We present a new model for generalized parton distributions (GPDs), based on the aligned jet model, which successfully describes the deeply virtual Compton scattering (DVCS) data from H1, ZEUS, HERMES and CLAS. We also present an easily implementable and flexible algorithm for their construction. This new model is necessary since the most widely used models for GPDs, which are based on factorized double distributions, cannot, in their current form, describe the DVCS data when employed in a full QCD analysis. We demonstrate explicitly the reason for the shortcoming in the data description. We also highlight several non-perturbative input parameters which could be used to tune the GPDs, and the t-dependence, to the DVCS data using a fitting procedure.

We use several diverse parameterizations of diffractive parton distributions, extracted in leading twist QCD analyses of the HERA diffractive deep inelastic scattering (DIS) data, to make predictions for leading twist nuclear shadowing of nuclear quark and gluon distributions in DIS on nuclei. We find that the HERA diffractive data are sufficiently precise to allow us to predict large nuclear shadowing for gluons and quarks, unambiguously. We performed detailed studies of nuclear shadowing for up and charm sea quarks and gluons within several scenarios of shadowing and diffractive slopes, as well as at central impact parameters. We compare these leading twist results with those obtained from the eikonal approach to nuclear shadowing (which is based on a very different space-time picture) and observe sharply contrasting predictions for the size and Q 2 -dependence of nuclear shadowing. The most striking differences arise for the interaction of small dipoles with nuclei, in particular for the longitudinal structure function F A L .

Predictions for the diffractive photoproduction of the Υ-family at HERA energies, within the framework of the analysis by Frankfurt, Koepf and Strikman, are presented. Two novel effects lead to a significant enhancement of the original calculation: the non-diagonal (or skewed) kinematics, calculated to leading-log(Q 2 ) accuracy, and the large magnitude of the real part of the amplitude. The resultant cross sections are found to agree fairly well with recent preliminary data from ZEUS and H1. A strong correlation between the mass of the diffractively produced state and the energy dependence of the cross section is found. In particular, a considerably stronger rise in energy is predicted than that found in J/ψ-production.

As a consequence of QCD factorization theorems, a wide variety of inclusive and exclusive cross sections may be formulated in terms of a universal colour dipole cross section at small $x$. It is well known that for small transverse size dipoles this cross section is related to the leading-log gluon density. Using the measured pion-proton cross section as a guide, we suggest a reasonable extrapolation of the dipole cross section to the large transverse size region. We point out that the observed magnitude and small $x$ rise of the gluon density from conventional fits implies that the DGLAP approximation has a restricted region of applicability. We found that `higher twist' or unitarity corrections are required in, or close to, the HERA kinematic region, even for small `perturbative' dipoles for scattering at central impact parameters. This means that the usual perturbative leading twist description, for moderate virtualities, $1 < Q^2 < 10$ GeV$^2$, has rather large `higher twist' corrections at small $x$. In addition, for these virtualities, we also find sizeable contributions from large non-perturbative dipoles ($b \gsim 0.4$ fm) to $F_2$, and also to $F_L$. This also leads to deviations from the standard leading twist DGLAP results, at small $x$ and moderate $Q^2$. Our model also describes the low $Q^2$ data very well without any further tuning. We generalize the Gribov unitarity limit for the structure functions of a hadron target to account for the blackening of the interaction at central impact parameters and to include scattering at peripheral impact parameters which dominate at extremely large energies.Comment: Final version, 38 pages, 16 figures, 1 table. A successful comparison to all low Q^2 HERA data is included. The discussion has been completely rewritten and extended to include both a detailed comparison with other models for the dipole cross section and also a new section on the approach to the black limit in DIS, including various new predictions. 23 new references have been added and several figures change

We present a detailed next-to-leading order (NLO) leading twist QCD analysis of deeply virtual Compton scattering (DVCS) observables, for several different input scenarios, in the MS scheme. We discuss the size of the NLO effects and the behavior of the observables in skewedness ζ, momentum transfer, t, and photon virtuality, q 2 = −Q 2 . We present results on the amplitude level for unpolarized and longitudinally polarized lepton probes, and unpolarized and longitudinally polarized proton targets. We make predictions for various asymmetries and for the DVCS cross section and compare with the available data.