Combustion Science and Technology

ISSN: 0010-2202 (Print) 1563-521X (Online) Journal homepage: https://www.tandfonline.com/loi/gcst20

AIR DENSITY EFFECT ON THE ATOMIZATION OF

LIQUID JETS IN CROSSFLOW

A. BELLOFIORE , A. CAVALIERE & R. RAGUCCI

To cite this article: A. BELLOFIORE , A. CAVALIERE & R. RAGUCCI (2007) AIR DENSITY EFFECT
ON THE ATOMIZATION OF LIQUID JETS IN CROSSFLOW, Combustion Science and Technology,
179:1-2, 319-342, DOI: 10.1080/00102200600809563

To link to this article: https://doi.org/10.1080/00102200600809563

Published online: 25 Jan 2007.

Submit your article to this journal

Article views: 588

View related articles

Citing articles: 11 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=gcst20
Combust. Sci. and Tech., 179: 319–342, 2007
Copyright Q Taylor & Francis Group, LLC
ISSN: 0010-2202 print/1563-521X online
DOI: 10.1080/00102200600809563

AIR DENSITY EFFECT ON THE ATOMIZATION OF

LIQUID JETS IN CROSSFLOW

A. BELLOFIORE
A. CAVALIERE
Dipartimento di Ingegneria Chimica, Università
Federico II, Napoli, Italy

R. RAGUCCI*
Istituto di Ricerche sulla Combustione, CNR,
Napoli, Italy

An experimental study of liquid injection in a transverse airflow at

high pressure and temperature was performed. Collected shadow-
graphs were processed by image statistical analysis tools. The break-
down point location was correlated to liquid-to-gas momentum
ratio, gas Reynolds number and aerodynamic Weber number. The
substantial uniformity of behavior of the normalized jet trajectories
allowed providing an empirical correlation validated in a range of
operating conditions of interest for LPP gas turbines. Liquid disper-
sion in the gas phase, atomization and evaporation were qualitatively
investigated by means of statistical parameters related to the beha-
vior of the spray plume. For liquid spraying and spreading, the rel-
evance of the interaction between aerodynamic and capillary
pressure was underscored, whereas in the range of operating con-
ditions studied, both experimental evidence and theoretical esti-
mation indicate that even at 600 K temperature the evaporation of
water drops is not significant within 100 diameters downstream
the injection point.

Keywords: atomization, crossflow, high-temperature–high pressure,

jet breakdown

Received 21 July 2005; accepted 13 January 2006.

Address correspondence to ragucci@irc.cnr.it

319
320 A. BELLOFIORE ET AL.

INTRODUCTION
Increase of efficiency and pollutant emission reduction are the main
guidelines in the development of combustion systems. In particular the
technology of gas turbines is raising interest, due both to the increasingly
volume of air transportation traffic and the massive resorting to gas tur-
bines in power production plants. In this technological frame of refer-
ence, the requirements are to put up the air pressure, in order to
improve the thermodynamic efficiency and contain the engine sizes,
and reduce the temperature peaks inside the combustor and hence the
formation of NOX. A promising way to match these requirements is
the Lean Premixed (and Prevaporized, in case of liquid fuel) combustor.
The LP concept exploits operating conditions near the lean extinction
limit, using the thermal inertia of large amounts of air to lower the level
of temperature inside the combustor, and pushes towards the require-
ment of a uniform air=fuel mixture to avoid temperature peaks.
The ability to produce, in a suitable premixing system, a mixture as
uniform as possible at the inlet section of the combustor is a critical step
in the development of efficient and reliable LPP gas turbines. In fact, the
realization of such a goal should be of great help to overcome problems
as flashback, self-ignition and combustion instabilities while ensuring a
good emission performance of the combustion system. For liquid fuel
gas turbines, in the premixing duct, the processes of atomization, evap-
oration and dispersion of the liquid phase in the gaseous means have to
be completed in time ranges as shorter as air pressure grows and ignition
delay goes down. Liquid injection in crossflow is an excellent way to pro-
duce a uniform mixture within the time constraints, by exploiting the
availability of large amounts of high-pressure, high-temperature air com-
ing from the compressor. This result can be achieved by means of a
simple and reliable system based on one or more plain nozzles injecting
the fuel normally to the airflow.
Because of the use of relatively low liquid injection pressure levels
and the absence of internal mixing with air, such as for airblast atomi-
zers, the liquid flows out of the nozzle as a compact column, initially
characterized by the presence of small disturbances on its surface, due
to turbulence or cavitation inside the injector or simply to the relaxation
of the liquid velocity profile when the boundary condition of adherence
to the nozzle wall suddenly lacks. Consequently, the interaction between
liquid and airflow is schematically referable to the so-called wind-induced
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 321

or turbulent regime rather than the atomization regime (Lefebvre, 1989).

