betaFeedback

WorksPhysics and Astronomy

On convergence properties of GPD expansion through Mellin/conformal moments and orthogonal polynomials

Abstract

We examine convergence properties of reconstructing the generalized parton distributions (GPDs) through the universal moment parameterization (GUMP). We provide a heuristic explanation for the connection between the formal summation/expansion and the Mellin-Barnes integral in the literature, and specify the exact convergence condition. We derive an asymptotic condition on the conformal moments of GPDs to satisfy the boundary condition at x=1x=1 and subsequently develop an approximate formula for GPDs when x>ξx>ξ. Since experimental observables constraining GPDs can be expressed in terms of double or even triple summations involving their moments, scale evolution factors, and Wilson coefficients, etc., we propose a method to handle the ordering of the multiple summations and convert them into multiple Mellin-Barnes integrals via analytical continuations of integer summation indices.

Introduction

Generalized parton distributions (GPDs) [1–3], which depict the three-dimensional (3D) structure of the nucleon [4–6] and contain information about the hadron state, such as the mass and angular momentum [2, 7, 8], have played a pivotal role in the non-perturbative regime of quantum chromodynamics (QCD) both theoretically and experimentally. As a generalization of forward parton distributions (PDFs), which only capture the one-dimensional longitudinal structure, GPDs, proposed as off-forward parton distributions [3], leverage nonzero momentum transfer to a hadron. This allows for the inclusion of the skewness parameter ξ\xi describing the fraction of the longitudinal momentum transfer, and the total momentum transfer squared tt, which enables probing the 3D structure of the nucleon. To be specific, GPDs are encapsulated as a function of (x,ξ,t)(x, \xi, t), denoted as F(x,ξ,t)F(x, \xi, t), where xx represents the regular parton momentum fraction. In contrast, PDFs, denoted as f(x)f(x), have only one parameter xx. The properties (including analyticity, symmetries, sum rules, polynomiality conditions, etc.) and dynamics of GPDs have been well studied over the past three decades and are comprehensively reviewed in [3, 9, 10].

On the other hand, significant efforts from experiments have yielded substantial datasets of exclusive measurements from HERA [11–18], Jefferson Lab (JLab) [24] [19, 20, 21, 22, 23], and are expected from the upcoming Electron-Ion Collider (EIC) [25, 26]. GPDs can be accessed through exclusive production processes, such as deeply virtual Compton scattering (DVCS) [27] and deeply virtual meson production (DVMP) [28], according to the factorization theorem [29, 30, 31]. However, extracting GPDs, which are high-dimensional quantities, from the amplitudes of these processes remains challenging, as the amplitudes typically only show tt dependence or ξ\xi & tt dependence from the form factors. This is known as the inverse problem. In addition, the advent of the large momentum effective theory (LaMET) [32, 33, 34] expanded the scope of lattice calculations and shed light on the first principle calculation of GPDs [35, 36, 37, 38].

Considering all these constraints, the program to parameterize GPDs through universal moment parameterization (GUMP) [39, 40] is proposed. This program aims to perform a global analysis by combining lattice calculations, PDFs, and experimental measurements for the form factors, in order to solve the inverse problem. In GUMP, an infinite number of conformal moments of GPDs are parameterized with a finite number of free parameters (called universal parameterization). These parameters are expected to fit the number of different constraints on the GPDs, allowing the GPDs to be fully determined and reconstructed using Mellin-Barnes integrals [41, 42]. However, the mathematical correlations between the formal summation and the Mellin-Barnes integral remain subtle and unclear. We also observed that the reconstructed GPDs do not vanish at ∣x∣=1|x| = 1 in previous literature [39]. Furthermore, developing a framework to convert the multiple summations into multiple Mellin-Barnes integrals is crucial for accurately incorporating the next-to-leading order (NLO) evolution in experimental observables. Thus, a solid mathematical treatment is needed to resolve these issues and support future programs.

The paper is structured as follows. In § 1.1, we review the Mellin moment expansion and conformal moment expansion of PDFs and GPDs in a logical approach. We heuristically argue the relationship between the formal summation and the Mellin-Barnes integral, connecting them with the extensions of δ(n)(x)\delta^{(n)}(x) and conformal partial wave functions pn(x,ξ)p_n(x,\xi). In Section III, we study the asymptotic expansion of the Mellin-Barnes integral in the complex conformal moment plane and find an approximate formula for F(x,ξ,t)F(x,\xi,t) when x>ξx>\xi. More importantly, we obtain an asymptotic condition for the conformal moments of GPDs to ensure F(x,ξ,t)F(x,\xi,t) vanish at ∣x∣=1|x|=1. In Section IV, we address the problem of double summations when considering the scale evolution operator Ejkii′(ξ,Q,μ0)E_{jk}^{ii'}(\xi,Q,\mu_0) containing non-trivial mixing in the conformal space at NLO [43, 44]. This involves the presence of non-diagonal terms in Ejkii′(ξ,Q,μ0)E_{jk}^{ii'}(\xi,Q,\mu_0) between jj and kk indices. We transform the double summations into double Mellin-Barnes integrals using a novel method to analytically continue the summation with variable bounds, followed by the application of the Sommerfeld-Watson transformation. The method can easily be generalized to handle multiple summations.

Reconstructing PDFs and GPDs from their moments

In this section, we provide some new insight on existing Mellin moment and conformal moment expansions for PDFs and GPDs, respectively [39, 40, 41, 42]. We reconstruct the inverse Mellin transformation and Mellin-Barnes integral from the formal summations by the extensions of δ(n)(x)\delta^{(n)}(x) and the conformal partial wave functions pn(x,ξ)p_n(x,\xi). That is, we begin with the case of PDFs, which possess a well-established inverse Mellin transformation. Through the anatomy of the inverse Mellin transformation, we can convert it into a summation using the Sommerfeld-Watson transformation, as briefly reviewed in Appendix A. During this process, we identify an extension of δ(n)(x)\delta^{(n)}(x), denoted as Dn(x)\mathcal{D}^n(x). This extension facilitates the formulation of a formal summation, involving a “formal” manipulation. For the case of GPDs, we employ the reverse procedure. Initially, we articulate the formal summation. Subsequently, by determining the appropriate extension for the conformal partial wave function pn(x,ξ)p_n(x,\xi) as the Schlӓfli integral, we arrive at the Mellin-Barnes integral.

Reconstruction of PDFs through the Dirac delta functions

We begin with the Mellin moment expansion of PDFs, a technique emerging from the analysis of deep inelastic scattering and extensively utilized in quantum field theory. This method finds applications ranging from factorization via operator product expansion (OPE) [45, 46, 47] to the Dokshitzer-Gribov-Lipatov-Altarelli-Parisi (DGLAP) equations [48, 49, 50]. In Mellin moment space, the DGLAP equations for PDFs become simple multiplicative expressions that can be easily solved to all orders in perturbation theory. The Mellin moment of PDFs f(x)f(x) is defined as

fn:=∫01dx xn−1f(x) ,(1)f_n := \int_0^1 \mathrm{d}x\,x^{n-1}f(x)\,, \tag*{(1)}

where n=1,2,…n=1,2,\ldots are positive integers. One can reconstruct the PDFs via the inverse Mellin transformation

f(x)=12πi∫c−i∞c+i∞x−jfj dj ,(2)f(x)=\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty}x^{-j}f_j\,\mathrm{d}j\,, \tag*{(2)}

where fjf_j is the analytic continuation of fnf_n in the complex jj-plane. Supposing that fjf_j is a meromorphic function on the complex plane with right bounded poles, we can always choose a sufficiently large cc such that the vertical line (c−i∞,c+i∞)(c-i\infty,c+i\infty) is to the right of all the poles (f∞=0f_\infty=0).

The physical condition f(x)=0f(x)=0 only for x≥1x\geq1 is satisfied as long as asymptotically,

fj=O(1j)as j→∞ .(3)f_j=\mathcal{O}\left(\frac{1}{j}\right)\quad\text{as }j\to\infty\,. \tag*{(3)}

In this case, we can construct a hypothetical semicircle with radius R→∞R\to\infty on the right half-plane, connecting with the vertical line (c−i∞,c+i∞)(c-i\infty,c+i\infty), contouring no poles, and there is no contribution to the contour integral on the semicircle. By Cauchy’s theorem, the original integral is 0. The integration contours are depicted in Figure 1. Note that the small-oo symbol we employ in this paper is slightly different from its strict mathematical definition. It means the function can only decay polynomially faster than 1/j1/j; it cannot decay exponentially, as this would cause the support smaller than [0,1][0,1]. As we can see in (2), there will be a rescaling of xx if there is an exponentially decaying term in fjf_j. In principle, there are many choices of functions to parametrize the Mellin moment fjf_j. In Table I, we show some models of Mellin moment fjf_j and the corresponding PDFs f(x)f(x). Observing (2), one can rewrite it in the following form

Image not extracted

Figure 1. Hypothetical integration contour for ∣x∣≥1|x| \ge1 (left) and ∣x∣<1|x| < 1 (right).

