ORIGINAL PAPER

Mimicking Spray Drying by Drying of Single Droplets
Deposited on a Flat Surface
Jimmy Perdana & Martijn B. Fox &
Maarten A. I. Schutyser & Remko M. Boom
Received: 18 August 2011 / Accepted: 20 December 2011
#The Author(s) 2012. This article is published with open access at Springerlink.com
Abstract The inactivation of bioactive ingredients during
spray drying is often matrix specific. Therefore, the design
of new processes or the optimisation of existing spray dry-
ing processes is usually highly product specific and requires
numerous experiments. Rapid experimentation methods that
facilitate fast data generation are therefore desired. A novel
method for drying single droplets to mimic spray drying is
proposed. The approach involves droplet deposition on a
hydrophobic flat surface followed by controlled drying. A
heat and mass transfer model is applied to predict the drying
history of the single droplets. The approach is successfully
evaluated through studying the inactivation of β-galactosidase
during drying. The heat and mass transfer model supple-
mented with inactivation kinetics provided reasonable predic-
tion of the residual enzyme activity after drying. In addition,
the inactivation kinetics could be directly extracted from sin-
gle droplet experiments rather than using the kinetics from
separate heating experiments. Finally, it was demonstrated
that the inactivation kinetics found with the single-drop
experiments could satisfactorily predict the residual activity
of β-galactosidase dried with a laboratory-scale spray dryer.
Keywords Enzyme
.
Temperature
.
Moisture content
.
Inactivation kinetics
.
Spray drying
.
Single droplet
Introduction
Spray drying belongs to the most common drying techniques
for liquid food products. The fine atomisation of the product
and the subsequent fast evaporation of water make it especial-
ly suitable to formulate heat-sensitive products (Xin Huang et
al. 2006; Fang and Bhandari 2010). Despite the relative mild
conditions during drying, (partial) inactivation of bioactive
compounds such as enzymes, antioxidants and vitamins can-
not be avoided. Optimisation of the spray drying conditions
and addition of stabilizers is often required to retain maximum
activity (Ré 1998; Peighambardoust et al. 2011; Sansone et al.
2011). In practice, numerous expensive pilot-scale trials are
executed to explore different product formulations and to find
optimum drying conditions.
Modelling tools are frequently used to accelerate process
development and optimisation in spray drying. The availabil-
ity of reliable inactivation kinetics—usually highly product
specific—is a prerequisite for this. Kinetic models require
experimental data from drying experiments for parameter
calibration. Pilot-scale experiments are not ideal for this as
they are expensive, time-consuming, and involve the evalua-
tion of complex process conditions. A more efficient alterna-
tive is the application of small-scale representative drying
experiments.
Once reliable inactivation kinetics are obtained, a process
model of the dryer can be used to predict the impact of the
drying on the remaining ingredient activity. Subsequently, the
model can be used to approximate optimal drying conditions
and product formulations in a more systematic way (Wijlhuizen
et al. 1979; Luyben et al. 1982; Meerdink and van't Riet 1991;
Millqvist-Fureby et al. 1999; Coumans 2000).
Small-scale spray drying has been carried out in
laboratory-scale spray dryers (Gianfrancesco et al. 2010;
Li et al. 2010; Wu et al. 2011). However, major differences
J. Perdana (*)
:
M. A. I. Schutyser
:
R. M. Boom
Food Process Engineering Group, Wageningen UR,
P.O. Box 8129, 6700 EV Wageningen, The Netherlands
e-mail: jimmy.perdana@wur.nl
M. B. Fox
NIZO Food Research,
P.O. Box 20, 6710 BA Ede, The Netherlands
e-mail: martijn.fox@nizo.nl
Food Bioprocess Technol
DOI 10.1007/s11947-011-0767-4
between laboratory and industrial spray dryers are the
smaller droplet size, caused by the different method of
atomisation, and the shorter residence time in the laboratory
spray dryer (Goula and Adamopoulos 2004; Filková et al.
2006). Since the laboratory spray dryer can (due to its
design and dimensions) only cover a limited range of drop-
let sizes and drying times and it produces droplets with a
size distribution (not monodisperse), it is virtually impossi-
ble to extract sufficient representative experimental data to
calibrate a kinetic model.
Another approach is the drying of single droplets under
well-controlled conditions. In this approach a small droplet
is generated, immobilised, and subsequently dried by con-
tacting it with well-defined drying air. Examples are drying
of a droplet that is non-intrusively levitated using for exam-
ple acoustic, aerodynamic or electromagnetic levitation
(Adhikari et al. 2000; Sloth et al. 2006; Schiffter and Lee
2007), intrusively levitated that is in pendant or filament
(Chuchottaworn et al. 1984; Ali Al Zaitone and Tropea
2011), or that is deposited on a flat surface (Perdana et al.
2011a).
In this study, the latter method is followed (Perdana et al.
2011a). The major advantages of this method are the possibil-
ity to vary droplet size and residence time, and to dry multiple
droplets at once to obtain higher volume of sample. The
minimum droplet diameter can be adjusted to 150 μm which
is only slightly bigger than the typical droplet size in industrial
spray dryers; a comparable droplet size is critical to accurately
mimic the kinetics in spray drying (Adhikari et al. 2000;
Vehring et al. 2007). The challenge in using this approach is
the presence of the flat surface. This surface for example
affects the spherical shape of the droplet. Therefore, a hydro-
phobic surface is used to minimize the contact between the
droplet and the surface and to retain the spherical shape. Other
differences introduced by the presence of the surface are the
air flow velocity and the air temperature near the droplet,
which deviate from the bulk conditions (Schlichting and
Gersten 2000). By monitoring the air temperature near the
droplet and by modelling the air flow and heat transfer across
the drying surface, the influence of the surface on the drying
conditions can be quantified.
This study focuses on the drying of the enzyme β-
galactosidase suspended in a maltodextrin matrix. The en-
zyme β-galactosidase is selected as the model enzyme in
this study, because it is an industrial relevant enzyme ap-
plied for production of lactose-hydrolysed milk and whey.
Additionally, its activity can be easily determined and its
inactivation kinetics has been studied before, though under
steady-state conditions, providing strong basis for this study
(Yamamoto and Sano 1992; Yoshioka et al. 1994; Perdana
et al. 2011b). The inactivation kinetics of β-galactosidase
from the previous studies were combined with the heat and
mass transfer model to predict the residual activity of β-
galactosidase during the single droplet drying experiments.
Finally, the predictive value of the model was also checked by
drying of β-galactosidase in a laboratory spray dryer system.
Modelling and Statistical Evaluation
This section consists of four parts: (1) model description for a
single droplet drying, (2) modelling temperature and air flow
distribution across the flat plate, (3) inactivation kinetics of β-
galactosidase, and (4) parameter optimisation and statistical
evaluation.
Model Description for Single Droplet Drying
In analogy to several previous studies, a heat and mass transfer
model is used to describe the drying of a sessile single droplet
(Sloth et al. 2006; Straatsma et al. 2007; Mezhericher et al.
2008). The sessile droplet is approximated as a spherical
droplet, and it is thus assumed that the diffusion occurs only
in the radial direction. For the small droplet sizes considered,
the temperature gradient inside the droplet could be neglected.
This assumption is valid if the Biot number (Bi) <0.1. Bi is
defined as the ratio of external heat transfer (between air and
droplet) and the internal heat transfer (in the droplet, Incropera
and De Witt 1985).
Bi ¼
hd
d
l
d
ð1Þ
In this study, Bi is around 0.02.
The following differential equation is used to describe the
droplet mass change in time
À
dm
d
dt
¼ k
c
4pR
2
d
M
w
R
a
w
P
sat
w
T
d
À
P
1
P
1
_ _
ð2Þ
where m
d
is the mass of the droplet, t is the time, k
c
is the mass
transfer coefficient, R
d
is the droplet radius, M
w
is the molec-
ular weight of water, R is the ideal gas constant, a
w
is the water
activity, P
w
sat
is the saturated vapour of water at T
d
, T
d
is the
temperature of the droplet, P

is the vapour partial pressure in
the bulk air and T

is the bulk air temperature.
The developing moisture gradient inside the droplet is
described by Fickian diffusion
@C
w
@t
¼ D
w;d
@
@r
@C
w
@r
_ _
ð3Þ
Where C
w
is the water concentration, D
w,d
is the
effective diffusion coefficient of water in the droplet and r is
Food Bioprocess Technol
the radial coordinate. The boundary conditions applied
here are
t ¼ 0; 0 r R
d
;
dC
w
dr
¼ 0
t ¼ t; r ¼ 0;
dC
w
dr
¼ 0
t ¼ t; r ¼ R
d
; ÀD
w;d
dC
w
dr
¼ k
c
4pR
d
2 M
w
R
a
w
P
sat
w
T
d
À
P
1
T
1
_ _
The temperature change of the droplet is described by the
following enthalpy balance:
m
w
c
p;w
þm
s
c
p;s
_ _
dT
d
dt
¼ h
d
4pR
d
2
T
1
ÀT
d
ð Þ À
dm
d
dt
H
lv
ð4Þ
where m
s
is the mass of the solute, c
p,w
and c
p,s
the heat
capacity of the water and the solute respectively, h
d
the heat
transfer coefficient to the droplet through convection and
H
lv
is the evaporation enthalpy of water.
The coupled heat and mass transfer models were solved
numerically using 100 radial layers in the droplet. In each
layer the same solute mass (m
s
) was calculated according to
m
s
¼ C
s;0
V
d;0
n
ð5Þ
where C
s,0
is the initial solute concentration and V
d,0
is the
initial volume of the droplet.
Since the initial solute concentration is assumed homoge-
nous, the initial solute mass distribution is approximated with
the initial volume distribution. The radial position of the i
th
layer from the centre of the droplet can be calculated as
follows
r
i
¼ i Á
R
d
3
n
_ _
1=3
ð6Þ
where n is the total number of layers for numerical integration.
The layer thickness Δr is
Δr
i
¼
r
i
; i ¼ 1
r
i
Àr
iÀ1
; i ! 2
_
ð7Þ
The volume of each layer V is then
V ¼
m
w
ρ
w
þ
m
s
ρ
s
ð8Þ
where ρ
w
and ρ
s
are the densities of water and solute, respec-
tively. The average water concentration is then defined as
C
w
¼
m
w
V
¼
m
w
m
w
ρ
w
þ
m
s
ρ
s
ð9Þ
Reorganizing Eq. 9. results in
m
w
¼
m
s
ρ
s
C
w
1 À
C
w
ρ
w
_ _ ð10Þ
The droplet shrinks upon drying due to the evaporation of
water. This shrinkage influences the heat and mass transfer
within the droplet. The model was therefore corrected to
describe the decrease of layer thickness due to shrinkage
(Farid 2003). This correction was based on the water concen-
tration in every layer at t>0 that is calculated from Eq. 3
(Crank 1990). From the water concentration, the new volume
of the layer was recalculated as
V ¼
m
s
ρ
s
1 þ
1
ρ
w
ÀC
w
ð Þ
_ _
ð11Þ
The radial coordinate of every partition was corrected
using this new volume. The radius of the ith partition is
r
i
¼
3

i
0
V
i
_ _
4p
_ _
1=3
ð12Þ
The thickness of every partition could be recalculated
using Eq. 7. To solve the model, several closure equations
like heat and mass transfer coefficient, water activity, mois-
ture diffusion coefficient, and vapour pressure are required,
which are tabulated in Table 1.
Modelling Temperature and Air Flow Distribution
Across the Flat Plate
In this study, the droplet was deposited on a flat plate. The
drawback of this configuration is that in addition to heat
transfer from the air to the droplet, also heat is transferred
from the air to the flat plate through convection. This heat is
further transferred from the plate to the droplet through
conduction. It is of major importance to quantify the heat
transfer between the flat plate and the droplet and compare
this to the convective heat transfer via the drying air. The
conductive heat transfer between the flat plate and the
droplet is preferably as low as possible, when one wants to
mimic the drying of a droplet suspended in air. By using a
hydrophobic surface, the contact area for conductive heat
transfer has already been reduced. Furthermore, the temper-
ature in the air near the flat plate deviates from the bulk air
temperature affecting the drying process. Therefore a model
description of the temperature gradients within the plate and
in the air near the plate was developed.
To minimize turbulence (eddy formation) near the edge
of the depositing plate, a thin flat plate is used rather than a
blunt-edged plate. However, the presence of the flat plate
still influences the flow pattern of the drying air above the
plate, i.e. near the droplet (Fig. 1). A boundary layer model
Food Bioprocess Technol
according to Mosaad (1999) is used to describe the air flow
and temperature distribution across the flat plate: a hydrody-
namic boundary layer is defined for the air flow distribution
and a thermal boundary layer for the temperature distribution
u
u
1
¼
3
2
z
d
_ _
À
1
2
z
d
_ _
3
; 0 z d ð13Þ
TÀT
1
T
p;x
ÀT
1
¼ 1 À
3
2
z
d
t
_ _
þ
1
2
z
d
t
_ _
3
; 0 z d
t
ð14Þ
where z is the height coordinate from the surface of the flat
plate, u is the air velocity at z, u

