\perp}^{h}a_{h} in Ai⊥A_{i}^{\perp} and the wave function contains the unstarred ϵ⊥h\epsilon_{\perp}^{h} [51, 52]. The conjugate vector multiplies the creation operator ah†a_{h}^{\dagger}, as in the field expansion of Freese and Cosyn [Eq. (10)] of Ref. [15]. The starred form printed in Refs. [7, 9, 50] uses ϵ⊥h∗=−ϵ⊥−h\epsilon_{\perp}^{h*}=-\epsilon_{\perp}^{-h} in this circular basis, so matching its numerator requires the helicity relabeling h→−hh\to-h and an overall phase. For a real photon the parent invariant mass vanishes. The vertex and light-front energy denominator contain the same longitudinal factor:

Kh/[x(1−x)]−(q⊥2+m2)/[x(1−x)]=−Khq⊥2+m2.(28)\frac{\mathcal{K}_{h}/[x(1-x)]}{-(q_{\perp}^{2}+m^{2})/[x(1-x)]}=-\frac{\mathcal{K}_{h}}{q_{\perp}^{2}+m^{2}}. \tag*{(28)}

This cancellation gives the denominator in Letter Eq. (3); mm remains the constituent mass.

For h=+1h=+1, the two OAM-carrying entries of Eq. (S1) give

∣N+,−+∣2=∣2xqR∣2=2x2q⊥2,\left|N^{+}_{+,-}\right|^{2}=\left|\sqrt{2}xq_{R}\right|^{2}=2x^{2}q_{\perp}^{2},
∣N−,++∣2=∣−2βqR∣2=2β2q⊥2.\left|N^{+}_{-,+}\right|^{2}=\left|-\sqrt{2}\beta q_{R}\right|^{2}=2\beta^{2}q_{\perp}^{2}.

Here ∣qR∣2=q⊥2\left|q_{R}\right|^{2}=q_{\perp}^{2}. The two wave functions share the same denominator and normalization, leaving longitudinal weights x2x^{2} and β2\beta^{2}. The minus sign in N−,++N^{+}_{-,+} drops out of its absolute square; the relative minus sign in the spin–orbit moment instead comes from the axial weight ss. Including the mass term, the channel sums are

∑ssˉ((s+sˉ)/2ℓzsℓz)∣Nssˉh∣2=2(m2+cq⊥2m2dq⊥2).(29)\sum_{s\bar{s}}\begin{pmatrix}(s+\bar{s})/2 \\ \ell_{z} \\ s\ell_{z}\end{pmatrix}\left|N^{h}_{s\bar{s}}\right|^{2}=2\begin{pmatrix}m^{2}+cq_{\perp}^{2} \\ m^{2} \\ dq_{\perp}^{2}\end{pmatrix}. \tag*{(29)}

Here c=x2+β2c=x^{2}+\beta^{2} and d=2x−1d=2x-1. Multiplication by Cf/[2(q⊥2+m2)2]C_{f}/[2(q_{\perp}^{2}+m^{2})^{2}] and integration give the pair probability PffˉP_{f\bar{f}}, spin SzffˉS^{f\bar{f}}_{z}, and orbital angular momentum LzffˉL^{f\bar{f}}_{z}; the fourth entry acquires the fermion orbital share β\beta for CzfC^{f}_{z}. Here Cf=Ncαe2/(2π2)C_{f}=N_{c}\alpha_{e}^{2}/(2\pi^{2}), af=πCfa_{f}=\pi C_{f}, α=e2/(4π)\alpha=e^{2}/(4\pi), and Nc=1N_{c}=1 for a lepton and 3 for a quark flavor.

At zero skewness, spinor completeness gives the helicity correlators

Wh′h[Γ]=Cf2D+D−tr⁡[S[Γ]Kh(q−)Kh′†(q+)].(30)W^{[\Gamma]}_{h'h}=\frac{C_{f}}{2D_{+}D_{-}}\operatorname{tr}\left[S^{[\Gamma]}\mathcal{K}_{h}(q_{-})\mathcal{K}^{\dagger}_{h'}(q_{+})\right]. \tag*{(30)}

Here Δ⊥\Delta_{\perp} is the transverse momentum transfer, q±=k⊥±βΔ⊥/2q_{\pm}=k_{\perp}\pm\beta\Delta_{\perp}/2 are the relative transverse momenta in the final (++) and initial (−-) wave functions, and D±=q±2+m2D_{\pm}=q_{\pm}^{2}+m^{2}. S[γ+]=1S^{[\gamma^{+}]}=1 and S[γ+γ5]=σ3S^{[\gamma^{+}\gamma^{5}]}=\sigma^{3}. The label h′h' denotes the final photon helicity. The helicity projections needed below are

Whh[γ+]=FS+ihk⊥×Δ⊥m2FA,W^{[\gamma^{+}]}_{hh}=\mathcal{F}_{S}+ih\frac{k_{\perp}\times\Delta_{\perp}}{m^{2}}\mathcal{F}_{A},
Whh[γ+γ5]=hGS+ik⊥×Δ⊥m2GA,W^{[\gamma^{+}\gamma^{5}]}_{hh}=h\mathcal{G}_{S}+i\frac{k_{\perp}\times\Delta_{\perp}}{m^{2}}\mathcal{G}_{A},
W+−[γ+]=−kL2FDK+m2FDΔm2.(31)W^{[\gamma^{+}]}_{+-}=-\frac{k_{L}^{2}\mathcal{F}_{DK}+m^{2}\mathcal{F}_{D\Delta}}{m^{2}}. \tag*{(31)}

where kL=kx−ikyk_{L}=k_{x}-ik_{y}, ΔL=Δx−iΔy\Delta_{L}=\Delta_{x}-i\Delta_{y}, and W−+[γ+]=(W+−[γ+])∗W^{[\gamma^{+}]}_{-+}=(W^{[\gamma^{+}]}_{+-})^{*}, with W+−[γ+γ5]=W−+[γ+γ5]=0W^{[\gamma^{+}\gamma^{5}]}_{+-}=W^{[\gamma^{+}\gamma^{5}]}_{-+}=0. The two opposite-fermion-helicity channels have equal helicity-flip products. Their opposite axial weights s=±1s=\pm1 cancel, giving GDk=GDΔ=0\mathcal{G}_{Dk}=\mathcal{G}_{D\Delta}=0 for this two-body overlap.