Mellin moment fjf_jPDF f(x)f(x)
(j−j0)−α(j-j_0)^{-\alpha}, α>1\alpha> 1(−1)α−1x−j0ln⁡α−1(x)θ(1−∣x∣)/Γ(α)(-1)^{\alpha-1}x^{-j_0}\ln^{\alpha-1}(x)\theta(1-|x|)/\Gamma(\alpha)
B(a+j,b+1)B(a+j,b+1), b>0b>0f(x)=xa(1−x)bθ(1−∣x∣)f(x)=x^a(1-x)^b\theta(1-|x|)
c−jj−αc^{-j}j^{-\alpha}, α>1\alpha>1(−1)α−1ln⁡α−1(cx)θ(1/c−∣x∣)/Γ(α)(-1)^{\alpha-1}\ln^{\alpha-1}(cx)\theta(1/c-|x|)/\Gamma(\alpha)
c−jB(a+j,b+1)c^{-j}B(a+j,b+1), b>0b>0(cx)a(1−cx)bθ(1/c−∣x∣)(cx)^a(1-cx)^b\theta(1/c-|x|)

Table 1. Some basic models of Mellin moments fjf_j and the corresponding PDFs f(x)f(x). The first row represents the simplest polynomial decaying moment, the second row represents the Beta function moment, which is generally used in parameterization [51, 52] for its simple PDFs f(x)=xa(1−x)bθ(1−∣x∣)f(x)=x^a(1-x)^b\theta(1-|x|). Note that θ(1−∣x∣)\theta(1-|x|) here ensures that the PDFs vanish when ∣x∣≥1|x|\ge1. The third and fourth rows represent variations of the first and second rows. As observed, the presence of an exponential decay term results in the PDFs having a more restricted support [0,1/c][0,1/c].

f(x)=12πi∫c−i∞c+i∞fjx−j dj,(4)f(x)=\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty} f_j x^{-j}\,dj , \tag*{(4)}
=−12i∮Cjfjsin⁡πjΓ(j)sin⁡πjπx−jΓ(j) dj,(5)=-\frac{1}{2i}\oint_{C_j}\frac{f_j}{\sin\pi j}\frac{\Gamma(j)\sin\pi j}{\pi}\frac{x^{-j}}{\Gamma(j)}\,dj , \tag*{(5)}

where cc is chosen to lie between 0 and 1, under the assumption that fjf_j has no poles with a real part exceeding cc, and the contour CjC_j encloses all the poles at j∈N+j\in\mathbb{N}^{+}, which are artificially introduced by the 1/sin⁡(πj)1/\sin(\pi j) term. Furthermore, we define

Dj(x)=Γ(j+1)sin⁡(π[j+1])πx−j−1,x>0,(6)\mathcal{D}^{j}(x)=\frac{\Gamma(j+1)\sin(\pi[j+1])}{\pi}x^{-j-1},\qquad x>0 , \tag*{(6)}

which is essentially the integral kernel of the inverse Mellin transform, up to a factor. Inserting (6) back into (5), we can further get

f(x)=−12i∮Cj1sin⁡πjfjDj−1(x)Γ(j) dj,(7)f(x)=-\frac{1}{2i}\oint_{C_j}\frac{1}{\sin\pi j}\frac{f_j\mathcal{D}^{j-1}(x)}{\Gamma(j)}\,dj , \tag*{(7)}
=∑n=1∞(−1)n−11(n−1)!fnDn−1(x),(8)= \sum_{n=1}^{\infty}(-1)^{n-1}\frac{1}{(n-1)!}f_n\mathcal{D}^{n-1}(x), \tag*{(8)}
=∑n=0∞(−1)nn!fn+1Dn(x).(9)= \sum_{n=0}^{\infty}\frac{(-1)^n}{n!}f_{n+1}\mathcal{D}^n(x). \tag*{(9)}

We applied the Sommerfeld-Watson transformation inversely in the second equality. However, the validity of the transformation here is subtle, rendering the expansion merely formal as we will see later. If we consider xmx^m as a basis functions, it appears that f(x)f(x) is expanded in the dual basis functions Dn(x)\mathcal{D}^n(x). We can verify the orthogonal relation between Dn(x)\mathcal{D}^n(x) and xmx^m for integers nn and $m,

∫dx Dn(x)xm=n!sin⁡(π[n+1])π∫dx x−n+m−1,(10)\int dx\,\mathcal{D}^n(x)x^m=\frac{n!\sin(\pi[n+1])}{\pi}\int dx\,x^{-n+m-1}, \tag*{(10)}
=n!sin⁡(π[n+1])π(m−n)xm−n,(11)=\frac{n!\sin(\pi[n+1])}{\pi(m-n)}x^{m-n}, \tag*{(11)}
={0if m≠n,(−1)nn!if m=n.(12)=\begin{cases} 0 & \text{if }m\ne n,\\ (-1)^n n! & \text{if }m=n. \end{cases} \tag*{(12)}

For the case m=nm=n, L’Hôpital’s rule has been applied. This result indicates that Dn(x)\mathcal{D}^n(x) behaves as the derivatives of the Dirac delta function δ(n)(x)\delta^{(n)}(x), which satisfies:

∫−11dx δ(n)(x)xm=(−1)nn!δmn.(13)\int_{-1}^{1}dx\,\delta^{(n)}(x)x^m=(-1)^n n!\delta_{mn}. \tag*{(13)}

Thus, we can identify Dn(x)\mathcal{D}^n(x) as δ(n)(x)\delta^{(n)}(x) due to their congruent properties. Therefore, equation (9) can be interpreted as the formal summation shown in [39]:

f(x)=∑n=0∞(−1)nn!fn+1δ(n)(x).(14)f(x)=\sum_{n=0}^{\infty}\frac{(-1)^n}{n!}f_{n+1}\delta^{(n)}(x). \tag*{(14)}

Note that the order of summation and integration cannot actually be interchanged when acting on (9) or (14) with xnx^n, making such manipulation merely “formal.” Nonetheless, both approaches yield the (n+1)(n+1)-th Mellin moment fn+1f_{n+1}. In this context, xmx^m and δ(n)(x)\delta^{(n)}(x) serve as the basis and dual basis in their respective function spaces, satisfying the orthogonal condition (13). The support of f(x)f(x) should be [0,1][0,1], whereas the support of δ(n)(x)\delta^{(n)}(x) or Dn(x)\mathcal{D}^n(x) is {0}\{0\}. This discrepancy contributes to the divergence observed in (14). Specifically, for the formal summation (14) with singular support at {0}\{0\}, no non-trivial Mellin moment fnf_n can converge: convergence only happens when the PDF has support only at x=0x=0, namely a delta function. In fact, we can use functional analysis and introduce the concepts of distributions and analytic functionals [53] to rigorously formalize the summation (14). We omit the detailed mathematical exposition of these concepts here to maintain our focus on the physical aspects.

Note that we can also express equation (6) as follows:

Dj(x)=Γ(j+1)2πi(eiπj−e−iπj)x−j−1,(15)\mathcal{D}^j(x)=\frac{\Gamma(j+1)}{2\pi i}\left(e^{i\pi j}-e^{-i\pi j}\right)x^{-j-1}, \tag*{(15)}
=−Γ(5/2+j)Γ(1+j)Γ(1/2)Γ(2+j)12πi∫−11du (u2−1)j+1xj+1,(16)=-\frac{\Gamma(5/2+j)\Gamma(1+j)}{\Gamma(1/2)\Gamma(2+j)}\frac{1}{2\pi i}\int_{-1}^{1}du\,\frac{(u^2-1)^{j+1}}{x^{j+1}}, \tag*{(16)}

where uu is a complex variable and the contour encircles the interval u∈[−1,1]u\in[-1,1]. The above integral expression is useful for analyzing GPD expansion in the next subsection.

Reconstructing GPDs from conformal moments and partial wave functions

For GPDs F(x,ξ,t)F(x,\xi,t), the reconstruction from moments can in principle be made in parallel to the previous subsection. One can start from Mellin moments Fj(ξ,t)F_j(\xi,t), and obtain F(x,ξ,t)F(x,\xi,t) through a similar inverse Mellin transformation. However, this straightforward generalization obscures a very important analytical property of GPDs. The variables of GPDs are defined as

x=2k+P++P′+,ξ=−Δ+P++P′+,t=Δ2,Δ≡P′−P,(17)x=\frac{2k^+}{P^++P^{\prime+}},\quad\xi=-\frac{\Delta^+}{P^++P^{\prime+}},\quad t=\Delta^2,\quad\Delta\equiv P'-P, \tag*{(17)}

where PP and P′P^\prime are momenta of the incoming and outgoing nucleon, kk is the momentum carried by the parton. Indeed, the physics of GPD functions can be divided into two separated regions as illustrated in the “phase diagram” Figure 2. These are the PDF-like region (x≥ξx \ge\xi and x≤−ξx \le-\xi) and the distribution amplitude (DA)-like region (−ξ<x<ξ-\xi< x < \xi), also referred to as the DGLAP and Efremov-Radyushkin-Brodsky-Lepage (ERBL) regions in the literature, respectively. These two regions connect at the lines x=±ξx = \pm\xi, where the GPDs are continuous but not analytic. This property of GPDs is ensured by the factorization theorem [30]. Physically, the non-analyticity arises from the differing interpretations in these two regions as shown in Figure 3. In the PDF-like region, GPDs represent the probability amplitude for the emission and absorption of a quark with different momentum fractions, analogous to the interpretation of PDFs. In contrast, in the DA-like region, GPDs represent the probability amplitude of a meson-like entity within the hadron, similar to the DA in meson production. At the lines x=±ξx = \pm\xi, one of the two partons has zero momentum, resulting in a “phase transition.”

Image not extracted

Figure 2. “Phase diagram” of GPDs, showing different kinematical regions on the (x,ξ)(x,\xi) plane. This figure is sourced from [39].

Image not extracted

Figure 3. Kinematics for GPDs in PDF-like (left) and DA-like (right) regions.

One can in principle reconstruct the above phase diagram through polynomiality condition [3] on the GPD moments. However, there is a simpler way to do this: Instead of expanding formally in terms of Mellin moments and δ(n)(x)\delta^{(n)}(x) with singular support of {0}\{0\}, one can introduce new class of moments through orthogonal polynomials in x∈[−1,1]x \in[-1,1] and expand the GPD in terms of the dual basis of these polynomials defined in the larger support region x∈[−ξ,ξ]x \in[-\xi,\xi], which reduces to the δ(n)(x)\delta^{(n)}(x) in the forward limit. The dual basis has the same analytic property as GPDs that they are continuous but not analytic at x=±ξx = \pm\xi. In this way, the phase-diagram of GPD is naturally incorporated from the beginning.

Although any choices of orthorgonal polynomials shall work, we follow [41] to use the conformal moments, as a generalization of Mellin moments, to perform the conformal partial wave expansion or conformal OPE [44]. The advantage of using conformal moments is that, when considering the DGLAP equations for GPDs, the convolution between the leading-order (LO) evolution kernel and GPDs can be diagonalized in the conformal moment space [39, 54]. However, at NLO and even higher orders, non-diagonal terms persist, causing non-trivial mixing of conformal indices between the evolution kernel and conformal moments, as we will examine in Sec. IV. In that case, double summations and double complex-plane integrals will be involved, which makes the above choice less obvious.

