Tuning Carbon-Based Fuel Cell Catalyst Support Structures via

ARTICLE pubs.acs.org/JPCC

Tuning Carbon-Based Fuel Cell Catalyst Support Structures via Nitrogen Functionalization. II. Investigation of Durability of PtRu Nanoparticles Supported on Highly Oriented Pyrolytic Graphite Model Catalyst Supports As a Function of Nitrogen Implantation Dose Svitlana Pylypenko,†,‡ Aimee Queen,† Tim S. Olson,‡ Arrelaine Dameron,‡ Kevin O’Neill,‡ K.C. Neyerlin,‡ Bryan Pivovar,‡ Huyen N. Dinh,‡ David S. Ginley,‡ Thomas Gennett,‡ and Ryan O’Hayre*,† †

Department of Metallurgical & Materials Engineering, Colorado School of Mines, 1500 Illinois Street, Golden, Colorado 80401, United States ‡ National Renewable Energy Laboratory, 1617 Cole Boulevard, Golden, Colorado 80401, United States

bS Supporting Information ABSTRACT: Nitrogen functionalization of carbon-based support materials for lowtemperature fuel cell catalysts has been shown to improve catalystsupport interactions and therefore enhance both the performance and durability of the supported electrocatalyst. While previous work has focused on pure Pt electrocatalysts, this work focuses on understanding the role of nitrogen functionalization on the durability of PtRu alloy nanoparticle catalysts. A well-defined model catalyst system approach is used by employing highly oriented pyrolytic graphite (HOPG) as a model graphitic carbon support, nitrogen ion beam implantation as the doping route, and magnetron sputtering from a single alloyed PtRu target for nanoparticle catalyst deposition. A series of PtRu/HOPG substrates with different levels of doped nitrogen but very similar initial PtRu nanoparticle coverage, size, and composition were evaluated using TEM and XPS before and after potential cycling. As compared to an undoped support, supports doped with low nitrogen dosage levels (both oxygen and nitrogen sites are present, oxygen sites are predominant, and nitrogen concentration is relatively low) appear to have a negative effect on PtRu nanoparticle stability. However, higher nitrogen doses (both oxygen and nitrogen sites are present, nitrogen concentration is relatively high) have a positive effect on durability, reaching an optimum with an implantation dosage of 4.7 1016 ions cm2. The improvement in durability can be directly related to nitrogen levels and specifically to the amount of pyridinic nitrogen. It is shown that strong positive tethering effects are related to formation of clustered multinitrogen defects, i.e., pyridinic rings in which more than one carbon is replaced with nitrogen, a condition that is met only when the nitrogen dosage is sufficiently high.

’ INTRODUCTION Degradation of fuel cell catalysts and supports has been extensively studied using a variety of techniques including electrochemical characterization, fuel cell testing, and spectroscopic and microscopic techniques;15 however, most studies have focused on hydrogen (rather than methanol) fuel cells. It has been shown that not only is long-term stability affected by weak bonding between catalyst and support but also weak interactions are responsible for low catalyst utilization resulting from detachment of catalyst nanoparticles from the carbon surface.5 Loss of active catalytic surface area has been attributed to a number of processes.68 Carbon corrosion is known to lead to detachment and agglomeration of nanoparticles. Beyond losses associated with carbon deterioration, electrochemically active surface area can be lost either through dissolutionredeposition or migration processes. Mechanistic studies of electrocatalyst degradation in the direct methanol fuel cell (DMFC) are more limited, especially studies involving performance degradation of PtRu electrocatalysts used r 2011 American Chemical Society

in the anodic electrooxidation of methanol. Nevertheless, reports have begun to examine issues associated with ruthenium crossover, methanol crossover, and crystallite growth.3,911 Ruthenium crossover has been observed under a variety of operating conditions and has been shown to depend on the anode potential and operating time of the DMFC.11 Even preleached anodes, where most unstable forms of ruthenium are depleted, have shown to further release ruthenium species that cross the membrane and contaminate the cathode active sites. The most defining results about the stability of PtRu alloys have been obtained from voltammetric cycling and analysis of the performance losses and corresponding structural and compositional changes in the catalyst in the MEA configuration. Improvements in catalystsupport interactions can potentially mitigate processes that lead to electrocatalyst degradation. Received: December 23, 2010 Revised: April 6, 2011 Published: May 17, 2011 13676

