Electrocatalytic Activity and Stability of Nitrogen-Containing Carbon

Jul 21, 2009 - (1) However, the slow kinetics of the oxygen reduction reaction ..... (49) The bond energies indicate that the strongest perturbation e...
0 downloads 0 Views 2MB Size
14302

J. Phys. Chem. C 2009, 113, 14302–14310

Electrocatalytic Activity and Stability of Nitrogen-Containing Carbon Nanotubes in the Oxygen Reduction Reaction Shankhamala Kundu,† Tharamani Chikka Nagaiah,‡ Wei Xia,† Yuemin Wang,† Stefan Van Dommele,§ Johannes Hendrik Bitter,§ Monika Santa,| Guido Grundmeier,| Michael Bron,†,‡ Wolfgang Schuhmann,‡ and Martin Muhler*,† Laboratory of Industrial Chemistry, Ruhr-UniVersity Bochum, D-44780 Bochum, Germany, Elektroanalytik & Sensorik, Ruhr-UniVersity Bochum, D-44780 Bochum, Germany, Inorganic Chemistry and Catalysis, Utrecht UniVersity, Utrecht, The Netherlands, and Max-Planck-Institut fu¨r Eisenforschung, D-40237 Du¨sseldorf, Germany ReceiVed: December 22, 2008; ReVised Manuscript ReceiVed: May 14, 2009

Nitrogen-containing carbon nanotubes (NCNTs) were prepared via pyrolysis of acetonitrile over cobalt catalysts at different temperatures to control the nitrogen content. The changes in the chemical and structural properties of undoped CNTs and NCNTs were investigated using high-resolution X-ray photoelectron and Raman spectroscopy. The NCNTs prepared at 550 °C had a higher amount of pyridinic groups and edge plane exposure than the ones prepared at 750 °C. The thermal stability and transformation of these nitrogen functional groups was studied using deconvoluted XP N 1s spectra. The NCNTs show a considerably higher activity in the oxygen reduction reaction in acidic electrolyte compared with undoped CNTs as demonstrated by cyclic voltammetry, rotating disk electrode measurements, and the redox-competition mode of scanning electrochemical microscopy (RC-SECM). Particularly, the NCNT sample prepared at 550 °C exhibited the highest activity, which was about 1 order of magnitude lower than that of a commercial Pt/C sample containing 20 wt % Pt. The oxygen reduction reaction (ORR) performance of this sample showed hardly any signs of deterioration after 3 days, as determined by voltammetric stability tests in H2SO4. 1. Introduction Proton exchange membrane fuel cells (PEMFCs) are considered to be an alternative high efficiency energy converting system for stationary and mobile applications. The development of a durable, low cost, highly active cathode electrocatalyst is one of the most critical challenges for the successful introduction of fuel cells into the market. At present, the most effective cathode electrocatalyst is composed of platinum metal or its alloys with carbon black as support.1 However, the slow kinetics of the oxygen reduction reaction (ORR) even on Pt catalysts is among the most limiting factors in the energy conversion efficiency of fuel cells.2 On the other hand, Pt is an expensive metal with low abundance. Hence, during the last few decades, numerous studies have been devoted to finding alternative nonnoble electrocatalysts for the cathode side of PEM fuel cells. Metal-N4 macrocycles, such as Fe- and Co-porphyrins and -phthalocyanines, have attracted great attention due to their high activity in the ORR.3-6 A major drawback of this kind of catalyst is its low stability under PEM fuel cell conditions, which is caused by several major problems related to the generation of H2O2 as reactive intermediate: oxidation of the carbon support, leaching of Fe and Co, and the degradation of the Nafion membrane.3,7 The drawback of low stability may partially be overcome by heat treatment or electropolymerization of carbon black supported metalloporphyrins and phthalocyanines.8-13 Depending on the applied heat treatment temperature, the macrocycle is partially or completely destroyed,11,12,14 and * To whom correspondence should be addressed. Telephone: +49 234 32 28754. Fax: +49 234 32 14115. E-mail: [email protected]. † Laboratory of Industrial Chemistry, Ruhr-University Bochum. ‡ Elektroanalytik & Sensorik, Ruhr-University Bochum. § Utrecht University. | Max-Planck-Institut fu¨r Eisenforschung.

species such as FeN2Cx and FeN4Cy have been proposed to form as active sites for the ORR.14,15,18 Experimental and theoretical results suggest that similar active sites are formed by heat treatment of much simpler iron, nitrogen, and carbon black compounds, avoiding the macrocycles, and that these are promising candidates as alternative catalysts for the cathode side in PEM fuel cells.14,16-19 However, as pointed out recently,1 the activity and stability of these noble-metal-free electrocatalysts is still too low for fuel cell applications, and further improvements are required. Recently, nitrogen-containing carbon nanotubes (NCNTs) have been prepared via pyrolysis of acetonitrile at 900 °C using iron particles supported on alumina as catalysts and tested with respect to the ORR.20 It was proposed that the role of the iron particles is likely that of a catalyst for both the CNT growth and the formation of catalytically more active carbon nanostructures during pyrolysis.20 However, it was concluded that iron is not part of the active site for the ORR.20 Several other publications suggest that the metal acts as a catalytic agent that helps to stabilize the nitrogen groups in the graphene sheets and to form carbon nanostructures with a higher degree of edge plane exposure.21-25 It was also demonstrated that there is neither a correlation between the activity and the metal phase present in the NCNTs, nor any evidence for the formation of MNx sites. However, edge plane exposure alone did not lead to improved activity in the ORR in acidic electrolytes using carbon without nitrogen doping.26,27 It is likely that the presence of nitrogen enhances the ability of the graphene sheets to donate electrons. Therefore, the nitrogen content in the carbon material is presumably an important parameter, and it is of great interest to identify which nitrogen functional groups are the main active sites for the ORR.