Polarization decomposition and tensor redundancy

The transverse polarization matrix decomposes as

Bh′hij=12δijIh′h+i2ϵijVh′h+12Th′hij.(32)B^{ij}_{h'h}=\frac{1}{2}\delta^{ij}\mathcal{I}_{h'h}+\frac{i}{2}\epsilon^{ij}\mathcal{V}_{h'h}+\frac{1}{2}\mathcal{T}^{ij}_{h'h}. \tag*{(32)}

Here Bh′hij=ϵih′∗ϵjhB^{ij}_{h'h}=\epsilon^{h'*}_{i}\epsilon^{h}_{j}, Ih′h=Bh′hii\mathcal{I}_{h'h}=B^{ii}_{h'h}, Vh′h=−iϵijBh′hij\mathcal{V}_{h'h}=-i\epsilon^{ij}B^{ij}_{h'h}, and Th′hij=Bh′hij+Bh′hji−δijIh′h\mathcal{T}^{ij}_{h'h}=B^{ij}_{h'h}+B^{ji}_{h'h}-\delta^{ij}\mathcal{I}_{h'h}, with ϵ⊥12=1\epsilon^{12}_{\perp}=1. A 2×22\times2 matrix has one trace component, one antisymmetric component, and two symmetric traceless components. These exhaust the transverse polarization structures. Under transverse rotations, Ih′h\mathcal{I}_{h'h} and Vh′h\mathcal{V}_{h'h} are scalars; Vh′h\mathcal{V}_{h'h} changes sign under a transverse reflection and represents longitudinal vector polarization (helicity). The tensor components describe linear polarization.

For a diagonal polarization state, Bhhji=(Bhhij)∗B^{ji}_{hh}=(B^{ij}_{hh})^{*}, so ϵ⊥ijBhhij\epsilon^{ij}_{\perp}B^{ij}_{hh} is purely imaginary. The factor −i-i makes Vhh\mathcal{V}_{hh} real and gives Vhh=h\mathcal{V}_{hh}=h for ϵ⊥±=(−h,−i)/2\epsilon_{\perp}^{\pm}=(-h,-i)/\sqrt{2}. The dual tensor and its helicity projections are

T~h′hij=iε⊥ikTh′hkj,Ih′h=δh′h,Vh′h=hδh′h,\widetilde{\mathcal{T}}^{ij}_{h'h}=i\varepsilon^{ik}_{\perp}\mathcal{T}^{kj}_{h'h}, \qquad\mathcal{I}_{h'h}=\delta_{h'h},\qquad\mathcal{V}_{h'h}=h\delta_{h'h},
Thh=T~hh=0,T+−=(−1ii1),\mathcal{T}_{hh}=\widetilde{\mathcal{T}}_{hh}=0,\qquad \mathcal{T}_{+-}=\begin{pmatrix}-1&i\\i&1\end{pmatrix},
T−+=T+−∗,T~+−=T+−,T~−+=−T−+.(33)\mathcal{T}_{-+}=\mathcal{T}^{*}_{+-},\qquad \widetilde{\mathcal{T}}_{+-}=\mathcal{T}_{+-},\qquad \widetilde{\mathcal{T}}_{-+}=-\mathcal{T}_{-+}. \tag*{(33)}

The subscripts denote final and initial photon helicities; transverse matrix indices are suppressed in the last two lines. Contracting with ε⊥ik\varepsilon^{ik}_{\perp} preserves symmetry and tracelessness in two dimensions and gives the opposite reflection parity needed in the axial correlator. The factor ii fixes the displayed helicity phases.

The symmetric traceless transverse contractions obey

Δ⊥2Ukk−2ωUkΔ+k⊥2UΔΔ=0.(34)\Delta_{\perp}^{2}U_{kk}-2\omega U_{k\Delta}+k_{\perp}^{2}U_{\Delta\Delta}=0. \tag*{(34)}

Here Uab=Uija⊥ib⊥jU_{ab}=U^{ij}a^{i}_{\perp}b^{j}_{\perp}, with UijU^{ij} symmetric and traceless, and ω=k⊥⋅Δ⊥\omega=\boldsymbol{k}_{\perp}\cdot\boldsymbol{\Delta}_{\perp}. The mixed contraction can be eliminated for ω≠0\omega\ne0. Extracting FDk\mathcal{F}_{Dk} and FDΔ\mathcal{F}_{D\Delta} from −m2W[γ+]-m^{2}W^{[\gamma^{+}]} has determinant −2ω(k⊥×Δ⊥)-2\omega(\boldsymbol{k}_{\perp}\times\boldsymbol{\Delta}_{\perp}). For the calculated overlap, the explicit GTMDs have positive denominators D+D−D_{+}D_{-} for m>0m>0 and no pole at ω=0\omega=0. Their forward limits are

FDk∣Δ⊥=0=−2Cfxβm2(k⊥2+m2)2,FDΔ∣Δ⊥=0=Cfxβ3m22(k⊥2+m2)2.(35)\left.\mathcal{F}_{Dk}\right|_{\Delta_{\perp}=0} =-\frac{2C_{f}x\beta m^{2}}{(k_{\perp}^{2}+m^{2})^{2}}, \qquad \left.\mathcal{F}_{D\Delta}\right|_{\Delta_{\perp}=0} =\frac{C_{f}x\beta^{3}m^{2}}{2(k_{\perp}^{2}+m^{2})^{2}}. \tag*{(35)}

These functions extend smoothly to parallel and orthogonal momenta. At Δ⊥=0\Delta_{\perp}=0, the tensor multiplying FDΔ\mathcal{F}_{D\Delta} vanishes; the coefficient in Eq. (35) is determined by continuity from nonzero Δ⊥\Delta_{\perp}. The eight-function parametrization is specific to the calculated time-even overlap.

