Sei sulla pagina 1di 52

ARTICLE IN PRESS

Progress in Energy and Combustion Science 34 (2008) 499–550


www.elsevier.com/locate/pecs

Flame acceleration and transition to detonation in ducts


G. Ciccarellia,, S. Dorofeevb
a
Queen’s University, 130 Stuart Street, Kingston, Ontario, Canada K7P 2M4
b
FM Global, 1151 Boston-Providence Turnpike, P.O. Box 910, Norwood, MA 02062, USA
Received 15 May 2007; accepted 13 November 2007
Available online 15 February 2008

Abstract

This paper reviews the state of knowledge on flame acceleration and deflagration-to-detonation transition (DDT) in smooth ducts and
ducts equipped with turbulence-producing obstacles. The objective is to bring to light the basic understanding of the phenomenon and its
application to explosion safety. The scope of the review is restricted to homogeneous gas-phase combustion with emphasis placed on
experimental investigation.
r 2007 Elsevier Ltd. All rights reserved.

Keywords: DDT; Flame; Detonation; Shock wave

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 500
1.1. Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 502
1.2. One-dimensional combustion waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 503
1.2.1. Steady detonation and deflagration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504
1.2.2. Unsteady double discontinuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 505
1.2.3. One-dimensional description of flame acceleration and DDT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 506
2. Deflagration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
2.1. Laminar flames. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 507
2.2. Flame instabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
2.2.1. Landau–Darrieus instability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
2.2.2. Diffusive instabilities. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
2.2.3. Cellular flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
2.2.4. Instabilities and flame stretch. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
2.2.5. Instabilities promoted by confinement and obstructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 510
2.3. Turbulent flames. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 511
2.3.1. Scales of turbulence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 511
2.3.2. Characteristic regimes of turbulent combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 512
2.3.3. Turbulent burning velocities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 513
2.3.4. Quenching of turbulent flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
3. Detonation waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
3.1. Historical review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
3.1.1. ZND detonation wave structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514
3.1.2. Multi-head detonation wave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515

Corresponding author. Tel.: +1 613 533 2586; fax: +1 613 533 6489.
E-mail address: ciccarel@me.queensu.ca (G. Ciccarelli).

0360-1285/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.pecs.2007.11.002
ARTICLE IN PRESS
500 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

3.1.3. Single-head spin detonation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518


3.2. Cellular detonation wave structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 518
3.2.1. Critical initiation energy and detonation cell size measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 520
3.2.2. Cell size correlations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
3.2.3. Critical tube diameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 522
4. Deflagration-to-detonation transition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
4.1. Phenomenology of DDT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
4.1.1. Basic studies of DDT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 523
4.1.2. Phases of DDT process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
4.2. FA in smooth tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
4.2.1. Process of FA in smooth tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
4.2.2. Run-up distances to supersonic flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 525
4.3. FA in obstructed channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
4.3.1. Process of FA in obstructed channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
4.3.2. Characteristic flame propagation regimes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529
4.3.3. Flame speeds as a function of distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 532
4.3.4. Criteria for weak and strong FA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533
4.3.5. Run-up distances to supersonic flames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 536
4.4. Onset of detonations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
4.4.1. Types of detonation onset phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 537
4.4.2. Underlying mechanism Shock Wave Amplification by Coherent Energy Release (SWACER) . . . . . . . . . . . 538
4.4.3. Criteria for onset of detonations in smooth tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 541
4.4.4. Criteria for onset of detonations in channels with obstacles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 542
4.5. Evaluation of potential for FA and DDT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 544
5. Concluding remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546

1. Introduction Explosions almost universally start by the ignition of a


flame from either an electrical spark, or the autoignition of
In most engineering applications, combustion occurs via the mixture in contact with a very hot surface. Under
deflagration where the molecular transport of heat and certain conditions, the flame can accelerate and undergo
mass are important. Deflagration represents the subsonic transition to a detonation wave. Collectively this process is
mode of combustion and the chemical reactions occur at often referred to as deflagration-to-detonation transition
roughly constant pressure. For stoichiometric gaseous (DDT). Strictly speaking the flame acceleration phenom-
fuel–air mixtures at atmospheric conditions the laminar enon leading up to detonation initiation is separate from
burning velocity is typically less than 1 m/s. For turbulent the actual process of the onset of detonation. The onset of
flames the transport properties are enhanced and as a result detonation is a local phenomenon that occurs in a very
the burning velocity can be significantly higher than the small volume of the unburned mixture whose thermo-
laminar value. The detonative mode of combustion is dynamic state has been conditioned by the flame accelera-
characterized by supersonic front propagation velocities on tion processes that take place over a much larger length
the order of a couple of thousand meters per second. In this and time scale. Under this DDT scenario the study of the
mode of combustion, chemical reactions occur as a result flame acceleration process is as important as the actual
of adiabatic shock heating and therefore the shock wave onset of the detonation event.
and the trailing reaction zone are coupled and propagate at There have been significant advances made over the
a constant velocity. For stoichiometric fuel–air mixtures years in the understanding of the flame acceleration and
the pressure ratio across the detonation wave is in the DDT processes that are outlined in several excellent review
range of 15–20. This is roughly twice the maximum papers [1–5]. Thanks to high spatial and temporal
possible pressure produced by a deflagration in the same resolution Schlieren photography the basic mechanisms
mixture under adiabatic, constant volume conditions. involved in flame acceleration in an obstructed channel
Because of the very high overpressures associated with a have been identified and the basic DDT process is fairly
detonation wave, historically the study of detonation wave well understood. Turbulent combustion plays a key role in
phenomena has been motivated by industries prone to the flame acceleration process, which to this day remains
uncontrolled explosions. Although a detonation wave can one of the most intensively studied but least understood
be initiated directly by the deposition of a large amount of phenomenon in combustion. The problem is made even
energy in a very small volume of the mixture, typically such more difficult since compressibility effects play an im-
ignition sources are not present in most industrial settings. portant role in the latter stage of the flame acceleration and
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 501

Nomenclature n1 reaction order of limiting component


nF reaction order for fuel
Latin nO reaction order for oxidant
P pressure
A ¼ l/DR proportionality constant P1 pre-shock pressure
AF elementary area of a flame front P2 post-shock pressure
a parameter of approximation P3 pressure behind reaction front
a1 pre-shock sound speed P pressure
a2 post-shock sound speed Pe Peclet number
a3 sound speed behind reaction front Q chemical energy release
ap sound speed in combustion products R gas constant
ar sound speed in reactants RF flame radius
b parameter of approximation ReT ¼ u0 LT/n ¼ (u0 /SL)(LT/d) turbulent Reynolds
b1, b2, b3, b4, and b5 correlation constants number
BR blockage ratio S obstacle spacing
C constant coefficient S1 second largest of transverse sizes of the volume
cp specific heat capacity at constant pressure confined between two consecutive obstacles
D tube diameter SL laminar burning velocity relative to unburned
Da ¼ (u0 /SL)1(LT/d1) turbulent Damkoehler number mixture
DL molecular diffusivity of limiting component SLeff effective laminar burning velocity relative to
DCJ Chapman-Jouguet detonation velocity unburned mixture
Djet orifice size of a turbulent jet Sn normal propagation velocity of curved flame
Dlim limiting tube diameter for detonation propaga- ST turbulent burning velocity relative to unburned
tion mixture
Dsp propagation velocity of spontaneous reaction T temperature
front T1 pre-shock temperature
d orifice diameter in an obstacle, size of unob- T2 post-shock temperature
structed passage T3 temperature behind reaction front
dc critical tube diameter Tb maximum flame temperature
Ea effective activation energy Tu initial mixture temperature
Ec critical initiation energy of detonation tK Kolmogorov time scale of turbulence
F function as defined tL integral time scale of turbulence
H enthalpy U shock velocity
H1 largest of transverse sizes of the volume u particle velocity
confined between two consecutive obstacles u1 pre-shock particle velocity
h wall roughness u2 post-shock particle velocity
K constant coefficient u3 particle velocity behind reaction front
Ka ¼ (u0 /SL)3/2(LT/d1)1/2 turbulent Karlovitz number ū mean flow speed
k turbulent kinetic energy u0 turbulent fluctuation velocity defined by inte-
kA pre-exponential factor in reaction rate constant gral scale
L characteristic geometrical size V flow speed
Lc detonation cell length W deflagration front velocity in double-discontinuity
Li induction length model
LM Markstein length X, x distance
LT integral length scale of turbulence XD run-up distance to detonation
Le Lewis number XS flame propagation distance where the flame
lK Kolmogorov length scale of turbulence speed reaches the sound speed in the combus-
Ma Markstein number tion products
Mab Markstein number defined relative to burned Y fraction of reactants in mixture of products and
mixture reactants
NF molar concentration of fuel
m parameter of approximation Greek
NL molar concentration of limiting component
NO molar concentration of oxidant a flame stretch rate
n overall order of reaction ac flame stretch rate due to curvature
ARTICLE IN PRESS
502 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

as flame strain rate due to inhomogeneities of the y dimensionless temperature


flow r density
b ¼ Ea(TbTu)/RTb2 Zeldovich number r1 pre-shock density
e Rdissipation rate of turbulent energy r2 post-shock density
1
Gnþ1  0 xn ex dx gamma function r3 density behind reaction front
g specific heat ratio rb density of combustion products
Z parameter of approximation ru density of reactants
D thickness of boundary layer s ratio of densities of reactants and products
DZND ZND induction length (expansion ratio)
d laminar flame thickness s* critical expansion ratio for effective FA
k constant coefficient x dummy variable of integration
l detonation cell width t ¼ d/SL characteristic chemical time scale
m constant coefficient tCJ induction time corresponding to the CJ detona-
n kinematic viscosity tion post-shock state
nF stoichiometric coefficient of fuel ti characteristic induction time
nL stoichiometric coefficient of limiting compo- tr characteristic reaction time
nent w temperature diffusivity
nO stoichiometric coefficient of oxidizer

DDT processes. Commercial computer programs com- conducive, the deflagration can transition to a detonation
monly used for explosion safety applications are based on wave.
Reynolds time-averaged conservation equations and simple In the chemical industry there are processes where
combustion models. These codes are missing the essential combustible gas mixture is transported and processed in
physics that are responsible for flame acceleration and piping and reactors. An example of such a process would
DDT and therefore can only be used for qualitative be one that relies on the partial-oxidation reaction. For
information. Computer hardware limitations make direct cost reasons, industrial process piping is typically designed
numerical simulation impossible at this time because of the to withstand moderate over-pressurisation and cata-
wide spectrum of time- and length-scales associated with strophic failure is mitigated by the implementation of
the chemistry and turbulent flow involved in the flame explosion venting devices [7]. Much like a mine shaft, the
acceleration phenomenon. Because of the limitations of piping and reactor vessels provide the lateral confinement
computational fluid dynamics in simulating this complex that is necessary for flame acceleration to occur and thus
phenomenon, experts in the field typically rely on experi- the potential for DDT. The explosion venting devices
ment-based correlations for quantitative predictions. This typically used are only effective for slow deflagration
paper presents the state-of-the-art in the field with phenomenon that occurs in low aspect ratio chambers
emphasis placed on experimental research with application under quiescent conditions. These devices are not effective
to industrial explosion safety. for limiting the rapid pressure rise associated with fast
flames and detonation waves. As a result, the chemical
1.1. Applications process industry has been involved for many years, directly
or indirectly, in the study of the DDT phenomenon from
The study of detonation phenomena goes back many the initiation prevention perspective. For example, for
years to the pioneering work of Mallard and Le Chatelier explosion prevention it is imperative to know the fuel
[6]. Working at the School of Mines clearly their interest in composition limits for flame acceleration and detonation
combustion stemmed from the problem of uncontrolled initiation and propagation for a given duct size.
explosions in coal mines that were very prevalent at the Following the accident at the Three Mile Island Unit 2
time. Unfortunately, over a century later, the problem of (TMI-2) nuclear power plant near Middletown, Pennsyl-
coal mine explosions remains a problem for the vania, on March 28, 1979 there was a surge in explosion
industry. In such explosions air in the mine shaft mixes prevention research, specifically involving hydrogen gas.
with methane gas, as well as small amounts of other During the accident, there was a partial meltdown of the
hydrocarbon volatiles that are released from the surround- reactor core that resulted in the production and release of a
ing coal. The accidental ignition of the combustible gas large amount of hydrogen gas into the containment
mixture results in a flame that accelerates down the mine building [8]. Hydrogen gas is produced as a result of
shaft producing damaging overpressures. Coal mine Zircaloy fuel clad oxidation that takes place at elevated
explosion phenomenon is further complicated by the temperatures in a steam atmosphere. During the accident
lofting of coal dust from the mine floor into the unburned several pressure spikes were recorded in the containment
gas ahead of the flame. In the worst case scenario, where building that were attributed to the combustion of the
the shaft size and the combustible mixture conditions are released hydrogen that mixed with the containment air
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 503

atmosphere. Fortunately, the pressure rise was not Environmental as well as economic concerns associated
significant enough to compromise the integrity of the with the use of hydrocarbon based fuels for transportation
power plant containment. However, subsequent accident and power generation has generated significant interest in
investigations indicated that if DDT were to occur the the use of hydrogen as an alternative energy carrier.
integrity of the containment could be jeopardized. This Hydrogen combustion does not produce any greenhouse
prompted many countries with nuclear power plants to gases, such as carbon dioxide, that are responsible for local
initiate intensive research programs investigating hydrogen and global environmental concerns. Because of these two
detonation related phenomena. The US Nuclear Regula- factors there is currently significant interest in the
tory Commission initiated such a program with large-scale introduction of hydrogen as an energy carrier of the
detonation studies performed at Sandia National Labora- future, and as a result the issue of hydrogen safety is once
tory. One of the most important objectives of the research again at the forefront. The main problem is that hydrogen
was to determine the minimum hydrogen concentration has very wide flammability concentration limits, e.g.,
required to support a detonation on a scale approaching 5–75% by volume in air. Any leakage of hydrogen into a
that of a nuclear power plant and its subcompartments. confined space frequented by motor vehicles, such as
The most vulnerable power plants have since been retro- tunnels and garages, poses a potential explosion hazard.
fitted with equipment, such as igniters and hydrogen This issue is further aggravated by the wide detonability
recombiners, that during a severe-accident maintain the limits of hydrogen–air mixtures and the propensity for a
overall containment hydrogen concentration below a flame propagating in a duct to rapidly accelerate due its
threshold level that detonation initiation and propagation very high laminar burning velocity. A review of the
is not possible. The hydrogen problem remains an hydrogen safety issue and current research can be found
important issue to the nuclear industry worldwide. The at the European Community funded Network of excellence
international Organization for Economic Co-operation HySafe web site (www.hysafe.org) and at the US Depart-
and Development (OECD) Nuclear Energy Agency ment of Energy Hydrogen Program web site (www.hydrogen.
(NEA) recently published a State-of-the-art-report on energy.gov).
flame acceleration and DDT for the nuclear industry [9].
The study of DDT has not been exclusively associated 1.2. One-dimensional combustion waves
with explosion safety. Over the last decade there has been a
renewed interest in using the detonation mode of combus- Up until the end of the 1950s detonations and
tion for propulsion purposes. The concept of a pulse deflagrations were treated theoretically as one-dimensional
detonation engine (PDE) in its simplest form involves the (1D) discontinuities. Since the flow associated with each
initiation of a detonation wave in a tube combustor with discontinuity was considered steady, the processes involved
the ignition-end closed and the opposite end open. In an could be analyzed by simple algebraic equations. Today it
ideal PDE the detonation wave is directly initiated and is well known that both detonation waves and deflagra-
propagates to the open end of the tube producing roughly a tions are locally multi-dimensional and unsteady phenom-
constant volume pressure at the ignition-end. This high ena. However, for propagation inside of a duct, if the
pressure eventually decays as the combustion products control volume is chosen such that the inlet and outlet
expand out the open-end of the tube. Therefore, the high- boundaries of the control volume are sufficiently far from
pressure acting on the closed end of the tube produces the front, they can be treated as 1D. Consider the steady
thrust for a finite duration. In order to get quasi-steady 1D control volume shown in Fig. 1 where the gas inflow
thrust, multiple tubes firing at a frequency of roughly and outflow are designated as states 1 and 3, respectively.
100 Hz is required. The key advantage of a PDE over a Based on this steady 1D control volume the inviscid
ramjet is that thrust can be produced at zero approach conservation of mass, momentum and energy are
speed up to hypersonic speeds. The proponents of PDEs
also claim better fuel consumption than both gas turbines r1 u1 ¼ r3 u3 , (1.1)
and ramjets over a wide Mach number range. There are
several technical challenges associated with PDEs, includ- P1 þ r1 u21 ¼ P3 þ r3 u23 , (1.2)
ing: overheating, noise, vibration, as well as air and exhaust
gas management and fuel–air mixing issues. Most im- H 1 þ Q þ 12u21 ¼ H 3 þ 12u23 , (1.3)
portantly, for an air-breathing PDE the challenge is how to
initiate a detonation consistently in a prototypic length
combustor without the use of oxygen sensitizing or a high- Combustion front
energy initiation source. One of the strategies investigated
involves using low energy ignition and DDT. Various
approaches have been investigated to minimize the run-up
distance required for the detonation to develop. A 3 1
comprehensive review of the various PDE design concepts
can be found in Roy et al. [10]. Fig. 1. One-dimensional control volume.
ARTICLE IN PRESS
504 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

where r, P, H, u, and Q correspond to density, pressure, 2


sensible enthalpy, particle velocity relative to the stationary Rayleigh Line
control volume and energy release per unit mass, respec-
tively. If we treat the gas as ideal H can be expressed as a P/P1
function of P and r Equilibrium
J
Hugoniot
g P
H¼ ; (1.4)
g1 r
V
where g is the ratio of the specific heats. Note, the control
volume shown in Fig. 1 may also contain a shock wave and 1 P
the same equations apply. Eqs. (1.1)–(1.4) are used in the 1
C
following Sections to analyze detonation waves, deflagra-
tions, and a combination of the two. The analyses serve to
familiarize the non-expert in the field with some of the 1 ρ1/ρ
detonation and DDT terminology used throughout the
Fig. 3. Hugoniot curve showing possible equilibrium end states.
paper, as well as to gain valuable insight into various
aspects of the true unsteady multi-dimensional DDT
phenomenon. and momentum equations. It can be shown that the slope
of the Rayleigh line is proportional to the square of the
1.2.1. Steady detonation and deflagration magnitude of the front velocity. The intersection of the
Consider the 1D detonation wave shown in Fig. 2 Rayleigh line and the Equilibrium Hugoniot curve
consisting of an inert planar shock wave followed by an represents the final equilibrium state for a given front
exothermic reaction zone. The flow velocity is shown in the velocity. Based on the detonation structure shown in Fig. 2,
reference frame of the shock wave. The chemical reactions since a particle first passes through the inert shock wave
are initiated in the mixture due to the elevated temperature before entering the reaction zone, it is common to show the
and pressure condition that exists after the shock wave and post shock state (point 2) that lies on the shock Hugoniot.
thus the deflagration and the shock wave propagate as a Two important states on the equilibrium Hugoniot are
coupled front. The front propagates at a constant velocity designated by the points V and P. The part of the Hugoniot
U with energy released per unit mass Q. The thermo- curve between points V and P are physically unattainable,
dynamic state ahead of the shock, post shock, and at the mathematically these states correspond to imaginary values
end of the reaction zone where chemical equilibrium is for the front velocity. Points on the upper (detonation)
achieved are designated as 1, 2, and 3, respectively. branch of the Hugoniot (above point V) represent super-
A complete derivation of the relevant equations and in sonic combustion where there is an increase in pressure and
depth discussion of the steady 1D analysis of the density across the front. Points on the lower (deflagration)
detonation wave that follows can be found in Glassman branch of the Hugoniot (below point P) represent subsonic
[11] as well as other basic textbooks in combustion. The combustion with a corresponding decrease in pressure and
equilibrium Hugoniot, obtained by combining the con- density across the front. Point V represents the final state
servation of mass and energy equations, represents all for an adiabatic constant volume process and point P
possible states at the end of the reaction zone for a given represents the final state for a constant pressure process.
energy release. The equilibrium Hugoniot curve, typically The Rayleigh line going through point V has an infinite
shown on a P versus 1/r plot, is shown in Fig. 3 with slope and thus represents the process by which the entire
several key states denoted. In the case of an inert shock mixture ignites at the same time, or effectively an infinite
wave, i.e., Q ¼ 0, the Hugoniot curve is shifted down from front velocity. Conversely the Rayleigh line going through
the equilibrium Hugoniot and passes through the initial point P has a zero slope which represents the limit of an
state 1. The Rayleigh lines shown in Fig. 3 between points infinitely slow combustion wave.
1–2 and 1-c, also commonly referred to as the Michelson On the upper branch the Rayleigh line is shown to be
line, are obtained by combining the conservation of mass tangent to the Hugoniot curve at point J. This state
corresponds to the minimum entropy increase across the
Reaction zone Shock wave detonation wave, and also represents the minimum possible
detonation velocity since a Rayleigh line with a smaller
u3 U slope does not intersect the Hugoniot curve. The state
Q represented by the point J is commonly referred to as the
P3 P1
CJ detonation state in recognition of the pioneering work
ρ3 ρ1
of Chapman and Jouguet in the field of detonation waves
[12,13]. Not only is it the minimum possible velocity but
Fig. 2. Structure of one-dimensional detonation wave consisting of shock also at point J the Rayleigh line is tangent to the isentrope
wave followed by reaction zone. that passes through the point, which can be shown to
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 505

indicate that the flow velocity at the end of the reaction tangent to the Hugoniot curve at point C. Similar to point
zone is sonic, i.e., u3=a3, where a is the speed of sound. J, this corresponds to a state where the flow velocity at the
This choked state, commonly referred to as the CJ end of the reaction zone is sonic (u3 ¼ a3). This is referred
condition, makes it possible for a detonation wave to to as the CJ deflagration and represents the maximum
propagate unaffected by any perturbations, that propagate possible velocity for a deflagration wave.
at the local speed of sound, originating downstream of the
reaction zone. For example, a detonation wave propagat- 1.2.2. Unsteady double discontinuity
ing in a tube away from a closed end is unaffected by the Transition from deflagration to detonation involves an
expansion waves that follow behind the wave in order to accelerating flame that produces upstream conditions that
satisfy the zero velocity closed-end boundary condition. are conducive for the onset of detonation. Let us consider
This expansion is centered at the end of the tube and is the unsteady propagation of a flame in a tube where the
commonly referred to as the Taylor expansion fan [14]. For ignition end is closed. After ignition the lateral confinement
a given initial thermodynamic state (point 1) there is a forces the combustion products to expand in the direction
unique CJ detonation wave velocity that is directly of the flame propagation and in doing so sets into motion
proportional to the square-root of the energy released per the unburned gas. The unburned gas motion is produced
unit mass of the mixture: by the action of compression waves that are generated at
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi the flame front and propagate ahead at a velocity equal to
DCJ ffi 2Qðg3 þ 1Þ, (1.5)
the sum of the local particle and acoustic velocities.
where g3 is the ratio of specific heats of the combustion Essentially the flame acts as a porous piston where the
products (state 3 in Fig. 2). Eq. (1.5) is valid for relatively piston velocity is equal to the unburned gas particle
high heat release, so that DCJba1. velocity immediately ahead of the flame. For an accelerat-
Experiments in large smooth tubes have shown that ing flame, the compression waves coalesce to form a shock
indeed self-sustained detonation waves propagate to within wave that not only sets into motion the unburned gas
1–2% of the theoretical CJ detonation velocity. A detailed ahead of the flame but also raises both the pressure and
review of detonation waves is provided in Section 2. temperature. It has been observed experimentally that
For a detonation velocity larger than the CJ value, the before the onset of DDT in a smooth tube a turbulent
Rayleigh line intersects the Hugoniot curve at two points, flame is always preceded by a shock wave.
one above point J and one between points J and V. Only The above mentioned universal shock–flame complex
the point above J is realizable for a fluid where reaction is can be easily analyzed if treated as a 1D phenomenon.
caused by a lead shock, since in order to reach the lower Shown in Fig. 4 is the classical unsteady double-disconti-
state a particle must pass through the upper point first nuity structure where the deflagration is moving at a
where all the energy available is released. A detonation constant velocity, W, preceded by a shock wave moving at
with an end state that lies above point J on the Hugoniot is a constant velocity, U.
commonly referred to as being overdriven. For points This double-discontinuity problem has been analyzed
above point J the flow at the end of the reaction zone is thoroughly over the years by many investigators [15–17].
subsonic (u3oa3), therefore in order for such a steady-state For initial conditions the unburned gas ahead of the shock
detonation wave to exist it must be supported from behind is taken to be at rest and the properties of the unburned gas
by a piston traveling at a velocity of u3. bounded by the shock wave and the deflagration are
A state between points J and V may be achieved, if a considered to be uniform. For subsonic deflagration
sequence of consecutive ignitions imposes the reaction relative to the burned gas (Woa3) the properties of the
front velocity given by the respective slope of the Rayleigh burned gas bounded by the closed end of the tube and the
line. In this situation, an end state on the Hugoniot curve deflagration products are uniform and at rest. The analysis
between point J and V is achieved from below by a gradual involves solving simultaneously the steady 1D conservation
change of the state along the Rayleigh line starting from equations for the deflagration and the inert shock wave
point 1. Such a wave is usually referred to as underdriven along with the perfect gas equation of state. The derivation
detonation. A transient analog of such an underdriven of the relevant algebraic equations will not be reproduced
detonation wave may be observed during detonation here but can be found in [16,17]. A sample calculation is
initiation via the induction time gradient mechanism
discussed in Section 4.4.2.
Deflagration Shock wave
Recall point P on the Hugoniot does not represent the
end state of a physically possible process but a point just
u3 u2 U
below point P on the Hugoniot curve represents the end u1 = 0
state for a laminar flame, which has a finite but very small q W P2
pressure drop across it and a very slow propagation P3, ρ3, a3 P1, ρ1, a1
ρ2
velocity. For a faster moving steady turbulent flame, the
end state would lie further down the lower branch of the Fig. 4. Double discontinuity model with velocities taken relative to the
Hugoniot curve, the limit being where the Rayleigh line is stationary tube.
ARTICLE IN PRESS
506 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

20 150
18 Equilibrium
100 Hugoniot

P/P1
16 P/P1 J
W=U
14 P2 50 Generalized
12 P3 Hugoniot
0 (Q-curve)
u/a1

10 I
8 W = a3 U
6 u2 1 P
a3 1
4 C
a2
2 u3
0
0 2 4 6 8 10 1 ρ1/ρ
W/a1
Fig. 6. Diagram showing conventional Hugoniot and Q-curve.
Fig. 5. Shock and flame parameters as a function of flame speed for closed
end tube (dashed curves correspond to condition Wu3 ¼ a3).

