Vector meson electroproduction at small Bjorken-x and generalized parton distributions

Bottom line: this is the paper that made the gluonic handbag approach quantitatively usable against HERA vector meson data, and the longitudinal channel is a genuine prediction. The transverse sector is normalized by fitting to the same data, and the paper says so only in passing. Worth a serious referee, not a desk reject either way. One correction to the record: the classifier says the authors state no limitations. That is wrong. Sects. 6 and 9 are explicit about the hand-inserted t-dependence, the neglected ~H^g and E_g, the dropped D-term, and the approximation of GPD evolution by gluon-distribution evolution. They are candid about what they did not compute. What is new: the complete set of gamma*g -> V g helicity amplitudes with quark transverse momenta retained and Sudakov suppression, the resulting spin density matrix elements for rho and phi, and the A_LL prediction. The load-bearing claim is that the collinear longitudinal result of Eq. (92) overpredicts HERA sigma_L by a large factor and that k_perp plus Sudakov brings it down to the data. That is a real result, and it is the reason people cite this paper. The soft spots, in proportion. First, the transverse normalization: f_VT, a_VT and B_VTT are all adjusted to the HERA data being described, and the paper itself notes the transverse cross section is controlled by the product (f_VT/M_V)²/B_VTT. So the T -> T agreement is partly true by construction. That is a real limitation, but it is stated, and the authors do not oversell the transverse channel. Second, the t-dependence is inserted as exponentials with slopes taken from data, not computed. Acknowledged, and honestly so. Third, the stress-test point about a_VL: the widths in Eq. (51) are fixed by an assumed Fock-state probability of 0.25, not by data, and the suppression that reconciles sigma_L with experiment depends on them. The paper never varies a_VL. I would not call this fatal, because the paper is demonstrating a mechanism rather than claiming precision, and it already quotes a 30% PDF uncertainty. But a one-line sensitivity check on a_VL would have cost nothing and would have made the central claim much harder to attack. Who should read it: anyone working on exclusive diffraction, GPD phenomenology, or interpreting r⁰_04 and the other spin density matrix elements at HERA and COMPASS. The method is specified in enough detail to reimplement, though the loop integrals and Sudakov implementation cannot be checked from the text alone. Recommendation: send it out. It is a real contribution with acknowledged limitations, and the referees should press on the a_VL sensitivity and the transverse fit.

The central quantitative claim is that the full handbag amplitude (Eq. 57) brings the collinear longitudinal cross section of Eqs. (92)/(94) — which overpredicts the HERA σ_L by a factor of roughly 3–5 at Q² ≈ 4–10 GeV² — down to the data, and that this is a genuine consequence of retaining quark transverse momentum and Sudakov suppression. The size of that suppression is controlled by the b-space width of the meson wavefunction. In Eq. (51) the widths a_ρL = 0.52 GeV⁻¹ and a_φL = 0.45 GeV⁻¹ are fixed not by data but by the assumption of 'equal probabilities for the vector meson and pion valence Fock states, namely 0.25.' The b-integrand of Eq. (57) is the product of the Gaussian Ψ̂(b) ∝ exp(−ττ̄b²/(4a_V²)), the propagator K₀(√d bQ), and exp[−S]. At Q² = 4–10 GeV² the perturbative cutoff 1/Q is comparable to a_V, so the Gaussian width materially controls how much weight sits at large b where the Sudakov factor bites; a 20% change in a_VL plausibly shifts σ_L by 30–50%, comparable to or larger than the 30% PDF error the paper does quote. The paper demonstrates agreement at one point in this parameter space and never varies a_VL, nor the Sudakov infrared prescriptions (exp[−S]=1 if ≥1, freeze at b = 1/Λ_QCD, s = 0 for b ≤ √2/(βQ)). The claim that the approach 'resolves' the normalization problem is therefore not shown to be robust; it could be a tuned cancellation. This is distinct from the reader's concern: it concerns the longitudinal channel, which the reader treats as the genuine prediction, not the fitted transverse sector.

