Parton distributions from deep-inelastic-scattering data

Abstract

We perform the analysis of existing light-targets deep-inelasticscattering (DIS) data in the leading-order (LO), next-to-leading-order (NLO), and next-to-next-to-leading-order (NNLO) QCD approximations and extract PDFs simultaneously with the value of the strong coupling constant α s and the high-twist contribution to the structure functions. The main theoretical uncertainties and experimental uncertainties due to all sources of experimental errors in data are estimated, the latter generally dominate for the obtained PDFs. The uncertainty in Higgs boson production cross section due to errors in PDFs is ∼ 2% for the LHC and varies from 2% to 10% for the Fermilab collider under variation of the Higgs boson mass from 100 GeV to 300 GeV. For the W -boson production cross section the uncertainty is ∼ 2% for the both colliders. The value of α NNLO s (M Z ) = 0.1143 ± 0.0014(exp.) is obtained, while the high-twist terms do not vanish up to the NNLO as required by comparison to data.

Neural network determination of parton distributions: the nonsinglet case

“…in a global PDF fit to deep-inelastic data only [4], but as soon as new data (specifically from the Drell-Yan process) are added to this fit, their errors must be suitably rescaled [2]. Global fits to all available data are instead typically constructed by determining a priori a suitable "tolerance criterion", by inspection of the quality of the fit to all available data [5].…”

“…Unlike in most parton fits (such as e.g. [6,5,4]), however, each parton distribution is parametrized by a neural network, rather than by a fixed functional form. Then, the N dat data in each replica are used to determine the N pdf neural networks of the corresponding set.…”

“…The χ 2 is computed by fully including the covariance matrix of the correlated experimental uncertainties (with a suitable treatment of normalization errors [14]), as we shall discuss in more detail in section 5.1 below. Unlike in conventional determinations of parton distributions with errors [6,5,4], however, the covariance matrices of the best-fit parameters are irrelevant and need not be computed. The uncertainty on the final result is found from the variance of the Monte Carlo sample.…”

“…In the same figure, we also show the structure function obtained from the results of the global fits of the CTEQ [5] and MRST [6] collaborations and the fit to deep-inelastic data by Alekhin [4].…”

Structure Functions and Parton Distributions

“…Strangeness can only be determined through weak current DIS, i.e., essentially neutrino data (and some HERA data), which are however scarce and subject to sizable uncertainties. As a consequence, in current fits [3,4,7] the shape of the strange distribution is not determined, and assumed to be related in a fixed way to that of the light quark sea; the s ands distributions have been determined only in dedicated analyses based on neutrino data [5]. Finally, because (see fig.…”