provided by Adams and Pack [17] for a gas with g ¼ 1.4, with time, i.e., there is an accumulation of mass and energy
Q ¼ 5.9 MJ/kg, r1 ¼ 1 kg/m3. The calculated values for P2, between the discontinuities. As a result of this unsteadiness,
P3, U, u2, a3 and a2 as a function of the normalized the Equilibrium Hugoniot is not capable of describing the
deflagration velocity W/a1 are provided in Fig. 5. It is end state 3 for the double-discontinuity structure. Shown in
important to note that in this analysis it is assumed that W Fig. 6 is the locus of possible end states (curve C–I–J) for
can take on any value. the double-discontinuity structure with the condition
One of the key findings resulting from the analysis is that Wu3 ¼ a3. Oppenheim [15] called this the Q-curve and
in general the shock wave travels faster than the deflagra- Troshin [18] referred to it as the generalized Hugoniot
tion, U4W. A critical flame velocity corresponds to the curve. Note the Q-curve is bounded by the CJ deflagration
condition where the flame velocity equals the speed of and detonation states that also meet this condition. At the
sound in the products (W ¼ a3). For the mixture values point I on the Q-curve the flow velocity immediately after
considered in [17] this condition corresponds to a normal- the deflagration is zero, i.e., W ¼ a3. Therefore, the
ized flame velocity of W/a1 ¼ 4.3. This flame velocity is segment of the Q-curve between points I and J represent
important because for values less than this the combustion the double discontinuity states shown in Fig. 5 between the
products remain at rest (u3 ¼ 0) as there is communication two critical normalized flame velocities 4.3 and 9.1. The
between the flame and the closed end where the gas must be lower segment of the Q-curve represents the end states
stationary. For flame velocities above this critical value an where the burned gas flow is in the opposite direction of
expansion fan must be present to drop the particle velocity flame propagation. This corresponds to the case of ignition
to zero at the closed end. For flame velocities greater than at the open end of a tube. The curve segment P–I in Fig. 6
W/a1 ¼ 4.3 the flame velocity is taken to be W ¼ a3+u3. represents the burned gas state corresponding to the case
This condition implies that the head of the expansion that where the flame propagates at a velocity lower than the
travels at a3+u3 follows directly behind the flame but is sound speed in the unburned gas, i.e., Woa3 [18]. Recall
unable to overtake it. Furthermore, the burned gas flow point P on the equilibrium Hugoniot represents the end
relative to the flame reference frame is sonic, i.e., state for a constant pressure process such as that associated
Wu3 ¼ a3. This is justifiable since heat addition to a with an infinitely slow moving flame.
subsonic compressible flow drives the flow to a choked
state. Note the solid lines in Fig. 5 are obtained assuming 1.2.3. One-dimensional description of flame acceleration and
u3 ¼ 0, which as discussed only applies for W/a1 ¼ 4.3. DDT
The error associated with this incorrect assumption is Troshin [18] asserts that, according to this simplified
represented by the deviation of the solid line from the model, the points on curve P–I–J in Fig. 6 represent the
corresponding dotted line, which takes into account the successive states that occur when flame acceleration leads
combustion product motion, for W/a144.3. to transition to detonation in a smooth tube. Specifically,
A second critical flame velocity corresponds to the case flame acceleration from a laminar flame up to a choked
where the flame and the shock wave propagates at the same flame, sonic relative to the combustion products, proceeds
velocity, i.e., W ¼ U ¼ 9.1a1. This corresponds to the CJ from point P to I. Flame acceleration continues from I to J
detonation state discussed in the context of Figs. 2 and 3. as the choked flame catches up with the preceding shock
Since in general U4W, the distance and thus the mass of wave to form a detonation wave. Although the shock and
the unburned gas between the two discontinuities increases flame velocities are the same at point J it does not mean
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 507

that the flame coalesces with the shock wave [17]. For a flame instabilities, shock–flame interactions as well as flow
detonation wave the chemical reaction is initiated by the shock distortion caused by turbulence. Small-scale turbulence
wave and thus the two are intimately coupled. For the double- also leads to the enhancement of the transport of mass and
discontinuity problem the chemical reactions associated with energy across the flame. These flame effects will be
the deflagration are diffusion driven and the leading shock discussed in detail in Section 2.
wave is supported by the expansion of the combustion
products. Therefore, for transition to detonation to occur 2. Deflagration
there must be a switch in the basic mode of combustion.
As pointed out by Adams and Pack [17] this switch can The processes following weak ignition in a combustible
occur even before state J is attained. As the flame mixture are characterized as deflagrations, which propa-
accelerates the shock wave strengthens and so the post gate at subsonic speed into fresh, unburnt mixture. Flow
shock temperature rises. At any given temperature ignition velocities associated with these flames are initially low
of the mixture will occur after an induction time, ti, where making compressible effects negligible in the burnt and
the natural log of the induction time is inversely propor- unburnt gases. If initial turbulence is not present in the
tional to the temperature via the Arhenius relation flow, the combustion front initiated from weak ignition
 
Ea appears to be initially laminar, and is known as a laminar
ti / exp ; (1.6) flame. Laminar flame propagation is controlled by the
RT 2
molecular transport of energy and species, and is discussed
where Ea is the activation energy and R is the gas constant. in Section 2.1.
This corresponds to the start of chemical reactions at a A freely expanding flame is intrinsically unstable. The
distance of Li ¼ (Uu2)ti behind the shock wave, commonly initially smooth surface of a laminar flame can become
referred to as the induction length. As the shock strength wrinkled due to Landau–Darrieus instability, which can be
increases during flame acceleration and correspondingly ti stabilized or destabilized by the thermal-diffusion effects. This
decreases, a critical point will be reached where the induction can result in cellular structure of a laminar flame. In cases
length equals the distance between the shock wave and the when confinement and/or obstacles are present, Richtmyer–
deflagration. From this point on ignition is triggered by the Meshkov (R–M), Kelvin–Helmholtz (K–H), and acoustic
shock and thus the shock and the deflagration are coupled instabilities may develop resulting in further increase of the
just as in a detonation. flame surface and the reaction rate. Some basic properties of
A breakthrough in detonation research occurred when it flame instabilities are discussed in Section 2.2.
was discovered that the deflagration does not simply merge Turbulence can be generated in the combustion-induced
with the leading shock wave to form a detonation wave, flow. Turbulence increases the surface area of the flame and
but instead a localized explosion occurs somewhere the transport of local mass and energy. Flame interactions
between the deflagration and the leading shock wave that with turbulence result in a variety of turbulent combustion
produces a detonation front that overtakes the original regimes, which will be discussed in Section 2.3. Flame
shock wave. The phenomenon was referred to as the instabilities, its interactions with confinement, obstruc-
‘‘explosion in the explosion’’ by Oppenheim [19] and will be tions, and turbulence may result in Flame Acceleration
discussed in detail in Section 4. (FA) and, under certain conditions, in the transition to
The Q-curve represents the locus of possible end states detonation. These processes are discussed in Section 4.
corresponding to a spectrum of flame velocities. The
mechanism required for the flame to achieve such velocities
2.1. Laminar flames
is not required for the analysis. Each point on the curve
corresponds to a unique flame speed that increases as one
An essential feature of a combustion process is that it is
progresses up the curve. The question is what is the
sustained by exothermic chemical reaction with the
mechanism responsible for the increase in flame velocity?
reaction rate being a strongly increasing function of
Troshin points out that in order for a 1D flame to
temperature. Most combustion systems are characterized
accelerate the burning velocity (u2W) must also increase.
by complicated multi-species chemistry with many elemen-
Using the Adams and Pack [17] example, if we consider the
tary steps. However, as shown in pioneering studies of
case where the flame velocity equals the speed of sound of
Zeldovich, Frank-Kamenetski, and Istratov Ya [20–23]
the burned gas, i.e., W/a1 ¼ 4.3, the unburned gas velocity
basic flame properties may be modeled, at least qualita-
relative to the flame is 0.5a1. Taking the speed of sound of
tively, by a one-step overall reaction:
the stationary unburned gas to be 340 m/s this gives a
burning velocity of 170 m/s. This magnitude burning nF Fuel þ nO Oxidant ¼ Products þ Q, (2.1)
velocity is 10–50 times the normal laminar burning velocity
where Q is heat released, and nF and nO are stoichiometric
for fuel–oxygen mixtures. Such high burning velocities are
coefficients. The reaction rate is governed by an Arrhenius law:
possible because the flame surface area is much larger than
that associated with a planar flame. There are various dN F
¼ kA N nFF N nOO eðE a =RTÞ , (2.2)
mechanisms for flame area enhancement including laminar dt
ARTICLE IN PRESS
508 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

where NF and NO are molar concentrations of fuel and and the main physical concepts to be used in the following
oxidant, nF and nO are orders of reaction for fuel and oxidant, discussion.
kA is pre-exponential factor, Ea is activation energy, and T is Generally there are two characteristic dimensional
temperature. parameters of the problem. The first one is a characteristic
The exponential Arrhenius reaction rate term in Eq. (2.2) reaction time, tr, and the second is the thermal diffusivity
creates difficulties for solving many flame problems. constant, w. Since chemical reaction is localized mainly in a
However, for cases of large activation energies, Ea/RTbb1, nearly isothermal reaction zone, the characteristic chemical
where Tb is the maximum flame temperature, the Arrhenius time, tr, is defined by the combustion temperature, Tb:
rate term in Eq. (2.2) may be approximated by a simple nL nF nO ðE a =RT b Þ
exponential function. This method is widely used in 1=tr ¼ kA N N e , (2.5)
NL F O
combustion theory following the work of Zeldovich and
Frank-Kamenetskii [24]: where nL and NL are stoichiometric coefficient and molar
concentration of the limiting component. For the same
eE a =RT  eE a =RT b ey , (2.3) reason of reaction localization, the thermal diffusivity
constant should be taken at the maximum flame tempera-
where y is dimensionless temperature:
ture. A general dimensional analysis suggests that the
E a ðT  T b Þ laminar burning velocity should be written as
y¼ . (2.4) sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
RT 2b
wðT b Þ
A laminar flame is obtained when the reaction proceeds SL ¼ F ðb; Le; sÞ, (2.6)
tr
as a subsonic wave propagating in the unburnt mixture by
diffusive transport of mass and energy. Fig. 7 shows where F is a function of dimensionless parameters of the
variations of fuel concentration, temperature, and reaction problem: the Zeldovich number, b, Lewis number, Le, and
rate in a laminar flame front. the ratio of densities between unburnt and burnt gases,
Thermal conductivity results in heat transport from hot s=ru/rb. The Zeldovich number is given by the following
combustion products into reactants and forms a preheat expression:
zone of characteristic width, d. Because of the strong E a ðT b  T u Þ
increase of the reaction rate with temperature, the reaction b¼ , (2.7)
RT 2b
mainly proceeds in a narrow zone where temperature is
close to Tb. The concentration of the reactants, Y, drops where Tu is unburnt gas temperature. The Lewis number is
through the reaction zone due to their consumption. The given by the ratio of mixture thermal diffusivity, w, to the
gradient of the reactant concentration results in a diffusion molecular diffusivity of the limiting component, DL:
flux of reactants and their transport to the reaction zone. Le=w/DL (limiting component is the one with concentra-
In the limit of large activation energy, the problem of tion below its stoichiometric value). The analytical solution
laminar flame propagation reduces to an eigenvalue for large activation energy, bb1, yields for the laminar
problem with one solution for the laminar burning velocity burning velocity:
SL, which first was provided by Zeldovich and Frank- sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Kamenetskii in 1938 [24]. Detailed derivation of the 1 wðT b Þ
SL ¼ 2Gnþ1 Len nþ1 , (2.8)
solution for the laminar burning velocity may be found s b tr
in [23,25]. Here we focus on qualitative relationships, ideas
where Rn=nO+nF is the overall reaction order,
1
Fraction of Reaction
Gnþ1  0 xn ex dx—Gamma function.
Y=1 reactants rate Tb The thermal thickness of the laminar flame (character-
istic width of the temperature gradient) may be estimated
by considering the balance of the heat flux from the
reaction zone with the energy flux at the cold boundary.
This gives the following approximate expression:
SL σSL
rb wðT b Þ wðT b Þ
d¼ ¼ . (2.9)
Reactants Temperature Pstcudor ru S L sS L
To illustrate the orders of magnitude of the laminar
burning velocity and flame thickness, we take typical values
Tu Y=0 of s=8, n=1, Le=1, w(Tb)=6  104 m2/s, b=10, and
tr=106 s. With these parameters, Eqs. (2.8) and (2.9)
give for the laminar burning velocity SL=0.4 m/s, for
Preheat zone, δ Reaction
the thermal thickness of laminar flame d=2  104 m,
zone, δ/β
and for the characteristic thickness of reaction zone
Fig. 7. Schematic of laminar flame structure. d/b=2  105 m.
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 509

In the area of turbulent combustion, to be discussed in tial development of the L–D instability is not observed.
Section 2.3, another approximate definition is often used Small wavelengths are stabilized by diffusion effects [28].
for the laminar flame thickness: Wavelengths much larger than the laminar flame thickness
n are stabilized at a certain level due to the difference in non-
d¼ , (2.10) linear evolution of convex and concave portions of the
SL
flame front [29]. The Landau–Darrieus instability has been
where n is kinematic viscosity. This is convenient because observed in freely propagating spherical flames [30], and in
the value of n is the main molecular transport coefficient the absence of other factors it results in the increase of the
used in the description of characteristic turbulent flow flame speed with the flame radius, RF, as RF1/3.
parameters. Generally Eq. (2.10) gives values for the flame
thickness that are somewhat smaller than those obtained
2.2.2. Diffusive instabilities
by Eq. (2.9) (with n ¼ 0.2  104 m2/s and SL ¼ 0.4 m/s,
For a slightly wrinkled flame front, in addition to
d ¼ 5  105 m).
unbalanced gas flows associated with the Landau–Darrieus
Eq. (2.8) gives a qualitative description for the depen-
instability, diffusive fluxes may also not be balanced. This is
dencies of the laminar burning velocity on temperature and
illustrated in Fig. 9, where diffusive fluxes of limiting
pressure [23]. As mentioned below, most practical combus-
component are shown with dashed lines (DL), and diffusive
tion systems are characterized by complicated multi-species
fluxes of heat—with solid lines (w). In cases where Leo1, the
chemistry with many elementary steps and quantitative
local combustion temperature, Tb, is increased behind the
aspects of Eq. (2.8) are difficult to discuss, because it is
convex part of the front, e.g., upper left part of Fig. 9, due to
based on 1-step overall global reaction. Qualitatively,
the increased mass diffusion of limiting component and
Eq. (2.8) suggests that the dependence of the laminar
decreased heat loss. This results in the local increase of the
burning velocity on pressure (P) is in the form SLpP(n2)/2
burning velocity. On the contrary, the local combustion
(because 1/trpPn1, and wpP1). The dependence of the
temperature is decreased behind the concave part of the front,
flame thickness is given by dpPn/2. For typical reaction
e.g., lower left part of Fig. 9, resulting in the local reduction of
orders of about n ¼ 2, the burning velocity does not
the burning velocity. All together this leads to further increase
depend on pressure, while the flame thickness is inversely
of the flame wrinkling. The opposite is true for cases where
proportional to pressure.
Le41, which are characterized by a tendency of smoothing
out flame wrinkles, which leads to stable flames.
2.2. Flame instabilities The thermal-diffusive effects were analyzed by Barenblat
et al. [31], Sivashinsky [32], and Joulin and Clavin [33]. It
2.2.1. Landau–Darrieus instability was shown that the onset of instability is actually given by
A planar flame front is intrinsically unstable due to b(Le1)o2. This is equivalent to Leo12/b and close to
gasdynamic effects connected with the expansion of the limit Leo1 for large activation energy, bb1.
combustion products. As illustrated by Fig. 8, if a flame
is slightly curved, the streamlines in the burnt gas converge
2.2.3. Cellular flames
behind the convex part of the front and diverge behind the
The thermal-diffusion effects reveal themselves as
concave parts. This creates an additional ‘‘push’’ for the
stabilizing of destabilizing factors of the underlying Landau–
convex part of the front, which increases the initial
Darrieus instability. This is illustrated by schlieren photos
curvature. This mechanism was revealed independently
of initial stage of flame propagation in Fig. 10 [34,35]. It is
by Landau [26] and Darrieus [27]. According to their linear
seen that unstable cellular flame is developed for mixtures
stability analysis, a planar laminar flame front is unstable
with Leo1 (10% H2/air), while smooth stable flame is
with respect to the Landau–Darrieus (L–D) instability for
all wavelengths of perturbations. In practice, the exponen-

t1 t2>t1 t2>t1 t1
t1 t2>t1 DL DL
χ χ
σSL
SL

Products Reactants Products Reactants


Products Reactants Le<1 Le>1

Fig. 8. Schematic of Landau–Darrieus instability. Flame shapes at two Fig. 9. Schematic of thermal-diffusive instability. Flame shapes at two
moments of time t1 and t2 are shown. moments of time t1 and t2 are shown. Flame propagates from left to right.
ARTICLE IN PRESS
510 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

suggested that the normal propagation velocity of a curved


flame, Sn, may be expressed as
LM
SL  Sn ¼ SL , (2.11)
RF
where LM is Markstein length and RF is the radius of flame
curvature. The ratio of the Markstein length to the thermal
thickness of the flame is known as the Markstein number,
Ma ¼ LM /d. According to the analysis of Markstein the
flame curvature is the main parameter that defines the
flame structure and the local flame speed. The values of
the Markstein length, or number, are determined experi-
mentally by measuring propagation speed of spherical
flames as a function of the flame radius (see, e.g., [38]).
Karlovitz et al. [39] proposed that it is the flame stretch
rate a ¼ ð1=AF ÞðdAF =dtÞ (where AF is the elementary area
of the flame front) that defines the local flame speed. The
stretch may be created by both the flame curvature (ac) and
inhomogeneities of the upstream flow (strain rate, as), and
yields the following more general version of Eq. (2.11) for
the normal propagation velocity of a stretched flame:
Sn d
1 ¼ Ma a. (2.12)
SL SL
If one separates the effects of curvature and strain rate,
as done by Buckmaster [40], Matalon [41], Chung and Law
[42], and Bradley et al. [43], Eq. (2.12) can be rewritten with
two different Markstein numbers: S L  S n ¼ Mac dac þ
Mas das .
The sign of the Markstein number in Eq. (2.12) defines
the response of a flame element to stretch and strain. For
Ma40 stretch leads to a decrease of the local burning
velocity, while for Mao0 stretch tends to increase the
burning velocity. As shown in detailed studies of Clavin,
Fig. 10. Schlieren photos of initial stage of flame propagation. Upper:
10%H2 in air; middle: 10%H2+5%O2+85%Ar; lower: 70%H2 in air
Williams and Joulin [33,36,44], positive Ma has a stabiliz-
[34,35]. ing effect on the flame wrinkling process and works against
hydrodynamic instability. It was also shown that within the
developed for mixtures with Le41 (70% H2/air). It is framework of the one-step Arrhenius reaction with large
interesting to notice that in the case where Le is close to activation energy the parameter b(Le1) defines the value
unity (10% H2+5%O2+85%Ar) a few large perturbations and even the sign of the Markstein number. Experimen-
are formed on an otherwise smooth flame surface. tally, flames with negative Markstein numbers, such as lean
Cellular flame propagation has been described by hydrogen–air mixtures, are known to be extremely unstable
Markstein [28] and theoretically analyzed by numerous [45,9]. As it will be discussed in Section 4, flame
authors, including Clavin and Williams [36], Joulin and acceleration is very pronounced in these mixtures. Qualita-
Clavin [33], Peclet and Clavin [37]. It should be noted that tively this may be explained by the destabilizing effect of
for expanding flames hydrodynamic instability can be turbulent flame stretch on flame wrinkling process.
damped when the characteristic flame radius, RF, is not too
large. This is controlled by the Peclet number, Pe, which is 2.2.5. Instabilities promoted by confinement and
defined as the ratio of the flame radius to the flame obstructions
thickness: Pe ¼ RF/d. The cellular structure (for unstable Flame propagation in an enclosure generates acoustic
flames with Leo1) must appear when the radius became waves that, after reflections from walls and obstacles, can
sufficiently large, so that the Peclet number exceeds a interact with the flame front and develop flame perturba-
critical value, Pecr [38]. tions through a variety of instability mechanisms. Such
instabilities have been observed by Guenoche [46] and
2.2.4. Instabilities and flame stretch Leyer and Manson [47] in open-ended and closed tubes, by
In the analysis of the effects of flame structure on the Kogarko and Ryzhkov [48] in closed spherical chambers,
hydrodynamic Landau–Darrieus instability Markstein [28] and by van Wingerden and Zeeuwen [49] and Tamanini
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 511

and Chaffee [50] in vented enclosures. For rich propane–air relatively slow flames in enclosures that are free of
mixtures, van Wingerden and Zeeuwen observed that these obstacles.
instabilities could result in a peak pressure enhancement
factor of 8. Tamanini and Chaffee observed enhancement 2.3. Turbulent flames
factors of 2–9 for stoichiometric methane–air and propa-
ne–air mixtures. Numerous experimental, analytical, and numerical
Flame acoustic instabilities are usually associated with studies have focused on the behavior of premixed turbulent
relatively slow flames in enclosures that are free of flames. Much detail can be found in comprehensive reviews
obstacles. Turbulence and turbulence inducing obstacles on turbulent flames [61–65] and others. Turbulent burning
have been shown to reduce relative contribution of acoustic velocities were systematically measured and correlations
instabilities on flame propagation and pressure build-up have been suggested by Abdel-Gayed, Bradley, Lau, and
[50]. It has also been shown that such instabilities can be Lawes [66–68], Bray [69], Gülder [70] and more recently by
successfully eliminated by lining the enclosure walls with Shy, Lin, Peng, and Yang [71,72]. These correlations allow
materials that can absorb acoustic waves. In the paper by the determination of turbulent burning velocities as a
Teodorczyk and Lee [51], flame acceleration experiments in function of dimensionless parameters characterizing tur-
tubes with repeated obstacles, with and without an bulence intensity, scale, and mixture properties. A detailed
absorbing material on the tube wall were reported. These review of studies of turbulent flames is beyond the scope of
experiments, performed with hydrogen–oxygen mixture, this paper. Our focus here is on the introduction of the
showed that the presence of an absorbing material reduced characteristic scales, turbulent combustion regimes, turbu-
the final flame velocity from 1000 m/s to 100 m/s. These lent burning velocity correlations, and concepts to be used
results would suggest that acoustic-flame instabilities might in the discussion of flame acceleration and transition to
in fact not be limited to slow flames in obstacle-free detonation.
environments.
The nature of the acoustic-flame instabilities has been 2.3.1. Scales of turbulence
reviewed by Oran and Gardner [52]. Searby and Rochwerger In this section, the definitions of the main length and
[53], Joulin [54], Jackson et al. [55], and Kansa and Perlee [56] time scales of turbulence are given. These scales will be
studied the mechanism of the instabilities in detail. These used in the following section for qualitative description of
mechanisms include flame distortion caused by the flame- turbulent combustion regimes. It is known that a gaseous
acoustic wave interaction, and wave amplification caused by flow becomes unstable at sufficiently high flow Reynolds
the coherence between the acoustic wave and the exothermic numbers:
energy release. The generation of acoustic waves by the rate
of change of heat release and instabilities is also nicely Re ¼ uL=n410  100, (2.13)
demonstrated by Al-Shahrany et al. [57]. where u is flow speed, L is the characteristic dimension of
Generally, if confinement and/or obstructions are pre- the flow, n is the kinematic viscosity. Flow instability
sent, several powerful instabilities may strongly influence results in the development of random oscillations in the
flame propagation. These are the well-known Kelvin– flow, which are superimposed on the mean flow variables.
Helmholtz (K–H) and Rayleigh–Taylor (R–T) instabilities. Following Reynolds’ idea one can separate statistical
The first of these two is associated with shear, and the means of the flow variables and their fluctuations. For
second one is develops when a lighter fluid is accelerated flow speed u this is illustrated by
towards a heavier fluid. In compressible flows this
instability is known as Richtmyer–Meshkov (R–M) in- u ¼ ū þ u0 , (2.14)
stability. Both K–H and R–T instabilities are triggered 0
where ū is mean speed, u is random mean square (r.m.s.)
when the flame is suddenly accelerated over an obstacle or velocity (by definition the mean of the random velocity
through a vent. ū0  0). The random mean square velocity is used to
Finally, sufficient fast flames can produce a shock wave characterize turbulent intensity. The turbulent kinetic
that can reflect off a surface and interact with the flame. As energy is a function of the r.m.s. velocity, and is usually
shown by Markstein and Somers [58], this can result in defined as k=3u0 2/2. In the following discussion we will
severe flame distortion which can induce flame acceleration assume that u0 is a function of k, and that u0 =k1/2 for
and, in extreme cases, cause transition to detonation simplicity.
[59,60]. The turbulent Reynolds number ReT is given by
While L–D and diffusive instabilities are relatively weak,
ReT ¼ u0 LT =n, (2.15)
K–H and R–T instabilities represent powerful mechanisms
that are mainly responsible for the increase in flame surface where LT is the characteristic length scale of turbulence
and generation of turbulence in channels with obstacles. which is associated with the largest dimensions of turbulent
L–D and diffusive instabilities may only play a role at the motions (large eddy size). At high turbulent Reynolds
initial stage of flame propagation, or in cases of unconfined numbers, a balance is quickly established between inviscid
flames. Acoustic instabilities may be important for transfer of turbulent energy through intermediate scales
ARTICLE IN PRESS
512 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

and its dissipation at the smallest scales. Thus, a wide range [61,73,74]. Such a diagram is usually referred to as the
of scales is an intrinsic feature of fully developed turbulent Borghi diagram, see Fig. 11. Note, that in the description
flows. of turbulent combustion regimes the laminar flame
The largest length LT and time tT scales (integral scales) thickness defined by Eq. (2.10) is usually assumed.
of turbulence are given by Characteristic regimes of turbulent combustion are listed
below:
LT ¼ k3=2 =; tT ¼ k=, (2.16)
totK Laminar flamelet regime.
where e is the dissipation rate of turbulent energy. The
(Kao1, Da41)
smallest scales, named after Kolmogorov, are given by
Chemical reaction remains fast and
3 1=4 1=2 turbulent eddies cannot penetrate
l K ¼ ðn =Þ ; tK ¼ ðn=Þ . (2.17)
into the flame. Burning occurs in
thin laminar flamelets, which are
2.3.2. Characteristic regimes of turbulent combustion stretched and curved by the flow.
Laminar flames in an initially quiescent mixture become Laminar flamelet regime is often
turbulent due to the development of flame instabilities subdivided into wrinkled flamelets
described above, and due to the growth of turbulence in the (u0 ESL) and corrugated flamelets
flame-generated flow. Turbulent motions randomly wrinkle (u0 4SL)
the flame surface resulting in flame area increase. At higher tKototT Thick flames and distributed
turbulent intensities, small eddies can interact with the (Ka41, Da41) reaction zone regime.
flame structure. Flame wrinkling and interactions of small Small scale turbulent eddies
eddies with the flame result in a variety of characteristic penetrate the flame and burning
turbulent combustion regimes. occurs in thick flames. Distributed
The interplay of characteristic scales is the key, which reaction zones are formed when the
defines the combustion regime. The flame itself provides smallest eddies penetrate the inner
the intrinsic length scale d (flame thickness) and the time reaction zone (which is estimated as
scale t=d/SL. Turbulence introduces the length scale, LT, d/b). This is approximately
the time scale, tT, and the smallest Kolmogorov scales, lK described by the condition: lKod/b,
and tK. These quantities define important dimensionless or Ka4b.
parameters known as the turbulent Damköhler number, tTot Well-stirred reactor.
Da, and the turbulent Karlovitz number, Ka: (Ka41, Dao1)
Da ¼ tT =t, (2.18) Turbulence mixes everything much
faster than reaction proceeds. There
Ka ¼ t=tK . (2.19) are no flames in this regime.
Parameters Da and Ka are used to define the character-
istic regimes of turbulent combustion. These regimes may It is convenient to use the expressions for the turbulent
be clearly shown on a diagram where velocity (u0 /SL) and Reynolds, Damköhler, and Karlovitz numbers, which are
scale (LT/d) ratios are used as independent variables given in terms of the variables of the Borghi diagram,
namely the velocity (u0 /SL) and scale (LT/d) ratios:
 0  
u LT
ReT ¼ , (2.20)
ReT=100 Da=1 SL d
Well stirred reactor Border of
100 quenching  1  
u0 LT
Da ¼ , (2.21)
SL d
Thick flames
u'/SL

