Materials Science and Engineering B 178 (2013) 15151522

Contents lists available at ScienceDirect

Materials Science and Engineering B

journal homepage: www.elsevier.com/locate/mseb

Physical and electrochemical properties of cobalt doped (Ti,Ru)O2 electrode coatings

Christine Hummelgrd a , Rasmus K.B. Karlsson b , Joakim Bckstrm a , Seikh M.H. Rahman c , Ann Cornell b , Sten Eriksson c , Hkan Olin a,
a b c

Department of Natural Sciences, Mid Sweden University, SE-851 70 Sundsvall, Sweden Applied Electrochemistry, School of Chemical Science and Engineering, KTH Royal Institute of Technology, SE-100 44 Stockholm, Sweden Department of Chemical and Biological Engineering, Chalmers University of Technology, SE-412 96 Gothenburg, Sweden

a r t i c l e

i n f o

a b s t r a c t
The physical and electrochemical properties of ternary oxides Ti0.7 Ru0.3x Cox O2 (x = 0.093 and x = 0) have been investigated and compared. Samples of three different thicknesses were prepared by spin-coating onto polished titanium to achieve uniform and well-dened coatings. The resulting electrodes were characterized with a variety of methods, including both physical and electrochemical methods. Doping with cobalt led to a larger number of micrometer-sized cracks in the coating, and coating grains half the size compared to the undoped samples (10 instead of 20 nm across). This is in agreement with a voltammetric charge twice as high, as estimated from cyclic voltammetry. There is no evidence of a Co3 O4 spinel phase, suggesting that the cobalt is mainly incorporated in the overall rutile structure of the (Ti,Ru)O2 . The doped electrodes exhibited a higher activity for cathodic hydrogen evolution compared to the undoped electrodes, despite the fact that one third of the active ruthenium was substituted with cobalt. For anodic chlorine evolution, the activity was similar for both electrode types. 2013 The Authors. Published by Elsevier B.V. All rights reserved.

Article history: Received 25 March 2013 Received in revised form 17 August 2013 Accepted 29 August 2013 Available online 11 September 2013 Keywords: DSA Cobaltdoping XRD Polarization curves Hydrogen evolution Chlorine evolution

1. Introduction Transition metal oxides are used for many electrochemical applications, such as gas sensors and chlorine or oxygen evolving electrodes [1]. One of the most commonly used transition metal oxide in catalytic electrode materials is ruthenium dioxide (RuO2 ). Ruthenium is a relatively expensive metal and is therefore often used as mixed with a variety of other metal oxides. The most common mixture is with titanium dioxide (TiO2 ) [2], where a purpose of the titanium dioxide also is to enhance the coating stability [3]. Coatings of the standard DSA composition (30 mol% RuO2 and 70 mol% TiO2 ) combine a high stability with a high activity and a high selectivity toward chlorine evolution [4,5], and have therefore been used as the primary catalyst in chlorine and chlorate production for over 40 years. Ruthenium dioxide has also been doped with for example Ce [6], Sn [7], Co [812] or Ni [13] to improve the catalytic activity, enhance the selectivity and/or increase the stability of the coating. Additions of a third oxide to the rutheniumtitanium

This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Corresponding author. Tel.: +46 60 148855. E-mail address: [email protected] (H. Olin).

mix have been performed to improve the coating properties even further [3,1417]. The interest for using Co3 O4 as an electrode coating component started in the late 1970s. Agapova and Kokhanov [18] studied the preparation and activity of Co3 O4 electrode coatings, and found a high selectivity for chlorine evolution, as well as Tafel slopes and overpotentials for chlorine evolution similar to those exhibited by DSA electrodes. A series of studies further examined the composition, morphology and chemical behavior of Co3 O4 coatings during the rst half of the 1980s [1921]. It was concluded that the high activity of Co3 O4 coatings was primarily governed by their large active surface area [21]. Boggio et al. [22] studied the anodic evolution of Cl2 on Co3 O4 and expressed the belief that the difference in activity between Co3 O4 and RuO2 is small enough to make the replacement economically favorable, even though RuO2 is intrinsically more active than the less expensive cobalt oxide. Boggio et al. [23] then further investigated the electrochemical surface properties of Co3 O4 coated electrodes. Although these early studies found that the activity of Co3 O4 was close to that of RuO2 , the instability of Co3 O4 in acidic solutions [1] was a serious drawback. Further studies were performed in the 1990s, looking at mixed oxides of Co3 O4 and RuO2 coated on Ni electrodes for hydrogen [24,25] and chlorine evolution [26]. At the beginning of the 2000s more studies appeared on surface characteristics, stability and activities for oxygen and chlorine evolution of mixed coatings of RuO2 and