dx.doi.org/10.1021/jp112236n | J. Phys. Chem. C 2011, 115, 13676–13684

The Journal of Physical Chemistry C Functionalization of support materials via doping with heteroatoms, such as nitrogen, boron, etc., appears to be a highly promising route.1228 The benefits of nitrogen-doped carbon-based materials are presented in Part I (10.1021/jp1122344) of this work. In summary, it is believed that introduction of nitrogen into the carbon support network can potentially increase utilization of noble-phase metal through improvements in dispersion, catalytic activity, and durability. Our previous work demonstrated these beneficial effects for Pt/N-doped highly oriented pyrolytic graphite (HOPG) model catalyst systems.12,29 The focus of our more recent work is to study the role of nitrogen on the durability of PtRu nanoparticle catalysts supported on model HOPG substrates.30 As shown in the scope of Part I (10.1021/jp1122344), model HOPG substrates were used to establish the nature of the modification of well-defined graphitic surfaces during implantation with nitrogen. We have shown that complex changes including formation of surface defects, incorporation of nitrogen functionalities, and surface oxidation are dose dependent. We have also identified that ion doses greater than 2.5 1016 ions cm2 result in saturation of the surface layers with implanted/incorporated nitrogen atoms. In Part II of this work, we examine the role of the nitrogen doping level on the stability of model magnetron-sputtered PtRu nanoparticle catalysts supported on HOPG substrates. Magnetron sputtering of PtRu from a single target enables tunable deposition of catalyst nanoparticles of well-controlled composition. Importantly, this synthesis route allows for deposition of a metal phase without preferential nucleation at defect sites, in contrast to the situation associated with most aqueous or electrochemical deposition methods. This approach provides an excellent pathway to study the role of the concentration and prevalence of specific surface groups on the stability of the metal nanophase without biasing the interpretation by preferential nucleation at defect sites. This study also extends the examination of N-doping effects from previously investigated pure Pt catalyst systems to PtRu catalyst systems which are more relevant to DMFC technology. Detailed analysis of the structural and chemical modification of HOPG as a function of nitrogen ion implantation conditions was first introduced in Part I (10.1021/jp1122344) of this study. Here, we discuss the composition and structure of a series of HOPG substrates modified by N implantation after deposition of the metal catalyst nanoparticles. Then, the stability of these model PtRu nanoparticle catalysts is studied by electrochemical cycling in a 1 M sulfuric acid solution (in the potential range between 0 and 1.1 V vs Ag/AgCl). X-ray photoelectron spectroscopy (XPS) is used to follow the compositional changes of the metal phase as well as changes in the nitrogen speciation of the modified HOPG support as a function of electrochemical cycling. Transmission electron microscopy (TEM) evaluation of the samples, conducted before and after potential cycling, enables a detailed analysis of the PtRu nanoparticle coverage and size distribution. Finally, a direct correlation between nitrogen content/speciation and the stability of the catalyst nanoparticles is presented and discussed.

’ EXPERIMENTAL SECTION Details regarding the HOPG substrates and ion implantation conditions were presented in Part I (10.1021/jp1122344). Selected samples from the Part I study were used as substrates in Part II. These include undoped HOPG and HOPG after the following nitrogen implantation doses (implantation times): 0.4 1016, 0.8 1016, 1.3 1016, 4.7 1016, and 9.6 1016 ions cm2 (5, 10, 15, 45, and 100 s).

ARTICLE

Pt1xRux nanoparticles were sputtered onto the undoped HOPG and N-doped HOPG substrates from a single Pt0.5Ru0.5 (ACI Alloys) alloyed target with a 2” Onyx magnetron sputter gun (Angstrom Sciences). Direct current (DC) magnetron sputtering was performed using an MDX 1.5 kW DC power supply (Advanced Energy). The samples were positioned at a distance of 1.5” directly above the target, and the deposition was done after establishing a base pressure below 4 106 Torr. Nanoparticles were deposited by sputtering for 5 s in an inert Argon environment at a constant power of 20 W and a constant pressure of 20 mTorr Ar. Material flux was limited to very short time frames (few seconds) by the use of a shutter between the sample and source. XPS analysis was performed on a Kratos Nova X-ray photoelectron spectrometer with a monochromatic Al K source operated at 300 W. Survey spectra were acquired at pass energies of 160 eV with high-resolution spectra of C 1s/Ru 3d, O 1s, N 1s, Ru 3p, and Pt 4f at 20 eV. XPS data were acquired for a minimum of three areas per sample. Data analysis was performed using CasaXPS software. A linear background was applied to O 1s and N 1s spectra, while Shirley background was used for other regions. The graphitic carbon peak at 284.6 eV was used for charge referencing. For HOPG substrates with deposited PtRu nanophase, the binding energy range for C 1s was extended to accommodate the Ru 3d peaks. Due to interference of these two lines, special attention was paid to the curve-fitting procedure (additional information can be found in Supporting Information).31 TEM micrographs of PtRu nanoparticles supported on HOPG substrates were obtained on a Philips CM200 Transmission Electron Microscope. TEM specimen preparation consisted of peeling off a thin top layer of HOPG graphitic sheets and positioning it between the two grids of a Cu double grid (Electron Microscopy Sciences). Evaluation of the coverage and particle size analysis was done using ImageJ.32 The coverage was determined by first using the threshold filter to separate the background HOPG from the PtRu particles. Once separated, the “analyze particles” function was used to calculate the area of the image covered by PtRu, thus leading to the area fraction. This method also quantified the areas of each individual particle which were then used to determine the average particle size (area). It is important to note that this procedure resulted in the analysis of agglomerations as well as single particles. To determine the size of only single nanoparticles, identification of spherical nanoparticles was completed manually. Depending on the image quality and scale, the areas of all or a majority of the single nanoparticles in each image were outlined and separated using the threshold filter which then allowed for use of the particle analysis function to calculate the area of each nanoparticle. The electrochemical durability cycling was done in an aqueous electrochemical cell using a three-electrode configuration. The surface-modified HOPG substrate comprised the working electrode, with the addition of a Ag/AgCl reference electrode and a Pt wire counter electrode. The experiments were conducted in 1.0 M H2SO4, and the potential was cycled from 0 to 1.1 V vs Ag/ AgCl at a scan rate of 250 mV/s for 300 cycles.