10  3=2  1=2
u0 LT
Ka ¼ . (2.22)
ReT=1 SL d
Ka=1 Flame
acceleration These expressions clearly show the slopes of the lines
1
in given
Laminar flamelets geometry given by constant Reynolds, Damköhler, and Karlovitz
numbers on the Borghi diagram. Note that among the
following five parameters characterizing the turbulence,
1 10 100 1000 (u0 /SL), (LT/d), ReT, Da, and Ka, only two are independent.
LT / In channels with obstacles, the integral length scale
of turbulence, LT, is primarily controlled by the geometry.
Fig. 11. Borghi diagram of characteristic regimes of turbulent combustion. In such situations, initially a laminar flame propagates
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 513

through the channel, and then the flame speed increases, as diagram. As a first approximation, the turbulent burning
well as the speed of the flame-generated flow and the velocity is expected to be a function of the two parameters
turbulence level. During flame acceleration, the character- characterizing turbulence, e.g., u0 /SL and LT/d, and the
istic combustion regime evolves from laminar flamelets value of the laminar burning velocity. There are also other
towards distributed reaction zones, and possibly towards parameters characterizing mixture properties in addition to
quenching, as shown schematically in Fig. 11. The border SL. Among these, the Lewis (or Ma) number is specifically
for quenching is illustrated by the results of direct expected to play a role [68]. It may also be argued, that
numerical simulations by Poinsot et al. [75]. It should also parameters characterizing the reaction rate and the
be noted that because an accelerating flame generates expansion of the combustion products, such as b, s, and
compression waves, and sometimes a shock ahead of it, the n, may play a role [78]. In general, the turbulent burning
flame finds itself in reactants that are at elevated velocity may be expressed as
temperature and pressure. In this situation, changes in  0 
the values of SL and d with temperature and pressure ST u LT
¼F ; ; Le; b; s; n . (2.23)
should be taken into account. This effect tends to increase SL SL d
the length scale ratio and decrease the velocity scale ratio. The values of the turbulent burning velocity are
Planar Laser Induced Fluorescence (PLIF) images of the measured experimentally in artificially created turbulent
reaction zones for various combustion regimes are pre- flows (see, e.g. [43,66,67,71,72]). This data is correlated
sented in Fig. 12. These images were obtained by Ardey with parameters characterizing turbulence and mixture
and Jordan at the Technical University of Munich [76,77]. properties. Here, we present several examples of the
turbulent burning velocity correlations, which are widely
used in models for turbulent flames. Although different
2.3.3. Turbulent burning velocities
parameters for characterization of turbulence are used in
The turbulent burning velocity, ST, is defined as a speed
various studies, all the correlations may be presented in the
of propagation of a turbulent reaction zone or turbulent
following universal form:
flame brush relative to the reactants. The structure of the
turbulent reaction zone is defined by the turbulent  0 b3  b4
ST u L
combustion regime. The introduction of the turbulent ¼ b1 þ b2 Leb5 , (2.24)
SL SL d
burning velocity is a simplification, which reduces the
complexity of the turbulent reaction zones, such as shown where b1, b2, b3, b4,and b5 are correlation constants. The
in Fig. 12 to a relatively thick turbulent flame that values of these constants are presented in Table 1. Gülder’s
propagates at a velocity ST. This velocity is defined by [70] correlation applies to wrinkled flames at relatively low
the rate of consumption of the reactants in the turbulent turbulence, e.g., Kao1.5, ReTo3200, while others extend
reaction zone, and, thus, mainly by the state on the Borghi the range of applicability to relatively high turbulence and

Fig. 12. PLIF images of flame structure for the various regimes [76,77].
ARTICLE IN PRESS
514 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

Table 1 an approach was investigated to address the problem of


Turbulent burning velocity correlation parameters quenching of products/reactants pockets mixed by turbu-
Reference b1 b2 b3 b4 b5
lence based on the analysis of thermal regimes of the
pocket, and not from the perspective of flame stretch. The
Gülder [70] 1 0.7 3/4 1/4 0 critical conditions for thermal quenching appeared to be in
Bray [69] 0 1.8 0.412 0.196 0 general accord with the models and analyses based on
Bradley et al. [68] 0 1.53 0.55 0.15 0.3
flame stretch.
Shy et al. [71] 1 0.05 0.39 0.61 0
Other aspects of quenching are connected with ignition
and flame kernels in turbulent flows. These issues were
addressed in studies of Ballal and Lefebvre [81], Checkel
thick flames, e.g., Ka Leo38, ReT/Le2o6000 [68]; Ka and Thomas [82], and Chomiak [83]. Phillips [84] and
Leo13, ReT/Le2o25,000 [71]. Larsen [85] studied quenching of flames propagating
These correlations, and especially those for high turbu- through critical gaps and holes.
lence, show a weak increase of ST/SL with u0 /SL and LT/d. Local quenching may be important to the flame
However, the domains of validity for these correlations in the acceleration process because, under certain conditions, it
Borghi diagram are limited by certain ReT and Ka numbers. can lead to violent secondary explosions and DDT. Global
At higher Reynolds numbers ReT41000 [68], an increase of quenching has been observed for lean flames propagating
u0 /SL above 20 may result in decrease of ST/SL. This shows in tubes filled with obstacles [86–88]. It has also been
that the increase of the turbulent burning velocity with observed when a flame propagates through an orifice into
turbulent intensity is limited and the maximum experimen- an unconfined area [89]. In this case, the minimum
tally observed values of ST/SL are of the order of 10 as quenching diameter increases with the magnitude of the
discussed in details in [68,71]. pressure differential generated across the orifice before the
The effect of the Lewis number introduced in [68, above], arrival of the flame front.
although conceptually well understood, does not explicitly
reveal itself in the available data, as found by Bray [69] and 3. Detonation waves
Shy et al. [71]. Generally, at relatively high turbulence
ST/SL tends to show a weaker dependency on u0 /SL a The fact that the CJ theory did such an exemplary job at
nd LT/d, compared to cases with relatively low turbulence predicting the measured detonation wave velocity led early
as in [70]. The compilation of the data obtained at researchers to believe that detonation waves were steady 1D
Leeds University used in [68,69] may also be approxi- phenomenon. Further verification of the theory lay in the
mated by a convenient expression with simple fractions in important fact that the measured detonation velocity was
exponents: found to be independent of the rate of chemical reactions.
 0 1=2  1=6 The CJ theory predicts that the propagation velocity is
ST u L governed by gasdynamic choking that is controlled by the
/ . (2.25)
SL SL d thermodynamics of the mixture. Gasdynamic choking
The coefficient of proportionality is intentionally occurs whereby heat addition, due to chemical energy
omitted in Eq. (2.25) because of the considerable un- release, accelerates the flow from a subsonic velocity
certainty of its value. This may be illustrated by the fact immediately behind the lead shock wave to the speed of
that the Leeds data [68] exceed the data of Shy et al. [71] by sound at the end of the reaction zone. This is in contrast to
a factor of 2 for the same values of ReT and u0 /SL. flames where the burning velocity is governed by the rate of
chemical reactions. The CJ detonation velocity and pressure
for hydrogen–air and methane–air mixtures initially at 20 1C
2.3.4. Quenching of turbulent flames and 1 atm are plotted as a function of equivalence ratio in
Turbulent mixing processes that produce a variety of the Fig. 13. The values in Fig. 13 were obtained from the
turbulent combustion regimes can also result in local or chemical equilibrium code STANJAN [90]. Note, in general
global quenching. This quenching can be considered as a the highest CJ pressure corresponds to a mixture slightly
process either due to excessive stretching of flames, or to rich of stoichiometric.
intensive mixing of cold unburned gas into the distributed
reaction zones. The flame-quenching process has been 3.1. Historical review
investigated by various authors. These studies included
analysis of Leeds experimental data [43,66,67,79], asymp- 3.1.1. ZND detonation wave structure
totic studies of stretched laminar flames (see e.g., [80]), One of the main problems with the basic CJ theory is
direct numerical simulations (see e.g., [75] and review [65]). that it does not provide a characteristic length scale, for the
The quenching was addressed in these studies as a result of detonation wave. It was not until the work of Zeldovich
flame stretch induced by turbulence. The critical conditions [91], von Neumann [92] and Döring [93] who independently
for flame quenching were determined using experimental came to the realization that a steady 1D detonation wave
correlations or direct numerical simulations. Recently [78], consists of a shock wave followed by a finite length
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 515

2500 30 1 2.5

CJ Detonation Pressure (atm)


CJ Detonation Velocity (m/s)

25 0.9
2000 2
0.8
20

Mach Number

T/T2, P/P2
1500 0.7 1.5
15
1000 0.6 1
10
0.5 Mach number
500 P/P2 0.5
Hydrogen-air 5 0.4 T/T2
Methane-air
0 0 0.3 0
0 0.5 1 1.5 2 2.5 3 0 0.05 0.1 0.15 0.2
Equivalence Ratio Distance from shock (cm)

Fig. 13. CJ detonation velocity as a function of equivalence ratio. Fig. 14. ZND detonation front structure for atmospheric hydrogen–air.

reaction zone that terminates at the CJ state, as described common definition for the induction distance is the
in Section 1.2.1. In recognition of the contributions distance from the shock to the point downstream of
of Zeldovich, von Neumann and Döring the steady 1D the shock where the rate of temperature increase is
detonation structure just described is commonly referred to maximum [94].
as the ZND structure. In the early 1960s, Erpenbeck [96,97] showed mathema-
The ZND structure provides a characteristic chemical tically that the steady-state ZND detonation structure is
length scale for the detonation wave that can be calculated unstable to infinitesimal longitudinal perturbations. The
with knowledge of the chemical kinetics. Finite rate stability analysis performed by Erpenbeck used the
reactions are initiated when the gas temperature and conservation equations linearized about the steady solution
pressure are instantaneously raised by the leading shock and thus did not follow the evolution of the applied
wave to state 2 on the equilibrium Hugoniot curve, see perturbation. The analysis was taken one step further by
Fig. 3. At the end of the reaction zone chemical equilibrium Fikett and Woods [98] who investigated the transient
is achieved and the mixture state is described by the evolution of a 1D, piston supported detonation wave by
point J on the equilibrium Hugoniot curve. The variation performing a numerical study of the time-dependent 1D
of the normalized temperature, normalized pressure and conservation equations. In the analysis the heat capacity is
local flow Mach number in the detonation front for a assumed constant and a single irreversible reaction obeying
hydrogen–air mixture at atmospheric initial conditions is the Arrhenius rate law is considered. The parameters used
provided in Fig. 14. The data is obtained using a computer in the study are similar to those used by Erpenbeck, they
program developed by Shepherd [94] that integrates the include the detonation overdrive, the reduced activation
steady 1D inviscid conservation equations with the de- energy, (Ea/RT1), as well as the normalized chemical energy
tonation post shock state taken as the initial condition. release (Q/RT1) and specific heat ratio that were kept
Essentially the conservation equations are integrated along constant at values of 50 and 1.2, respectively. Both analyses
the Rayleigh line from state 2 to J in Fig. 3. For this showed that for a fixed detonation overdrive there was a
mixture the detonation velocity is 1977 m/s and the post- critical activation energy below which the detonation was
shock temperature and pressure are 1533 1C and 27.7 atm, stable and above which it was unstable. For example, for
respectively. The calculation was performed with the an overdrive of 1.6 the critical reduced activation energy
chemical kinetics mechanism of Konnov [95]. Since the was roughly 45. Conversely, increasing the overdrive tends
reactions follow the Arrhenius rate law, the leading shock to stabilize the detonation wave. For the stable case, after a
wave is immediately followed by an energy neutral short transient associated with the starting of the piston,
induction zone where the thermodynamic properties the shock pressure asymptotes to the steady-state value.
remain roughly constant. During this time chemical For the unstable case, the shock pressure undergoes large
reactions proceed such that free radical species are oscillations about the steady-state value with swings in
produced. Once recombination reactions start to take pressure reaching up to 50% above the steady value. These
affect energy is released and the temperature rises. The swings in shock strength are due to the non-linear coupling
particle velocity after the shock is subsonic and as energy is between the gasdynamics and chemical kinetics.
released it accelerates towards the CJ choking condition
while the pressure drops. The flow Mach number 3.1.2. Multi-head detonation wave
asymptotes to a value of one, therefore, the CJ condition Around the same time that Erpenbeck showed theoretically
cannot be used to define an induction distance. The most that the ZND detonation structure is unstable, experimental
ARTICLE IN PRESS
516 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

I
R

M
Triple point
trajectory s

Fig. 16. Mach reflection producing triple-point configuration, I—incident


shock, R—reflected shock, M—Mach stem, s—slipstream.

Fig. 15. Interferograms of detonations in 2H2+O2+0.92Xe at an initial


pressure of (a) 0.075 atm and (b) 0.04 atm [99]. Reused with permission
from Donald R. White, Physics of Fluids, 4, 465 (1961). Copyright 1961,
American Institute of Physics.
B
A
K
evidence started to surface that showed that a detonation
wave was not a planar steady phenomenon as was believed
for many years. One of the first indications that detonation
waves are actually multi-dimensional phenomenon came K
A
from the experimental work of White [99] who used
B
interferometry to visualize the detonation structure. A couple
of White’s interferograms are reproduced in Fig. 15 in which Fig. 17. Soot imprint of a detonation wave propagating from left to right
it is clear that the gas density behind the shock wave is not in an 11 mm diameter tube in stoichiometric hydrogen–oxygen at
uniform. Based on his observations he concluded that 130 mmHg and schematic showing the two triple-point configurations
that produces the imprint lines [100].
‘‘turbulence is a common property of detonations’’. The
interferograms in Fig. 15 show the presence of secondary
waves oriented perpendicular to the detonation front, and the surface. In two dimensions the triple point is actually the
leading shock wave is observed to be slightly curved where projection of a line where three shock waves meet. A
the transverse waves merge with the leading front. White also slipstream (shear layer) also extends from the triple point
noted that as the initial mixture pressure is increased the back into the compressed region between the reflected
number of transverse waves present increases. shock wave and the reflecting surface. The streamlines on
The multi-dimensional structure of a detonation wave either side of the slipstream are parallel and the pressure
was first reported by Denisov and Troshin [100] based on across the slipstream is equal. Since the pressure across the
high-speed streak photographs and soot imprints. They slipstream is equal it is clear that the Mach stem is a
made a distinction between the previously observed, and stronger shock than the incident shock since the particles
well documented ‘‘single-head spin’’ mode; to be discussed passing through the incident shock wave also pass through
later; and the so-called ‘‘multi-head spin’’, or ‘‘pulsating’’, the reflected shock. Note, a double-Mach stem configura-
mode. The novelty of the experimental work was the soot tion is also possible whereby a second triple-point
imprint technique that was used to unravel the detonation configuration forms on the reflected shock some distance
wave structure. The soot imprint technique was first used from the primary triple point [102].
by Mach and Sommer [101] in the study of shock wave The soot, or smoked, foil technique commonly used in
reflection. When a shock wave reflects obliquely from a detonation research involves coating a metal or plastic foil
surface, a Mach reflection can result if the angle of with a thin layer of carbon soot and placing the foil inside
incidence is larger than a critical value that depends on the the detonation tube with the sooted side facing inwards.
shock Mach number and heat capacity ratio [102]. In a During the test the detonation wave initiated in the tube
Mach reflection a third shock wave, referred to as a Mach passes over the inside of the foil. As a triple point moves
stem, forms that is perpendicular to the reflecting surface across the foil, soot is displaced leaving a line that
and propagates parallel to the surface. Shown schemati- corresponds to the triple-point trajectory. An example of
cally in Fig. 16 is the interaction of an incident planar such a soot foil obtained by Denisov and Troshin [100] is
shock wave, I, with a ramp producing a reflected shock provided in Fig. 17. Based on the converging and diverging
wave, R, and a Mach stem, M. The three shock waves lines observed on the imprint they described the structure
intersect at the so-called triple point that moves away from of the detonation wave as two triple-point configurations
the reflecting surface as the Mach stem moves along the (ABK in Fig. 17) propagating first towards each other and
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 517

then away from each other on the detonation front. The Fig. 18 for stoichiometric hydrogen–oxygen with different
turbulent shear layers extending back from the front can diluents. The smoked foil records show that the triple-point
clearly be seen in White’s interferograms [99]. Denisov and trajectories form a cellular pattern with varying degree
Troshin used the term ‘‘head’’ to describe a triple point on of uniformity, or ‘‘regularity.’’ Strehlow found that the
the detonation front, so for the detonation wave shown in addition of at least 40% argon to the hydrogen–oxygen
Fig. 17 there are two heads. Denisov and Troshin pointed made the cellular pattern very regular in contrast to the
out that upon collision of the two transverse waves AK, the addition of nitrogen, which tends to make the structure
unreacted gas immediately behind the shock wave AA more ‘‘irregular’’. Similar effects of diluent addition on the
would be raised to a high pressure and temperature cellular structure regularity were found for ethylene–
resulting in autoignition after an induction time. Before oxygen and acetylene–oxygen. In general methane–oxygen
the collision of the triple points the shock wave AA is gave much poorer detonation structure regularity. It was
relatively flat and after the collision the autoignition of the later shown by Ul’yanitskii that the regularity of the
gas produces a highly curved shock wave traveling faster detonation cellular structure correlated qualitatively with
than the flatter shock wave before the collision. They point the reduced activation energy [105]. Specifically the higher
out that the shock wave AA thus undergoes periodic the value Ea/RT2, recall T2 is the post shock temperature,
variations in velocity that on average equals the CJ the more irregular the cellular pattern. The reduced
detonation velocity. Denisov and Troshin also point out activation energy for argon-diluted mixtures is low because
that the frequency of the pulsations depends on the of the low heat capacity of the argon, relative to say
chemical kinetics of the mixture. nitrogen, and thus higher post shock temperature. Erpen-
By the mid 1960s the structure of a self-sustaining beck’s 1D stability analysis showed that detonation
detonation wave was universally accepted to consist of a stability correlated in the same manner with reduced
forward moving strong shock wave followed by a reaction activation energy [97]. The detonation cellular structure
zone that contained transverse waves, as depicted in Fig. 17. observed in a real multi-dimension detonation is simply the
Strehlow and Fernandes [103] showed using an acoustic manifestation of the detonation front instability. In some
model that high frequency perturbations occurring in the mixtures the detonation cellular pattern recorded on
ZND reaction zone are amplified giving rise to transverse smoked foils shows secondary much finer lines commonly
waves propagating parallel to the leading shock wave. The referred to as fine substructure. The substructure emanates
model was supported by shock tube experiments performed from the triple-point collision site and extends to about
with combustible gas where a detonation wave was half the cell length. Manzhalei showed a correlation
initiated by reflecting a planar shock wave off the shock between the presence of fine structure and the reduced
tube end wall. The roughly 1D initiation process resulted in activation energy, i.e., for mixtures with a reduced
the spontaneous formation of transverse waves behind the activation energy of less than 6.2 fine structure is absent
detonation front. and for values greater than 6.5 fine structure appears [106].
In the 1960s Strehlow made significant contributions to By analyzing the nonreactive shock wave interaction that
the understanding of the detonation wave structure, and occurs at the collision of the two triple points the transverse
specifically the role of transverse waves [104]. Strehlow shock wave Mach number can be approximated. Critical to
used smoked foils extensively to record the triple-point this analysis is the included angle between the triple-point
trajectories for detonation waves in different fuel–oxygen– trajectories just before collision that is obtained from the
diluent mixtures. A sample of these foils is provided in smoked foil records. For hydrogen–oxygen–argon detonations

Fig. 18. Smoked foils obtained for detonations in stoichiometric hydrogen–oxygen: (a) 150 Torr, 50% nitrogen dilution; (b) 120 Torr, no dilution;
(c) 150 Torr, 40% argon dilution [104]. This article was published in Astronautica Acta, Vol. 14, R.A. Strehlow, The nature of transverse waves in
detonations, pp. 539–48, Copyright Elsevier 1969.
ARTICLE IN PRESS
518 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

the transverse wave Mach number for ordinary detona- the shock front. The fine transverse structure associated
tions was approximated to be 1.25 and for marginal with the strong detonation is responsible for the ‘‘rope’’-
detonations it was 1.35 [107]. Note a marginal detonation like thick line that appears on the soot imprint for a
wave is obtained near the detonability limit where in a spinning detonation wave [100]. Therefore, a spinning
rectangular channel only two triple points exist forming a detonation wave is simply a special case of a multi-head
single cell across the channel, see Fig. 17. In some detonation that occurs for mixtures near the detonability
experiments in rectangular channels the width of the limits.
channel is chosen to be smaller than the triple-point
spacing in order to eliminate the transverse wave propaga- 3.2. Cellular detonation wave structure
tion in the direction of the smaller dimension. Under this
condition a 2D detonation wave structure is obtained but Based on soot imprints, interferograms and Schlieren
the characteristics are different from an ordinary detona- photographic evidence obtained for self-sustained detona-
tion wave obtained in the same mixture in a larger channel. tion waves it is clear that the front is actually an unsteady
In a round tube a marginal detonation wave is called a 3D structure. The instability of the front manifests itself as
‘‘spinning wave’’ and is briefly discussed in the next section. perturbations that propagate transversely to the front
propagation direction. These transverse perturbations
3.1.3. Single-head spin detonation include finite amplitude shock waves that stretch from
A detonation wave in a gas mixture near the detonability the leading shock wave deep into the combustion zone
limit behaves much differently than the pulsating detona- resulting in a discontinuous reaction zone. The presence of
tion wave described above. The so-called spinning, or the transverse waves is essential to the propagation of a
single-head spin, detonation wave was first reported by self-sustained detonation wave. The schematic in Fig. 19
Campbell and Woodhead [108] in 1927 and was studied shows the detonation front structure at several instants in
extensively by many investigators over the next 40 years. time as well as the trajectories of the triple points [112]. The
The spinning detonation wave appeared to be very triple-point trajectories trace out a cellular pattern where
different from the 1D detonation wave that was satisfacto- the transverse dimension of an individual cell, e.g., ABCD,
rily described by the CJ theory. The most outstanding is denoted by l and the length of the cell is Lc.
characteristic observed in the early studies performed in The collision of a pair of neighboring triple points at
round tubes is that a luminous head on the detonation point A in Fig. 19 results in the formation of a strong
front moves in a helical path where the pitch is roughly leading wave that initiates chemical reactions after a very
three times the tube diameter [103]. The first theories to short induction time. The leading shock wave propagating
describe the propagation mechanism of a spinning detona- through the cell from point A to D decays in strength. It is
tion wave were proposed by Manson [109] and indepen- common to use Mach reflection terminology in describing
dently by Fay [110]. These were acoustic theories that the progression of the leading shock. In the first part of the
showed that the measured spin detonation frequency was cell when the strong leading shock is created from the
equal to the natural frequency of (transverse) vibration of triple-point collision it is referred to as the Mach stem. In
the hot combustion products behind the detonation wave.
It was proposed that the nonuniform combustion at the
detonation front excites a transverse vibration in the wake
gas in the form of a rotating compression wave.
Denisov and Troshin [100] used their soot imprint
technique to investigate the structure of the spinning λ
detonation front. In a multi-head detonation there are two
or more triple points that collide, in a spinning detonation
there is only a single triple-point with no chance for
collision so it simply rotates circumferentially around the
front. As a result the soot imprint taken of a spinning
detonation only has a single oblique line that repeats itself
periodically in the axial direction, in the actual tube the
triple-point trajectory is a helix. The most complete and
accurate picture of the spinning detonation front was first
provided by Schott [111] who mapped the complex
structure consisting of a non-planar shock wave followed
by a reaction zone with a nonuniform induction distance.
Schott showed that the single-head that rotates over the
front consists of a CJ detonation wave propagating Fig. 19. Schematic of the propagation of a cellular detonation front,
obliquely to the front. This CJ detonation propagates into showing the trajectories of the triple points (dashed line represent start of
a gas region that is compressed by the weakest portion of exothermic reactions) [112].
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 519

the latter part of the cell leading up to the triple-point below the triple point pass through the incident wave and
collision it is referred to as the incident shock. For then the transverse wave. In general, the induction time for
marginal detonation waves Strehlow and Crooker [113] the particles passing through the Mach stem is shorter than
report a centerline shock velocity of roughly 1.6 times the those propagating through the incident and transverse
CJ detonation velocity at the start of the cell and 0.6 times waves. As a result of this difference in induction time the
the CJ detonation velocity at the end of the cell. The shear layer represents a discontinuity in the reaction zone.
decaying shock wave achieves the CJ velocity in the first For marginal detonation waves that have stronger trans-
25% of the cell length after which the decay rate is much verse waves it has been observed experimentally that the
slower. Experiments performed by Austin et al. [114] chemical reactions can occur immediately behind the
suggest that the excursions in the shock strength increase transverse wave, much like in a single-head spin [116,113].
with increased reduced activation energy, in agreement As the centerline shock wave strength decays through the
with numerical predictions [115]. cell the induction time increases. As shown in Fig. 19 this
In detonation waves the triple point can take on a results in a progressive widening of the distance between
‘‘weak’’ configuration consisting of a single triple point, or the leading shock wave and the start of energy release in
a ‘‘strong’’ configuration where there is a secondary triple the reaction zone denoted in the figure by dotted lines. By
point at the end of the reflected wave corresponding to the the end of the cell the shock wave strength decays to a level
primary triple point, also known as a double Mach where the induction time is very long and the reaction zone
configuration. Fig. 20 shows details of an ideal ‘‘weak’’ decouples from the shock wave [114]. This leaves a
triple point propagating along the trajectory AC in Fig. 19. substantial amount of unburned gas behind the incident
Specifically it shows various particle paths crossing the shock wave that takes on a characteristic ‘‘keystone-shaped
shock front from a reference point relative to the triple region’’ [117]. Simultaneous Schlieren and planar laser
point. All the gas particles above the triple point pass induced fluorescence techniques were used in a narrow
through the Mach stem and below the triple point pass channel (less than 0.5l wide) by Austin et al. [118] to image
through the incident shock. The particles entering just both the shock configurations and the OH reaction front.
An example of such images clearly showing the keystone
feature is provided in Fig. 21. The detonation front images
Shear layer Mach stem
in Fig. 21 were taken in a heavily argon-diluted hydro-
gen–oxygen mixture that exhibit a regular structure. In
such a detonation wave the shock and reaction fronts are
E
Transverse smooth and free of any localized explosions.
wave Heavily argon-diluted mixtures have a low reduced
activation energy and thus a 1D detonation stability
Incident wave
analysis predicts that such mixtures produce ‘‘stable
detonations’’. In a 1D analysis the terms stable and
unstable correspond to the regularity in the periodicity of
the oscillation observed in the shock velocity and pressure.
Fig. 20. Schematic showing particle trajectories relative to the triple point
for a single Mach stem configuration. The shaded area represents the Austin et al. extend the use of the term stable and unstable
burned gas region. See Fig. 19 for the location on the detonation front of detonation to describe the regularity of the cellular
the triple-point labeled E. structure of an actual detonation wave. The argon-diluted

Fig. 21. Images of detonation propagation from left to right in 2H2+O2+12Ar, Pi ¼ 20 kPa in narrow channel. (a) Schlieren image. The box shows the
location of the corresponding OH fluorescence image shown in (b). (c) Superimposed Schlieren and OH fluorescence images, FLIF image 60 mm high
[118].
ARTICLE IN PRESS
520 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

Fig. 22. Images of detonation propagation from left to right in 2H2+O2+3.5N2, Pi ¼ 20 kPa in narrow channel. (a) Schlieren image. The box shows the
location of the corresponding OH fluorescence image shown in (b). (c) Superimposed Schlieren and OH fluorescence images, FLIF image 30 mm high
[118].