0921-5107/$ see front matter 2013 The Authors. Published by Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.mseb.2013.08.018

1516

C. Hummelgrd et al. / Materials Science and Engineering B 178 (2013) 15151522

Co3 O4 [911,27]. More recently, nanocrystalline coatings of RuO2 and Ru0.8 Co0.2 O2x were investigated using a variety of surface characterization techniques and differential electrochemical mass spectrometry (DEMS) experiments [12,28]. Thus far, no conclusive proof have been found of any synergistic effects regarding activities for chlorine or hydrogen evolution of mixed oxides. Increased activities of mixed coatings have instead been associated with morphological effects [10,11,25]. However, for the more kinetically hindered reaction of oxygen evolution, mixed oxides of Co3 O4 and RuO2 have shown synergistic catalytic effects [9]. In this work, we have studied the kinetic properties of ternary oxide coatings containing the standard 70 mol% TiO2 fraction together with a mixture of ruthenium and cobalt oxides. Since the standard DSA composition is known to yield coatings of high stability, it is interesting to study mixtures containing both stabilizing TiO2 and active oxides of ruthenium and cobalt. The possible advantages of adding a third oxide could be lower cost and increased stability with a retained activity. To the best of our knowledge, there are no reports on this ternary mixed oxide. Studies concerning the effect of cobalt addition on selectivity and stability of DSA electrodes are planned for future work.

2. Material and methods We have replaced one third of the ruthenium in the 30:70 Ru:Ti mixture by cobalt, which after calcination resulted in a mixed oxide hereafter referred to as Ti0.7 Ru0.3x Cox O2 , where x = 0.093 for the doped samples and 0 for the undoped. Electrodes were made by spin coating onto polished titanium in order to obtain well dened and uniform coating layers [30]. Two coating solutions were prepared using titanium(IV)nbutoxide (TBOT), RuCl3 nH2 O (35.51 wt% Ru) and CoCl2 6H2 O as metal precursors. The rst solution contained 14.929 g of TBOT and 5.351 g RuCl3 nH2 O dissolved in 1-propanol to a nal volume of 50 ml, giving a molar ratio between Ti and Ru of 7:3. For the second solution some of the Ru was replaced by an equimolar amount of Co by dissolving 14.929 g TBOT, 3.692 g RuCl3 nH2 O and 1.387 g CoCl2 6H2 O in 1-propanol to 50 ml nal volume. This gave a molar ratio of Ti:Ru:Co = 7:2.07:0.93. Both solutions were stirred for 48 h to assure complete dissolution of the salts. These solutions were used to prepare electrodes on polished titanium substrates, and were also investigated by differential scanning calorimetry. Titanium disks with a diameter of 59 mm, thickness 2.4 mm, were polished to a mirror nish using a Struers Tegra system following a procedure suitable for metallographic preparation of titanium specimens. First, the disks were polished with SiC paper (grit 320) for 2 min, followed by polishing with 9 m diamond suspension for 5 min, and nished by 40 nm silica suspension for 7 min. After polishing, the substrates were carefully cleaned in an ultrasonic bath and thoroughly rinsed in milli-Q water. The disks were prior to coating pickled in dilute hydrouoric acid (20 ml concentrated HF per liter H2 O) at room temperature for 2 min and again thoroughly rinsed in milli-Q water. To assure that the titanium surfaces were free of water, they were dried at 80 C for several minutes. To produce the samples, the prepared substrates were placed in a standard photo-resist spinner (Electronics Micro Systems model 4000). Coating solution was applied to the surface and the substrate was spun with 1400 rpm to distribute an even layer of the solution. The samples were dried at 80 C for 10 min and calcined at 470 C. The spin coating procedure was repeated one, three or seven times resulting in 7, 21 or 48 mg oxide on the disk (500, 1450 or 3400 nm thick oxide layers, estimated from weight increase [30]). For the intermediate layers, the calcination of the samples lasted