10.1021/jp811320d CCC: $40.75  2009 American Chemical Society Published on Web 07/21/2009

Electrocatalytic Activity and Stability of NCNTs Nitrogen can be present in different forms in NCNTs including pyridine-like, pyrrole-like, imine-like, or graphitic nitrogen. The amount of each species depends on the synthesis conditions such as temperature, the feed gas composition, the nitrogen precursor used for the growth, and the metal catalysts. It is believed that either pyridinic- or pyrrole/pyridine-type nitrogen is responsible for the ORR activity of NCNTs.24,28 However, a recent study suggests that more graphitic nitrogen rather than pyridinic nitrogen is important for the ORR.29 It is also reported that a heat treatment leads to the formation of more stable nitrogen functional groups.30,31 Stanczyk et al.31 showed that at low treatment temperatures pyrrolic structures transform into pyridinic structures and vice versa via dynamic surface rearrangements. With the increase of treatment temperature, both pyrrolic and pyridinic structures transform into more stable quaternary nitrogen. The reaction pathways of the transformation of different nitrogen functional groups are complicated. Given the presence of a large variety of carbon structures and different nitrogen doping methods, detailed studies on the thermal stability of nitrogen-containing functional groups are still not available. In the present study, we investigated nitrogen-containing CNTs directly grown from acetonitrile at two different temperatures using Co as metal catalyst. The amount and type of nitrogen functional groups were identified, and the thermal stability of these groups was investigated using high-resolution X-ray photoelectron spectroscopy (XPS). The catalytic activity of undoped and nitrogen doped CNTs in the ORR was studied electrochemically aiming at providing better insight into the nature of the active sites. 2. Experimental Section The nitrogen-containing CNTs used in this work were synthesized according to a procedure described before.32 Briefly, the NCNTs, of varying N/C composition, were prepared by chemical vapor deposition of acetonitrile over a supported Co catalyst. The NCNTs were grown at 550 °C (NCNT-L) and 750 °C (NCNT-H) for 16 h using a helium flow (50 sccm) saturated by bubbling through the liquid precursor at 30 °C. The as-grown NCNTs were first refluxed in a 1 M KOH solution for 1 h to remove the silica support and then, after washing, refluxed in concentrated HCl for 2 h to dissolve the metal particles. The undoped CNTs (Baytubes, Bayer MaterialScience, Leverkusen) used for comparison have inner diameters of 4-6 nm and outer diameters of 13-16 nm. 2.1. XPS Measurements. The XPS measurements were carried out in an ultrahigh vacuum (UHV) setup equipped with a high-resolution Gammadata-Scienta SES 2002 analyzer. A monochromatic Al KR X-ray source (1486.3 eV; anode operating at 14.5 kV and 50 mA) was used as incident radiation. The analyzer slit width was set at 0.3 mm and the pass energy was fixed at 200 eV for all the measurements. The overall energy resolution was better than 0.5 eV. A flood gun was used to compensate for the charging effects. The binding energies were calibrated based on the graphitic C 1s peak at 284.5 eV. The samples for the XPS measurements were prepared by gently pressing the CNT powders into the cavity of the XPS sample holder in order to achieve a smooth flat surface. The thermal treatment of the NCNTs in UHV (2 × 10-9 mbar) was carried out in the preparation chamber of the XPS setup for 30 min at 300, 440, 590, and 720 °C. Prolonged treatment of 120 min was performed at 720 C to monitor any further effect of the heat treatment. Prior to the elemental scans, a survey scan was measured for all the samples in order to detect all the elements

J. Phys. Chem. C, Vol. 113, No. 32, 2009 14303 present. The XP spectra were deconvoluted using the CASA XPS program with a Gaussian-Lorentzian mix function and Shirley background subtraction. The full width at half-maximum (FWHM) values were fixed at a maximum limit of 1.2 eV for all the peaks during fitting. The peak positions were reproducible by applying the same Gaussian to Lorentz ratio (30:70) and FWHM. All the fitted peaks of the deconvoluted spectra were self-consistent, allowing quantitative comparisons. 2.2. Raman Spectroscopy. Raman spectroscopy measurements were performed using a modular system (Dilor, LabRAM, ISA Instruments, SA, INC) that consisted of a green Ar+ laser (514.532 nm) with adjustable output power, a 100 × 0.9 microscope objective, an Olympus BX40 confocal microscope, and a charge-coupled device camera (cooled down to -70 °C by liquid nitrogen). Raman spectra for all the CNT samples were recorded under the following conditions: the objective of the Raman spectrometer was 10×, the laser power was set at 10 mW, and the spectra acquisition time was 100 s. 2.3. CV and RDE Measurements. For cyclic voltammetry (CV) and rotating disk electrode (RDE) measurements, glassy carbon electrodes (Ø 3 mm, HTW Hochtemperatur-Werkstoffe GmbH, Germany) embedded in Teflon were used as working electrodes, which were polished on polishing cloth using different grades of alumina paste (3, 1, 0.3, and 0.05 µm, Leco, Michigan) to obtain a mirror finish. The RDE measurements were carried out using an analytical rotator (model ASR2) from Pine Instruments. The CV and RDE measurements were carried out using an Autolab potentiostat/galvanostat (PGSTAT12) in a single compartment glass cell using a three electrode arrangement. The working electrode was prepared as follows: a total of 1.25 mg of CNTs was dispersed in 1 mL water and was sonicated for 30-45 min. Then, 2 µL of this suspension was dropped onto the glassy carbon electrode, and the electrode was dried at room temperature. A platinum grid was used as counter electrode, and a double junction Ag/AgCl/3 M KCl electrode served as reference electrode. All CV and RDE experiments were carried out at a scan rate of 5 mV/s in the potential range of +0.7 to -0.2 V in 0.5 M H2SO4. Argon or oxygen was used to purge the solution to achieve oxygen-free or oxygen-saturated electrolyte solution. Commercial Pt/C catalysts (20 wt %) were purchased from BASF Fuel Cell Inc. (New Jersey, earlier E-TEK) and used for comparison. The voltammetric stability test of NCNT-L was performed in O2saturated 0.5 M H2SO4 using CV in the potential range from 0.8 to -0.2 V at a scan rate of 5 mV/s for up to 3 days. All measurements except for the stability test were repeated at least four times for verification. 2.4. RC-SECM Measurements. Redox competition scanning electrochemical microscopy (RC-SECM)33 was carried out using a SECM setup (Sensolytics, Bochum, Germany) in combination with a bipotentiostat (PG 100, Jaissle, Waiblingen, Germany). In order to avoid background current shift caused by uncompensated tilt angles between the scanning plane of the SECM tip and the sample, a software based tilt correction procedure was used. The details of the instrumentation have been described previously.34 Homemade Pt (Goodfellow, Bad Nauheim, Germany) microelectrodes with 25 µm diameter and a glass to metal ratio (RG) of 100:25 (4.0) were used as SECM tips. The electrochemical cell was composed of four electrodes with the SECM tip connected as working electrode 2 and the sample (modified glassy carbon electrode) as working electrode 1. The catalyst spots on the glassy carbon plate (SigradurG, HTW, Thierhaupten, Germany) were prepared by dispersing 300 µg nanotube samples in 2 mL of water followed by sonication

14304

J. Phys. Chem. C, Vol. 113, No. 32, 2009

Kundu et al.

