Pharmaceutical Development and Technology, 13:93–104, 2008

Copyright © Informa Healthcare USA, Inc.

ISSN: 1083-7450 print / 1097-9867 online
DOI: 10.1080/10837450701830957

LPDT

Droplet Size Measurements for Spray Dryer Scale-Up

Pia Thybo and Lars Hovgaard

Droplet Size Measurements for Spray Dryer Scale-Up

Department of Pharmaceutics and Analytical Chemistry, Faculty of Pharmaceutical Sciences,

University of Copenhagen, Copenhagen, Denmark
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13

Sune Klint Andersen and Jesper Sæderup Lindeløv

Niro A/S, Soeborg, Denmark

guideline for selecting appropriate operating conditions when

This study was dedicated to facilitate scale-up in spray scaling up the spray-drying process.
drying from an atomization standpoint. The purpose was to
investigate differences in operating conditions between a pilot Keywords spraytec, two-fluid nozzle, atomization, scale-up,
and a production scale nozzle. The intension was to identify the droplet size, spray drying
operating ranges in which the two nozzles produced similar
droplet size distributions. Furthermore, method optimization
and validation were also covered. Externally mixing two-fluid
For personal use only.

nozzles of similar designs were used in this study. Both nozzles
are typically used in commercially available spray dryers, and
they have been characterized with respect to droplet size distri-
butions as a function of liquid type, liquid flow rate, atomization
gas flow rate, liquid orifice diameter, and atomization gas ori-
fice diameter. All droplet size measurements were carried out by
using the Malvern Spraytec with nozzle operating conditions
corresponding to typical settings for spray drying. This gave
droplets with Sauter Mean Diameters less than 40 μm and typi-
cally 5–20 μm. A model previously proposed by Mansour and
Chigier was used to correlate the droplet size to the operating
parameters. It was possible to make a correlation for water
incorporating the droplet sizes for both the pilot scale and the
production scale nozzle. However, a single correlation was not
able to account properly for the physical properties of the liquid
to be atomized. Therefore, the droplet size distributions of etha-
nol could not be adequately predicted on the basis of the water
data. This study has shown that it was possible to scale up from
a pilot to production scale nozzle in a systematic fashion.
However, a prerequisite was that the nozzles were geometrically
similar. When externally mixing two-fluid nozzles are used as
atomizers, the results obtained from this study could be a useful
Received 18 June 2007, Accepted 7 September 2007.
Address correspondence to Lars Hovgaard, Novo Nordisk A/S
Novo Nordisk Park, DK-2760, Maaloev, Denmark; E-mail:
lhvg@novonordisk.com
INTRODUCTION
Atomization is the process by which a liquid is
disintegrated into droplets. It is considered as one of the
key steps in several pharmaceutical unit operations such as
spray drying, spray-freeze drying, coating, and granula-
tion. A successful outcome of these processes will indeed
rely on a controlled and efficient atomization resulting in
reproducible and controllable droplet size distributions.
This is of great importance to the production of certain
pharmaceuticals, where a specific particle size distribution
and aerodynamic diameter is required (e.g., in the production
of inhalable particles).[1] Spray drying is a well-established
process in the production of solid particles in the pharma-
ceutical industry.[2–4] In spray drying, the final particle
size distribution is in general directly proportional to the
droplet size distribution created by the atomizer.[5] There-
fore, determination of the droplet size distribution,
coupled with a powder mass balance, will usually give a
good estimate of the final particle size of the spray-dried
product. Thus, it gives an idea of the operating conditions
for a given spray-drying process when it is desired to scale
up from laboratory scale to pilot or production scale spray
dryer. In spray drying, scaling up will often require the use
of an atomizer with larger dimensions. Few studies have
been carried out with respect to droplet size measurements
of differently sized atomizers used for spray drying.[6]
Furthermore, the measurements can be used to identify

