ARTICLE pubs.acs.org/Langmuir
Using Ordered Carbon Nanomaterials for Shedding Light on the Mechanism of the Cathodic Oxygen Reduction Reaction Pavel S. Ruvinskiy,† Antoine Bonnefont,‡ Cuong Pham-Huu,† and Elena R. Savinova*,† †
Laboratoire des Materiaux, Surfaces et Procedes pour la Catalyse, UMR 7515 du CNRS-UdS, Ecole de Chimie, Polymeres et Materiaux, Universite de Strasbourg, 25 rue Becquerel, 67087 Strasbourg, France ‡ Institut de Chimie de Strasbourg, UMR 7177 du CNRS-UdS, 4 rue Blaise Pascal, Universite de Strasbourg, 67070 Strasbourg, France
bS Supporting Information ABSTRACT: Insufficient understanding of the mechanism of the cathodic oxygen reduction reaction puts constraints on the improvement of the efficiency of polymer electrolyte fuel cells (PEMFCs). We apply ordered catalytic layers based on vertically aligned carbon nanofilaments and combine experimental rotating ringdisk studies with mathematical modeling for shedding light on the mechanism of the oxygen reduction reaction on Pt nanoparticles. Based on the experimental and simulation evidence we propose a dual path ORR mechanism which comprises a “direct 4e” and a “series 2e þ 2e” pathway and explains switching between the two. For the first time we show that below 0.8 V the “direct” path may be discarded and the ORR predominantly occurs via H2O2 mediated pathway, while in the potential interval between ca. 0.8 V and the onset of the ORR the “direct” path is dominating.
1. INTRODUCTION The oxygen reduction reaction (ORR, eq 1) is a key step in the aerobic energy production in living organisms, the O2 molecule serving as the terminal acceptor of electrons resulting from glucose oxidation.1,2 Biological O2 reduction is catalyzed by Cytochrome C oxidase which comprises a multinuclear redox center capable of multielectron transfer and couples exergonic O2 reduction to endergonic proton transmembrane transport to drive ATP synthesis.3 The inorganic ORR is the cathode reaction occurring in the polymer electrolyte membrane fuel cells (PEMFC) which are considered the power sources of tomorrow. Pt or its alloys are currently employed at the cathode of a PEMFC as the oxygen reduction catalysts.4 The electrochemical ORR on Pt electrodes proceeds through a complex multistep mechanism,510 and contrary to the biological O2 reduction1113 is a slow reaction which largely limits the efficiency of a PEMFC and boosts its cost. The kinetics of the ORR is significantly influenced by the Pt loading at the working electrode and currently requires up to 0.5 gPt cm2, which needs to be reduced at least 5 times, while maintaining MEA power density, in order to meet requirements for vehicle applications.14 One of the major challenges for the large scale commercialization of PEMFCs is the development of highly efficient catalytic layers, containing ultralow amount of Pt (and/or other noble metals) and having optimized architecture for efficient mass transport of the reactants and the reaction products. We have recently proposed novel approach to the design of structurally ordered catalytic layers based on Pt nanoparticles r 2011 American Chemical Society
attached to vertically aligned carbon nanotubes/nanofilaments (Pt/VACNF) for improving the efficiency of the PEMFC catalytic layers.15 This approach is inspired by the work of Debe et al.16 on nanostructured thin film (NSTF) electrodes. However, the advantages of our approach are (i) the higher specific surface of Pt which is supported on conductive carbon support, (ii) the thickness of the catalytic layer can be easily varied in a wide range from 1 to tens of μm (which is of special interest for model studies presented in this manuscript), and (iii) the preparation procedure (for details see experimental and ref.15) involves inexpensive techniques and chemicals, with the ease of a scale-up possibility. Given that carbon nanofilaments possess unique mechanical strength,17 high electrical conductivity,18 and corrosion stability,19 a layer composed of parallel, evenly spaced nanofilaments has potential of improving mass transport, electrical conductivity, Pt utilization and robustness of the membrane-electrode assembly.15 On the other hand, the concept of three-dimensionally (3D) ordered catalytic layers based on Pt/VACNF can be employed as a novel tool to unveil the kinetics of complex electrochemical reactions.15,20 In this paper we apply Pt/VACNF for shedding light on the mechanism of the electrocatalytic ORR on Pt nanoparticles. Received: February 17, 2011 Revised: April 18, 2011 Published: June 14, 2011 9018
dx.doi.org/10.1021/la2006343 | Langmuir 2011, 27, 9018–9027
Langmuir
ARTICLE
Figure 1. (a) SEM image of a VACNF layer grown on TiOx/Ti/Si substrate with corresponding TEM-based diameter distribution diagram (b). (d) TEM image of a carbon nanofilament modified with Pt nanoparticles prepared using polyol method (ΓPt =0.04) with corresponding Pt nanoparticle size distribution diagram (e). (c) Schematic representing system geometry with Rf being the radius of the filament, Rc radius of the diffusion domain, H height of the cylindrical diffusion domain, which is a sum of L (nanofilament length) and δ (diffusion layer thickness). Note that for modeling a homogeneous distribution of active sites over the filaments was considered.
