WorksPhysics and Astronomy
Deeply virtual pion production through two-loop order
Abstract
Deeply virtual meson production (DVMP) is among the most prominent channels to extract the nucleon's generalized parton distributions (GPDs) at scattering facilities such as JLab and the upcoming EIC/EicC experiments, which plays a vital role in unravelling the three-dimensional internal structure of nucleon. In this work we calculate for the first time the next-to-next-to-leading order (NNLO) QCD radiative corrections to the DVP processes and in the generalized Bjorken limit , accurate at the leading twist within collinear factorization framework. The impact of the two-loop QCD corrections appears to be positive and substantial, including which considerably improves the agreement between the perturbative QCD prediction and the available JLab data. In addition to the differential longitudinal DVP cross section, we also study the impact of the two-loop QCD corrections on the transverse single-spin asymmetries (TSSA) in some benchmark kinematics at JLab, EIC and EicC.
Introduction
Study of generalized parton distributions (GPDs) of nucleons [3] [1, 2] stands as one of the central goals of contemporary hadron physics. By correlating partonic longitudinal momentum with transverse spatial position [4], GPDs provide a unique tool for accessing three-dimensional partonic structure of nucleons. In addition, GPDs are intimately linked with the proton spin decomposition [5, 6, 7, 8], gravitational form factors [9, 10, 11, 12, 13, 14, 15, 16, 17, 18], and internal mechanical distributions [19, 20, 21, 22, 23, 24, 25, 26] (for comprehensive review on various aspects of GPDs, see Refs. [27, 28, 29, 30]).
In sharp contrast to the familiar one-dimensional PDFs of proton, to date our theoretical knowledge toward the nucleon GPDs, which are functions of three variables, are still rather limited. Some influential parameterized models of nucleon GPDs are available in the market [31, 32, 33, 34, 35, 36, 37, 38, 39, 40]. Recently there also appear the numerical calculation of GPDs of pion and nucleon at some specific value of skewness from lattice QCD simulation [41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51]. On the experimental side, GPDs can be accessed through the gold-plated hard exclusive processes such as the deeply virtual Compton scattering (DVCS) [3] [1, 2] and deeply virtual meson production (DVMP) [31, 52, 53, 54, 55, 56]. Over the past decade, various experimental facilities such as HERA, COMPASS and JLab have measured the DVCS [57, 58, 59, 60, 61, 62, 63] and DVMP [64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74] in kinematics of . This progress sets the stage for a new exciting era of high-precision tomography at the JLab 22 GeV facility [75] and Electron-Ion Colliders in the US (EIC) [76] and China (EicC) [77] in future.
Reliable extractions of nucleon GPDs from experimental measurements require a combined analysis of multiple exclusive channels. Whereas DVCS probes only flavor-weighted combinations of GPDs, DVMP provides complementary sensitivity that facilitates the separation of individual quark flavor contributions. Consequently, the measurements of DVMP are regarded as an indispensable complements to those of DVCS.
To match the experimental advancements, it is compulsory to provide the most precise predictions for the DVCS and DVMP processes, demanding the inclusion of higher-order corrections. Very recently the coefficient functions for DVCS [78, 79, 80] and double deeply virtual Compton scattering [81] have been computed at next-to-next-to-leading order (NNLO) within the collinear factorization framework. In contrast, phenomenological analyses of DVMP [82, 83, 84, 85] is still based on the next-to-leading order (NLO) QCD predictions, which were first available about two decades ago [86, 87]. Since the NLO corrections turn out to be sizable for DVMP processes, one naturally wonders whether the NNLO corrections in DVMP may also play an important role in the precise extraction of GPDs.
In this paper, we fill the long-missing gap by presenting the first calculation of the NNLO QCD corrections to a class of important DVMP processes, and production. We find the two-loop corrections are positive and substantial, including which yields a considerably improved description of the available JLab data. This implies that NNLO QCD correction is an indispensable ingredient for a reliable description of the DVP cross sections in the upcoming EIC/EicC programs and the precise extraction of nucleon GPDs.
Observables related with DVP.
Let us specialize to the DVP process
where the momenta and helicities are indicated in parentheses. Concretely speaking, we concentrate on the leading-twist processes, and , where the incoming virtual photon is longitudinally polarized.1
The differential longitudinal cross section of DVP can be expressed in terms of two helicity amplitudes [32, 68]:
with signifying helicity amplitudes. The phase space factor is defined as , with Källen function .
In the collinear factorization regime , the helicity amplitudes in (2) can be expressed in terms of the transition FFs (TFFs) [33]:
These TFFs and (collectively ) factorize into a convolution of a perturbative hard-scattering kernel
with the twist-2 pion DA and the axial-vector GPDs or . This factorized structure for DVP is shown schematically in Figure 1.