The paper is a careful extension of the GPD handbag program, and its longitudinal-channel result is the part that is not fitted to the same data: f_VL comes from measured decay constants, the GPD from CTEQ5M, and the Sudakov factor from the published Li–Sterman/Dahm–Jakob–Kroll formalism. That is why the reader's verdict of CONDITIONAL rather than ACCEPT is right. The reader's weakest_assumption correctly identifies unquantified model dependence, but locates it in the transverse sector (fitted f_VT, a_VT, B_VTT) and in the hand-inserted t-exponentials of Eqs. (79)/(84). I agree those are real and acknowledged in Sects. 6 and 9. The concern I raise is in the longitudinal sector, which the reader treats as the genuine prediction: the suppression factor that reconciles Eq. (92) with data is set by a_VL, fixed by an untested Fock-probability assumption and never varied, and the Sudakov infrared treatment is likewise not probed. This is a correctness-risk concern (the agreement may be tuned), not an internal inconsistency, and it does not overturn the paper — 'fair agreement' with a factor-level theory uncertainty could survive a ±20% variation in a_VL. But the paper's error budget quotes only the ~30% PDF uncertainty, so the robustness of the headline resolution is asserted rather than demonstrated. The concrete test above would settle it. Because the concern sharpens rather than reverses the reader's CONDITIONAL verdict, I recommend UNCHANGED, with agreement_with_reader = partial: same category of unquantified model dependence, different sector.

The paper is a careful and, for its time, genuinely new attempt to confront the GPD-based handbag approach with the full body of small-$x$ vector-meson electroproduction data rather than with the longitudinal cross section alone. Its real strength is the longitudinal channel: $f_{VL}$ is taken from decay constants, $a_{VL}$ is fixed by an independent Fock-state argument, and $H^g$ is tied to the CTEQ gluon density, so the agreement in Figs. 5 and 6 is a meaningful test of the $k_perp$-plus-Sudakov mechanism and of the skewing of $H^g(xi,xi)$ relative to $2xi g(2xi)$. The explicit comparison with the collinear result of Eq. (92) and with the $ln(1/x)$ approximation (Fig. 10) is a useful, falsifiable statement about where the approximations diverge. The paper is also commendably explicit about its own model dependence, both for the infrared regularization of the $T to T$ amplitude and for the $t$-dependence, which it inserts as exponentials with slopes taken from data. The transverse sector is weaker as evidence, since its normalization is fixed by parameters fitted to the same data, and the paper says so in Sect. 6; the abstract and conclusions should carry that distinction more leg

The paper asks whether the handbag picture of diffractive vector-meson electroproduction — a hard subprocess $gamma^* g to V g$ glued to a gluonic generalized parton distribution (GPD) of the proton — can be made to give the right *numbers*, not just the right $Q^2$ scaling. The authors show that the strict collinear leading-twist version badly overpredicts the longitudinal $rho$ and $phi$ cross sections measured at HERA, and that keeping the transverse momenta of the quark and antiquark forming the meson, together with a Sudakov factor that damps large quark-antiquark separations, brings the normalization down to the data for $Q^2$ above about 4 GeV$^2$. The same machinery yields all helicity amplitudes, so the authors can also compute the spin density matrix elements of the produced mesons, the longitudinal-to-transverse ratio $R$, the $W$ dependence $sigma_L propto W^{4delta(Q^2)}$, and the small double-spin asymmetry $A_{LL}$. What makes the result worth attention is that a single soft input — the gluon GPD, modelled from the ordinary gluon distribution through double distributions — plus a perturbatively calculated subprocess accounts simultaneously for cross sections and for the angular decay distributions that probe the relative sizes and phases of the three retained helicity amplitudes.

What carries the argument. The load-bearing object is the gluonic generalized parton distribution $H^g(bar x, xi, t)$, which encodes the proton's response to emitting and reabsorbing a gluon, combined with a $k_perp$-dependent subprocess amplitude in impact-parameter space. The GPD is not calculated: it is generated from a double distribution $f(beta, alpha)$ proportional to the ordinary gluon distribution, with the skewing effect that makes $H^g(xi,xi)$ differ from $xi g(xi)$ by a factor growing to about 1.6–2 at $xi = 0.1$. The Sudakov factor is what does the quantitative work of suppressing the end-point regions where one constituent carries almost none of the meson momentum; without it the leading-twist longitudinal cross section is far too large. The amplitude hierarchy $H^V_{0lambda,0lambda} propto 1$, $H^V_{0lambda,+lambda} propto sqrt{-t}/Q$, $H^V_{+lambda,+lambda} propto langle k_perp^2 rangle^{1/2}/Q$ then propagates into every observable the paper computes.

