You are on page 1of 21

Correlations to predict droplet size in ultrasonic atomisation

R. Rajan
a,
*
, A.B. Pandit
b
a
Hindustan Lever Research Centre, 64, Main Road, Whiteeld, Bangalore 560 066, India
b
Chemical Engineering Section, Department of Chemical Technology, University of Mumbai, Matunga, Mumbai 400 019, India
Received 2 April 2000; accepted 2 November 2000
Abstract
In conventional two uid nozzles, the high velocity air imparts its energy to the liquid and disrupts the liquid sheet into droplets.
If the energy for liquid sheet fragmentation can be supplied by the use of ultrasonic energy, ner droplets with high sphericity and
uniform size distribution can be achieved. The other advantage of ultrasound induced atomisation process is the lower momentum
associated with ejected droplets compared to the momentum carried by the droplets formed using conventional nozzles. This has
advantage in coating and granulation processes. An ultrasonic probe sonicator was designed with a facility for liquid feed ar-
rangement and was used to atomise the liquid into droplets. An ingenious method of droplet measurement was attempted by
capturing the droplets on a lter paper (size variation with regard to wicking was uniform in all cases) and these are subjected to
image analysis to obtain the droplet sizes. This procedure was evaluated by high-speed photography of droplets ejected at one
particular experimental condition and these were image analysed. The correlations proposed in the literature to predict droplet sizes
using ultrasound do not take into account all the relevant parameters. In this work, a truly universal correlation is proposed which
accounts for the eects of physico-chemical properties of the liquid (ow rate, viscosity, density and surface tension), and ultrasonic
properties like amplitude, frequency and the area of vibrating surface. The signicant contribution of this work is to dene di-
mensionless numbers incorporating ultrasonic parameters, taking cue from the conventional numbers that dene the signicance of
dierent forces involved in droplet formation. The universal correlations proposed are robust and can be used for designing ul-
trasonic atomisers for dierent applications. Among the correlations proposed here, those ones that are based on the dimensionless
numbers and Davies approach predict droplet sizes within acceptable limits of deviation. Also, an empirical correlation from ex-
perimental data has been proposed in this work. 2001 Elsevier Science B.V. All rights reserved.
Keywords: Ultrasound; Droplets; Sonication; Atomisation; Nozzle; Imaging; Emulsication
1. Introduction
The conventional way to atomise a liquid is to force it
at high velocity through a small aperture. The liquid
must have suciently high velocity of ejection in order
to form droplets. In the conventional nozzles, the orice
size greatly aects the resulting droplet size. The orice
can be plugged up if any impurity and solid particulates
are present in liquid. This limitation does not allow the
orice to be too small and hence it is dicult to produce
very ne droplets using conventional nozzles. As the
frequency of ultrasonic vibration determines the droplet
size in ultrasonic atomisation, very small droplets (15
lm) can be produced by increasing the frequency. The
thickness of lm formed on the vibrating surface also
has an eect on droplet size which can be controlled by
adjusting the liquid ow rate. The liquid lm thickness is
known to get aected by the physico-chemical properties
of liquid mainly surface tension and viscosity.
Based on the frequency of operation, the ultrasonic
transducers are classied as high frequency and low
frequency transducers. The dimension of transducer
in high frequency atomiser needs to be small in order
to reduce multimode operation. The atomisers using
ultrasonic vibrations are further categorised into two
types: Gas driven (whistle type) and electrically driven
atomisers. Wilcox and Tate [1] studied the operation of
whistle type sonicator. The electrically driven atomisers
operate solely on ultrasonic vibrations whose energy is
derived from piezoelectric or magnetostrictive trans-
ducers. As the eciency of latter ones decrease with
Ultrasonics 39 (2001) 235255
www.elsevier.nl/locate/ultras
*
Corresponding author. Tel.: +91-80-845-1505; fax: +91-80-845-
3086.
E-mail address: rajan.raghavachari@unilever.com (R. Rajan).
0041-624X/01/$ - see front matter 2001 Elsevier Science B.V. All rights reserved.
PII: S0041- 624X( 01) 00054- 3
increasing frequencies, they cannot be used above 125
kHz. The optimal operation of magnetostrictive trans-
ducers is between 20 and 40 kHz. Piezoelectric devices
made of synthetic ceramic material operate eectively
over a wide range of frequencies, from 30 kHz to 5 MHz.
The major advantages of ultrasonic atomisers are
smaller drop sizes which can be accurately controlled,
more uniform size distribution, high sphericity of drop-
lets, relatively large liquid ow passages, and low re-
quired liquid feed pressure. An added advantage when
ultrasonic atomisers are used for sensitive coating is the
fact that the spray contains droplets with little mo-
mentum when formed. Due to lower droplet velocities,
there will be less bounce back from the source on im-
pact. Also the spray pattern can be varied by appro-
priately designing the geometry of the vibrating surface.
They can be used for viscous materials and slurries.
These are single phase atomisers where there is no need
for compressed air to transfer its kinetic energy to liquid
to form droplets as in the case of conventional two
phase nozzles. The limitation is throughput rate and as
it is determined by the area of vibrating surface. As the
high frequency transducers are tiny, their volume han-
dling capacity is small and hence the industrial ultra-
Nomenclature
A constant in correlation A
A rate of creation of new surface
Am amplitude of sound wave
Am
crit
threshold amplitude
C velocity of sound in liquid medium
D pipe diameter/diameter of vibrating surface
D
sur
diameter of solid surface
d
32
Sauter mean diameter of droplet
d
j
jet diameter
d
max
maximum droplet diameter
d
ori
nozzle diameter
d
p
droplet diameter
E
s
rate of generation of surface energy
f excitation frequency
fr friction factor
g acceleration due to gravity
H lm thickness on solid surface
I intensity of ultrasound
k proportionality constant in the denition of
N; (consequently k appears in correlation B
and C and assumes dierent values)
k
1
constant in Davies equation for drop
break-up in turbulent emulsion system and
consequently appears as a constant in corre-
lation D
L length of pipe/length of vibrator
l turbulent length scale ~0.08D
m
l
, m
g
mass of liquid and gas respectively
N number of points or jets per unit surface area
for drop ejection
N
f
frequency of droplet ejection
n total number of droplets generated
n number of drops generated per second
DP pressure drop (sux ``g'' and ``l'' refers to gas
and liquid respectively)
DP
+
g
reduced pressure drop
P
a
pressure amplitude in N/m
2
P
m
power dissipated per unit mass
Q, Q
l
volumetric ow rate of liquid
Q
crit
critical volumetric ow rate
Q
e
volumetric rate of droplet ejection
Q
f
volume supplied per unit area per unit time
Q
g
volumetric ow rate of gas
Q
max
maximum volumetric ow rate
Q
opt
optimal volumetric ow rate
Q
v
volumetric displacement rate of vibrator
R radius of droplet
U
/
uctuating turbulent velocity
V
j
liquid volume in a jet
V
t
total volume of liquid in all jets per unit area
of vibrating surface
v or v
l
mean liquid velocity
v
g
mean gas velocity
Greeks
a
e
attenuation coecient
e rate of energy dissipation per unit mass
g liquid viscosity
g
crit
critical liquid viscosity
k wavelength
q, q
1
liquid density
q
f
falling liquid density
q
c
density of surrounding medium (air)
q
g
air or gas density
r surface/interfacial tension
l
p
ratio of mass ow rates of liquid to gas
s liquid loading per unit width of surface
h angle of inclination with respect to horizontal
Dimensionless numbers
We Weber number
conventional v
2
d
p
q=r
modied fQq=r
Oh Ohnesorge number
conventional g=(qrd)
0:5
modied g=(f Am
2
q)
power intensity (P
I
) P
2
a
=2qC
I
N
, intensity number f
2
Am
4
=CQ
Reynolds number Dvq=g
236 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
sonic atomisers cannot be operated with high frequency
transducers. Although piezoelectric transducers can be
operated over a wide range of frequencies (from 30 kHz
to 5 MHz), they cannot be operated at a very low fre-
quency in order to have high throughputs as at low
frequencies the amplitude is insucient to achieve liquid
sheet or ligament disintegration. Hence there needs a
compromise between amplitude and throughput rate.
Horn atomisers amplify the vibrations and hence are
best suited for atomising liquid lm with minimal power
requirement. The size of a 40 kHz horn device is about
10 mm in diameter but using waveguides the size can be
increased or reduced keeping the total power delivered
same aecting the intensity (W/cm
2
) and hence the am-
plitude, limited by the fatigue stress of the material of
construction used for the vibrating surface.
The interaction of cavitation produced shocks with
capillary waves is very complex. A lack of quantitative
understanding of this mechanism of disintegration is
the bottleneck in translating the concept of ultrasonic
modulated atomisation into industrial practise and
in designing an appropriate ultrasonic atomiser for a
specic application. Although numerous reports have
appeared in literature on the mechanism of droplet
formation due to disintegration of liquid lm on a vi-
brating surface, the complication involved in theoretical
analysis of integrated eect of cavitating transient bub-
bles eludes complete understanding of the cavitation
zone in an atomiser.
The objective of this work is to assess the impact of
various physico-chemical properties of liquid, its ow
rate, the amplitude and frequency of ultrasound and the
area and geometry of the vibrating surface on the
droplet size distribution. A more rigorous correlation is
proposed to predict the droplet size formed using an
ultrasonic atomiser taking into consideration the eect
of liquid ow rate and viscosity. Also, the eect of ul-
trasonic amplitude is incorporated when the atomisation
is carried out below a critical amplitude. The proposed
correlation is evaluated against the experimental obser-
vations and also against the predictions made by other
correlations. The behaviour of spray pattern and the
momentum associated with it would also be discussed.
Translating the eect of ultrasonic parameters to the air
and liquid feed pressure, feed rates and orice size to
obtain a particular droplet size distribution is also at-
tempted by comparing the energy requirement to pro-
duce a droplet of a specic size.
2. Literature background
2.1. Conventional atomisation
Before the theory of ultrasonic atomisation is dis-
cussed, a brief sketch on the available knowledge of the
eects of liquid properties on liquid atomisation through
the nozzles and also in liquidliquid dispersion systems
is made here. The disintegration of liquid sheet or liga-
ment into ne droplets in a gas phase is known as
atomisation. Basically the atomisation is possible by
disintegration of either a liquid jet or a liquid sheet/lm.
The disintegration of liquid sheets produce ne droplets
or in other words the sheet disintegration requires much
lower pressures compared to jets in order to produce
droplets of a particular size. Rim contraction, aerody-
namic wave formation, and turbulent break up are dis-
tinguished during liquid sheet disintegration. The liquid
sheets are stable compared to liquid jets. This is due to
the fact that any deformation to liquid sheet results in an
increase in surface area which helps in restoring forces.
The forces acting in a turbulent ow eld during liquid
break up are buoyancy, gravity, drag or external viscous
force, inertial or dynamic pressure forces from velocity
uctuations, surface forces, and internal viscous force.
When the deformation caused by these forces becomes
too large, the instability arises and the liquid sheet
breaks up into droplets.
In the conventional process whereby a liquid is ato-
mised by passing it through a nozzle, the velocity of
emerging liquid jet is of considerable importance. It is
the mean kinetic energy of issuing jet which determines
the droplet sizes. Masato [2] studied the eect of droplet
size variation in a stirred tank and conrmed the role of
mean and root mean square (rms) velocities in such
dispersion systems also. In a nozzle system, the kinetic
energy of the turbulent motion of the liquid can be
enhanced by the use of another phase which can be
compressed air or an immiscible liquid. The density
dierence between the liquid phase and the gas is many
orders of magnitude higher compared to the density
dierence between two liquids as in the case of a dis-
persion system. Hence the mechanisms of droplet for-
mation in an atomiser and in a dispersion system dier
from each other. In the two phase atomisers, the liquid
pressure can be low and the high velocity of gas ow
leads to a high relative velocity between the two phases.
