We present a next-to-leading order (NLO) global DGLAP analysis of nuclear parton distribution functions (nPDFs) and their uncertainties. Carrying out an NLO nPDF analysis for the first time with three different types of experimental input -deep inelastic ℓ+A scattering, Drell-Yan dilepton production in p+A collisions, and inclusive pion production in d+Au and p+p collisions at RHIC -we find that these data can well be described in a conventional collinear factorization framework. Although the pion production has not been traditionally included in the global analyses, we find that the shape of the nuclear modification factor R dAu of the pion p T -spectrum at midrapidity retains sensitivity to the gluon distributions, providing evidence for shadowing and EMC-effect in the nuclear gluons. We use the Hessian method to quantify the nPDF uncertainties which originate from the uncertainties in the data. In this method the sensitivity of χ 2 to the variations of the fitting parameters is mapped out to orthogonal error sets which provide a user-friendly way to calculate how the nPDF uncertainties propagate to any factorizable nuclear cross-section. The obtained NLO and LO nPDFs and the corresponding error sets are collected in our new release called EPS09. These results should find applications in precision analyses of the signatures and properties of QCD matter at the LHC and RHIC.

The scale dependence of the ratios of parton distributions in a proton of a nucleus A and in the free proton,, is studied within the framework of the lowest order leading-twist DGLAP evolution. By evolving the initial nuclear distributions obtained with the GRV-LO and CTEQ4L sets at a scale Q 2 0 , we show that the ratios R A i (x, Q 2 ) are only moderately sensitive to the choice of a specific modern set of free parton distributions. We propose that to a good first approximation, this parton distribution setdependence of the nuclear ratios R A i (x, Q 2 ) can be neglected in practical applications. With this result, we offer a numerical parametrization of R A i (x, Q 2 ) for all parton flavours i in any A > 2, and at any 10 −6 ≤ x ≤ 1 and any Q 2 ≥ 2.25 GeV 2 for computing cross sections of hard processes in nuclear collisions.

We introduce a global analysis of collinearly factorized nuclear parton distribution functions (PDFs) including, for the first time, data constraints from LHC proton–lead collisions. In comparison to our previous analysis, EPS09, where data only from charged-lepton–nucleus deep inelastic scattering (DIS), Drell–Yan (DY) dilepton production in proton–nucleus collisions and inclusive pion production in deuteron–nucleus collisions were the input, we now increase the variety of data constraints to cover also neutrino–nucleus DIS and low-mass DY production in pion–nucleus collisions. The new LHC data significantly extend the kinematic reach of the data constraints. We now allow much more freedom for the flavor dependence of nuclear effects than in other currently available analyses. As a result, especially the uncertainty estimates are more objective flavor by flavor. The neutrino DIS plays a pivotal role in obtaining a mutually consistent behavior for both up and down valence quarks, and the LHC dijet data clearly constrain gluons at large momentum fraction. Mainly for insufficient statistics, the pion–nucleus DY and heavy-gauge-boson production in proton–lead collisions impose less visible constraints. The outcome – a new set of next-to-leading order nuclear PDFs called EPPS16 – is made available for applications in high-energy nuclear collisions.