The electrochemical ORR has been subject of many original papers and review chapters,510,21 yet its mechanism is still poorly understood. Insufficient understanding of the mechanism of the cathodic ORR puts constraints on the improvement of the PEMFC efficiency. It is generally accepted that depending on the electrode material and on the experimental conditions the ORR proceeds either directly to water or via formation of the hydrogen peroxide as a stable intermediate (eq 2). The latter can be either further reduced to water in a 2e process (eq 3) or decomposed to H2O and O2 (eq 4). O2 þ 4Hþ þ 4e ¼ 2H2 O E° ¼ 1:23 V
ð1Þ
O2 þ 2Hþ þ 2e ¼ H2 O2 E° ¼ 0:70 V
ð2Þ
H2 O2 þ 2Hþ þ 2e ¼ 2H2 O E° ¼ 1:78 V
ð3Þ
1 H2 O2 ¼ H2 O þ O2 ð4Þ 2 The formation of H2O2 is a major concern for fuel cell application since it attacks the polymer electrolyte membrane and the catalytic layers, strongly affecting the durability of the PEMFC.22 Despite numerous investigations there is no general agreement not only on the detailed mechanism of the ORR but also on the predominant pathway. Polycrystalline and single crystal Pt electrodes are known to produce negligible amounts of H2O2 in acid electrolytes outside the Hupd (hydrogen underpotential deposition) region, which is often considered as a proof of the quasi-direct, 4e reduction mechanism. Some authors believe that the ORR on Pt proceeds through the step of dissociative O2 adsorption followed by the reduction of the adsorbed oxygen atoms,23,24 while others suggest that the reaction predominantly occurs via series 2e þ 2e path.9 There is also no agreement on whether the H2O2 reduction occurs in an electrochemical step (3) or is preceded by a chemical disproportionation step (4). The rate determining step (rds) of the ORR on Pt has also been widely debated. Numerous studies support the first electron (or proton-coupled electron) transfer to the O2 molecule as the rds.6,2527 However, it was also suggested that the rds can be the rupture of the OO bond via various dual-site mechanisms,28,29 initial adsorption of O2 on the surface of the electrocatalyst,30 or desorption of O or OH from the surface.31 Wang et al.24 suggested that the ORR on Pt is desorption-limited in the potential interval of OH adsorption, and electron-transfer limited
in the potential interval where Pt is free from OHads. Overall this controversy probably manifests the existence of a complex multistep mechanism comprising a combination of various reaction pathways with the rds depending on the electrode potential, and significant influence of the pretreatment of the electrode on the reaction pathways. Recently significant progress has been achieved in the detection of the H2O2 intermediate using either ultramicroelectrodes composed of isolated Pt nanoparticles supported on a single carbon filament,32 or two-dimensional arrays of ca. 100 nm Pt particles deposited on glassy carbon substrate.3335 It has been shown that under the enhanced mass-transport conditions the effective number of electrons drops noticeably below 4 (e.g., Chen and Kucernak32 observed neff ∼ 3.4 corresponding to 25% selectivity in the H2O2 production), while the H2O2 yield strongly increases (e.g., Schneider et al.34 observed H2O2 yield of ca. 20% for low density Pt arrays on glassy carbon). These studies strongly affected the accepted paradigm of the ORR on Pt and created a bias toward “series” 2e þ 2e pathway, even if a contribution of the direct path was not excluded. Increased H2O2 production has been also observed by Inaba et al.36 for low Pt/C loadings on the RRDE. In this work we discuss the influence of the 3D spatial distribution of Pt nanoparticles on the ORR kinetics with a special emphasis on the amount of H2O2 escaping the catalytic layer. For this purpose, we are using 3D array of Pt nanoparticles supported on vertically aligned carbon nanofilaments (VACNF, see Figure 1), mounted on a disk insert of a rotating ring-disk electrode (RRDE).15 This strategy was already successfully applied to the studies of CO and H2 electrooxidation in acid solutions.15,20 We show that the analysis of the RRDE data on Pt/ VACNFs combined with a finite element modeling taking into account the influence of the spatial architecture of the layer on the current potential curves delivers substantial insights into the ORR mechanism. The progress achieved in this work in the understanding of the ORR may be further applied to the design of better catalysts for the cathodes of PEMFCs.
