This paper describes a methodology that has been developed to facilitate a detailed study of molecular composition effects on particulate matter emissions in advanced diesel engines. This includes a sophisticated numerical optimization algorithm to formulate well-characterized diesel fuel blends and an analytical method to characterize diesel fuels more accurately than previously possible. These tools are described, together with application to the formulation of test fuels with identical boiling point distribution and cetane number, but differing molecular composition. Test results are discussed from an advanced high-speed direct injection diesel engine for several of these fuels, demonstrating the improved insight and understanding available from these combined techniques.