Powder Technology 162 (2006) 133  137
www.elsevier.com/locate/powtec
Mapping of temperature distribution in pharmaceutical
microwave vacuum drying
a, b a c d
*
Á. Kelen , S. Ress , T. Nagy , E. Pallai , K. Pintye-Hódi
a
Formulation Development, Richter Gedeon Ltd., Budapest 10., P.O. B.27 H-1475 Budapest, Hungary
b
Department of Electronic Devices, Technical University of Budapest, H-1521 Budapest, Hungary
c
University Veszprém, Research Institute of Chemical and Process Engineering, H-8200 Veszprém, Hungary
d
Department of Pharmaceutical Technology, University of Szeged, H-6720 Szeged, Hungary
Received 29 November 2004; received in revised form 14 September 2005
Available online 7 February 2006
Abstract
Microwave vacuum drying is getting more and more popular thanks to its known advantageous features. In spite of its uniqueness, there is a
rightful resistance and mistrust because of the nonhomogeneous electric field that may cause nonhomogeneous temperature distribution in the
workload. In practice the best uniformity of power density and the shortest drying time are sought simultaneously, thus the drying method is close
to its secure limit. Control and monitoring of a running process remains unsolved but even experimental mapping is rather circuitous. The
dielectric and thermal properties of a complex pharmaceutical composition are rarely known, and moreover, they change during a drying process,
which makes accurate mathematical modelling rather uncertain. For that very reason preliminary tests can never be neglected. The aim of our
study is to experimentally map and evaluate the heat distribution quantitatively. To get a 3D overview of a free-flowing bulk, the workload was
divided with Teflon layers to form cross-sectional surfaces. After dissipation of microwave energy, IR thermocartograms were taken and the
temperature distribution was evaluated quantitatively. The   3D layered thermography  method offers reliable and workload-specific information,
via a simple executable technique, for optimization of a microwave assisted drying process.
© 2005 Elsevier B.V. All rights reserved.
Keywords: Pharmaceutical; Microwave; Homogeneity; Thermography
1. Introduction and the temperature gradient. The use of microwaves can be
risky if the thermal diffusivity of the workload is low, in which
1.1. Dielectric heating case the heat flow is slower than the rate of energy dissipation.
When drying of extremely fragile corn starch-based
The benefits and drawbacks of microwave drying in the granules (6.3 kg) was carried out under vacuum (50 T 5 mbar)
pharmaceutical industry have been well known for decades. In and accelerated by microwaves (1.2 kW, 2450 MHz) in a
spite of the fact that dielectric drying offers unique advantages single/one pot unit (Collette Ultima 25 l; a high-shear
[1] the biggest resistance to widespread use may be the non- granulator that incorporates vacuum and microwave drying
uniformity of the electromagnetic field (E-field), which results options), local burning was experienced after 25 min of
in a nonhomogeneous temperature pattern [2,3]. The origin and microwave heating. In the case of dielectric heating the
result of a generated hot-spot is influenced by the electromag- location and temperature of hot spots are unpredictable,
netic and thermodynamic features of the microwave system because of many factors, which influence the uniformity of
and the workload. Hotter areas are cooled by heat diffusion to the E-field [4].
the surrounding material, determined by the thermal diffusivity To avoid the undesirable unequal temperature distribution
there are several possible solutions, e.g. intensification of the
mixer motion, and/or reduction of the microwave power. The
former would change the grain-size distribution unacceptably,
* Corresponding author.
E-mail address: [e-mail] (Á. Kelen). while the latter would considerably increase the process time.