Figure 1. A cartoon illustrating the factorized structure of the amplitude for the deeply virtual pion production.
Complementary to the longitudinal cross section in (2), the transverse single-spin asymmetry (TSSA) is also a leading-twist observable relevant to the transversely polarized proton targets, which originates from the interference of the helicity amplitudes in (3). At leading twist, this observable is defined as [54, 55, 82]
The TSSA provides direct access to the relative phase between the TFFs and , offering unique sensitivity to the poorly constrained GPD .
Factorization formula for TFF.
We start with the production. In the collinear factorization regime, the TFF admits the following factorized form at leading twist:
where (with ) signifies the isovector combination of the quark GPDs of nucleon [53]. For brevity, the renormalization scale and the factorization scales in GPD and DA have been suppressed in (6) and thereafter. It is worth remarking that the factorization theorem (6) has been proven to hold for all order of [88].
The hard-scattering kernel can be expressed in terms of the basic building block:
with . Owing to asymptotic freedom of QCD, the coefficient function can be computed in perturbation theory order by order in ,
The hard-scattering kernel can be determined through the perturbative matching procedure, e.g., by computing the non-singlet (NS) partonic reaction with order by order in . Some representative Feynman diagrams through two-loop order are shown in Figure 2.

Figure 2. Some representative parton-level diagrams for the deeply virtual production through two-loop order. Note that the rightmost two-loop diagram signifies the pure singlet contribution for leptoproduction.
Fortunately, it turns out that we do not need conduct the two-loop calculation from the scratch, instead we can directly infer the desired coefficient function from the counterpart in the electromagnetic form factor (EMFF) by making the following replacement:
This shortcut works because in the Efremov-Radyushkin-Brodsky-Lepage (ERBL) region (), both the NS quark amplitudes and the evolution kernels [89, 90] in two processes are identical at any prescribed perturbative order, so are the coefficient functions by iteratively solving the matching equations. The coefficient function for can then be analytically continued into the Dokshitzer-Gribov-Lipatov-Altarelli-Parisi (DGLAP) regime ( or ). To circumvent the singularities located at , we utilize the causal prescription by making the substitution [82, 83, 84].
The LO coefficient function in spacetime dimension reads [31, 53, 54, 55, 56]
The NLO coefficient function has been deduced from the counterpart in the NLO correction to EMFF long ago [82, 86]. Consequently, following the recipe specified in (8), we obtain the intended NNLO coefficient function, , from the recently available NNLO perturbative correction to EMFF [91, 92].
Factorization formula for TFF
In the generalized Bjorken limit, the TFF for can be expressed in terms of the sum of the non-singlet and pure singlet (PS) contributions 2:
The NS channel contribution is intimately connected with :
with
which specifically probes the C-odd combinations of nucleon GPDs [83]. Since the sea quark distributions cancel within the C-odd axial-vector GPDs, consequently we only need consider the valence quark contributions in the NS channel.
The coefficient function in (12) is identical to that introduced in (7) for , which has already been deduced from the higher order corrections to EMFF by employing (8).
An interesting new type of contribution for stems from the singlet quark channel. The corresponding singlet GPD in (10) is defined by
Due to color and -parity conservation, there receive no tree-level and one-loop contributions to the hard-scattering kernel in the PS channel. As represented in the rightmost diagram in Figure 2, there arises a nonvanishing contribution starting at two-loop order. One of the major technical challenge of this work is to compute analytically.
Description of the calculation. As aforementioned, the main challenge is to ascertain the two-loop PS hard-scattering kernel. We use the packages QGRAF [94] and HepLib [95] to generate Feynman diagrams and the corresponding amplitudes for the quark process . There are in total 430 two-loop diagrams, and one of which is depicted in the rightmost Feynman diagram of Figure 2. After employing Apart [96] and FIRE [97] for partial fraction and integration-by-part reduction, we end up with 271 master integrals (MIs). They are calculated by using the method of differential equations [98, 99, 100], with the boundary conditions fixed by asymptotically expanding the MIs around with the aid of the method of region [101, 102, 103]. The boundary integrals are calculated by using the methods developed in Refs. [104, 105]. The analytic expressions of all MIs are obtained by using the automated package AmpRed [106, 107], which have also been numerically verified by AMFlow [108].
It is worth stressing that, due to the presence of in each quark line, the naive use of covariant projector approach [109] in dimensional regularization would lead to incorrect result. We therefore give up taking the Dirac trace and keep all the spinor indices of external quarks open. By enforcing Ward-Takahashi identity to hold, we are able to obtain the UV and IR-finite result for .
We devoted Appendix § 3 for a comprehensive description of our strategy to deduce .
