- Email: [email protected]
Electrocatalysis in water electrolysis with solid polymer electrolyte
Electrochimica Acta 48 (2003) 3945 /3952 www.elsevier.com/locate/electacta
Electrocatalysis in water electrolysis with solid polymer electrolyte Egil Rasten 1, Georg Hagen 2, Reidar Tunold 2,* Department of Materials Technology and Electrochemistry, Norwegian University of Science and Technology, NO 7491 Trondheim, Norway Received 20 January 2003; received in revised form 16 April 2003; accepted 19 April 2003
Abstract Powders of IrO2 were used as anode catalysts in water electrolysis cells with solid polymer electrolyte (SPE). The catalyst was prepared by a pyrolysis process in a nitrate melt at 340 8C and then annealed at different temperatures from 440 to 540 8C. The catalyst materials were applied to an electrode membrane assembly (MEA) and studied in situ in an electrolysis cell using cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS) and stationary current density /potential relations. The impedance results clearly show that the non-annealed material possesses the highest electrocatalytic activity and that this activity deteriorates with increasing annealing temperature. The electrical conductivity, however, increases by increasing annealing temperature. Optimum annealing conditions were found at 490 8C, where the total polarisation reaches a minimum in the high current density range (1 /2 A cm 2), at the actual conditions. Increased crystallinity and size of the particles with increasing annealing temperature was observed by transmission electron microscopy (TEM). A very high electrochemical performance was obtained in a water electrolysis cell using IrO2 on the anode side and Pt on Vulcan XC-72 on the cathode side, corresponding to a current density of 1 A cm 2 at 1.60 V. # 2003 Elsevier Ireland Ltd. All rights reserved. Keywords: Electrocatalysis; Water electrolysis; Oxygen electrode; Iridium oxide; Solid polymer electrolyte
1. Introduction In a future energy scenario based on renewable energy, hydrogen is an attractive energy carrier and advanced water electrolysis will be the most efficient and practical production process of hydrogen. It is very important to reduce production costs by improving the specific performance of electrode catalysts, in order to increase specific production rate at lowered specific electric power consumption. Water electrolysis with a solid polymer electrolyte (SPE)-cell possesses certain advantages compared with the classical alkaline process like: increased energy efficiency and specific production
From a lecture given at the 4th International Conference ‘Electrocatalysis: From Theory to Industrial Applications’, 22 /25 September 2002, Como, Italy. * Corresponding author. Tel.: /47-73-594-043; fax: /47-73-3594083. E-mail address: [email protected] (R. Tunold). 1 Present address: Norsk Hydro ASA, NO 3907 Porsgrunn, Norway. 2 ISE member.
capacity and a simple system with a solid electrolyte at a low temperature. The first electrolysers using polymer membranes as electrolyte were developed by General Electric Co. in 1966 for space applications [1]. Extensive research and development on SPE electrolysers has been performed within the Japanese WE-NET program [2 /4]. A cell voltage of 1.68 V at 1 A cm 2 was obtained on a single cell of 50 cm2 at 80 8C. Kondoh et al. [5] tested two cell stacks of 0.25 m2 /10 for 2500 h at 1 A cm 2 with an average cell voltage of 1.74 V at 1 A cm 2 and 1.9 V at 2 A cm 2, respectively. ABB developed the commercial Membrel electrolyser in the period 1976 /1989. Two units of 100 kW were long term tested with an average voltage 1.75 V at 1 A cm 2 and 80 8C [6]. The oxygen electrode is the main source of overpotential in this system. In contact with the acidic Nafion† membrane non-noble catalytic metals like Ni and Co will corrode and Pt will be covered by a lowconducting oxide film. After Henry B. Beer obtained his British patent on the dimensionally stable anodes (DSA) in 1965 [7], oxides of Ir and Ru have been used as catalysts on titanium substrates in DSA. For the chlor-
0013-4686/03/$ - see front matter # 2003 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.electacta.2003.04.001
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
3946
alkali process RuO2 has taken over the whole market. For oxygen evolving electrodes IrO2 based catalytic layers are preferred because of the higher stability [8,9]. RuO2 and IrO2 are often mixed with inert components like Ta2O5, TiO2 or SnO2 to stabilize the structure [7,10]. As oxygen evolving catalyst in SPE electrolysis cells, IrO2, either pure [11,12] or mixed with RuO2 [13], and even IrO2 /Pt [3,14,15] composites, have been used. The Ir and Ru materials were in oxidic form, mostly produced by thermal oxidation, since metallic Ir and Ru have been observe to be unstable [14]. On the hydrogen side metallic Pt is commonly used [3,14,15]. The scope of the present work has been to develop a SPE system with high energy efficiency combined with a high production capacity at a low loading of noble metal oxide catalyst. The work includes development of processes for catalyst production and application on electrode membrane assemblies (MEAs). Focus is on the oxygen electrode, which is the main source of overpotential in this system, and pure IrO2 was chosen as the active catalyst for the electrocatalytic process.
2. Experimental Chemicals used for preparation of metal oxide catalysts and for preparation of MEAs were pa quality; H2IrCl6 from Johnson Matthey via Alfa Aesar, Nafion† membrane and Nafion† solution from Du Pont via ElectroChem Inc. and 10/40 wt.%Pt on carbon from Johnson Matthey, E-TEK, Inc. All the water used was tapped from a Milli-Q Ultra Pure Water system with a water resistivity of 18.2 MV cm. A sketch of the single circular SPE cell is shown in Fig. 1. The cell body was made of Ti with flow fields machined into the Ti-plates. Porous Ti discs functioned as backings and were pressed by the Ti-plates against the MEAs. The total ohmic resistance of the cell hardware
Fig. 1. Sketch of the SPE cell.