0032-5910/$ - see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2005.12.001
134 Á. Kelen et al. / Powder Technology 162 (2006) 133  137
As a result of experimental process optimization the best distribution and its effect after starting the microwave
uniformity and highest acceptable power density, as well as the treatment. With the help of mathematical models based on
shortest drying time are sought, without any corresponding Maxwell s equations the theoretical electric and magnetic field
damage of the workload. configuration within the product can be calculated [8,10] even
The aim of the present study is to experimentally map and to in 3D [11,12], if the configuration of the cavity, the dielectric
evaluate the forming stereoscopic temperature pattern of a free- properties of the workload and the granule geometry, etc., are
flowing bulk workload. This paper focuses on the macroscopic exactly known. The following dielectric heating equation is
temperature distribution, and not on the molecular/microscopic also used to calculate the dissipated microwave power ( Pd,
level [5]. [W]) [8]:
2
Pd ź 2pf e0eVVEi ð1Þ
1.2. Thermography
where f = microwave frequency [Hz]; (0 = free-space or abso-
Among the non-perturbing, and non-intrusive temperature-
lute permittivity 8.854 10 12 [F/m]; eVV= loss factor of the
monitoring alternatives, infrared imaging is known as one of
dielectric material [dimensionless]; Ei= electric-field strength
the most promising. Thermal imaging relies on the fact that all
within the dielectric [V/m].
bodies emit electromagnetic radiation due to electronic
The internal energy (U, [J]) of the product being dried in
oscillation and the radiated energy is proportional to the
microwave oven changes with the absorbed (dissipated)
temperature of the object. The unique advantage of IR
microwave energy (Ed, [J]). Based on the first law of
monitoring is that it does not disturb the drying and a huge
thermodynamics, temperature is considered as an indicator of
quantity of data can be recorded digitally and displayed
E-field. The change in the internal energy can be expressed by
instantly [6]. The limitation of IR monitoring is that it provides
the following relations:
information exclusively about the monitored surface. Ohlsson
et al. [7] made cross sections of solid objects and used thermal DU ź RQ RW ð2Þ
imaging to get 3D information about their temperature
distributions.
DU ź Qsolventðt YtBPÞ þ Qsteamðt YtendÞ þ Qsolidðt YtendÞ
start BP start
Wvol þ Wevap ð3Þ
2. Theoretical aspects
DU ź csolventmsolventDTðt YtBPÞ þ csteammsteamDTðt YtendÞ
2.1. Nonhomogeneity of microwaves
start BP
þ csolidmsolidDTðt YtendÞ pDV þ Lvmsolvent ð4Þ
start
Field concentration of standing waves at close proximity to
the power-feed-point can cause non-uniform distribution of the
DU ź csolventmsolventDTðt YtBPÞ þ csolidmsolidDTðt YtendÞ
microwave field [8]. Many factors influence the uniformity of start start
the E-field. They can be divided roughly into two groups:
þ Lvmsolvent ð5Þ
cavity effects (design limitation, location of the microwave
inlet point, shape of the cavity, hanging parts such as spray where Q = quantity of heat [J]; W = work [J]; Qxx(t tzz) =
yy Y
gun, mixer, chopper, thermometer) and workload interactions quantity of heat of the indicated material in the given
(loss factor, penetration depth and thickness of the workload, temperature range [J]; Wvol = volumetric work [J]; Wevap = eva-
particle features, etc.), that are different from product to poration work [J]; cxx = specific heat capacity of the indicated
product and from equipment to equipment [4]. material [J/kg K]; mxx = mass of the indicated material [kg];
Inter alia   mechanical moving mode stirrers  or   wave- DT(t tzz)= temperature difference between the indexed events
yy Y
guide rotating joints  or simple agitation of the workload are [K]; Lv = heat of vaporization [J/kg].
used to assure more uniform E-field distribution, and thus The energy dissipation of the steam, which is present in the
heating. Adequate homogeneity can be achieved, e.g. in a cavity, is considered negligible due to its small amount
developed microwave applicator, characterised by cylindrical (msteam å 1 g). There is no volumetric work (DV å 0). The
shape and adjusted with several magnetrons [9]. In the case of change in the internal energy during microwave drying can be
special single/one pot pharmaceutical microwave equipment, calculated on the basis of the dissipated microwave power ( Pd,
the number and position of magnetrons is very restricted due to [W]) and the microwave treatment time (t, [s]).
its primary functional purpose, thus agitation of the workload is
DU ź Ed ź Pdt ð6Þ
preferred. In the case of drying of extremely fragile granules,
any type of mechanical movements endanger the quality of the
DT¨DU¨Pd¨Ei ð7Þ
product.
Microwaves are not forms of heat, but rather forms of
2.2. Theoretical models energy that are manifested as heat through their interaction with
materials. There is a two-step energy conversion: electric field
An inherent deficiency of dielectric drying is that there is no is converted to induced ordered kinetic energy, which in turn is
common method to control, or properly monitor, the E-field converted to disordered kinetic energy, at which point it may be
Á. Kelen et al. / Powder Technology 162 (2006) 133  137 135
120
regarded as heat within the material [13]. In accordance with
100
the aforementioned, it can be stated that the change in the
product temperature is proportional to the change in the
80
internal energy of the material and to the dissipated power
60
thus to the electric-field strength within the dielectric.
40
For determination of the dissipated microwave power ( Pd,
20
[W]), a special instrumental set-up [14] would be required to
0:00 0:07 0:14 0:21 0:28
measure the magnetron output power ( Pm, [W]), the reflected
time [min]
power ( Pr, [W]) and all the losses that are evolved in the set-up
(e.g. losses by the direction coupling, by fitting attenuation).
Fig. 2. Temperature  time curve of cornstarch. The contact thermometer fitted
into the chopper arm (Pt100) (>) was operating during the running process,
Based on the measured reflected microwave power, the
hot-spot surrounding temperature (?) is measured immediately after micro-
dissipated microwave power could be calculated by the
waves were switched off (6.3 kg, 50 mbar, 1.2 kW, 2450 MHz). (Three
following equation:
replications were used to generate each data point.)
Pd ź Pm BPr ð8Þ
controlled at 6 T 1 -C during the drying processes. Circulation
where B = comprises the different attenuations and losses
in the cooling system was stopped, thus the double-jacket was
[dimensionless].
heated up exclusively by thermal conduction.
Theoretical models are always limited by generalization and
After 25 min of microwave radiation carbonised dots were
simplification. Calculations consider the workload homoge-
detected on the surface. Because of carbonisation, the temper-
neous from dielectric, thermal and other point of views
ature of the local hot spots must have been over 200 -C,
although it is generally not the case. Especially when drying
according to the thermogravimetric analysis of cornstarch.
complex pharmaceutical compositions, the workload may
During the study the samples were heated slowly and
consist of several ingredients, characterised by different and
continuously during ¨100 min, from 20 -C to 250 -C (T curve
often unknown dielectric and thermal properties. Moreover
in Fig. 1). According to the TG (thermogravimetry) and DTG
these are continuously changing during a drying process, not
(differential thermogravimetry) curves there are three tempera-
only in time but also in 3D, and depending on many factors
tures (T1, T2 and T3) where the mass of the sample changes
(e.g. moisture content and temperature) [15,16]. For that very
because of thermodynamic effects (Fig. 1). Knowing the
reason, experimental tests are much more reliable.
structure and chemical behaviour of corn starch it can be stated
that the first two peaks correspond to loss of free water (from
3. Experiments (  3D layered thermography 
T1: ¨40 -C to T2: ¨145 -C). The third peak (T3: over 200 -C)
mapping method)
refers to the conspicuous carbonization (decomposition).
At 25 min radiation time the contact thermometer fitted into
In the experiments the steady-state workload was corn
the chopper arm (Pt100 in Fig. 4) measured around 60 -C while
starch (Ph.Eur., Roquette GmbH, Germany), a common
there were carbonized spots (Fig. 2).
pharmaceutical diluent. Its density was found ~560 kg/m3 that
The measured temperature difference between the two areas
does not influence considerably the drying process [17].
was significant. Therefore more detailed mapping was essential
During the drying process the initial moisture content was
in order to get reliable information about the temperature
13% (wet weight based) and the temperature was 25 -C.
distribution within the entire workload. Teflon (PTFE) disks of
The workload (6.3 kg, 2/3 of the total capacity) was heated
1 cm thickness were used to divide the workload horizontally
by microwaves at 1.2 kW (2450 MHz) under a pressure of
into six layers of 2 cm thickness. Teflon was chosen because it
50 T 5 mbar in a single pot system (Collette Ultima 25 l,
does not absorb microwave energy [8]. Between the layers tiny
Collette NV, Belgium) (Fig. 4). Initially, the temperatures of the
double-jacket of the cavity and the workload were tempered at
100
25 T 1 -C for 60 min. The temperature of the condenser was
90
80
70
60
50
40
0,0 0,5 1,0 1,5 2,0 2,5 3,0
time, [min]
Fig. 3. Transient temperature of cornstarch at different initial temperatures. The
long-wave emission constant of corn starch is found at 0.95 (in harmony with
Fig. 1. T (=temperature) and TG (=thermogravimetry) curves of corn starch. the data given in the user manual of the IR camera). (Three replications were
T1, T2 and T3 indicate the temperatures were the mass of the sample changed. used to generate each data point.)
T [
°
C]
T, [
°
C]
136 Á. Kelen et al. / Powder Technology 162 (2006) 133  137
Table 1
Teflon distance pieces assured an even density of cornstarch,
The quantitative evaluation of the temperature distribution based on   layered
because it is known that the loss factor depends on the
thermography 
characteristic bulk density [18]. After 25 min drying time, the 6
nx, layer no.
cornstarch layers were immediately monitored one by one by
Ti * [C] 1 2 3 4 5 6 Mi * [%]
an infrared camera (AGA782 Infrared Imaging System, y y
Infrared System AB, Sweden). The six snapshots were taken mn iy* [%]
x
within 1 min of the microwave being switched off. Transient IR
<25 16.3 3.2     3.2
25  30 21.9 33.9 37.3 34.9 31.3 13.9 28.9
30  35 19.9 39.1 39.1 37.7 34.9 39.1 34.9
35  40 20.7 10.5 18.0 20.7 26.4 38.3 22.4
40  45 6.1 7.2 3.1 5.0 5.7 8.6 5.9
45  50 3.2 3.5 0.7 1.0 1.8  1.7
50  55 2.2 0.7 0.5 0.5   0.6
55  60 1.6 0.6 0.5 0.2   0.4
60  65 1.4 0.4 0.6    0.4
65  70 1.2 0.3 0.1    0.2
70  75 1.0 0.2     0.2
75  80 0.8 0.2     0.2
80  85 0.6 0.2     0.1
85  90 0.6 0.4     0.2
90  95 0.6      0.1
>95 2.1      0.4
Å» Å»
Tn [-C] 36.7 33.2 32.4 32.6 33.1 34.6 T: 33.7
x
Å»
nx =number of a layer; Ti = temperature range [-C]; Mi * = percent of the total
y y
material in the iy temperature range within the whole workload [%];
mn iy* = percent of the material in the iy temperature range within the nx layer
x
Å» Å»
[%]; Tn =average temperature of the nx layer [-C]; T = average temperature of
x
the total workload [-C]). (Three replications were used to generate each data.)
snapshots [19] prove that the temperature decrease between the
end of the microwave treatment and taking the thermograms is
negligible (DTmax found 1 2 -C) (Fig. 3).
Thermography offers immediate coloured 2D images (Fig. 4)
about the heating pattern of each layer that can be characterised
and quantified.
4. Results and discussion
After 25 min of microwave treatment carbonised hot spots
are detected. Their surrounding temperature was measured and
found to be around 100 -C meanwhile the temperature of
another area is found around 60 -C (Pt100 thermometer).
According to the thermal analysis of corn starch (Fig. 1) it can
be stated that the carbonised hot spots are heated up to a
minimum of 200 -C during the drying process. A significant
difference can be found between the temperature in the vicinity
of the hot-spots and the contact thermometer fitted into the
chopper arm (Fig. 2). The experiments prove the non-uniform
temperature distribution that originates from the non-uniform
E-field distribution.
The colours of the thermocartograms refer to the tempera-
ture of the area, in accordance with the given colour scale,
which makes qualitative analysis possible. Two hotter areas are
seen directly under the microwave inlet window. The asym-
metric temperature pattern on the surface is the consequence of
Fig. 4. The single/one pot equipment. The workload is divided by Teflon disks
the microwave inlet position and the vertical temperature
into six horizontal layers. The six thermocartograms are in order of location:
decrease is in agreement with penetration depth concepts
layer No. 1 is the upper and No. 6 is the base of the pot. Diameters of the disks
(Fig. 4). The amplitude of microwave diminishes owing to
follow the geometry of the bowl: Nos. 1, 2, 3 and 4 are 0.37 m, No. 5 is 0.34 m
absorption of power as heat in the material [8].
and No. 6 is 0.22 m.
Á. Kelen et al. / Powder Technology 162 (2006) 133  137 137
To get more detailed 3D information about the temperature ature differences but the prevention of its developing is more
thus E-field distribution of a free flowing workload the serviceable from the product quality point of view.
obtained thermocartograms are evaluated with the under- The presented   3D layered thermography method  makes
mentioned   layered thermography  technique (Table 1). One possible to map the temperature distribution in a free-flowing
layer is represented by some 24 000 pixels in the thermo- bulk and it also provides quantitative 3D information. Based on
cartograms. One pixel corresponds to 1.2 mm2 of the corn the known temperatures of the identified hottest areas the
starch layer. The mass (mn iy, kg) of bulk characterised by a drying process can be regulated to keep the temperature of the
x
chosen temperature range (iy: i25  30, . . ., i95  100) in a layer endangered areas below the critical limit.
(nx: n1, . . ., n6) can be calculated by the camera-detected
surface area (An iy, m2) (the sum of the surfaces according to
Acknowledgements
x
the number of the pixels), the known thickness (dn = d = con-
x
stant, 0.02 m) and the density (qn = q = constant, ¨560 kg/
The authors thank László CserÅ„ak, Attila Bódis and
x
m3) of the product layer. Based on preliminary tests, the
Andrásné Kucsera (Gedeon Richter Ltd.).
vertical temperature difference within the corn starch layers of
0.02 m thickness was never greater than 1 2 -C, thus the
References
temperature of the whole amount under a given surface can
be characterised by the surface temperature. [1] M.E. Aulton, Radiation drying of wet solid, Pharmaceutics, 2nd edR,
Churchill Livingstone, London, 2002, pp. 386  388.
mn iy ź An iydq ð9Þ
[2] A.F. Harvey, Microwave Engineering, Academic Press, New York, 1963.
x x
[3] A.C. Metaxas, R.J. Meredith, Dielectric properties, Industrial Microwave
The percentage of material (mn iy*, %) characterized by an
x Heating, 2nd edR, Peter Peregrinus Ltd., London, 1988, p. 57.
iy temperature range within the nx layer can be calculated from
[4] R. Meredith, Introduction and fundamental concepts, in: Engineers
handbook of industrial microwave heating, The Institution of Electrical
the mass in question (mn iy, kg) and the mass of the entire layer
x
Engineers, London, 1998, pp. 7  13, pp. 19  35. and pp. 163  176.
(mn , kg) (in other words from the number of pixels in question
x
[5] J. Berlan, Microwaves in chemistry another way of heating reaction
and the sum number of pixels in the entire layer):
mixture, Radiat. Phys. Chem. 45 (4) (1995) 581  589.
mn iy [6] J. Bows, K. Joshi, Infrared imaging feels the heat in microwave ovens,
x
m* ź I100 ð10Þ
nxiy Phys. World 5 (1992) 21  22.
mn
x
[7] T. Ohlsson, P.O. Risman, Temperature distribution of microwave heating
spheres and cylinders, J. Microw. Power Electromagn. Energy 13 (4)
The percentage of total material (M*, %) characterized by an
iy
(1987) 303 310.
iy temperature range within the whole workload (M, %) is:
[8] G. Duschler, W. Carius, K.H. Bauer, Single-step granulation method with
microwaves preliminary studies and pilot scale results, Drug Dev. Ind.
Mi
y
M* ź I100 ð11Þ
Pharm. 21 (14) (1995) 1599  1610.
iy
M
[9] J. Suhm, M. Möller, Aufbau von Mikrowellenanlagen Proc. OTTI-
Å» Profiforum Mikrowellen-Thermoprozesstechnik Grundlagen, Anlagen
The average temperature of the nx layer (Tn , -C) can be
x
und Anwendungen, Deutschland, November, 2003.
calculated from the percentage of the material characterized by
[10] S. Zhu, Y. Zhang, T.R. Marchant, A DRBEM model for microwave
an iy temperature range within the nx layer (mn iy*, %)
x
heating problems, Appl. Math. Model. 19 (5) (1995) 287  297.
multiplied by the mean temperature of the incidental iy [11] V.A. Mechenova, V.V. Yakovlev, Efficiency optimization for systems and
Å»
temperature range (Ti , -C).
components in microwave power engineering, J. Microw. Power Electro-
y
magn. Energy: Publication of The International Microwave Power
Å» Å»
Institute 39 (1) (2004) 15  29.
T T
T ź R m* IT =100 ð12Þ
nx nxiy iy
[12] A. Hallac, A.C. Metaxas, Finite element time domain analysis of
microwave heating applicators using higher order vector finite elements,
The presented   3D layered thermography  technique offers
Proc. 9th International Conference on Microwave and High Frequency
quantitative information about the temperature distribution of a
Heating, UK, Sept., 2003, p. 021.
free-flowing workload in a simple way. It is not suitable for
[13] R.F. Schiffmann, Microwave and dielectric drying, Handbook of Industrial
monitoring the deeper layers directly during a running process,
Drying, 2nd edR, Marcel Dekker, Inc., New York, 1955, pp. 345  372.
[14] A. Göllei, L. Ludányi, E. Pallai, A. Vass, E. Szijjártó, Power engineering
but based on the confirmed reproducibility of the E-field
study of combined microwave and convective drying, Proc. 6th
distribution (analysis of variance proves the reproducibility;
International Conference on Food-Science, Hungary, May, 2004, (CD).
p > 0.05), it makes the modelling of a real microwave assisted
[15] M.K. Ndife, G. Sumnu, L. Bayindrili, Dielectric properties of six different
drying process possible even in the case of a pharmaceutical
species of starch at 2450MHz, Food Res. Int. 31 (1) (1998) 43  52.
formula characterised by unknown and continuously changing
[16] C.M. McLoughlin, W.A.M. McMinn, T.R.A. Magee, Physical and
dielectric properties of pharmaceutical powders, Powder Technol. 134
features.
(2003) 40  51.
[17] C. Péré, E. Rodier, Microwave vacuum drying of porous media
5. Conclusions experimental study and qualitative considerations of internal transfers,
Chem. Eng. Process. 41 (2002) 427  436.
[18] S. Nelson, Density dependence of dielectric properties of particulate
In the case of dielectric drying the non-uniform temperature
materials, Trans. ASAE 26 (1983) 1823  1825.
distribution in the workload is the consequence of the
[19] V. Szekely, S. Ress, A. Poppe, S. Torok, D. Magyari, Zs. Benedek,
electromagnetic field interaction with the different dielectric
K. Torki, B. Courtois, M. Rencz, New approaches in the transient
materials. Some techniques are known to lessen this temper- thermal measurements, Microelectron. J. 31 (2000) 727  733.