’ RESULTS AND DISCUSSION For comparison purposes, micrographs and graphs in Figures 14 show data for pre- and postcycled samples. Discussion of the results, 13677

dx.doi.org/10.1021/jp112236n |J. Phys. Chem. C 2011, 115, 13676–13684

The Journal of Physical Chemistry C

ARTICLE

Figure 1. TEM micrographs representative of sputtered PtRu deposited on HOPG substrates (1) precycled, demonstrating coverage and range of nanoparticle sizes, and (2) postcycled, demonstrating changes in the coverage, particle size, and formation of agglomerations. Shown for samples with dosages of (a) 0.0 1016, (b) 0.4 1016, (d) 1.3 1016, (e) 4.7 1016, and (f) 9.6 1016 ions cm2 (0, 5, 15, 45, and 100 s). (c) 0.8 1016 (10 s) is not shown here as the TEM image of a precycled sample is not available; a micrograph of this sample after 300 cycles can be found in Figure 2 in the Supporting Information. The Supporting Information also contains images at lower magnification showing coverage over larger areas.

however, starts with initial characterization of the precycled samples followed by the analysis and discussion of the postcycled samples.

Initial Characterization of PtRu Nanoparticles Sputtered on HOPG Substrates. Initial evaluation of the samples after 13678



ARTICLE

Figure 2. Particle size statistics for PtRu on HOPG: (1) distributions of nanoparticle diameters for single nanoparticles in precycled samples, (2) distributions of nanoparticle diameters for single nanoparticles in postcycled samples, (3) distributions of nanoparticle areas for all nanoparticles in postcycled samples, (4) mean values and standard deviations for data shown in (1) and (2), and (5) mean values and standard deviations for data shown in (3). Shown for samples with dosages of (a) 0.0 1016, (b) 0.4 1016, (c) 0.8 1016, (d) 1.3 1016, (e) 4.7 1016, and (f) 9.6 1016 ions cm2 (0, 5, 10, 15, 45, and 100 s). Number of particles: (1a) 99, (1b) 116, (1d) 132, (1e) 49, (1f) 198; (2a) 106, (2b) 40, (2c) 65, (2d) 123, (2e) 239, (2f) 182; (3a) 5306, (3b) 664, (3c) 994, (3d) 1727, (3e) 11,175, (3f) 8019. 13679



Figure 3. PtRu coverage as a function of implantation dosage for N-doped series before cycling and after 300 cycles. Average and standard deviation were calculated from a minimum of four areas per sample.

metal deposition consisted of TEM and XPS characterization. Figure 1(1a1f) provides representative TEM micrographs obtained after metal deposition. Visual analysis of TEM images of HOPG substrates with various implantation doses revealed even distribution of the metal phase on the surface of HOPG. High coverage with particle sizes ranging from 2 to 3 nm was obtained for all samples essentially independent of HOPG doping conditions. Sputter deposition is a line of sight technique, and therefore, the entire surface of HOPG is nucleated with PtRu catalyst nanoparticles. To provide a baseline for comparison of samples before and after potential cycling, particle size and coverage were further evaluated statistically and are reported in Figures 2 (1a1f) and 3. As a result of the highly reproducible PtRu sputtering conditions used for all samples, the coverage prior to cycling remains nearly constant regardless of the implantation dosage at 3440% (Figure 3). The individual particle areas determined from the particle size analysis in ImageJ were converted to equivalent circular diameters, di, using the following equation rffiffiffi 4 Ai di ¼ π The particle size distributions in Figure 2 (1a1f) suggest that the ion implantation dosage has at most a modest effect on the initial PtRu particle diameters. N modification resulted in a slightly narrower distribution of particle size than the undoped control. Lower N dosages tended toward larger particles, while the higher N dosages produced slightly smaller particles. Relative elemental concentrations measured for the series of HOPG substrates after metal deposition are reported in Table 1 of the Supporting Information. Due to challenges arising when fitting interfering lines, the concentration of carbon reported here may deviate from the real concentration by several percent. Though this affects the relative amounts of all elements, this deviation is expected to be fairly consistent from sample to sample allowing for reliable comparison within a series of samples. Metal deposition leads to attenuation (and hence a decrease) in the signal from the elements present in the underlying HOPG layers. Thus, the postmetalized nitrogen concentrations range from 3 to 5 atom %, which are below the levels reported in Part I (10.1021/jp1122344) for implanted samples before metal deposition. Nitrogen detected on undoped HOPG