was determined by mounting the cell together without the MEA, applying a current of 10 A between the end plates and measuring the voltage drop. When pristine Ti-sinters were applied, a voltage drop of 5 mV was obtained, corresponding to a total resistance of 0.5 mV. When the resistance increased to twice this value during use, the Ti-sinters were replaced by new ones. A Hg/ Hg2SO4 reference electrode was used for measuring the anode and cathode potentials and was made by a glass capillary, filled with Nafion† membrane material and sealed with epoxy on the top with a Pt wire for electrical contact. 2.1. Preparation of metal oxides Preparation of noble metal oxides like platinum oxide, by fusion of chloroplatinic acid and sodium nitrate was first described by Adams and Shriner [16]. This method has successfully been used for preparation of anode catalysts in SPE electrolysis systems by several groups [17 /19] and was found to be fast and convenient for screening of different catalysts. During the fusion process nitrogen dioxide is evolved and noble metal oxide is precipitated. In our preparations 1 g of metal chloride precursor was mixed in an excess of 50 g of NaNO3 and dissolved in water. The water was carefully evaporated from the mixture and the salt mixture was introduced into a ceramic furnace, pre-heated at 340 8C, and kept at these conditions for 30 min until evaporation of the dark brown nitrous gas ceased. The salt mixture was thereafter cooled to room temperature and washed in water to remove excess salt. The resulting metal oxide was then dried and annealed in an electric furnace at a predetermined time and temperature. The active area was 5 cm2 of all samples. 2.2. Preparation of MEAs A short screening process of different MEA-preparation methods was carried out at the beginning of this study and a spraying technique was chosen. Spraying the catalyst layer either directly onto the membrane [20,21] or onto a backing/current collector substrate [21,22] is a commonly applied technique for preparation of catalytic layers in PEM-fuel cells. The spraying technique was preferred due to its simplicity and easy control of the catalyst loading. The catalyst was sprayed from an ink composed of nano-sized catalyst particles in a mixture of water, ethanol, glycerol and ionomer, ultrasonically mixed beforehand. The spraying process was carried out by using an air driven spray gun where an exact composition of the slurry was deposited to a loading of 2 mg IrO2 cm 2 and a Nafion† ionomer content of 20 wt.%. Before spraying, the membrane was treated according to a procedure described by Srinivasan and co-workers [23]. Membranes were first pre-
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
3947
heated in purified water, treated by H2O2 to remove organic impurities, further treated in hot sulphuric acid solution to ion exchange the membrane completely, then boiled in purified water and finally stored in purified water. 2.3. Experimental methods and procedures The catalyst particles were examined in a JEOL 2010 transmission electron microscope and catalytic layers in a Zeiss DSM 950 scanning electron microscope. Electrode potentials were determined using an in situ Hg/ Hg2SO4 electrode connected to the membrane by a glass capillary filled with wetted polymer membrane material. The reference electrode was placed in ionic contact with a part of the membrane protruding from the gaskets between the cell body and the membrane. All measurements using the reference electrode were carried out only at room temperature as the contact point dried out too fast at higher temperatures. The electrochemical performance of both electrodes in the SPE-cell was determined by stationary polarisation curves at galvanostatic conditions up to 2 A cm 2, with cyclic voltammetry (CV) and with electrochemical impedance spectroscopy (EIS) in the frequency range of 10 mHz /10 kHz. Impedance data were modelled by using ZView software from Scribner Associates.
Fig. 2. TEM and diffraction image of: (a) non-annealed sample; and (b) sample annealed at 490 8C.
3. Results and discussion 3.1. Effect of annealing temperature on the IrO2 catalyst Transmission electron microscopy (TEM) images of a non-annealed sample and a sample annealed at 490 8C are shown in Fig. 2 a and b, respectively. Corresponding diffraction images are shown in the lower right corner of the pictures. The average particle size of the nonannealed sample is observed to be in the range of 5/ 20 nm. These particles, however, seem to consist of agglomerates of smaller particles of 1 /2 nm. The material is probably amorphous, recognized by the diffuse and broad rings and the absence of speckles in the images. The annealed samples exhibit a wider range of particle sizes, from 5 to 100 nm, where the larger particles seem to consist of clusters of particles with diameters of 5/10 nm. Diffraction images of the annealed samples reveal a high degree of crystallinity. Annealing then leads to crystallization and to crystal growth. 3.2. Cyclic voltammetry Fig. 3 shows voltammograms obtained at 20 mV s 1 on samples with different annealing conditions. The non-annealed sample in Fig. 3 exhibits a large area in
Fig. 3. Voltammograms of IrO2 catalysts annealed at different temperatures for 8 h.
the oxygen region and with a very low hydrogen adsorption. The voltammetric charge in the oxygen region of the annealed samples decreases with increasing annealing temperature as also is the case with the capacitance values shown in Table 1. The two peaks at /0.9 and 1.2 VRHE have earlier been attributed to redox couples Ir(III)/Ir(IV) and Ir(IV)/Ir(VI), respectively [24,25]. The voltammetric current and therefore Table 1 Capacitance from voltammetry as a function of annealing conditions Sample (8C)
Capacitance/mF cm 2
Non-annealed 440 490 510 540
283 128 110 88 66
3948
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
also the voltammetric capacitance contains contributions from double layer capacitance and pseudocapacitances due to adsorption and redox processes. All these processes, however, depend on surface area. The capacitance values were calculated from the anodic redox peak in the potential region 1.1 /1.2 VRHE by using the relation C /i/(dE /dt) at scan rates between 10 and 100 mV s 1 where C includes both the pseudocapacitance and the dobbel layer capacitance. The very distinct and narrow peaks at about 0.6 V and especially at 0.9 /1.0 V on the non-annealed sample, are especially difficult to explain. These peaks tend to disappear for higher annealing temperatures. One explanation could be impurities, disappearing with annealing. Another and a rather suggestive explanation are changes in electronic energy levels due to the noncrystallinity and nanosize of the non-annealed particles. The results show that the active surface decreases with increasing annealing temperature. If the capacitance is a direct measure of the surface area it decreases to about 25% of the non-annealed sample when annealing at 540 8C. The voltammogram of the non-annealed sample resembles that of an anodically grown iridium oxide identified by the narrow hydrogen region and a relatively broad oxygen region [24]. Anodically formed hydroxides/oxides are more porous than thermally prepared oxides and a larger part of the interior area is available for the electrochemical processes [26]. This might explain the larger voltammetric charge for the non-annealed sample. The decrease in voltammetric charge after annealing can partly be due to area reduction associated with particle growth and coalescence and partly due to the dehydration of the oxide with stronger oxide bridge bonding [8]. 3.3. Impedance spectroscopy The impedance plots in Fig. 4 reveal that the annealing conditions have a considerable effect on the properties of the catalytic layers. The plots are determined at a potential of 1.49 VRHE which corresponds to current densities of 1 /50 mA cm 2 (see Fig. 12). The high frequency intercept with the real axis is very high for the non-annealed sample and decreases significantly
Fig. 4. AC-impedance of IrO2 catalysts as a function of annealing temperature, measured at 1.49 VRHE during oxygen evolution.
when the catalyst is annealed, up to 490 8C, from where it becomes constant with temperature. The diameter of the low frequency arc, which should be a measure of the polarisation resistance and thereby of the catalytic activity of the electrode, is lowest for the non-annealed sample and increases continuously with increasing annealing temperature. These results are probably connected with the crystal growth of IrO2 with increasing annealing temperature followed by a lower electrocatalytic activity. The corresponding increased crystallinity of the material, however, leads to an increase in the electrical conductivity. The combined effect of annealing conditions and potential is shown in Figs. 5 /7 for the non-annealed sample and the samples annealed at 440 and 490 8C, respectively, where the impedance is measured for each sample between 1.54 and 1.61 VRHE at 10 mV increments. The impedance arc of the non-annealed sample in Fig. 5 shifts towards lower values along the real axis as the potential increases above 1.50 VRHE, which corresponds to a decrease in the electric resistance as the potential increases. For the material annealed at 440 8C a second, high frequency arc appears and takes over with increasing potential at 1.58 VRHE and above. This is even more visible in the Bode plot in Fig. 8. The time constant for the low frequency process (t1) disappears and the time constant for the high frequency process (t2) appears more clearly with increasing potential. The impedance data was modelled by using the circuit RV(R1Q1)(RctQdl), where RV is the ohmic resistance, Rct is the charge transfer resistance. The capacitive elements are represented by Q , a constant phase element often used to model depressed semi circles due to heterogeneities and rough surfaces. The impedance of a capacitive element is given by Q /1/(B (iv ))n where B is a frequency independent constant, v is the angular pffiffiffiffiffiffi frequency, i 1 and n is a factor ranging between 0 and 1. Here, Qdl includes both the double layer capacitance and the pseudocapacitance. This type of model has been used in the literature to interpret the
Fig. 5. Impedance of non-annealed sample as a function of potential in 10 mV steps from 1.54 to 1.61 VRHE during oxygen evolution.