is the bulk velocity of air, υ
is the kinematics viscosity of air, T
p,x
is the temperature of
the surface of the flat plate at distance x (at the front edge of
the flat plate, x00), T

is the bulk air temperature, δ is the
hydrodynamic boundary layer at distance x and δ
t
is the
thermal boundary layer. The hydrodynamic boundary layer
for u00.99u

is (Schlichting and Gersten 2000)
d ¼ 5
ffiffiffiffiffiffiffi
ux
u
1
_
ð15Þ
and δ
t
is thermal boundary layer at distance x
d
t
¼
Pr
À1=3
; Pr 1
Pr; Pr > 1
_
ð16Þ
The heat transfer coefficient between the air and the flat
plate h
p
is approximated by (Thirumaleshwar 2009)
h
p
¼ 0:332
l
a
x
1=2
u
1
v
_ _
Pr
1=3
; Pr ¼
c
p;a
μ
a
l
a
ð17Þ
where 1
a
is the thermal conductivity of air, Pr is the Prandtl
number of air, c
p,a
is the heat capacity of air and μ
a
is the
dynamic viscosity of air. The value of 1
a
, c
p,a
and μ
a
are
temperature dependent; in Eq. 17, the value are evaluated at
T0(T
p
+T

)/2.
The heat transfer coefficient decreases as a function of
the x coordinate (Eq. 17) and is calculated locally. It was
Table 1 The closure equations used in drying model
Equation name Equation Reference
Saturated vapour pressure log
p
sat
w
100
_ _
¼ À7:90298
Tst
T
À1
_ _
þ5:02808 log
Tst
T
_ _
À1:3816 Á 10
À7
10
11:344 1À
1
st
T
ð Þ
À1
_ _
þ8:1328 Á 10
À3
10
À3:49149
T
st
T
À1 ð Þ
À1
_ _
þlog e
st
ð Þ
Goff and Gratch (1946)
For vapour at total pressure 1 atm, T
st
is 373.15 K
and e
st
is 1013.25 (in hPa)
Water activity (Guggenheim-
Anderson-de Boer model)
Xw
Xwm
¼
CgÁKgÁaw
1ÀKÁaw ð Þ 1ÀKgÁawþCgÁKgÁaw ð Þ
Quirijns et al. (2005)
Heat and mass transfer coefficients For a droplet suspended in air, used in the drying
model for laboratory-scale spray dryer:
Perdana et al. (2011b),
Ranz and Marshall (1952)
hdd
la
¼ 2 þ0:6Re
0:51
Pr
0:33
kcdd
Dw;a
¼ 2 þ0:6Re
0:51
Sc
0:33
For a sessile droplet, used in the drying model
for single droplet drying:
hdd
la
¼ 0:24 þ0:63Re
0:51
Pr
0:33
kcdd
Dw;a
¼ 0:24 þ0:63Re
0:51
Sc
0:33
where Re ¼
ρ
a
ddu1
μ
a
; Pr ¼
cp;aμ
a
la
; Sc ¼
μ
a
ρ
a
Dw;a
Diffusion coefficient of water
in maltodextrin solution
D
w;d;T¼30
o
C
¼ exp À
35:8þ215Xw
1þ10:2Xw
_ _
Räderer et al. (2002)
D
w;d
¼ D
w;d;T¼30
o
C
exp À
Ea
R
1
T
À
1
303:15
_ _ _ _
where E
a
¼
dþ190Xw
1þ10Xw
Diffusion coefficient of water in air D
w;a
¼ À2:775 Á 10
À6
þ4:479 Á 10
À8
T þ1:656 Á 10
À10
T
2
Bolz and Tuve (1973)
Fig. 1 Sketch of a single droplet drying on a flat plate with the air flow
pattern and governing heat transfer processes indicated. The sessile
droplet is dried on a thin plate consisting of a hydrophobic membrane
(0.15 mm) on top of a stainless steel platform (1.00 mm)
Food Bioprocess Technol
estimated numerically; i.e. assuming a Δx at distance L from
the front side of the plate, the average heat transfer coeffi-
cient is obtained by integrating h
p
along Δx
h
p;avg
¼ 0:664
l
a
v
1=2
Pr
1=3
u
1
L
_ _
1=2
À
u
1
L þΔx
_ _
1=2
_ _
ð18Þ
where h
p,avg
is the average local heat transfer coefficient.
The non-uniform heat transfer rate across the plate con-
tributes to the development of a temperature gradient within
the plate. The temperature distribution within the plate is
described with a one-dimensional partial differential equation
(Kakac and Yener 1993). It is assumed that no temperature
gradient in the y direction (perpendicular to the air flow
direction) and in the z direction (within the plate; because of
the small thickness, i.e. 150 μm)
@T
p
@t
¼
l
p
ρ
p
c
p;p
@
2
T
p
@x
2
þ
h
p;avg
T
1
ÀT
p
_ _
Zl
p
_ _
ð19Þ
where ρ
p
is the density of the plate, Z is the thickness of the flat
plate and l
p
is the thermal conductivity of the flat plate. The
boundary layer applied here are that near the droplet the flat
plate temperature is equal to the droplet temperature and at the
edge of the flat plate (far from the droplet) the temperature
gradient of the flat plate is 0.
Equations 14 and 19 were solved numerically to estimate
the temperature distribution history within and above the flat
plate. The predictions were validated by simple temperature
measurements at different heights above the plate.
To estimate the conductive heat transfer between the
surface and the droplet and compare it to the convective
heat transfer between the air and the droplet, the conductive
heat flux Q through the contact surface of the droplet was
estimated according to
Q ¼ l
p
2pR
d
z
ΔT
cond
Δx
cond
ð20Þ
where ΔT
cond
/Δx
cond
is the temperature gradient in the
partition layer of the flat plate that is closest the droplet. It
was found that the conductive heat flux was smaller than 5%
of the total heat transferred to the droplet regardless of its
position. Therefore, it can be safely assumed that most heat
is transferred via convective heat transfer.
Inactivation Kinetics of β-galactosidase
The inactivation rate of the β-galactosidase is affected by
the temperature and moisture content of the droplet, which
both change with time. A kinetic inactivation model for β-
galactosidase is developed in earlier work and calibrated
using constant heating experiments (Perdana et al. 2011b).
In the latter experiments, the suspended enzyme is exposed
to various temperature–moisture value combinations for a
specific time and the remaining enzyme activity is mea-
sured. The inactivation kinetics are described by a two-
step inactivation process and the observed inactivation co-
efficient (k
obs
) is described as
k
obs
¼
K
1
1 þK
1
_ _
k
2
ð21Þ
where
K
1
¼ exp
ΔΔS
z
1;w
R
À
ΔΔH
z
1;w
RT
ref
_ _
exp À
ΔΔH
z
1;w
R
1
T
À
1
T
ref
_ _
_ _
exp À
f 1 Àx
w
ð Þ
RT
_ _
ð22Þ
k
2
¼
k
B
T
h
exp
ΔS
z
2;w
R
À
ΔH
z
2;w
RT
ref
_ _
exp
ΔH
z
2;w
RT
1
T
À
1
T
ref
_ _
_ _
exp À
ΔH
z
2;w
ÀΔH
z
2;w
R
1
T
À
1
T
int
_ _
exp Àg
x
w
1 Àc
w
_ _
_ _
ð23Þ
where K
1
is the reversible unfolding equilibrium constant, k
2
is the complete denaturation rate constant, ΔΔS
1,w

is the
activation entropy difference between the unfolding and
refolding reactions in pure water, ΔΔH
1,w

is the activation
enthalpy difference between the unfolding and refolding
reactions in pure water, k
B
is Boltzmann’s constant, h is
the Planck’s constant, ΔS
2,w