ARTICLE

Figure 4. Relative amount of metallic platinum (Pt1), obtained from the quantification of Pt 4f spectra, shown as a function of implantation dosage for N-doped series before cycling and after 300 cycles. Relative amounts and standard deviations are reported for each sample as an average from at least three areas.

is probably a result of nitrogen chemisorption occurring upon exposure to ambient immediately after metal deposition. This is also consistent with the fact that HOPG after implantation with Arþ has also been found to contain small levels of nitrogen, comparable to the levels observed after sputtering. Because the range of PtRu nanoparticle sizes determined from TEM image analysis is smaller than the XPS analysis depth, the composition of the metal phase determined by XPS can be attributed to a true “bulk-averaged value” for the nanoparticles. Elemental concentrations range from 4.3 to 4.9 atom % for platinum and from 6.6 to 7.9 atom % for ruthenium, resulting in fairly consistent Pt/Ru ratios in the range of 0.590.67. Thus, all samples, despite being deposited from a single target with a Pt:Ru composition of 1:1, have a ruthenium-rich metal phase.33 Further assessment of the composition of the metal phase is achieved through curve-fitting of the Pt 4f and Ru 3d spectra (discussion and spectra are available in Supporting Information). The ratio of metallic (i.e., nonoxide) platinum to metallic ruthenium, after accounting for their elemental concentrations, is assessed at ∼0.9. This ratio is significantly higher than the overall elemental Pt:Ru ratio and is fairly close to that of state of the art PtRu electrocatalysts. XPS analysis indicates that approximately 40% of the PtRu is in the metallic state with an atomic ratio close to 1:1; another 20% is platinum in the oxide state; and the remaining 40% is ruthenium oxides. Structural Stability of PtRu Nanoparticles Sputtered on HOPG Substrates. Initial Evaluation. After potential cycling, all PtRu/HOPG substrates were reevaluated using TEM and XPS. Figure 1 (2a2f) shows that the coverage of the metal phase changes dramatically after cycling and that these changes vary greatly depending on the implantation dosage. Visual analysis of the images (Figure 1, and Figure 1 in Supporting Information) and estimated coverages (Figure 3) demonstrates that despite nearly constant initial coverage all samples show a significant decrease in catalyst coverage after cycling. The loss in coverage area is particularly severe for the low-nitrogen dose samples (0.4 10161.3 1016 ions cm2). For example, the sample dosed with 1.3 1016 ions cm2 retains less than 10% of its initial catalyst coverage after cycling. Intriguingly, the dosage of 4.7 1016 ions cm2 (45 s), previously used as an optimum 13680


The Journal of Physical Chemistry C implantation condition in prior Pt/N-HOPG studies,13 also appears to provide the optimum durability performance for this PtRu/N-HOPG study. This sample retains 40% of its original coverage after cycling. As was noted earlier, the sputter deposition method results in catalyst deposition without preference to nitrogen/defect sites, and therefore significant loss of the metal phase is expected for all the samples. The difference in the loss for various implantation doses however can be directly related to the amount and speciation of the nitrogen introduced during implantation, vide infra. The loss of metal phase during cycling is accompanied by a significant commensurate decrease in the XPS signal from both platinum and ruthenium (Table 2 in Supporting Information). Estimated concentrations of these elements after cycling are less than 1 atom % for each metal. For areas where a ruthenium signal was obtained, the postcycled spectrum is dominated by metallic ruthenium; contributions from ruthenium oxide were below detection limits. Curve-fitting of the Pt 4f spectra was done in a similar fashion as discussed previously, and the percent of metallic Pt detected after cycling is reported in Figure 4 against the trend observed for the same set of samples before cycling. While the overall decrease in platinum is due to the decrease in catalyst coverage, the fluctuation in the metallic vs oxide content is directly related to agglomeration (and hence catalyst nanoparticle surface/volume changes) during cycling (discussed later in the section on degradation processes). Careful examination of the TEM micrographs in Figure 1 reveals that in addition to coverage loss the spatial distribution and morphology of the metal phase also changes upon cycling. The samples can be broadly divided into two categories: (1) the undoped and low-dosage samples (ad) which show significant coverage loss and substantial morphological changes vs (2) the highly N-doped samples (2e2f) which show less coverage loss and minimal morphological changes. As can be readily observed, it is clear that only the higher levels of nitrogen implantation (e.g., e,f) result in a significant positive durability effect. The potential underlying reasons for this behavior will be discussed in the paragraphs that follow. A detailed assessment of changes in the size and morphology of the metal phase for all samples was conducted using a collection of TEM images and corresponding particle distribution analyses shown in Figure 2. For the postcycled samples, the first set of particle distribution statistics (middle column, 2a2f) provides the distribution of equivalent spherical particle diameters in nanometers and only includes the analysis of isolated single spherical particles (i.e., agglomerated particles are not included in these distributions). The second set (far right column, 3a3f, summarized in Figure 2(5)) provides the distribution of all particle areas in nanometers squared. On the basis of these analyses (statistics obtained from a minimum of four images per sample), the undoped and low-dosage samples (ac) show broad particle area distributions with a significant tail toward higher values, indicative of particle agglomeration. With further increase in the implantation dosage the distribution narrows, and although some agglomerations are still observed for the sample with the dose of 1.3 1016 ions cm2 (d), the samples with higher N-dosages (ef) show remarkably consistent, agglomerate-free particle coverages. Degradation Processes. There are several fundamental processes frequently cited in the fuel cell literature that could contribute to the changes in the particle size distributions shown above: (1) coarsening based on Oswald ripening; (2)