93
 94 P. Thybo et al.
adequate atomizers for either producing the same average
droplet size or a different average droplet size. This is an
important issue because laboratory scale spray dryers are
often limited to particle diameters less than 15 μm,
whereas often much larger particles are desired for, e.g.,
tabletting.
Various methods have been developed for the
determination of droplet sizes. Several of these techniques
are based on laser diffraction. The use of the Malvern
Spraytec for measuring particle or droplet sizes is cur-
rently very limited, and only a few pharmaceutical articles
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13
could be found in the literature.[7–13] In addition, many of
these studies use the Malvern Spraytec installed with the
inhalation cell attachment, and they do not evaluate the
setup of the instrument. None of these reports covers the
aspect of atomization scale-up.
The mechanism of atomization of liquid jets is a
complex process that is still not fully understood despite
numerous theoretical and experimental studies performed
by Lefebvre, Rayleigh, Tyler, Weber, Haenlein, Ohnesorge,
and Castleman. For a comprehensive review please, see
Lefebvre 1989.[16] The model that has found the widest
acceptance is the wave mechanism, which is illustrated in
Figure 1. According to this model, interaction between
an atomization gas and a liquid jet causes waves to form
For personal use only.
on the surface of the liquid jet. This causes the liquid jet
to shed ligaments that rapidly form droplets. The waves
are generated by pressure fluctuations or turbulence in the
gas stream. If aerodynamic pressure forces dominate, as
they do at very high atomization gas velocity, the waves
will continue as acceleration waves and the droplet
diameter is proportional to (Z/We)2/3, where Z is the
Ohnesorge number and We is the Weber number. If sur-
face tension dominate, the waves continue as capillary
waves and the droplet diameter will be proportional with
(Z2/We).[18,19]
This classical mechanism satisfactorily predicts mean
droplet size by using equations that express mean droplet
Figure 1. Disintegration of sheet.[17]
diameter divided by a characteristic nozzle dimension in
terms of the We, Z, and the gas/liquid jet ratio (ALR).[20]
Nevertheless, Lefebvre[20] questioned if Z should be
calculated by using the liquid or gas density. For low vis-
cosity liquids, such as water, the term including Z is sig-
nificantly smaller than the term including the We number.
Therefore, the density of the gas has little importance.
This is not the case for highly viscous liquids and high
gas densities.
According to Lefebvre,[20] the atomization mecha-
nism in cases where the atomization gas velocity is very
high, or the gas impinges on the liquid jet at an appreciable
angle, can be described as occurring by the “prompt”
mechanism. According to this model, droplet size is nearly
independent of liquid viscosity and initial liquid jet diame-
ter. Both mechanisms of atomization will occur for a two-
fluid nozzle with the “prompt” mechanism being favored
at high We.
The first part of this article deals with the optimiza-
tion and validation of a method to evaluate the influence
on spray droplet size distribution and atomization quality.
Variables examined include atomization gas flow rate,
liquid flow rate, liquid orifice diameter, atomization gas
orifice diameter, and liquid properties. The second part
deals with up-scaling. A semiempirical model correlating
droplet size to operational parameters is proposed.
MATERIALS AND METHODS
Materials
Demineralized water from in-house supply was used.
Ethanol (CAS 64-17-5) 99.9% was supplied by V&S
Distillers (Aalborg, Denmark), acetaminophen (CAS
103-90-2) from Mallinckrodt Inc (NC, USA), and polyvi-
nylpyrrolidone K-30 (PVP K-30; CAS 9003-39-8) was
obtained from ISP Switzerland AG (Baar, Switzerland).
Experiments were carried out with demineralized water,
ethanol 99.9%, and solutions of acetaminophen and PVP
K-30 in ethanol. The solutions covered acetaminophen/
povidone w/w ratios of 1:2 and 1:9 at concentrations of 10
and 20 w/w%. Nitrogen (≥ 99.5%, pharmacopoeia grade)
was used as atomization gas.
Two-Fluid Nozzles
Two externally mixing two-fluid nozzles were used in
this study. One is used primarily in pilot scale spray dryers
with water evaporation capacities of 0.5–7.0 kg/hr
(hereafter referred to as pilot scale nozzle). The other is
 Droplet Size Measurements for Spray Dryer Scale-Up
mainly used in production scale spray dryers with water
evaporation capacities of 20–100 kg/hr (hereafter referred
to as production scale nozzle).
The pilot scale nozzle was a Niro Mobile Minor two-
fluid Nozzle (Niro A/S, Soeborg Denmark) mounted with
a 0.5, 0.8, or 1.0 mm liquid orifice diameter (3.0 mm in
outer diameter) and 5.0 mm atomization gas orifice diameter.
The production scale nozzle was a standard Schlick-05
two-fluid nozzle (Düsen-Schlick, Untersiemau, Germany)
with a liquid orifice diameter of 4.0 mm (6.0 mm in outer
diameter) and 9.0, 10.0, or 11.0 mm atomization gas
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13
orifice diameter. The designs of the nozzles are illustrated
schematically in Figure 2, from which it can be observed
that even though the two nozzles come from different
manufacturers, they are similar in design.
Methods
Determination of Droplet Size Distributions
The experimental setup is shown schematically in
Figure 3 and consists of a RTS 5114 Malvern Spraytec
(Malvern Instruments, Worcestershire, UK), an adjust-
able holder for the nozzle, a ventilation unit for collec-
For personal use only.
tion of the spray, a liquid system, and an atomization
system. The liquid system includes a pump to control
the liquid flow rate [peristaltic pump for low flow rates
(Watson Marlow, Cornwall, England) or a mono-pump
for high flow rates (Netzsch Mohnopumpen GMBH,
Waldkraiburg, Germany)] and a balance (Mettler PC
4400, Mettler instruments, Zürich, Switzerland) or
flowmeter (Endress + Hauser Promag 50, Endress Hauser,
Figure 2. Configuration of the co-axial externally mixing two-fluid nozzles. The number of vanes in the gas distributors are 8 for
the MobileMinor and 12 for the Schlick 05 two-fluid nozzle.
95
Germany) for measuring the liquid flow rate. The
atomization gas system includes a rotameter for con-
trolling the atomization gas flow rate (model D10A11;
tube 1/2″-27-G-10 or 1″-35-G-10; float 1/2″-GSVT-48
or 1″-GSVT-65; ABB Automation, Taastrup, Denmark)
and membrane valves for regulating the pressure inside
the system.
The Malvern Spraytec was equipped with an air
purge system for the lenses to avoid contaminating the
lens by incoming droplets. In all experiments a lens
with a focal length of 200 mm was used. The operating
range for this lens was 1–460 μm. Even though nitro-
gen was supplied as atomization gas, the refractive
index of air of 1.00 + 0.00i was used in all calculations.
This was done because of the atomization nitrogen gas