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
Fig. 6. Impedance of sample annealed at 440 8C as a function of potential in 10 mV steps from 1.54 to 1.61 VRHE during oxygen evolution.
Fig. 7. Impedance of sample annealed at 490 8C as a function of potential in 10 mV steps from 1.54 to 1.61 VRHE.
3949
phenomenon determining the first time constant, the circuit above was used to model the impedance data. The n value was found to be close to 1 for Qdl whereas it ranged between 0.6 and 0.8 for Q1. Fig. 9 shows 1/R1 and 1/Rct as a function of potential on a logarithmic scale, found by modelling the impedance data of the sample annealed at 440 8C. It is observed that the obtained values for both parameters can be represented by straight lines. Straight lines of only one decade of current density is, however, somewhat insufficient to determine a real Tafelian behaviour. Reliable data was difficult to obtain at higher potentials due to the very low impedances of these systems and the limitation of the potentiostat. Additional importance, however, is given by the fact that the slopes were found to correspond well with the slopes found by stationary polarisation measurements (see Table 2), indicating that both time constants are related to electrochemical processes. The line for the high frequency process intercepts the line for the low frequency process at about 1.50 VRHE. In Figs. 10 and 11 Mott /Schottky plots are shown for non-annealed material and material annealed at 490 8C, respectively, based on the Qdl values found by fitting the impedance data to the model presented above. The plots reveal that the non-annealed material probably is a semiconductor which changes from n -type conduction to p -type conduction with increasing potential. This could be connected to an oxidation of iridium to a higher valence state, probably hexavalent, at a certain potential. The oxidation then takes place at a higher potential when the material is stabilized by increased annealing temperature. 3.4. Steady state polarisation curves
Fig. 8. Bode plot, face angle u vs. frequency f of sample annealed at 440 8C, as function of potential.
impedance data on different noble metal oxides and their compositions, where the R1Q1 circuit has been attributed to different phenomena, e.g. diffusion of protons along the oxide grains on thermally prepared RuO2 [27], diffusion of reduced oxide sites on anodically grown IrO2 [24] and also to the interlaying oxide layer on the titanium substrate [24,28,29]. From the Bode plots in Fig. 8 it is clear that there are at least two time constants present, however, since there is no interlaying oxide layer of titanium, both time constants are more likely to be related to steps in the oxygen evolution reaction (OER). Regardless of what
Steady state polarisation curves for non-annealed material and for samples annealed at temperatures ranging from 440 to 540 8C, are shown in Fig. 12. The potentials are not corrected for IR drops. At low current densities the non-annealed material exhibits the best
Fig. 9. Reaction admittances 1/R1 and 1/Rct on a logarithmic scale as a function of potential of the sample annealed at 440 8C by fitting of impedance data to the circuit model RV(R1Q1)(RctQdl).
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
3950
Table 2 Tafel coefficients from polarisation curves and Tafel coefficients and ohmic resistance values from modelling of EIS data Annealing conditions (8C)
Non-annealed 440 490 510 540
Polarisation curves
AC impedance
ba, low cd (mV dec 1)
ba, high cd (mV dec 1)
ba, low/high cd (mV dec 1)
RV (V cm2)
36 43 46 47 49
89 145 151 146
36 38/95 42 44 48
0.15 /0.34 0.09 /0.11 0.05 0.05 0.05
Fig. 10. Mott /Schottky plot of non-annealed sample.
Fig. 11. Mott /Schottky plot of sample annealed at 490 8C.
Fig. 12. Steady state anodic polarisation curves as a function of different annealing conditions of IrO2 catalyst during oxygen evolution.
catalytic properties with the lowest total overpotential. The potential then increases with increasing annealing temperature. At high cds , above 200 mA cm 2, the curves cross each other and with a very steep raise of the slope of the curves for the non-annealed sample and the sample annealed at 440 8C. These results confirm the findings both by voltammetry and EIS. At low cds the catalytic properties of the material are governing the
potential, whereas the ohmic resistance of the catalyst layer determines the potential at high cds . This means that an optimum is found at 490 8C where the total anodic potential in the high cd range of 1 /2 A cm 2, is lowest. There seems to exist three factors related to the effect of the particle size on the properties of the catalysts. The most obvious is the effect on the active surface, which increases with decreasing size. A second effect is on the electronic properties of the material, affecting both electronic conductivity and electrocatalytic properties. The contact resistance through the catalytic layer will also depend on the particle size, where larger particles could create less interfacial contact resistance. Such an effect of the particle size can also have influenced the results in this work. However, the electrical conductivity properties of the non-annealed sample and the sample annealed at 440 8C is typical for semiconductors and cannot be related to this effect. Problems with resistance control have not been emphasised in the use of DSA electrodes. The reason for this is rather obvious. In the DSA technology the catalytic oxide mixture is formed on a dense, solid metal substrate. The Ti backing/current collector in the present system is very porous with pore openings ranging from 10 to 50 mm whereas the thickness of the catalytic anode layer is about 10 mm. This results in a considerable lateral resistance in the catalytic layer depending on the electronic conductivity. 3.5. Possible mechanism of oxygen evolution on iridium oxide in SPE systems Tafel constants from IR compensated steady state polarisation curves are shown in Table 2. Values for both low and high cds are shown. In addition values of the slope of E /log (1/R ) diagrams from EIS are shown. Here RV is the series ohmic resistance. The impedance data suggest that the oxygen evolution process consists of two succeeding charge transfer steps. The Tafel gradient of about 40 mV/decade for the low cd range demonstrates that the second step is rate determining at these conditions and that a primary one electron charge transfer is rate determining at high cd, with a Tafel slope
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
varying from about 90 mV/decade for non-annealed material and the sample annealed at 440 8C, to 140 /150 mV/decade for materials annealed at higher temperatures. This is in rather good accordance with the process following the so called ‘Electrochemical Oxide Path’ [30] in three steps, where the first step is a charge transfer step with the formation of an adsorbed hydroxy species on an active site S : SH2 O 0 SOHads H e
(1)
The next step is a second charge transfer step that deprotonates the adsorbent: SOHads 0 SOads H e
(2)
The third step is then the formation of an oxygen molecule and the back formation of two active sites: 2SO 0 2SO2(g)
(3)
That the Tafel slopes increases by increasing annealing temperature might be explained by the increasing stability of the metal /oxygen bond. One possible explanation is that this could increase the activation energy leading to a change in the charge transfer coefficient a and thereby changing the Tafel slope.