Numerical Studies. The leading-twist pion DA is conveniently expanded in the Gegenbauer polynomial basis:
where nonperturbative dynamics is encoded in the Gegenbauer moments . The recent NNLO study of pion EMFF [91] suggests that the lattice QCD prediction by RQCD Collaboration, [110], gives a better account of the data. Therefore we will adopt this value in our phenomenological exploration.
With the aid of the package PolyLogTools [111], the convolution between the hard-scattering kernel and pion DA over can be done analytically once the latter is expressed in the expanded form of (14). Thus the pion TFFs in (5) and (10) can be reduced into one-dimensional integrals, which can be computed with high numerical precision by employing the numerical integrator HCubature [112] with multiple-precision floating-point support.
Although there have been impressive recent progress for the first-principle determination of nucleon GPDs from lattice QCD, so far the lattice predictions for nucleon axial-vector GPDs are available only for very limited values of [44, 45]. Therefore, we turn to two popular parameterizations of nucleon axial-vector GPDs for phenomenological purpose: Goloskokov-Kroll (GK) model [32, 33, 34, 35] and the GPD through universal moment parametrization (GUMP) program [40]. In the GK model, the isospin symmetry implies that the contributions from the - and -sea quarks cancel with each other inside . Consequently, only the valence-quark GPDs contribute to DVP [33, 86]. On the other hand, the GUMP 1.0 parametrization represents the first global extraction to consistently synthesize experimental data and lattice QCD constraints at NLO accuracy in the conformal moment expansion framework.
In the numerical analysis, we fix and . We use the package FAPT [113] to evaluate the running QCD coupling constant to three-loop accuracy. Three-loop evolution effect is incorporated in pion DA [114, 115]. We also use the package tiktaalik [116] to evolve the GPD from 2 GeV to any intended scale with two-loop accuracy.
We present the leading-twist predictions to DVP at various perturbative order in Figure 3, choosing three benchmark kinematic points typical of the JLab (, ; upper panel), EicC (, ; middle panel), and EIC (, ; lower panel) facilities. The nonperturbative inputs are chosen as the GK+RQCD (left column) and GUMP+RQCD (right column). The uncertainties are estimated by varying the factorization and renormalization scales in the range .
Figure 3. The predicted for DVP at various level of perturbative accuracy. The experimental data come from [65].
As can be clearly visualized in Figure 3, the NNLO perturbative correction to DVP is positive and substantial, which is even as significant as the NLO correction at JLab kinematic range [65]. Interestingly, the relative importance of the NNLO correction appears to decrease as increases. The perturbative convergence behavior, especially within the GK model, appears to be satisfactory at EIC energy.
Within the GK model parametrization, the dominant contribution of electroproduction stems from the pion pole term in the term. As shown in the upper panel of Figure 3, including the NNLO perturbative correction leads to an improved description of the measured longitudinal cross section [65], though the theoretical uncertainties remain sizable. However, within the GUMP parametrization, the NNLO prediction seems to fail to capture the measured -dependence of the . This may be partly attributed to the fact that the truncation procedure of implemented by GUMP may not work well for [65].
Figure 4 displays the predicted for DVP at various levels of perturbative accuracy, again taking the same kinematics at JLab, EicC and EIC. Similar to leptoproduction, the NNLO correction is also positive and significant for DVP, whose relative importance with respect to the NLO correction appears to decrease with the increasing .
Figure 4. The predicted for DVP at various level of perturbative accuracy. The experimental data come from [71].
Due to the experimental incapability of separating the longitudinal DVP cross section, we compare our predictions with the measured unpolarized cross section [71], which is juxtaposed in the upper panel of Figure 4. Where is the ratio of longitudinal and transverse photon flux.
However, even after including the NNLO corrections, the GK model predictions are still significantly lower than the measured (note the data points have been shrunk by a factor of 3 for the sake of visibility). In contrast, the NLO and NNLO predictions based on the GUMP parametrization appear to be compatible with the data.
In Figure 5 we also predict the -dependence of the TSSA for DVP at various level of perturbative order. Here we select representative and values that span the kinematic reach of JLab, EicC, and EIC, mirroring the preceding longitudinal cross section analysis. We find that including NNLO corrections does not significantly alter the TSSA, echoing the conclusions drawn for NLO corrections in Refs. [82, 86]. Nevertheless, our results demonstrate that the TSSA remains sizable across this broad range, a feature that is robust across different GPD parametrizations.
Figure 5. The predicted TSSA for DVP at various level of perturbative accuracy.
Interestingly, the predicted TSSA at low and small diverges significantly between the GK and GUMP models, as shown in the lower panel of Figure 5. This discrepancy might originate from the dominant pion pole contribution from the term in the GK model, which is absent in the GUMP parametrization. The pion
pole dictates a enhancement at small [117], generating a pronounced asymmetry. Consequently, future high-precision measurements of the TSSA at low will serve a crucial probe to test the picture of the pion-pole dominance.