is the activation entropy of
complete denaturation in pure water, ΔH
2,w

is the activation
enthalpy in pure water, ΔH
2,m

is the activation enthalpy in
pure solid form (i.e. no moisture), f describes the effect of
moisture content on conformational stability of the enzyme,
g describes the effect of moisture content on the irreversible
inactivation (second step) kinetic constant of the enzyme
and T
in
is the intercept temperature at which k
2,w
/k
2,s
01.
Alternatively, the enzyme inactivation can also be regarded
as a one-step inactivation process in which the native enzyme
is irreversibly inactivated. Then
k
obs
¼ k
2
ð24Þ
An advantage is that this approach involves only five
parameters to fit instead of eight in the two-step model.
Food Bioprocess Technol
The effect of the temperature is then still described using the
transitional state theory as shown in Eq. 23.
Parameter Optimization and Statistical Evaluation
To directly extract the kinetics from single droplet drying
experiments the following approach was taken:
1. The temperature and moisture content histories of sev-
eral droplets were predicted using the drying model
(Eqs. 3 and 4).
2. The inactivation kinetic constants were optimised to fit
the measured residual enzyme activities after the drying
for all droplets simultaneously.
The parameter optimisation was carried out using a non-
linear least square method solved using the Levenberg–
Marquardt algorithm (Seber and Wild 2005). The confi-
dence interval of the parameters (p00.95) were estimated
using the Hessian matrix, which was again derived from the
Jacobian matrix of the solution (Richard et al. 2005). All
calculations were performed with MATLAB version 7.10.
Materials and Methods
Sample Preparation
The enzyme, β-galactosidase from Aspergillus oryzae (Sigma-
Aldrich, Germany) was dissolved in a buffer solution. The
solution was then filtered with a 0.2-μm Minisart sterile sieve
(Sartorius Stedim Biotech SA, Germany) and stored overnight
in a refrigerator. The maltodextrin, with DE 4-7 (Sigma-
Aldrich, Germany), was dissolved in a buffer. The buffer was
prepared from 0.2 M Na
2
HPO
4
(Sigma-Aldrich, Germany)
and 0.1 M citric acid (C
6
H
8
O
7
) solutions (Sigma-Aldrich,
Germany). The pH of the buffer was adjusted to 6.00±0.01.
The feed for single droplet drying experiments was pre-
pared by mixing 600 μL 2.5% w/w enzyme solution and
2,400 μL 25% maltodextrin solution. The feed for the
laboratory-scale spray dryer was prepared by mixing 4 mL
2% w/w enzyme solution and 96 mL 20.8% w/w maltodex-
trin solution. A lower enzyme concentration is used during
the laboratory-scale spray dryer experiments. The major
reason was that a larger sample volume could be easily
obtained for performing the enzyme activity test. This is
allowed, since at low enzyme concentration, the inactivation
kinetics of β-galactosidase is not affected by its concentra-
tion (Yamamoto and Sano 1992).
Deposited Droplet Drying Experiments
The single droplet drying experiments involved the subse-
quent steps: droplet generation, droplet drying, rehydration
and enzyme activity measurement. A micro-dispenser was
used to generate the droplet as described by Perdana et al.
(2011a). The droplet was deposited on a polypropylene
membrane (Akzo Nobel Faser Ag., The Netherlands) posi-
tioned on a platform from stainless steel slab; the thickness
of the membrane was 0.15 mm, and the stainless steel slab
was 1 mm. The droplet was then positioned in a drying
tunnel (Fig. 2). By guiding the drying air through a porous
medium, a uniform flow distribution could be achieved
within the tunnel. The tunnel was insulated and heated with
heating oil to ensure that the air temperature was constant.
The temperature and relative humidity of the air bulk air was
monitored using SHT75 temperature and humidity sensor
(Sensiron AG, Switzerland). The temperature near the droplet
was monitored using a thermocouple Type K (NiCr–NiAl; RS
Component, United Kingdom) with probe diameter of
250 μm. Furthermore, the setup was equipped with a μEye
1480ME CCD camera with a lens with ×9 magnification ratio
(Imaging Development Systems GMBH, Germany) to moni-
tor the droplet geometry evolution during drying.
The single droplet drying experiments were performed
using dry air (RH00.0%), preheated to a temperature between
80 and 110 °C and a bulk air velocity of 0.20 m/s. After
drying, the resulting powder particle was dissolved in 50 μL
buffer solution, stored overnight in the refrigerator, and then
the enzyme activity was measured.
Laboratory-Scale Spray Drying Experiments
The enzyme solution was dried with a Buchi B-190 spray
dryer (Buchi Labortechnik AG, Switzerland). The residence
Fig. 2 Schematic drawing of
the drying tunnel, side view (S)
and top view (T)
Food Bioprocess Technol
time of the particle inside this small spray dryer can be as
short as 1 s. Based on the equipment specification, the
particle diameter after drying was between 2 and 25 μm.
The inlet air temperature was 180 °C, and the outlet air
temperature varied between 80 and 130 °C. Subsequently,
around 1 g of powder was taken and dried further in a
heating chamber at 105 °C for 72 h to determine the
moisture content of the powder after spray drying. Addi-
tionally, a 0.200-g powder sample was reconstituted to
3.80 mL buffer solution and analysed for its residual enzyme
activity.
Measurement of the Residual Activity of β-galactosidase
The residual activity of β-galactosidase was measured using an
o-nitrophenyl-β-D-galactopyranoside (ONPG) assay (Sigma-
Aldrich, Germany) according to Perdana et al. (2011b). The
absorbance of the samples incubated in ONPG solution was
immediately measured after incubation at a wavelength
420 nm using a spectrophotometer (Beckman Coulter, Inc.,
USA).
Particle Size Measurement
Approximately 1 g of the spray-dried sample was dried
further in a heating chamber at 105 °C for 72 h to reduce
the moisture content and to avoid particle agglomeration.
Afterwards, the particle size distribution of the sample was
measured using Mastersizer Scirocco 2000 (Malvern Instru-
ment LTD, England).
Results and Discussion
Droplet Geometry Evolution
Snapshots of a deposited droplet during the drying process
are shown in Fig. 3. The droplet shrinks uniformly before
30 s, and then starts to develop wrinkles. The uneven
shrinkage after 60 s suggests that a thin non-flowing layer
is developed, which cannot accommodate a homogenous
shrinkage (Walton and Mumford 1999). Instead, it collapses
which leads to an irregular shape of the particle.
Drying Model
In the heat and mass transfer model, the sessile droplet is
assumed to be a perfect sphere. This assumption reduces the
complexity of the model and is acceptable since the initial
sessile droplet has a very high contact angle (>130 °C) and
also the final powder particle remains approximately spheri-
cal, as shown in Fig. 3. The predicted temperature and mois-
ture profiles are shown in Fig. 4.
The model predicts that the moisture content at the sur-
face decreases faster than in the centre of the droplet and
reaches a moisture content close to 0% at around 35 s. This
is in line with the visual observations, i.e. the occurrence of
wrinkles indicating the presence of a thin dry layer. As
shown in Fig. 4 (bottom), at t 031 s, the moisture content
gradient near the surface of the droplet is very steep, indi-
cating that a very thin layer near the droplet surface is very
dry. The model also predicts that the droplet radius does not
change any more after approximately 60 s of drying. This is
also in line with the visual observations.
The droplet temperature at the beginning (t <40 s) of the
drying process is equal to the corresponding wet bulb tem-
perature of the heating air (Fig. 4, top). When the moisture
content near the surface of the droplet decreases to less than
100 kg/m
3
, the droplet temperature starts to increase. At this
point the water evaporation rate decreases together with the
lower vapour pressure at the droplet surface (a
w
<1). While
the evaporation rate decreases, the heat transfer into the
droplet continues and causes an accumulation of heat, ob-
served as an increase in droplet temperature.
Predicting the Residual Enzyme Activity After Drying
The single droplet drying method is applied to dry β-
galactosidase suspended in a maltodextrin matrix. The inac-
tivation kinetics of β-galactosidase determined with inde-
pendent experimental data was combined with the heat and
mass transfer model to predict the residual enzyme activity
after drying (Perdana et al. 2011b). Subsequently, the pre-
dicted enzyme activity is compared to the drying results. In
modelling the inactivation of β-galactosidase, the pH change
due to decreasing moisture content was neglected. This is
allowed since most of the enzyme activity loss takes place at
high moisture content where the presence of the added buffer
stabilizes the pH. Furthermore, at lower moisture content, the
enzyme activity loss is negligible at the time scale of drying
applied. The model predictions and the experimental data are
shown in Fig. 5. As shown in Fig. 5, the heat and mass transfer
model using the independent kinetic inactivation model for β-
galactosidase is in reasonable agreement with the experimen-
tal data.
Both the predictions and the experimental data show that
the enzyme activity does not decrease during the initial
Fig. 3 Droplet geometry change during a single droplet drying exper-
iment at an air temperature of 80 °C, an absolute air humidity of 0 g/kg
dry air, a bulk air velocity of 0.20 m/s, an initial droplet height of
800 μm and an initial droplet moisture content of 80% w/w
Food Bioprocess Technol
drying period (<40 s) because the temperature of the droplet,
which is near the (low) wet bulb temperature. Then, a rapid
inactivation rate is observed, which slowly declines to result
in a final enzyme activity. The subsequent rapid inactivation
can be explained by an increase in particle temperature. At
Fig. 4 Top temperature and moisture content history of a droplet dried
during a single droplet drying experiment, T
a
is the air temperature
contacting the droplet, averaged as the air temperature at half droplet
height; T
d
is the droplet temperature; and x
w
is the moisture content.
Middle droplet radius change during drying from the visual monitoring
(symbol) and the model (line). Bottom moisture content distribution
inside the droplet at t 032 s, the symbols show the border of the
partition for each layer. The drying is carried out at an air temperature
of 80 °C, a bulk air velocity of 0.20 m/s and an absolute air humidity of
0 kg/kg dry air. The initial moisture content is 80% w/w and the initial
droplet height is 800 μm
Food Bioprocess Technol
this point the heat transfer is not compensated by sufficient
water evaporation. The increase in temperature has especial-
ly impact on the enzyme present in the centre of the droplet
as the moisture content in the centre is still high as also
reported for drying of other heat-sensitive products (Chen
and Patel 2007; Langrish 2009). After a while the enzyme
inactivation rate decreases and the residual enzyme activity
in the particle is obtained. The enzyme inactivation rate
decreases because of the decreasing moisture content, espe-
cially near the surface of the droplet and finally also in the
centre of the droplet. Most of the residual active enzyme is
located near the surface of the particle as shown in Fig. 5
(bottom). It can be concluded that the different moisture
content history at different locations in the droplet deter-
mines the large differences in residual enzyme activity
across the droplet radius. A strategy that might be followed
to retain maximum enzyme activity is to minimize the
presence of enzyme in the centre of the droplet. This can
for example be achieved by drying the enzyme in a droplet
that forms a hollow sphere upon drying (Etzel et al. 1996) or
by applying a coating of a concentrated enzyme solution on
pre-dried particles.
The results also show that the initial droplet diameter
determines the residual enzyme activity to a large extent.
Figure 5 indicates that the residual activity is higher for
smaller droplets. Although inactivation starts earlier in
a smaller droplet, the critical region is shorter (i.e. the
combination of high temperature and high moisture con-
tent in the centre of the droplet) compared to that in the
larger droplet. This implies that with respect to enzyme
inactivation, a smaller droplet size is preferred for spray
drying.
Direct Extraction of Inactivation Kinetic Parameters
from Drying Experiments
The parameters describing the specific inactivation kinetics of
β-galactosidase in maltodextrin may be extracted directly from
the single droplet drying results. This is preferred as these
experiments are much less labour intensive and can be scaled
out to facilitate a high throughput approach. If the method is
reliable, the parameters for β-galactosidase inactivation direct-
ly obtained from single droplet drying experiments should be
similar to the ones obtained from the separate heating experi-
ments (Perdana et al. 2011b).
Both the one-step and the two-step inactivation models
were evaluated to describe the experimental drying data.
The results of the parameter optimisation are compared to
the parameters from the earlier study (Table 2).
It can be observed that the parameters obtained from the
drying experiments differ from the parameter values from
the heating experiments (Table 2). This can be explained by
the fact that the inactivation during single droplet drying
experiments occurs primarily during a very short critical
period. This critical period is dictated by the drying history
and involves temperatures and moisture content values that
lead to rapid inactivation. Therefore, the parameter values
are optimised such that they specifically describe the inacti-
vation during this critical period. This explanation is further
confirmed by implementing the one-step inactivation model,
which is also able to describe the inactivation during drying
experiments well. The data fromthe heating experiments were
obtained at many different temperature and moisture content
Fig. 5 Top the residual enzyme activity after drying of a single sessile
droplet at air temperatures of 80 °C (empty circles), 87 °C (empty
squares), 95 °C (empty triangles) and 110 °C (empty diamonds) for
an initial droplet diameter of 800 μm. Middle the effect of initial
droplet size on residual enzyme activity for an initial droplet diameter
of 800 μm (empty triangles) and 1,400 μm (empty circles). The error
bar shows the standard deviation of the data. The predictions (solid
line) are based on the inactivation kinetics of β-galactosidase from the
constant heating experiments. Bottom the residual activity of β-
galactosidase as a function of the radial coordinate (solid line) and of
the volumetric coordinate (symbols) after drying of a droplet with an
initial diameter of 800 μm at an air temperature of 95 °C for 300 s
R
Table 2 The parameter values
for the inactivation kinetics of
β-galactosidase estimated from
heating experiments and from
single droplet drying
experiments
a
The uncertainty of the parame-
ter is provided within 95% con-
fidence interval
b
The values of the parameters
are according to Perdana et al.
(2011b)
c
Reference value, not fitted
Estimated parameter
a
Heating experiments
b
Drying experiments
Two-step inactivation model
(Eqs. 21, 22, and 23)
One-step inactivation
model (Eq. 23)
ΔΔS
1

(J mol
−1
K
−1
) 1.08∙10
3
±0.25∙10
3
1.31∙10
3
±0.0063∙10
3
NA
ΔΔH
1

(J mol
−1
) 3.57∙10
5
±0.81∙10
5
4.86∙10
5
±0.023∙10
5
NA
ΔS
2,w

(J mol
−1
K
−1
) 6.75∙10
2
±1.17∙10
2
5.49∙10
2
±0.097∙10
2
1.88∙10
3
±0.00010∙10
3
ΔH
2,w

(J mol
−1
) 3.28∙10
5
±0.40∙10
5
2.76∙10
5
±0.063∙10
5
7.74∙10
5
±0.00023∙10
5
ΔH
2,s