mixing with the ventilation air in the measurement
zone. Furthermore, the refractive indexes of air and
nitrogen gas are virtually identical.[21] Therefore, it
was found reasonable to use the value for air. For the
pure solvents, the standard values 1.33 + 0.00i and 1.36 +
0.00i were used for water and ethanol, respectively.
The refractive indexes for the solutions were measured
and used. Droplet size measurements were recorded
continuously for 60 s and analyzed in accordance with
the Mie theory. The RTsizer software V.5.40 (Malvern
Instruments, Worcestershire, UK) was used for analyz-
ing the droplet size data, and results were reported in
terms of Sauter Mean Diameter (SMD), Volumetric
Mean Diameter (VMD), and span.
The position of the nozzle and other settings will
vary according to the experiment performed. Therefore,
the exact operating conditions will be specified in each
section.
 96 P. Thybo et al.
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13
Figure 3. Schematic illustration of the setup for the droplet
size measurements by laser diffraction. In the setup, the vertical
position is restricted (in and out of the plan in this illustration)
(i.e., in all experiments the laser beam is aligned with the center
For personal use only.
of nozzle).
Surface Tension
The measurements of surface tension were performed
at 20°C by the Wilhelmy plate method on a K10T Krüss
tensiometer (Krüss GmbH, Hamburg, Germany).
Viscosity
The dynamic viscosities of the liquids were measured
on a ARG2 Rheometer (TA Instruments Ltd, West Sussex,
UK), installed with a double gap peltier cylinder system
and a double gap rotor for concentric cylinder system (TA
Instruments Ltd). The rheology advantage software
V.5.4.0 (TA Instruments Ltd) was used for data analysis.
All measurements were carried out with the steady-state
flow procedure at 20°C.
Refractive Index
The refractive indexes of the solutions were determined
by using a RL 3 laboratory refractometer (Polish Optical
Company, Warsaw, Poland). The refractive indexes of all
solutions were measured to 1.38. The imaginary part was
not determined, and a value of 0.00i (similar to water and
ethanol) was used for the spraytec measurements.
RESULTS AND DISCUSSION
Characterization of Solvents and Solutions
The solvents and solutions were characterized in
terms of dynamic viscosity and surface tension because
these values are important for the atomization process. All
solutions tested exhibit Newtonian flow behavior, because
they gave linear shear rate versus shear stress plots that
pass through the origin. The measured viscosities and
surface tensions of the two solvents (water and ethanol)
and all the solutions (PVP K-30 and Acetaminophen in
ethanol) are shown in Table 1. As expected, an increase in
polymer concentration resulted in a higher viscosity. How-
ever, the addition of up to 18% PVP K-30 did not change
the surface tension of the solutions (i.e., PVP K-30 does
not change the surface tension of ethanol). Therefore, the
effect of PVP K-30 on ethanol solution viscosity and
droplet size distribution could be studied without the sur-
face tension being a variable.
Droplet Size Measurements–Pilot Scale Nozzle
Initially, a set of experiments were carried out for the
position of the nozzle (horizontal and vertical distance, and
height) relative to the Spraytec laser beam and detector as
well as the ventilation rate for spray collection. All measure-
ments were carried out at a room temperature of 15–20°C.
Ventilation Rate
A ventilation system was installed to keep the ethanol
vapor concentration at least a factor 10 below the 3.3 vol%
explosive limit of ethanol.
The influence of the ventilation rate on the droplet size
measurements was investigated by using water, because it
was assumed that ethanol would behave like water as the
ventilation rate varied. No variation in droplet size could be
observed for ventilation rates varying from 0 to 5000 kg/hr.
A ventilation rate of 1500 kg/hr was used for all the
following measurements, which ensured safe operation.
Repeatability
The repeatability of the Spraytec measurements was
tested with water. The pilot scale nozzle with a 1.0-mm
orifice diameter, a liquid flow rate of 1.5 kg/hr, and a gas
flow rate of 7.5 kg/hr was used for these tests.
The test included turning off the liquid flow, the
atomization gas flow, and the ventilation system, and then
turning them on again with the same settings as before.
A couple of tests were also carried out where, in addition
 Droplet Size Measurements for Spray Dryer Scale-Up
Table 1
Dynamic viscosity and surface tension of solvents and solutions
Total solid Acetaminophen
conc. [w/w%] conc. [w/w%]
Dem. water NA NA
Ethanol 99.9% NA NA
Solution (1:2) 10 3.3
Solution (1:9) 10 1.0
Solution (1:2) 20 6.7
Solution (1:9) 20 2.0
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13
to the above, the nozzle was removed, all parts detached,
put together again, and the nozzle repositioned. A total of
10 tests were carried out by using operating conditions that
ensured good atomization.
Under the given circumstances, an average SMD of
14.4 μm (standard deviation 0.8) was found when turning
off the liquid flow, the atomization gas flow, and the
ventilation system, and an SMD of 14.3 μm (standard
deviation 1.4) when turning off and disassembling the
whole system. On the basis of these tests it was concluded
that the repeatability was good.
For personal use only.
Nozzle Position-Vertical Position Relative to
the Laser Beam
The instrument setup used in this study does not easily
facilitate variation of the vertical position of the nozzle rela-
tive to the laser beam. In the available setup of the instru-
ment, the nozzle holder was mounted on an unmovable
horizontal bar restricting the capability to move the nozzle
up and down with respect to the laser beam. Studying the
effect of the vertical position would require a major recon-
struction of the entire setup. It was therefore decided to
carry out all measurements in this study at the center of the
spray i.e. the laser beam aligned with the nozzle center. In a
previous study by Simmons and Hanratty[9] the droplet size
varied with vertical measurement position which was
explained by gravitational settling of the droplets. However,
as the droplets in this study come out of the nozzle with high
velocity (estimated from momentum balance to ≥ 50 m/s),
the distance between nozzle exit and laser beam is short.
Therefore, gravitational settling is not expected to have a
significant influence on the measurements under the given
operating conditions.
Nozzle Position–Horizontal Position Relative to the
Detector (d1 in Figure 3)
The horizontal position of the nozzle is defined as the
distance from the center of the nozzle to the detector.
97
PVP K-30 Viscosity Surface tension
conc. [w/w%] [10− 3 Pa·s] [10− 3 N/m]
NA 0.9 72
NA 1.2 23
6.7 4.0 23
9.0 5.3 23
13.3 14.1 23
18.0 20.2 23
Horizontal positions of 100–200 mm made it possible to