TABLE 1: Surface Composition of NCNTs and Undoped CNTs Derived from the XP Spectra sample

C%

O%

N%

Co%

NCNT-H NCNT-L undoped CNTs

90.3 86.4 96.6

2.0 6.9 3.4

7.5 6.3 -

0.2 0.4 -

for 30 min. A total of 0.1 µL of each suspension was dropped on the glassy carbon plate using a capillary tube with a sharp edge (diameter 200-500 µm). Ag/AgCl/3 M KCl and Pt foil were used as the reference and counter electrode, respectively. The catalyst-modified sample was positioned at the bottom of the electrochemical cell horizontally with the catalyst deposited surface facing the tip. SECM measurements were carried out in 0.5 M H2SO4. Area scans were obtained with a scan rate of 50 µm/s at a tip-tosample distance of 10 µm. During scanning, a pulse potential profile was applied to the SECM tip which consisted of 2 steps: (1) a resting potential (P1, 550 mV, 500 ms), where the diffusion controlled restoration of equilibrium occurred without invoking a redox process; (2) a measuring potential (P2, -100 mV, 400 ms), where oxygen was reduced at the tip. During a SECM measurement, the sample was continuously polarized at a constant potential (ES) at which O2 can be catalytically reduced. Measurements have been performed at Es values of -100, +100, and +200 mV. As a result, a competition of O2 reduction occurs between the Pt microelectrode used as SECM tip and the sample during P2. Thus, a low oxygen reduction current at the SECM tip reflects a high activity of the underlying sample. 3. Results 3.1. Compositional Analysis by XPS. The NCNTs were found to consist of hollow, tubular-like structures of comparable outer diameter between 12 and 17 nm. The elemental surface composition of NCNTs and undoped CNTs was investigated using XPS. The XPS survey scan of NCNTs revealed the presence of carbon, nitrogen, oxygen, and a small amount of Co. Only carbon and oxygen were detected in the undoped CNTs. The surface atomic concentrations of C, O, N, and Co were calculated from the corresponding peak areas of the XP spectra (Table 1). NCNT-H grown at 750 °C contains 7.5% nitrogen, which is higher than the 6.3% nitrogen of NCNT-L grown at 550 °C. On the other hand, NCNT-L has a larger amount of oxygen on the surface than NCNT-H, which can be attributed to the low thermal stability of surface oxygen species, suggesting that more surface oxygen species decompose at higher temperatures.35 The total Co surface content of NCNT-H and NCNT-L was 0.2% and 0.4%, respectively, which remained in the sample after several purification steps including refluxing in KOH and HCl. Obviously, it is difficult to remove the catalyst particles from the sample entirely, since some of the Co particles are fully covered by a carbon layer during NCNT growth.32 It has been shown that the average tube diameter of NCNTs increases gradually with increasing growth temperature.32 This results in a lower contribution in the XP Co 2p spectra from the Co particles located inside the tubes because of the limited sampling depth of XPS. At the same time, a higher growth temperature leads to an increased Co particle size and, correspondingly, to a lower Co XPS surface atomic concentration. These two factors may account for the lower amount of Co detected by XPS in NCNT-H than in NCNT-L. Figure 1 represents typical XP C 1s spectra of NCNT-H, NCNT-L, and undoped CNT. The C 1s spectrum of undoped CNTs has a peak at 284.5 eV which mainly represents graphitic

Figure 1. XP C 1s spectra of NCNT-H, NCNT-L, and undoped CNTs.

Figure 2. Deconvoluted XP N 1s spectra of (a) NCNT-L and (b) NCNT-H.

carbon. According to literature, the presence of C-N bonds gives rise to a peak in the range from 286.0 to 287.0 eV depending on the type of nitrogen functional groups present on the surface.36,37 The C 1s spectrum of undoped CNTs has almost no shoulder at around 286.5 eV. With increasing nitrogen content, the C-N peak intensity increases, which is visible as a shoulder in the C 1s spectrum of NCNT-L. The shoulder increases further for the NCNT-H sample, which has the highest nitrogen concentration of about 7.5%. Figure 2 shows the deconvoluted XP N 1s spectra of the NCNT-L and NCNT-H samples. The contributions of different nitrogen species to the N 1s peaks are summarized in Table 2. Fitting of the N 1s spectra yields peaks at the following binding energies: peak N1 at 398.4 ( 0.2 eV can be attributed to the pyridinic nitrogen; peak N2 at 400.1 ( 0.1 eV is mainly assigned to pyrrole-type nitrogen; peak N3 at 401.0 ( 0.1 eV arises due to the presence of nitrogen substitutes in the aromatic graphene

Electrocatalytic Activity and Stability of NCNTs

J. Phys. Chem. C, Vol. 113, No. 32, 2009 14305

TABLE 2: Concentrations of Different N-Containing Functional Groups Present in the NCNTs as a Function of the Pretreatment Temperature Obtained by Deconvolution of the XP N 1s Spectraa pretreatment temperature (°C)

N1

N2

N3

N4

N5

NCNT-H

23 300 450 600 720

20.3 20.3 19.2 18.8 20.3

11.7 9.8 8.5 9.5 6.0

46.3 41.1 46.0 46.1 48.8

13.7 18.1 16.5 18.4 16.9

8.0 10.7 9.8 7.2 8.0

NCNT-L

23 300 450 600 720

42.8 41.9 50.3 49.6 49.4

16.6 11.4 6.2 6.6 5.9

28.2 31.0 30.3 34.5 35.8

7.1 8.6 5.7 4.7 4.1

5.3 7.1 3.5 4.6 3.8

sample

a The thermal treatment was performed in the UHV chamber of the XPS setup. N1 at 398.4 eV (pyridinic); N2 at 400.1 eV (pyrrole-type); N3 at 401.0 eV (quaternary); N4 at 402.3 (pyridine-N-oxide); N5 at 403.5 eV (nitrogen oxides).