(J mol
−1
) 1.28∙10
5
±0.20∙10
5
1.57∙10
5
±0.027∙10
5
4.77∙10
5
±0.0010∙10
5
m 2.60∙10
4
±0.00094∙10
4
8.92∙10
3
±0.0010∙10
3
NA
p 1.16±0.0056 5.37±0.011 5.67±0.00038
T
int
(°C) 33.85±0.065 −49.65±8.80 8.25±0.026
T
ref
(°C)
c
68.50 68.50 68.50
Food Bioprocess Technol
combinations and could be varied independently. Therefore,
the values are based on a much wider data set, but having less
data in the specific range that are relevant for spray drying.
Another reason for the difference between the kinetic
parameters is that the parameter optimisation procedure
forces the model to describe the residual enzyme activity
after the drying. By doing so, any uncertainties in the heat
and mass transfer model are neglected. However, from
earlier observations, it was concluded that the heat and mass
transfer model could predict the droplet shrinkage and the
residual enzyme activity using the kinetic parameters from
the heating experiments reasonably accurate. Therefore, it is
believed that the uncertainties involved are not very large
and that the kinetic parameters of the single droplet experi-
ments remain valid. This conclusion is supported by labo-
ratory spray drying experiments.
In Fig. 6 (bottom), the effect of the initial droplet diameter
on the residual β-galactosidase activity after drying is shown.
The figure shows that the model accurately predicts the ex-
perimental data. Therefore, it can be concluded that although
the values for the parameters in the model for inactivation
kinetics obtained from the drying experiments may be less
precise, the inactivation kinetics are accurate enough to pre-
dict the effect of drying on the inactivation of the enzyme.
This is because the critical drying period dictates the end
product properties.
Table 2 shows that the values of the confidence
intervals of the parameters are smaller for the one-
step inactivation compared to the two-step inactivation
model. Therefore, it may be concluded that the one-
step inactivation model is more favourable than the
two-step inactivation model to describe the effect of
Fig. 6 Top the fitted residual enzyme activity obtained from single
droplet drying experiments at varying air temperatures of 80 °C (empty
circles), 87 °C (empty squares), 95 °C (empty triangles) and 110 °C
(empty diamonds) for an initial droplet diameter of 800 μm. Bottom the
residual enzyme activity predicted using the inactivation kinetics
extracted from drying data at various heating air temperatures com-
pared to the drying results from different initial droplet diameter:
500 μm (empty squares) and 1,400 μm (empty circles). The β-
galactosidase inactivation is described by the two-step inactivation
model (left) and the one-step inactivation model (right)
Food Bioprocess Technol
drying on enzyme activity loss. From practical point of view,
the one-step inactivation model may be sufficient to translate
the measured inactivation from single droplet experiments
into useful inactivation kinetics.
Laboratory-Scale Spray Drying
The models were also applied to predict the residual enzyme
activity after drying on a laboratory-scale spray dryer. The
results are shown in Fig. 7. These results indicate that at an
outlet air temperature below 100 °C, the enzyme is hardly
inactivated upon drying, while with an outlet air temperature
higher than 100 °C, the enzyme is increasingly inactivated
with higher temperature as also observed by Yamamoto and
Sano (1992).
The volume-weighted mean diameter (d
4,3
) of the spray-
dried particle is 6.94 μm, fed into the drying model to
predict the residual enzyme activity. By assuming that the
volumetric droplet shrinkage is equal to the amount of water
removed, the initial droplet diameter was estimated at 11.8 μm
(initial moisture content 80% w/w, Langrish 2009).
The model can predict the drying results reasonably
accurate, but at temperatures between 100 and 120 °C, the
predicted value of the residual activity of β-galactosidase is
slightly lower than the experimental results. This may be
due to the lower initial droplet size in reality due to non-
ideal shrinkage leading to more rapid drying and thus lower
inactivation (see also Fig. 3).
Furthermore, Fig. 7 shows that there is only a small differ-
ence between the two inactivation models at temperatures
larger than 100 °C. This is probably because the inactiva-
tion kinetic constant of the one-step inactivation model
increases more quickly with temperature. This may lead
to an overestimation of the inactivation kinetic constant at
high temperatures.
Conclusions
A newly developed, small-scale experimental setup to mim-
ic spray drying was presented which involves the drying of
single droplets deposited on a hydrophobic flat plate. This
approach ensures rapid and inexpensive trials while main-
taining key parameters similar to the process condition on
spray drying. In this study, the setup is used to evaluate the
inactivation of β-galactosidase during drying.
To describe the physical and chemical changes during
drying, a model based on heat and mass transfer is pre-
sented. The model, combined with the inactivation kinetics
of β-galactosidase from separate heating experiments, was
used to predict the loss in enzyme activity after drying. It is
found that the model can provide a reasonably accurate
prediction on the residual enzyme activity.
It was also shown that the inactivation kinetics of β-
galactosidase can be directly extracted fromthe single droplets
drying experiments rather than using the kinetics from sepa-
rate heating experiments. The parameters that were obtained
in this way were used to predict other experimental results
both from other single droplet drying experiments and from a
laboratory-scale spray dryer. The new inactivation kinetics
provides reasonably accurate prediction on the residual en-
zyme activity for both procedures.
Fig. 7 Left the residual activity of β-galactosidase after spray drying
in a laboratory-scale spray dryer. The symbols represent the experi-
mental results at an inlet air temperature of 180 °C and varied outlet air
temperature. The lines represent the predicted enzyme activity using
the inactivation kinetics obtained from the single droplet drying experi-
ments and from the constant heating experiments. Right the particle
size distribution of the powder obtained from the laboratory-scale
spray dryer
Food Bioprocess Technol
Nomenclature
A Area m
2
a
w
Water activity –
c
p
Heat capacity J kg
−1
K
−1
C Concentration kg m
–3
C
g
Constant in GAB sorption model –
d Diameter (2R
d
) m
D Diffusion coefficient m
2
s
−1
E
a
Arrhenius-type diffusion
activation energy
J mol
−1
f Parameter to describe the
effect of moisture content
on conformational
stability of the enzyme

g Parameter to describe the effect
of moisture content on the
irreversible inactivation
(second step) kinetic
constant of the enzyme

h Planck’s constant
(6.626×10
−34
)
J s
−1
h Convective heat
transfer coefficient
J s
−1
m
−2
K
−1
ΔH
lv
Enthalpy of evaporation J kg
−1
ΔH

Activation enthalpy J mol
−1
ΔΔH

Activation enthalpy difference
between unfolding and
refolding reaction
J mol
−1
J Evaporation rate kg s
−1
k
B
Boltzmann’s constant
(1.380×10
−23
)
J K
−1
k
obs
Observed inactivation kinetic
constant of β-galactosidase
s
−1
k
2
Irreversible complete
denaturation rate constant
of β-galactosidase
s
−1
k
c
Mass transfer coefficient m s
−1
K
g
Constant in GAB sorption model –
K
1
Unfolding equilibrium constant
of β-galactosidase
s s
−1
L Length m
m Mass kg
m
u
Parameter to describe the
effect of moisture content
on unfolding equilibrium
of β-galactosidase

M
w
Molecular weight of water
(18∙10
−3
)
kg mol
−1
n Number of partition –
p Uncertainty of the parameters –
p
i
Parameter to describe the
effect of moisture content
on irreversible complete
denaturation rate constant
of β-galactosidase

P Pressure Pa
Pr Prandtl number –
Q Heat J s
−1
r Distance in radial coordinate m
Nomenclature
R Ideal gas constant (8.314) J mol
−1
K
−1
Re Reynolds number –
R
d
Droplet radius m
ΔS

Activation entropy J mol
−1
K
−1
ΔΔS

Activation entropy difference
between unfolding and
refolding reaction
J mol
−1
K
−1
Sc Schmidt number –
t Time s
T Temperature °C
u Velocity m s
−1
V Volume m
3
x Distance in Cartesian coordinate m
x
w
Mass fraction of water kg kg
−1
X
w
Moisture content kg kg
−1
dry matter
X
wm
Monolayer moisture content in
GAB sorption model
kg kg
−1
dry matter
y Distance in Cartesian coordinate m
z Distance in Cartesian coordinate m
Z Thickness m
Greek symbols
δ Hydrodynamic boundary
layer thickness
m
δ
t
Thermal boundary
layer thickness
m
l Thermal conductivity J s
−1
m
−1
K
−1
μ Dynamic viscosity Pa s
ρ Density kg m
−3
υ Kinematic viscosity m
2
s
−1
Subscript
a Air
avg Average
cond Conduction
d Droplet
int Intercept
m Very dry matrix
obs Observed
p Flat plate for
droplet deposition
ref Reference
s Solute
sat Saturated
w Water
0 Initial
∞ Bulk air
Acknowledgement The authors thank Advanced Chemical Technol-
ogies for Sustainability (ACTS)–Netherlands Organization for Scien-
tific Research (NWO) for financial aid through Process on a Chip
(PoaC) program.
Open Access This article is distributed under the terms of the Crea-
tive Commons Attribution Noncommercial License which permits any
Food Bioprocess Technol
noncommercial use, distribution, and reproduction in any medium,
provided the original author(s) and source are credited.
References
Adhikari, B., Howes, T., Bhandari, B. R., & Truong, V. (2000).
Experimental studies and kinetics of single drop drying and their
relevance in drying of sugar-rich foods: A review. International
Journal of Food Properties, 3(3), 323–351.
Ali Al Zaitone, B., & Tropea, C. (2011). Evaporation of pure liquid
droplets: Comparison of droplet evaporation in an acoustic field
versus glass-filament. Chemical Engineering Science, 66(17),
3914–3921.
Bolz, R. E. & Tuve, G. L. (1973). CRC handbook of tables for applied
engineering science. CRC Press, Cleveland, Ohio, USA.
Chen, X. D., & Patel, K. C. (2007). Micro-organism inactivation
during drying of small droplets or thin-layer slabs—A critical
review of existing kinetics models and an appraisal of the drying
rate dependent model. Journal of Food Engineering, 82(1), 1–10.
Chuchottaworn, P., Fujinami, A., & Asano, K. (1984). Experimental
study of evaporation of a volatile pendant drop under high mass
flux conditions. Journal of Chemical Engineering Japan, 17, 7–13.
Coumans, W. J. (2000). Models for drying kinetics based on drying curves
of slabs. Chemical Engineering and Processing, 39(1), 53–68.
Crank, J. (1990). The mathematics of diffusion. London, UK: Oxford
University Press.
Etzel, M. R., Suen, S. Y., Halverson, S. L., & Budijono, S. (1996).
Enzyme inactivation in a droplet forming a bubble during drying.
Journal of Food Engineering, 27(1), 17–34.
Fang, Z., & Bhandari, B. (2010). Encapsulation of polyphenols—A
review. Trends in Food Science & Technology, 21(10), 510–523.
Farid, M. (2003). A new approach to modelling of single droplet
drying. Chemical Engineering Science, 58(13), 2985–2993.
Filková, I., Huang, L.X. & Mujumdar, A. (2006). Industrial spray
drying systems. In: Handbook of Industrial Drying, 3rd Ed. Boca
Raton: CRC Press.
Gianfrancesco, A., Turchiuli, C., Flick, D., & Dumoulin, E. (2010).
CFD modeling and simulation of maltodextrin solutions spray
drying to control stickiness. Food and Bioprocess Technology, 3
(6), 946–955.
Goff, J. A., & Gratch, S. (1946). Low-pressure properties of water from -
160 to 212 F, in Transactions of the American Society of heating and
ventilating engineers. The 52
nd
annual meeting of the American
society of heating and ventilating engineers. New York: 95–122.
Goula, A. M., & Adamopoulos, K. G. (2004). Influence of spray
drying conditions on residue accumulation-simulation using
CFD. Drying Technology, 22(5), 1107–1128.
Incropera, F. P., & De Witt, D. P. (1985). Fundamentals of heat and
mass transfer (2nd ed.). New York: Wiley.
Kakac, S., & Yener, Y. (1993). Heat conduction. Washington: Taylor &
Francis.
Langrish, T. (2009). Degradation of vitamin C in spray dryers and
temperature and moisture content profiles in these dryers. Food
and Bioprocess Technology, 2(4), 400–408.
Li, X., Anton, N., Arpagaus, C., Belleteix, F., & Vandamme, T. F.
(2010). Nanoparticles by spray drying using innovative new tech-
nology: The Büchi Nano Spray Dryer B-90. Journal of Controlled
Release, 147(2), 304–310.
Luyben, K. C. A. M., Liou, J. K., & Bruin, S. (1982). Enzyme
degradation during drying. Biotechnology and Bioengineering,
24(3), 533–552.
Meerdink, G., & van't Riet, K. (1991). Inactivation of thermostable
[alpha]-amylase during drying. Journal of Food Engineering, 14
(2), 83–102.
Mezhericher, M., Levy, A., & Borde, I. (2008). Heat and mass transfer
of single droplet/wet particle drying. Chemical Engineering Sci-
ence, 63(1), 12–23.
Millqvist-Fureby, A., Malmsten, M., & Bergenståhl, B. (1999). Spray-
drying of trypsin—surface characterisation and activity preserva-
tion. International Journal of Pharmaceutics, 188(2), 243–253.
Mosaad, M. (1999). Laminar forced convection conjugate heat transfer
over a flat plate. Heat and Mass Transfer, 35(5), 371–375.
Peighambardoust, S. H., Golshan Tafti, A., & Hesari, J. (2011). Ap-
plication of spray drying for preservation of lactic acid starter
cultures: A review. Trends in Food Science & Technology, 22(5),
215–224.
Perdana, J., Fox, M. B., Schutyser, M. A. I., & Boom, R. M. (2011a).
Single-droplet experimentation on spray drying: Evaporation of a
sessile droplet. Chemical Engineering and Technology, 34(7),
1151–1158.
Perdana, J., Fox, M. B., Schutyser, M. A. I. & Boom, R. M. (2011b).
Enzyme inactivation kinetics: Coupled effects of temperature and
moisture content. Food Chemistry (in press).
Quirijns, E. J., Boxtel, A. J. B., Loon, W. K. P., & Straten, G. (2005).
Sorption isotherms, GAB parameters and isosteric heat of sorp-
tion. Journal of the Science of Food and Agriculture, 85(11),
1805–1814.
Ranz, W. E., & Marshall, J. M. (1952). Evaporation from drops, Part II.
Chem. Eng. Prog., 48(3), 141–146 (Part I); 48(4), 173–l 80 {Part II).
Räderer, M., Besson, A., & Sommer, K. (2002). A thin film dryer ap-
proach for the determination of water diffusion coefficients in viscous
products. Chemical Engineering Journal, 86(1–2), 185–191.
Ré, M. I. (1998). Microencapsulation by spray drying. Drying Tech-
nology: An International Journal, 16(6), 1195–1236.
Richard, C., Aster, B. B. & Clifford, H. T. (2005). Appendix C. Review
of vector calculus. In: International Geophysics, pp. 273–280.
New York: Academic Press.
Sansone, F., Mencherini, T., Picerno, P., d'Amore, M., Aquino, R. P., &
Lauro, M. R. (2011). Maltodextrin/pectin microparticles by spray
drying as carrier for nutraceutical extracts. Journal of Food En-
gineering, 105(3), 468–476.
Schiffter, H., & Lee, G. (2007). Single-droplet evaporation kinetics and
particle formation in an acoustic levitator. Part 1: Evaporation of
water microdroplets assessed using boundary-layer and acoustic
levitation theories. Journal of Pharmaceutical Sciences, 96(9),
2274–2283.
Schlichting, H., & Gersten, K. (2000). Boundary-layer theory (8th ed.).
Berlin: Spinger.
Seber, G. A. F. & Wild, C. J. (2005). Computational methods for
nonlinear least squares. In: Nonlinear regression, pp. 619–660.
New York: Wiley.
Sloth, J., Kiil, S., Jensen, A. D., Andersen, S. K., Jørgensen, K.,
Schiffter, H., & Lee, G. (2006). Model based analysis of the
drying of a single solution droplet in an ultrasonic levitator.
Chemical Engineering Science, 61(8), 2701–2709.
Straatsma, H., Verschueren, M., Gunsing, M., Verdurmen, R. & Jong,
P. d. (2007). CFD simulation of spray drying of food products. In:
Computational fluid dynamics in food processing, pp. 249–286.
New York: CRC Press.
Thirumaleshwar, M. (2009). Fundamentals of heat and mass transfer.
New Delhi: Dorling Kindersley.
Vehring, R., Foss, W. R., & Lechuga-Ballesteros, D. (2007). Particle forma-
tion in spray drying. Journal of Aerosol Science, 38(7), 728–746.
Walton, D. E., & Mumford, C. J. (1999). The morphology of spray-
dried particles: The effect of process variables upon the morphol-
ogy of spray-dried particles. Chemical Engineering Research and
Design, 77(5), 442–460.
Wijlhuizen, A. E., Kerkhof, P. J. A. M., & Bruin, S. (1979). Theoretical
study of the inactivation of phosphatase during spray drying of
skim-milk. Chemical Engineering Science, 34(5), 651–660.
Food Bioprocess Technol
Wu, W. D., Lin, S. X., & Chen, X. D. (2011). Monodisperse droplet
formation through a continuous jet break-up using glass nozzles
operated with piezoelectric pulsation. AICHE Journal, 57(6),
1386–1392.
Xin Huang, L., Filkova, I. & Mujumdar, A. (2006). Industrial spray
drying systems. In: Handbook of industrial drying, 3rd ed. New
York: CRC Press.
Yamamoto, S., & Sano, Y. (1992). Drying of enzymes: enzyme reten-
tion during drying of a single droplet. Chemical Engineering
Science, 47(1), 177–183.
Yoshioka, S., Aso, Y., Izutsu, K-i, & Kojima, S. (1994). Is stability
prediction possible for protein drugs? Denaturation kinetics of β-
galactosidase in solution. Pharmaceutical Research, 11(12),
1721–1725.
Food Bioprocess Technol