The superimposition of aerodynamically enhanced waves spreads the
wave frequency over a wide range. The further effect of the high air den-
sity is to increase the growth rate of instabilities and also to produce a
quite strong Eulerian buckling load, sustaining the development of asym-
metric instabilities, even more amplified by the transverse energy transfer
from the airflow. The complex propagation of these interface oscillations
couples with the global bending of the liquid column due to the drag
force exerted by the air crossflow.
Schetz et al. (1980) hypothesized that the velocity variation of liquid
windward surface due to the bending can also be responsible of the
development of acceleration waves. As a result the pattern of wave
propagation over the jet surface loses any regularity, showing a compli-
cated three-dimensional structure (Inamura and Nagai, 1997). The wave
amplitude growth promotes the formation of ligaments protruding from
jet surface. The length of ligaments has been proven to depend on liquid
viscosity that delays their rupture (Nejad and Schetz, 1983). The liga-
ment cross-sectional size determines the initial sizes of liquid fragments
detaching from the jet basically by the Rayleigh breakup mechanism (Wu
and Faeth, 1993). The size distribution of liquid fragments detached
from ligaments is expected to be quite wide, spread as the wave frequency
range. In case of high air density, the stronger aerodynamic effects
induce a merging of Rayleigh and secondary breakup, resulting in an
overall shifting of the size distribution toward lower values (Wu and
Faeth, 1993).
The study of liquid jets in crossflow has to face several problems.
The first difficulty is the fact that the process cannot be reduced to a
one-dimensional scheme, so that the liquid bending induces a greater
complexity than the injection in still or co-flowing gas. Moreover, the tur-
bulent regime places midway between the extreme cases of Rayleigh
breakup and fully developed atomization, in a range of relative velocity
conditions have not been studied extensively. The lack of understanding
of the mechanisms characterizing this regime reflects on the lack of mod-
els for the atomization in the turbulent regime. The non-linear atomi-
zation models, developed in the last years, managed to describe liquid
dynamics and breakup in the wind-induced regimes (Eggers, 1997; Span-
gler et al., 1995; Yoon and Heister, 2004), but only under very controlled
conditions, faraway from the framework of this paper. On the opposite
side in the fully developed atomization regime the consistent hypothesis
322 A. BELLOFIORE ET AL.

that the liquid jets collapses just after the nozzle outlet, due to the high
relative velocity, allowed the successful development of blob models
(Reitz, 1987), which should be considered unsuitable for the turbulent
regime and yet are largely adopted for crossflow atomization.
Another approach to overcome the lack of comprehension of this
process is to presume the existence of an analogy with the injection of
gas jets in an air crossflow (Heister et al., 1989; Nguyen and Karagozian,
1992; Tambe et al., 2005), even though it is proved that submerged jets
are controlled by diffusivity and concentration gradients, while in the
case of liquid injection, the controlling parameter is the surface tension
sustaining the existence of a phase discontinuity at the interface. One of
the reasons why the process of liquid injection in crossflow is poorly
understood is the difficulty of experimentally studying the dynamics
and breakup of the jet, because of the occurrence of the atomization pro-
duces large fragments and a dense aerosol of drops, which envelop the
continuous jet in the near field. This fact creates serious problems for
diagnostic investigation, based on the collection of light scattering or
extinction signal from liquid interface. In fact, the dense drop cloud that
interposes between jet and observer optically obscures the liquid column.
As a matter of fact, the overall amount of interface of the drop cloud,
which is much larger than the liquid corrugated surface, results in a very
high scattering efficiency that masks any scattering signal coming from
the liquid column interface. For the same reason, diagnostics based on
the detection of the extinction signal, like the shadowgraphic techniques,
can only furnish a picture of the external plume of drops surroundings
the jet. These problems appear to be even more important for high air
density conditions, where the onset of fragment detachment takes place
closer to the nozzle outlet, preventing the observation of the liquid jet
at all (Cavaliere et al., 2003; Ragucci et al., 2000, 2003; Ragucci and
Cavaliere, 2002). Consequently, up to date information about jet actual
behavior is really exiguous, except in some particular cases where oper-
ating conditions or liquid properties reduce the atomization effectiveness
to a minimum, allowing direct observation of the liquid jet (Inamura and
Nagai, 1997; Kihm et al, 1995; Wu et al., 1997, 1998).
For the near field, the study of liquid jets has been usually based on
image collection by means of high-speed digital cameras (Becker and
Hassa, 1999, 2000, 2002; Schetz et al., 1980; Wu et al., 1997) and on
Mie scattering data obtained by slicing the spray with a laser sheet
(Oda et al., 1994; Ragucci et al., 2000). The use of diagnostic techniques
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 323

for flow field data or single-point drop sizing is possible only outside of
the dense spray region (Becker and Hassa, 2002; Wu et al., 1998). In this
paper, the interest is focused on the study of liquid jet trajectory and dis-
ruption; thus, a flash shadowgraphy diagnostic setup has been adopted.
Several approaches have been suggested for the prediction of jet tra-
jectory in airstreams. Chen et al. (1993) built an empirical model of the
jet trajectory based on the assumption of existence of three different
zones (near field, ligament region and droplet region), whose evolution
needs three separate mathematical descriptions. The accurate tuning
of three exponential terms allowed for the use of a single composite func-
tional form. Reference experiments were carried out by injecting an
undisclosed fuel in an up to 0.2 MPa subsonic airflow. Wu et al. (1997)
used an even lower air pressure and injected different liquids but no fuel.
The development of a simple phenomenological analysis led to empirical
correlations pointing out that the jet trajectory well resembles a square
root behavior. Moreover, there is a dependence of the jet penetration,
in the liquid-streamwise direction, on the square root of the liquid-
to-air momentum ratio q. A similar dependence on q was also proposed
by Chen et al. (1993) and Becker and Hassa (1999, 2002), even though
the exponent of q was found to be slightly lower than 0.5. The latter
authors also proposed a logarithmic functionality to fit the path followed
by the liquid jet, as did Tambe et al. (2005) recently, relying on the non-
linear regression of data collected from water, jet-A and n-heptane jets in
transverse airstream at atmospheric pressure and temperature.
Jet breakup is considered a main concept for the development of
physically consistent modeling tools for spray behavior prediction.
Nevertheless this parameter is still enveloped by a definition ambiguity,
since in the atomization process more threshold events can be connected
with the onset of discontinuities in the liquid phase. A similar abundance
of breakup conditions has been already pointed out by Pilch and Erdman
(1987) in their review on fragmentation mechanisms for liquid drops. In
the case of liquid jets in crossflow, the simplest definitions of breakup are
related to either the total penetration of the spray in the liquid-stream
wise direction (Becker and Hassa, 2002; Chen et al., 1993; Schetz
et al., 1980) or the occurrence of the first drop detachment from the
jet column. Between these two extreme conditions, great interest has
been devoted to the penetration of the liquid jet as a continuous medium,
so the breakup point is placed in the zone where the jet, already bent,
stressed and having undergone liquid stripping from its surface as
324 A. BELLOFIORE ET AL.