capture all droplets within the measurement window with-
out droplets hitting the lens. The measured SMDs for these
positions were similar. For externally mixing two-fluid
nozzles, as used in this study, the angle of the spray cone is
narrow (70–80°) and relatively independent of operational
parameters such as the atomization gas, liquid velocity, and
the liquid orifice diameter.[2] Therefore, it is expected that a
horizontal position of 150 mm between the center of the
nozzle and the detector facilitate valid measurements at
several positions downstream from the nozzle (see next
section). This position was therefore chosen as standard.
Nozzle Position–Downstream Distance Between
Nozzle and Laser Beam (d2 in Figure 3)
The downstream distance is defined as the axial
distance from the nozzle exit to the laser beam. The
influence of this parameter was evaluated by measuring
the droplet size distribution in the center of the spray at
distances ranging from 20 to 150 mm. The atomization gas
flow rate was kept constant at 9.0 kg/hr, and both water
and ethanol were sprayed. Because similar tendencies
were observed for the two fluids, only the measurements
with water are presented in Figure 4, which shows that,
initially, SMD increases linearly with distance from about
6.5 to 10 μm. After approximately 50 mm SMD is rela-
tively constant at 10 μm. Elverson et al.[22] also measured
the development of the droplet size distribution as a
function of distance from a two-fluid nozzle, but their
findings contrasted with the results obtained in this study.
Their measurements were done almost similar to this study
for the distances between the laser beam and the nozzle.
The nozzle, though, was much smaller than the one used
in this study, with much less liquid (0.3 kg/hr), a lower
atomization gas flow rate (2.2 kg/hr), and larger orifices
(1.5 mm and 2.0 mm). Furthermore, it is unknown if the
nozzle was of the internally mixing, externally mixing, or
air-blast type. It is hypothesized that the reason for the
increase in droplet size followed by a constant size with
 98 P. Thybo et al.
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13
Figure 4. Influence of the downstream distance from nozzle to
laser beam on the SMD. Atomizing gas flow = 9 kg/hr, flow rate
of water 1.5 or 2.5 kg/hr.
distance from the nozzle is caused by the turbulent nature
of the spray (Reynolds number of 23.000 for the atomiza-
tion gas). If the spray is turbulent, collisions in the vicinity
of the nozzle resulting in droplet growth is more likely.
But if the droplet concentration is sufficiently low,
For personal use only.
collisions could be avoided. The longer away from the
nozzle tip the lower is the concentration of the spray.
Therefore, the measured droplet diameters are similar, when
the measurements are performed at a certain distance away
from the nozzle, here approximately 50 mm. The reason
Elverson et al.[22] did not observe this phenomenon was
probably their low atomization gas flow rates, which did not
result in a turbulent flow or at least not a very turbulent one.
Other authors have found that droplet size distribu-
tions were influenced by droplet evaporation[11], but as the
droplet size increases for both water and ethanol and the
room temperature was kept at 15–20°C. Droplet evapora-
tion was regarded as less important in this study.
The spatial distribution of the spray and the droplet
concentration are not similar at all measurement positions.
A more dense spray could lead to multiple scattering and
incorrectly skew the distribution towards smaller
droplets.[8] To eliminate this effect, all data were corrected
for multiple scattering.
In conclusion, a change in the downstream distance
from nozzle to laser beam influences the droplet size
distribution; Therefore, it is necessary to specify the
exact axial position of the nozzle in relation to the laser
beam. For the resulting droplet size measurements on
the pilot scale nozzle, a downstream distance of 70 mm
from the nozzle to the laser beam was chosen as the
droplet size distribution for the given operating condi-
tions did not evolve further.
Influence of Atomisation-Gas-Liquid-Mass-Ratio
(ALR), Liquid Orifice Diameter, and Liquid
Flow Rate
According to Masters,[2] the most important variable
involved in the control of droplet size for a two-fluid
nozzle is the mass ratio of atomization gas flow rate to liq-
uid flow rate (ALR). An increase in this ratio should cause
a decrease in droplet size. As seen in Figure 5 all the
results obtained in this study are in agreement with this
statement. At a critical ALR value the atomization gas
reaches its maximum velocity, which is the speed of sound
for a convergent nozzle, and subsequent increases in the
atomization gas flow rate only increases the atomization
gas pressure (density). This will have a much smaller
effect on atomization than the atomization gas velocity
(i.e., the SMD is constant).
From Figure 5 it can be seen that ALR is not the
only factor controlling droplet size. A change in the liq-
uid exit orifice diameter could also alter the droplet size.
Figure 5 shows a general increase in SMD as the liquid
orifice diameter is increased and is consistent for atomi-
zation of both water and ethanol. The effect is more pro-
nounced when working with lower ALR. Thus, the
findings from this study are in agreement with the
results of Liu et al.[23] and Rick and Lefebvre,[24]
whereas other studies showed only a small influence of
liquid orifice diameter on droplet size distribution for
liquids of low viscosity.[25] From Figure 5 it can also be
seen that atomization of ethanol at the same operating
conditions as water results in substantially smaller drop-
lets (50% smaller). Because the surface tension of etha-
nol is only one third of water, this liquid is easier broken
into droplets.
Another important factor is the liquid flow rate,
especially at the lower values of ALR. At a given ratio,
larger droplets could be produced by operating with lower
liquid flow rates. An important point here is that two
identical ALRs achieved with different liquid and atomi-
zation gas flow rates do not result in the same droplet size,
which is easily seen in Figure 5. The effect of atomization
gas flow rate was investigated for liquid (water and etha-
nol) flow rates of 0.8, 1.5, and 2.5 kg/hr, and the results
are presented in Figure 6. At the lower atomization gas
flow rates, the SMD was found to increase with faster liq-
uid flow. At the higher operating gas flows, no difference
in SMD was found among the investigated liquid flows.
The degree of atomization is determined by the velocity
difference between the atomization gas and the liquid.
Therefore, the liquid flow rate has lower importance when
operating at higher atomization gas flow rates. These find-
ings were thus in agreement with the results obtained by
Liu et al.[23]
 Droplet Size Measurements for Spray Dryer Scale-Up 99
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13

