Effect of metal particle size on sulfur tolerance of Ni catalysts during autothermal reforming of isooctane

Applied Catalysis A: General 400 (2011) 203–214

Contents lists available at ScienceDirect

Applied Catalysis A: General journal homepage: www.elsevier.com/locate/apcata

Effect of metal particle size on sulfur tolerance of Ni catalysts during autothermal reforming of isooctane Joseph M. Mayne, Kevin A. Dahlberg, Thomas A. Westrich, Andrew R. Tadd, Johannes W. Schwank ∗ Transportation Energy Center, Department of Chemical Engineering, University of Michigan, 2300 Hayward, Ann Arbor, MI 48109, USA

a r t i c l e

i n f o

Article history: Received 11 January 2011 Received in revised form 21 April 2011 Accepted 25 April 2011 Available online 30 April 2011 Keywords: Nickel Particle size Sulfur poisoning Autothermal reforming Isooctane Thiophene

a b s t r a c t This paper describes to what extent Ni particle size affects the sulfur-tolerance of ceria-zirconia supported Ni catalysts during autothermal reforming (ATR) of isooctane. Particle size was isolated as an experimental variable by preparing catalysts with a range of Ni loadings that had nearly identical Ni surface areas. Under sulfur-free conditions, isooctane conversion and synthesis gas yield increased as the Ni particle size increased, contrary to the expectation that smaller particle sizes with lower Ni coordination would be more active. However, larger Ni particles proved to be more vulnerable to sulfur poisoning. The poor ATR activity of small Ni particles can be attributed either to a lack of sufficiently large nickel surface ensembles, or to their higher propensity to form nickel oxides under reaction conditions. This contribution suggests that, under typical ATR conditions, more highly dispersed Ni catalysts will not result in elevated sulfur tolerance. © 2011 Elsevier B.V. All rights reserved.

1. Introduction The influence of metal particle size on catalyst performance has been demonstrated for a broad range of systems. In most cases, the goal of catalyst synthesis is to formulate a material with a stable particle size distribution skewed towards smaller particles [1]. Smaller particles maximize the number of surface metal atoms available for catalytic chemistry and additionally have active sites that have been shown to be chemically different compared to larger particles. Such chemical differences arise from an increased relative concentration of higher energy crystallographic planes in smaller particles. These surfaces feature a lower average coordination number and altered electronic structure [2,3], thus binding many adsorbing molecules more strongly. The effect of particle size was investigated when producing hydrogen from hydrocarbon fuels via autothermal reforming (ATR) over Ni-based catalysts. Ideally, this system couples oxidation and steam reforming pathways to produce an equilibrium-limited mixture of H2 , CO, CO2 , and CH4 [4–6]. This reformate has several applications, including power generation via solid oxide fuel cells and automotive emissions control [7,8]. In practice, the long-term stability and activity of noble- and non-noble-metal based catalysts

∗ Corresponding author. Tel.: +1 734 764 3374; fax: +1 734 763 0459. E-mail address: [email protected] (J.W. Schwank). 0926-860X/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.apcata.2011.04.039

are limited by side reactions that produce carbon deposits and poisoning of active sites by sulfur-containing compounds commonly found in hydrocarbon fuels [9–11]. Investigations into particle size effects have been applied previously to certain aspects of this chemistry. For example, it has been well demonstrated that steam reforming reactions preferentially occur at lower coordinated kink and step sites. By contrast, these reactions proceed at much slower rates on highly coordinated terrace sites [12–17]. In addition, there is evidence that carbon deposition rates are dependent on metal particle size [18–20]. However, it is often difficult to control particle size independently of other variables such as metal loading and total number of surface sites. Therefore, many previous studies have failed to differentiate or account for these confounding effects. This study attempts to isolate the effect of particle size on the sulfur tolerance of Ni-based catalysts during the ATR of isooctane by varying particle size independently of the total number of active surface sites. To accomplish this a novel experimental strategy was employed, whereby a series of catalysts with different weight loadings were selectively pretreated to yield samples with various particle sizes but similar numbers of active surface sites as indicated by H2 chemisorption measurements. The activities of these catalysts were then measured and compared for ATR of sulfur-free and thiophene-containing isooctane. These experiments clarify the role of particle size effects in the ATR system, and therefore provide guidance for the development of Ni catalysts that are more tolerant to sulfur exposure under ATR conditions.