ligaments and drops, ultimately loses its continuity by breaking up in

large fragments.
Along with the conceptual ambiguities, the aforementioned inability
of diagnostic techniques to capture the actual jet dynamics represents a
further problem to the study of jet breakup. The substantial inaccessi-
bility of the liquid jet, mostly in the so called surface breakup regime
(Wu et al., 1997) promoted by high air density typical of LPP conditions,
strongly hardens the task of individuating a column fracture point by
means of direct inspection of spray images. Moreover even in the case
of weak atomization (column breakup regime) the practice of evaluate
the location of breakup by directly observing collected spray pictures
appears to be largely unsatisfactory since it suffers from subjectivity
and non-repeatability. Furthermore, it is necessarily bounded to the
study of very few pictures, usually less than 10, so that experimental
uncertainty due to the unsteady nature of the process is not overcome
by a statistically significant sampling. An additional source of mistakes
in the evaluation of breakup is provided by the presence of asymmetric
waves along the jet, so that the liquid column is deeply distorted and
its apparent section, such as it is captured by the observing device, can
appear remarkably thinned, at some point resembling the occurrence
of a discontinuity. Wu et al. (1997) used shadowgraph images to evaluate
the occurrence of liquid column fractures and suggested that the liquid-
streamwise coordinate of the breakup location only depends on the
liquid-to-air momentum ratio q, while the gas-streamwise coordinate is
independent of both the operating condition and the properties of the
two phases. Tambe et al. (2005) came to the same conclusion, although
they pointed out that since the jet breakup location was determined by
means of direct, non-rigorous observation of the liquid column fracture
as captured by shadowgraph images, only images of sprays in the column
breakup regime were useful to the purpose. The database in the surface
breakup regime was therefore neglected and so the proposed correlation
suffers from lack of extensive validation.
The liquid jet momentum coherence breakdown concept was firstly
introduced by Ragucci and Cavaliere (2002) as an attempt to overcome
both conceptual ambiguities and experimental uncertainties. The non-
theoretical definition is based on the implementation of an image statisti-
cal analysis procedure, which evaluates the occurrence of a sudden
collapse of jet resilience to the unsettling action of propagating interface
waves. The present work presents a fully automatic, and then repeatable,
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 325

routine developed to evaluate the occurrence and location of the coher-

ence breakdown point. A further advantage is the possibility of using
a suitably large image sample, reducing uncertainties due to statistical
fluctuations.
The developed statistical procedures have already been successfully
used to study the behavior of jet-A1 and water jets in air crossflow at
0.1 MPa and room temperature. Extensive ranges of air and liquid velo-
city were explored using a flash shadowgraphy technique. This paper
aims to extend those results to higher air pressure and temperature,
resembling conditions close to the ones of a gas turbine premixing duct.
To evaluate the dynamics and disruption of the liquid jet, the results pre-
sented here concern the coherence breakdown point distances, in both
gas and liquid-streamwise direction, from the injection point. These
two parameters are shown to be dependent on operating conditions
and surface tension in a quite simple way. With respect to the previous
set of data, the dependence on air density is accounted for. In addition
the availability of data at higher temperature indicates a slight influence
of air viscosity on jet penetration. The breakdown point coordinates are
used as effective normalizing factors for jet trajectory allowing for writ-
ing them in a very simple mathematical form. With respect to atomi-
zation and spray placement in the premixing duct, this paper presents
results on two statistical indexes introduced to evaluate the extent of
the spray plume in the plane normal to the optical path of the diagnostic
setup and the dispersion of the spray in the liquid-streamwise direction.

EXPERIMENTAL SETUP
An experimental facility designed to reproduce geometry and operating
conditions of the premixing channel of a LPP gas turbine engine, already
described (Ragucci et al., 2004), was used to study the evolution of a
liquid jet when injected orthogonally to an airflow. The test rig consisted
of a fully accessible chamber with a square cross-section of 25  25 mm,
capable of resisting high pressures, up to 10 MPa, and high temperatures,
up to 1000 K. The test section was designed so that three of its walls are
quartz windows that ensure the observation of the whole channel. On the
fourth side, a plain nozzle, with a recessed hole of 500 mm, was mounted
with the axis normal to the channel one. A 45 taper introduces the liquid
flow to the terminal straight section of the nozzle having an L=D ratio
equal to 4. Some authors cared about reducing to a minimum liquid
326 A. BELLOFIORE ET AL.