Transverse integrals

The common denominator is

D+D−=(m2+k⊥2+β2Δ⊥24)2−β2(k⊥⋅Δ⊥)2.(36)D_{+}D_{-}=\left(m^{2}+k_{\perp}^{2}+\frac{\beta^{2}\Delta_{\perp}^{2}}{4}\right)^{2} -\beta^{2}(\boldsymbol{k}_{\perp}\cdot\boldsymbol{\Delta}_{\perp})^{2}. \tag*{(36)}

Feynman parametrization gives

1D+D−=∫01du[k⊥2+μu2]2,μu2=m2+u(1−u)β2Δ⊥2.(37)\frac{1}{D_{+}D_{-}}=\int_{0}^{1}\frac{\mathrm{d}u}{[k_{\perp}^{2}+\mu_{u}^{2}]^{2}}, \qquad \mu_{u}^{2}=m^{2}+u(1-u)\beta^{2}\Delta_{\perp}^{2}. \tag*{(37)}

Here κ⊥=k⊥+(2u−1)βΔ⊥/2\boldsymbol{\kappa}_{\perp}=\boldsymbol{k}_{\perp}+(2u-1)\beta\boldsymbol{\Delta}_{\perp}/2 is the shifted integration momentum. The full-plane integral is ∫d2κ⊥/(κ⊥2+μu2)2=π/μu2\int\mathrm{d}^{2}\kappa_{\perp}/(\kappa_{\perp}^{2}+\mu_{u}^{2})^{2}=\pi/\mu_{u}^{2}. With the shifted circular cutoff κ⊥2<ΛUV2\kappa_{\perp}^{2}<\Lambda_{\mathrm{UV}}^{2},

∫κ⊥2<ΛUV2d2κ⊥ κ⊥2[κ⊥2+μu2]2=π[log⁡(1+ΛUV2μu2)−ΛUV2ΛUV2+μu2].(38)\int_{\kappa_{\perp}^{2}<\Lambda_{\mathrm{UV}}^{2}} \frac{\mathrm{d}^{2}\kappa_{\perp}\,\kappa_{\perp}^{2}} {[\kappa_{\perp}^{2}+\mu_{u}^{2}]^{2}} =\pi\left[ \log\left(1+\frac{\Lambda_{\mathrm{UV}}^{2}}{\mu_{u}^{2}}\right) -\frac{\Lambda_{\mathrm{UV}}^{2}}{\Lambda_{\mathrm{UV}}^{2}+\mu_{u}^{2}} \right]. \tag*{(38)}

The parameter integrals give

∫01duμu2=S(βτ)m2,\int_{0}^{1}\frac{\mathrm{d}u}{\mu_{u}^{2}}=\frac{S(\beta\sqrt{\tau})}{m^{2}},
β2Δ⊥2∫01u(1−u)duμu2=1−S(βτ).(39)\beta^{2}\Delta_{\perp}^{2}\int_{0}^{1}\frac{u(1-u)\mathrm{d}u}{\mu_{u}^{2}} =1-S(\beta\sqrt{\tau}). \tag*{(39)}

Here the transfer variable is τ=−t/(4m2)\tau=-t/(4m^{2}), with t=−Δ⊥2t=-\Delta_{\perp}^{2}, and the function is S(z)=arsinh⁡z/[z1+z2]S(z)=\operatorname{arsinh}z/[z\sqrt{1+z^{2}}], with S(0)=1S(0)=1. The exact finite-cutoff parton densities are

(qγ,Δqγ)(x,ΛUV2)=af[(c,d)L+2β(x,1)R].(40)(q^{\gamma},\Delta q^{\gamma})(x,\Lambda_{\mathrm{UV}}^{2}) =a_{f}\left[(c,d)\mathcal{L}+2\beta(x,1)\mathcal{R}\right]. \tag*{(40)}

Here L=log⁡(1+ΛUV2/m2)\mathcal{L}=\log(1+\Lambda_{\mathrm{UV}}^{2}/m^{2}) and R=ΛUV2/(ΛUV2+m2)\mathcal{R}=\Lambda_{\mathrm{UV}}^{2}/(\Lambda_{\mathrm{UV}}^{2}+m^{2}). Their zeroth moments give PffˉP_{f\bar f} and SzffˉS_{z}^{f\bar f}, respectively. Here SzffˉS_{z}^{f\bar f} is the constituent spin in a regulated Fock sector. The real-photon sum rule ∫01g1γ(x) dx=0\int_{0}^{1}g_{1}^{\gamma}(x)\,\mathrm{d}x=0 [5] concerns the physical polarized structure function, with its factorization prescription and coefficient functions. Identifying it with the cutoff density Δqγ\Delta q^{\gamma} requires matching these definitions. The finite GPD terms use the shifted cutoff in Eq. (38); the forward orbital moments use k⊥2<ΛUV2k_{\perp}^{2}<\Lambda_{\mathrm{UV}}^{2}. The F2F_{2} integrals below retain the original relative-momentum disk. For the logarithmically divergent GPD integrals, changing between disks in the original and Feynman-shifted momenta changes only terms that vanish as ΛUV→∞\Lambda_{\mathrm{UV}}\to\infty. The bare GPDs retain their ultraviolet logarithms. The quadratically divergent F2F_{2} kernel requires the original disk and the subtraction specified below.

