Académique Documents
Professionnel Documents
Culture Documents
electrical properties and particle sizes. In this way we extend The probability of dielectric failure is usually determined
the DBM to take into account material inhomogeneities from as a function of the propagation time t, measured as the
the point of view of electrical properties. number of channels incorporated into the tree; the incorpo-
The extension of the DBM model presented in this paper ration of a new channel represents a unit of time.
also allows us to describe dielectric breakdown patterns by The cumulative probability of failure, P(t), of a family of
means of their fractal dimension and by their Weibull distri- trees generated by computer simulations satisfies a two-
bution parameters. parameter Weibull distribution 关22兴, such as those observed
In Sec. II we present a description of the DBM, and in in experimental studies, given by
Sec. III the new model is introduced. Results are presented in
Sec. IV, and our discussion and conclusions are summarized P 共 t 兲 ⫽1⫺exp关 ⫺ 共 t/ ␣ 兲  兴 , 共3兲
in Secs. V and VI.
where ␣ is the characteristic propagation time and  is a
II. DIELECTRIC BREAKDOWN MODEL „DBM… shape parameter.
In the DBM 关17兴 the dielectric is represented by a rectan- III. COMBINED MODEL „CM…
gular lattice where each site corresponds to a point in the
dielectric. Microscopic examination of electrical tree growth Previous models of dielectric breakdown 关17–19兴 were
shows that branch extension occurs in increments typically developed for homogeneous materials. We now want to dis-
of 5 –10 m, while the interelectrode gap is 1–2 mm, 关18兴. cuss the basic aspects that should be taken into account to
This implies that a gap of 100 lattice units will represent the study composite materials. These kinds of material could be
experimental situation adequately and, accordingly, 100 represented by a matrix with randomly distributed inhomo-
⫻100 lattices were employed in this work 共therefore, the geneities. In a real material the matrix could be represented
separation between nodes represents a distance L by a polymer, and the inhomogeneities by carbon black, alu-
⫽10 m). The DBM assumes that the tree grows stepwise, minum, or titanium dioxide, i.e., a highly insulating matrix
starting in an electrode with electric potential ⫽0, and surrounding conductor inhomogeneities.
ending in the counterelectrode where ⫽1. The discharge In order to build up a model for composite materials, we
structure has zero internal resistance, i.e., at each point of the should first define some characteristics of inhomogeneities,
structure the electric potential is ⫽0. The tree channel such as electrical properties, shape, size, etc.
growth is governed stochastically by the electric field. The As a first approximation we assume conducting inhomo-
probability P of a tree channel growth at each site of the geneities of circular shape 共with a diameter not much less
electrical tree neighborhood is chosen to be proportional to a than the length L 0 of a breakdown channel兲 randomly dis-
power of the electric field E at such site ( P⬀E ). The tributed in a two-dimensional geometry. Note that inhomo-
electric field E can be written from , and therefore geneities are centered at the matrix nodes, and therefore, in-
homogeneities do not form equipotential clusters. This
共 i ⬘ ,k ⬘ 兲 assumption simplifies the calculation of the electric potential
P 共 i,k→i ⬘ ,k ⬘ 兲 ⫽ . 共1兲 during tree growth.
兺 共 i ⬘ ,k ⬘ 兲 In the DBM and according to Eq. 共1兲, the probability P of
breakdown channel growth between two nodes is chosen to
The sum in the denominator refers to all of the possible be proportional to a power of the electric field, and
growth sites (i ⬘ ,k ⬘ ) adjacent to the electrical tree. therefore
冉 冊
The electric field distribution is obtained by solving the
Laplace equation considering that the tree structure has the i ⬘ ,k ⬘
electric potential of the electrode ( ⫽0). Li
兺冉 冊
P 共 i,k→i ⬘ ,k ⬘ 兲 ⫽ , 共4兲
Breakdown patterns generated by this model have a frac- i ⬘ ,k ⬘
tal structure that has broadly been dealt with in the literature
Li
关17–21兴. The fractal structure of the trees is highly depen-
dent on the value of the exponent . where we have explicitly introduced L i as a breakdown
Experimental and simulated electrical trees can be char- channel length 共Note that in the DBM all channels have the
acterized by their fractal dimension D and failure probability. same length.兲
The fractal dimension is defined from the correlation The extension of the DBM presented in this work intro-
function C(r), which is the quotient of the average number duces inhomogeneity characteristics, assigning different
of lattice sites that belong to the tree, divided by the total probabilities P to the breakdown channel formation, accord-
number of lattice points that can be found within a circle of ing to the conducting characteristics at each site. The situa-
radius r. The average is performed over the set of circles of tion can be rationalized introducing different values of L i in
radius r centered on every point of the electrical tree. The Eq. 共4兲. We note that this modification affects only the prob-
scaling behavior of C(r) with r is given by the following ability P assigned to each site adjacent to the electrical tree.