2. METHODS 2.1. Sample Preparation. The preparation procedure of Pt/ VACNF arrays on Ti discs (3D samples, throughout the paper) can be found elsewhere.15 Briefly, the precision machined Ti inserts (5 mm diameter) were used as the substrates for the growth of the aligned carbon nanofilaments, the latter were adjusted in terms of the height of the layer and modified with Pt nanoparticles. Pt deposition was performed 9019
dx.doi.org/10.1021/la2006343 |Langmuir 2011, 27, 9018–9027
Langmuir
ARTICLE
Table 1. Characteristics of the Samples sample number layer thickness L/μm Pt EASA/cm2 a.1
ΓPtc
N/% d
1
∼0.01
∼0.008 26.8 ( 1.0 ∼0.005 24.8 ( 2.1
a.2
2
∼0.01
a.3
5
0.17
a.4
14
0.09
a.5
20
0.07
a.6a
21
7.0
0.31
26.6 ( 2.8
VACNFb
21
0
0
24.7 ( 3.4
a
0.04
26.6 ( 0.8
0.01
22.3 ( 2.4
∼0.003 20.8 ( 2.8
a
The samples prepared using ethylene glycol synthesis. The rest of the samples were prepared using the electrochemical deposition method. b This is sample a.6 prior to the deposition of Pt nanoparticles. c ΓPt is the ratio of the electrochemically active surface area of Pt to the surface area of VACNF. d Collection efficiency obtained for each sample using 10 mM K3FeCN6 in 0.1 M NaOH. For more details, see the Methods section. using two methods: polyol37 and electrochemical deposition,38 the particular details of these approaches are described elsewhere.15 Table 1. shows the characteristics of 3D Pt/VACNF samples. The Pt coverage ΓPt is defined as a ratio between the electrochemically active Pt surface area (EASA) and the surface area of the VACNF. Thus obtained samples were prepared for the electrochemical measurements and mounted into the standard RRDE chuck for an MSR rotator (PINE, USA) with a Pt ring (outer diameter 7.5 mm, inner diameter 6.5 mm) insulated from the disk by a PTFE insert (gap 0.75 mm). Pt polycrystalline electrode (Ptpc) was a standard disk-insert from PINE (USA) and was used as received, the standard electrochemical cleaning procedure was applied. 2.2. Electrochemical Measurements. Electrochemical studies were performed at 30 ( 0.1 °C in a three-electrode cell in 0.1 M H2SO4 supporting electrolyte, using Autolab PGSTAT30 (Eco Chemie, The Netherlands) potentiostat equipped with an analogue linear sweep generator at the sweep rate of 5 mV s1. As the reference, mercury sulfate (MSE) electrode was used. Unless otherwise stated, all potentials hereinafter are referred to the reversible hydrogen electrode (RHE). All aqueous solutions were prepared using ultrapure water (18 MΩ cm, < 3 ppb TOC) and supra-pure H2SO4 (Sigma-Aldrich). The EASA of Pt in the samples was measured by CO-stripping experiment. For details the reader is referred to ref.15 In O2-reduction experiments O2 was constantly bubbled through the solution in order to maintain the saturation level and the ring potential was set at 1.2 V RHE in accordance with previous studies.39,40 Collection efficiency was calculated from the experimental data obtained in 10 mM K3FeCN6 in 0.1 M NaOH at standard measurement conditions (potential sweep rate 5 mV s1, 30 °C). These experiments were performed for each sample individually, the obtained data is presented in Table 1. The collection efficiency for the Ptpc electrode was found to be 20.5 ( 2.4% which is in agreement with the literature data.40 2.3. Material Characterization. Scanning electron microscopy (SEM) was applied to examine the morphology of 3D arrays and performed using Jeol JSM-6700F (Japan) electron microscope with the lattice resolution 1 nm, at accelerating voltage 10 kV. Transmission electron microscopy (TEM) was applied to determine the size of Pt particles supported on VACNF arrays as described in ref.15 and performed with Topcon 002B (Japan) apparatus with the lattice resolution 0.17 nm at 200 kV accelerating voltage. The surface average size of Pt particles was calculated from multiple TEM images as = Σnidi3/Σnidi2. was estimated as 3 nm for ethylene glycol synthesis and 4 nm for electrochemical deposition.15 Other conditions being equal, we have not observed differences in the electrochemical performance of the samples obtained by these two synthetic approaches. This is in agreement with previous studies which confirm negligible differences in the ORR electrocatalytic activity within the interval of the particle sizes
from 3 to 4 nm (see review chapter ref.41 and references therein). Height estimation and roughness studies of the VACNF composites were performed using Dektak 6 M profilometer (Veeco, USA).