This high relative velocity between the two phases cre-
ates a drag on the liquid and the shear induced due to
aerodynamics disrupts the liquid sheet and droplets are
ejected from its edges. The disintegration of liquid oc-
curs when the dynamic air pressure exceeds the pressure
inside the liquid. The air can be mixed with the liquid jet
internally or externally. This creates a suction in the
liquid channel and gravity feed of liquid is possible.
Dombrowski and Fraser [3] have shown that the
liquid sheet rst disintegrates into liquid threads and
these are further broken up into droplets. The process of
formation of liquid threads, an essential step before the
drop production, is shown to be intermittent. Also an
increase in the ow rate increases the pendant drop size
and at high ow rates a free liquid sheet bounded by the
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 237
rim is observed. Threads are thrown out from liquid
edge in a more irregular fashion leading to wider droplet
size distribution. The stability of this sheet is dependent
on the local oscillatory disturbances, due to turbulence
on its surface. Swartz and Kessler [4] observed that the
number of drops formed during drop breakage in a
turbulent pipe ow to be proportional to Weber num-
ber. Recently Clark [5] studied the eects of density,
viscosity, interfacial tension, and the relative uid ve-
locity on drop deformation under turbulence in a mixing
vessel. They suggested that the critical Weber number
may be dependent on the parent droplet size, interfacial
tension, viscosity, and the magnitude and duration of
the disruptive force known as the impressed frequency
of the disturbing force. Hinze [6] suggested the increas-
ing importance of viscosity as interfacial tension de-
creases which is taken care of in the viscosity number.
Entov and Yarin [7] observed the rate of growth of
(amplitude) disturbance to be aected by the liquid
viscosity. Bellman and Pennington [8] also studied the
eect of viscosity and surface tension on Taylor insta-
bility. They remarked that the viscosity will not remove
instability but it would dampen the rate of growth of
disturbance at any particular frequency. As the fre-
quency increases (i.e. for small wavelength) this reduc-
tion in the amplitude is more pronounced. It was shown
that an increase in disperse phase viscosity has a sig-
nicant eect on maximum deformation and this was
not predicted by Kolmogoro or Taylor model [5].
Davies [9] studied turbulence as the basic mechanism of
drop break up in an emulsion system and identied the
viscosity of the dispersed phase as the important prop-
erty in determining droplet size. Kitscha and Koc-
amustafaogullari [10] developed a general correlation
based on the combination of KelvinHelmholtz and
RayleighTaylor instability theories to predict the
maximum droplet size in a stagnant uid.
The photographic investigation [3] of the liquid dis-
integration showed the break up of threads to conform
to Rayleighs theory of stability of liquid columns ac-
cording to which a free column of liquid becomes un-
stable when its length exceeds its circumference. It has
been shown that the rapidly growing wavelength of
disturbance, in the case of nonviscous liquids, is 4.5
times its diameter. In the jet break up model [11], the
drop diameter was predicted to be 1.89 times the jet
diameter. It was also showed that the optimum or
dominant wavelength of the disturbance that leads to
liquid break up into droplets is equal to 4.5 times the jet
diameter [12]. By equating the drop volume to the vol-
ume of the (capillary) disturbance, they arrived at the
same ratio of (1.89) drop to jet diameter.
In the two phase external mix type nozzles, the sec-
ondary gas ow in the vicinity of the nozzle causes the
liquid to spread out as a lm on the outow surface.
This is carried to the separating edge where actual
atomisation takes place. The original annular jet con-
tracts to a cylindrical jet [13]. Thus the prerequisite for
the formation of droplets is the achievement of a thin
liquid sheet (lm) with a liquid thread at its edge. The
liquid threads are intrinsically unstable and centrally
symmetric surface waves form on their surface. Also by
imposing pressure or mechanical vibrations on the liq-
uid sheets, pressure waves and surface waves within the
liquid can be created or existing waves amplied. These
surface waves grow very fast when their wavelength
exceeds thrice the jet diameter. When the instability of
liquid lm and its surface waves overcome the restoring
surface tension forces droplets are ejected from its edges.
The surface waves are thus blown over the edges due to
disturbances in the ow or due to the drag caused by the
velocity dierence between the liquid and the gas, to
form many tiny droplets.
In general, droplets are formed as a consequence of
two subsequent phenomena: the liquid ow is dispersed
in the form of a liquid sheet/lm which decreases in
thickness and disintegrates into ligaments followed by
the break-up of these ligaments into droplets. Although
many correlations are available in the literature for
droplet size prediction based on the liquid loading and
the ratio of energy present in the liquid to surface en-
ergy, the chaotic and complex disintegration process
makes it dicult to predict accurately the size and its
distribution. Use of a vibrating nozzle or a pulsed
discharge of liquid through the nozzle at a specic fre-
quency leads to monodisperse spray. The forces in-
volved in droplet formation are taken into account
through the use of two dimensionless numbers Weber
number (We) which is the ratio between the aerody-
namic drag and the surface tension force of the jet
and a viscosity number called Ohnesorge number (Oh)
which is the ratio between internal liquid viscosity (vis-
cous force) to surface tension and the nozzle diameter. It
was also found that the transition from laminar into
intermediate sinusoidal disintegration takes place when
We=Oh
0:4
is about 2500 and it further transforms to
turbulent disintegration when this ratio crosses 60,000.
For a particular ow pattern, above a certain critical
Weber number, secondary droplet formation also oc-
curs.
Pandit and Davidson predicted the rupture velocity
of spherical or at lms [14]. The number of liquid
threads and hence the number of droplets formed was
observed to be a strong function of liquid viscosity. As
the liquid viscosity increases, the number of liquid
threads decreases. Also, the rupture velocity (rate of
droplet formation) was found to be aected only by the
surface tension and not the viscosity but the surface
tension aected the droplet size. It is clearly established
that the physico-chemical properties of the liquid, ow
pattern, relative velocity and density dierence between
the two phases, nozzle type and its dimension deter-
238 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
mines the droplets size. Though considerable literature
exists on single uid atomisation and droplet size cor-
relation, there appears to be a scant literature on two
uid atomisation.
2.2. Ultrasonic atomisation
When a liquid is subjected to suciently high inten-
sity of ultrasonic eld, the splitting of liquid occurs to
form droplets and they are ejected from the liquid-
ultrasonic source interface into the surrounding air as a
very ne dense fog. The major advantage of ultrasonic
atomisation is that the size of the fog particle and its
density can be independently controlled unlike in the
case of two uid (pneumatic) atomisation. The fog
particle size is controlled by the ultrasonic frequency for
a liquid at specic ow rate having specied physico-
chemical properties such as density, viscosity, and sur-
face tension. The fog density is adjusted by varying the
air ow past the liquid surface.
There are two major hypotheses that explain the
mechanism of liquid disintegration during ultrasonic
atomisation: cavitation hypothesis and capillary wave
hypothesis. Cavitation hypothesis is generally applied to
high frequency and high energy intensity systems. When
a liquid lm is sonicated, cavitation bubbles are formed.
During the implosive collapse of these bubbles, espe-
cially nonlinearly oscillating small bubbles near the
surface of the liquid, high intensity hydraulic shocks are
generated. These hydraulic shocks initiate the disinte-
gration of liquid lm and cause the direct ejection of
droplets. Capillary wave hypothesis is based on Taylor
instability. The liquid capillary waves are composed
of crests and troughs. Atomisation takes place when
unstable oscillations tear the crests (peaks) of surface
capillary waves away from the bulk liquid. Thus the
drops are produced at the crests whose size is propor-
tional to wavelength. The capillary wavelength decreases
with increasing frequency of ultrasound and thus lead to
the production of ne droplets at higher frequencies.
The dependence of ultrasound induced atomisation on
liquid vapour pressure, gas content in the liquid, static
pressure and also the existence of sonoluminescence in
the zone of atomisation lends credence to the cavitation
theory. On the other hand, a strong correlation between
mean droplet size and capillary wavelength favours
capillary wave theory. Lang [15] measured the sur-
face disturbances by photographing them. This study
showed that the disturbances on the liquid surface is
nothing but the crossed capillary standing waves. It was
reported that the major dierence is with respect to their
wavelength when operated at varying frequencies and
wavelengths were smaller at high frequencies. Bougu-
slavskii and Eknadiosyants [16] coupled these two the-
ories and proposed a conjunction theory according to
which the periodic hydraulic shock from the cavitation
disturbance interacts with the nite amplitude capillary
waves and excite them to form droplets.
It was assumed that the exponential growth of surface
disturbances result in drop formation whose diameters
are proportional to the wavelength of the most rapidly
growing initial disturbance [17]. The correlation predicts
the drop size to be half the wavelength of the distur-
bance. According to Rayleighs instability criteria, the
wavelength is 4.5 times the diameter of the jet and
the diameter of the jet is 0.53 times the diameter of
the droplet. Thus, Rayleighs theory also predicts the
diameter of the droplet to be approximately half the
wavelength. By analysing the hydrodynamic stability of
liquid lm, they determine the rapidly growing initial
disturbance. The liquid motion is assumed to be irro-
tational. Cavitation being a random phenomenon, this
interaction is likely to introduce certain random varia-
tions in the droplet size, though no quantitative results
have been reported in the literature.
The capillary waves are more random in the con-
ventional atomisation process whereas the capillary
waves modulated by ultrasonic vibration are more reg-
ular in nature which is the reason for uniform droplet
size distribution in the later case. It should also be noted
that the cavitation shock often leads to irregular random
disintegration and the observed uniformity of droplet
distribution adds support to the conjunction theory. The
role of cavitation induced shock waves is believed to be
that of exciting the capillary waves. The conjunction
theory explains the correlation between mean droplet
size and capillary wavelength while explaining the de-
pendence of atomisation on liquid vapour pressure,
dissolved gas content, and static pressure. Ultrasonic
atomisation using a stroboscope and a video camera
with close-up lens was studied [18]. These studies
strengthened the understanding of ultrasound induced
liquid atomisation.
Although there are many categories of atomisers in-
duced by ultrasound, capillary wave atomisers are most
widely used. The amplitude requirement of such low
frequency (<100 kHz) atomisers is in excess of 3 lm,
corresponding to 10100 W power output. These are
helpful in producing ne droplets from low viscosity
liquids (<20 mPa s) without using pressure. As a thin
lm of liquid wets an oscillating surface at a xed fre-
quency and oscillates in the direction normal to the
liquid surface, capillary waves are produced which on
excitation disrupts into ne uniform droplets from its
peaks, for a wide range of liquid ow rates. Because of
cavitational disturbances there is a broadening of the
droplet size distribution [19]. Topp et al. have listed ef-
fect of dierent frequencies but there was no informa-
tion about the eects of viscosity, density, and liquid
ow rates [19]. When the majority of the droplets are
formed by the capillary wave mechanism, the droplet
size can be estimated using Lang equation [15]. For
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 239
sonically generated capillary waves whose wavelength
can be calculated using Kelvins equation, the surface
wave frequency in this equation is equal to one half the
exciting sound frequency. Hence the droplet produced
should be less than the capillary wavelength. Generally
the number mean diameter should correspond close to
half the capillary wavelength as only a part of the wave
peak is ejected as droplets. Lang also reported that
roughly 90% of the droplets were found to be smaller
than twice the number-median diameter. A comparison
of number mean diameter with capillary wavelength
indicated that the drop size is a constant fraction of the
capillary wavelength and this turned out to be 0.34.