Derivation of the orbital and spin–orbit moments

For h=+1h=+1 and k⊥≠0k_{\perp}\ne0, the overlap expands, up to O(Δ⊥2)O(\Delta_{\perp}^{2}), as

Whh[Γ]=Nc∑ssˉwssˉΓAssˉ2[1−iℓzβ(k⊥×Δ⊥)k⊥2].(41)W_{hh}^{[\Gamma]}=N_{c}\sum_{s\bar{s}}w_{s\bar{s}}^{\Gamma}A_{s\bar{s}}^{2} \left[1-i\ell_{z}\frac{\beta(\boldsymbol{k}_{\perp}\times\boldsymbol{\Delta}_{\perp})}{k_{\perp}^{2}}\right]. \tag*{(41)}

where Ψssˉ+=Assˉ(x,q⊥2)eiℓzϕq\Psi^{+}_{s\bar{s}}=A_{s\bar{s}}(x,q_{\perp}^{2})e^{i\ell_{z}\phi_{q}}, with real AssˉA_{s\bar{s}} after absorbing constant channel phases; w=1w=1 for Γ=γ+\Gamma=\gamma^{+} and w=sw=s for Γ=γ+γ5\Gamma=\gamma^{+}\gamma_{5}. The azimuths ϕ±\phi_{\pm} of q±q_{\pm} obey ϕ+−ϕ−=β(k⊥×Δ⊥)/k⊥2+O(Δ⊥2)\phi_{+}-\phi_{-}=\beta(\boldsymbol{k}_{\perp}\times\boldsymbol{\Delta}_{\perp})/k_{\perp}^{2}+O(\Delta_{\perp}^{2}). At Δ⊥=0\Delta_{\perp}=\boldsymbol{0}, matching Eq. (31) gives

−k⊥2m2(FA,GA)=Ncβ∑ssˉ(1,s)ℓzAssˉ2.(42)-\frac{k_{\perp}^{2}}{m^{2}}(\mathcal{F}_{A},\mathcal{G}_{A}) =N_{c}\beta\sum_{s\bar{s}}(1,s)\ell_{z}A_{s\bar{s}}^{2}. \tag*{(42)}

with FS(x,k⊥,0⊥)=Nc∑A2\mathcal{F}_{S}(x,k_{\perp},\boldsymbol{0}_{\perp})=N_{c}\sum A^{2}. Equation (S16) identifies the weighted GTMDs at each (x,k⊥)(x,k_{\perp}) with the wave-function probabilities weighted by the fermion’s orbital share βℓz\beta\ell_{z} and by sβℓzs\beta\ell_{z}, respectively. Integrating over xx and k⊥k_{\perp} gives the moments in Letter Eq. (14). Pair overlaps use Zγ=1Z_{\gamma}=1 at O(α)O(\alpha).

Gravitational moments

The fermionic energy-momentum tensor (EMT) moment is

∫−11dx xHhh′[γ+](x,0,t)=⟨p′,h′∣T++∣p,h⟩2(P+)2.(43)\int_{-1}^{1} dx\,x H_{hh'}^{[\gamma^+]}(x,0,t)=\frac{\langle p',h'|T^{++}|p,h\rangle}{2(P^+)^2}. \tag*{(43)}

Here H[γ+]=∫d2k⊥W[γ+]H^{[\gamma^+]}=\int d^2k_\perp W^{[\gamma^+]} and P+=p+=p′+P^+=p^+=p^{\prime+}. We use Milton’s photon tensors [57], crossed to the scattering channel and normalized to the positive free-photon T++T^{++} matrix element. The tensors multiplying F2F_2 and F3F_3 are

Θ2μν=−(ΔμΔν−tgμν)\Theta_2^{\mu\nu}=-(\Delta^\mu\Delta^\nu-tg^{\mu\nu})
×[2(Δ⋅ϵ)(Δ⋅ϵ′∗)−t(ϵ⋅ϵ′∗)],{}\times\left[2(\Delta\cdot\epsilon)(\Delta\cdot\epsilon^{\prime*})-t(\epsilon\cdot\epsilon^{\prime*})\right],
Θ3μν=−4PμPν[2(Δ⋅ϵ)(Δ⋅ϵ′∗)−t(ϵ⋅ϵ′∗)].\Theta_3^{\mu\nu}=-4P^\mu P^\nu\left[2(\Delta\cdot\epsilon)(\Delta\cdot\epsilon^{\prime*}) -t(\epsilon\cdot\epsilon^{\prime*})\right].

Here P=(p′+p)/2P=(p'+p)/2, Δ=p′−p\Delta=p'-p, t=Δ2t=\Delta^2, and ϵ\epsilon and ϵ′∗\epsilon^{\prime*} are the initial and final photon polarization vectors. For Δ⊥=Δ⊥x^\Delta_\perp=\Delta_\perp\hat{x}, the full EMT obeys ⟨T++⟩hh/[2(P+)2]=F1\langle T^{++}\rangle_{hh}/[2(P^+)^2]=F_1 and ⟨T++⟩+−/[2(P+)2]=−2tF3\langle T^{++}\rangle_{+-}/[2(P^+)^2]=-2tF_3. The equal antifermion moment doubles the fermion term; the one-photon normalization fixes F1(0)=1F_1(0)=1 and has no helicity flip. The resulting finite integrals are

F1−1af=2∫01dx x[(1−c−2cβ2τ)S(βτ)−(1−c)],\frac{F_1-1}{a_f}=2\int_0^1 dx\,x\left[(1-c-2c\beta^2\tau)\mathcal{S}(\beta\sqrt{\tau})-(1-c)\right],
2tF3af=4∫01dx x2β[1−S(βτ)].(44)\frac{2tF_3}{a_f}=4\int_0^1 dx\,x^2\beta\left[1-\mathcal{S}(\beta\sqrt{\tau})\right]. \tag*{(44)}

Transverse-EMT overlap for F2F_2

Overlap and forward subtraction

For on-shell photons, we use the symmetric frame p⊥′=−p⊥=Δ⊥/2p'_\perp=-p_\perp=\Delta_\perp/2, with Δ⊥=Qx^\Delta_\perp=Q\hat{x}, t=−Q2t=-Q^2, and Q>0Q>0. The transverse component perpendicular to the transfer gives

⟨p′,+∣Tyy∣p,−⟩=−t2F2(t),⟨p′,+∣Txx∣p,−⟩=0.(45)\langle p',+|T^{yy}|p,-\rangle=-t^2F_2(t),\qquad\langle p',+|T^{xx}|p,-\rangle=0. \tag*{(45)}

The minus sign follows from the tensors in Sec. S4 and the polarization vectors used in Letter Eq. (3).

