khan2017

Experimental investigation of bound and free water transport process
during drying of hygroscopic food material
Md Imran H. Khan
a , b
, R. Mark Wellard
a
, Szilvia Anett Nagy
c , d
, M.U.H. Joardder
e
,
M.A. Karim
a , *
a
Science and Engineering Faculty, Queensland University of Technology (QUT), Brisbane, Queensland, Australia
b
Department of Mechanical Engineering, Dhaka University of Engineering &Technology, Gazipur, 1700, Bangladesh
c
P ecs Diagnostics Center, H-7623, P ecs, R et Street 2, Hungary
d
MTA - PTE Neurobiology of Stress Research Group, H-7624, P ecs, Ifjús ag Street 20, Hungary
e
Department of Mechanical Engineering, Rajshahi University of Engineering and Technology, Bangladesh
article info
Article history:
Received 15 July 2016
Received in revised form
30 March 2017
Accepted 4 April 2017
Keywords:
Bound water
Free water
Drying
Cell rupture
Transport
abstract
Hygroscopic food materials contain free (FW) and bound (BW) water in different cellular environments.
In-depth understanding of the mechanisms of moisture migration from different cellular environments
during drying is crucial for optimising heat and mass transfer as well as for obtaining better quality dried
foods. Therefore, the main aim of the present work is to investigate the transportation mechanisms of
FW and BW during drying. Experiments were performed on the potato tissue using
1
H-NMR T
2
relax-
ometry to uncover the mechanisms involved in FW and BW transportation. The results have confirmed
the view that BW migrates after the rupture of the cell membranes. It is interesting to highlight that the
cell membranes rupture at different stages of drying rather than collapsing at one time. The membrane
collapse depends predominantly on the penetration rate of heat energy and the pressure gradient be-
tween intracellular and intercellular environments. All test results suggest that most of the cell mem-
branes rupture at the middle stage of drying where the moisture content is about 2e4 kg/kg (db.).
Furthermore, the moisture distribution profile confirmed that some moisture remained around the
centre of the dried sample although the surface of the sample became dry.
©2017 Elsevier Masson SAS. All rights reserved.
1. Introduction
Drying is an excellent method of food preservation. However,
the dehydration of foodstuffs is a very complex process because of
its complexity in internal structure and simultaneous heat and
mass transfer during drying [1]. Understanding of the actual heat
and mass transfer during drying is crucial for optimising energy
efficiency of the drying process and preserving the quality attri-
butes of food materials [2]. Many physical and chemical changes
take place in food tissue during drying and moisture and temper-
ature distributions significantly contribute to these changes [3].
Therefore, understanding of moisture and temperature distribu-
tions and transport mechanism during drying is important. The rate
of moisture transport depends on the size and orientation of cells as
well as the types of cellular water in the samples [4]. Plant-based
foods materials are hygroscopic and porous in nature and contain
two types of water, namely free water (FW) and bound water (BW)
[5]. FW is present in capillaries or intercellular spaces; whereas, the
water in the intracellular space is referred to as BW [6], as shown in
Fig. 1. The migration pathways of FW and BW are also different. It is
assumed that most of the BW migrate after the collapse of the cells.
When cells are collapsed, BW from cells moves to intercellular
spaces and shows the characteristics of FW [7]. Due to insufficient
knowledge about the migration pathways and characteristics of FW
and BW during drying, the current food drying models consider
bulk water transport mechanisms in the mathematical modelling
[8e12].
Compared to single phase models, multi-phase models consid-
ering transport of liquid water, water vapour and air inside the food
materials are more realistic [5]. These three phases (water, gas and
solid matrix) inside plant-based food structures, as shown in Fig. 2,
represent the domain which is commonly used in most multiphase
drying models. According to the assumptions presented in
Refs. [8e12], there is no water inside the solid matrix.
*Corresponding author.
E-mail address: [email protected] (M.A. Karim).
Contents lists available at ScienceDirect
International Journal of Thermal Sciences
journal homepage: www.elsevier.com/locate/ijts
http://dx.doi.org/10.1016/j.ijthermalsci.2017.04.006
1290-0729/©2017 Elsevier Masson SAS. All rights reserved.
International Journal of Thermal Sciences 117 (2017) 266e273

Although this assumption leads to the simplified process of
multiphase transport, the real transport process remains unclear
[7,13]. It is reported that migration of free water has a minimum
effect on the food quality. Migration of BW causes cellular
shrinkage, pore formation and the collapse of the cells and pore
structure and hence has a major effect on food quality [14,15].