3.6. Performance of a PEM electrolysis cell Fig. 13 shows the performance of a total PEM electrolysis cell based on an IrO2 anode and a cathode with 10% Pt on Vulcan XC-72. The total noble metal loading was 2.4 mg cm 2 and the temperature 80 8C and a cell voltage of 1.65 V at 1 A cm 2 (10 kA m 2) was obtained. Interesting is that no sign of transport hindrance is observed up to a current density of 2 A cm 2. This diagram also includes the cathodic overpotential, since it was impossible to use the reference electrode due to drying out of the protruding membrane lip at such high temperatures. The hydrogen overvoltage is less than 150 mV up to cds of nearly 1 A ×/cm 2 [31].
Fig. 13. Cell voltage as a function of current density, measured at 80 8C.
3951
4. Conclusions Iridium oxide is an interesting electrocatalyst for oxygen evolution in a water electrolysis cell using a polymer membrane electrolyte. The catalytic and electronic properties of the catalyst vary considerably with the heat treatment of the material, after producing it in a pyrolysis process in a nitrate melt at 340 8C. The nonannealed material consists of nanosized particles with low crystallinity, as observed by TEM. This material possesses a high electrocatalytic activity but a low electronic conductivity. The TEM study reveals that the particle size increases and the material becomes more crystalline with increasing annealing temperature in the range from 440 to 540 8C, leading to lower electrocatalytic activity and higher electronic conductivity. It is likely that the non-annealed material is a semiconductor that changes from n- to p-type conduction mode in the actual potential range. When increasing the crystallinity of the material by increasing the annealing temperature, the electronic conduction increases and reaches a constant value (see Table 2). The total polarisation of the oxygen electrode then reaches an optimum value, at these conditions, at a temperature of 490 8C. Combined results from EIS and polarisation curves indicate that the reaction mechanism of the oxygen electrode follows the so-called ‘Electrochemical Oxide Path’. Very high performance of a total electrolysis cell can be obtained using a polymer electrolyte with iridium oxide catalyst for the oxygen electrode and platinum for the hydrogen electrode, using totally 2.5 mg of noble metal catalyst per cm2 MEA area.
Acknowledgements The authors acknowledge the financial support from The Norwegian Research Council.
References [1] W. Kreuter, H. Hofmann, Hydrogen Energy Progress XI, 1 (1996) 537. [2] C. Mitsugi, A. Harumi, F. Kenzo, Int. J. Hydrogen Energy 23 (1998) 159. [3] H. Mori, C. Inazumi, M. Kato, S. Maezawa, K. Oguro, H. Takenaka, E. Torikai, Hydrogen Energy Progress XI, 1 (1996) 579. [4] C. Inazumi, H. Mori, M. Kato, S. Meazawa, N. Sawai, K. Oguro, H. Takenaka, Hydrogen Energy Progress XI, 1 (1996) 579. [5] M. Kondoh, N. Yokoyama, C. Inazumi, S. Maezawa, N. Fujiwara, Y. Nishimura, K. Oguro, H. Takenaka, J. New Mat. Electrochem. Syst. 3 (2000) 61. [6] S. Stucki, G.G. Scherer, S. Schlagowski, E. Fischer, J. Al. Electrochem. 28 (1998) 1041. [7] H.B. Beer, British Pat. no. 1,147,442 ,1969.
3952
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
[8] L.M. Da Silva, J.F.C. Boodts, L.A. De Faria, Electrochim. Acta 45 (2000) 2719. [9] R. Ko¨tz, S. Stucki, Electrochim. Acta 31 (1986) 1311. [10] O. De Nora, G. Bianchi, A. Nidola, G. Trisoglio, US Pat. no. 3,926,751, 1975. [11] F. Andolfatto, R. Durand, A. Michas, P. Millet, P. Stevens, Int. J. Hydrogen Energy 19 (1994) 421. [12] K. Ledjeff, F. Mahlendorf, V. Peinecke, A. Heinzel, Electrochim. Acta 40 (1995) 315. [13] F. Mahlendorf, PhD thesis, Albert-Ludwigs-Universita¨t Freiburg im Breisgau, 1995. [14] P. Millet, T. Alleau, R. Durand, J. Appl. Electrochem. 23 (1993) 322. [15] T. Ioroi, N. Kitazawa, K. Yasuda, Y. Yamamoto, H. Takenaka, J. Electrochem. Soc. 147 (2000) 2018. [16] R. Adams, R.L. Shriner, J. Am. Chem. Soc. 45 (1923) 2171. [17] J.L. Weininger, R.R. Russel, J. Electrochem. Soc. 9 (1978) 1482. [18] R. Hutchings, K. Mu¨ller, R. Ko¨tz, S. Stucki, J. Matet. Sci. 19 (1984) 3987. [19] L.L. Swette, A.B. LaConti, A. McCatty, J. Power Sources 47 (1994) 343.
[20] S. Møller Holst, PhD thesis, Norwegian University of Science and Technology, 1996. [21] M.S. Wilson, S. Gottesfeld, J. Electrochem. Soc. 139 (1992) L28. [22] E. Passalacqua, F. Lufrano, G. Squadrito, A. Patti, L. Giorgi, Electrochim. Acta 46 (1996) 799. [23] E.A. Ticianelli, C.R. Derouin, A. Redondo, S. Srinivasan, J. Electrochem. Soc. 135 (1988) 2209. [24] L.A. da Silva, V.A. Alves, M.A.P. da Silva, S. Trasatti, J.F.C. Boodts, Electrochim. Acta 42 (1997) 271. [25] F.I. Mattos-Costa, P. De Lima-Neto, S.A.S. Machado, L.A. Avaca, Electrochim. Acta 44 (1998) 1515. [26] S. Ardizzone, G. Fregonara, S. Trasatti, Electrochim. Acta 35 (1990) 263. [27] S.H. Glarum, J.H. Marshall, J. Electrochem. Soc. 127 (1980) 2727. [28] T.A.F. Lassali, J.F.C. Boodts, L.O.S. Buoho˜es, Electrochim. Acta 44 (1999) 4203. [29] V.A. Alves, L.A. Da Silva, J.F.C. Boodts, J. Appl. Electrochem. 28 (1998) 899. [30] J.O’M. Bockris, J. Chem. Phys. 24 (1956) 817. [31] E. Rasten, G. Hagen, R. Tunold, To be published.
