Investigating a Methanol Spray Flame Interacting With an

Annular Air Jet Using Phase-Doppler Interferometry and

Planar Laser-Induced Fluorescence
JACOB A. FRIEDMAN AND METIN RENKSIZBULUT*

University of Waterloo, Department of Mechanical Engineering, Waterloo, Ontario, Canada, N2L 3G1
An experimental investigation of the interaction of an annular air jet with a methanol spray flame was
conducted using phase Doppler interferometry (PDI) to measure fuel droplet size and velocity as well as
gas-phase velocity, and planar laser-induced fluorescence (PLIF) to determine instantaneous and timeaveraged reaction zone location, as well as quantitative OH concentration. Temperature measurements were
made using Pt/Pt-10% Rh thermocouples. PLIF imaging of the OH radical showed that the spray flame had a
dual reaction zone structure with annular air off but a single reaction zone with annular air on. Peak OH
concentrations were measured to be 5400 PPM, regardless of the annular air flow rate. The OH concentration
peaks always occurred on the fuel-lean side of the reaction zone, while temperature peaks occurred on the
fuel-rich side. Gas-phase velocity vector fields show that annular air tends to channel the flow towards the
centreline, and large droplet velocity vector fields show that the large droplets follow this trend as well,
suggesting that the annular air jet assists in confining large spray droplets. Overall visible flame length is
reduced by over 50% with annular air, providing a flame well-suited to compact combustion chambers. 1999
by The Combustion Institute

NOMENCLATURE
Roman
A 5 Area viewed by collection
optics, (cm2)
A11, A10 , A01, A00 5 Einstein coefficients for
spontaneous
emission/
absorption (1/s)
Ap 5 Area of laser sheet imaged by a pixel (cm2)
B10 5 Einstein coefficient for
stimulated
absorption
(cm3/J-s2)
Bv 5 Rotational
constant
(cm21)
c 5 Speed of light (3 3 1010
cm/s)
D 5 Droplet diameter (mm)
D10 5 Arithmetic mean diameter (mm)
D32 5 Sauter mean diameter
(mm)
E 5 Laser pulse energy (J)
E(n) 5 Laser energy lineshape
function (J/Hz)
Ei 5 Incident laser energy (J)
*Corresponding author. E-mail: metin@sunwise.uwaterloo.ca
COMBUSTION AND FLAME 117:661 684 (1999)
1999 by The Combustion Institute
Published by Elsevier Science Inc.

ERay 5 Rayleigh scattered energy (J)

FJ 5 Boltzmann fraction in absorbing rotational state
Fv 5 Boltzmann fraction in absorbing vibrational state
g(n) 5 Absorption lineshape (1/
Hz)
h 5 Plancks constant (6.63 3
10234 J-s)
hL 5 Laser sheet height (cm)
J 5 Rotational
quantum
number
k 5 Boltzmann constant (1.38
3 10223 J/K)
L 5 Length of imaged volume
(cm)
m 5 Rosin-Rammler
mass
fraction
n 5 Number of laser pulses
during exposure
n0 5 Number density of molecule being probed (1/cm3)
n1 5 Number of molecules in
the excited state
npp 5 Number of photons directed at a pixel
NA 5 Number density of air
molecules (1/cm3)
0010-2180/99/$see front matter
PII 0010-2180(98)00136-9

662

J. A. FRIEDMAN AND M. RENKSIZBULUT

Q0 5 Quench rate from n9 5 0

(1/s)
Q1 5 Quench rate from n9 5 1
(1/s)
t 5 Laser sheet thickness (cm)
T 5 Temperature (K)
T11, T10 , T01, T00 5 Filter transmission at the
transition wavelength
n9 5 Vibrational
quantum
number in excited state
n0 5 Vibrational
quantum
number in ground state
V 5 Vibrational transfer rate
from n9 5 1 to n9 5 0 (1/s)
X 5 Rosin-Rammler mean diameter (mm)
Greek
Dn 5 Laser linewidth (Hz)
d 5 Rosin-Rammler exponent
V 5 Solid angle of collection
of imaging optics (sr)
sRay 5 Rayleigh scattering crosssection (6.605 3 10227
cm2/sr at 312 nm)
vv 5 Molecular vibrational energy constant (cm21)
INTRODUCTION
The combustion of liquid fuels currently provides the energy used by many stationary and
most mobile powered systems, such as industrial
and domestic furnaces and boilers, and automotive and aerospace engines. In virtually all cases,
the liquid fuel is sprayed and mixed with an
oxidizer to ensure complete combustion and
sufficient energy density to efficiently power the
system. The flame characteristics for a given
fuel depend not only on the fuel droplet size
distribution but on the spray spatial distribution
and its interaction with the oxidizer flow field.
These interactions include turbulent heat, mass
and momentum transfer, as well as complicated
chemical kinetics. Each of these phenomena is
extremely complicated in its own right, and,
when coupled together in a real flow situation,
form a formidable challenge for the combustion
researcher.

The spray flame environment is a very hostile

one in which to make measurements. Temperature extremes range from ambient to beyond
3000 K, and fuel droplets abound, making intrusive measurements difficult to obtain. Recent
developments in non-intrusive laser-based diagnostic systems, reviewed by Chigier [1], have
provided new tools to extend knowledge of the
combustion process in spray flames along several fronts. Phase-Doppler interferometry
(PDI) allows researchers to obtain spatially
resolved information on fuel droplet size and
velocity at high data rates, even in hostile environments, and can be extended to track the
gas-phase velocity as well, thus allowing velocities in the two-phase flow field to be resolved.
Planar laser-induced fluorescence (PLIF) allows two-dimensional imaging of the OH radical (or other species) concentration in the spray
flame; a good marker of the reaction and recombination zones [2].
The work reported herein applies PDI and
PLIF techniques to a spray combustion system
consisting of a standard pressure-swirl nozzle
co-axially located within an annular air jet. This
configuration has been previously studied under
both non-reacting [3, 4] and reacting [5] conditions using PDI and conventional photographic
techniques. Combining PDI and PLIF allows
correlation of fuel droplet size and velocity
distribution as well as gas-phase velocity fields
with reaction zone location, providing considerable insight into the combustion phenomenon.
The results may also constitute a valuable data
bank for computational studies.
EXPERIMENTAL APPARATUS
Burner System
The burner used in this work consists of a
standard pressure-swirl nozzle (Delavan 0.75
60A) mounted at the end of a hollow cylinder,
fitted with a cylindrical bluff body as shown in
Fig. 1. This assembly is coaxially mounted
within an outer shroud, forming an annular
passage of 64 mm outer diameter and 50 mm
inner diameter. Air is supplied to the burner by
a turbo-blower through a calibrated rotameter
and controlled by a manual valve. Methanol is

METHANOL SPRAY FLAME INVESTIGATION

663
TABLE 1
PDI System Optical Parameters Used in the Present
Work

Laser
Argon ion laser
Transmitting optics
Beam expanding telescope ratio
Beam separation
Transmitting lens focal length
Probe volume waist (1/e2)
Receiving optics
Receiving lens focal length
Lens diameter
Spatial filter slit width
Collimating lens focal length
Magnification
Separation between detectors
A, B, and C
Orientation

Fig. 1. Burner system cross-section (all dimensions in mm).

supplied to the atomizer nozzle from a pressurized holding tank at approximately 6 bar, and
metered through a calibrated rotameter. The
entire burner assembly is mounted in the vertically up-firing position on a manual three-axis
traversing system. The burner was operated at a
fixed fuel flow rate of 0.42 g/s, with annular air
flow rates of 0, 4.77, and 9.52 l/s, corresponding
to Reynolds numbers (based on the bluff body
diameter of 50 mm) of 0, 13,000, and 26,000,
respectively.
Phase-Doppler Interferometry System
PDI allows simultaneous determination of
droplet size and velocity by measuring the phase
difference and frequency of light scattered when
a droplet passes through the intersection of two
laser beams (the Doppler burst). In the basic
PDI arrangement, the Doppler bursts are ob-

