University of Cambridge Logo
Department of Engineering Logo
Horizontal Line
Autoignition in Turbulent Flows Logo Autoignition in Turbulent Flows Autoignition in Turbulent Flows Logo

Christos N. Markides

Horizontal Line


Home Autoignition Archive Members Search Contact Information


Background Knowledge: Autoignition Theory

Autoignition (or auto-ignition), self-ignition or spontaneous ignition (not to be confused with spontaneous combustion) is the phenomenon by which a combustible mixture of fuel and oxidizer reacts in a self-accelerating manner, and eventually explosively, leading towards fully fledged combustion, such as in the form of a flame. Premixed or homogeneous autoignition occurs when the fuel and oxidizer have been previously mixed, down to the molecular level. Conversely, non-premixed or inhomogeneous autoignition occurs when the, initially separated, (usually colder) fuel and oxidizer (usually hot air) mix and react simultaneously, either in a stagnant field, or in a flow field that can be laminar or in most applications turbulent. Ultimately, inhomogeneous autoignition is a result of both (slow pre-ignition) chemical reactions, the fluid mechanical processes (including potentially the turbulence) that bring the reactants and the 'radical pool' necessary for reaction together (or the transport of mass of reactants and reacting species), and the thermodynamical fate of the heat that is released from these reactions (or the transport of the heat that results from the reactions). By definition, the overall (or global) autoignition chemistry is exothermic, with most of the important elementary reactions also being exothermic. However, some key elementary reactions can be endothermic depending on the fuel-oxidizer system and conditions. The temporal and spatial location of the resulting autoignition 'kernels', where autoignition first appears, will be determined by the struggle and interplay between the aforementioned processes.

In order to begin to understand the fundamentals of the theory of this phenomenon, an elementary Autoignition Tutorial has been prepared. In addition, you will find further information split into two pages, one on each of the following two relevant topics: Chemistry, and, Mixing (or Heat and Mass Transfer), by clicking on these two links. After you have explored these (or indeed, if you are aware of the fundamentals already), you can return to this page and examine our work and some of our results that are outlined in the following two sections. For more information and the latest updates contact One can also see the Members page to contact specific people involved in this wider project.


Autoignition in Turbulent Flows: An Experimental Investigation

It has been mentioned that although the problem of autoignition has been approached from many angles, the effect of turbulent mixing on autoignition chemistry and the statistical nature of the random appearance of the phenomena in the turbulent inhomogeneous case remain unclear. This situation is compounded by the facts that, on the one hand, the most important underlying processes of this phenomenon, namely oxidation kinetics and turbulence, cannot be said to be closed problems in themselves, and on the other, in most applications of interest autoignition occurs in situations where the two processes cannot be decoupled and in which the final outcome is a compounded result of so-called 'chemistry-turbulence interactions', or placed more simply, of their 'mutual interference'. Accurate predictions of autoignition can only be made for a limited (and special) number of cases.

This document presents results from an experimental effort to observe autoignition in a turbulent inhomogeneous flow of a parabolic nature. It describes experiments in which various pure, or nitrogen-diluted, gaseous fuels were injected concentrically and continuously into a co-flow of preheated, turbulent air. Hydrogen, acetylene (ethyne), ethylene (ethene) and prevaporized n-heptane have been used. With the exception of the more complex situation of evaporation, mixing and autoignition of liquid fuel sprays that are not considered here, this study has been specifically designed to consider the problem from the most generalized point of view; that in which the phenomenon occurs as a consequence of turbulent mixing, real chemistry and such that the chemical and turbulent timescales are of the same order (and hence equally important).