turbulence and unsteadiness, by accurately designing a laminar flow in a

supercavitating nozzle (Sallam et al., 2004; Wu et al., 1997). That way
they expect to have very little instability at the nozzle outlet, mainly
due to the relaxation of liquid velocity profile out of the injection tube,
so that it should be possible to study the pure crossflow atomization
mechanism. In this work liquid is injected in turbulent conditions com-
parable to the operation of gas turbines. The liquid was supplied to the
nozzle by means of a nitrogen-pressurized vessel and regulated by a digi-
tal pressure control valve.
The diagnostic setup was a simple shadowgraphic scheme using a
low-pressure xenon flash lamp, with a light pulse duration of 15 mS,
and an 8-bit digital camera with maximum resolution of 640x480 pixels
and minimum shutter time equal to 30 ms. For each test condition a
1,000-frame sample was collected. For each measurement a background
image was collected and then used to clean up the spray images, reducing
the electronic noise. The spray windward profile, assumed as representa-
tive of the compact jet trajectory, was extracted by applying standard
image morphological algorithms based on the evaluation of the steepest
light dampening gradients. The statistical evaluation of the local fluctua-
tions along the jet was exploited to individuate the occurrence of jet
breakdown (Ragucci and Cavaliere, 2002; Ragucci et al., 2004). Further
image analysis procedures were implemented to study the behavior of the
spray plume. This paper investigates two statistical parameters; normal-
ized plume width and spray extent. The former definition refers to the
maximum spread of the spray cloud in the liquid-streamwise direction,
averaged on the 1,000 frames sample and normalized with respect to
the observable channel width, which is about 22.5 mm, due to the stray
light coming from the edges of the quartz window. The spray extent is
the area, expressed in mm2, where the light dampening due to the
interposition of the spray is more intense than a cutoff threshold, chosen
as 20%. The various statistical parameters introduced can be seen in
Figure 1, which is a sketch of a typical spray, as observed by means of
the flash shadowgraphy.

TEST CONDITIONS
Tests have been performed in the above-described facility at two refer-
ence values of air temperature and pressure. Air pressure was set at
1.0 and 2.0 MPa, as measured by an electronic transducer placed just
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 327

Figure 1. Sketch of the liquid jet in crossflow. The statistical parameters evaluated and dis-
cussed in this paper are pointed out.

upstream of the liquid injection point. Air temperature, measured by a

thermocouple placed near the pressure transducer, was set at reference
values of 300 and 600 K by means of an electric heater and a PID con-
troller. Air flow rate was varied by regulating the area of the sonic throttle
placed downstream the test section. Air velocity was measured by means
of a Pitot tube and it ranged between 20 and 60 m=s. Test liquids were
water and jet-A1. Their properties of interest are summarized in Table 1.
In particular it must be noted that surface tension was about three times
lower for jet-A1, covering the range of values typical of atomization pro-
blems. Liquid velocity was evaluated from the pressure drop at the nozzle
and the discharge coefficient of the atomizer, experimentally assessed as
0.69. Liquid velocity ranged between 15 and 55 m=s.

Table 1. Relevant properties of test liquids

Liquid Density [Kg=m3] Viscosity [Kg=m s] Surface tension [N=m]

Kerosene 790 0.0015 0.027

Water 1000 0.001 0.073
328 A. BELLOFIORE ET AL.

Table 2. Definition of the dimensionless parameters of interest

q V2
Liquid-to-air momentum ratio q ¼ qL VL2
G G

Gas Reynolds number ReG ¼ qGlVGG D

qG VL2 D
Aerodynamic Weber number Weaero ¼ r

The results presented next refer to some dimensionless parameters

that were found to influence the behavior of the liquid jet and the spray
plume. Table 2 reports the definitions of these parameters. q, l and r are
the symbols adopted for density, viscosity and surface tension; V for velo-
city, D is the diameter of the nozzle outlet and the subscripts G and L are
for the gas and liquid phases, respectively.
The entire database used here consists of about 90 experimental
conditions. Part of these measurements were collected previously and
presented by Ragucci et al. (2004). The newly collected data are listed
in Table 3, where for each test condition the operating parameters along
with dimensionless parameters are reported.

RESULTS AND DISCUSSION

Results are presented here as they were evaluated by means of flash sha-
dowgraphy measurements and image statistical analysis procedures.
Data include jet breakdown and trajectory and spray ensemble features
such as spray extent and plume width. For jet breakdown, for each test
condition, the point where the liquid column is assumed to lose its coher-
ence is evaluated in terms of liquid and gas-streamwise coordinates,
respectively indicated as zjb and xjb . In Figure 2, zjb is plotted as a func-
tion of the liquid-to-air momentum ratio q. On a logarithmic scale,
experimental points appear aligned and are seen to increase less than lin-
early with q. Nevertheless the trend line calculated on the whole set of
data (not reported in Figure 2) has a Pearson correlation coefficient
lower than the trend line evaluated on measurements taken in cold con-
ditions. Consequently, a dependence was supposed on the gas Reynolds
number ReG, accounting for the increase of air viscosity with tempera-
ture. ReG ranges from about 5,000 to 26,000 in the data set presented
and it has been varied by acting on gas density, velocity and viscosity.
In Figure 2 data for jet-A1 and water were divided into two sets, by
choosing the threshold ReG ¼ 13,000 because in this way all the mea-
surements at 600 K range in the same group, i.e., below the threshold
 Table 3. Experimental test conditions

Nozzle Air reference Air reference Liquid Liquid-to-Air Gas

diameter pressure temperature Air velocity velocity momentum Aerodynamic Reynolds
Liquid [mm] [MPa] [K] [m=s] [m=s] ratio Weber number number