The reduced one-fermion overlap kernel is

Of;s′syy(k′,k)=⟨f;k′,s′∣Tfyy(0)∣f;k,s⟩k′+k+=2qy2+isqyQxP+δs′s.(46)\begin{aligned} \mathcal{O}^{yy}_{f;s's}(k',k)=\frac{\langle f;k',s'|T_f^{yy}(0)|f;k,s\rangle}{\sqrt{k^{\prime+}k^+}} \\ &=\frac{2q_y^2+isq_yQ}{xP^+}\delta_{s's}. \tag*{(46)} \end{aligned}

Here k′+=k+=xP+k^{\prime+}=k^+=xP^+, ky′=ky=qyk'_y=k_y=q_y, and s/2s/2 is the fermion light-front helicity. The two orbital channels in Eq. (S1) have equal helicity-flip wave-function products, so their sum cancels the ss-dependent term in the kernel. The pair overlap, including both the fermion and antifermion, is

Mpairyy(Q;ΛUV)=2NcP+∫01dx∫ΛUVd2q⊥∑s,sˉΨf;ssˉ∗(x,q⊥′)Of;ssyyΨf;ssˉ(x,q⊥)=8afπ∫01dy y K(yQ;ΛUV).(47)\begin{aligned} \mathcal{M}^{yy}_{\mathrm{pair}}(Q;\Lambda_{\mathrm{UV}})=2N_cP^+\int_0^1 dx\int_{\Lambda_{\mathrm{UV}}}d^2q_\perp \sum_{s,\bar{s}}\Psi^{*}_{f;s\bar{s}}(x,q'_\perp)\mathcal{O}^{yy}_{f;ss}\Psi_{f;s\bar{s}}(x,q_\perp) \\ &=\frac{8a_f}{\pi}\int_0^1dy\,y\,K(yQ;\Lambda_{\mathrm{UV}}). \tag*{(47)} \end{aligned}

Here q⊥′=q⊥+(1−x)Qx^q'_\perp=q_\perp+(1-x)Q\hat{x}, y=1−xy=1-x, and ∫ΛUVd2q⊥\int_{\Lambda_{\mathrm{UV}}}d^2q_\perp denotes the disk qx2+qy2<ΛUV2q_x^2+q_y^2<\Lambda_{\mathrm{UV}}^2 in the original relative momentum. The antifermion term equals the fermion term after x↔1−xx\leftrightarrow1-x and qx→−qxq_x\to-q_x, giving the factor of two. Terms odd in qyq_y integrate to zero on this disk. The transverse combination is

K(a;ΛUV)=I(a;ΛUV)+aJ(a;ΛUV)=∫ΛUVd2q⊥qy2(qx2−qy2+aqx)D0Da,(48)\begin{aligned} K(a;\Lambda_{\mathrm{UV}})=I(a;\Lambda_{\mathrm{UV}})+aJ(a;\Lambda_{\mathrm{UV}}) \\ &=\int_{\Lambda_{\mathrm{UV}}}d^2q_\perp\frac{q_y^2(q_x^2-q_y^2+aq_x)}{D_0D_a}, \tag*{(48)} \end{aligned}

where a=yQa=yQ, D0=q⊥2+m2D_0=q_\perp^2+m^2, and Da=(qx+a)2+qy2+m2D_a=(q_x+a)^2+q_y^2+m^2. The two required integrals are

I(a;ΛUV)=∫ΛUVd2q⊥qy2(qx2−qy2)D0Da,I(a;\Lambda_{\mathrm{UV}})=\int_{\Lambda_{\mathrm{UV}}}d^2q_\perp\frac{q_y^2(q_x^2-q_y^2)}{D_0D_a},
J(a;ΛUV)=∫ΛUVd2q⊥qy2qxD0Da.(49)J(a;\Lambda_{\mathrm{UV}})=\int_{\Lambda_{\mathrm{UV}}}d^2q_\perp\frac{q_y^2q_x}{D_0D_a}. \tag*{(49)}

We subtract the forward pair overlap at the same finite cutoff:

F2(t)=−8afπt2lim⁡ΛUV→∞∫01dy y[K(yQ;ΛUV)−K(0;ΛUV)].(50)F_2(t)=-\frac{8a_f}{\pi t^2}\lim_{\Lambda_{\mathrm{UV}}\to\infty}\int_0^1dy\,y\left[K(yQ;\Lambda_{\mathrm{UV}})-K(0;\Lambda_{\mathrm{UV}})\right]. \tag*{(50)}

For m>0m>0, F2(t)F_2(t) approaches a finite value as t→0t\to0. Equation (45) therefore requires the transverse helicity-flip matrix element to be proportional to t2=Q4t^2=Q^4 near zero transfer. The pair overlap contains a transfer-independent forward contribution, which is canceled in the full EMT. We implement this cancellation by subtracting K(0;ΛUV)K(0;\Lambda_{\mathrm{UV}}) at the same finite cutoff, before dividing by t2t^2 and taking ΛUV→∞\Lambda_{\mathrm{UV}}\to\infty. Here K(0;ΛUV)K(0;\Lambda_{\mathrm{UV}}) sets the transfer to zero while retaining the integration over q⊥q_\perp.