and it is thus assumed that the diffusion occurs only in the radial direction. which deviate from the bulk conditions (Schlichting and Gersten 2000). and the shorter residence time in the laboratory spray dryer (Goula and Adamopoulos 2004. Additionally. Rd is the droplet radius. a comparable droplet size is critical to accurately mimic the kinetics in spray drying (Adhikari et al. This surface for example affects the spherical shape of the droplet. the predictive value of the model was also checked by drying of β-galactosidase in a laboratory spray dryer system. 2011a). The major advantages of this method are the possibility to vary droplet size and residence time. Bi ¼ hdd ld ð1Þ In this study. (3) inactivation kinetics of βgalactosidase. The minimum droplet diameter can be adjusted to 150 μm which is only slightly bigger than the typical droplet size in industrial spray dryers. providing strong basis for this study (Yamamoto and Sano 1992. The following differential equation is used to describe the droplet mass change in time À dmd Mw ¼ kc 4pR2 d dt R  sat aw P w P1 À Td P1  ð2Þ where md is the mass of the droplet. Filková et al.d @r @t  @Cw @r  ð3Þ Where Cw is the water concentration. Straatsma et al. Ali Al Zaitone and Tropea 2011). Therefore. Td is the temperature of the droplet. Other differences introduced by the presence of the surface are the air flow velocity and the air temperature near the droplet. caused by the different method of atomisation. Finally. 2006). Schiffter and Lee 2007). By monitoring the air temperature near the droplet and by modelling the air flow and heat transfer across the drying surface. 2008). and (4) parameter optimisation and statistical evaluation. In this study. the temperature gradient inside the droplet could be neglected. Incropera and De Witt 1985). immobilised. 1994.Food Bioprocess Technol between laboratory and industrial spray dryers are the smaller droplet size. 2011b). 2007). 2007. intrusively levitated that is in pendant or filament (Chuchottaworn et al. 2000. the latter method is followed (Perdana et al. R is the ideal gas constant. it is virtually impossible to extract sufficient representative experimental data to calibrate a kinetic model. Perdana et al. Bi is defined as the ratio of external heat transfer (between air and droplet) and the internal heat transfer (in the droplet. Bi is around 0. Modelling and Statistical Evaluation This section consists of four parts: (1) model description for a single droplet drying. because it is an industrial relevant enzyme applied for production of lactose-hydrolysed milk and whey. 2000. and subsequently dried by contacting it with well-defined drying air. and to dry multiple droplets at once to obtain higher volume of sample. (2) modelling temperature and air flow distribution across the flat plate. aerodynamic or electromagnetic levitation (Adhikari et al. the influence of the surface on the drying conditions can be quantified. The inactivation kinetics of β-galactosidase from the previous studies were combined with the heat and mass transfer model to predict the residual activity of β- galactosidase during the single droplet drying experiments. The enzyme β-galactosidase is selected as the model enzyme in this study. Yoshioka et al. t is the time. Mezhericher et al. Dw. Since the laboratory spray dryer can (due to its design and dimensions) only cover a limited range of droplet sizes and drying times and it produces droplets with a size distribution (not monodisperse).02. Model Description for Single Droplet Drying In analogy to several previous studies. 2011a). Examples are drying of a droplet that is non-intrusively levitated using for example acoustic. 2006. 1984. For the small droplet sizes considered. Vehring et al. 2006. The developing moisture gradient inside the droplet is described by Fickian diffusion @Cw @ ¼ Dw.d is the effective diffusion coefficient of water in the droplet and r is . or that is deposited on a flat surface (Perdana et al. The sessile droplet is approximated as a spherical droplet. kc is the mass transfer coefficient. The challenge in using this approach is the presence of the flat surface. Pwsat is the saturated vapour of water at Td. In this approach a small droplet is generated. P∞ is the vapour partial pressure in the bulk air and T∞ is the bulk air temperature. a hydrophobic surface is used to minimize the contact between the droplet and the surface and to retain the spherical shape.1. Another approach is the drying of single droplets under well-controlled conditions. though under steady-state conditions. This study focuses on the drying of the enzyme βgalactosidase suspended in a maltodextrin matrix. Sloth et al. a heat and mass transfer model is used to describe the drying of a sessile single droplet (Sloth et al. This assumption is valid if the Biot number (Bi) <0. Mw is the molecular weight of water. its activity can be easily determined and its inactivation kinetics has been studied before. aw is the water activity.

0 Vd.0 is the initial solute concentration and Vd. which are tabulated in Table 1. The coupled heat and mass transfer models were solved numerically using 100 radial layers in the droplet. The conductive heat transfer between the flat plate and the droplet is preferably as low as possible. Furthermore. The layer thickness Δr is & Δri ¼ ri . 1). By using a hydrophobic surface. To minimize turbulence (eddy formation) near the edge of the depositing plate. cp. To solve the model. The model was therefore corrected to describe the decrease of layer thickness due to shrinkage (Farid 2003). Modelling Temperature and Air Flow Distribution Across the Flat Plate In this study. 9.s the heat capacity of the water and the solute respectively. when one wants to mimic the drying of a droplet suspended in air. results in mw ¼ ms C  w  ρ s 1 À Cw ρ w ¼0 ð10Þ t ¼ t. dCw dr ¼0   t ¼ t. The radius of the ith partition is ÀP Á !1=3 3 i0 Vi ð12Þ ri ¼ 4p The thickness of every partition could be recalculated using Eq. a thin flat plate is used rather than a blunt-edged plate. This shrinkage influences the heat and mass transfer within the droplet.0 n ð5Þ where Cs.s Á dTd dt ¼ hd 4pRd 2 ðT1 À Td Þ À dmd Hlv dt The droplet shrinks upon drying due to the evaporation of water.d dCw ¼ kc 4pRd 2 Mw dr R sat aw P w Td 1 À P1 T The temperature change of the droplet is described by the following enthalpy balance: À mw cp. i ¼ 1 ri À riÀ1 . The average water concentration is then defined as Cw ¼ mw ¼ V mw ρw mw þ ms ρ s ð9Þ . r ¼ Rd .Food Bioprocess Technol the radial coordinate. the new volume of the layer was recalculated as   ms 1 V ¼ 1þ ð11Þ ρs ðρw À Cw Þ The radial coordinate of every partition was corrected using this new volume. ÀDw.w þ ms cp. 0 r Rd . several closure equations like heat and mass transfer coefficient. the contact area for conductive heat transfer has already been reduced.0 is the initial volume of the droplet. and vapour pressure are required.w and cp. the presence of the flat plate still influences the flow pattern of the drying air above the plate. the temperature in the air near the flat plate deviates from the bulk air temperature affecting the drying process. In each layer the same solute mass (ms) was calculated according to ms ¼ Cs. water activity. r ¼ 0. This heat is further transferred from the plate to the droplet through conduction. It is of major importance to quantify the heat transfer between the flat plate and the droplet and compare this to the convective heat transfer via the drying air. respectively. From the water concentration. the droplet was deposited on a flat plate. 3 (Crank 1990). The radial position of the ith layer from the centre of the droplet can be calculated as follows  ri ¼ iÁ Rd 3 n 1=3 ð6Þ where n is the total number of layers for numerical integration. i. The drawback of this configuration is that in addition to heat transfer from the air to the droplet. also heat is transferred from the air to the flat plate through convection. the initial solute mass distribution is approximated with the initial volume distribution. i ! 2 ð7Þ The volume of each layer V is then V ¼ mw ms þ ρw ρs ð8Þ where ρw and ρs are the densities of water and solute. However. The boundary conditions applied here are t ¼ 0. This correction was based on the water concentration in every layer at t>0 that is calculated from Eq. 7. near the droplet (Fig. A boundary layer model ð4Þ where ms is the mass of the solute. hd the heat transfer coefficient to the droplet through convection and Hlv is the evaporation enthalpy of water.e. dCw dr Reorganizing Eq. Since the initial solute concentration is assumed homogenous. moisture diffusion coefficient. Therefore a model description of the temperature gradients within the plate and in the air near the plate was developed.

