Vous êtes sur la page 1sur 14

ARTICLE IN PRESS

Physica A 350 (2005) 393406 www.elsevier.com/locate/physa

Electrocrystallization under magnetic elds: experiment and model


J.C. Mansur Filho, A.G. Silva, A.T.G. Carvalho, M.L. Martins
sica, Universidade Federal de Vic Departamento de F - osa, 36570-000, Vic - osa, MG, Brazil Received 27 August 2004; received in revised form 22 October 2004 Available online 28 November 2004

Abstract We report some experimental results for quasi-two-dimensional electrocrystallization of copper under magnetic elds. Such results are theoretically investigated by large scale simulations of a DLA-like model in which random walkers can move along circular vortices enhanced by the Lorentz force. In addition, a sticking probability is used to take into account the complex reaction dynamics at the cathode surface. Our results indicate that the convective motion does not change the nature of the normal diffusive regime, but increases dramatically the diffusion constant by a factor of up to six. The characteristic features (morphology and scaling laws) of both random walks and growing electrodeposits under a perpendicular magnetic eld are determined. r 2004 Elsevier B.V. All rights reserved.
PACS: 81.15.Pq; 61.43.Hv; 05.40.Fb Keywords: Electrocrystallization; Magnetic effects; DLA aggregates; Random walks

1. Introduction Electrochemical phenomena are widespread in nature and technology, playing important roles in batteries, corrosion, electrocrystallization, macrowiring of copper
Corresponding author.

E-mail address: mmartins@ufv.br (M.L. Martins). 0378-4371/$ - see front matter r 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.physa.2004.11.009

ARTICLE IN PRESS
394 J.C. Mansur Filho et al. / Physica A 350 (2005) 393406

deposits between two electrodes [1] and microwiring of metallic nanoparticles [2]. Electron transfer processes are also central in many biochemical reactions [3]. From the physicist point of view, quasi-two-dimensional electrochemical deposition (ECD) constitutes the most experimentally tractable paradigm for diffusive growth phenomena under non-equilibrium conditions [4]. The rich phenomenology observed in thin-layer ECD, plenty of fractal structures and morphological transitions [5] inspired the diffusion-limited aggregation model (DLA) of Witten and Sander [6], the equivalent of the Ising model in the eld of purely diffusive growth phenomena. Pattern formation in diffusive systems includes crystal growth, viscous ngering, dielectric breakdown, chemical dissolution and many biological growths ranging from bacterial colonies [7] to primary cancer [8]. Introducing a magnetic eld into an electrolytic cell can be expected to control, through the magnetohydrodynamic effect and diamagnetic moment orientation, the charge transfer and mass transport processes. Consequently, the electrodeposit morphology, not only of metals but also of organic polymers [9,10], can be manipulated using a magnetic eld. In electrochemistry it is known that magnetic elds induce dramatic modications in the rate of mass transport [11] whose magnitude depends on electrolyte concentration, solution viscosity and on the nature of the supporting electrolyte. Moreover, magnetic elds change the morphology of the ECD patterns of metal leaves (silver, lead, zinc [12] and copper [13]) modifying their fractal dimension and chirality, i.e., slightly bending the branches of the electrodeposits. Some extensions of the DLA model have been proposed to take into account the effects of a magnetic eld on ECD [15,16]. These models essentially assume that particle diffusion in the medium is strongly affected by the magnetic eld. So, distinct strategies were designed to treat the cyclotron motion of the ions subjected to the Lorentz force. In addition, the concentration gradient established across the diffusion layer around the deposit was modeled through a particle drift toward the center or multiple aggregation in which the Coulomb forces between the ions are considered. However, the problem is to understand how a magnetic eld can have impressive inuence on ECD since the direct effect of the Lorentz force on ions moving with low velocities ( 10 mm=s) and short mean free paths ( 1 nm) should be utterly negligible. The way out of this paradox is to assume that the eld acts to enhance eddies or convection cells in the diffusion layer near the cathode. Indeed, as pointed out in Ref. [11], eld gradient, Lorentz and electrokinetic forces, comparable in magnitude depending on the experimental conditions, always contribute to induce convection, in addition to the natural convection due to the gravitational eld. In 1995, Huth et al. made the rst direct measurements of convective motions in ECD. In a conspicuous experiment they mapped the velocity eld of the uid electrolyte due to buoyancy forces and electric eld, revealing that the buoyancy-driven convection occurs in a vertical plane whereas electroconvective vortices rotate in the quasi-plane of the growth. In this paper some experimental results for quasi-two-dimensional ECD under magnetic elds are reported and a modied DLA model to describe the observed patterns is proposed. In Section 2, the experimental system and the obtained results

