Journal of Fluid Mechanics

Download as pdf or txt
Download as pdf or txt
You are on page 1of 33

This draft was prepared using the LaTeX style file belonging to the Journal of Fluid Mechanics 1

Dynamics of cellular flame deformation after


a head-on interaction with a shock wave:
arXiv:2009.02546v1 [physics.flu-dyn] 5 Sep 2020

reactive Richtmyer-Meshkov instability


Hongxia Yang1,2 †, and Matei Ioan Radulescu2
1
Fire & Explosion Protection Laboratory, Northeastern University, Shenyang, 110819, China
2
Department of Mechanical Engineering, University of Ottawa, 161 Louis-Pasteur, Ottawa,
K1N 6N5, Canada

(Received xx; revised xx; accepted xx)

Shock flame interactions are fundamental problems in many combustion applications


ranging from flame acceleration to flame control in supersonic propulsion applications.
The present paper seeks to quantify the rate of deformation of the flame surface and
burning velocity caused by the interaction and to clarify the underlying mechanisms.
The interaction of a single shock wave with a cellular flame in a Hele-Shaw shock
tube configuration was studied experimentally, numerically, and theoretically. A mixture
of stoichiometric hydrogen-air at sub-atmospheric pressure was chosen such that large
cells can be isolated and their deformation studied with precision subsequent to the
interaction. Following passage of the incident shock, the flame cusps are flattened and
reversed backwards into the burnt gas. The reversed flame then goes through four
stages. At times significantly less than the characteristic flame burning time, the flame
front deforms as an inert interface due to the Ricthmyer-Meshkov instability with non-
linear effects becoming noticeable. At times comparable to the laminar flame time,
dilatation due to chemical energy release amplifies the growth rate of Ricthmyer-Meshkov
instability. This stage is abruptly terminated by the transverse burnout of the resulting
flame funnels, followed by a longer front re-adjustment to a new cellular flame evolving
on the cellular time scale of the flame. The proposed flame evolution model permits to
predict the evolution of the flame geometry and burning rate for arbitrary shock strength
and flame properties (burning velocity and expansion ratio) in two-dimensions.

Key words:

1. Introduction
The interaction of shocks with flames is a fundamental problem of reactive compressible
flows with applications to flame acceleration and flame control in propulsion systems.
Experimental evidence suggests that the passage of a shock over the flame results in
important deformation of the flame surface and subsequent enhancement of the burning
rate (Markstein 1957; Rudinger 1958; Urtiew & Oppenheim 1968; Scarinci et al. 1993;
Ciccarelli et al. 2010; Rakotoarison et al. 2019; Wei et al. 2017). Also, the shock reflected
from walls or other reflective surfaces may result in multiple interactions, which would
lead to more considerable flame acceleration and possibly the occurrence of deflagration-
† Email address for correspondence: yang.hongxia@foxmail.com
2 H. Yang and M. I. Radulescu
to-detonation transition (Thomas et al. 2001). The generally very rapid interactions make
it hard to quantify the enhancement of the burning rate, which depends intimately on
the flame front surface and the thermodynamic properties of the flow field.
Efforts have focused on laminar flames, due to the well-posedness of initial conditions.
An example is the classic work of Markstein (1957), which investigated the head-on
collision between an initially curved laminar flame and a planar shock. He attributed the
mechanism for the flame reversion to be a modified version of Taylor instability. More
recent studies also suggest the Richtmyer-Meshkov (RM) instability, which is generally
referred to the impulsive counterpart of Rayleigh-Taylor instability, to be the main factor
associated to the increase of the total burning velocity by the extension of the flame front
after the shock-flame interaction (Ju et al. 1998; Khokhlov et al. 1999; Oran & Gamezo
2007; Dong et al. 2008; Zhu et al. 2013; Jiang et al. 2016; Massa & Jha 2012).
The linear and non-linear growth of the inert RM instability is now relatively well
understood. A quick review is given here, since the reactive growth explored in the present
paper builds upon this. The RM instability is known to amplify the perturbation on the
interface through the baroclinic torque mechanism, which results from the misalignment
of the pressure gradient of the shock and the density gradient across the interface.
Since the pioneering work of Richtmyer (1960), extensive experiments, simulations and
theoretical models have been formulated to elucidate the interface deformation subject
to the inert RM instability. Reviews of these studies can be found in Brouillette (2002)
and Zhou (2017a,b). The basic configuration of a shock passing a sinusoidal interface
is generally adopted in these studies. The amplitude growth is typically recognized as
a key indicator to represent the interface deformation. For the case of a shock passing
a sinusoidal interface separating a light and a heavy fluid, the initial linear growth of
the interface amplitude is well predicted by the impulsive model of Richtmyer (1960) for
sufficiently small perturbation amplitudes. The growth rate of the interface amplitude
η 0 is given by the impulsive model and takes the form η 0 = kAt [u]η0 , where k is the
wavenumber, [u] is the velocity jump associated to the passage of the shock, η0 is the
initial amplitude and At = ρ2 −ρ1/ρ2 −ρ1 is the Atwood number with ρ1 and ρ2 refer to
the light and heavy fluid density. Conversely, when the shock wave is directed from
the heavy fluid to the light one, the interface amplitude decreases before reversing its
phase and growing linearly. This linear growth relation has been proposed by Meyer &
Blewett (1972) with an improvement to the impulsive model by using the average of the
amplitudes before and after the interaction along with the post-shock Atwood number.
With the amplification of the amplitude progresses to be comparable to the wavelength,
non-linear effects become significant. The non-linear mechanism leads to a decrease in the
growth rate. Numerous models (Sadot et al. 1998; Mikaelian 2003; Zhang & Sohn 1999;
Kilchyk et al. 2013; Zhang & Guo 2016) have been developed to predict the amplitude
growth in the nonlinear stage for arbitrary Atwood numbers, the late time non-linear
growth rate of the interface was found to decay at the rate of O(1/t). Mikaelian (2003)
proposed a model to join the linear and non-linear regime by making an abrupt change
in the growth rate for arbitrary Atwood numbers. The model was found to show good
agreement to other studies (Goncharov 1999; Oron et al. 2001).
Studies on the reactive RM instability have demonstrated that the flame deformation
induced by the single shock-laminar-flame interaction can be assumed to be in the laminar
flamelet regime (Ju et al. 1998; Khokhlov et al. 1999; Lutoschkin et al. 2013). The
main features and the time scale of the initial flame evolution are similar to that of
the inert case. While profound differences were noted in the later non-linear stages,
the perturbation growth of the interface was found to be larger than that of the inert
interface instability case due to both the interface stretching and flame propagation
Shock cellular flame interaction 3
mechanism. As a result, the non-linear model for the inert RM instability would fail
to predict the interface evolution for the reactive cases. Furthermore, the competition
between the initial torque by the shock and the chemistry-induced baroclinic effect
leads to the disappearance of small scale perturbations in the reaction cases than in
the inert shock interface interaction cases (Khokhlov et al. 1999; Massa & Jha 2012;
Wijeyakulasuriya & Mitra 2014). Scarinci et al. (1993) evaluated the burning rate increase
for corrugated flames with not too large initial amplitude through a series of experiments
by passing the shock wave to a pre-defined corrugated flame front. By the inspection of
the pressure records and conservation laws arguments, the different mechanisms of the
thermodynamic dependence of flame speed, the increase of transport rates by interaction
generated turbulence and deformation of the flame surface that might have influenced
the resultant burning rate increase was examined. In the analysis, they attributed the
dominant mechanism to the RM instability through the increase of the burning interface
and the turbulent energy acting on the flame interface. The effect of chemistry was found
to be more important at the later stage of the interaction, especially for higher Mach
numbers (Chen et al. 2018). The results from the quasi-one-dimensional investigation
of the problem by separating the one-dimensional effects of gas compression and two-
dimensional flame front distortion also found the gas compression to be the dominant
effect in influencing the total burning rate increase for stronger incident shocks (Scarinci
et al. 1993; Kilchyk et al. 2013; Lutoschkin et al. 2013). However, the mechanism on the
competition between the aforementioned potential effects at different stages is still not
well understood.
While previous work on shock-flame interactions reviewed above treated either a
laminar flame with small perturbations for well-posedness, or the turbulent or strained
flames constrained by the tube geometry for practical considerations as in the work of
Markstein (1957) and Thomas et al. (2001), the present study addresses cellular flames.
Indeed, it is well known that planar laminar flames are unstable to both hydrodynamic
(Landau-Darrieus) and thermo-diffusive instabilities (Sivashinsky 1983), which make
them saturate to a cellular structure of finite amplitude (Sharpe & Falle 2006). Flames
in practice where shock-flame interactions may be of importance, as in astrophysical,
engine or large-scale accidental explosion scenarios are likely cellular due to the much
larger characteristic propagation distance compared with the laminar flame thickness.
These cellular shapes are taken also conveniently as the initial seeds of the RM-type
instability induced by the passage of the planar shock. The objective of this paper is
thus to quantitatively clarify the controlling mechanism on the flame front deformation
for the fundamental problem of head-on interaction of a shock with a cellular unstable
flame experimentally, numerically and theoretically. The experiments are performed at
low pressure in stoichiometric hydrogen-air, such that large cells can be isolated and their
deformation studied with precision after passage of a shock wave. Also, the interaction
can be readily amenable for the numerical simulation and theoretical modeling. Similar
to the inert RM instability studies, in the experiments, we take the flame front amplitude
as an indicator for the evolution of the flame surface, under the impulsive action of the
shock wave leading to RM instability. Numerical simulations are conducted to extend the
observational time limited in the experiments. A simplified model is then proposed to
predict the evolution of the flame geometry and burning rate for arbitrary shock strength
and flame properties.
The remainder of the present paper is organized as follows. Section 2 provides a brief
description of the experimental details. The experimental results of the stoichiometric
hydrogen-air flame and the head-on interactions of the cellular flame with various incident
shocks are given in section 3. Sections 4 and 5 present the numerical method and physical
4 H. Yang and M. I. Radulescu
3400mm
Hot wire ignition
p1 p2 p3 p4 p5 p6 p7
Obstacles to promote l1 l2 l2 l3 l2 l2 l2
Ignition detonation formation Diaphragm

Incoming Cellular End


203mm Shock Flame wall

Driver section Test section 140mm

Figure 1: Schematic of the experiment setup in the shock tube, l1 = 203 mm, l2 = 102
mm, l3 = 406 mm.

model to reconstruct the observed experimental flow field. Subsequent discussion on the
evolution of the flame deformation following the interaction with the incident shock is
presented in section 6. Finally, conclusions are given in section 7.

2. Experimental procedure
The experiments were conducted in a shock tube with dimensions of 3400 mm × 19.1
mm × 203.2 mm, as shown in figure 1. The flame was ignited by a 0.15-mm-thick hot
tungsten wire mounted at the right end of the shock tube, which consisted of the test
mixture of stoichiometric hydrogen-air. A reactive driver gas (C2H4+3O2) generated a
shock wave propagating to the right, while the flame that was ignited at the opposite
end has propagated a sufficient distance to acquire the desired cellular structure in the
test mixture. The driver gas was separated from the test gas by a layer of aluminum
diaphragm. The shock was initiated by triggering a detonation in the driver gas. This
detonation passed the aluminum diaphragm and transmitted the shock wave into the
test gas, followed by a much weaker flame that did not participate in the experiment.
The transmitted shock then traveled toward the test flame and interacted head-on. The
strength of the incident shock was controlled by varying the initial pressure of the driver
gas.
A series of high-frequency piezoelectric PCB pressure sensors (p1-p7) were mounted
flush on the top wall of the shock tube to collect pressure signals and the arrival of the
shock. The pressure signals were sampled at a rate of 1.538 MHz, and low-pass filtered at
100 kHz. The incident shock Mach number was evaluated using the pressure amplitude
measured at each gauge and confirmed by the time of arrival. For each experiment, the
shock tube was evacuated to a pressure less than 80 Pa. Then the C2H4+3O2 mixture was
filled into the driver section before filling the rest part with 2H2/air mixture. The gases
were prepared in mixing tanks by the method of partial pressure and left to mix for more
than 24 hours. A pair of optical quality glass window was installed at the test section
in order to visualize the phenomenon. A Z-type Schlieren system with a field of view
of 317.5 mm was implemented to capture the shock-flame interaction and the ensuing
evolution. The image sequence was recorded using a high-speed camera (Phantom v1210).
The frame rate was 59,590 frames per second (fps) with a resolution of 512 × 320, and
the exposure time was set to 0.468 µs.
Shock cellular flame interaction 5

Figure 2: Superposition of flame fronts development at different time for stoichiometric


hydrogen air ignited by hot tungsten wire. Pressure of 20.7 kPa. Recorded at 59,590 fps.
The time interval between the flame front is 0.25 ms.

3. Experimental results
3.1. Flame evolution before the interaction
Figure 2 shows an example of the cellular flame establishment prior to its interaction
with the shock. The superimposed Schlieren image illustrates the stoichiometric 2H2 /air
flame fronts evolution for the very early stage at an initial pressure of 20.7 kPa with a time
interval of 0.25 ms. The flame propagated from right towards left. It can be seen from the
variation of the distance between the flame fronts acquired in sequential frames in figure
2 that, once ignited by the vertical hot tungsten wire in the right end, the flame first
slowly accelerated as the perturbation curvature increased, and then decelerated to form
two cells. As the newly formed cells propagated forward, the increase of their amplitudes
is noticeable. Figure 3 further demonstrates the birth of new cells from the bottom large
cell ensuing the formation of two cells, with an initial pressure of 17.2 kPa. The flame
front evolution was tracked every 10th pixels horizontally in each of the sequential frames,
as shown in the space-time diagram in figure 3b. Evident from the nearly linear slope of
the flame front streaks in figure 3b is that the flame propagated with practically steady
speed before the formation of the third cell, and it showed the same trend after the fully
development of three cells. The interactions of shock and three-cells-flame with the initial
pressure of 17.2 kPa are reported as follows.

3.2. Interaction of Ms = 1.9 shock and the flame


Figure 4 shows the detailed evolution of the interaction of a shock with propagation
Mach number Ms = 1.9 and a stoichiometric 2H2 /air flame, which has an initial pressure
of 17.2 kPa before ignition. The first two frames illustrate the cellular flame structure
and evolution before the interaction. Note that comparing to the middle cell, the top and
bottom cell are highly asymmetric because of the top and bottom walls of the channel.
The right propagating shock can be seen in the second frame to reach the flame interface.
In the third frame, the incident shock wave (ISW) passed half the flame front, and gave
rise to a curved transmitted shock wave (TSW) and a reflected expansion wave. Note
6 H. Yang and M. I. Radulescu
5.50ms 10.54ms

1 4

7.18ms 12.22ms

2 5
8.86ms 13.90ms

3 6

(a)

Forming the third cell

t
(b)

Figure 3: (a) Schlieren image sequence, (b) space-time diagram obtained every 10th
pixels for stoichiometric hydrogen-air cellular flame development. Pressure of 17.2 kPa.
Recorded at 59,590 fps.

that the incident shock reached the bottom cell first, and led to the earlier formation
of the transmitted shock than the rest of the flame. Following the passage of the shock,
the flame was flattened and pushed backwards to the burned gas. The fourth to eighth
frames show the further reversion of the flame cusps and the growth of the flame cells
amplitudes. Here we define the amplitude to be the maximum horizontal distance from
the tip to the root linking the intersection point of the cell with the adjacent cells, as
marked as η in figure 5. With the elongation of the flame cusps, one can also observe
the closing-up of the flame funnels, which implies the combustion toward the inside of
the funnel. Note that the closing-up effects are more evident for smaller cells. Another
Shock cellular flame interaction 7
-1.22ms 0.12ms

TSW
MF

1 F 5

-0.03ms 0.15ms

ISW

2 6

0.05ms 0.18ms

3 7

0.08ms 0.22ms

TSW

F
4 8

Figure 4: Schlieren image sequence of the interaction of a Ms = 1.9 incident shock


wave(ISW) with stoichiometric hydrogen-air flame(F). TSW is the transmitted shock
wave, MF is the middle flange of the shock tube. Time 0 corresponds to the beginning
of the interaction.

point to notice is that although before the interaction the initial amplitudes of the cells
are evidently different, the reversed funnels developed to lengths similar to each other in
the last three frames.
To quantify the enhancement effect of the incident shock wave on the flame front, the
cell amplitudes growth were measured and shown in figure 6a. Before the measurement
of the cell amplitudes, an edge detection method was adopted to process the schlieren
images. The leading edge of the flame was chosen to be at the front gradient band. Here
we assume each half cell to be 1/4 of a sine wave, thus the wavelength can be defined as 4
times the vertical height of a half cell. In consideration of the cell asymmetry, as shown in
8 H. Yang and M. I. Radulescu
a
b

c 1/4λ
η

Figure 5: Sketch of the flame cells configuration. The amplitude and wavelength for cell
a, b, c, d, e and f are measured separately.

the sketch in figure 5, the amplitude was measured separately for the top and bottom half
of each cell. To compare the amplitudes evolution of the different cells, we measured the
amplitudes evolution in terms of the relative amplitude difference and normalized by the
wavenumber. As mentioned in the MB theory, the average of the amplitudes before and
after the interaction is a parameter that influences the growth rate of the amplitudes. The
amplitudes were thus also divided by (η0 − +η0 + ) for comparison. Figure 6a shows the time
+

evolution of the normalized amplitudes with the expression of η = k(ηη−η −
0
0 +η0 )
+ , where η0

and η0+ refer to the flame amplitude at time -0.03 ms and 0.05 ms in this experiment. The
error bar represents the uncertainty from the thick band in the measurement. As shown
in figure 6a, the shock first passed the cell (f) at approximately 0.034 ms and traversed all
the cells at 0.075 ms, and led to the decrease of the flame amplitude. Subsequent to the
interaction, the compressed flame interface grew nearly linearly with time for all the 6
half-cells. Note the growth rate of the normalized amplitudes for all the cells are in good
agreement with each other. Analysis of the sequential frames in the experiments also
permits us to evaluate the average flame propagation speed following the interaction,
as shown in figure 6b. Assuming the flame is two-dimensional, the flame propagation
speed is determined by the rate of change of the volume weighted flame position. By
taking the middle flange as a reference, we measured the volume between the middle
flange to the flame front in each consecutive frames, the volume weighed distance was
then evaluated by dividing the volume by the channel height and width. During the
passage of the shock, the left moving flame was pushed back and acquired a maximum
speed of 742 m/s. Then the propagation speed went through a decrease with fluctuation
within the range of visualization after the shock has fully passed the flame, implying
an increase in the burning velocity opposite to the propagation direction. The average
propagation speed was used for the reconstruction of the flow field to evaluate the average
burning velocity increase within the visualization time, as reported in Appendix A. Over
the time interval available in the experiments, the burning velocity was evaluated to be
7.1±4.5m/s for the case of the Ms =1.9 shock-flame interaction.

3.3. Effects of incident shock strength


Figure 7 shows the evolution of the flame with a weaker incident shock of Mach number
1.75. Similar to the case in figure 4, the first two frames show the structure of the cellular
flame before the interaction. The shock reached the flame in the second frame, formed a
transmitted shock wave and a rarefaction wave after passing the flame in the third frame.
The fourth and fifth frames show the time interval between 0.18 ms and 0.22 ms after
the interaction and are to be compared to the seventh and eighth frames in figure 4. One
Shock cellular flame interaction 9
0.045
a
0.04
b
0.035 c
d
0.03
e
0.025 f

η (m)
0.02

_
0.015
0.01
0.005
0
-0.005
0 0.05 0.1 0.15 0.2 0.25
t (ms)

(a)
800
700
600 Shock passed

500
Sf (m/s)

400
300
200
100 Before
interaction
0
-100
-0.05 0 0.05 0.1 0.15 0.2 0.25
t (ms)

(b)

Figure 6: Time evolution of (a) normalized cell amplitudes and (b) average flame
propagation speeds following the interaction of a Ms =1.9 incident shock wave with the
stoichiometric hydrogen-air flame.

can notice that with lower incident shock strength, the flame reversion and propagation
is slower than the case of the shock of Ms = 1.9. The flame amplitude also increased
slower for the case with lower incident shock, while the closing up of the funnel shows
no obvious difference between the two cases. The flame cusps further reversed in the
sixth to eighth frames with larger amplitude burning rate toward the inside of the flame
funnel. In the last frame, the very thin funnel neck implies that the mushroom caps were
separating from the main structure.
Figure 8 shows the time evolution of the normalized amplitude for the case of Ms =
1.75. Similarly, the shock passed the cell (f) before the rest of the cells. The amplitudes of
the compressed flame interface, which formed during the passage of the incident shock,
showed consistent nearly linear increase with time until approximately 0.2 ms for all
the cells. Then nonlinear growth of the amplitudes can be observed from 0.2 ms. In
the non-linear stage, the growth rate of the amplitude gradually decay and slightly
diverged from each other for the cells. The average flame propagation speed following the
interaction is plotted in figure 8b. Limited by the middle flange and the Schlieren system,
10 H. Yang and M. I. Radulescu
-2.47ms 0.22ms

MF

1 5

0.0ms 0.25ms

ISW
2 6

0.13ms 0.28ms

TSW

F 3 7

0.18ms 0.32ms

4 8

Figure 7: Schlieren image sequence of the interaction of a Ms = 1.75 incident shock


wave with stoichiometric hydrogen-air flame. Time 0 corresponds to the beginning of the
interaction. Flame initial pressure of 17.2 kPa. Recorded at 59,590 fps.

we plotted only the flame displacement speed after the shock has passed. Subsequent to
the interaction with the shock, the flame propagation speed gradually decreases with
a fluctuation through the examined channel, which implies the increase in the burning
velocity opposite to the flame propagation direction.
In figure 9, we present the interaction of the flame with further decreased incident
shock Mach number of 1.53. The first stage of the interaction until 0.25 ms are similar
to the results in figures 4 and 7, although with moderate reversion, slower amplitude
growth and more smooth flame front than the other cases due to the lower strength of
the incident shock. As shown in the 5th to 8th frame, the flame tip gradually separated
Shock cellular flame interaction 11
0.09
a
0.08 b
c
0.07 d
e
0.06 f
0.05

η (m)
0.04
_ 0.03
0.02
0.01
0
-0.01
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35
t (ms)

(a)
650

600

550
Sf (m/s)

500

450

400

350
0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3
t (m/s)

(b)

Figure 8: Time evolution of (a) normalized cell amplitude and (b) average flame
propagation speed following the interaction of a Ms =1.75 incident shock wave with
stoichiometric hydrogen-air flame.

from the main structure 0.3 ms after the interaction because of the burn-out of the
unburned gas inside the cell funnels.
Figure 10 compared the evolution of cell amplitudes and flame propagation speeds
after the shock fully passed the flame for all the incident Mach numbers considered.
For simplicity, the middle-down cell was chosen for comparison. The first to notice is
that the basic structure for the growth of the amplitude and the development of the
flame propagation speed are similar for the cases considered. Whereas the growth of
the amplitude and the flame propagation speed are more pronounced for higher Mach
numbers. After the passage of the incident shock, the amplitudes grew nearly linearly.
Then as the flame evolved, the growth of the amplitude gradually slowed down. For the
case with weak incident shocks, the decay of growth rates is more obvious.
12 H. Yang and M. I. Radulescu
-4.13ms 0.30ms

MF

1 5

0.0ms 0.35ms

ISW
2 6

0.20ms 0.40ms

F
TSW

3 7

0.25ms 0.44ms

4 8

Figure 9: Schlieren image sequence of the interaction of a Ms = 1.53 incident shock


wave with stoichiometric hydrogen-air flame. Time 0 corresponds to the beginning of the
interaction. Flame initial pressure of 17.2 kPa. Recorded at 59,590 fps.

4. Numerical method and physical model


In the experiments, the duration of the flame evolution is limited by the length of
the apparatus and the dimension of the Schlieren visualization section. The evolution of
the flame to later times was thus addressed numerically. We consider a two-dimensional
problem in a rectangular domain, for which we model the flame evolution and interaction
Shock cellular flame interaction 13
60

50

40

η/η0
30

20

10 Ms=1.9
Ms=1.75
Ms=1.53
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
t (ms)

(a)
800
Ms=1.9
Ms=1.75
700 Ms=1.53

600
Sf (m/s)

500

400

300

200

100
0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
t (ms)

(b)

Figure 10: Time evolution of (a) amplitudes of cell d and (b) propagation speeds
subsequent to the interaction for different incident shock strength of Ms = 1.9, 1.75
and 1.53, η0 corresponds to the respective initial amplitude before the interaction.

by the reactive Navier-Stokes equations.


∂ρ
+ ∇ · (ρu) = 0, (4.1)
∂t
∂ρu 2
+ ∇ · (ρuu) + ∇p = ∇ · (µ(∇u + (∇u)T − (∇ · u)I)), (4.2)
∂t 3
∂ρe
+ ∇ · ((ρe + p)u) = ∇ · (u · τ ) − ∇ · (K∇T ) + Qω̇, (4.3)
∂t
∂ρY
+ ∇ · (ρuY ) + ∇ · (ρD∇Y ) = ω̇, (4.4)
∂t
where ρ, u, p, e and Y refer to mass density, velocity vector, pressure, specific total
energy and mass fraction of the product, respectively. The equation of state satisfies the
ideal gas law as follows:
p p u2
T = , e= + , (4.5)
Rs ρ ρ(γ − 1) 2
14 H. Yang and M. I. Radulescu

ρ0 0.1474 kg/m3 SL 1.98 m/s γ 1.4013 Lf 0.0034 m


K/(ρcp ) 2.6 ×10−4 m2 /s µ 1.8 ×10−5 pa s Ea /R 27390 K−1 Q 3.0 ×106 J/kg
A 4.43 ×1011 s−1 Pr 0.4655 Le 1.0

Table 1: Thermo-chemical properties and model parameters for the stoichiometric


hydrogen-air combustion at T=294 K and p0 = 17.24 kPa.

where Rs is the specific gas constant. The reaction rate ω̇ is described by the single
step Arrhenius kinetics,
Ea
ω̇ = ρA(1 − Y ) exp(− ). (4.6)
RT
Other properties to note are the isentropic index γ, the viscosity µ, the thermal
conduction coefficient K, the mass diffusion coefficient D, the activation energy Ea , the
pre-exponential factor A and the universal gas constant R. Also, I refers to the unit
matrix.
The system of equations is non-dimensionalized as follows

ρ u v P P̃ Rs T 
ρ̃ = , ũ = , ṽ = , P̃ = , T̃ = = , 
ρ0 (SL ) 2 (SL ) 2

ρ0 SL SL ρ̃ (4.7)
e ρ0 SL cp ρ0 SL cp ρ0 (SL ) 2 cp
ẽ = , x̃ = x, ỹ = y, t̃ = t,


ρ0 (SL ) 2

K K K
where the characteristic density ρ0 is the initial density of the fresh gas, the characteristic
speed SL is the one-dimensional laminar flame speed, the characteristic flame length
scale Lf s = ρ0 SKL cp gives a measure of the pre-heat zone thickness of steady planar flame
L
(Sharpe & Falle 2006; Strehlow 1984), tf s = SfLs is the characteristic flame time. Here,
cp denotes the specific heat at constant pressure. Furthermore, it is useful to mention the
non-dimensional Lewis (Le ) and Prandtl (Pr ) numbers:
K µcp
Le = , Pr = . (4.8)
cp D K
The input parameters for the model are given in table 1 to fit the laminar free
flame of stoichiometric hydrogen-air mixture at 17.2 kPa initial pressure and 294 K
initial temperature. The initial flame density, isentropic index, fluid viscosity, and heat
conductivity were evaluated for the unburned gas composition using Cantera thermal-
chemical tools and the Li mechanism (Li et al. 2004). The laminar flame speed and the
flame thickness Lf were calculated from the one-dimensional free flame. The heat release
was evaluated by the enthalpy difference between the fresh and burned gases. The Lewis
number was set to be 1.0 following Jomaas et al. (2007). The global activation energy
was determined from  
∂(ln(ρ0 SL ))
Ea = −2R (4.9)
∂(1/Tad ) p0 ,φ
following Egolfopoulos & Law (1990), by calculating the mass burning rate (ρ0 SL ) for the
given initial pressure(p0 ) and equivalence ratio (φ), then slightly varying its value through
the substitution of a small quantity of nitrogen by argon. Here, Tad is the adiabatic flame
temperature.
The pre-exponential factor was calculated by solving a system of equations for the
laminar flame speed satisfying the compressible one-dimensional steady flame structure
Shock cellular flame interaction 15
9 1
8
7 0.99995

6
0.9999
5

P/P0
T/T0

4
0.99985
3
2 0.9998
1
0 0.99975
0 5 10 15 20 25 0 5 10 15 20 25
x/Lfs x/Lfs

(a) (b)
1 1
0.9 0.9

0.8 0.8
0.7
0.7
0.6
0.6
0.5
ρ/ρ0

Y
0.5
0.4
0.4
0.3
0.3 0.2
0.2 0.1
0.1 0
0 5 10 15 20 25 0 5 10 15 20 25
x/Lfs x/Lfs

(c) (d)

Figure 11: Steady 1D flame profile of temperature, pressure, density and reaction progress
variable.

using the shooting method proposed by Travnikov et al. (1997). Figure 11 shows the
results of the structure for temperature, pressure, density and reaction progress variable
of the 1D steady flame in the frame of reference the flame. Note that the pressure only
changes very little, given that the flame burning velocity is much smaller than the sound
speed in the fresh gases. The laminar flame thickness (Lf ), which was determined as
dx
(Tb − T0 )( dT )max , was evaluated to be 1.48Lf s and later used to compare with the
L
experiments. For comparison with the experiments, we also define tf = SLf to be the
laminar flame time.
The cellular flame before the interaction was developed by imposing a small pertur-
bation on the evaluated 1D flame in the longitudinal direction. The perturbed flame
was placed in the middle of a rectangular channel with a domain length of 1000 flame
lengths and a height of 16.5 Lf s that represents the width of the cell (c) in figure 5 or
the dashed rectangle in figure 4 (2) before the interaction. The perturbation was set to
be A0 cos(2πy)/λ where A0 = 0.01Lf s . The flame then freely evolved to form a steady
cellular structure, which helps to prevent to deal with the initial unsteady evolution effect
from subsequent dynamics. When the flame has propagated sufficient enough to form the
steady cellular flame structure, the jumps in the fluid state parameters corresponding to
a shock with a Mach number in the range of 1.53 to 2.5 were imposed in the left, 50 flame
lengths from the flame, with the post shock parameters determined from the Rankine-
Hugoniot relations. The length to the right of the flame was then varied within a range
16 H. Yang and M. I. Radulescu
between 500 and 2000 Lf s to investigate the flame evolution after the interaction. The
top and bottom boundaries were set to be symmetric while the left and right boundaries
had imposed zero-gradient conditions.
The numerical model detailed above was solved using a finite-volume code developed
by S. Falle at the University of Leeds (Maxwell et al. 2018) with second-order accurate
Godunov Riemann solver and adaptive mesh refinement (Falle et al. 1993). The method
has been demonstrated in the papers of Sharpe & Falle (2006) and Sharpe et al. (2008) for
the simulation of flame propagation using the Navier-Stokes equations. In the simulation,
a base grid of 3 points per flame thickness with five refinement levels was used. The
reaction and diffusion zone were enforced to have the highest refinement level, giving an
effective of 48 points per flame thickness which is higher than that suggested in Sharpe &
Falle (2006) for properly resolve the cellular flame. The refinement also ensured that all
shock waves formed by the interaction were properly captured. The results of resolution
studies are given in Appendix B.

5. Numerical results
5.1. Establishment of a cellular flame
Figure 12a shows the evolution of an initially planar flame, to which an initial pertur-
bation of half cosine wave with an amplitude of 0.01Lf s was applied. The contour was
obtained from the locus of maximum heat release along the flame. The interface of the
initially sinusoidally perturbed flame increase exponentially until approximately 35 tf s .
The exponential increase can better be observed in the plot of the amplitude and the
burning velocity in figure 12. Here, we define the amplitude as the largest x-direction
distance on the flame front, and the burning velocity as
RR
ω̇dxdy
Su = s , (5.1)
H ρ̄u
where ρ̄u is the average density of the unburned gas in the section of the channel occupied
by the flame, H is the domain height. With the development of the flame, the nonlinear
effects became important, and the flame began to accelerate and moved forward. The
steady cellular flame was formed at approximately 35tf s , when the flame has acquired
a speed of 1.3SL and an amplitude of 11.6Lf s . The flame structure at 49.9tf s is then
compared with the reference cell from the experiment, as shown in figure 13. Note that
though the flame structure in the simulation is achieved by a perturbation development,
the acquired structure is in very good agreement with the experimentally determined
shape, with some differences observed at the cell corners. Once such a steady flame was
developed, shocks with varying Mach numbers were generated 50 flame lengths from the
flame at time 49.9tf s to have the interaction.

5.2. Interaction of Ms =1.9 shock with flame


Figure 14a illustrates the detailed evolution of the interaction of the cellular flame and
a shock with an initial Mach number of 1.9. The first frame shows the incoming shock
and the flame interface before the interaction. Similar to the experimental results, the
interface was flattened (2nd frame) and pushed backwards (3rd frame to the last) to
the burned gas. In the 3rd frame, the incident shock wave traversed the interface and
formed a transmitted shock wave, a rarefaction wave and a series of transverse waves.
A small bump can be observed in the middle of the interface. As the funnel shaped
interface further reversed, the small bump punctured down and pushed the interface tip
Shock cellular flame interaction 17
16 12 1.35
η/Lfs
14 10 1.3
Su/SL
12 1.25
8
10
1.2
y/Lfs

Su/SL
η/Lfs
8 6
1.15
6 4
1.1
4
2 1.05
2

0 0 1
4 6 8 10 12 14 16 18 20 22 24 0 5 10 15 20 25 30 35 40 45 50
x/Lfs t/tfs

(a) (b)

Figure 12: Time evolution of: (a) maximum heat release contour at times t/tf s = 0, 10,
20, 30, 35, 40, 45 and 49.9 illustrating the initial setup and the cellular flame development,
(b) flame shape amplitude and burning velocity.

Figure 13: Comparison of the flame shape from the top-half middle cell marked in figure 4
(2) before the interaction (top) and the simulation (bottom), the red dashed line and the
black line denote the profile of the simulation, the blue dashed line shows the difference.

to propagated further, more transverse waves can be observed as the interface propagated
from the 4th to 5th frames. The first 5 frames show the same trend of flame flattened
and reversed back as the experiment in figure 4. As the flame funnel stretched further
in the 6th to 9th frames, it progressively developed into a very long and narrow neck
that is close to the bottom boundary. Whereas no sign of small structures along the
flame front can be observed as in the Schlieren images in figure 4. In the 10th frame, the
flame funnel burned out the fresh gas along its neck and left only the cellular root. The
remained flame front then gradually decayed to another cellular flame, as shown in the
11th to 14th frame.
To investigate the influence of the inert RM instability on the shock-flame interaction,
the reaction was artificially turned off right before the shock reached the flame, as shown
in the first frame in figure 14b. The first 5 frames show the same trend of the interface
flattened and reversed back as the reactive simulation and the experiment. In the 6th
frame, the concave structure formed by the initial small bump merged with the interface
tip and further evolved to a spike topped off with a mushroom cap, implying the classic
non-linear RM instability structure. Unlike the smooth flame front of the reactive case,
the development of the vortex at the tip of the mushroom cap can be observed in the
18 H. Yang and M. I. Radulescu
(a) (b)
t/tf = 0 t/tf = 0

t/tf = 0.02 t/tf = 0.02

t/tf = 0.067 t/tf = 0.067

t/tf = 0.108 t/tf = 0.108

t/tf = 0.128 t/tf = 0.128

t/tf = 0.357 t/tf = 0.357

t/tf = 0.619 t/tf = 0.619

y/Lfs

t/tf = 0.781 t/tf = 0.781

t/tf = 0.996 t/tf = 0.996

t/tf = 1.252 t/tf = 1.252

t/tf = 1.373 t/tf = 1.373

t/tf = 2.046 t/tf = 1.69

t/tf = 2.72 t/tf = 1.804

x/Lfs
t/tf = 3.39

ρ/ρ0

Figure 14: Density profiles illustrating the (a) flame and (b) inert interface evolution
subsequent to the interaction with the Ms =1.9 shock.

6th to the last frame. During the interface propagation, the growing vortex dissipated
the top of the interface. In the 10th to 13th frame, the growth of the interface become
less obvious.
Figure 15a illustrates the amplitude development of the interface from the reactive and
Shock cellular flame interaction 19
10 3.5
Experiment Experiment
9 Simulation (Inert) Simulation (Inert)
Mikaelian model
3 Simulation (Flame)
8 Simulation (Flame) Mikaelian model

7 Sim.(Inert) + flame model 2.5 MB model


Proposed model
6 MB model
2

η/η0
η/η0

5
4 1.5

3 1
2
0.5
1
0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 -0.05 0 0.05 0.1 0.15 0.2
t/tf t/tf

(a) (b)

Figure 15: (a) Comparison of the interface amplitude as a function of time for
the experiment, the simulation of the reactive and non-reactive interaction, and the
prediction of Mikaelian’s model, combined model of the inert simulation and the
flame model, and the proposed model, (b) zoomed-in view of interface amplitude for
the experiment, the simulation of the reactive and non-reactive interaction, and the
prediction of the Mikaelian model and the MB model within the time range of 0 to 0.2
tf subsequent to the interaction with the Mach 1.9 incident shock.

inert simulation of the RM instability. To track the interface evolution in the simulation,
we imposed a scalar Y in the flow field assuming Y = 0.0 and 1.0 to be the heavier
and lower density region and Y = 0.0 to 1.0 along the interface. The amplitudes were
then acquired by tracking the contour of the scalar value of Y = 0.2. For comparison,
the time evolution of the flame amplitude for the experiment of the interaction of the
same incident shock Mach number is also plotted here. By looking at figure 15a, we can
conclude that the flame amplitude went through four distinct stages after the reversion
of the interface. The first stage can be recognized as the inert evolution of the interface
according to the inert RM instability, before approximately 0.12tf . Evident from the
zoomed-in figure of the amplitude evolution until 0.15tf in figure 15b is that the reactive
and the inert simulations are nearly identical from 0 to 0.12tf . Moreover, the amplitude
evolution from the simulations is in excellent agreement with the experiment. In the
second stage, the volumetric expansion caused by the combustion along the flame front
gradually became important and resulted in a longer amplitude comparing to the inert
case up to 1 laminar flame time (tf ). In this stage, the amplitude growth rate gradually
decreased and became smaller than the first stage. Note that the small drop at 0.77tf
and 1.05tf are due to the separation of the mushroom cap from the main structure. The
third stage started from 1.07tf when the flame consumed the fresh gas inside the funnel
and gave rise to an abrupt collapse of amplitude. In the last stage, the remaining flame
spike went through a gradual decay of amplitude to reach the new steady state in the
post shock flow subjected to the Landau-Darrieus instability.
Figure 16 shows the evolution of the flame burning velocity and a short time average
value corresponding to the observation time in the experiments. The passage of the
incident shock quickly increased the burning velocity to approximately twice the initial
value, then it gradually increased with the growth of flame area in the inert evolution
stage. Within the experimental visualization time, the burning velocity increase is about
2.76 times the laminar flame speed. In the reactive evolution stage, the burning velocity
developed with a smaller rate and went through a change with an even lower increase rate
20 H. Yang and M. I. Radulescu

Simulation
10 Simulation average

6
Su/SL

0
0 0.5 1 1.5 2 2.5 3 3.5 4
t/tf

Figure 16: Time evolution of the flame burning velocity after the interaction of a Mach
1.9 incident shock.

at 0.5tf when it formed the long neck along the bottom line. It peaked at approximately
0.75 laminar flame time while the mushroom cap started to separate from the main spike,
then developed with mild decrease at about 8.5 times the laminar burning velocity until
the flame finally burned all the fresh gases along the funnel neck, demonstrated by the
sharp decrease in the burning velocity at about 1 laminar flame time. In the last stage,
the burning velocity gradually asymptoted to a value close to the flame velocity before
the interaction.
The average burning velocity in the simulations is found to be in agreement with the
experiment, within the uncertainties of the model used to infer Su described in Appendix
A. The experimental value was found 1.3 times larger, possibly suggesting 3D effects in
the experiments, but this difference remains within the uncertainty of the method.

5.3. Influence of the incident shock strength


The effect of shock strength is illustrated in figure 17 where the evolution of the flame
contours was obtained numerically as explained above for shock Mach numbers of 1.53,
1.75, 1.9 and 2.5. For comparison, we rearranged and superimposed the flame contour of
Y = 0.2 for cases with various incident shock strength at each time step such that they all
start from the same coordinate. It is evident that the overall flame deformation followed
the same trend of interface reversion, funnel elongation and area increase, mushroom-cap
separation and burn-out of the funnel neck, and new stage cellular flame development.
With an increase in shock Mach number, the flame front area and shape amplitude are
more pronounced.
We then examined the flame shape amplitudes and their growth rate for all the cases
considered to determine the effect of shock strength on the flame front deformation. As
shown in figure 18, the overall evolution were qualitatively similar for the various shock
Mach numbers investigated. Larger flame shape amplitudes were achieved with stronger
shocks. The flame amplitude went through the same trend of decrease during the traverse
of the incident shock, and increase due to the RM instability and the chemical reaction.
Hereafter, the amplitudes went through a drastic decrease due to the burn-out of the
fresh gas along the flame funnel neck, and then gradually decay to reach the new steady
Shock cellular flame interaction 21
t/tf = 0 t/tf = 0.619

t/tf = 0.067 t/tf = 0.781

t/tf = 0.128 t/tf = 1.373

t/tf = 0.357 t/tf = 2.047

Figure 17: Time evolution of the superimposed flame contour of Y = 0.2 for the
interaction with different incident shock strengths illustrated with the same size as of
figure 14. The red, green, blue and black lines refer to the cases of shock Mach number
of 1.53, 1.75, 1.9 and 2.5.

state for the case of incident Mach number of 1.53, 1.75 and 1.9, while the amplitude of
the flame that interacted with the Ms =2.5 shock gradually increased. To examine the
flame evolution with more detail, figure 19 show the enlarged view of the amplitudes and
growth rates evolution at early times before the collapse of the interface for the cases
considered. As shown in figure 19, the flame interfaces were compressed to the minimum
values faster for higher incident Mach numbers. The initial post-shock amplitude growth
toward the burnt gas also showed the same tendency with the increase of Mach number at
time before 0.075 tf . Also, as shown in figure 19a, the growth rate mildly fluctuated within
0.075 tf for small Mach numbers comparing to the larger Mach numbers. Surprisingly,
the growth rate of all cases considered decayed with the rate close to each other from
approximately 0.075tf until when the mushroom-cap started to disconnect from the main
spike. The sharp decrease of amplitude for example at 0.73tf for Ms =1.53 implying the
first separation of the flame tip. Then amplitude of the remained structure kept growing
until the flame burnt out the gas along the funnel neck as shown in figure 14 (10th to 11th
frame) at approximately 1tf for all the cases, denoted by the sharp decrease of amplitude
starting from the case of higher Mach numbers. The fact that the amplitude growth
is weakly influenced by the incident Mach number might suggest that the later flame
development is not solely controlled by an inert RM instability. At last, as demonstrated
in figure 20, the flame amplitudes all decayed and asymptoted to a value about half the
initial state while the rates asymptoted to approximately zero for the cases of Ms =1.53,
1.75 and 1.9. The only exception is the Ms =2.5 case, where the decay rate of the flame
amplitude gradually passed 0 and asymptoted to 12 m/s over the time of the calculation.
The saturation of the amplitudes growth is indicative of the saturation to a new cellular
structure controlled Landau-Darrieus instability, while the non-steadyness observed at
Ms =2.5 is attributed to the typical chaotic dynamics of cellular flames (Sivashinsky
1983), since for a stronger shock, the higher compression of the gas ahead of the flame
leads to thinner flames, allowing them to evolve in an essentially larger channel.
The differences in flame shape evolution correlate very well with the evolution of the
burning velocity shown in figure 21. The development of all the flame propagation speeds
and burning velocities are qualitatively consistent to the aforementioned case of Ms =1.9,
although the flame propagated with larger speed for the case with stronger incident Mach
numbers.
22 H. Yang and M. I. Radulescu
9
Ms=2.5
8 Ms=1.9
Ms=1.75
7 Ms=1.53

6
5
η/η0
4
3
2
1
0
0 1 2 3 4 5 6
t/tf

Figure 18: Comparison of the time evolution of amplitudes of the flame subsequent to
the interaction with the different incident shock Mach number of 1.53, 1.75, 1.9 and 2.5.

7 550
Ms=2.5 Ms=2.5
Ms=1.9 500 Ms=1.9
6 Ms=1.75
Ms=1.75
450 Ms=1.53
Ms=1.53
5 400

4 350
η'(m/s)

300
η/η0

3 250
2 200
150
1
100
0 50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
t/tf t/tf

(a) (b)

Figure 19: Zoomed-in view of (a) the time evolution of the flame amplitudes from 0 to
0.7tf and (b) their growth rate in the amplification stages for the interaction with the
different incident shock Mach number of 1.53, 1.75, 1.9 and 2.5.

6. Discussion
6.1. Four evolution stages of the flame deformation
The results presented in figures 14, 15, and 18 suggest there are four distinct evolution
stages of the flame amplitude after the passage of the shock and flame front inversion.
The overall flame evolution follows the trend of inert evolution stage dictated by the
non-linear Ricthmyer-Meshkov instability, the reactive growth stage where volumetric
expansion due to energy release enhances the growth of the inert RM instability, the flame
funnel collapse stage and the new cellular flame development stage. These are discussed
in more detail below and a novel model for the reactively assisted RM instability is
formulated.
Shock cellular flame interaction 23
20
10
0
-10
-20
η' (m/s) -30
-40
-50
Ms=2.5
-60 Ms=1.9
-70 Ms=1.75
Ms=1.53
-80
1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6
t/tf

Figure 20: Comparison of the time evolution of amplitudes growth of the flame in the
new cellular flame formation stage for the cases of different incident shock Mach numbers
of 1.53, 1.75, 1.9 and 2.5.

12
Ms=2.5
Ms=1.9
10 Ms=1.75
Ms=1.53

8
Su/SL

0
0 1 2 3 4 5 6
t/tf

Figure 21: Evolution of the the flame burning velocity following the interaction with the
different incident shock Mach number of 1.53, 1.75, 1.9 and 2.5.

6.1.1. Inert evolution stage


The first stage of the flame evolution after the flame is flattened by the incident shock
can be recognized as the inert evolution of the interface, at times significantly less than
the characteristic flame time, i.e., smaller than 0.12tf for the case with the incident Mach
number of 1.9, as shown in figures 15 and 19a. In this stage, the growth rate of the flame
interface is controlled by the inert RM instability.
When the interface amplitude is sufficiently small, the growth of the amplitude can
be predicted by the linear model (Richtmyer 1960; Meyer & Blewett 1972). For our case
of shock passing from a fresh gas (heavy fluid) to a burned gas (light fluid), (Meyer &
Blewett 1972) proposed the following expression (MB model) for the initial linear growth
24 H. Yang and M. I. Radulescu
rate:
dη η + + η0−
= 0 k[u]A+
t , (6.1)
dt 2
where η is the amplitude of the perturbation, η0− and η0+ are the amplitudes before and
immediately after the shock passage, and A+t the Atwood number after the interaction. As
the amplitude develops to be comparable to or larger than the wavelength, the growth
of the amplitude becomes nonlinear. Mikaelian (2003) suggested a model that joined
the linear and non-linear regime for predicting the amplitude growth rate for arbitrary
Atwood numbers. The Mikaelian model gives:

0 η00
ηM ik = , (6.2)
1 + 3η00 kt (1+A t)
(3+At )

where η00 = 21 (η0− + η0+ )[u]kA+


t is the amplitude growth rate in the linear stage (Meyer
& Blewett 1972).
In figures 15 and 23, we compared the experiment and numerical simulations results
with the MB and the Mikaelian models. As shown in figure 15a,the MB model shows
good agreement with the initial amplitude evolution for the inert interface, while the
interface evolution after the shock-inert interface interaction can be well approximate by
the Mikaelian model. Relatively good agreement can also be observed for the two models
in the inert evolution stage in figure 23, the discrepancy between the model and the
experimental and simulation results might caused by the finite thickness of the interface
rather than the discontinuity. It is obvious that for stronger incident shock Mach numbers,
the inert evolution stage lasts for longer times. Also, the Mikaelian model shows better
performance at later times for the case with the higher incident shock Mach number.
An interesting aspect of these results is that the shock-flame experiment provides
a simple and promising method to further study the RM instability. For observation
times an order of magnitude less than the flame consumption time, the interface remains
essentially inert. This method provides a unique opportunity to study sharp interfaces
devoid of separating membranes or non-desired mixing from injected gas curtains or
sliding valves.

6.1.2. Reactive growth stage


As illustrated in figure 15, when the time is comparable with the characteristic flame
time tf , the growth rate of the flame shape amplitude can no longer be captured by the
inert RM instability dynamics. The gases processed by the flame expands and occupies
a non-negligible volume affecting the motion of the flame shape.
A model for the growth rate of the flame accounting from this supplementary volume
expansion of the material crossing the flame, which lengthens the flame and increases its
surface area can be formulated as a linear correction to the inert non-linear growth rate
model:
dη dηrmi dηq
= + , (6.3)
dt dt dt

where we define dηdt ,
dηrmi
dt and dtq to be the total flame interface growth rate, the growth
rate caused by the inert RM instability and the heat release. Here, we expected the
correction to be independent of the inert model.
We assume a flame front with a simple structure of a plane area, as shown schematically
in figure 22. According to the mass conservation, the rate of volume increase of the burned
gas is equal to the mass consumption rate of the fresh gas during the development of the
Shock cellular flame interaction 25
δ

A(
H A( f t+∆
f t) t)

η Δη

Figure 22: Simplified geometrical model of flame front expansion.

flame:
d H
(ρb ηq δ) = ρu Af Su δ, (6.4)
q dt 2
where Af = ηq2 + H 2 is the flame area, H is the channel height, δ is the channel width.
Equation 6.4 thus can be re-written as:
d H q
(ρb ηq ) = ρu ηq2 + H 2 Su , (6.5)
dt 2
Equation 6.5 implicitly gives the growth rate of the flame amplitude. It is thus
integrated to yield
1 q
2 + H 2 )exp( 2ρu Su t) − H2 2ρu Su
ηq = (ηq0 + ηq0 q exp(− t). (6.6)
2 ρb H 2 2
2(ηq0 + ηq0 + H ) ρb H

The constant of integration ηq0 is the initial flame amplitude. Hence, the growth rate of
the flame length (and surface area) depends on the surface of the flame itself, yielding
an exponential feedback. The reactive correction to be applied in expression (6.3) is the

time derivative of (6.6), dtq .

The first test of the model proposed is to combine the result obtained for dtq with
the numerical result for the inert evolution shown in figure 15a. The result, also shown
in figure 15a, labeled ”Sim. (Inert) + flame model” recovers very well the reactive
simulation.
A closed form expression for the reactive RM instability can be written using (6.3), the

correction dtq and an algebraic model for the inert RM instability. Since the Mikaelian
model can well predict the non-reactive evolution of the inert interface, as shown in figure
15a and 23, the growth rate in the second stage can be modeled as:
dη 0 dηq
= ηM ik + . (6.7)
dt dt
The time evolution of the flame amplitude predicted by the proposed model was then
compared with the result from the simulation and the combined model of the inert RMI
interface evolution with the flame expansion model for Ms =1.9 shock interacts with the
cellular flame, excellent agreement can be observed in figure 15a.
We then verified the proposed model by the comparison of the amplitude and its
growth rate evolution with the simulation results of the interaction of the flame with
different incident shock strengths. As shown in figure 23, the overall flame evolution can
be very well predicted by the proposed model. Note that for stronger incident shocks,
the inert RM instability plays dominant roles for longer time following the shock-flame
interaction, i.e. the inert Mikaelian model is in good agreement with the simulation until
0.3tf for the case of Ms =2.5 incident shock flame interaction. While for weaker incident
26 H. Yang and M. I. Radulescu
shocks, the chemical reaction of the flame starts to play important roles earlier. Note
that the chemical reaction seems to start to play roles only 0.05tf after the passage of
the shock Ms =1.53, whereas the inert RM instability controlled the flame evolution for
more than 0.25tf after the Ms =2.5 shock. Thus, the inert RM instability plays dominant
roles following the shock-flame interaction. As the shock propagates far from the flame,
the baroclinic effect is weakened with a rate of O(1/t), the exponentially increase of the
flame interface gradually takes control in the total flame deformation.

6.1.3. Flame funnel collapse stage


The non-linear reactively enhanced growth of the flame terminates when the funnels
created burn out by the transverse reaction fronts. This occurs on a time scale associated
with the flame’s displacement velocity and the transverse length scale of the cellular
flame, ρρubSHu . In the simulation, the post-interaction ratio of ρρub is found to be within the
range of 6-8. With the non-dimensionalized H=16.5Su tf s , the burn-out time thus is 2.06
to 2.75tf s , which is 1.39 to 1.85tf . As can be observed in figure 18, the subsequent flame
relaxation to a new cellular structure occurs on this same time scale.

6.1.4. New cellular flame stage


Subsequent to the collapse of the flame interface, it saturated to a new stage of cellular
flame evolution. In this stage, the transmitted shock has propagated far from the flame,
the baroclinic effect, which decays at a rate of 1/t, plays minor roles in influencing the
flame deformation. Thus, similar to the formation of the cellular structure, the flame
evolution in this stage is only controlled by the intrinsic Landau-Darrieus instability. As
shown in figures 18, 21 and 24, the growth rate of the amplitude progressively asymptoted
to values close to zero except for the case of Ms = 2.5 incident shock, which shows the
gradual increase of positive growth rate at the end of this stage. This implies that the
flames evolved to a new cellular flame evolution stage.