In Figure 6, we predict the TSSA for DVP at various perturbative order, adopting the same kinematic setup as in Figure 5. To align with the case, including higher order radiative corrections does not significantly alter the asymmetry. Due to the absence of the pion pole contribution, the TSSA predicted by the GK model (left panel) is notably small, posing a significant challenge for future experimental measurements. Conversely, the GUMP parametrization (right panel) yields a sizable asymmetry, but only at intermediate value of .
Figure 6. The predicted TSSA for DVP at various level of perturbative accuracy.
Summary. In this work, we have presented the first calculation of NNLO QCD corrections to the deeply virtual pion production () at leading twist in the collinear factorization approach. By relating the hard-scattering kernel in the non-singlet channel of the DVP to that of the EMFF, and employing modern multi-loop computational techniques, we have derived the complete two-loop coefficient functions for both charged and neutral pion production channels. For the DVP process, we have also accomplished the first calculation of the pure singlet contribution, whose phenomenological impact appears to be modest.
Employing two influential GPD parametrization models, we find that the NNLO perturbative corrections are positive and substantial. For typical kinematics setup at JLab, EicC, and the EIC, the inclusion of NNLO effects significantly enhances the longitudinal cross section compared to NLO predictions, even reaching more than a 100% correction in the intermediate region. These higher-order contributions considerably improve
the agreement between perturbative QCD predictions and the available JLab data for DVP. Furthermore, we have also investigated the transverse single-spin asymmetry for both and leptoproduction channels. While the magnitude of TSSA appears stable against the inclusion of the NNLO QCD correction, the -dependence of TSSA exhibits a strong sensitivity to the underlying GPD parametrizations (notably the pion-pole contribution in the GK model).
Our findings underscore the critical necessity of NNLO accuracy for a reliable extraction of nucleon GPDs from the vast amount of DVMP data expected at current and upcoming facilities. Furthermore, our work inaugurates a new era for exploring the effects of NNLO QCD corrections across a broader class of DVMP processes, including neutrino-induced meson production () and vector meson leptoproduction (, where ).
A: Calculation of the pure singlet contribution for DVP
As aforementioned in the main text, due to the presence of in the quark line, some technical nuisance is encountered in projecting out the pure singlet (PS) contribution to DVP in dimensional regularization. The purpose of this Appendix is to elaborate on a consistent recipe to compute the two-loop PS contribution.
Let us take the following partonic reaction as an explicit example:
where two distinct quark lines are connected via gluons, as represented by the rightmost sample diagram in Figure 2. Note that the one-gluon exchange (tree) diagram yields a vanishing contribution due to the color-singlet requirement. Had we naively employed the covariant projection technique [109], we would end up with the two-loop PS amplitude which bears wrong Lorentz structure and also contains the unwanted divergences. Unsurprisingly, the origin of this symptom can be traced to the fact that taking a Dirac trace containing a single is a rather subtle issue in dimensional regularization.
In passing, we note that similar symptom has already been encountered in the previous calculation of NLO QCD correction to the DVP process, where the ‘t Hooft-Veltman-Breitenlohner-Maison (HVBM) scheme was employed there [84]. In this work, we adopt a simpler method to overcome the problem.
With that said, at the very beginning we refrain from taking the Dirac trace and retain all the spinor indices open, and expand the amplitude of (A1) in terms of a set of linearly-independent Dirac matrix bases. The corresponding
Acknowledgments
We acknowledge Yuxun Guo, Siwei Hu, Rong Wang and Yaping Xie for helpful discussions. We are also indebted to Bernard Pire and Jian Zhou for valuable comments on the manuscript. The work of W. C. is supported in part by the NNSFC under Grant No. 11975200. The work of F. F. is supported in part by the NNSFC under Grant No. 12275353. The work of Y. J., G. T. and Z. Y. W. is supported in part by the NNSFC under Grant No. 12475090. The work of Q. T. S. is supported in part by the NNSFC under Grant No. 12005191 and by the Natural Science Foundation of Henan Province under Grant No. 252300423011. amplitude can then be written as and
where
with signifying the Dirac spinor indices. Concretely speaking, in (A3) signifies the shorthand for the totally anti-symmetrized product of Dirac -matrices:
Note that the Lorentz index in (A2) may be affiliated with either the coefficient or the anti-symmetrized product of Dirac -matrices.
It is natural to demand that the amplitude obeys the Ward-Takahashi identity (current conservation), which holds at any prescribed order in . Expanding the amplitude by , we thus expect . Explicit manipulation over the one- and two-loop amplitudes yield
To condense the notation, we often replace a Lorentz index of the totally anti-symmetrized product of Dirac matrices with the external momentum. It should be understood as a contraction of the Lorentz index, say, .