500750 mW
2:1
40 mm
250 mm
117 mm
300 mm
72 mm (f/4.2)
150 mm
250 mm
0.833
16.32 mm (A-B)
49.96 mm (A-C)
30 off forward axis

served by two photodetectors arranged at different points in space. The two detectors observe the same Doppler burst but with a phase
difference due to their spatial separation. This
phase difference is directly proportional to the
droplet diameter, and the frequency of the
Doppler burst is directly proportional to the
droplet velocity. The theory of operation of the
phase-Doppler system is well established (e.g.,
[6 8]) and therefore will not be repeated here.
The PDI system used in this work is an
Aerometrics DSA fibre-optics based system
with two velocity component capability. The
system is operated in first order refraction
mode, with the receiver located at a 30 off-axis
angle. A Coherent Innova 90 water-cooled argon ion laser is used, operating at approximately
600 mW output. System optical parameters are
summarized in Table 1.
Planar Laser-Induced Fluorescence System
PLIF allows non-intrusive instantaneous 2-D
imaging of species concentration. According to
the principles of quantum mechanics, a molecule can be raised to an excited state by interaction with a photon of a frequency corresponding to the energy difference between the initial
lower energy state and the desired excited state.
The molecule can then relax either by interaction with neighbouring species (quenching) or
by the emission of a photon corresponding to

664

Fig. 2. Planar laser-induced fluorescence system configuration.

the energy difference between the excited state

and the relaxed state, which may be the same as
the original state before excitation (resonant
fluorescence) or a different state (non-resonant
fluorescence). By using a tuneable laser to excite a population of selected molecules and
observing the intensity of fluorescence emitted,
the concentration of that species can be determined.
In the present work, imaging OH fluorescence was accomplished by exciting the (1,0)
band of the A2 2 X2II transition and detection
of the resulting fluorescence from the (1,1) and
(0,0) bands. The Q1(6) rotational transition at
35,334.35 cm21 (283.01 nm) [9] was selected for
excitation due to the low dependence of fluorescence on temperature at this rotational level,
and the large ground state population available
for excitation [10]. Excitation of the (1,0) band
and detection of the resulting (1,1) and (0,0)
fluorescence allows sufficient spectral distance
between the excitation and detection frequencies that most elastically scattered light from
spray droplets can effectively be filtered out,
although regions of high particle density in the
immediate vicinity of the fuel nozzle scatter
sufficient light that not all can be filtered out.
The PLIF system used in this work consisted
of a Lumonics HY750 Nd:YAG pump laser
operated at 532 nm, a tunable Lumonics HyperDye 300 dye laser and a Lumonics HyperTrack
1000 second harmonic generator, as illustrated
in Fig. 2. This combination allowed generation
of ultraviolet laser pulses of approximately 6 mJ

J. A. FRIEDMAN AND M. RENKSIZBULUT

at 35334.35 cm21. The laser pulse was spread
into a sheet by using a cylindrical lens followed
by a long focal length spherical lens to produce
a sheet approximately 70 mm high and 0.5 mm
thick at the focal plane of the spherical lens. A
Princeton Instruments intensified and cooled
CCD camera with a resolution of 576 3 384
pixels was arranged at 90 to the laser sheet to
image the OH fluorescence. The ICCD camera
was equipped with a 105 mm Nikor UV-grade
lens, fitted with a 10 nm band pass interference
filter centered at 313 nm to pass (0,0) and (1,1)
fluorescence at 306 320 nm while blocking essentially all light scattered at the laser excitation
wavelength of 283 nm.
In order to obtain quantitative OH number
density from planar laser-induced fluorescence
intensity measurements, as obtained from the
ICCD pixel intensity data, it is necessary to first
determine the number of photons emitted from
a volume in space as a function of laser fluence,
local fluorescing species number density, and
the spectral characteristics of the laser itself, as
well as the species absorption and emission
characteristics. The number of fluorescence
photons arising from a given volume in space is
a function of the number of fluorescing molecules in the excited state. This number is in turn
dependent on the total population of the fluorescing species and the exciting radiation. The
number of molecules in the excited state (for
weak laser excitation) can be determined by [11]
n1 5

S DF E
B 10
c

z ~ At!.

1
h Lt

E~ n ! g~ n ! d n ~ f n f Jn 0!
(1)

The first term in parenthesis represents the

Einstein coefficient for stimulated absorption.
The second term is the overlap integral between
the laser energy density spectral lineshape and
the molecular absorption lineshape, which represents the amount of laser energy that can
potentially be absorbed by molecules in the
absorbing quantum states. Typically, for OH at
combustion temperatures and using a laser with
a linewidth of 0.2 cm21, the value of the overlap
integral ranges from 0.7 0.85 E/Dn, where E is
the total laser pulse energy and Dn is the laser
linewidth in Hz. Details on the overlap integral

METHANOL SPRAY FLAME INVESTIGATION

calculation can be found in [11]. The third term
represents the number density of molecules in
the ground state that are in the appropriate
quantum state for absorption, where the Boltzmann fractions f n and f J can be found from the
expressions below, and the final term is the
volume being probed. It should be noted that
the laser sheet thickness, appearing in the second and fourth terms, cancels out, and hence
detailed knowledge of laser sheet thickness is
not necessary in this portion of the calculation:

DF

n hc v n
f n~T! 5 exp 2
kT

f J~T! 5

DG

hc v n
1 2 exp 2
kT

665
TABLE 2
Spectroscopic Constants for OH
OH (A2 2 X2P (1,0))

Constant

1.4 3 106 sec21 (0,0) [30]

5.11 3 105 sec21 (1,0) [30]
8.38 3 105 sec21 (1,1) [30]
5.6 3 108 sec21 (Q 0 and Q 1 ) [31][32]
1.12 3 109 sec21 [33]
1.9 3 1024 cm3 J21 s22 [11][34]
18.513 cm21 2P(v0 5 0)
16.126 cm21 2(v0 5 1) [9]
3569.59 cm21 2P [9]
2792.92 cm21 2 [9]

A 00
A 10
A 11
Q 0, Q 1
V
B10
Bv

vv

(2)

hcB n
B nhcJ~ J 1 1!
~2J 1 1! exp 2
.
kT
kT
(3)

In order to minimize the dependence of excited

state population on temperature, it is necessary
to select a rotational ground state quantum
level J that does not vary strongly with temperature. This J level can be found by differentiating Eq. 3 with respect to T and setting to zero.
For OH, J 5 6.5 or 7.5 are found to be the least
sensitive for temperatures in the 1500 K range.
Molecules in the excited state can relax
through photon emission (fluorescence) and
collision with other molecules. Collision with
other molecules can result in complete relaxation (quenching), downward energy transfer to
lower excited states, or upward energy transfer
to higher energy states. Downward and upward
energy transfer result in population of many
rotational levels in the excited state, and in the
case of (1,0) excitation, population of the n9 5 0
level as well. Generally, upward energy transfer
resulting in population of the n9 5 2 level is
negligible. Since many excited state levels are
populated, the resulting fluorescence will be at
many frequencies, some of which will pass
through collection optics (which generally include narrow band interference filters) at different efficiencies. A fluorescence yield model
which gives the fraction of excited state molecules that will result in photons that pass
through the collection optics filter and accounts
for downward vibrational excited state transfer
is given by [12]:

Fy

A 01
A 00
1 T 00
Q0
Q0

T 11A 11 1 T 10 A 10 1 V T 01
5

Q1 1 V
(4)

In this work, the interference filter used had a

10 nm bandpass, centered at 314 nm. The
transmission factors for this filter were [13]: T 11
5 0.17 (313 nm), T 00 5 0.11 (308 nm), T 10 5
T 01 5 0 (283 and 346 nm). Spectroscopic
constants used in the above equations are listed
in Table 2.
Each pixel on the ICCD receives photons
emitted by a volume in space defined by the
area of the laser sheet imaged by the pixel
multiplied by the laser sheet thickness. The
number of photons emitted by this volume in
space that will be directed at the imaging pixel
and will pass through the interference filter is
obtained by multiplying the number of molecules in the excited state, given by Eq. 1, by the
fluorescence yield, Eq. 4, then multiplying by
the fraction of total radiation emitted in all
directions collected by the imaging optics, resulting in the following:
n pp 5

S DF E
SD
B 10
c

1
hL

z ~ A p!~F y!