hydrogen–oxygen mixture in which the stable detonation very irregular structure similar to the observations made by
shown in Fig. 21 was observed has a reduced activation Austin et al. [118]. Radulescu et al. observed an unburned
energy, Ea/RT2, of roughly 5. This reduced activation gas ‘‘tongue-like structure’’ extending from the triple point
energy value lies just below the neutral stability boundary deep into the burned gas. This is essentially a highly
corresponding to the CJ Mach number [119]. If nitrogen is distorted half of the keystone feature observed by Austin
used in place of argon the reduced activation energy is et al. [118]. The Schlieren images clearly show that the
increased to a value above the neutral stability value and shear layer extending from the triple point is K–H
thus the detonation is expected to be unstable and display unstable. It is proposed that consumption of the unburned
an irregular detonation structure. The simultaneous gas in the tongue-like structure occurs via turbulent
Schlieren and planar laser induced fluorescence images combustion at the shear layer as opposed to adiabatic
obtained for such a mixture is shown in Fig. 22. There is a shock compression. Radulescu et al. also observed local
dramatic difference in the structure of the detonation front explosions occurring within the tongue-shaped structure
compared to the stable detonation wave shown in Fig. 21 behind the leading shock. It is argued that the intense
that displays a regular cellular structure. As pointed out by mixing of the hot and cold gases across the shear layer may
Austin et al. [118] the unstable detonation fronts are lead to the observed explosion.
characterized by wrinkling of the shock and OH fronts
over a wide range of spatial scales. The irregularity of the 3.2.1. Critical initiation energy and detonation cell size
structure makes it more difficult to discern the keystone measurements
feature that is very evident for the stable case. They showed It is of interest to characterize the detonation sensitivity
that the further the mixture reduced activation energy of a fuel–oxidizer mixture at a given initial condition. The
deviates from the neutral stability boundary the more critical, or minimum, energy required for direct initiation
irregular the detonation structure becomes. The most of a spherical detonation is essentially a measure of the
unstable detonations were obtained in hydrocarbon–oxygen detonation sensitivity of a mixture at a given initial
mixtures diluted with nitrogen. It was also observed that in thermodynamic condition. Ideally a point energy source
the narrow channel the stable detonation waves propa- should be used for this measurement such that the
gated at a velocity up to 8% below the CJ value and for the initiation process is independent of the source size and
least stable detonations the velocity deficit was less than energy release rate. Experimentally the energy is deposited
3%. This dependency of the detonation velocity deficit is into the mixture using an electric spark for fuel–oxygen
consistent with observations made in larger tubes by Moen mixtures and high explosives for fuel–air mixtures. A
et al. [120]. comprehensive set of critical initiation energy data in
Radulescu et al. [121] performed narrow channel fuel–air mixtures at atmospheric conditions was obtained
experiments investigating the ignition mechanism in an by Bull et al. [122], see Fig. 23 for a plot of initiation
irregular structure detonation wave. The experiments were energy, in grams of tetryl, versus equivalence ratio.
performed in a narrow channel (o0.25l wide) using If sufficient energy is instantaneously released at a point
stoichiometric methane–oxygen, which is known to display source a reactive spherical blast wave is produced that
a very irregular cell structure. The channel height was decays asymptotically to a CJ detonation wave. If
chosen such that only a single triple point existed and thus insufficient energy is released the shock wave and reaction
the detonation waves are marginal propagating with a zone permanently decouple and no detonation is formed.
velocity deficit of as much as 20%. Schlieren images show a At the limit, where a critical amount of energy is deposited,
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 521

Fig. 23. Comparison of critical initiation energy for fuel–air mixtures with
surface energy theory (1 g tetryl ¼ 4520 J) [112].
Fig. 24. Detonation cell size of fuel–air mixtures at atmospheric pressure,
curves represent l ¼ ADR anchored at stoichiometric condition [112].

the shock wave and the reaction zone start to decouple and The detonation cell width can be considered a funda-
enter a quasi-steady period where the shock wave and the mental property of the combustible mixture that can be
reaction front propagate at roughly the CJ detonation used to characterize the detonability of the mixture.
velocity. This quasi-steady period ends when a localized Strehlow and Engel [125] made detailed measurements of
explosion occurs in the compressed gas between the shock the spacing between parallel triple-point trajectory lines
and the reaction front [123]. This localized explosion which is essentially the detonation cell width l shown in
immediately forms into a multi-head detonation that Fig. 19. They showed that the spacing was governed by the
quickly sweeps across the compressed gas layer behind reactivity of the mixture and that for heavily argon-diluted
the blast wave. fuel–oxygen mixtures the line spacing was proportional to
Zeldovich first theorized that for successful detonation the CJ detonation induction zone length. Strehlow and
initiation the time required for the blast wave to decay to the other researchers of the time restricted their cell width
CJ shock strength must be of the order of the 1D induction measurements to fuel–oxygen mixtures heavily diluted with
time [124]. This theory predicts that the critical initiation monatomic gases. The main reason for this is that the
energy is proportional to the cube of the induction time, e.g., cellular imprints for these mixtures are very regular and
the minimum initiation energy corresponds to the most thus appealing for investigating the structure of the
reactive composition for a mixture at a given initial detonation front. However, from a practical perspective
thermodynamic condition and increases sharply as one most industrial explosion hazards involve fuel–air mix-
approaches the lean and rich detonation limits. This is tures, for which very little detonation cell size data was
corroborated by the Bull et al. [122] experimental data available at the time. The experimental measurement of
shown in Fig. 23. Note the data follows a U-shaped curve detonation cell size in fuel–air mixtures requires a larger
that is typical of all detonation properties. Lee [112] apparatus and a more energetic initiation source, such as
developed a series of theories that showed that the critical high explosives. Bull et al. [126] reported detonation cell
initiation energy is proportional to the cube of the size data for stoichiometric fuel–air mixtures at initial
detonation cell width, Ecl3. The solid lines shown in pressures up to one atmosphere. The general trend in the
Fig. 23 are predictions based on Lee’s semi-empirical data shows that for a given mixture composition the
‘‘surface-energy theory’’ that requires an experimentally detonation cell size decreases with increasing initial
measured chemical length scale, typically taken to be the pressure. The following year Knystautas et al. [127]
cell width. This semi-empirical relationship is more practical published a comprehensive set of detonation cell size data
than the Zeldovich theory as it is based on a measured for atmospheric fuel–air mixtures over a wide equivalence
chemical length scale and not a theoretical induction ratio range. Fig. 24 shows a plot of the measured
time. detonation cell size as a function of equivalence ratio for
ARTICLE IN PRESS
522 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

atmospheric fuel–air mixtures [112]. The minimum cell size mixture equivalence range. Using the maximum rate of
is typically obtained for a stoichiometric composition temperature rise to define the induction distance he found
(equivalence ratio equal to one). The cell size increases that A varied from 5 to 50 over the full equivalence ratio
rapidly as the mixture becomes leaner or richer, thus the range. These values of A for hydrogen–air at atmospheric
data takes on a U-shaped curve similar to the critical conditions are consistent with the findings of Ciccarelli
initiation energy curve shown in Fig. 23. The effect of et al. [129]. Ciccarelli et al. [129] also showed that the value
initial temperature on the detonation cell size is not as clear of A changes significantly with initial temperature, e.g., for
cut. Detonation cell size data obtained in hydrogen–air– initial temperatures of 500 and 650 K the range of A is
steam mixtures at atmospheric pressure shows a decrease in 20–100. Similar findings were reported by Knystautas et al.
cell size with increasing initial temperature over the range [134] for benzene–air mixtures. These studies indicate that
of 300–650 K [128,129]. The effect of initial temperature is a correlation involving only the induction distance is not
very pronounced for the leaner mixtures. The opposite sufficient. Gavrikov et al. proposed a correlation, where
effect of initial temperature on detonation cell size was the proportionality constant A is a function of the stability
observed by Auffret et al. for mixtures of acetylene– parameters, namely the reduced effective activation energy
oxygen–argon [130] and ethylene–oxygen–argon [131]. and the reduced heat release, that gives better results [135].
Although the simple linear correlation is of limited use
3.2.2. Cell size correlations for accurately predicting cell size, it does capture the basic
Much like for the critical initiation energy no theory dependency of cell size on the initial thermodynamic
based on first principles exists to predict detonation cell properties of the mixture. In the cell size versus equivalence
size. Direct modeling of the detonation phenomenon using ratio plot in Fig. 24 the lines correspond to the correlation
CFD has progressed significantly in the last few decades anchored at the stoichiometric condition for each fuel–air
but we are still not at a point where the cell size can be mixture. The induction zone length was obtained from
reliably predicted by numerical simulation. Shchelkin and Westbrook’s calculations [133]. In general the correlation
Troshin [132] first proposed a linear relationship between does a good job of predicting the general trend in cell size
the detonation cell width and the ZND induction length data with mixture composition. The apparent agreement
DZND, e.g., l ¼ ADZND. In order to use this relationship the between the measured cell size and the correlation on the
constant of proportionality A must be established by lean size is deceiving considering the very steep nature of
anchoring the correlation using a known cell size data the curve. The linear correlation can be used satisfactorily
point. for interpolating between measured cell size data points but
Westbrook and Urtiew [133] calculated the induction care should be taken in extrapolating existing cell size data.
distance for a 1D detonation wave in hydrocarbon–air
mixtures using an 87 equation reaction mechanism. The 3.2.3. Critical tube diameter
induction time tCJ was calculated for a constant volume Experimentally it is found that when a detonation wave
reaction with the initial condition corresponding to the CJ emerges from a round tube into unconfined space contain-
detonation post shock state. The induction time was ing the same combustible gas mixture the detonation either
defined as the time corresponding to the point of maximum fails or continues as a spherical detonation in the
rate of temperature increase. The induction distance was unconfined volume. This phenomenon was first reported
calculated by taking the product of this induction time and by Zeldovich et al. [136] in their study of spherical
the post shock particle velocity relative to the shock wave detonation initiation. They performed experiments in
DZND ¼ tCJ(Uu1). In the calculation they showed that the various gas mixtures in different size round tubes. They
detonation cell size and the reaction zone for stoichiometric found that for each gas mixture there is a critical tube
mixtures at atmospheric conditions were correlated by diameter below which the detonation wave emerging from
lffi20DZND for fuel–air mixtures, and lffi35DZND for the tube fails and above which a detonation is transmitted
fuel–oxygen mixtures. They also compared the calculated into the unconfined space. They concluded that the critical
induction distance with the experimental critical initiation tube diameter was related to the 1D detonation wave
energy and showed a cubic dependency, EcDZND3, as induction zone. As we now know, the detonation cell size is
expected based on the empirical correlation Ecl3 dis- also related to the induction zone length so it is of no
cussed above. surprise that the critical tube diameter is governed by the
Shepherd [94] performed 1D detonation induction zone detonation cell size. Experiments performed with oxyace-
calculations for hydrogen–air–steam mixtures where the tylene mixtures in a round tube by Mitrofanov and
governing equations were integrated along the Rayleigh Soloukhin [137] showed that the critical tube diameter
line. In this way a precise history of the gas element corresponds to the condition where thirteen cells exists
through the detonation front is obtained as shown in across the tube diameter, i.e., dc ¼ 13l. Knystautas et al.
Fig. 14. Shepherd used a 23 equation and 11 species [138] later showed via an extensive series of experiments
kinetics model to show that the ratio of the measured cell where both the critical tube diameter and the detonation
size and the calculated induction zone length, i.e., the cell size were simultaneously measured that this correlation
proportionality constant A, varies significantly over the full applies for all common fuel–air mixtures and thus can be
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 523

considered a universal detonation property. Desbordes and where the temperature time-rate of change of fluid elements
Vachon [139] showed that the critical tube correlation in the flow are tracked through the reaction zone. They
applies for overdriven detonation waves as well. Moen found that decoupling of the shock and the reaction zone
et al. [120] showed that the dc ¼ 13l correlation does not occurs when the Lagrangian derivative of the temperature
apply for detonations that display a regular cellular vanishes after the passage of the shock. This condition is
structure such as heavily argon-diluted mixtures. For these satisfied when the energy release rate and the flow
mixtures they found experimentally that the critical tube unsteadiness due to the deceleration of the leading shock
diameter fell in the range of 13–26l depending on the initial wave are balanced. Their calculations indicate that for the
pressure. Experiments performed by Ciccarelli et al. [140] case of failed transmission front curvature does not play a
with hydrogen–air mixtures at initial temperatures up to critical role. It should be pointed out that in these
600 K reported critical tube measurements in the range of calculations the detonation cellular structure was sup-
18–26lU pressed by the choice of appropriate initial conditions.
As the detonation wave emerges from the tube it Clearly this is a simplification to the actual phenomenon.
diffracts around the corner and in doing so it decays in
strength. The reduction in shock strength results in a drop
in post shock temperature and a corresponding increase in 4. Deflagration-to-detonation transition
the chemical induction time. If the diffraction is severe
enough the leading shock wave decouples from the reaction 4.1. Phenomenology of DDT
zone. The diffraction is most severe for the part of the
detonation wave near the tube wall because of the 901 turn. 4.1.1. Basic studies of DDT
As a result that is the first part of the initial planar Many of the early detonation studies were performed in
detonation wave that fails. As the expansion progresses smooth tubes using a weak ignition source. As a result,
further towards the centerline of the tube more of the detonation initiation was the result of flame acceleration
detonation front is affected. Eventually the expansion leading to DDT. In these studies the objective was the
reaches the centerline of the tube at which point the entire study of the steady detonation wave produced at the end of
detonation wave fails. For mixture near the critical the acceleration process. The flame run-up distance
condition re-initiation occurs behind the diffracted shock required to form a detonation was considered a property
wave before the expansion affects the entire detonation of the mixture. Considering the mathematical expertize of
wave leading to complete failure. This re-initiation takes the time it was much easier to analyze the steady
the form of a local explosion that forms a detonation wave propagation of the detonation wave than the transient
that sweeps across the shocked gas in the diffracted part of flame acceleration phenomenon. This led to the develop-
the detonation wave [141,142]. This local re-initiation ment of the CJ theory for detonation waves. Very little was
phenomenon is similar to what happens at the critical condi- known about the flame acceleration process, as the study of
tion during direct initiation, as discussed in Section 3.2.1. turbulence was in its infancy and the concept of turbulent
Based on photographic and soot foil records Lee [143] flame propagation was even less developed at the time.
proposed that for regular cellular structure detonation Chapman and Wheeler [147] were the first to place
waves failure occurs globally as a result of overall wave obstacles (orifice plates) in a smooth walled tube in order to
curvature whereas for irregular cellular structure detona- promote flame acceleration. Without the orifice plates the
tion waves failure is due to local wave diffraction on the maximum flame velocity achieved in a methane–air mixture
scale of the detonation cell. in a 5 cm inner-diameter tube was on the order of 10 m/s.
There is currently no theory that can predict the critical With the orifice plates placed at one tube diameter spacing
tube correlation dc ¼ 13l that is observed experimentally they measured a maximum flame velocity of over 400 m/s.
for irregular cellular structured detonation waves. Over the Due to the limited diameter of the tube they did not
years there have been several analytical analyses performed observe transition to detonation.
largely based on applying the Shchelkin criterion [144] Shchelkin and his colleagues did very significant work on
and Whitham’s [145] theory of shock diffraction, e.g., see DDT in the next few decades [148–150]. Shchelkin
Ref. [141]. In such analyses it is assumed that the shock proposed that flame acceleration was governed by turbu-
diffraction proceeds independent of the chemical reactions. lent fluctuations in the unburned gas ahead of the flame
Because of the strong dependence of the induction time that led to an increase in flame area. Since the unburned
with the shock strength, based on Schelkin’s criterion [144] gas velocity is related to the flame velocity there is a
local decoupling of the detonation front occurs with a drop feedback loop between the flame velocity and the flame
in shock velocity of roughly 10% of the CJ detonation area that results in efficient flame acceleration. In order to
velocity. Such an analysis neglects the effect of unsteadi- enhance the turbulence in the unburned gas ahead of the
ness on the induction time. More recently high-resolution flame Shchelkin [148] artificially roughened the tube wall
numerical simulations have been used to study the critical by placing a wire coil helix inside of the tube. The so-called
tube diffraction problem. Arienti and Shepherd [146] used Shchelkin spiral is commonly used to reduce the detonation
a novel approach to investigate local detonation failure run-up distance in such applications as PDEs [10].
ARTICLE IN PRESS
524 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

In early studies of run-up distances to DDT in smooth shock wave. The detonation wave sweeps across the
tubes [151–153], flame acceleration and transition to channel and overtakes the shock waves ahead. Also evident
detonation was characterized by discrete flame velocity is the retonation wave that propagates back into the
measurements made using flame time-of-arrival instrumen- combustion products.
tation, or continuous flame and shock velocity measure- An example of a detonation wave formed as a result of
ments using Schlieren streak photography. Introduction of two shock waves coalescing is shown in Fig. 25b. In the
stroboscopic flash Schlieren photography [154–156] made uppermost photograph the flame is just visible on the left
it possible to gain significant insights into the phenomena. top corner. In the previous frame (not shown) two shock
The publication of high-resolution stroboscopic Schlieren waves coalesce to form the leading shock wave correspond-
photographs taken of the onset of transition to detonation ing to the rightmost discontinuity in the uppermost
by Urtiew and Oppenheim [157] represents a milestone in photograph. The coalescing of two shock waves forms a
the study of DDT phenomenon. The photographs were contact discontinuity which in the photograph is located
taken with a light source consisting of a ruby laser with an immediately behind the leading shock wave. In the third
extremely short pulse width (o108 s) and ultrahigh photograph from the top, ignition is observed at the
repetition rate (up to 106 frames per second). The contact discontinuity adjacent to the upper wall. Note the
photographs clearly show the initiation of a detonation contact discontinuity is at an elevated temperature
from a local explosion within the shock–flame complex, the compared to the unburned gas in any other region ahead
so-called ‘‘explosion in the explosion’’. The photographs of the flame. A roughly planar detonation wave forms from
show that this local explosion may occur at the flame front the ignition of the discontinuity.
or at the shock front. After a nearly a sixty year hiatus since the pioneering
An example of detonation initiation occurring at the experiments of Chapman and Wheeler [147] the effect of
flame front is provided in the sequence of photographs in obstacles on flame acceleration leading to DDT received
Fig. 25a. The upper-most photograph shows multiple significant attention in the early 1980s in connection with
shock waves ahead of the highly corrugated turbulent industrial explosion safety. Several papers dealing with
flame. In the preceding photographs, not shown in the flame acceleration in obstacle-filled tubes were presented at
figure, the curved shock immediately ahead of the flame the First International Specialist Meeting on Fuel–Air
originates at the tip of the flame next to the upper wall. The Explosions held in Montreal, Canada in 1981 [158,159]. A
curved shock reflects off the lower wall producing a Mach comprehensive study of flame acceleration and DDT in
stem and reflected shock wave. In the next photograph tubes filled with orifice plates spaced at one tube diameter
from the top, taken 5 ms later, an explosion occurs at the was undertaken at McGill University [86,160]. The
upper tip of the flame that develops into a detonation wave influence of fuel type, mixture composition, tube diameter
complete with closely spaced transverse waves. This and orifice plate blockage ratio (defined as the ratio of the
explosion occurs in an environment of higher pressure orifice plate blockage area and the tube cross-sectional
and temperature produced by the passage of the curved area) on the flame velocity was investigated. The results of

Fig. 25. Stroboscopic Schlieren photographs taken of explosion in the explosion in an equimolar hydrogen–oxygen mixture at initial pressure of
82.7 mmHg in a 1  1.5 in2 channel. Time between frames is 5 ms, distance between marks at bottom lower photograph is 5 cm [157].
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 525

these studies indicated that the transition process requires shape of the laminar flame. This is different from a ‘‘tulip’’
that (i) the flame should accelerate to a sufficiently high shaped turbulent flame described in the following discus-
speed prior to the onset of detonation, i.e., at least of the sion.
order of the speed of sound in the combustion products; It has been demonstrated by Shchelkin [150], Salamandra
and (ii) the orifice opening diameter must be sufficiently et al. [154] and Soloukhin [155] that flame acceleration in
large, i.e., of the order of the detonation cell width of the relatively smooth tubes is strongly affected by wall rough-
mixture in the tube. ness. Fig. 26 shows qualitatively the evolution of the flow
ahead of the accelerated flame. Initially, the flow may
4.1.2. Phases of DDT process have a smooth axial velocity distribution V(x), as shown in
Following the detailed reviews of characteristic features Fig. 26a. In sufficiently long tubes, the velocity profile is
of DDT processes by Lee and Moen [2] and more recently expected to steepen with time (Fig. 26b) and at some stage
by Shepherd and Lee [3], the DDT phenomena is usually a shock forms ahead of the flame. Depending on the details
divided into two separate phases (1) the creation of of the flame acceleration, the shock may form at the head
conditions for the onset of detonation by processes of flame of the compression wave, which propagates with the sound
acceleration, vorticity production, formation of jets, and speed, ar, as shown in Fig. 26b, or somewhere closer to the
mixing of products and reactants, (2) the actual formation flame. At the same time the flow interaction with tube walls
of the detonation wave itself or the onset of detonation. results in the formation of a turbulent boundary layer. The
Processes in the first phase are particular to the specific thickness of the layer grows with time. The boundary layer
initial and boundary conditions of the problem. Different starts to appear in the compression wave and continues to
physical mechanisms dominate the process of flame grow after the lead shock is formed. At a later stage, a
acceleration in obstructed channels, smooth tubes or large situation shown in Fig. 26c is developed with the lead
volumes filled with combustible mixtures. However, the shock and growing boundary layer behind it. Interactions
actual formation of the detonation appears to be a more of the flame with boundary layer results in a significant
universal phenomenon. increase of the burning rate near the tube walls, and a
The following description is divided into three parts, the characteristic ‘‘tulip’’ flame shape is formed, as first
first part deals with flame acceleration in smooth tubes, the reported by Salamandra et al. [154]. A sequence of
second focuses on the same phenomenon in channels photographs showing the propagation of a ‘‘tulip-shaped’’
equipped with obstacles, and the third one deals with the flame is provided in Fig. 27 [165].
onset of detonation. This is followed by a summary of While the flame propagates along the tube, turbulence is
the framework that may be used for the evaluation of the also generated in the core flow. All these processes may
potential for FA and the onset of detonation. result in a variety of flame shapes in the tube depending on
the mixture properties and the tube size. The tulip shape
4.2. FA in smooth tubes may be formed and destroyed at a later stage of flame
propagation.
4.2.1. Process of FA in smooth tubes The larger is the roughness of the tube wall, the faster is
Following weak ignition, the initially smooth surface of the growth of the boundary layer, with other conditions
a laminar flame can be wrinkled due to the L–D instability, being the same. For this reason the roughness of the tube
which can be stabilized or destabilized by the thermal- wall is an important parameter that controls the rate of
diffusion effects. This can result in the formation of a flame acceleration. In obstructed tubes with relatively small
cellular flame. With the development of the cellular blockage ratio (BR), i.e., less than roughly 10%, the
instability (see Fig. 10) the flame surface grows and the characteristic features of the flame acceleration process and
flow generated due to the expansion of the combustion the flame structures are similar to those in smooth tubes.
products accelerates. However, the growth of perturbations Obstacles increase the roughness of the tube wall resulting
arising from the L–D instability is limited, and the flame in relatively fast growth of the boundary layer.
surface can only have a slow growth due to this instability.
Thermal expansion of the combustion products produces
movement in the unburned gas. The flow interaction with 4.2.2. Run-up distances to supersonic flames
the confinement causes an increase of the flame surface. In sufficiently long tubes and in typical fuel–air mixtures,
This results in the moderate increase of the flow velocity flames accelerate up to the speed of about 600–1000 m/s [2].
and flame speed relative to a fixed observer. This, initially If the tube diameter is sufficiently large, as described in
laminar, phase of FA was analyzed by Bychkov et al. Section 4.4.4, transition to detonation can be observed as
[161–163] and Akkerman et al. [164]. The effects related to soon as the flame reaches a speed of the order of the sound
the flame surface evolution may lead to a slight decelera- speed of the combustion products. In smooth tubes the
tion of the flame when the flame reaches the tube walls. The onset of detonation is often observed as soon as the flame
heat losses to the tube walls may result in additional flame reaches a speed of about 600–1000 m/s. Although there are
deceleration [154]. The same effects may also be responsible situations when additional propagation distance is neces-
for the temporary development of a characteristic ‘‘tulip’’ sary to establish the detonation regime, most of the data
ARTICLE IN PRESS
526 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

V(x)

Flame
Δ
b. l.

V(x)
SW

Flame
Δ

V(x)

SW

Flame
Δ
b. l.

Flame(t1) Flame(t2) Flame(t3)

Δ(t3)
Δ(t1) Δ(t2) Δ(x)

b. l. b. l. b. l.