d. Tst is 373. Pr is the Prandtl number of air. (2002) Diffusion coefficient of water in air Bolz and Tuve (1973) according to Mosaad (1999) is used to describe the air flow and temperature distribution across the flat plate: a hydrodynamic boundary layer is defined for the air flow distribution and a thermal boundary layer for the temperature distribution À Á 1 À z Á3 u 3 z ð13Þ z d u1 ¼ 2 d À 2 d . T∞ is the bulk air temperature.a is the heat capacity of air and μa is the dynamic viscosity of air. used in the drying model for single droplet drying: hdd 0:51 0:33 Pr la ¼ 0:24 þ 0:63Re kc dd ¼ 0:24 þ 0:63Re0:51 Sc0:33 Dw.a For a sessile droplet.T ¼30o C ¼ exp À 35:8þ215Xw 1þ10:2Xw À E À ÁÁ 1 o exp À a 1 À Dw.aa a .25 (in hPa) Cg ÁKg Áaw Xw Xwm ¼ ð1ÀKÁaw Þð1ÀKg Áaw þCg ÁKg Áaw Þ For a droplet suspended in air.x ÀT1 ¼1À3 2 þ1 2  3 z dt . 1 Sketch of a single droplet drying on a flat plate with the air flow pattern and governing heat transfer processes indicated. The value of 1a.a and μa are temperature dependent.d ¼ Dw. 17) and is calculated locally.T ¼30 C R T 303:15 where Ea ¼ dþ190Xww 1þ10X Dw.aa a ð17Þ 1=2 l where 1a is the thermal conductivity of air.d.15 mm) on top of a stainless steel platform (1. Pr 1 dt ¼ Pr. Pr > 1 ð16Þ The heat transfer coefficient between the air and the flat plate hp is approximated by (Thirumaleshwar 2009) À 1Á c μ hp ¼ 0:332 xla uv Pr1=3 . υ and δt is thermal boundary layer at distance x & PrÀ1=3 .a c μ ρ where Re ¼ ρa dd u1 . cp.x is the temperature of the surface of the flat plate at distance x (at the front edge of the flat plate.00 mm) .15 K and est is 1013. The hydrodynamic boundary layer for u00. The heat transfer coefficient decreases as a function of the x coordinate (Eq. 0 z dt ð14Þ where z is the height coordinate from the surface of the flat plate. (2011b). cp. Tp. (2005) Perdana et al. u∞ is the bulk velocity of air. δ is the hydrodynamic boundary layer at distance x and δt is the thermal boundary layer. It was Fig. in Eq. used in the drying model for laboratory-scale spray dryer: hdd 0:51 0:33 Pr la ¼ 2 þ 0:6Re kc dd ¼ 2 þ 0:6Re0:51 Sc0:33 Dw.a ¼ À2:775 Á 10À6 þ 4:479 Á 10À8 T þ 1:656 Á 10À10 T 2 Water activity (GuggenheimAnderson-de Boer model) Heat and mass transfer coefficients Quirijns et al.99u∞ is (Schlichting and Gersten 2000) d¼5 rffiffiffiffiffiffiffi ux u1 ð15Þ T ÀT1 Tp. Pr ¼ p. Pr ¼ p. 0   z dt is the kinematics viscosity of air. The sessile droplet is dried on a thin plate consisting of a hydrophobic membrane (0. x00). the value are evaluated at T0(Tp +T∞)/2. Ranz and Marshall (1952) Diffusion coefficient of water in maltodextrin solution Räderer et al.Food Bioprocess Technol Table 1 The closure equations used in drying model Equation name Saturated vapour pressure Equation Reference Goff and Gratch (1946)  sat  À Á À Á pw log 100 ¼ À7:90298 Tst À 1 þ 5:02808 log Tst T T   1st 11:344ð1À T Þ À7 10 À1:3816 Á 10 À1   Tst þ8:1328 Á 10À3 10À3:49149ð T À1Þ À 1 þ logðest Þ For vapour at total pressure 1 atm. 17. Sc ¼ μaw. u is the air velocity at z.aa μa l D   Dw.

2011b). In the latter experiments. which An advantage is that this approach involves only five parameters to fit instead of eight in the two-step model. the suspended enzyme is exposed to various temperature–moisture value combinations for a specific time and the remaining enzyme activity is measured.w exp À R    ! 1 1 xw exp Àg À 1 À cw T Tint ð23Þ where K1 is the reversible unfolding equilibrium constant. h is the Planck’s constant.avg ð18Þ where hp. the enzyme inactivation can also be regarded as a one-step inactivation process in which the native enzyme is irreversibly inactivated. 150 μm) @Tp lp ¼ @t ρp cp. the conductive heat flux Q through the contact surface of the droplet was estimated according to Q ¼ lp 2pRd z ΔTcond Δxcond ð20Þ k2 ¼ !  ! kB T ΔS z 2.m‡ is the activation enthalpy in pure solid form (i.w ΔH z 2. Alternatively.w exp À R  !  !  f ð1 À x w Þ exp À RT ð22Þ  K1 k2 1 þ K1 ð21Þ ð19Þ 1 1 À T Tref where ρp is the density of the plate. Inactivation Kinetics of β-galactosidase The inactivation rate of the β-galactosidase is affected by the temperature and moisture content of the droplet. The boundary layer applied here are that near the droplet the flat plate temperature is equal to the droplet temperature and at the edge of the flat plate (far from the droplet) the temperature gradient of the flat plate is 0. f describes the effect of moisture content on conformational stability of the enzyme. no moisture). kB is Boltzmann’s constant.s 01.w À ΔH z 2. Therefore. ΔH2. The predictions were validated by simple temperature measurements at different heights above the plate.w ΔH z 2. A kinetic inactivation model for βgalactosidase is developed in earlier work and calibrated using constant heating experiments (Perdana et al.w‡ is the activation entropy difference between the unfolding and refolding reactions in pure water.Food Bioprocess Technol estimated numerically. Z is the thickness of the flat plate and lp is the thermal conductivity of the flat plate.w 1 1 exp À exp À R RTref RT h T Tref ΔH z 2.w À K1 ¼ exp R RTref ΔΔH z 1. i. because of the small thickness. g describes the effect of moisture content on the irreversible inactivation (second step) kinetic constant of the enzyme and Tin is the intercept temperature at which k2.e. ΔH2. The temperature distribution within the plate is described with a one-dimensional partial differential equation (Kakac and Yener 1993). Then kobs ¼ k2 ð24Þ where ΔTcond/Δxcond is the temperature gradient in the partition layer of the flat plate that is closest the droplet.e. It was found that the conductive heat flux was smaller than 5% of the total heat transferred to the droplet regardless of its position. the average heat transfer coefficient is obtained by integrating hp along Δx 1=2  u 1=2  u la 1 1 1=3 ¼ 0:664 1=2 Pr À L L þ Δx v ! hp.w‡ is the activation entropy of complete denaturation in pure water.w‡ is the activation enthalpy in pure water. It is assumed that no temperature gradient in the y direction (perpendicular to the air flow direction) and in the z direction (within the plate.w/k2. Equations 14 and 19 were solved numerically to estimate the temperature distribution history within and above the flat plate.avg T1 À Tp þ @x2 Zlp 2 both change with time. The inactivation kinetics are described by a twostep inactivation process and the observed inactivation coefficient (kobs) is described as  kobs ¼ where ΔΔS z 1.avg is the average local heat transfer coefficient.w‡ is the activation enthalpy difference between the unfolding and refolding reactions in pure water. ΔΔH1. ΔΔS1.e. The non-uniform heat transfer rate across the plate contributes to the development of a temperature gradient within the plate. To estimate the conductive heat transfer between the surface and the droplet and compare it to the convective heat transfer between the air and the droplet.w ΔΔH z 1. ΔS2. . k2 is the complete denaturation rate constant.p  À Á @ Tp hp. assuming a Δx at distance L from the front side of the plate. it can be safely assumed that most heat is transferred via convective heat transfer. i.

the laboratory-scale spray dryer experiments.0%). Deposited Droplet Drying Experiments The single droplet drying experiments involved the subsequent steps: droplet generation. The temperature near the droplet was monitored using a thermocouple Type K (NiCr–NiAl. Parameter Optimization and Statistical Evaluation To directly extract the kinetics from single droplet drying experiments the following approach was taken: 1. Switzerland). 2 Schematic drawing of the drying tunnel. Germany) and stored overnight in a refrigerator.10. was dissolved in a buffer. droplet drying.95) were estimated using the Hessian matrix. The droplet was then positioned in a drying tunnel (Fig.. the resulting powder particle was dissolved in 50 μL buffer solution. The temperature and moisture content histories of several droplets were predicted using the drying model (Eqs. Germany) and 0. The pH of the buffer was adjusted to 6. preheated to a temperature between 80 and 110 °C and a bulk air velocity of 0. with DE 4-7 (SigmaAldrich. The confidence interval of the parameters (p00. The major reason was that a larger sample volume could be easily obtained for performing the enzyme activity test. 3 and 4). The residence Materials and Methods Sample Preparation The enzyme.2 M Na2HPO4 (Sigma-Aldrich.2-μm Minisart sterile sieve (Sartorius Stedim Biotech SA. 2. The maltodextrin. the inactivation kinetics of β-galactosidase is not affected by its concentration (Yamamoto and Sano 1992). The single droplet drying experiments were performed using dry air (RH00. a uniform flow distribution could be achieved within the tunnel.01. and then the enzyme activity was measured. Laboratory-Scale Spray Drying Experiments The enzyme solution was dried with a Buchi B-190 spray dryer (Buchi Labortechnik AG. The feed for single droplet drying experiments was prepared by mixing 600 μL 2.15 mm. (2011a). The Netherlands) positioned on a platform from stainless steel slab.Food Bioprocess Technol The effect of the temperature is then still described using the transitional state theory as shown in Eq. Germany) to monitor the droplet geometry evolution during drying. and the stainless steel slab was 1 mm. This is allowed. stored overnight in the refrigerator. since at low enzyme concentration. 2). Germany).1 M citric acid (C6H8O7) solutions (Sigma-Aldrich.8% w/w maltodextrin solution. the thickness of the membrane was 0. A micro-dispenser was used to generate the droplet as described by Perdana et al.00±0. The inactivation kinetic constants were optimised to fit the measured residual enzyme activities after the drying for all droplets simultaneously.400 μL 25% maltodextrin solution. Germany). The buffer was prepared from 0. Switzerland). side view (S) and top view (T) . which was again derived from the Jacobian matrix of the solution (Richard et al. 2005).20 m/s. RS Component. The parameter optimisation was carried out using a nonlinear least square method solved using the Levenberg– Marquardt algorithm (Seber and Wild 2005).5% w/w enzyme solution and 2. All calculations were performed with MATLAB version 7. The feed for the laboratory-scale spray dryer was prepared by mixing 4 mL 2% w/w enzyme solution and 96 mL 20. United Kingdom) with probe diameter of 250 μm. The temperature and relative humidity of the air bulk air was monitored using SHT75 temperature and humidity sensor (Sensiron AG. The droplet was deposited on a polypropylene membrane (Akzo Nobel Faser Ag. A lower enzyme concentration is used during Fig. By guiding the drying air through a porous medium. the setup was equipped with a μEye 1480ME CCD camera with a lens with ×9 magnification ratio (Imaging Development Systems GMBH. rehydration and enzyme activity measurement. The solution was then filtered with a 0. Furthermore. β-galactosidase from Aspergillus oryzae (SigmaAldrich. Germany) was dissolved in a buffer solution. After drying. 23. The tunnel was insulated and heated with heating oil to ensure that the air temperature was constant.