E~ n ! g~ n ! d n ~ fn f Jn 0!

V
.
4p

(5)

It should be noted that the above equation is

independent of laser sheet thickness. In order to

666

J. A. FRIEDMAN AND M. RENKSIZBULUT

use this equation to obtain the ground state

population n0, it is necessary to determine the
number of photons incident on a given pixel
from the ICCD output, which is generally a
unitless intensity count. In order to convert the
intensity count into a photon count, a calibration factor must be obtained.
In this work, the calibration factor for the
ICCD camera was obtained following a method
proposed by Luque and Crosley [14] by using
Rayleigh scattering of laser light by air to provide a known number of photons incident on a
particular pixel. The Rayleigh scattered power
from a given volume of air is:

S D

E Ray 5 n

Ei
~N As RayV!~ AL!.
A

(6)

The first term in parenthesis is the incident

energy per unit area, and the second term is the
Rayleigh scattering rate per unit volume. The
third term is the volume being imaged. It should
be noted that the cross-sectional area of the
laser beam, A, cancels out of the first and third
terms, and hence knowledge of the laser beam
dimensions is not required.
The incident energy arising from viewing a
length L of laser beam at right angles in space
can be converted to a photon count by dividing
by h n , Plancks constant times frequency. The
number of photons arriving at the lens can then
be determined as shown below. It should be
noted that the incident pulse energy is the time
averaged power divided by the number of pulses
per second:

S D

n pp 5 n

Ei
~N AsRayV! L.
hn

(7)

If L is selected to correspond to the imaged

laser beam length on one pixel, then all photons
arising from the volume in space defined by this
laser beam length arriving at the lens will be
destined for a column of pixels imaging this
beam segment. If the intensity counts of the
pixels in the column of pixels are summed up, a
calibration factor corresponding to intensity
count per photon incident at the lens can be
obtained. This calibration factor can then be
used to obtain n pp required in the fluorescence
calculation from an intensity count for a pixel.

Fig. 3. Photographic image of flame with annular air off

(1/8 sec @f/8, ISO 400 film).

Temperature Measurements
Temperature measurements were made using a
75 mm type S (Pt/Pt 2 10%Rh) thermocouple
mounted on a stainless steel and ceramic probe.
No corrections for radiative losses, catalytic
effects or droplet impacts were made to the
mean temperature measurements obtained.
RESULTS AND DISCUSSION
Photographic images, made using laser sheet
illumination through the centreline of the flame
to highlight spray particles, are presented in
Figs. 35. The imaged height of the flame in
these figures was 150 mm, with laser sheet
lighting through the lower 70 mm. Figure 3,
obtained using 1/8 sec exposure time, shows the
spray flame with annular air off. The flame
stabilized approximately 12 mm above the spray
nozzle, and the stabilization point was steady
and symmetric about the spray axis. An inner
reacting core approximately 13 mm in diameter
was visible inside the flame, with little variation

METHANOL SPRAY FLAME INVESTIGATION

Fig. 4. Photographic image of flame with annular air 5 4.77

l/s (1/8 sec @ f/8, ISO 400 film).

in diameter with distance from the nozzle. This

inner core was observed to be quite steady in
the region within 150 mm of the nozzle, with
transition to turbulence occurring approximately 100 mm downstream. Considerable
amounts of fuel spray were visible outside the
flame sheath, which would contribute to low
combustion efficiency, although some fuel
would undoubtedly be re-entrained into the
flame further downstream. The overall flame
length was approximately 55 cm.
When the annular air flow rate was increased
to 4.77 l/s, the flame structure changed considerably, as seen in Fig. 4. There was no longer a
distinct inner core and outer flame sheath, but
rather a single reaction zone. In this case, the
entire spray cone passed through reaction zones
before emerging into the surrounding flow. The
reaction zone in the region near the nozzle had
a tulip shape, with a secondary reacting flow
apparent near the outer edge of the bluff body.
The overall flame length under these conditions
was approximately 35 cm.
Increasing the annular air flow rate further to

667

Fig. 5. Photographic image of flame with annular air 5 9.52

l/s (1/8 sec @ f/8, ISO 400 film).

9.52 l/s, as shown in Fig. 5, resulted in a much

shorter, more intense flame. The flame stabilized just above the bluff body, with a tapered
shape and a base diameter equal to the bluff
body diameter. The overall flame length reduced to approximately 24 cm, less than half the
length of the flame with annular air off. In this
case, the entire spray cone trajectory passed
through flame, and once through was deflected
downstream somewhat by the annular air jet.
Figures 6, 8, and 10 are composite images of
OH fluorescence in the spray flame obtained by
stacking three time averaged OH fluorescence
images obtained in three planes. The total flame
height imaged is 195 mm. Each time-averaged
image was produced by superposing OH fluorescence from 100 laser shots onto a single
exposure. Darkest areas correspond to regions
of highest OH fluorescence, and thus highest
OH concentration, and mark the time-averaged
locations of the reaction zones quite clearly.
In Fig. 6, with annular air off, the flame is
seen to consist of inner and outer reaction

668

J. A. FRIEDMAN AND M. RENKSIZBULUT

Fig. 6. One hundred shot average composite OH fluorescence image, annular air off.

zones, as suggested in the photographic images.

The outer reaction zone appears to begin approximately 12 mm above the nozzle. Interestingly, the inner reaction zone does not appear to
begin until further downstream, approximately
24 mm from the nozzle, and has a much less
distinct anchor point than that in the outer
zone. Near the nozzle, the emerging spray cone
is visible due to very strong elastic scattering
that cannot be entirely suppressed by the interference filter. Comparison of the time-averaged
image with instantaneous single shot images
obtained in the same three planes, though at
different times, given in Fig. 7, show that the
outer reaction zone appears to be essentially
laminar, with a slowly varying structure, while

Fig. 7. Annular air off, single shot OH fluorescence images.

the inner reaction zone appears quite turbulent,

with a thicker wavy structure containing fairly
small scale eddies on the order of a few millimeters. It can also be seen in the instantaneous
image (Fig. 7) that the inner and outer reaction
zones appear continuous, arising from the same
point approximately 12 mm above the spray
nozzle, contrary to the impression given by the
time-averaged image. The first few centimeters
of the inner reaction zone do not produce

METHANOL SPRAY FLAME INVESTIGATION

strong instantaneous fluorescence, and the unsteady nature of this region produces quite a
diffuse and weak image in the time-averaged
shot. There appears to be a closed region within
the inner reaction zone that produces little OH
fluorescence, suggesting the presence of an inner air or fuel vapour core extending approximately 100 mm downstream of the nozzle. Immediately beyond this inner core, across a
distinct boundary, are hot products of combustion producing considerable OH fluorescence.
Dark spots in the region of the nozzle, visible in
Figures 6 and 7, are due to elastically scattered
light from droplets that leaks through the interference filter, which are present in high concentration in this region (see Figs. 19 and 20).
Increasing the annular air flow rate to 4.77 l/s
shortens the flame considerably, and leads to a
tulip-shaped flame structure that encloses the
emerging spray for approximately 21 mm downstream, before the spray cone crosses the reaction zone into the surrounding air stream, as
seen in Fig. 8. The figure also shows considerable amounts of OH fluorescence in the region
between the emerging spray cone and the bluff
body surface. This region appears diffuse, suggesting a fair degree of unsteadiness. In contrast, the reaction zone just downstream appears stationary for the first 25 mm or so, until
it rapidly becomes diffuse. Figure 9, showing
single shot images of the same flame, indicates a
structure similar to that seen in the time-averaged fluorescence images in the region near the
nozzle, but rapidly becomes chaotic further
downstream. The length scale of the structures
seen in this figure is much larger than those seen
in the air-off case, and isolated structures not
associated with a continuous reaction zone are
apparent. Some flow symmetry is apparent in
the region near the nozzle, but beyond approximately 130 mm, instantaneous symmetry is lost.
The symmetry of the structures, particularly
those seen in the z 5 65130 mm plane, suggest
that torroidal vortices on the outside of the
reaction zone are being formed. There appears
to be some flame impingement on the surface of
the bluff body, as well as localized torroidal
vortices which are quite unsteady in nature.
Figure 10 presents a time-averaged composite image of the flame with an annular air flow
rate of 9.52 l/s. The flame has lifted off the bluff

669

Fig. 8. One hundred shot average composite OH fluorescence image, annular air 5 4.77 l/s.

body surface by about 10 mm, and has a base

diameter equal to the bluff body diameter,
suggesting that the reaction zone has stabilized
in a region of strong shear along the inner
surface of the annular jet. The reaction zone in
the vicinity of the emerging spray cone appears
quite steady, as evidenced by the appearance of
the OH fluorescence. Approximately 50 mm
downstream of the nozzle, the structure appears
much more diffuse, suggesting a high degree of
unsteadiness in this region. This interpretation
is supported by the instantaneous fluorescence
measurements shown in Fig. 11, which indicate

670

J. A. FRIEDMAN AND M. RENKSIZBULUT

Fig. 10. One hundred shot average composite OH fluorescence image, annular air 5 9.52 l/s.

