Vous êtes sur la page 1sur 15

Drying Technology, 24: 917930, 2006

Copyright # 2006 Taylor & Francis Group, LLC


ISSN: 0737-3937 print/1532-2300 online
DOI: 10.1080/07373930600775979

Models for Sorption Isotherms for Foods: A Review


Santanu Basu,1 U. S. Shivhare,1 and A. S. Mujumdar2
1

Department of Chemical Engineering & Technology, Panjab University, Chandigarh, India


Department of Mechanical Engineering, National University of Singapore, Singapore

This article presents basic concepts related to the thermodynamics of sorption of water and measurement of sorption isotherm
for food materials. A comprehensive review of the widely used sorption models is presented. Various statistical techniques used to
ascertain the effectiveness of a model to describe the sorption data
are discussed. It is anticipated that this article will provide useful
information to researchers pursuing work on sorption behavior of
food materials as well as modeling of drying processes.
Keywords Heat of sorption; Isosteric heat; Model; Sorption isotherm; Statistical analysis

INTRODUCTION
The state of water plays a crucial role in food preservation. The quality of preserved foods depends upon the
moisture content, moisture migration, or moisture uptake
by the food material during storage. Extent of sorption
of water by or desorption from a food product depends
on vapor pressure of water present in the food sample
and that in the surroundings. Moisture content at which
vapor pressure of water present in the food equals that of
the surroundings is referred to as equilibrium moisture content (EMC).[1] Relationship between EMC and corresponding relative humidity at constant temperature yields
the so-called moisture sorption isotherm. For a given
material the EMC increases with relative humidity but
decreases with increase in temperature. The phenomenon
where the EMC during the adsorption and that during
the desorption process is different is called hysterisis.
Water activity is commonly used to characterize the food
quality and is defined as
aw

P
ERH

P0
100

where P is vapor pressure of water in the food material at


any given temperature (Pa), P0 is the vapor pressure of
pure water at that temperature (Pa), aw is the water activity
Correspondence: U. S. Shivhare, Department of Chemical
Engineering & Technology, Panjab University, Chandigarh,
160014 India; E-mail: usshiv@yahoo.com

(dimensionless), and ERH is the equilibrium relative


humidity (%).
Five types of isotherms were described by Brunauer
et al.[2] (Fig. 1). Type 1 is the well-known Langmuir isotherm, obtained assuming monomolecular adsorption of
gas by the porous solids in a finite volume of voids. Type 2
is the sigmoid isotherm obtained for soluble products,
which exhibits asymptotic trend as water activity
approaches 1. Type 3, known as the Flory-Higgins isotherm, accounts for a solvent or plasticizer such as glycerol
above the glass transition temperature. Type 4 isotherm
describes adsorption by a swellable hydrophilic solid until
a maximum of hydration sites are reached. Type 5 is the
BET[3] multilayer adsorption isotherm, observed for
adsorption of water vapor on charcoal; it is related to types
2 and 3 isotherms. The two isotherms most commonly
found in food products are types 2 and 4.[4]
Water adsorption by foods is a process wherein water
molecules progressively and reversibly combine with the
food solids via chemisorption, physical adsorption, and
multilayer condensation.[5] An isotherm can typically be
divided into three regions as shown in Fig. 2. The water
in region A represents strongly bound water with enthalpy
of vaporization considerably higher than that of pure
water. The bound water includes structural water
(H-bonded water) and monolayer water,[6] which is sorbed
by hydrophilic and polar groups of the food components
(polysaccharides, proteins, etc.). Bound water is unfreezable and is not available for chemical reactions or as a plasticizer. In region B, water molecules bind less firmly than in
the first zone. The vaporization enthalpy is slightly higher
than that for pure water. This class of constituent water
can be looked upon as the continuous transition of the
bound to the free water. Properties of water in region
C are close to those of the free water that is held in
voids, capillaries, crevices, and loosely binds to the food
materials.[1,7]
Measurement and modeling of sorption isotherms of
food materials has attracted numerous researchers because
of their value in industrial practice. Comprehensive reviews
on sorption behavior of foods have been published.[813]

917

918

BASU, SHIVHARE, AND MUJUMDAR

More than 1000 scientific papers have been published


covering a very wide range of products. This article gives
a comprehensive review of the main theories of the sorption phenomenon in foods and different mathematical
models commonly used to describe the sorption behavior.
The article elaborately discusses various statistical analysis
tools that should be considered for drawing conclusions
from any set of experimental and predicted values. Note
that in drying it is the desorption isotherm that is of greater
interest than the adsorption isotherm. The word sorption covers both adsorption and desorption.
THERMODYNAMICS OF SORPTION
When water is removed from a food product, heat is
absorbed because water has to be removed against a water
activity gradient or against increasing osmotic pressure.
The isosteric heat of sorption, also called differential
enthalpy, is an indicator of the state of water held by the
solid material. Net isosteric heat of sorption (DHS ) is the
difference of total heat of sorption (DHd ) in the food and
the heat of vaporization of water (DHvap ) associated with
the sorption process and can be computed from experimental data using Clausius-Clapeyron equation


dln aw
DHS
2

d1=T M
R

FIG. 1. Five types of isotherms.[4]

FIG. 2. A typical sorption isotherm showing the phenomenon of


hysteresis.[1]

where the isosteric heat of sorption is associated with


sorbed molecules at a particular moisture content (M).
Integrating Eq. (2) using the associated boundary conditions yields
 
Z aw2
Z
DHS T2 1
dln aw 
d
T
R T1
aw1
3
 


aw2
DHS 1
1
ln


aw1
R T1 T2
where aw1 and aw2 represent water activities at temperatures T1 and T2, respectively, and R is the universal gas
law constant (8.314 kJ=molK)
Several assumptions are made implicitly in applying the
Clausius-Clapeyron equation. First, the heat of vaporization of pure water (DHvap ) and the excess heat of sorption
DHS are assumed not to change with temperature. Secondly, the equation applies only when the moisture content
of the system remains constant with respect to time.[14]
From the isotherms determined at least at 10C apart, aw
at any other temperature should be predictable from
Eq. (3) if the assumptions are correct. Plotting the experimental sorption isotherm in the form ln(aw) vs. 1=T for a
specific moisture content, DHS is determined from the
slope DHs =R.
An empirical relationship between the net isosteric heat
of sorption (DHS ) and the moisture content (M, dry basis)
is expressed as[14]