The conformal moment of GPD F(x,ξ,t)F(x,\xi,t) is defined as

Fn(ξ,t)=∫−11dx cn(x,ξ)F(x,ξ,t),(18)F_n(\xi,t)=\int_{-1}^{1}dx\,c_n(x,\xi)F(x,\xi,t), \tag*{(18)}

where cn(x,ξ)c_n(x,\xi) are Gegenbauer polynomials Cn32(x/ξ)C_n^{\frac{3}{2}}(x/\xi) up to a conventional normalization factor, that is [41]

cn(x,ξ)=ξnΓ(32)Γ(1+n)2nΓ(32+n)Cn32(xξ),(19)c_n(x,\xi)=\xi^n\frac{\Gamma\left(\frac{3}{2}\right)\Gamma(1+n)}{2^n\Gamma\left(\frac{3}{2}+n\right)}C_n^{\frac{3}{2}}\left(\frac{x}{\xi}\right), \tag*{(19)}

where nn are also positive integers. Especially, it reduces to xnx^n for Mellin moment in the forward limit, that is lim⁡ξ→0cn(x,ξ)=xn\lim_{\xi\to0}c_n(x,\xi)=x^n. If we select cn(x,ξ)c_n(x,\xi) as the basis of the function space, the corresponding dual basis pn(x,ξ)p_n(x,\xi), known as the conformal partial wave function, can be determined through the orthogonality condition:

∫−ξξpm(x,ξ)cn(x,ξ) dx=(−1)nδmn.(20)\int_{-\xi}^{\xi}p_m(x,\xi)c_n(x,\xi)\,dx=(-1)^n\delta_{mn}. \tag*{(20)}

Note that the integral bounds are x∈[−ξ,ξ]x\in[-\xi,\xi]. This constraint is necessary because this relationship holds only for x/ξ∈[−1,1]x/\xi\in[-1,1], owing to the presence of the weighting function w(x/ξ)=[1−(x/ξ)2]w(x/\xi)=[1-(x/\xi)^2] in the orthogonal relation for the Gegenbauer polynomials:

∫−ξξCm32(xξ)Cn32(xξ)w(xξ) dx=ξΓ(n+3)(n+32)Γ(n+1)δmn.(21)\int_{-\xi}^{\xi}C_m^{\frac{3}{2}}\left(\frac{x}{\xi}\right)C_n^{\frac{3}{2}}\left(\frac{x}{\xi}\right)w\left(\frac{x}{\xi}\right)\,dx=\xi\frac{\Gamma(n+3)}{\left(n+\frac{3}{2}\right)\Gamma(n+1)}\delta_{mn}. \tag*{(21)}

The conformal partial wave function pn(x,ξ)p_n(x,\xi) is then defined as follows:

pn(x,ξ)=(−1)nξ−n−12nΓ(52+n)Γ(32)Γ(3+n)[1−(xξ)2]Cn32(xξ)θ(ξ−∣x∣)for ∣x∣<ξ,(22)p_n(x,\xi)=(-1)^n\xi^{-n-1}\frac{2^n\Gamma\left(\frac{5}{2}+n\right)}{\Gamma\left(\frac{3}{2}\right)\Gamma(3+n)} \left[1-\left(\frac{x}{\xi}\right)^2\right]C_n^{\frac{3}{2}}\left(\frac{x}{\xi}\right)\theta(\xi-|x|)\quad\text{for }|x|<\xi, \tag*{(22)}

where θ(ξ−∣x∣)\theta(\xi-|x|) is the Heaviside step function, ensuring the function’s definition is restricted to the interval ∣x∣<ξ|x|<\xi. In analogy with Eqs. (9) and (14), the formal expansion/summation for GPDs now reads:

F(x,ξ)=∑n=0∞(−1)nFn(ξ,t)pn(x,ξ).(23)F(x,\xi)=\sum_{n=0}^{\infty}(-1)^nF_n(\xi,t)p_n(x,\xi). \tag*{(23)}

The above properties of pn(x,ξ)p_n(x,\xi) result in the exclusion of information for x>ξx>\xi and x<−ξx<-\xi, leading to the divergence of the formal summation. The reduction of support necessitates a stronger decaying condition on the conformal moments FnF_n of GPDs to make the formal summation converge, as discussed in the case of PDFs. More specifically, if GPDs are non-vanishing only in the DA-like region, the resulting moments will make the above expansion convergent. Any non-vanishing contributions in the PDF-like region will generate conformal moment series rendering the above expansion divergent. When convoluting the formal summation (23) with cn(x,ξ)c_n(x,\xi), we anticipate it to yield the nn-th conformal moment Fj(ξ,t)F_j(\xi,t). Nonetheless, the legitimacy of interchanging between the integral and summation in this context is subtle, rendering this also a “formal” manipulation. An appropriate resummation is needed to extract physical results. There are several resummation techniques proposed in the literatures [54, 55, 56, 57, 58]. Here, we reconstruct GPDs using the Mellin-Barnes integral suggested in [41, 42], which is based on the Sommerfeld-Watson transformation. In this case, one has to analytically continue the moments and pn(x,ξ)p_n(x,\xi) to complex jj plane.

It is easy to see that in the forward limit, the above formalism reduces to that for PDFs. Indeed, the conformal partial wave function pn(x,ξ)p_n(x,\xi) can also be written as

pn(x,ξ)=Γ(52+n)n!Γ(12)Γ(2+n)∫−11du (1−u2)n+1δ(n)(x−uξ),(24)p_n(x,\xi)=\frac{\Gamma\left(\frac{5}{2}+n\right)}{n!\Gamma\left(\frac{1}{2}\right)\Gamma(2+n)}\int_{-1}^{1}du\,(1-u^2)^{n+1}\delta^{(n)}(x-u\xi), \tag*{(24)}

which makes the orthogonality condition (20) easy to verify using:

∫−ξξδ(n)(x−uξ)Cn32(xξ) dx=(−1)ndndxnCn32(xξ)∣x=uξ=(−1)nξ−n2nΓ(32+n)Γ(32).(25)\int_{-\xi}^{\xi}\delta^{(n)}(x-u\xi)C_n^{\frac{3}{2}}\left(\frac{x}{\xi}\right)\,dx=(-1)^n\left.\frac{d^n}{dx^n}C_n^{\frac{3}{2}}\left(\frac{x}{\xi}\right)\right|_{x=u\xi}=(-1)^n\xi^{-n}2^n\frac{\Gamma\left(\frac{3}{2}+n\right)}{\Gamma\left(\frac{3}{2}\right)}. \tag*{(25)}

Note that due to the presence of δ(n)(x−uξ)\delta^{(n)}(x-u\xi), the support of pn(x,ξ)p_n(x,\xi) is also confined to [−ξ,ξ][-\xi,\xi]. In the forward limit, lim⁡ξ→0pn(x,ξ)=δ(n)(x)/n!\lim_{\xi\to0}p_n(x,\xi)=\delta^{(n)}(x)/n!, which reduces to the case for PDFs.

However, to make the analytical continuation of pn(x,ξ)p_n(x,\xi), we need another expression, analogous but not identical to the extension of δ(n)(x)\delta^{(n)}(x) to Dj(x)\mathcal{D}^j(x). It turns out that we can generalize (16) by employing the Schlӓfli integral [59] to properly address poles and branch cuts, as detailed in references [41, 42]. Specifically, in our context, this involves modifying xx to x+uξx+u\xi in (16),

pj(x,ξ)=−Γ(5/2+j)Γ(1/2)Γ(2+j)12πi∫−11du (u2−1)j+1(x+uξ)j+1.(26)p_j(x,\xi)=-\frac{\Gamma(5/2+j)}{\Gamma(1/2)\Gamma(2+j)}\frac{1}{2\pi i}\int_{-1}^{1}du\,\frac{(u^2-1)^{j+1}}{(x+u\xi)^{j+1}} . \tag*{(26)}

The integral can be expressed in terms of hypergeometric functions after carefully considering the poles and branch cuts. We summarize the results as follows:

pj(x>ξ,ξ)=sin⁡(π[j+1])πx−j−1 2F1(j+12,j+22;52+j∣ξ2x2),(27)p_j(x>\xi,\xi)=\frac{\sin(\pi[j+1])}{\pi}x^{-j-1}\,{}_2F_1\left(\frac{j+1}{2},\frac{j+2}{2};\frac{5}{2}+j\mathrel{\Big|}\frac{\xi^2}{x^2}\right) , \tag*{(27)}
pj(∣x∣≤ξ,ξ)=2j+1Γ(5/2+j)ξ−j−1Γ(1/2)Γ(1+j)(1+x/ξ)2F1(−1−j,j+2,2∣ξ+x2ξ),(28)p_j(|x|\leq\xi,\xi)=\frac{2^{j+1}\Gamma(5/2+j)\xi^{-j-1}}{\Gamma(1/2)\Gamma(1+j)}(1+x/\xi){}_2F_1\left(-1-j,j+2,2\mathrel{\Big|}\frac{\xi+x}{2\xi}\right) , \tag*{(28)}
pj(x≤−ξ,ξ)=0.(29)p_j(x\leq-\xi,\xi)=0 . \tag*{(29)}

It has been verified that pjp_j is continuous at x=±ξx=\pm\xi, but not analytic at these points. In the forward limit, equation (27) reduces to equation (6). In the DA-like region, equation (28) reduces to equations (24) and (22). Observe that for integer jj and x=±ξx=\pm\xi, the conformal partial wave function pj(x,ξ)p_j(x,\xi) is zero.

To do the resummation, applying the Sommerfeld-Watson transformation to equation (23), the Mellin-Barnes integral in Ref. [41, 42] is obtained:

F(x,ξ,t)=12πi∫c−i∞c+i∞dj pj(x,ξ)sin⁡(π[j+1])Fj(ξ,t).(30)F(x,\xi,t)=\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty}dj\,\frac{p_j(x,\xi)}{\sin(\pi[j+1])}F_j(\xi,t) . \tag*{(30)}

In the forward limit, it reduces to the inverse Mellin transformation as the PDFs. Furthermore, when x=±ξx=\pm\xi, the GPDs are not zero, as the evaluation of the integral along the imaginary axis ensures that the sine term in (27) remains non-zero. Further studies on its analytic properties will be conducted in the next section. Note that in this resummation, GPDs F(x,ξ,t)F(x,\xi,t) are zero for x<−ξx<-\xi. This situation is analogous to quark PDFs, where we set f(x)=0f(x)=0 for x<0x<0, with the negative xx region included in the antiquark PDFs. Since GPDs do not inherently distinguish between quark and antiquark components, they are formulated explicitly as a linear combination of various contributions: quark GPDs, antiquark GPDs, and an additional DA term, as discussed in [40].