2.4. Modeling of the Oxygen Reduction Reaction at ThreeDimensional Nanostructured and Flat Electrodes. For the simulation of the currentpotential curves, the VACNF layer geometry was modeled (see Figure 1, c) as an array of vertically aligned cylindrical rods (radius Rf 100 nm) regularly distributed on a planar electrode surface with an axis-to-axis distance of approximately 600 nm.15,20 Using the diffusion domain approximation, the mass transport equations of oxygen and hydrogen peroxide are solved in a cylindrical unit cell containing one single rod.15 Under steady state conditions and using the cylindrical symmetry of the diffusion domain, the concentration profile of oxygen and hydrogen peroxide in the unit cell is obtained by solving the following set of equations: ! D2 ½O2 1 D½O2 D2 ½O2 þ ¼0 ð5Þ þ DO2 Dr 2 r Dr Dz2 D2 ½H2 O2 1 D½H2 O2 D2 ½H2 O2 þ þ Dr 2 r Dr Dz2
DH2 O2
! ¼0
ð6Þ
with the following boundary conditions: (i) In the bulk solution, for z = 0: [O2] = [O2]bulk and [H2O2] = 0 (ii) At the nanofiber surface, r = Rf, δ e z e H: DO2
DH2 O2
D½O2 Dr
¼ ΓPt St υ1
D½H2 O2 ¼ ΓPt St υ5 Dr
ð7Þ
ð8Þ
Here υ1 and υ5 stand for the O2 and H2O2 reaction rate, respectively, and will be considered in detail in the Annex. (iii) At r = Rc, 0 e z e H the concentration gradient in the radial direction was set to zero: DO2
DH2 O2
D½O2 Dr
¼0
D½H2 O2 ¼0 Dr
ð9Þ
ð10Þ
(iv) At z = H, Rfe r e Rc, and z=δ, 0e r e Rf the concentration gradient in the perpendicular direction was set to zero: DO2
DH2 O2
D½O2 Dz
¼0
D½H2 O2 ¼0 Dz
ð11Þ
ð12Þ
Under stationary conditions the ORR disk and the H2O2 escape current densities for the 3D electrodes calculated at z = 0 (see Figure 1):
jORR
4FDO2 ¼ Rc 2
Z 0
Rc
Z D½O2 DH2 O2 Rc D½H2 O2 dr þ 2F dr Dz z ¼ 0 Dz Rc 2 0 z¼0 ð13Þ
jH2 O2 ¼ 2F 9020
Z DH2 O2 Rc D½H2 O2 dr Dz Rc 2 0 z¼0
ð14Þ
dx.doi.org/10.1021/la2006343 |Langmuir 2011, 27, 9018–9027
Langmuir
ARTICLE
Figure 2. Ring (a, b) and disk (c, d) currents for the ORR obtained in experiment at rotation rate 900 rpm (a, c solid symbols) and by modeling (b, d open symbols). Note that ring currents are normalized to the corresponding collection efficiency. For further details the reader is referred to Experimental section. Currents are given per geometric surface area of the disk electrode. For demonstration purposes, forward scans starting from 0.05 V are shown. The modeling is performed for 3D electrodes as specified in the text. The symbols do not represent a complete set of data points. In case of a flat electrode geometry, solving of the mass transport equations can be simplified assuming a linear concentration gradient under steady state conditions and the mass transport equations for O2 and H2O2 can be written as: D½O2 DO2 ð½O2 bulk ½O2 x ¼ 0 Þ ¼ ΓPt St υ1 ð15Þ ¼ DO2 δ Dx x ¼ 0 DH2 O2
D½H2 O2 DH O ¼ 2 2 ½H2 O2 x ¼ 0 ¼ ΓPt St υ5 δ Dx x¼0
ð16Þ
, where x is the spatial coordinate normal to the electrode and x = 0 indicates in this case the electrode surface. Under stationary conditions the ORR and the H2O2 escape current densities of the flat electrodes are given by: jORR ¼ 4 F
DO2 DH O ð½O2 bulk ½O2 x ¼ 0 Þ þ 2F 2 2 ð½H2 O2 x ¼ 0 Þ δ δ ð17Þ jH2 O2 ¼ 2F
DH2 O2 ð½H2 O2 x ¼ 0 Þ δ
Figure 3. Effective number of electrons n participating in the ORR (a, b) and xH2O2 (c, d) obtained in experiment (a, c) and by modeling (b, d) as a function of the disk potential. Panel (e) shows a close-up of the xH2O2 vs E plot for experimental data. For experimental conditions see Figure 2. The modeling is performed for 3D electrodes as specified in the text. Symbols do not represent a complete set of data points.