Figure 4. XPS surface atomic concentrations of different nitrogen groups as a function of the pretreatment temperature. The pretreatment was performed in the UHV chamber of the XPS setup.

Figure 3. XP Co 2p spectra of NCNT-H and NCNT-L.

structure, that is, quaternary nitrogen; peak N4 at 402.2 ( 0.1 eV can be attributed to the pyridine-N-oxide; and peak N5 at 403.7 ( 0.1 eV corresponds to the chemisorbed nitrogen oxide species.38,39 The quantitative analysis of the N 1s spectra reveals that the content of the different nitrogen groups is different for the NCNT-H and NCNT-L samples. Table 2 shows that the amount of nitrogen species present as pyridinic groups in NCNT-L (42.8%) is much higher than that in NCNT-H (20.3%). However, more quaternary nitrogen was identified in NCNTH. The difference could be related to the transformation of pyridinic groups to quaternary nitrogen groups at high temperatures. In addition, the pyrrolic-type nitrogen species are another major group present on the carbon surface. The amount of chemisorbed NOx species is more than 5% in both samples. Figure 3 shows the Co 2p spectra of the NCNT-H and NCNT-L samples consisting of Co 2p1/2 and 2p3/2 doublet peaks. Due to the very small amount of Co in both samples, only a broad Co 2p3/2 peak was obtained at around 779.3 eV. The presence of a shakeup satellite peak is observed at about 786.0 eV for both the samples. The peak at 779.3 eV corresponds to the quoted value for cobalt oxides (mainly Co3O4).37,40 While the Co XP spectra may be satisfactorily explained by the presence of oxides, the presence of Co in a CoN4- or CoN2like environment cannot be excluded in our samples, which would give rise to a peak at about 779.0 ( 0.2 eV. 3.2. Thermal Stability. To gain an improved insight into the thermal stability of different nitrogen functional groups, XPS measurements of NCNTs after heat treatment at different

temperatures were performed. After peak deconvolution based on the above assignments, changes of different nitrogencontaining functional groups were derived. The changes of nitrogen concentrations in NCNT-H and NCNT-L were calculated, and they are shown in Figure 4 as a function of the preheating temperature. In the case of NCNT-L, nitrogen shows a slight but gradual decrease from 6.3% to 5.8% upon heating to 720 °C in UHV. Correspondingly, there is an increase of carbon. The NCNT-H sample shows almost no change in the total amount of nitrogen even after heating at 720 °C for 120 min. As a whole, there is no significant change of N amount upon heating for both NCNT-H and NCNT-L. We further investigated whether different nitrogen species transform into each other upon heating, so that samples with a higher amount of certain nitrogen species may be obtained by heat treatment. Based on the deconvoluted XP N 1s spectra, the relative concentrations of different nitrogen functional groups were derived as a function of the heating temperature as summarized in Table 2. It can be seen that there are no significant changes in any of the nitrogen functional groups in NCNT-H and NCNT-L upon heating. Minor changes of all functional groups were observed upon heating, which, however, are not significant to draw any meaningful conclusion. Hence, for electrochemical studies, NCNTs without additional thermal treatment were used. 3.3. Raman Spectroscopy. Differences in the crystal structure of undoped CNTs and NCNTs were assessed by means of Raman spectroscopy. Figure 5 displays the Raman spectra of undoped CNTs (Figure 5a) and NCNT-L (Figure 5b). Both spectra show two bands at about 1580 and 1333 cm-1, which can be assigned to the G-band and D-band of carbon structures.41 The G-band, arising from the Raman-active E2g in-plane vibration, is characteristic of the graphitic phase indicating the presence of crystalline graphitic carbon. The D-band originates

14306

J. Phys. Chem. C, Vol. 113, No. 32, 2009

Kundu et al.

Figure 6. Cyclic voltammograms measured in oxygen-saturated 0.5 M H2SO4 at a scan rate of 5 mV/s: (1) bare glassy carbon electrode; (2) undoped CNTs; (3) NCNT-H; (4) NCNT-L.

Figure 5. First-order Raman spectra of (a) undoped CNTs, and (b) NCNT-L.

from the A1g in-plane breathing vibration and corresponds to disorder. The intensity ratio of the two bands (ID/IG) is considered as a key parameter to assess the degree of disorder of CNTs.42,43 Clearly, NCNT-L exhibits a higher amount of defects with an ID/IG ratio of 1.25 as compared to 1.05 in case of the undoped CNTs. In addition, both the D-band and G-band of NCNT-L are broadened, showing larger FWHMs. This is in agreement with earlier studies by Tascon et al.44,45 showing that the first-order band broadening correlates strongly with the disorder in the basal plane of graphite. Moreover, the G-band of NCNT-L is shifted to higher wavenumbers (1598 cm-1) compared with the undoped CNTs (1580 cm-1). This shift can be explained by the contraction of the lattice spacing due to the formation of C-N bonds, which increases the size of the Brillouin zone.46 Therefore, the stretching of in-plane covalent bonds is stronger which shifts the G-band to higher frequencies. The ID/IG ratio of NCNT-H was determined to be 1.17, which is slightly lower than that of NCNT-L (1.25). The XPS measurements revealed that the nitrogen content is somewhat higher in NCNT-H than in NCNT-L. The deconvoluted N 1s spectra disclosed that the NCNT-L sample has a much higher amount of pyridinic groups, which obviously contribute significantly to the disorder and hence to a surface with more defects. 3.4. CV and RDE Measurements. Cyclic voltammetry was carried out in 0.5 M H2SO4 in the potential range from 0.8 to -0.2 V at a scan rate of 5 mV/s. First, a CV was recorded in the absence of oxygen by purging the electrolyte solution with argon. Subsequently, oxygen was dissolved in the solution before the measurements and applied above the solution during the measurements. The obtained voltammograms did not indicate any redox reaction in the absence of oxygen. Figure 6 shows the CVs of undoped CNTs, NCNT-H, and NCNT-L measured in the presence of oxygen. The voltammogram obtained from a bare glassy carbon electrode under the same conditions is shown additionally for comparison. The reduction of oxygen occurred at potentials of 0.4-0.6 V at the NCNT samples, whereas no reduction current was generated in the same potential range at the undoped CNTs. However, all three samples appeared active for the ORR at more negative potentials, resulting in similar slopes of the reduction current with decreasing potential at potentials of 0.2 to -0.2 V despite that