Hence Lang proposed a correlation as below:
d
p
= 0:34
8pr
qf
2
_ _
1=3
where r is surface tension (dyne/cm), q is density (g/ml),
f is excitation frequency (Hz), and d
p
is the droplet size
(cm). The droplets produced would have certain distri-
bution as there is a variation as to how much portion of
surface wave peak is ejected as droplets. The nonuni-
formity is also due to the fact that collision and ag-
glomeration of droplets after being ejected from the
liquid surface. Langs correlation indicates no depen-
dence of liquid phase viscosity and the volumetric liquid
ow rate which is contrary to the experimental obser-
vations. An empirical equation to predict the droplet
size (d
p
) at high liquid ow rates was proposed [20]
which is written as
d
p
= 31:7
r
q
_ _
0:354
g
0:303
Q
0:139
here g is the liquid viscosity (cP) and Q is the volu-
metric ow rate (lpm). This correlation is dimensional in
nature and is less than satisfactory as it does not account
for the excitation frequency and the amplitude of os-
cillation. An another prediction [21] was made for drops
produced from thick lm. Based on the numerical so-
lution of hydrodynamic stability equations [17], a cor-
relation was proposed between drop size, transducer
frequency, transducer amplitude, and liquid-lm thick-
ness using three nondimensional groups (including the
ratio of inertial forces to surface tension forces) which is
very similar to a correlation proposed in this work. A
mathematical analysis of surface capillary waves excited
by ultrasonic cavitation in an incompressible liquid has
been reported [22]. While predicting the threshold con-
ditions, periodic excitations due to cavitation in the
form of hydraulic shocks is assumed to occur. Mir [22]
has shown that as the change in uid velocity across an
impulse increases, the number of capillary modes also
increases. This is suggested as a reason for decrease of
monodispersity of atomised droplets. It was also shown
that capillary waves oscillating at half the frequency of
sound eld are easier to excite. The capillary wavelength
and thus the droplet size increases with the viscosity of
liquid. There exists a good correspondence between cap-
illary wavelength and droplet size. The capillary wave-
length increases slowly (nonlinearly) with liquid phase
viscosity. Capillary waves are assumed to be produced
due to cavitation. However, no explicit correlation was
proposed.
The above discussions indicate the lack of either the-
oretical or empirical correlation available to predict the
droplet size formed by ultrasonic atomisation, account-
ing for the observed dependence of liquid phase viscosity
and the volumetric liquid ow rate. Due to the non-
availability of such a correlation, there are no guidelines
for the design of the ultrasonic atomisers for any specic
application. In the subsequent sections, such an attempt
will be made and it will be shown that Langs equation is
a special case of the proposed generalised equation and is
valid over a limited range.
The modications often made while designing ultra-
sonic atomiser are related to the surface geometry and
the way the liquid lm is formed on the oscillating sur-
face. Spray patterns can be inuenced by the geometry of
the surface prole. A convex geometry would produce a
wide spray cone. Use of hollow horns helps in increasing
the throughput. The eect of transducer geometry was
also studied [23]. Careful attention must be given to the
mechanism by which the liquid lm forms on the oscil-
lating surface. It could be formed by impinging the liq-
uid, feeding the liquid through the centre of the horn, or
by partially immersing the horn. Also ``pre-lming'' can
be achieved on a nonoscillating surface and this could
then be allowed to ow onto the vibrating surface [24]. A
soviet design with a capacity of 500 lph based on ``pre-
lming'' technique has been reported [25]. Atomisers
using the impingement technique can handle high vis-
cosity liquids upto 300 cP. Higher ow rates need to be
avoided as it would lead to thick lms and hence larger
droplets. In the feed through device, there is an internal
channel through which liquid is supplied to a liquid lm
which spreads over the oscillating surface. When the vi-
brating surface is immersed partially, the liquid meniscus
forms a capillary lm over the surface of ultrasonic horn
which is then atomised. This can produce droplets of the
order of 100200 lm.
The conventional ultrasonic atomisers are based on
the use of capillary waves or cavitation. Ultrasonic
atomisations of a liquid layer and in a fountain have been
reported by many investigators [26,27]. A simple and
compact device to jet and atomise liquids as uniform
droplets with lower power consumption was developed.
This device has a circular ring piezoelectric element, a
pinhole plate, and a body. A cavity with a diameter of 10
mm and a depth of 2 mm is connected to a tank. It also
consists of an air suction fan. The 50 lm thick stainless
240 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
steel pinhole plate, with 37 regularly spaced pinholes of
size 80 lm, and the piezoelectric element are adhesively
joined to one another to make up a circular vibrator and
this is supported around the cavity where the liquid is
lled. The bending vibration of the pinhole plate gener-
ates strong concentric ultrasonic waves in the liquid ad-
jacent to the pinholes. This leads to the jetting out of
multiple trains of droplets of the same diameter as that of
pinholes. The droplets are estimated to travel at a speed
of 5.6 m/s. It was reported that stable and continuous
jetting and atomisation of liquid occur in the vicinity
of second resonance frequency. Also, atomisation rate
was not proportional to the amplitude of voltage applied
to the piezoelectric element and is determined by the
number of pinhole plates and the vibration frequency
[28].
A similar study on the mechanism of droplet forma-
tion using a pinhole plate with seven pinholes of 80 lm
diameter vibrating at 34.4 kHz was conducted. The ef-
fects of distribution in displacement amplitude of the
pinhole plate, threshold pressure of atomisation, phase
of ejection, diameter and amount of droplets as a func-
tion of applied voltage were studied. It was reported that
when there is no liquid in the cavity, the atomiser
resonated at two frequencies of 38.5 and 45.3 kHz.
When the cavity is lled with liquid, it was found that
the vibration at the lower resonance frequency was
damped and the upper resonance frequency went down
to 34.41 kHz. The bubbles due to vibration of liquid at
lower frequency gave rise to unstable and intermittent
atomisation [29]. Extending this study further, an opti-
mum design procedure for the multi-pinhole vibrating
plate atomiser was developed. This vibrating system has
a thin plate with many pinholes and a circular ring
piezoelectric element. A great eciency of atomisation
was reported when there is a coincidence among the rst
resonance frequency of the inner portion and the outer
portion of the vibrating system and the second harmonic
resonance frequency of the entire system. Also, greater
eciency was reported when the inner diameter of the
element was equal to that of the second nodal circle of a
circular plate [30].
The performance of bimorph vibrator structure with
regard to number of pinholes and shape of the plate was
examined. Changing the shape of the pinhole plate
easily controlled the spray pattern. The formation of a
curved portion on the plate and the arrangement of
pinholes on the curved portion help in controlling the
spray pattern [3133].
3. Analysis of energy distribution
It is quite important to gain information on the ways
of energy transformation and dissipation in an atomi-
sation device in order to appreciate its ecacy as an
atomisation equipment. In pipe ow and conventional
atomisation (pipe ow with static mixers like nozzles),
the mean velocity (mean kinetic energy) is dominant
compared to the uctuating velocity (turbulent kinetic
energy). After the energy is transferred to the bulk liq-
uid, the energy is dissipated in bulk and the interface due
to viscous dissipation and the remaining energy is dis-
tributed as the mean and turbulent kinetic energies. The
ratio of turbulent to mean kinetic energy is enhanced
by the static mixers which increase localised turbulent
shears. Actually the turbulent energy is utilised in liquid
sheet disintegration. There are ways to enhance the
turbulent shear in a pipe ow while keeping the velocity
constant. This is achieved by increasing the pressure
drop due to form friction as against only the skin fric-
tion. On the other hand, maintaining the pressure drop
and lowering the velocity which is actually the Taylors
experiment, led to eective atomisation suggesting that
the droplets are primarily formed due to turbulent shear
and not due to the mean kinetic energy. It is not always
possible to increase the turbulent shear without aecting
the mean velocity. Also beyond a certain mean velocity
which corresponds to the liquid velocity which gives rise
to droplets on Kolmogoro length scale, the droplet size
cannot be decreased by increasing the mean velocity.
Only the number of drops and its formation rate gets
aected. Also it is not desirable to increase the mean
velocity beyond a certain limit as this would associate
a higher momentum with the droplets. This may not
augur well for applications like coating and granula-
tion where impaction with higher momentum is detri-
mental. Thus it is important to somehow increase the
turbulent kinetic energy component of energy dissipa-
tion for overall eective energy utilisation and the pos-
sibilities are the use of secondary uid, and designing
nozzles to eect swirling action while passing liq-
uid through it have already proved the utility of this
strategy.
It can be shown that the ratio of turbulent to mean
kinetic energy is signicantly higher atleast by an order
of magnitude in ultrasonic atomisers. The uctuating
velocity is atleast an order of magnitude higher in ul-
trasonic atomiser as compared to the conventional pipe
ow with the nozzle. The mean kinetic energy is still the
dominant form even with the ultrasonic atomisers. It
should be clearly understood that the mean kinetic en-
ergy in an ultrasonic atomiser is not the same as the
mean kinetic energy in a pipe ow. Here the mean ki-
netic energy is nothing but a periodic turbulent energy
because of very high frequencies involved. Thus in an
ultrasound induced atomiser, apart from the increased
contribution of uctuating component of turbulence,
the periodic component of turbulence (which is referred
to as the mean kinetic energy) is also available for the
process of atomisation. In the nozzle system, the tur-
bulent energy is not only a meagre percentage (0.072%)
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 241
of mean kinetic energy but also there is a high degree of
randomness associated with it which is the cause of wide
distribution in the observed droplet size. The turbulent
shear can be enhanced in an ultrasonic atomiser without
aecting the periodic kinetic energy.
If the energy transferred to the bulk liquid is taken as
the basis, it can be proved that the ultrasonic atomiser is
more energy ecient. Even if one considers the spent
electrical energy as the basis, assuming 50% eciency
for pumps and compressors and 30% eciency for the
piezoelectric device, the ultrasonic atomisers prove to be
more energy economical apart from other advantages
such as lower momentum associated with droplets, uni-
form size distribution, absence of the secondary ato-
mising uid phase, and no restrictions on the lowest
liquid ow rate and feed pressure (please refer to the
Appendix A for energy analysis). Referring to the Ap-
pendix A, it is clear that the estimated energy require-
ments show that for a typical droplet size production
with a particular liquid ow rate, the energy consumed
in an ultrasonic atomiser is 6% of the total electri-
cal energy required in a two phase external mix type
nozzle. By maintaining a required critical lm thick-
ness on the vibrating surface, eective energy utilization
and uniformity in size distribution can be further im-
proved.
3.1. Process of emulsication
In the emulsication studies conducted by Majumdar
et al. [34], it has been reported that the droplet size
(equilibrium) increased with an increase of the distance
of the irradiating surface from the oilwater interface.
They also showed that the maximum droplet diame-
ter could be correlated by the following expression based
on the energy balance
e =
r
0:6
q
0:2
c
d
max
_ _
2:5
where r is the interfacial tension and e is the rate of
energy dissipation per unit mass.
In their work, they have also shown that the eect of
the distance of the interface from the irradiating surface
can be attributed to the attenuation of the ultrasound
and the eect of the liquid phase viscosity can be cor-
related with the following dependence:
Attenuation coecient a
e
~ g
0:724
i.e. the intensity of
the ultrasound reaching the interface will decrease with
g
0:724
.
From the above discussion, it appears that the eec-
tive energy transfer takes place with the proper coupling
of the liquid with ultrasound horn and thus eects of
viscosity and liquid ow rate in the case of atomisation
are likely to be more severe.