The inactivation kinetics of β-galactosidase determined with independent experimental data was combined with the heat and mass transfer model to predict the residual enzyme activity after drying (Perdana et al. In modelling the inactivation of β-galactosidase.20 m/s. When the moisture content near the surface of the droplet decreases to less than 100 kg/m3. Subsequently. indicating that a very thin layer near the droplet surface is very dry. The absorbance of the samples incubated in ONPG solution was immediately measured after incubation at a wavelength 420 nm using a spectrophotometer (Beckman Coulter. 4. a 0. i. Fig. Afterwards. USA). 5.Food Bioprocess Technol time of the particle inside this small spray dryer can be as short as 1 s. The model predicts that the moisture content at the surface decreases faster than in the centre of the droplet and reaches a moisture content close to 0% at around 35 s. The model predictions and the experimental data are shown in Fig. 4 (bottom). the heat transfer into the droplet continues and causes an accumulation of heat. 3 Droplet geometry change during a single droplet drying experiment at an air temperature of 80 °C.80 mL buffer solution and analysed for its residual enzyme activity. 3. the sessile droplet is assumed to be a perfect sphere. The droplet shrinks uniformly before 30 s. observed as an increase in droplet temperature. Particle Size Measurement Approximately 1 g of the spray-dried sample was dried further in a heating chamber at 105 °C for 72 h to reduce the moisture content and to avoid particle agglomeration.e. and the outlet air temperature varied between 80 and 130 °C. Instead. The uneven shrinkage after 60 s suggests that a thin non-flowing layer is developed. and then starts to develop wrinkles. top). 4. This assumption reduces the complexity of the model and is acceptable since the initial sessile droplet has a very high contact angle (>130 °C) and also the final powder particle remains approximately spherical. an initial droplet height of 800 μm and an initial droplet moisture content of 80% w/w The single droplet drying method is applied to dry βgalactosidase suspended in a maltodextrin matrix. 5. 2011b). it collapses which leads to an irregular shape of the particle. the particle size distribution of the sample was measured using Mastersizer Scirocco 2000 (Malvern Instrument LTD. the predicted enzyme activity is compared to the drying results. This is allowed since most of the enzyme activity loss takes place at high moisture content where the presence of the added buffer stabilizes the pH.200-g powder sample was reconstituted to 3. the particle diameter after drying was between 2 and 25 μm. Measurement of the Residual Activity of β-galactosidase The residual activity of β-galactosidase was measured using an o-nitrophenyl-β-D-galactopyranoside (ONPG) assay (SigmaAldrich. Additionally. the enzyme activity loss is negligible at the time scale of drying applied. This is also in line with the visual observations. Germany) according to Perdana et al. an absolute air humidity of 0 g/kg dry air. The model also predicts that the droplet radius does not change any more after approximately 60 s of drying. At this point the water evaporation rate decreases together with the lower vapour pressure at the droplet surface (aw <1). as shown in Fig. As shown in Fig. Predicting the Residual Enzyme Activity After Drying Results and Discussion Droplet Geometry Evolution Snapshots of a deposited droplet during the drying process are shown in Fig. While the evaporation rate decreases. Both the predictions and the experimental data show that the enzyme activity does not decrease during the initial .. a bulk air velocity of 0. at lower moisture content. Inc. The predicted temperature and moisture profiles are shown in Fig. the moisture content gradient near the surface of the droplet is very steep. England). The inlet air temperature was 180 °C. which cannot accommodate a homogenous shrinkage (Walton and Mumford 1999). around 1 g of powder was taken and dried further in a heating chamber at 105 °C for 72 h to determine the moisture content of the powder after spray drying. The droplet temperature at the beginning (t<40 s) of the drying process is equal to the corresponding wet bulb temperature of the heating air (Fig. This is in line with the visual observations. 3. As shown in Fig. the pH change due to decreasing moisture content was neglected. Drying Model In the heat and mass transfer model. Furthermore. the droplet temperature starts to increase. the heat and mass transfer model using the independent kinetic inactivation model for βgalactosidase is in reasonable agreement with the experimental data. at t031 s. the occurrence of wrinkles indicating the presence of a thin dry layer. Subsequently. (2011b). Based on the equipment specification.

4 Top temperature and moisture content history of a droplet dried during a single droplet drying experiment. Bottom moisture content distribution inside the droplet at t032 s. the symbols show the border of the partition for each layer. The initial moisture content is 80% w/w and the initial droplet height is 800 μm drying period (<40 s) because the temperature of the droplet. The drying is carried out at an air temperature of 80 °C. averaged as the air temperature at half droplet height. which is near the (low) wet bulb temperature. Ta is the air temperature contacting the droplet. Middle droplet radius change during drying from the visual monitoring (symbol) and the model (line). which slowly declines to result in a final enzyme activity.20 m/s and an absolute air humidity of 0 kg/kg dry air. The subsequent rapid inactivation can be explained by an increase in particle temperature. Td is the droplet temperature. a rapid inactivation rate is observed. and xw is the moisture content. At .Food Bioprocess Technol Fig. a bulk air velocity of 0. Then.

0063∙103 4. The enzyme inactivation rate decreases because of the decreasing moisture content.0010∙105 NA 5.31∙103 ±0.w‡ (J mol−1) ΔH2.027∙105 8.60∙104 ±0.81∙105 6. The data from the heating experiments were obtained at many different temperature and moisture content this point the heat transfer is not compensated by sufficient water evaporation. the combination of high temperature and high moisture content in the centre of the droplet) compared to that in the larger droplet.57∙105 ±0.25∙103 3.80 68. The results also show that the initial droplet diameter determines the residual enzyme activity to a large extent.023∙105 5.00023∙105 4.e.065 68.86∙105 ±0.28∙105 ±0. 22. not fitted ΔΔS1‡ (J mol−1 K−1) ΔΔH1‡ (J mol−1) ΔS2.50 1.50 a The uncertainty of the parameter is provided within 95% confidence interval b The values of the parameters are according to Perdana et al. It can be observed that the parameters obtained from the drying experiments differ from the parameter values from the heating experiments (Table 2).097∙102 2. a smaller droplet size is preferred for spray drying. 5 Top the residual enzyme activity after drying of a single sessile droplet at air temperatures of 80 °C (empty circles). 23) NA NA 1. (2011b) c Reference value. This critical period is dictated by the drying history and involves temperatures and moisture content values that lead to rapid inactivation.00010∙103 7. especially near the surface of the droplet and finally also in the centre of the droplet.50 .08∙103 ±0.026 68. This is preferred as these experiments are much less labour intensive and can be scaled out to facilitate a high throughput approach.011 −49. After a while the enzyme inactivation rate decreases and the residual enzyme activity in the particle is obtained. Both the one-step and the two-step inactivation models were evaluated to describe the experimental drying data.49∙102 ±0.77∙105 ±0.w‡ (J mol−1 K−1) ΔH2. and 23) One-step inactivation model (Eq. Bottom the residual activity of βgalactosidase as a function of the radial coordinate (solid line) and of the volumetric coordinate (symbols) after drying of a droplet with an initial diameter of 800 μm at an air temperature of 95 °C for 300 s a smaller droplet. The predictions (solid line) are based on the inactivation kinetics of β-galactosidase from the constant heating experiments. the parameters for β-galactosidase inactivation directly obtained from single droplet drying experiments should be similar to the ones obtained from the separate heating experiments (Perdana et al. Most of the residual active enzyme is located near the surface of the particle as shown in Fig. the parameter values are optimised such that they specifically describe the inactivation during this critical period. 1996) or by applying a coating of a concentrated enzyme solution on pre-dried particles. 5 (bottom). The error bar shows the standard deviation of the data.65±8. which is also able to describe the inactivation during drying experiments well.40∙105 1.0010∙103 5.25±0. This can for example be achieved by drying the enzyme in a droplet that forms a hollow sphere upon drying (Etzel et al.74∙105 ±0.67±0. This implies that with respect to enzyme inactivation. Middle the effect of initial droplet size on residual enzyme activity for an initial droplet diameter of 800 μm (empty triangles) and 1. 95 °C (empty triangles) and 110 °C (empty diamonds) for an initial droplet diameter of 800 μm.16±0. Therefore.76∙105 ±0.Food Bioprocess Technol R Fig.57∙105 ±0.85±0. This can be explained by the fact that the inactivation during single droplet drying experiments occurs primarily during a very short critical period.400 μm (empty circles).17∙102 3. the critical region is shorter (i. 21.0056 33.37±0.75∙102 ±1. A strategy that might be followed to retain maximum enzyme activity is to minimize the presence of enzyme in the centre of the droplet. Although inactivation starts earlier in Table 2 The parameter values for the inactivation kinetics of β-galactosidase estimated from heating experiments and from single droplet drying experiments Estimated parametera Heating experimentsb Drying experiments Two-step inactivation model (Eqs. The increase in temperature has especially impact on the enzyme present in the centre of the droplet as the moisture content in the centre is still high as also reported for drying of other heat-sensitive products (Chen and Patel 2007. The results of the parameter optimisation are compared to the parameters from the earlier study (Table 2).28∙105 ±0. This explanation is further confirmed by implementing the one-step inactivation model.20∙105 2. Direct Extraction of Inactivation Kinetic Parameters from Drying Experiments The parameters describing the specific inactivation kinetics of β-galactosidase in maltodextrin may be extracted directly from the single droplet drying results. If the method is reliable.88∙103 ±0.063∙105 1.00038 8. Figure 5 indicates that the residual activity is higher for smaller droplets. 2011b). Langrish 2009). 87 °C (empty squares).00094∙104 1.92∙103 ±0.s‡ (J mol−1) m p Tint (°C) Tref (°C)c 1. It can be concluded that the different moisture content history at different locations in the droplet determines the large differences in residual enzyme activity across the droplet radius.

it may be concluded that the onestep inactivation model is more favourable than the two-step inactivation model to describe the effect of Fig. it can be concluded that although the values for the parameters in the model for inactivation kinetics obtained from the drying experiments may be less precise. Therefore. Therefore. the values are based on a much wider data set. 6 (bottom). However. Therefore. it is believed that the uncertainties involved are not very large and that the kinetic parameters of the single droplet experiments remain valid. This conclusion is supported by laboratory spray drying experiments. Therefore.400 μm (empty circles). the effect of the initial droplet diameter on the residual β-galactosidase activity after drying is shown. but having less data in the specific range that are relevant for spray drying. The βgalactosidase inactivation is described by the two-step inactivation model (left) and the one-step inactivation model (right) . The figure shows that the model accurately predicts the experimental data. Table 2 shows that the values of the confidence intervals of the parameters are smaller for the onestep inactivation compared to the two-step inactivation model. 95 °C (empty triangles) and 110 °C (empty diamonds) for an initial droplet diameter of 800 μm. Bottom the residual enzyme activity predicted using the inactivation kinetics extracted from drying data at various heating air temperatures compared to the drying results from different initial droplet diameter: 500 μm (empty squares) and 1. it was concluded that the heat and mass transfer model could predict the droplet shrinkage and the residual enzyme activity using the kinetic parameters from the heating experiments reasonably accurate. 87 °C (empty squares). any uncertainties in the heat and mass transfer model are neglected. By doing so. the inactivation kinetics are accurate enough to predict the effect of drying on the inactivation of the enzyme. Another reason for the difference between the kinetic parameters is that the parameter optimisation procedure forces the model to describe the residual enzyme activity after the drying. This is because the critical drying period dictates the end product properties.Food Bioprocess Technol combinations and could be varied independently. 6 Top the fitted residual enzyme activity obtained from single droplet drying experiments at varying air temperatures of 80 °C (empty circles). from earlier observations. In Fig.

The results are shown in Fig. The model can predict the drying results reasonably accurate. These results indicate that at an outlet air temperature below 100 °C. The volume-weighted mean diameter (d4. but at temperatures between 100 and 120 °C. 7.8 μm (initial moisture content 80% w/w. 7 Left the residual activity of β-galactosidase after spray drying in a laboratory-scale spray dryer. The model.3) of the spraydried particle is 6. This may be due to the lower initial droplet size in reality due to nonideal shrinkage leading to more rapid drying and thus lower inactivation (see also Fig.94 μm. The parameters that were obtained in this way were used to predict other experimental results both from other single droplet drying experiments and from a laboratory-scale spray dryer. while with an outlet air temperature higher than 100 °C. the one-step inactivation model may be sufficient to translate the measured inactivation from single droplet experiments into useful inactivation kinetics. Conclusions The models were also applied to predict the residual enzyme activity after drying on a laboratory-scale spray dryer. The symbols represent the experimental results at an inlet air temperature of 180 °C and varied outlet air temperature. The new inactivation kinetics provides reasonably accurate prediction on the residual enzyme activity for both procedures. 7 shows that there is only a small difference between the two inactivation models at temperatures larger than 100 °C. This may lead to an overestimation of the inactivation kinetic constant at high temperatures. Langrish 2009). From practical point of view. 3). Fig. It was also shown that the inactivation kinetics of βgalactosidase can be directly extracted from the single droplets drying experiments rather than using the kinetics from separate heating experiments. By assuming that the volumetric droplet shrinkage is equal to the amount of water removed. To describe the physical and chemical changes during drying. small-scale experimental setup to mimic spray drying was presented which involves the drying of single droplets deposited on a hydrophobic flat plate. a model based on heat and mass transfer is presented. the initial droplet diameter was estimated at 11. Laboratory-Scale Spray Drying tion kinetic constant of the one-step inactivation model increases more quickly with temperature. It is found that the model can provide a reasonably accurate prediction on the residual enzyme activity. Furthermore. Fig. Right the particle size distribution of the powder obtained from the laboratory-scale spray dryer . fed into the drying model to predict the residual enzyme activity. the setup is used to evaluate the inactivation of β-galactosidase during drying. This approach ensures rapid and inexpensive trials while maintaining key parameters similar to the process condition on spray drying. the enzyme is increasingly inactivated with higher temperature as also observed by Yamamoto and Sano (1992).Food Bioprocess Technol drying on enzyme activity loss. the enzyme is hardly inactivated upon drying. The lines represent the predicted enzyme activity using the inactivation kinetics obtained from the single droplet drying experiments and from the constant heating experiments. This is probably because the inactivaA newly developed. the predicted value of the residual activity of β-galactosidase is slightly lower than the experimental results. combined with the inactivation kinetics of β-galactosidase from separate heating experiments. was used to predict the loss in enzyme activity after drying. In this study.