Water 0.5 1.0 300 24.3 18.7 49.2 28.9 7932

Water 0.5 1.0 300 29.1 19.3 36.9 30.5 9406
Water 0.5 1.0 300 29.6 23.6 52.8 46.0 9662
Water 0.5 1.0 300 35.1 24.3 40.6 47.8 11233
Water 0.5 1.0 300 35.8 30.5 61.2 75.7 11514
Water 0.5 1.0 300 38.5 23.7 32.2 45.4 12297
Water 0.5 1.0 300 39.0 26.6 39.1 57.7 12569
Water 0.5 1.0 300 39.9 34.4 59.6 101.3 13495
Water 0.5 1.0 300 54.2 33.1 31.7 88.5 17294
Water 0.5 1.0 300 55.4 40.0 45.2 126.6 17323
Water 0.5 1.0 300 57.0 46.3 59.0 164.5 17278
Water 0.5 1.0 300 62.8 43.9 48.8 132.3 17032
Water 0.5 1.0 300 63.7 47.5 60.5 142.1 15854
Water 0.5 2.0 300 28.8 17.9 15.5 54.9 19482
Water 0.5 2.0 300 28.9 34.6 60.0 196.3 18719
Water 0.5 2.0 300 33.6 22.7 19.0 84.9 21871
Water 0.5 2.0 300 34.0 34.1 44.0 182.2 21046
Water 0.5 2.0 300 34.2 41.7 66.5 266.8 20734
Water 0.5 2.0 300 38.8 30.8 26.8 153.2 24761
Water 0.5 2.0 300 38.9 49.2 64.7 410.5 26066
Water 0.5 2.0 300 39.0 27.3 22.9 109.3 22598

(Continued)

329
330
Table 3. Continued

Nozzle Air reference Air reference Liquid Liquid-to-Air Gas

diameter pressure temperature Air velocity velocity momentum Aerodynamic Reynolds
Liquid [mm] [MPa] [K] [m=s] [m=s] ratio Weber number number

Water 0.5 2.0 300 39.2 20.8 12.2 68.4 24503

Water 0.5 2.0 300 39.2 46.6 60.0 350.7 25016
Water 0.5 2.0 300 39.3 43.8 63.6 257.2 20816
Water 0.5 2.0 300 39.3 33.1 35.7 149.3 21167
Water 0.5 2.0 600 27.6 12.9 18.5 13.4 5250
Water 0.5 2.0 600 27.7 26.0 81.0 50.4 4840
Water 0.5 2.0 600 28.2 14.8 23.8 17.3 5310
Water 0.5 2.0 600 28.3 22.3 54.5 39.2 5280
Water 0.5 2.0 600 28.4 18.1 35.4 25.8 5310
Water 0.5 2.0 600 28.4 19.7 41.9 30.9 5350
Water 0.5 2.0 600 28.5 11.5 14.2 10.4 5310
Water 0.5 2.0 600 28.5 20.5 45.1 32.9 5320
Water 0.5 2.0 600 28.5 22.3 53.5 39.2 5330
Water 0.5 2.0 600 28.6 23.9 60.7 45.4 5390
 Water 0.5 2.0 600 28.6 25.2 67.7 49.7 5330
Water 0.5 2.0 600 37.2 36.2 71.4 119.8 9080
Water 0.5 2.0 600 37.8 32.4 56.0 93.9 8880
Water 0.5 2.0 600 38.4 19.3 20.4 31.9 8320
Water 0.5 2.0 600 39.1 20.9 22.4 38.4 8910
Water 0.5 2.0 600 39.3 25.3 34.7 52.7 7990
Water 0.5 2.0 600 39.5 15.8 12.7 21.7 8790
Water 0.5 2.0 600 40.1 25.7 33.4 55.5 8550
Water 0.5 2.0 600 40.7 31.5 52.1 78.5 7810
Water 0.5 2.0 600 40.9 32.6 52.3 88.4 8400
Water 0.5 2.0 600 51.2 21.5 12.9 43.1 12900
Water 0.5 2.0 600 51.4 28.8 24.1 74.8 12100
Water 0.5 2.0 600 52.1 36.2 39.3 110.1 11300
Water 0.5 2.0 600 52.8 45.0 60.3 167.2 11200

331
332 A. BELLOFIORE ET AL.

Figure 2. Liquid-streamwise coordinate of the breakdown point plotted against the q

number.

value. The choice of such a value of Reynolds is then purely arbitrary and
has no other scope than pointing out the effect of a variation of ReG .
The result of this segregation is that the data at lower ReG, repre-
sented by hollow circles and triangles in Figure 2, place somewhat below
the other group of point, as clearly shown by the two trend lines
reported. Anyway, the influence of the Reynolds number on zjb is not
so strong as the one of q and can be attributed to the larger drag force
of the airflow on liquid jet as gas viscosity increases. A non-linear
regression was performed on the whole data set and the result is an
empirical correlation between zjb and both q and ReG :
zjb
¼ 1:449 q0:476 ReG0:135 ð1Þ
D
In order to point out the agreement of this correlation with experi-
mental data, Figure 3 plots the behavior of zjb against the functional
group q0:476 ReG0:135. The calculated value of the Pearson correlation coef-
ficient is 0.949. Different from Figure 2, in this case the several sets of
data are clearly categorized. It is evident that both data at higher air
pressure and data at higher air temperature are well predicted by the pro-
posed correlation, which accounts for both air density and viscosity
effect on penetration.
Figure 4 shows results regarding xjb . In this case the parameter in the
abscissa is the aerodynamic Weber number. In logarithmic scale points
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 333