4. Theoretical predictions
4.1. Correlations to predict droplet size taking into
account the eects of liquid volumetric ow rate, liquid
viscosity and ultrasonic amplitude
As pointed out earlier, Langs correlation is applicable
only when the liquid phase viscosity and the volumetric
liquid ow rate have no eect on droplet size. This is not
always true as there are experimental observations indi-
cating a dependence on liquid phase viscosity and volu-
metric ow rate. The correlation proposed by Mochida
[20] takes into account the eect of physical properties
and volumetric ow rate of liquid. It also correlates the
radial distances of the spray to the liquid ow rate, am-
plitude, and the density. The complete spray was ob-
served to fall in a parabolic orbit using high speed
photography. Their observations gave evidence to the
capillary wave mechanism of droplet formation. How-
ever, his correlation is less than satisfactory as it does not
account for the excitation frequency and the amplitude
of oscillation. Also while making a remark on the mo-
mentum with which the drops are carried further from
the atomiser, they discuss about the eects of liquid
density but ignored the inuence of amplitude on mo-
mentum associated with the issued droplets. There is a
need for an universal correlation to predict the droplet
sizes during ultrasonic atomisation considering the de-
pendence on all physico-chemical properties of liquid
atomised and the ultrasonic parameters. Such a corre-
lation is proposed here based on certain established
logics in conventional nozzle atomisation and the ex-
perimental observations using ultrasound.
(1) The studies on droplet formation and breakage
have shown that the numbers which dictate droplet size
are the Weber number and the Ohnesorge number. It is
also shown that depending on the ow pattern, above a
certain critical Weber number the maximum stable drop
breaks into many daughter droplets. This concept is ex-
tended to ultrasonic atomisation to suggest that there is
a maximum possible liquid ow rate below which the
droplet size formed is independent of ow rate. Tsai et al.
[35] studied the resonant eect of air ow and ultrasound
to amplify the growth of the instability. The experi-
mental observations of Tsai et al. also indicates such a
possibility and the droplet size dependence is observed to
be Q
0:250:30
. It is proposed that by equating the external
(inertial) forces to restoring surface tension forces i.e. by
taking We equal to unity, the maximum possible ow
rate which will not have inuence on droplet size can be
predicted. The Weber number is modied to include the
ow rate and ultrasonic frequency as
We =
fQq
r
242 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
We
crit
=
fQ
crit
q
r
= 1
Q
crit
=
r
f q
Various denitions of Weber number have been used
in the literature. The basic paper of Hinze (6) gives the
origin and the genesis for the same. It can be dened in
terms of average velocities if the break up is due to
laminar shear or in terms of RMS uctuating velocities
for turbulent break up. This concept is made use of
to dene the Weber number for ultrasound induced
atomisation.
(2) Experimental observations have indicated the
viscosity dependence as g
0:1660:303
. The rate of growth of
(amplitude) disturbance is aected by the liquid viscosity
as there would be more viscous energy dissipation. As
the droplet size is dependent on the energy available for
the growth of the instability which is the energy supply
less the viscous dissipation, higher viscous dissipation
would increase the droplet size [5,7,36]. This eect is
observed to be more pronounced at higher frequencies
[8]. The conventional viscous number (Oh) is dened
as the ratio of liquid viscosity to the square root of
the product of the nozzle diameter, surface tension and
density. In ultrasonic atomisation, it is the ultrasonic
amplitude which decides the growth of instability, as is
decided by the nozzle diameter in the conventional
process. Hence, the viscous number (Ohnesorge num-
ber) is modied to
Oh =
g
f Am
2
q
The growth of the instability has the dimensions of
the nozzle as the dominant eddy size in the energy
spectrum is shown to be a function of the ow area di-
mension. In the case of ultrasound the amplitude of the
wavy structure is dependent on the amplitude of the
perturbation/vibration and hence it has been consid-
ered in the above denition of Ohnesorge number. This
modication for Oh number is based on the intuition or
at the most heuristic arguments to include the eect of
viscosity and with the knowledge that the viscosity is
known to aect the parameters included in the modi-
cation.
(3) The geometry of the vibrating surface and its area
determines the intensity (W/m
2
) of ultrasonic power
delivered. This energy intensity aects the droplet size.
This is analogous to the power dissipation per unit
volume in nozzle atomisation which inuences the
length scale of turbulence and hence the turbulence in-
tensity governing the maximum stable diameter [9].
Thus, apart from the above two numbers, another di-
mensionless number is dened to take into account the
eect of energy density on droplet size. This number is
called Intensity number (I
N
) dened as
I
N
=
f
2
Am
4
CQ
Here C is the velocity of sound in liquid medium. The
power intensity which is nondimensionalised in the In-
tensity number is equivalent of Power number in the
case of mechanically agitated contactor or pressure drop
in the case of nozzle atomisation. The basis of these
dimensionless numbers and their modications are as
follows. The above denitions of the dimensionless
numbers include the eect of liquid ow rate, ultrasound
intensity (as energy dissipation per unit mass aects the
droplet diameter), and liquid phase viscosity (as this
aects the wavelength of instability due to the changes in
the proportion of viscous energy dissipation). The above
arguments were implemented in the form of dimen-
sionless numbers as dened above giving the net de-
pendence as
d
p
~ Q
0:248
~ g
0:166
~ Am
0:443
and the correlation is presented as
d
p
=
pr
qf
2
_ _
0:33
1
_
A(We)
0:22
(Oh)
0:166
(I
N
)
0:0277
_
A
The right-hand side of the above equation indicates
the change in the diameter of the drops due to the in-
uence of parameters other than only r and f as
considered by Lang. Rationale for choosing the expo-
nents in this correlation are based on the experimental
observations reported in literature where few of the
parameters have been studied whereas we are trying to
include all the parameters to oer an universal correla-
tion.
It should be mentioned here that in order to break the
liquid into drops using ultrasonic vibrations, the am-
plitude must be above a threshold value [37]. This
threshold amplitude is given as
Am
crit
=
2g
q
_ _
q
prf
_ _
0:333
If the amplitude is below ``Am
crit
'' then atomisation
would not take place and the correlations of either
Langs or proposed here cannot be used. They found the
critical amplitude for aluminium melt to be greater than
40 lm. The instability equations developed by Peskin
and Raco [17] also shows the existence of a minimum
transducer amplitude for instability and they mentioned
that this minimum amplitude is less than 2 lm for most
applications.
Our correlation needs to be curtailed by eliminating
the modied Weber number if the liquid ow rate is
below Q
crit
which is given by
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 243
Q
crit
=
r
f q
and under such circumstances, only the additional in-
uence of viscosity and amplitude needs to be consid-
ered. This is the minimum ow rate above which the
inuence of ow rate is predicted by our correlation.
The maximum ow rate for the correlation to be ap-
plicable is the volumetric displacement rate of vibrating
surface which is given by the product of frequency,
amplitude, and area of the vibrating surface. When the
liquid ow rate exceeds the displacement rate, then
dripping due to gravity takes place and larger drops are
formed.
4.2. Eect of Q
crit
on number and size of droplets
1. Below Q
crit
size is independent of Q and only the
number of drops will be proportional to Q.
2. Above Q
crit
but Q < Q
max
, size is proportional to
Q
0:250:33
and thus number will be decided by volume
balance.
3. The increase in d
p
with liquid ow rate above Q
crit
is
possibly due to an increase in the liquid lm thickness
with linear loading. According to Peskin and Raco,
the lm thickness is directly proportional to the liquid
ow rate. A correlation needs to be found for the
minimum liquid lm thickness (at minimum wetting
rate) as a function of solid surface characteristics, liq-
uid viscosity, surface tension, and liquid loading rate.
Although the weak inuence of viscosity is continu-
ous, there would be a critical value above which the
inuence cannot be ignored. Although there is no way to
dene the value of the critical viscosity the studies have
shown that the inuence of viscosity is marked when it is
greater than 10 cP. Only when the parameters are above
Am
crit
, Q
crit
, and g
crit
the right-hand side of the proposed
correlation will contribute signicantly over the droplet
size as predicted by Langs equation.
5. Alternate correlations
5.1. Using Rayleigh's instability criteria
(1) The droplet size prediction can be made based on
the number of drops formed on the vibrating surface
area. Let us consider a unit surface area on a at vi-
brating surface with frequency f . One drop is ejected
from one node/wave rather the drop ejection takes place
at every one wavelength. Therefore, number of points
per unit surface area for drop ejection is N = (1=k)
2
;
here, k is the wavelength; the notation in the picture
``ae'' is the same as the threshold amplitude ``Am
crit
''.
(2) According to Rayleighs instability criteria [38],
the wavelength of the disturbance is 4.5 times the di-
ameter of the jet (d
j
) and the diameter of the jet depends
on the frequency, liquid surface tension and its density.
N ~
1
d
j
_ _
2
;
N = k
qf
2
pr
_ _
0:66
d
j
~
pr
qf
2
_ _
0:33
Here k is the proportionality constant. This gives rise to
a rough estimate of 10
8
number of water drops ejected
from a unit surface area when the vibrating frequency is
28 kHz.
The Rayleighs instability criteria shows that the di-
ameter of jet to be 0.53 times the diameter of the droplet.
Drop volume =4.5 (length) (cross-sectional area of
the jet)
p
6
d
3
p
= 4:5d
j
p
4
d
2
j
d
p
= 1:89d
j
; or d
j
= 0:53d
p
:
(3) Pohlman and Heisler [37] proposed a criterion
to predict the maximum length of the jet issuing from
the points before it breaks-up into droplets (criteria of
threshold amplitude). According to this criteria,
Am
crit
=
2g
q
q
prf
_ _
1=3
(4) Assuming that sucient liquid lm thickness is
available on the vibrating surface, one can impose the
criteria of mass balance in order to get the frequency of
droplet ejected from the edges of each jet. The cylin-
drical jets are assumed and the droplets are assumed to
be spherical. The deviations from cylindrical structure of
244 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
the jets and sphericity of droplets are absorbed in the
constant k in the nal expression.
Liquid volume occupied in each jet,
V
j
=
p
4
d
2
j
Am
crit
Number of jets per unit area,
N = k
qf
2
pr
_ _
0:66
Hence, the volume of liquid in all jets per unit area of
vibrating surface is
V
t
(m
3
=m
2
) = NV
j
V
t
= k
qf
2
pr
_ _
0:66
p
4
d
2
j
_ _
Am
crit
V
t
= k
qf
2
pr
_ _
0:66
p
4
d
2
j
_ _
2g
q
q
prf
_ _
1=3
At steady state, if the volumetric supply rate is Q (m
3
/s),
then the volume retained on the unit area of vibrating
surface is V
t
(m
3
/m
2
). The volumetric supply rate (Q)
must equal the rate of ejected volume (Q
e
) in the form of
droplets. Q
e
per unit area can be obtained by multiply-
ing V
t
and N
f
, the frequency of droplet ejection.
Q
e
= V
t
N
f
; the volume supplied per unit area per unit
time is (m
3
/m
2
s) Q
f
= f Am; at steady state, on unit area
basis, Q
f
= Q
e
; hence,
N
f
= (f Am)=V
t
;
k
qf
2
pr
_ _
0:66
p
4
d
2
j
_ _
2g
q
q
prf
_ _
1=3
N
f
= f Am
Now, if the droplet size is d
p
and the frequency of
ejection is N
f
, the volumetric rate of droplet ejection
(Q
e
) is the product of total liquid in droplets and the
frequency of their ejection.