204

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

Table 1 Summary of the hydrothermal reduction treatment used to obtain the pre-reaction catalysts, the pretreatment H2 chemisorption SA , the post-treatment SA , and the particle size estimations from H2 chemisorption, XRD and STEM/EDS. All catalysts were calcined at 500 ◦ C in air prior to any aging treatment. Ni loading (wt%)

Initial SA (m2 /g)

Age time (h)

Age temperature (◦ C)

Final SA (m2 /g)

H2 chemisorption dp (nm)

XRD Ni [1 1 1] dp (nm)

XRD Ni [2 0 0] dp (nm)

STEM/EDS dp (nm)

0.7 3 5 10

0.29 0.94 1.15 1.49

N/A 4 5 8

N/A 700 800 830

0.29 0.33 0.26 0.33

16.1 62.1 137.7 204.5

N/A 20.1 38.7 62.4

N/A N/A 22.7 45.1

N/A 18.9 26.7 51.3

2. Experimental 2.1. Catalyst preparation and hydrothermal treatment The catalysts used for this study were Ni supported on zirconium-modified ceria. The support, Ce0.75 Zr0.25 O2 (CZO), was prepared by coprecipitation of the precursor salts Ce(NO3 )3 ·6H2 O and ZrOCl2 ·8H2 O dissolved in 4 M NH4 OH. The precipitate was recovered by filtration and dried at 100 ◦ C. The dried material was ground and sieved to yield particles in the range of 250–420 ␮m. The support was then calcined in air at 900 ◦ C for 2 h. Ni was impregnated onto the support using the incipient wetness technique with the metal precursor Ni(NO3 )2 ·6H2 O dissolved in deionized water. Catalysts with four different nominal loadings of Ni (0.7 wt%, 3 wt%, 5 wt%, and 10 wt%) were prepared. Following impregnation, each catalyst was calcined in air at 500 ◦ C for 30 min in order to decompose the nickel precursor. Prior studies have shown that Ni particle size increases with increasing Ni loading [21–23], concurrently with an increase in active surface area (SA ). To separate these competing effects, increasingly severe hydrothermal treatments were performed to sinter the catalysts as Ni loading was increased, such that the final catalysts exhibited equivalent active Ni SA . The sintering behavior of Ni-based materials has been characterized previously by Sehested et al. [24–26] under simulated steam reforming conditions. They reported simple power law rate expressions and Arrhenius-type temperature dependence, and these results guided the hydrothermal treatment procedures applied in this study. Hydrothermal treatments were conducted in a 1/2 in. o.d. 316SS vessel. Prior to treatment the catalysts were reduced at 600 ◦ C for 30 min under flow of 5% H2 in N2 . Following reduction the catalysts were either heated or cooled to the desired treatment temperature in pure N2 . The catalysts were then exposed to a gas stream with a 19:10:1 ratio of N2 to H2 O to H2 . A needle valve was used to set the pressure in the vessel to approximately 20 psig. Various aging times and temperatures were considered for the 3%, 5%, and 10% Ni loading catalysts to characterize the influence of aging conditions on the SA of the catalyst. To match the SA of the 0.7% Ni catalyst, the 3% Ni sample was treated at 700 ◦ C for 4 h, the 5% sample at 800 ◦ C for 5 h, and the 10% sample at 830 ◦ C for 8 h. The catalysts used for reactor studies are summarized in Table 1. The synthesis and subsequent aging procedures were repeated to create a second batch of catalysts to determine experimental reproducibility. 2.2. ATR behavior The above described catalysts with similar Ni SA and different Ni loadings were tested for their ATR performance with sulfur-free and thiophene-containing isooctane. Previously published results have shown that ATR operating conditions have a significant impact on the effects of thiophene [27]. For example, at higher steamto-carbon ratios, low rates of carbon deposition and relatively high production of H2 lead to stable catalytic performance even in the presence of thiophene. A steam-to-carbon ratio of 3.0, an oxygen-to-carbon ratio of 0.75, and a gas-hourly-space-velocity of

