Journal of Food Engineering 44 (2000) 213±223

www.elsevier.com/locate/jfoodeng

Simulation and experimental investigation of food material breakage

using pulsed electric ®eld treatment
N.I. Lebovka a,b, M.I. Bazhal a,c, E. Vorobiev a,*
a
Departement de G enie Chimique, Universit
e de Technologie de Compiegne, Centre de Recherche de Royallieu, B.P. 20529-60205,
Compiegne cedex, France
b
Institute of Biocolloidal Chemistry named after F.D. Ovcharenko, NAS of Ukraine, 42, blvr. Vernadskogo, Kyiv 252142, Ukraine
c
Ukrainian State University of Food Technologies, 68, Volodymyrska str., Kyiv 252033, Ukraine
Received 16 July 1999; accepted 17 January 2000

Abstract
We consider the simpli®ed dielectric breakage model used for simulation of the kinetics of food material breakage under pulsed
electric ®eld (PEF) treatment. The model is based on an e€ective media approximation, which includes equations with the same
morphology parameters as in percolation theory. The probability of a whole cell breakage by a pulse of ti duration is estimated on
the basis of electroporation theory. We account for the bridging e€ect resulting from the deviations of the local conductivity near the
selected cell from the average e€ective media conductivity. The most important feature of the proposed model is the existence of the
``jamming'' behaviour occurring sometimes in experimental observations of the biological tissue breakage. The di€erent transitions
corresponding to the ``jamming'' steps are identi®ed. Experimental results were obtained for thin apple slices treated with electric
pulses at ®eld strengths E ˆ 0.2±2.2 kV cmÿ1 , pulse durations ti ˆ 10±100 ls, pulse repetition times t ˆ 10±100 ms and the number of
pulses n ˆ 1±100 000. The model gives results consistent in general with the experimental observations. We discuss the correlation
between the degree of food material destruction, ®eld strength, time of PEF treatment and power consumption. Ó 2000 Elsevier
Science Ltd. All rights reserved.

Notation Nt total number of cells

Cm speci®c capacity of membrane, F mÿ2 Nx ; Ny ; Nz dimensions of a lattice
C ˆ Cm …ew =em ÿ 1†=…2c† P degree of biological tissue destruction
C speci®c heat of food material, J kgÿ1 Kÿ1 Pl local degree of cell destruction
dm membrane thickness, m Pp critical percolation degree of destruction
dc cell diameter, m rl averaged local cell resistance, X
E electric ®eld strength, kV cmÿ1 rm;c;i ˆ dm;c;c =…Sm rm;c;i †, resistances of membrane,
f ˆ u=dc E (at a ˆ 1), normalised cell intracellular ¯uid and intact cell,
transmembrane voltage respectively, X
DF  ˆ px2 =…kT c†, reduced critical free energy of R total resistance of a linear chain, X
pore formation Sm  dc2 , cross-section area of a cell or a
g breakage probability for a membrane membrane, m2
j current density, A mÿ2 ti pulse duration, ls
k Boltzmann constant, Dt pulse repetition time, ms
1:3806581210ÿ23 J Kÿ1 T temperature, K
L ˆ dc Nz , total width of a sample DT temperature increase over initial value, °C
m ˆ dc =dm u transmembrane voltage, V
n number of pulses u0 midpoint of a probability transition function
nbr number of pulses at the moment of the total g, V
dielectric breakdown of material Du width of a probability transition function g, V
Ni number of intact cells U external voltage, V
Nl number of intact cells neighbouring a given W moisture content, %
cell
Greek letters
a geometry factor
*
Corresponding author. b scaling exponent
E-mail address: eugene.vorobiev@utc.fr (E. Vorobiev). c surface tension of membrane, N mÿ1

0260-8774/00/$ - see front matter Ó 2000 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 0 - 8 7 7 4 ( 0 0 ) 0 0 0 2 9 - 7
214 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223

ew ˆ 80, dielectric constant of water for increasing the product quality in juice production
em ˆ 2, dielectric constant of membrane (Papchenko et al., 1988; McLellan, Kime & Lind, 1991),
u ˆ P =…1 ÿ P †
k ˆ dc rm =dm rc
winemaking (Kalmykova, 1993), and sugar production
q density of food material, kg mÿ3 (Gulyi et al., 1994; Jemai, 1997).
r conductivity, S mÿ1 However, at present, increased use of PEF treatment
rc conductivity of intracellular ¯uid, S mÿ1 for nonthermal processing of heterogeneous food ma-
ri ˆ rc k=…1 ‡ k†, conductivity of an intact cell terials is restrained by the poor reproducibility of ex-
R ˆ Re =Rc
R0 ˆ Ri =Rc
perimental data, the unclear mechanism of dielectric
Re;c;i ˆ r1=b
e;c;i
breakdown in the cellular systems, and the absence of
s lifetime of a membrane criteria for choosing optimal parameters of PEF treat-
s1 parameter, lifetime of a membrane at T ˆ 1 ment. Computer simulation methods may be extremely
w ˆ 1=Pp ÿ 1 useful for elucidation of the breakdown mechanism and
x linear tension of membrane, N
for optimisation of PEF treatment in cellular materials.