Fig. 26. Sketch showing development of boundary layer ahead of accelerated flame with time (a, b, and c). Curves V(x) show qualitatively the distribution
of flow velocity ahead of the flame; SW designates shock wave; b. l.—boundary layer; D—boundary layer thickness at flame position. Part ‘d’ shows flame
positions and boundary layers at three different moments of time t1, t2, and t3. Curve D(x) shows boundary layer thickness at flame positions as a function
of distance.

available on the run-up distances refers to the run-up observed near the end of the tubes. In these cases mixture
distance to detonation. pre-compression effects may be important, so it is difficult
There are some early experimental data on the effects of to describe the mixture properties as that of the initial
tube diameter, initial pressure, and temperature on the run- mixture. Also, as it was found by Salamandra et al. [169],
up distance to detonation, XD, for tubes without obstacles the transition to detonation in short tubes can be facilitated
[149,151–153,166–168]. These data show a decrease of the by the flame interaction with the pressure waves reflected
run-up distance with increase of the initial mixture from tube endwall. In addition, the tube roughness, which
pressure, P. If such a dependence is expressed in a power plays an important role, was not always sufficiently
form, i.e., XDpPm, the exponent, m, depends on mixture characterized.
properties and lies in the range from 0.4 to 0.8 for various More recent studies of Lindstedt and Michels [170] and
mixtures for the pressure range from 0.1 to 6.5 bar. Kuznetsov et al. [88,165,171,172] were made in sufficiently
In some of these studies, an increase of the run-up long tubes and with the tube roughness characterized.
distance with tube diameter was reported [151,167]. The These studies give a detailed description of the flame
ratio of the run-up distance to the tube diameter XD/D was acceleration process so that flame speed as a function of
found to be in the range from 15 to 40. On the contrary, distance is available. This data makes it possible to analyze
Bollinger et al. [167] noted that only for rich mixtures of the effect of mixture properties, tube size and roughness on
hydrogen the run-up distance depends on the tube the run-up distance to supersonic flames. Such an analysis
diameter. Somewhat ambiguous data on the effect of the was recently presented in [173], where the distance
tube diameter may be due to the hidden influence of other necessary for the flame to reach a speed equal to the
factors, such as tube roughness and tube length. In many of sound speed in combustion products, ap, was analyzed in
the tests mentioned above the onset of detonation was the framework of a simple analytical model.
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 527

turbulent velocity correlation of Bradley [43] (see


Eq. (2.25)), and the description for the thickness of the
boundary layer, D, at the flame position along the tube
[165]. The run-up distance, XS, is defined as the flame
propagation distance where the flame speed reaches the
sound speed in the combustion products. This approach
results in the following expression for the run-up distance:
   
XS g 1 D
¼ ln g þK , (4.1)
D C k h
where k, K and C are constants (independent of mixture
composition) taken to be: k ¼ 0.4; K ¼ 5.5; and C ¼ 0.2
[165,171]; D/hd can be expressed through the blockage
ratio: D/h=2/(1(1BR)1/2); and g=D/D is given by
"  1=3 #1=ð2mþ7=3Þ
ap d
g¼ , (4.2)
Zðs  1Þ2 SL D
where SL is the laminar burning velocity; d=n/SL is the
laminar flame thickness; n the kinematic viscosity; Z and m
are two unknown parameters, which were determined using
the set of experimental data [88,165,170–172]: Z=2.1 and
Fig. 27. Sequence of shadow photographs (0.1 ms between frames) m=0.18. The model was shown to describe the experi-
showing boundary layers ahead of accelerated flame. Flame propagates mental data with an accuracy of prediction for the run-up
from left to right; speed of lead flame edges is about 320 m/s. Boundary
layers are seen as dark regions on the top wall and as lighter regions in the
distances of about 725%.
bottom wall of the channel. Wall roughness is 1 mm; mixture is According to Eqs. (4.1) and (4.2) the dimensionless run-
stoichiometric H2–O2 at 0.6 bar initial pressure. up distance, XS/D, is a function of the tube diameter due to
the term d/D in Eq. (4.2). Fig. 29 shows, as an illustration,
the dimensionless run-up distances for stoichiometric
ST + V mixtures of methane, propane, ethylene, and hydrogen
with air versus tube diameters computed from Eqs. (4.1)
and (4.2). Fig. 29 shows that in smooth tubes the run-up
D V
distances in methane and propane mixtures are about twice
as high as that in ethylene and hydrogen mixtures. This is
why there is no data available on the run-up distances for
Boundary layer these mixtures in smooth tubes. There is some data on FA
in ‘‘turbulent gas/air mixtures’’ of hydrogen, methane, and
X Δ d propane, where turbulence preexists in the mixture at the
time of ignition [174]. The data cannot be compared this
Fig. 28. Schematic of the problem.
data directly with the model predictions. However, if one
assumes that the preexisting turbulence leads to the
Fig. 28 shows the schematic of a flame in a tube with increase of the initial burning velocity to some effective
diameter D and wall roughness h at a distance X from the
200
ignition point, which was the basis for the model in [173]. At
180 BR = 0.01 H2 BR<0.1
the stage of flame propagation shown in Fig. 28 the boundary C2H4 BR<0.1
160
layer is formed ahead of the flame with a thickness D. The C3H8 BR<0.1
140 CH4 BR<0.1
flame propagates in the boundary layer with a turbulent
XS/D model

120
velocity ST relative to the unburned mixture and with a 100
velocity ST+V in the laboratory frame, where V is the flow 80
speed ahead of the flame. The burning velocity in the core of 60
the flow is lower than the value of ST in the boundary layer. 40
The thickness of the boundary layer grows with time while 20
the flow interacts with the wall, resulting in an increase of the 0
0.01 0.1 1 10
boundary layer thickness as measured at various flame
D, m
positions along the tube, as shown in [165,171].
The model utilizes a general expression for the flow Fig. 29. Run-up distances over tube diameter as a function of tube
balance in the tube with two unknown parameters, the diameter for BR ¼ 0.01.
ARTICLE IN PRESS
528 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

value, SLeff, such a comparison suggests that SLeff ¼ 2.5SL. combustion products produces movement in the unburned
This is a reasonable value and so the ‘‘turbulent’’ data [174] gas. The flow around the obstacles leads to a rapid increase
is also well described by the model. of the flame surface (see Figs. 30 and 31). This effect results
in the further increase of the flow velocity, flame speed
4.3. FA in obstructed channels relative to a fixed observer, and flame surface. This
feedback loop can result in continuous flame acceleration
4.3.1. Process of FA in obstructed channels (note that this loop is not connected to any turbulence
Qualitatively, the mechanism for flame acceleration in effect). In addition to ‘‘geometrical’’ increase of the flame
obstructed channels is well understood. The processes surface, K–H and R–M instabilities lead to additional
involved in FA in obstructed channels is shown in Figs. 30 flame surface increase.
and 31 using shadow photographs taken of the flame at A novel technique was used by Johansen and Ciccarelli
different times [34,35]. [175] to visualize the development of the unburned gas flow
Initially a smooth, or wrinkled, laminar flame develops field ahead of the flame propagating in a square duct
depending on the mixture properties. This is due to the equipped with 2D obstacles. Before ignition at the closed
L–D instability, which can be stabilized or destabilized by end of the duct, helium gas is injected between adjacent
the thermal-diffusion effects. While the cellular instability obstacles, as shown in Fig. 32a. This creates a pocket of
(if developed as in Fig. 30) provide slow growth of the low-density gas mixture consisting of the test gas and
flame surface with flame propagation, obstacles located helium. Once the test gas is ignited on the left end of
along the path of the expanding flame can cause rapid the duct, a flow is induced in the entire duct causing the
increase of the flame surface. Thermal expansion of the hot helium-diluted gas to be convected downstream. At the

Fig. 30. Sequence of shadow photographs of flame propagation in 10% H2–air mixture with BR ¼ 0.6 [34,35]. Channel width is 80 mm. Combined from
images taken from three different windows along the channel. Times after ignition in ms are shown on the right.

Fig. 31. Sequence of shadow photographs of flame propagation in 70% H2–air mixture with BR=0.6 [34,35]. Channel width is 80 mm. Combined from
images taken from three different windows along the channel. Times after ignition in ms are shown on the right.
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 529

Helium
i) Unburned Jet Unburned Vunburned = 0 m/s
Gas Gas

0 ms

ii) Vunburned

10 ms

iii) Vflame

14 ms

iv) Vflame

18 ms

Fig. 32. Unburned gas flow field visualization: (a) schematic showing helium gas injection ahead of the flame front, and (b) a sequence of Schlieren images
employing this technique [175].

same time undiluted test gas is convected into the volume Depending on mixture properties and boundary condi-
originally occupied by the helium-diluted gas. The sudden tions, the interaction of the flame with turbulence in the
onset of flow produces a pair of laminar vortices on the unburned gas can lead to either weak FA resulting in
downstream edge of each obstacle, as shown schematically relatively slow unstable turbulent flame regimes, see
in Fig. 32a and in the Schlieren photographs provided in Fig. 33, or to strong FA resulting in fast flames propagating
Fig. 32b. These eddies grow into large recirculation zones with supersonic speed relative to a fixed observer,
that eventually occupy the volume between adjacent see Fig. 34.
obstacles on the same side of the duct. In time, a turbulent The slow unstable turbulent flames that are shown in
shear layer develops separating the recirculation zone and Fig. 33 propagate through highly turbulent media at
the core flow as seen in the bottom photograph in Fig. 32b. conditions close to quenching. Often incomplete combus-
If the flame reaches a eddy in its early development phase, tion is observed, or the reaction quenches globally at some
e.g., for obstacles near the ignition end, the flame is stage of flame propagation. Also these flames often move
entrained into the vortex and subsequently burns it out, see back and forth, locally consuming unreacted material left
the uppermost Schlieren photographs in Figs. 30 and 31. If behind as shown in Fig. 33.
the flame reaches an vortex after the shear layer has The fast turbulent flames, e.g., as shown in Fig. 34,
developed the flame propagates in the core flow and then which result from strong FA are supersonic combustion
burns into the recirculation zone. At a later stage of flame waves consisting of a lead shock, or a system of shocks,
propagation, a significant randomization of the flow ahead followed by a turbulent flame brush. Although the burning
of the flame can occur. This results in the development of rate in the flame brush is controlled by the molecular
intense turbulence in the flow as shown in the upper-most transport of energy and species typical of deflagrations,
photo in Fig. 33. overall energy release rate is high enough to result in the
Turbulence increases the local burning rate by increasing supersonic flame propagation relative to a fixed observer.
both the surface area of the flame and the transport of local The flame speed in this regime is often close to the sound
mass and energy. An overall higher burning rate, in turn, speed in combustion products [86,160], although there are
produces a higher flow velocity in the unburned gas. This cases of unstable propagation with strong velocity oscilla-
‘‘turbulent’’ feedback loop can result in further acceleration tions [45,88]. Under appropriate conditions, a transition
of the propagating flame. At the same time, the ‘‘geometrical’’ from this fast turbulent deflagration regime to detonation
increase of the flame surface continues to play a role. The can be observed.
increase of the burning velocity, ST, with turbulence level is
generally limited and saturation of ST is usually reached at a 4.3.2. Characteristic flame propagation regimes
level of 10–20 times the laminar burning velocity, SL [68,71]. It was first shown by Lee et al. [86] and Peraldi et al.
Further increase of turbulence intensity can cause local [160] that several characteristic flame propagation regimes
quenching of the combustion process, which results in a exist for tubes filled with orifice plates. A typical data set
decrease in the effective energy release rate. for propane–air at atmospheric conditions in different size
ARTICLE IN PRESS
530 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

Fig. 34. Shadow photographs showing structure of fast deflagration in


70% H2–air mixture with BR=0.6 [34,35]. Turbulent flame is preceded
with system of shocks. Channel width is 80 mm. Left boundary is 2.4 m
from ignition. Top and bottom images are taken in the same test (times
after ignition in ms are shown on the right), and average speed of flame
propagation is about 800 m/s.

In the quenching regime the flame accelerates for a


certain distance and is then extinguished. As suggested in
[86], this regime features flame propagation as a result of
the sequential ignition of the gas in the connected
‘‘chambers’’ created by the higher blockage ratio orifice
plates. Ignition in a chamber occurs as a result of the jetting
of the combustion products from the preceding chamber
into the unburned gas contained in the following chamber.
Lee et al. [86] proposed that flame propagation terminates
in a chamber when quenching occurs as a result of the
Fig. 33. Shadow photographs showing structure of slow deflagration in mixing time being shorter than the chemical time. In more
10% H2–air mixture with BR ¼ 0.6 [34,35]. Flame tong propagates trough
recent studies [34,35], this quenching behavior was
the central part of the channel (with a speed of about 100 m/s relative to
fixed observer) leaving unreacted material behind. Then flame propagates observed in channels with BR=0.9. However, for BR of
back through partially unreacted material. Channel width is 80 mm. Left 0.6 and 0.3 there was no jetting observed, instead chamber-
boundary is 1.6 m from ignition, times after ignition in ms are shown on to-chamber combustion was observed and the flame
the right. structure corresponded to that shown in Fig. 33, where
local quenching of the flame is observed. Such a behavior
resulted in incomplete combustion and/or global flame
tubes with similar blockage ratio orifice plates is provided quenching.
in Fig. 35. Based on the measured terminal flame velocity, If flame quenching does not occur the flame typically
four propagation regimes were observed: the quenching, accelerates to a steady velocity on the order of the speed of
the choking, the quasi-detonation and the CJ detonation sound of the combustion products, e.g., 600–1000 m/s. In a
regimes. smooth tube a flame that accelerates to such velocities
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 531

substantial velocity deficit compared to the CJ detonation


velocity, this is referred to as the ‘‘quasi-detonation’’
regime. The results show that the larger the blockage ratio
the more significant the velocity deficit. Experiments
performed by Gu et al. [179] with orifice plate spacings of
0.5, 1, and 2 tube diameters showed that the velocity deficit
is inversely proportional to the orifice plate spacing. In
smooth tubes a very small velocity deficit, typically 1–2%,
is incurred by the detonation due to boundary layer effects.
Therefore, one may deduce that the much larger velocity
deficit observed in obstacle-filled tubes is due to enhanced
momentum and heat losses associated with the significant
wall roughness [179]. However, for large blockage ratio
orifice plates it has been observed that the velocity deficit is
due to the detonation wave failing and re-initiating as it
propagates through the orifice plates [176]. Detonation re-
initiation occurs as a result of shock reflection off the
orifice plate face or the tube wall [176]. If the orifice plate
diameter is more than the critical tube diameter (dX13l)
then the detonation wave propagates at roughly the CJ
velocity. Recall the critical tube diameter criterion implies
that the detonation wave propagates independent of the
tube wall boundary.
The terms ‘‘chocked flame’’ or ‘‘choking regime’’ are
often used to characterize fast flames in ducts with
obstacles that travel at a speed just below the isobaric
sound speed in the combustion products. There is
considerable uncertainty in the understanding of the
propagation mechanism of a high-speed turbulent defla-
gration in an obstacle-laden tube and in the meaning of
the term ‘‘choking’’. It should first be pointed out that the
Fig. 35. Variation of the terminal flame velocity with mixture composition
for the propane–air system [160]. larger the blockage ratio (up to about BR ¼ 0.75), the
more distinctive is the choking regime [88]. This regime is
also more pronounced for orifice plate obstacles, than with
normally transits to a detonation. However, in an obstacle- less regular obstacle configurations, as suggested by Chao
laden tube turbulence can sustain the flame at such a and Lee [180]. Although there may be a dependence of the
velocity indefinitely, Lee et al. refers to this as the choking fast flame propagation velocity on the obstacle geometry,
regime. At these high velocities compression waves form this fast flame regime can be distinguished from quasi-
ahead of the turbulent flame brush, see Fig. 34. Unlike in a detonation by a relatively low strength lead shock which is
smooth tube where the flame and shock structure is not sufficient to cause mixture autoignition, as shown by
relatively orderly, in an obstacle-laden tube the shock Teodorczyk et al. [176,177]. According to the authors of
waves ahead of the flame undergo periodic reflections off [176,180,177], the mechanism responsible for ignition in
the obstacles and the tube wall and thus are dispersed in all high-speed deflagrations is turbulent mixing of hot
directions (see Fig. 25 and photographs by Teodorczyk et combustion products and reactants promoted by the
al. [176,177]). Chue et al. [178] proposed that this complex obstacles. It is also suggested [176,180,177] that a regular
shock–flame structure can be modeled as a CJ deflagration. turbulent flame is too slow to be responsible for the fast
They argue that in the choked regime flame propagation is burning rate required for the propagation at high super-
governed more by the mixture energy content than the sonic speeds.
turbulent transport rates. Their argument is supported by Another point of view concerning the fast flame
experimental observations that the final steady flame propagation mechanism was suggested by Veser et al.
velocity does not vary with tube diameter and wall [181]. They argued that the relatively low burning velocity
roughness. of a regular turbulent flame (i.e., STE 10SL) can provide
If the orifice plate diameter is larger than the mixture cell the required burning rate because of the development of
size (dXl) then DDT can occur [160]. For propane–air in a sufficiently large flame surface area (see the flame brush in
5 cm tube, see Fig. 35, this criterion is not satisfied and thus Figs. 30, 31, and 33). Moreover, 3D numerical simulations
there is no detonation propagation regime. The detonation of flame propagation with constant prescribed turbulent
produced in the 15 cm and 30 cm tubes propagates at a burning velocity, ST, in a tube with orifice plate obstacles
ARTICLE IN PRESS
532 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

[181] showed that the surface of the flame adjusts itself in and its deficit, as compared to the value of ap, is expected
such a way that the flame surface is higher for lower ST be mixture-specific for a given geometry in accordance with
values and vise versa. This results in similar values of the the experimental observations [88].
flame speed in the laboratory frame (close to ap).
What effect is responsible for the observation that the 4.3.3. Flame speeds as a function of distance
flame speed is close to the isobaric sound speed in combustion A series of experimental studies were made recently by
products? What is the meaning of term choking? Chao and Kuznetsov et al. investigating the behavior of turbulent
Lee [180] argued that in a tube with orifice plate obstacles, a flames in mixtures based on hydrogen fuel [88], and
high-speed flame propagates as consequence of the sequential hydrocarbon fuels [182]. It was found that in tubes containing
venting of the combustion products through choked orifices different configurations of obstacles, a well-defined difference
and subsequently travels at the local sound speed of the in flame behavior can be observed between slow, subsonic
combustion products. The emphasis on the orifice plates, flames and fast combustion phenomena such as choked
however, cannot explain some of their experimental data flames and detonations. Figs. 36 and 37 show flame
[180], where pronounced choked flames were observed with propagation velocities versus distance along the tubes with
an array of cylindrical obstacles. Chao and Lee [180] go BR=0.3 and 0.6 for hydrogen–air mixtures. Figs. 38 and 39
further suggesting that the choking regime with orifices show similar plots for hydrocarbon mixtures.
cannot be considered to be an actual propagation regime, The data presented in Figs. 36–39 show that flame
since the mechanism is controlled by gasdynamic jetting and acceleration results in one of three different propagation
not by turbulent combustion. regimes: slow subsonic flame, fast supersonic flame (choked
From the point of view suggested by Veser et al. [181], flame), and quasi-detonation. It is seen that in some cases
the turbulent flame brush surface increases during flame of slow flames global quenching is observed. Generally this
propagation in an obstacle field following the ignition at is in qualitative accord with the classification suggested by
the closed end of a duct. This results in an increase of the Peraldi et al. [160]. The most pronounced difference is
speed of the leading tip of the flame in the laboratory observed between cases of slow and fast flames. This is true
flame. The flow velocity of the combustion products in the for all mixtures and blockage ratios tested in [88,182]. The
direction of flame propagation relative to a fixed observer, difference between fast flames and quasi-detonations is not
u3, reaches a maximum at the flame tip and decreases to very well defined for BR=0.3. It was found [88,182] that
zero at some distance behind the flame. As soon as the for large BR of 0.6 and 0.9 more unstable slow regimes
combustion products flow velocity reaches the local sound resulted as compared to BRo0.6. Global quenching was
speed, the flow behind the flame tip is choked. Any not observed for BRp0.1. Constant velocity fast flames,
pressure perturbations generated due to the combustion i.e., chocked flames, were not observed for BRp0.1.
processes behind the surface where the difference between
the local flow speed, u3, and its maximum at the flame tip is
equal to the local sound speed, cannot reach the flame tip.
Thus, these perturbations cannot contribute to any further 2000
increase of the flame surface area. As a result, the turbulent 174 mm 80 mm
BR=0.3 (air)
10%H2 10%H2
flame brush located between the lead flame tip and the 11%H2 13%H2 520 mm
sonic surface propagates as a whole with both the speed 1600 12%H2 10%H2
and the flame surface being nearly constant. This explana- 13%H2 13%H2

tion is similar to the one suggested by Chao and Lee [180], 15%H2
17.5%H2
in the sense that it is the flow of combustion products
behind the lead flame tip that is chocked in the high-speed 1200
v, m/s

turbulent flames, although without the emphasis on the


orifice plates. The important difference is that Veser et al.
[181] suggest that turbulent combustion does play a role in 800
the propagation mechanism of a choked flame by affecting
the surface area of the turbulent flame brush, i.e., the flame
Fast flames
area is higher for lower values of ST and vise versa.
400
Therefore, the structure of the high-speed flame, such as
shown in Fig. 34, appears to be affected by the range of Slow flames
values of ST for a given mixture. Spatial distribution of
flow velocities and local sound speeds behind the flame tip 0
depends on the flame structure. Because the choking occurs 0 10 20 30 40 50 60 70
when the difference of flow velocities (as explained above) x/D
equates to the local sound speed, the maximum flow speed Fig. 36. Flame velocities for lean hydrogen–air mixtures versus dimen-
at choking conditions in its turn depends on wave structure sionless distance along tubes of 174, and 520 mm id, and square channel of
and on ST. Consequently, the actual propagation velocity 80  80 mm2 cross section (BR ¼ 0.3).
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 533

1600 1000
H2- air C3H8-air
BR = 0.6 Quasi-detonations D=174 mm
BR=0.6
800 Fast flames
1200
80 mm 174 mm
10%H2 10%H2

Flame Velocity (m/s)


11%H2 11%H2
13%H2 15%H2 600
25%H2
v, m/s

520 mm 45%H2
800 10%H2
11%H2
Fast flames 2%
400
2.2%
3%
400 4.03%
200 6.5%
Slow flames 7%
7.5%
Slow flames
0 0
0 10 20 30 40 50 60 70 0 4 8 12
x/D Distance (m)

Fig. 37. Flame velocities versus distance in obstructed tubes filled with Fig. 39. Flame velocities versus distance in propane–air mixtures in
hydrogen–air mixtures in tubes of 174, and 520 mm id, and square channel 174-mm tube (BR ¼ 0.6).
of 80  80 mm2 cross section (BR ¼ 0.6).
well below the sound velocity in reactants. This means
that an energy release rate, defined by the flame surface
2500
2% area, chemistry and turbulence, remains limited in cases
C3H8-air
2.2% D=174 mm of weak FA, and the flame is not able to generate signi-
2.5%
3%
BR=0.3 ficant compression waves during the process. In the
2000 4.03% Quasi-detonations cases of strong FA, the energy release rate appears
6.5%
7% to be high enough that the flames are able to generate
7.5% strong compression and shock waves ahead, resulting in
Flame Velocity (m/s)

8%
1500
fast supersonic deflagrations and, possibly, transition to
detonation.
The possibility of transition from fast deflagrations to
detonations is controlled by processes that are typical of
1000 fast flames and detonations, including shock waves and
chemical reactions induced by compression. The critical
condition for the onset of detonation includes the affect of
Fast flames
500 the mixture detonation sensitivity, such as the detonation
cell width. These critical conditions will be discussed in
Slow flames Section 4.4. As distinct from the onset of detonations, the
nature of the boundary between weak FA and strong FA
0
cannot depend on processes typical of strong deflagrations
0 4 8 12
Distance from Ignition Point (m)
and detonations. The ‘‘decision’’ on whether a flame
accelerates strongly or not occurs at rather low, subsonic
Fig. 38. Flame velocities versus distance in propane–air mixtures in flame speeds. The important mechanisms for this stage of
174-mm tube (BR ¼ 0.3). FA are basically the turbulent transport of energy and
species typical for slow deflagrations. Characteristic para-
Instead continuous FA was observed, making the flame meters of the mixture detonation sensitivity, i.e., the
behavior similar to that in smooth tubes. detonation cell width, is not an important contributor to
While there is a sharp difference between cases of slow the critical condition for strong FA.
and fast flames, there is also a difference in the process of
flame acceleration itself, but not the final steady propaga-
tion regime, which can be characterized as weak or strong. 4.3.4. Criteria for weak and strong FA
In the cases of weak FA a significant increase of a flame Dorofeev et al. [45] analyzed the experimental data on
speed from initial values characteristic to the laminar FA in hydrogen combustibles obtained by Sherman et al.
flame is observed, but the propagation velocity remains [183,184], Dorofeev et al. [185,186], Ciccarelli et al. [187],
ARTICLE IN PRESS
534 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

and Kuznetsov et al. [88]. A set of parameters was ratio, s, and b. It is seen that the critical s increases with b
considered that can influence the flame-unburned gas flow for mixtures with positive and negative Mab. Correspond-
feedback mechanism in obstructed ducts. Among these are ing critical conditions were suggested in the form: s4s*
parameters given by properties of the mixture and (b, LT/d), where b, is the Zeldovich number, LT is the
geometry, such as; the laminar burning rate and flame integral scale of turbulence and d is the laminar flame
thickness (i.e., SL and d), the gas expansion ratio (i.e., ratio thickness. It was also found that for a sufficiently large-
of reactant density to product density, s), characteristic scale ratio, LT/d, exceeding two orders of magnitude, the
sound speeds, ratio of specific heats, Lewis number, critical conditions become independent of scale and can be
Markstein number, and Zeldovich number. These para- reduced to a function of the mixture composition only
meters, which can be determined a priori, define a sort of s4s*(b).
potential for flame acceleration. It was found that a The correlations presented in Figs. 41 and 42 show
sufficiently large mixture expansion ratio is necessary for qualitatively that the critical expansion ratio for strong FA
the development of fast flames. increases with dimensionless activation energy. It should be
The experimental data [88,183–187] are summarized in noted that the values of b and Mab, depend on the model
Fig. 40 in the form of a graph with the expansion ratio and used for their extraction, e.g., in [45] these values were
the Markstein number defining the mixture initial state (see extracted using experimental laminar flame velocity mea-
[45] for details). The data points denoted by black circles surements. One should use the same model to apply data of
correspond to tests that resulted in strong FA and data Figs. 41 and 42 to back-calculate the critical mixture
points denoted by gray circles correspond to weak FA. It is composition for strong FA. It should be noted that the
seen that the critical mixture expansion ratio depends on difference in the critical condition between stable and
the value of Ma number (see Section 2.2.4), which is unstable flames, see Figs. 41 and 42, may also include the
defined relative to burnt mixture Mab. While a variation of Lewis number effect. This effect was not addressed in
the critical s with negative Mab is seen in the plot, the sufficient detail in [45], because of uncertainty in the model
critical s was nearly constant for positive Mab. This was used for determination of b.
suggested to be due to the similar values of Zeldovich Probably the most convincing experimental argument
numbers, b, for all mixtures with positive Mab, and a for the importance of the expansion ratio as a parameter
strong variation of b for negative Mab. characterizing a potential for strong FA was presented by
In order to increase the range of b for mixtures with Alekseev et al. [188]. In a series of tests FA was studied in
positive Mab, hydrogen fuels were amended in analysis [45] obstructed tubes with variable venting on the lateral
with hydrocarbons from Peraldi et al. [160]. Figs. 41 and 42 surface of the tubes. In these tests the natural expansion
show the combustion regimes as a function of expansion of the combustion products, quantified by the mixture

6 6
Ma < 0
slow flames
choked flames
5 5 and detonations

4 4
σ

3 3

2 2
slow flames
choked flames
and detonations

1 1
-4 -2 0 2 4 6 8 10 12 2 3 4 5 6
Mab β

Fig. 40. Resulting combustion regime as a function of expansion ratio, s, Fig. 41. Resulting combustion regime as a function of expansion ratio s
and Mab. Black points—fast (strong FA) and gray points—slow and b for mixtures with Mabo0. Black points—fast (strong FA) and gray
combustion regimes (weak FA). points—slow combustion regimes (weak FA).
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 535

9 generated flow of the unburned gas and the flame itself.