Now let us state a key observation. As a sufficient condition for the right-hand sides of (A5) to vanish, the following identity should hold true in -dimensional spacetime:
Here we attempt to present a proof to (A6). It is convenient to adopt the light-cone coordinate . Lorentz covariance requires that the Lorentz indices can enter only through the form (which nevertheless is forbidden due to the anti-symmetric feature under in ), or or (Similar to ’t Hooft-Veltman prescription [118], here we also assume the totally-antisymmetric -tensor is a four-dimensional object). Therefore, one is forced to conclude that there should be at least one and only one index to align with the direction of (Recall that and are two conjugate light-like vectors and ). Consequently, one can rewrite
with for and zero otherwise.
In a similar fashion, one can show that there should be at least one and only one index in to align with the direction of .
We thus arrive at
The current conservation indicates that . To facilitate the extraction of the intended PS hard kernel, it is convenient to further contract the with :
After substituting the expressions of all the MIs, we find that all the divergences are contained in , at both one-loop and two-loop order. Fortunately, the accompanying spinor cluster generates a vanishing result after projecting the final state quark-antiquark pair into the quantum number.
Since and are both UV and IR finite, from now on we can safely go back to dimension. For the process, the virtual photon can be attached either to the initial state quark line or final state quark line. The coefficient can be decomposed as . However, in order to infer the complete PS hard kernel, we must also take into account the PS amplitude for the process, in line with the flavor superposition of neutral pion . Therefore, the short-distance coefficient reads . We should also take into account the contributions of the partonic channels and , whose effect is reflected in the isospin-singlet combination of nucleon GPDs in (15).
At one-loop level, it turns out that due to the conservation of parity, since the -even cannot couple to a virtual photon plus two gluons.
At two-loop order, the last term in (A9b) contains , which trivially vanishes in four spacetime dimension, thus can be dropped. It turns out that only the term survives. Applying the covariant projection technique, we then obtain the desired two-loop PS hard-scattering kernel:
It is worth stressing that, the phenomenological impact of the PS contribution for DVP appears to be rather modest with respect to the two-loop NS contribution. As one readily observes from Figure 7, including the PS contribution only decreases the NNLO prediction for by about 1%, in both JLab and EicC experiments. In comparison, the effect of the PS contribution for TSSA is somewhat more important. As shown in Figure 8, one can see at large regime, including the PS contribution could change the NNLO predictions for by about 10%, in both JLab and EicC experiments.
Figure 7. The predicted at various level of perturbative accuracy. The curve labelled with NNLO implies that only the two-loop NS contribution is considered, while the curve labelled with NNLO implies that both the NS and PS contributions are included.
Figure 8. The predicted TSSA at various level of perturbative accuracy. The meaning of the labels NNLO’ and NNLO is the same as in Figure 7.
References
- [3]A. V. Radyushkin, Phys. Lett. B 380 (1996), 417-425.
- [4]M. Burkardt, Phys. Rev. D 62 (2000), 071503 [erratum: Phys. Rev. D 66 (2002), 119003].
- [5]X. D. Ji, Phys. Rev. Lett. 78 (1997), 610-613.
- [6]E. Leader and C. Lorcé, Phys. Rept. 541 (2014) no.3, 163-248
- [7]C. A. Aidala, S. D. Bass, D. Hasch and G. K. Mallot, Rev. Mod. Phys. 85 (2013), 655-691
- [8]X. Ji, F. Yuan and Y. Zhao, Nature Rev. Phys. 3 (2021) no.1, 27-38
- [9]V. D. Burkert, L. Elouadhiri, F. X. Girod, C. Lorcé, P. Schweitzer and P. E. Shanahan, Rev. Mod. Phys. 95 (2023) no.4, 041002.
- [10]S. Kumano, Q. T. Song and O. V. Teryaev, Phys. Rev. D 97 (2018) no.1, 014020.
- [11]A. Freese and I. C. Cloët, Phys. Rev. C 100 (2019) no.1, 015201 [erratum: Phys. Rev. C 105 (2022) no.5, 059901]
- [12]B. D. Sun and Y. B. Dong, Phys. Rev. D 101 (2020) no.9, 096008.
- [13]B. Duran, Z. E. Meziani, S. Joosten, M. K. Jones, S. Prasad, C. Peng, W. Armstrong, H. Atac, E. Chudakov and H. Bhatt, et al. Nature 615 (2023) no.7954, 813-816.
- [14]D. C. Hackett, D. A. Pefkou and P. E. Shanahan, Phys. Rev. Lett. 132 (2024) no.25, 251904.