It will be shown that a variety of autoignition phenomena are possible, as are related phenomena such as post-ignition flame propagation and flashback or extinction, that are not necessarily the same for all fuels. Furthermore, it will be demonstrated that for a certain, relatively wide range of conditions, autoignition appears in the form of random events, or 'spots', that manifest both visually and acoustically in a well-defined manner. It will be shown that the randomness of these spots cannot be neglected and that for a complete description both the mean behaviour and statistical fluctuations must be considered. In this study autoignition will been examined in terms of:

1. The mean location of its appearance relative to injection, as well as an estimated mean residence/delay time associated with this location. The effects of the choice of fuel, the degree of fuel dilution with nitrogen, the temperature and velocity, and of the flow geometry will be examined, with specific reference to the discrepancy between the DNS and counterflow experiments concerning the effect of turbulent mixing on autoignition chemistry.

2. The randomness, or spread, in the location and delay time of autoignition.

3. Topological and spatial details of the manifestation of the instantaneous, explosive autoignition spots, as well as details of the temporal evolution of the resulting autoignition spots into propagating flames and the fate of these post-ignition flames.

4. The frequency of appearance of the random spots and its relation to the autoignition location.

5. The spectral and acoustic signature of the spots. The former will be investigated with spectroscopy and the latter with acoustic measurements.

6. Possible differences in the overall behaviour between the various fuels.

To the best knowledge of the author experiments investigating these aspects of autoignition have not been previously reported in the literature. Similar configurations have been explored, but in those experiments the phenomena were either additionally affected by such issues as the two-phase physics of liquid fuel spray evaporation and/or were done in more complex and less well-known mixing fields than the steady, axisymmetric, plume-in-co-flow chosen in this work. More importantly, all earlier studies reported a delay time only, with no reference to the location of autoignition relative to the turbulent mixing field, or a proper characterization of this field to ascertain any effects of relevant mixing quantities on the phenomena.

The statistics of autoignition have not been previously explored experimentally and could be the outcome of calculations, such as those by Probability Density Function (PDF) or Large Eddy Simulation (LES) methods. Therefore the present data can serve as an ideal test-bed for the validation of advanced Computational Fluid Dynamic (CFD) and turbulent reacting flow models. Hence, in addition to the previously mentioned points and from a more practical point of view, the work that is reported in this document aims to:

1. Provide well-characterized data of autoignition lengths and delay times for various fuels in the presence of velocity fluctuations and mixture inhomogeneities for the purposes of modelling.

2. Uncover any connection between the ensemble-mean behaviour (e.g. the average ignition timing measured from many cycles in a diesel engine) and the possibility of relatively rare events causing dangerous autoignition (e.g. in a gas turbine premix duct).

3. Provide information on the chemiluminescence emissions and sound characteristics of phenomena that arise.

A novel and very promising approach for the exploration of the phenomenon of turbulent autoignition is based on a concentric, non-premixed combustion configuration. Hot air and cold fuel are supplied separately and allowed to mix due to turbulence as they are advected in the flow-field. The interest lies mostly in the effects of turbulence on the topological features (location, structure and propagation mechanisms) of the autoignition spots. The ultimate goal is to be able to predict the spatial and temporal characteristics (location, spread and subsequent development) of autoignition in the turbulent mixing flow field, given the characteristics of this field.

We choose a simple background flow field: turbulent flow, downstream of a grid confined in a pipe. For the purposes of the current study the exploration of the sensitivity of autoignition to the aforementioned parameters is best attempted in conditions for which the chemical time scales are of the same order, or close to the order of, the fluid-mechanical time scales. The direct effect of turbulent mixing is most significant for the non-uniformities of non-premixed autoignition, but the effect of partial premixing is something also worth examining. Confirmation of the results obtained from DNS can only come from experiment. Experiments on the effects of turbulence on autoignition are virtually non-existent.

This experimental configuration has been used successfully to observe the phenomenon and preliminary results show that, indeed, the turbulent flow field must strongly influence autoignition, especially the spread of the autoignition sites. In the autoignition experiments, the sites are clearly seen to exhibit a wide-spread randomness in location, even though the bulk flow field is uniform and steady.

It is considered that the project has advanced significantly over the course of last four years. Follow the link to the Archive page to see papers, posters, presentations and other documents related to these experiments. The experiments mentioned there were performed between September 2001 and October 2004.