for 10 min, while the nal layer was calcined for 1 h. Fig. 1 demonstrates the good reproducibility with the spin-coating technique by showing the weight increase of the samples for each applied coating layer. The surface structures of the electrodes were studied in scanning electron microscopy (SEM) using a LEO 1450 EP and atomic force microscopy (AFM) using a Digital Instruments Dimension 3100. The morphology of the coatings were examined in transmission electron microscopy (TEM) using a JEOL 2000fx. TEM samples were prepared by scraping the coating off the substrate using a sharp knife and collecting the powder on a grid (Carbon grid 200 mesh). The X-ray diffraction (XRD) measurements were carried out on the doped and undoped sample disk surface at ambient temperature using a Siemens D5000 diffractometer in a grazing incidence set-up with a Gbel mirror and an energy dispersive radiation was used. Scans were perdetector. Cu-K ( = 1.5418 A) formed for the 2 range 1575 , with a step size of 0.0092 and 4 s collection time per step. The resulting pattern was imported in the evaluation software package DIFFRACplus EVA with SEARCH, where phases were identied, after background subtraction, with the aid of the ICDD database [31]. Average crystallite size was determined by the Scherrer equation, and measurements in TEM and AFM images. All electrochemical measurements used Milli-Q water from a Millipore Direct-Q3 water purication system, a Princeton Applied Research PAR 273A potentiostat and a thermostatted cell (25 C), with a Pt mesh counter electrode. The cell was washed with Milli-Q water between each measurement, and periodically cleaned with a dilute HNO3 solution, and the electrolyte was continuously deaerated with nitrogen gas starting 15 min before each experiment. The reference electrode was connected to the cell via a Luggin capillary. Rotating disk electrodes with an area of 1 cm2 were used, rotating at 3000 rpm during all polarization curve measurements and with no rotation during the cyclic voltammetry. Reference electrodes of two kinds, Ag/AgCl (saturated KCl) (XR300) and Hg/HgO (1 M KOH) (XR400), from Radiometer analytical were used in chloride electrolyte and in NaOH, respectively. Cyclic voltammetry was performed in 1 M NaOH (Merck, pro analysi). Voltammograms for two electrodes, one containing cobalt and one without cobalt, both with three layers of coating, were measured by sweeping the potential from 1.60 to +1.60 V vs Hg/HgO at a sweep rate of 50 mV/s. Cyclic voltammetry for determination of the surface charge q* was performed by sweeping the electrode potential from 0.67 to +0.28 V vs Hg/HgO at a sweep rate of 20 mV/s. Numerical integration, using the trapezoidal method, of data for the 10th sweep was used to determine q*. Polarization curve measurements were performed for anodic or cathodic current densities from 1 to 1000 A/m2 , from higher to lower current densities. The current densities were logarithmically spaced, with ten current densities per decade. Anodic current densities were used to yield chlorine evolution from 5 M NaCl (Merck, pro analysi) (adjusted to pH 2 using HCl (Merck, pro analysi)), and cathodic current densities to yield hydrogen evolution from 1 M NaOH. The chloride electrolyte pH was checked using a Metrohm 827 pH lab pH-meter. Before the start of each polarization curve measurement, the working electrode was polarized at 1000 A/m2 for between 3 and 5 min for anodic measurements and for 15 min for cathodic measurements. Then, anodic or cathodic currents were applied, and current interrupt was used to correct for the resistive voltage drops in the system. The overpotential decay transient during 450 s after each interrupt was recorded using either a Nicolet Pro 10 oscilloscope or a National Instruments (NI) cDAQ 9178 chassis with a NI9223 voltage digitizer. For each applied current density, the mean of 24 transients (when the Nicolet instrument was used) or 72 transients (when the NI instrument was used) were registered. The

C. Hummelgrd et al. / Materials Science and Engineering B 178 (2013) 15151522

1517

50x10

-3

Samples without Co Co-doped samples

Weight increase / g

40

30

20

10

0 0 1 2 3 4 5 6 7 0 1 2 3 0 1

Coat number
Fig. 1. Weight increase as a function of number of coating layers. The weight increase of the samples for each coating layer during the preparation.

following overpotential decay expression, rst developed by Morley and Wetmore [32], E = E (t = 0) b ln 1 + t (1)

Differential scanning calorimetry (DSC) measurements were performed using a DSC 822e from Metler Toledo. 10 l of solution was placed in a 40 l gold crucible, the lid was pierced a few times before sealing. The measurement were performed from 25 C to

was then tted to the overpotential decay transient using (E, t) data from each current density. In this expression, b is the Tafel slope of the electrode reaction, t is the time after the current interrupt and is a time constant for the overpotential decay. This tting yields E(t = 0), which is the electrode potential corrected for the ohmic drop in the electrolyte. However, for very low ohmic drops, where the difference between the uncorrected and the corrected potential was less than 5 mV, the uncorrected electrode potential was used. The logarithmic tting was performed using the open source EzyFit toolbox for MATLAB [33], which interfaces the NelderMead simplex function FMINSEARCH in MATLAB. In this way, the corrected electrode potential for each applied current density was found. These potentials were then used to construct polarization curves. The standard deviation in voltage between two duplicate samples measured in this way was generally less than 10 mV, for both cathodic and anodic measurements. The only exception was for the thinnest coatings, where the standard deviation reached 1050 mV at current densities lower than 10 A/m2 .

Fig. 2. Differential scanning calorimetry. DSC curves of coating solution (RuCl3 + TBOT in 1-propanol) with and without CoCl2 doping, as a function of reference temperature. Measurement performed using a gold crucible, exotherm reaction upwards.

Fig. 3. Cracked mud structure. SEM image of cracked mud structure of coating (A) without cobalt and (B) doped with cobalt. There were more cracks per area unit in the cobalt-containing sample. These are one-layer samples (coating thickness approximately 500 nm) and the scratches from the grinding/polishing procedure (visible as faint straight lines) on the substrate was not fully masked by the coating.

1518

C. Hummelgrd et al. / Materials Science and Engineering B 178 (2013) 15151522

Fig. 4. Crystal structure and crystallite size. (A) XRD patterns of Ti0.7 Ru0.3x Cox O2 (x = 0 and 0.093) spin coated onto titanium disks. The curves are vertically offset for clarity. Coating thickness was approximately 3400 nm (7 layers). The magenta vertical lines marks the angles where the Co3 O4 spinel peaks should have appeared. (B) and (C) TEM images of coating oxide with and without cobalt doping (grain size approximately 10 and 20 nm, in agreement with estimates from XRD data), respectively. The black elliptic dot in the center is an imaging artifact.

600 C at a heating rate of 5 C/min. As reaction gas 50 ml/min of air was used (inlet temperature: 21 C; relative humidity: 23.8%). 3. Results and discussion 3.1. Differential scanning calorimetry Fig. 2 shows the DSC curves from the two coating solutions used in the study. The results have been veried using both gold and aluminum crucibles, so it is unlikely that any part of the curve can be explained as reactions between the solution and the crucible material. The endothermic feature below 100 C is present in both curves, and could be straightforwardly attributed to the evaporation of solvent. The main features of the curves was found in the temperature interval 270380 C. These were attributed to oxidation of the precursors and formation of the oxide coating material. When comparing the two curves, it can be seen that the oxidation of the solution doped with cobalt peaked at a slightly higher temperature than the oxidation of the undoped coating solution (Tc increased from 338 C to 341 C). Also to be noted is that the oxidation took

place over a larger temperature interval and showed lower heat release in the presence of cobalt. No reactions were detected in the temperature range of 380600 C. For comparison, we have also performed DSC experiments on a solution of only cobalt chloride in 1-propanol (1.387 g CoCl2 6H2 O dissolved in 1-propanol to 50 ml nal volume) using the same procedure (data not shown). This curve displayed the same initial solvent evaporation, but was then completely at for all temperatures up to 600 C. In other words, CoCl2 in propanol did not oxidize in the range of less than 600 C, which is in contrast to Louardi et al. [34] who have obtained Co3 O4 from CoCl2 in water by spray pyrolysis at 350 C, and Garavaglia et al. [19] who have obtained Co3 O4 by thermal decomposition of CoCl2 6H2 O at 400 C when the duration time was long enough (>10 h). The lower heat release of the oxidation of the cobalt containing solution was not surprising. The standard enthalpies of formation for the oxides and chlorides of the involved metals can be found in standard reference tables [35] to be 205 kJ/mol for RuCl3 , 205 kJ/mol for RuO2 , 312 kJ/mol for CoCl2 , 238 kJ/mol for CoO and 891 kJ/mol for Co3 O4 . In other words, the oxidation reaction 3CoCl2 + 2O2 Co3 O4 + 3Cl2 is endothermic