We would like to emphasize again that the strategy in this subsection is quite general. In principle, any complete set of polynomials can serve as basis functions with the argument x/ξx/\xi to expand the GPDs and derive their respective moments. By applying the orthogonality condition, we can find the corresponding dual basis functions with a restricted support, which is smaller than [−1,1][-1,1]. Due to discrepancies of the supports, the formal summation will be divergent. However, after identifying a suitable generalization of the dual basis functions, we can employ the Sommerfeld-Watson transformation to resum the moments and reconstruct the GPDs accurately over the entire range.

Mellin-Barnes integrals and asymptotic behavior of GPDs

In this section, we study the analytic properties of the Mellin-Barnes integral on the complex jj-plane. We begin with the concept of residue at infinity, then use this concept to examine the inverse Mellin transformation (2) and the Mellin-Barnes integral (30) by studying their asymptotic behaviors. Next, we derive the asymptotic conditions on GPDs in the complex plane and test the formulation with a minimal example of GUMP. The asymptotic conditions are of great importance, as it ensures that the GPDs vanish at ∣x∣=1|x|=1, a property that was violated in previous parameterizations [39]. Finally, we derive an approximate formula for GPDs when x>ξx>\xi.

Residue at infinity

It is useful to introduce the extended complex plane as C∗=C∪{∞}\mathbb{C}^*=\mathbb{C}\cup\{\infty\}, which is compact and isomorphic to the Riemann sphere. Define the residue at infinity for a function f(z)f(z) as [60]

Res⁡(f,∞)=−Res⁡(1z2f(1z),0).(31)\operatorname{Res}(f,\infty)=-\operatorname{Res}\left(\frac{1}{z^2}f\left(\frac{1}{z}\right),0\right) . \tag*{(31)}

For a meromorphic function on the Riemann sphere, the sum of the residues (including the one at infinity) must be zero. This can be regarded as another definition of the residue at infinity:

Res⁡(f,∞)=−∑k=1nRes⁡(f,ak),(32)\operatorname{Res}(f,\infty)=-\sum_{k=1}^{n}\operatorname{Res}(f,a_k), \tag*{(32)}

where aka_k are poles of f(z)f(z) in C\mathbb{C}. We start with the inverse Mellin transformation (2) reconstructing PDFs to elaborate on this concept. We have studied the properties of this integral when ∣x∣≥1\lvert x\rvert\ge1 to ensure it yields zero. When ∣x∣<1\lvert x\rvert< 1, we construct a contour to encircle all the poles as in Figure 1, that is

f(x)=12πi∫c−i∞c+i∞x−jfj dj,(33)f(x)=\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty}x^{-j}f_j\,\mathrm{d}j, \tag*{(33)}
=∑k=1nRes⁡(x−jfj,ak),(34)=\sum_{k=1}^{n}\operatorname{Res}(x^{-j}f_j,a_k), \tag*{(34)}
=−Res⁡(x−jfj,∞).(35)=-\operatorname{Res}(x^{-j}f_j,\infty). \tag*{(35)}

Denoting the inverse Mellin transformation from jj space to xx space as f(x)={M−1fj}(x)f(x)=\{\mathcal{M}^{-1}f_j\}(x), we have

{M−1fj}(x)=−Res⁡(x−jfj,∞).(36)\{\mathcal{M}^{-1}f_j\}(x)=-\operatorname{Res}(x^{-j}f_j,\infty). \tag*{(36)}

One can easily verify this for a simple PDF model f(x)=xα(1−x)βf(x)=x^\alpha(1-x)^\beta if the Mellin moment fj=B(j+α,β+1)f_j=B(j+\alpha,\beta+1). The situation for GPDs is alike when x>ξx>\xi, which is in the PDF-like region. From (30) and (27), we have

F(x,ξ,t)=12πi∫c−i∞c+i∞dj x−j−1 2F1((j+1)/2,(j+2)/25/2+j|ξ2x2)Fj(ξ,t),(37)F(x,\xi,t)=\frac{1}{2\pi i}\int_{c-i\infty}^{c+i\infty}\mathrm{d}j\,x^{-j-1}\,{}_2F_1\left(\begin{matrix}(j+1)/2,(j+2)/2\\5/2+j\end{matrix}\middle|\frac{\xi^2}{x^2}\right)\mathcal{F}_j(\xi,t), \tag*{(37)}

with cc chosen to be to the right of all the poles. Analyzing this integral, we find that:

  1. When x=1x=1, the integral should be zero. We impose conditions on Fj(ξ,t)\mathcal{F}_j(\xi,t) such that 2F1(…)Fj(ξ,t){}_2F_1(\ldots)\mathcal{F}_j(\xi,t) asymptotically decays faster than 1/j1/j. Then by adding a right half infinite hemicircle, whose contribution to the integral is zero, the original integral equals the sum of residues in the contour. However, after carefully chosen cc, Fj(ξ,t)\mathcal{F}_j(\xi,t) and the hypergeometric function both have no poles there. Therefore, the integral is zero as expected. The asymptotic expansion for the hypergeometric function can be found in [61],

2F1((j+1)/2,(j+2)/25/2+j|z:=ξ2x2)∼22C(z)[21+1−z]jas ∣j∣→∞, ∣arg⁡j∣<π, z<1,(38){}_2F_1\left(\begin{matrix}(j+1)/2,(j+2)/2\\5/2+j\end{matrix}\middle|z:=\frac{\xi^2}{x^2}\right)\sim2\sqrt{2}C(z)\left[\frac{2}{1+\sqrt{1-z}}\right]^j\quad\text{as }|j|\to\infty,\ |\arg j|<\pi,\ z<1, \tag*{(38)}

where C(z)=(1−1−z)1/2(1+1−z)3/2C(z)=\frac{(1-\sqrt{1-z})^{1/2}}{(1+\sqrt{1-z})^{3/2}} is a constant for given zz. It shows that to make 2F1(…)Fj(ξ,t)=o(1/j){}_2F_1(\ldots)\mathcal{F}_j(\xi,t)=o(1/j), as ∣j∣→∞|j|\to\infty, ∣arg⁡j∣<π/2|\arg j|<\pi/2 when x=1x=1, the conformal moment must satisfy

Fj=(21+1−ξ2)−jo(1j)as ∣j∣→∞, ∣arg⁡j∣<π2.(39)\mathcal{F}_j=\left(\frac{2}{1+\sqrt{1-\xi^2}}\right)^{-j}o\left(\frac{1}{j}\right)\quad\text{as }|j|\to\infty,\ |\arg j|<\frac{\pi}{2}. \tag*{(39)}

Furthermore, we can substantiate the assertion made in the previous section: if we aim to restrict the support of GPDs to [−ξ,ξ][-\xi,\xi] as the support of dual basis functions and ensure the convergence of the formal summation (23), the asymptotic condition of GPDs will be

Fj′=(2/ξ1+1−ξ2)−jo(1j)as ∣j∣→∞, ∣arg⁡j∣<π2,(40)\mathcal{F}'_j=\left(\frac{2/\xi}{1+\sqrt{1-\xi^2}}\right)^{-j}o\left(\frac{1}{j}\right)\quad\text{as }|j|\to\infty,\ |\arg j|<\frac{\pi}{2}, \tag*{(40)}

which actually represents a rescaling of xx in (37). This condition is stricter than (39), explaining why the formal summation of the moments derived from the GPDs expansion over the range [−1,1][-1,1] is typically divergent.

  1. When x>1x>1, the factor x−j−1→0x^{-j-1}\to0 as ∣j∣→∞|j|\to\infty, ∣arg⁡j∣<π/2|\arg j|<\pi/2, causing the integrand to decay even faster than in the case for x=1x=1. Hence, the integral is also zero if it is zero at x=1x=1, as anticipated.

  1. When ξ<x<1\xi< x < 1, similarly, the factor x−j−1→0x^{-j-1} \to0 as ∣j∣→∞\lvert j\rvert\to\infty, π/2<∣arg⁡j∣<π\pi/2 < \lvert\arg j\rvert< \pi, and the integral on the left hemicircle is zero. The original integral equals the sum of residues in the contour and is equal to negative of the residue at infinity. That is

F(x,ξ,t)=∑k=0nRes⁡[x−j−12F1((j+1)/2,(j+2)/25/2+j∣ξ2x2)Fj(ξ,t);ak],(41)F(x,\xi,t)=\sum_{k=0}^{n}\operatorname{Res}\left[x^{-j-1}{}_2F_1\left(\left.\frac{(j+1)/2,(j+2)/2}{5/2+j}\right|\frac{\xi^2}{x^2}\right)F_j(\xi,t);a_k\right], \tag*{(41)}
=−Res⁡[x−j−12F1((j+1)/2,(j+2)/25/2+j∣ξ2x2)Fj(ξ,t);∞].(42)=-\operatorname{Res}\left[x^{-j-1}{}_2F_1\left(\left.\frac{(j+1)/2,(j+2)/2}{5/2+j}\right|\frac{\xi^2}{x^2}\right)F_j(\xi,t);\infty\right]. \tag*{(42)}

This suggests that the GPDs in the PDF-like region are exclusively linked to the asymptotic behaviors of both the hypergeometric function 2F1(…){}_2F_1(\ldots) and the conformal moment Fj(ξ,t)F_j(\xi,t) in the complex-jj plane.

A minimal example of GUMP model

We then show the numerical calculations of a minimal example of GUMP. Because of the polynomiality condition of conformal moments of the GPDs [39] [3], we parametrize the conformal moment at small ξ\xi as [62]

Fj(ξ,t)=Fj,0(t)+ξ2Fj,2(t)+O(ξ4),(43)F_j(\xi,t)=F_{j,0}(t)+\xi^2F_{j,2}(t)+O\left(\xi^4\right), \tag*{(43)}

and the coefficient before the skewness parameter ξ\xi is

Fj,k(t)=NkB(j+1−αk,1+βk)j+1−k−αkj+1−k−αk−αk′tβ(t),(44)F_{j,k}(t)=N_kB(j+1-\alpha_k,1+\beta_k)\frac{j+1-k-\alpha_k}{j+1-k-\alpha_k-\alpha'_kt}\beta(t), \tag*{(44)}

in which NkN_k, αk\alpha_k, βk\beta_k, αk′\alpha'_k are parameters determined by global fitting, and the tt dependence is incorporated by Regge trajectory term [63, 64] and the extra residual term β(t)\beta(t) [40]. For simplicity, we only consider the leading term Fj,0F_{j,0}, since ξ=xB/(2−xB)+O(Q−2)\xi=x_B/(2-x_B)+O(Q^{-2}) is small in the high energy limit [62]. We neglect the tt dependence, as it can always be recovered by simply multiplying Regge terms, and focus solely on the jj dependence. We have the minimal model

Fj,original(ξ,0)=B(j+1−α,1+β).(45)F_{j,\mathrm{original}}(\xi,0)=B(j+1-\alpha,1+\beta). \tag*{(45)}

Note that the j+1j+1 index in the conformal moment corresponds to the jj index in the Mellin moment conventionally, so Fj,originalF_{j,\mathrm{original}} corresponds to generalization of the PDF f(x)=x−α(1−x)βf(x)=x^{-\alpha}(1-x)^\beta, which is the most basic ansatz in the CTEQ global analysis [65]. Since this moment does not satisfy the condition (39), we observe that the GPDs reconstructed from it do not vanish at x=1x=1 [39]. However, for relatively not large values of ξ≲0.4\xi\lesssim0.4 (most available experimental data fall into this range), the deviation is negligible.