MODELS FOR SORPTION ISOTHERMS: A REVIEW

DHS DH0 expM=MC

where DH0 is the net isosteric heat of sorption when moisture content (M) 0. The constant MC is a characteristic
moisture content (kg water per kg dry matter) of a food
material at which the net isosteric heat of sorption (DHS )
has been reduced by 63%. Due to the exponential decay,
the net isosteric heat of sorption becomes very small at high
moisture content; e.g., at M 3 MC, DHS is less than 5%
of DH0 .
The net heat of sorption of water Q (kJ per kg dry matter) from dryness (M 0) to moisture content M is
expressed as:[14]
Z M
Q
DHS dM
5
0

Substituting DHS from Eq. (4) into Eq. (5) and integrating
yields
Q DH0 MC 1  expbM=MC c

For very high moisture content (M ! 1), Eq. (6) gives the
total net isosteric heat in terms of heat of sorption of water
(QT).
QT DH0 MC

The net heat of sorption of water for a change of moisture


content from M1 to M2 (Q12) is expressed as
Q12 DH0 MC bexpM1 =MC  expM2 =MC c

Equation (8) can be used to estimate the energy required for


drying a food material from moisture content M1 to M2.
In drying, M1 > M2, and the net heat of desorption (Q12)
is negative, because an extra amount of energy is required,
in addition to heat of vaporization of water (DHvap ), to
remove the adsorbed water from the food product.
The relationship between the isosteric heat (DHd ) and
differential entropy (DSd ) of sorption is expressed as:[15]
lnaw M DHd =RT  DSd =R

Applying Eq. (9) at different moisture contents, the


dependence of DHd and DSd on moisture may be
determined.[16,17]
The compensation theory proposes a linear relationship
between DHd and DSd [15,18]
DHd Tb DSd a

10

The isokinetic temperature (Tb ) and constant a are


computed using linear regression of Eq. (10). Parameter a
may be neglected due to its negligible contribution to the
enthalpy change. The compensation theory may be further
used to evaluate the effect of temperature on the sorption
behavior by applying Eq. (11).[16]
 lnaw DHd =R1=T  1=Tb

11

919

MEASUREMENT OF SORPTION ISOTHERMS


Many methods are available for determination of water
sorption isotherm. These methods may be classified into
three categories: (1) gravimetric, (2) manometric, and (3)
hygrometric. The gravimetric method involves the measurement of mass changes that can be measured both continuously and discontinuously in dynamic and static systems.
Manometric methods involve sensitive manometers to
measure vapor pressure of water in equilibrium with a food
material of given moisture content. Hygrometric methods
measure the equilibrium relative humidity of air in contact
with a food material at given moisture content. Dew point
hygrometers detect the condensation of cooling water
vapor. Electronic hygrometers measure the change of
conductance or capacitance of hygrosensors.[19]
The most common technique, for which a recommended
procedure has been defined in the European project COST
90, uses thermostatted jars filled at the bottom with supersaturated salt solutions to maintain the desired air relative
humidity.[20] However, this method encounters some problem at high humidity ranges due to (a) excessive equilibration times and (b) its inability to produce and control
high relative humidities.
Baucour and Daudin[20] developed a rapid but accurate
method to measure moisture sorption isotherms of solid
foods in the water activity range 0.91. This method avoids
the drawbacks of saturated salt solution method by blowing calibrated air along thin slices of material at a high
velocity to impose an intensive water vapor exchange
between air and samples.
MATHEMATICAL MODELS OF ISOTHERMS
Numerous attempts have been made to describe the
sorption isotherms mathematically. While some models
have been derived theoretically based on thermodynamic
concepts, others are an extended or modified form of these
models. Some of the widely used mathematical models are
presented in the following paragraphs.
LANGMUIR EQUATION
On the basis of monomolecular layers with identical,
independent sorption sites, Langmuir[21] proposed the following physical adsorption model:


1
1
1

12
aw

M M0
CM0
where M0 is monolayer sorbate constant and C is a constant.
The relation described by Eq. (12) yields the type I isotherm.
BRUNAUER-EMMETT-TELLER (BET) EQUATION
The BET isotherm equation[3] is one of the most widely
used models and gives good fit for a variety of foods over
the region 0.05 < aw < 0.45.[22] It provides an estimate of

920

BASU, SHIVHARE, AND MUJUMDAR

monolayer value of moisture adsorbed on the surface. The


BET equation is expressed as
M
Caw

M0 1  aw 1 C  1aw 

13

or
aw
1
C1

aw
1  aw M M0 C M0 C

14

where M is the equilibrium moisture content (kg water=


kg dry matter), M0 is the monolayer moisture content on
the internal surface (kg water=kg dry matter), and C is a
dimensionless parameter related to heat of sorption of
monolayer region.
The theory behind the development of the BET equation
has been questioned due to the assumptions that (a) the
rate of condensation on the first layer is equal to the rate
of evaporation from the second layer, (b) binding energy
of all of the adsorbate on the first layer is same, and (c)
binding energy of the other layers is equal to that of pure
adsorbate. The assumptions of uniform adsorbent surface
and absence of lateral interactions between adsorbed
molecules are incorrect in view of the heterogeneous food
surface interactions.[9]
MODIFIED OSWIN EQUATION
The Oswin[23] equation is a mathematical series expansion for a sigmoid-shaped curve and is represented as

MK

aw
1  aw

N
15

Here, K and N are constants.


The parameter K was found to be a linear function of
temperature.[24] The Oswin equation may be modified as

N
aw
M A BT
16
1  aw
or

"
aw

M
A BT

#1

C
1

17

with C 1=N.
MODIFIED HALSEY EQUATION
Halsey[25] developed an equation to describe condensation of multilayers, assuming that the potential energy
of a molecule varies inversely as the Cth power of its
distance from the surface. The equation is


AM C
aw exp 
18
RT

Because the use of the term RT does not eliminate the temperature dependence of constants A and C, Iglesias and
Chirife[26] simplified it to the form
aw expA0 M C

19

where A0 is a new constant.


Iglesias and Chirife[27] analyzed the parameter A of the
Halsey equation and found that it could be related to temperature by an empirical exponential function. A new
modified Halsey equation was therefore proposed:


 expA BT
aw exp
20
MC
where A, B, and C are constants.
Iglesias and Chirife[28] reported that the Halsey equation
described 220 experimental sorption isotherms of 69 different foods in the range of 0.1 < aw < 0.8.
MODIFIED HENDERSON EQUATION
Henderson[29] presented an equation of the form
aw 1  expbATM Cc

21

Thompson et al.[30] modified the Henderson equation by


adding another constant to the temperature term. The
modified Henderson equation becomes
aw 1  expbAT BM Cc

22

where A, B, and C are constants.