Q =
p
6
d
3
p
N
f
Q =
p
6
d
3
p
(f Am)
k
qf
2
pr
_ _
0:66
p
4
d
2
j
_ _
2g
q
q
prf
_ _
1=3
Assuming d
j
= 0:53d
p
; we can get d
p
in terms of Q and
and liquid viscosity.
Q =
p
6
d
3
p
(f Am)
k
qf
2
pr
_ _
0:66
p
4
0:28d
2
p
_ _
2g
q
q
prf
_ _
1=3
On simplication, we get,
Q =
3:74
k
d
p
rAm
g
; hence d
p
=
gQk
3:74rAm
B
The value of the constant k can be taken as 0.3. This
expression is independent of liquid density and the fre-
quency of oscillation and is applicable only at steady
state conditions.
5.2. Using Walzel's relation
Walzel [13] proposed a relation between drop size and
diameter of jet incorporating the viscous eect which is
written as
p
6
d
3
p
= pd
j
(2 0:6Oh)
0:5
p
4
d
2
j
Here Ohnesorge number is given by (g=f Am
2
q). We
have,
p
6
d
3
p
=
p
2
4
d
3
j
2
_
0:6
g
f Am
2
q
_
0:5
d
p
= 1:68 2
_
0:6
g
f Am
2
q
_
1=6
d
j
d
j
=
d
p
1:68 2 0:6
g
f Am
2
q
_ _
1=6
Substituting the above expression for d
j
in
Q =
p
6
d
3
p
(f Am)
k
qf
2
pr
_ _
0:66
p
4
d
2
j
_ _
2g
q
q
prf
_ _
1=3
On simplication and rearrangement, we get,
Q =
p
6
d
3
p
f Am ( )
k
qf
2
pr
_ _
0:66
p
4
d
2
p
2:822
_ _
2 0:6
g
f Am
2
q
_ _
1=3
2g
q
q
prf
_ _
1=3
d
p
=
1:06kQ
g
q
q
prf
_ _
0:33
qf
2
pr
_ _
0:66
f Am ( ) 2 0:6
g
f Am
2
q
_ _
1=3
C
The value of k is taken as 0.1. This correlation takes
into account the inuence of frequency, viscosity, ow
rate apart from the liquid properties and ultrasound
parameters.
5.3. Using Davies approach
Davies [9] proposed an equation to predict the max-
imum stable droplet size in an emulsion system based on
the turbulent energy dissipation rates as the turbulence
is the basic mechanism of drop break-up [39]. He con-
sidered the eect of dispersed phase viscosity by adding
an extra resistance term of dispersed phase viscosity
divided by the relevant energy dissipation for drop
breakage. This was given as (gv=d
max
). Thus along with
the interfacial tension pressure (4r=d
max
) resisting drop
breakage, the viscous resistance term was also added.
His equation for drop break up in a turbulent emulsion
system is
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 245
d
max
= k
1
(r gv=4)
0:6
q
0:6
P
0:4
m
Here P
m
is the power dissipated per unit mass and v is
the mean velocity. Extending this power dissipation
concept to the ultrasonic atomiser, we can arrive at an
expression for the prediction of droplet size. However,
the above expression was meant to predict the equilib-
rium droplet size after breakage and coalescence due to
turbulence. This may not be the case in ultrasonic
atomisation as the breakage process may not be com-
plete due to short residence times on the atomising
surface. This has been proved by the two-uid atomi-
sation experiments [35] modulated by ultrasound where
the superimposition of ultrasonic eect on the two-
uid nozzle atomisation led to decreased droplet sizes,
although the distributions were widened. The above
equation when suitably modied for power dissipated
per unit mass in terms of ultrasonic parameters would
give a correlation which would predict the minimum
equilibrium size that can be obtained. Taking into
consideration that the short residence times would not
allow the liquid jets/sheets to break up to the equilib-
rium size, the constant k
1
is introduced.
Let us now consider the power dissipated per unit
mass in a sonication system. The power intensity is given
by 1=2qCAm
2
(2pf )
2
. This is multiplied by area to get
the power and then divided by volume (Area Am) to
get the power per unit volume of liquid which is given by
power=volume = 1=2qCAm(2pf )
2
:
This is divided by the liquid density to get the power per
unit mass, P
m
.
P
m
= 1=2CAm(2pf )
2
; W=kg;
The above expression for P
m
is substituted in the
Davies equation for d
max
to get a correlation for droplet
size by introducing a constant and v (mean velocity) is
substituted by the product of Am and f.
d
p
= k
1
r
_

g Amf ( )
4
_
0:6
q
0:6
1
2
CAm 2pf ( )
2
_ _
0:4
D
This correlation is dimensional in nature. It predicts the
droplet size by considering the power dissipated per unit
mass. It considers the inuence of ultrasonic parameters
and physico-chemical properties of the liquid. This
correlation predicts dierent dependence of droplet size
on these properties compared to previous correlations.
Although the term included to account for viscous re-
sistance indicates a dependence on amplitude and fre-
quency, it is very insignicant unless the interfacial
tension is too small. Hence, excluding this term, the
exponent value of frequency is 0.8 (slightly high
compared to the Langs exponent value of 0.67) and
for the amplitude the exponent value is 0.4.
The major drawback of this correlation is that it does
not explicitly take into account the inuence of liquid
ow rate on the droplet size. It has been amply proved in
this paper that the eect of ow rate cannot be ignored.
In Table 1, one can see that the predictions made by this
correlation is qualitatively in line as far as the amplitude
eect is considered. However, as this correlation predicts
independent of liquid ow rate, the predicted values
deviate from the experimental values observed by Tsai
et al. [35]. The prediction of correlation (D) can be im-
proved to accommodate the variations in the droplet
size due to variation in the liquid ow rate by dening
the volume of liquid used in the energy dissipation per
unit volume in terms of the liquid lm thickness and the
horn area rather than horn area and amplitude as is
done. It is known that the liquid lm thickness is a
function of liquid ow rate and the physico-chemical
properties of the liquid and thus volumetric ow rate
can be accounted for.
Another point to be noted is that the viscosity and
surface tension cannot be considered in isolation. If
surface tension is high then the viscosity needs to be that
much higher to see an observable eect. As the droplet
formed due to ultrasonic atomisation is far from the
equilibrium value due to short residence times, the value
of k
1
is taken as 30 to t one experimental point and
using this value droplet sizes for other conditions are
predicted.
Table 1
Comparison between droplet size predicted by the correlations (A, B, C and D) and the experimental observation of Tsai et al.
Sl. no. Power
(W)
Frequency,
f (kHz)
Amplitude
(lm)
Q (m
3
/s)
10
8
d
p
(lm)
Observed Ref.
[15]
Correlation
A B C D
A = 0:1 k = 0:3 k = 0:3 k
1
= 30
1 1.0 54 2.02 2.16 50 29 60 12 9 50
2 1.5 54 2.48 2.16 45 29 58 10 8 46
3 1.8 54 2.72 8.47 80 29 65 35 29 44
4 2.0 54 2.86 27.6 90 29 83 108 91 43
5 2.5 54 3.20 8.47 60 29 60 30 26
27.6 75/80 78.5 96 83 41
6 3.5 54 3.78 27.6 70 29 72 81 73 38
246 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
Comment: Tsai et al. [35] conducted the experiments
with water and the observed droplet size tabulated here
was taken from their size distribution curves cor-
responding to the maximum volume percentage. The
volumetric ow rate, frequency, and power used in their
experiments are reported. The area of irradiation is ta-
ken to be 2:83 10
6
m
2
(L = D = 0:95 mm). The am-
plitude is calculated from the power and the nozzle area
reported in their paper. The constant A in our correla-
tion A is taken as 0.1. The value of constant k in both B
and C correlations are taken as 0.3. It is clear from the
above table that the proposed correlation predicts dif-
ferently compared to the predictions of the Langs cor-
relation as our correlations took into account the
inuence of ow rate, amplitude, and viscosity. The
correlation B does not consider the inuence of fre-
quency and density. Both B and C correlations predict
linear dependence of ow rate on the drop size which is
not observed through the experiments. The overesti-
mated dependence of ow rate led to the deviation in
droplet size predicted by these correlations. This com-
parison highlights the signicance of incorporating ow
rate, amplitude, and viscosity eects apart from the
liquid density, vibration frequency, and surface tension
as considered by Lang. The correlation A predicts well
and it is truly an universal correlation as it includes all
the eects and also the dependence of each variable is
matching with the experimental observations.
It is to be seen whether the suggested correlation can
predict the resulting droplet size if the droplets already
formed from a conventional system is made to fall on an
ultrasonic vibrating surface to get further atomised. In
such a case, the correlations need to be modied and the
conventional denition of Weber number as v
2
d
p
q=r is
to be used.
The proposed correlation can be extended to predict
the size of droplets produced in an ultrasonic liquid
liquid emulsication system. The energy dissipation
pattern would be dierent in emulsication system
compared to atomiser system because the density and
viscosity of the continuous phase is higher and of the
same order of magnitude as that of the dispersed phase.
Also the interfacial tension in an emulsier is signi-
cantly lower compared to the surface tension in an
atomiser. The drag and shear created due to the velocity
dierence and density dierence are also dierent in
these two systems. However, if one extrapolates the
knowledge and understanding of energy transformation
in an agitated emulsication system in terms of ultra-
sonic parameters, the genesis of the proposed correla-
tion can be extended.
5.4. Constraints on ow rate
The correlation indicates that if the volumetric ow
rate is low, the resulting drop sizes are small. It is not
true for ow rates below the critical ow rate. Moreover
if the ow rate is too low for sucient lm thickness to
be maintained on the vibrating surface, intermittently
the surface goes dry and gets exposed to air. Under such
situations, ultrasonic energy is wasted as no eective
atomisation can take place during such intermittent
exposure to air. Thus, one needs to maintain a certain
minimum ow rate in order to eectively utilise the en-
ergy. It is shown in Appendix B that when the liquid
ow rate is Q
crit
, there is sucient thickness of liquid
lm available in order to cover the vibrating area all the
time so that energy is not wasted due to the intermittent
exposure of vibrating surface to air. From throughput
view point also, Q
crit
is the minimum ow rate that one
should operate as any reduced ow than this would not
yield drops smaller than what would be obtained with
the critical ow rate. When operated above this ow
rate, one can have higher throughput but with increased
droplet size. But one should not operate the ultrasonic
atomiser with the ow rate above the volumetric dis-
placement rate of the vibrating surface because as
mentioned earlier dripping would occur and size distri-
bution becomes broader. The volumetric displacement
rate is the maximum allowable throughput rate for ef-
cient ultrasonic atomisation with a compromise on the
droplet size and Q
crit
calculated based on critical Weber
number value being unity appears to be the minimum
ow rate which would give maximum energy utilization
with minimum drop size.
6. Experimental
6.1. Set-up
A prototype 20 kHz ultrasonic atomiser with a mi-
croprocessor control was designed with a facility to
change the power input to the horn device (see Fig. 1).
Fig. 1. Ultrasonic atomiser (horn device) with microprocessor control.
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 247
The microprocessor based controller can be pro-
grammed for continuous or pulse mode of operation.
The material of construction is SS 304 grade. The elec-
tric supply is single phase, 230 V AC and 50 Hz. The full
power of the ultrasonicator is 80 W. An auto-trans-
former was used to vary the power fed to the atomiser
(graduated in % of full power).
A provision was made to attach probes of varying
diameters (1 and 2 cm) of vibrating surface. The probe
dimensions are straight with 2 cm diameter and another
one tapered and stepped to 1 cm diameter. In order to
vary the ow rate a reservoir was xed to provide liquid
at a determined ow rate while the probe is sonicated.