- [15]X. H. Cao, F. K. Guo, Q. Z. Li and D. L. Yao, Nature Commun. 16 (2025), 6979.
- [16]Y. Guo, F. Yuan and W. Zhao, Phys. Rev. Lett. 135 (2025) no.11, 119002.
- [17]M. Tanaka, D. Fujii and M. Kawaguchi, Phys. Rev. D 112 (2025) no.5, 054048.
- [18]Y. Hatta and J. Schoenleber, Phys. Rev. Lett. 134 (2025) no.25, 251901.
- [19]M. V. Polyakov, Phys. Lett. B 555 (2003), 57-62.
- [20]M. V. Polyakov and P. Schweitzer, Int. J. Mod. Phys. A 33 (2018) no.26, 1830025.
- [21]V. D. Burkert, L. Elouadrihi and F. X. Girod, Nature 557 (2018) no.7705, 396-399.
- [22]C. Lorcé, H. Moutarde and A. P. Trawiński, Eur. Phys. J. C 79 (2019) no.1, 89.
- [23]K. Kumerički, Nature 570 (2019) no.7759, E1-E2.
- [24]P. E. Shanahan and W. Detmold, Phys. Rev. Lett. 122 (2019) no.7, 072003.
- [25]C. Lorcé and Q. T. Song, Phys. Lett. B 864 (2025), 139433.
- [26]A. Freese and G. A. Miller, Phys. Rev. D 103 (2021), 094023.
- [27]M. Diehl, Phys. Rept. 388 (2003), 41-277.
- [28]A. V. Belitsky and A. V. Radyushkin, Phys. Rept. 418 (2005), 1-387.
- [29]S. Boffi and B. Pasquini, Riv. Nuovo Cim. 30 (2007) no.9, 387-448.
- [30]K. Goeke, M. V. Polyakov and M. Vanderhaeghen, Prog. Part. Nucl. Phys. 47 (2001), 401-515.
- [31]M. Vanderhaeghen, P. A. M. Guichon and M. Guidal, Phys. Rev. D 60 (1999), 094017.
- [32]S. V. Goloskokov and P. Kroll, Eur. Phys. J. C 65 (2010), 137-151.
- [33]S. V. Goloskokov and P. Kroll, Eur. Phys. J. A 47 (2011), 112.
- [34]P. Kroll, H. Moutarde and F. Sabatie, Eur. Phys. J. C 73 (2013) no.1, 2278.
- [35]S. V. Goloskokov and P. Kroll, Eur. Phys. J. C 53 (2008), 367-384.
- [36]K. Kumerički and D. Müller, EPJ Web Conf. 112 (2016), 01012.
- [37]K. Kumerički, S. Liuti and H. Moutarde, Eur. Phys. J. A 52 (2016) no.6, 157.
- [38]H. Moutarde, P. Sznajder and J. Wagner, Eur. Phys. J. C 79 (2019) no.7, 614.
- [39]B. Kriester, P. Velie, E. Yeats, F. Y. Lopez and S. Liuti, Phys. Rev. D 105 (2022) no.5, 056022.
- [40]Y. Guo, F. P. Aslan, X. Ji and M. G. Santiago, Phys. Rev. Lett. 135 (2025) no.26, 261903.
- [41]H. T. Ding, X. Gao, S. Mukherjee, P. Petreczky, Q. Shi, S. Syritsyn and Y. Zhao, JHEP 02 (2025), 056.
- [42]X. Gao, S. Mukherjee, Q. Shi, F. Yao and Y. Zhao, Phys. Rev. D 113 (2026) no.1, 014505.
- [43]H. W. Lin, Phys. Lett. B 846 (2023), 138181.
- [44]S. Bhattacharya, K. Cichy, M. Constantinou, J. Dodson, X. Gao, A. Metz, J. Miller, S. Mukherjee, P. Petreczky and F. Steffens, et al. Phys. Rev. D 109 (2024) no.3, 034508.
- [45]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.
- [46]M. H. Chu, M. Colaço, S. Bhattacharya, K. Cichy, M. Constantinou, A. Metz and F. Steffens, Phys. Rev. D 112 (2025) no.9, 094510.
- [47]K. Cichy, S. Bhattacharya, M. Constantinou, J. Dodson, X. Gao, A. Metz, J. Miller, S. Mukherjee, A. Scapellato and F. Steffens, et al. Acta Phys. Polon. Supp. 16 (2023) no.7, 7-A6.
- [48]C. Alexandrou, K. Cichy, M. Constantinou, K. Hadjiyiannakou, K. Jansen, A. Scapellato and F. Steffens, Phys. Rev. Lett. 125 (2020) no.26, 262001.