Zjb
Figure 3. Breakdown coordinate D as a function of q and ReG .

are quite aligned on a straight line, so a power law dependence of xjb on

Weaero can be assumed. The result of a non-linear regression on all data
is the following correlation:
xjb 0:366
¼ 3:794 Weaero ð2Þ
D
The straight line in Figure 4 represents this correlation. The agree-
ment with experimental data is good, as testified by the Pearson corre-
lation coefficient, assessed as 0.933. This behavior confirms the result
already pointed out (Ragucci et al., 2004) that xjb depends on the operat-
ing conditions, differently from what has been reported by other authors
(e.g., Wu et al., 1997 and more recently Tambe et al., 2005). The intro-
duction of data at different air pressure and temperature helped to define
the functional dependence for xjb, and so Eq. (2) can be considered a sig-
nificant improvement on the previous correlation (Ragucci et al., 2004).
The aerodynamic Weber number accounts for the interaction between
334 A. BELLOFIORE ET AL.

aerodynamic and capillary pressure in the transverse direction. The rea-

sons of this somehow unexpected dependence are not clear, and it appears
evident that the various active mechanisms operate and interact in a com-
plex way. The absence of an explicit dependence on gas velocity can be
roughly explained as suggested by Wu et al. (1997) by supposing that the
larger bending at higher air velocity is counterbalanced by a more intense
atomization that reduces the lifetime of the continuous liquid column.
Chen et al. (1993) suggested the possibility to use the breakdown
point coordinates as normalizing factor to compare jet trajectories eval-
uated at different operating conditions. The adoption of such a method
allowed Ragucci et al. (2004) to state that the normalized behavior of the
liquid jet is independent of both liquid properties and operating con-
ditions. The same behavior was observed also for data at higher air press-
ure and temperature, as shown in Figures 5a and 5b, which respectively
report the normalized trajectory for the set of measurements collected at
higher pressure and at higher temperature. Along with data at 2 MPa

Figure 4. Gas-streamwise coordinate of the breakdown point as a function of the aero-

dynamic Weber number.
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 335

Figure 5. Normalized trajectories for water at high air pressure (a) and high air temperature
measurements (b). The solid line is the average behavior, whereas the single trajectories are
plotted in grayed lines. The error bars represent a measure of the local standard deviation.

pressure, Figure 5a also contains a set of measurements on water at

1 MPa air pressure not reported by Ragucci et al. (2004). The dispersion
with respect to the average trajectory was evaluated by computing the
average standard deviation, which resulted to be 1.95% for data in
Figure 5a and 2.19% for Figure 5b.
The average trajectories in Figures 5a and 5b are plotted again in
Figure 6, along with analogous curves derived for previously collected
sets of data. All these trajectories do not depend on operating con-
ditions, and are well described by the power law already proposed by
Ragucci et al. (2004):
 0:35
z x
¼ ð3Þ
zjb xjb

The Pearson correlation coefficient was evaluated by Ragucci et al.

(2004) to estimate the agreement of Eq. (3) with average normalized tra-
jectory, and it resulted to be 0.954 for jet-A1 and 0.963 for water data.
For the new sets of data the agreement is even better; in fact, the corre-
lation coefficient is equal to 0.997 for the set labeled as water at 1 and
2 MPa air pressure at room temperature and 0.996 for water at 2 MPa
air pressure and 600 K temperature. By substituting Eqs. (1) and (2) into
Eq. (3), an empirical expression can be found for the trajectory of a
336 A. BELLOFIORE ET AL.

Figure 6. Generalized trajectory (solid line) as predicted by the empirical correlation (3),
compared with the average trajectories of the various sets of data.

liquid jet in crossflow:

z  0:35
0:128 x
¼ 0:909 q0:476 ReG0:135 Weaero ð4Þ
D D
For the spray plume ensemble features, two statistical indexes were
evaluated, i.e., the overall surface of the pixels where light extinction is
above the noise threshold, named spray extent and measured in mm2,
and the normalized plume width in the liquid-streamwise direction.
The data used for these calculations are the same used for breakdown
point computations, and other measurements for jet-A1 at 1 and
2 MPa. Figure 7 presents the normalized plume width, plotted against
the aerodynamic Weber number corrected by the square root of the den-
sity ratio. Although points appear relatively scattered, it is possible to
deduce that the filling of the premixing duct gets better as the corrected
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 337

Figure 7. Normalized plume width as a function of the aerodynamic Weber number cor-
rected by the square root of the density ratio.

Weber number increases. It is evident that a higher liquid velocity means

larger penetration in the channel, but the dependence on the Weber
number also accounts for the effect of the atomization promoting the
removal of larger amounts of droplets, thus letting the spray cover the
zone near the injection wall as well. The growth of the plume width stops
when the channel width is fully filled. The saturation of the channel, evi-
dent in Figure 7 above a value of the abscissa of about 1500, sets an
upper limit to the liquid flow rate, over which there are no further ben-
efits on liquid dispersion in the channel. Data at either higher air press-
ure or temperature do not seem to differ from the other data sets,
showing that the modification of these parameters only affects the spray
dispersion through the variation of the air density. The presence of qL in
the abscissa essentially aims to let the parameter remain dimensionless.
338 A. BELLOFIORE ET AL.