The liquid at a constant ow rate slips through the horn
and getting sucked towards the vibrating surface. The
whole probe set-up is levelled using adjustable screws
and a level indicator attached to the liquid feeder.
6.2. Experimental conditions
Experiments were conducted in pulse mode. The
liquid used to study droplet formation was water with
little amount of blue dye (methylene blue). The amount
of dye added was very low that the change in physical
properties of water such as density, surface tension and
viscosity are insignicant to be considered. Fig. 2 shows
the droplet ejection from the vibrating surface. Droplet
formation was carried out by changing the ow rate of
water from 0.44 to 3 lph. The variation in power input
was from 60% to 100% of rated power. The surface
tension eect was studied using 0.6 to 3 gpl sodium
lauryl sulphate solution. In order to study the eect of
liquid viscosity on droplet size, IP grade glycerine was
added to dye solution. The glycerine content was varied
from 20 to 50 vol.% in water. The density of solution
was varied by dissolving dierent amounts of NaCl (4, 8,
12, and 16 wt.%). The properties of dierent solutions
studied are given in Table 2. With respect to surfactant
addition, only the change in surface tension is taken into
account. Similarly, while adding NaCl the change in
density alone was considered. However, when glycerine
solutions were atomised, the changes in viscosity and
density were taken into consideration. The procedure
used here is to isolate the eect of surface tension, vis-
cosity, density, and the volumetric ow rate.
The intensity variation was studied using another
probe with dierent diameter of vibrating surface. The
calorimetric method was used to obtain the intensity of
ultrasonic waves. 20 cm
3
of water was taken in an in-
sulated beaker. This was sonicated for dierent time
periods. The increase in temperature was noted. From
this, the increase in thermal energy (W) was calculated.
The relationship between intensity (I) and the amplitude
(Am) is given as
I (W=m
2
) = 0:5[qC(2pf Am)
2
[:
Here q is the liquid density, C is velocity of sound, f is
frequency of sound (Hz), and Am is the amplitude of
sound (m). This relationship is true for planar wave
which should be true in this case. The pressure ampli-
tude (P
a
) can be calculated using the relation, I (W/
m
2
) =P
2
a
=2qC, and P
a
is given in N/m
2
. The intensity
values for dierent power input are given in Table 3. The
observed random variation in the intensity with respect
to power rating is possibly due to generator and horn
resonating eect which delivers maximum power at
some power rating for a particular horn.
The droplets were collected at dierent distances from
the vibrating surface (3, 7 and 12 cm from vibrating
surface) and also on the right and left side of the spray
cone. The droplets were captured on Whatmann 1 lter
Fig. 2. Droplet ejection from vibrating surface.
Table 2
Properties of solutions
SLS (gpl) concen-
tration
Surface tension, r
(kg/s
2
)
Vol.%
glycerine
Viscosity, g
(kg/m/s)
Density, q
(kg/m
3
)
NaCl concentration
(wt.%)
Density, q
(kg/m
3
)
0 0.072 0 0.001 1.0 0 1.0
0.6 0.0513 20 0.01 1.052 4 1.025
1.2 0.0358 28 0.016 1.073 8 1.054
1.8 0.0327 30 0.017 1.078 12 1.084
2.4 0.0341 40 0.024 1.104 16 1.114
3.0 0.0367 50 0.03 1.13
248 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
paper by quickly exposing the lter paper to the ato-
mising cone. The details of the range of operating pa-
rameters varied are as follows: (a) liquid ow rate 0.443
l/h; (b) surfactant (sodium lauryl sulphate) concentra-
tion 0.63 gpl; (c) 2050% glycerol solutions for study-
ing the eect of viscosity and density; (d) 416 wt.%
NaCl solution for studying the eect of liquid density;
(e) vibrating horn of diameter 1 and 2 cm; and (f) power
ratings of 60100%; (g) varying positions at which the
drops were collected.
The range of physico-chemical properties and the
ultrasound intensities used are indicated in Tables 2 and
3.
Table 3 gives the data for the small probe with di-
ameter 1 cm. The intensity values for the bigger probe
with 2 cm diameter would be four times less than the
ones reported above for corresponding power ratings
and the pressure amplitude and amplitude in length unit
would be half of those reported in Table 3.
6.3. Measurement
The droplets were collected on Whatmann 1 lter
papers by quickly exposing them to the atomising cone
under dierent experimental conditions. The drops were
collected in pulse mode operation when there was no
dripping of liquid. It should be mentioned here that
there would be some overlap of drops on the lter paper
and also the drops after impinging on the lter paper
would wick to some extent that may lead to erroneous
results. The wicking of drops with known volume of
liquid added to the lter paper was carefully studied and
this is relatively insignicant compared to drop sizes.
The error due to wicking was ignored as the qualitative
results are not aected and its eect on absolute values
of drop sizes is very insignicant.
The lter papers were measured for droplet sizes us-
ing image analysis technique. The images of droplets on
the lter paper were captured using a high resolution
camera and then digitised using a software KONTRON KONTRON
400 400. The software enables to enhance the resolution of
the images by way of adjusting the contrast and by using
the provision to label the images, overlapping droplets
were ignored. The pixel calibration is carried out using a
XY graph sheet in terms of absolute units (mm
2
) and
thus the area of droplet images are obtained in absolute
units. The software also calculates the feret ratio of the
droplets which is an indication of the sphericity of the
droplets. Assuming the droplets to be spherical, from
the area of the droplets obtained from image analysis,
the diameter of the droplets were calculated.
An independent calibration of droplet size was car-
ried out by taking high speed photographs of the ato-
mising droplets for one particular condition using a high
speed camera with a shutter speed of 1/5000, 5.6 aper-
ture size, high focal length, and Ilford 800 XP lm. The
background was kept completely dark and high intense
lights were shone from behind the object. The focus was
made on the probe tip. Both black and white and colour
photographs were taken. The image to object ratio was
kept at unity. This photograph was subjected to image
analysis and the absolute projected area of the droplets
was obtained. From this, the size of the droplets was
calculated assuming them to be spherical. Using this
value, a scaling factor was arrived at for the size calcu-
lated using the image analysis technique for the same
experimental condition. This scaling factor was found to
be ~2 and this was applied to all other experimental
conditions to get the droplet sizes.
7. Results and discussion
7.1. Feret ratio of droplets formed
One interesting observation was when we operated
the atomiser with 100% power rating (about 80 W) the
droplet size observed was higher than when operated at
say 70% of full power. It is contrary to the expectation
that with higher power rating (higher intensity of ul-
trasound) the droplet size would become smaller. It
seems that the drops were thrown with higher velocities
before they had the opportunity to atomise when oper-
ated at higher intensity. Thus, there exists an optimum
value of power intensity in order to get smaller droplet
sizes.
The feret ratio must be unity for spherical droplets. It
is a bit surprising that the feret ratios calculated by the
image analysis software are 0.70.76 indicating some
nonsphericity. This might be due to the lter paper not
being exactly horizontal. This should not matter in the
analysis of the qualitative trends but might be worth-
while to interpret these values of feret ratios in the light
of the fact that due to the propulsion of the liquid from
the probe the drops are likely to be deformed margin-
ally, giving rise to lower feret ratios as observed.
The droplet sizes obtained are as expected even from
the analysis based on the liquid held on the probe sur-
face. Let us calculate the liquid held on the probe sur-
face from the area of the vibrating surface, pulsations
per second, and liquid thickness on probe surface. The
Table 3
Amplitude data for 1 cm diameter probe
Power rating
(%)
Intensity
(W/m
2
)
Pressure
amplitude
(10
5
N/m
2
)
Amplitude
(lm)
100 14,954 2.12 1.124
90 14,331 2.07 1.1
80 15,842 2.18 1.157
70 13,739 2.03 1.077
60 17,500 2.29 1.22
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 249
sonication frequency is 20 kHz i.e. 20,000 cycles per
second. Each cycle comprises of a positive and negative
phase. It is assumed that during each negative phase of a
cycle the liquid is sucked onto the vibrating surface and
during each subsequent positive phase the liquid drop is
ejected. Thus, in one second, 20,000 pulsations would
occur leading to the formation of 20,000 ``drop forma-
tion events''. Also, according to Rayleighs criteria,
1:21 10
8
drops formed per unit surface. Hence, for
the probe with the area of 7:85 10
5
m
2
(0.01 m
dia probe), the number of drops formed per second is
(1:21 10
8
)(7:85 10
5
) = 9498 drops/s. This proves
the order of magnitude calculation to be correct.
Let us consider a liquid ow rate of 1:22 10
7
m
3
/s.
Volume of liquid in a single drop = (total liquid vol-
ume ejected=s)=(number of drops=s) = 1:22 10
7
(m
3
/
s)/20,000 (no:=s) = 6:1 10
12
m
3
; assuming the drops
to be spherical, the diameter of each droplet ejected is
d
p
=
6
p
6:1
_
10
12
_
0:33
= 226 lm
Depending upon the liquid wettability the thickness of
the liquid lm on vibrating surface is determined and
thus the above calculation shows that the order of
magnitude of droplet size is within the observed range of
droplet sizes.
7.2. Comments on correlations
The parity plots between the droplet sizes predicted
by correlations A and D and experimentally observed
droplet sizes has been given in Figs. 3 and 4. As can be
seen from these plots, it is clear that though the corre-
lations A and D correlate the data with average stan-
dard deviations of 0.18 and 0.27 and maximum
deviations of 0.48 and 0.61 respectively, the observed
trends are not correctly indicated. The constants used in
correlations A and D are 0.025 and 20 respectively
which are dierent from the constants used to predict
the data reported by Tsai et al. (Table 1). These con-
stants are applicable when all the units are in SI system.
The predictions of correlations B and C which are based
on RayleighTaylor type of instability for the formation
of drops do not predict the droplet sizes, neither the
trend correctly, indicating the natural ow related in-
stabilities have little role in ultrasonically driven atom-
isation process. The dierence between correlations A
and D mainly arises because in correlation A critical
ow rate is assumed, below which the dimensionless
groups relating to the ow and physico-chemical prop-
erty parameters do not show any dependence and hence
the correlation is of the form (1 eect of other vari-
ables). Whereas correlation D does not pose any such
limit on the operating parameters and the eect of
physico-chemical properties and ow rate is assumed to
be observed from the beginning.
7.3. Empirical correlation using experimental data
The experimental data were analysed to get the de-
pendency of droplet size on individual parameters. This
was carried out to check whether the dependencies stay
constant or change over a range of parametric values.
The following dependencies emerged from the plot of
droplet diameter against a specic variable while keep-
ing all other variables constant:
d
p
~ Q
0:207
(where Q is in m
3
/s)
~ r
0:11
(up to 0.035 N/m beyond which droplet size is
independent of surface tension)
~ g
0:166
(this is similar to a system where viscosity
shows an exponent of 0.11)
Fig. 3. Comparison between predictions of correlation A and experi-
mental droplet sizes.
Fig. 4. Comparison between predictions of correlation D and experi-
mental droplet sizes.
250 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
~ q
0:274
(it should be directly proportional and hence
this dependency is acceptable)
~ (power/area)
0:4
(this exponent is consistent with
the correlation reported in literature).
Based on the above dependencies arrived at from the
experimental data, a totally empirical correlation to
predict the droplet size is proposed:
d
p
= constant(f )
0:66
(Q)
0:207
(r)
0:11
(q)
0:274
(g)
0:166
(power=area)
0:4
All the units are in SI system and Q is given in m
3
/s.