Figure 7. (a) RDE linear sweep voltammograms and (b) Tafel plots for ORR in oxygen-saturated 0.5 M H2SO4 at a rotation speed of 900 rpm and a scan rate of 5 mV/s with (1) undoped CNT, (2) NCNT-H, (3) NCNT-L, and (4) E-TEK Pt/C (20 wt %) as catalyst.

the absolute values of the current were different. It is believed that the ORR activity of the NCNTs at more positive potentials (0.2-0.6 V) results from N-containing active sites which do not exist in undoped CNTs. Obviously, the oxygen reduction current is limited by the availability of these active sites and not by the diffusional mass transport of oxygen to the electrode surface. Thus, at more negative potentials, also the carbon surface itself becomes active for the ORR. A closer inspection of the potential region of 0.4-0.6 V discloses that the onset potential for the ORR at NCNT-L is about 0.54 V and at NCNT-H about 0.5 V. This further indicates that the active sites generated at NCNT-L exhibit higher activity for the reduction of oxygen in acidic medium. RDE experiments were carried out at 100, 400, and 900 rpm (see the Supporting Information, Figure S1). Figure 7a compares the RDE linear sweep voltammograms of the NCNT-L and NCNT-H samples in oxygen-saturated H2SO4 at 900 rpm. For both NCNTs, the ORR onset potential is shifted to more positive values as compared to the undoped CNTs, demonstrating the catalytic activity of the formed active sites for the ORR at the NCNTs. Furthermore, it is obvious that the catalytic activity of

Electrocatalytic Activity and Stability of NCNTs

J. Phys. Chem. C, Vol. 113, No. 32, 2009 14307

NCNT-L is higher than that of NCNT-H. For comparison, a commercial Pt/C catalyst from E-TEK with a high Pt loading of 20% was investigated under the same experimental conditions. The Pt/C catalyst shows an onset potential for the ORR at 900 rpm of 0.65 V, which is about 0.1 V more positive than that of NCNT-L. Considering the nonmetallic and especially the noble-metal-free nature of the catalytically active sites formed at the NCNT surfaces, their activity in the ORR is surprisingly high. For a more quantitative evaluation of the electrocatalytic activities of the NCNTs with respect to the ORR, rotation-ratedependent RDE experiments were analyzed using the first-order Koutecky-Levich equation47

1/j ) 1/jk + 1/Bω1/2

(1)

where jk is the kinetic current density and B is related to the diffusion-limited current density through the expression jd ) Bω1/2. Here, ω is the angular velocity of rotation. The parameter B is defined as

B ) 0.62nFγ-1/6DO22/3CO2

Figure 8. Electrochemical stability measurements of NCNT-L by means of continuous cyclic voltammetry in oxygen-saturated 0.5 M H2SO4 at a scan rate of 5 mV/s.

(2)

where n is the (average) number of electrons exchanged per molecule of O2; F is the Faraday constant; γ is the kinematic viscosity of the electrolyte; DO2 is the oxygen diffusion coefficient; CO2 is the bulk oxygen concentration in the electrolyte; and ω is the angular velocity of the electrode. The values used in this work were 1.15 × 10-5 cm2/s for the diffusion coefficient of oxygen and 1.4 × 10-6 mol/cm3 for the bulk oxygen concentration under the given conditions.48 The kinetic current density (jk) and the Koutecky-Levich slope (1/B) can be obtained from a plot of 1/j versus 1/ω1/2. The Koutecky-Levich plots were plotted at -0.1 V for all investigated catalysts (see the Supporting Information, Figure S2). Parallel and straight lines of 1/i versus 1/ω1/2 indicate a first-order reaction with respect to dissolved oxygen. From the value of 1/Bexp, an average number of electrons transferred per oxygen molecule can be calculated using eq 2. The n value of NCNT-L is derived to be 3.92, indicating that most of the O2 is reduced to H2O either directly by a simultaneous four-electron reduction or via H2O2 as primarily formed intermediate followed by further reduction of H2O2 to H2O. However, the n value obtained for NCNT-H is only 2.48, which suggests that the active sites introduced at the NCNT-H surface catalyze oxygen reduction mainly via two-electron reduction of O2 to H2O2. Tafel plots (Figure 7b) reveal that both NCNT samples exhibit much higher kinetic currents as compared with undoped CNTs. In particular, in H2SO4, NCNT-L has an about 100-fold higher kinetic current compared with undoped CNTs. This current is only 1 order of magnitude lower than that of the expensive commercial 20 wt % Pt/C catalyst at a potential of 0.5 V versus Ag/AgCl/3 M KCl, which is the relevant potential for fuel cell applications. The XPS studies showed that the surface N-containing groups on NCNTs were very stable upon thermal treatments at different temperatures. In addition to the thermal stability measurements, the electrochemical stability under ORR conditions of NCNT-L was studied in O2-saturated 0.5 M H2SO4 for 3 days using sequential cyclic voltammetry. As can be seen in Figure 8, there is hardly any change in the onset potential or in the kinetic current after continuous ORR reaction for 3 days, indicating