Moreover, the energy requirements for drying of a particular food
material depend on its cellular structure and cellular level moisture
distributions [16]. Transport of bound water requires more energy
compared to transport of free water [17]. In most of the cases, en-
ergy and time required for first 90% of water is almost equal to the
energy and time required for removing last 10% of water present in
a food sample [14]. Therefore, in order to accurately predict the heat
and mass transport during drying of food material, better under-
stand of the migration mechanisms of BW water is necessary.
There are several techniques available for analysing bound wa-
ter transport including bioelectric impedance analysis (BIA), dif-
ferential scanning calorimetry (DSC), differential thermal analysis
(DTA), centrifugal settling method (CSM) and nuclear magnetic
resonance (NMR) methods. However, not all these methods are
suitable for investigating moisture migration mechanisms while
drying is in progress. BIA and NMR methods are evaluated as the
most appropriate for this type of investigations. BIA is a very simple
and established technique for assessing body compositions by
measuring the resistance of tissues to the flow of electrical current
[6]. The proportion of different contents in a tissue can be calcu-
lated as the current flows more easily through the parts of the
material that contain more water. This method was used in the
study of Halder et al. [4] for calculation of the migration pathways
of intracellular water in the plant-based food materials. The tem-
perature influence on the cellular water migration was examined in
the same study. It was stated that all of the membranes of the cells
would collapse at once after specific temperature is reached. They
suggested that below 50
C the cells remain intact and therefore the
conversion of the intracellular water to free water remains un-
changed during the drying. As cells remain intact, the intracellular
water (BW) moves to intercellular space only through micro-
capillaries, denoting this moisture transport as the slow one.
However, their argument is not justified because cell collapse
depends on internal thermal stress [15] that first develops near the
surface and gradually penetrates to the centre of the sample during
convective drying. In other words, entire food sample does not
reach ‘cell rupture temperature’at a time. Therefore, it is logical
that the cells may collapse progressively from the surface to centre
[14,15]. Moreover, BIA is mainly used for analysing animal fat
composition. It may not be a sufficiently accurate technique for
predicting moisture migration pathways in plant tissue as BIA
cannot detect the position of different types of water (FW and BW).
Nuclear magnetic resonance (NMR) is a widely used method for
investigating the distribution of different types of water in various
locations inside plant-based food material [18]. Proton nuclear
magnetic resonance (
1
H-NMR) relaxometry study has been proven
to be a viable method in the study of plants and plant-based food
materials submitted to stress reflecting anatomical details of the
entire tissue and the water status in particular [19,20].
1
H-NMR
signals, which are an average over the whole sample, provide in-
formation about the water content of the plant tissue since the
proton signal is dominated by water protons and the proton NMR
signal intensity is directly proportional to the proton density of the
tissue [21,22]. The water proton relaxation behaviour strongly de-
pends on the water mobility in the microscopic environment of the
tissue, local magnetic field fluctuations (related to the molecular
environment) and the strength of the applied magnetic field. The
spin-spin (T
2
) relaxation is the transverse component of the
magnetization vector, which exponentially decays towards its
equilibrium value after excitation by radio frequency energy.
Some researchers have tried to investigate the changes in water
compartmentation during drying of plant tissue [23,24]. These
studies found a strong relationship between different water T
2
relaxation times and the percentage of moisture loss during drying.
However, they did not investigate the migration mechanisms of
bound water and free water separately. Moreover, it is not clear
when the cell exactly collapses and what are the consequence for
the migration of bound water.
Therefore, the primary aim of this study is to investigate the free
and bound water migration mechanism in plant-based food ma-
terial during drying.
Fig. 1. General water distribution inside Plant-based food materials [5].
Fig. 2. Domains considered in existing literature [8e12].
M.I.H. Khan et al. / International Journal of Thermal Sciences 117 (2017) 266e273 267

2. Material and methods
2.1. Sample preparation and drying
Experiments were performed on the potato (Eureka) samples
collected from a local market in Brisbane, Australia. Samples were
stored in a refrigerator at 4
C until the start of drying experiments,
which were performed in a cabinet dryer. The dryer was started
about 30 min before drying experiments to achieve steady state
conditions before each drying run. At the start of each experiment,
the materials were washed and cut into cylindrical slices of 30 mm
thickness and 20 mm diameter. Experiments were performed at a
temperature of 60
C and a constant air velocity of 0.7 m/s. The
relative humidity of the inlet air was in the range of 60%e65%.
During the drying process, the tray was taken out at 30-min in-
tervals and weighted using a digital electronic balance (model
BB3000; Mettler-Toledo AG, Grefensee, Switzerland). The mea-
surement range of the balance was 0e100 g with an accuracy of
0.01 g. The temperature was measured by a thermal imaging
camera Flir-i7. The temperature range of the thermal camera
was 20
C to 250 C with 140 140 pixels resolution. In order to
better interpret and analyse the results, a simulated 3D tempera-
ture distribution profile was developed (Fig. 7) using a validated
heat transfer model. The details of the theoretical model, from
where simulation results were generated, can be found in authors'
previous publications [25,26].
2.2. NMR measurements
At different stages of the drying process, the samples were taken
out of the dryer and immediately placed in a 25 mm diameter NMR
tube. To protect the sample from oxidation, it was immersed in
Fomblin PFPE (Grade 06/6, USA) oil to provide a barrier for mass
transfer between samples and surroundings and to assist with
sample shimming. The tube was sealed with a standard NMR tube
cap and incubated at 22
C for 5 min to reach thermal equilibrium.
Measurements were made with a Bruker DRX wide-bore spec-
trometer (Bruker Bio-spin, Karlsruhe, Germany) operating at
300 MHz for hydrogen and fitted with a micro-imaging (micro 120)
gradient set and a birdcage coil. Data were collected and processed
using Paravision 4 software (Bruker). Regional T
2
relaxation times
were measured from relaxation maps acquired with a multi-slice-
multi-echo (MSME) sequence using the following acquisition pa-
rameters: 64 averages, 1000 echoes with 10 ms echo time and 5.0 s
repetition time. The slice thickness and the matrix size were 3 mm
and 6464, respectively. The spatial resolution of the scans was
468
m m.
2.3. Mathematical analysis
2.3.1. Theory
The water mobility in potato tissue was investigated quantita-
tively using multicomponent analysis of T
2
relaxation decay curves.
A nonlinear least-squares method was applied for data analysis
[27]. Generally, the free induction decay of the proton relaxation
follows an exponential decay. For multiple environments, there will
be corresponding T
2
relaxation time constants. A multi-exponential
equation can describe these functions. Each tissue compartment
corresponds to a different environment and will have a distinct
relaxation time constant (T
2
). It was assumed that these compart-
ments were not inter-reliant at the time of the measurements such
that the multi-exponential nature of the T
2
decay curve relates to
the different water compartments in the tissue, and the water
molecules do not undergo rapid exchange between compartments
on the NMR timescale. However, when water proton relaxation
follows a pattern of mono-exponential decay, there is a fast ex-
change of protons between tissue water and macromolecules,
showing that the water compartments are symbiotic [28e30].
2.3.2. Data analysis
For each sample, two regions-of-interest (ROI) were defined
manually on a given slice of the mid transverse section of the MSME
images [6]. The mean value of these ROIs was computed for each
sample. The accuracy of the multi-exponential parameters
considerably depends on the noise level; therefore, a third ROI was
assessed outside the sample for determining the signal-to-noise
ratio (SNR, average signal intensity over the standard deviation of
the noise). Noisy T
2
-signals were eliminated from the original
signal, as described in authors' previous publications [6,31], and
data with appropriate SNR (SNR>5) were used for parameter
fitting. In order to achieve SNR >5, a constant cut-off of TE <600 ms
was used for all datasets [32]. The mean signal of the two ROIs at
each echo time were measured and plotted as a function of time on
a semi-log scale where curvature was determined [33]. The number
of inflection points was used to determine the number of T
2
com-
ponents. The different T
2
relaxation times comprising the signal
from each sample was determined by bi-exponentially fitting the
mean signal intensity using the following equation.
Y¼A1expt=T
1
2
þA2expt=T
2
2
(1)
where, Yis the function of T
2
relaxation time constant, A
1
and A
2
represent the relative contributions of the two proton environ-
ments, and T
1
2
,T
2
2
are the relaxation time constants of the different
components. For T
2
measurements, odd echoes were excluded to
minimize error due to the influence of stimulated echoes [6,34,35].
T
2
relaxation time data processing was carried out with a non-
negative least squares algorithm using self-written program code
in the curve fitting toolbox of Matlab
®
software (The MathWorks,
Inc., Natick, MA). The experiments were replicated three times for
each sample, and the average of these measurements was used for
analysis. The details of the experimental procedure can be found in
the authors' previous publication [6].
2.4. Statistical analysis
Data are expressed as mean ±SD of the mean. For statistical
analysis least-squares linear regression analysis was used. A 95%
confidence level was regarded as significant, as shown in Table 1.
3. Result and discussions
The water migration mechanisms at different stages of drying
were investigated using
1
H-NMR relaxometry. The T
2
relaxation
decay curves obtained from samples at different drying times are
presented in Fig. 3. Each curve was fitted using the multi-
exponential decay equation (1). After fitting the bi-exponential
decay curve with different T
2
relaxation intensity data, two
different proton components of water relaxation (long and short)
were determined, as shown in Table 1, with their corresponding
standard deviation and statistical confidence level. Depending on
the water mobility and pore size, two signal components namely,
long and short were categorised as intracellular environmental
water (BW) and intercellular environmental water (FW) [6].
3.1. Collapse of the cell membrane during drying
Fig. 4 shows the change in the proportion of BW over time
during the drying process, calculated on the assumption that the
total moisture present at that time is considered 100%. The curve is
M.I.H. Khan et al. / International Journal of Thermal Sciences 117 (2017) 266e273268

divided into four zones according to the cell membrane collapse
phenomenon. In Fig. 5, these zones are presented separately for
clearly showing the cell membrane rupture points. For each zone,
the peak of the BW curves is represented as cell membrane
collapse point. It can be seen in the figure that the duration of the
Zone - 1 is approximately 100 min. In this zone, BW is about
80e85% and this proportion remains constant up to 55 min of
drying, as shown in Fig. 5a. This is because only free water mi-
grates at the initial stage of drying. After 55 min of drying, the
curve drops sharply to a BW level of 50e55%. In order to interpret
the zone, an average surface temperature profile and 3D tem-
perature distribution profile have been presented in Figs. 6 and 7
respectively.
A 3D coupled heat and mass transfer model for convective
drying of food materials was developed earlier by the present
authors [25,26]. COMSOL Multiphysics, finite element-based en-
gineering simulation software was used to simulate the model in
2D and 3D and to solve the coupled heat and mass transfer
equations. That model was validated using comprehensive
experimental data. The details of the model can be found in au-
thors' previous publications [25,26].Fig. 7 presents simulated
temperature profile of the sample, which was developed using
above validated heat transfer model.
From Fig. 7a, it can be observed that after 55 min of drying the
temperature at the surface rose to 53e54
C, and the heat energy
penetrated slowly towards the centre of the sample. It has been
reported that cell membranes start to collapse when the drying
temperatures are at or above 50
C[4]. It has also been stated that
the cell membrane collapses due to the thermal stress which are
induced by the temperature and pressure gradients [15].As
mentioned earlier, it can be assumed that when the temperature
reaches 52e53 C, the cell membranes close to the surface of the
samples will start to rupture first. As a result, the exposed BW
(intracellular water) is able to readjust rapidly to become part of
the intercellular space (i.e. FW) which is registered as a rapid BW
drop (Fig. 5a).
Within the next drying Zone, the FW is gradually decreasing
due to its migration up to the surface. According to Vasi c et al.
[36], this internal mass transport is controlled by several drying
mechanisms. Consequently, the relative proportion of BW
(compared to FW) increases during this period. The total water at
any point (instantaneous moisture) is assumed as 100%, although
the average moisture content is decreased gradually (Fig. 4, zone
2). This process continues up to 170 min of drying. Simulta-
neously, with the previously mentioned process, heat is contin-
uously propagated towards the centre of the drying sample
(Fig. 7). At that time the surface temperature increased to about
58
C(Fig. 6), whereas inside the sample the temperature was
about 56e57
C as shown in Fig. 7c. The sample surface temper-
ature increased up to 58
C at the end of the Zone 2, while the
temperature inside the sample was 2
C lower (see Figs. 7c and 6).
As a result of this temperature difference, the cell membrane
ruptured again and the BW curve dropped rapidly as it was
converted to FW (Fig. 5b).
Within the next drying Zone (Zone 3) the internal moisture
transport of the FW to the surface is continued. In other words,
FW is gradually decreasing (see Fig. 4). Consequently, the relative
proportion of BW (compared to FW) increases during this period.
The majority of cell membranes are damaged at the end of the
Zone 3 (after 180 min) when the temperature of the entire sample
has reached 59.5
C (above cell collapse temperature). This point
is registered at Fig. 5c as the cell membrane collapse point. From
this point up to the end of the Zone 3 gradual decrease of BW is
registered (see Fig. 4) as BW is transferred into FW. The final cell
collapse is registered in the Zone 4 (after 350 min), as shown in
Table 1 The different T
2
component at different drying times, each measurement is the mean and SD of 3 samples.
Drying time (min) Long component Short component Goodness of fit(R
2
)
T
2
values Contribution (A) T
2
values Contribution (A)
T
2
(ms) 95% CI (%) 95% CI T
2
(ms) 95% CI % 95% CI
Fresh Potato 98.92 ±10.2 97.02 79.92 ±6.8 100.80 26.47 ±6.2 22.29 20.08 ±4.2 30.65 0.9999
30 100.75 ±11.2 82.50 81.98 ±8.2 87.01 24.25 ±5.8 28.07 18.02 ±3.5 32.92 0.9999
60 92.32 ±9.5 90.84 77.25 ±7.3 93.80 29.25 ±3.9 18.47 22.75 ±3.8 24.62 0.9999
90 118.21 ±10.8 104.70 50.2 ±6.4 131.70 65.72 ±9.9 56.66 49.8 ±5.6 74.81 0.9999 120 98.87 ±10.3 97.23 76.22 ±5.5 100.50 38.87 ±7.5 35.90 23.78 ±3.5 41.85 0.9999
150 109.92 ±12.2 106.20 85.49 ±7.5 113.70 37.99 ±6.8 27.27 14.51 ±2.8 48.72 0.9998
180 138.87 ±14.4 137.00 84.58 ±7.2 140.70 32.83 ±6.4 28.40 15.42 ±2.7 37.28 0.9997
210 127.02 ±10.8 124.30 50.72 ±5.3 129.70 48.69 ±5.9 47.16 49.28 ±6.2 50.21 0.9994
240 150.84 ±15.3 148.20 82.03 ±6.8 153.40 60.80 ±8.9 55.09 17.97 ±2.9 66.50 0.998
270 104.78 ±11.3 101.70 41.77 ±4.8 107.90 32.57 ±7.7 31.27 58.23 ±6.8 33.86 0.9997
300 149.92 ±13.2 139.80 46.22 ±5.7 160.00 74.87 ±11.2 69.88 53.78 ±5.8 79.87 0.9998
360 107.74 ±10.4 105.80 47.05 ±5.3 109.60 43.50 ±8.5 42.50 52.95 ±4.9 44.50 0.9998
420 332.75 ±16.2 76.30 7.02 ±3.5 318.70 95.42 ±10.2 86.15 92.98 ±8.2 102.30 0.9996 450 532.70 ±19.4 103.20 0.1 ±2.4 134.30 98.21 ±12.4 43.71 99.9 ±10.2 53.97 0.9995
M.I.H. Khan et al. / International Journal of Thermal Sciences 117 (2017) 266e273 269

Fig. 5d. The sample surface and internal temperature have reached
60
C (see Fig. 7d) in the final collapse point. From this moment up
to the end of the Zone 4 progressive decrease of BW is registered
(see Fig. 4).
Detail analysis of Figs. 4 and 5 showed that the Zone 2 lasts
longer than other three Zones. Initially, cell membranes close to the
sample surface will be the first to rupture (zone 1). The rest of the
tissue (i.e. cells far away from the surface) gains energy slowly prior
to developing thermal stresses. As a result, in zone 2, the cell
membrane takes more time to collapse. On the other hand, at zone
3, the cell membranes inside the sample have already absorbed
sufficient heat energy to develop higher thermal stresses. Conse-
quently, the cell membranes in zone 3 will collapse faster (Fig. 5c).
As a final conclusion, it can be stated that the cell membranes in
plant-based food material collapse in different stages of drying
rather than collapsing altogether. In the case of conventional drying
the cell rupture mainly depends on the rate of penetration of heat
energy as well as the resulting pressure gradient between intra-
cellular and intercellular spaces for conventional drying.
3.2. Moisture distribution
Several NMR images showing the moisture distribution profile
at different drying times are presented in Fig. 8. The lighter parts of
the NMR images are related to the higher water content while
darker parts are associated with the dried sample segments. It is
interesting to note that the moisture distribution in fresh potato
tissue is uniform throughout the whole sample (see Fig. 8a). At the
Fig. 4. The change in percentage of free and bound water with drying time.
0
5000000
10000000
15000000
20000000
25000000
30000000
15 65 115 165 215 265 315 365 415 465 515 565 615
yti
sn
etn
I
Echo time (ms)
Fresh
After drying of 30 min
60 min
90 min
120 min
150 min
180 min
210 min
240 min
270 min
300 min
360 min
420 min
450 min
Fig. 3. T
2
relaxation at different drying time.
M.I.H. Khan et al. / International Journal of Thermal Sciences 117 (2017) 266e273270