Transverse integrals

Set qx=ucos⁡ϕq_x=\sqrt{u}\cos\phi, qy=usin⁡ϕq_y=\sqrt{u}\sin\phi, so that d2q⊥=du dϕ/2d^2q_\perp=du\,d\phi/2, 0≤u≤ΛUV20\leq u\leq\Lambda_{\mathrm{UV}}^2, and 0≤ϕ≤2π0\leq\phi\leq2\pi. The angular denominator is Da=A+bcos⁡ϕD_a=A+b\cos\phi, where

A=u+a2+m2,b=2au,A=u+a^2+m^2,\qquad b=2a\sqrt{u},
S=A2−b2=(u+m2−a2)2+4a2m2.(51)S=\sqrt{A^2-b^2}=\sqrt{(u+m^2-a^2)^2+4a^2m^2}. \tag*{(51)}

The angular identities used are

∫02πdϕA+bcos⁡ϕ=2πS,\int_{0}^{2\pi}\frac{d\phi}{A+b\cos\phi}=\frac{2\pi}{S},
∫02πsin⁡2ϕ dϕA+bcos⁡ϕ=2πA+S,\int_{0}^{2\pi}\frac{\sin^{2}\phi\,d\phi}{A+b\cos\phi}=\frac{2\pi}{A+S},
∫02πsin⁡2ϕcos⁡ϕ dϕA+bcos⁡ϕ=−πb(A+S)2,\int_{0}^{2\pi}\frac{\sin^{2}\phi\cos\phi\,d\phi}{A+b\cos\phi}=-\frac{\pi b}{(A+S)^{2}},
∫02πsin⁡2ϕcos⁡2ϕ dϕA+bcos⁡ϕ=πA(A+S)2.(52)\int_{0}^{2\pi}\frac{\sin^{2}\phi\cos^{2}\phi\,d\phi}{A+b\cos\phi}=\frac{\pi A}{(A+S)^{2}}. \tag*{(52)}

The last three follow by dividing powers of cos⁡ϕ\cos\phi by A+bcos⁡ϕA+b\cos\phi and using the first identity. Since sin⁡2ϕ(cos⁡2ϕ−sin⁡2ϕ)=2sin⁡2ϕcos⁡2ϕ−sin⁡2ϕ\sin^{2}\phi(\cos^{2}\phi-\sin^{2}\phi)=2\sin^{2}\phi\cos^{2}\phi-\sin^{2}\phi, Eq. (49) becomes

I(a;ΛUV)=−π∫0ΛUV2du u2S(u+m2)(A+S)2,I(a;\Lambda_{\mathrm{UV}})=-\pi\int_{0}^{\Lambda_{\mathrm{UV}}^{2}}du\,\frac{u^{2}S}{(u+m^{2})(A+S)^{2}},
J(a;ΛUV)=−πa∫0ΛUV2du u2(u+m2)(A+S)2.(53)J(a;\Lambda_{\mathrm{UV}})=-\pi a\int_{0}^{\Lambda_{\mathrm{UV}}^{2}}du\,\frac{u^{2}}{(u+m^{2})(A+S)^{2}}. \tag*{(53)}

For the radial integrations, use

w=u+m2−a2+S,w=u+m^{2}-a^{2}+S,
u=w2+a2−m2−2a2m2w,u=\frac{w}{2}+a^{2}-m^{2}-\frac{2a^{2}m^{2}}{w},
du=(12+2a2m2w2)dw,du=\left(\frac{1}{2}+\frac{2a^{2}m^{2}}{w^{2}}\right)dw,
S=w2+2a2m2w,A+S=w+2a2.S=\frac{w}{2}+\frac{2a^{2}m^{2}}{w},\qquad A+S=w+2a^{2}.

The integrands are then rational functions of ww, with limits w(0)=2m2w(0)=2m^{2} and w(ΛUV2)w(\Lambda_{\mathrm{UV}}^{2}). Integrating gives

I(a;ΛUV)≃π[−ΛUV24+a2+2m24L+3m24+5a224−(a2+2m2)Ra2aζ],(54)I(a;\Lambda_{\mathrm{UV}})\simeq\pi\left[-\frac{\Lambda_{\mathrm{UV}}^{2}}{4}+\frac{a^{2}+2m^{2}}{4}\mathcal{L}+\frac{3m^{2}}{4}+\frac{5a^{2}}{24}-\frac{(a^{2}+2m^{2})R_{a}}{2a}\zeta\right], \tag*{(54)}
J(a;ΛUV)≃π[−a4L−a8+Ra2ζ].(55)J(a;\Lambda_{\mathrm{UV}})\simeq\pi\left[-\frac{a}{4}\mathcal{L}-\frac{a}{8}+\frac{R_{a}}{2}\zeta\right]. \tag*{(55)}

Here Ra=a2+4m2R_{a}=\sqrt{a^{2}+4m^{2}}, ζ=arsinh⁡(a/2m)\zeta=\operatorname{arsinh}(a/2m), and L=log⁡(1+ΛUV2/m2)\mathcal{L}=\log(1+\Lambda_{\mathrm{UV}}^{2}/m^{2}). The symbol ≃\simeq retains every divergent and finite term and omits only terms that vanish as ΛUV→∞\Lambda_{\mathrm{UV}}\to\infty at fixed a,ma,m. Combining the integrals gives

K(a;ΛUV)≃π[−ΛUV24+m22L+3m24+a212−m2Raaζ].(56)K(a;\Lambda_{\mathrm{UV}})\simeq\pi\left[-\frac{\Lambda_{\mathrm{UV}}^{2}}{4}+\frac{m^{2}}{2}\mathcal{L}+\frac{3m^{2}}{4}+\frac{a^{2}}{12}-\frac{m^{2}R_{a}}{a}\zeta\right]. \tag*{(56)}

At zero transfer,

K(0;ΛUV)=−π4∫0ΛUV2du u2(u+m2)2=−π4[ΛUV2+m2R−2m2L].(57)K(0;\Lambda_{\mathrm{UV}})=-\frac{\pi}{4}\int_{0}^{\Lambda_{\mathrm{UV}}^{2}}du\,\frac{u^{2}}{(u+m^{2})^{2}} =-\frac{\pi}{4}\left[\Lambda_{\mathrm{UV}}^{2}+m^{2}R-2m^{2}\mathcal{L}\right]. \tag*{(57)}

Here R=ΛUV2/(ΛUV2+m2)R=\Lambda_{\mathrm{UV}}^{2}/(\Lambda_{\mathrm{UV}}^{2}+m^{2}). The common terms −πΛUV2/4-\pi\Lambda_{\mathrm{UV}}^{2}/4 and πm2L/2\pi m^{2}\mathcal{L}/2 cancel between Eqs. (56) and (S31), giving