that the catalytic sites of NCNT-L are rather stable in acidic medium. The results were supported by long-term chronoamperometric studies for NCNT-L conducted in O2-saturated H2SO4 using RDE at a rotation speed of 900 rpm (see the Supporting Information, Figure S3). 3.5. RC-SECM Studies. RC-SECM experiments were carried out in 0.5 M H2SO4 using a pulse potential profile as described in the Experimental Section for the visualization of the local ORR activity of different catalyst spots on a single glassy carbon plate. When the SECM tip is scanned over an active region of the sample in the redox competition mode, where the underlying catalyst consumes oxygen, the oxygen reduction current at the tip will decrease due to the decrease in the oxygen concentration within the gap between the accurately positioned SECM tip and the catalyst surface. Figure 9 shows the RC-SECM images after background current subtraction at sample potentials Es ) -0.1, +0.1, and +0.2 V. The local catalytic activity is illustrated by the brightness, that is, a bright color indicates a smaller tip current, which corresponds to a higher local ORR activity of the sample. As shown in Figure 9a, undoped CNTs exhibit almost no catalytic ORR activity even at Es ) -0.1 V. In contrast, NCNT-modified spots show a significant decrease in local oxygen concentration with NCNT-L clearly exhibiting higher catalytic activity toward ORR as compared with NCNT-H-modified surface areas. At a more positive potential of Es ) +0.1 V, the difference in ORR activity between NCNT-L and NCNT-H is also clearly visible (Figure 9b). NCNT-L still exhibits a considerable ORR activity even at Es ) +0.2 V, whereas NCNT-H-modified areas do not show a pronounced catalytic activity at this Es value (Figure 9c). The normalized line scans (current as a function of the x displacement) along the most active area of the two NCNT samples at y ) 1300 µm (Figure 9d) are clearly demonstrating the differences in local catalytic ORR activity of the two NCNT samples at all three sample potentials. The RC-SECM results agree well with the results from the CV and RDE measurements. 4. Discussion In contrast to the RDE experiments, the RC-SECM experiments show a direct comparison of different catalysts in a single experiment and moreover the local distribution of catalytic activity on surfaces. Thus, the RC-SECM measurements additionally confirm the surprisingly high catalytic ORR activity of the NCNTs in acidic electrolyte as compared with CNTs or other carbon materials. There are several possible reasons for this enhancement of catalytic ORR activity. Edge planes are known to be the active sites for a variety of different catalytic reactions including ORR in acidic electrolytes. Nitrogen is known to be able to create defects

14308

J. Phys. Chem. C, Vol. 113, No. 32, 2009

Figure 9. RC-SECM images of NCNT-L, NCNT-H, and undoped CNTs over glassy carbon for ORR in 0.5 M H2SO4 at sample potentials Es of (a) -0.1 V, (b) +0.1 V, and (c) +0.2 V. (d) x-Line scans obtained along the most active area of the NCNT samples at y ) 1300 µm at different sample potentials.

on carbon surfaces, which could lead to an increase in edge plane exposure and thus enhanced catalytic activity.24,25 However, Ozkan and co-workers20,22 showed that increased edge plane exposure without simultaneous nitrogen doping did not lead to improved activities for ORR. Quantum mechanical calculations show that the bond strength of carbon adjacent to nitrogen,

Kundu et al. which is centrally located in the aromatic ring, is only slightly increased. However, the bond strengths at the edge sites are increased more.49 The bond energies indicate that the strongest perturbation effect of nitrogen is local. When the nitrogen atom is close to an edge, however, it magnifies the reactivity of edge sites toward bonding of intermediate radicals. Hence, it is important to obtain information on the optimal content of different nitrogen functional groups on the surface in order to obtain highest catalytic activity in the ORR. The investigated NCNT-L and NCNT-H samples have a high overall nitrogen content amounting to 7.5% for NCNT-H, which is higher than that of NCNT-L amounting to 6.3% as determined by the XPS measurements. However, NCNT-L contains a much higher amount of pyridinic groups, which are believed to be responsible for the observed high ORR activity. Raman measurements additionally prove that NCNT-L has a higher amount of defect sites with a higher edge plane exposure compared with NCNT-H. While there seems to be a consensus in literature that pyridine-type nitrogen can be considered as responsible for high activity with respect to the ORR, the role of metals is still a matter of controversy. Recent studies have shown that only a very small amount of iron in carbon is necessary to obtain high ORR activities.14 The same work also suggested that the active sites in Fe/N-modified carbon black and related materials were FeN2Cx and FeN4Cy.14 Thus, given the analogy of Co and Fe porphyrins with respect to catalytic oxygen reduction, the relation between the cobalt phases present in the investigated NCNTs and the observed catalytic ORR activity has to be discussed. The XPS study revealed that a very small amount of cobalt remains in the NCNTs after purification. A detailed analysis of the Co 2p spectra and N 1s spectra of NCNTs indicates that a very small amount of cobalt nitride is likely to be present in the sample. On the other hand, we observed no voltammetric peak that would correspond to the oxidation of metallic cobalt or reduction of cobalt oxide on the surface of the CNTs. Transmission electron microscopy studies of the NCNTs showed that the cobalt particles were covered by at least a few graphene layers,32 indicating that the cobalt nanoparticles are not accessible by the electrolyte, which is in agreement with the CV measurements. It is interesting to note that Jaouen et al.50 obtained strong evidence that the proposed active sites are hosted in micropores. However, such micropores are certainly not present in our NCNTs. Based on these facts, we would like to tentatively exclude the contribution of metalcontaining active sites such as CoN2Cx or CoN4Cy to the high ORR activities of the NCNTs. It is more likely that the cobalt particles catalyzed the growth of NCNTs with a high edge plane exposure, which leads to the high ORR activities. Therefore, our further discussion will be restricted to NCNTs assuming that cobalt is not part of the active sites. It seems that both exposed edge planes and nitrogen doping are necessary for high ORR activities. Maldonado and Stevenson24 observed a strong correlation between the oxygen reduction peak potential and the nitrogen dopant level. The oxygen reduction peak potential for a given scan rate shifts by (30 mV per 1 atom % of nitrogen incorporated into carbon. However, our study suggests that the shift in reduction peak potential is not directly related to the total nitrogen content. From RDE experiments, it is obvious that NCNTs containing a higher amount of pyridinic groups, that is, NCNT-L, show a higher catalytic activity for ORR than NCNT-H. Hence, it seems to be important to find optimal conditions for incorporating nitrogen in pyridinic form during the NCNT formation.

