Parton distribution functions (PDFs) f_i(x, Q^2) give the probability density of finding parton i (quark flavor or gluon) carrying momentum fraction x inside the proton at resolution scale Q^2. PDFs cannot be calculated perturbatively and must be extracted from experimental data, but their evolution with Q^2 is predicted by the DGLAP equations and serves as a rigorous test of QCD.
Parton distribution functions are the bridge between the fundamental QCD Lagrangian and observable hadron-level cross sections. The factorization theorem of QCD states that a hadronic cross section can be written as a convolution of perturbatively calculable partonic cross sections with non-perturbative PDFs: sigma(pp -> X) = sum_{i,j} integral f_i(x_1, Q^2) * f_j(x_2, Q^2) * sigma-hat(ij -> X) dx_1 dx_2. This separation into short-distance (calculable) and long-distance (universal but non-perturbative) components is the foundation of collider phenomenology.
The DGLAP evolution equations describe how PDFs change with the resolution scale Q^2. As Q^2 increases, the probe resolves finer structure and sees more parton splittings. The equations are integro-differential equations involving the splitting functions P_{ab}(z), which are calculated perturbatively and now known to three-loop accuracy (NNLO). The evolution predicts that at high Q^2, the gluon and sea quark distributions grow rapidly at small x while the valence quark distributions shift toward smaller x. This has been verified over four decades in Q^2 by combining DIS data from HERA with hadron collider data from the Tevatron and LHC.
Modern global PDF fits (CT18, MSHT20, NNPDF4.0) extract PDFs by fitting to thousands of data points from DIS, Drell-Yan, jet production, W/Z production, top quark production, and other processes. The input data span a wide range of x (from ~10^{-5} at HERA to ~0.7 from fixed-target DIS) and Q^2 (from a few GeV^2 to ~10^5 GeV^2 at the LHC). The fits determine not only central values but also uncertainty bands, propagated using either the Hessian method (CT, MSHT) or Monte Carlo replicas (NNPDF). PDF uncertainties are typically the dominant theoretical uncertainty for LHC precision measurements.
At very small x (below ~10^{-3}), the rapid growth of the gluon PDF raises the question of saturation: at some point, the gluon density becomes so high that nonlinear recombination effects (gluon merging) must balance the splitting, taming the growth. This regime is described by the BFKL and BK/JIMWLK evolution equations and is a major target for future electron-ion collider (EIC) experiments. Understanding the transition from the dilute (DGLAP) to the saturated regime is one of the frontiers of QCD.