Hendersons equation has been applied to many
foods,[3134] but compared to the Halsey equation, its applicability has been rather limited.
CHUNG-PFOST EQUATION
Chung and Pfost[35,36] developed a model based on the
assumption that the change in free energy for sorption is
related to the moisture content


A
expBM
23
aw exp 
RT
Later, in order to obtain a better fit, Pfost et al.[37] added a
new parameter to the temperature term and combined the
R parameter in constant term. The modified Chung-Pfost
equation is expressed as:


A
expBM
24
aw exp
T C
FERRO FONTAN EQUATION
The Ferro Fontan equation[38] is represented as
 
c
ln
aM r
aw

25

921

MODELS FOR SORPTION ISOTHERMS: A REVIEW

where c is the parameter that accounts for the structure of


sorbed water, a is a constant, and r is a constant that
involves net isosteric heat with moisture content adjustable
parameters. Iglesias and Chirife[39] compiled sorption data
for 156 food items and documented that Ferro-Fontan
equation is an accurate tool for the mathematical description of food isotherms. The Ferro-Fontan equation accurately represents the sorption isotherm in the range of
water activity 0.10.9 with only 24% error in the predicted moisture content.
GAB EQUATION
van den Berg[40] refined the Langmuir and BET theories
and proposed a new equation with three parameters having
physical meanings
M
CKaw

M0 1  Kaw 1  Kaw CKaw

26

M
C  1Kaw
Kaw

M0 1  Kaw CKaw 1  Kaw

27

where C is a dimensionless GAB parameter related to heat


of sorption of monolayer region and K is a dimensionless
GAB parameter related to heat of sorption of multilayer
region.
Equation (27) is divided into two additive terms, the first
one describes the classical mono-molecular layer
expression in Langmuirs adsorption isotherms and the
second term describes the multilayer adsorption corresponding to Raoults law.[40]
Parameters K and C can be expressed by Arrhenius type
equations:


DHc
0
C C exp
28
RT


DHk
0
K K exp
29
RT
where DHc DHm  DHq ; DHk DHl  DHq (kJ mol 1);
DHl is the heat of condensation of pure water
(kJ mol 1); DHm is the total heat of sorption of mono layer
(kJ mol 1); DHq is the total heat of sorption of multilayer
covering the monolayer (kJ mol 1); and C 0 , K 0 are the
constants of entropic character.
The GAB model underestimates the water content values
at high water activities (aw > 0.93). The discrepancy underlines two facts: (a) this type of model is unsuitable for high
humidity range, and (b) the saturated salt solution method
does not afford sufficient information to get a complete sorption curve. The GAB model was refined for higher water
activities by Timmerman and Chirife;[41] and Viollaz and
Rovedo.[42] They modified the GAB model by inserting a
new parameter to add a third sorption stage without changing
the values of the usual parameters defined in the GAB model.

ANALYSIS OF GAB ISOTHERM


Equation (27) can be rewritten in one-parameter form:
w

1
1

1  X 1  1  CX

30

where w M=M0 and X Kaw are new coordinates.


Equation (30) can be described for five conditions: (1) for
X 0, (2) for w 1 for the so-called mono-molecular
layer, (3) for X K (i.e., for aw 1), (4) at the point of
inflection w  X, and (5) potential singular points. The
characteristic properties of the isotherm at these points give
the limits on the parameters K and C.[43]
1. For X 0; Eq. (30) gives the value w 0 identically.
The shape of the sigmoid sorption isotherm may be
obtained by differentiating Eq. (30) corresponding to
dw
dw
the condition
 0:
K
daw
dX
dw
1
1C


2
dX 1  X
1  1  C X 2

31

Substituting X 0 into Eq. (31), we obtain dw=dX C,


so the mentioned condition leads to the final form
KC  0. It infers that both parameters K and C have
to be of the same sign.
2. The existence of a mono-molecular layer for w 1 leads
to the quadratic equation Xm2 1  C  2Xm 1 0
where Xm is the value of variable X in the monomolecular state. The equation has the following solution
for 0 < X < 1:
Xm

1
p
1 C

for C  0:

32

In combination with condition (1) it means that both C


and K have to be non-negative.
3. Substituting X K into Eq. (31) and rearranging, the
following equation is obtained:
w

KC
1  K1  K1  C

for X K

33

that has to be greater than 0. This equation shows that


both parameters K and C have to be greater than zero.
Moreover, the following condition
1  K1  K1  C > 0
has to be fulfilled. This inequality has two solutions:
K 1
; i:e:; C > 0
K
K 1
b=K > 1; 0 < C <
K

a=0 < K;

C>

34a
34b

922

BASU, SHIVHARE, AND MUJUMDAR

4. Points of inflection: the point of inflection of the sorption isotherm is given by the zero point of the second
derivative of w  X plot. The following equation is
obtained for the X coordinate (Xi) of such a point:
Xi3

3Xi
C2

0
C  1 C  1 2

35

where k1, k2, n1, and n2 are constants (n1 < 1) and n2 > 1).
The model expressed by Eq. (43) has no monolayer incorporated in it. However, plots of aw =M1  aw  vs. aw used
to determine the monolayer value with the BET model are
still expected to be practically linear over a water activity
range of up to about 0.4. This is because at this range of
aw Eq. (43) can be approximated by

which has the positive solution

M k1 anw1

Z1
Xi
for C  2
36
Z2

p
where Z 3 C  1. The relative moisture content corresponding to the point of inflection wr is given by inserting Xi from Eq. (36) into Eq. (30)
wr

Z2  1
A

37

A Z2  Z 1

38

5. Singular points: No singular points can appear in


Eq. (30) for 0 < X < 1. This condition is fulfilled
for all K  1 but it is not fulfilled for K > 1.[43]
PELEG MODEL
Peleg[44] developed a two-parameter model to describe
the sorption curves
Mt Mi

t
k1 k2 t

39

where M(t) moisture content after time t (% dry basis),


Mi initial moisture content (% dry basis) and k1, k2 are
constants. According to this model, equilibrium moisture
content ME when t ! 1 is given by
M E Mi

1
k2

40

Similarly, the instantaneous sorption rate dM(t)=dt is given


by
dMt
k1

dt
k1 k2 2

41

and the initial rate (at t 0) by 1=k1.