Theoretically, to make it precisely vanish at x=1x=1, we multiply the moment by the exponential decaying term as specified in (39). Thus, the moment now becomes

Fj,revised(ξ,0)=B(j+1−α,1+β)(21+1−ξ2)−j−1.(46)F_{j,\mathrm{revised}}(\xi,0)=B(j+1-\alpha,1+\beta)\left(\frac{2}{1+\sqrt{1-\xi^2}}\right)^{-j-1}. \tag*{(46)}

Actually, this revision will not cause much difference for moderate skewness, as shown in Figure 4. Only when ξ\xi is large does it bring a noticeable deviation, causing the GPDs to decay faster towards x=1x=1 and become significantly smaller. An obvious drawback of the revised moment is that it does not obey the polynomiality conditions. One can remedy this by expanding the above ξ\xi factor at ξ=0\xi=0 and truncate to an appropriate order, which can be done analytically by using the Mellin-Barnes technique introduced in Sec. IV B to deal with the finite summation. The explicit formula for the moment that satisfies the polynomiality condition is given by

Image not extracted

Figure 4. A plot showing the rescaled GPDs, xF(x,ξ,0)xF(x,\xi,0), versus the parton momentum fraction xx for different skewness parameters ξ=0.2,0.4,0.6,0.8\xi=0.2,0.4,0.6,0.8. The solid lines represent the GPDs reconstructed from original moments, while the dashed lines correspond to the ones reconstructed from revised moments at t=0t=0. The lines x=ξx=\xi are marked with dotted black lines. The plot demonstrates that for moderate skewness, the deviation between the original and revised moments is negligible, whereas at larger skewness values, the deviation becomes more pronounced, with the GPDs decaying more rapidly as xx approaches 1.

Fj,polynomial(ξ,0)=B(j+1−α,1+β)(21+1−ξ2)−j−1−2−j−4(j+1)ξj+3{[1+(−1)j]ξ3F~2(1,32;4−j2,j+62;ξ2)+[1−(−1)j]F~2(1,1,32;2−j+12,j+12+2;ξ2)}B(j+1−α,1+β).(47)\begin{aligned} F_{j,\mathrm{polynomial}}(\xi,0)={}&B(j+1-\alpha,1+\beta)\left(\frac{2}{1+\sqrt{1-\xi^2}}\right)^{-j-1}-2^{-j-4}(j+1)\xi^{j+3}\left\{[1+(-1)^j]\xi_3\widetilde{F}_2\left(1,\frac{3}{2};\frac{4-j}{2},\frac{j+6}{2};\xi^2\right)\right.\\ &\left.+[1-(-1)^j]\widetilde{F}_2\left(1,1,\frac{3}{2};2-\frac{j+1}{2},\frac{j+1}{2}+2;\xi^2\right)\right\}B(j+1-\alpha,1+\beta). \tag*{(47)} \end{aligned}

It can be verified that for integer values of jj, the formula above produces a polynomial of ξ\xi that satisfies the polynomiality condition. As discussed in the previous subsection, since the GPDs only have to do with the asymptotic behavior of the moment, truncation is not necessary since it will not affect the GPDs in the PDF-like region, and in the DA-like region, the difference remains negligible for the same reason. Therefore, we can consistently use the asymptotic moment (46) and disregard the polynomiality condition, as it will not result in any noticeable deviation. Besides, we have verified the consistency of the Mellin-Barnes integral formulation. Starting from the GPDs F(x,ξ,t)F(x,\xi,t) reconstructed from (46) and (30), we transform it again using the definition of the conformal moment (18). The result yields exactly the same conformal moment we used initially.

An approximation formula for F(x,ξ,t)F(x,\xi,t) when x>ξx>\xi

Using the asymptotic expansion of the hypergeometric function (38), we can further approximate (42) as:

F(x,ξ,0)≈−22C(z)Res⁡[x−j−1(21+1−z)jFj(ξ,t);∞](48)F(x,\xi,0) \approx-2\sqrt{2}C(z)\operatorname{Res}\left[x^{-j-1}\left(\frac{2}{1+\sqrt{1-z}}\right)^j F_j(\xi,t);\infty\right] \tag*{(48)}
=−Cp(z)Res⁡[xp(z)−(j+1)Fj(ξ,t);∞](49)=-C_p(z)\operatorname{Res}\left[x_p(z)^{-(j+1)}F_j(\xi,t);\infty\right] \tag*{(49)}

where xp(z)=1+1−z2x_p(z)=\frac{1+\sqrt{1-z}}{2}, Cp(z)=(21−z1+1−z)1/2C_p(z)=\left(\frac{2\sqrt{1-z}}{1+\sqrt{1-z}}\right)^{1/2}. We find that this is essentially another Mellin transformation of Fj(ξ,t)F_j(\xi,t), but transform Fj(ξ,t)F_j(\xi,t) from moment space to xp(z)x_p(z) space instead of the normal xx space. Therefore, employing the notation as (36), we have

F(x,ξ,0)=Cp(z){M−1Fj(ξ,t)}(xp(z))(50)F(x,\xi,0)=C_p(z)\left\{\mathcal{M}^{-1}F_j(\xi,t)\right\}(x_p(z)) \tag*{(50)}

in the case of ξ=0\xi=0, it reduces to the PDFs case (36). For the moment (46), we can derive the approximation formula for GPDs when x>ξx>\xi

F(x,ξ,0)=(21−ξ2/x21+1−ξ2/x2)1/2(1+1−ξ2/x21+1−ξ2)−α(1−1+1−ξ2/x21+1−ξ2)β,x>ξ(51)F(x,\xi,0)=\left(\frac{2\sqrt{1-\xi^2/x^2}}{1+\sqrt{1-\xi^2/x^2}}\right)^{1/2}\left(\frac{1+\sqrt{1-\xi^2/x^2}}{1+\sqrt{1-\xi^2}}\right)^{-\alpha}\left(1-\frac{1+\sqrt{1-\xi^2/x^2}}{1+\sqrt{1-\xi^2}}\right)^\beta,\quad x>\xi \tag*{(51)}

which reduces to x−α(1−x)βx^{-\alpha}(1-x)^\beta in the forward limit of ξ=0\xi=0. Again, we caution that no polynomiality condition has been imposed on the model moments. In Figure 5, we show the comparison between approximation formula and Mellin-Barnes integrals. The approximation formula is only valid when x>ξx>\xi and xx is not too close to ξ\xi. This limitation arises due to the steepest descent method used in [61]. When x∼ξx\sim\xi, the effect from the double saddle will invalidate the asymptotic expansion. Here, we only focus on analyzing the analytic properties of the Mellin-Barnes integral for x>ξx > \xi, corresponding to the PDF-like region. In the DA-like region, the situation becomes more intricate, and the Mellin-Barnes integral lacks a straightforward residue interpretation.

Image not extracted

Figure 5. A plot showing the comparison between the approximation formula (solid lines) for PDF-like GPDs (x>ξx > \xi), derived from the asymptotic expansion of hypergeometric functions in conformal partial waves, and the numerical reconstruction of GPDs using the Mellin-Barnes (MB) integral (dashed lines), plotted against the parton momentum fraction xx. The lines at x=ξx = \xi are indicated by dotted black lines. The plot shows that the approximation formula is quite accurate for values of xx that are not too close to ξ\xi, with the deviation becoming minimal as xx moves further away from ξ\xi.

Analytic continuation of summation indices and multiple Mellin-Barnes integrals

In this section, we first review the framework of double summations to calculate the amplitudes when considering the non-diagonal scale evolution operator contracting with the conformal moment or the Wilson coefficient at NLO in DVCS and DVMP. Then, we develop the technique of analytic continuation for summations indices. Finally, we convert the double summations into double Mellin-Barnes integrals to express the amplitudes, which will be beneficial for NLO calculations in future GUMP programs. This technique can be straightforwardly extended to the case of multiple summations and multiple Mellin-Barnes integrals.

Amplitudes as double summations for DVCS and DVMP

In the global analysis of GPDs, exclusive processes like DVCS and DVMP are crucial due to their direct correlation with the experimental quantities such as the Compton form factor and transition form factor. To achieve a universal moment parametrization as GUMP underlined, we mainly work in the conformal moment space. Schematically, we can write the amplitudes of DVCS [43] and DVMP [44] as

A∝∑i=q,qˉ,g∑j=0∞[1∓(−1)j]ξ−j−1Cji(Q)Fji(ξ,t,Q)(52)\mathcal{A} \propto\sum_{i=q,\bar q,g} \sum_{j=0}^{\infty} \left[1 \mp(-1)^j\right] \xi^{-j-1} C_j^i(Q) \mathcal{F}_j^i(\xi,t,Q) \tag*{(52)}

where ii is the flavor index, QQ is the hard scale, Cji(Q)C_j^i(Q) is the jj-th conformal moment of the Wilson coefficient, and the sign ∓\mp corresponds to vector GPDs and axial-vector GPDs, respectively. We have chosen both the renormalization and factorization scales to be QQ. When considering the evolution of the moment or the Wilson coefficient, there will be another summation for the conformal index. The evolved GPD moment or Wilson coefficient at scale QQ can be written in terms of the evolution operator Ejkij(ξ,Q,μ0)E_{jk}^{ij}(\xi,Q,\mu_0) and the unevolved ones at scale μ0\mu_0.