Subscripts There exist many types of models for simulation of dy-
a after treatment
namic breakage processes in random disordered sys-
b before treatment
c cell tems. The most popular is the model of random resistor
e e€ective (or random conductor) networks (De Arcangelis,
i intact cell Redner & Herrmann, 1985; Duxbury & Beale, 1995).
j juice The models of this type are applied for simulation of the
l local
physical behavior of granular superconductors, metal-
m membrane
t total insulator composites and various other disordered
materials. Usually, the breakdown process kinetics is
Abbreviations
investigated using network models of di€erent types
PEF pulsed electric ®eld
LC linear chain through the iterative solving of Ohm±Kirchho€'s system
MF mean ®eld of equations with allowance for boundary conditions
RSA random sequential adsorption and by choosing relevant initial conditions for break-
RSO random sequential occupation down initialisation (Lebovka & Mank, 1992). But all the
above-mentioned models are highly time consuming and
require a lot of computation time for the real cellular
1. Introduction material simulation (Press, Teukolsky, Vetterling &
Flannery, 1997).
Pulsed electric ®eld (PEF) treatment is a new and The objective of this study is to develop a simulation
promising nonthermal processing method for heteroge- model describing the breakage kinetics under condi-
neous food materials. PEF methods are based on tions of PEF treatment and to correlate experimental
the e€ect of cell transformation or rupture under an data for heterogeneous food materials with simulation
external electric ®eld, which results in increase in the results. We consider a very simpli®ed model of food
electric conductivity and permeability of cellular material breakage under the PEF, which is based on a
material. This e€ect, known as dielectric breakdown generalised approximation of the e€ective medium.
(Zimmermann, Pilwat & Riemann, 1974), or electro- This model takes into account the percolation proper-
plasmolysis (Scheglov, 1983), can be explained generally ties of the medium and local bridging e€ects. This
by electroporation, i.e., electric ®eld-induced formation approach is, certainly, rather crude, as are all the other
and growth of pores in biological membranes resulting e€ective medium approaches, but it leads to a realistic
from their polarisation. The method of electroporation description of the breakage processes and kinetics
became very popular because it was found to be an trends.
exceptionally practical way of transferring drugs, genetic
material (e.g. DNA), or other molecules inside the cells
(Chang, Chassy, Saunders & Sowers, 1992). Recently, 2. Background theory and simulation model
scientists began to use the PEF methods for treatment of
liquid food (fruit juices, milk etc.), as an alternative to 2.1. Field induced on a cell membrane in an external ®eld
high-temperature preservation (Barbosa-Canovas,
Pothakamury, Palou & Swanson, 1998; Hulsheger, Potel We consider at ®rst the case of an individual cell in-
& Niemann, 1983). PEF protocols also have been re- side the homogeneous medium having a low e€ective
cently introduced for cellular tissue treatment (Knorr, conductivity re placed into an external electric ®eld of
Geulen, Grahl & Sitzmann, 1994; Knorr, 1997). In food strength E. The maximum voltage u between the internal
technologies, the electric ®eld treatment was demon- and external cellular surfaces of membrane (or trans-
strated to be good for intensi®cation of juice yield and membrane voltage) induced on cell poles by the external
 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223
®eld, is equal to (Neumann, Sprafke, Boldt & Wolf,
1992)
u ˆ afdc E; …
where a is the geometry factor equal to 0.75 for the
spherical geometry of cell and to 1 for a cell of cubic
geometry, and f depends on the electrical and geomet-
rical properties of the cell (Kotnik, Miklavcic & Slivnik,
1998)
re …6mre ÿ …12m2 ÿ 8m3 †…rm ÿ rc ††
f ˆ 3
…rm ‡ 2re †…rm ‡ rc =2† ÿ …1 ÿ 2m† …re ÿ rm †…rc ÿ rm †
…2†
Taking into account the low conductivity of mem-
brane, rm =rc  1 and expanding Eq. (2) as a power of
rm =rc , we obtain
  h i
rm
f ˆ 1‡
2re K
= K ‡ O …rm =rc †2 ; …3†
2
where K ˆ 1 ‡ k…2 ‡ rc =re †=4, the O‰…rm =rc † Š term in-
2
dicates that the truncation error is of order …rm =rc † ,
and under the usual conditions (standard values of dc ,
dm , rm , rc and re parameters for this problem were
collected by Kotnik eet al., 1998) k  10ÿ1 ÿ 10ÿ2  1
and f  1. For the purposes of estimation we shall
consider further on only cells of cubic geometry, so
a ˆ 1. In this approximation, f is equal to the norma-
lised transmembrane voltage u=dc E.
Now we consider a problem of transmembrane volt-
age …u† estimation for a cell inside a biological tissue. In
this case, the external media is not homogeneous, so,
Eqs. (1) and (2) are inaccurate. They can be used only as
an approximate estimation by assuming a certain e€ec-
tive value of re . This is just the mean ®eld (MF) type of
approximation, which does not take local conductivity
¯uctuations into account. It can be assumed, as a rough
approximation, that in an inhomogeneous medium,
Eq. (1) may also be used for determination of the
u values, however, f ˆ f …r† will depend on the space
co-ordinate r and will vary from cell to cell.
As follows from Eq. (3), the value of f gradually in-
creases with the increase in external medium e€ective
conductivity re , which depends on the biological tissue
destruction degree, P. Here we de®ne the biological
tissue destruction degree as:
P ˆ 1 ÿ Ni =Nt : …
As a more precise estimation of re versus P depen-
dence, we have used the generalised e€ective media
equation with the same morphology parameters as in
percolation theory (McLachlan, 1989)
u…1 ÿ R† R0 ÿ R
‡ ˆ 0: …
1 ‡ wR R0 ‡ wR
This equation reduces to the well-known Brugg-
eman's symmetric equation when b ˆ 1 (Landauer,
215
1978). Everywhere throughout this work we use the
values of b ˆ 2:3, Pp ˆ 0:307, which are typical for the
1† 3d simple cubic lattice (Sahimi, 1994). In the general
case we obtain the following solution of Eq. (5):
2
re …P † ˆ rc ‰…u…w ÿ R0 † ‡ R0 w ÿ 1 ‡ …u2 …w ‡ R0 †
2
‡ 2u‰R0 …w ‡ 1† ÿ w…R0 ÿ 1† Š
2
‡ …wR0 ‡ 1†2 †1=2 †=…2w…u ‡ 1††Šb : …6†
which can be easily used for the numerical estimation of
: re versus P dependency.