The frequency dependence (f
0:66
) arises from the logic
as used in deriving correlation A. All previous investi-
gations also indicate that the droplet size is inversely
proportional to the operating frequency with an expo-
nent of 0.66. The same dependence is assumed to be
valid here. The value of constant was obtained by tting
the predicted droplet sizes to the experimental drop
sizes. Thus, the value of constant was found to be 1600
(corresponding to SI units). This value will change if the
parameters used in the correlation are substituted in
units other than SI units. This constant is dierent from
those used in correlations A and D and is dimensional in
nature. A parity plot between the predicted droplet size
(using the recommended empirical correlation) and ob-
served droplet size was drawn (Fig. 5) and the values of
standard deviation and maximum deviation was calcu-
lated. The standard deviation and maximum deviation
were found to be 0.27 and 0.62 respectively. The corre-
lation does predict the trends observed with the vari-
ables tested out in this work. A further improvement can
be made by optimising the exponents.
The earlier correlations reviewed in the text predict
the dependence of one or two parameters explicitly i.e.
eect of viscosity or ow rate. No correlation is uni-
versal in character. They do not explain the eect of all
the variables such as liquid phase physico-chemical
properties (density, surface tension, and viscosity), liq-
uid ow rate, intensity and frequency of ultrasound. The
word ``universal'' characterises all the variables being
included in the correlation.
Although the correlation A explains the observed
data with some standard deviation the trends are not
correctly indicated. The empirical correlation is dierent
in form from correlation A, however the exponent val-
ues are more or less the same to those in correlation A.
The empirical correlation predicts the data and trends
more precisely than the other correlations. Thus, it can
be concluded that with the present state of knowledge
the empirical correlation has been recommended for use
for the prediction of droplet sizes in an ultrasonic ato-
miser.
7.4. Calculation of momentum and energy utilisation to
create new surfaces
The momentum carried by a single drop in the case of
ultrasonic atomisation is almost three orders of magni-
tude less compared to the momentum associated with a
drop ejected in a conventional two phase nozzles (see
Appendix C.1). This is a major advantage with atomi-
sation using ultrasound.
The total surface area created by formation of one
droplet is given by 4pR
2
. Here R is the radius of the
droplet. The total surface area generated is (n4pR
2
) and
n is the total number of droplets generated. When the
total surface area generated is multiplied by the surface
tension, this product would give the total surface energy.
The kinetic energy associated with the moving droplet is
given by [1=2qQ(f Am)
2
[. The sum of total surface en-
ergy and the kinetic energy associated with all droplets is
compared with the total energy delivered to the system
in order to nd the eciency of the device. It turns out
to be that the percentage of dissipated energy utilised in
creating new surface is in the range of 0.03% to 0.2%
depending on the liquid ow rate.
A similar analysis for a two uid nozzle indicates that
the % utilisation on the basis of dissipated energy turns
out to be two orders of magnitude lower then the per-
centage utilisation in ultra sound induced atomisation.
The above gures indicate that considerably more scope
exists to optimise the energy utilisation and improve
these numbers. The details of calculation are given in
Appendix C.2.
More interesting insights can be obtained on droplet
formation using ultrasonic energy by carrying out a
rigorous study on the following:
1. A robust method has to be developed for analysing
the experimental droplet sizes and to study the eect
of dierent geometry for liquid feed, spread on the vi-
brating surface and the ultrasonic frequency.
Fig. 5. Comparison between predictions of empirical correlation and
experimental droplet sizes.
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 251
2. The limiting value of viscosity of the liquid to be ato-
mised and the maximum intensity available for atom-
isation for the system of our interest (atomisation of
high viscous LAS) has to be investigated. A larger
viscosity range is to be covered for which we may re-
quire/have to design an equipment with higher power
rating and frequency.
3. Creation of Taylor instability is the primary reason
for droplet formation. The wavelength of the instabil-
ity needs to be of the same order as that of expected
droplet diameter and hence will be of fairly high fre-
quency (possibly in the ultrasonic range again).
4. A two stage system can be experimented wherein a
very thin lm is fed on the vibrating surface (possibly
force feed for viscous liquids) using low intensity ul-
trasound in the rst stage to thin the liquid and then
atomise this thin lm using high intensity ultrasound
in the second stage. This would work eectively if the
liquid of high viscosity is of shear thinning type.
5. Mixing of air can be studied as an option to increase
the eciency of atomisation process. However, this
may not be an eective way as the ultrasonic coupling
would be adversely aected and the actual energy de-
livered for liquid sheet fragmentation would be re-
duced. Nevertheless, air can be used to reduce lm
thickness on the vibrating surface by spreading the
liquid and better atomisation can be achieved. This
would require much lesser pressure as compared to
that required in the two-phase atomisation using
the conventional nozzle.
6. Based on the momentum balance calculations it is
clear that considerable amount of energy gets dis-
sipated for generation and growth of instabilities
before the atomisation actually occurs. Thus there
exists a possibility of separating these two stages of
the growth of instabilities and the atomisation for
better energy utilisation. This can be achieved by al-
teration of the equipment involving two stages.
8. Conclusions
The correlations proposed in this work capture the
inuence of physico-chemical properties of the liquid
and ultrasonic properties reasonably well. Apart from
the inuence of ultrasonic frequency, the proposed
correlations predict the inuence of ultrasonic intensity
also. The correlations also take into consideration the
eect of viscosity and liquid ow rate, which are very
relevant for industrial applications. The signicant con-
tribution of this work is to dene dimensionless numbers
incorporating ultrasonic parameters, taking cue from
the conventional numbers which dene the signicance
of dierent forces involved in droplet formation. The
universal correlation proposed are robust and can be
used for designing ultrasonic atomisers for dierent ap-
plications.
Acknowledgements
The authors sincerely thank Dr. C.V. Natraj, Direc-
tor, HLRC and Dr. V.R. Dhanuka for their support and
encouragement. Thanks are due to Dr. Girish Rao for
his valuable suggestions and help in image analysis of
droplets. RR and ABP thank Hindustan Lever Research
Centre and UDCT, Mumbai for the supports rendered
for this collaborative work.
Appendix A. Comparison of energy requirements in a
conventional two phase (air assisted) nozzle and in an
ultrasonic atomiser
Let us consider a two phase (air assisted) external mix
type nozzle for drop formation. The liquid feed rate is 11
lph and the air ow rate is 79 lpm. The feed pressures of
liquid and gas are 3 and 2.5 atm respectively. The nozzle
diameter is 2 mm giving liquid and gas velocities
through the nozzle as v
l
= 1:03 m/s and v
g
= 55:28 m/s
respectively. The ratio of gas loading to liquid loading is
taken to be 430.
Walzel correlation is used to determine the droplet
size for this two phase nozzle atomisation. Appropriate
constants are taken from literature and the values of m
and j are taken as 0.4 and unity. After inserting the
values of constants, and exponent values m and j, we
have
d
32
d
j
= 0:35
DP
g
1 l
p
_ _
2
_ _
0:4
1 ( 2:5Oh)
Here l
p
is the ratio of mass ow rates of liquid to gas
and its value is calculated to be 1.785 and d
j
is the di-
ameter of exit liquid jet which is taken to be equal to the
nozzle diameter (d
ori
).
The reduced pressure drop DP
+
g
is given by
DP
+
g
= DP d
ori
=r;
here d
ori
is the nozzle diameter and r is the surface
tension. Assuming that all the liquid and gas pressure
drops while passing through the nozzle, we can calculate
DP
+
g
. The Ohnesorge number is calculated as
Oh =
g

qrd
ori
_ = 0:00208
Now, we substitute the values in Walzel correlation to
get the Sauter mean diameter of the droplet.
d
32
=0:002 = 0:35[(2:5 10
5
0:002=0:072)
=(1 1:785)
2
[
0:4
[1 2:5 0:00208[
252 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
The value of d
32
is found to be 46.44 lm.
Now, let us calculate the total energy supplied to
form the droplets of size 46.44 lm.
(1) Energy associated with liquid:
(DP
1
Q
1
) (1=2q
l
v
2
l
) = 3 10
5
11=3600 10
3

0:5 1000 v
2
l

Substituting the value of v
l
in the energy equation, we
get the energy supplied through the liquid as 531.37 W.
(2) Similarly the energy associated with the gas ow is
to be calculated.
(DP
g
Q
g
) (1=2q
g
v
2
g
) = 2:5 10
5
79=60 10
3

0:5 1:3 v
2
g

Here, v
g
= 55:28 v
l
and from the value of v
l
, net v
g
is
calculated to be 56.94 m/s. Substituting the value of v
g
in
the energy equation, we get the energy supplied through
the gas as 2436.57 W. The total energy associated
with the gas and liquid ow is 2967.94 W.
Thus, a total energy of 2967.94 W (~3 kW) is utilised
to disintegrate the liquid sheets to produce droplets of
size 46.44 lm, for the assumed ow rates and nozzle
dimensions.
Let us now calculate the energy that would be re-
quired to split the same liquid ow rate into droplets of
size 46.44 lm in an ultrasonic atomiser. Let the me-
chanical displacement of the vibrating surface produce a
volumetric displacement rate which is three times the
liquid ow rate used in the process.
For an energy consideration, we can assume the
amplitude of vibrations to be 5 lm. Making use of the
Lang correlation, the frequency required for producing
a droplet of 46.44 lm is found to be 26.65 kHz. In order
to get a volumetric displacement rate of vibrating sur-
face as 33 lph (three times the liquid ow rate to be
disintegrated), the area of the vibrating surface to be
Volume displacement rate = f Am (area)
33=36 10
5
= (26:65 10
3
)(5 10
6
) (area)
area = 6:88 10
5
m
2
(0:69 cm
2
)
The rate of energy dissipation required
= 1=2qC(2pf Am)
2
(area)
= 500 1500 (2 p 26:65 10
3
5 10
6
)
2
(6:88 10
5
)
= 36:17 W
The net power to be expended is 36.17 W whereas the
power requirement in the conventional nozzle system
considered here is 2967.94 W. Assuming 50% eciency
for pumps and compressors in conventional two uid
atomisation for the set of constraints and conditions
described in this analysis, the amount of electrical en-
ergy required is 2 (2967:94) = 5935:88 W (~6 kW).
For ultrasonic atomisation using an equipment which is
10% ecient (typical) the requirement of electrical en-
ergy is 10 36:17 = 361:7 W (~0.4 kW). This is about
6% of the electrical energy required for a conventional
two uid nozzle (atomisation) system.
Appendix B. Calculation of lm thickness for a specic
liquid loading rate
The minimum lm thickness must be few times
(~three times) the threshold amplitude of the ultrasonic
wave in order that the complete growth of instabilities
occur before droplet is ejected from the edges of the
liquid sheets/ligaments. One can calculate the liquid
loading rate that should be maintained for maintaining
sucient lm thickness on the vibrating surface using
the equation for falling lm. One would think that this
should decide the optimal ow rate Q
opt
. If one oper-
ates below Q
opt
, sucient lm thickness would not be
maintained and energy is not eectively used and above
this Q
opt
throughput rate is made higher but the droplet
size would be bigger. Let us nd out Q
opt
for a typical
situation which would maintain a liquid lm having
thickness three times the threshold amplitude.
Let the threshold amplitude be Am
crit
and the lm
thickness is 3 Am
crit
.
Am
crit
= (2g=q)(q=prf )
1=3
For water with 28 kHz frequency, Am
crit
turns out to be
1.1 lm. Now, let us calculate the liquid loading rate
required to maintain the liquid lm with a thickness of
3.3 lm using the falling lm equation [40].
