RANS methodology to assess how resolving the near-nozzle flow of an underexpanded H ₂ jet impacts mixture stratification

Hydrogen (H 2 ) direct injection (DI) engines hold great promise as a future sustainable option for high power-output applications, where in-cylinder mixture formation is critical and actively investigated by means of CFD. This includes simulating the underexpanded jets produced by current DI systems, containing near-nozzle shocks that require fine meshes to be numerically resolved, incompatible with practical CFD. The subject of this work is exploring whether the use of too coarse meshes as in most practical RANS simulations critically impacts predicting mixture stratifications. To that end, the jet studied experimentally by Ruggles and Ekoto is simulated in the CONVERGE CFD solver, employing a range of mesh resolutions. Comparisons with theoretical references and experimental findings reveal that the ideal gas law suffices to predict flow properties around the near-nozzle Mach disk, and that low nozzle inflow turbulence intensity allows for better resolving this region. On the other hand, a compromise for this parameter is required to improve predictions of H 2 stratifications downstream, combined with a Reynolds Stress model and turbulent Schmidt number equal to 0.5. It is found that accurate H 2 stratifications with too coarse meshes can only be obtained far enough away from the nozzle. However, in the near-nozzle field, they differ from those obtained using a fine mesh. With the latter, the Mach disk is resolved, and high-momentum flow paths emerge around it and away from the centreline. These lead to enhanced turbulent mixing within the emerging shear layer and to a reduction in H 2 concentrations along the jet axis further downstream. Resolving the near-nozzle flow does affect turbulent mixing and H 2 stratifications at least in the near field for this jet in a quiescent ambience. Thus, these effects are expected to impact engine in-cylinder flow interactions and should be reproduced with either fine meshes or dedicated models.