From Eq. (6) we can easily obtain re ˆ ri ˆ
rc k=…1 ‡ k† at P ˆ 0 and re ˆ rc at P ˆ 1.
The example of f versus P dependence in the mean
®eld approximation obtained with the help of Eqs. (2)
and (6) is presented in Fig. 1 (line 1). At low degrees of
biological tissue destruction, P ! 0, the e€ective con-
ductivity approximately equals to re ˆ ri ˆ rc k=…1 ‡ k†.
At complete destruction of a biological tissue, P ! 1, we
have re  rc . Thus, we obtain the following limiting
relations in MF approximation:
f …P ! 0† ˆ 1=…2 ‡ 3k†  1=2;
…7†
f …P ! 1† ˆ 1=…1 ‡ 3k†  1;
i.e., the value of f approximately doubles with the bio-
logical tissue destruction increase within the range of
P ˆ 0±1.
In fact, the change in f with increasing biological
tissue destruction degree (P) is impossible to describe
within the frames of such a simpli®ed approach, because
of substantial local ¯uctuations in the external medium
4†
Fig. 1. Normalised transmembrane voltage f ˆ u=dc E versus media
destruction degree P dependencies calculated for the MF approxima-
tion using Eqs. (2) and (6) (curve 1), for the LC approximation using
5†
Eqs. (6) and (12) (curve 2) and for the RSO model using results of
simulation (curve 3). The dashed curve 4 corresponds to f dispersion
Df in RSO model. The calculations were performed at dc ˆ 10ÿ5 m,
dm ˆ 10ÿ8 m, rc ˆ 1 S mÿ1 , rm ˆ 10ÿ5 S mÿ1 , and k ˆ 0:01.
216 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223
conductivity existing in the real inhomogeneous biolog-
ical tissues. Such ¯uctuations are especially high near
the punctured cells, where the local values of conductivity
can exceed the e€ective average value re by some orders of
magnitude. As we can see from Fig. 1, the MF approxi-
mation works especially badly at low degrees of biological
tissue destruction, P / 0:1, when the local conductivity
¯uctuations are high. In order to demonstrate the com-
plex behaviour of f versus P dependence in a real bio-
logical tissue, we have considered the oversimpli®ed
linear chains (LC) approximation for simulation of a real
tissue structure. The tissue is simulated by linear chains
consisting of Nt cells of a cubic-like geometry (a ˆ 1). The
total resistance of such linear chain consisting of serially
connected membrane resistances (rm ˆ dm =…rm Sm †)
and intracellular ¯uid resistances (rc ˆ dc =…rc Sm †) can be
estimated using the following relation:
R ˆ …Nt rc ‡ Ni rm †:
The density of current running through such a linear
chain is equal to
j ˆ U =…RSm † ˆ U =…Nt dc =rc ‡ Ni dm =rm †
ˆ re U =…Nt dc † ˆ re E;
where
re ˆ krc =…k ‡ 1 ÿ P †; …
is an e€ective chain conductivity, and E ˆ U =L ˆ
U =…Nt dc †.
The percolation point is observed in LC-approxima-
tion at P ˆ Pp ˆ 1, and it can be easily shown, that the
more general Eq. (6) reduces to Eq. (10) in the case of
b ˆ 1.
The voltage drop at a single membrane is equal to
u ˆ jSm rm ˆ jdm =rm ˆ dc Ere =…krc †; …
or we obtain for f ˆ u=dc E (at a ˆ 1)
f ˆ re …P †=krc : …
The example of f versus P dependence for the linear
chain approximation obtained with the help of Eqs. (6)
and (12) is presented in Fig. 1 (line 2).
We obtain from Eqs. (6) and (12) an LC approxi-
mation which accounts for k  1
f …P ! 0† ˆ 1=…1 ‡ k†  1;
…
f …P ! 1† ˆ 1=k  1;
i.e., in this case the value of f can considerably increase
in the course of biological tissue destruction, and it is in
contradiction with the conclusion which follows from
Eq. (7) on the basis of MF approximation.
Moreover, MF approximation gives us f …P ! 1†  1,
but in LC approximation we have f …P ! 0† ˆ
1=…1 ‡ k†  1. How can we clarify this contradiction?
MF approximation is good only at large values of
P ' 0:9, when all of the tissue cells are destroyed and the
media is practically homogeneous. On the other side, LC
approximation works well only at small values of P ! 0
(see Fig. 1). In this latter case we can neglect any
bridging e€ects, which are very important in the middle
of the range of P values, 0:1 < P < 0:9. The bridging
e€ect results from the deviations of local conductivity
near the selected cell from the average e€ective con-
ductivity of the whole media. Analysing the parallel
electrical circuits in the near-neighbours environments
of a given cell, we can easily approximate the averaged
local cell resistance rl as
rl  1=……1 ÿ Pl †=ri ‡ Pl =rc †: …14†
For the cells arranged in the sites of a simple cubic
lattice we have Pl ˆ …1 ÿ Nl =6†, where Nl is the discrete
number of intact cells neighbouring a given cell. Then
the voltage drop at a single membrane equals to
…8† dc Ere
u ˆ jrl Sm rm =ri ˆ ; …15†
krc …1 ‡ Pl =k†
and we obtain the following generalised equation for
estimation of the local value fl ˆ u=dc E:
…9† fl …P ; Pl † ˆ re …P †=…k ‡ Pl †rc ; …16†
where in the general case the value of re at given P may
be determined from Eq. (6).