Schematic Diagram of the Apparatus:

In the figure directly below, a qualitative and not-to-scale representation of the main experimental layout can be seen. With no fuel injection and in normal, high temperature operation, hot air flows non-reactively through the inner tube in a vertically upwards direction. By hovering your mouse over the image you can see the resulting mixing pattern, i.e. turbulent mixing plumes, formed by fuel being continuously injected via the injector at the bottom end of the tube, into the hot air. Under certain favourable conditions, autoignition occurs inside the tube, downstream of the injector. Typical examples of resulting autoignition spots are shown. By clicking on the image you can also see a photograph of the apparatus.


Autoignition Visualization:

By clicking on the image links below you can watch movies of typical examples of turbulent, non-premixed autoignition in a pipe flow, as observed in this experimental apparatus. The first three animations show the apparatus in its initial form, which utilized a ceramic, insulating blanket wrapped around the quartz tube through which the hot air flowed, instead of a two-tube vacuum insulated 'jacketed' design. The white 'object' in the video is this blanket. A vertical slit was made through which we can see the autoignition spots, which appear as short 'spotty flashes', running down the centre of the camera's optical field. The fuel/nitrogen mixture injector is located approximately at the lowest point of the slit. In the fourth video (with audio), the current state of the apparatus is demonstrated in operation. To improve the viewing resolution during playback try decreasing the window size of your browser. The autoignition spots can be heard, with each 'spotty flash' being associated with a 'popping' sound. You can e-mail me ( for a better quality version of this (or any other) video.

Acetylene (Ethyne, C2H2):

1st Animation 2nd Animation 3rd Animation


Our very first images of autoignition in the original apparatus.
(1): Image sequencing including an instance of 'Flashback' (near the end) with acetylene in the default apparatus geometry.
(2): Image sequencing of the 'Random Spots' autoignition behaviour of acetylene in the default apparatus geometry and at 'slow spotting' conditions, i.e. low temperatures and higher velocities (both in the co-flow and of injection).
(3): Image sequencing of the 'Random Spots' autoignition behaviour of acetylene in the default apparatus geometry.

4th Animation


(4): A mini-tour of the experiment running in the new apparatus, containing a number of samples of acetylene autoignition as the conditions vary in the default geometry, with audio. Including an instance of 'Flashback' (near the end).

5th Animation 6th Animation 7th Animation


(5): High-speed capture of single autoignition 'spot' realization advected by the mean flow. Autoignition kernel initiation, propagation and extinction.
(6): High-speed, enlarged image of autoignition kernel initiation, flame front propagation and extinction.
(7): Lower frame speed, low exposure time autoignition spots.

8th Animation 9th Animation


(8) and (9): Unsteady 'Random Spots' behaviour with acetylene. Background conditions are steady as before, and fuel flow is constant.

Normal Heptane (n-C7H16):

10th Animation


(10): High-speed movie of autoignition of n-heptane 'Random Spots' autoignition in the default apparatus.

Acetylene (Ethene, C2H2) with Small Injector and Ethylene (Ethene, C2H4) with Bluff Body:

11th Animation 12th Animation


(11): High-speed movie of autoignition of acetylene in 'Random Spots' with a smaller injector. Comparison of two conditions, 'slow spotting' on the left at low temperature and high velocity, and, 'fast spotting' on the right at high temperature and low velocity.
(12): Unsteady 'Spot-Wake Interactions' autoignition behaviour of ethylene with bluff-body.


Closure: Conclusions from this Work

Experiments of turbulent, inhomogeneous autoignition have been carried out in concentric, co-flowing fuel plumes confined in a well-insulated quartz tube, formed by the continuous injection of a pure or nitrogen-diluted fuel into a uniform, turbulent, preheated co-flow of air, without (referred to as CTHC burner) and with (referred to as CTHAJ burner) a 45o bluff-body at the end of the injector. Hydrogen, acetylene, ethylene and n-heptane have been used as fuels. Two injector nozzles with different inner diameters were used, as were two bluff-body/quartz tube combinations. In addition, for certain experiments the quartz tube was further insulated by placing a heat exchanger over its length in order to ascertain any effect of heat losses. The conditions of the experiments were such that the autoignition, or chemical, timescales were of the order of the outer (integral) turbulent timescales.