ARTICLE IN PRESS
J.C. Mansur Filho et al. / Physica A 350 (2005) 393406 395

are presented. In Section 3, the properties of a random walk proposed to take into account the electroconvection vortices enhanced by the applied magnetic eld are investigated. Such a random walk is used to model the trajectories of the ions near the double layer in a DLA-like model for the ECD under transverse magnetic elds. The growth patterns are characterized through their fractal dimension and chirality. Finally, we conclude in the last section.

2. Experimental setup Experiments were carried out in a quasi-two-dimensional electrochemical cell composed of two electrodes; a thin carbon ring counter electrode (thickness 0:5 mm and inner diameter 30 mm) pasted at an acrylic sheet, and a copper wire working electrode (1 mm in diameter) whose point crosses over the acrylic plate and is fastened to the center of the carbon ring. The copper was electrodeposited under a 2:0 V applied voltage from a 0:4 mol=l aqueous solution of CuSO4 and initial pH 0:85: The depth h of the aqueous lm can be varied by regulating the total volume of the CuSO4 solution. The electrical current as a function of time was measured using an automatic data acquisition system. An outline of the experimental setup is shown in Fig. 1. The morphology of the electrolytic crystal deposited was observed and analyzed using an optical microscope. A uniform magnetic eld up to 0:5 T; oriented perpendicularly to the plane of the cell, was produced by an electromagnet. A leveled support allowed a precise horizontal positioning of the cell when it was inserted between the poles of the electromagnet. The direction of the magnetic eld was carefully adjusted in order to generate eld lines perpendicular to the plane of the cell. At the end of each experiment, losses due to evaporation of the electrolyte was not observed. The temperature of the system was xed at 25  C: 2.1. Experimental results As previously reported on the literature [9,12,13], the magnetic eld effects on ECD are related to mass transport and deposit morphology. Indeed the limiting

Fig. 1. Schematic diagram of the experimental setup of ECD in magnetic elds.

ARTICLE IN PRESS
396 J.C. Mansur Filho et al. / Physica A 350 (2005) 393406

current increases in the presence of the magnetic eld revealing a negative magnetoresistence. Consequently, the electrodeposition rate under a magnetic eld is signicantly enhanced as shown in Fig. 2. In our experiments, the deposited mass M was indirectly measured by integrating in time the curves of currents I t: From the total charge Q transferred to the cathode during a time interval t, we determine M through the Faraday law for electrodeposition [14]. Concerning the morphology of electrodeposits, the growing branches become progressively more thick as the intensity of the magnetic eld increases (see Fig. 3a). Such an increase in the branch thickness indicates that the kinetics of the surface processes is also affected by the eld. It seems that the eld reduces the screening effect typical of purely diffusive growth regimes. In turn, Fig. 3b neatly shows the growth of spiral branches in electrodeposits under a magnetic eld. The chirality of the patterns, i.e., the bending of the branches, is in agreement with that expected from the Lorentz force. These spiral branches indicate that the eld induces convective ow on the cell plane. Indeed, it is claimed [11] that the eld effect is equivalent to rotating the electrode or stirring the electrolyte. Furthermore, it is important to note that only the thickness of the electrolyte solution has been varied from Fig. 3a to b, changing the growth regime from a quasi-two-dimensional to a three-dimensional one. Now, the magnetic eld induces a morphological transition from a compact to a dense branched pattern in which a macroscopic texture is engraved in the branches.

Fig. 2. Ratio between mass transfer at the cathode with (M B ) and without (M 0 ) an applied magnetic eld. A tension V 2:0 V between the electrodes, an aqueous solution of CuSO4 at 0:4 mol=L; and an aqueous solution with depth h 0:6 mm were used.

ARTICLE IN PRESS
J.C. Mansur Filho et al. / Physica A 350 (2005) 393406 397

Fig. 3. ECD patterns under a magnetic eld applied perpendicularly to the electrolytic cell plane. The experimental conditions are those described in Fig. 2, but depths h 0:4 mm (upper) and h 0:8 mm (bottom) of the electrolyte solution were used. In (a) an increase of deposited mass (thicker branches) and in (b) the growth of spiral branches due to the application of a magnetic eld are evident. Also, a morphological transition from compact to dense branched pattern occurs in (b).

The chirality of copper electrodeposits in a thick electrolyte solution is easily observed at moderate eld intensities (Fig. 3b), but the same is not true for thin (quasi-two-dimensional) cells which exhibit high electrical resistances and, consequently, low currents (Fig. 3a). As we shall see, the chirality of simulated patterns under weak magnetic elds (modeled by small turn probability pg and electroconvection radius rg ) is very small. The simulated patterns shown in Fig. 8, in which rg is small (rg 5), exhibit neat spiral branches only for sticking probabilities ps p0:5: In contrast, spiral aggregates are evident, even for ps 1; if the radius of the electroconvective rolls is greater (rg 15), as shown in Fig. 11.

ARTICLE IN PRESS
398 J.C. Mansur Filho et al. / Physica A 350 (2005) 393406

3. Model In any ECD experiment, both gravity and electrically driven convections are relevant in ion transport. Convective effects, due to Coulombic forces concentrated at the tips of the growing branches (electroconvection) and buoyancy forces generated by concentration gradients near each electrode (gravitoconvection) are inevitable even in the absence of an applied magnetic eld [17]. As demonstrated experimentally [17] and through computer simulations [18], when gravitoconvection prevails, the gravity-driven rolls rotate in a vertical plane parallel to the direction of gravity for a cell oriented in the usual horizontal conguration. In turn, when electroconvection is dominant, the convective ow rotates in the quasi-plane of the growth forming a pair of counterrotating vortices around each growing tip. Thus, for a horizontally oriented cell, the ow is normal to the gravity. Moreover, the transition between electro or gravitoconvection regimes is determined by the 2 Grashof numbers, namely, G e ezC F0 =r0 u2 0 (electric) and G g x0 Cga=u0 (gravity) [18]. Here, e is the electronic charge, g is the gravitational acceleration, C is the dimensionless concentration of the ionic species, and x0 ; u0 ; r0 ; and F0 are reference values of length, velocity, uid density, and electrostatic potential, respectively. Also, a 1=r0 qr=qC (Boussinesq-like approximation). So, a transition from a gravitoconvection regime to a electroconvection one can be obtained varying the cell thickness, the electrolyte concentration and the current density. In particular, we estimate that the Grashof numbers set our experiment in the absence of a magnetic eld in the electroconvective prevailing regime. Indeed, the ratio l G g =G e r0 gax0 =ezF0 assumes a value l 2 108 in our zero-eld experiments. This value should be compared with l 11 108 (gravitoconvective regime) and l 5:5 108 (electroconvective regime) corresponding to the experiments reported in Ref. [17]. Here, we associate a typical length scale x0 h; the cell thickness, and estimate the characteristic potential drop over this length scale as F0 0:244V ln1 x0 =r0 ; where V is the applied tension and r0 1 mm is the radius of the cathode surface when the tips begin to grow. Also, we used the value for a calculated from Ref. [17], since our experiment has practically the same ion concentration as that one for the electroconvective regime. This complex scenario for ECD, whatsoever the balance between electro and gravitoconvection, is altered by an applied magnetic eld. Several experiments rmly support that magnetic elds additionally induce convection in electrolyte solutions at the cell plane [19]. In our view, what happens is the following. A magnetic eld applied perpendicularly to a horizontal cell does not affect the gravitoconvection rolls, but strongly distorts the electroconvection vortices. For each pair of counterrotating vortices around any growing tip, one of them is shrinked and the other one is enlarged. Thereby, a net rotating ow oriented accordingly to the Lorentz force is generated near the cathode surface. Such a net ow conguration in the quasi-plane of the growth emerges, independently of gravito- or electroconvective prevailing regimes, in ECD under a magnetic eld normal to the cell plane. From the previous discussion, it can be inferred that even at low ionic concentrations and applied voltages, the purely diffusive DLA model becomes

ARTICLE IN PRESS
J.C. Mansur Filho et al. / Physica A 350 (2005) 393406 399

inadequate to describe ECD patterns under a magnetic eld, and electroconvective processes must be taken into account. So, in order to understand some of the experimental ECD patterns previously described, a modied DLA model is proposed. The main hypothesis underlying our 2D DLA-like model is that the net electroconvective ow, altered by the magnetic eld applied perpendicularly to the electrolytic cell, can be effectively represented by circular vortices near the cathode. Clearly, it is known that a DLA-like model is adequate to describe ECD in low electrolyte concentrations and applied currents. Thereby, our approach can be useful to simulate ECD patterns grown under magnetic elds as shown in Fig. 3a. 3.1. Random walk The electroconvection vortices enhanced by the applied magnetic eld affect the trajectories of the ions nearby the cathode surface. So, the unbiased random walk formerly used to simulate the primarily diffusive growth is altered. It is proposed that near the electrodeposit an ion can move along vortices, with a probability pg ; or executes a unitary step of an unbiased random walk, with probability 1 pg : The motion along a vortex consists of a circular trajectory of length l rg y: The center of this arc is dened as x0 x rg cos f and y0 y rg sin f; where x; y is the particle vector position and f is the angle between the direction of the last unitary step performed by the particle and the positive x-axis. The radius rg of the vortex and the rotation angle y in each turn are parameters of the model somehow inuenced by the magnetic eld intensity. Both turns and steps are off-lattice. Typical trajectories generated by this random walk are shown in Fig. 4. As one can readily notice, the walk visits larger regions and lls the plane more densely as the turn probability pg increases. Hence, its statistical properties and overall shape should be analyzed. Through computer simulations we investigated the time dependence of the mean square displacement of a walker, the number of distinct sites visited by him, the square of the principal radii of gyration, and the asphericity or asymmetry [20] of its random trajectory. The last two quantities characterize the

Fig. 4. Sample random walks with n 104 steps or turns. The turn probability are (a) pg 0; (b) pg 0:8; and (c) pg 0:99: In these simulations a vortex radius rg 5 and a rotation angle y 0:4 rad ( 23 ) were used.

ARTICLE IN PRESS
400 J.C. Mansur Filho et al. / Physica A 350 (2005) 393406

shape of the random walk, whereas the rst ones inform about the nature of the diffusive process. Our simulations indicate that the mean square displacement has a linear time dependence, i.e., /r2 tS 2Dt; independently of the pg value. Therefore, the diffusion process corresponds to a normal Brownian motion. However, as shown in Fig. 5, the diffusion constant D exhibits a non-monotonic behavior. It increases signicantly (up to six times) attaining a maximum at pg  0:80; but decreases for larger values vanishing at pg 1: So, the particle is essentially arrested in a circular trajectory whose center practically does not diffuse for very large turn probabilities. Such result is consistent with the experimental ndings obtained through impedance spectroscopy [13]. Those measurements demonstrate that the eld increases the effective diffusion coefcient D, thereby enhancing mass transport during ECD as evidenced by our own results shown in Fig. 2. Concerning the number of distinct sites S t that the random walker visits after t steps or turns, the following procedure was implemented in our simulations. Since the random walk is off-lattice, the embedding space was covered by a square lattice with a unitary lattice constant and, after each step or turn, the lattice site visited by the walker is that whose integer coordinates are the nearest to the walker position. Our simulations indicate that for every pg the scaling of S t with the number of steps t is consistent with S t  t= log t; valid in d 2 (the critical dimensionality [21]). In fact, as shown in Fig. 6, this asymptotic behavior is attained as a power law of the number of steps t.

Fig. 5. Diffusion constant D as a function of the turn probability pg estimated from simulated random walks with n 105 steps averaged over 106 samples. The parameters rg 5 and y 0:4 rad were used.

ARTICLE IN PRESS
J.C. Mansur Filho et al. / Physica A 350 (2005) 393406 401

Fig. 6. The number of distinct sites Sn that the random walker visits after t n turns or steps for pg 0:8: The data correspond to averages over 106 samples with n 105 ; and parameters rg 5 and y 0:4 rad: The curve is a nonlinear tting y A Bna of the data.

Finally, our results indicate that the shape of this modied random walk also remains unchanged. Indeed, the scaling laws for its principal radii of gyration and asphericity are controlled by the same exponents obtained for a normal (unitary steps) random walk [20]. Specically, the squares of the principal components of the radius of gyration scales with n as R2 i  n; i 1; 2; and the random walk asphericity or asymmetry Ad  0:6; independently of the turn probability pg : 3.2. DLA-like model The effects of electroconvection on the electrodeposits are modeled through a DLA-like growth process. The previously described random walk and a sticking probability are used to model the trajectories of the ions near the growing electrodeposit and the complex reaction dynamics on the growing interface, respectively. Specically, the sticking probability is related to the chemical activation energy necessary to reduce irreversibly the cations at the surface of the aggregate. In our modied DLA model, the particles, released from a distant circle of radius RL and centered at the initial seed, start normal off-lattice random trajectories constituted of steps with length a (assumed unitary). If a particle wanders away from the growing cluster by a distance greater than RK it is discarded and another particle is launched. In contrast, if the wandering particle approaches the cluster and crosses a circle of radius RV centered on the initial seed, it can move along a convection vortex, with probability pg ; or it can perform a normal unitary step, with probability

ARTICLE IN PRESS
402 J.C. Mansur Filho et al. / Physica A 350 (2005) 393406

1 pg : Finally, if the particle reaches a nearest neighbor site of the aggregate, either along its convective motion or after a unitary step, it can stick to this site with probability ps : A launching circle of radius RL d max 400; a convective zone of radius RV d max 300 and a killing circle of radius RK 103 d max were used in the simulations (see Fig. 7). Here, d max is the distance from the seed of the farthest site of the growing aggregate. A large value for RV must be chosen, since, as experimentally revealed in Ref. [17], the electroconvection vortices have a characteristic length scale of about  104 m: In order to optimize the algorithm, particles outside the convective zone of radius RV can perform long steps with a length l out maxfR RV 1; 1g if, after such step, its distance R to the seed remains greater than RV : So, a long step outside the region of vortices cannot lead the particle to the interior of the convection zone of radius RV : Finally, even inside the convective zone, the search for neighbor sites belonging to the cluster, performed at each step or turn, is not done if the particle starts at a distance RXd max rg from the seed. Typical simulated patterns are shown in Fig. 8. As one can see, for a xed pg value, the electrodeposits become more dense as the sticking probability ps decreases. Therefore, the screening effect characteristic of a DLA process is reduced as long as

RK

d max

RV

RL

Fig. 7. Scheme of the optimized DLA-like model showing all the radius dened in the algorithm. The seed particle (center of the lattice) is depicted in black and the other sites of the growing aggregate are the gray squares.

ARTICLE IN PRESS
J.C. Mansur Filho et al. / Physica A 350 (2005) 393406 403

Fig. 8. Simulated patterns with N 3 105 particles, rg 5 and y 0:4 rad: On the top row pg 0 (absence of a magnetic eld) whereas pg 0:8 (electroconvection enhanced by a magnetic eld) on the bottom row. In both rows ps 1 (left), ps 0:5 (middle), ps 0:2 (right).

Fig. 9. Fractal dimension DF of simulated patterns as a function of (a) ps (pg 0:8 xed) and (b) pg (ps 0:5 xed). The data correspond to averages over 10 samples with N 3 105 particles, rg 5 and y 0:4 rad: The mass radius method was used to estimate the fractal dimensions.

the growth dynamics on the cathode interface progressively becomes a reactionlimited one. Accordingly, the fractal dimension DF of the ECD patterns increases as depicted in Fig. 9a. In contrast, Fig. 9b, drawn for xed ps ; shows that the fractal dimension of the electrodeposits does not change with the turn probability pg : This is an expected result since the nature of the diffusive process is not altered by the

ARTICLE IN PRESS
404 J.C. Mansur Filho et al. / Physica A 350 (2005) 393406

vortices. Thus, electroconvection corresponds to a homogeneous perturbation to the DLA model and the screening associated to the diffusive growth is unaffected [4]. However, the chirality of the patterns depends on pg : In the absence of an enhanced electroconvection (pg 0), the ECD clusters are not chiral, independently of the ps value. The chirality of the simulated patterns with pg a0 is also affected by the sticking probability ps : Indeed, as shown in Fig. 10, for different radii of electroconvection eddies, the chirality seems to increase slowly as ps decreases. Clearly, not only the curvature of the main branches of the growing electrodeposit increases, but the minor (internal) tips become more spiraled and space lling, as shown in Fig. 11. Again, the basic feature involved is the progressive screening breakdown in the DLA due to the decreasing sticking probability ps : Finally, it is evident that larger rg and/or y also strengthens the chirality of the patterns, as one can see in Fig. 10. At this point it is important to mention that for very large chiral aggregates the DLA scaling properties must be recovered. Indeed, it is known [4] that the sticking probability introduces a correlation length l below which the patterns are compact (length scales l 5l), but fractal (DLA-like) at longer length scales (l bl). This correlation length l increases as ps decreases. Such effect is present in all DLA-like models in which the diffusive screening is weakened in some way [22]. It is evident that in ECD experiments, the usual size of the electrolytic cells is not very larger than l and, consequently, the crossover to DLA-like scaling regime is rarely observed. In fact, DLA patterns only appear in quasi-two-dimensional ECD under very low electrolyte concentrations or current densities.

Fig. 10. Chirality of simulated patterns as a function of ps (pg 0:8 xed). The data correspond to averages over 10 samples with N 3 105 particles and y 0:4 rad: As in Ref. [13], chirality is dened as the inverse radius of curvature of the main branches in the patterns. The continuous lines are linear ttings to the data with slopes 0.24 and 0.46 for rg 5 and rg 15; respectively.

ARTICLE IN PRESS
J.C. Mansur Filho et al. / Physica A 350 (2005) 393406 405

Fig. 11. Effect of the sticking probability ps on the simulated patterns with N 3 105 particles. In (a) ps 1; (b) ps 0:5; and (c) ps 0:1: The parameters pg 0:8; rg 15 and y 0:4 rad were used.

In summary, these results indicate that the combined action of pg a0 and ps o1 is necessary to explain two of the main features observed in ECD experiments under a magnetic eld, namely, the increasing chirality and fractal dimension of the patterns as the eld magnitude increases [13]. This means that an enhanced electroconvection and a crossover to a reaction-limited dynamics at the cathode interface are basic and simultaneous effects of the magnetic eld.

4. Conclusions A 2D DLA-like model has been proposed to describe the copper ECD patterns under a magnetic eld applied perpendicularly to the plane of a thin electrolytic cell. The model assumes that the magnetic eld does not affect gravity-driven convection rolls, but enhances electroconvection characterized by vortex pairs emerging from each branch tip and rotating in the plane of the growth. The trajectories of the ions in such electroconvection-dominated ow is modeled through an off-lattice random walk in which an ion can move along circular vortices, with a probability pg ; or execute a unitary step of an unbiased random walk, with probability 1 pg : In addition, a sticking probability is used to model the complex reaction dynamics on the growing interface. Our simulation results show that the diffusive regime (scaling laws) and shape (asymmetry) of the random walk are the same as those for a normal Brownian motion. However, the eld increases the effective diffusion coefcient D, thereby enhancing mass transport during ECD, a main result consistent with experimental ndings. Moreover, it is demonstrated that enhanced electroconvection induces chirality on simulated patterns and that a reaction-limited growth dynamics on the aggregate interface progressively increases the fractal dimension of the patterns. Again, these results are in agreement with two of the central observations from experiments of copper ECD.

ARTICLE IN PRESS
406 J.C. Mansur Filho et al. / Physica A 350 (2005) 393406

Acknowledgements We would like to thank S.C. Ferreira Jr. for helpful discussions on the simulation procedures and DLA scaling. This research was partially supported by the Brazilian agencies CNPq and FAPEMIG. References
[1] J.C. Bradley, H.M. Chen, J. Crawford, J. Eckert, K. Enazarova, T. Kurzeja, M. Lin, M. McGee, W. Nadler, S.M. Stephens, Nature 389 (1997) 269. [2] K.D. Hermanson, S.O. Lumsdon, J.P. Williams, E.W. Kaler, O.D. Velev, Science 294 (2001) 1082. [3] S.R. Caplan, I.R. Miller, G. Milazzo (Eds.), Bioelectrochemistry: General Introduction, Birkha user, Basel, 1995. [4] P. Meakin, Fractals, Scaling and Growth Far from Equilibrium, Cambridge University Press, Cambridge, 1998. s, M.Q. Lo pez-Salvans, J. Claret, Phys. Rep. 337 (2000) 97. [5] F. Sague [6] T.A. Witten, L.M. Sander, Phys. Rev. Lett. 47 (1981) 1400. k, E. Ben-Jacob, Physica A 233 (1996) 678. [7] I. Cohen, A. Cziro [8] S.C. Ferreira Jr., M.L. Martins, M.J. Vilela, Phys. Rev. E 65 (2002) 021907. [9] I. Mogi, M. Kamiko, J. Cryst. Growth 166 (1996) 276. [10] I. Mogi, Physica B 216 (1996) 396. [11] G. Hinds, J.M.D. Coey, M.E.G. Lyons, Electrochem. Commun. 3 (2001) 215. [12] I. Mogi, M. Kamiko, S. Okubo, Physica B 211 (1995) 319. [13] J.M.D. Coey, G. Hinds, M.E.G. Lyons, Europhys. Lett. 47 (1999) 267. [14] A.J. Bard, L.R. Faulkner, Electrochemical Methods: Fundamentals and Applications, Wiley, New York, 1980. [15] H. Mizuseki, K. Tanaka, K. Kikuchi, K. Ohno, Y. Kawazoe, Comput. Mater. Sci. 10 (1998) 46. in, J.M.D. Coey, J. Magn. Magn. Mater. 226230 (2001) 1281. [16] T.R. N Mh ocha [17] J.M. Huth, H.L. Swinney, W.D. McCornick, Phys. Rev. E 51 (4) (1995) 3444. [18] G. Marshall, E. Mocskos, F.V. Molina, S. Dengra, Phys. Rev. E 68 (2003) 021607. [19] R.A. Tacken, L.J.J. Janssen, J. Appl. Electrochem. 25 (1995) 1. [20] J. Rudnick, G. Gaspari, Science 237 (1987) 384. [21] M.N. Barber, B.W. Ninham, Random and Restricted Walks, Gordon and Breach, New York, 1970. [22] S.C. Ferreira Jr., cond-mat/0403527 (2004).

Vous aimerez peut-être aussi