Electrocatalysis in water electrolysis with solid polymer electrolyte Egil Rasten 1, Georg Hagen 2, Reidar Tunold 2,* Department of Materials Technology and Electrochemistry, Norwegian University of Science and Technology, NO 7491 Trondheim, Norway Received 20 January 2003; received in revised form 16 April 2003; accepted 19 April 2003
Abstract Powders of IrO2 were used as anode catalysts in water electrolysis cells with solid polymer electrolyte (SPE). The catalyst was prepared by a pyrolysis process in a nitrate melt at 340 8C and then annealed at different temperatures from 440 to 540 8C. The catalyst materials were applied to an electrode membrane assembly (MEA) and studied in situ in an electrolysis cell using cyclic voltammetry (CV), electrochemical impedance spectroscopy (EIS) and stationary current density /potential relations. The impedance results clearly show that the non-annealed material possesses the highest electrocatalytic activity and that this activity deteriorates with increasing annealing temperature. The electrical conductivity, however, increases by increasing annealing temperature. Optimum annealing conditions were found at 490 8C, where the total polarisation reaches a minimum in the high current density range (1 /2 A cm 2), at the actual conditions. Increased crystallinity and size of the particles with increasing annealing temperature was observed by transmission electron microscopy (TEM). A very high electrochemical performance was obtained in a water electrolysis cell using IrO2 on the anode side and Pt on Vulcan XC-72 on the cathode side, corresponding to a current density of 1 A cm 2 at 1.60 V. # 2003 Elsevier Ireland Ltd. All rights reserved. Keywords: Electrocatalysis; Water electrolysis; Oxygen electrode; Iridium oxide; Solid polymer electrolyte
1. Introduction In a future energy scenario based on renewable energy, hydrogen is an attractive energy carrier and advanced water electrolysis will be the most efficient and practical production process of hydrogen. It is very important to reduce production costs by improving the specific performance of electrode catalysts, in order to increase specific production rate at lowered specific electric power consumption. Water electrolysis with a solid polymer electrolyte (SPE)-cell possesses certain advantages compared with the classical alkaline process like: increased energy efficiency and specific production
From a lecture given at the 4th International Conference ‘Electrocatalysis: From Theory to Industrial Applications’, 22 /25 September 2002, Como, Italy. * Corresponding author. Tel.: /47-73-594-043; fax: /47-73-3594083. E-mail address: [email protected] (R. Tunold). 1 Present address: Norsk Hydro ASA, NO 3907 Porsgrunn, Norway. 2 ISE member.
capacity and a simple system with a solid electrolyte at a low temperature. The first electrolysers using polymer membranes as electrolyte were developed by General Electric Co. in 1966 for space applications [1]. Extensive research and development on SPE electrolysers has been performed within the Japanese WE-NET program [2 /4]. A cell voltage of 1.68 V at 1 A cm 2 was obtained on a single cell of 50 cm2 at 80 8C. Kondoh et al. [5] tested two cell stacks of 0.25 m2 /10 for 2500 h at 1 A cm 2 with an average cell voltage of 1.74 V at 1 A cm 2 and 1.9 V at 2 A cm 2, respectively. ABB developed the commercial Membrel electrolyser in the period 1976 /1989. Two units of 100 kW were long term tested with an average voltage 1.75 V at 1 A cm 2 and 80 8C [6]. The oxygen electrode is the main source of overpotential in this system. In contact with the acidic Nafion† membrane non-noble catalytic metals like Ni and Co will corrode and Pt will be covered by a lowconducting oxide film. After Henry B. Beer obtained his British patent on the dimensionally stable anodes (DSA) in 1965 [7], oxides of Ir and Ru have been used as catalysts on titanium substrates in DSA. For the chlor-
0013-4686/03/$ - see front matter # 2003 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.electacta.2003.04.001
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
3946
alkali process RuO2 has taken over the whole market. For oxygen evolving electrodes IrO2 based catalytic layers are preferred because of the higher stability [8,9]. RuO2 and IrO2 are often mixed with inert components like Ta2O5, TiO2 or SnO2 to stabilize the structure [7,10]. As oxygen evolving catalyst in SPE electrolysis cells, IrO2, either pure [11,12] or mixed with RuO2 [13], and even IrO2 /Pt [3,14,15] composites, have been used. The Ir and Ru materials were in oxidic form, mostly produced by thermal oxidation, since metallic Ir and Ru have been observe to be unstable [14]. On the hydrogen side metallic Pt is commonly used [3,14,15]. The scope of the present work has been to develop a SPE system with high energy efficiency combined with a high production capacity at a low loading of noble metal oxide catalyst. The work includes development of processes for catalyst production and application on electrode membrane assemblies (MEAs). Focus is on the oxygen electrode, which is the main source of overpotential in this system, and pure IrO2 was chosen as the active catalyst for the electrocatalytic process.
2. Experimental Chemicals used for preparation of metal oxide catalysts and for preparation of MEAs were pa quality; H2IrCl6 from Johnson Matthey via Alfa Aesar, Nafion† membrane and Nafion† solution from Du Pont via ElectroChem Inc. and 10/40 wt.%Pt on carbon from Johnson Matthey, E-TEK, Inc. All the water used was tapped from a Milli-Q Ultra Pure Water system with a water resistivity of 18.2 MV cm. A sketch of the single circular SPE cell is shown in Fig. 1. The cell body was made of Ti with flow fields machined into the Ti-plates. Porous Ti discs functioned as backings and were pressed by the Ti-plates against the MEAs. The total ohmic resistance of the cell hardware
Fig. 1. Sketch of the SPE cell.