In the case of moment evolution,

Fji(ξ,t,μ)=∑i′=q,qˉ,g∑k=0j[1+(−1)k2]Ejkii′(ξ,Q,μ0)Fki′(ξ,t,μ0),(53)F_j^i(\xi,t,\mu)=\sum_{i'=q,\bar{q},g}\sum_{k=0}^{j}\left[\frac{1+(-1)^k}{2}\right]E_{jk}^{ii'}(\xi,Q,\mu_0)F_k^{i'}(\xi,t,\mu_0), \tag*{(53)}

where the evolution operator Ejkii′(ξ,Q,μ0)E_{jk}^{ii'}(\xi,Q,\mu_0) includes non-trivial mixing in both the flavor space and conformal moment space at NLO [43]. Inserting (53) into (52), we obtain the double summation of conformal moment space indices jj and kk. We neglect flavor mixing and focus solely on the summation in conformal moment spaces, which is the primary subject of our interest here. We take the example for vector GPDs

A∝∑j=0∞∑k=0j[1−(−1)j]ξ−j−1Cji(Q)[1−(−1)k2]Ejkii′(ξ,Q,μ0)Fki′(ξ,t,μ0),(54)\mathcal{A}\propto\sum_{j=0}^{\infty}\sum_{k=0}^{j}\left[1-(-1)^j\right]\xi^{-j-1}C_j^i(Q)\left[\frac{1-(-1)^k}{2}\right]E_{jk}^{ii'}(\xi,Q,\mu_0)F_k^{i'}(\xi,t,\mu_0), \tag*{(54)}
=12∑j=2∞[1−(−1)j]Cji(Q)∑k=1j−1[1−(−1)k]ξ−k−1Ejkii′(Q,μ0)Fki′(ξ,t,μ0).(55)=\frac{1}{2}\sum_{j=2}^{\infty}\left[1-(-1)^j\right]C_j^i(Q)\sum_{k=1}^{j-1}\left[1-(-1)^k\right]\xi^{-k-1}E_{jk}^{ii'}(Q,\mu_0)F_k^{i'}(\xi,t,\mu_0). \tag*{(55)}

In the second equality, we sum kk from 1 to j−1j-1 to avoid the pole at k=jk=j from the Regge term of the GPD moment and k=0k=0 from anomalous dimension in the NLO evolution operator [43]. Besides, we have factorized out the ξ\xi dependence in the evolution kernel, which is valid at least to the NLO,

Ejkii′(ξ,Q,μ0)=ξj−kEjkii′(Q,μ0).(56)E_{jk}^{ii'}(\xi,Q,\mu_0)=\xi^{j-k}E_{jk}^{ii'}(Q,\mu_0). \tag*{(56)}

Alternatively, we can change the order of summation and relabel the indices as:

A∝∑j=0∞∑k=j∞[1−(−1)k]ξ−k−1Cki(Q)[1−(−1)j2]Ekjii′(ξ,Q,μ0)Fji′(ξ,t,μ0),(57)\mathcal{A}\propto\sum_{j=0}^{\infty}\sum_{k=j}^{\infty}\left[1-(-1)^k\right]\xi^{-k-1}C_k^i(Q)\left[\frac{1-(-1)^j}{2}\right]E_{kj}^{ii'}(\xi,Q,\mu_0)F_j^{i'}(\xi,t,\mu_0), \tag*{(57)}
=12∑j=1∞[1−(−1)j]ξ−j−1Fji′(ξ,t,μ0)∑k=j+1∞[1−(−1)k]Cki(Q)Ekjii′(Q,μ0),(58)=\frac{1}{2}\sum_{j=1}^{\infty}\left[1-(-1)^j\right]\xi^{-j-1}F_j^{i'}(\xi,t,\mu_0)\sum_{k=j+1}^{\infty}\left[1-(-1)^k\right]C_k^i(Q)E_{kj}^{ii'}(Q,\mu_0), \tag*{(58)}

which physically corresponds to the evolution of Wilson coefficient physically. In principle, these two summations should produce identical results after resummation. A proper resummation technique is needed to first resum the inner summation with variable bounds, then the outer infinite summation. For the single summation as (52) without considering evolution in conformal moment space, it is straightforward to transform into the Mellin-Barnes integral as shown in ref.[43, 66]. However, for the case of double summations, the transformation to the double Mellin-Barnes integrals remains unclear in the literature.

Analytic continuation technique and double Mellin-Barnes integrals

In this scenario, we cannot simply regard the bounds of summation as integers. Instead, for the first part, we must determine the analytic continuation for the summation indices in (55) and (58). These sums are then converted into Mellin-Barnes integrals using the Sommerfeld-Watson transformation. For instance, we need to generalize the inner summation of equation (77) as follows:

FG(j):=∑→k=1j−1[1−(−1)k]ξ−k−1Ejkii′Fki′,(59)F_G(j):=\overrightarrow{\sum}_{k=1}^{j-1}\left[1-(-1)^k\right]\xi^{-k-1}E_{jk}^{ii'}F_k^{i'}, \tag*{(59)}

where ∑→\overrightarrow{\sum} is an formal notation indicating the analytic continuation of summation indices, so that jj can take complex values. This requires finding an analytic function whose values at integer jj match the finite summation. For notation simplicity, we have abbreviated Ejkii′E_{jk}^{ii'} for Ejkii′(Q,μ0)E_{jk}^{ii'}(Q,\mu_0), Fki′F_k^{i'} for Fki′(ξ,t,μ0)F_k^{i'}(\xi,t,\mu_0), and CjiC_j^i for Cji(Q)C_j^i(Q). This generalization is thoroughly studied in the monograph [67]. That is,

∑→k=1j−1[1−(−1)k]ξ−k−1Ejkii′Fki′,(60)\overrightarrow{\sum}_{k=1}^{j-1}\left[1-(-1)^k\right]\xi^{-k-1}E_{jk}^{ii'}F_k^{i'}, \tag*{(60)}
∑k=0∞{[1+(−1)k]ξk−2Ej,k+1ii′Fk+1i′−[1−(−1)k+j]ξ−k−j−1Ej,k+jii′Fk+ji′},(61)\sum_{k=0}^{\infty}\left\{\left[1+(-1)^k\right]\xi^{k-2}E^{ii'}_{j,k+1}\mathcal{F}^{i'}_{k+1}-\left[1-(-1)^{k+j}\right]\xi^{-k-j-1}E^{ii'}_{j,k+j}\mathcal{F}^{i'}_{k+j}\right\}, \tag*{(61)}
−12i∫ck−i∞ck+i∞dk{cot⁡(πk2)ξk−2Ej,k+1ii′Fk+1i′−[cot⁡(πk)−1sin⁡(πk)(−1)j]ξ−k−j−1Ej,k+jii′Fk+ji′},(62)-\frac{1}{2i}\int_{c_k-i\infty}^{c_k+i\infty}dk\left\{\cot\left(\frac{\pi k}{2}\right)\xi^{k-2}E^{ii'}_{j,k+1}\mathcal{F}^{i'}_{k+1}-\left[\cot(\pi k)-\frac{1}{\sin(\pi k)}(-1)^j\right]\xi^{-k-j-1}E^{ii'}_{j,k+j}\mathcal{F}^{i'}_{k+j}\right\}, \tag*{(62)}

analogously, for the infinite inner summation of (58), we have

∑k=j+1∞[1−(−1)k]CkiEkjii′,(63)\sum_{k=j+1}^{\infty}\left[1-(-1)^k\right]C^i_kE^{ii'}_{kj}, \tag*{(63)}
=∑k=0∞[1+(−1)k(−1)j]Ck+j+1iEk+j+1,jii′,(64)=\sum_{k=0}^{\infty}\left[1+(-1)^k(-1)^j\right]C^i_{k+j+1}E^{ii'}_{k+j+1,j}, \tag*{(64)}
=−12i∫ck−i∞ck+i∞dk[cot⁡(πk)+1sin⁡(πk)(−1)j]Ck+j+1iEk+j+1,jii′.(65)=-\frac{1}{2i}\int_{c_k-i\infty}^{c_k+i\infty}dk\left[\cot(\pi k)+\frac{1}{\sin(\pi k)}(-1)^j\right]C^i_{k+j+1}E^{ii'}_{k+j+1,j}. \tag*{(65)}

The second part should then be straightforward, as we have extended the first summation as an ordinary analytic function. We have for (55) the moment evolution case

12∑j=2∞[1−(−1)j]Cji∑k=1j−1[1−(−1)k]ξ−k−1Ejkii′Fki′,(66)\frac{1}{2}\sum_{j=2}^{\infty}\left[1-(-1)^j\right]C^i_j\sum_{k=1}^{j-1}\left[1-(-1)^k\right]\xi^{-k-1}E^{ii'}_{jk}\mathcal{F}^{i'}_k, \tag*{(66)}
=12∑j=2∞[1−(−1)j]Cji∑k=1j−1[1−(−1)k]ξ−k−1Ejkii′Fki′,(67)=\frac{1}{2}\sum_{j=2}^{\infty}\left[1-(-1)^j\right]C^i_j\sum_{k=1}^{j-1}\left[1-(-1)^k\right]\xi^{-k-1}E^{ii'}_{jk}\mathcal{F}^{i'}_k, \tag*{(67)}
=−14i∑j=2∞[1−(−1)j]Cji∫ck−i∞ck+i∞dk{cot⁡(πk2)ξk−2Ej,k+1ii′Fk+1i′−[cot⁡(πk)−1sin⁡(πk)(−1)j]ξ−k−j−1Ej,k+jii′Fk+ji′},(68)=-\frac{1}{4i}\sum_{j=2}^{\infty}\left[1-(-1)^j\right]C^i_j\int_{c_k-i\infty}^{c_k+i\infty}dk\left\{\cot\left(\frac{\pi k}{2}\right)\xi^{k-2}E^{ii'}_{j,k+1}\mathcal{F}^{i'}_{k+1}-\left[\cot(\pi k)-\frac{1}{\sin(\pi k)}(-1)^j\right]\xi^{-k-j-1}E^{ii'}_{j,k+j}\mathcal{F}^{i'}_{k+j}\right\}, \tag*{(68)}
=−14i∑j=2∞[1−(−1)j]Cji∫ck−i∞ck+i∞dk{cot⁡(πk2)ξk−2Ej,k+1ii′Fk+1i′−[cot⁡(πk)+1sin⁡(πk)]ξ−k−j−1Ej,k+jii′Fk+ji′},(69)=-\frac{1}{4i}\sum_{j=2}^{\infty}\left[1-(-1)^j\right]C^i_j\int_{c_k-i\infty}^{c_k+i\infty}dk\left\{\cot\left(\frac{\pi k}{2}\right)\xi^{k-2}E^{ii'}_{j,k+1}\mathcal{F}^{i'}_{k+1}-\left[\cot(\pi k)+\frac{1}{\sin(\pi k)}\right]\xi^{-k-j-1}E^{ii'}_{j,k+j}\mathcal{F}^{i'}_{k+j}\right\}, \tag*{(69)}
=18∫cj−i∞cj+i∞djtan⁡(πj2)Cj+2i∫ck−i∞ck+i∞dkcot⁡(πk2)(ξ−k−2Ej+2,k+1ii′Fk+1i′−ξ−k−j−3Ej+2,k+j+2ii′Fk+j+2i′),(70)=\frac{1}{8}\int_{c_j-i\infty}^{c_j+i\infty}dj\tan\left(\frac{\pi j}{2}\right)C^i_{j+2}\int_{c_k-i\infty}^{c_k+i\infty}dk\cot\left(\frac{\pi k}{2}\right)\left(\xi^{-k-2}E^{ii'}_{j+2,k+1}\mathcal{F}^{i'}_{k+1}-\xi^{-k-j-3}E^{ii'}_{j+2,k+j+2}\mathcal{F}^{i'}_{k+j+2}\right), \tag*{(70)}

