DNS Study of Differential Diffusion in a Hydrogen Jet in Cross Flow Configuration Gauss Centre for Supercomputing e.V.


DNS Study of Differential Diffusion in a Hydrogen Jet in Cross Flow Configuration

Principal Investigator:
Christian Hasse

Simulation of reactive Thermo-Fluid Systems, Technical University of Darmstadt

Local Project ID:

HPC Platform used:
SuperMUC and SuperMUC-NG of LRZ

Date published:


As part of the changing energy landscape, the influence of hydrogen as fuel will further increase. Along with that, fuel flexibility of common combustion applications is a crucial issue for their development when using e.g. blends of hydrogen and fossil fuels or even pure hydrogen due to different combustion characteristics and especially the higher diffusivity of hydrogen. Thus, increasing amounts of hydrogen lead to differential diffusion effects, which alter the molecular transport mechanisms towards the reaction zone. In practical combustion applications, both molecular and turbulent transport processes play a significant role and affect the overall flame structure, flame stabilization mechanisms and pollutant formation.

Within this project, a DNS study concerning differential diffusion effects and mixing characteristics during hydrogen combustion, using a canonical jet in cross flow (JICF) flame configuration, has been performed. The investigations in the hydrogen JICF configurations are twofold. First, a detailed analysis of the DNS data could yield a fundamental understanding of mixing characteristics in the JICF configuration and differential diffusion effects, e.g. their relation to the location of the turbulent/non-turbulent interface (T/NT) [1]. Second, commonly applied tabulated chemistry approaches and their capability of predicting differential diffusion can be validated against the DNS data. The latter, which is of highly practical interest for a related project dealing with hydrogen combustion in a novel micro gas turbine, using a JICF configuration for fuel injection [2], will be the final target of the current project.

As a first step of the study, the jet in cross flow results were compared against experimental and simulation results, published by Sandia National Laboratories [4]. The typical jet visualization of the JICF simulation is presented by volume rendering of temperature, H2O mass fraction and mixture fraction iso-surfaces, given in Fig. 1.

Results and Methods

For the JICF simulations, the direct numerical simulation code DINO [3] was used. DINO is a three-dimensional low-Mach number DNS solver code with a 6th-order finite-difference spatial discretization for reacting and multiphase turbulent flows. The code is parallelized in two dimensions using the 2DECOMP&FFT library that acts on top of standard MPI and FFTW. The Poisson equation for pressure is solved by means of FFT for both periodic and non-periodic boundary conditions, but with dedicated preand post-processing FFT techniques in the latter case.

Although in any low-Mach number solver the time step restriction associated with acoustic waves is removed, the restriction of time step due to chemistry stiffness is still present. For that reason, an implicit time integration of the stiff chemical source terms has been implemented, relying on a semi-implicit Runge-Kutta 3rd-order. By default, the chemical source terms are computed using the opensource Cantera library. The transport properties are computed either with the Cantera library or with the EGlib-3.4 library.

The initial turbulent field is generated by inverse Fourier transform of an analytically prescribed energy spectrum (Passot-Pouquet or Von Karman-Pao). In DINO, input/output operations rely on MPI-I/O routines provided by the 2DECOMP&FFT library. These files are used for restarting the simulations while DINO uses parallel HDF5 saving for actual post-processing data. The code is already under GIT version control, which helps all users to quickly and safely carry out changes or updates, if needed. As build environment DINO uses cmake and it can be compiled with both GNU and Intel Fortran compilers.

After the validation, the project has been divided into two subprojects: (1) studying the mixing characteristics and (2) analyzing differential diffusion in the JICF configuration. The first one is almost finished and the publication is currently under review. In this manuscript different zones, highlighted in Fig. 2, are investigated in detail. Especially local mixing effects are assessed by spatial and temporal statistics at the corresponding locations.

During the project, typical minimum number of cores was 512, which was used for test jobs. The maximum number of cores, used during the strong scaling tests, was 65,536 cores. Two typical common numbers of cores which were often used for production cases are 2048 and 4096. The required resources of the project are summarized in Tab. 1.

Table 1: Summary of required resourdes
Total CPU-hOverall storageTypical # cores# genrated files
17.7 million20 TB204/40966000


On-going Research / Outlook

Even though the end of the project is almost reached, several methods are still under development. The most important modification is implementing a novel 2nd-order immersed boundary algorithm based on a ghost-cell level-set method. In this method, a local directional extrapolation scheme is employed, leading to an accurate representation of the boundaries on the DNS grid. This will help to implement the inflow condition in more accurate way, instead of using mathematical functions to implement the velocity and species profiles at the jet inlet. A first application, tested on SuperMUC, was the transient processes controlling ignition by a hot jet issued from a prechamber as it can be observed from Fig. 3.

References and Links

[1] F. Hunger, M. Gauding, C. Hasse. On the impact of the turbulent/non-turbulent interface on differential diffusion in a turbulent jet flow. Journal of Fluid Mechanics 802 (2016), R5

[2] P. Jeschke, A. Penkner. A novel gas Generator concept for jet engines using a rotating combustion chamber. In: ASME J. Turbomach.: Vol. 137, Nr. 7 (2015), 071010-1-8

[3] A. Abdelsamie, G. Fru, T. Oster, F. Dietzsch, J. Gábor, D. Thévenin. Towards direct numerical simulations of low-Mach number turbulent reacting and two-phase flows using immersed boundaries. Computers & Fluids 131 (2016), 123-141

[4] S. Lyra, B. Wilde, H. Kolla, J.M. Seitzman, T.C. Lieuwen, J.H. Chen. Structure of hydrogen-rich transverse jets in a vitiated turbulent flow, Combustion and Flame 162 (2015), 1234-1248

Research Team

Abouelmagd Abdelsamie2, Cheng Chi2, Christian Hasse1 (PI), Sebastian Popp1

1Simulation of reactive Thermo-Fluid Systems (STFS), Technical University (TU) of Darmstadt
2The Laboratory of Fluid Dynamics and Technical Flows (LSS: Lehrstuhl für Strömungsmechanik und Strömungstechnik), Otto von Guericke University Magdeburg 

Scientific Contact

Prof. Dr. Christian Hasse
Technische Universität Darmstadt
Simulation of reactive Thermo-Fluid Systems
Otto-Berndt-Str. 2, D-64287 Darmstadt (Germany)
e-mail: hasse [@] stfs.tu-darmstadt.de

LRZ project ID: pr74xi

December 2020

Tags: LRZ Technische Universität Darmstadt CFD CSE