Fig. 9. Annular air 5 4.77 l/s single shot OH fluorescence

images.

increasingly chaotic structures with distance

from the nozzle. The reaction zone no longer
appears to be continuous in the z 5 130 195
mm plane, with isolated structures apparent.
Figure 12 presents the instantaneous fluorescence intensity (ICCD pixel intensity count)
measured along a line 25 mm downstream of
the nozzle, for the air-off and annular air 5 9.52

l/s cases. Although these flow conditions are

very different, the instantaneous OH fluorescence intensities in the reaction zones are almost identical, suggesting that the detailed combustion chemistry of the spray flame is
independent of flow conditions. The peak OH
number density in both these cases is approximately 2.2 3 1016/cm3, or 5400 PPM assuming
1800 K local temperature which is 22 times
larger than the equilibrium concentration expected from stoichiometric combustion of
methanol at this temperature. This super-equilibrium condition within the reaction zone is

METHANOL SPRAY FLAME INVESTIGATION

671

Fig. 12. Instantaneous OH fluorescence intensity in the z 5

25 mm plane with annular air off and annular air 5 9.52 l/s.

Fig. 11. Annular air 5 9.52 l/s single shot OH fluorescence

images.

consistent with other measurements and computations for laminar and turbulent diffusion
flames reported in the literature [15, 16]. Noise
appearing as isolated intensity spikes in the
centre region for the annular air-off case is due
to elastically-scattered light from droplets
present in high concentrations in this region.
Figures 1315 present measurements of mean

Fig. 13. Mean OH concentration and temperature in three

planes with annular air off.

672

J. A. FRIEDMAN AND M. RENKSIZBULUT

Fig. 14. Mean OH concentration and temperature in three

planes with annular 5 4.77 l/s.

Fig. 15. Mean OH concentration and temperature in three

planes with annular 5 9.52 l/s.

OH concentration and mean temperature in

three planes 25, 60, and 100 mm downstream of
the nozzle for the three annular air flow rates
tested. Figure 13, for the annular air-off case,
clearly shows two OH concentration and temperature peaks in each plane, consistent with
the observation of two reaction zones. The OH
concentration peaks occur just inside the location of the mean temperature peaks. Figure 14,
for the annular air 5 4.77 l/s case shows a
distinct OH concentration peak in the z 5 25
mm plane occurring just outside (fuel lean side)
of the temperature peak. This observation is
consistent with other investigations [17, 18]
which show that peak temperatures in a diffusion flame occur on the fuel rich side of the
reaction zone, while peak OH concentrations
occur on the fuel lean side. Mean temperature
and OH peaks are not as apparent in the
downstream planes (60 and 100 mm downstream of the nozzle) as turbulent fluctuations
spread out the region over which the reaction
zone can be located at any instant. This is also

true for the annular air 5 9.52 l/s case, Fig. 15,
which shows distinct mean temperature and OH
concentration peaks only in the plane 25 mm
downstream of the nozzle. Again, the temperature peak occurs slightly inside (fuel rich side)
of the mean OH concentration peak. It is
apparent that there is little difference in centreline temperature for the two annular air-on
cases tested, although both these cases result in
considerably hotter centerline temperatures
than the air-off case.
Measurements of droplet Sauter mean diameter obtained in planes 25, 60, and 100 mm downstream of the nozzle are presented in Figs. 1618.
Included in these figures are best fit RosinRammler parameters at locations corresponding to the peak volume flux. At these locations,
the Rosin-Rammler distribution model provides reasonable agreement with data, generally
with the arithmetic mean diameter D 10 within
10% and the Sauter mean diameter D 32 within
5% of the measured values. The RosinRammler size distribution model is given by

METHANOL SPRAY FLAME INVESTIGATION

Fig. 16. Variation of Sauter mean diameter D 32 in the z 5

25 mm plane.

S S DD

D
m 5 1 2 exp 2
X

(8)

where m is the mass fraction of droplets less

than diameter D, X is the Rosin-Rammler
mean diameter, and d is the Rosin-Rammler
exponent.
In a plane 25 mm downstream of the nozzle,
Fig. 16, D 32 trends for the air-off and air 5 9.52
l/s case are similar, but the air 5 4.77 l/s case is

Fig. 17. Variation of Sauter mean diameter D 32 in the z 5

60 mm plane.

673

Fig. 18. Variation of Sauter mean diameter D 32 in the z 5

100 mm plane.

quite different, with a much smaller Sauter

mean diameter on centreline, and a much faster
increase with radius. This is likely due to the
presence of combusting gasses in the immediate
vicinity of the emerging fuel spray, as seen in
Fig. 8 and 9, an influence not felt in the other
two cases. At a location 60 mm downstream of
the nozzle, as shown in Fig. 17, the mean
droplet diameter does not appear to vary substantially with annular air flow rate, although in
the region near the centreline, there was insufficient data to obtain mean diameter for the
annular air-on cases. In a plane 100 mm downstream of the nozzle, shown in Fig. 18, there is
little difference in mean droplet diameter for
the two annular air-on cases. With annular air
off, there was insufficient data to obtain mean
droplet diameter except near centreline, as most
droplets that escaped the flame envelope had
settled back down and never reached this plane
without an air jet to transport them. The centreline mean droplet diameter in this plane
increased with increasing annular air flow rate,
although there is little difference between the
two annular air-on cases.
Figures 19 21 present the radial distribution
of droplet volume flux in planes 25, 60, and 100
mm downstream of the nozzle for the three
annular air flow rates tested. In the plane 25 mm
downstream of the nozzle, Fig. 19, peak volume
fluxes for all air flow rates occur at a radial

674

Fig. 19. Variation of volume flux with radial position in the

z 5 25 mm plane.

position approximately 15 mm off centreline.

The centreline volume flux is essentially zero
with annular air on. However, as seen in Fig. 20,
in a plane 60 mm downstream of the nozzle,
there is a volume flux peak on centreline as well
as a smaller peak at a radial location of 35 mm
for the annular air-off case. Since the Sauter
mean diameter D 32 is small on centreline, the
number density must be high to produce a large
volume flux. The centreline peak disappears
when annular air is turned on, which results in
essentially zero particle flux in the centreline

Fig. 20. Variation of volume flux with radial position in the

z 5 60 mm plane.

