The NNPDF1.2 parton set: implications for the LHC

Abstract

Recently a new set of Parton Distribution Functions (NNPDF1.2) has been produced and released by the NNPDF Collaboration. The inclusion of dimuon data in the analysis allows a determination of the strange content of the proton with faithful uncertainty estimation together with a precision determination of electroweak parameters. In this contribution, we discuss some of the implications of the NNPDF1.2 set, and in particular of its uncertainty determination of the strange PDFs, for LHC phenomenology. First of all, we study the impact on the electroweak boson production cross-section, with special attention to the �(Z)/�(W) ratio. Then we revisit the top pair production cross-section, and perform a comparison of partonic fluxes between various PDF sets. Finally, we discuss the potential of using associated production of W with a charm quark at the Tevatron and the LHC to constrain the proton strangeness. 1 The NNPDF1.2 parton set The determination of the strange and antistrange quark distributions of the nucleon is considerably interesting from the phenomenological point of view. However, till very recently, the bulk of data included in parton determinations, namely neutral-current deep-inelastic scattering, had minimal sensitivity to flavour separation, and no sensitivity at all to the separation of quarks and antiquark contributions. As a consequence, in standard parton fits the strange and antistrange quark distributions were not determined directly: rather, they were assumed to be equal and proportional to the total light antiquark sea distribution. Due to the availability of the new deep-inelastic neutrino and anti-neutrino charm production data, which is directly sensitive to the strange and antistrange parton distributions, independent parametrizations of the strange and antistrange distributions have been included in most recent parton fits. However, the standard method for determining parton distributions, based on fitting the parameters of a fixed functional form, is known to be hard to handle when the experiments are relatively unconstraining. An alternative approach to parton determination which overcomes this difficulty has been developed by the NNPDF Collaboration in a series of papers [1, 2, 3, 4, 5, 6] a . The method is based on the use of neural networks for parton parametrization, and a Monte Carlo method supplemented by a suitable training and stopping algorithm for the construction of the parton fit. In this approach, parton distributions are given as a Monte Carlo sample which represents their probability distributions as inferred from the data. a See also Ref. [7] for a series of benchmark comparisons between the NNPDF approach and the standard approach.