314) Reynolds number Droplet radius Activation entropy Activation entropy difference between unfolding and refolding reaction Schmidt number Time Temperature Velocity Volume Distance in Cartesian coordinate Mass fraction of water Moisture content Monolayer moisture content in GAB sorption model Distance in Cartesian coordinate Distance in Cartesian coordinate J mol−1 K−1 – m J mol−1 K−1 J mol−1 K−1 – s °C m s−1 m3 m kg kg−1 kg kg−1 dry matter kg kg−1 dry matter m m m m m J s−1 m−1 K−1 Pa s kg m−3 m2 s−1 g – h h ΔHlv ΔH‡ ΔΔH‡ J kB kobs k2 J s−1 Js −1 m K −2 −1 J kg−1 J mol−1 J mol−1 kg s−1 J K−1 s−1 s−1 m s−1 – s s−1 m kg – kc Kg K1 L m mu Mw n p pi kg mol−1 – – – Z Thickness Greek symbols δ Hydrodynamic boundary layer thickness δt Thermal boundary layer thickness l Thermal conductivity μ Dynamic viscosity ρ Density υ Kinematic viscosity Subscript a Air avg Average cond Conduction d Droplet int Intercept m Very dry matrix obs Observed p Flat plate for droplet deposition ref Reference s Solute sat Saturated w Water 0 Initial ∞ Bulk air P Pr Q r Pa – J s−1 m Acknowledgement The authors thank Advanced Chemical Technologies for Sustainability (ACTS)–Netherlands Organization for Scientific Research (NWO) for financial aid through Process on a Chip (PoaC) program.380×10−23) Observed inactivation kinetic constant of β-galactosidase Irreversible complete denaturation rate constant of β-galactosidase Mass transfer coefficient Constant in GAB sorption model Unfolding equilibrium constant of β-galactosidase Length Mass Parameter to describe the effect of moisture content on unfolding equilibrium of β-galactosidase Molecular weight of water (18∙10−3) Number of partition Uncertainty of the parameters Parameter to describe the effect of moisture content on irreversible complete denaturation rate constant of β-galactosidase Pressure Prandtl number Heat Distance in radial coordinate m2 – J kg−1 K−1 kg m–3 – m m2 s−1 J mol−1 – Nomenclature R Re Rd ΔS‡ ΔΔS‡ Sc t T u V x xw Xw Xwm y z Ideal gas constant (8. Open Access This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any .626×10−34) Convective heat transfer coefficient Enthalpy of evaporation Activation enthalpy Activation enthalpy difference between unfolding and refolding reaction Evaporation rate Boltzmann’s constant (1.Food Bioprocess Technol Nomenclature A aw cp C Cg d D Ea f Area Water activity Heat capacity Concentration Constant in GAB sorption model Diameter (2Rd) Diffusion coefficient Arrhenius-type diffusion activation energy Parameter to describe the effect of moisture content on conformational stability of the enzyme Parameter to describe the effect of moisture content on the irreversible inactivation (second step) kinetic constant of the enzyme Planck’s constant (6.

(1993). K.X. Micro-organism inactivation during drying of small droplets or thin-layer slabs—A critical review of existing kinetics models and an appraisal of the drying rate dependent model. The morphology of spraydried particles: The effect of process variables upon the morphology of spray-dried particles. Fang. (2007). CRC handbook of tables for applied engineering science. The 52nd annual meeting of the American society of heating and ventilating engineers. Nanoparticles by spray drying using innovative new technology: The Büchi Nano Spray Dryer B-90. 83–102. Particle formation in spray drying. Perdana. 2274–2283. Arpagaus. 185–191. 22(5). 48(3). Anton. & Adamopoulos. The mathematics of diffusion. Chemical Engineering Journal. Heat and Mass Transfer. Schiffter. J. K. 442–460. Li. Food Chemistry (in press).. Thirumaleshwar. Trends in Food Science & Technology. E. B.. In: Computational fluid dynamics in food processing. K. K. Cleveland. & Gratch. 17–34. Straatsma. & Bruin.... 3914–3921. Washington: Taylor & Francis.. A. Walton. In: Handbook of Industrial Drying. K. Evaporation from drops. Biotechnology and Bioengineering. Quirijns. F. L. S. Boxtel. Model based analysis of the drying of a single solution droplet in an ultrasonic levitator. & Jong.. (1952). (2007). 86(1–2). S. E. B.. K. M. 1195–1236. Foss. Schutyser. Verdurmen. & Sommer. P. R.. Farid. Gianfrancesco.. J. M. Goula. USA. 16(6)... New York: Wiley. Heat and mass transfer of single droplet/wet particle drying. Chuchottaworn. (2006). UK: Oxford University Press. P. Schlichting. M. G. J. T. B. 215–224. Influence of spray drying conditions on residue accumulation-simulation using CFD. M. Single-droplet evaporation kinetics and particle formation in an acoustic levitator. H. Eng. 22(5). M. 14 (2). Boca Raton: CRC Press. 2985–2993. Journal of Chemical Engineering Japan.. & Truong.. (1996).. A. J... New Delhi: Dorling Kindersley. CRC Press. 1805–1814. M. 61(8).. Single-droplet experimentation on spray drying: Evaporation of a sessile droplet. 66(17). 273–280. Bhandari. Enzyme inactivation kinetics: Coupled effects of temperature and moisture content. (1973). T. (2006). Huang. (1991). K. S. Trends in Food Science & Technology.. Belleteix. (2009). H. & Bergenståhl. Journal of Controlled Release. 188(2). New York: CRC Press. Kerkhof. J. G. T. Millqvist-Fureby. 38(7). Maltodextrin/pectin microparticles by spray drying as carrier for nutraceutical extracts. 58(13). (2011). (2010). Loon. B. A thin film dryer approach for the determination of water diffusion coefficients in viscous products. 371–375. B. Journal of Aerosol Science. I. B. 728–746. I. Crank. Bolz. S. 323–351. F. A. M... & Marshall. S. Theoretical study of the inactivation of phosphatase during spray drying of skim-milk. A. Microencapsulation by spray drying. New York: Wiley. (2000). V. In: International Geophysics.. Chen. C. (1998).. Sansone. & Vandamme. R. C. 34(7). 651–660. G. Fox. (2009). Y. (2011b). Kakac.. C. X. 173–l 80 {Part II). Filková. Peighambardoust.. 96(9). R. K. A. Laminar forced convection conjugate heat transfer over a flat plate.. K. Mencherini. A. G. Halverson.. Journal of Food Engineering. 3(3). B. 141–146 (Part I). Liou.Food Bioprocess Technol noncommercial use. B. Fundamentals of heat and mass transfer.. J. (2000). Suen.). (2002). GAB parameters and isosteric heat of sorption. Journal of Food Engineering. & Clifford. A. References Adhikari. (2005). 510–523. X. 3 (6).. K.. & Bruin. N. E. Chemical Engineering Research and Design. 12–23. W.. & Dumoulin. Review of vector calculus. 35(5). M. Howes. 85(11). G. Jørgensen. (2011). & Boom. S. 24(3). Incropera. M. J. S. 147(2). (1990). Picerno. H. Chemical Engineering Science. R. Fox. Chemical Engineering and Technology. 34(5).. & Hesari. H. L. R. London. Kiil. 3rd Ed. 400–408. 243–253. S. New York: 95–122. pp. I. Chemical Engineering Science. and reproduction in any medium. Langrish. & De Witt. (1946). A. (1999). C. Ohio. H. D. T... Ali Al Zaitone. & Borde. Spraydrying of trypsin—surface characterisation and activity preservation. 2701–2709. Gunsing. 105(3). R. Y. J. 48(4). International Journal of Food Properties. R. Appendix C. & Boom. R. M. provided the original author(s) and source are credited. Flick. Computational methods for nonlinear least squares. J. R. Richard. Chemical Engineering Science. & Lee. 619–660. (2008).. I. & Lee. 249–286. A. Chemical Engineering Science. D. Mezhericher. Prog. & van't Riet. & Gersten. M. J. Sorption isotherms. Ré. W.. M. Levy. A. 21(10). Aquino. H.. Golshan Tafti. A. (2007). B. P. Turchiuli. & Mumford. 1151–1158. P.. (2005). Goff. J. D. Z. M. Berlin: Spinger. . A. & Yener. S. G.. T. & Mujumdar. L... & Tropea. (1979). Journal of Food Engineering. A. Degradation of vitamin C in spray dryers and temperature and moisture content profiles in these dryers. & Straten. Part 1: Evaporation of water microdroplets assessed using boundary-layer and acoustic levitation theories. 468–476. C. 82(1). (2000). & Patel. Etzel. 7–13. Schiffter.. Journal of the Science of Food and Agriculture. (1984). Mosaad. 27(1). M.. Räderer. Chem. M... J. Evaporation of pure liquid droplets: Comparison of droplet evaporation in an acoustic field versus glass-filament. P. W.. (1999). d'Amore. pp. C. K. (2010).. Part II. 1107–1128. d. (1999). & Tuve. & Asano. Experimental study of evaporation of a volatile pendant drop under high mass flux conditions. (2011a). E. M. Meerdink. Industrial spray drying systems. 2(4).. & Bhandari. Boundary-layer theory (8th ed. D. M. D. (1982).. Malmsten. 53–68. Enzyme inactivation in a droplet forming a bubble during drying. & Wild.. 946–955. Heat conduction. S. B. Inactivation of thermostable [alpha]-amylase during drying.. M. P. & Budijono. Journal of Pharmaceutical Sciences. (2003). 304–310... Low-pressure properties of water from 160 to 212 F. M. Experimental studies and kinetics of single drop drying and their relevance in drying of sugar-rich foods: A review. Fujinami. New York: Academic Press.. I.. F.. Perdana. 17.. 533–552. Wijlhuizen. F. Seber. M. pp. (2004). & Lauro. 1–10. C. (2005).. Food and Bioprocess Technology. P. D. 77(5). (2007). A.. Models for drying kinetics based on drying curves of slabs. P. F. distribution. W.. 63(1). Schutyser. Food and Bioprocess Technology. Coumans.. A new approach to modelling of single droplet drying. In: Nonlinear regression. Encapsulation of polyphenols—A review. Chemical Engineering and Processing. 39(1). Vehring. Ranz. Drying Technology: An International Journal. Enzyme degradation during drying. Luyben. & Lechuga-Ballesteros. Fundamentals of heat and mass transfer (2nd ed. M. J.. Andersen. J. Besson. International Journal of Pharmaceutics. CFD modeling and simulation of maltodextrin solutions spray drying to control stickiness. E. Verschueren.).. E. Aster. A. A. A. (2010). M.. in Transactions of the American Society of heating and ventilating engineers.. R. (1985). Sloth. (2011). Drying Technology. Journal of Food Engineering. Jensen. CFD simulation of spray drying of food products. Application of spray drying for preservation of lactic acid starter cultures: A review. C.. G. Chemical Engineering Science.

W. Yoshioka. Industrial spray drying systems.. Lin. (1994). (1992). D.. Y. S. & Chen... & Sano. S. 177–183. Xin Huang. D. I. S. Y. 1386–1392. & Kojima. Drying of enzymes: enzyme retention during drying of a single droplet. Filkova. AICHE Journal. S. Is stability prediction possible for protein drugs? Denaturation kinetics of βgalactosidase in solution. (2011). A. 57(6). Yamamoto. 3rd ed.Food Bioprocess Technol Wu. & Mujumdar. Pharmaceutical Research. Aso. X. Izutsu. Monodisperse droplet formation through a continuous jet break-up using glass nozzles operated with piezoelectric pulsation. New York: CRC Press.. K-i. L. Chemical Engineering Science. .. 11(12). (2006). In: Handbook of industrial drying. X. 1721–1725. 47(1).

Sign up to vote on this title
UsefulNot useful