was determined by mounting the cell together without the MEA, applying a current of 10 A between the end plates and measuring the voltage drop. When pristine Ti-sinters were applied, a voltage drop of 5 mV was obtained, corresponding to a total resistance of 0.5 mV. When the resistance increased to twice this value during use, the Ti-sinters were replaced by new ones. A Hg/ Hg2SO4 reference electrode was used for measuring the anode and cathode potentials and was made by a glass capillary, filled with Nafion† membrane material and sealed with epoxy on the top with a Pt wire for electrical contact. 2.1. Preparation of metal oxides Preparation of noble metal oxides like platinum oxide, by fusion of chloroplatinic acid and sodium nitrate was first described by Adams and Shriner [16]. This method has successfully been used for preparation of anode catalysts in SPE electrolysis systems by several groups [17 /19] and was found to be fast and convenient for screening of different catalysts. During the fusion process nitrogen dioxide is evolved and noble metal oxide is precipitated. In our preparations 1 g of metal chloride precursor was mixed in an excess of 50 g of NaNO3 and dissolved in water. The water was carefully evaporated from the mixture and the salt mixture was introduced into a ceramic furnace, pre-heated at 340 8C, and kept at these conditions for 30 min until evaporation of the dark brown nitrous gas ceased. The salt mixture was thereafter cooled to room temperature and washed in water to remove excess salt. The resulting metal oxide was then dried and annealed in an electric furnace at a predetermined time and temperature. The active area was 5 cm2 of all samples. 2.2. Preparation of MEAs A short screening process of different MEA-preparation methods was carried out at the beginning of this study and a spraying technique was chosen. Spraying the catalyst layer either directly onto the membrane [20,21] or onto a backing/current collector substrate [21,22] is a commonly applied technique for preparation of catalytic layers in PEM-fuel cells. The spraying technique was preferred due to its simplicity and easy control of the catalyst loading. The catalyst was sprayed from an ink composed of nano-sized catalyst particles in a mixture of water, ethanol, glycerol and ionomer, ultrasonically mixed beforehand. The spraying process was carried out by using an air driven spray gun where an exact composition of the slurry was deposited to a loading of 2 mg IrO2 cm 2 and a Nafion† ionomer content of 20 wt.%. Before spraying, the membrane was treated according to a procedure described by Srinivasan and co-workers [23]. Membranes were first pre-
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
3947
heated in purified water, treated by H2O2 to remove organic impurities, further treated in hot sulphuric acid solution to ion exchange the membrane completely, then boiled in purified water and finally stored in purified water. 2.3. Experimental methods and procedures The catalyst particles were examined in a JEOL 2010 transmission electron microscope and catalytic layers in a Zeiss DSM 950 scanning electron microscope. Electrode potentials were determined using an in situ Hg/ Hg2SO4 electrode connected to the membrane by a glass capillary filled with wetted polymer membrane material. The reference electrode was placed in ionic contact with a part of the membrane protruding from the gaskets between the cell body and the membrane. All measurements using the reference electrode were carried out only at room temperature as the contact point dried out too fast at higher temperatures. The electrochemical performance of both electrodes in the SPE-cell was determined by stationary polarisation curves at galvanostatic conditions up to 2 A cm 2, with cyclic voltammetry (CV) and with electrochemical impedance spectroscopy (EIS) in the frequency range of 10 mHz /10 kHz. Impedance data were modelled by using ZView software from Scribner Associates.
Fig. 2. TEM and diffraction image of: (a) non-annealed sample; and (b) sample annealed at 490 8C.
3. Results and discussion 3.1. Effect of annealing temperature on the IrO2 catalyst Transmission electron microscopy (TEM) images of a non-annealed sample and a sample annealed at 490 8C are shown in Fig. 2 a and b, respectively. Corresponding diffraction images are shown in the lower right corner of the pictures. The average particle size of the nonannealed sample is observed to be in the range of 5/ 20 nm. These particles, however, seem to consist of agglomerates of smaller particles of 1 /2 nm. The material is probably amorphous, recognized by the diffuse and broad rings and the absence of speckles in the images. The annealed samples exhibit a wider range of particle sizes, from 5 to 100 nm, where the larger particles seem to consist of clusters of particles with diameters of 5/10 nm. Diffraction images of the annealed samples reveal a high degree of crystallinity. Annealing then leads to crystallization and to crystal growth. 3.2. Cyclic voltammetry Fig. 3 shows voltammograms obtained at 20 mV s 1 on samples with different annealing conditions. The non-annealed sample in Fig. 3 exhibits a large area in
Fig. 3. Voltammograms of IrO2 catalysts annealed at different temperatures for 8 h.
the oxygen region and with a very low hydrogen adsorption. The voltammetric charge in the oxygen region of the annealed samples decreases with increasing annealing temperature as also is the case with the capacitance values shown in Table 1. The two peaks at /0.9 and 1.2 VRHE have earlier been attributed to redox couples Ir(III)/Ir(IV) and Ir(IV)/Ir(VI), respectively [24,25]. The voltammetric current and therefore Table 1 Capacitance from voltammetry as a function of annealing conditions Sample (8C)
Capacitance/mF cm 2
Non-annealed 440 490 510 540
283 128 110 88 66
3948
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
also the voltammetric capacitance contains contributions from double layer capacitance and pseudocapacitances due to adsorption and redox processes. All these processes, however, depend on surface area. The capacitance values were calculated from the anodic redox peak in the potential region 1.1 /1.2 VRHE by using the relation C /i/(dE /dt) at scan rates between 10 and 100 mV s 1 where C includes both the pseudocapacitance and the dobbel layer capacitance. The very distinct and narrow peaks at about 0.6 V and especially at 0.9 /1.0 V on the non-annealed sample, are especially difficult to explain. These peaks tend to disappear for higher annealing temperatures. One explanation could be impurities, disappearing with annealing. Another and a rather suggestive explanation are changes in electronic energy levels due to the noncrystallinity and nanosize of the non-annealed particles. The results show that the active surface decreases with increasing annealing temperature. If the capacitance is a direct measure of the surface area it decreases to about 25% of the non-annealed sample when annealing at 540 8C. The voltammogram of the non-annealed sample resembles that of an anodically grown iridium oxide identified by the narrow hydrogen region and a relatively broad oxygen region [24]. Anodically formed hydroxides/oxides are more porous than thermally prepared oxides and a larger part of the interior area is available for the electrochemical processes [26]. This might explain the larger voltammetric charge for the non-annealed sample. The decrease in voltammetric charge after annealing can partly be due to area reduction associated with particle growth and coalescence and partly due to the dehydration of the oxide with stronger oxide bridge bonding [8]. 3.3. Impedance spectroscopy The impedance plots in Fig. 4 reveal that the annealing conditions have a considerable effect on the properties of the catalytic layers. The plots are determined at a potential of 1.49 VRHE which corresponds to current densities of 1 /50 mA cm 2 (see Fig. 12). The high frequency intercept with the real axis is very high for the non-annealed sample and decreases significantly
Fig. 4. AC-impedance of IrO2 catalysts as a function of annealing temperature, measured at 1.49 VRHE during oxygen evolution.
when the catalyst is annealed, up to 490 8C, from where it becomes constant with temperature. The diameter of the low frequency arc, which should be a measure of the polarisation resistance and thereby of the catalytic activity of the electrode, is lowest for the non-annealed sample and increases continuously with increasing annealing temperature. These results are probably connected with the crystal growth of IrO2 with increasing annealing temperature followed by a lower electrocatalytic activity. The corresponding increased crystallinity of the material, however, leads to an increase in the electrical conductivity. The combined effect of annealing conditions and potential is shown in Figs. 5 /7 for the non-annealed sample and the samples annealed at 440 and 490 8C, respectively, where the impedance is measured for each sample between 1.54 and 1.61 VRHE at 10 mV increments. The impedance arc of the non-annealed sample in Fig. 5 shifts towards lower values along the real axis as the potential increases above 1.50 VRHE, which corresponds to a decrease in the electric resistance as the potential increases. For the material annealed at 440 8C a second, high frequency arc appears and takes over with increasing potential at 1.58 VRHE and above. This is even more visible in the Bode plot in Fig. 8. The time constant for the low frequency process (t1) disappears and the time constant for the high frequency process (t2) appears more clearly with increasing potential. The impedance data was modelled by using the circuit RV(R1Q1)(RctQdl), where RV is the ohmic resistance, Rct is the charge transfer resistance. The capacitive elements are represented by Q , a constant phase element often used to model depressed semi circles due to heterogeneities and rough surfaces. The impedance of a capacitive element is given by Q /1/(B (iv ))n where B is a frequency independent constant, v is the angular pffiffiffiffiffiffi frequency, i 1 and n is a factor ranging between 0 and 1. Here, Qdl includes both the double layer capacitance and the pseudocapacitance. This type of model has been used in the literature to interpret the
Fig. 5. Impedance of non-annealed sample as a function of potential in 10 mV steps from 1.54 to 1.61 VRHE during oxygen evolution.