C. Hummelgrd et al. / Materials Science and Engineering B 178 (2013) 15151522

1519

( H = +46 kJ/mol), as is the reaction CoCl2 + 0.5O2 CoO + Cl2 ( H = +74 kJ/mol). On the contrary, the corresponding reaction for ruthenium, RuCl3 + O2 RuO2 + 1.5Cl2 is exothermic ( H = 100 kJ/mol). Therefore, it is expected that the formation of a mixture of Ti, Ru and Co oxides should release less heat than the formation of a mixture of only Ti and Ru oxides. 3.2. Surface morphology Doping with cobalt affected the cracked-mud structure of the (Ti,Ru)O2 mixed oxide so that the crack-density increased (Fig. 3). The effect was most distinct for the thinnest (0.5 m) coating and were hardly present for the thickest (3 m) coating. Also, in the intersections of the cracks of the doped coating, there were some very small plaquettes that were not seen on the undoped samples. Da Silva et al. [11] characterized the surface properties of the (RuO2 )x + (Co3 O4 )(1x) system and found that if the amount of Co3 O4 is less than 50 mol%, the electrode is indistinguishable from the pure RuO2 electrode using SEM. However, they did nd an increase in the crack-density for higher cobalt concentrations. 3.3. Crystallinity The (Ru,Ti)O2 mixed oxide doped with cobalt showed smaller grains compared to the undoped oxide. Average crystallite size, estimated from XRD (the pattern is displayed in Fig. 4) using the Scherrer equation for the main peaks at 2 = 27.67 and 27.53 respectively, was 10 nm for the sample with cobalt and 20 nm for the sample without. The size estimation was conrmed by measurements in TEM (Fig. 4) and AFM images (Fig. A.1 in supplementary information). This gave a real area for the doped sample that was about twice the area of the undoped (Eq. (2)). Atot = 1 1 1 A 3 r2 = V r r (2)
q* / C cm-2

0.02

With Co Without Co

0.015

0.01

0.005

0 0 1 2 3 4 5 6 7 8

N
Fig. 5. Voltammetric charge. The voltammetric charge, q*, versus the number of applied coating layers.

mixtures (>70 mol%) by the highly amorphous nature of their lm, and interpreted the amorphous state as an indication of insolubility of Co3 O4 in RuO2 . Since our oxide lm was not highly amorphous that explanation of the non-existing spinel peaks is unlikely, and we have shown that it is possible to solve cobalt in rutheniumtitanium et al. [28] dioxide in another crystalline form than Co3 O4 . Jirkovsky found spinel in samples annealed at 400 C only when the Co:Ru ratio was larger than 2:1. Their less cobalt-containing samples were all of a single phase (rutile). 3.4. Cyclic voltammetry The calculated voltammetric charge, q*, versus the number of layers applied is shown in Fig. 5. Data from the 10th sweep, where the voltammograms indicated that the electrodes had reached steady state, was used in all cases. The voltammetric charge for the doped coatings is about twice as high as that of the undoped samples, which is in agreement with results on coatings of RuO2 and of RuO2 + Co3 O4 having the same molar ratio as in the present study (Ru:Co of 2:1) [10] and indicates a higher active surface area for the doped electrodes.
0.002

The area enlargement was calculated under the assumption of monocrystalline symmetrical grains of homogeneous size [36] in each sample. The smaller grains of the doped sample was in contra et al. [28] who found that RuO2 doped with diction with Jirkovsky cobalt in fact have larger grains than the undoped RuO2 . Their study is however of a binary oxide without titanium and with nitrate precursors which could explain the difference. Garavaglia et al. [19] studied the physical properties of Co3 O4 prepared from CoCl2 and Co(NO3 )2 and concluded that CoCl2 is not a good starting point if one is to obtain small crystals, in analogy with the RuO2 case [36]. The effect of the precursor on the ternary mixed oxide have not been investigated further. In the cobalt doped TiO2 binary oxide case, Li et al. [37] found that the doping does not affect the particle size (calculated from BrunauerEmmettTeller (BET) surface area). The undoped (Ti,Ru)O2 oxide displayed mainly rutile structure in agreement with Comninellis and Vercesi [38]. Co3 O4 (spinel structure) is the thermodynamically preferred cobalt oxide in lms prepared from solutions [39], but no spinel peak was observed in the XRD data. The only difference between the two samples is the existence of TiO2 in anatase (PDF 01-086-1157 [31]) in the cobalt doped sample. This led to the conclusion that the cobalt must have entered the overall rutile structure of the (Ti,Ru)O2 as suggested by Li et al. [37]. It might also exist as an amorphous compound. This did not agree with most reports on mixed oxides of ruthenium and cobalt [811] as they refer to the spinel structure together with the rutile, but agreed with Makarova et al. [12] who only found the spinel structure in their sample when the annealing temperature had been very high (900 C). Da Silva et al. [11] explains the lack of spinel peaks in the diffractogram for high ruthenium concentration

