Particle Flow at CMS and the ILC

This thesis describes hadron reconstruction at the Compact Muon Solenoid (CMS) experiment at the Large Hadron Collider (LHC) at CERN, Geneva. The focus is on the particle flow reconstruction of these objects. This thesis revisits the subject of the CMS calorimeters' non-linear response to hadrons. Data from testbeam experiments conducted in 2006 & 2007 is compared with simulations and substantial differences are found. A particle flow calibration to correct the energy response of the testbeam data is evaluated. The reconstructed jet response is found to change by ~ 5% when a data-driven calibration is used in place of the calibration derived from simulation. Collision data taken at the early stage of CMS' commissioning is also presented. The hadron response in data is determined to be compatible with testbeam results presented in this thesis. This thesis also details the use of neural networks to improve the energy measurement of hadrons at CMS. The networks are implemented in a functional and concurrent language (Erlang). The advantages to using a concurrent language to solve problems which can be parallelized are demonstrated. The Monolithic Active Pixel Sensor (MAPS) is described. This device, with 50 x 50 um^2 pitch and binary readout, is designed for sampling calorimeters at the proposed International Linear Collider (ILC). Such calorimeters are optimized for a particle flow reconstruction. The MAPS benefits from a novel industrial process whereby a deep p-well implant separates the epitaxial silicon layer from parasitic pixel electronics, to increase the charge collection efficiency. Data acquired from testbeam in 2007 is analysed. The low efficiency of the prototype is attributable to operating the sensor at an incorrect working point; this has subsequently been addressed.