to vary independently and in a wide range two parameters of the catalytic layer: the density of the active Pt surface sites and the diffusion path, while keeping the Pt particle size of 34 nm fairly constant. The density of the active sites is varied by changing the coverage of carbon filaments with Pt nanoparticles ΓPt, while the diffusion path is adjusted by varying the thickness of the catalytic layers L. Figure 2 represents typical RRDE data obtained for 3D structured samples showing an evolution of the ORR response as a function of the electrode geometry at constant rotation rate. While the disk currents are related to the overall rate of the ORR, the ring currents are proportional to the rate of the H2O2 escape from the diffusion layer. In the absence of Pt the VACNF layer (even at layer thickness of 21 μm), does not demonstrate neither significant ring nor disk currents, therefore ensuring low contribution of intrinsic activity of the substrate into the overall response of the system. Obtained data was further processed using standard algorithm in order to obtain the average number of electrons (n, eq 19) per O2 molecule participating in the reaction and the hydrogen peroxide molar (xH2O2, eq 20) fraction.25 n¼
ð18Þ
For the modeling of the experimental RRDE curves, we will consider three different ORR mechanisms which are discussed in detail in the Results and Discussion, while the kinetic equations are given in the Annex.
x H2 O2
3. RESULTS AND DISCUSSION 3.1. RRDE Study of the ORR on Pt/VACNF. VACNF layer (Figure 1, a) grown by the catalytic chemical vapor deposition features fairly good alignment in the lower part of the growth which was utilized for the preparation of the working electrodes (see ref.15). The active sites for the cathodic ORR are provided by Pt nanoparticles deposited on the VACNF (see TEM image in Figure 1, d). The overall structural organization of the working electrode is visualized in Figure 1, c. It is essential that the preparation method of the 3D nanostructured electrodes allows
4ID IR ID þ N IR 2 N ¼ IR ID þ N
ð19Þ
ð20Þ
Here, ID is a disk current, IR is a ring current, N is a collection efficiency. The values of the latter were obtained separately for each sample and can be found in Table 1. The evolution of n and xH2O2 with potential for various samples is presented in Figure 3. The typical values of n lie between 2 and 4 corresponding to two extreme cases, a value of 4 meaning that all the reacting O2 molecules are finally reduced into water even if H2O2 might be an intermediate specie of the reaction, whereas n = 2 means that the 9021
dx.doi.org/10.1021/la2006343 |Langmuir 2011, 27, 9018–9027
Langmuir
ARTICLE
Figure 4. Panels (a), (b) and (c) show H2O2 escape current, ORR disk current and H2O2 yield, respectively, simulated for mechanisms 1, 2, and 3 represented in panels (d) through (f). The rate constant values ki can be found in Table 2. Modeling results are presented for a flat electrode with active site density 2.2 109 mol cm2.
reduction of O2 molecules proceeds only to H2O2, which escapes from the catalyst layer without being reduced. In the first case, the hydrogen peroxide molar ratio xH2O2 is 0, in the second case xH2O2 = 1. In Figure 3, one can distinguish four potential intervals (Figure 3, a, c, see also Modeling part) corresponding to various ORR regimes. At low reaction overpotential (0.8 V < E < 1.0 V, interval I), the fraction of H2O2 produced is almost nil, even for very low amount of Pt in the catalytic layer. As the potential is shifted negative (0.4 V < E < 0.8 V, interval II), xH2O2 gradually increases suggesting potential activated H2O2 formation. As the potential is further shifted negative (interval III, 0.25 V < E
