Advancement in the understanding and design of such important gas-phase processes as light hydrocarbon cracking, combustion, and partial oxidation hinges, in part, on the development of correct, detailed chemical kinetic models. These kinetic models may have thousands of individual chemical reaction steps, and many hundreds of species, along with attendant information on thermochemistry and rates of reaction. But the very size and complexity of the required chemical mechanisms makes them extremely difficult to construct correctly by hand. Chemists and engineers have thus turned to software tools that attempt to build these large mechanisms automatically. Unfortunately, none of these tools can properly treat the pressure-dependence of certain reaction steps — and a great many systems of industrial importance have at least some pressure-dependent reactions. This has technically limited large, automatically-generated chemical mechanisms to a narrow region called the "high-pressure limit". Furthermore, most of these tools do not systematically terminate the otherwise combinatorial growth of a computer-generated chemical mechanism, requiring the use of arbitrary criteria for deciding when mechanism growth is "complete".

We present a new, elementary-step-based mechanism generation algorithm which combines an integrated approach to pressure-dependent reactions with a rational, flux-based criteria for truncating mechanism growth. Applications to difficult problems in methane and "high-conversion" ethane pyrolysis, which could not be addressed by hand, and which include important pressure-dependent phenomena, reveal important new pathways not previously considered by other researchers. These reaction pathways can explain the experimentally observed behavior. An example including steam-cracking of ethane reveals certain pathways to minor species (thought to be important for the formation of coke and soot) to be pressure-dependent. The rational criteria for limiting mechanism growth gives us confidence that all the important pathways have been discovered and captured by our software.