Ksub(a)≡lim⁡ΛUV→∞[K(a;ΛUV)−K(0;ΛUV)]=π[m2+a212−m2Raaζ].(58)K_{\mathrm{sub}}(a)\equiv\lim_{\Lambda_{\mathrm{UV}}\to\infty}\left[K(a;\Lambda_{\mathrm{UV}})-K(0;\Lambda_{\mathrm{UV}})\right] =\pi\left[m^{2}+\frac{a^{2}}{12}-\frac{m^{2}R_{a}}{a}\zeta\right]. \tag*{(58)}

An independent dimensionally regulated evaluation of the conserved transverse projection gives the same finite kernel:

Ksub(a)=π2∫01dv[v(1−v)a2−m2log⁡(1+v(1−v)a2m2)].(59)K_{\mathrm{sub}}(a)=\frac{\pi}{2}\int_{0}^{1}dv\left[v(1-v)a^{2}-m^{2}\log\left(1+\frac{v(1-v)a^{2}}{m^{2}}\right)\right]. \tag*{(59)}

Expanding Eq. (58) at small aa gives Ksub(a)=πa4/(120m2)+O(a6/m4)K_{\mathrm{sub}}(a)=\pi a^{4}/(120m^{2})+O(a^{6}/m^{4}). Since a=yQa=yQ, the subtracted overlap is proportional to Q4=t2Q^{4}=t^{2} near zero transfer, so dividing by t2t^{2} gives a finite F2(0)F_{2}(0). The parallel projection ⟨Txx⟩+−\langle T^{xx}\rangle_{+-} vanishes in the same conserved-EMT prescription.

Substituting Eq. (58) into Eq. (50) gives

2tF2(t)af=4τ∫01dy[y+τy33−1+τy2τarsinh⁡(yτ)].(60)\frac{2tF_{2}(t)}{a_{f}}=\frac{4}{\tau}\int_{0}^{1}dy\left[y+\frac{\tau y^{3}}{3}-\frac{\sqrt{1+\tau y^{2}}}{\sqrt{\tau}}\operatorname{arsinh}(y\sqrt{\tau})\right]. \tag*{(60)}

Here τ=Q2/(4m2)\tau=Q^{2}/(4m^{2}), η=arsinh⁡τ\eta=\operatorname{arsinh}\sqrt{\tau}, and r=1+1/τr=\sqrt{1+1/\tau}. Evaluating the integral gives

2tF2(t)af=13+3τ−2rητ−η2τ2.(61)\frac{2tF_{2}(t)}{a_{f}}=\frac{1}{3}+\frac{3}{\tau}-\frac{2r\eta}{\tau}-\frac{\eta^{2}}{\tau^{2}}. \tag*{(61)}

with continuous limit F2(0)=−af/(90m2)F_{2}(0)=-a_{f}/(90m^{2}).