ARTICLE

Figure 5. Representative SEM image of PtRu coverage (including geometrically shaped PtRu precipitates) on a N-doped HOPG substrate after cycling.

coalescence via migration, and (3) agglomeration induced by corrosion of carbon support.6,8 Both Oswald ripening (which occurs through atomic-scale dissolution and redeposition) and migration/coalescence (which occurs by the macroscopic movement of nanoparticles on the carbon support) are known to lead to growth of the mean particle size and changes in the particle size distribution. Widening of the particle size distribution and the formation of a tail toward larger particle sizes is expected for the migration/coalescence mechanism. Oswald ripening should result in widening of the distribution of large particles and a narrowing of the distribution of small particle sizes. Platinum dissolution has been identified as one of the major reasons for the loss of active surface area for electrodes exposed to potentials higher than 0.8 V vs RHE.8 It has been shown that the rate of dissolution increases rapidly for particles less than 3 nm in size. This behavior is also well supported by the observation that particles with sizes greater than 5 nm survive high potentials, while particles less than 2 nm in size dissolve almost immediately. In addition, it has been shown that a large number of particles in typical fuel cell catalyst layers are not sufficiently bonded to the carbon support and thus show increased mobility that allows for rapid coalescence during potential cycling.34 For the undoped and low-dosage samples examined in this study (Figure 1 and 2(a)(c)), the clearly observed disappearance of small nanoparticles and a shift in the distribution toward larger nanoparticle sizes could be typical of Oswald ripening or migration/coalescence processes. The agglomeration seen on these samples is also consistent with the relative increase in the metallic platinum XPS signal since the decreasing surface/volume ratio associated with the larger nanoparticle agglomerates results in an increase in the bulk metallic Pt vs surface Pt-oxide character of the catalyst. The formation of clear particle necking/sintering structures in the cycled samples likely indicates that a combination of particle migration and dissolution/reprecipitation is occurring in these samples. Larger area evaluation of the postcycled HOPG substrate after cycling provides further evidence for metal dissolution processes. As shown in Figure 5 (a representative SEM image of a postcycled sample), in addition to the small-scale PtRu agglomerates previously observed in the high-magnification TEM investigations, additional large-size (∼50 nm) 13681



ARTICLE

Table 1. Quantification of the high-resolution N1s XPS spectra for a series of N-implanted HOPG substrates after 300 cycles. Representative fitted N 1s spectra can be found in the Supplementary section ID

pyridinic

amine, pyrrolic

pyrrolic

pyrrolic, graphytic, queternary

graphytic, quaternary, N-oxide

BE, eV

398398.9

399.5

400.1

400.9

401.6403

dosage, ions cm2

Relative Amount, %

Total, %

16

0.0 10

0.0

11.5

23.5

18.3

46.6

100

0.4 1016

7.1

12.7

24.1

21.5

34.6

100

0.8 1016

8.5

18.1

25.1

17.0

31.4

100

1.3 1016

14.5

20.1

22.2

15.7

27.5

100

4.7 1016

15.0

18.9

27.9

17.5

20.7

100

9.6 1016

12.4

16.8

23.8

18.3

28.6

100

Normalized to Total Nitrogen, atom %

Total, atom %

0.0 1016

0.00

0.18

0.36

0.28

0.72

1.6

0.4 1016

0.22

0.40

0.76

0.68

1.09

3.2

0.8 1016

0.36

0.76

1.05

0.71

1.32

4.2

1.3 1016

0.53

0.74

0.81

0.58

1.01

3.7

4.7 1016

0.63

0.80

1.17

0.74

0.87

4.2

9.6 1016

0.62

0.83

1.18

0.91

1.41

4.9

geometrically shaped noble metal particles are also observed after cycling. These larger-scale particles are clear evidence for the occurrence of dissolution/reprecipitation processes during cycling; similar shapes have been reported by other researchers examining the membrane interface of cycled fuel cell MEAs.7 For the samples with higher implantation doses (beginning with (d), then more dramatically with (e)(f)), the particle distributions narrow significantly and shift to lower mean sizes. Importantly, for all three of these samples, the average nanoparticle size decreases after cycling (compare Figure 2 (1d1f vs 2d2f, summarized in Figure 2(4)). This behavior, particularly in the presence of large-size (∼50 nm) geometrically shaped metal particles indicates that metal-phase dissolution is likely still occurring in these samples, albeit in the absence of significant agglomerate formation. Because agglomeration is not observed, it is likely that particle migration is minimized in these samples— perhaps indicating a stabilizing influence due to the increased nitrogen functionalization associated with the higher nitrogen implantation dosage. Considering the high potential limit during voltage cycling, carbon corrosion processes potentially add another level of complexity in interpretation of the results. It is possible that improved resistance to agglomeration is also related to improved corrosion resistance, but currently, we are lacking evidence either for or against the benefits of nitrogen doping on carbon corrosion and will try to address this question in our future work. Roughness of the substrates is another factor that, to a certain degree, can affect both migration and corrosion processes. Increased roughness of HOPG substrates irradiated with Nþ or Arþ perhaps could potentially mitigate

migration/coalescence processes but would also result in higher corrosion currents.13 Several studies16,35 have suggested that defect sites on the graphitic carbon support, step edges, in particular, can serve as trapping sites, thereby decreasing nanoparticle migration. DFT studies also suggest a binding or tethering influence between CN functional groups (such as pyridinic defects) and Pt nanoparticles.16,35 Such suggestions are consistent with the observations in this study, which reveal a trend toward decreased nanoparticle migration with increasing HOPG surface modification. The cycling potential used in this study can also result in accelerated corrosion of the carbon support, which can in turn lead to nanoparticle agglomeration.3639 It is well documented that carbon supports with higher graphitic content are more corrosion resistant, while excess oxygen groups can lead to an acceleration of the carbon support corrosion rate.40 Comparison of the elemental content before and after potential cycling shows an increased amount of oxygen for all samples (Table 2, Supporting Information), but the oxygen content is especially high for the low-dosage samples, indicating that these samples may be most susceptible to carbon corrosion. As discussed in Part I (10.1021/jp1122344) of this study, these low-dose samples sustain significant implant damage but only a small amount of nitrogen functionalization—most of the implant-induced defects are oxidized. Indeed, such low nitrogen dose samples are chemically similar to samples exposed to Arþ irradiation,13 where implantation induces both physical damage and oxygen functionalization but does not provide for significant nitrogen functionalization. Both experimental and modeling studies of 13682


The Journal of Physical Chemistry C Arþ HOPG have previously shown that the physical defects (such as vacancies) and oxygen sites associated with Arþ implantation result in enhanced particle ripening and poor durability. Thus, low nitrogen implantation dosages for the HOPG model system appear to have an overall negative effect on catalyst durability because the small degree of nitrogen functionalization associated with these low dosages is outweighed by the loss in stability and carbon corrosion issues introduced by implant-induced physical damage and oxygen functionalization. At higher nitrogen implantation dosages, both the amount and speciation of the nitrogen functionalization change appreciably (c.f. Table 1, obtained via N1s XPS spectral deconvolution as described in Part I (10.1021/jp1122344) of this study; for spectra see Supporting Information Figure 4), and this is accompanied by a significant improvement in catalyst durability. This is especially apparent by comparing the cycled N1s XPS data (Table 1) for the three samples with the lowest nitrogen implantation dosages (0, 0.42 1016, and 0.7 1016, respectively, corresponding to Figures 1(a)(c)) to the three samples with the highest nitrogen implantation dosages (1.3 1016, 4.7 1016, and 9.6 1016 ions cm2, respectively, corresponding to Figures 1(d)(f)). While only small variation in the amount of pyrrolic N is detected between sample groups, there is a significant difference in the amount of pyridinic N. Low dosage samples have a high percentage of graphitic/quaternary N and a low percentage of pyridinic N. As the nitrogen implantation dosage increases, the relative percentage of pyridinic nitrogen increases at the expense of the graphitic/quaternary nitrogen content. This trend is consistent with the increase in the overall total amount of nitrogen, as multinitrogen defects become more probable. Indeed, the sample with the highest amount of pyridinic N (implant dosage of 4.7 1016 ions cm2) also showed the best durability, as evident from morphological and structural analysis. It is interesting to note that a further increase in the nitrogen dose to 9.6 1016 ions cm2 increased the amount of graphitic/quaternary N but did not further increase the amount of pyridinic N, and the tendency toward agglomerations in the last two samples was very similar.

