Phenomenology in the next decade will be dominated by data from the
experiments at the Large Hadron Collider (LHC) at CERN. In order to
fully exploit the outcome of a hadron collider like the LHC, a
detailed description of the structure of the nucleon is needed. Such a
description incorporates non-perturbative QCD dynamics and can only be
computed from first principles on the lattice, or extracted from
experimental data. This project focuses on obtaining a faithful and unbiased
determination of the parton distribution functions that describe the
inner structure of the nucleon.
In order to compute the errors and correlations in a reliable manner,
a Monte carlo sample of artificial datasets is generated, in such a way
that it reproduces the statistical features of the real experimental
data. A Monte Carlo sample of PDFs is then generated by fitting a PDF
for each set of artificial data. The ensemble of fitted PDF is then used to
define a probability measure in the space of functions.
The fitting task can be naturally run in parallel on a cluster of PCs,
or on the Grid. Each fit only requires the data corresponding to one
given replica of the original dataset. The aim of the project is to develop
a framework which
allows the user to run the fitting, monitor the progress, collect the
results, in parallel for a large number of replicas. The use of
existing tools on different platforms should be investigated, and
expanded if needed. The statistical analysis and some of the physical
implications should also be studied.