- [49]H. W. Lin, Phys. Lett. B 824 (2022), 136821.
- [50]H. Dutriex et al. [HadStruc], JHEP 08 (2024), 162.
- [51]H. Dutriex, R. G. Edwards, J. Karpie, C. Mezrag, C. Monahan, K. Orginos, A. Radyushkin, D. Richards, E. Romero and S. Zafeiropoulos, [arXiv:2604.21476 [hep-lat]].
- [52]A. V. Radyushkin, Phys. Lett. B 385 (1996), 333-342.
- [53]L. Mankiewicz, G. Piller and T. Weigl, Eur. Phys. J. C 5 (1998), 119-128.
- [54]L. L. Frankfurt, M. V. Polyakov, M. Strikman and M. Vanderhaeghen, Phys. Rev. Lett. 84 (2000), 2589-2592.
- [55]L. L. Frankfurt, P. V. Pobylitsa, M. V. Polyakov and M. Strikman, Phys. Rev. D 60 (1999), 014010.
- [56]L. Mankiewicz, G. Piller and A. Radyushkin, Eur. Phys. J. C 10 (1999), 307-312.
- [57]F. D. Aaron et al. [H1], Phys. Lett. B 681 (2009), 391-399.
- [58]S. Chekanov et al. [ZEUS], JHEP 05 (2009), 108.
- [59]A. Airapetian et al. [HERMES], JHEP 10 (2012), 042.
- [60]N. Hirlinger Saylor et al. [CLAS], Phys. Rev. C 98 (2018) no.4, 045203.
- [61]M. Hattawy et al. [CLAS], Phys. Rev. Lett. 123 (2019) no.3, 032504.
- [62]R. Akhunzyanov et al. [COMPASS], Phys. Lett. B 793 (2019), 188-194 [erratum: Phys. Lett. B 800 (2020), 135129].
- [63]A. Hobart et al. [CLAS], Phys. Rev. Lett. 133 (2024) no.21, 211903.
- [64]A. Airapetian et al. [HERMES], Phys. Lett. B 659 (2008), 486-492.
- [65]T. Horn, X. Qian, J. Arrington, R. Asaturyan, F. Benmokhtar, W. Boeglin, P. Bosted, A. Bruell, M. E. Christy and E. Cisbouk, et al. Phys. Rev. C 78 (2008), 058201.
- [66]A. Airapetian et al. [HERMES], Phys. Lett. B 682 (2010), 345-350.
- [67]S. A. Morrow et al. [CLAS], Eur. Phys. J. A 39 (2009), 5-31.
- [68]I. Bedlinsky et al. [CLAS], Phys. Rev. C 90 (2014) no.2, 025205.
- [69]M. Defurne et al. [Jefferson Lab Hall A], Phys. Rev. Lett. 117 (2016) no.26, 262001.
- [70]M. Mazouz et al. [Jefferson Lab Hall A], Phys. Rev. Lett. 118 (2017) no.22, 222002.
- [71]M. Diamini et al. [Jefferson Lab Hall A], Phys. Rev. Lett. 127 (2021) no.15, 152301.
- [72]G. D. Alexeev et al. [COMPASS], Eur. Phys. J. C 83 (2023), no.10, 924.
- [73]S. Diehl et al. [CLAS], Phys. Lett. B 839 (2023), 137761.
- [74]G. D. Alexeev et al. [COMPASS], Phys. Lett. B 870 (2025), 139832.
- [75]A. Accardi, P. Achenbach, D. Adhikari, A. Afanasev, C. S. Akondi, N. Akopov, M. Albaladejo, H. Albataineh, M. Albrecht and B. Almeida-Zamora, et al. Eur. Phys. J. A 60 (2024) no.9, 173.
- [76]R. Abdul Khalek, A. Accardi, J. Adam, D. Adamiak, W. Akers, M. Albaladejo, A. Al-batainah, M. G. Alexeev, F. Ameli and P. Antonioli, et al. Nucl. Phys. A 1026 (2022), 122447.
- [77]D. P. Anderle, V. Bertone, X. Cao, L. Chang, N. Chang, G. Chen, X. Chen, Z. Chen, Z. Cui and L. Dai, et al. Front. Phys. (Beijing) 16 (2021) no.6, 64701.
- [78]V. M. Braun, Y. Ji and J. Schoenleber, Phys. Rev. Lett. 129 (2022) no.17, 172001.
- [79]Y. Ji and J. Schoenleber, JHEP 01 (2024), 053.
- [80]V. M. Braun, P. Gotzler and A. N. Manashov,
- [81]V. M. Braun, H. Y. Jiang, A. N. Manashov and A. von Manteuffel, JHEP 01 (2025), 069.
- [82]M. Diehl and W. Kugler, Eur. Phys. J. C 52 (2007), 933-966.