Core claim. The central claim is that the gluonic handbag approach, evaluated beyond the collinear approximation, describes light vector-meson electroproduction at small Bjorken-$x$ quantitatively. Concretely: the subprocess $gamma^* g to (qbar q) g$ with $qbar q to V$ is computed to lowest order in perturbative QCD while retaining the quark transverse momentum $k_perp$ in the propagator denominators; the amplitude is transformed to impact-parameter space, where a Sudakov factor $exp[-S(tau,b,Q)]$ resums gluonic radiative corrections and suppresses large $b$; the meson is described by a Gaussian light-cone wavefunction matched to the asymptotic distribution amplitude; and the gluon GPD $H^g$ is built from a double-distribution ansatz fed by the CTEQ5M gluon distribution. With this, the collinear leading-twist estimate of Eq. (94), which the authors find to be well above experiment, is reduced to agreement with the H1 and ZEUS data on $sigma(gamma^* p to rho p)$, $sigma(gamma^* p to phi p)$, $sigma_L$ and $R$ for $Q^2 > 4$ GeV$^2$. The approach also reproduces the observed hierarchy of helicity amplitudes, $L to L$ dominant with $T to T$ at roughly a quarter of it near $t=0$, and gives spin density matrix elements in fair agreement with the measured $Q^2$ and $t$ dependence.

The premise it rests on. The momentum-transfer dependence is not derived but inserted by hand as exponentials with slopes taken from data, and the transverse-sector normalization is fixed by fitting the decay constant $f_{V T}$ and the slope $B_{VTT}$ to the same HERA measurements the calculation is then compared with.

The longitudinal cross section ratio between $phi$ and $rho$ production should approach $(f_{phi L} C_phi / f_{rho L} C_rho)^2$ as $Q^2$ grows, which the authors' curves indeed do.

The energy dependence of $sigma_L$ at fixed $Q^2$ should follow $W^{4delta(Q^2)}$ with the same $delta$ that governs the small-$x$ behaviour of the gluon distribution, so measuring the power tests the GPD input directly.

The real-to-imaginary ratio of the dominant $L to L$ amplitude should rise with $Q^2$ roughly as $pi delta(Q^2)/2$, reaching 0.2–0.4 over the HERA range, which is what the computed Argand diagram shows.

The helicity correlation $A_{LL}$ should be small, of order $langle k_perp^2 rangle/Q^2$ times the ratio of polarized to unpolarized gluon density, because it arises only from interference between $H^g$ and $tilde H^g$.

Spin density matrix elements should follow relations identical to those obtained under the assumption of natural-parity t-channel exchange with negligible proton helicity flip, because the $tilde H^g$ and $E^g$ terms enter only as bilinears.

If the Sudakov-plus-$k_perp$ mechanism is really what fixes the normalization, then the same correction should be needed in other exclusive processes whose collinear leading-twist predictions overshoot — deeply virtual Compton scattering and exclusive $pi^0$ or $J/psi$ production are the natural places to check whether one universal suppression recipe works.

The paper's own freedom in trading $M_V$ against $f_{V T}$ and $B_{VTT}$ at fixed $(f_{V T}/M_V)^2/B_{VTT}$ suggests that the transverse sector is effectively a one-parameter fit; a genuinely independent determination of $f_{rho T}$ from lattice or QCD sum rules would convert the transverse agreement into a prediction rather than a calibration.

Because the $t$ dependence is inserted as an exponential rather than derived, a future GPD model that produces the slope from the proton's transverse structure would turn the spin density matrix elements versus $t$ into a sharp discriminator between models, and would test whether $B_{VTT} = B_{VLL}/2$ or $B_{VTT} = B_{VLL}$.

The skewing factor of 1.6–2 at $xi = 0.1$ implies that analyses using the $ln(1/x)$ approximation at forward energies may underestimate the gluonic contribution by a factor of a few, which is a checkable statement against fixed-target and COMPASS kinematics where $xi$ is not small.

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review