Figure 5. SMD plotted against ALR for water (left) and ethanol (right) for various liquid flow rates (0.8, 1.5, and 2.5 kg/hr) and
liquid orifice diameters (0.8 and 1.0 mm). Error bars were omitted for clarity.
For personal use only.

Figure 6. SMD plotted against atomization gas flow rate for water (left) and ethanol (right) at various liquid flows (0.8, 1.5, and
2.5 kg/hr). Liquid orifice 1.0 mm. Error bars were omitted for clarity.

Effect of Adding Polymer and Drug to the Solvent Table 2

Sauter mean diameters (μm) of sprays generated with
In Table 2 a comparison of the droplet size distribu- pilot scale nozzle
tions of 99.9% ethanol and the drug/polymer solutions is
Liquid flow rate [kg/hr]
given. The addition of solids up to a total concentration
of 20 w/w% resulted in small increases of the measured 1.5 2.5
droplet sizes. The change in the liquid properties
Ethanol 99.9% 4.7 ± 0.1 6.1 ± 0.3
could explain the increases, as the addition of polymer
Solution 1:2 10w/w% 5.4 ± 0.2 7.2 ± 0.4
and drug resulted in an increased viscosity of the liquid
Solution 1:9 10w/w% 5.8 ± 0.3 8.0 ± 0.4
(see Table 1). Solution 1:2 20w/w% 6.3 ± 0.3 8.4 ± 0.5
Solution 1:9 20w/w% 6.7 ± 0.3 8.8 ± 0.5

Up-Scaling
The experiments done with the pilot scale nozzle
showed that the downstream distance from nozzle to the
laser beam influenced the measured droplet size. It was
therefore chosen to inspect this parameter for the produc-
tion scale nozzle as well. The results are presented in
Figure 7. For the given test conditions, positions closer
Atomization gas flow rate of 9.0 kg/hr. Liquid orifice diam-
eter 1.0.
than 65 mm caused wetting of the lens and were therefore
not included in the figure. Considering the uncertainty of
the Spraytec, the droplet size was constant at all inspected
downstream distances. Of course these values are only
 100 P. Thybo et al.
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13

Figure 7. Influence of the downstream distance from nozzle to Figure 8. SMD for water droplets plotted against ALR. Flow
laser beam on the SMD for water droplets. Atomizing gas flow rate of water 45 kg/hr, liquid orifice diameter 4.0 mm, and
47 kg/hr, flow rate of water 45 kg/hr, liquid orifice diameter atomization gas orifice diameters (9.0, 10.0, and 11.0 mm).
4.0 mm, and atomization gas cap diameters 9.0 mm.

the production scale nozzle relative to the used liquid

valid under the given set of test conditions. For further
measurements 120 mm was chosen as standard down-
stream distance between nozzle and laser beam.
For personal use only.
flow rate. This results in much higher atomization gas
velocities and thereby smaller droplets for identical
ALR. That it is possible to produce similar droplet size
distributions with the nozzles is illustrated in Figure 9.

