The LHC at CERN, which is the most powerful collider in the world, was designed to search for the Higgs boson and new physics signals. But besides these discoveries, the LHC experiments have collected (and will collect) a large amount of data that can be used to improve our knowledge about Quantum ChromoDynamics (QCD). This thesis is in this former line of research, It contains two parts which use correlation variables to constrain non perturbative inputs. The first part concerns the constraints which can be put on fragmentation functions (FFs) using momentum correlation variables in hadron+jet production. The non perturbative inputs for the recent FFs were extracted from the LEP e^+e^- collision data. These LEP data put constraints on the FFs at low fragmentation variable x (x < 0.7), but do not give constraints at larger x. Because of that, the behaviors of the FF-sets at high x differ strongly from one another. A next-to-leading order (NLO) analysis of hadron-jet momentum correlations in p-p collisions at the LHC is carried out. We consider two cases: the correlation between a hadron and an away-side jet, and the correlation between a hadron and the jet to whom the hadron belongs. These two cases give similar results. We show that the inclusive charged hadron momentum distribution inside jets is a very sensitive observable which allows one to disentangle among various fragmentation function sets presently available. Correlations using identified hadrons (kaons, protons) are investigated as well. The second part concerns the study of the associated production of photon + heavy-flavor tagging in hadronic collisions at the Tevatron and the LHC. This cross section is used to constrain the heavy-flavor partonic densities inside the proton. A previous theoretical calculation showed a large discrepancy between the predictions and the experimental measurements for associated production of photon + charm at the Tevatron. We present three ways to compute this cross section depending on how the heavy flavor is tagged in experiments. The first way, called the invariant mass approach, has been used to recover the previous results, the second one, called the kt-flavor algorithm, uses a flavor jet algorithm which leads to infra-red an safe observable. The third way, called FF approach, uses fragmentation functions of partons into heavy-flavor hadrons, the jet size being taken into account in the final state factorization scale. The predictions for the three ways are given in the NLO approximation of perturbative QCD. They have been implemented into a code which has been built from the JetPhox program. By using the invariant mass approach, the results of the previous theoretical calculation have been re-obtained. These results are comparable with the one using kt-flavor algorithm. These two approaches give theoretical predictions which are too low for the charm flavor compared to Tevatron results. The charm hadron FF approach predicts a differential cross section with respect to the photon transverse momentum (pt) about 1.8 -- 2.7 times greater than the one obtained using the invariant mass approach when pt ∼ 80 -- 110 (GeV) at the Tevatron. This result agrees well with the experimental data. Predictions are given at LHC energy using the different ways for charm flavor and bottom flavor.