200,000 h−1 were used in this study to avoid carbon deposition as much as possible, and to focus on the relationship between Ni particle size and sulfur tolerance. To better observe the effects of thiophene, a higher concentration was used than that used previously [27]. A thiophene concentration of approximately 230 ppmw was used in the isooctane feed, equating to 88 ppmw S in the isooctane, or 7.9 ppmv S in the entire reactor feed stream. Flow controllers (MKS, Bronkhorst) fed reactants into sulfurresistant lines which were heated to 180 ◦ C to vaporize the feed. The reactor consisted of a 1/2 in. quartz tube inside a furnace (Thermolyne). A thermocouple within a closed-ended 1/8 in. quartz tube was used to measure in situ temperature profiles along the center-axis of the reactor tube. Online gas chromatography (Varian CP-3800 GC) was used to analyze the products of the reactor. A single-stage condenser was used to prevent water from flooding the GC columns. Two thermal conductivity detectors and a pulsed flame photometric detector identified and quantified the species within the product stream (H2 , N2 , CO, CO2 , CH4 , ethane, ethylene, propane, propylene, isobutylene, H2 S, CH3 S, and thiophene) that were separated on packed columns in the GC. This GC analysis provided compositional data at an interval of approximately 35 min. Additional details of the reactor system were previously reported [27]. Prior to reaction, 500 mg of catalyst were reduced in situ at 600 ◦ C for 30 min using a 5% H2 in N2 feed. Previous studies have demonstrated that high-temperature light-off causes excessive sintering [22]. Therefore, a lower light-off temperature was used for these experiments. Following reduction, the catalysts were cooled in N2 to 300 ◦ C. The reaction was started by switching the feed to the reactant flow (1.82 mmol/min iC8 H18 containing 230 ppmw thiophene, 45.2 mmol/min H2 O, and 25.8 mmol/min air), and then ramping the furnace to the desired ‘feed’ temperature of 500 ◦ C. For each experiment the reaction was carried out for 400 min time-onstream. An axial temperature profile of the catalyst bed was taken in the last 30 min of the reaction. Following the experiment, the catalysts were cooled under N2 -flow, and prior to any post-reaction

Table 2 Metrics of isooctane ATR performance and their corresponding formulae. Performance metric

Symbol Formula

Synthesis gas yield

YSG

H2 yield

YH2

CO yield

YCO

CO2 yield

YCO2

CH4 yield

YCH4

Yield of other hydrocarbons Y2−4

(FH +FCO ) 2

exit

17×FC H ,inlet 8 18 FH ,exit 2

9×FC H ,inlet 8 18 FCO,exit 8×FC H ,inlet 8 18 FCO ,exit 2

8×FC H ,inlet 8 18 FCH ,exit 4

8×FC H ,inlet 8 18 (2×FC H +2×FC H +3×FC H +3×FC H +4×FC H ) 2 4 2 6 3 6 3 8 4 8 exit 8×FC H ,inlet 8 18

Conversion of isooctane

X

YCO + YCO2 + YCH4 + Y2−4 FH S,exit 2

H2 S yield

YH2 S

CH3 S yield

YCH3 S

Thiophene yield

YC4 H4 S 1 − XC4 H4 S ≈

FC H S,inlet 4 4 FCH S,exit 3

FC H S,inlet 4 4

FC H S,exit 4 4 max(FC H S,exit ) 4 4

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

analysis, the spent catalysts were finely ground to minimize sampling errors. The catalytic activity was described in terms of the yields of H2 , CO, and the combined performance metric, synthesis gas yield (YSG ), as defined in Table 2. Additionally, the distribution of carbon between CO, CO2 , methane, and paraffins, olefins (C2 H4 –C4 H8 ), and the distribution of sulfur (when thiophene was present) between H2 S, CH3 S and thiophene were determined. The yields of these various species were calculated by first determining the flowrates of each species, Fx , in the product. This was achieved by treating N2 as non-reactive species and as internal standard in the interpretation of the GC data. Since previous work has demonstrated a carbon balance closure within 5% [27], summing the yields of carbon-containing species provided a good estimate of isooctane conversion. 2.3. Catalyst characterization The fresh and spent catalysts were characterized by a variety of techniques. H2 chemisorption was used to verify equivalent Ni SA for the pre-reaction catalysts. Ni particle size distributions were obtained by scanning transmission electron microscopy (STEM), and the average Ni particle sizes were obtained by X-ray diffraction (XRD). The nature of the CO–Ni bonding for each of the catalysts was investigated by diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) of CO chemisorption. The post-reaction catalysts were characterized by temperature programmed reduction (TPR), H2 chemisorption, and XRD. H2 chemisorption experiments were carried out using a Micromeritics ASAP 2020 instrument. Prior to analysis, the samples were first evacuated at 100 ◦ C, and then reduced under pure H2 flow at 450 ◦ C for 2 h. Next, the samples were evacuated for 2 h at 450 ◦ C before being cooled to 35 ◦ C for analysis. The H2 uptake was measured at pure H2 doses ranging from 50 to 300 mmHg, at 50 mmHg intervals. The analysis was repeated in order to correct for physically adsorbed H2 that was not stoichiometrically bound to the Ni surface atoms. Accordingly, the uptake reported in this study is the difference between the two measurements extrapolated to 0 mmHg. The H2 uptake was used to calculate an apparent SA by assuming a 1:2 H2 -to-Ni molecular ratio and a cross-sectional atomic surface area of 0.0649 nm2 for Ni [28]. The SA was used to approximate a surface-averaged particle diameter (dp ) assuming a spherical particle shape. STEM coupled with energy dispersive spectroscopy (EDS) elemental mapping was used to measure the Ni particle size distribution in the catalysts and calculate a population-averaged dp . Powder samples were prepared for analysis by dispersing approximately 10 mg of catalyst powder in 5 mL ethanol followed by ultrasonicating the suspension for 1 h to break up agglomerates. Catalyst particles were then deposited on carbon films that were supported on copper grids by submerging the carbon-covered grids in the suspension. A JOEL 2010F analytical electron microscope was used for STEM and EDS studies. The instrument was equipped with a zirconated tungsten [1 0 0] thermal field tip filament and was operated at 200 kV at a pressure of 1.5 × 10−7 Torr. A probe size of 0.5 nm was used for STEM. High-angle annular dark field (HAADF) STEM images were collected with a Gatan retractable CCD camera. Elemental maps for Ni, Ce, and Zr were obtained via an EDAX r-TEM detector with EDAX acquisition software. The effects of image drift were corrected using a drift correction mode. To measure metal particle sizes, the area of each clearly resolved particle in the Ni EDS map was measured and converted to the diameter of a circle having the same area. XRD provided a volume-averaged dp measurement. Phase identification and crystallite size measurements for fresh catalyst

205

samples via XRD were accomplished with a Rigaku Rotating Anode X-Ray Diffractometer. The instrument was equipped with a Rigaku 12 kW rotary anode generator with a Cu K␣ source. K␤ X-ray wavelengths were filtered using a graphite diffracted beam monochromator. Diffraction patterns were obtained over a 2 range of 20–80◦ , at a scan speed of 2◦ /min and a 0.01◦ resolution. Ni particle sizes were calculated using the Scherrer equation (1). The broadening, ˇ, of each profile line of maximum intensity, Ip , was calculated according to an integral broadening expression (2). dp = ˇ=

2dhkl tan ˇ 1 Ip

(1)



I(2)d 2

(2)