Influence of Atomization Gas Orifice Diameter

Correlation of Droplet Size to Operational
For the production scale nozzle, only one liquid
Parameters
orifice diameter was used as the influence of this had
already been studied for the pilot scale nozzle and by
The intention was to determine a correlation, which was
Liu et al.[23] Instead, the atomization gas orifice diame-
able to predict the mean droplet diameter for an externally
ter was chosen as parameter. The results for atomiza-
mixing two-fluid nozzle. On the basis of the wave mecha-
tion gas orifice diameters of 9.0, 10.0, and 11.0 mm are
nism, Mansour and Chigier[26] proposed several semiempiri-
presented in Figure 8. The figure shows that droplet
cal models for air-blast atomization, where the droplet size
size distributions with SMDs less than 20 μm predominate
was a function of the Weber number, the Ohnesorge number,
when ALRs range from 1 to 4. The figure also shows that
and the dimensionless group (1 + 1/ALR). Mansour and
the SMD decreases with increasing ALR and that an
Chigier found that for both Newtonian and non-Newtonian
almost constant SMD was reached for high ALR
liquids, the following correlation was especially good (con-
regardless of the orifice diameter of the atomization
sidering it had only two parameters, where others had three)
gas cap used. The usage of smaller atomization gas ori-
relating the SMD to operational parameters:
fice diameters facilitates formation of droplets with
smaller SMD for a given ratio because they increase
SMD ⎛ M ⎞
x
the atomization gas velocity. Operation at the lower ⎛ Z ⎞ 1
ALR values (< 1) was not possible with the larger = ⎜ 1 + l ⎟ ⋅ X2 ⋅ ⎜ ⎟
d0 ⎝ Mg ⎠ ⎝ We ⎠
atomization gas orifice diameters due to poor atomiza-
tion (dripping from the nozzle or large spans in droplet
size distribution). Comparing Figure 5 with Figure 8 where d0 is the liquid orifice diameter, Ml is the liquid
shows that the production scale nozzle produces the flow rate, Mg is the atomization gas flow rate, Z is the
same SMDs as the pilot scale nozzle but that it does so dimensionless Ohnesorge number, We is the dimension-
at much lower ALRs. The main reason for this is that less Weber number, and X1 and X2 are parameters.
while the two nozzles are of similar design, the area of The exponent and the coefficient to the above equation
the atomization gas exit is disproportionately smaller for were determined by the best least-square fit to the
 Droplet Size Measurements for Spray Dryer Scale-Up 101
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13
For personal use only.

Figure 9. Example of droplet size distribution after atomization of water. Pilot scale nozzle (A): liquid flow 9 kg/hr, gas flow
2.5 kg/hr, liquid orifice 1.0 mm, D10 6.4 μm, D50 17.0 μm, D90 32.5 μm, and SMD 10.6 μm Production scale nozzles (B): liquid flow
45 kg/hr, gas flow 66 kg/hr, atomization gas cap diameter 10.0 mm, D10 6.1 μm, D50 16.6 μm, D90 35.7 μm, and SMD 10.1 μm.

experimental data using linear curve fitting. The same was
also performed by using the volumetric mean diameter
(VMD) instead of SMD. Because of the convergent design
of the nozzles used in this study, the maximum velocity that
the atomization gas can attain is the velocity of sound at the
particular temperature.[16] So when this velocity is reached,
the gas density will start to increase. Therefore, when
calculating the Weber number, the density of the gas at
atmospheric pressure and 288 K was used unless the atomi-
zation gas velocity (calculated at the nozzle exit) exceeds
the velocity of sound (346 m/s for Nitrogen at 288 K).
When the atomization gas velocity reaches the velocity of
sound, the density was calculated by using the velocity of
• •
sound and mass conservation: m = ρ g ⋅ ug ⋅ AAt , g , where m
is the atomization gas flow rate, rg is the gas density, ug is
velocity of sound, and AAt,g is the nozzle exit area of the
atomization gas. The results of the least-square analysis
based on the droplet size measurements are given in Table 3
(SMD) and Table 4 (VMD), which shows the estimated
parameter values for the various liquids and nozzles together
with the correlation coefficients.
In general, there is a good fit in all four cases as
evidenced by the high correlation coefficients (Tables 3
and 4). It is interesting that the calculated exponents of
 102 P. Thybo et al.

Table 3
Estimated parameters and correlation coefficients (SMD)

Liquid Correlation
Nozzle atomized X1 X2 coefficient

Pilot scale (74) Water 0.606 1.19·107 0.973

Pilot scale (64) Ethanol 0.649 1.31·107 0.964
Production scale Water 0.595 1.22·107 0.974
(28)
Pilot & production Water 0.566 7.42·106 0.987
scale (102)
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13

Numbers in parentheses are the number of data points.