10†
Fig. 2 shows an example of fl versus P dependencies
obtained at various Nl with the help of Eqs. (6) and (16).
We see that the local behaviour of fl …P ; Pl † function can
be rather complex and it can re¯ect the local conduc-
11†
12†
13†
Fig. 2. Normalised local transmembrane voltage fl ˆ …u=dc E†l versus
media destruction degree P at various Nl (the number of intact cells
neighbouring a given cell). The dashed line corresponds to the example
of fl behaviour with increase of the number of destroyed near-neigh-
bours in the course of the media breakage. The calculations were
performed at dc ˆ 10ÿ5 m, dm ˆ 10ÿ8 m, rc ˆ 1 S mÿ1 , rm ˆ 10ÿ5 S mÿ1 ,
and k ˆ 0:01 for cells arranged in the nodes of a simple cubic
lattice.
 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223
tivity changes near the chosen cell. We have shown with
the dashed curve the imaginary ¯uctuations of fl with P
increase and successive breakdown of neighbouring
cells. We understand that this approach proposed for
estimation of the fl …P ; Pl † behaviour is very simpli®ed. In
particular, we overestimate the importance of the lattice
contribution to the fl …P ; Pl † behaviour due to the dis-
creteness of this problem, and in the case of a continuum
a more smooth behaviour of fl …P ; Pl † can be observed for
the continual parameter Pl .
The behaviour of the averaged value f …P † ˆ
hfl …P ; Pl †i (here h


i implies averaging over all the cells
of a system) will depend on the spatial distribution of
broken cells in such a system. Let us consider as an
example, the f …P † behaviour for the model of random
sequential occupation (RSO) of the lattice sites with the
broken cells. This problem is very complex and has no
exact solution.
Here, the Monte Carlo simulation is a rather simple
and useful method allowing us to get the f …P † function
for the given type of cell distribution in the system. We
have considered the cells located in the nodes of a simple
cubic lattice, where one lattice unit was corresponding to
one cell. We used the periodical boundary conditions
along all the x, y and z directions in order to reduce the
in¯uence of boundaries. Lattices of a rather large di-
mension (100  100  100) were studied, which allowed
us to neglect usual ®nite size scaling e€ects similar to
those described by Watanabe ((1995). The simulation
consisted of successive random choice of an intact cell,
its further destruction and averaging of the calculated f-
values over the whole lattice. The results were averaged
over 10 di€erent initial con®gurations.
The example of f versus P dependence obtained as
described above is presented in Fig. 1 (line 3). We see
that on increase of the destruction degree (P) the values
of f ®rst decrease, then pass through a minimum and
increase again. The dashed line 4 corresponds to dis-
persion (Df ) and characterises ¯uctuations of f in the
system. The minimum point of f …P † approximately
corresponds to maximal ¯uctuations of the local fl val-
ues in a system, i.e., to the maximum of Df . The im-
portant positive feature of this method of f …P †
estimation is as follows: we get f  1 in both limit cases
of P ! 0 and P ! 1, i.e., we remove here the con-
tradiction between MF and LC approximations.
2.2. Probability of a single cell destruction
The average lifetime of a membrane in the external
electric ®eld can be estimated with the help of the fol-
lowing equation (Weaver & Chismadzhev, 1996):
s…T ; u† ˆ s1 exp…DF  =…1 ‡ u2 C  ††: …17†
Lebedeva ((1997) has presented the following estimates
for the general lipid membranes: s1 u 0:37  10ÿ6 s,
217
x u 1:69  10ÿ11 N, c u 2  10ÿ3 N mÿ2 , Cm u 3:5
10ÿ3 F mÿ2 at T ˆ 298 K.
Then the breakage probability for a membrane (as a
whole cell) during the impact period of a pulse with
duration of ti may be estimated as
g ˆ 1 ÿ exp…ÿti =s…T ; u††: …18†
Taking Eq. (17) into account, we can rewrite Eq. (18)
in the following convenient dimensionless form:
g…u † ˆ 1 ÿ exp…ÿ ln 2= exp a…‰1 ÿ …1 ÿ u2 †
=…aDu ln 2†Š
ÿ1
ÿ 1††; …19†
p 
where u ˆ u=u0 , Du ˆ Du=u0 , u0 ˆ …DF  =a ÿ 1†=C  ,
Du ˆ u0 =……1 ÿ a=DF  †a ln 2†, Du P Duc ˆ 1=…a ln 2†,
and a ˆ ln…ti =…s1 ln 2†† is a parameter.
We see that g…u † is a kind of probability transition
function and u ˆ 1 (u ˆ u0 ) corresponds to the mid-
point, where g…u† ˆ 1=2 and Du is the width of this
transition function. In fact, the value of u0 may serve as
an estimate for the critical value of transmembrane
voltage, which causes the abrupt decrease of the mem-
brane lifetime.
We can estimate the following actual values for the
general lipid membranes using the experimental data of
Lebedeva ((1997): u0 u 0:92 V, Du u 0:41, a u 3:66 and
Duc u 0:39 (at ti ˆ 10 ls) and u0 u 0:71 V, Du u 0:26,
a u 5:97 and Duc u 0:24 (at ti ˆ 100 ls).
2.3. Description of the simulation model
To achieve the best performance, we have used here a
hybrid of MF and cellular automaton strategies in order
to reduce the computational complexity. First we con-
struct an array of intact cells located at the nodes of a
simple cubic lattice. The nodes with intact cells are
marked as unoccupied.
The system has an Nx  Ny  Nz dimension and the
external electric ®eld of E strength is applied along the z-
axis (Fig. 3). We consider the idealised square pulse
sequence with a pulse duration ti , and a pulse repetition
time Dt. We use Nz ˆ 1000 in our simulation and the
total sample width is L ˆ dc Nz (1 cm when
dc ˆ 10 lm). Because of the computation time restric-
tions we use Nx ˆ Ny ˆ 10 and periodical boundary
conditions are applied along x and y directions in order
to decrease the ®nite size e€ects.
The simulation procedure is done at each time step
according to the following scheme:
(a) choose the next occupied site within a lattice using
the linear search procedure;
(b) determine the number of occupied nodes (Nl )
among all of its six near-neighbours, determine Pl ,
and calcule the transmembrane voltage u at a given
cell with the help of Eqs. (1), (6) and (16);
218 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223
Fig. 3. The model of the food material structure used in the computer
simulation.
(c) calculate the probability of a given cell destruction
using Eq. (19) and occupate the node in a case when
the cell is destroyed;
(d) repeat the step (a) until all the occupied lattice
nodes appear to be tested.
The total destruction degree P and values of re and
hfl …P ; Pl †i are calculated after each pulse. These time
steps run until P reaches its asymptotic value. This
model is a simple cellular automaton for the purposes of
simulation of the food material breakage kinetics. In-
deed, at step (b) we test the near-neighbour environment
of a given cell and then we calculate the probability of
destruction.
We use in all calculations: dc ˆ 10ÿ5 m, dm ˆ 10ÿ8 m,
rc ˆ 0:1 ÿ 1 S mÿ1 , rm ˆ 10ÿ4 ÿ 10ÿ6 S mÿ1 , ti ˆ 10±
100 ls.
The elevation of the temperature after each pulse was
calculated by considering Joule heating of the material.
The ohmic temperature increase DT …ti † after each
pulse of ti duration was calculated using the following
equation:
dDT E2 re ti
DT …ti † ˆ ˆ : …
dn Cq
For estimation purposes we put here C ˆ 3:93 kJ
kgÿ1 Kÿ1 and q ˆ 0:81  103 kg mÿ3 , which are char-
acteristic values for apples at 25°C (Losano, Urbicain &
Rotstein, 1979).
3. Materials and methods
3.1. Materials
Freshly harvested apples of the Golden Delicious
variety were selected for investigation and stored at 4°C
until required. A moisture content of apples W was
within 80±85%. We estimated the cell destruction degree
as a ratio of e€ective conductivities measured before and
after pulsed electric ®eld treatment. The speci®c con-
ductivities of samples, both initial (before treatment)
and ®nal (after the full treatment at highest voltages)
were within rb ˆ 0.004±0.008 S mÿ1 and ra ˆ 0.1±
0.2 S mÿ1 , respectively. The speci®c conductivity of the
apple juice extracted from the sample apples was within
rj ˆ 0.1±0.3 S mÿ1 . In all the cases the value of rj for the
apple juice extracted from a treated sample was higher
than the value of ra for the same sample.
3.2. Methods
Thin slices (6  0:2 mm thickness and 45  0:5 mm
diameter) were cut from an apple pap. The conductivi-
ties were measured by contacting electrode method with
an LCR Meter HP 4284A (Hewlett Packard, 38 mm
guarded/guard Electrode-A HP 16451B) for the thin
apple slice samples at a frequency of 1000 Hz and with a
Conductimetre HI8820N (Hanna Instruments, Portu-
gal) for the apple juice samples at a frequency of 50 Hz
(this frequencies were selected as optimal in order to
remove the in¯uence of the polarising e€ects on elec-
trodes and inside the samples). The value of rb was
measured 3±5 s after ®nishing the treatment procedure.
Fig. 4 is a schematic representation of the experi-
mental pulsed electric ®eld treatment set-up. The tem-
perature was recorded in the on-line mode by a
thermocouple THERMOCOAX type 2 (AB 25 NN,
0.1°C). A high-voltage pulse generator, 1500 V±15 A
(Service Electronique UTC, France) allowed ti to vary
within the interval of 10±1000 ls (to precision 2 ls), Dt
within the interval of 1±100 ms (to precision 0.1 ms)
and n within the interval of 1±100 000. All the experi-
ments were repeated at least ®ve times. Pulse protocols
and all the output data (current, voltage, impedance,
and temperature) were controlled using a data logger
and special software HPVEE v.4.01 (Hewlett-Packard)
adopted by Service Electronique UTC, France.
20†
4. Results and discussion
Fig. 5 presents the experimental curves of relative
conductivity ra =rb versus number of pulses n for dif-
ferent values of the electric ®eld strength E with ti ˆ
100ls and Dt ˆ 10 ms. We have obtained practically
same (experimental data coincidence within the limits of
measuring errors) results for ra =rb versus n dependen-
cies at di€erent values of Dt within the interval of 1±
100 ms. Hence, pulse repetition time does not in¯uence
our data and we have used Dt ˆ 10 ms in all the ex-
periments. We have observed similar results for ra =rb
 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223 219

versus nti (equivalent time of electrical treatment) with ti
variation.
We can conclude out of the data obtained that there
exist, at least, two di€erent stages of the material
breakdown process. We have divided these stages con-
ditionally by a horizontal dashed line in Fig. 5. In the
®rst stage, the sequential and the correlated breakdowns
of cells develop in the system. The time of the ®rst stage
changes drastically with E increase. At low values of E
we observe a very slow evolution of the material
breakdown. The second stage of material breakdown
(over the horizontal dashed line) ¯ows more rapidly
until terminated by an abrupt total dielectric breakdown
of the material. For cases presented in Fig. 5 the further
increase of n leads to the uncontrolled dielectric break-
down when E ˆ 0.6±2.0 kV cmÿ1 . At this moment, we
observe the over¯ow value of the out-of-limit current
(15 A) and stop the further treatment of material.
The kinetics of temperature evolution DT against n is
presented in Fig. 6. The inset in Fig. 6 shows that in
practically all the cases we observe approximately linear
T versus n dependencies. Some deviation from this

Fig. 4. Schematic representation of the experimental set-up used in the study of pulsed electric ®eld treatment of apple slices.

Fig. 6. Relative ohmic temperature increase DT (°C) versus number of

Fig. 5. Relative conductivity ra =rb versus number of pulses n at dif- pulses n at di€erent values of the electric ®eld strength E, ti ˆ 100 ls
ferent values of the electric ®eld strength E, ti ˆ 100 ls and Dt ˆ 10 ms and Dt ˆ 10 ms for thin apple slices at 25°C. Insert shows the same
for thin apple slices at 25°C. data for small n values in linear co-ordinates T versus n.
220 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223
Fig. 7. Dependencies of the number of pulses in the total breakdown
point nbr (a) and of the mean ohmic temperature increase after each
pulse dDT …ti †=dn (b) against the electric strength E at ti ˆ 100 ls and
Dt ˆ 10 ms for thin apple slices at 25°C.
linear behaviour can be seen only for small E values
(E ˆ 0:2 kV cmÿ1 ) and at large n > 104 ; we can explain
it by the heat exchange with the outside media.
We observe for each E a certain value of nbr , which
corresponds to the number of pulses at the moment of
the total dielectric breakdown of material. Fig. 7(a)
presents nbr versus E (electric ®eld strength) dependence.
The slopes of the near-linear T versus n curves corre-
spond to the mean values of DT …ti † ˆ dDT =dn averaged
over the total interval of n. They are shown in Fig. 7(b)
for the di€erent values of E. Here the solid line corre-
sponds to the square-law DT …ti † versus E increase in
accordance with relation
DT …ti † ˆ a ‡ bE ‡ cE2 ; …
which is consistent with Eq. (20). Here a ˆ 1:116  10ÿ4 ,
b ˆ ÿ1:118  10ÿ5 , c ˆ 4:946  10ÿ8 are the values ob-
tained from root mean square treatment of the experi-
mental data for the interval E < 1 kV cmÿ1 . At large
values of ®eld strength, E > 1 kV cmÿ1 , we observed a
considerable deviation from the parabolic law of Eq. (20)
in the dDT =dn versus E dependency. The reasons of such
deviation are still unclear.
The computer simulation results for the e€ective
conductivity (re ) versus number of pulses (n) and
breakage degree (P) versus number of pulses (n) de-
pendencies at the di€erent values of E, k ˆ 0:1 and
k ˆ 0:01 are presented in Fig. 8(a) and (b). We observe
the obvious step-like behaviour of these P …n† and re …n†
curves describing the breakage kinetics. This behaviour
re¯ects the ``jamming'' e€ects present in the systems
under investigation and has a pure geometric origin.
The ``jamming'' problem is very important and oc-
curs in a number of situations, such as irreversible sur-
face deposition of extended objects, random sequential
adsorption (RSA problem), the polymer physics prob-
lem or the car-parking problem (Evans, 1993; Nielaba,
Privman & Wang, 1993). The basic characteristic of this
problem is the ``jamming'' coverage that depends on the
type of the lattice and system dimensionality. In the case
of deposition of the de®nite size particles, the ``jamming''
limit is reached when it is impossible to place any further
objects without overlapping those previously deposited.
The ``jamming'' e€ects in the dielectric breakdown
problem have the following origin: when a cell gets
punctured, there appear bridging e€ects, which cause an
abrupt fall in destruction probability for cells sur-
rounding the punctured one. In the frame of our model,
this bridging e€ect results from deviations of the local
conductivity near the selected cell rl (see Eq. (14)) from
the average e€ective conductivity of the whole media re .
In the RSA problem, the cell puncture is equivalent to
the lattice site occupation, thus, occupation of a de®nite
site results in abrupt fall of the occupation probability
for nodes neighbouring the punctured one. The dielec-
tric breakdown problem in such formulation is very
similar to the RSA problem, except for successive
``jamming'' limits, which correspond to the site occu-
pation with one, two, and more near-neighbours. We
have carried out the Monte±Carlo simulation for the
random sequential occupation of nodes in the simple
cubic lattice with j neighbour punctured cells in order to
®nd the ``jamming'' limits P …j†. We have obtained the
following values for the lattice of 100  100  100 di-
mension, same as was used in Section 2.1, by averaging
the calculation results over the 10 di€erent initial
con®gurations: P …0† ˆ 0:3051 . . . ; P …1† ˆ 0:4203 . . . ;
P …2† ˆ 0:4937 . . . ; P …3† ˆ 0:5444 . . . ; P …4† ˆ 0:5932 . . . ;
P …5† ˆ 0:6435 . . . ; and P …6† ˆ 1.
The ``jamming'' e€ects in the dielectric breakdown
21†
problem result in existence of the saturation regimes in
the breakdown kinetics. The ®rst ``jamming'' step is
clearly observed in Fig. 8 for the small values of E <1 at
P ˆ P …0†  0:3. The next ``jamming'' steps are less
pronounced and we have clearly observed them only at
small values of Du  0:25 ÿ 0:3 (data not shown). The
initial increase in the P versus n curves corresponds to
the breakdown development without any near-neigh-
bour bridging e€ects. At large values of E > 1:5 we
observe the jump on the next ``jamming'' steps (P …1†,
P …2†) with the subsequent saturation regime.
We have observed similar saturation regimes in ex-
perimental observations of the kinetics of dielectric
breakage in thin apple slices (see Fig. 5), particularly, at
small values of E < 0:5 kV cmÿ1 . The step-like behav-
iour is not so pronounced in experimental results as in
computer simulation data presented in Fig. 8; this dif-
ference results from the simulation model restrictions
and, partially, from the lattice nature.
Fig. 8(c) presents the curves of the calculated break-
age degree P versus relative ohmic temperature increase
 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223 221

Fig. 8. Calculated e€ective conductivity re , breakage degree P versus number of pulses n dependencies (a,b) and breakage degree P versus relative
ohmic temperature increase DT (°C) dependencies (c) at di€erent values of electric ®eld strength E for k ˆ 0:1 and k ˆ 0:01. The calculation are
performed at Du ˆ 0:35, u0 ˆ 1 V, ti ˆ 100 ls, dc ˆ 10ÿ5 m, dm ˆ 10ÿ8 m, rc ˆ 0:5 S mÿ1 , rm ˆ 5  10ÿ5 S mÿ1 (k ˆ 0:1) , and rc ˆ 1 S mÿ1 ,
rm ˆ 10ÿ5 S mÿ1 (k ˆ 0:01).
222 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223
DT …°C† at di€erent values of the electric strength E for
k ˆ 0:1 and k ˆ 0:01. These calculations were done us-
ing Eq. (20). These data allow us to understand the
correlation existing between the achieved breakdown
degree and the power consumption, which is propor-
tional to DT .
When the electric strength (E) values are low, the high
destruction degree (P) can be achieved only on account
of high power consumption and, correspondingly, with
high overheat of the surrounding medium. However, the
weakness of the electric ®eld treatment results in the
possibility of a ``jamming'' regime, when the increase in
power consumption fails to result in P increase, and
poor control over the course of process. As far as it is
dicult to choose precisely the required electric treat-
ment mode for the given ®eld strength values, here the
transition to the mode of the over¯owing out-of-limit
current value occurs readily. In this case, we stop the
further treatment of material and actually do not
achieve high values of P.
Fig. 9 presents the curves of the number of pulses in
the total breakdown point (nbr ) versus reduced electric
strength (E=u0 ) for di€erent values of Du=u0 and k. We
see that the character of nbr versus E curves depends on
the values of Du=u0 and k. The general behaviour of
calculated nbr vs E dependencies (Fig. 9) correlates with
experimental data presented in Fig. 7(a). The nbr in-
creases with decrease of E at a given Du=u0 value, ap-
proximately, in accordance with relation of type
ln nbr  E. Unfortunately, we are unable here to make
more strict comparison between theoretical and experi-
Fig. 9. Calculated number of pulses in total breakdown point nbr
versus reduced electric strength E=u0 for di€erent values of Du=u0
and k. The calculation are performed at u0 ˆ 1 V, ti ˆ 100ls,
dc ˆ 10ÿ5 m, dm ˆ 10ÿ8 m, rc ˆ 1 S mÿ1 , rm ˆ 10ÿ5 S mÿ1 (k ˆ 0:01),
and rc ˆ 0:5 S mÿ1 , rm ˆ 5  10ÿ5 S mÿ1 (k ˆ 0:1).
mental data as long as we have no precise data for u0 , Du
and k parameters used in computer model.
5. Conclusion
A simpli®ed dielectric breakage model based on ef-
fective media approximation is proposed. This approx-
imation includes equations with the same morphology
parameters as in percolation theory. The normalised
transmembrane voltage f ˆ u=dc E versus media break-
age degree P dependence is obtained; this dependence is
useful for estimating the transmembrane voltage and the
cell breakdown probability under the PEF treatment of
food material. The most important feature of the pro-
posed model is the existence of the ``jamming'' behav-
iour occurring in experimental observations of the
biological tissue breakage. The di€erent transitions
corresponding to the di€erent ``jamming'' steps are
identi®ed. The experimental study has yielded informa-
tion about the material destruction degree and temper-
ature elevation versus the time of PEF treatment at
di€erent values of ®eld strength E in the interval of 0.2±
2.2 kV cmÿ1 . The proposed simulation model gives re-
sults consistent with the general trends observed in ex-
perimental breakage kinetics. The electric treatment is
most ecient at high values of the electric ®eld strength
E. However, the eciency of such electric treatment is
limited by manifestations of the ``jamming'' e€ect, as
well as by the processes of uncontrolled dielectric
breakdown, current over¯ow, etc. Reduction in E values
allows more precise control of the treatment process and
prevents uncontrolled breakdown but it is accompanied
with increase of electric power consumption.
Acknowledgements
The authors would like to thank the ``Pole Regional
Genie des Procedes`` (Picardie, France) for providing
®nancial support. Authors also thank Dr. N.S. Piv-
ovarova for help with the preparation of the manuscript.
References
De Arcangelis, L., Redner, S., & Herrmann, H. (1985). A random fuse
model for breaking processes. Journal de Physique (Paris) Lett, 46,
L585.
Barbosa-Canovas, G. V., Pothakamury, U. R., Palou, E., & Swanson,
B. G. (1998). Non-thermal preservation of foods (pp. 53±72).
NewYork: Marcel Dekker.
Chang, D. C., Chassy, B. M., Saunders, J. A., & Sowers, A. E. (1992).
Overview of electroporation and electrofusion. In D. C. Chang, B.
M. Chassy, J. A. Saunders, & A.E. Sowers, Guide to electroporation
and electrofusion (pp. 1±6). San Diego: Academic Press.
Duxbury, P. M., & Beale, P. D. (1995). Breakdown of two-phase
random resistor networks. Physical Review B, 51, 3476±3481.
 N.I. Lebovka et al. / Journal of Food Engineering 44 (2000) 213±223
Evans, J. W. (1993). Random and cooperative sequential adsorption.
Review of Modern Physics, 65, 1281±1344.
Gulyi, I. S., Lebovka, N. I., Mank, V. V., Kupchik, M. P.,
Bazhal, M. I., Matvienko, A. B., & Papchenko, A. Y. (1994).
Scienti®c and practical principles of electrical treatment of food
products and materials. Kiev: UkrINTEI (in Russian).
Hulsheger, H., Potel, J., & Niemann, E. G. (1983). Electric ®eld e€ects
on bacteria and yeast cells. Radiation and Environmental Biophysics,
22, 149±162.
Jemai, A. B. (1997). Contribution a l'etude de l'e€et d'un traitement
electrique sur les cossettes de betterave a sucre. Incidence sur le
procede d'extraction. These de Doctorat, Universite de Technol-
ogie de Compiegne, Compiegne, France.
Kalmykova, I. S. (1993). Application of electroplasmolysis for intensi-
®cation of phenols extracting from the grapes in the technologies
of red table wines and natural juice. Ph.D. Thesis, Odessa
Technological Institute of Food Industry, Odessa, Ukraine (in
Russian).
Knorr, D., Geulen, M., Grahl, T., & Sitzmann, W. (1994). Food
application of high electric ®eld pulses. Trends in Food Science &
Technology, 5, 71±75.
Knorr, D. (1997). Impact of high intensity electric ®eld pulses on plant
cells and tissues. In Proceedings of the Conference of Food
Engineering, Los Angeles, California, pp. 66.
Kotnik, T., Miklavcic, D., & Slivnik, T. (1998). Time course of
transmembrane voltage induced by time-varing electric ®elds ± a
method for theoretical analysis and its application. Bioelectro-
chemistry and Bioenergetics, 45, 3±16.
Landauer, R. (1978). Electrical transport and optical properties of
inhomogeneous media. American Institute of Physics, 40, 2.
Lebedeva, N. E. (1987). Electric breakdown of bilayer lipid membranes
at short times of voltage e€ect. Biologicheskiye Membrany, 4(9),
994±998 (in Russian).
Lebovka, N. I., & Mank, V. V. (1992). Phase diagram and kinetics of
inhomogeneous square lattice brittle fracture. Physica A, 181,
346±363.
223
Losano, J. E., Urbicain, M. J., & Rotstein, E. (1979). Thermal
conductivity of apples as a function of moisture content. Journal of
Food Science, 44, 198±199.
McLachlan, D. S. (1989). The complex permitivity of emulsions: an
e€ective media-percolation equation. Solid State Communications,
72 (8), 831±834.
McLellan, M. R., Kime, R. L., & Lind, L. R. (1991). Electroplasm-
olysis and other treatments to improve apple juice yield. Journal of
Science Food Agriculture, 57, 303±306.
Neumann, E., Sprafke, A., Boldt, E., & Wolf, H. (1992). Biophysical
considerations of membrane electroporation. In D. C. Chang,
B. M. Chassy, J. A. Saunders, & A. E. Sowers, Guide to
electroporation and electrofusion (pp. 77±117). San Diego: Aca-
demic Press.
Nielaba, P., Privman, V., & Wang, J.-S. (1993). Irreversible multilayer
adsorption. In D. P. Landau, K. K. Mon, & H.-B. Sch uttler,
Computer simulation studies in condensed-matter physics VI (p. 143).
Springer Proceedings in Physics, vol. 76, Heidelberg, Berlin:
Springer.
Papchenko, A. Y., Bologa, M.K., Berzoi, S. E., Paukov, J. N.,
Chebanu, V. G., & Rudkovskaya, G. V. (1988). Electroplasmolyzer
for processing vegetable stock, US Patent no. 4723483.
Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P.
(1997). Numerical recipes in Fortran 77: the art of scienti®c
computing (vol. 1). Cambridge: Cambridge University Press.
Sahimi, M. (1994). Applications of percolation theory. London: Taylor
and Francis.
Scheglov, Y. A. (1983). Electroplasmolysis: a new method for treating
fruits and vegetables. Journal of Panorama. Licensintorg, 9, 30.
Watanabe, M. S. (1995). Percolation with a periodic boundary
conditions: the e€ect of system size for crystallisation in molecular
dynamics. Physical Review E, 51 (5), 3945±3951.
Weaver, J. C., & Chismadzhev, Y. A. (1996). Theory of electropora-
tion: a review. Bioelectrochemistry and Bioenergetics, 41, 135±160.
Zimmermann, U., Pilwat, G., & Riemann, F. (1974). Dielectric
breakdown of cell membranes. Biophysical Journal, 14, 881±889.