Revealing the Dynamics of Flames by Tracking Material Points in Highly Resolved Simulations

Principal Investigator:
Feichi Zhang

Engler-Bunte-Institute, Karlsruhe Institute of Technology

Local Project ID:

HPC Platform used:
Hazel Hen and Hawk of HLRS

Date published:


Most of the world’s primary energy is generated by combustion–from fossil fuels like coal and oil to bio-fuels like methane. Because of this, it is an important task to reduce the release of greenhouse gases as well as other pollutants from these processes. But due to the physical complexity of combustion as well as the many time and length scales involved, an in-depth study is challenging. Highly resolved simulations of relevant combustion cases have only been made possible with the advance in high performance computing. In this report, we show two examples of such highly-resolved simulations only possible on today’s largest supercomputers: The first is a simulation of a hydrogen flame which features so called thermo-diffusive instabilities. As the current energy economy shifts toward hydrogen due to its carbon-free combustion, these types of flames currently gain more interest in the scientific community. Secondly, a simulation of the interaction between a flame and turbulent flow is presented since almost all technically relevant combustion applications take place in a turbulent flow field.


All simulations shown here are conducted with an in-house code [1,2] based on OpenFOAM. It has been developed for high-performance computing applications and features a number of performance optimizations to utilize the parallel hardware of modern supercomputers efficiently and saves up to 70 % of total simulation times compared with OpenFOAM’s standard solvers [3,4]. More information about the code, its performance and accuracy can be found in [2].

Thermo-diffusive Instabilities in Lean Hydrogen Flames

Hydrogen is regarded as a future-oriented, carbon emission-free fuel. However, hydrogen flames behave differently than many other carbon-based fuels, yielding a much higher burning velocity. If hydrogen is burned with excess air, the resulting flames tend to form strong thermo-diffusive instabilities leading to the formation of cellular structures. An example of this is shown in Fig. 1. The flame starts out flat and burns from the top to bottom. During the propagation, instabilities develop on the flame surface and “flame fingers” protrude into the unburnt fuel/air mixture, creating cellular structures. These structures are the result of locally changing flame speeds due to the diffusion of oxygen and hydrogen. The simulation consists of 100 million cells in two dimensions and has been carried out taking detailed molecular diffusion for each chemical species as well as intermediates into account. The setup is based on the work of Berger et al. [5].

Interaction of Turbulent Flow with Flames

Turbulent flow is characterized by fluctuations in the fluid flow which interact with the thin reaction layers of flames. This interaction takes place on length and time scales with span several orders of magnitudes. Because of this, detailed simulations of these phenomena require enormous computational resources and are only possible on today’s supercomputers. Figure 2 shows an example of a direct numerical simulation of a laboratory-scale, mixed-mode turbulent flame [2,6]. The combustion process is characterized by a novel burning strategy combining both premixed and non-premixed combustion in order to achieve an improved flame stability. The simulation case consists of 150 million cells. 19 chemical species and more than 300 chemical reactions are considered for the simulation of the flame. The two-dimensional cutting plane shows the temperature field, where the high-temperature regions shows the flame. The structures in the center visualize the turbulent flow with an iso-surface of the vorticity. By performing this detailed simulation, the interaction of the turbulent flow structures and the chemical reactions in the flame can be studied to improve future combustion models.

Further Insight into Flame Dynamics with Flame Particle Tracking

A novel way of studying the physical complexity of flames has been developed in recent works, which employs a Lagrangian viewpoint of the flame. In this method, called the flame particle tracking method [7], virtual particles are seeded onto the flame. They track specific points on the flame over time, called material points. In this way, not only instantaneous states of the flame can be analyzed but also the time history of important quantities, thus revealing the underlying flame dynamics [8]. Figure 3 shows an example of the flame particle method in a numerical simulation. Depicted is the surface of a turbulent flame, colored by its heat release rate. The red spheres show the virtual flame particles which track the flame and generate additional information about the flame dynamics during the large-scale simulations.