J. A. FRIEDMAN AND M. RENKSIZBULUT

Fig. 21. Variation of volume flux with radial position in the

z 5 100 mm plane.

region. However, the volume flux peaks in the

off-centreline regions increase substantially,
suggesting that a redistribution of the spray
occurs when the annular air jet is on. In the
plane 100 mm downstream of the nozzle, shown
in Fig. 21, only small amounts of droplets are
detectible with annular air off, and these are
near the centreline. With annular air on, the
volume flux peak occurs at a radial position
approximately 30 mm from centreline, with the
4.77 l/s annular air flow rate producing a peak
volume flux approximately half that of the 9.53
l/s case. Integration of the volume flux curves
results in the total volume flow rate across a
plane as a function of distance from the nozzle,
and is shown in Fig. 22. The total volume flow
rate of methanol supplied to the nozzle is 0.53
cm3/s. Although it appears that the integrated
volume flux with annular air-off results in the
fastest reduction of droplet flow rate, suggesting
better combustion efficiency, it should be noted
that considerable amounts of fuel spray escape
the flame envelope and quickly settle back down
due to gravity, as there is no annular jet to
transport them. This results in few droplets
having sufficient momentum to reach planes far
downstream of the nozzle, and hence the apparent rapid disappearance of fuel. With annular
air on, it appears that the total volume flux past
a given plane increases somewhat with annular
air flow rate. This would be expected, since the

METHANOL SPRAY FLAME INVESTIGATION

Fig. 22. Integrated volume flux versus distance from the

nozzle for all air flow rates.

higher velocities associated with increased annular air flow rate would mean that droplets
require less time to travel downstream, which
means they would have less time to evaporate
before travelling a given distance. Other factors
which affect the size and volume flux distribution are the changes in flame structure that arise
with varying annular air flow rates, which affect
local evaporation rates, and hence droplet size
and trajectory histories.
Gas-phase mean velocity was obtained from
raw data sets by filtering the data to generate a
file consisting of only small droplet data. The
cut-off diameter was typically 5 mm, resulting in
a turbulent Stokes number of approximately 40
(defined as the turbulent time scale divided by
the droplet drag response timescale) in a region
with an integral time scale of 2 ms and relative
velocity of 10 m/s. Figure 23 presents a diameter-velocity correlation plot, obtained in the z 5
25 mm plane, approximately 18 mm off centreline in a region containing a wide spectrum of
droplet sizes under highly turbulent conditions.
As can be seen, there appears to be a clear
cut-off in measured velocity for diameters near
10 mm, below which there is essentially no
change in droplet velocity, suggesting that droplets in this size range are accurately following
the gas-phase flow.
Figures 24 to 26 present the gas-phase velocity vector fields for all annular air flow rates

675

Fig. 23. Typical droplet diameter/axial velocity correlation,

obtained at z 5 25 mm r 5 18 mm with an annular air flow
rate of 4.77 l/s.

tested, obtained by filtering the raw data set to

eliminate velocity data from all particles larger
than 510 mm, depending upon location. Figure
24, showing the velocity vector plot with annular
air off, shows a smoothly developing velocity
field, with peak velocities occurring on centreline, with a magnitude of approximately 5 m/s.
The inner reaction zone, with a nearly constant
diameter of 16 mm, is located in a region of

Fig. 24. Gas-phase velocity vector plot with annular air off.

676

J. A. FRIEDMAN AND M. RENKSIZBULUT

Fig. 25. Gas-phase velocity vector plot with annular air 5

4.77 l/s.

Fig. 27. Large droplet (D . 30 mm) velocity vector plot

with annular air off.

relatively high velocity in each plane, ranging

from 2.5 to 4 m/s. The outer reaction zone, by
contrast, is located in regions of relatively low
velocity, below 1 m/s. In all cases, the velocity
vectors are nearly parallel to the reaction zone.
There are no regions of large velocity gradients
present, and no evidence of recirculating flow.

The velocity field changes quite a bit when the

annular air flow rate is increased to 4.77 l/s, as
seen in Fig. 25. In this case, there is a channeling effect evident inside the reaction zone,
where the gas flow is noted to be directed
towards the centreline. Outside the reaction
zone, the velocity vectors are directed away
from centreline at all locations measured. The
centreline velocity in the z 5 100 mm plane is
around 3.75 m/s. Increasing the annular air flow
rate further to 9.53 l/s, as shown in Fig. 26,
reveals clear evidence of recirculating flow
within the reaction zone in the z 5 10 and 25
mm planes, with negative axial gas-phase velocity apparent at the flame front. Interestingly, in
the z 5 25 mm plane, it appears that the axial
velocity is negative on the inner side of the
reaction zone, and positive on the outside,
suggesting strong shear across the flame front.
The channeling effect noted in the 4.77 l/s air
flow case is even more evident at the higher air
flow rate. As before, there is no evidence of any
re-entrainment occurring.
Figures 2729 present velocity vector plots of
large (.30 mm) droplet trajectories for all air
flow rates tested. As can be seen in Fig. 27, with
no annular air flow, the trajectories are essentially ballistic with an origin at the nozzle. The
cone formed by the peak flux trajectories en-

Fig. 26. Gas-phase velocity vector plot with annular air 5

9.52 l/s.

METHANOL SPRAY FLAME INVESTIGATION

677

Fig. 28. Large droplet (D . 30 mm) velocity vector plot

with annular air 5 4.77 l/s.

Fig. 29. Large droplet (D . 30 mm) velocity vector plot

with annular air 5 9.52 l/s.

closes an angle of approximately 55, very close

to the expected 60 from this type of nozzle.
With 4.77 l/s annular air flow rate, shown in Fig.
28, the droplet velocity field no longer appears
ballistic. The initial cone defined by the peak
flux velocity vectors remained unaffected at 55,
but narrowed considerably beyond z 5 40 mm.
In the z 5 100 mm plane, there was some
evidence that the droplet trajectories within the
reaction zone tended towards the centreline,
producing a focussing effect that would concentrate fuel flow to the centre. When the annular
air flow rate is increased to 9.53 l/s, the focussing effect is very pronounced, as seen in Fig. 29.
There is still no evidence of large droplet reentrainment, and no evidence of large droplet
recirculation. The initial cone angle, defined by
the location of the volume flux peak, has widened noticeably to approximately 65 in the 9.53
l/s annular air flow rate case. Beyond 60 mm,
there is no further widening of the peak flux
location.
In the flows tested, relative velocities between
droplets and the gas-phase flow can be in excess
of 25 m/s. However, the Weber number for a 50
mm methanol droplet subjected to this relative
velocity is less than 2, so droplet break-up does
not occur, although droplet distortion, resulting
in non-spherical droplets, is likely.

ERRORS AND UNCERTAINTIES

PDI Uncertainties
The PDI system consists of an optical system and
a processing unit. Uncertainties associated with
individual droplet size and velocity measurement
are relatively easy to assess, and are related to
the system hardware and processing technique.
Uncertainties associated with statistical calculations such as mean diameter and volume flux are
much more difficult to quantify, as are uncertainties associated with user-selected parameters such as photomultiplier voltage, threshold,
and filter settings.
All PDI measurements depend ultimately on
the geometry of the system. The velocity measurement is dependent on the fringe spacing in
the probe volume, which is in turn dependent
on the laser beam intersection angle and laser
wavelength. The laser wavelength is known to a
high degree of precision, and does not contribute to instrument uncertainties. The beam intersection angle is determined by the transmitting lens focal length and beam separation, each
of which is known and/or measurable to better
than 0.5% accuracy. The constant of proportionality between phase difference and droplet
diameter is dependent on the system geometry,

678
and its uncertainty is estimated to be approximately 1.5%.
The other term in the phase/diameter relationship is the relative index of refraction m
between the fuel and air. Although this is known
to a high degree of precision, it is a function of
droplet temperature. Since the fuel droplet can
be at any temperature from below ambient to
near its boiling point, the index of refraction is
variable between these limits. For methanol, the
index of refraction ranges from 1.3108 to 1.3329
over the temperature range from 10C to near
boiling at 64.5C [19], resulting in a sizing
uncertainty of approximately 4%. Selection of
index of refraction at the mid-point of the range
limits this error to 62%. Recent work by
Schneider and Hirleman [20] examined the effect of index of refraction gradients on particle
sizing, as larger particles would not be isothermal in a combusting environment and radial
gradients in temperature and hence refractive
index would be present. It was concluded that,
for atmospheric pressure flames, there is insufficient temperature and refractive index range
to cause a significant sizing error. However, in
high-pressure combustion systems with fuels
able to reach several hundred degrees, this
effect should be addressed.
The diameter-phase relationship assumes
that the particles are spherical, and at least of
the same order as the incident light wavelength.
As the particle size decreases, oscillations in the
phase/diameter relationship can occur, and care
must be taken in selecting the appropriate collection angle in order to size small particles
accurately. Naqwi and Durst [21, 22] performed
a detailed theoretical analysis of the scattering
process, and concluded that a PDI system could
be configured to accurately size particles as
small as 0.5 mm, at the expense of overall size
range. Rather than examining this problem
from a theoretical point of view, Ceman et al.
[23] and OHern et al. [24] performed a set of
experiments using a vibrating orifice aerosol
generator capable of accurately producing droplets as small as 3.7 mm, and assessing the
Aerometrics Phase Doppler Particle Analyzers
sizing ability directly. Oscillations in response
were found to occur at the two commonly
employed collection angles tested. Sizing errors
were found to increase dramatically below 10