References

  1. [1]E. Witten, Nucl. Phys. B 120, 189 (1977).
  2. [2]R. Nisius, Phys. Rep. 332, 165 (2000), arXiv:hep-ex/9912049.DOI
  3. [3]M. Krawczyk, A. Zembrzuski, and M. Staszel, Phys. Rep. 345, 265 (2001), arXiv:hep-ph/0011083.DOI
  4. [4]M. Stratmann and W. Vogelsang, Phys. Lett. B 386, 370 (1996), arXiv:hep-ph/9606346.DOI
  5. [5]S. D. Bass, S. J. Brodsky, and I. Schmidt, Phys. Lett. B 437, 417 (1998), arXiv:hep-ph/9805316.DOI
  6. [6]S. Friot, B. Pire, and L. Szymanowski, Phys. Lett. B 645, 153 (2007), arXiv:hep-ph/0611176.DOI
  7. [7]A. Mukherjee and S. Nair, Phys. Lett. B 706, 77 (2011), arXiv:1105.5299.DOI
  8. [8]A. Mukherjee and S. Nair, Phys. Lett. B 707, 99 (2012), arXiv:1110.5242.DOI
  9. [9]A. Mukherjee, S. Nair, and V. K. Ojha, Phys. Lett. B 721, 284 (2013), arXiv:1301.6895.DOI
  10. [10]I. R. Gabdrakhmanov and O. V. Teryaev, Phys. Lett. B 716, 417 (2012), arXiv:1204.6471.DOI
  11. [11]S. Nair, C. Mondal, X. Zhao, A. Mukherjee, and J. P. Vary, Phys. Lett. B 827, 137005 (2022), arXiv:2201.12770.DOI
  12. [12]S. Nair, C. Mondal, X. Zhao, A. Mukherjee, and J. P. Vary (BLFQ), Phys. Rev. D 108, 116005 (2023), arXiv:2302.13645.arxiv.org/abs/2302.13645
  13. [13]S. Puhan, N. Kumar, and H. Dahiya, Eur. Phys. J. A 61, 56 (2025), arXiv:2408.07714.arxiv.org/abs/2408.07714
  14. [14]S. Puhan, N. Kumar, and H. Dahiya, Understanding photon TMDs using light-front wave function (2025), arXiv:2504.01419.arxiv.org/abs/2504.01419
  15. [15]A. Freese and W. Cosyn, Phys. Rev. D 106, 114014 (2022), arXiv:2207.10788.DOI
  16. [16]J. More, A. Mukherjee, S. Nair, and S. Saha, Phys. Rev. D 105, 056017 (2022), arXiv:2112.06550.DOI
  17. [17]J. More, A. Mukherjee, S. Nair, and S. Saha, Few-Body Syst. 64, 68 (2023), arXiv:2306.08683.DOI
  18. [18]S. Meissner, A. Metz, and M. Schlegel, JHEP 08 (2009), 056, arXiv:0906.5323.DOI
  19. [19]C. Lorcé and B. Pasquini, JHEP 09 (2013), 138, arXiv:1307.4497.
  20. [20]X. Ji, Phys. Rev. Lett. 91, 062001 (2003), arXiv:hep-ph/0304037.DOI
  21. [21]A. V. Belitsky, X. Ji, and F. Yuan, Phys. Rev. D 69, 074014 (2004), arXiv:hep-ph/0307383.DOI
  22. [22]C. Lorcé, B. Pasquini, and M. Vanderhaeghen, JHEP 05 (2011), 041, arXiv:1102.4704.DOI
  23. [23]R. L. Jaffe and A. Manohar, Nucl. Phys. B 337, 509 (1990).
  24. [24]X. Ji, Phys. Rev. Lett. 78, 610 (1997), arXiv:hep-ph/9603249.DOI
  25. [25]E. Leader and C. Lorcé, Phys. Rep. 541, 163 (2014), arXiv:1309.4235.DOI
  26. [26]C. Lorcé and B. Pasquini, Phys. Rev. D 84, 014015 (2011), arXiv:1106.0139.DOI
  27. [27]C. Lorcé, B. Pasquini, X. Xiong, and F. Yuan, Phys. Rev. D 85, 114006 (2012), arXiv:1111.4827.
  28. [28]Y. Hatta, Phys. Lett. B 708, 186 (2012), arXiv:1111.3547.DOI
  29. [29]X. Ji, X. Xiong, and F. Yuan, Phys. Rev. Lett. 109, 152005 (2012), arXiv:1202.2843.DOI
  30. [30]C. Lorcé, Phys. Lett. B 735, 344 (2014), arXiv:1401.7784.
  31. [31]K. Kanazawa, C. Lorcé, A. Metz, B. Pasquini, and M. Schlegel, Phys. Rev. D 90, 014028 (2014), arXiv:1403.5226.DOI
  32. [32]Y. Hatta, B.-W. Xiao, and F. Yuan, Phys. Rev. Lett. 116, 202301 (2016), arXiv:1601.01585.
  33. [33]S. Bhattacharya, A. Metz, and J. Zhou, Phys. Lett. B 771, 396 (2017), arXiv:1702.04387.DOI
  34. [34]S. Bhattacharya, R. Boussarie, and Y. Hatta, Phys. Rev. Lett. 128, 182002 (2022), arXiv:2201.08709.DOI
  35. [35]M. Engelhardt, Phys. Rev. D 95, 094505 (2017), arXiv:1701.01536.DOI
  36. [36]S. Bhattacharya, K. Cichy, M. Constantinou, X. Gao, A. Metz, J. Miller, S. Mukherjee, P. Petreczky, F. Steffens, and Y. Zhao, JHEP 01 (2025), 146, arXiv:2410.03539.DOI
  37. [37]S. Bhattacharya, R. Boussarie, and Y. Hatta, Phys. Rev. D 111, 034019 (2025), arXiv:2404.04209.arxiv.org/abs/2404.04209
  38. [38]Y. Hatta and J. Schoenleber, JHEP 09 (2024), 154, arXiv:2404.18872.arxiv.org/abs/2404.18872
  39. [39]C. Lorcé and Q.-T. Song, Phys. Lett. B 864, 139433 (2025), arXiv:2501.05092.DOI
  40. [40]P. Choudhary, A. Mukherjee, and R. Singh, Spin-orbit correlations in the pion and the role of quark-gluon interaction (2026), arXiv:2601.02223.arxiv.org/abs/2601.02223
  41. [41]A. Acharyya, S. Puhan, and H. Dahiya, Phys. Rev. D 110, 034020 (2024), arXiv:2405.00446.DOI
  42. [42]T. Hu, X. Cao, S. Xu, W. Du, Q.-T. Song, and Y. Li, Spin-orbit correlation of quarks within quarkonium (2026), arXiv:2606.06263.arxiv.org/abs/2606.06263
  43. [43]A. Mukherjee, S. Nair, and V. K. Ojha, Phys. Rev. D 90, 014024 (2014), arXiv:1403.6233.DOI
  44. [44]J. More, A. Mukherjee, and S. Nair, Phys. Rev. D 95, 074039 (2017), arXiv:1701.00339.DOI
  45. [45]J. More, A. Mukherjee, and S. Nair, Eur. Phys. J. C 78, 389 (2018), arXiv:1709.00943.DOI
  46. [46]A. Bacchetta and P. J. Mulders, Phys. Rev. D 62, 114004 (2000), arXiv:hep-ph/0007120.DOI
  47. [47]S. Kumano and Q.-T. Song, Phys. Rev. D 103, 014025 (2021), arXiv:2011.08583.DOI
  48. [48]W. Cosyn, S. Cotogno, A. Freese, and C. Lorcé, Eur. Phys. J. C 79, 476 (2019), arXiv:1903.00408.DOI
  49. [49]H. K. Kim and J.-Y. Kim, Phys. Lett. B 883, 141109 (2026), arXiv:2608.09012.arxiv.org/abs/2608.09012
  50. [50]A. Harindranath, R. Kundu, and W.-M. Zhang, Phys. Rev. D 59, 094013 (1999), arXiv:hep-ph/9806221.DOI
  51. [51]T. Lappi, H. Mäntysaari, and J. Penttala, Phys. Rev. D 102, 054020 (2020), arXiv:2006.02830 [hep-ph].DOI
  52. [52]G. Beuf, Phys. Rev. D 94, 050416 (2016), arXiv:1606.00777 [hep-ph].DOI
  53. [53]S. J. Brodsky, D. S. Hwang, B.-Q. Ma, and I. Schmidt, Nucl. Phys. B 593, 311 (2001), arXiv:hep-ph/0003082.
  54. [54]W.-M. Zhang and A. Harindranath, Phys. Rev. D 48, 4881 (1993).
  55. [55]A. Harindranath and R. Kundu, Phys. Rev. D 59, 116013 (1999), arXiv:hep-ph/9802406.DOI
  56. [56]M. V. Polyakov and P. Schweitzer, Int. J. Mod. Phys. A 33, 1830025 (2018), arXiv:1805.06596.DOI
  57. [57]K. A. Milton, Phys. Rev. D 15, 2149 (1977).
  58. [58]F. A. Berends and R. Gastmans, Ann. Phys. (N.Y.) 98, 225 (1976).

Paper details

vlnc:W_sxrv2nfw
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

Anthropic’s Claude Opus 5.5

View statement
Looking for related papers…
Contents