similarly for (58) the coefficient evolution case, we can derive

12∑j=1∞[1−(−1)j]ξ−j−1Fji′∑k=j+1∞[1−(−1)k]CkiEkjii′,(71)\frac{1}{2}\sum_{j=1}^{\infty}\left[1-(-1)^j\right]\xi^{-j-1}\mathcal{F}^{i'}_j\sum_{k=j+1}^{\infty}\left[1-(-1)^k\right]C^i_kE^{ii'}_{kj}, \tag*{(71)}
=12∑j=1∞[1−(−1)j]ξ−j−1Fji′∑k=j+1∞[1−(−1)k]CkiEkjii′,(72)=\frac{1}{2}\sum_{j=1}^{\infty}\left[1-(-1)^j\right]\xi^{-j-1}\mathcal{F}^{i'}_j\sum_{k=j+1}^{\infty}\left[1-(-1)^k\right]C^i_kE^{ii'}_{kj}, \tag*{(72)}
=−14i∑j=1∞[1−(−1)j]ξ−j−1Fji′∫ck−i∞ck+i∞dk[cot⁡(πk)+1sin⁡(πk)(−1)j]Ck+j+1iEk+j+1,jii′,(73)=-\frac{1}{4i}\sum_{j=1}^{\infty}\left[1-(-1)^j\right]\xi^{-j-1}\mathcal{F}^{i'}_j\int_{c_k-i\infty}^{c_k+i\infty}dk\left[\cot(\pi k)+\frac{1}{\sin(\pi k)}(-1)^j\right]C^i_{k+j+1}E^{ii'}_{k+j+1,j}, \tag*{(73)}
=−14i∑j=1∞[1−(−1)j]ξ−j−1Fji′∫ck−i∞ck+i∞dk[cot⁡(πk)−1sin⁡(πk)]Ck+j+1iEk+j+1,jii′,(74)=-\frac{1}{4i}\sum_{j=1}^{\infty}\left[1-(-1)^j\right]\xi^{-j-1}\mathcal{F}^{i'}_j\int_{c_k-i\infty}^{c_k+i\infty}dk\left[\cot(\pi k)-\frac{1}{\sin(\pi k)}\right]C^i_{k+j+1}E^{ii'}_{k+j+1,j}, \tag*{(74)}
=14i∑j=0∞[1+(−1)j]ξ−j−2Fj+1i′∫ck−i∞ck+i∞dktan⁡(πj2)Ck+j+2iEk+j+2,j+1ii′,(75)=\frac{1}{4i}\sum_{j=0}^{\infty}\left[1+(-1)^j\right]\xi^{-j-2}\mathcal{F}^{i'}_{j+1}\int_{c_k-i\infty}^{c_k+i\infty}dk\tan\left(\frac{\pi j}{2}\right)C^i_{k+j+2}E^{ii'}_{k+j+2,j+1}, \tag*{(75)}
=18∫cj−i∞cj+i∞djcot⁡(πj2)ξ−j−2Fj+1i′∫ck−i∞ck+i∞dktan⁡(πk2)Ck+j+2iEk+j+2,j+1ii′.(76)=\frac{1}{8}\int_{c_j-i\infty}^{c_j+i\infty}dj\cot\left(\frac{\pi j}{2}\right)\xi^{-j-2}\mathcal{F}^{i'}_{j+1}\int_{c_k-i\infty}^{c_k+i\infty}dk\tan\left(\frac{\pi k}{2}\right)C^i_{k+j+2}E^{ii'}_{k+j+2,j+1}. \tag*{(76)}

To summarize, at the end of the day, we have for the vector GPDs

AV∝18∫cj−i∞cj+i∞dj[i+tan⁡(πj2)]Cj+2i∫ck−i∞ck+i∞dkcot⁡(πk2)(ξ−k−2Ej+2,k+1ii′Fk+1i′−ξ−k−j−3Ej+2,k+j+2ii′Fk+j+2i′)(77)\mathcal{A}_{V} \propto\frac{1}{8}\int_{c_j-i\infty}^{c_j+i\infty} \mathrm{d}j\left[i+\tan\left(\frac{\pi j}{2}\right)\right]\mathcal{C}^{i}_{j+2}\int_{c_k-i\infty}^{c_k+i\infty}\mathrm{d}k\cot\left(\frac{\pi k}{2}\right)\left(\xi^{-k-2}\mathcal{E}^{ii'}_{j+2,k+1}\mathcal{F}^{i'}_{k+1}-\xi^{-k-j-3}\mathcal{E}^{ii'}_{j+2,k+j+2}\mathcal{F}^{i'}_{k+j+2}\right) \tag*{(77)}

as well as

AV∝−18∫cj−i∞cj+i∞dj[i−cot⁡(πj2)]ξ−j−2Fj+1i′∫ck−i∞ck+i∞dktan⁡(πk2)Ck+j+2iEk+j+2,j+1ii′(78)\mathcal{A}_{V} \propto-\frac{1}{8}\int_{c_j-i\infty}^{c_j+i\infty}\mathrm{d}j\left[i-\cot\left(\frac{\pi j}{2}\right)\right]\xi^{-j-2}\mathcal{F}^{i'}_{j+1}\int_{c_k-i\infty}^{c_k+i\infty}\mathrm{d}k\tan\left(\frac{\pi k}{2}\right)\mathcal{C}^{i}_{k+j+2}\mathcal{E}^{ii'}_{k+j+2,j+1} \tag*{(78)}

where cjc_j and ckc_k are both chosen to be in the range (−1,0)(-1,0) and to the right of all the poles of Fj/ki′\mathcal{F}^{i'}_{j/k}. Note that we have manually added an extra ii to the jj integral in accordance with [43, 66] to account for the imaginary part of the amplitude. For axial-vector GPDs, the treatment is alike

AP∝−18∫cj−i∞cj+i∞dj[i−cot⁡(πj2)]Cj+2i∫ck−i∞ck+i∞dk{tan⁡(πk2)ξ−k−2Ej+2,k+1ii′Fk+1i′+cot⁡(πk2)ξ−k−j−3Ej+2,k+j+2ii′Fk+j+2i′}(79)\mathcal{A}_{P} \propto-\frac{1}{8}\int_{c_j-i\infty}^{c_j+i\infty}\mathrm{d}j\left[i-\cot\left(\frac{\pi j}{2}\right)\right]\mathcal{C}^{i}_{j+2}\int_{c_k-i\infty}^{c_k+i\infty}\mathrm{d}k\left\{\tan\left(\frac{\pi k}{2}\right)\xi^{-k-2}\mathcal{E}^{ii'}_{j+2,k+1}\mathcal{F}^{i'}_{k+1}+\cot\left(\frac{\pi k}{2}\right)\xi^{-k-j-3}\mathcal{E}^{ii'}_{j+2,k+j+2}\mathcal{F}^{i'}_{k+j+2}\right\} \tag*{(79)}

and

AP∝−18∫cj−i∞cj+i∞dj[i+tan⁡(πj2)]ξ−j−2Fj+1i′∫ck−i∞ck+i∞dktan⁡(πk2)Ck+j+2iEk+j+2,j+1ii′.(80)\mathcal{A}_{P} \propto-\frac{1}{8}\int_{c_j-i\infty}^{c_j+i\infty}\mathrm{d}j\left[i+\tan\left(\frac{\pi j}{2}\right)\right]\xi^{-j-2}\mathcal{F}^{i'}_{j+1}\int_{c_k-i\infty}^{c_k+i\infty}\mathrm{d}k\tan\left(\frac{\pi k}{2}\right)\mathcal{C}^{i}_{k+j+2}\mathcal{E}^{ii'}_{k+j+2,j+1}. \tag*{(80)}

We have tested that, providing numerical implementation ensuring proper evaluation and reliable convergence, these two double Mellin-Barnes integrals for the evolution of either the conformal moment or the Wilson coefficient both yield the same result. These double integrals can be utilized in future GUMP programs.

For the case of multiple summations, this approach remains available. Similar methods of analytically continuing the summation indices and applying the Sommerfeld-Watson transformation can effectively convert these multiple summations into multiple Mellin-Barnes integrals.

Conclusion

In this work, we investigate the convergence properties of the conformal moment expansion of GPDs useful for their global analysis from various constraints, elucidating the relationship between formal summation and the Mellin-Barnes integral, which had previously been suggested in earlier GUMP studies [39, 40]. Furthermore, we examine the asymptotic properties of the Mellin-Barnes integral and derive asymptotic conditions on the conformal moment Fj\mathcal{F}_j that had not been revealed in prior research. A limitation of this formulation is that the derived asymptotic conditions appear to contradict the polynomiality condition of conformal moments which can be remedied easily.

Furthermore, we propose an approximate formula for the GPDs F(x,ξ,t)F(x,\xi,t) for x>ξx>\xi. Note that xx cannot be too close to ξ\xi due to the limitations of the steepest descent method used to find the asymptotic expansion of hypergeometric functions. Identifying the uniform asymptotic expansion of this hypergeometric function would enable us to overcome this limitation, although achieving this would necessitate considerable mathematical rigor.