’ SUMMARY In summary, we have utilized a model catalyst system to explore the effect of nitrogen doping on the durability of PtRu nanoparticle catalysts. Magnetron sputtering of the PtRu enabled deposition of the metal nanophase without preference to defect sites or nitrogen functionalities, in contrast to electrodepostion routes, which result in preferential deposition on defects/nitrogen sites. The sputtering conditions used here resulted in high metal coverages in the range of 3440% and particle sizes in the range of 23 nm. While all samples were somewhat ruthenium-rich, the behavior and morphology were consistent across the series of samples independent of the nitrogen implantation dose. After potential cycling, the coverage and morphology of the metal phase changed dramatically and were shown to be greatly dependent on the implantation dosage. The undoped (only carbon sites) and low-dosage samples (oxygen and nitrogen sites are present, nitrogen concentration is low) showed significant coverage loss and substantial morphological changes and had an overall negative effect on PtRu nanoparticle stability. Highly N-doped samples, on the other hand (oxygen and nitrogen sites are present, nitrogen

ARTICLE

concentration is high), showed less coverage loss and minimal morphological changes. The optimum stability of the PtRu nanoparticles was obtained with an implantation dosage of 4.7 1016 ions cm2. Formation of agglomerations, with a consequent widening of the particle size distribution and the formation of a tail toward larger particles typical of migration/coalescence, was only observed for the undoped and low dosage samples. Samples with higher doses appeared to have significantly improved stability due to mitigation of the migration/coalescence process. Such improved stability is most likely associated with a higher amount of nitrogen (at saturation level), specifically pyridinic nitrogen, and also a higher degree of clustering of the multinitrogen defects.

’ ASSOCIATED CONTENT

bS

Supporting Information. TEM micrograph demonstrating PtRu nanoparticles in the postcycled sample with dosage of 0.77 1016 ions cm2. Lower magnification TEM micrographs showing a representative of precycled samples, demonstrating remarkable initial coverage, and postcycled samples demonstrating dose-dependent changes in the coverage. XPS analysis of platinum and ruthenium including experimental details, highresolution Pt 4f and Ru 3d þ C 1s, and discussion of their fits. Elemental composition of PtRu/HOPG before and after potential cycling, and XPS N 1s spectra representative of low and high dosage postcycled samples. This material is available free of charge via the Internet at http://pubs.acs.org.

’ AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected].

’ ACKNOWLEDGMENT This work was supported by the Army Research Office under grant #W911NF-09-1-0528 and the U.S. Department of Energy EERE, FCT Program, under Contract No. DE-AC36-08GO28308 with the National Renewable Energy Laboratory. The authors acknowledge G. Zito, John Chandler, and Electron Microscopy Laboratory at CSM for assistance with SEM and TEM analysis and Surface Analysis group at NREL for support and access to Kratos NOVA XPS. ’ REFERENCES (1) Arico, A. S.; Creti, P.; Poltarzewski, Z.; Mantegna, R.; Kim, H.; Giordano, N.; Antonucci, V. Mater. Chem. Phys. 1997, 47, 257. (2) Ascarelli, P.; Contini, V.; Giorgi, R. J. Appl. Phys. 2002, 91, 4556. (3) Shyam, B.; Arruda, T. M.; Mukerjee, S.; Ramaker, D. E. J. Phys. Chem. C 2009, 113, 19713. (4) Wilson, M. S.; Garzon, F. H.; Sickafus, K. E.; Gottesfeld, S. J. Electrochem. Soc. 1993, 140, 2872. (5) Xie, J.; Wood, D. L.; More, K. L.; Atanassov, P.; Borup, R. L. J. Electrochem. Soc. 2005, 152, A1011. (6) Borup, R.; Meyers, J.; Pivovar, B.; Kim, Y. S.; Mukundan, R.; Garland, N.; Myers, D.; Wilson, M.; Garzon, F.; Wood, D.; Zelenay, P.; More, K.; Stroh, K.; Zawodzinski, T.; Boncella, J.; McGrath, J. E.; Inaba, M.; Miyatake, K.; Hori, M.; Ota, K.; Ogumi, Z.; Miyata, S.; Nishikata, A.; Siroma, Z.; Uchimoto, Y.; Yasuda, K.; Kimijima, K. I.; Iwashita, N. Chem. Rev. 2007, 107, 3904. (7) Ferreira, P. J.; la O’, G. J.; Shao-Horn, Y.; Morgan, D.; Makharia, R.; Kocha, S.; Gasteiger, H. A. J. Electrochem. Soc. 2005, 152, A2256. 13683