Equation (39) can be easily transformed to a linear
relationship
t
k1 k2 t
Mt  Mi 
Peleg

aw
aw

M1  aw  k1 anw1 1  aw 

42

also developed a four-parameter model


M k1 anw1 k2 anw2

43

44
45



The plot of aw = anw1 1  aw vs. aw (k1 1) is linear in this
water activity range and for a range of n1 values. The intercept is positive when n1 is greater than 0.3.
aw
aw =k1
n1
M1  aw  aw 1 k1 =k2 anw1 n2 1  aw 

where

[45]

and

n1 < 1

46

the added term 1 k1 =k2 anw2 n1  in the denominator


tends to lower the magnitude of aw =M1  aw  as the
magnitude of aw increases.
TIMMERMANN GAB MODEL
Compared to the classical BET isotherm, the GAB
isotherm contains a third constant, K, which measures the
differences between the standard chemical potential of the
molecules in the second sorption stage and in the pure
liquid. Some researchers[46,47] noted that at very high water
activities some systems showed a sorption stage larger than
the predicted by the GAB model. Timmermann[41,47] suggested that the second sorption stage introduced by the
GAB model may be limited to a certain number of layers
and that, thereafter, a third stage becomes available
for the sorbate molecules, which has true liquid-like properties, as postulated by the original BET model. The
so-called third sorption stage isotherm developed by
Timmermann[47] is
M
CKaw HH 0

M0 1  Kaw 1 CH  1Kaw 

47

where H and H0 are functions containing a fourth dimensionless parameter h.


1  K Kaw h
K 1  aw


H  1 1  Kaw
0
H 1
h 1  haw 
H
1  aw
H 1

48
49

For h ! 1, H; H 0 ! 1 and with this result Eq. (47) transforms into a GAB equation.

923

MODELS FOR SORPTION ISOTHERMS: A REVIEW

VIOLLAZ GAB MODEL


Equation (27) was modified by Viollaz and Rovedo[42] in
an empirical way by adding a term
M
CKaw
CKK2 a2w

M0 1  Kaw 1 C  1Kaw  1  Kaw 1  aw


50
K2 is an additional dimensionless parameter.
The second term of the RHS allows the necessary flexibility to obtain a good fitting for high aw values. This term
has a very low weight for low values of aw, so the values of
M0, C, and K are not substantially affected by the addition
of this new term. It can be observed that if the value of K2
is equal to zero, the GAB isotherm is obtained. The predictive capacity of Eq. (50) was evaluated using the sorption
data for starch and gluten by Viollaz and Rovedo.[42]
LEWICKI MODEL
Lewicki[48] developed a model on the basis of Raoults
law. The model assumed that water present in food occurs
in two states, as free water with properties of the bulk
water and as water of hydration. The Lewicki model of
food sorption isotherm is

b1
1
MA
1
51
aw
where A and b are constants. Equation (51) described the
water sorption isotherms of 38 products and 31 model
mechanical mixtures.[48]
WATER ACTIVITY OF SOLUTION/MIXTURES
For water solution, Raoults law shows that the water
activity in ideal solution is equal to mole fraction of water
(xw); that is, aw xw.
For nonideal solutions, the activity coefficient c corrects
for deviation from nonideality.
aw xw c

52

Norrish[49] proposed a semiempirical correction to Raoults


law for the nonideal solution
aw xw kx2s

53

where xs is mole fraction of solute and k is an empirical


constant.
Ross[50] extended the Norrish equation for multi component systems by using the Gibbs-Duhem equation with the
assumption that the activity coefficient of individual components with water are equal to their activity coefficients
in binary mixtures in water.
The Ross equation is represented as:
lnaw

k
X
j1

lna0wj

54

Here, awj is the water activity of the binary mixture with the
j component.
The Ross equation has been applied by several researchers, e.g., Chirife et al.,[51] Herman et al.,[52] and Roman
et al.[34] The water activities of non-solute parts were also
evaluated with the Henderson equation based on literature
data for starch, protein, and fiber by Roman et al.[34]
ARTIFICIAL NEURAL NETWORK MODELING
Artificial neural network (ANN) modeling is a method
to describe relations between independent and dependent
variables when the explicit form of mapping is not known.
In recent years, the concept of ANN has gained wide
acceptance in food engineering for predictive modeling.
In case of sorption isotherm prediction, Myhara et al.,[53]
Myhara and Sablani,[54] and Kaminski and Tomczak[55]
used ANN modeling for different food materials. Myhara
et al.[53] demonstrated that when chemical composition
data is combined with physical data through an ANN
approach, significant improvements in the prediction of
water sorption behavior can be achieved.
COMPARISON OF BET AND GAB MODELS
The BET and the GAB isotherms are closely related as
they are derived from the same statistical model. The
GAB model represents a refinement over the BET model
and shares with it two original BET constants (M0, the
monolayer capacity, and C, the energy constant) and owes
its versatility due to the introduction of a third constant
(K). The regression of an experimental sorption data by
each of these two isotherms will give two sets of values of
M0 and C. The same type of differences between both sets
of values of M0 and C have always been observed:
M0B < M0G ;

CB > CG

55

where subscript B and G represent BET and GAB,


respectively.
The M0 value given by BET isotherms is always lower
than the monolayer value derived from GAB and, reversibly, the BET value of energy constant is always higher than
the GAB value. It has been demonstrated that there exist
mathematical and physical reasons for the inequalities set
by Eq. (55).[56,57]
The BET equation may be rewritten as
M

CB aw M0B
1  aw 1 CB  1aw 

56

To obtain the two characteristic constants, the BET equation is linearized by the following function
F BET

aw
1
CB  1

aw
1  aw M CB M0B CB M0B

57

924

BASU, SHIVHARE, AND MUJUMDAR

The so-called F(BET) vs. aw plot usually gives a linear part


at low activities (0.05 < aw < 0.30.5), after which always
an upward curvature is observed. This deviation shows that
at higher activities, less gas or vapor is sorbed than that
indicated by the BET equation using the values of the constants corresponding to the low activity range.
The GAB equation is rewritten as
M

CG Kaw M0G
1  Kaw 1 CG  1Kaw 

where
1
CGG M0G

64a

CGG  1 21  K
CGG M0G

64b

1  KCGG 1  K
CGG M0G

64c

A
B

58

and
CG K CGG

59

Here, M0G is the GAB monolayer capacity and CG(G) is the


analogue of this formulation to the BET energy constant
CB. The constant (K) is a measure of the difference of free
enthalpy of the sorbate molecules in the pure liquid and
the second sorption stage, the layer above the monolayer.
It is always found that K < 1. Finally, with K 1, the
GAB isotherm reduces to the original BET equation
(M0B M0G; CB CG(G)).
To determine the three constants of the GAB equation,
it is linearized like the BET equation and the function
F(GAB) is introduced:
F GAB

aw
1
CG  1

aw
1  Kaw M CG KM0G CG M0G

60

F(GAB) vs. aw should be linear if the correct K value is


used. From the two linear regression coefficients, M0G
and CG are obtained.
Equation (57) can be rewritten in a modified form by
incorporating F(GAB)
aw
1  Kaw
F BET