The background co-flow has been inspected for and found to be uniform in terms of the mean and turbulent fluctuations of the velocity (at cold conditions) and the mean and fluctuations of temperature. The heat losses from the tube were measured, both by axial and radial profiles of the mean temperature. It was found that the presence of the heat exchanger halved the axial decay of mean temperature in the tube, but no considerable differences in the results concerning autoignition were found. The fluctuations of temperature were found to be small. Uniform profiles of mean velocity were also obtained at hot conditions, as was the magnitude of the rms of the turbulent fluctuations of the velocity. The turbulent character of the velocity field was inspected further, by measuring the integral and Kolmogorov lengthscales and by compiling pdfs and power spectra of the velocity fluctuations (at cold conditions). The measured Kolmogorov lengthscales from the turbulent dissipation were close to the estimates of this quantity from the integral lengthscale. Both the integral timescale and Kolmogorov lengthscale were found to be reasonably uniform in the radial direction and to increase slightly in the axial direction. The pdfs and power spectra were as expected for a confined pipe flow, at regions downstream of a turbulence generating grid. Pdfs and power spectra of the temperature fluctuations showed that these were small and well characterized by a random process, suggesting that the temperature at the inlet is steady and not turbulent. This was confirmed by measurements of the cross-correlation of the fluctuations of velocity and temperature, which were also found to be small.

The mixing field was measured by acetone PLIF at low (but not ambient) temperatures. It was found that, for injection of fuel with equal bulk velocity as that of the air, the mean mixing patterns resembled well the behaviour expected from the analytic treatment of diffusion from a point source. The radial profiles of mean mixture fraction were Gaussian, scaling as expected with downstream distance and bulk velocity. The axial profiles tended to a '-2' power-law decay downstream of the (finite sized) injector and were unchanged by changes to the bulk velocities. The variance of the mixture fraction and mean scalar dissipation rate were also considered and these were found to be well related by a timescale ratio of '2.0' along the centreline, except close to the injector. The effect of increasing the air bulk velocity or the non-dimensional fuel injection velocity, was to increase both the mixture fraction variance and mean scalar dissipation. The effect of increasing the non-dimensional fuel injection velocity was to shift the mixing patterns (mean and fluctuations of the mixture fraction and mean scalar dissipation rate) downstream. Detailed considerations of the point statistics of these variables revealed that, for the equal velocity case, the probability of having a most reactive mixture rises from zero at the injector to a maximum value and decays downstream. The effect of increasing the bulk velocities, or a richer choice of most reactive mixture, is to decrease the probability of having the most reactive mixture. The mean conditional scalar dissipation rate at the most reactive mixture decays monotonically downstream. An increase in air velocity increases the unconditional and conditional mean of the scalar dissipation rate at the same location and decreases the probability of having low conditional scalar dissipation rate at the most reactive mixture.

Autoignition was achieved in the tube with all fuels and in all configurations. Depending on the inlet velocity and temperature of the air and fuel and the degree of dilution of the fuel stream with nitrogen, a variety of phenomena have been observed in the CTHC that have been grouped into operation regimes, namely: 'No Ignition', 'Random Spots', 'Flashback' and 'Lifted Flame'. In all cases autoignition occurred followed by inevitable post-ignition flames that propagated in the tube. The conditions for which transition occurred between regimes was briefly investigated. Generally, 'No Ignition' was observed at lower temperature and higher velocities, 'Flashback' was possible at higher temperatures and lower velocities and 'Lifted Flames' at high temperatures and high velocities. In the 'Random Spots' regime autoignition manifested visually in the form of repeated 'spotty flashes', each one of which was audibly associated with a 'popping' sound. In addition, complex behaviour was possible in the 'Random Spots' regime in the CTHAJ that involved unsteady, igniting and extinguishing pulsed-combustor type lifted flames, perhaps due to interactions of autoignition with acoustics or the turbulent flow, but more probably due to a strong interaction between autoignition and the recirculation zone in bluff-body wake and organized shear flow in the annular jet.