Table 4
Estimated parameters and correlation coefficients (VMD)

Liquid Correlation Figure 10. Calculated SMD versus experimental measured

Nozzle atomized X1 X2 coefficient SMD for water droplets. The solid line illustrates the optimal
agreement between model and experimental data.
Pilot scale (74) Water 0.533 8.83·106 0.988
Pilot scale (64) Ethanol 0.639 2.07·107 0.975
Production scale Water 0.556 1.13·107 0.961 gas orifice, the agreement between the proposed model and
(28) the experimental data was very good, especially for the
Pilot & production Water 0.557 1.17·107 0.994 smaller droplets (< 20 μm). The large deviations observed
scale (102) for droplets greater than 20 μm diameter is most likely
caused by poor atomization, which was indeed observed
For personal use only.

Numbers in parentheses are the number of data points.
0.566 (SMD) and 0.557 (VMD) are close to the theoreti-
cally predicted 0.667 for the wave mechanism, supporting
the notion that for these low-viscosity Newtonian liquids
the atomization occurs via the wave mechanism. It was
found that the correlation was not able to account properly
for the physical properties of the liquid. An average
deviation of 60% was observed when applying the parame-
ters estimated from water droplet diameters to prediction of
ethanol droplet diameters for the pilot scale nozzle. Lefeb-
vre also studied the factors influencing the mean droplet
size[20,27] and stressed that no single equation can correlate
properly all the experimental data from a nozzle.
Because the pilot scale and the production scale
nozzle were similar in design, it was interesting to investi-
gate if one correlation could predict the droplet sizes for
both and as the tables show; it was indeed possible to
combine the two nozzles into one correlation with a high
correlation coefficient. Figure 10 displays the agreement
between the obtained experimental SMDs and the SMDs
calculated by using the estimated parameters for the pilot
and production scale nozzle combined for water droplets.
On average, the proposed model was found to match the
measured droplet diameters within 9%. An even better
match was obtained between experimental VMDs and
calculated VMDs. Taking into account the variation in
both nozzle size, diameter of liquid orifice, and atomization
for both the pilot and production scale nozzle when oper-
ating at low atomization gas flow rates and high liquid
flow rates. Another contributing factor is the operating
conditions used in this study with measured mean droplet
diameters primarily being less than 20 μm (~90% of the
data). This means that the small droplets determine the
parameters; as a consequence, it is obvious that the best
agreement was observed in this range.
CONCLUSIONS
To scale up from pilot to production scale spray dryers,
the performance of 2 two-fluid nozzles of similar design,
but of different capacity, were evaluated by measurement
of the droplet size distributions by laser diffraction.
Various liquid properties and setup considerations for
measurement of the droplet size distribution were evalu-
ated. The study showed that two-fluid nozzle atomization
was dependent on the following variables: atomization gas
flow rate, liquid flow rate, liquid viscosity, liquid surface
tension, diameter of liquid orifice, and outlet area of the
atomization gas. Furthermore, for the droplet size
measurements the downstream distance from nozzle
orifice to laser beam was of importance. The mean droplet
size increased with distance from the nozzle exit until
reaching a constant value some distance from the nozzle
exit, with the exact distance depending on the nozzle and
 Droplet Size Measurements for Spray Dryer Scale-Up
nozzle operating parameters. The study showed the pos-
sibility of scaling up from a liquid flow of 2.5 kg/hr for
the pilot scale nozzle to 45 kg/hr on the production scale
nozzle while achieving the same droplet size distribution.
It is interesting that the atomization gas to liquid flow
rate ratio leading to the same droplet size distribution
was substantially lower for the production scale nozzle
than the pilot scale nozzle. However, this could be
explained by higher atomization gas velocities for the
production scale nozzle when compared to the pilot scale
nozzle resulting in smaller droplets for identical atomiza-
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13
tion gas to liquid flow ratio. Therefore, scaling up just
using the ALR is not possible, which was also observed
when operating the pilot scale nozzle with different liq-
uid flow rates.
On the basis of the droplet size measurements a
semiempirical model was fitted to the water droplet size
data by way of a correlating equation developed by
Mansour and Chigier. The model based on data from both
the pilot and the production scale nozzle showed excellent
agreement between calculated and experimental values.
However, it was found that the correlation was not able to
account properly for the physical properties of the liquid,
because an average deviation of 60% was observed when
applying the parameters estimated from water droplet
For personal use only.