F GAB
1  aw M
1  aw

A  C B  Caw

C
1  aw

65

By minimizing the sum over the n experimental points


n 
X

F BETi  a0 a1 awi

2

minimum

66

i1

Here, awi stands for aw at the point i. The coefficients a0 and a1


are given by the solutions of the system of normal equations
associated to the extreme condition (Eqs. (67a) and (67b).
The least square estimates of a0 and a1 , a00 and a01 are
given directly by the BET relations by which the BET constants are usually computed.
BET: a00
a01

1
CB M0B

CB  1
CB M0B

67a
67b

and hereafter

The second expression for F (BET) shows that, if K < 1,


F  (BET) will not be linear in aw but will present a hyperbolic behavior.
Caw
1  aw

PBET a0 a1 awi

61

Substituting the value of F(GAB) from Eq. (60) into


Eq. (61) a new relationship represented by F (BET) is
obtained, in terms of the three constants of the GAB
isotherm:


1
CG  1
1  Kaw

F BET

aw
CG KM0G CG M0G
1  aw
CGG  1 21  K
1

aw
CGG M0G
CGG M0G


1  KCGG 1  K
aw

62
CGG M0G
1  aw

F  BET A Baw

QUANTITATIVE EXPRESSION OF INEQUALITIES


The experimental values of F(BET) are adjusted by the
linear polynomial

63

M0B

CB

1
a00 a01

a00 a01
a00

68a

68b

It should be noted that the energy constant CB is inversely


proportional to the intercept a00 of the linear regression
polynomial P(BET) of Eq. (68b) and therefore CB is highly
sensitive to the value of a00 , which is usually very low.
The second expression of F (BET) may also be adjusted
by the same linear polynomial (Eq. (65)), but now using an
analytical formulation as F (BET) is known as a function
of aw and not by a set of numerical data. The calculation
implies the adjustment of a function of a known functional
dependence of a higher degree than one to a straight line.
In the discrete procedure, F (BET) given by Eq. (63) is put
into condition (Eq. (66)), which becomes

925

MODELS FOR SORPTION ISOTHERMS: A REVIEW


n
X
fA  C B  Cawi C=1  awi a0 a1 awi g2
i1

minimum

69

This expression can be solved analytically for a0 and


a1 using the least squares technique. Coefficients a0 and
a1 become functions of the constants A, B, and C of
Eq. (63) on one side, and of regression sums over the values
of the independent variable awi on the other.
The final expressions are the following:
a0

A  C Cd=d0

70a

a1 B  C Cd1 =d

70b

where a0 and a1 are the minimum squares estimates in


terms of F (BET). The functions d, d0 , and d1 contain only
the regression sums of aw over the employed regression
interval with the following signs: d0 =d < 0, d1 =d < 0, and
d0 =d d1 =d < 0 for aw < 1.[56,57]
The corresponding relations where Gaussian brackets
have been used are
d0  1=1  aw a2w   aw =1  aw aw 

71a

d1  aw =1  aw n  1=1  aw aw 

71b

d  na2w   aw 2

71c

where n is the number of data included in the regression.


As Eqs. (65) and (66) remain valid, the BET constants
are now given by
M0B

1
1=fA B C d0 d1 =d  2g 72a
a00 a01

and
CB

These functions are always greater than unity.[56] Therefore, Eqs. (73a) and (73b) reproduce the inequalities
(Eq. (55)) if K<1.
The marked differences between the BET and GAB constants are noteworthy. The GAB isotherm, is an improved
version of the BET model, by the introduction of the third
parameter, K, characterizing the state of sorbed molecules
beyond the first layer.
RESEARCH RESULTS FOR SORPTION ISOTHERMS
OF FOOD MATERIALS
Modeling of sorption behavior of a variety of food
materials has been carried out by many researchers. A list
of recent research done after 2000 on different types
of food materials is presented in Table 1. This list is
not exhaustive; comprehensive information of sorption
isotherms for different food materials has been
published.[39,5860]
STATISTICAL TESTING OF SORPTION MODELS
Statistical analysis of an isotherm may be quantified
through five standards: the coefficient of determination
(R2), the residual sum of squares (RSS), the standard error
of the estimate (SE), the mean relative deviation (MRD),
and the plot of residuals. The coefficient of determination
is a statistical measure of how closely two variables are
related. The coefficient of determination is a dimensionless
index that ranges from 0 to 1 and reflects the extent of linear relationship between two data sets.
The residual sum of squares (RSS) is defined as
RSS

n
X

M  MCal 2

75

i1

a00 a01
a00

fABC d0 d1 =d2g=fAC d0 =d1g

72b

Finally, by Eqs. (71a)(71c), A, B, C f M0G ;CG ;K


and an algebraically explicit expression for M0B and CB
in terms of the three GAB constants are obtained:
M0G

M0B 
121K=CGG Rm


21K 
CB CGG 1
RC
CGG

73a
73b

where the functions Rm and RC are given by




Rm 11K CGG 1K=CGG 21K
d0 d1 =d2


RC Rm = 11KCGG 1Kd0 =d1

74a
74b

where M is the measured value, MCal is the value estimated


through the fitting equation and n is the number of data
points.
The standard error of estimate (SE) is the conditional
standard deviation of the dependent variable and has the
form
s s
Pn
2
RSS
i1 M  MCal
SE
76

df
df
Here, df represents the degrees of freedom of the fitting
equation. If a large data set is available, Eq. (76) is simplified to
r
RSS
SE
77
n
The mean relative deviation (MRD) is an absolute value
because it gives the mean divergence of the estimated data
from the measured data

926

BASU, SHIVHARE, AND MUJUMDAR

TABLE 1
Recent works on sorption isotherms of food materials
Temperature (C)

Food product
Berries
Mushroom

4, 13, 27
4, 13, 27
3070
25, 30, 35

Cocoa beans
Starch, potato
Pumpkin seed flour
Cured beef and pork
Chicken meat
Hazelnuts
Rough rice

3060
3060
10, 25, 40
10, 20, 25, 35, 49
430
25
0, 5, 10, 15, 20, 25, 30, 35

Yogurt powder
Gelatine gel
Crystalline lactose powder

35
20
12, 20, 30, 40

MRD

n
1X
jM  MCal j
M
n i1

aw

Best fitted model

Reference

0.110.87
0.120.75
0.200.80
0.080.94

GAB
GAB
Chung-Pfost
BET (aw < 0.5)
Harkins-Jura (aw > 0.5)
GAB, Halsey
GAB, Ferro-Fontan, Peleg
GAB
Peleg, GAB
Ferro-Fontan, GAB
GAB
Modified Chung-Pfost,
modified Henderson
Modified Chung-Pfost
Ferro-Fontan
Timmermann-GAB, GAB