ARTICLE

(8) Shao-Horn, Y.; Sheng, W. C.; Chen, S.; Ferreira, P. J.; Holby, E. F.; Morgan, D. Top. Catal. 2007, 46, 285. (9) Cheng, T. T. H.; Jia, N. Y.; He, P. J. Electrochem. Soc. 2010, 157, B714. (10) Park, G. S.; Pak, C.; Chung, Y. S.; Kim, J. R.; Jeon, W. S.; Lee, Y. H.; Kim, K.; Chang, H.; Seung, D. J. Power Sources 2008, 176, 484. (11) Piela, P.; Eickes, C.; Brosha, E.; Garzon, F.; Zelenay, P. J. Electrochem. Soc. 2004, 151, A2053. (12) Zhou, Y. K.; Neyerlin, K.; Olson, T. S.; Pylypenko, S.; Bult, J.; Dinh, H. N.; Gennett, T.; Shao, Z. P.; O’Hayre, R. Energy Environ. Sci. 2010, 3, 1437. (13) Zhou, Y. K.; Pasquarelli, R.; Holme, T.; Berry, J.; Ginley, D.; O’Hayre, R. J. Mater. Chem. 2009, 19, 7830. (14) Chen, Y. G.; Wang, J. J.; Liu, H.; Li, R. Y.; Sun, X. L.; Ye, S. Y.; Knights, S. Electrochem. Commun. 2009, 11, 2071. (15) Chetty, R.; Kundu, S.; Xia, W.; Bron, M.; Schuhmann, W.; Chirila, V.; Brandl, W.; Reinecke, T.; Muhler, M. Electrochim. Acta 2009, 54, 4208. (16) Groves, M. N.; Chan, A. S. W.; Malardier-Jugroot, C.; Jugroot, M. Chem. Phys. Lett. 2009, 481, 214. (17) Kundu, S.; Nagaiah, T. C.; Xia, W.; Wang, Y. M.; Van Dommele, S.; Bitter, J. H.; Santa, M.; Grundmeier, G.; Bron, M.; Schuhmann, W.; Muhler, M. J. Phys. Chem. C 2009, 113, 14302. (18) Maiyalagan, T. Appl. Catal., B 2008, 80, 286. (19) Maldonado, S.; Stevenson, K. J. J. Phys. Chem. B 2005, 109, 4707. (20) Ozaki, J.; Anahara, T.; Kimura, N.; Oya, A. Carbon 2006, 44, 3358. (21) Prehn, K.; Warburg, A.; Schilling, T.; Bron, M.; Schulte, K. Compos. Sci. Technol. 2009, 69, 1570. (22) Roy, S. C.; Christensen, P. A.; Hamnett, A.; Thomas, K. M.; Trapp, V. J. Electrochem. Soc. 1996, 143, 3073. (23) Shukla, A. K.; Ravikumar, M. K.; Roy, A.; Barman, S. R.; Sarma, D. D.; Arico, A. S.; Antonucci, V.; Pino, L.; Giordano, N. J. Electrochem. Soc. 1994, 141, 1517. (24) Sidik, R. A.; Anderson, A. B.; Subramanian, N. P.; Kumaraguru, S. P.; Popov, B. N. J. Phys. Chem. B 2006, 110, 1787. (25) Sun, C. L.; Chen, L. C.; Su, M. C.; Hong, L. S.; Chyan, O.; Hsu, C. Y.; Chen, K. H.; Chang, T. F.; Chang, L. Chem. Mater. 2005, 17, 3749. (26) Wu, G.; Li, D. Y.; Dai, C. S.; Wang, D. L.; Li, N. Langmuir 2008, 24, 3566. (27) Yu, X. W.; Ye, S. Y. J. Power Sources 2007, 172, 133. (28) Yu, X. W.; Ye, S. Y. J. Power Sources 2007, 172, 145. (29) Zhou, Y. K.; Holme, T.; Berry, J.; Ohno, T. R.; Ginley, D.; O’Hayre, R. J. Phys. Chem. C 2010, 114, 506. (30) Pylypenko, S.; Queen, A.; Neyerlin, K. C.; Olson, T.; Dameron, A.; O’Neill, K.; Ginley, D.; Gorman, B.; Kocha, S.; Dinh, H. N.; Gennett, T.; O’Hayre, R. ECS Trans. 33, 351. (31) Pylypenko, S.; Blizanac, B. B.; Olson, T. S.; Konopka, D.; Atanassov, P. ACS Appl. Mater. Interfaces 2009, 1, 604. (32) http://rsbweb.nih.gov/ij/. (33) Dameron, A. A. L., J. E.; Dinh H. N.; Pylypenko S.; Hurst K.; Gillaspie, D.; O’Hayre, R.; Ginley, D. S.; Gennett, T. Acs Appl. Mater. Interfaces 2010, submitted. (34) Borup, R. L.; Davey, J. R.; Garzon, F. H.; Wood, D. L.; Inbody, M. A. J. Power Sources 2006, 163, 76. (35) Holme, T.; Zhou, Y. K.; Pasquarelli, R.; O’Hayre, R. Phys. Chem. Chem. Phys. 2010, 12, 9461. (36) Avasarala, B.; Moore, R.; Haldar, P. Electrochim. Acta 2010, 55, 4765. (37) Choo, H. S.; Kinumoto, T.; Jeong, S. K.; Iriyama, Y.; Abe, T.; Ogumi, Z. J. Electrochem. Soc. 2007, 154, B1017. (38) Nose, M.; Kinumoto, T.; Choo, H. S.; Miyazaki, K.; Abe, T.; Ogumi, Z. Fuel Cells 2009, 9, 284. (39) Siroma, Z.; Ishii, K.; Yasuda, K.; Miyazaki, Y.; Inaba, M.; Tasaka, A. Electrochem. Commun. 2005, 7, 1153. (40) Kinoshita, K. Carbon: Electrochemical and Physicochemical Properties; Wiley: New York, 1988. 13684


Tuning Carbon-Based Fuel Cell Catalyst Support Structures via

Recommend Documents