Finally, we introduce a method to effectively manage double summations arising from the NLO non-diagonal evolution kernel in DVCS and DVMP. By converting these summations into multiple Mellin-Barnes integrals through the analytic continuation of summations with variable bounds, we establish a mathematical framework. This technique is anticipated to be beneficial for future research within the GUMP program, as it provides a foundation for accurately calculating amplitudes at NLO, which is crucial for comprehensive global analysis.

Acknowledgments

We thank Yuxun Guo, M. Gabriel Santiago and Yushan Su for useful discussions. XJ is partially supported by Maryland Center for Fundamental Physics (MCFP), and HCZ acknowledges the hospitality and support of MCFP and U. Maryland for the summer intern program.

Sommerfeld-Watson transformation

The Sommerfeld-Watson transformation is a technique for converting series into complex integrals, based on the residue theorem, proposed in [68, 69]. For our usage, we have the following formulas:

∑k=0∞(−1)kfk=−12i∫c−i∞c+i∞1sin⁡(πs)fs ds,(81)\sum_{k=0}^{\infty}(-1)^k f_k=-\frac{1}{2i}\int_{c-i\infty}^{c+i\infty}\frac{1}{\sin(\pi s)}f_s\,ds, \tag*{(81)}

and

∑k=0∞fk=−12i∫c−i∞c+i∞cot⁡(πs)fs ds,(82)\sum_{k=0}^{\infty}f_k=-\frac{1}{2i}\int_{c-i\infty}^{c+i\infty}\cot(\pi s)f_s\,ds, \tag*{(82)}

where fsf_s is the analytic continuation of fkf_k, having no poles on the right half-plane. Here, cc is chosen such that −1<c<0-1<c<0 and to be to the right of all the poles of fsf_s. Furthermore, we have

∑k=0∞[1−(−1)k]fk=12i∫c−i∞c+i∞tan⁡(πs2)fs ds,(83)\sum_{k=0}^{\infty}\left[1-(-1)^k\right]f_k=\frac{1}{2i}\int_{c-i\infty}^{c+i\infty}\tan\left(\frac{\pi s}{2}\right)f_s\,ds, \tag*{(83)}

and

∑k=0∞[1+(−1)k]fk=−12i∫c−i∞c+i∞cot⁡(πs2)fs ds.(84)\sum_{k=0}^{\infty}\left[1+(-1)^k\right]f_k=-\frac{1}{2i}\int_{c-i\infty}^{c+i\infty}\cot\left(\frac{\pi s}{2}\right)f_s\,ds. \tag*{(84)}

References

References

  1. [24]Jefferson Lab Hall A collaboration, Deeply virtual compton scattering cross section at high bjorken xB, 2201.03714.
  2. [25]A. Accardi, J.L. Albacete, M. Anselmino, N. Armesto, E.C. Aschenauer, A. Bacchetta et al., Electron Ion Collider: The Next QCD Frontier - Understanding the glue that binds us all, 1212.1701.arxiv.org/abs/1212.1701
  3. [26]R. Abdul Khalek et al., Science Requirements and Detector Concepts for the Electron-Ion Collider: EIC Yellow Report, Nucl. Phys. A 1026 (2022) 122447 [2103.05419].
  4. [27]X. Ji, Deeply virtual Compton scattering, hep-ph/9609381.
  5. [28]A.V. Radyushkin, Asymmetric gluon distributions and hard diffractive electroproduction, hep-ph/9605431.DOI
  6. [29]X.-D. Ji and J. Osborne, One loop corrections and all order factorization in deeply virtual Compton scattering, Phys. Rev. D 58 (1998) 094018 [hep-ph/9801260].DOI
  7. [30]J.C. Collins and A. Freund, Proof of factorization for deeply virtual Compton scattering in QCD, Phys. Rev. D 59 (1999) 074009 [hep-ph/9801262].DOI
  8. [31]J.C. Collins, L. Frankfurt and M. Strikman, Factorization for hard exclusive electroproduction of mesons in QCD, hep-ph/9611433.DOI
  9. [32]X. Ji, Parton physics on a euclidean lattice, 1305.1539.pubmed.ncbi.nlm.nih.gov/23848864
  10. [33]X. Ji, Parton Physics from Large-Momentum Effective Field Theory, Sci. China Phys. Mech. Astron. 57 (2014) 1407 [1404.6680].arxiv.org/abs/1404.6680
  11. [34]X. Ji, Y.-S. Liu, Y. Liu, J.-H. Zhang and Y. Zhao, Large-momentum effective theory, 2004.03543.
  12. [35]C. Alexandrou, S. Bacchio, M. Constantinou, J. Finkenrath, K. Hadjiyiannakou, K. Jansen et al., Nucleon form factors from Nf=2+1+1 twisted mass QCD at the physical point, 2112.06750.
  13. [36]M. Constantinou and Others, Parton distributions and lattice-QCD calculations: Toward 3D structure, 2006.0836.
  14. [37]H.-W. Lin, Nucleon helicity generalized parton distribution at physical pion mass from lattice QCD, 2112.07519.DOI
  15. [38]H.-W. Lin, Nucleon tomography and generalized parton distribution at physical pion mass from lattice QCD, 2008.12474.
  16. [39]Y. Guo, X. Ji and K. Shiels, Generalized parton distributions through universal moment parameterization: Zero skewness case, 2307.05768.
  17. [40]Y. Guo, X. Ji, M.G. Santiago, K. Shiells and J. Yang, Generalized parton distributions through universal moment parameterization: Non-zero skewness case, 2302.07279.
  18. [41]D. Mueller and A. Schafer, Complex conformal spin partial wave expansion of generalized parton distributions and distribution amplitudes, hep-ph/0509204.
  19. [42]D. Müller, M.V. Polyakov and K.M. Semenov-Tian-Shansky, Dual parametrization of generalized parton distributions in two equivalent representations, 1412.4165.
  20. [43]K. Kumericki, D. Mueller and K. Passek-Kumericki, Towards a fitting procedure for deeply virtual Compton scattering at next-to-leading order and beyond, hep-ph/0703179.DOI
  21. [44]D. Müller, T. Lautenschlager, K. Passek-Kumericki and A. Schaefer, Towards a fitting procedure to deeply virtual meson production - the next-to-leading order case, Nucl. Phys. B 884 (2014) 438 [1310.5394].DOI
  22. [45]B.E. White, Factorisation in Deeply Virtual Compton Scattering: Local OPE Formalism and Structure Functions, hep-ph/0102121.
  23. [46]J.C. Collins and D.E. Soper, Parton Distribution and Decay Functions, Nucl. Phys. B 194 (1982) 445.DOI
  24. [47]K.G. Wilson and W. Zimmermann, Operator product expansions and composite field operators in the general framework of quantum field theory, Commun. Math. Phys. 24 (1972) 87.DOI
  25. [48]G. Altarelli and G. Parisi, Asymptotic Freedom in Parton Language, Nucl. Phys. B 126 (1977) 298.DOI
  26. [49]Y.L. Dokshitzer, Calculation of the Structure Functions for Deep Inelastic Scattering and e+ e- Annihilation by Perturbation Theory in Quantum Chromodynamics, Sov. Phys. JETP 46 (1977) 641.
  27. [50]V.N. Gribov and L.N. Lipatov, Deep inelastic e p scattering in perturbation theory, Sov. J. Nucl. Phys. 15 (1972) 438.
  28. [51]D. Graudenz, M. Hampel, A. Vogt and C. Berger, The mellin transform technique for the extraction of the gluon density, hep-ph/9506333.DOI
  29. [52]P. Stratmann and W. Vogelsang, Towards a global analysis of polarized parton distributions, hep-ph/0107064.
  30. [53]Y. Tsuno, Analytic functionals and distributions, Hiroshima Mathematical Journal 34 (1970) .DOI
  31. [54]A.V. Belitsky, B. Geyer, D. Mueller and A. Schafer, On the leading logarithmic evolution of the off forward distributions, Phys. Lett. B 421 (1998) 312 [hep-ph/9710427].DOI
  32. [55]L. Mankiewicz, G. Piller and T. Weigl, Hard exclusive meson production and nonforward parton distributions, Eur. Phys. J. C 5 (1998) 119 [hep-ph/9711227].DOI
  33. [56]A. Manashov, M. Kirch and A. Schafer, Solving the leading order evolution equation for GPDs, Phys. Rev. Lett. 95 (2005) 012002 [hep-ph/0503109].DOI
  34. [57]A. Shuvaev, Solution of the off forward leading logarithmic evolution equation based on the Gegenbauer moments inversion, Phys. Rev. D 60 (1999) 116005 [hep-ph/9902318].DOI
  35. [58]M.V. Polyakov and A.G. Shuvaev, On dual parametrizations of generalized parton distributions, hep-ph/0207153.
  36. [59]L. Schläfli, Über die zwei Heine’schen Kugelfunktionen mit beliebigem Parameter und ihre ausnahmslose Darstellung durch bestimmte Integrale (1881).
  37. [60]M.J. Ablowitz and A.S. Fokas, Complex Variables: Introduction and Applications, Cambridge Texts in Applied Mathematics, Cambridge University Press, 2nd ed ed. (2003).
  38. [61]R.B. Paris, Asymptotics of the gauss hypergeometric function with large parameters, II, Journal of Classical Analysis (2013) 1.DOI
  39. [62]K. Kumerički and D. Mueller, Deeply virtual Compton scattering at small xB and the access to the GPD H, 0904.0458.
  40. [63]T. Regge, Introduction to complex orbital momenta, Nuovo Cim. 14 (1959) 951.DOI
  41. [64]T. Regge, Bound states, shadow states and Mandelstam representation, Nuovo Cim. 18 (1960) 947.DOI
  42. [65]T.-J. Hou and Others, New CTEQ global analysis of quantum chromodynamics with high-precision data from the LHC, 1912.10053.DOI
  43. [66]M. Č. G. Duplančić, K. Kumerički and K. Passek-K, NLO corrections to the deeply virtual meson production revisited: Impact on the extraction of generalized parton distributions, 2310.13837.arxiv.org/abs/2310.13837
  44. [67]I.M. Alabdulmohsin, Summability Calculus, Springer International Publishing (2018), 10.1007/978-3-319-74648-7.
  45. [68]A. Sommerfeld and E.G. Straus, Partial Differential Equations in Physics, no. 6 in Lectures in Theoretical Physics, Academic Press (1964).
  46. [69]G.N. Watson, The diffraction of electric waves by the earth, Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character 95 (1918) 83.DOI

Paper details

vlnc:W_tjw2s4py
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents