Combustion and Flame 138 (2004) 97–107

www.elsevier.com/locate/jnlabr/cnf

A CFD study of propane/air microflame stability

D.G. Norton and D.G. Vlachos ∗
Department of Chemical Engineering and Center for Catalytic Science and Technology (CCST), University of Delaware,
Newark, DE 19716-3110, USA
Received 7 October 2003; received in revised form 26 March 2004; accepted 5 April 2004
Available online 8 May 2004

Abstract
A two-dimensional elliptic computational fluid dynamics model of a microburner is solved to study the effects
of microburner wall conductivity, external heat losses, burner dimensions, and operating conditions on combus-
tion characteristics and the steady-state, self-sustained flame stability of propane/air mixtures. Large gradients are
observed, despite the small scales of the microburners. It is found that the wall thermal conductivity is vital in
determining the flame stability of the system, as the walls are responsible for the majority of the upstream heat
transfer as well as the external heat losses. Furthermore, there exists a range of flow velocities that allow stabilized
combustion in microburners. It is found that the microburner dimensions strongly affect thermal stability. Engi-
neering maps denoting flame stability are constructed and design recommendations are made. Finally, comparisons
with methane/air systems are made.
 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Propane; Microburners; Computational fluid dynamics; Flame stability; Extinction; Thermal management

1. Introduction the production of hydrogen for fuel cell applica-

Microburners may play a vital role in the portable
production of energy. Hydrocarbons have an energy
density significantly higher than that of the traditional
Li batteries (40 vs 0.5 MJ/kg) that are currently used
in laptops, cellular phones, and other portable elec-
tronics devices [1]. The small scales in microburners
result in lower combustion temperatures due to en-
hanced heat-transfer coefficients. Thus, we propose
that microburners could possibly reduce the gas-phase
production of NO [2]. Finally, microburners can also
serve as efficient sources of heat for endothermic re-
actions, such as steam reforming and ammonia de-
composition, in integrated microchemical systems for
* Corresponding author. Fax: (302)-831-1048.
E-mail address: vlachos@che.udel.edu
(D.G. Vlachos).
tions [3].
Unfortunately, the benefits arising at the mi-
croscale are overshadowed by major difficulties in
creating working microburners. In 1817, Davy per-
formed experiments showing the inability of flames
to propagate between gaps of submillimeter scale [4].
His work was followed by many other groups, whose
work confirmed that depending on geometry, com-
position, and flow rate, hydrocarbon/air flames are
typically quenched when confined within spaces with
critical dimensions < 1–2 mm [5–9]. The two pri-
mary mechanisms for quenching in these systems are
thermal and radical quenching [2,10,11]. Increased
heat-transfer coefficients are inherent to microscales,
because for a fixed Nusselt number, the heat-transfer
coefficient scales with the inverse of the length scale.
The high heat-transfer rates increase the heat lost
from the reaction, reducing the operating tempera-
tures and causing the combustion to extinguish. At
the same time, the increased mass transfer within

0010-2180/$ – see front matter  2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.combustflame.2004.04.004
98 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
the system, coupled with the high surface-area-to-
volume ratio, causes radical adsorption onto the walls,
followed by radical recombination. This dearth of
radicals quenches the homogeneous chemistry [12].
Another mechanism for loss of stability is blowout,
which occurs when the burner exit velocity exceeds
the flame burning velocity [7]. In this case, the reac-
tion front shifts downstream with increasing velocity
and eventually exits the burner. The competition be-
tween the shifting of the reaction zone and thermal
quenching has been observed in elliptic models of
microscale systems for methane [13] and mesoscale
systems for propane [14].
Recent experiments have demonstrated that it is
feasible to stabilize homogeneous methane/oxygen
flames between parallel plates with gaps smaller than
1 mm [15,16]. This is accomplished by modifying
the surface to make it chemically inactive, to elimi-
nate radical quenching, and insulating the burner, to
reduce thermal quenching. The chemical inactivation
process involves high-temperature annealing to heal
crystal defects and surface cleaning with deionized
water, hydrochloric acid, and hydrogen peroxide to
remove ionic and heavy metal contaminants.
In our recent computational fluid dynamics (CFD)
work [13], overall heat management was shown to
play a critical role in determining homogeneous flame
stability of methane/air mixtures in microburners. The
thermal conductivity of the wall allowed both the vi-
tal upstream heat transfer for preheating the feed to
the ignition temperature and detrimental heat losses
to the exterior. When the thermal conductivity is too
low, the upstream heat transfer through the walls is
choked, and the system blows out. At the other ex-
treme of high thermal conductivity, the wall and fluid
temperature profiles flatten, causing delocalized reac-
tion fronts and an increased external hot area for heat
losses. As a result extinction is observed.
The significant role of upstream heating in sta-
bilizing combustion in a tube was in fact realized
several years ago by Churchill [17], who developed
thermally stabilized burners (TSB). These TSB con-
sisted of ceramic cylindrical tubes > 7 mm in diam-
eter into which fuel and air were fed. Flame fronts
were stabilized within the tubes by thermal feedback
caused by conduction through the walls and radia-
tion between the walls to preheat the incoming feed.
To maximize the heat transfer between the fluid and
the walls, the flow rates were such that the flow was
turbulent upstream of the combustion. In the reac-
tion zone further downstream, the viscosity increased
due to the increased temperature, resulting in lami-
nar flow. These systems were found to be very sta-
ble to minor disturbances, but the range of flow rates
that allowed stabilized combustion was very limited.
These experimental findings are in general qualita-
tive agreement with our simulations [13]. However,
it is unclear how much of the TSB experimental re-
sults transfer to the microscale, where the flow is
severely confined. Another idea for stabilizing com-
bustion within mesoscale (∼ 1 mm) and macroscale
( 1 mm) tubes revolves around the work of Lloyd
and Weinberg on heat-recirculating burners, where
hot combustion products are used to preheat the in-
coming feed [18,19]. Ronney and co-workers have
extended this idea on mesoscale “Swiss roll” heat-
recirculating burners, with gap sizes on the order of
3 mm, for the homogeneous and heterogeneous com-
bustion of propane in air [1,20]. In recent work, Ron-
ney modeled a countercurrent heat-recirculating com-
bustor [21]. The system consisted of a cold reactant
inlet that fed into a well-stirred reactor (WSR). The
products from the WSR then flowed countercurren-
twise past the incoming feed to preheat it. The hot
and cold tubes were modeled by a one-dimensional
model, with convective heat-transfer coefficients to
model the heat transfer between them and the exterior.
The reaction was assumed to take place within the
WSR. It was shown that the axial conduction within
the wall had a major effect on the operating limits of
the combustor. Even a small wall thermal conductiv-
ity results in significantly higher required flow rates to
maintain stabilized combustion, whereas when the ax-
ial wall conduction is ignored, the minimum required
flow rate disappears. Ronney’s analysis emphasizes
the importance of heat transfer as our CFD modeling
did [13].
Despite previous work, there are a number of an-
swered questions regarding flame stability at the mi-
croscale. Examples include the roles of fuel and mi-
croburner dimensions in flame stability. In this work
we extend our previous modeling work to the steady-
state self-sustained microcombustion of propane/air.
Two-dimensional (2D), fully elliptic simulations are
performed that explicitly treat heat and mass trans-
fer in the fluid as well as heat transfer in the wall.
The effects of wall thermal conductivity, external heat
losses, operating conditions, and microburner dimen-
sions on flame stability and combustion characteris-
tics are discussed. In order to understand how dif-
ferent hydrocarbons behave, the differences between
propane/air and methane/air systems are also investi-
gated.
2. Model
The burner is modeled as two parallel plates that
are infinitely wide, 1 cm long, and distance L apart.
The plate thickness is Lw . For most simulations L =
600 µm and Lw = 200 µm (unless otherwise noted).
Premixed, nonpreheated propane/air mixtures are fed
 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
Fig. 1. Schematic of the computational domain (not to scale
for ease of visualization).
to the inlet of the microburner, and hot product gases
exit the microburner. Due to the aspect ratio, the sys-
tem is simplified to a 2D one and the plane of symme-
try between the two plates allows simulations of only
half of the microburner. A schematic of the system is
shown in Fig. 1.
The steady-state 2D continuity, momentum, en-
ergy, and species equations in the fluid phase and the
steady-state 2D energy equation in the solid phase are
discretized using a finite-volume method. Fluent 6.0
is used to perform these calculations [22]. The flow
is laminar. Previous studies of methane/air mixtures
utilizing surface and gas-phase radiation have shown
that the ignition distances for combustion reactions in
microburners were only slightly increased by radia-
tion [13]. The aspect ratio of the system is so high that
any surface-to-surface radiation is most likely emit-
ted and absorbed at nearly the same axial location.
Therefore, radiation is omitted from the simulations
performed in this work to focus on the effect of diffu-
sive and convective heat transport on flame stability.
The boundary conditions used in this model are
as follows. At the inlet a fixed flat velocity profile
is assumed. For the species and energy equations,
Danckwerts boundary conditions are employed; i.e.,
the convective portions of the equations are fixed, and
the diffusive portions are calculated implicitly. At the
interface between the fluid and the solid, no slip and
no normal species diffusive flux boundary conditions
are applied. The heat flux at this interface is calcu-
lated using Fourier’s law and continuity in tempera-
ture and heat flux is ensured. A symmetry boundary
condition is applied at the centerline between the two
plates. At the exit, the pressure is specified and the
remaining variables are calculated assuming far-field
conditions, i.e., zero diffusive flux of species or en-
ergy normal to the exit. In the bulk of the wall the
2D energy equation is solved. The exterior/top sur-
face of the wall is assumed to obey Newton’s law of
cooling, q  = h(Tw − Ta ), where q  is the heat flux,
h is the exterior convective heat-transfer coefficient,
Tw is the temperature at the exterior surface (an un-
known of the problem), and Ta is the ambient temper-
ature, which is assumed to be 300 K. It is important
to note that all the 2D internal heat transfer within
99
the fluid and the solid are calculated explicitly with
the 2D elliptic models without any further simplifica-
tions. The exterior convective heat-transfer coefficient
is only used for the calculation of the heat flux of
the exterior wall edge boundary condition. This heat-
transfer coefficient lumps the details of heat loss from
the microburner and of the process that utilizes the
heat generated by the burner. The left and right wall
edges are taken to be insulated (zero flux boundary
condition).
Nonuniform node spacing is employed in this
work, with more nodes in the reaction zone. The num-
ber of nodes varies depending on dimensions, but
the simplest one consists of 120 axial nodes by 60
transverse nodes, totaling approximately 7200 nodes.
Meshes in excess of 20,000 nodes are utilized for
the largest dimensions. Typical fluid node spacing is
50 µm in the axial direction and 6 µm in the trans-
verse direction. Typical wall node spacing is 50 µm in
the axial direction and 20 µm in the transverse direc-
tion, where the temperature does not vary (only a few
nodes are placed in the transverse direction within the
wall). However, as the mesh is nonuniform, these are
just representative values.
The fluid viscosity, specific heat, and thermal con-
ductivity are calculated by a mass-fraction-weighted
average of species properties. The species specific
heat is calculated using a piecewise polynomial fit of
temperature [22]. The fluid density is calculated using
the ideal gas law.
It has been shown that for typical materials of
construction, radical quenching is important in de-
termining flame stability in microburners. In partic-
ular, simulations that were performed with complex
gaseous chemistry have shown that radical sticking
coefficients that are larger than 0.001 severely de-
crease the flame stability of microburners [11]. Recent
experimental work has produced a nearly “quench-
less” wall material that is resistant to radical quench-
ing [15,16]. Thus, our focus here is on understanding
the overall heat-transfer characteristics in microburn-
ers and developing guidelines for appropriate thermal
management that can result in more robust flame be-
havior. One-step chemistries, determined from flame
speeds, are a useful tool for describing flame dynam-
ics and flame responses to external perturbations [23].
In this work a reduced one-step propane combustion
chemistry is used that assumes the irreversible com-
bustion of propane:
C3 H8 + 5O2 → 3CO2 + 4H2 O. (1)
Consequently five species, namely C3 H8 , O2 , N2 ,
CO2 , and H2 O, are modeled with corresponding con-
servation equations. Specifically, the mechanism used
is the one proposed by Westbrook and Dryer [24],
100 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107


rC3 H8 kgmol/(m3 s)
 
1.256 × 108 J/kgmol
= 4.836 × 109 · exp −
RT
· [CC3 H8 ]0.1 · [CO2 ]1.65 , (2)

where the concentrations are in units of kgmol/m3 .

We should remark that the one-step approximation
fails to describe several aspects, such as the possibility
for formation of partial oxidation products CO and H2
and the resulting superadiabatic flame temperature, or
the ignition temperature when ignition is kinetically
controlled [25], etc. As a result of this approxima-
tion, the flame location may be somewhat inaccurate.
Therefore, the results presented here should be used
as a guide to understand trends rather than an accu-
rate prediction.
The conservation equations were solved implic-
itly with a 2D steady-state segregated solver using an
under-relaxation method. The segregated solver first
solves the momentum equation, then the continuity
equation, and then updates the pressure and mass flow
rate. The conservation equations are then checked
for convergence. Convergence is determined from the
residuals of the conservation equations as well as the
difference (the L2 norm) between subsequent itera-
tions of the solution. The pressure was discretized
using a “Standard” method. The pressure–velocity
coupling was discretized using the “Simple” method.
The momentum, species, and energy equations were
discretized using a first-order upwind approximation.
Details about these schemes can be found in [22].
The simulations were performed on a Beowulf
cluster consisting of 58 Pentium IV processors and
58 GB of RAM. When parallel processing was used,
the message passing interface (MPI) was used to
transmit information between nodes. In order to
achieve convergence as well as compute extinction
points, natural parameter continuation was imple-
mented. The calculation time of each simulation var-
ied between 30 min and several hours, depending on
the difficulty of the problem and the initial guess.
3. Microflame characteristics
Fig. 2 shows contour plots of the temperature, re-
action rate, and propane conversion for a typical set
of operating parameters. The entire microburner is
shown. The flame stabilizes in the center between the
two plates. The reaction starts at the wall and trav-
els towards the center as the flow goes downstream.
Combustion occurs very rapidly, consuming most of
the propane in a very small region. Complete con-
version is achieved, and a significant temperature rise
is observed due to the exothermicity of the reaction.
The narrow flame front observed in the simulations
Fig. 2. Contours of (a) temperature [K], (b) reaction rate
[kgmol/(m3 s)], and (c) conversion for L = 600 µm, Lw =
200 µm, kw = 3 (W/m)/K, h = 10 (W/m2 )/K, Vinlet =
0.5 m/s, and a stoichiometric feed (not to scale, and reflec-
tion of symmetry is used for easier visualization).
is consistent with the experimental observations by
Churchill of a large-diameter TSB [17].
Despite the small scale, there are significant tem-
perature and species gradients within the fluid near
the reaction zone. These gradients necessitate the use
of an elliptic model, as axial diffusion of species and
energy cannot be neglected. Within the walls there
exist axial temperature gradients but near-zero trans-
verse temperature gradients (only a handful of nodes
are used within the wall in the transverse direction).
Thus, when the wall temperature is displayed in the
graphs below, only the exterior wall temperature is
shown.
In order to understand how to design microburners
with enhanced stability and robustness, it is necessary
to understand the extinction and blowout processes.
Fig. 3 shows reaction-rate profiles on the centerline
for three different cases, a stable microflame, one near
blowout, and one near extinction. Qualitatively, as a
microburner approaches extinction the reaction zone
shifts downstream slightly and is broadened, whereas
the maximum reaction rate decreases. The reaction
is quenched without leaving the microburner. On the
other hand, when blowout occurs, the reaction zone
shifts significantly downstream. In both cases, loss of
flame stability is caused by a lack of upstream heat
transfer to the incoming reactants. The primary dif-
ference between the two behaviors is that in blowout
more heat leaves in the form of a hot exit gas, whereas
in extinction, the excess heat loss occurs through the
walls to the surroundings. This distinction is not al-
ways sharp. We should note that once the reaction
zone shifts approximately past half the length of the
burner, the far-field boundary conditions at the exit
 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
Fig. 3. Reaction rate vs axial displacement for three typi-
cal cases, a stabilized microflame, one near blowout, and
one near extinction from thermal losses. Thermal quenching
shifts the reaction downstream slightly, broadens the reac-
tion zone a bit, and reduces the maximum reaction rate,
whereas blowout shifts the reaction zone downstream sig-
nificantly without decreasing the reaction rate.
may no longer describe the system properly. As a
result, the blowout critical conditions are less accu-
rate. To overcome the accuracy problem for a fixed
microburner length, one needs to experimentally mea-
sure the exit conditions and impose them as boundary
conditions.
4. Role of wall thermal conductivity and external
heat loss in flame stability
The location of the flame shifts as a function of op-
erating parameters. In [13] we introduced the flame
location as a convenient criterion for the stability or
robustness of a microburner, defined as the axial po-
sition with the highest reaction rate. Fig. 4a shows the
flame location as a function of wall thermal conduc-
tivity for different external heat-transfer coefficients.
As the wall thermal conductivity decreases to low
values, the flame location shifts downstream for all
external heat-transfer coefficients. For high wall ther-
mal conductivity and low external-heat-loss coeffi-
cients, increasing wall thermal conductivity to high
values has a minor effect on the flame location. On
the other hand, for high external-heat-loss coefficients
in systems with 600-µm gaps, increasing wall ther-
mal conductivity shifts the reaction downstream. This
nonlinear behavior is caused by the interaction be-
tween two competing modes of heat transfer, namely
upstream heat transfer through the walls to preheat
the feed, and transverse heat transfer resulting in heat
loss to the surroundings. The former is critical for
ignition and flame stabilization in microchannels, as
it allows preheating of the feed without the need for
an external preheater. If the upstream heat transfer is
insufficient to increase the fluid temperature to the ig-
101
Fig. 4. (a) Flame location vs wall thermal conductivity for
different external heat-transfer coefficients. Low wall ther-
mal conductivities cause the flame to shift downstream. In-
creasing wall thermal conductivity has little effect on flame
location unless there are significant external heat losses.
The parameters used are Vinlet = 0.5 m/s, L = 600 µm,
Lw = 200 µm, and a stoichiometric feed. (b) Flame loca-
tion vs wall thermal conductivity for different microburner
dimensions with Vinlet = 0.5 m/s, h = 35 (W/m2 )/K, and
a stoichiometric feed.
nition temperature, a flame is not stabilized within
the microburner [11]. Since the conductivity of the
walls is orders of magnitude higher than that of the
fluid, heat conduction through the walls is the pri-
mary mechanism of upstream heat transfer. When this
upstream heat transfer is limited by low wall ther-
mal conductivity, it takes a greater distance to achieve
the preheating, resulting in the reaction zone shifting
downstream (left part of the curves in Fig. 4a). This
makes the flame less stable. For a given wall thermal
conductivity, increasing the external heat loss coeffi-
cient shifts the reaction zone downstream as more of
the heat generated is lost to the surroundings.
The wall thermal conductivity alone does not de-
termine the relative upstream heat transfer in the sys-
tem. The wall thickness and the gap distance also
play an important role. Fig. 4b shows the effect of
conductivity on the flame location for different gap
distances and wall thickness. The flame location for a
1200-µm gap distance, which borders on mesoscale,
shows behavior qualitatively different from that of
102 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
the 600-µm gap distance. As the gap distance in-
creases, the time scales for heat transfer from the
reaction zone to the walls and from the hot walls to
the inlet reactants increases because of the increased
length scale. As a result of the latter, the flame lo-
cation occurs further downstream and more conduc-
tive materials are needed for stable operation. As a
result of the former, the system is more robust to ex-
terior heat losses. In particular, for highly conductive
materials and large external heat-transfer coefficients
(e.g., 35 (W/m2 )/K), the flame location does not shift
downstream with increasing wall thermal conductiv-
ity (see flat region for the 1200-µm case); i.e., the
larger gap makes the burner very robust with respect
to heat losses.
Aside from the gap distance, the wall thickness
is another key factor in microburner design. Fig. 4b
shows that with 400-µm-thick walls, the minimum
wall thermal conductivity allowable is approximately
half of the minimum wall thermal conductivity allow-
able with 200-µm-thick walls, while the burner is very
robust even for highly conductive materials. When
the wall thickness is doubled from 200 to 400 µm,
the amount of heat transferred upstream for a given
wall thermal conductivity is roughly doubled. Over-
all, thicker walls add mechanical and thermal stability
(at the expense, of course, of increased weight).
The flame locations with respect to the entrance
calculated in these simulations are significantly short-
er than those observed in the TSB of Churchill, i.e.,
O(1 mm) here vs O(150 mm) in [17]. This can be
attributed to the substantially faster thermal feedback
loop of microscale systems. The primary difference
likely stems from the change in the transverse time
scales for energy diffusion. If so, the ratio of time
scales would be (10 mm)2 /(0.6 mm)2 ∼ 280, which
is comparable to the aforementioned ratio of flame lo-
cations.
Aside from the flame location, the material ther-
mal conductivity affects the temperature profile within
the wall and the possibility of hot spots. Fig. 5 shows
the temperature profiles for the outer edge of the wall
for different material thermal conductivities. For low-
wall-thermal-conductivity materials, significant axial
temperature gradients are observed. Hotspot temper-
atures in excess of 2000 K can occur, an undesir-
able situation, as it exceeds the maximum operating
temperatures of most materials of construction. Ex-
ceedingly high wall temperatures are characteristic
of both micro- and macroscale thermally stabilized
burners [17,26]. As the wall thermal conductivity in-
creases, the wall temperature profiles become more
uniform and the wall hot spot is eliminated. Despite
the apparent advantages of a higher wall thermal con-
ductivity for material stability, most materials that of-
fer high conductance are metals, and therefore would
Fig. 5. Wall outer edge temperature profiles for different wall
thermal conductivities. Low wall thermal conductivities re-
sult in large axial wall-temperature gradients and high max-
imum temperatures. High wall thermal conductivity leads
to uniform temperature profiles without hotspots. The pa-
rameters are L = 600 µm, Lw = 200 µm, Vinlet = 0.5 m/s,
a stoichiometric feed, and h = 10 (W/m2 )/K.
not be inert to radical quenching. A more reasonable
solution would be thicker walls of a more inert mate-
rial that may have a lower thermal conductivity.
Parametric continuation is used to move from
one stationary solution to another. When the solu-
tion reaches a turning point or blowout occurs, this is
denoted as a critical point. Knowledge of critical para-
meter values of the external heat transfer coefficient,
wall thermal conductivity, feed composition, and flow
velocity gives a better understanding of the important
factors controlling flame stability. These critical val-
ues are useful as guides, but actual values will vary
depending on the system (e.g., dimensions) of inter-
est.
Fig. 6 shows the critical external heat loss co-
efficient as a function of wall thermal conductivity.
These bell-shaped envelopes separate the region of
self-sustained combustion below the curve from the
region above the curve where combustion cannot be
self-sustained. The conductivity of several materials
is also indicated by arrows. There exists a critical
wall thermal conductivity for propane/air mixtures,
at ∼ 0.1 (W/m)/K, below which combustion cannot
be self-sustained, even with insulating walls. When
the wall thermal conductivity increases from low val-
ues, the allowable-heat-loss coefficient first increases
quickly, and then decreases and levels off in the range
of metals or high-thermal-conductivity ceramics such
as SiC. The allowable-heat-loss coefficient reaches a
maximum for insulating ceramics such as silica and
alumina. The behavior seen for low-conductivity ma-
terials is at first counterintuitive. Highly insulating
materials are poor for flame stability due to the lack
of a continuous ignition source, needed to preheat the
cold incoming gases.
 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
Fig. 6. Critical external heat loss coefficient vs wall ther-
mal conductivity. Typical ceramics allow maximum external
heat loss coefficients. Materials with lower wall thermal con-
ductivities limit the upstream heat transfer. Materials with
higher wall thermal conductivities result in enhanced heat
transfer to the surroundings. Propane allows self-sustained
combustion for higher external heat loss coefficients and
more insulating materials than methane. The rest of the pa-
rameters are the same as in Fig. 4.
5. Effect of fuel on flame stability
Propane’s mechanism for the loss of stabilized
combustion is qualitatively similar to that of methane,
discussed in previous work [13]. For low wall thermal
conductivities the primary mode of burner instability
is blowout, whereas for high wall thermal conductiv-
ities it is extinction. However, propane microflames
are more robust than methane microflames, as shown
in Fig. 6. Note that the methane map is a subset of the
propane map. Lower wall thermal conductivities and
higher exterior heat-loss coefficients are possible.
In order to better understand the differences be-
tween propane and methane microflames, a theoret-
ical fuel, denoted as “pseudo-propane,” was defined.
This is simply a sensitivity analysis or numerical ex-
periment aiming at delineating the differences be-
tween fuels. Pseudo-propane has all of the properties
of propane except for a parameter that is changed
to a value that is identical to that for methane. The
methane/air and propane/air mixture properties, such
as thermal conductivity, specific heats, and viscosi-
ties, are similar, as the primary component is nitrogen
in both cases. Therefore, the properties of primary in-
terest are the heats of reaction and the reaction-rate
constants.
Fig. 7 shows the centerline temperature profile for
propane, methane, and two types of pseudo-propane.
Note that due to the difference in molecular weights
and densities of various fuel/air mixtures, the mass-
flow rates of different fuel/air mixtures are different
when the residence time is kept constant, as happens
in our simulations. Therefore, instead of simply re-
103
Fig. 7. Sensitivity analysis of the primary differences be-
tween propane and methane microflames. The reaction-rate
constant has a larger effect on the solution than the heat of
reaction. The parameters are L = 600 µm, Lw = 200 µm,
Vinlet = 0.5 m/s, kw = 1 (W/m)/K, h = 9 (W/m2 )/K, and
a stoichiometric feed.
placing one heat of reaction with the other, we have
also accounted for the difference in densities. The
way we accomplish this in the first numerical exper-
iment is by changing the heat of reaction of pseudo-
propane so that the power generated upon complete
combustion of pseudo-propane in the microburner
matches that of methane. For this case, the maxi-
mum temperature decreases, and the reaction location
shifts downstream. However, these changes are small
in comparison to the difference between propane and
methane microflames.
Next, the reaction-rate constant parameters of
pseudo-propane are changed to those of methane. For
this case, the maximum temperature increases, and
the flame location shifts significantly downstream.
This solution is closer to the methane solution. The
lower apparent activation energy of propane com-
bustion (∼ 126 kJ/mol for propane compared to
∼ 203 kJ/mol for methane [24]) causes easier ig-
nition and upstream flame stabilization. From this
analysis we conclude that the reaction-rate constant
parameters have the largest effect on flame location
and stability between different fuels. Higher hydro-
carbons, such as octane, generally have lower ignition
temperatures than methane. It is expected that they
would also exhibit increased stability compared to
methane.
6. Role of inlet velocity in flame stability and
fuel-lean operation limit
The inlet velocity plays a key role in determining
the location of the flame in the burner [13]. Fig. 8a
shows the flame location as a function of inlet veloc-
ity for several wall thermal conductivities. For high
104 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
Fig. 8. Flame location vs inlet velocity. An optimum flow
rate exists for flame stability. Higher wall thermal conduc-
tivities allow higher flow rates. The laminar flame speed for
a stoichiometric mixture of propane/air initially at 25 ◦ C is
denoted as Vlam . The results for different wall thermal con-
ductivities are shown in (a). The parameters are L = 600 µm,
Lw = 200 µm, h = 10 (W/m2 )/K, and a stoichiomet-
ric feed. The results for different burner dimensions are
shown in (b). The parameters are h = 10 (W/m2 )/K, kw =
7.5 (W/m)/K, and a stoichiometric feed.
inlet velocities the location of the flame shifts down-
stream with increasing flow rate due to the decrease
in the convective timescale (shorter residence times).
For low inlet velocities, a sharp shift of the reaction
zone downstream occurs with decreasing flow rate.
This is due to the decrease in the heat generation rate.
The external heat loss rate does not decrease as fast
as the heat generation rate, resulting in a reduced up-
stream heat-transfer rate. As a result of competition
between increased volumetric heat released and de-
creased residence time with increasing flow rate, there
is a minimum in the flame location between 0.3 and
0.5 m/s, depending on the wall thermal conductiv-
ity. This minimum is near the unconfined flame speed
for the same composition, experimentally determined
by Dugger and others to be ∼ 0.4 m/s [27,28]. The
minimum shifts slightly toward higher flow rates for
higher wall thermal conductivity since the latter al-
lows greater upstream heat transfer to compete with
the faster convective flow.
As shown above, when conductivity is the primary
variable, the gap distance and wall thickness play a vi-
tal role in the stabilization of the flame also when the
Fig. 9. Critical velocity vs wall thermal conductivity. The
lower curve ( –"– ) represents stability loss due to insufficient
heat generation. The upper curve ( –2– ) represents blowout.
The shaded region allows stabilized combustion. The exper-
imentally determined laminar flame speed [27] is plotted as
a dashed line. Higher wall thermal conductivity allows faster
flows. The parameters are the same as in Fig. 8.
flow velocity changes (see Fig. 8b). When the gap dis-
tance is doubled from 600 to 1200 µm the blowout
velocity decreases from ∼ 1.7 to ∼ 0.8 m/s. This
decrease in stability with respect to flow is due to
the increased timescales for energy diffusion between
the gas and the walls, resulting in the relative slow-
ing of the upstream preheating process, which shifts
the flame location downstream. In contrast, increas-
ing the wall thickness from 200 to 400 µm increases
the blowout velocity from ∼ 1.7 to ∼ 2.5 m/s while
leaving unaffected the flame stability for slow flows.
This increased flame stability is due to the increased
area for heat flux, doubling in our example the up-
stream preheating rate. When flow velocities greater
than the unconfined flame speed are required, the gap
distance must be small, and the wall thermal con-
ductivity and thickness must be sufficiently high to
provide adequate thermal feedback to preheat the in-
coming reactants.
Fig. 9 shows the critical velocity envelope vs the
wall thermal conductivity for a fixed external heat
loss coefficient. The upper curve represents the high-
velocity limit, resulting in blowout due to decreased
convective timescales. The lower curve represents the
low-velocity limit, resulting in flame stability loss
due to reduced heat generation. Between these curves
stabilized combustion is allowed, whereas outside
the envelope, self-sustained combustion is impossi-
ble. Smaller wall thermal conductivities allow stabi-
lized combustion for lower flow rates. Lower flow
rates require less upstream heating and more insu-
lation against exterior heat losses. At the other ex-
treme, higher wall thermal conductivities result in
maximum allowable flow rates (upper curve), but the
 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
increased heat losses prohibit low flow rates (lower
curve). This relationship is important when designing
devices. When a low-power device is desired, more
insulating materials should be preferred. On the other
hand, when a high-power device is desired, more con-
ductive materials should be chosen. Our focus here
has been on thermal stability. However, other material
properties, such as allowable operating temperatures,
radical sticking, and mechanical strength, along with
the microburner efficiency (complete conversion is
found here for stoichiometric mixtures using the one-
step chemistry) should be considered when choosing
a material for construction and designing microburner
dimensions.
The ability of a burner to operate under lean con-
ditions is beneficial, as it may reduce unwanted prod-
ucts such as coke, carbon monoxide, and nitric oxide.
It also reduces the operating temperature, which in
turn can increase burner lifetime. Fig. 10 shows the
lean equivalence ratio operation limit for the burner
as a function of Reynolds (Re) number calculated
based on the inlet conditions and the gap distance
half-height. There are two regimes of low (< 12) and
high (> 12) Re. For low Re, stability is lost with de-
creasing Re because of the diminished heat generation
rate. On the other hand, for high Re, blowout happens
because of decreased convective time scales. At the
transition between these two regimes, there appears
to be a deep minimum in the fuel-lean operation limit
and possibly a turning point over a narrow regime
of Re. Arc-length continuation is however needed to
fully characterize this situation.
Experimental work by Ahn et al. using a Swiss roll
burner with gap width 3.5 mm showed an optimum
Re at a relatively shallow minimum equivalence ra-
tio [20]. However, this optimum Re is ∼ 1000, two
Fig. 10. Minimum allowable equivalence ratio vs Reynolds
number. The Reynolds number was calculated using the ve-
locity, density, and kinematic viscosity at the inlet and the
gap half-width. The parameters are L = 600 µm, Lw =
200 µm, h = 10 (W/m2 )/K, and kw = 3 (W/m)/K.
105
orders of magnitude higher than the one calculated in
this work. Furthermore, the experimental minimum
equivalence ratio is approximately 0.2, compared to
the ∼ 0.56 found in this work. The differences be-
tween the experimental and our computational results
may lie in the enhanced preheating and insulation that
are achieved with the Swiss roll design and the gap
distance, which is considerably larger in the experi-
ments. More work is needed to fully understand these
differences.
7. Nusselt number analysis
To better understand how these microscale sys-
tems relate to their better-understood macroscale
counterparts, a Nusselt number (Nu) analysis was
performed. The Nu or dimensionless heat transfer co-
efficient is calculated as
hL (kf ∂Tf
dy )|wall L
Nu = = (3)
kf,cm (Tw − Tf,cm )kf,cm
and is evaluated at a given axial displacement. Here
Tf is the fluid temperature, Tf,cm is the cup mixing
fluid temperature, Tw is the wall temperature at the
wall–fluid interface, and kf,cm is the fluid cup mixing
thermal conductivity.
Fig. 11 shows Nu versus axial displacement for
two different cases, one with a flame stabilized near
the entrance and another at a higher velocity near
blowout. Nu exhibits strongly nonmonotonic behavior
with an oscillation at the reaction zone that finger-
prints the heat source, namely walls upstream trans-
ferring heat to the cold incoming gases and combus-
tion chemistry downstream of the entrance heating the
walls. In both examples, Nu approaches ∼ 4, which is
Fig. 11. Nusselt number vs axial displacement (see text) for
two inlet flow velocities indicated. The Nusselt number is a
strongly nonmonotonic function of position. The parameters
are L = 600 µm, Lw = 200 µm, h = 10 (W/m2 )/K, kw =
1 (W/m)/K, and a stoichiometric feed.
106 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
between the constant temperature and constant flux
values for circular tubes.
Groppi and Tronconi used 3D parabolic energy/
species balances to study methane catalytic combus-
tion [29]. They found very high Nu and Sh near the
entrance that eventually reach the fully developed
flow values predicted by Shah [30]. Their entrance
behavior is reminiscent of ours, but the downstream
behavior differs substantially between the catalytic
combustion studied previously and the homogeneous
combustion studied here. Comparison of our results
with the solutions of the Graetz–Nusselt problem for
two parallel plates for the special cases of constant
wall temperature and constant wall heat flux [31]
show that the solutions of the Graetz–Nusselt prob-
lem overestimate Nu. Note that existing correlations
do not take into account fluid chemistry and ignore
axial energy diffusion.
8. Conclusions
The characteristics of premixed propane/air mi-
crocombustion and stability envelopes were studied.
We have found that propane/air flames can be sta-
bilized in narrow channels but very careful design
is necessary. The wall material thermal conductivity
plays a competing role in flame stability. Walls trans-
fer heat upstream for ignition of the cold incoming
gases but at the same time are responsible for heat
losses. Consequently, there is an optimum wall ther-
mal conductivity in terms of flame stability, which ap-
pears to be that of common ceramics such as alumina
and silica. Despite the small scales of these systems,
large transverse gradients in temperature and species
mass fractions exist in the fluid and large axial gradi-
ents in temperature may exist in the walls. Regarding
material lifetimes, higher wall thermal conductivities
reduce the wall temperature gradients and hotspots
and should be preferred. It was also shown that the
burner size plays a significant role. Thicker walls en-
able more upstream heat propagation and faster flows
before blowout occurs and allow less conductive ma-
terials to be used. On the other hand, increasing the
gap distance from the micro- to the mesoscale offers
the advantage of higher stability for very conductive
materials but decreases the stability with respect to
blowout, and one can hardly use ceramics. These find-
ings point to the advantage of microscale combustion
and the need for sufficiently thick walls.
It has been shown that the inlet flow velocity plays
a competing role in flame stability. Low flow ve-
locities result in reduced power generation. On the
other hand, high flow velocities decrease the convec-
tive timescale below that of the upstream heat transfer
through the walls. As a result, there is only a relatively
narrow envelope of flow rates within which combus-
tion can be stabilized. When a low-power device is
being designed, more insulating materials should be
favored to minimize external heat losses. Conversely,
a high-power device would favor more conductive
materials.
Overall, propane/air microflames are more robust
than methane/air ones. They allow a wider range of
wall thermal conductivities as well as higher external-
heat-loss coefficients. This enhanced stability appears
to be due to propane’s lower ignition temperature,
which causes the reaction front to stabilize further
upstream than for methane. It is expected that larger
hydrocarbons will behave more like propane than
methane.
Finally, the available heat-transfer correlations are
inadequate when homogeneous reactions are present.
To accurately capture heat transfer in microchemical
systems, an elliptic model, such as the one presented
in this work, is necessary.
Acknowledgments
This work was supported by the Army Research
Office under Contract DAAD19-01-1-0582. Any
opinions, findings, and conclusions or recommenda-
tions expressed are those of the authors and do not
necessarily reflect the views of the Army Research
Office.
References
[1] L. Sitzki, K. Borer, E. Schuster, P.D. Ronney, S. Wus-
sow (Eds.), The Third Asia–Pacific Conference on
Combustion, Seoul, Korea, 2001.
[2] P. Aghalayam, D.G. Vlachos, AIChE J. 44 (9) (1998)
2025–2034.
[3] J.C. Ganley, E.G. Seebauer, R.I. Masel, AIChE J. 50 (4)
(2004) 829–834.
[4] H. Davy, Trans. R. Soc. London 107 (1817) 45–76.
[5] M. Fukuda, K. Koji, M. Sakamoto, Bull.
JSME 24 (193) (1981) 1192–1197.
[6] B. Lewis, G. von Elbe, Combustion, Flames and Explo-
sions of Gases, Academic Press, Orlando, FL, 1987.
[7] A. Linan, F.A. Williams, Fundamental Aspects of Com-
bustion, Oxford Univ. Press, New York, 1993.
[8] M. Maekawa, Combust. Sci. Technol. 11 (1975) 141–
145.
[9] S. Ono, Y. Wakuri, Bull. JSME 20 (147) (1977) 1191–
1198.
[10] D.G. Vlachos, L.D. Schmidt, R. Aris, AIChE J. 40 (6)
(1994) 1005–1017.
[11] S. Raimondeau, D. Norton, D.G. Vlachos, R.I. Masel,
Proc. Combust. Inst. 29 (2003) 901–907.
[12] P. Aghalayam, P.-A. Bui, D.G. Vlachos, Combust. The-
ory Modeling 2 (1998) 515–530.
 D.G. Norton, D.G. Vlachos / Combustion and Flame 138 (2004) 97–107
[13] D.G. Norton, D.G. Vlachos, Chem. Eng. Sci. 58 (2003)
4871–4882.
[14] C.H. Kuo, C. Eastwood, L. Sitzki, K. Borer, P.D. Ron-
ney (Eds.), Proceedings of the Third Joint Meeting of
the U.S. Sections of the Combustion Institute, 2003.
[15] C. Jensen, R.I. Masel, G.V. Moore, M. Shannon, Com-
bust. Sci. Technol. (2003), in press.
[16] R.I. Masel, M. Shannon, Microcombustor Having
Submillimeter Critical Dimensions, US06193501; The
Board of Trustees of the University of Illinois, Urbana,
IL, 2001.
[17] S.W. Churchill, Chem. Eng. Technol. 12 (1989) 249–
254.
[18] S.A. Lloyd, F.J. Weinberg, Nature 251 (1974) 47–49.
[19] S.A. Lloyd, F.J. Weinberg, Nature 257 (1975) 367–370.
[20] J. Ahn, C. Eastwood, L. Sitzki, K. Borer, P.D. Ron-
ney, in: Proceedings of the Third Joint Meeting of the
U.S. Sections of the Combustion Institute, Chicago, IL,
2003.
[21] P.D. Ronney, Combust. Flame 135 (4) (2003) 421–439.
[22] Fluent. Fluent 6.0, Lebanon, NH, 2002.
107
[23] F.A. Williams, in: M.D. Smooke (Ed.), Overview
of Asymptotics for Methane Flame, Springer-Verlag,
Berlin, 1991.
[24] C.K. Westbrook, F.L. Dryer, Combust. Sci. Technol. 27
(1981) 31–43.
[25] D.G. Vlachos, L.D. Schmidt, R. Aris, Combust.
Flame 95 (1993) 313–335.
[26] C.M. Miesse, R.I. Masel, M. Short, M.A. Shannon,
Combust. Sci. Technol. (2003), in press.
[27] G.L. Dugger, NACA Report 1061 (1952) 105–116.
[28] R.M. Fristrom, Flame structure and Processes, Oxford
Univ. Press, New York, 1995.
[29] G. Groppi, E. Tronconi, Chem. Eng. Sci. 52 (1997)
3521–3526.
[30] R.K. Shah, in: S. Kakaç, R.K. Shah, A.E. Bergles
(Eds.), Low Reynolds Number Flow Heat Exchang-
ers (Fully developed laminar flow forced convection
in channels), Hemisphere Publishing Company, New
York, 1983.
[31] R.K. Shah, A.L. London, Laminar Flow Forced Con-
vection in Ducts, Academic Press, New York, 1978.