diameters to prediction of ethanol droplet diameters for the
pilot scale nozzle.
Knowledge of the droplet size distribution for the
pilot scale nozzle in combination with the model could be
used to predict nozzle operation conditions resulting in
same droplet size for a larger nozzle of similar design.
Therefore, the results obtained from this study could be a
useful guideline for selection of nozzle operating condi-
tions for up-scaling of a spray-drying process from a droplet/
particle size point of view.
ACKNOWLEDGMENTS
The authors thank The Drug Research Academy,
GEA Pharmaceutical A/S, and Niro A/S, for financial
support, and Martha Zielinski, for carrying out a substan-
tial number of droplet size measurements on the pilot
scale nozzle.
REFERENCES
1. Dunbar C. Dry powder formulations for inhalation. Drug
Deliv. Syst. Sci. 2002;2(3):78–80.
2. Masters, K. Spray Drying Handbook. Essex: Longman
Scientific & Technical; 1991.
103
3. Wendel S, Celik M. An overview of spray-drying applica-
tions. Pharm. Technol. 1997;21(10):24–144.
4. Giunchedi P, Conte U. Spray-drying as a preparation method
of microparticulate drug-delivery systems: an overview.
S. T. P. Pharm. Sci. 1995;5(4):276–290.
5. Gibson SG. How to optimize your spray dryer’s perfor-
mance. Powder Bulk Eng. 2001;15:31–41.
6. Schick RJ, Brown K. Spray dryer scale-up: from laboratory to
production. http://service.spray.com/Literature_PDFs/asp/view_
lit_file.asp?code=WP010 (accessed 23-05-2007).
7. Nielsen AF, Bertelsen P, Kristensen HG, Kristensen J,
Hovgaard L. Investigation and comparison of performance
of effervescent and standard pneumatic atomizer intended
for soluble aqueous coating. Pharm. Dev. Technol. 2006;11
(2):243–253.
8. Triballier K, Dumouchel C, Cousin J. A technical study on
the Spraytec performances: Influence of multiple light scat-
tering and multi-modal drop-size distribution measurements.
Exp. Fluids 2003;35(4):347–356.
9. Simmons MJH, Hanratty TJ. Droplet size measurements in
horizontal annular gas-liquid flow. Int. J. Multiphase Flow
2001;27(5):861–883.
10. Priol L, Baudel P, Louste C, Romat, H. Theoretical and
experimental study (linear stability and Malvern granulome-
try) on electrified jets of diesel oil in atomization regime.
J. Electrostat. 2006;64(7–9):591–596.
11. Haynes A, Shaik MS, Krarup H, Singh M. Evaluation of the
Malvern Spraytec® with inhalation cell for the measurement
of particle size distribution from metered dose inhalers.
J. Pharm. Sci. 2004;93(2):349–363.
12. Adi H, Larson I, Chiou H, Young P, Traini D, Stewart P.
Agglomerate strength and dispersion of salmeterol xinafoate
from powder mixtures for inhalation. Pharm. Res. 2006;23
(11):2556–2565.
13. Dayal P, Shaik MS, Singh M. Evaluation of different param-
eters that affect droplet-size distribution from nasal sprays
using the Malvern Spraytec®. J. Pharm. Sci. 2004;93(7):
1725–1742.
14. Dayal P, Pillay V, Babu RJ, Singh M. Box-Behnken
experimental design in the development of a nasal drug
delivery system of model drug hydroxyurea: Characteriza-
tion of viscosity, in vitro drug release, droplet size, and
dynamic surface tension. Aaps Pharmscitech. 2005;6(4):
573–585.
15. Hornby JA, Robinson J, Opp W, Sterling M. Laser-diffraction
characterization of flat-fan nozzles used to develop aerosol
clouds of aerially applied mosquito adulticides. J. Am. Mosq.
Control Assoc. 2006;22(4):702–706.
16. Lefebvre AH. Atomization and Sprays. New York: Hemi-
sphere Publishing Corporation; 1989.
17. Dombrowski N, Johns WR. The aerodynamic instability and
disintegration of viscous liquid sheets. Chem. Eng. Sci.
1963;18:203–214.
18. Adelberg M. Mean drop size resulting from the injection of a
liquid jet into a high-speed gas stream. AIAA J. 1968;6(6):
1143–1147.
19. Adelberg M. Breakup rate and penetration of a liquid jet in a
gas stream. AIAA J. 1967;5(8):1408–1415.
 104 P. Thybo et al.
20. Lefebvre AH. Some recent developments in twin-
fluid atomization. Part. Part. Syst. Char. 1996;13(3):
205–216.
21. Weast RC. CRC Handbook of Chemistry and Physics. West
Palm Beach, FL: CRC Press, Inc.;1979.
22. Elversson J, Millqvist-Fureby A, Alderborn G, Elofsson U.
Droplet and particle size relationship and shell thickness of
inhalable lactose particles during spray drying. J. Pharm.
Sci. 2003;92(4):900–910.
23. Liu HF, Li, WF, Gong X, Cao XK, Xu JL, Chen XL, Wang
YF, Yu GS, Wang FC, Yu ZH. Effect of liquid jet diameter
Pharmaceutical Development and Technology Downloaded from informahealthcare.com by University of Toronto on 03/01/13
For personal use only.
on performance of coaxial two-fluid airblast atomizers.
Chem. Eng. Process. 2006;45(4):240–245.
24. Rizk NK, Lefebvre AH. Spray characteristics of plain-jet air-
blast atomizers. Trans ASME J. Eng. Gas Turbines Power
1984;106;639–644.
25. Lorenzetto GE, Lefebrre AH. Measurements of drop size on
a plain jet airblast atomizer. AIAAJ. 1997;15:1006–1010.
26. Mansour A, Chigier N. Air-blast atomization of non-Newtonian
liquids. J. Non-Newton. Fluid Mech. 1995;58(2–3):161–194.
27. Lefebvre AH. Twin-fluid atomization: Factors influencing
mean drop size. Atom. Sprays 1992;2:101–119.