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
Fig. 6. Impedance of sample annealed at 440 8C as a function of potential in 10 mV steps from 1.54 to 1.61 VRHE during oxygen evolution.
Fig. 7. Impedance of sample annealed at 490 8C as a function of potential in 10 mV steps from 1.54 to 1.61 VRHE.
3949
phenomenon determining the first time constant, the circuit above was used to model the impedance data. The n value was found to be close to 1 for Qdl whereas it ranged between 0.6 and 0.8 for Q1. Fig. 9 shows 1/R1 and 1/Rct as a function of potential on a logarithmic scale, found by modelling the impedance data of the sample annealed at 440 8C. It is observed that the obtained values for both parameters can be represented by straight lines. Straight lines of only one decade of current density is, however, somewhat insufficient to determine a real Tafelian behaviour. Reliable data was difficult to obtain at higher potentials due to the very low impedances of these systems and the limitation of the potentiostat. Additional importance, however, is given by the fact that the slopes were found to correspond well with the slopes found by stationary polarisation measurements (see Table 2), indicating that both time constants are related to electrochemical processes. The line for the high frequency process intercepts the line for the low frequency process at about 1.50 VRHE. In Figs. 10 and 11 Mott /Schottky plots are shown for non-annealed material and material annealed at 490 8C, respectively, based on the Qdl values found by fitting the impedance data to the model presented above. The plots reveal that the non-annealed material probably is a semiconductor which changes from n -type conduction to p -type conduction with increasing potential. This could be connected to an oxidation of iridium to a higher valence state, probably hexavalent, at a certain potential. The oxidation then takes place at a higher potential when the material is stabilized by increased annealing temperature. 3.4. Steady state polarisation curves
Fig. 8. Bode plot, face angle u vs. frequency f of sample annealed at 440 8C, as function of potential.
impedance data on different noble metal oxides and their compositions, where the R1Q1 circuit has been attributed to different phenomena, e.g. diffusion of protons along the oxide grains on thermally prepared RuO2 [27], diffusion of reduced oxide sites on anodically grown IrO2 [24] and also to the interlaying oxide layer on the titanium substrate [24,28,29]. From the Bode plots in Fig. 8 it is clear that there are at least two time constants present, however, since there is no interlaying oxide layer of titanium, both time constants are more likely to be related to steps in the oxygen evolution reaction (OER). Regardless of what
Steady state polarisation curves for non-annealed material and for samples annealed at temperatures ranging from 440 to 540 8C, are shown in Fig. 12. The potentials are not corrected for IR drops. At low current densities the non-annealed material exhibits the best
Fig. 9. Reaction admittances 1/R1 and 1/Rct on a logarithmic scale as a function of potential of the sample annealed at 440 8C by fitting of impedance data to the circuit model RV(R1Q1)(RctQdl).
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
3950
Table 2 Tafel coefficients from polarisation curves and Tafel coefficients and ohmic resistance values from modelling of EIS data Annealing conditions (8C)
Non-annealed 440 490 510 540
Polarisation curves
AC impedance
ba, low cd (mV dec 1)
ba, high cd (mV dec 1)
ba, low/high cd (mV dec 1)
RV (V cm2)
36 43 46 47 49
89 145 151 146
36 38/95 42 44 48
0.15 /0.34 0.09 /0.11 0.05 0.05 0.05
Fig. 10. Mott /Schottky plot of non-annealed sample.
Fig. 11. Mott /Schottky plot of sample annealed at 490 8C.
Fig. 12. Steady state anodic polarisation curves as a function of different annealing conditions of IrO2 catalyst during oxygen evolution.
catalytic properties with the lowest total overpotential. The potential then increases with increasing annealing temperature. At high cds , above 200 mA cm 2, the curves cross each other and with a very steep raise of the slope of the curves for the non-annealed sample and the sample annealed at 440 8C. These results confirm the findings both by voltammetry and EIS. At low cds the catalytic properties of the material are governing the
potential, whereas the ohmic resistance of the catalyst layer determines the potential at high cds . This means that an optimum is found at 490 8C where the total anodic potential in the high cd range of 1 /2 A cm 2, is lowest. There seems to exist three factors related to the effect of the particle size on the properties of the catalysts. The most obvious is the effect on the active surface, which increases with decreasing size. A second effect is on the electronic properties of the material, affecting both electronic conductivity and electrocatalytic properties. The contact resistance through the catalytic layer will also depend on the particle size, where larger particles could create less interfacial contact resistance. Such an effect of the particle size can also have influenced the results in this work. However, the electrical conductivity properties of the non-annealed sample and the sample annealed at 440 8C is typical for semiconductors and cannot be related to this effect. Problems with resistance control have not been emphasised in the use of DSA electrodes. The reason for this is rather obvious. In the DSA technology the catalytic oxide mixture is formed on a dense, solid metal substrate. The Ti backing/current collector in the present system is very porous with pore openings ranging from 10 to 50 mm whereas the thickness of the catalytic anode layer is about 10 mm. This results in a considerable lateral resistance in the catalytic layer depending on the electronic conductivity. 3.5. Possible mechanism of oxygen evolution on iridium oxide in SPE systems Tafel constants from IR compensated steady state polarisation curves are shown in Table 2. Values for both low and high cds are shown. In addition values of the slope of E /log (1/R ) diagrams from EIS are shown. Here RV is the series ohmic resistance. The impedance data suggest that the oxygen evolution process consists of two succeeding charge transfer steps. The Tafel gradient of about 40 mV/decade for the low cd range demonstrates that the second step is rate determining at these conditions and that a primary one electron charge transfer is rate determining at high cd, with a Tafel slope
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
varying from about 90 mV/decade for non-annealed material and the sample annealed at 440 8C, to 140 /150 mV/decade for materials annealed at higher temperatures. This is in rather good accordance with the process following the so called ‘Electrochemical Oxide Path’ [30] in three steps, where the first step is a charge transfer step with the formation of an adsorbed hydroxy species on an active site S : SH2 O 0 SOHads H e
(1)
The next step is a second charge transfer step that deprotonates the adsorbent: SOHads 0 SOads H e
(2)
The third step is then the formation of an oxygen molecule and the back formation of two active sites: 2SO 0 2SO2(g)
(3)
That the Tafel slopes increases by increasing annealing temperature might be explained by the increasing stability of the metal /oxygen bond. One possible explanation is that this could increase the activation energy leading to a change in the charge transfer coefficient a and thereby changing the Tafel slope.