The flame particle tracking method has been implemented in the in-house solver described above [8,9]. The implementation is able to handle millions of concurrently tracked particles on thousands of parallel processes without significant overhead. Figure 4 shows parallel scaling results on up to 1000 CPU cores for a smaller case. On the left, the simulation is performed without particle tracking and on the right with the newly implemented flame particle tracking approach, showing very good performance.


[1] Zhang, F., Bonart, H., Zirwes, T., Habisreuther, P., Bockhorn, H., Zarzalis, N. Direct Numerical Simulation of Chemically Reacting Flows with the Public Domain Code OpenFOAM. in: Nagel, W.E., Kröner, D.H., Resch, M.M. High Performance Computing in Science and Engineering ’14. Springer. 2014: 221–236; (

[2] T. Zirwes, F. Zhang, P. Habisreuther, M. Hansinger, H. Bockhorn, M. Pfitzner, and D. Trimis, “Quasi-DNS dataset of a piloted flame with inhomogeneous inlet conditions,” Flow, Turbulence and Combustion, vol. 104, pp. 997–1027, 2020 (

[3] T. Zirwes, F. Zhang, J.A. Denev, P. Habisreuther, and H. Bockhorn, “Automated Code Generation for Maximizing Performance of Detailed Chemistry Calculations in OpenFOAM,” in High Performance Computing in Science and Engineering ’17 (W. Nagel, D. Kröner, and M. Resch, eds.), pp. 189–204, Springer, 2017 (

[4] T. Zirwes, F. Zhang, J.A. Denev, P. Habisreuther, H. Bockhorn, and D. Trimis, “Improved Vectorization for Efficient Chemistry Computations in OpenFOAM for Large Scale Combustion Simulations,” in High Performance Computing in Science and Engineering ’18 (W. Nagel, D. Kröner, and M. Resch, eds.), pp. 209–224, Springer, 2018 (

[5] L. Berger, K. Kleinheinz, A. Attili, H. Pitsch, H. “Characteristic patterns of thermodiffusively unstable premixed lean hydrogen flames”. Proceedings of the Combustion Institute, 37(2), 1879-1886, 2019

[6]Barlow, R., Meares, S., Magnotti, G., Cutcher, H., Masri, A. Local extinction and near-field structure in piloted turbulent CH4/air jet flames with inhomogeneous inlets, Combustion and Flame 162 (10), 2015, 3516–3540.

[7] S. Chaudhuri. “Life of flame particles embedded in premixed flames interacting with near isotropic turbulence”. Proceedings of the Combustion Institute, 35(2), 1305-1312, 2015

[8] T. Zirwes, F. Zhang, Y. Wang, P. Habisreuther, J.A. Denev, Z. Chen, H. Bockhorn, and D. Trimis, “In-situ Flame Particle Tracking Based on Barycentric Coordinates for Studying Local Flame Dynamics in Pulsating Bunsen Flames,” in Proceedings of the Combustion Institute, vol. 38, Elsevier, 2020 (

[9] T. Zirwes, F. Zhang, J.A. Denev, P. Habisreuther, H. Bockhorn, and D. Trimis, “Implementation of Lagrangian Surface Tracking for High Performance Computing,” in High Performance Computing in Science and Engineering ’20 (W. Nagel, D. Kröner, and M. Resch, eds.), Springer, 2020

Research Team

Thorsten Zirwes1,2, Feichi Zhang2, Peter Habisreuther2, Jordan A. Denev1, Henning Bockhorn2, Dimosthenis Trimis2

1 Steinbuch Centre for Computing, Karlsruhe Institute of Technology
2 Engler-Bunte-Institute, Chair of Combustion Technology, Karlsruhe Institute of Technology

Scientific Contact:

Thorsten Zirwes
Karlsruhe Institute of Technology
Steinbuch Centre for Computing
Hermann-von-Helmholtz-Platz 1, D- 76344 Eggenstein-Leopoldshafen (Germany)
e-mail: thorsten.zirwes [@]

HLRS project ID: DNSbomb

January 2021