H =
3sg
gq
f
(q
f
q
c
) sinh
_ _
1=3
Here H is the lm thickness on the solid surface, s is the
loading per unit width of the surface, g is the liquid
viscosity, g is the acceleration due to gravity, q
f
is the
falling liquid density, q
c
is the surrounding uid (air)
density, and h is the angle of inclination with respect to
horizontal. Let us assume that h is 90. For a lm
thickness of 3.3 lm, the liquid loading per unit width
must be 4:23 10
4
kg/mh. Thus in volumetric terms,
assuming the surface to be cylindrical (horn) with a di-
ameter of 10 mm,
Q
opt
= (s=q
f
)(pD
sur
)
= 4:23 10
4
p 0:01=1000 m
3
=h
= 1:33 10
8
m
3
=h:
Now, one must check whether this Q
opt
is below Q
crit
as it is not advisable to operate below Q
crit
because
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 253
neither one would get reduced drop size nor the energy
eciency is signicant compared to what would be
obtained with Q
crit
. The Q
crit
for water with 28 kHz
frequency is (by equating critical Weber number to
unity)
Q
crit
= r=qf = 9:26 10
6
m
3
=h:
As Q
opt
Q
crit
, one should operate with critical ow
rate. In other words, when the liquid ow rate is Q
crit
, it
is made sure that sucient thickness of liquid lm is
maintained on the vibrating surface.
Using the falling lm equation, the liquid lm
thickness corresponding to Q
crit
and Q
max
(Q
max
is given
by the volumetric displacement rate of the vibrating
surface) are found to be 29.27 and 138.68 lm respec-
tively assuming the amplitude of 1.1 lm at 28 kHz
frequency. Although Q
opt
decides the minimum lm
thickness required, as Q
opt
would always be lower than
Q
crit
, in practical applications the minimum lm thick-
ness that one should maintain would be dictated by the
critical liquid ow rate and the maximum thickness that
can be allowed is determined by the maximum liquid
ow rate i.e. the volumetric displacement rate of the
vibrating surface above which dripping of liquid would
occur.
Appendix C
C.1. Momentum associated with the drops formed
Assumption: No slip condition
For two phase nozzle:
The momentum associated with a single drop =(mass
of drop) (ejection velocity): Here the ejection velocity is
taken as v
g
v
l
(see Appendix A) =58 m/s; the size of
the droplet (d
p
) is 46.5 lm; mass of drop = q
1
(p=6)d
3
p
=
5:26 10
11
kg. Thus, the momentum associated with a
single drop is 3:05 10
9
kg m/s.
If the same size of droplet is obtained using ultra-
sound with a frequency of 20 kHz and an amplitude of
1.12 lm (Table 3), the momentum associated with a
single drop in this case works out to be q
l
(p=6)d
3
p
=
5:26 10
11
kg; the ejection velocity is the product of
frequency and amplitude of ultrasonic wave which is
2.24 10
2
m/s. Thus, the momentum associated with a
single drop atomised using ultrasound is 1.18 10
12
kg m/s.
The calculation clearly proves a point that the mo-
mentum carried by a drop is three orders of magnitude
less in the case of ultrasonic atomisation. This is of great
advantage in situations like coating and granulation
where the impact force has to be minimal.
C.2. New surface creation eciency
C.2.1. Sample calculation: two phase nozzle (Data as
given in Appendix A)
Flow rate (Q) 11 lph = 3:05 10
6
m
3
/s; d
p
= 46:5
lm;
Number of drops being generated per second (n) =
Q=(p=6)d
3
p
= 58,070,147 drops/s.
Rate of creation of new surface (A) = npd
2
p
= 0:394
m
2
/s. Thus, rate of generation of surface energy (E
s
) =
(A)(r) = 0:0284 W. Here, r is the surface tension of
water which is taken as 0.072 N/m.
Similarly, we can calculate the kinetic energy associ-
ated with the drops which turns out to be four orders of
magnitude less compared to the surface energy. Energy
dissipation rate (refer Appendix A) 2968 W. Thus, the
percentage energy utilisation as surface energy 100
(0:0284=2968) = 9:57 10
4
%:
C.2.2. Sample calculation: ultrasonic atomisation
Flow rate (Q) 0.44 lph = 1:22 10
7
m
3
/s; d
p
= 150
lm.
Number of drops being generated per second (n) =
Q=(p=6)d
3
p
= 69,199 drops/s.
Rate of creation of new surface (A) = npd
2
p
=
4:88 10
3
m
2
=s.
Thus, rate of generation of surface energy (E

s
) =
(A

) (r) = 3:52 10
4
W.
Here, r is the surface tension of water which is taken
as 0.072 N/m.
Similarly, we can calculate the kinetic energy associ-
ated with the drops which turns out to be four orders of
magnitude less compared to the surface energy.
Energy dissipation rate (Table 3) 14,954 7:5
10
5
= 1:12 W; Thus, the percentage energy utilisation
as surface energy 100 (3:52 10
4
=1:12) = 0:0314%:
From the above sample calculations, it can be con-
cluded that although the energy utilisation in creating
new surface is low in both the cases, the % utilisation in
ultrasonic atomisation is 35 times more than that in the
two phase nozzle.
References
[1] R.L. Wilcox, R.W. Tate, Liquid atomization in a high intensity
sound eld, AIChE. J. 11 (1) (1965) 6972.
[2] T. Masato, Local droplet diameter variation in a stirred tank,
Can. J. Chem. Engng. 63 (1985) 723727.
[3] N. Dombrowski, R.P. Fraser, A photographic investigation into
the disintegration of liquid sheets, Phil. Trans. A. 247 (1954) 101
130.
[4] J.E. Swartz, D.P. Kessler, Single drop break-up in developing
turbulent pipe ow, AIChE. J. 16 (2) (1970) 254260.
254 R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255
[5] M.M. Clark, Drop break up in a turbulent ow I. Conceptual
and modelling considerations, CES 43 (3) (1988) 671679.
[6] J.O. Hinze, Fundamentals of the hydrodynamic mechanism of
splitting in dispersion forces, AIChE. J. 1 (1955) 289295.
[7] V.M. Entov, A.L. Yarin, The dynamics of thin liquid jets in air,
J. Fluid Mech. 140 (1984) 91111.
[8] R. Bellman, R.H. Pennington, Eects of surface tension and
viscosity on Taylor instability, Quant. Appl. Math. 12, 151 (1954).
[9] J.T. Davies, Drop sizes of emulsion related to turbulent energy
dissipation rates, CES 40 (5) (1985) 839842.
[10] J. Kitscha, G. Kocamustafaogullari, Breakup criteria for liquid
particles, Int. J. Multiphase Flow 15 (4) (1989) 573588.
[11] D.M. Newitt, A.I.B.P. Dombrowski, F.H. Knelman, Liquid
entrainment, Trans. Instn. Chem. Engrs. 32 (1954) 244261.
[12] G.E Charles, S.G. Mason, The mechanism of practical coales-
cence of liquid drops at liquid/liquid interfaces, J. Colloid. Sci. 15
(1960) 105122.
[13] P. Walzel, Liquid atomization, Int. Chem. Engng. 33 (1) (1993)
4660.
[14] A.B. Pandit, J.F. Davidson, Hydrodynamics of the rupture of thin
liquid lms, J. Fluid Mech. 212 (1990) 11.
[15] R.J. Lang, Ultrasonic atomization of liquids, J. Acoust. Soc. Am.
34 (1962) 6.
[16] Y.Y. Bouguslavskii, O.K. Eknadiosyants, Physical mechanism of
the acoustic atomization of a liquid, Soviet Phys. Acoust. 15
(1969) 14.
[17] R.L. Peskin, R.J. Raco, J. Acoust. Soc. Am. 35 (9) (1963) 1378.
[18] B.W. Hansen, High-speed photographic studies of droplet
formation at 20 kHz ultrasonic atomization of oil, Ultrasonics 8
(2) (1970) 9799.
[19] M.N. Topp, P. Eisenklam, Industrial and medical uses of ul-
trasonic atomizers, Ultrasonics 10 (1972) 127133.
[20] T. Mochida, Proc. ICLASS-78, 1978, p. 193.
[21] R.L. Peskin, R.J. Raco, J. Acoust. Soc. Am. 43 (1967) 229.
[22] J.M. Mir, Cavitation-induced capillary waves in ultrasonic
atomization, J. Acoust. Soc. Am. 67 (1) (1980) 201205.
[23] D. Sindayihebura, L. Bolle, A. Cornet, L. Joannes, Theoretical
and experimental study of transducers aimed at low-frequency
ultrasonic atomization of liquids, J. Acoust. Soc. Am. 103 (3)
(1998) 14421448.
[24] A. Morgan, Ultrasonic Atomization, Advances in Sonochemistry,
vol. 3, 1993, pp. 145164.
[25] M.N. Topp, P. Eisenklam, Proc. Ultra. Int. Conf., Imperial
College, Butterworths, 1969, p. 24.
[26] H.S. Fogler, K.D. Timmerhaus, Ultrasonic atomization study,
J. Acoust. Soc. Am. 39 (1966) 515518.
[27] H.H. Hunter, A small ultrasonic atomizer for liquid fuels,
Ultrasonics 7 (1) (1969) 6364.
[28] N. Maehara, S. Nakane, K. Yamamoto, K. Iga, A pinhole-plate
ultrasonic atomizer, Ultrasonics 22 (6) (1984) 259260.
[29] S. Ueha, N. Maehara, E. Mori, Mechanism of ultrasonic at-
omization using a multi-pinhole plate, J. Acoust. Soc. Jpn. (E) 6, 1
(1985) 2126.
[30] N. Maehara, S. Ueha, E. Mori, Optimum design procedure for
multi-pinhole plate ultrasonic atomizer, Jpn. J. Appl. Phys. 26
(Suppl. 26-1) (1987) 215217.
[31] N. Maehara, S. Ueha, E. Mori, Inuence of the vibrating system
of a multipinhole-plate ultrasonic nebulizer on its performance,
Rev. Sci. Instrum. 57 (11) (1986) 28702876.
[32] N. Maehara, S. Ueha, E. Mori, Atomizing rate control of a multi-
pinhole plate ultrasonic atomizer, J. Acoust. Soc. Jpn. 44(2) (1988)
116121 (in Japanese).
[33] N. Maehara, S. Ueha, E. Mori, Inuences of liquids physical
properties on the characteristics of a pinhole-plate ultrasonic
atomizer, J. Acoust. Soc. Jpn. 44(6) (1988) 425431 (in Japa-
nese).
[34] S. Majumdar, P.R. Senthilkumar, A.B. Pandit, Emulsication by
ultrasound, relation between intensity and ultrasound quality,
Ind. J. Chem. Tech. 4 (1998) 277284.
[35] S.C. Tsai, P. Childs, P. Luu, Ultrasound-modulated two-uid
atomization of a water jet, AIChE. J. 42 (12) (1996) 33403350.
[36] M.M. Clark, Drop breakup in a turbulent ow II. Experiments
in a small mixing vessel, CES 43 (3) (1988) 681692.
[37] R. Pohlman, K. Heisler, M. Cichos, Powdering aluminium and
aluminium alloys by ultrasound, Ultrasonics 12 (1) (1974) 1115.
[38] L. Rayleigh, Proc. London Math. Soc. 10, 4 (1878).
[39] J.T. Davies, Turbulence Phenomena, Academic Publishers, Lon-
don, 1972.
[40] R.H. Perry, D. Green (Eds.), Perrys Chemical Engineers Hand-
book, sixth ed., McGraw Hill, New York, 1984, p. 5.59.
R. Rajan, A.B. Pandit / Ultrasonics 39 (2001) 235255 255

You might also like