In the 'Random Spots' regime autoignition spots appeared both as a consequence of independent events of autoignition and from events which were caused by propagating flames originating from earlier events elsewhere in the flow. Due to the extinction of the post-ignition flamelets in relative closeness to the original autoignition sites, autoignition was inspected in the 'Random Spots' regime, in terms of chemiluminescence emissions, frequency of occurrence, acoustics, location and an estimate of the autoignition delay time from injection. From the study of chemiluminescence emissions from the autoigniting regions in the flow, the hydroxyl (OH) radical was chosen as a suitable candidate for optical measurements for all fuels and both in the CTHC and CTHAJ.

In the CTHC the autoignition lengths decreased with increasing temperature, decreasing air velocity, decreasing non-dimensional fuel injection velocity and decreasing dilution (except for hydrogen that was not affected in the tested envelope of conditions and n-heptane that showed a non-linearity). A decrease was also observed for autoignition with the smaller diameter injector nozzles and for injection farther downstream of the grid (i.e. at lower turbulence levels). With the exception of small injector results concerning the effect of air velocity, all aforementioned conclusions were unchanged even when viewed in terms of an estimated mean residence time until autoignition (autoignition delay). Specifically, the autoignition delay time decreased with increasing temperature, decreasing air velocity, decreasing non-dimensional fuel injection velocity and decreasing dilution. It also decreased for autoignition with the smaller injector and for injection farther downstream of the grid. With the small injectors, the air velocity was found not to affect the autoignition delay time. In the CTHAJ, the autoignition lengths decreased with increasing temperature and decreasing Reynolds number. The effect of the non-dimensional fuel injection velocity was not found to be significant for the tested conditions.

In the CTHC the mean frequency of occurrence of autoignition was found to decrease exponentially with increasing autoignition length from the injector and considerable spread was observed in the autoignition locations with all conditions fixed. The randomness in the occurrence of autoignition, both temporally but also spatially, cannot be completely explained in terms of the (minimal) randomness in the initial temperature, but must be attributed to the randomness in the scalar mixing field, including the turbulent dispersion of the scalar, and possibly, the variation in the conditional scalar dissipation.

The most important conclusions from this work are that:
(1) Most of these phenomena can be explained in terms of mixing and simple chemistry and that, in general, turbulent, inhomogeneous autoignition cannot be predicted from chemical delay times obtained from homogeneous experiments. The ability to understand and predict the majority of autoignition phenomena in these flows does not rest primarily on an improved understanding of the complex chemical kinetics of autoignition chemistry, but rather on a deeper knowledge surrounding the effect of turbulent mixing on the pre-ignition chemistry.
(2) The effect of enhanced turbulence (increased fluctuations of the velocity) is to cause a decrease in the probability of having low conditional scalar dissipation rates at the most reactive mixture, which in turn causes a delay in autoignition. From a broader point of view, both these experiments and previous DNS studies agree on the basic finding that:
(a) For inhomogeneous autoignition, the effect of turbulence in determining the eventual emergence of autoignition can be understood in terms of pdf(ξ = ξMR) and the history of χ|ξ = ξMR, and that,
(b) The randomness in the location of autoignition can be understood in terms of the spatial randomness of the regions containing finite pdf(ξ = ξMR), and perhaps, of χ|ξ = ξMR.

Back to Top

Horizontal Line

Home | Autoignition | Archive | Members | Search | Contact Information

For problems or questions regarding this web contact

Last updated: Wednesday, 29 March, 2006 02:06 AM
Christos N. Markides

This page was checked and found to be valid HTML 4.01 Transitional. Validated at:

Valid HTML 4.01 Transitional