J. A. FRIEDMAN AND M. RENKSIZBULUT

mm, and were on the order of 60% at 5 mm. The

threshold for sizing errors over 10% was found
to be approximately 17 mm at a 30 collection
angle.
The Aerometrics system records individual
particle data (size, velocity, arrival time, and
gate time), then presents a statistical summary.
Generally, the system is set to measure a fixed
number of particles, then compute statistical
mean diameters as well as mean and RMS
velocity and volume flux, correcting for probe
volume size variation with droplet diameter.
The statistical uncertainties associated with calculations of means from finite sample sizes can
be easily determined. However, the errors associated with instrument uncertainties are difficult
to quantify, since they are dependent on the
actual distribution measured. As discussed
above, the uncertainty of a given measurement
is dependent on the particle size, increasing as
size decreases. Therefore, it would be expected
that a data set consisting of mostly small particles would have a higher level of uncertainty
associated with its calculated means than one
that is skewed more to larger particles. In
addition, in any experiment, the system only
validates a certain percentage of Doppler
bursts, while others are rejected for a number of
reasons, including low signal to noise ratio,
velocity and/or diameter outside pre-set bounds
and non-agreement between detector pairs.
Typically, the validation rate of Doppler bursts
that pass through initial filters (low pass, high
pass, threshold, etc.) ranges between 85 and
95%, thus at least 5 to 15% of the droplets
passing through the probe volume are not measured. If these unmeasured droplets are randomly distributed with respect to velocity and
diameter, no bias in mean measurements will be
induced. However, if there is a bias in rejections, then the data set itself will be biased.
These types of uncertainties have been acknowledged in the literature, but there does not seem
to be any work directed at quantifying these
effects.
Volume flux measurements made by the system are particularly uncertain. The instrument
calculates the volume flux by determining the
effective probe area for each size class, then
multiplies the count in each size class by the
mean droplet volume in that size class, divided

METHANOL SPRAY FLAME INVESTIGATION

by the probe area and time span of acquisition.
The major source of uncertainty in this calculation arises from the determination of the probe
area. The effective width of the probe area is
dependent on droplet size, droplet velocity,
intensity distribution and photomultiplier gain.
The Aerometrics software includes a routine to
determine the probe area in each size class
based on the maximum transit time of particles
through the probe volume in each size class.
The assumption is made that, in a size class
containing many counts, the longest transit time
of particles in that size class corresponds to the
particle whose trajectory carried it through the
centre of the probe volume, defining the probe
volume width when the transit time is multiplied
by the particles velocity. In many sprays this
provides an adequate result. However, in the
case of sprays with a wide size distribution and
limited system memory (thus limiting the total
number of samples that may be obtained), each
size class may only have a few particles associated with it, and the assumption that the longest
transit time corresponds to a trajectory passing
through the centre of the probe volume may be
erroneous, leading to underestimation of probe
volume area and overestimation of volume flux.
In dense sprays, it is also possible that some
measurements would involve multiple particles
passing through the probe volume. If one particle enters the probe volume before the preceding one exits, and if they are of similar size
and velocity, the instrument will not reset after
the first particle leaves, but will only count a
single particle, with a transit time corresponding
to that of the sum of the two particles passage,
resulting in a large overprediction of probe
area. It is possible to reduce the likelihood of
multiple particle detection by reducing photomultiplier gain and increasing the threshold
setting, but this would result in reduced detectibility of small particles. McDonell and Samuelsen [25] found that flux measurements
showed a strong dependence on photomultiplier gain setting which did not show asymptotic
behavior, unlike diameter measurements which
tended to stabilize after photomultiplier gain
was set sufficiently high to detect the small
particles. Integration of their flux data over the
spray field yielded a volume flow rate more than
double the injected amount, which they attrib-

679

uted to difficulties with the software calculation

algorithm. In light of the above, it would seem
that volume flux measurements should be
treated carefully, and used in qualitative fashion
rather than quantitatively.
PLIF Uncertainties
In some respects, there are few uncertainties
associated with PLIF imaging of OH. If the
intent is to image reaction zone location, uncertainties are limited to the physical set-up of the
system and camera characteristics. If the intent
is to determine quantitatively the hydroxyl radical concentration, the sources of uncertainty
are many. The work reported herein uses PLIF
both to image the reaction zone, and to measure
the mean and peak OH concentration, and
hence a detailed uncertainty assessment is warranted.
As discussed previously, the induced fluorescence at low excitation levels is proportional to
the overlap integral between the laser lineshape
and the absorption lineshape of the molecule
being excited, the coefficient of spontaneous
emission, and the quench rate, as well as the
ground state population of the molecules being
excited.
The overlap integral between the laser pulse
and the absorption linewidth can be determined
with a reasonable degree of accuracy. The laser
linewidth is typically measured by using an
etalon, which projects an annular fringe pattern
whose fringe spacing and width can be used to
determine the linewidth. In this work, the measured linewidth was found to be 0.20 0.25
cm21. Uncertainties in this measurement are
mainly due to the limited resolution of the
ICCD camera used to resolve the width and
spacing of the fringe pattern. This uncertainty is
estimated to be approximately 20%. However,
for a typical OH measurement at 1500K, an
uncertainty of 20% on laser linewidth results in
a total uncertainty of only 10% for the combined lineshape overlap and spectral irradiance
terms. This is not a general case, but arises
because the laser line width is of the same order
as the absorption line width. Hence, an overestimation of laser linewidth leads to an underestimation of the spectral irradiance, and an overestimation of the overlap integral. It should also

680
be noted that there is a temperature dependence of the absorption lineshape due to Doppler and pressure broadening. In the range of
1000 to 2000K, the Doppler linewidth ranges
from 0.20 to 0.27 cm21, and the pressure broadened linewidth ranges from 0.11 to .08 cm21 and
the resulting overlap integrals range from 0.84
to 0.68 E/Dn respectively. Hence, any uncertainty in temperature can lead to some uncertainty in the overlap integral.
Another uncertainty in the laser fluence
arises from the laser itself in terms of shot to
shot variation of total and point laser power.
Since laser power measurements were made
with a time-averaging instrument, and LIF images were made using a single laser pulse, it was
necessary to assess the variability of the laser
power on a shot to shot basis. This was accomplished by imaging the light scattered by the
laser sheet when directed onto a smooth surface, and recording the total intensity count of
all pixels in the sheet image. The variation of
this total count on a shot-to-shot basis represents the variation of total sheet energy, again
on a shot-to-shot basis, which was found to be
approximately 10%. A similar quantification,
looking at a single pixel in the sheet image,
showed a 15% variation, suggesting some variability in laser light distribution across the sheet
on a shot-to-shot basis.
Some uncertainties are associated with the
energy distribution in the laser sheet across its
width and along its height. As long as the
intensity distribution across the width of the
sheet does not result in localized partial saturation, the resulting fluorescence is linear with
intensity and the total fluorescence arising from
an excited volume in space is related to the total
laser energy exciting that volume, regardless of
energy distribution. The absence of significant
localized saturation was verified by measuring
the fluorescence intensity in a steady flame
while increasing laser power, and ensuring that
fluorescence intensity increased linearly. Variations in intensity along the height of the sheet
were approximately 610%, and were compensated for in the data processing stage.
Perhaps the most significant source of uncertainty in the estimation of OH concentration is
the value used for the quenching rate, Q. The
quench rate is dependent on knowledge of the