Turbulence generated in the flow ahead of the flame
8 CH - fuels interacts with the flame. Initially, this interaction results in
a positive feedback through increase of the flame surface,
flow speed, and turbulence level. At a certain stage, the
7
Ka-number may become greater than unity, and mixed
reactants/products pockets may be formed in the flow.
6 Then, if quenching of the largest mixed pockets dominates
(left-hand side of Eq. (4.3) less than unity), all mixed
5 pockets are quenched, the overall rate of energy release is
σ

suppressed, and the positive feedback is destroyed. If


4 re-ignition of the largest mixed pockets dominates, mixing
of products and reactants by turbulence does not affect
significantly the rate of energy release, and the positive
3 Ma > 0 feedback is sustained. This argument supports the view
H2 mixtures slow flames
choked flames that the critical conditions defined by Eq. (4.3) may be
2 and detonations interpreted as the borderline between the cases of weak and
+- 8% deviation strong FA.
1 The critical conditions given by Eq. (4.3) are plotted in
3 4 5 6 7 8 9 10 11 12 Fig. 43, which can be compared with experimental data in
β Figs. 43 and 42. The first observation is that the order of
magnitude and the actual values for the critical mixture
Fig. 42. Resulting combustion regime as a function of expansion ratio s
and b for mixtures with Mab40. Black points—fast (strong FA) and gray
expansion ratio are close to those obtained experimentally.
points—slow combustion regimes (weak FA). The general trend and the order of magnitude of the
function of s on b, presented in Fig. 43 are very similar to
those given in Figs. 41 and 42.
expansion ratio, and the combustion product venting were It is seen that the model predicts reasonable values and
the competing processes. It was found that to provide correct trends for the critical boundary between the cases
conditions for strong FA in vented tubes s should be of weak and strong FA. A detailed comparison would
increased in proportion to the loss of expansion due to require a more reliable estimate for the mixture properties
venting of combustion products. This observation suggests such as s, b, n, and Le.
that s as a measure of expansion is an important parameter
for description of the critical conditions for strong FA. n=1
The effect of the dimensionless activation energy b on 11
the critical conditions for strong FA was related by Le = 0.3
9
Dorofeev et al. [45] with quenching of products/reactants Le = 1
pockets mixed by turbulence. Qualitatively, a high b 7 Le = 2
σ

provides a stronger ability of turbulent mixing to quench 5


locally the combustion process. More recently this problem
was revisited [78] with more detailed analysis of quenching 3
for a range of sizes of products/reactants pockets mixed by 1
turbulence. It was found that the critical conditions for 6 9 12 15
quenching/re-ignition of the largest mixed eddies appear to β
be independent of any parameters characterizing turbulence:
Le = 1
s2 b2 ðb=2  1Þn e1b=2 11
¼ 1, (4.3) n = 0.5
6Len Gnþ1 m 9
n=1
where mE20 is a constant. Only mixture properties, such as 7 n=2
s, b, n, and Le, affect the thermal regime of the largest mixed
σ

pockets. This is a remarkable result, which shows that 5


mixture properties may prescribe certain types of flame 3
behavior in various turbulent flows. It was also shown [78]
that all mixed pockets are expected to be quenched in sub- 1
critical mixtures (the left-hand side of Eq. (4.3) less than 6 9 12 15
β
unity) in any turbulent flows with KaX1.
If one considers the process of flame acceleration, its Fig. 43. Critical conditions for quenching and re-ignition of the largest
effectiveness depends on the feedback between flame mixed pockets expressed as s versus b for various n and Le [78].
ARTICLE IN PRESS
536 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

Both the model and the data show that the critical Turbulent flame brush
mixture expansion ratio increases with b and Le. This
means that unstable mixtures (Leo1) and mixtures with ST
low dimensionless effective activation energy (such as lean Ω U
hydrogen mixtures at high initial temperatures) will show
strong FA in compositions with low energy content (s of
R
about 2 and more). At the same time, stable mixtures and
mixtures with high dimensionless effective activation
energy (such as rich methane mixtures at low initial
temperatures) can only sustain strong FA in compositions X
with high-energy content (s of about 7 and more).
Fig. 44. Schematic for estimation of flame surface O in tube with orifice
4.3.5. Run-up distances to supersonic flames plates.
If strong FA and the development of fast deflagrations
are possible, an important problem to be solved is the velocity reaches its maximum saturation value of the order
evaluation of the minimum run-up distance for the of STE10SL at an initial stage of FA, from that point
development of such fast combustion waves. Most of on flame acceleration is mainly due to the increase of
the studies of FA and DDT are related to the problem the surface of the turbulent flame brush as it is shown in
of the run-up distance in one way or another. While Figs. 30, 31 and 44. The dimensionless flame acceleration
in smooth tubes it is not always possible to distinguish distance was expressed as a function of BR:
the run-up distances to supersonic flames and to X S 10S L ðs  1Þ 1  BR
detonation, in obstructed channels, these are often a , (4.4)
R ap 1 þ b  BR
distinguishable (see, for example, Fig. 38). Historically,
the run-up distances to detonation was of prime interest in where R is tube radius, and a and b are unknown
most of the DDT studies mentioned above, including parameters of the model. The scaling of the run-up dis-
[86,88,148,160,170,174]. tance with mixture properties as in Eq. (4.4) was evaluated
Veser et al. [181] were the first who specifically addressed using a wide range of experimental data [88,182,187] and
the run-up distances to supersonic flames. The problem of results of 3D numerical simulations in the range of BR
the minimum run-up distance for the flame acceleration to from 0.3 to 0.75. It was shown that the grouping on the
supersonic combustion regimes in tubes with obstacles was left-hand side of Eq. (4.4) collapses the data to within
studied both experimentally and numerically in [181]. 725%. The effect of BR given by the right-hand side of
Experiments were made in an explosion tube equipped Eq. (4.4) was found to correlate well with the data for a ¼ 2
with orifice plate obstacles. The tube was 12-m long with and b ¼ 1.5, for the BR range of 0.3–0.75.
internal tube diameter of 0.35 m. Blockage ratios of the Recently, Card et al. [189] investigated flame acceleration
orifice plates were 0.3, 0.45, 0.6 and 0.75. Various mixtures and DDT in fuel–air mixtures at initial temperatures up to
of hydrogen were used in the tests. In addition, data from 573 K and pressures up to 2 atmospheres. The fuels
Ciccarelli et al. [187] and Kuznetsov et al. [88,182] were investigated include hydrogen, ethylene, acetylene and
used in the analysis. JP-10 aviation fuel. The experiments were performed in a
A simple model was proposed by Veser et al. [181], which 3.1-m long, 10-cm inner-diameter heated detonation
describes the evolution of the flame shape in a channel tube equipped with equally spaced orifice plates. It was
containing obstacles with relatively high BR. The model found that hydrocarbon fuels successfully correlated with
assumes that the flame has the shape of a deformed cone, Eq. (4.4). Hydrogen mixtures at high temperatures of 473 K
which stretches until the moment when the speed of the and especially 573 K did not show satisfactory correlation,
flame tip reaches the sound speed with respect to the although the high temperature hydrogen data of Ciccarelli
combustion products. The position of the flame tip at this et al. [187] were correlated successfully in [181].
moment gives an estimate for the run-up distance. Then the Ciccarelli et al. [190] studied the influence of obstacle size
flame cone cannot stretch any further and moves down the and spacing on the initial phase of flame acceleration in an
tube at a quasi-steady velocity. An illustration of the flame obstacle-laden tube. It was found that for various
shape in a tube with orifice plates is shown in Fig. 44 that propane–air mixtures the correlation given in Eq. (4.4)
can be compared with photographs taken of the flame collapsed the run-up distance data to within 750%. It was
shown in Figs. 30 and 31. also found that for a BR of 0.6 and 0.75 the run-up
The dimensionless flame acceleration distance was distance decreases slightly with obstacle spacing, S, while
determined in the model, taking into account mixture for BR ¼ 0.43 the opposite trend is observed. For the
properties, such as the laminar burning velocity, SL, the larger blockage ratio orifice plates it is proposed that flame
ratio of densities between products and reactants, s, and folding is the key mechanism for the early stage of the
the sound speed in the combustion products, ap. It was flame acceleration process. It was also shown that the
assumed that for relatively high BR, the turbulent burning optimum flame acceleration corresponds to the condition
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 537

where the predicted length of the unburned gas recircula- 90 H2 BR<0.1


C2H4 BR<0.1
tion zone downstream of the orifice plate is roughly equal 80
D=1m C3H8 BR<0.1
to the orifice plate spacing. It is argued that if the 70 CH4 BR<0.1
recirculation zone is short compared to the plate spacing, 60 H2 BR>0.3

XS/D model
a portion of the flame would come into contact with the 50 C2H4 BR>0.3
C3H8 BR>0.3
tube wall and flame area would be compromised. If the 40
CH4 BR>0.3
plate spacing is too small the recirculation zone would 30
extend completely between the plates compromising the 20
core flow expansion and contraction which is responsible 10
for flame folding. 0
Cases when the obstacle spacing, S, is not equal to the 00.1 0.1 1
tube diameter, D, were not analyzed by Veser et al. [181], BR
however, the same approach to the estimation of the flame
Fig. 45. Run-up distances over tube diameter as a function of BR for
surface with D6¼S would require the term b  D/S instead of D ¼ 1 m.
b in the right-hand side of Eq. (4.4). This would result in a
weak increase of XS/D with S/D. In the range of S/D from 4.4. Onset of detonations
0.5 to 1.5, the change of XS/D would be within 720%,
which is a small variation considering the uncertainty range 4.4.1. Types of detonation onset phenomena
of the model. In the sense that the effect of S/D is weak, As soon as the first phase of the DDT phenomena—the
this extension is in agreement with the data [190]. creation of conditions for the onset of detonation—is
Sorin et al. [191] studied run-up distances to DDT, XD, accomplished through flame acceleration in a duct, the
in mixtures of hydrocarbons and hydrogen with air, which actual formation of the detonation wave itself or the onset
were sensitized by increased oxygen content in oxidizer, of detonation becomes possible. The key feature associated
compared to that in air. Tests were made in a 26 mm with the onset of detonation is the formation of localized
diameter tube equipped with a Shchelkin spiral. It was explosion somewhere in the mixture [157], which is pre-
clearly shown that the measured XD was significantly conditioned by compression and/or by turbulent mixing of
different from the run-up distances to supersonic flames, products and reactants. Although the processes of the
XS. An attempt was made to correlate XD and XS using onset of detonation can be observed in a wide variety of
Eq. (4.4), which showed the ratio XDDT/XS to vary from situations, it can be classified into two categories:
3.4 to 6.8. The laminar burning velocities necessary for the
evaluation of XS was calculated using the PREMIX code. 1. Detonation initiation resulting from shock reflection or
There were no details of the status of validation of this shock focusing;
model. Unknown SL for the mixtures used in [191] make it 2. Onset of detonation caused by instabilities and mixing
difficult to make reliable conclusions concerning the processes:
accuracy of correlation (4.4). a. instabilities near the flame front;
In summary, Eq. (4.4) provides a simple scaling relation b. flame interactions with a pressure wave, another
for the run-up distances to supersonic flame, which takes flame or a wall;
into account the duct geometry, i.e., tube diameter and BR, c. explosion of a previously quenched pocket of
and mixture properties i.e., laminar burning velocity, combustible gas;
expansion ratio and isobaric sound speed in the combus- d. pressure and temperature fluctuations in the flow and
tion products. The accuracy of the predictions of Eq. (4.4) boundary layer.
is estimated to be within a factor of 2. In cases where the
value of SL is unknown, i.e., at elevated initial tempera- The first category essentially involves a direct initiation
tures or sensitized mixtures, one should be cautious with process where the shock strength is sufficient to auto-ignite
estimates based on Eq. (4.4). the gas and promote detonation. For cases of flame
A comparison of the run-up distances predicted using acceleration in ducts where the shock is produced by an
Eqs. (4.1) and (4.2) (BRp0.1) with those predicted by accelerating flame, this process becomes much more
Eq. (4.4) (BRX0.3) for stoichiometric mixtures of methane, probable when the shock interacts with a corner or a
propane, ethylene, and hydrogen with air is presented in concave wall that can produce shock focusing. Shock
Fig. 45. It is seen that the run-up distances in smooth tubes are initiation is an important mechanism in maintaining the
significantly higher than the run-up distances in obstructed propagation of quasi-detonations in a channel or a tube
tubes. This is qualitatively in accord with the observations that filled with obstacles [160,176,192]. It has also been found to
FA is strongly promoted by the obstructions. In the range be very efficient in promoting detonation onset for flames
of BR between 0.1 and 0.3, neither of the two models propagating towards an orifice, a corner, or a concave wall
is applicable. For practical applications, one may ‘‘bridge’’ [193–199]. Fig. 46 shows an example of the onset of
the range of BR from 0.1 to 0.3 as shown by dashed lines detonation triggered by Mach reflection of a shock wave
in Fig. 45. diffracting around an obstacle.
ARTICLE IN PRESS
538 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

increasing evidence that these phenomena may be con-


trolled by a single underlying mechanism. It has been
suggested by Zeldovich et al. [200] in their theoretical study
on spontaneous flames, and by Lee et al. [201] and
Yoshikawa [202] in the experimental study on photoche-
mical initiation of detonation, that induction time gradi-
ents associated with temperature and concentration
nonuniformities can led to detonation initiation. It was
further suggested by Lee and Moen [2], Shepherd and Lee
[3], and by Zeldovich and colleagues [203,204] that the
induction time gradients may be ultimately responsible for
a wide range of detonation initiation observations. The
proposed mechanism is based on the formation of an
induction time gradient that can produce a spatial time
sequence of energy release. As illustrated schematically in
Fig. 48, if a gradient of induction time, ti, with distance, X,
is created the sequential ignition of portions of the mixture
results in formation of spontaneous flame, which propa-
gates through the mixture at a speed of Dsp ¼ (dti/dX)1.
This spontaneous flame can then produce a compression
wave as shown schematically in Fig. 49. In the case of a
coherence between the energy release in spontaneous flame
propagating at speed Dsp and gasdynamic process, i.e.,
characterized by the speed of sound, the compression wave
is gradually amplified into a strong shock wave that can
auto-ignite the mixture and produce DDT. This process,
commonly referred to as the SWACER mechanism, was
used by Lee et al. [201] and Yoshikawa [202] to explain the
photo-initiation of gaseous detonations, where detonation
develops without an action of strong shock from an
external source.
Because temperature and concentration gradients are
formed by a wide variety of processes, the SWACER
mechanism is believed to be an underlying mechanism for a
wide variety of detonation phenomena including:

Fig. 46. Shadow photographs showing onset of detonation resulting from (1) direct detonation initiation that is due to the tempera-
Mach reflection of lead shock of fast deflagration. Reflection at the top ture gradient behind the leading shock,
wall occurs when the shock enters from obstructed channel into a wider
volume.
(2) detonation initiation that is due to shock focusing,
(3) DDT in tubes that is caused by the temperature
gradient in the boundary layer or between a fast flame
The second category of the detonation onset processes is and the leading shock,
considerably more complex because it involves a variety of (4) DDT that is due to pre-compression at the end of a
instabilities and mixing processes. It can be easily isolated channel by a slow flame,
for DDT in smooth tubes where the onset of detonation (5) DDT in rough or obstacle-filled tubes,
can occur ahead of the turbulent flame brush, near the lead (6) jet initiation caused by temperature and concentration
shock, inside the flame brush [158], in the boundary layer gradients in the flame/vortex structure,
[155,171], or resulting from the interaction between the (7) DDT in multi-phase systems resulting from the
flame front with a reflected shock wave [60]. Fig. 25a shows temperature relaxation caused by the particles.
an example of the onset of detonation inside the flame
brush. Fig. 47 shows example of the onset of detonation at Although some cases in the list above (direct initiation,
the boundary layer. shock focusing, DDT in obstacle filled tubes) involve
strong shocks, it was observed (see, e.g., [2–4]), that at
4.4.2. Underlying mechanism Shock Wave Amplification by nearly critical conditions a secondary explosion occurs
Coherent Energy Release (SWACER) behind the initial decaying shocks. This secondary explo-
Although the onset of detonation occurs through a sion initiates from the ignition of pre-conditioned gas with
variety of seemingly unrelated phenomena, there is an induction time gradient. The explosion develops into
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 539

Fig. 47. Shadow photographs showing onset of detonation triggered by complicated interactions of pressure waves with the lead part of the flame near the
boundary layer. The origin of the detonation wave in the second frame seems to be located at a distance from the bottom wall (assumed to be in the center
of the spherical wave).

Numerous calculations and analysis have been presented


Dsp = (dτi /dX)-1
in the literature to demonstrate the SWACER or Zeldovich
gradient process. These include the early studies by
Zeldovich et al. [200,203,204], Lee et al. [201], Yoshikawa
[202] and those by Yoshikawa, Thibault and Hassam that
Induction were discussed in the review papers [2,3]. More recent
calculations have been performed by Clark [205], Frolov
τ

time
et al. [206], Dorofeev et al. [207–209], Nikiforakis and
Clarke [210], Gelfand et al. [211], Short and Dold [212],
Sequential
Dold et al. [213], Smijanovski and Klein [214], Khokhlov
ignition
et al. [215,216], Bartenev and Gelfand [217], Blumental
X
et al. [218], He and Clavin [219], Gu et al. [220] and
others. These authors have established a strong theoretical
Fig. 48. Schematic illustration of spontaneous flame concept. foundation for the amplification process in the different
modes of propagation of autoignitive fronts, and have
demonstrated the role of the SWACER or Zeldovich
gradient mechanism for various DDT and direct initiation
Reaction front
problems.
In spite of the number of SWACER calculations that can
be found in the literature, few of these can be directly and
Spontaneous
flame
Shock convincingly linked to a particular experimental result.
wave This is particularly true for problems involving turbulent
t

mixing for which calculations usually assume a sponta-


Reaction length neous formation of a temperature or concentration
Compression wave gradient or both. The actual formation of such gradients
involves a variety of turbulent mixing and combustion
Sensitized mixture Unperturbed mixture mechanisms. These mechanisms introduce additional
X
instabilities that must be addressed by the calculations in
order to be truly predictive. Such difficulties have recently
Fig. 49. X–t diagram illustrating propagation of spontaneous flame, been addressed by Khokhlov et al. [216], who investigated
formation of compression and shock wave, and reaction front, which is
the very difficult problem of DDT caused by shock–flame
initially due to spontaneous flame and transforms then to shock induced
reaction front. Characteristic reaction length increases as the wave interaction. This type of DDT, which has been observed by
propagates from sensitized region with induction time gradient to Thomas et al. [60], involves a Meshov–Richmeyer in-
unperturbed, less sensitive, mixture. stability, where the reflected shock interacts with the flame
front. The severe flame distortion produced by this
coupled shock–flame structure through the SWACER instability then produces K–H instabilities, which increase
mechanism, overtakes the lead shock and results in the the flame area and the rate of combustion. The very high-
onset of detonation. resolution calculations of Khokhlov et al. [216] are able to
ARTICLE IN PRESS
540 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

resolve these instabilities and capture the DDT process that boundary conditions. This work is summarized and
occurs inside a temperature gradient. These Navier–Stokes reviewed by Oran and Gamezo in [5].
calculations remain limited in that they do not directly Studies of the SWACER mechanism described above
account for the fine-scale turbulent mixing. Nevertheless, show that several factors could be important for sponta-
they represent one of the most successful efforts so far to neous formation of a detonation. First, a local distribution
isolate the SWACER mechanism for a particular experi- of mixture properties (autoignition delay time) in the
mental DDT observation. Fig. 50 shows an example of hot sensitized region should provide coupling of chemical and
spot ignition near the flame brush simulated numerically in gasdynamic processes that results in the formation of an
[216]. Fig. 51 shows a map of induction time and explosion wave (coupled shock/reaction zone structure)
spontaneous flame speed in this hot spot. This illustration [200,210,212]. Second, this wave should survive propagat-
suggests that the SWACER mechanism was responsible for ing from the sensitized to the unperturbed, less sensitive,
the onset of detonations in these simulations. Following mixture [207–209,215]. Finally, the explosion wave should
the pioneering study of Khohlov et al. [216], highly adjust itself to the chemical length scale, i.e., reaction zone
resolved simulations were made for various initial and length, of the ambient mixture as shown in Fig. 49. The
latter gives a measure of the minimum size of the sensitized
region that is necessary for detonation formation. Such a
measure of the minimum size could be useful for establish-
ing a necessary criterion for the onset of detonation.
Numerical and analytical studies have been made to
determine the minimum size of the sensitized region that is
necessary for detonation formation. Critical conditions for
detonation formation in locally sensitized mixtures were
studied. The sensitized region was modeled by temperature
distributions, addition of a fast reactive component
[207–209], and mixed products and reactants [215]. The
problem of explosion wave propagation through reactivity
gradients was studied also analytically as a separate
problem by Kryuchkov et al. [221].
The main results from the studies looking into the
minimum size of the sensitized region may be summarized
as follows: a characteristic size of the sensitized region
should exceed a minimum value for the onset of a
detonation; this minimum size depends on the properties
of the mixture surrounding the sensitized region; a
characteristic length scale for this process is of the order
of 10l (l is detonation cell width of the surrounding
mixture). Details of the detonation formation process may
influence the minimum size. A decrease of the volumetric
energy content in the sensitized region (e.g., for tempera-
ture nonuniformities, where the density decreases with
Fig. 50. Hot spot ignition near the flame brush simulated numerically in temperature) results in the minimum size increase. In the
[216]. opposite case of the detonation onset in a locally

Fig. 51. Map of induction time (left) and spontaneous flame speed (right) in a hot spot [216].
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 541

pre-compressed region (e.g., for autoignition from shock the absence of confining structure effects, may be expressed
reflections), detonation onset is facilitated. In cases of as Djet424l, where Djet is the orifice size of the initiating
detonation formation as an expanding wave (e.g., spherical jet. This is about 3 times as much as what should be
symmetry), the minimum size was found to be increased expected with presence of strong confining structures
due to curvature effects. A combination of the above (about 10l), where development of initially planar detona-
factors is typical for DDT events. It is hardly possible, tions is possible.
thus, to obtain a universal criterion for minimum size of This data supports the results of the analyses of the main
the sensitized region for the onset of detonations. An features of spontaneous onset of detonations in a sensitized
estimate suggested in [208,209] assumed that a detonation mixture. They show once more that some minimum size of
is developed initially as a planar wave, volumetric energy the sensitized, or gradient, region is required for detonation
content is uniform, and typical fuel–air mixtures are onset; and that 10l is the order of magnitude for the
considered. With these assumptions, the minimum macro- minimum size; and that this size can vary from several
scopic size of sensitized mixture for detonation onset was l to several tenths of l depending on particular conditions.
estimated to be about 7l. The results of these numerical In such a situation, experimental data and their correla-
studies were obtained using 1D calculations of detonation tions, which are described in the following section, becomes
formation in nonuniform mixtures. Limitations of 1D important in order to derive criteria for the onset of
models limit the reliability of their predictions, especially detonations.
their quantitative predictions. In view of this, it is Available criteria for onset of detonations are based on
important to mention recent experimental results, comparing characteristic geometrical sizes with the chemi-
which generally confirm the main conclusions of these cal length scale of the mixture l. The definition of
calculations. characteristic geometrical sizes of an enclosure is necessary
A series of experiments was made by Kuznetsov et al. to formulate necessary criteria for the onset of detonations.
[222] to study the critical condition for propagation of This size should be large enough compared to l to allow
explosion waves through reactivity gradients. Propagation the onset of detonations. In the following sections, these
of a detonation wave from a donor mixture through a criteria will be considered separately for smooth and
gradient region to a less reactive acceptor mixture was obstructed channels.
studied. The length of the donor mixture, the width of the
gradient region Dx, and the reactivity of acceptor mixture 4.4.3. Criteria for onset of detonations in smooth tubes
were varied in the tests. It was shown, that a critical Once the minimum flame speed, which is of the order of
sensitivity gradient (Dl/Dx)* (Dl is the difference in cell the sound speed in the combustion products, is established
sizes between acceptor and donor mixtures) may be in a smooth tube, the onset of detonation becomes
defined, which determines the possibility of detonation possible, if the tube diameter exceeds a certain minimum
decay in the gradient region. Detonations decayed in the value. It was Kogarko and Zeldovich [230] who proposed
gradient region, in the cases of (Dl/Dx) 4 (Dl/Dx)*. It was that the onset of the single spin would represent an
found that the critical value of (Dl/Dx)* depends approximate criteria for the onset of detonation. Studies of
significantly on the difference in energy densities of donor Peraldi et al. [160] and Guirao et al. [231], however, showed
and acceptor mixtures. The more energetic was the donor that the minimum tube diameter, D, for the onset of
mixture compared the acceptor one, the sharper (greater detonation should be greater than the detonation cell width
(Dl/Dx)*) was the critical gradient. Extrapolation of the of the mixture: D4l. More recently, much attention was
experimental results to the uniform energy density resulted given to the problem of whether the hydraulic diameter pD
in critical values of (Dx/Dl)*E10. This data, thus, appears or the actual diameter D, should be used to relate
to be in qualitative accord with the results of numerical confinement dimensions to the detonation cell structure.
calculations described earlier in this section. The critical The limiting tube diameter for detonation propagation is
values of the gradient determined experimentally for defined as Dlim ¼ l/p, with the single-head spin detonation
hydrogen–air mixtures also appeared to be in a reasonable mode as the limiting propagation regime. Peraldi et al.
quantitative agreement with the calculations. [160] and Guirao et al. [231] argued that the minimum size
Another aspect of the numerical and analytical predic- for the onset of detonation as a result of DDT is larger and
tions for the minimum size of a gradient region for is given by D4l. The results of experimental study of
detonation initiation concerns the effect of symmetry Lindstedt and Michels [170], in agreement with Kogarko
(initiation of spherical detonations). Results of turbulent and Zeldovich, showed that D4l/p should be used as a
jet initiation experiments [223–229] have shown a signifi- criterion for the onset of detonation in smooth tubes.
cant difference in critical initiation conditions between The difference between the results [160,231] and [170]
cases of detonation onset directly in the jet of combustion may be explained by the fact that relatively insensitive
products and initiation influenced by confining structures. mixtures in a long tube (200 tube diameters) were used by
This difference was especially addressed in tests [227,228]. Lindstedt and Michels [170]. In this situation, DDT was
These studies have shown that the minimum requirement observed at the end of the tube in part of the mixture that
for initiation of spherical detonations by turbulent jet, in was more compressed and heated during the process of
ARTICLE IN PRESS
542 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

flame acceleration than in cases of DDT in beginning or in As a conservative approach for practical applications,
the middle of the tube. Thus the detonation cell size of the the requirement of the minimum flame speed (or run-up
mixture, where DDT was observed, might be significantly distance to supersonic flames as described in Section 4.2.2)
smaller than the cell size of original mixture. If the onset of and the minimum scale requirement D4l/p may be used
detonation is observed at the initial part of the tube as it is as a set of two necessary conditions for DDT in smooth
more typical in sensitive mixtures, the effect of pre- tubes.
compression is less significant, and the cell size of the
mixture at the time of the onset of detonation is close to the
original cell size. Although this explanation may remove 4.4.4. Criteria for onset of detonations in channels with
the contradiction in the results described here, for practical obstacles
purposes, the more conservative criterion D4l/p should Minimum tube diameter criterion (d4l). The results of
be used. the detailed study of DDT in tubes with orifice plates by
Critical conditions for the onset of detonations in a very Peraldi et al. [160] and Guirao et al. [231], suggested that
sensitive mixture such as stoichiometric hydrogen–oxygen the size of unobstructed passage d, has to be 4l to observe
and relatively large tubes were considered by Kuznetsov the onset of detonation. Follow-on studies of DDT in tubes
et al. [165,171]. In these cases an important problem to and channels with obstacles [86,176,177,88,232] have
be solved is the determination of where the onset of shown that the critical value of d/l appeared to depend
detonation occurs, rather than can it be expected or on the obstacle configuration. The critical ratio d/l
not. Part of the problem is related to the minimum increases with decrease of obstacle spacing (from 4.28d to
run-up distance to supersonic flames as described in 1.07d [176,177]) and with increase of BR (from 0.1 to 0.75
Section 4.2.2. As soon as fast deflagration is developed, [88]). DDT is easier achieved for smaller BR and for larger
the onset of detonation is possible. However, it is not distances between obstacles. Variations of the critical d/l
always the case that the onset of detonation is observed can be quite large, ranging from 0.8 to 5.1 for BR from 0.3
immediately after the flame reached the supersonic speed. to 0.6. The critical ratio d/l was found to decrease further
There are cases when additional flame propagation to 0.7 for Shchelkin spirals [170]. This data shows that the
distance is required for the onset of detonation to occur, transverse size of the channel is not the only defining
even in a sensitive mixture such as stoichiometric hydro- parameter for onset of detonation.
gen–oxygen [165,171]. Minimum scale requirement (L/l-criterion). A different
Kuznetsov et al. [165], suggested that the flow ahead of approach was also suggested, which requires the minimum
the flame results in formation of the turbulent boundary distance L for detonation formation [207–209]. The size L
layer. This boundary layer not only plays an important role should give a measure of the possible macroscopic size of
in the flame acceleration, as described in Section 4.2.2, but the sensitized mixture in which detonations might originate
also controls the scale of turbulent motions in the flow. The and develop. Originally, such a criterion was formulated as
maximum scale of the turbulent pulsations given by the L47l, where L was defined as a characteristic size of a
boundary layer thickness at flame positions along the tube room filled with combustible mixture, or the size of a
was estimated using a simple model with an accuracy of mixture cloud.
about a factor of 2. This scale controls the characteristic Despite a general agreement of the L47l criterion with
nonuniformity size, or the size of pre-conditioned regions experimental data, definitions for the characteristic size L
of mixture that can trigger the onset of detonation. The used in [233,234] were not always unambiguous, especially
latter was shown in Section 4.4.2 to be 4l by an order of for practical applications. An appropriate definition of L
magnitude for detonation onset to be possible. for chains of connected rooms or tubes with obstacles was
It was found [165], that in very sensitive mixtures (l5D) derived in [186]. For a channel with repeating obstacles the
the onset of detonation was observed as soon as the scale of characteristic size L is given by
the turbulent pulsations increased during flame accelera- ðH 1 þ S 1 Þ=2
tion up to about 15 times the local detonation cell size, or L¼ , (4.5)
1  d=H 1
about 10 times the cell size of the initial mixture. This
observation was shown to be valid over a wide where H1 and S1 are two largest transverse sizes of the
range of the detonation cell widths from 0.1 to 10 mm volume confined between two consecutive obstacles, e.g.,
within the limits of accuracy of the cell size data. Although channel height and distance between obstacles, and d is the
these results give a reasonable criterion for the onset of size of unobstructed passage. For tubes with repeated
detonation in very sensitive mixtures and smooth tubes, in obstacles this yields:
many tests presented in [165] the difference between the ðD þ SÞ=2
run-up distance to supersonic flame and to detonation was L¼ , (4.6)
1  d=D
not clearly distinguished. For this reason additional
experimental data is required to evaluate the range of where D is tube diameter, and S is the obstacle spacing.
applicability of the detonation onset criteria proposed Eqs. (4.5) and (4.6) may be applied to channels and tubes
in [165]. with BR40.1.
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 543

