The minimum pressures required for ethylene to propagate a decomposition flame were found at 40, 120 and 280°C in six-inch diameter pipe. Pressure limits found using a anovel “starburst” pyrotechnic igniter array were confirmed as true values using a 75:25 ethylene-oxygen flame cannon. By producing simultaneous initiation across the pipe cross-section, these ignition systems simulated extensive heat sources in process equipment such as adiabatic compression of a trapped diatomic gas, external fire exposure, and runaway reaction in a purifier bed. Heat losses from the nascent flame caused by flame stretch, buoyancy and other effects were minimized. The limits at 40°C, 120°C and 280°C (440 ± 5, 368 ± 8 and 250 ± 10 psia) are about one-half the lowest literature values. The pressure limit varies linearly with temperature and is given by P (psia) = 468 - 0.784* T where T is temperature (°C). The limit was extrapolaed to the derived autodecomposition curve, yielding a complete ignition diagram for ethylene decomposition. Flame speeds were less than 2 ft/s with a typical value of 1 ft/s. The new data explain several reported incidents [15] in this pressure range and underline the potential for decomposition in ethylene purification systems. The data support the practice of blowing down intermediate pressure ethylene to about 150 psi to extinguish flames in hot equipment while avoiding metal embrittlement caused by decompression cooling. Addition of up to 50 mol% hydrogen to ethylene at 120°C progressively lowered the ignition energy; the pressure limit found using 5 kJ “point” ignition sources fell with increased hydrogen. Conversely, the “true” limit found using the starburst array remained similar to ethylene alone (380 ± 10 psia) at up to 25% hydrogen but increased to 490 psia at 50% hydrogen, showing that the latter is a weak diluent. Maximum pressures, pressure rise rates and flame speeds were slightly lower than for pure ethylene. Carbon, hydrogen and methane decomposition products tended to an equilibrium composition with only about 1% ethane. The limit data help explain reported explosions in ethylene-hydrogen mixtures at initial pressures less than 450 psi [15]. Numerous process materials including bed packings and high surface area iron may create hot spots via catalytic hydrogenation of ethylene, hence the lowering of ignition energy by hydrogen can be a hazardous feature. Nitrogen proved to be an effective diluent forethylene-hydrogen mixtures; cases examined with respect to low pressure polymerization reactors displayed pressure limits above 700 psia. Propylene is an effective diluent in binary mixtures with ethylene despite its small positive heat of formation; the pressure limit at 120°C was greater than 600 psia at 57% ethylene. Literature values for the upper flammability limit (UFL) of ethylene in oxygen were critically reviewed. UFL literature data are increasingly unreliable above 90% ethylene owing to igniter limitations. The condition that the UFL must equal 100% ethylene at the observed pressure limit enabled the UFL to be estimated at about 120°C and pressures above 50 psia, where the UFL variation was approximately linear. Ignition sources having insufficient ignition energy may give results that are not only non-conservative, but also have the wrong trend. “Limits” from such studies are not true limits of flammability but only loci at which the igniter system dissipates sufficient power density for ignition. Statistically-based limit modeling of high ignition energy systems may be particularly prone to error, especially if data are extrapolated. An example, based on the UFL findings, is prediction of the limiting oxidant concentration in high pressure process streams based on “low ignition energy” test data. In order to determine true propagation limits for high ignition energy systems, careful attention must be given both to the ignition source and test vessel. Increased ignition energy should not progressively change the observed limit. Especially for slowly developing flames such as ethylene, igniter arrays are more efficient than “point” igniters and can simulate the most severe cases for plant equipment. The test vessel should be large enough to prevent “overdriving” by the ignition source and minimize quenching of the developing flame. The starburst igniter array described in this paper revealed true limits in a six foot long, six inch diameter test pipe, but had an upper operability limit of 125°C and required a significant correction for the overpressure it produced. At temperatures above 125°C, flame initiation was necessary owing to the thermal limitations of pyrotechnics. A 26 foot long test pipe was used to minimize overpressure effects of the flame cannon, which were similar to those observed for the starburst in the 6 foot long pipe. Over-pressure effects were negligible using the starburst in the 26 foot long pipe and this was the preferred system for test temperatures up to 125°C.