The scattering of points prevents inferring any dependence on a param-

eter such qL , which ranges in a too-narrow interval.
Figure 8 reports the behavior of the spray extent, again plotted as a
function of the aerodynamic Weber number corrected by the square root
of the density ratio. As could be expected, this parameter has a behavior
similar to the plume width, even though in this case the experimental
points are less scattered and show a lower level of saturation, due to
the fact that the spray extent accounts for both plume width and down-
stream displacement of the spray. Anyway in this case there is also a
reduction in the slope of the line suggested by the points in correspon-
dence of the abscissa value where the plume width reaches its top and
stops growing. The dependence on the corrected Weber number is not
easy to explain. Probably the presence of a squared liquid velocity is

Figure 8. Spray extent as a function of the aerodynamic Weber number corrected by the
square root of the density ratio.
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 339

attributable to the double effect of this parameter on both liquid flow

rate, increasing the amount of liquid instantaneously interposing
between the flash lamp and the digital camera, and atomization intensity,
as already observed for xjb.
Also in this case, a higher temperature does not apparently have any
other effect than reducing air density. One could expect that calculations
on spray extent would point out an additional influence of air tempera-
ture connected to the higher level of evaporation. Along with the evalu-
ation of the spray extent, a possible effect of evaporation on the spray
features has been further investigated by comparing images of water jets
at the same liquid and gas velocity and the same air density but different
temperature. The comparison of both single and average frames did not
point out any substantial difference attributable to a more intense evap-
oration at higher air temperature, as well as the comparison of the histo-
grams of the distribution of light extinction intensity. From a theoretical
point of view, the potential relevance of evaporation can be assessed by
comparing the time characteristic of evaporation with the time charac-
teristic of the fluid-dynamic transport. Starting from the point where they
detach from the liquid jet, drops travel with a velocity that can be roughly
assumed equal to the undisturbed air velocity.
Another hypothesis is that the visible path they follow is about as
long as a straight line linking the plane x ¼ 0, corresponding to the liquid
injection section with the downstream boundary of the collected image,
that is no more than 50 mm. Consequently, each drop is visible by the
camera for an order-of-magnitude time of about 1 ms. In order to esti-
mate the evaporation time, the steady state evaporation model is
assumed. As is well known, this model predicts that the drop shrinks
with a constant rate (Lefebvre, 1989), which has been assessed for water,
in air at 20 bar and 600 K, as about 0.23 mm2=s. With this value a water
drop with an initial diameter of 10 mm completely disappears in 0.43 ms,
while a drop of 100 mm lasts 43 ms. Since liquid is injected cold, this esti-
mation of the evaporation time does not account for the heating-up time,
and so is lower than the actual time required by the drop to disappear.
The agreement between experimental evidence and theoretical esti-
mation seems to justify the hypothesis that in the case of water injected
in airflow at temperature up to 600 K, drops travel too fast to signifi-
cantly evaporate within 100 diameters downstream the injection
point, i.e., before leaving the region of interest of the adopted diagnostic
setup.
340 A. BELLOFIORE ET AL.

CONCLUSIONS
The behavior of a liquid jet in crossflow was experimentally investigated,
choosing conditions comparable with the actual operation of a gas
turbine. Liquid and air velocity, liquid properties and air pressure and
temperature were varied in order to explore how some critical features
of jet and spray depend on operating conditions. The investigation was
carried out by adopting a flash shadowgraphy scheme and then post-pro-
cessing the collected images to extract information on the trajectory and
breakdown point of the jet and on the atomization and dispersion of the
liquid phase in the gaseous means.
The trajectory of the liquid jet has always the same shape, well
described by a power law functional expression, whatever the liquid
injected or the operating conditions, whereas the length and aspect ratio
of this curve scale as the location of the breakdown point. The coordi-
nates of the breakdown point were empirically correlated with some
dimensionless parameters. In particular zjb strongly depends on the
liquid-to-air momentum ratio, as already largely pointed out in literature.
The variation of air temperature allowed finding a further dependence on
the gas Reynolds number, accounting for the effect of air viscosity on jet
bending. Quite surprisingly, the stronger bending at higher ReG seems to
affect zjb but not xjb , therefore resulting in a shorter lifetime of the jet.
This could be attributable to the probable influence of ReG on the defor-
mation and flattening of the jet cross-section, so promoting a larger
liquid stripping and a quicker consumption of the jet. As discussed
before, xjb apparently depends only on the aerodynamic Weber number.
With respect to the spray plume, both the plume width and the spray
extent show a linear dependence on a modified form of the Weber num-
ber, at least for low values of this parameter. After a certain value of the
abscissa, the limited transverse length of the channel blocks the increase
of the dispersion parameter and hence reduces the growth slope of the
spray extent. For the geometry of the premixing duct adopted here the
channel saturation threshold is at about ðqL =qG Þ0:5 Weaero ¼ 1500. Above
this value, the spray starts hitting the opposite wall and the system works
in an undesired way due to the formation of a non-atomized film carrying
liquid fuel in the combustor and moreover coking on the channel wall at
high temperature. Finally pressure and temperature seem to affect the
spray behavior only as they modify the air density and therefore the
atomization level. In the range of operating conditions explored, both
AIR DENSITY EFFECT ON THE ATOMIZATION IN CROSSFLOW 341

experimental evidence and theoretical estimation indicate that even at

600 K, the evaporation of water drops is not significant within 100 dia-
meters downstream of the injection point. This result has to be further
investigated to widen its validity limits, in particular by studying the
behavior of more volatile liquids, such as jet-A1 itself, injected in high-
temperature airstreams.