equation: As indicated in Fig. 1, L i is written as
C 共 r 兲 ⫽C 0 r D⫺2 , 共2兲 L0
L i ⫽L 0 ⫺ , 共5兲
where D is the fractal dimension. ai
066121-2
DIELECTRIC BREAKDOWN MODEL FOR COMPOSITE . . . PHYSICAL REVIEW E 67, 066121 共2003兲
066121-3
PERUANI et al. PHYSICAL REVIEW E 67, 066121 共2003兲
continue with the results derived from the CM. Finally, a A comparison of the fractal dimension of electrical trees
comparison of results obtained from the CM and the SCM simulated with either SCM or CM 共Figs. 5 and 7, respec-
will be performed. tively兲 shows that there is a remarkable agreement in
Figure 2 shows four electrical trees simulated on lattices the fractal dimension of electrical trees in the interval
with growing concentration p of conducting particles and 0⬍ p⬍0.3.
⫽1. The propagation time of these electrical trees also fol-
lows a Weibull distribution like those grown on lattices with V. DISCUSSION
p⫽0, see Ref. 关21兴 and references therein. Figure 3 shows
the dependence of the cumulative probability of failure, To understand the dependence of the fractal dimension D
P(t), on the propagation time t of a set of 100 electrical trees of electrical trees on the fraction p of conducting particles
grown with the set of parameters p⫽0.45 and ⫽1. The 共see Figs. 5 and 7兲, we will resort to some elements from the
Weibull distribution parameters ␣ and  , see Eq. 共3兲, depend percolation theory.
on the concentration of conducting particles p and on the As particles are randomly added to the lattice, nearest
parameter 共see Fig. 4兲. It is interesting to point out that neighbor particles will form clusters as in the percolation
shape parameter  decreases monotonically by increasing p, model. The size of clusters grows by increasing the fraction
p of conducting particles. Therefore, for a sufficiently large
therefore as p→p c , the Weibull distribution is narrowed
value of p, the cluster size will be of the order of the inter-
around ␣ .
electrode gap. This cluster size limit is known as the perco-
Electrical trees are characterized by their fractal dimen-
lation cluster. From percolation theory it is well known that
sion D obtained by a log-log plot of their average correlation
the greatest cluster size N scales with the fraction p of con-
function C(r) versus r 关see Eq. 共2兲兴. Figure 5 shows the
ducting particles as follows:
再
dependence of fractal dimension D on the set of parameters p
and . ln共 p 兲 for p⬍ p c
The corresponding results obtained from CM are shown Dp
in Figs. 6 and 7. In Fig. 6 we show the Weibull parameters ␣ N⬀ p for p⫽ p c 共7兲
and  as a function of p and . In Fig. 7 we show the p⬁ for p⬎ p c ,
dependence of the fractal dimension D 共of a set of 100 elec-
trical trees兲 on the fraction of conducting particles p for dif- where p c is a critical concentration. If p⬍ p c , there exist
ferent values of the parameter . only clusters of finite size, whereas if p⭓p c there exists a
066121-4
DIELECTRIC BREAKDOWN MODEL FOR COMPOSITE . . . PHYSICAL REVIEW E 67, 066121 共2003兲
cluster that bridges the interelectrode gap. The percolation 共1兲 An electrical tree is simulated using the dielectric
limit observed in our simulations with a matrix size 100 breakdown model 共Sec. II兲 with a given value.
⫻100 was p c ⫽(0.59⫾0.01), in remarkable agreement with 共2兲 A fraction p of conducting particles is then ‘‘randomly
the values found in the literature 关23兴, whereas the percola- added’’ to the lattice employed to perform the simulation.
tion dimension of the clusters was D p ⫽(1.89⫾0.03). Figure 共3兲 Those clusters of conducting particles that are nearest
8 shows the dependence of percolation probability P(p) neighbors to the electrical tree are added to it 共see particles A
共i.e., the fraction of percolating clusters兲 on the fraction p of and C in Fig. 1兲.
conducting particles. The critical fraction p c was estimated 共4兲 Once step 共3兲 is fulfilled, the correlation function of
by adjusting P(p) to the function G(p) defined as follows: this ‘‘new’’ electrical tree is determined. For every p value
investigated, 100 simulations were performed. From the cor-
relation function C(r) a fractal dimension D is evaluated.
p␣ The procedure described in steps 共1兲–共4兲 is repeated for
G共 p 兲⫽ . 共8兲
共 p c 兲 ␣ ⫹p ␣ every value investigated, the results obtained are shown in
Fig. 9.
In the SCM, described by Eq. 共6兲, when a growing elec- We learn from Fig. 9 that for rather low p values (p
trical tree incorporates a conducting particle, it will also in- ⬍0.3) there is a remarkable agreement with the results
corporate all their conducting particle nearest neighbors, see shown in Figs. 5 共CM兲 and 7 共SCM兲. Agreement between
particles A and B in Fig. 1. From Figs. 5 and 7 we learn that these figures is also observed when the fraction of conduct-
the fractal dimension of our simulated electrical trees obeys ing particles approaches the critical value p c .
the expected percolation behavior when the fraction of con-
ducting particles approaches the critical fraction p c . VI. CONCLUSIONS
To investigate the dependence of the fractal dimension D
In this paper, we generalized the DBM to describe dielec-
of electrical trees on the fraction of conducting particles in
tric breakdown patterns in conductor-loaded composites.
the limit p→0 共see Figs. 5 and 7兲, we study how strong the
Conducting particles are distributed at random in the insulat-
perturbation produced by the conducting particles on the
ing matrix, and the dielectric breakdown propagates accord-
electrical tree structure is. The following procedure was
ing to new rules to take into account electrical properties and
applied.
particle size.
066121-5
PERUANI et al. PHYSICAL REVIEW E 67, 066121 共2003兲
关1兴 V. E. Gul, Structure and Properties of Conducting Polymer Report—Conference on Electrical Insulation and Dielectric
Composites 共VSP, Utrecht, The Netherlands, 1996兲. Phenomena, Minneapolis, edited by IEEE Dielectrics and
关2兴 G. Schwartz, Ph.D. thesis, University of Buenos Aires, Argen- Electrical Insulation Society 共Preney Print & Litho, Inc.,
tina, 2001 共unpublished兲. Windsor, Ontario, Canada, 1997兲, p. 173.
关3兴 E. K. Sichel, J. I. Gittleman, and P. Sheng, in Carbon Black- 关14兴 C. C. Ku and R. Liepins, Dielectric Breakdown in Polymers in
Polymer Composites, edited by E. K. Sichel 共Dekker, New Electrical Properties of Polymers-Chemical Principles
York, 1982兲. 共Hanser, Munich, 1987兲, p. 102–199.
关4兴 A. I. Medalia, Rubber Chem. Technol. 51, 437 共1978兲. 关15兴 R. Gachter and H. Muller, Plastic Additives Handbook, 3rd ed.
关5兴 A. I. Medalia, in Carbon Black-Polymer Composites, edited by 共Hanser, Munich, 1990兲, p. 970.
E. K. Sichel, 共Dekker, New York, 1982兲. 关16兴 R. W. Coppard, J. Bowman, L. A. Dissado, S. M. Rowland,
and R. T. Rakowski, J. Phys. D 23, 1554 共1990兲.
关6兴 A. I. Medalia, Rubber Chem. Technol. 59, 432 共1989兲.
关17兴 L. Niemeyer, L. Pietronero, and H. J. Wiesmann, Phys. Rev.
关7兴 D. Roig Fernandez and A. J. Marzocca, Rubber Chem. Tech-
Lett. 52, 1033 共1984兲.
nol. 64, 501 共1991兲.
关18兴 L. A. Dissado and P. J. J. Sweeney, Phys. Rev. B 48, 16 261
关8兴 J. C. Maxwell, A Treatise on Electricity and Magnetism 共Clar- 共1993兲; L. A. Dissado, J. C. Fothergill, N. Wise, A. Willby, and
endon Press, Oxford, 1881兲, Vol. 1, p. 403. J. Cooper, J. Phys. D 33, L1 共2000兲.
关9兴 S. Kirkpatrick, Rev. Mod. Phys. 45, 574 共1973兲. 关19兴 I. M. Irurzun, J. L. Vicente, M. C. Cordero, and E. E. Mola,
关10兴 G. E. Pike and C. H. Seager, Phys. Rev. B 10, 1421 共1974兲. Phys. Rev. E 63, 016110 共2001兲.
关11兴 S. Nakamura, T. Minami, G. Sawa, and K. Kitagawa, in Pro- 关20兴 J. L. Vicente, A. C. Razzitte, M. C. Cordero, and E. E. Mola,
ceedings of the IEEE Annual Report—Conference on Electri- Phys. Rev. E 57, R1 共1998兲.
cal Insulation and Dielectric Phenomena, Minneapolis, edited 关21兴 I. M. Irurzun, P. Bergero, V. Mola, M. C. Cordero, J. L.
by IEEE Dielectrics and Electrical Insulation Society 共Preney Vicente, and E. E. Mola, Chaos, Solitons Fractals 13, 1333
Print & Litho, Inc., Windsor, Ontario, Canada, 1997兲, p. 201. 共2002兲.
关12兴 T. Prasse, Appl. Phys. Lett. 72, 2903 共1998兲. 关22兴 W. Weibull, J. Appl. Mech. 18, 293 共1951兲
关13兴 M. M. Ueki and M. Zanin, in Proceedings of the IEEE Annual 关23兴 Jens Feder, Fractals 共Plenum, New York, 1998兲.
066121-6