63
63
64
65

0.110.83
0.050.95
0.110.85
0.100.94
0.250.94
0.110.90
0.2590.935
0.40.99
0.750.98
0.110.98

78

Plotting of the residuals (MMCal) against the independent variable is also used as a measure of the distribution of errors. If the model is correct, then the
residuals should be only random independent errors with
a zero mean, constant variance and arranged in a normal

15
66
67
68
69
70
71
72
20
73

distribution. If the residual plots indicate a clear pattern,


the model should not be accepted.
In general, low values of the correlation coefficient, high
values of RSS, SE, and MRD, and clear patterns in the
residual plots mean that the model is not able to explain
the variation in the experimental data. It is also evident
that a single statistical parameter cannot be used to select
the best model and the model must always be assessed
based on multiple statistical criteria.[61]

TABLE 2
Data on goodness of fit of selected models for sardine sheet at selected temperatures (25, 40, and 50C) and water activity
(0.110.84)
Coefficients of linear regression
Isotherm
BET
Oswin
Smith
Hasley
Henderson
Chung and Pfost
Iglesias and Chirife
GAB
Peleg
Modified Oswin
Modified Henderson
Modified Chung-Pfost

0.9390.978
0.9630.984
0.9540.974
0.9590.994
0.9310.974
0.8850.906
0.9900.993

Coefficients of nonlinear regression

S.E.

R2

S.E.

0.6480.968
0.1010.139
0.0120.022
0.0650.164
0.1520.237
0.250.299
0.0060.008

0.9830.996
0.9860.997
0.9540.974
0.9700.994
0.9480.978
0.9310.953
0.990.993
0.9890.996
0.9360.968
0.8130.995
0.9480.978
0.9310.953

0.0040.009
0.0040.009
0.0120.021
0.0200.042
0.360.054
0.060.067
0.0060.008
0.0040.007
0.0140.026
0.0050.029
0.0360.054
0.0600.067

MODELS FOR SORPTION ISOTHERMS: A REVIEW

927

TABLE 3
Pattern of residual plots of selected models
Model

Residual plot pattern

BET
Peleg
Oswin
Hasley
GAB
Modified Henderson
Modified Chung-Pfost
Iglesias and Chirife

Patterned
Patterned
Scattered
Patterned
Scattered
Scattered
Patterned
Patterned

FIG. 4. Residual plot of sorption data for sardine fish at 40C (GAB
isotherm).

AN EXAMPLE
Published sorption data[62] (water activity range 0.11
0.84) of freeze-dried sardine fish at 25, 40, and 50C
was selected for comparison of the different sorption isotherm models. Both linear and nonlinear approaches were
used to estimate the model coefficients. Standard error
and R2 values of different isotherm models for both linear
and nonlinear forms are given in Table 2. The R2 values
of the nonlinear models were always higher than that of
the linear equations. Similarly, standard error values were
also lower in nonlinear cases compared to the linearized
form, which suggests that nonlinear forms should be used
for estimation of model parameters. After choosing few
sorption models with higher R2 and lower SE values, their
residual plots were checked. Residual plot patterns of different models are presented in Table 3. Among all the
TABLE 4
Calculated GAB parameters at selected temperatures

Temperature
(C)
25
40
50

Monolayer moisture
content
(kg water=kg
dry matter)

2.0699
0.6834
7.3495

1.1172
1.0697
1.2121

0.0768
0.1857
0.1857

models, the GAB and Oswin models were close because


of their high correlation coefficient values and low standard error values. Both these models gave a scattered
residual plot. Although the Oswin model gave higher R2
(0.9860.997) and lower SE (0.0040.009) values, it does
not include temperature parameter. It may therefore be
inferred that GAB isotherm describes the sorption data
best over the experimental range of temperature and
water activity among all the isotherm models considered.
The monolayer moisture content (M0) and dimensionless
GAB parameters related to heat of sorption for the
monolayer and multilayer region (C and K) determined
from GAB equation are reported in Table 4. Figure 3
represents the sorption isotherm for sardine fish based
on GAB model and Fig. 4 shows a typical residual plot
of sorption data at 40C.

CONCLUSION
Theoretical aspects of sorption of water and commonly
used mathematical models to describe sorption isotherms
are discussed. Statistical criteria for selection of a model
have also been presented. Both the GAB and Oswin equations presented high R2 and low standard error values and
scattered residual pattern for freeze-dried sardine fish. But
because of lack of a temperature term in the Oswin equation, the GAB model is recommended for fitting the sorption data. From this discrimination approach based on
statistical measures, it is possible to select the best model
equation for describing experimental sorption data.
NOMENCLATURE
aw
A, A0 , B, r, k1, k2, n1, n2
C

FIG. 3. Sorption isotherm for sardine fish at 25, 40, and 50 C (GAB
isotherm).

df
ERH

Water activity
Constants
Dimensionless parameter
related to heat of sorption
of monolayer region
Degrees of freedom
Equilibrium relative
humidity

928
H and H0

K2
M, M1, M2
MC
MCal
MRD
M0
M0B(B), M0G(G), CB(B), CG(G)
P

P0
Q
Q12
R
RSS
SE
T
w
X
Greek Letters
a, a0
c

DHd
DHS
DHvap
d; d0 ; d1 ; a0 ; a1 ; a0 ; a1

BASU, SHIVHARE, AND MUJUMDAR

Functions containing a
fourth dimensionless
parameter h
Dimensionless GAB
parameter related to heat of
sorption in multilayer
region
Dimensionless parameter
Moisture content
Characteristic moisture
content
Estimated value
Mean relative deviation
Monolayer moisture content on the internal surface
Constants
Water vapor pressure in the
food material at any given
temperature
Vapor pressure of pure
water at that temperature
Net heat of sorption of
water
Net heat of desorption
Universal gas constant
Residual sum of squares
Standard error of the
estimate
Absolute temperature
M=M0
Kaw
Constants
Parameter which accounts
for the structure of sorbed
water
Total heat of sorption in
the food
Net isosteric heat of sorption
Heat of vaporization of
water
Functions

REFERENCES
1. Mujumdar, A.S.; Devahastin, S. Fundamental principles of drying. In
Mujumdars Practical Guide to Industrial Drying; Devahastin, S., Ed.;
Exergex Corp: Montreal, 2000; 122.
2. Brunauer, S.; Deming, L.S.; Teller, E. On a theory of Van der Waals
adsorption of gases. Journal of the American Chemical Society 1940,
62, 17231732.
3. Brunauer, S.; Emmet, P.H.; Teller, E. Adsorption of gases in multimolecular layers. Journal of the American Chemical Society 1938, 60,
309319.

4. Mathlouthi, M.; Roge, B. Water vapour sorption isotherms and the


caking of food powders. Food Chemistry 2003, 82, 6171.
5. van den Berg, C.; Bruin, S. Water activity and its estimation in food
systems: Theoretical aspects. In Water Activity: Influences on Food
Quality; Rockland, L.B., Stewart, G.F., Eds.; Academic Press: New
York, 1981; 161.
6. Kinsella, J.E.; Fox, P.F. Water sorption by proteins: Milk and whey
proteins. CRC Critical Reviews in Food Science & Nutrition 1986,
24, 91139.
7. Mujumdar, A.S. Ed. Handbook of Industrial Drying, 3rd Enhanced
Ed; CRC Press: Boca Raton, FL, 2006.
8. Iglesias, H.A.; Chirife, J. Handbook of Food Isotherms; Academic
Press: New York, 1982.
9. Rizvi, S.S.H. Thrmodynamic properties of foods in dehydration. In
Engineering Properties of Foods; Rao, M.A., Rizvi, S.S.H., Eds.;
Marcell Dekker: New York, 1986; 133214.
10. Rahman, M.; Labuza, T.P. Water activity and food preservation. In
Handbook of Food Preservation; Shafiur Rahman, M., Ed.; Marcel
Dekker: New York, 1997; 339382.
11. Labuza, T.P.; Hyman, C.R. Moisture migration and control in multidomain foods. Trends in Food Science & Technology 1998, 9, 4755.
12. Bell, L.N.; Labuza, T.P. Practical Aspects of Moisture Sorption Isotherm Measurement and Use, 2nd Ed; AACC Egan Press: Egan,
MN, 2000.
13. Al-Muhtaseb, A.H.; McMinn, W.A.M.; Magee, T.R.A. Moisture
sorption isotherm characteristics of food products: A review. Transaction of IChemE, Part C, (Food and Bioproducts Processing) 2002,
80, 118128.
14. Tsami, E.; Maroulis, B.; Marinos-Louris, D.; Saravacos, G.D. Heat of
sorption of water in dried fruits. International Journal of Food science
and Technology 1990, 25, 350359.
15. McMinn, W.A.M.; Magee, T.R.A. Thermodynamic properties of
moisture sorption of potato. Journal of Food Engineering 2003, 60,
157163.
16. Aguerre, R.J.; Suarez, C.; Viollaz, P.E. Enthalpy-entropy compensation in sorption phenomena application to the prediction of the
effect of temperature on food isotherms. Journal of Food Science
1986, 51 (6), 15471549.
17. McMinn, W.A.M.; Al-Muhtaseb, A.H.; Magee, T.R.A. Enthalpyentropy compensation in sorption phenomena of starch materials.
Food Research International 2005, 38, 505510.
18. Leffer, J.E.; Grunwald, E. Rates and Equilibria of Organic Reactions;
Wiley: New York, 1963.
19. Troller, J.A. Statistical analysis of aw measurements obtained with the
sinascope. Journal of Food Science 1977, 42, 8690.
20. Baucour, P.; Daudin, J.D. Development of anew method for fast
measurement of water sorption isotherms in the high humidity range
validation on gelatine gel. Journal of Food Engineering 2000, 44,
97107.
21. Langmuir, I. The adsorption of gases on plane surfaces of glass, mica
and platinum. Journal of the American Chemical Society 1918, 40,
13611402.
22. Chirife, J.; Iglesias, H.A. Equations for fitting water sorption isotherms of foods: Part 1. A review. Journal of Food Technology
1978, 13, 159174.
23. Oswin, C.R. The kinetics of package life. III. Isotherm. Journal of
Society Chemical Industry 1946, 65, 419421.
24. Chen, C. A study of equilibrium relative humidity for yellow-dent
corn kernels. Ph.D. thesis, University of Minnesota: St. Paul, 1988.
25. Halsey, G. Physical adsorption on non-uniform surfaces. Journal of
Chemical Physics 1948, 16, 931937.
26. Iglesias, H.A.; Chirife, J. Equilibrium moisture content of air dried
beef: Dependence on drying temperature. Journal of Food Technology 1976, 11, 567573.

MODELS FOR SORPTION ISOTHERMS: A REVIEW


27. Iglesias, H.A.; Chirife, J. Prediction of effect of temperature on water
sorption of food materials. Journal of Food Technology 1976, 11,
109116.
28. Iglesias, H.A.; Chirife, J. A model for describing the water sorption
behavior of foods. Journal of Food Science 1976, 41, 984992.
29. Henderson, S.M. A basic concept of equilibrium moisture. Agricultural Engineering 1952, 33, 2932.
30. Thompson, T.L.; Peart, R.M.; Foster, G.H. Mathematical simulation
of corn drying-a new model. Transactions of the ASAE 1968, 24 (3),
582586.
31. Aggarwal, K.K.; Calry, B.L.; Nelson, G.L. Investigation into the theories of desorption isotherms for rough rice peanuts. Journal of Food
Science 1971, 36, 919924.
32. Chen, C.S.; Clayton, J.T. The effect of temperature on sorption isotherms of biological materials. Transactions of the ASAE 1971, 14,
927929.
33. Young, J.H. Evaluation of models to describe sorption and desorption
equilibrium moisture content isotherms of Virginia type peanuts.
Transactions of the ASAE 1976, 19, 146150.
34. Roman, A.D.; Herman-y-Lara, E.; Salgado-Cervantes, M.A.; GarcaAlvarado, M.A. Food sorption isotherms prediction using the Ross
equation. Drying Technology 2004, 22 (8), 18291843.
35. Chung, D.S.; Pfost, H.B. Adsorption and desorption of water vapour
by cereal grains and their products. Part I. Heat and free energy
changes of adsorption and desorption. Transactions of the ASAE
1967, 10, 549551, 555.
36. Chung, D.S.; Pfost, H.B. Adsorption and desorption of water vapour
by cereal grains and their products. Part II. Hypothesis for explaining
the hysteresis effect. Transactions of the ASAE 1967, 10, 552555.
37. Pfost, H.B.; Mourer, S.G.; Chung, D.S.; Milliken, G.A. Summarizing
and reporting equilibrium moisture data for grains. ASAE Paper No.
763520; ASAE: St. Joseph, MI, 1976.
38. Ferro Fontan, C.; Chirife, J.; Sancho, E.; Iglesias, H.A. Analysis of a
model for water sorption phenomena in foods. Journal of Food
Science 1982, 47, 15901594.
39. Iglesias, H.A.; Chirife, J. An alternative to the Guggenheim, Anderson
and De Boer for the mathematical descrioption of moisture sorption
isotherms of foods. Food Research International 1995, 28 (3),
317321.
40. van den Berg, C. Desorption of water activity of foods for engineering
purposes by means of the GAB model of sorption. In Engineering and
Foods; Mckenna, B.M., Ed.; Elsevier: London, U.K., 1984; 311321.
41. Timmermann, E.O.; Chirife, J. The physical state of water sorbed at
high activities in starch in terms of the GAB sorption equation. Journal of Food Engineering 1991, 13, 171179.
42. Viollaz, P.; Rovedo, C.O. Equilibrium sorption isother and thermodynamic properties of starch and gluten. Journal of Food Engineering
1999, 40, 287292.
43. Blahovec, J. Sorption isotherms in materials of biological origin mathematical and physical approach. Journal of Food Engineering 2004,
65, 489495.
44. Peleg, M. An empirical model for the description of moisture sorption
curves. Journal of Food Science 1988, 53, 12491251.
45. Peleg, M. Assessment of a semi-empirical four parameter general
model for sigmoid moisture sorption isotherms. Journal of Food Process Engineering 1993, 16, 2137.
46. van den Berg, C. Sorption theories. In Vapour Sorption Equilibria and
other Water-Starch Interactions; a Physicochemical Approach. Ph.D.
thesis, Wageningen Agricultural University: Netherlands, 1981;
100126.
47. Timmermann, E.O. A B.E.T. like three sorption stage isotherm. Journal
of the Chemical Society, Farday Transactions I 1989, 85, 16311645.
48. Lewicki, P.P. Raoults law based food water sorption isotherm. Journal of Food Engineering 2000, 43 (1), 3140.

929

49. Norrish, R.S. An equation for the activity coefficients and equilibrium
relative humidities of water in confectionary syrups. Journal of Food
Technology 1966, 1, 2539.
50. Ross, K.D. Estimation of water activity in intermediate moisture
foods. Food Technology 1975, 3, 2654.
51. Chirife, J.; Resnik, S.L.; Ferro Fontan, C. Application of Ross equation for prediction of water activity in intermediate moisture food system containing a nonsolute solids. Journal of Food Technology 1985,
20, 773779.
52. Herman, E.; De la Cruz, J.; Garcia, M.A. Prediction of pineapple isotherms using the Ross equation. Drying Technology 1999, 17 (4&5),
11611172.
53. Myhara, R.M.; Sablani, S.S.; Al-Alawi, S.M.; Taylor, M.S. Water
sorption isotherm of dates: Modeling using GAB equation and artificial neural network approaches. Lebensmittel Wissenschaft undTechnologie 1998, 31, 699706.
54. Myhara, R.M.; Sablani, S.S. Unification of fruit water sorption
isotherms using artificial neural networks. Drying Technology 2001,
19 (8), 15431554.
55. Kaminski, W.; Tomczak, E. Description of sorption isotherms as a
function of water activity, composition and temperature using MLP.
In Drying 2004Proceeding of the 14th International Drying Symposium
(IDS 2004); Sao Paolo, Brazil, 2225 August 2004; vol. A, 201208.
56. Timmermann, E.O.; Chirife, J.; Iglesias, H. Water sorption isotherms
of foods and foodstuffs: BET or GAB parameters? Journal of Food
Engineering 2001, 4 (1), 1931.
57. Timmermann, E.O. Multilayer sorption parameters: BET or GAB
parameters? Colloids and Surfaces A: Physicochemical and Engineering Aspects 2003, 220, 235260.
58. Chen, C.C.; Vance Morey, R. Comparision of four EMC=ERH equation. Transactions of the ASAE 1989, 32 (3), 983990.
59. Aguerre, R.J.; Viollaz, P.E.; Suarez, C. A fractal isotherm for multilayer adsorption in foods. Journal of Food Engineering 1996, 30,
227238.
60. Chen, C.; Jayas, D. Evaluation of the GAB equation for the isotherms
of agricultural products. Transactions of the ASAE 1998, 41 (6),
17551760.
61. Jayas, D.S.; Mazza, G. Comparision of five, three-parameter equation
for description of adsorption data of oats. Transactions of the ASAE
1993, 36 (1), 119125.
62. Sablani, S.S.; Myhara, R.M.; Mahgoub; Al-Attabi, Z.H.;
Al-Mugheiry, M.M. Water sorption isotherms of freeze dried fish
sardines. Drying Technology 2001, 19 (3&4), 673680.
63. Khalloufi, S.; Giasson, J.; Ratti, C. Water activity of freeze dried
mushrooms and berries. Canadian Agricultural Engineering 2000,
42 (1), 5156.
64. Shivhare, U.S.; Arora, S.; Ahmed, J.; Raghavan, G.S.V. Moisture
adsorption isotherm for mushroom. Lebensmittel Technologie 2004,
37, 133137.
65. Sandoval, A.J.; Barreiro, J.A. Water sorption isotherm of nonfermented cocoa benas (Theobroma cacao). Journal of Food Engineering
2002, 51, 119123.
66. Al-Muhtaseb, A.H.; McMinn, W.A.M.; Magee, T.R.A. Water sorption isotherms of starch powders: Part 1: Mathematical description
of experimental data. Journal of Food Engineering 2004, 61 (3),
297307.
67. Menkov, N.D.; Durakova, A.G. Equilibrium moisture content of semi
fatted pumpkin seed flour. International Journal of Food Engineering
2005, 1 (3), Article 1. http:==www.bepress.com=ijfe=vol1=iss3/art1
68. Delgado, A.E.; Sun, D.W. Desorption isotherm for cooked and
cured beef and pork. Journal of Food Engineering 2002, 51, 163170.
69. Delgado, A.E.; Sun, D.W. Desorption isotherms and glass transition
temperature for chicken meat. Journal of Food Engineering 2002,
55, 18.

930

BASU, SHIVHARE, AND MUJUMDAR

70. Guzey, D.; Ozdemir, M.; Seyham, F.G.; Dogan, H.; Devres. Adsorption isotherms of raw and roasted hazelnuts. Drying Technology 2001,
19 (3&4), 691699.
71. San Martin, M.B.; Mate J.I.; Fernandez, T.; Virseda, P. Modelling
adsorption equilibrium moisture characteristics of rough rice. Drying
Technology 2001, 19 (3&4), 681690.

72. Stencl, J. Modelling the water sorption isotherms of yoghurt powder


spray. Mathematics and Computers in Simulation 2004, 65 (1& 2),
157164.
73. Bronlund, J.; Paterson, T. Moisture sorption isotherms for crystalline,
amorphous and predominantly crystalline lactose powders. International Dairy Journal 2004, 14 (3), 247254.