- [83]D. Müller, T. Lautenschlager, K. Passek-Kumericki and A. Schaefer, Nucl. Phys. B 884 (2014), 438-546.
- [84]G. Duplančić, D. Müller and K. Passek-Kumerički, Phys. Lett. B 771 (2017), 603-610.
- [85]M. Čučić, G. Duplančić, K. Kumerički and K. Passek-K., JHEP 12 (2023), 192 [erratum: JHEP 02 (2024), 225].
- [86]A. V. Belitsky and D. Mueller, Phys. Lett. B 513 (2001), 349-360.
- [87]D. Y. Ivanov, L. Szymanowski and G. Krasnikov, JETP Lett. 80 (2004), 226-230 [erratum: JETP Lett. 101 (2015) no.12, 844].
- [88]J. C. Collins, L. Frankfurt and M. Strikman, Phys. Rev. D 56 (1997), 2982-3006.
- [89]A. V. Belitsky, A. Freund and D. Mueller, Nucl. Phys. B 574 (2000), 347-406 doi:10.1016/S0550-3213(00)00012-2 [arXiv:hep-ph/9912379 [hep-ph]].DOI
- [90]M. H. Sarmadi, Phys. Lett. B 143 (1984), 471 doi:10.1016/0370-2693(84)91504-1
- [91]L. B. Chen, W. Chen, F. Feng and Y. Jia, Phys. Rev. Lett. 132 (2024) no.20, 201901 [erratum: Phys. Rev. Lett. 134 (2025) no.22, 229901].
- [92]Y. Ji, B. X. Shi, J. Wang, Y. F. Wang, Y. M. Wang and H. X. Xu, Phys. Rev. Lett. 134 (2025) no.22, 221901.
- [93]S. Bhattacharya, D. Zheng and J. Zhou, Phys. Rev. Lett. 133 (2024) no.5, 051901.
- [94]P. Nogueira, J. Comput. Phys. 105 (1993), 279-289.
- [95]F. Feng, Y. F. Xie, Q. C. Zhou and S. R. Tang, Comput. Phys. Commun. 265 (2021), 107982.
- [96]F. Feng, Comput. Phys. Commun. 183 (2012), 2158-2164.
- [97]A. V. Smirnov and F. S. Chukharev, Comput. Phys. Commun. 247 (2020), 106887.
- [98]A. V. Kotikov, Phys. Lett. B 254 (1991), 158-164.
- [99]E. Remiddi, Nuovo Cim. A 110 (1997), 1435-1452.
- [100]T. Gehrmann and E. Remiddi, Nucl. Phys. B 580 (2000), 485-518.
- [101]M. Beneke and V. A. Smirnov, Nucl. Phys. B 522 (1998), 321-344.
- [102]A. Pak and A. Smirnov, Eur. Phys. J. C 71 (2011), 1626.
- [103]B. Jantzen, JHEP 12 (2011), 076.
- [104]W. Chen, JHEP 02 (2020), 115.
- [105]W. Chen, M. x. Luo, T. Z. Yang and H. X. Zhu, JHEP 01 (2024), 131.
- [106]W. Chen, Comput. Phys. Commun. 312 (2025), 109607.
- [107]W. Chen, [arXiv:2505.13540 [hep-ph]].arxiv.org/abs/2505.13540
- [108]X. Liu and Y. Q. Ma, Comput. Phys. Commun. 283 (2023), 108565.
- [109]V. L. Chernyak and A. R. Zhitnitsky, Phys. Rept. 112 (1984), 173.
- [110]G. S. Bali et al. [RQCD], JHEP 08 (2019), 065.
- [111]C. Duhr and F. Dulat, JHEP 08 (2019), 135.
- [112]For Hcubature website, see: https://github.com/stevengj/cubature.
- [113]A. P. Bakulev and V. L. Khandramai, Comput. Phys. Commun. 184 (2013) no.1, 183-193.
- [114]V. M. Braun, A. N. Manashov, S. Moch and M. Strohmaier, JHEP 06 (2017), 037.
- [115]M. Strohmaier, “Conformal symmetry breaking and evolution equations in Quantum Chromodynamics,” 2018.DOI
- [116]A. Freese, D. Adamiak, I. Cloët, W. Melnitchouk, J. W. Qiu, N. Sato and M. Zaccheddu, Comput. Phys. Commun. 311 (2025), 109552.
- [117]C. Bechler and D. Mueller, [arXiv:0906.2571 [hep-ph]].arxiv.org/abs/0906.2571
- [118]G. ’t Hooft and M. J. G. Veltman, Nucl. Phys. B 44 (1972), 189-213 doi:10.1016/0550-3213(72)90279-9DOI