6.2. Evolution of the burning velocity


For practical considerations, it is of interest to quantify the burning rate enhancement
of the flame due to the interaction with the shock waves, and elucidate the controlling
parameters. To a good approximation, the increase of the burning rate of the flame is
given by the increase in its surface area (Sivashinsky 1983), as shown in figure 12b for
example. For η  H, the flame surface area in 2D is proportional to η. When can thus
expect the characteristic rate of burning rate increase, i.e.,
d ln Su
RSu ≡ | | (6.8)
d(t/tf )
to be well approximated by the characteristic rate of flame shape amplitude increase,
i.e.,
d ln η
Rη ≡ | | (6.9)
d(t/tf )
Figure 24 shows the comparison of these two characteristic rates evaluated by taking
time derivatives of the numerical signals for flame shape amplitude and burning velocity.
The two rates are found in very good agreement over the times of interest, even for
conditions where the assumption η  H is not valid (at early and late times). We have
not attempted to further model the burning rate increase with higher precision given
more accurate flame shape functions parametrized by the evolution of η. This is left for
future study.
Shock cellular flame interaction 27
7 350
Simulation Ms = 1.53 Simulation Ms = 1.53
6 MB model Proposed model
Proposed model 300 Mikaelian model
Mikaelian model
5
250

η' (m/s)
4
200
η/η0

3
150
2

1 100

0 50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t/tf t/tf

(a) (b)
7 400
Simulation Ms = 1.75 Simulation Ms = 1.75
MB model Proposed model
6 Proposed model 350
Mikaelian model
Mikaelian model
5 300

η' (m/s)
4 250
η/η0

3 200

2 150

1 100

0 50
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t/tf t/tf

(c) (d)
7 400
Simulation Ms = 1.9 Simulation Ms = 1.9
MB model 350 Proposed model
6 Proposed model Mikaelian model
Mikaelian model
300
5
250
4
η' (m/s)
η/η0

200
3
150
2
100
1 50

0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
t/tf t/tf

(e) (f)
9 500
Simulation Ms = 2.5 Simulation Ms = 2.5
8 MB model 450 Proposed model
Proposed model Mikaelian model
7 Mikaelian model
400
350
6
300
η' (m/s)

5
η/η0

250
4
200
3
150
2 100
1 50
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
t/tf t/tf

(g) (h)

Figure 23: Comparison of the time evolution of amplitude and growth rate of the
simulation with proposed model for the interaction of flame with different incident shock
Mach number of 1.53, 1.75, 1.9 and 2.5.
28 H. Yang and M. I. Radulescu
100 100
Rη Rη
RSu RSu
10
10

1
1

RSu, Rη
RSu, Rη

0.1
0.1
0.01

0.01
0.001

0.0001 0.001
0 1 2 3 4 5 6 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
t/tf t/tf

(a) Ms =1.53 (b) Ms =1.75


100 100
Rη Rη
RSu RSu
10 10

1 1

RSu, Rη
RSu, Rη

0.1 0.1

0.01 0.01

0.001 0.001
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3
t/tf t/tf

(c) Ms =1.9 (d) Ms =2.5

Figure 24: Comparison of the time evolution of burning velocity and amplitude growth
rate in new cellular flame formation stage for the interaction of flame with different
incident shock Mach number of 1.53, 1.75, 1.9 and 2.5.

7. Conclusion
In this study, single shock wave colliding with a cellular flame in a shock tube was
studied experimentally, numerically, and theoretically. The experiments performed in
the shock tube at low pressure in stoichiometric hydrogen-air permitted to isolate and
monitor the flame deformation with precision after interaction. Following the passage
of the incident shock, the flame cusps were flattened and reversed backwards to the
burnt gas. The flame front amplitude was taken to be an indicator for the flame surface
evolution. The growth of the amplitude increased weakly with the strength of the incident
Mach number. The flame propagation speed for the interaction of different Mach numbers
considered with the cellular flame mildly decreased within the time of measurement.
The simulation extended the observation time for the flame subsequent to the interac-
tion of the experiments. Following the passage, the reversed flame interface development
was found to go through the following stages: inert evolution due to the RM instability
at times significantly less than the laminar flame time, nonlinear increment result from
the amplification of chemical energy release on growth rate of inert RM instability,
transverse burn-out of the neck of flame funnel and re-adjustment to a new cellular flame
on the laminar flame time scale. A proposed model can well approximate the evolution
of the flame geometry and burning rate increase for arbitrary shock strength and flame
Shock cellular flame interaction 29
properties in 2D. The model provides a simple way to estimate the flame burning rate
increase upon passage of a shock and duration of this event.
The authors thank NSERC for the financial support and Compute Canada for the
computation resources. H.Y. thanks China Scholarship Council (CSC) and NSFC grant
(51774068) for the financial support. H.Y. also wants to thank Andre Fecteau for the
help in calculating the 1D steady flame profile. The authors wish to thank Sam Falle for
the use of the MG code. The authors report no conflict of interest.

Appendix A. Extraction of the burning velocity from experiment


A simple self-similar one-dimensional gasdynamic model was formulated in order to
extract the effective burning velocity increase from the interaction of a shock with a
flame. The model uses the experimentally measured flame displacement speed to infer
the corresponding burning velocity.
The model assumes gasdynamic discontinuities for the incident and transmitted shocks,
the flame brush and a contact surface, as illustrated in figure 25. The incident shock
always transmits as a shock. If the flame surface were inert, since the acoustic impedance
ρc of the burned gases is lower than that of the unburned fresh gases owing to the large
density changes, the incident shock reflects as an expansion wave. For the reactive case, a
flame surface is introduced, such that the layer of gas 5 is the gas consumed by the flame
since the interaction. All waves are assumed to have a constant speed and all regions
separating the discontinuities are assumed uniform. As such, the model is effectively a
time-averaged approximation.
Our flow reconstruction uses the knowledge of the initial states, the initial flame
propagation speed and shape, the incident shock propagation speed and the displacement
velocity of the flame to infer all states of interest and the flame burning velocity. State
2 across the flame was computed from the adiabatic flame model using Cantera and
the thermo-chemical database of Li et al. (2004). The cellular flame burning velocity
was corrected from the free flame speed from Cantera by considering the ratio of the
flame area and channel cross-sectional area, whereas the pressure increase in state 2 was
evaluated by the method of characteristics. The flame propagation velocity before the
interaction was measured from the mean average moving distance per unit time after the
flame formed 3 cells. State 2 is then calculated by applying the mass and momentum
conservation balance. The flow from state 2 to state 3 and state 1 to state 6 across
the shock obey the usual shock jump equations. The expansion from state 4 to state
3 is assumed to be isentropic. For all states, a two-gamma approximation was used,
where the non-reacted and reacted gas are assumed to be polytropic gases with constant
thermodynamics properties (Chue et al. 1993). A system of equations is then built from
state 3 to state 6, in which the average post-shock flame propagation speed was matched
with the experimentally measured value in order to infer the other flow parameters such
as the flame burning velocity.
Over the time interval available in the experiments, the burning velocity was evaluated
to be 7.1±4.5m/s for the case of the Ms =1.9 shock-flame interaction. Note that a minor
difference of propagation speed can largely influence the 1D evaluation of the burning
velocity. The mean evaluated burning velocity is 3.58 times the laminar flame speed and
1.3 times the one from the simulation, which shows relatively good agreement with the
simulation result.
30 H. Yang and M. I. Radulescu
1

0.8
Expansion fan
0.6

0.4 F
4 5
3
0.2 CS 6

t (ms)
0 ISW TSW

-0.2 2 1

F
-0.4

-0.6

-0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25


x (m)

Figure 25: The space-time diagram evolution of interaction of figure 4 reconstructed from
the one-dimensional gas dynamic model.

Appendix B. Numerical resolution study


In order to test the effectiveness of the numerical resolution on the results presented
earlier, we have performed a resolution study by varying the minimum grid spacing. As
suggested in Sharpe & Falle (2006), roughly 30 grid points in the reaction zone of the
steady, planar flame is the requirement to properly resolve the cellular flame. To make
sure our grid points of 48 points per flame thickness (Level 5) is enough to resolve the
cellular flame evolution after the interaction with the incident shock. We have performed
a simulation with the minimum grid spacing of 96 points per flame thickness (Level 6)
for convergence comparison. As shown in figure 26a, the flame front is approximately
at the same location, and displays the same global characteristics of the front. A more
quantitative appraisal of numerical convergence was conducted on the flame burning
velocity. As illustrated in figure 26b, excellent convergence can be observed for the two
simulations conducted with 48 and 96 points per flame thickness, except for the drastic
decay in the flame funnel collapse stage. As the amplitude of the flame simulated with
48 grid points per flame thickness evolved also in good agreement with the experiment,
we thus conclude that the grid resolution we employed is enough to resolve the flame.