Table 2
Experimental data used in L/l-correlation for onset of detonations

Data source Label Blockage ratio, Tube or channel Initial Mixture type Equivalence ratio,
BR size, mm temperature, K f

Chan et al. [235] a1 0.31 280 373 H2/air/H2O


Ciccarelli et al. [187] b1 0.43 273 300 H2/air o1
Ciccarelli et al. [187] b2 0.43 273 500 H2/air o1
Ciccarelli et al. [187] B3 0.43 273 650 H2/air o1
Ciccarelli et al. [187] b4 0.43 273 400 H2/air/H2O o1
Ciccarelli et al. [187] b5 0.43 273 500 H2/air/H2O o1
Ciccarelli et al. [187] b6 0.43 273 650 H2/air/H2O o1
Kuznetsov et al. [88] c1 0.1 80 293 H2/air o1; 41
Kuznetsov et al. [88] C2 0.3 80 293 H2/air o1; 41
Kuznetsov et al. [88] c3 0.6 80 293 H2/air o1; 41
Kuznetsov et al. [88] d1 0.09 174 293 H2/air o1; 41
Kuznetsov et al. [88] d2 0.3 174 293 H2/air o1; 41
Kuznetsov et al. [88] d3 0.6 174 293 H2/air o1; 41
Kuznetsov et al. [88] d4 0.9 174 293 H2/air o1
Sherman et al. [184] f1 0.33 1830 293 H2/air o1
Sherman et al. [184] f3 0.33 150 293 H2/air o1
Kuznetsov et al. [88] g1 0.6 350 293 H2/air o1; 41
Kuznetsov et al. [88] g2 0.6 350 293 H2/O2/N2 1
Kuznetsov et al. [88] g3 0.3 350 293 H2/air 1
Kuznetsov et al. [88] g6 0.6 350 293 H2/air/CO2 .5
Kuznetsov et al. [88] g7 0.6 350 293 H2/air/CO2 1
Kuznetsov et al. [88] g8 0.6 350 293 H2/air/CO2 2
Kuznetsov et al. [88] g9 0.6 350 293 H2/air/CO2 4
Teodorczyk et al. [176] m1 0.44 16  57  50 293 H2/air o1
Teodorczyk et al. [176] m2 0.44 16  57  100 293 H2/air o1
Lee et al. [86] m3 0.43 50 293 H2, CH-fuels/air o1
Lee et al. [86] m4 0.43 150 293 H2, CH-fuels/air o1
Lee et al. [86] m5 0.43 300 293 H2, CH-fuels/air o1
Teodorczyk et al. [177] m6 0.44 65  52  32 293 H2, CH-fuels/air o1
Teodorczyk et al. [177] m7 0.44 65  52  64 293 H2, CH-fuels/air o1
Teodorczyk et al. [177] m8 0.44 65  52  128 293 H2, CH-fuels/air o1
Dorofeev et al. [185] r1 0.6 2250 293 H2/air o1
Dorofeev et al. [185] r2 0.3 2250 293 H2/air o1
Dorofeev et al. [185] r3 Room 10.5  6  2.3 m3 293 H2/air o1
Dorofeev et al. [186] r4 Room 10.5  6  2.3 m3 375 H2/air/H2O p1
Dorofeev et al. [186] r5 0.3 2250 375 H2/air/H2O p1
Dorofeev et al. [186] r6 0.3 2250 293 H2/air/CO2 o1
Dorofeev et al. [186] r7 Room 10.5  6  2.3 m3 293 H2/air/CO2 o1
Dorofeev et al. [236] ri Room 15  6  2.3 m3 293 H2-injection p1
Sherman et al. [183] S1 0.6 406 383 H2/air 41
Sherman et al. [183] S2 0.3 406 383 H2/air/H2O 41
Kuznetsov et al. [88] t1 0.6 520 293 H2/air o1;41
Kuznetsov et al. [88] T3 0.3 520 293 H2/air o1;41
Kuznetsov et al. [88] t4 0.1 520 293 H2/air o1;41
Dorofeev et al. [186] v1 0.3 46 293 H2/air o1
Dorofeev et al. [186] v2 Room 210  120  50 293 H2/air o1

A considerable database on limiting conditions for DDT from the general borderline L/lE7 is within 730%, which
was used in [186] to correlate the characteristic geometrical is much smaller than inaccuracy of the cell size data
size L with the cell size l. A partial summary of the data is estimated by a factor of 2. The data presented in Fig. 52
given in Table 2. The results from the experiments listed in that corresponds to a variety of different geometrical
Table 2 are shown graphically in Fig. 52 as GO and NO configurations shows that the L/l47 correlation for
GO data with L/l and L as the independent parameters. predicting DDT is applicable. This correlation can be
Those tests that resulted in DDT are shown as black solid used as necessary criterion for onset of detonations within
symbols and a gray solid symbol is used for those tests for the range of mixtures, initial conditions and scales used in
which DDT did not occur. this correlation. It should be mentioned, however, that
A good correlation is observed for L/lE7 as the possible uncertainties in the determination of L and
necessary condition for DDT within the accuracy of the especially l should be taken into account in practical
cell size data over a wide range of scales. The deviation applications.
ARTICLE IN PRESS
544 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

100 comparison of the critical conditions in mixtures with


No DDT, BR<0.5 regular and irregular structure was made. The critical
No DDT, BR>0.5
& rooms
conditions for the onset of detonations (expressed both in
5
DDT, BR<0.5 terms of hydrogen concentrations and detonation cell
DDT, BR>0.5
& rooms
widths l) were shown to be strongly influenced by the
3
L = 7λ regularity of the detonation cellular structure of the test
2
λ accuracy limits mixtures. The critical values of the cell sizes in Ar- and He-
diluted mixtures have been found to be significantly smaller
than those in air mixtures and oxygen mixtures with N2
L/λ

10 dilution for the same tube and obstacle geometry. This


means that systems with highly regular detonation
5
cellular structures are much less capable of undergoing
DDT compared to irregular ones with the same values of
3
the detonation cell sizes. The difference in the critical
conditions between regular and irregular systems was
2 shown to have the same trend in the case of DDT and in
the problems of self-sustained detonation propagation
[238].
1
100 2 3 5 1000 2 3 5 10000
Geometrical size L, mm
4.5. Evaluation of potential for FA and DDT

Fig. 52. Summary of detonation onset conditions. Resulting explosion The combustion processes that follow weak ignition in a
mode as a function of L/l ratio and geometrical size L. Only cases of fast combustible mixture are very complex as illustrated by the
deflagrations and DDT are shown [186].
description of FA and DDT presented here. A detailed
description of all these processes is extremely difficult at
% CH4 % H2 present because of the complicated interactions of com-
18 90
pressible flow, turbulence, and chemical reactions, which
16 80 should be described at high spatial and temporal resolu-
slow tion. As a result, our knowledge of FA and DDT is
slow
14 70 fast essentially based on experimental data. Analysis of the
fast existing data permits the formulation of a framework of
12 60
ideas and criteria that can be used to evaluate initial and
DDT 50 boundary conditions under which FA and DDT may be
10
possible. It is important to note that only necessary
8 fast 40 DDT conditions for FA and DDT can be suggested at the
present time. Thus, in many situations we can suggest that
6 slow 30 effective flame acceleration and DDT is extremely unlikely,
but it is much more difficult to predict whether these will
4 20
actually occur. The criteria for the evaluation of the
fast
2 10 potential for FA and DDT listed below are based on the
slow
experiments conducted to date and are still the subject of
0 0 ongoing research.
174 mm 520 mm 80 174 350 520
1. FA follows weak ignition of combustible gas in ducts
Tube diameter, mm with or without obstacles. Obstructions provide the
Fig. 53. Composition limits for slow flame, fast flames, and DDT in most effective environment for FA. However, there are
hydrogen–air and methane–air mixtures in different tubes. Vertical scale in cases when FA is weak and supersonic combustion
the plots corresponds to vol% of fuel in air. DDT limits are shown for regimes cannot be developed.
tubes with BR ¼ 0.3. 2. In order for FA to be strong, a sufficiently large
expansion ratio s ¼ ru/rb must exist between the
Applicability of the L47l criterion was evaluated in unburned and burned gas. The critical expansion ratio
more recent studies of DDT in tubes with obstacles depends on the mixture composition and initial tem-
[232,237]. Kuznetsov et al. [237] showed that this correla- perature and pressure as described in Section 4.3.4.
tion is valid for insensitive methane–air mixtures in 3. If the expansion ratio is sufficiently large, the transverse
sufficiently large tubes. size of the duct should be at least two orders of
An interesting aspect of the critical conditions for the magnitude larger than the laminar flame thickness for
onset of detonation was addressed in [232], where a strong FA to be possible. This limitation may be
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 545

important, especially for rich mixtures of hydrocarbons follows weak ignition in a combustible mixture, is
due to relatively large flame thickness. described as intrinsically unstable combustion phenomen-
4. If strong FA is possible (expansion ratio and scale on. The instabilities of the flame and its interactions
exceed critical values), a sufficiently large run-up with confinement and obstructions result in growth of the
distance is necessary for actual development of super- flame surface and the burning rate. Eventually the flame
sonic combustion regimes as described in Sections 4.2.2 becomes turbulent and proceeds through a sequence of
and 4.3.5. The minimum run-up distances depends on turbulent combustion regimes with generally increasing
mixture properties (such as the laminar burning velocity, speed relative to a fixed observer. This process, which is
laminar flame thickness, and isobaric sound speed in referred to as flame acceleration, under certain initial and
combustion products) initial conditions, and duct size. boundary conditions may trigger a local explosion some-
5. Supersonic combustion regimes should be developed before where in the flow, which in turn may result in the onset of
the conditions for the onset of detonation can be reached. detonation.
6. If the supersonic combustion regime is developed, then There are many spatial and temporal physical scales that
detonation initiation may only occur if the physical size are involved in the processes of flame and detonation
of a duct or mixture volume is sufficiently large propagation. These scales are given by chemistry, turbu-
compared with a length that characterizes the reactivity lence, and confinement. The interplay of these scales
of the mixture. The usual choice of the reactivity length control major features of the combustion regimes and
scale is the detonation cell size l, which is a function of thresholds for various transition phenomena, such as onset
mixture composition and thermodynamic conditions. of instabilities, flame structure, onset, and structure of
Results of numerous DDT experiments presented in detonations. The wide range of the scales involved makes it
Section 4.4 have demonstrated that the physical dimen- extremely difficult to resolve all the phenomena involved
sion of the compartment or duct has to be greater than from first principles. However, it is the comparison of
some multiple of the cell width in order for DDT to be scales that give us a way to approach a number of practical
possible. This leads to the necessary criteria for DDT problems.
a. dXl, where d is the transverse dimension of the The prediction of flame acceleration and DDT in
unobstructed passage in a channel with obstacles; a duct is very important for estimating potential
b. LX7l, where L is a more general characteristic size explosion hazards. Considering the present state of knowl-
defined for interconnected rooms or channels with edge it is very difficult to predict in detail this very
various configurations of obstacles (see Eq. (4.5)); complicated phenomenon. The basic elements of the
c. DXl/p, where D is internal diameter of a smooth tube. processes of flame acceleration and DDT have been
understood for many years. Now we are able to add a
As an illustration of the framework of criteria described more quantitative framework for evaluation for the
above, Fig. 53 shows the combustion regimes that can be potential of flame acceleration and DDT. At the same
observed in sufficiently long tubes equipped with 0.3 BR time a range of uncertainties still exists in such an
orifice plates filled with methane–air and hydrogen–air evaluation.
mixtures at initial 300 K and 1 bar. It is seen that within the Critical conditions for strong flame acceleration and
flammable range, there is a region of mixture compositions DDT described in Section 4 are formulated as only
where only slow flames are possible. Inside of this region, necessary criteria. At the present time, it is only possible
fast flames can be formed and within the region of fast to specify the necessary conditions for FA and DDT.
flames there is a region where DDT is possible. It is Uncertainties in the determination of the critical condi-
important to note that the boundaries between combustion tions, including critical values of mixture expansion ratio,
regimes are not simply fixed concentration values, but they uncertainties in the detonation cell size data, the laminar
are functions of physical scale as shown in Fig. 53. It is also burning velocity, and the laminar flame thickness must be
important to note that the fast flame and DDT boundaries taken into account in practical applications. There are
depends on the initial thermodynamic state of the mixtures, issues in respect to changes of thermodynamic state of
e.g., temperature and pressure, that are not taken into unburned mixture during flame acceleration, which can
account in Fig. 53. Geometrical details can be also change the critical conditions, both for flame acceleration
important. DDT limits for BR ¼ 0.6, e.g., are significantly and DDT, and should also be taken into account in
narrower than those shown in Fig. 53, e.g., no DDT was practical applications.
observed in methane–air mixtures in 174 and 520-mm tubes
with BR ¼ 0.6 [237].
Acknowledgments
5. Concluding remarks
Authors are grateful to I. Matsukov, V. Alekseev and
We have presented a review of the state of knowledge on M. Kuznetsov, who took many excellent shadow photos
flame acceleration and DDT in smooth ducts and ducts used in this paper. One of us (SD) is thankful to FM
equipped with obstacles. The laminar flame, which Global for the support provided for preparing this review.
ARTICLE IN PRESS
546 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

References [27] Darrieus G. Unpublished works presented at La Technique


Moderne (1938) and at Congres de Mechanique Appliquee Paris,
[1] Oppenheim AK, Soloukhin RI. Experiments in gasdynamics of 1945.
explosions. Ann Rev Fluid Mech 1973;3:31–58. [28] Markstein GH. Nonsteady flame propagation. Oxford, UK:
[2] Lee JHS, Moen IO. The mechanism of transition from deflagration Pergamon Press; 1964.
to detonation in vapour cloud explosion. Prog Energy Combust Sci [29] Zeldovich YaB. An effect which stabilizes the curved front of a
1980;6:359–89. laminar flame. J Appl Mech Tech Phys 1966;7:68.
[3] Shepherd JE, Lee JHS. On the transition from deflagration to [30] Gostintsev YuA, Istratov AG, Shulenin YuV. Self-similar propaga-
detonation. In: Hussaini MY, Kumar A, Voit RG, editors. Major tion of a free turbulent flame in mixed gas mixtures. Combust Expl
research topics in combustion. Berlin: Springer; 1991. Shock Waves 1988;24:563.
[4] Gelfand BE, Frolov SM, Nettleton MA. Gaseous detonations—a [31] Barenblat GI, Zeldovich YaB, Istratov AG. Diffusive-thermal
selective review. Prog Energy Combust Sci 1991;17:327–71. stability of a laminar flame. Prikl Mekh Fiz 1962;2:21.
[5] Oran ES, Gamezo VN. Origins of the deflagration-to detonation [32] Sivashinsky GI. Diffusional-thermal theory of cellular flames.
transition in gas phase combustion. Combust Flame 2007; Combust Sci Technol 1977;15:137.
148:4–47. [33] Joulin G, Clavin P. Linear stability analysis of nonadiabatic
[6] Mallard E, Le Chatelier H. Recherches Experimentales et Theori- flames—diffusional-thermal model. Combust Flame 1979;35:139.
ques sur la Combustion des Melanges Gaseoux Explosifs. Ann [34] Matsukov I, Kuznetsov M, Alekseev V, Dorofeev S. Photographic
Mines 1883;8(4):274–568. study of the characteristic regimes of turbulent flame propagation,
[7] Bradley D, Mitcheson A. The venting of gaseous explosions in local and global quenching in obstructed areas. Report No. 315/
spherical vessels. Combust Flame 1978;32:221–55. 20088504/INR, prepared for FZK-INR, 1998.
[8] Henrie JO, Postma AK. Analysis of the Three Mile Island (TM-2) [35] Matsukov I, Kuznetsov M, Alekseev V, Dorofeev S. Photographic
hydrogen burn. In: Second international meeting of nuclear reactor study of unstable turbulent flames in obstructed channel. In:
thermal hydraulics, Santa Barbara, CA January 1983. Proceedings of the 17th international colloquium on the dynamics
[9] OECD Nuclear Energy Agency. Flame acceleration and deflagra- of explosions and reactive systems, July 25–30, 1999, Heidelberg,
tion to detonation transition in nuclear industry. State of the Art Germany, ISBN 3-932217-01-2, Universitat Heidelberg, 1999.
Report NEA/CSNI/R report 7, 2000. [36] Clavin P, Williams FA. Effects of molecular diffusion and thermal
[10] Roy GD, Frolov SM, Borisov AA, Netzer DW. Pulse detonation expansion on the structure and dynamics of premixed flames in
propulsion: challenges, current status, and future perspective. Prog turbulent flows of large scale and low intensity. J Fluid Mech
Energy Combust Sci 2004;30(6):545–672. 1981;116:252–82.
[11] Glassman II. Combustion. 3rd ed. New York: Academic Press; [37] Peclet P, Clavin P. Influence of hydrodynamics and diffusion upon
1996. the stability of laminar premixed flames. J Fluid Mech 1982;124:
[12] Chapman DL. On the rate of explosions in gases. Phil Mag 47 Ser. 219–37.
5, No. 284,1899. p. 90–104. [38] Gu XJ, Haq MZ, Lawes M, Woolley R. Laminar burning velocity
[13] Jouguet E. On the propagation of chemical reactions in gases. J Pure and Markstein lengths of methane–air mixtures. Combust Flame
Appl Math 1947;1(Series 6):347. 2000;121:41–58.
[14] Taylor GI. The dynamics of the combustion products behind plane [39] Karlovitz B, Denission JR, Knapschaffer DH, Wells FE. Studies on
and spherical detonation fronts in explosives. Proc Roy Soc Lond A turbulent flames. Proc Combust Inst 1953;4:613–20.
1950;200:235–47. [40] Buckmaster JD. The quenching of two-dimensional premixed
[15] Oppenheim AK. Gasdynamic analysis of the development of flames. Acta Astronaut 1979;6:741.
gaseous detonation and its hydraulic analogy. Proc Combust Inst [41] Matalon M. On flame stretch. Combust Sci Technol 1983;31:169.
1953;4:471–80. [42] Chung SH, Law CK. An invariant derivation of flame stretch.
[16] Taylor GI, Tankin RS. In: Emmons HW, editor. Fundamentals Combust Flame 1984;55:123.
of gasdynamics. Princeton: Princeton University Press; 1958. [43] Bradley D, Gaskell PH, Gu XJ. Burning velocities. Markstein
p. 622–86. lengths, and flame quenching for spherical methane–air flames: a
[17] Adams GK, Pack DC. Some observations on the problem of computational study. Combust Flame 1996;104:176–98.
transition between deflagration and detonation. Proc Combust Inst [44] Clavin P, Joulin G. Premixed flames in large scale and high intensity
1959;7:108–819. turbulent flow. J Phys Lett 1983;44:L-1.
[18] Troshin YaK. The generalized Hugoniot adiabatic curve. Proc [45] Dorofeev SB, Kuznetsov MS, Alekseev VI, Efimenko AA, Breitung
Combust Inst 1959;7:789–98. W. Evaluation of limits for effective flame acceleration in hydrogen
[19] Meyer JW, Urtiew PA, Oppemheim AK. On the inadequacy of mixtures. J Loss Prevent Process Ind 2001;14(6):583–9.
gasdynamics processes for triggering the transition to detonation. [46] Guenoche H. The detonation and deflagration of gas mixtures. Rev
Combust Flame 1970;14:13–20. Inst Franc Petrole 1949;4:48.
[20] Zeldovich YaB. Chain reactions in flames—approximate theory of [47] Leyer JC, Manson N. Development of vibratory flame propagation
flame velocity. Kinet Katal 1961;2:305. in short closed tubes and vessels. Proc Combust Inst 1971;13:551–7.
[21] Istratov AG, Librovitch VB. On theory of flame speed in systems [48] Kogarko SM, Ryzhkov DL. A study of the amplification of
with chain reactions. Zh Prikl Mat Teckhn Fiz v 1962;8(1):68–75. compression waves during combustion. Sov Phys Tech Phys
[22] Frank-Kamenetskii DA. Diffusion and heat transfer in chemical 1961;31:211.
kinetics. New York: Plenum Press; 1969. p. 445. [49] Van Wingerden CJM, Zeeuwen JP. On the role of acoustically
[23] Zeldovich YaB, Barenblatt GI, Librovich VB, Makhviladze GM. driven flame instabilities in vented gas explosions and elimination.
Mathematical theory of combustion and explosion. Moscow: Combust Flame 1983;51:109–11.
Nauka; 1980. [50] Tamanini F, Chaffee JL. Turbulent vented gas explosions with and
[24] Zeldovich YaB, Frank-Kamenetskii DA. A theory of thermal without acoustic waves. Proc Combust Inst 1992;42:1845–51.
propagation of flames. Acta Phys—Chem URSS IX 1938;2:348. [51] Teodorczyk A, Lee JHS. Detonation attenuation by foams and wire
[25] Bush B, Fendell FE. Asymptotic analysis of laminar flame meshes lining the walls. Shock Waves 1995;4(4):225–36.
propagation for general lewis numbers. Combust Sci Technol [52] Oran ES, Gardner JH. Chemical-acoustic interactions in combus-
1970;1:421. tion systems. Prog Energy Combust Sci 1985;11:253–76.
[26] Landau LD. Theory of slow combustion. Zh Exp Teor Fiz [53] Searby G, Rochwerger D. A parametric acoustic instability in
1944;14:240. flames. J Fluid Mech 1991;231:529–54.
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 547

[54] Joulin G. On the response of premixed flames to time dependent [81] Ballal DR, Lefebvre AH. The influence of flow parameters on
stretch and curvature. Combust Sci Technol 1994;97:219. minimum ignition energy and quenching distance. Proc Combust
[55] Jackson TL, Macaraeg MG, Hussaini MY. The role of acoustics in Inst 1975;15:1473.
flame/vortex interactions. J Fluid Mech 1993;254:259–603. [82] Checkel MD, Thomas A. Turbulent combustion of premixed flames
[56] Kansa E, Perlee HE. Constant-volume flame propagation: finite- in closed vessels. Combust Flame 1994;96:351–70.
sound-speed theory. Report of investigations 8163, United States [83] Chomiak J. Flame development from an ignition kernel in laminar
Department of the Interior, Bureau of Mines, 1976. and turbulent homogeneous mixtures. Proc Combust Inst Pitts-
[57] Al-Shahrany AS, Bradley D, Lawes M, Liu K, Woolley R. Darrieus- burgh 1979;17:25.
Landau and thermo-acoustic instabilities in closed vessel explosions. [84] Phillips H. The save gap revisited. In: Kuhl AL, et al., editors.
Combust Sci Technol 2006;178:1771–802. Dynamics of explosions, vol. 114. Washington DC: AIAA progress
[58] Markstein GH, Somers LM. Cellular flame structure and vibratory in astronautics and aeronautics; 1988. p. 77–96.
flame movement in N-butane–methane mixtures. Proc Combust Inst [85] Larsen O. A study of critical dimensions of holes for trans-
1953;4:527–35. mission of gas explosions and development and testing of a
[59] Scarincini T, Lee JH, Thomas GO, Brambrey R, Edwards DH. In: Schlieren system for studying jets of hot combustion products,
Kuhl AL, et al., editors. Dynamics of heterogeneous combustion Thesis for the degree of Cand Scient, University of Bergen, Norway,
and reacting systems, vol. 152. Washington, DC: AIAA progress in 1998.
astronautics and aeronautics; 1993. p. 3–24. [86] Lee JH, Knystautas R, Chan CK. Turbulent flame propagation in
[60] Thomas GO, Sands CJ, Brambrey RJ, Jones SA. Experimental obstacle-filled tubes. Proc Combust Inst 1985;20:1663–72.
observations of the onset of turbulent combustion following [87] Knystautas R, Lee JHS, Peraldi O, Chan CK. Transmission of a
shock–flame interaction. In: Proceedings of the 16th international flame from a rough to a smooth wall tube. In: Bowen JR, et al.,
colloquium on the dynamics of explosions and reactive systems, editors. Dynamics of explosions, vol. 106. Washington DC: Progress
Cracow, 1997. p. 2–5. in Astronautics and Aeronautics; 1986. p. 37–52.
[61] Peters N. Laminar flamelet concepts in turbulent combustion. Proc [88] Kuznetsov M, Alekseev V, Bezmelnitsyn A, Breitung W, Dorofeev
Combust Inst 1986;2:1231–50. S, Matsukov I, Veser A, Yankin Yu. Effect of obstacle geometry on
[62] Clavin P. Dynamic behavior of premixed flame fronts in laminar behavior of turbulent flames, Report No. FZKA-6328, Forschungs-
and turbulent flows. Prog Energy Combust Sci 1985;11:1–59. zentrum Karlsruhe/Preprint No. IAE-6137/3, RRC ‘‘Kurchatov
[63] Bradley D. How fast can we burn? Proc Combust Inst 1992;25: Institute,’’ 1999.
247–62. [89] Thibault P, Liu YK, Chan CK, Lee JH, Knystautas R, Guirao C, et
[64] Libby PA, Williams FA, editors. Turbulent reacting flows. London: al. Transmission of an explosion through an orifice. Proc Combust
Academic Press; 1994. Inst 1982;19:599–606.
[65] Bray KNC. The challenge of turbulent combustion. Proc Combust [90] Reynolds WC. The element potential method for chemical
Inst 1996;26:1–26. equilibrium analysis: implementation in the interactive program
[66] Abdel-Gayed RG, Al-Khishali KJ, Bradley D. Turbulent burning STANJAN. 3rd ed. Mechanical Engineering Department, Stanford
velocities and flame straining in explosions. Proc R Soc Lond A University; 1986.
1984;391:393–414. [91] Zeldovich YaB. On the theory of the propagation of detonation in
[67] Abdel-Gayed RG, Bradley D. Criteria for turbulent propagation gaseous systems. J Exp Theor Phys 1940;10(5):543–68.
limits of premixed flames. Combust Flame 1985;62:61–8. [92] von Neumann J. Progress report on the theory of detonation waves.
[68] Bradley D, Lau AKC, Lawes M. Phil Trans Proc R Soc Lond A Office of scientific research and development, Report No. 549, 1942.
1992;338:359–87. 24p.
[69] Bray KNC. Studies of turbulent burning velocities. Proc R Soc [93] Döring W. On the detonation process in a gas. Annal Phys 5e Folge
Lond A 1990;431:315–35. 1943;43:421–36.
[70] Gülder OL. Turbulent premixed combustion modelling using fractal [94] Shepherd JE. Chemical kinetics of hydrogen–air–diluent detona-
geometry. Proc Combust Inst 1990;23:83. tions. In: Bowen JR, editor. Dynamics of explosions, vol. 106.
[71] Shy SS, Lin WJ, Peng KZ. High intensity turbulent premixed Washington DC: Progress in Astronautics and Aeronautics; 1986.
combustion: general correlations of turbulent burning velocities in a p. 263–93.
new cruciform burner. Proc Combust Inst 2000;28:561–8. [95] Konnov AA. Detailed reaction mechanism for small hydrocarbons
[72] Yang SI, Shy SS. Global quenching of premixed CH4/air flames: combustion. Release 0.4. /http://homepages.vub.ac.be/akonnov/S.
effects of turbulent straining, equivalence ratio, and radiative heat [96] Erpenbeck JJ. Stability of steady state equilibrium detonations. Phys
loss. Proc Combust Inst 2002;29:1841–7. Fluids 1962;5:604–14.
[73] Borghi R. In: Bruno C, Casci C, editors. Recent advances in [97] Erpenbeck JJ. Stability of steady state equilibrium detonations. Phys
aerospace science. NY: Plenum Press; 1985. p. 117. Fluids 1964;7(5):684–96.
[74] Williams F. Combustion theory. 2nd ed. Menlo Park, CA: Benjamin [98] Fikett W, Woods WW. Flow calculation for pulsating one-
Cummings; 1985. dimensional detonations. Phys Fluids 1966;9(5):903–16.
[75] Poinsot T, Veynant D, Candel S. Quenching processes and premixed [99] White DR. Turbulent structure of gaseous detonations. Phys Fluids
turbulent combustion diagrams. J Fluid Mech 1991;228:561–606. 1961;4(4):465–80.
[76] Ardey N. Structure and acceleration of turbulent hydrogen–air- [100] Denisov YN, Troshin YK. On the mechanism of detonative
flames within obstructed confinements, PhD thesis, Technical combustion. Proc Combust Inst 1961;8:62–3.
University of Munich; 1998. [101] Mach E, Sommer J. Akademie der Eissenschaften, Wien, Math.-
[77] Jordan M. Ignition and combustion of premixed turbulent jets, PhD Naturw-Klasse Sitzungsberichte Teil II; 1877. p. 101–30.
thesis, Technical University of Munich; 1999. [102] Hornung H. Regular and Mach reflection of shock waves. Ann Rev
[78] Dorofeev S. Thermal quenching and re-ignition of mixed pockets of Fluid Mech 1986;18:33–58.
reactants and products in gas explosions. Proc Combust Inst [103] Strehlow RA, Fernandes FD. Transverse waves in detonations.
2007;31:2371–9. Combust Flame 1965;9:109–19.
[79] Bradley D, Lawes M, Liu Kexin, Woolley R. The quenching of [104] Strehlow RA. The nature of transverse waves in detonations.
premixed turbulent flames of iso-octane, methane and hydrogen at Astronautica Acta 1969;14:539–48.
high pressures. Proc Combust Inst 2007;31:1393–400. [105] Ul’yanitskii VY. Role of flashing and transverse wave collisions in
[80] Libby P, Liñan A, Williams F. Strained laminar flames with non- the evolution of a multifront detonation wave structure in gases.
unity Lewis numbers. Combust Sci Technol 1983;34:257. Fizika Goreniya I Vzryva 1981;17(2):227–32.
ARTICLE IN PRESS
548 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

