Fixed target Drell-Yan data and NNLO QCD fits of parton distribution functions

Abstract

We discuss the influence of fixed target Drell-Yan data on the extraction of parton distribution functions at next-to-next-to-leading order (NNLO) in QCD. When used in a parton distribution fit, the Drell-Yan (DY) data constrain sea quark distributions at large values of Bjorken x. We find that not all available DY data are useful for improving the precision of parton distribution functions (PDFs) obtained from a fit to the deep inelastic scattering (DIS) data. In particular, some inconsistencies between DIS-based parton distribution functions and DY data for large values of dilepton rapidity are found. However, by selecting a sample of the DY data that is both representative and consistent with the DIS data, we are able to perform a combined PDF fit that significantly improves the precision of non-strange quark distributions at large values of x. The NNLO QCD corrections to the DY process are crucial for improving the precision. They reduce the uncertainty of the theoretical prediction, making it comparable to the experimental uncertainty in DY cross-sections over a broad range of x.2

Dynamical next-to-next-to-leading order parton distributions

“…Our dynamical analysis indeed yields a smaller value at NNLO in Table 1 [61] which was performed for a restricted set of (mainly small-x) DIS data.) The same holds for our NNLO 'standard' fit result in Table 2 Table 2 agrees, within errors, with the results of [15,16] and [17], 0.1143 ± 0.0014 and 0.1128 ± 0.0015, respectively, and is compatible with the one obtained from a 'standard' fit [60] to a restricted set of small-x DIS data. Data for high-p T jet production in hadron-hadron scattering should not be included in a consistent NNLO analysis, since such processes are theoretically known only up to NLO.…”

“…Since experimentally the dilepton p T is small (below about 1.5 GeV) as compared to the dilepton invariant mass M > ∼ 5 GeV, we have checked that it can be safely neglected and the two distributions can be related using leading order kinematics, as has been done in [17]:…”

“…Furthermore, since all our fits did not require the additional polynomial in (17) for the gluon distribution, we have set A g = B g = 0. This left us with a total of 21 independent fit parameters, including α s .…”

“…1 Therefore, the heavy flavor contributions F h 2,L are taken as given by fixed order NLO perturbation theory [55,56] as in our previous more restricted NNLO analyses [60,61]. This is also common in the literature [15,16,17] and the error in the resulting parton distributions due to NNLO corrections to heavy quark production is expected [15] to be less than their experimental errors. These contributions are gluon g(x, µ 2 F ) dominated and the factorization scale, also of the remaining parton distributions, should preferably chosen [62] to be µ 2 F = 4m 2 h , although a much larger choice like µ 2 F = 4(Q 2 + 4m 2 h ) leaves the NLO results essentially unchanged [7,8].…”

“…[9][10][11][12][13][14][15][16][17][18][19][20][21], the input scale is fixed at some arbitrarily chosen Q 0 > 1 GeV and the corresponding input distributions are less restricted. For example, the observed steep small-x behavior (a f < 0) of the gluon and sea distributions has to be fitted, allowing even for negative gluon distributions [12,13,14,18,19], i.e.…”