3.6. Performance of a PEM electrolysis cell Fig. 13 shows the performance of a total PEM electrolysis cell based on an IrO2 anode and a cathode with 10% Pt on Vulcan XC-72. The total noble metal loading was 2.4 mg cm 2 and the temperature 80 8C and a cell voltage of 1.65 V at 1 A cm 2 (10 kA m 2) was obtained. Interesting is that no sign of transport hindrance is observed up to a current density of 2 A cm 2. This diagram also includes the cathodic overpotential, since it was impossible to use the reference electrode due to drying out of the protruding membrane lip at such high temperatures. The hydrogen overvoltage is less than 150 mV up to cds of nearly 1 A ×/cm 2 [31].
Fig. 13. Cell voltage as a function of current density, measured at 80 8C.
3951
4. Conclusions Iridium oxide is an interesting electrocatalyst for oxygen evolution in a water electrolysis cell using a polymer membrane electrolyte. The catalytic and electronic properties of the catalyst vary considerably with the heat treatment of the material, after producing it in a pyrolysis process in a nitrate melt at 340 8C. The nonannealed material consists of nanosized particles with low crystallinity, as observed by TEM. This material possesses a high electrocatalytic activity but a low electronic conductivity. The TEM study reveals that the particle size increases and the material becomes more crystalline with increasing annealing temperature in the range from 440 to 540 8C, leading to lower electrocatalytic activity and higher electronic conductivity. It is likely that the non-annealed material is a semiconductor that changes from n- to p-type conduction mode in the actual potential range. When increasing the crystallinity of the material by increasing the annealing temperature, the electronic conduction increases and reaches a constant value (see Table 2). The total polarisation of the oxygen electrode then reaches an optimum value, at these conditions, at a temperature of 490 8C. Combined results from EIS and polarisation curves indicate that the reaction mechanism of the oxygen electrode follows the so-called ‘Electrochemical Oxide Path’. Very high performance of a total electrolysis cell can be obtained using a polymer electrolyte with iridium oxide catalyst for the oxygen electrode and platinum for the hydrogen electrode, using totally 2.5 mg of noble metal catalyst per cm2 MEA area.
Acknowledgements The authors acknowledge the financial support from The Norwegian Research Council.
References [1] W. Kreuter, H. Hofmann, Hydrogen Energy Progress XI, 1 (1996) 537. [2] C. Mitsugi, A. Harumi, F. Kenzo, Int. J. Hydrogen Energy 23 (1998) 159. [3] H. Mori, C. Inazumi, M. Kato, S. Maezawa, K. Oguro, H. Takenaka, E. Torikai, Hydrogen Energy Progress XI, 1 (1996) 579. [4] C. Inazumi, H. Mori, M. Kato, S. Meazawa, N. Sawai, K. Oguro, H. Takenaka, Hydrogen Energy Progress XI, 1 (1996) 579. [5] M. Kondoh, N. Yokoyama, C. Inazumi, S. Maezawa, N. Fujiwara, Y. Nishimura, K. Oguro, H. Takenaka, J. New Mat. Electrochem. Syst. 3 (2000) 61. [6] S. Stucki, G.G. Scherer, S. Schlagowski, E. Fischer, J. Al. Electrochem. 28 (1998) 1041. [7] H.B. Beer, British Pat. no. 1,147,442 ,1969.
3952
E. Rasten et al. / Electrochimica Acta 48 (2003) 3945 /3952
[8] L.M. Da Silva, J.F.C. Boodts, L.A. De Faria, Electrochim. Acta 45 (2000) 2719. [9] R. Ko¨tz, S. Stucki, Electrochim. Acta 31 (1986) 1311. [10] O. De Nora, G. Bianchi, A. Nidola, G. Trisoglio, US Pat. no. 3,926,751, 1975. [11] F. Andolfatto, R. Durand, A. Michas, P. Millet, P. Stevens, Int. J. Hydrogen Energy 19 (1994) 421. [12] K. Ledjeff, F. Mahlendorf, V. Peinecke, A. Heinzel, Electrochim. Acta 40 (1995) 315. [13] F. Mahlendorf, PhD thesis, Albert-Ludwigs-Universita¨t Freiburg im Breisgau, 1995. [14] P. Millet, T. Alleau, R. Durand, J. Appl. Electrochem. 23 (1993) 322. [15] T. Ioroi, N. Kitazawa, K. Yasuda, Y. Yamamoto, H. Takenaka, J. Electrochem. Soc. 147 (2000) 2018. [16] R. Adams, R.L. Shriner, J. Am. Chem. Soc. 45 (1923) 2171. [17] J.L. Weininger, R.R. Russel, J. Electrochem. Soc. 9 (1978) 1482. [18] R. Hutchings, K. Mu¨ller, R. Ko¨tz, S. Stucki, J. Matet. Sci. 19 (1984) 3987. [19] L.L. Swette, A.B. LaConti, A. McCatty, J. Power Sources 47 (1994) 343.
[20] S. Møller Holst, PhD thesis, Norwegian University of Science and Technology, 1996. [21] M.S. Wilson, S. Gottesfeld, J. Electrochem. Soc. 139 (1992) L28. [22] E. Passalacqua, F. Lufrano, G. Squadrito, A. Patti, L. Giorgi, Electrochim. Acta 46 (1996) 799. [23] E.A. Ticianelli, C.R. Derouin, A. Redondo, S. Srinivasan, J. Electrochem. Soc. 135 (1988) 2209. [24] L.A. da Silva, V.A. Alves, M.A.P. da Silva, S. Trasatti, J.F.C. Boodts, Electrochim. Acta 42 (1997) 271. [25] F.I. Mattos-Costa, P. De Lima-Neto, S.A.S. Machado, L.A. Avaca, Electrochim. Acta 44 (1998) 1515. [26] S. Ardizzone, G. Fregonara, S. Trasatti, Electrochim. Acta 35 (1990) 263. [27] S.H. Glarum, J.H. Marshall, J. Electrochem. Soc. 127 (1980) 2727. [28] T.A.F. Lassali, J.F.C. Boodts, L.O.S. Buoho˜es, Electrochim. Acta 44 (1999) 4203. [29] V.A. Alves, L.A. Da Silva, J.F.C. Boodts, J. Appl. Electrochem. 28 (1998) 899. [30] J.O’M. Bockris, J. Chem. Phys. 24 (1956) 817. [31] E. Rasten, G. Hagen, R. Tunold, To be published.