[106] Manzhalei VI. Fine structure of the leading front of a gas [131] Auffret Y, Desbordes D, Presles HN. Detonation structure and
detonation. Fizika Goreniya I Vzryva 1977;13(3):402–4. detonability of C2H2–O2 mixtures at elevated initial temperature.
[107] Strehlow RA. Multi-dimensional detonation wave structure. Astro- Shock Waves 2001;11:89–96.
nautica Acta 1970;15:345–57. [132] Shchelkin KI, Troshin YK. Non-stationary phenomena in the
[108] Campbell C, Woodhead DW. The ignition of gases by an explosion gaseous detonation front. Combust Flame 1963;7:143–51.
wave I: carbon monoxide and carbon mixtures. J Chem Soc [133] Westbrook CK, Urtiew PA. Use of chemical kinetics to predict
1926;301:3010–21. critical parameters of gaseous detonations. Fizika Goreniya I
[109] Manson N. Propagation des detonations et des deflagration dans les Vzryva 1983;19(6):65–76.
mélanges gazeux. Compt Rend 1946;222:46. [134] Knystautas R, Lee JHS, Shepherd J, Teodorczyk A. Flame
[110] Fay JA. A mechanical theory of spinning detonation. J Chem Phys acceleration and transition to detonation in benzene–air mixtures.
1952;20(6):942–50. Combust Flame 1998;11:424–36.
[111] Schott GL. Observation of the structure of spinning detonations. [135] Gavrikov AI, Efimenko AA, Dorofeev SB. A model for detonation
Phys Fluids 1965;8:850–65. cell size prediction from chemical kinetics. Combust Flame
[112] Lee JHS. Dynamic parameters of gaseous detonations. Ann Rev 2000;120(1):19–33.
Fluid Mech 1984;16:311–36. [136] Zeldovich YB, Kogarko SM, Simonov NN. An experimental
[113] Strehlow R, Crooker A. The structure of marginal detonation investigation of spherical detonation in gases. Sov Phys Tech Phys
waves. Acta Astronautica 1974;1:303–15. 1956;1:1689.
[114] Austin JM, Pintgen F, Shepherd JE. Lead shock oscillation and [137] Mitrofanov VV, Soloukhin RI. The diffraction of multifront
decoupling in propagating detonations. 43rd AIAA Aerospace detonation waves. Sov Phys Dokl 1964;9:1055.
Sciences Meeting and Exhibit, Reno, AIAA-2005-1170, 2005. [138] Knystautas R, Lee JHS, Guirao C. The critical tube diameter for
[115] Gamezo VN, Desbordes D, Oran ES. Two dimensional reactive flow detonation failure in hydrocarbon–air mixtures. Combust Flame
dynamics in cellular detonation waves. Shock Waves 1999;9:11–7. 1982;48:63–83.
[116] Subbotin VA. Two kinds of transverse wave structures in multifront [139] Desbordes D, Vachon M. Critical diameter of diffraction for strong
detonation. Fizika Goreniya I Vzryva 1975;11(1):96–102. plane detonations. In: Bowen JR, et al., editors. Dynamics of
[117] Pintgen F, Eckett CA, Austin JM, Shepherd JE. Direct observation explosions, vol. 106. Washington, DC: AIAA Progress in Astro-
of reaction zone structure in propagating detonations. Combust nautics and Aeronautics; 1986. p. 131–43.
Flame 2003;133(3):211–29. [140] Ciccarelli G. Critical tube measurements at elevated initial mixture
[118] Austin JM, Pintgen F, Shepherd JE. Reaction zones in highly temperatures. Combust Sci Tech 2002;174:173–83.
unstable detonations. Proc Combust Inst 2005;30:1849–57. [141] Soloukhin RI, Ragland KW. Ignition process in expanding
[119] Lee HI, Stewart DS. Calculation of linear detonation instability: detonations. Combust Flame 1969;13:295–302.
one-dimensional instability of plane detonation. J Fluid Mech [142] Edwards DH, Thomas GO, Nettleton MA. The diffraction of a planar
1990;216:103–32. detonation wave at an abrupt area change. J Fluid Mech 1979;95:79–96.
[120] Moen IO, Sulmitras A, Thomas CO, Bjerketvedt D, Thibault PA. [143] Lee JHS. Dynamics of exothermicity. In: Bowen R, editor.
Influence of cellular regularity on the behaviour of gaseous Combustion science and technology book series, vol. 2. New York:
detonations. In: Bowen JR, et al., editors. Dynamics of Explosions, Gordon Breach Publishers; 1996.
vol. 106. Washington DC: AIAA Progress in Astronautics and [144] Shchelkin KI. Two cases of unstable combustion. Sov Phys JETP
Aeronautics; p. 220–43. 1959;9:415.
[121] Radulescu MI, Sharpe GJ, Lee JHS, Kiyanda CB, Higgins AJ, [145] Whitham GB. A new approach to the problem of shock dynamics.
Hanson RK. The ignition mechanism in irregular structure Part I. Two dimensional problems. J Fluid Mech 1957;2:145.
detonations. Proc Combust Inst 2005;30:1859–67. [146] Arienti M, Shepherd JE. A numerical study of detonation
[122] Bull DC, Elsworth JE, Hooper G. Initiation of spherical detonation diffraction. J Fluid Mech 2005;529:117–46.
in hydrocarbon–air mixtures. Astronautica Acta 1978;5:997–1008. [147] Chapman WR, Wheeler RN. The propagation of flame in mixtures
[123] Lee JHS. Initiation of gaseous detonation. Ann Rev Phys Chem of methane and air. J Chem Soc (London) 1926:2139.
1977;28:75–104. [148] Shchelkin KI. Influence of tube roughness on the formation and
[124] Zeldovich I, Kogarko SM, Simonov NI. Experimental investigation detonation propagation in gas. J Exp Theor Phys 1940;10:823–7.
of a spherical gas detonation. Sov Phys Tech Phys 1956;1(8): [149] Shchelkin KI, Sokolik AS. The effect of chemical presentation on
1689–713. the initiation of the detonation wave. Acta Physicochim URSS
[125] Strehlow RA, Engel C. Transverse waves in detonations: II. 1937;7:589–96.
Structure and spacing in H2–O2, C2H2–O2, C2H4–O2 and CH4–O2 [150] Shchelkin KI. Occurance of detonation in gases in rough-walled
systems. AIAA J 1969;7(3):492–6. tubes. Sov J Tech Phys 1947;17(5):613.
[126] Bull DC, Elsworth JE, Shuff PJ, Metcalfe E. Detonation cell [151] Laffitte P, Dumanois P. Influence of pressure on the formation of
structures in fuel/air mixtures. Combust Flame 1982;45:7–22. the explosive wave. Compt Rend Acad Sci Paris 1926;183:284.
[127] Knystautas R, Guirao C, Lee JHS, Sulmistras A. Measurement of [152] Lafitte P. Influence of temperature on the formation of explosive
cell size in hydrocarbon–air mixtures and predictions of critical tube waves. Compt Rend Acad Sci Paris 1928;186:951.
diameter, critical tube initiation energy and detonability limits. In: [153] Egerton A, Gates SF. Further experiments on explosion of gaseous
Bowen JR, editor. Dynamics of shock waves, explosions, and mixtures of acetylene, of hydrogen, and of pentane. Proc Roy Soc A
detonations, vol. 94. Washington, DC: AIAA Progress in Astro- 1927;114:152.
nautics and Aeronautics; 1983. p. 23–37. [154] Salamandra GD, Bazhenova TV, Naboko IM. Formation of
[128] Ciccarelli G, Ginsberg T, Boccio J, Economos C, Sato K, Kinoshita detonation wave during combustion of gas in combustion tube.
M. Detonation cell size measurements and predictions in hydro- Proc Combust Inst 1959;7:851–5.
gen–air–steam mixtures at elevated temperature. Combust Flame [155] Soloukhin RI. Deflagration to detonation transition in gases. Soviet
1994;99:212–20. Prikladn Mech i Techn Phys (Appl Mech Tech Phys) 1961;4:128.
[129] Ciccarelli G, Ginsberg T, Boccio J. The influence of initial [156] Salamandra GD, Bazhenova TV, Zaicev SG, Soloukhin RI, et al.
temperature on the detonability characteristics of hydrogen–air– Some methods for investigation of fast-running processes. Moscow:
steam mixtures. Combust Sci Tech 1997;128:181–96. AN SSSR; 1962.
[130] Auffret Y, Desbordes D, Presles HN. Detonation structure of [157] Urtiew PA, Oppenheim AK. Experimental observation of the
C2H4–O2–Ar mixtures at elevated initial temperature. Shock Waves transition to detonation in an explosive gas. Proc of Roy Soc A
1998;9:107–11. 1966;295:1328.
ARTICLE IN PRESS
G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550 549

[158] Moen IO. The influence of turbulence on flame propagation in Dynamics of explosions, vol. 104. Washington, DC: AIAA Progress
obstacle environments. First international specialist meeting on of Astronautics and Aeronautics; 1988. p. 230–47.
fuel–air explosions, Montreal, 1981: University of Waterloo Press [180] Chao J, Lee JHS. The propagation mechanism of high speed
SM study no. 16, 1982. p. 101–13. turbulent deflagrations. Shock Waves 2003;12:277–89.
[159] Wagner HGg. Some experiments about flame acceleration, First [181] Veser A, Breitung W, Dorofeev SB. Run-up distances to supersonic
international specialist meeting on fuel–air explosions, Montreal, flames in obstacle-laden tubes, IV ISHPMI. J Phys IV France
1981: University of Waterloo Press SM study no. 16, 1982. 2002;12(7):333–40.
p. 101–35. [182] Kuznetsov M, Alekseev V, Yankin Yu, Dorofeev S. Slow and fast
[160] Peraldi O, Knystautas R, Lee JHS. Criteria for transition to deflagrations in hydrocarbon–air mixtures. Combust Sci Technol
detonation in tubes. Proc Combust Inst 1986;21:1629–37. 2002;174(5–6):157–72.
[161] Bychkov V, Petchenko A, Akkerman V, Eriksson L-E. Theory and [183] Sherman M, Berman M, Beyer R. Experimental investigation of
modeling of accelerating flames in tubes. Phys Rev E 2005;72:046307. pressure and blockage effects on combustion limits in H2–air–steam
[162] Bychkov V, Akkerman V. Explosion triggering by an accelerating mixtures. Report SAND 91-0252, 1993.
flame. Phys Rev E 2006;73:066305. [184] Sherman M, Tieszen S, Benedick W. The effect of obstacles and
[163] Bychkov V, Akkerman V, Fru G, Petchenko A, Eriksson LE. Flame transverse venting on flame acceleration and transition to detona-
acceleration at the early stages of burning in tubes. Combust Flame tion for hydrogen–air mixtures at large scale. US Nuclear
2007;150:263–7. Regulatory Commission NUREG/CR-5275, SAND 85-1264, 1989.
[164] Akkerman V, Bychkov V, Petchenko A, Eriksson L-E. Accelerating [185] Dorofeev SB, Sidorov VP, Dvoinishnikov AE, Breitung W.
flames in cylindrical tubes with non-slip at the walls. Combust Deflagration to detonation transition in large confined volume of
Flame 2006;145:206. lean hydrogen–air mixtures. Combust Flame 1996;104:95–110.
[165] Kuznetsov M, Alekseev V, Matsukov I, Dorofeev S. DDT in a [186] Dorofeev SB, Sidorov VP, Kuznetsov MS, Matsukov ID, Alekseev
smooth tube filled with a hydrogen–oxygen mixture. Shock Waves VI. Effect of scale on the onset of detonations. Shock Waves
2005;14(3):205–15. 2000;10(2):137–49.
[166] Campbell GA, Rutledge PV. Detonation of hydrogen peroxide [187] Ciccarelli G, Boccio J, Ginsberg T, Finfrock C, Gerlach L, Tagawa
vapor. Inst Chem Eng Symp Series No 33. Inst Chem Eng London; H, Malliakos A. The effect of initial temperature on flame
1972. p. 37. acceleration and deflagration-to-detonation transition phenomenon.
[167] Bollinger LE, Fong MC, Edse R. Experimental measurements and US Nuclear Regulatory Commission NUREG/CR-6509, BNL-
theoretical analysis of detonation induction distance. Am Rocket NUREG-5251, 1998.
Soc J 1961;31:588. [188] Alekseev VI, Kuznetsov MS, Yankin YuG, Dorofeev SB. Experi-
[168] Bollinger LE, Laughrey JA, Edse R. Experimental detonation mental study of flame acceleration and DDT under conditions of
velocities and induction distances in hydrogen–nitrous oxide transverse venting. J Loss Prevent Process Ind 2001;14(6):591–6.
mixture. Am Rocket Soc J 1962;32:81. [189] Card J, Rival D, Ciccarelli G. DDT in fuel–air mixtures at elevated
[169] Salamandra GD, Bazhenova TV, Zaicev SG, Soloukhin RI, et al. temperatures and pressures. Shock Waves 2005;14(3):167–73.
Some methods for investigation of fast-running processes. Moscow: [190] Ciccarelli G, Fowler CJ, Bardon M. Effect of obstacle size and
AN SSSR; 1963. spacing on the initial stage of flame acceleration in a rough tube.
[170] Lindstedt RP, Michels HJ. Deflagration to detonation transitions Shock Waves 2005;14(3):161–6.
and strong deflagrations in alkane and alkene air mixtures. Combust [191] Sorin R, Zitoun R, Desbordes D. Optimization of the deflagration
Flame 1989;76:169–81. to detonation transition: reduction of length and time of transition.
[171] Kuznetsov M, Matsukov I, Alekseev V, Breitung W, Dorofeev S. Shock Waves 2006;15(2):137–45.
Effect of boundary layer on flame acceleration and DDT. In: CD- [192] Chan CK, Greig DR. The structures of fast deflagrations and quasi-
ROM. Proceedings of 20th international colloquium on the detonations. Proc Combust Inst 1988;22:1733–9.
dynamics of explosions and reactive systems, Montreal, Canada, [193] Ciccarelli G, de Witt B. Detonation initiation by shock reflection
2005. from an orifice plate. Shock Waves 2006;15(3–4):259–65.
[172] Kuznetsov M, Singh RK, Breitung W, Stern G, Grune J, Friedrich [194] Borisov AA, Gelfand BE, Skachkov GI, et al. Selfignition of
A, Sempert K, Veser A. Evaluation of structural integrity of typical gaseous mixtures by focusing of reflected shock waves. Chimiches-
DN15 tubes under detonation loads. Report Forschungszentrum kaya Phisika 1988;7(12):1387.
Karlsruhe, December, 2003. [195] Borisov A, Gelfand B, Skatchkov G, et al. Ignition of gaseous
[173] Dorofeev SB. Run-up distances to supersonic flames in smooth combustible mixtures in focused shock waves, current topics in
tubes. In: CD-ROM. Proceedings of the 21st international shock waves. In: Proceedings of the 17th ISSW, AIP, NY, 1990.
colloquium on the dynamics of explosions and reactive systems, p. 696–701.
Poitiers, France, 2007. [196] Chan CK, Lau D, Thibault PA, Penrose JD. Ignition and
[174] Bartknecht W. Explosions. Berlin, Heidelberg, New York: Springer; detonation initiation by shock focusing. In: 17th International
1981. symposium on shock waves and shock tubes, Lehigh University,
[175] Johansen C, Ciccarelli G. Characterization of the flow field ahead of Bethlehem, Pennsylvania, USA, July 17–21, 1989; AIP Proceedings
a flame propagating in an obstructed channel. In: CD-ROM. 208, 1990. p. 161–6.
Proceedings of the 21st colloquium on the dynamics of explosions [197] Gelfand BE, Frolov SM, Medvedev SP, Tsyganov SA. Three cases
and reactive systems, Poitiers, France, 2007. of shock wave focusing in a two phase combustible medium, shock
[176] Teodorczyk A, Lee JHS, Knystautas R. Photographic study of the waves, proceedings, Sendai, Japan, 1991, vol. II. Berlin: Springer;
structure and propagation mechanism of quasidetonation in rough 1992. p. 837–42.
tubes. In: Kuhl AL, et al., editors. Dynamics of explosions, vol. 133. [198] Achasov OV, Labuda AA, Penzijakov OG, Pushkin RM. Shock
Washington, DC: AIAA Progress of Astronautics and Aeronautics; waves initiation of detonation in semiclosed cavity. Chimicheskaya
1990. p. 223–40. Phisika 1993;12(5):714–6.
[177] Teodorczyk A, Lee JHS, Knystautas R. Propagation mechanism of [199] Rose M, Roth P, Uphoff U. Ignition of reactive gas by focusing
quasi-detonations. Proc Combust Inst 1988;22:1723–31. of shock wave. In: Proceedings of the 16th ICDERS, Krakow, 1997.
[178] Chue RS, Clarke JF, Lee JHS. Chapman-Jouget deflagrations. Proc p. 554–6.
Roy Soc London 1993;441:607–23. [200] Zeldovich YaB, Librovich VB, Makhviladze GM, Sivashinsky GI.
[179] Gu LS, Knystautas R, Lee JHS. Influence of obstacle spacing on the On the development of detonation in a non-uniformly preheated
propagation of quasi-detonation. In: Kuhl AL, et al., editors. gas. Astronautica Acta 1970;15:313–21.
ARTICLE IN PRESS
550 G. Ciccarelli, S. Dorofeev / Progress in Energy and Combustion Science 34 (2008) 499–550

[201] Lee JHS, Knystautas R, Yoshikawa N. Photochemical initiation Experimental, modelling and computation in flow, turbulence and
and gaseous detonations. Acta Astronautica 1978;5:971–2. combustion, vol. 1, 1996. p. 137–57.
[202] Yoshikawa N. Coherent shock wave amplification in photo- [220] Gu XJ, Emerson DR, Bradley D. Modes of reaction front
chemical initiation of gaseous detonations, PhD thesis, Department propagation from hot spots. Combust Flame 2003;133:63–74.
of Mechanical Engineering, McGill University, Montreal, Quebec, [221] Kryuchkov SI, Dorofeev SB, Efimenko AA. Critical conditions for
Canada, 1980. detonation propagation through mixture with decreasing reaction
[203] Zeldovich YaB, Gelfand BE, Tsyganov SA, Frolov SM, Polenov rate. Proc Combust Inst 1996;26:2965–72.
AN. Concentration and temperature non-uniformities (CTN) of [222] Kuznetsov MS, Alekseev VI, Dorofeev SB, et al. Detonation
Combustible mixtures as a reason of pressure generation, vol. 114. propagation, decay, and reinitiation in nonuniform gaseous
Washington DC: AIAA Progress of Astronautics and Aeronautics; mixtures. Proc Combust Inst 1998;27:2241–7.
1988. p. 99–123. [223] Schildknecht M, Geiger W, Stock M. Flame propagation and
[204] Zeldovich YaB. Regime classification of an exothermic reaction with pressure buildup in a free gas–air mixture due to jet ignition. In:
non-uniform initial conditions. Combust Flame 1990;39:211–4. Kuhl AL, et al., editors. Dynamics of shock waves, explosions, and
[205] Clark JF. Fast flames, waves and detonations. Prog Energy detonations, vol. 94. Washington, DC: AIAA Progress in Astro-
Combust Sci 1989;15:241–71. nautics and Aeronautics; 1985. p. 474–90.
[206] Frolov SM, Gelfand BE, Tsygonov SA. Spontaneous combustion [224] Moen IO, Bjerketvedt D, Jenssen A, Thibault PA. Transition to
regimes. Sov J Explos Combust Shock Wave 1992;28(5):13–27. detonation in a large fuel–air cloud. Combust Flame 1985;61:
[207] Dorofeev SB, Kochurko AS, Chaivanov BB. Detonation onset 285–91.
conditions in spatially nonuniform combustible mixtures. Preprint [225] Moen IO, Bjerketvedt D, Engenbretsen T, Jenssen A. Transition to
IAE4871/13, Moscow, Atominform, 1989. detonation in a flame jet. Combust Flame 1989;75:297–308.
[208] Dorofeev SB, Kochurko AS, Efimenko AA, Chaivanov BB. [226] Carnasciali F, Lee JH, Knystautas R, Fineschi F. Turbulent jet
Evaluation of hydrogen explosion hazard. Nucl Eng Design initiation of detonation. Combust Flame 1991;84:170–80.
1994;148(2):305–10. [227] Bezmelnitsyn AV, Dorofeev SB, Yankin YuG. Direct comparison of
[209] Dorofeev SB, Sidorov VP, Dvoinishnikov AE, Breitung W. detonation initiation by turbulent jet under confined and unconfined
Deflagration to detonation transition in large confined volume of conditions. In: Proceedings of the 16th international colloquium on
lean hydrogen–air mixtures. Combust Flame 1996;104:95–110. dynamics of explosions and reactive systems. Cracow, Poland, 1997.
[210] Nikiforakis N, Clarke JF. Numerical studies of the evolution of p. 222–5.
detonations. Math Comput Model 1996;24:149–64. [228] Bezmelnitsyn AV. Experimental study of conditions for detonation
[211] Gelfand BE, Medvedev SP, Polenov AN, Khomik SV, Bartenev initiation by a jet of combustion products. PhD dissertation, RRC
AM. Basic selfignition regimes and conditions for their realization in ‘‘Kurchatov Institute,’’ Moscow, 1998.
combustible gas mixtures, combustion. Explos Shock Waves [229] Pfahl UJ, Shepherd JE. Jet initiation of deflagration and detonation
1997;33(2):127–33. in stoichiometric H2–O2–N2 mixtures. GALCIT report FM 99-1.
[212] Short M, Dold JW. Corrections to Zeldovich’s ‘‘Spontaneous California Institute of Technology, Pasadena, CA, USA, 1999.
Flame’’ and the onset of detonation via nonuniform preheating. In: [230] Kogarko SM, Zeldovich YaB. Dokl Akad Nauk SSSR 1948;63:553.
Kuhl AL, et al., editors. Dynamic aspects of explosion phenomena, [231] Guirao CM, Knystautas R, Lee JH. A summary of hydrogen–air
vol. 154. Washington, DC: AIAA Progress in Astronautics and detonations for reactor safety. Sandia National Laboratories/
Aeronautics; 1993. p. 59–75. McGill University, Report NUREG/CR–4961, 1989.
[213] Dold JW, Short M, Clark JF, Nikiforakis N. Accumulating [232] Kuznetsov MS, Alekseev VI, Dorofeev SB. Comparison of critical
sequence of ignitions from a propagating pulse. Combust Flame conditions for DDT in regular and irregular cellular detonation
1995;100:465–73. systems. Shock Waves 2000;10(3):217–23.
[214] Smijanovski V, Klein R. Flame front tracking via in-cell reconstruc- [233] Dorofeev SB, Sidorov VP, Breitung W, Kotchourko AS. Large-scale
tion. In: Proceedings of the fifth international conference on combustion tests in the RUT facility. In: Proceedings of the 14th
hyperbolic systems, SUNY, Stony Brook, June 1994. international conference on structural mechanics in reactor technol-
[215] Khokhlov AM, Oran SE, Wheeler JC. A theory of deflagration-to- ogy. Lyon, France, vol. 10, 1997. p. 275–83.
detonation transition in unconfined flames. Combust Flame [234] Sidorov VP, Dorofeev SB. Influence of initial temperature, dilution,
1997;108:503–17. and scale on DDT conditions in hydrogen–air mixtures. Arhivum
[216] Khokhlov AM, Oran ES, Thomas GO. Numerical simulation of Combustionis 1998;18:87–103.
deflagration-to-detonation transition: the role of shock–flame [235] Chan CK, Dewit WA. Deflagration-to-detonation transition in end
interactions in turbulent flames. Combust Flame 1999;117:323–39. gases. Proc Combust Inst 1996;26:2679–84.
[217] Bartenev AM, Gelfand BE. Weak and strong ignition within the [236] Dorofeev SB, Sidorov VP, Breitung W. Explosions resulted from the
scope of spontaneous flame concept. Proc Combust Inst accidental release and ignition of hydrogen at dynamic conditions.
1994;25:61–4. Second international specialist meeting on fuel–air explosions.
[218] Blumental R, Fieweger K, Komp KH, Adomeit G. Gas dynamic Bergen, Norway, 1996. p. 3.36–3.48.
features of self ignition of nondiluted fuel/air mixtures at high [237] Kuznetsov M, Ciccarelli G, Dorofeev S, Alekseev V, Yankin Y, Kim
pressure. Combust Sci Technol 1996;137:113–4. TH. DDT in methane–air mixtures. Shock Waves 2002;12:215–20.
[219] He L, Clavin P. Numerical and analytical studies of the initiation of [238] Dupre G, Joannon J, Knystautas R, Lee JH. Unstable detonations in
combustion wave by hot pockets. In: Desileri J, Wiley J, editors. the near-limit regime in tubes. Proc Combust Inst 1991;23:1649–58.

Potrebbero piacerti anche