0.001

j / A cm-2

0.001 Without Co With Co

0.002

0 0.5 E / V vs Hg/HgO (1M KOH)

0.5

Fig. 6. Broad cyclic voltammograms. Broad cyclic voltammogram in 1 M NaOH. There are signicant differences between coatings with and without cobalt. Specifically, the doped coating exhibited an anodic peak at 0.3 V vs Hg/HgO. Hydrogen evolution occurs at potentials below 1 V vs Hg/HgO and oxygen evolution at potentials above 0.6 V vs Hg/HgO. The potential was swept between 1.6 V and 1.6 V vs Hg/HgO at 50 mV/s.

1520

C. Hummelgrd et al. / Materials Science and Engineering B 178 (2013) 15151522

0.9

1 E / V vs Hg/HgO (1M KOH)

1.1 7 layers, no Co 7 layers, Co 3 layers, no Co 3 layers, Co 1 layer, no Co 1 layer, Co

1.2

Number of layers

1.3

1.4 1 10 -j / Am-2 100 1000

1.2 1.18 1.16 1.14 1.12 1.1 1.08 1.06 1.04

E / V vs Ag/AgCl [sat. KCl]

7 7 3 3 1 1

layers, no Co layers, Co layers, no Co layers, Co layer,no Co layer,Co

Number of layers

10

100

1000

j / Am-2
Fig. 7. Polarization curves. (A) Cathodic polarization curves for hydrogen evolution from 1 M NaOH. (B) Anodic polarization curves for chlorine evolution from 5 M NaCl solution.

The magnitude of the voltammetric charge is hard to compare with previous work, since different parts of the voltammograms and Trasatti have been used to calculate q*. Trasatti [26] and Krstajic [25] found voltammetric charges for binary rutheniumcobalt oxide coatings of the same order of magnitude as those of the thickest electrode coatings examined here. However, the binary coatings examined by Da Silva et al. [11] exhibited voltammetric charges about ten times as large as those of the present coatings. All previous studies of binary rutheniumcobalt oxides have used electrodes prepared using brush- or dip coating, but not spin coating as in the present work. Cyclic voltammograms recorded between the onset potentials for hydrogen evolution and oxygen evolution are shown in Fig. 6. The voltammogram for the undoped sample is similar to that reported for RuO2 [40], whereas the CV for the cobalt doped coating show an additional anodic peak at around 0.3 V vs Hg/HgO. This peak is partially included in the voltage range used for the

calculation of q* (see Fig. 5), inuencing the calculated value of the voltammetric charge. Previous studies [23,25] performed in alkaline solutions have, based on Pourbaix diagrams for cobalt [42], identied a peak at a potential corresponding to 0.5 V vs Hg/HgO. Boggio et al. [23] related this peak to the oxidation of surface cobalt compounds to Co(IV) according to CoOOH CoO2 + H+ + e (3)

It has been found, for binary rutheniumcobalt oxide coatings, that increasing the content of ruthenium decreases the peak related to reaction (3) without changing the potential at which it occurs [25]. In the present case, the voltammograms show a peak unique to the ternary cobalt doped coating at 0.3 V vs Hg/HgO, likely related to Co oxidation.

C. Hummelgrd et al. / Materials Science and Engineering B 178 (2013) 15151522

1521

3.5. Polarization curves Polarization curves for hydrogen and chlorine evolution are seen in Fig. 7 where each marker represents the mean of two measurements on duplicate electrodes. The polarization curves for hydrogen evolution (Fig. 7A) exhibit two distinct regions, as has also been seen in previous studies of RuO2 [40]. The Tafel slopes in region I (at lower applied current densities) are 55 to 60 mV/dec, irrespective of doping, which is in the range of 4060 mV/dec typical for hydrogen evolution on ruthenium dioxide in alkaline electrolyte [24,40,41]. Note that the corresponding slope on coatings of 100% Co3 O4 was found to be higher, 105 mV/dec [24]. In region II (at current densities from about 100 A/m2 ), the cobalt doped electrodes show lower Tafel slopes than the undoped electrodes, in particular for thinner coatings. The slope for undoped electrodes in region II changes from 0.22 V/dec for electrodes with 7 layers of coating, to 0.26 V/dec for 3 layer coatings and 0.27 V/dec for the thinnest coating. For doped coatings, the Tafel slope for both 7 and 3 layer electrodes was almost the same, at around 0.19 V/dec, increasing to 0.21 V/dec for 1 layer coatings. This difference in Tafel slope between doped and undoped electrodes yields a signicantly lower overvoltage for the doped coatings at current densities approaching industrial levels. For chlorine evolution, the Tafel slopes and overpotentials are similar for all electrodes. There is a general tendency of lower overpotentials for thicker electrode coatings, in agreement with previous work [2,39,40]. The Tafel slopes between 10 and 1000 A/m2 are, regardless of doping and coating thicknesses, 35 to 45 mV/dec, which is typical for both Co3 O4 [22] and RuO2 [1] in acidic solution. To examine if a synergistic effect of the doping is present, polarization curves normalized with respect to surface area as measured by the voltammetric charge have been prepared for both hydrogen evolution and chlorine evolution (curves are presented in the supplementary information). These curves indicate that the active surface area of the electrodes have a key inuence on the overpotentials, and that for electrodes with the same coating thickness, doped electrodes exhibit lower normalized current densities at a certain electrode potential. Thus the activity, when corrected for the active surface area, is lower for doped electrodes. Synergistic effects between Co, Ru and Ti in ternary mixed coatings do not seem to be present and the lower overpotentials for hydrogen evolution in region II exhibited by doped electrodes could thus be explained by the increased electrochemical surface area of these samples. The same factor can explain why doped coatings showed essentially the same Tafel slope and overpotential in region I of the hydrogen evolution polarization curve, and in the chlorine evolution polarization curves, as undoped samples. These results give further support to the interesting possibility of reducing the amount of Ru in DSA coatings by doping with cobalt. Except for adhesion tests using normal ofce Scotch tape, we did not perform any regular stability measurements. The tape tests did however show excellent adhesion for all samples. Moreover, in SEM and AFM images the electrodes appeared to have similar morphology regardless of the doping. The increased porosity is not visible in any other way than the smaller grains of the doped samples (that is, no larger pores were observed). The packing of the grains was similar in all electrodes and in none of the samples the grains were detached or displaced by the AFM tip.

thirds as much active ruthenium. The activity for chlorine evolution was similar for the two coating types. Doping with cobalt approximately doubled the surface area of the electrodes, as estimated from measurements of grain sizes and voltammetric charges, and this increased area explained the high activity of the cobalt containing coatings. The crack density on the micrometer scale also increased by the cobalt doping, at least for thin coatings, but the excellent coating adhesion was not affected. The cobalt does not form Co3 O4 phase, but is rather incorporated in the overall rutile structure of the (Ru,Ti)O2 . Authors contribution Authors CH and RKBK have contributed equally. Conceived and designed the study: CH, JB. Sample preparation: RKBK, JB. Designed and performed the experiments: CH, RKBK, SMHR. Analyzed the data: CH, RKBK, JB, SMHR, AC, SE, HO. Wrote the paper: CH, RKBK, AC, JB, HO. Role of the funding sources This work is part-nanced by the European Regional Development Fund through the project Energywise, the Swedish Research Council, the Swedish Energy Agency, Permascand AB and Lnstyrelsen Vsternorrland. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Acknowledgements Susanne Holmin is acknowledged for fruitful discussions. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.mseb.2013. 08.018. References
[1] S. Trasatti (Ed.), Electrodes of Conductive Metallic Oxides: Part A, Elsevier, Amsterdam, 1980. [2] S. Trasatti, Electrochim. Acta 45 (2000) 23772385. [3] L.A. DeFaria, J.F.C. Boodts, S. Trasatti, Electrochim. Acta 37 (1992) 25112518. [4] N. Bun, V.V. Losev, M.F. Reznik, E. Zaripova, Soviet Electrochem. 22 (1986) 365368. [5] H.B. Beer, British Patent 1,195,871 (1967). [6] S. Hosokawa, H. Kanai, K. Utani, Y. Taniguchi, Y. Saito, S. Imamura, Appl. Catal. B - Environ. 45 (2003) 181187. [7] M. Ito, Y. Murakami, H. Kaji, K. Yahikozawa, Y. Takasu, J. Electrochem. Soc. 143 (1996) 3236. [8] Y. Liu, W. Zhao, X. Zhang, Electrochim. Acta 53 (2008) 32963304. [9] L.M. Da Silva, J.F.C. Boodts, L.A. De Faria, Electrochim. Acta 46 (2001) 13691375. [10] L.M. Da Silva, J.F.C. Boodts, L.A. De Faria, J. Braz. Chem. Soc. 14 (2003) 388395. [11] L.M. Da Silva, J.F.C. Boodts, L.A. De Faria, Electrochim. Acta 45 (2000) 27192727. M. Klementov, I. Jirka, K. Macounov, P. Krtil, [12] M.V. Makarova, J. Jirkovsky, Electrochim. Acta 53 (2008) 26562664. M.V. Makarova, J. Franc, P. Krtil, J. Solid State Elec[13] K. Macounov, J. Jirkovsky, trochem. 13 (2009) 959965. [14] Y. Takasu, S. Onoue, K. Kameyama, Y. Murakami, K. Yahikozawa, Electrochim. Acta 39 (1994) 19931997. [15] D.T. Shieh, B.J. Hwang, Electrochim. Acta 38 (1993) 22392246. [16] T.A.F. Lassali, J.F.C. Boodts, S. Trasatti, Electrochim. Acta 39 (1994) 15451549. [17] J.L. Fernandez, M.R.G. De Chialvo, A.C. Chialvo, J. Appl. Electrochem. 32 (2002) 513520. [18] R.A. Agapova, G.N. Kokhanov, Elektrokhimiya 12 (1976) 16491653. [19] R. Garavaglia, C.M. Mari, S. Trasatti, Surf. Technol. 19 (1983) 197215. [20] R. Garavaglia, C.M. Mari, S. Trasatti, Surf. Technol. 23 (1984) 4147. [21] S. Trasatti, Electrochim. Acta 29 (1984) 15031512. [22] R. Boggio, A. Carugati, G. Lodi, S. Trasatti, J. Appl. Electrochem. 15 (1985) 335349. [23] R. Boggio, A. Carugati, S. Trasatti, J. Appl. Electrochem. 17 (1987) 828840. S. Trasatti, J. Appl. Electrochem. 28 (1998) 12911297. [24] N. Krstajic,

4. Conclusions Cobalt doped DSA electrodes exhibited higher activity for hydrogen evolution than undoped electrodes at current densities approaching industrial levels, even though containing only two

1522

C. Hummelgrd et al. / Materials Science and Engineering B 178 (2013) 15151522 [34] A. Louardi, A. Rmili, F. Ouachtari, A. Bouaoud, B. Elidrissi, H. Erguig, J. Alloys Compd. 509 (2011) 91839189. [35] D.R. Lide (Ed.), CRC Handbook of Chemistry and Physics, 89th edition, CRC Press, Boca Raton, 2008. [36] C. Malmgren, A.K. Eriksson, A. Cornell, J. Bckstrm, S. Eriksson, H. Olin, Thin Solid Films 518 (2010) 36153618. [37] J.G. Li, R. Bchel, M. Isobe, T. Mori, T. Ishigaki, J. Phys. Chem. C 113 (2009) 80098015. [38] C. Comninellis, G.P. Vercesi, J. Appl. Electrochem. 21 (1991) 335345. [39] S. Ardizzone, G. Fregonara, S. Trasatti, Electrochim. Acta 35 (1990) 263 267. [40] A. Cornell, D. Simonsson, J. Electrochem. Soc. 140 (1993) 31233129. [41] D. Miousse, A. Lasia, J. New Mater. Electrochem. Syst. 2 (1999) 7178. [42] E. Deltombe, M. Pourbaix, Cobalt, Pergamon Press Ltd., 1966, pp. 322329.

S. Trasatti, J. Electrochem. Soc. 142 (1995) 26752681. [25] N. Krstajic, [26] S. Trasatti, in: R.W. Curry (Ed.), Modern Chlor-Alkali Technology, vol. 6, Royal Society of Chemistry, Cambridge, 1995. [27] L.M. Da Silva, L.A. De Faria, J.F.C. Boodts, J. Electroanal. Chem. 532 (2002) 141150. M. Makarova, P. Krtil, Electrochem. Commun. 8 (2006) [28] J. Jirkovsky, 14171422. [30] C. Hummelgrd, J. Gustavsson, A. Cornell, H. Olin, J. Bckstrm, Thin Solid Films 536 (2013) 7480. [31] S. Kabekkodu (Ed.), ICDD. PDF-4+2010 (Database), International Centre for Diffraction Data, Newtown Square, 2010. [32] H. Morley, F. Wetmore, Can. J. Chem. 34 (1956) 359363. [33] F. Moisy, EzyFit, a free curve tting toolbox for MATLAB, Version 2.4. http://www.fast.u-psud.fr/ezyt/