Electrocatalytic Activity and Stability of NCNTs In an attempt to influence the amount and ratio of different nitrogen functional groups, a heat treatment of the samples was carried out. Thermal stability investigations revealed that all nitrogen functional groups were more stable than surface oxygen functional groups. Our previous study showed that almost all of the oxygen-containing functional groups were decomposed upon heating to 720 °C, leaving some ether groups presumably below the CNT surface.35 In contrast, there is almost no decrease in the total content of nitrogen functional groups upon heating to 720 °C. This is in contrast to recent findings concerning N-containing functional groups introduced by post-treatment of CNTs with NH3, which decreased significantly upon heating.51 Thus, it appears that the thermal stability of different nitrogen functional groups present in the CNT structure strongly depends on the method of nitrogen incorporation in the graphite structure. Since the desorption energy of these groups is directly related to their bond strength to the graphitic structure, the thermal stability study indirectly reflects how strongly the nitrogen atoms are bound to the basal plane of carbon. In addition, since almost no interconversion of the nitrogen species was observed by means of XPS, an additional thermal treatment seems not to be beneficial for improving the catalytic ORR activity. The pyrolysis temperature during NCNT growth proved to be the most important parameter determining the maximum amount of active sites. At the same time, the electrochemical stability tests indicate that the incorporation of nitrogen functional groups within the carbon structure enhances the stability against oxidation under ORR conditions even in acidic medium. The value for n, that is, the average or effective number of electrons transferred per oxygen molecule, as derived from Koutecky-Levich plots shows that NCNT-L (n ) 3.92) is rather selective toward complete reduction of O2 to H2O. NCNT-H shows a lower n value (2.48), indicating that O2 is mainly reduced to H2O2. In literature, it has been shown that the final product of the ORR (H2O versus H2O2) is influenced by the plane at which O2 reacts. At the basal planes of nitrogen-doped carbon, O2 could only be reduced to H2O2.26,27 However, further reduction to H2O is possible at the edge planes.26,27,32 Additionally, it has been reported that, in both neutral and alkaline media, N-doped CNTs exhibit an over 100-fold increase in catalytic activity for H2O2 decomposition.24 Thus, it is likely that the ORR at NCNTs follows a 2 × 2 electron reduction pathway, where in a first step O2 is reduced to H2O2, which is then further reduced to H2O. The higher edge plane exposure of NCNT-L is therefore responsible for the nearly complete reduction of O2 to H2O in a two-step pathway. The electrochemical studies confirmed that incorporation of nitrogen within the CNT structure enhanced dramatically the catalytic ORR activity compared with commercial, nitrogenfree CNTs even in acidic electrolyte medium. In particular, NCNT-L showed a 100-fold increase in the kinetic current as derived from the Tafel plots. This increase in kinetic current can be attributed to the increase in stability of the carbon structure against oxidation and to an increase in the basic nature of the graphite structure due to incorporation of nitrogen. The most active nanostructures contain nitrogen and have a higher edge plane exposure. The selectivity of the catalysts for complete reduction of O2 to H2O follows a similar trend. 5. Conclusions Nitrogen was incorporated in carbon nanotubes via pyrolysis of acetonitrile at different temperatures. The NCNTs prepared at lower temperature had a higher amount of pyridinic groups with a high edge plane exposure as shown by the XPS and

J. Phys. Chem. C, Vol. 113, No. 32, 2009 14309 Raman measurements. Different nitrogen functional groups were found to be more thermally stable than oxygen functional groups, and almost no transformation between these nitrogen species was observed even after heating to 720 °C. The high activity of NCNTs containing only negligible traces of embedded cobalt suggests that the metal is not required for the ORR in acidic electrolytes. RDE and SECM studies demonstrated that the NCNTs containing a higher amount of pyridinic groups had the highest catalytic activity and selectivity in the ORR with a kinetic current, which was only 1 order of magnitude lower than that of a commercial Pt/C sample with 20 wt % Pt loading. Acknowledgment. S.K. thanks the International Max Planck Research School Surface and Interface Engineering in AdVanced Materials (SurMat) for a research grant. The authors acknowledge the financial support by the MIWFT-NRW in the framework of “Nachwuchsgruppe: Mikroelektrochemie zur Optimierung von Heterogenkatalysatoren fu¨r PEM-Brennstoffzellen”. Supporting Information Available: An extended analytical characterization was performed to investigate the electrocatalytic activity of the NCNTs. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Gasteiger, H. A.; Kocha, S. S.; Sompalli, B.; Wagner, F. T. Appl. Catal. B 2005, 56, 9. (2) Wang, B. J. Power Sources 2005, 152, 1. (3) Fierro, C.; Anderson, A. B.; Scherson, D. A. J. Phys. Chem. A 1988, 92, 6902. (4) Jahnke, H.; Schonborn, M.; Zimmerman, G. Top. Curr. Chem. 1976, 61, 133. (5) Ohms, D.; Herzog, S.; Franke, R.; Neumann, V.; Wiesener, K. J. Power Sources 1992, 38, 327. (6) Zagal, J. H.; Paez, M. A.; Silva, J. F. In N4-Macrocyclic Metal Complexes; Zagal, J. H., Bedioui, F.; Dodelet, J.-P., Eds.; Springer Science+Business Media, Inc.: New York, 2006; pp 44-82. (7) Liu, H.; Song, C.; Tang, Y; Liu, H.; Zhang, J. Electrochim. Acta 2007, 52, 4532. (8) Bae, I. T.; Tryk, D. A.; Scherson, D. A. J. Phys. Chem. B 1998, 102, 4114. (9) Bouwkamp-Wijnoltz, A. L.; Visscher, W.; Van Veen, J. A. R. Electrochim. Acta 1998, 43, 3141. (10) Faubert, G.; Lalande, G.; Cote, R.; Guay, D.; Dodelet, J.-P.; Wenig, L. T.; Bertrand, P.; Denes, G. Electrochim. Acta 1996, 41, 1689. (11) Van Veen, J. A. R.; Colijn, H. A.; Van Baar, J. F. Electrochim. Acta 1988, 33, 801. (12) Widelo¨v, A. Electrochim. Acta 1993, 38, 2493. (13) Okunola, A.; Kowalewska, B.; Bron, M.; Kulesza, P. J.; Schuhmann, W. Electrochim. Acta 2009, 54, 1954. (14) Dodelet, J.-P. In N4-Macrocyclic Metal Complexes; Zagal, J. H.; Bedioui, F.; Dodelet J.-P., Eds.; Springer Science+Business Media, Inc.: New York, 2006; pp 83-147. (15) Lefe`vre, M.; Dodelet, J.-P.; Bertrand, P. J. Phys. Chem. B 2002, 106, 8705. (16) Wang, H.; Cote, R.; Faubert, G.; Guay, D.; Dodelet, J.-P. J. Phys. Chem. B 1999, 103, 2042. (17) Bron, M.; Fiechter, S.; Hilgendorff, M.; Bogdanoff, P. J. Appl. Electrochem. 2002, 32, 211. (18) Bron, M.; Radnik, J.; Fieber-Erdmann, M.; Bogdanoff, P.; Fiechter, S. J. Electroanal. Chem. 2002, 535, 113. (19) Bron, M.; Fiechter, S.; Bogdanoff, P.; Tributsch, H. Fuel Cells 2002, 2, 137. (20) Matter, P. H.; Wang, E.; Arias, M.; Biddinger, E. J.; Ozkan, U. S. J. Phys. Chem B 2006, 110, 18374. (21) Yeager, E. Electrochim. Acta 1984, 29, 1527. (22) Matter, P. H.; Zhang, L.; Ozkan, U. S. J. Catal. 2006, 239, 83. (23) Matter, P. H.; Wang, E.; Arias, M.; Millet, J.-M. M.; Ozkan, U. S. J. Phys. Chem. C 2007, 111, 1444. (24) Maldonado, S.; Stevenson, K. J. J. Phys. Chem. B 2005, 109, 4707. (25) Trasobares, S.; Stephan, O.; Colliex, C.; Hsu, W. K.; Kroto, H. W.; Walton, D. R. M. J. Chem. Phys. 2002, 116, 8966. (26) Strelko, V. V.; Kuts, V. S.; Thrower, P. A. Carbon 2000, 38, 1499.