J. A. FRIEDMAN AND M. RENKSIZBULUT

concentration and collisional cross-section of
each species present in the region of interest, as
well as the temperature. Particularly in the
reaction zone itself, a region of non-equilibrium, this is not well known at all. Some experimental measurements have been made of the
quench rate in various types of flames, as well as
some theoretical predictions. For methanol, the
quench rate has been reported to be (5 6 1) 3
108 s21 at 2000K [12, 26]. As with the absorption
lineshape, the quench rate is temperature dependent, hence uncertainties in temperature
translate into uncertainties in quench rate. Garland and Crosley [31] summarized the state of
the art as . . . OH quenching can be estimated
to within 30 50% in many cases . . . .
The spectroscopic constant for spontaneous
emission Aij is well known and is not a major
source of uncertainty. Copeland et al. [30] reported values for the transition probabilities
with uncertainties on the order of 9% for the
transitions of interest in this work.
In addition to the above uncertainties associated with fluorescence emission, there are other
possible sources of error. One source of error
not considered in the data analysis is effects
resulting from excitation beam absorption
through the flame, and fluorescence reabsorption prior to arrival at the collection optics. The
first effect, arising from laser absorption from
OH and outscattering by fuel particles as the
laser sheet passes through the flame, was assessed by measuring the beam energy before
and after the flame, and was found to be
negligibly small. The effect of fluorescence trapping was not measured, but would be expected
to be of the same order as beam absorption or
somewhat larger, due to the larger Einstein
coefficients for (0,0) and (1,1) fluorescence.
The other major source of uncertainty associated with fluorescence measurements is in the
fluorescence detection system, the ICCD, and
associated optics. The uncertainties associated
with the ICCD camera depend on the signal
being measured. For low level signals, the uncertainty is related to photon shot noise, readout noise, and dark charge buildup. Photon shot
noise is related to the statistics of photon counting, described by a Poisson distribution. The
readout noise is related to the ICCD electronics
and the uncertainty associated in measuring the

METHANOL SPRAY FLAME INVESTIGATION

electron count of the pixel well. Dark charge
build-up is related to the slow accumulation of
electrons in the pixel well in the absence of
incident signal, which can be minimized by
cooling the ICCD.
The total readout noise of the ICCD is typically around ten electrons, or, at a gain setting
of 1 count per photoelectron, ten counts. The
dark charge build-up during a typical 100 ms
exposure, with the ICCD cooled to 235C is
approximately one electron, or one count at the
above gain setting. The typical incoming signal
produces a peak count reading on the order of
3000 counts, and the associated shot noise is the
square root of the signal, or 55 counts. Hence,
for the typically high signal levels encountered
in this work, the predominant source of uncertainty of the above is shot noise, at approximately 2%, or an SNR of 50. The corresponding
OH concentration at this signal level in the
configuration used in this work was approximately 5400 PPM. The lowest OH concentration measurable with reasonable accuracy
would be approximately 20 PPM, with a shot
noise limited SNR of around 3.
Perhaps the most significant source of error
associated with the fluorescence detection system with respect to quantifying peak OH concentrations as presented in this work arises from
the 12% pixel to pixel non-uniformity arising
from the image intensifier. There is a 12%
variation in output signal for a uniform input
signal. Since the system was calibrated on the
basis of a single proportionality constant between incident photon flux and pixel intensity
count for all pixels, no attempt was made to
calibrate each individual pixel. Hence, any photon count derived from a given pixel has a 12%
uncertainty associated with it. In addition, the
technique employed to calibrate the system
used the sum of intensity count over a few pixels
to determine the above-mentioned proportionality constant, and hence the uncertainty associated with this count would introduce a bias
error estimated at around 5%.
It should be noted that many other possible
sources of error have been removed by the
technique employed to calibrate the system. For
example, bias errors associated with the power
meter used to measure laser energy, which can
be large, are eliminated since the same instru-

681

ment is used for Rayleigh calibration and LIF

imaging. Geometric errors as well as optical and
quantum efficiency uncertainties are also eliminated in the same fashion, as the same geometric and optical set-up is used for calibration and
OH imaging. The only optical factor not calibrated in this fashion is the interference filter
efficiency at the various fluorescence wavelengths, and these are supplied by the manufacturer to a high degree of certainty.
Combining all the above errors, the uncertainty associated with OH concentration at a
given point is on the order of 50%, quite a high
level of uncertainty. Although the accuracy of
the concentration measurements can be improved by more precise calibration, the major
source of error is uncertainty associated with
the quench rate.
Temperature Measurement Uncertainties
Uncertainties associated with thermocouple
measurements have been quite well documented, as outlined in [27]. However, methods
of quantifying and minimizing these uncertainties are less well developed. The uncertainties
associated with thermocouple measurements
can be divided into two broad categories: timeresponse uncertainties and thermal uncertainties. Time response uncertainties are associated
with the thermal response time of the thermocouple junction to changes in medium temperature, while thermal uncertainties are associated with heat loss/gain effects such as radiation
and conduction losses from the thermocouple
junction, as well as catalytic effects on the
thermocouple surface. In measurements of
mean temperature, as presented in this work,
the second class of uncertainty is dominant.
In order to compensate a time-averaged measurement for conduction and radiation losses,
many terms must be known. The convection
coefficient, dependent on local fluid properties,
velocity, and shape and dimensions of the junction, can be estimated using correlations. Temperature gradients along the leadwires, which
must be evaluated to assess conduction losses,
depend on local flow conditions as well as
convection coefficients and are difficult to assess. Radiation measurements depend on
knowledge of temperatures of all participants

682
temperatures and relative importance. In some
cases, such as lean combustion of non-sooting
fuels, radiation can be considered as radiant
exchange between the thermocouple junction
and the far surroundings, neglecting exchange
between hot gases. However, in other flames
where particulate loadings are high, or where
hot gases may be participating, these losses are
extremely difficult to quantify. Attya and
Whitelaw [28] estimated that the radiation correction is less than 5% for an 80 mm thermocouple wire where the burning mode of the
spray flame was primarily a diffusion flame.
Uncertainties and errors associated with fuel
droplet strikes on the thermocouple junction
itself, or on the lead wires near the junction, do
not appear to have been quantified. Conduction
losses can be minimized by using long leadwires,
and orienting the thermocouple in the flow to
minimize temperature gradients in the vicinity
of the junction.
Although the effect is not well quantified,
catalytic effects involving reactions at the thermocouple surface can produce large errors.
Bare platinum thermocouples are particularly
susceptible, since platinum is a particularly effective catalyst for a large number of reactions.
Pita and Nina [29] investigated the catalytic
effect in a premixed propane/air flame, as well
as in a 20C hydrogen/air non-reacting flow, and
concluded that substantial errors associated
with catalytic effects are present in radical-rich
flame zones. The magnitude of the error was
not quantified, depending strongly on local
composition. It was also found that, within the
ranges tested, the effect was independent of
thermocouple size and local flow velocity.
In this work, severe catalytic action was initiated upon shut-down after operating with an
annular air flow rate of 4.77 l/s, a condition
which caused heating of the bluff body due to
flame impingement. On shut-down, methanol
left in the nozzle vaporized and emerged from
the nozzle as a vapour jet which caused sufficient catalytic action on contact with the thermocouple that it was seen to glow. Turning on
the fuel spray and directing it at the thermocouple enhanced this reaction, and caused a temperature reading over 1600K, Hence, in this
case, the error induced by the catalytic effect is
on the order of 1300K when the fuel spray is at

J. A. FRIEDMAN AND M. RENKSIZBULUT

300K, or 430%! It should be noted that an
attempt to induce this catalytic reaction using a
type K (Ni-Cr/Ni-Al) thermocouple was unsuccessful, even though nickel is itself a good
catalyst. Although there was no evidence of
catalytic effects of this degree in the reacting
flow, measurements made in regions of high
fuel vapour concentration and in reaction zones
must be considered suspect. Measurements
made in post-combustion regions are likely free
of catalytic effects.
It would appear that uncertainties associated
with conduction and radiation losses dominate
uncertainties in temperature measurements in
regions of the flow where catalytic effects and
droplet strikes are unlikely, and are on the order
of 5%, while in regions where catalytic effects
and/or droplet strikes are likely, uncertainties
are much higher, and unquantified.
SUMMARY AND CONCLUSIONS
The photographs presented in Figs. 3 6 provide
a good overall picture of the flame and show a
strong influence of the annular air jet on the
flame structure. Overall, the length of the flame
is seen to shorten considerably with increasing
annular air flow rate, and the flame structure
alter considerably from a two reaction zone
system to a single reaction zone system. Evidence of flame impingement on the bluff body
surface is seen in Fig. 4. It is also evident that
some of the fuel spray escapes the flame sheath
in all cases, although it appears that the annular
air jet may be redirecting some of the escaping
spray in a direction parallel to the flame, keeping fuel vapour in a region that may be reentrained further downstream.
The PLIF images, Figs. 6 11, provided good
detail on both the time-averaged and instantaneous structure of the reaction zones, and confirmed the presence of an inner and outer
reaction zone system for the annular air-off
case. Increasing annular air flow rates produced
a single reaction zone tulip-shaped structure in
the region near the nozzle, which appeared to
become more steady with increasing annular air
flow rates. Symmetry of the reaction zone in the
region near the nozzle was strong, but instantaneous symmetry in the far field tended to dis-