REFERENCES
Becker, J. and Hassa, C. (1999) Breakup and atomization of a kerosene jet in
crossflow of air at elevated pressure. Proceedings of the 15th ILASS-Europe
Conference, Toulouse, France, July 5–7.
Becker, J. and Hassa, C. (2000) Plain jet kerosene injection into high tempera-
ture, high pressure crossflow with and without filmer plate. Proceedings of
the 8th ICLASS Conference, Pasadena, CA, July 16–20.
Becker, J. and Hassa, C. (2002) Breakup and atomization of a kerosene jet in
crossflow at elevated pressure. Atomization Sprays, 11, 49–67.
Cavaliere, A., Ragucci, R., and Noviello, C. (2003) Bending and break-up of a
liquid jet in a high pressure airflow. Exp. Therm Fluid Sci., 27, 449–454.
Chen, T.H., Smith, C.R., Schommer, D.G., and Nejad, A.S. (1993) Multi-zone
behavior of transverse liquid jet in high-speed flow. Proceedings of the 31st
AIAA Aerospace Sciences Meeting and Exhibit, Reno, NV, January 11–14.
Eggers, J. (1997) Nonlinear dynamics and breakup of free-surface flows. Rev.
Mod. Phys., 69, 865–929.
Heister, S.D., Nguyen, T.T., and Karagozian, A.R. (1989) Modeling of liquid jets
injected transversely into a supersonic crossflow. AIAA J., 27, 1727–1734.
Inamura, T. and Nagai, N. (1997) Spray characteristics of liquid jet traversing
subsonic airstreams. J. Propul. Power, 13, 250–256.
Kihm, K.D., Lyn, G.M., and Son, S.Y. (1995) Atomization of cross-injecting
sprays into convective air stream. Atomization Sprays, 5, 417–433.
Lefebvre, A.H. (1989) Atomization and Sprays, Hemisphere Publishing Corp,
New York.
Nejad, A.S. and Schetz, J.A. (1983) Effects of viscosity and surface tension on a
jet plume in supersonic flow. AIAA J., 22, 458–459.
Nguyen, T.T. and Karagozian, A.R. (1992) Liquid fuel jet in subsonic crossflow.
J. Propul. Power, 8, 21–29.
Oda, T., Nishida, K., and Hiroyasu, H. (1994) Charaterization of liquid jet atomi-
zation across a high speed airstream by laser sheet tomography. Proceedings
of the 6th ICLASS Conference, Rouen, France, July 18–22.
342 A. BELLOFIORE ET AL.

Pilch, M. and Erdman, C.A. (1987) Use of breakup time data and velocity history
data to predict the maximum size of stable fragments for acceleration-
induced breakup of a liquid drop. Int. J. Multiphase Flow, 13, 741–757.
Ragucci, R., Bellofiore, A., Carulli, G., and Cavaliere, A. (2003) Momentum
coherence breakdown of bending atomizing liquid jet. Proceedings of the
9th ICLASS Conference, Sorrento, Italy, July 13–17.
Ragucci, R., Bellofiore, A., and Cavaliere, A. (2004) Statistical evaluation of
dynamics and coherence breakdown of kerosene and water jets in crossflow.
Proceedings of the 19th ILASS-Europe Conference, Nottingham, UK, Septem-
ber 6–8.
Ragucci, R. and Cavaliere, A. (2002) Identification of cross-flow liquid-jet struc-
tures by means of statistical image evaluation. Proceedings of the 18th ILASS-
Europe Conference, Zaragoza, Spain, September 9–11.
Ragucci, R., Cavaliere, A., and D’Amico, R. (2000) Atomization of a liquid jet
in gas-turbine configuration. Proceedings of the 8th ICLASS Conference,
Pasadena, CA, July 16–20.
Reitz, R.D. (1987) Modelling atomization processes in high pressure vaporizing
sprays. Atomization Spray Technol., 3, 309–337.
Sallam, K.A., Aalburg, C., and Faeth, G.M. (2004) Primary breakup of nontur-
bulent round liquid jets in uniform gaseous crossflows. Proceedings of the
19th ILASS-Europe Conference, Nottingham, UK, September 6–8.
Schetz, J.A., Kush, E.A. Jr., and Joshi, P.B. (1980) Wave phenomena in liquid jet
breakup in a supersonic crossflow. AIAA J., 18, 774–778.
Spangler, C.A., Hilbing, J.H., and Heister, S.D. (1995) Nonlinear modeling of jet
atomization in the wind-induced regime. Phys. Fluids, 7, 964–971.
Tambe, S.B., Jeng, S.-M., Mongia, H., and Hsiao, G. (2005) Liquid jets in sub-
sonic crossflow. Proceedings of the 43rd AIAA Aerospace Sciences Meeting and
Exhibit, Reno, NV, January 10–13.
Wu, P.K. and Faeth, G.M. (1993) Aerodynamic effects on primary breakup of
turbulent liquids. Atomization Sprays, 3, 265–289.
Wu, P.K., Kirkendall, K.A., and Fuller, R.P. (1998) Spray structures of liquid jets
atomized in subsonic crossflows. J. Propul. Power, 14, 173–182.
Wu, P.K., Kirkendall, K.A., Fuller, R.P., and Nejad, A.S. (1997) Breakup pro-
cesses of liquid jets in subsonic crossflows. J. Propul. Power, 13, 64–73.
Yoon, S.S. and Heister, S.D. (2004) A nonlinear atomization model based on a
boundary layer instability mechanism. Phys Fluids, 16, 47–61.