The replacement of hydrocarbon fuels by ammonia in industrial systems is challenging due to its low burning velocity, its narrow flammability range, and a large production of nitric oxide and nitrogen dioxide when burned close to stoichiometric conditions. Cracking a fraction of ammonia into hydrogen and nitrogen prior to injection in the combustion chamber is considered a promising strategy to overcome these issues. This paper focuses on evaluating how different levels of ammonia cracking affect the overall burning velocity, the lean blow-off limit, the concentration of nitric oxide and nitrogen dioxide, and the flame response to acoustic perturbations. Swirl stabilized premixed flames of pure ammonia–air and ammonia–hydrogen–nitrogen–air mixtures mimicking 10%, 20%, and 28% of cracking are experimentally investigated. The results show that even though ammonia cracking is beneficial for enhancing the lean blow-off limit and the overall burning velocity, its impact on pollutant emissions and flame stability is detrimental for a percentage of cracking as low as 20%. Based on an analysis of the flame dynamics, reasons for these results are proposed.

Prediction of thermoacoustic combustion instabilities via low-order modeling approaches requires a model for flame dynamics, which can be inferred from large eddy simulation (LES). Modeling errors associated with the thickened flame model for LES(TFLES), a combustion model widely used in this context, are benchmarked in the present study. In particular, the thickening factor—a fundamental parameter of the TFLES model—is varied firstly by changing the grid, while keeping the flame resolution constant. Secondly, the thickening factor is changed while using a constant computational grid. A fully flame-resolved simulation without the use of a turbulence–chemistry interaction model serves as a reference. A global chemical mechanism is optimized for use in LES of a partially cracked ammonia flame in order to facilitate cost-effective simulation. Its performance in canonical one-dimensional configurations is assessed versus a detailed chemical scheme. From the LES time series data, flame transfer functions (FTFs) are identified and compared to experiment qualitatively as well as to each other. It is demonstrated that the number of grid cells used to resolve the flame front, rather than the thickening factor, is the TFLES model parameter with the most impact on predictions when using a static modeling constant in the efficiency function. It is also demonstrated that thickening introduces damping of high-frequency heat release fluctuations in turbulent flames. However, the FTFs of the present configuration approach zero gain already at lower frequencies.