METHANOL SPRAY FLAME INVESTIGATION

appear with increasing annular air flow rates,
although time-averaged symmetry was still
present.
Quantitative OH concentration measurements obtained from the PLIF images showed
that peak time-averaged OH concentrations
corresponded well with the mean temperature
peaks, although a temperature peak in and of
itself did not necessarily indicate a region of
high OH concentration. Peak instantaneous OH
concentrations were on the order of 5400 PPM,
consistent with other workers measurements in
similar flames. The peak OH concentration did
not appear to vary with annular air flow rate.
Measurements of droplet size showed that
the Sauter mean diameter increased monotonically with increasing radial position in all
planes. In the regions downstream of the nozzle,
no large difference in the droplet diameter
distribution with annular air flow rate is observed, except for the air 5 4.77 l/s case, where
quite different distributions are evident, due to
the flame impingement phenomenon discussed
earlier.
Volume flux measurements were made in
several planes for all annular air flow rates
tested. In the regions 40 mm downstream of the
nozzle and further, a volume flux peak on
centreline was observed for the annular air-off
case. This centreline peak disappeared with
increasing annular air flow rate. Integration of
the volume flux curves suggested that the air-off
case resulted in the fastest evaporation of the
fuel, but this was likely due to the lack of
re-entrainment velocity and settling of droplets
due to gravity.
Mean velocity maps of the gas-phase flow as
well as the large droplet field showed that the
annular air jet produced some recirculation in
the region near the nozzle, outside the spray
cone. This recirculation entrained small droplets and returned them to the hot region where
the flame was stabilized. Increasing annular air
flow rates also produced a channeling effect in
the downstream region, with both gas-phase and
large droplet vectors directed towards the flow
centreline, suggesting a confinement effect.
Temperature measurements made by using
thermocouples showed that peak mean temperature decreased with the annular air flow rate.
The location of temperature peaks tended to

683

coincide with reaction zone location except for

the centreline temperature peaks, where PLIF
measurements showed low OH concentrations
and hence no reaction zone.
Overall, the annular air jet appears to shorten
the spray flame considerably and confine the
spray by directing the flow to centerline. These
attributes are desirable in space-limited combustion chambers where compact, clearly defined flames are necessary.
This work was made possible by a research
grant to Prof. M. Renksizbulut from the Natural
Sciences and Engineering Research Council of
Canada and by a post-graduate scholarship from
the Government of Ontario to Dr. J.A. Friedman.
The authors also thank Prof. E. Weckman for
providing the PLIF system used in the present
work and Prof. J. Hepburn for technical advice
related to PLIF.

REFERENCES
1.
2.
3.
4.
5.
6.

7.
8.

9.
10.

11.

12.
13.
14.
15.

Chigier, N., Atomization Sprays 3:365 (1993).

Kohse-Hoinghaus, K., Prog. Energy Combust. Sci. 20:
203 (1994).
Li, X., Tankin, R. S., Combust. Sci. Technol. 64:141
(1989).
Friedman, J. A., and Renksizbulut, M., Part. Part. Syst.
Charact. 11:442 (1994).
Hardalupas, Y., Liu, C. H., and Whitelaw, J. H.,
Combustion Science and Technology 97:157 (1994).
Durst, F., Zare, M., Proceedings of the Laser Doppler
Anemometry Symposium, Hemisphere, Washington,
D.C., 1977, p. 403.
Bachalo, W. D., Applied Optics 19:363 (1980).
Saffman, M., Buchhave, P., and Tanger, H., Second
International Symposium on Applications of Laser Anemometry to Fluid Mechanics, Lisbon, 1984, p. 8.1.
Dieke, G. H., and Crosswhite, H. M., J. Quant. Spectrosc. Radiat. Transfer 2:97 (1962).
Allen, M. G., and Hanson, R. K., Twenty-First Symposium (International) on Combustion, The Combustion
Institute, Pittsburgh, 1986, p. 1755.
Eckbreth, A. C., Laser Diagnostics for Combustion
Temperature and Species (2nd Ed.) Overseas Publishers Association, Amsterdam, the Netherlands, 1996.
Allen, M. G., McManus, K. R., Sonnenfroh, D. M.,
and Paul, P. H., Applied Optics 34:6287 (1995).
Melles Griot, Spectral Data-313FS10-50-01 Interference
Filter, Melles Griot, Ottawa, Ontario, 1995.
Luque, J., and Crosley, D. R., Applied Phys. B 63:91
(1996).
Drake, M. C., Pitz, R. W., Lapp, M., Fenimore, C. P.,
Lucht, R. P., Sweeney, D. W., and Laurendeau, N. M.,

684
Twentieth Symposium (International) on Combustion,
The Combustion Institute, Pittsburgh, 1984, p. 327.
16. Smooke, M. D., Xu, Y., Zurn, R. M., Lin, P., Frank,
J. H., and Long, M. B., Twenty-fourth Symposium (Intl)
on Combustion, The Combustion Institute, Pittsburgh,
1992, p. 813.
17. Bilger, R. W., Joint Intl Conference Australia/New
Zealand and Japanese Sections, The Combustion Institute, 1989.
18. Keyes, D. E., and Smooke, M. D., Comb. Sci. Technol.
67:85 (1989).
19. Dean, J. A., Langes Handbook of Chemistry 14th
Edition, McGraw Hill, New York, p. 5.130, 1992.
20. Schneider, M. and Hirleman, E. D., Applied Optics
33:2379 (1994).
21. Naqwi, A. A., and Durst, F., Part. Part. Syst. Charact.
8:245 (1991).
22. Naqwi, A. A., and Durst, F., Part. Part. Syst. Charact.
9:66 (1992).
23. Ceman, D. L., OHern, T. J., and Rader, D. J., ASME
Fluids Engineering Division FED V.95, ASME, 1990, p.
61.
24. OHern, T. J., Rader, D. J., and Ceman, D. L.,
Proceedings of the ASME Fluid Measurements and
Instrumentation Forum, San Diego, 1989, p. 49.
25. McDonell, V. G., and Samuelsen, S., Liquid Particle

J. A. FRIEDMAN AND M. RENKSIZBULUT

26.
27.

28.
29.

30.
31.

32.
33.
34.

Size Measuring Techniques: 2nd Volume, ASTM STP

1083, 1990, p. 170.
Paul, P. H., J. Quant. Spectrosc. Radiat. Transfer 51:511
(1994).
Baker, H. D., Ryder, E. A., and Baker, N. H., Temperature Measurement in Engineering Omega Engineering,
Ltd., Stamford, CT, 1975.
Attya, A. M., and Whitelaw, J. H., ASME paper No.
81-WA/HT-47, 1981.
Pita, G. P. A., and Nina, M. N. R., International
Congress on Instrumentation in Aerospace Simulation
Facilities, IEEE Publication 89CH2762-3, 1989.
Copeland, R. A., Jeffries, J. B., and Crosley, D. R.,
Chemical Phys. Lett. 138:425 (1987).
Garland, N. L., and Crosley, D. R., Twenty-first Symposium (International) on Combustion 1986, The Combustion Institute, Pittsburgh, 1986, p. 1693.
Tsujishita, M., and Hirano, A., Applied Physics B
62:255 (1996).
Joklik, R. G., Combust. Sci. Tech. 87:109 (1992).
Atkins, P. W., Physical Chemistry, 2nd Edition Oxford
University Press, 1982.

Received 19 June 1998; revised 2 September 1998; accepted 24

September 1998