14310

J. Phys. Chem. C, Vol. 113, No. 32, 2009

(27) Strelko, V. V.; Kartel, N. T.; Dukhno, I. N.; Kuts, V. S.; Clarkson, R. B.; Odintsov, B. M. Surf. Sci. 2004, 548, 281. (28) Ozakiv, J. I.; Tanifuji, S. I.; Kimura, N.; Furuichi, A.; Oya, A. Carbon 2006, 44, 1324. (29) Niwa, H.; Horiba, K.; Harada, Y.; Oshima, M.; Ikeda, T.; Terakura, K.; Ozaki, J.; Miyata, S. J. Power Sources 2009, 187, 93. (30) Roy, S. S.; Papakonstantinou, P.; Okpalugo, T. I. T.; Murphy, H. J. Appl. Phys. 2006, 100, 053703. (31) Stanczyk, K.; Dziembaj, R.; Piwowarska, Z.; Witkowski, S. Carbon 1995, 33, 1383. (32) Dommele, S. V.; Romero-Izquirdo, A.; Brydson, R.; de Jong, K. P.; Bitter, J. H. Carbon 2008, 46, 138. (33) Eckhard, K.; Chen, X.; Turcu, F.; Schuhmann, W. Phys. Chem. Chem. Phys. 2006, 8, 5359. (34) Kranz, C.; Ludwig, M.; Gaub, H. E.; Schuhmann, W. AdV. Mater. 1995, 7, 568. (35) Kundu, S.; Wang, Y.; Xia, W.; Muhler, M. J. Phys. Chem. C 2008, 112, 16869. (36) Bhattacharyya, S.; Cardinaud, C.; Turban, G. J. Appl. Phys. 1998, 83, 4491. (37) Ruby, C.; Zhou, J. N.; Du, J.; Street, S. C.; Barnard, J. Surf. Interface Anal. 2000, 29, 38. (38) Kapteijn, F.; Moulijn, J. A.; Matzner, S.; Boehm, H. P. Carbon 1999, 37, 1143. (39) Pietrzak, R.; Wachowska, H.; Nowicki, P. Energy Fuels 2006, 20, 1275.

Kundu et al. (40) Singh, A.; Derry, T. E.; Luyckx, S. B.; Sellschop, J. P. F. J. Mater. Sci. Lett. 1990, 9, 1101. (41) Murphy, H.; Papakonstantinou, P.; Okpalugo, T. I. T. J. Vac. Sci. Technol. 2006, 24, 715. (42) Tuinstra, F.; Koenig, J. L. J. Chem. Phys. 1970, 53, 1126. (43) Tuinstra, F.; Koenig, J. L. J. Compos. Mater. 1970, 4, 492. (44) Cuesta, A.; Dhamelincourt, P.; Laureyns, J.; Martinez-Alonso, A.; Tascon, J. M. D. Carbon 1994, 32, 1523. (45) Cuesta, A.; Dhamelincourt, P.; Laureyns, J.; Martinez-Alonso, A.; Tascon, J. M. D. J. Mater. Chem. 1998, 8, 2875. (46) Sjostrom, H.; Stafstrom, S.; Boman, M.; Sundgren, J. E. Phys. ReV. Lett. 1995, 75, 1336. (47) Pleskov, V.; Filinovsky, V. Y.; Wroblowa, H. V.; Conway, B. E. The Rotating Disk Electrode; New York Consultants Bureau: New York, 1976. (48) Hsueh, K.; Gonzalez, E.; Srinivasan, S. Electrochim. Acta 1983, 28, 691. (49) Sidik, R. A.; Anderson, A. B.; Subramanian, N. P.; Kumaraguru, S. P.; Popov, B. N. J. Phys. Chem. B 2006, 110, 1787. (50) Jaouen, F.; Lefe`vre, M.; Dodelet, J.-P.; Cai, M. J. Phys. Chem. B 2006, 110, 5553. (51) Arrigo, R.; Havecker, M.; Schlo¨gl, R.; Su, D. S. Chem. Commun. 2008, 40, 4891.

JP811320D