REFERENCES
Brouillette, Martin 2002 The Richtmyer-Meshkov instability. Annual Review of Fluid
Mechanics 34 (1), 445–468.
Chen, Xiao, Dong, Gang & Li, Baoming 2018 Numerical study of three-dimensional
developments of premixed flame induced by multiple shock waves. Acta Mechanica Sinica
34 (6), 1035–1047.
Chue, RS, Clarke, JF & Lee, JH 1993 Chapman-Jouguet deflagrations. Proceedings of the
Royal Society of London. Series A: Mathematical and Physical Sciences 441 (1913), 607–
623.
Ciccarelli, Gaby, Johansen, Craig T & Parravani, Michael 2010 The role of shock–
flame interactions on flame acceleration in an obstacle laden channel. Combustion and
Flame 157 (11), 2125–2136.
Dong, Gang, Fan, Baochun & Ye, Jingfang 2008 Numerical investigation of ethylene flame
bubble instability induced by shock waves. Shock Waves 17 (6), 409–419.
Shock cellular flame interaction 31
(a) t/tf = 0 (c) t/tf = 0.619
xmin=1/48Lf xmin=1/48Lf

y/Lfs y/Lfs
xmin=1/96Lf xmin=1/96Lf

x/Lfs x/Lfs
(b) t/tf = 0.101 (d) t/tf = 0.963
xmin=1/48Lf xmin=1/48Lf

y/Lfs y/Lfs
xmin=1/96Lf xmin=1/96Lf

x/Lfs x/Lfs

ρ/ρ0

(a)
9
8 Level 5
Level 6
7
6
5
Su/SL

4
3
2
1
0
0 0.5 1 1.5 2 2.5 3 3.5 4
t/tf

(b)

Figure 26: Comparison of (a) density profile and (b) time evolution of burning velocity
for different resolutions considered.

Egolfopoulos, FN & Law, Chung King 1990 Chain mechanisms in the overall reaction
orders in laminar flame propagation. Combustion and Flame 80 (1), 7–16.
Falle, SAEG, Giddings, JR, Morton, KW & Baines, MJ 1993 Numerical methods for fluid
dynamics 4.
Goncharov, VN 1999 Theory of the ablative Richtmyer-Meshkov instability. Physical review
letters 82 (10), 2091.
Jiang, Hua, Dong, Gang, Chen, Xiao & Li, Baoming 2016 A parameterization of the
Richtmyer–Meshkov instability on a premixed flame interface induced by the successive
passages of shock waves. Combustion and Flame 169, 229–241.
Jomaas, Grunde, Law, Chung King & Bechtold, JK 2007 On transition to cellularity in
expanding spherical flames. Journal of fluid mechanics 583, 1–26.
Ju, Yiguang, Shimano, Akishi & Inoue, Osamu 1998 Vorticity generation and flame
distortion induced by shock flame interaction. In Symposium (International) on
Combustion, , vol. 27, pp. 735–741. Elsevier.
Khokhlov, Alexei M, Oran, Elaine S, Chtchelkanova, Almadena Yu & Wheeler,
J Craig 1999 Interaction of a shock with a sinusoidally perturbed flame. Combustion and
flame 117 (1-2), 99–116.
32 H. Yang and M. I. Radulescu
Kilchyk, V, Nalim, R & Merkle, C 2013 Scaling interface length increase rates in Richtmyer–
Meshkov instabilities. Journal of fluids engineering 135 (3).
Li, Juan, Zhao, Zhenwei, Kazakov, Andrei & Dryer, Frederick L 2004 An updated
comprehensive kinetic model of hydrogen combustion. International journal of chemical
kinetics 36 (10), 566–575.
Lutoschkin, E, Rose, MG & Staudacher, S 2013 Pressure-gain combustion using shock–
flame interaction. Journal of Propulsion and Power 29 (5), 1181–1193.
Markstein, GH 1957 A shock-tube study of flame front-pressure wave interaction. In
Symposium (International) on Combustion, , vol. 6, pp. 387–398. Elsevier.
Massa, L & Jha, P 2012 Linear analysis of the Richtmyer-Meshkov instability in shock-flame
interactions. Physics of Fluids 24 (5), 056101.
Maxwell, Brian, Pekalski, Andrzej & Radulescu, Matei 2018 Modelling of the transition
of a turbulent shock-flame complex to detonation using the linear eddy model. Combustion
and Flame 192, 340–357.
Meyer, KA & Blewett, PJ 1972 Numerical investigation of the stability of a shock-
accelerated interface between two fluids. The Physics of Fluids 15 (5), 753–759.
Mikaelian, Karnig O 2003 Explicit expressions for the evolution of single-mode Rayleigh-
Taylor and Richtmyer-Meshkov instabilities at arbitrary atwood numbers. Physical Review
E 67 (2), 026319.
Oran, Elaine S & Gamezo, Vadim N 2007 Origins of the deflagration-to-detonation transition
in gas-phase combustion. Combustion and Flame 148 (1-2), 4–47.
Oron, Dan, Arazi, L, Kartoon, D, Rikanati, A, Alon, U & Shvarts, D 2001
Dimensionality dependence of the Rayleigh–Taylor and Richtmyer–Meshkov instability
late-time scaling laws. Physics of Plasmas 8 (6), 2883–2889.
Rakotoarison, Willstrong, Maxwell, Brian, Pekalski, Andrzej & Radulescu,
Matei I 2019 Mechanism of flame acceleration and detonation transition from the
interaction of a supersonic turbulent flame with an obstruction: Experiments in low
pressure propane–oxygen mixtures. Proceedings of the Combustion Institute 37 (3), 3713–
3721.
Richtmyer, Robert D 1960 Taylor instability in shock acceleration of compressible fluids.
Communications on pure and applied mathematics 13 (2), 297–319.
Rudinger, George 1958 Shock wave and flame interactions. In Combustion and Propulsion,
Third AGARD Colloquium, p. 153. Pergamon Press London.
Sadot, O, Erez, L, Alon, U, Oron, Dan, Levin, LA, Erez, G, Ben-Dor, G & Shvarts,
D 1998 Study of nonlinear evolution of single-mode and two-bubble interaction under
Richtmyer-Meshkov instability. Physical review letters 80 (8), 1654.
Scarinci, T, Lee, JH, Thomas, GO, Bambrey, R & Edwards, DH 1993 Amplification
of a pressure wave by its passage through a flame front. Progress in Astronautics and
Aeronautics 152, 3–3.
Sharpe, GJ & Falle, SAEG 2006 Nonlinear cellular instabilities of planar premixed flames:
numerical simulations of the reactive Navier–Stokes equations. Combustion Theory and
Modelling 10 (3), 483–514.
Sharpe, GJ, Falle, SAEG & Billingham, J 2008 Numerical solutions of a model for the
propagation of a surface-catalysed flame in a tube. IMA journal of applied mathematics
73 (1), 107–122.
Sivashinsky, G I 1983 Instabilities, pattern formation, and turbulence in
flames. Annual Review of Fluid Mechanics 15 (1), 179–199, arXiv:
https://doi.org/10.1146/annurev.fl.15.010183.001143.
Strehlow, Roger A 1984 Combustion fundamentals. McGraw-Hill College.
Thomas, Geraint, Bambrey, Richard & Brown, Caren 2001 Experimental observations
of flame acceleration and transition to detonation following shock-flame interaction.
Combustion Theory and Modelling 5 (4), 573–594.
Travnikov, O Yu, Liberman, MA & Bychkov, VV 1997 Stability of a planar flame front in
a compressible flow. Physics of Fluids 9 (12), 3935–3937.
Urtiew, PA & Oppenheim, AK 1968 Transverse flame-shock interactions in an explosive gas.
Proceedings of the Royal Society of London. Series A. Mathematical and Physical Sciences
304 (1478), 379–385.
Shock cellular flame interaction 33
Wei, Haiqiao, Zhao, Jianfu, Zhou, Lei, Gao, Dongzhi & Xu, Zailong 2017 Effects of the
equivalence ratio on turbulent flame–shock interactions in a confined space. Combustion
and Flame 186, 247–262.
Wijeyakulasuriya, Sameera D & Mitra, Saurav 2014 Analyzing three-dimensional
multiple shock-flame interactions in a constant-volume combustion channel. Combustion
Science and Technology 186 (12), 1907–1927.
Zhang, Qiang & Guo, Wenxuan 2016 Universality of finger growth in two-dimensional
Rayleigh–Taylor and Richtmyer–Meshkov instabilities with all density ratios. Journal of
Fluid Mechanics 786, 47–61.
Zhang, Qiang & Sohn, S-I 1999 Quantitative theory of Richtmyer-Meshkov instability in three
dimensions. Zeitschrift für angewandte Mathematik und Physik ZAMP 50 (1), 1–46.
Zhou, Ye 2017a Rayleigh–Taylor and Richtmyer–Meshkov instability induced flow, turbulence,
and mixing. ii. Physics Reports 723, 1–160.
Zhou, Ye 2017b Rayleigh-Taylor and Richtmyer-Meshkov instability induced flow, turbulence,
and mixing. i. Physics Reports 720, 1–136.
Zhu, Yuejin, Dong, Gang & Liu, Yixin 2013 Three-dimensional numerical simulations of
spherical flame evolutions in shock and reshock accelerated flows. Combustion science
and technology 185 (10), 1415–1440.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy