Journal of Electroanalytical Chemistry 536 (2002) 11 /18 www.elsevier.com/locate/jelechem Kramers Kronig transformation, dc behaviour and steady state response of the Warburg impedance for a disk electrode inlaid in an insulating surface / J. Navarro-Laboulais a,, J.J. Garcı́a-Jareño b, F. Vicente b a Depto. Ingenierı́a Quı́mica y Nuclear, Escuela Politécnica Superior de Alcoy, Universidad Politécnica de Valencia, Paseo del Viaducto 1, 03801 Alcoy (Alicante), Spain b Depto. Quı́mica Fı́sica, Universitat de València, C/Dr. Moliner 50, 46100 Burjassot (València), Spain Received 11 March 2002; received in revised form 17 July 2002; accepted 6 September 2002 Abstract As the frequency approaches zero, the impedance described by the Warburg function tends to infinity. This means that the resistance of the equivalent circuit representing the electrochemical process has an infinite resistor and then the current cannot flow through it. This asymptotic behaviour also prevents the application of the Kramers /Kronig transformations, a set of integrals which should be fulfilled by any linear system. Using a more general expression of the impedance for a disk electrode inlaid in an insulating surface developed by Fleischmann and Pons (J. Electroanal. Chem. 250 (1988) 277), the Warburg impedance can be deduced and the Kramers /Kronig transformation is possible. An expression is also deduced for the mass-transfer resistance and the frequency at which the Warburg function fails for the representation of the impedance for a disk. A fast algorithm for the calculation of the generalised impedance function is outlined for its implementation in non-linear impedance fitting programs. # 2002 Elsevier Science B.V. All rights reserved. Keywords: Microelectrode impedance; Mass-transport resistance; Composite electrodes 1. Introduction The development of microelectrode technology for electrochemistry and the disagreement between the theoretical predictions based on one-dimensional diffusion equations and the experimental observations [1], led to theoretical efforts to study the inlaid disk microelectrode problem. Several approaches have been used to treat this mathematical problem with mixed boundary conditions. Some of them are based on the properties of the discontinuous Bessel functions, or the more general method of Neumann’s integral as a complete solution of the differential equation describing the inlaid microelectrode [2 /8]. Furthermore, numerical methods and techniques have been developed to solve a great variety of geometrical and electrochemical problems in two and Corresponding author. Tel.: /34-96-652-8479; fax: /34-96-6528409 E-mail address: jnavarla@iqn.upv.es (J. Navarro-Laboulais). three-dimensions. Besides these methods, other semianalytical methods leading to multidimensional integrals have been developed and should be considered to calculate the electrochemical response of an electrode of arbitrary shape. Although the purpose of the works cited above was to solve the non-stationary transport equations for microelectrodes, no mention about the size of the electrodes is made in order to solve the problem mathematically. In fact, the differential equations and boundary conditions are not affected or modified by the size of the system, and therefore the solutions must remain valid for macroelectrodes. In other words, if a complete solution for the non-stationary transport problem of a diskshaped electrode inlaid in an insulating surface is obtained, the steady state (microelectrode behaviour), the macroscopic and all the possible intermediate states will be included in that solution. Numerical methods in electrochemistry are irreplaceable methods to study and analyse the experimental data 0022-0728/02/$ - see front matter # 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 2 - 0 7 2 8 ( 0 2 ) 0 1 1 7 3 - 7 12 J. Navarro-Laboulais et al. / Journal of Electroanalytical Chemistry 536 (2002) 11 /18 in an engineering or in a technical context. Here, the geometry usually is not regular, failing the conventional mathematical methods to solve the diffusion equations with a complex chemical reaction scheme, leading to a coupled system of non-linear partial differential equations that usually does not have an analytical solution. Scanning electrochemical microscopy (SECM) [9], hydrodynamic voltammetry [10], or the current distribution in an electrochemical reactor [11], are examples where numerical methods are fundamental to understand the information provided by experiments. On the other hand, analytical expressions for the data interpretation are fundamental to the electrochemist when factors such as the geometry of the electrode are known. Simple experiments lead in these cases to knowledge of physical quantities of interest such as the diffusion coefficient from a chronoamperometric experiment or the heretogeneous rate constant for an electrochemical reaction from impedance measurements. Although the validity of these equations is here beyond all doubt, it should be noticed that these expressions are the result of a few simplifications that do not make them universal in the sense that they cannot be used for all situations for all electrochemical systems. This is the case of the Warburg impedance studied here. The present work was initially inspired by previous work carried out by the authors on the electrochemistry of composite materials [12 /15] and some explanation of this problem to understand why the Warburg equation should be revised, is advisable. From this starting point, the results obtained in this analysis are general and could be applicable to other electrochemical systems. The electrochemical behaviour of conductor/insulator composite materials is the result of a few interacting factors. On the electrode surface there could be electrochemically active and inactive conducting particles surrounded by the insulating phase. These particles are connected through the solid, so the percolation process governs the bulk electrical conduction properties in the solid. This implies the existence of a potential distribution on the surface because the ohmic drop of each conducting particle across the solid could not be the same [16 /21]. Furthermore, composites are usually prepared by dispersing in the continuous insulating phase the conducting particles which have different shapes and sizes that could be represented by a probability density function. All these phenomena imply that the observed electrochemical properties are the mean value of a set of distributed electrical and geometrical properties. In order to solve this problem, different approaches have been tried. One is to consider the composite electrode as a set of independent microelectrodes. This model is known as the multimicroelectrode hypothesis [22 /27]. Owing to the difficulties in accomplishing the independence of electrochemically active sites on a composite electrode, this hypothesis has not been corroborated yet [28,29]. The first stage to undertake this problem is the selection of an impedance function which fulfils the physical restrictions of the problem, and then, the application of a probability distribution function to contemplate the effect of the microelectrode collective. The Warburg equation seems an appropriate first option to try an impedance function for the composite electrodes but presents two disadvantages which make it unacceptable. First, using a function of this kind assumes that the diffusion field near the electrode is one-dimensional and the existence of a microelectrode distribution with different sizes leads to a Warburg impedance, for which the parameters are the mean values of this distribution [30]. And second, the Warburg impedance is not Kramers /Kronig transformable. From a theoretical point of view this opposes the conditions of linearity, causality and stability that a diffusion controlled process must fulfil and from a practical point of view this means that an equivalent circuit which uses the Warburg function for a blocking electrode could be inappropriate [31]. Additionally, one problem associated with the Kramers /Kronig transformation of experimental data is the evaluation of the impedance in the frequency range from 0 to . If the data set is modelled with a Warburg function, its value diverges when the frequency approaches to zero which contradicts the experimental evidence obtained by potentiodynamic techniques. Few alternatives have been proposed to avoid the problem of the evaluation of the limits of the impedance function, but these solutions, which have in general a mathematical nature, do not help the explanation of the Warburg impedance behaviour related to other electrochemical techniques. The aim of this work is to show these physical inconsistencies associated with the Warburg impedance, which make it difficult to relate impedance experiments with potentiodynamic ones. These problems could be resolved using a more general description of the impedance for an inlaid electrode proposed by Fleischmann and Pons [5]. The relation between these functions and the Warburg impedance is demonstrated and the derivation of an expression for the steady state mass transport resistance for an electrode is obtained. Results from other electrochemical techniques than impedance and from some experimental observations are also discussed. 2. Physical inconsistencies of the Warburg impedance Let SE be the region defined by the electrode and SI the insulating region which surrounds it. The electrode surface could have an arbitrary shape but, for simplicity, J. Navarro-Laboulais et al. / Journal of Electroanalytical Chemistry 536 (2002) 11 /18 let us consider a disk-shaped electrode of radius r . The electrochemical reaction takes place on the electrode surface characterised by the forward and backward heterogeneous rate constants, kf and kb, respectively. Under these conditions the diffusion problem can be described by Fick’s equation: @cox (q; t) @t @cred (q; t) @t Dox 92q cox (q; t) z 0; q SE Dred 92q cred (q; (1) where j0 is the exchange current density and the other symbols have their usual meaning [32,33]. This solution (5) is the transfer function of an electrochemical system, and then, must accomplish the physical restrictions of any transfer function for a lineal system, that is, linearity, causality, stability and being a finite function. In other words, the Warburg impedance must be Kramers /Kronig transformable [31,34 /36], that is, it must fulfil the equations: Re[Z(v)] Re[Z()] t) with initial conditions: cox (q; t)c+ox q; t5 0 + cred (q; t)cred 2 p g w Im[Z(w)] v Im[Z(v)] (2) 2v Im[Z(v)] p (3a) z0; q SE @n @cred (q; t) @n z0; q SI (3b) The subscripts ‘ox’ and ‘red’ refer to oxidised and reduced species, respectively, D is the diffusion coefficient, t the time, q the coordinate vector, q /q (x , y, z), n the normal vector to the electrode surface and i(q, t) the local current on the electrode that should be integrated over the whole electrode surface to obtain the measured current, I (t): I(t) g i(q; t) dq (4) SE The Warburg impedance can be evaluated from the previous equations considering a small amplitude sinusoidal perturbation and some extreme simplifications. For an electrode of infinite radius (r 0/) the Eq. (1) can be rewritten as one-dimensional equations and the boundary condition (3b) and Eq. (4) disappears because it is not applicable for this kind of problem. Another required simplification used in electrochemistry, even if the three-dimensional problem is considered, is to assume Butler /Volmer kinetics in Eq. (3a). Finally, the assumption of equal initial concentration and equal diffusion coefficient for the oxidised and reduced species, allows the achievement of a compact analytical solution of the previous system of differential equations. These simplifications led us to the well-known equation for the impedance of a macroscopic electrode called the Warburg equation or the Warburg impedance: ZW (v) RT nF pr2 j g 2 RT pffiffiffi (1j)v1=2 2 2 2 1=2 + 2 n F pr D c 0 (5) dw Re[Z(w)] Re[Z(v)] dw w2 v2 (6a) (6b) In order to determine the above integrals, it can be demonstrated that the impedance function should previously fulfil the following mathematical conditions: lim Z(v)Z?0 jZƒ0 (7a) lim Z(v) Z? (7b) v00 0 0 @cox (q; t) @c (q; t) Dred red i(q; t) @n @n @cox (q; t) w 2 v2 0 and boundary conditions: Dox 13 v0 where Z?0 ; Zƒ0 and Z? are constants. In other words, the real and imaginary parts of the impedance should be bounded between two known values. If the Warburg impedance is considered, this function diverges when the frequency approaches zero. Thus, the Warburg impedance function would not be Kramers /Kronig transformable. Some mathematical transformations have been proposed in order to avoid this problem. One method was to add a shunt resistance in the impedance calculations in order to elude the asymptotic behaviour of the function at null frequency [36]. Another proposal was to use the KK-transforms in the admittance domain where the function is well-behaved. Although these alternatives could work from a practical point of view, they cannot explain or help to explain why the Warburg impedance behaves as it does and why it is not KKtransformable. This mathematical analysis does not seem to be consistent with the empirical evidence of the usefulness of the Warburg impedance which explains many electrochemical systems. In fact, the reason for this apparent physical inconsistency is found in the way Eq. (5) has been deduced. The simplification of Fick’s second law, a second order partial differential equation involving spatial and time variables, to a spatial one-dimensional partial differential equation is a common mathematical operation in electrochemistry. This first step leads to the Warburg equation, useful from an empirical and practical point of view but inconsistent from a mathematical or physical one, due to the loss of information in the simplification of the original differential equation. J. Navarro-Laboulais et al. / Journal of Electroanalytical Chemistry 536 (2002) 11 /18 14 Another disagreement between the Warburg equation and the empirical observations is the inability of the former to explain the dc behaviour of any electrochemical system. It is assumed that Randles’ equivalent circuit could be used as a representation of an electrochemical process when an alternating potential is applied at the electrode. If the system is linear, and the Warburg impedance or the double layer capacitance are deduced assuming this hypothesis, the electrochemical process will always be represented by this equivalent circuit for any potentiodynamic perturbation. The current /potential curves obtained experimentally (i.e. polarography, steady-state voltammetry, etc.) and the impedance experiences, cannot be connected through the Warburg impedance. As was shown before, the real part of the Warburg impedance tends to infinity as the frequency tends to zero. This could be interpreted as an infinite resistance to the charge transfer, and then, the current should not be able to flow through the electrode. Because the equations used to derive the current / potential dependence in steady-state voltammetry are the same as those used to derive the Warburg impedance, a conflict between the two techniques occurs. The mathematical justification of this discrepancy is again that the Warburg equation is not KK-transformable, and as will be shown in Section 3, the description of a more complete definition of the diffusion impedance will need to explain these apparent inconsistencies. F5 (x) g [J (bx 1=2 1 )]2 0 sin(u=2) b(1 b4 )1=4 db (10) where J1(x ) is the Bessel function of the first kind and order 1, and u /arctan b1. Next we will analyse the behaviour of this function and its connection with the Warburg impedance will be shown. A representation of Eq. (8) in a Nyquist diagram leads to a flattened semicircle as is shown in Fig. 1. Defining a new variable x /r2v /D and using the properties of the trigonometric functions, we can rewrite the impedance function as: Z(x)Rct s x1=2 (F4 (x)jF5 (x)) (11) with x1=2 F4 (x) pffiffiffi 2 g [J1 (y)]2 0 1=2 x F5 (x) pffiffiffi 2 g 0 [J1 (y)]2 [y2 (x2 y4 )1=2 ]1=2 dy (12) [(x2 y4 )1=2 y2 ]1=2 dy y(x2 y4 )1=2 (13) y(x2 y4 )1=2 where Rct RT pnFr2 j0 (14) and 3. Electrochemical impedance of an inlaid disk electrode In the early 1990s the problem of the inlaid disk electrode stated in Eqs.(1) /(4) was treated, solving the steady and unsteady state conditions for relatively simple electrochemical mechanisms [2 /8]. All this work led to similar analytical or semi-analytical solutions involving multidimensional integral equations [8]. One of these, adequate to our purposes, was provided by Fleishmann and Pons [5] who proposed the following equation for the impedance of a disk-microelectrode inlaid in an insulating surface, assuming Butler /Volmer kinetics and equal bulk concentrations for the two redox components: 2 RT 4RT rv Re(Z) (8a) F4 pnFr2 j0 pn2 F 2 D1=2 v1=2 r2 c+ D 2 4RT rv Im(Z) (8b) F5 2 2 1=2 1=2 2 + pn F D v r c D where r is the radius of the disk, and F4(x ) and F5(x ) are defined as: F4 (x) g [J (bx 1 0 1=2 )]2 cos(u=2) b(1 b4 )1=4 db (9) s 4RT prn2 F 2 Dc+ (15) Note that the variable x , relates the characteristic time of the electrochemical system, r2/D , with the experimental time connected through the angular frequency, 2p/v . As was stated in Section 2, a well-behaved impedance function must be Kramers /Kronig transformable, so, the limits of (8a) and (8b) should be evaluated in order to prove that Eq. (11) is bounded. From Eqs. (12) and (13) we obtain (see Appendix A for details): Fig. 1. Nyquist diagram for a Butler /Volmer kinetic process in a inlaid disk electrode. This graph corresponds to Eq. (11) in the text. For high frequencies, Warburg behaviour is observed and plotted by a line with a 458 slope. For low frequencies, diffusive transport resistance is observed. The apex of the graph is evaluated numerically as vmax / 2.5119(D /r2). J. Navarro-Laboulais et al. / Journal of Electroanalytical Chemistry 536 (2002) 11 /18 1 lim F4 (x) pffiffiffi 2 2 1 lim F5 (x) pffiffiffi x0 2 2 x0 (16) (17) 4 1=2 x 3p (18) 1 lim F5 (x) pffiffiffi x x00 4 2 (19) lim F4 (x) x00 The first two limits are related to the high frequency and the macroelectrode response, while the other two are related to the dc response and the steady state microelectrode behaviour. Let us analyse them separately. No simplification or mention of the size of the electrode has been made in the mathematical procedure used to obtain Eq. (8). So, the equation remains valid for micro- and macro-electrodes, where the distinction between both types depends on the characteristic time (r2/D ) and the experimental time. As the radius of the disk electrode increases, the variable x approaches infinity. Then, using Eqs. (16) and (17), the impedance function (Eq. (11)) approaches: s lim Z(x)Rct pffiffiffi x1=2 (1j) 2 2 x0 (20) Using the definition of s and x , we can prove that: 2 RT Z(v)Rct pffiffiffi (1j)v1=2 2 n2 F 2 pr2 D1=2 c+ From this equation we deduce an important consequence because it is related to the dc behaviour of a microelectrode. The limit of Eq. (22) is a real value, which is the sum of two contributions, the charge transfer resistance and the diffusive transport resistance: 4s lim Z(x)Rct Rct RDt x00 3p RT 3RT 2 2 pnFj0 r p rn2 F 2 Dc+ Eq. (23) shows that for low frequencies the impedance of an electrochemical system can be considered as two resistances connected in series, one related to the electron transfer kinetics, and the other to the mass transport towards the electrode. Eq. (23) is also the lower bound of the impedance (11), which then allows the Kramers /Kronig transformation. The value of the mass transport resistance, RDt, is several orders of magnitude higher than the charge transport resistance. This value is experimentally available only for low frequencies using small electrodes. For example, considering a bulk concentration equal to 106 mol cm 3, and a diffusion coefficient of 10 5 cm s 2, at 298.15 K the mass transport resistance is 8094 V for a disk electrode with a radius of 1 cm and 8094 kV for a microelectrode of 10 mm. Summarising, the impedance function (11) which may be called the ‘generalised Warburg impedance’ is bounded and reproduces the micro- and macroelectrode behaviour. This function is a more general impedance function that the Warburg impedance and proves the importance of the geometrical aspects in the electrochemical response of an electrode. Finally, the deduction of Eq. (5) from Eq. (11) explains the apparent anomaly of the Warburg impedance against KK-transformation. The complete function (11) is KK-transformable but its asymptotic behaviour for macroelectrodes is not. (21) This equation matches the Warburg impedance exactly, confirming that the asymptotic behaviour of Eq. (11) is a more general form for the impedance of a macroelectrode. In this situation, the impedance approaches Rct as the frequency tends to zero, which is the higher bound of Eq. (11). Conversely, in the limit when x approaches zero, the expression matches the microelectrode behaviour. Using Eqs. (18) and (19), the impedance (11) is reduced to: 4 x1=2 lim Z(x)Rct s j pffiffiffi (22) x00 3p 4 2 15 4. Frequency limit of the Warburg impedance In Section 3, the connection between the general impedance function (Eq. (11)) and the microelectrode and Warburg impedances has been shown. Thus, it is advisable to have a criterion to distinguish between both types of extreme behaviour. This rule could be deduced from the value of the apex in Fig. 1 which depends on the adimensional frequency, x , defined in Eq. (11), or also from the maximum of the imaginary part of the function (11). Using this last procedure, the problem consists in calculating the frequency at the maximum, vmax, solving the equation: dF5 (x) 2x0 F5 (x0 )0 (24) dx xx0 where x0 (23) This consequence agrees with the empirical observations and the steady state behaviour of the microelectrodes. r2 vmax D (25) Eq. (24) has been deduced by applying the maximum condition to the imaginary part of Eq. (11), and its solution can be obtained by series expansion. Eq. (10) can be rewritten as: 16 J. Navarro-Laboulais et al. / Journal of Electroanalytical Chemistry 536 (2002) 11 /18 0 F5 (x0 ) g 1=2 J21 (bx0 ) sin pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 u tan(u) sin(u) du 2 (26) p=2 with u /arctan b1. Making use of the properties of the derivatives of the Bessel function, the solution of Eq. (24) is the solution of: 1=2 F5 (x0 )2x0 0 g 1=2 1=2 J1 (bx0 ) J0 (bx0 ) sin pffiffiffiffiffiffiffiffiffiffiffiffi 1 u sin(u) du 2 p=2 (27) 0 The previous equation can be solved by expanding the Bessel functions in series and evaluating the resulting integrals. For the lower order terms of these expansions, the integrals can be easily evaluated but this is not the case for higher order terms. Separately, the programming of Eq. (27) allows its numerical evaluation accurately enough. The details of both methods are given in Appendix B leading to the following result: vmax 2:5119 D r2 (28) For an electroactive substance with a coefficient of diffusion of 105 cm2 s 1, the apex of the impedance in a Nyquist diagram is located at 4 mHz for an electrode of 1.0 cm of radius, and 4 mHz for an electrode of 10 mm. Eq. (28) allows the prediction of the effect of the diffusion at the edges of the electrode because this geometrical phenomenon is responsible to some extent for the experimental deviation of macroelectrodes from Warburg behaviour. Furthermore, the frequencies calculated from Eq. (28) are, from an experimental point of view, very low to be attained with sufficient reliability to avoid factors such as the unsteady state or natural mass convection towards the electrode. equation derived from the three-dimensional one to describe the Warburg impedance is the consequence that makes the KK-transformation impossible. The derivation of the Warburg function from the generalised one allows also the obtention of other magnitudes useful from a practical point of view. The use of this function is advisable when small electrodes, not necessarily microelecrodes are used or when the impedance is measured at very low frequencies. A fast algorithm to be used in non-linear fitting programs has been outlined in appendices. The use of the Fleischmann and Pons impedance functions allows the calculation of magnitudes of interest such as the charge transfer resistance, Rct, the constant related to the mass transport, s , and the relaxation time that relates the size of the electrode and the diffusion coefficient, r2/D . An expression for a mass-transfer resistance using the limiting properties of the generalised impedance function has also been obtained. This resistance is several orders of magnitude greater than the charge transfer resistance and was not always experimentally accessible. An estimate of the order of magnitude of the angular frequency at which this resistance is attainable or at which the Warburg impedance begins to fail has also been derived. This magnitude depends on the mass transport towards the electrode and the size of the electrode also. Acknowledgements J.J. Garcia-Jareño acknowledges financial support from the Ministerio de Ciencia y Tecnologı́a (Programa Ramon y Cajal). Financial support from the Ministerio de Ciencia y Tecnologı́a (Project ‘MUVE’ MAT20000100-P4-03) is gratefully acknowledged by the authors. Appendix A: Limits of functions F4(x ) and F5(x) 5. Conclusions Any linear system can be characterised by a transfer function resulting from the system of differential equations, that includes the initial and boundary conditions, describing the physical problem. If the transfer function is lineal, causal, stable and bounded, this function is Kramers /Kronig transformable. In this work the difficulties in carrying out these transformations on the Warburg impedance have been demonstrated. Using a more general model for the impedance of a disk-shaped electrode inlaid in an insulating surface, it was possible to demonstrate the Warburg impedance as a limiting case of this generalised electrochemical impedance with the integral transformations. Then, the loss of information in the one-dimensional differential In order to calculate the Kramers /Kronig transform of the impedance function (11), the limits of functions F4(x ) and F5(x ) must be evaluated. Using the definition of F4(x ) (Eq. (11)) the limit of the argument of the integral for x 0/ is: 2 2 2 1=2 1=2 ] x1=2 2 [y (x y ) (A1) [J1 (y)]2 lim [J1 (y)] 1=2 x0 y(x2 y2 ) yx Then, the limit of F4(x ) is: x1=2 lim F4 (x)$ pffiffiffi x0 2 g 0 1 pffiffiffi 2 2 x1=2 1 dy pffiffiffi [J1 (y)] 2 yx 2 g [J1 (y)]2 dy y 0 (A2) J. Navarro-Laboulais et al. / Journal of Electroanalytical Chemistry 536 (2002) 11 /18 In a similar way, the limit of F5(x ) can be calculated: (A3) The limits of both functions for x 0/0 are calculated using the same properties of the limits of the arguments in the integrals. Using Eq. (11) again, we have: [y2 (x2 y2 )1=2 ]1=2 [J1 (y)]2 lim [J1 (y)]2 (A4) x00 y(x2 y2 )1=2 y2 Thus, the limit of F4(x ) is: lim F4 (x)x 1=2 x00 g [J1 (y)]2 y2 dy 4 3p 1=2 x (A5) Using the same procedure for F5(x ), this function tends to zero for x 0/0, but a more interesting asymptotic behaviour of this function can be obtained if the Bessel function is expanded as a series: x2 x4 5x6 4 16 768 (A6) As x approaches zero, the faster the higher terms of the above expression approach zero and become negligible. Making use of the trigonometric properties, the integral (10), an equivalent equation to Eq. (13), can be written as: 1 F5 (x) pffiffiffi 2 g 4 1=2 [J1 (bx1=2 )]2 0 [(1 b ) 2 1=2 b ] b(1 b4 )1=2 db (A7) Thus, substituting the first term of Eq. (A6) and neglecting the others terms, we have: x F5 (x): pffiffiffi 4 2 g b[(1 b4 )1=2 b2 ]1=2 0 g b2 b4 2 x0 x0 4 16 sin u (tan(u) sin(u))1=2 du 2 p=2 1 4 u sin(u) 1=2 sin du 2 tan(u) 0 x0 g p=2 1 16 0 x20 g sin u sin(u) 1=2 2 tan3 (u) du (A10) p=2 0 [J1 (x)]2 : can be expressed as: 0 1 lim F5 (x) pffiffiffi x0 2 2 17 (1 b4 )1=2 db (A8) The integral in the right hand side of this equation can be solved and is equal to unity. So the asymptotic behaviour of F5(x ) for x 0/0 is: 1 lim F5 (x) pffiffiffi x x00 4 2 (A9) Appendix B: Solution of Eq. (24) The solution of Eq. (24), or its equivalent Eq. (27), implies the evaluation of the integrals which include the unknow, x0. If the Bessel functions are expanded in series, we obtain a series of rational integrals that can be solved and then, a polynomial in x0 is obtained. Using the first two terms of the expression (A6) and neglecting the rest, the integral involved in the definition of F5(x ) where the relation u /arctan b1 has been used. The integrals in Eq. (A11) can be calculated leading to the following expression in x0: pffiffiffi 1 1 1 2 x0 F5 (x0 ): 1 pffiffiffi ln pffiffiffi 4 2 2 21 1 pffiffiffi ( 2 2)x20 (A11) 64 For the evaluation of the integral in the second term in the left hand side of Eq. (27), the expansion of the following function was used: J1 (x) J0 (x) 1 3 3 5 x x x5 2 16 192 (A12) Carrying out the same procedure as before and rewriting Eq. (27), we obtain an algebraic equation that allows us to calculate the apex of the impedance function (11): pffiffiffi pffiffiffi 3 1 1 2 22 1 pffiffiffi ln pffiffiffi x0 4 2 2 21 64 pffiffiffi 3( 2 2) 2 x0 0 (A13) 32 This equation has two solutions, one of them has no physical meaning because it is negative, and the other one is: x0 2:35325 . . . If the number of terms used in the series expansion in Eq. (A10) or in the integral in Eq. (27) is increased, the accuracy of this result will be increased. Unfortunately, the resulting integrals are too complicated and the convergence of the resulting series should be tested. So, an alternative and faster solution is the calculation of x0 by numerical methods. Then, an efficient algorithm for the calculation of F4(x), F5(x ) or any integral involved in the problem is necessary. The Gaussian quadrature is a numerical method that allows the calculation of integrals with high precision using only a few values of the abcissae. The choice of one or other quadrature scheme depends on the weighting function and the limits of the integral. In our case, we have adopted the Gauss /Legendre quadrature J. Navarro-Laboulais et al. / Journal of Electroanalytical Chemistry 536 (2002) 11 /18 18 scheme where the integration interval extends from 0 to 1. For this reason, the function F4(x ) must be written in a more convenient form: F4 (x) g [J1 (bx1=2 )]2 0 1 g [J (bx 1 0 1 g 0 1=2 )]2 cos(u=2) b(1 b4 )1=4 cos(u=2) b(1 b4 )1=4 db db 1=2 2 x cos(tan1 (1 g)2 =2) J1 dg [1 (1 g)4 ]1=4 1g (A14) The gaussian quadrature methods are very efficient numerical methods for continuous and monotonic functions, needing a small number of abcissae evaluations, ranging from 10 to 20. In our case, the integrands of the functions F4(x) and F5(x ) are oscillating functions and the selection of a small number of intervals of integration will lead to a lack of accuracy. So, in order to avoid this inconvenience a number of points large enough has been chosen in the interval. In particular our integration interval, [0, 1] has been divided into 10 000 steps. The evaluation of these integrals is time consuming and the procedure given before is not appropriate for its implementation in non-linear least-squares fitting programs for impedance. One solution of this problem is to tabulate a number of values of F4(x ) and F5(x) large enough for different values of x . Then the calculation of both functions at any value of x could be performed by rational interpolation. Once the algorithm is built, its implementation in a program to obtain the solution of Eq. (24) is straightforward. The algorithm written in standard C can be requested via e-mail from the authors. References [1] J.A. Alden, F. Hutchinson, R.G. Compton, J. Phys. Chem. Sect. B 101 (1997) 949. [2] A.M. Bond, K.B. Oldham, C.G. Zoski, J. Electroanal. Chem. 245 (1988) 71. [3] M. Fleischmann, S. Pons, J. Electroanal. Chem. 250 (1988) 257. [4] M. Fleischmann, J. Daschbach, S. Pons, J. Electroanal. Chem. 250 (1988) 269. [5] M. Fleischmann, S. Pons, J. Electroanal. Chem. 250 (1988) 277. [6] M. Fleischmann, S. Pons, J. Daschbachnn, J. Electroanal. Chem. 317 (1991) 1. [7] M.A. Bender, H.A. Stone, J. Electroanal. Chem. 351 (1993) 29. [8] M.V. Mirkin, A.J. Bard, J. Electroanal. Chem. 323 (1992) 1. [9] A.J. Bard, M.V. Mirkin (Eds.), Scanning Electrochemical Microscopy, Marcel Dekker, New York, 2001. [10] J. Booth, R.G. Compton, J.A. Cooper, R.A.W. Dryfe, A.C. Fisher, C.L. Davies, M.K. Walters, J. Phys. Chem. 99 (1995) 10942. [11] D. Pletcher, F.C. Walsh, Industrial Electrochemistry, 2nd ed., Chapman & Hall, London, 1990. [12] J. Navarro-Laboulais, J. Trijueque, J.J. Garcı́a-Jareño, F. Vicente, J. Electroanal. Chem. 399 (1995) 115. [13] J. Navarro-Laboulais, J. Trijueque, J.J. Garcı́a-Jareño, F. Vicente, J. Electroanal. Chem. 442 (1997) 91. [14] J. Navarro-Laboulais, J. Trijueque, J.J. Garcı́a-Jareño, D. Benito, F. Vicente, J. Electroanal. Chem. 443 (1998) 41. [15] J. Navarro-Laboulais, J. Vilaplana, J. López, J.J. Garcı́a-Jareño, D. Benito, F. Vicente, J. Electroanal. Chem. 484 (2000) 33. [16] K.W. Yu, P.Y. Tong, Phys. Rev. B 16 (1992) 11487. [17] P.Y. Tong, K.W. Yu, J. Phys. A. Math. Gen. 26 (1993) L119. [18] J. Machta, R.A. Gruyer, S.M. Moore, Phys. Rev. B 33 (1986) 4818. [19] L. De Arcangelis, S. Redner, A. Coniglio, Phys. Rev. B 31 (1985) 4725. [20] L. De Arcangelis, S. Redner, A. Coniglio, Phys. Rev. B 34 (1986) 4656. [21] R.F. Angulo, E. Medina, J. Stat. Phys. 75 (1994) 135. [22] H. Reller, E. Kirowa-Eisner, E. Gileadi, J. Electroanal. Chem. 138 (1982) 65. [23] H. Reller, E. Kirowa-Eisner, E. Gileadi, J. Electroanal. Chem. 161 (1984) 247. [24] E. Gileadi, Electrode Kinetics, VCH, New York, 1993. [25] B.R. Scharifker, J. Electroanal. Chem. 240 (1988) 61. [26] C. Amatore, J.M. Savéant, D. Tessier, J. Electroanal. Chem. 147 (1983) 39. [27] D. Shoup, A. Szabo, J. Electroanal. Chem. 160 (1984) 19. [28] J. Navarro-Laboulais, J. Trijueque, J.J. Garcı́a-Jareño, F. Vicente, J. Electroanal. Chem. 399 (1995) 115. [29] J. Navarro-Laboulais, J. Trijueque, J.J. Garcı́a-Jareño, D. Benito, F. Vicente, J. Electroanal. Chem. 444 (1998) 173. [30] L. Beaunier, M. Keddam, J.J. Garcı́a-Jareño, F. Vicente, J. Navarro-Laboulais, 14ème FORUM sur les Impedances Electrochimiques, Paris, 2002. [31] B.A. Boukamp, J. Electroanal. Chem. 142 (1995) 1885. [32] J.R. Macdonald, Impedance Spectroscopy, John Wiley & Sons, New York, 1987. [33] A.J. Bard, L.R. Faulkner, Electrochemical Methods, John Wiley & Sons, New York, 1980. [34] J.R. Macdonald, Electrochim. Acta 38 (1993) 1883. [35] B.A. Boukamp, J.R. Macdonald, Solid State Ionics 74 (1994) 85. [36] G. Làng, L. Kocsis, G. Inzelt, Electrochim. Acta 38 (1993) 1047.