For these calculations slower scans (0.1◦ /min) were collected in the 2 regions of Ni [1 1 1] (43–46◦ ) and Ni [2 0 0] (50.4–53.4◦ ) at a 0.2◦ resolution to obtain clean experimental profiles of these relatively weak lines. The contribution of broadening due to instrumental aberrations to the overall broadening of these lines was measured by repeating the procedure for a LaB6 standard powder in the 2 regions of [2 0 0] and [2 1 0] lines, corresponding to Ni [1 1 1] and [2 0 0] lines, respectively. All patterns were smoothed and stripped of their background using JADE software routines. Pure profile broadening values used in particle size calculations were obtained by decomposition of experimental profiles using the Fourier expansion procedure of Rafaja [29]. Infrared signatures of adsorbed CO species can provide information about the bonding character of heterogeneous catalysts. DRIFTS experiments were conducted in a Bruker Optics Tensor 27 Infrared Spectrometer equipped with a Harrick Scientific Praying Mantis IR accessory and high temperature reaction vessel. Catalyst samples were diluted with potassium bromide (KBr, Spectrophotometric Grade, Sigma-Aldrich) to 35.0 wt%, finely ground, and loaded into the DRIFTS sample cell. The sample was oxidized in air at 600 ◦ C for 30 min to clean the sample of any carbonaceous species. Argon gas (UHP, 99.999% Ar) was used to purge the DRIFTS accessory of oxygen; the sample temperature was held at 600 ◦ C during the argon purge. A mixture of 50% H2 in Ar was used to reduce the sample for 60 min at 600 ◦ C. After reduction, UHP Argon was used to purge the DRIFTS accessory and remove chemisorbed H2 for a period of 60 min at 400 ◦ C. The sample was then cooled to 20 ◦ C under UHP Argon gas flow. Once the sample temperature reached 20 ◦ C, a background FTIR spectrum was collected; all FTIR measurements were acquired using 2048 scans at 2 cm−1 of resolution. Carbon monoxide (10% CO in argon) was introduced into the DRIFTS accessory for 20 min and allowed to adsorb on the reduced Ni metal. UHP Argon was used to flush the DRIFTS accessory of gas-phase and weakly bound CO for 30 min, at which point the spectrum of interest was acquired. TPR was used in this study to approximate the mass-averaged oxidation state of spent catalyst materials. Temperature programmed oxidation (TPO) of fresh catalysts followed by TPR provided information about the ease to which metallic Ni particles may be oxidized or that NiO particles may be reduced. Both TPR and TPO analyses were performed using thermogravimetric analysis (TGA) using a TA TGA Q500 instrument. Samples were heated from 120 to 800 ◦ C under a flow of 60 mL/min N2 and 40 mL/min air for TPO and 90 mL/min N2 and 10 mL/min H2 for TPR. In order to corroborate the oxidation state approximations obtained from TPR analysis, the spent catalysts were also characterized using X-ray photoelectron spectroscopy (XPS). The spent catalysts were finely ground and pressed into indium foil. The samples were then evacuated at less than 5 × 10−7 Torr for 8 h prior to XPS analysis. Spectra were collected at approximately 1 × 10−9 Torr

206

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

background pressure. A Kratos Axis Ultra XPS was used with a monochromated Al X-ray source. Ni 2p core scans were collected in the binding energy range between 880 and 846 eV with a 200 ms dwell time at each 0.1 eV interval. The detector was set to analyze electrons at pass energy of 40 eV. Forty scans of a given region were collected in order to improve the signal-to-noise ratio. Binding energy calibration was achieved by setting the adventitious carbon 1s peak position to 284 eV. Data processing, including background removal and peak fitting, was performed using the CasaXPS software package. 3. Results and discussion Previous work has demonstrated that stable activity for ATR of thiophene-doped isooctane could be achieved under conditions where thiophene was completely converted to H2 S, though the activity was reduced compared to sulfur-free cases [27]. Experimental evidence suggested that steam reforming reaction pathways were most impacted by thiophene, and it was hypothesized that the Ni sites most active for steam reforming were also most vulnerable to deactivation by thiophene or H2 S. This hypothesis presents the possibility of enhancing the sulfur-tolerance of Ni catalysts by balancing sites with higher activity and sulfursensitivity and sites with less activity but higher sulfur tolerance. Investigations of model systems have demonstrated that key elementary steps of the steam reforming reaction and the poisoning of sites by sulfur are both structure-sensitive processes; their rates depend strongly on the identity and chemical environment of the active sites [30]. Based on these findings, and the knowledge that higher metal dispersions yield metal particle surfaces with a higher fraction of lower coordinated atoms or sites, Ni dispersion and reforming activity should be closely related. Indeed, the turn-over-frequency (reaction rate per active site) of methane steam reforming has been shown to increase with dispersion for several transition metal catalysts, including Ni [17]. Investigation of structure effects for catalytic ATR in the presence of sulfur is challenging due to the coinciding structure sensitivities of steam reforming and sulfur poisoning, and also due to temperature and composition gradients in the catalyst bed, characteristic of ATR. The goal of this research was to determine if low-coordinated Ni sites, which are more highly concentrated on small Ni particles, were most responsible for deactivation under thiophene exposure. 3.1. Catalyst preparation and hydrothermal reduction treatments One of the easiest methods used to increase the average metal particle size is by increasing the metal loading, which consequently also alters the proportion of different types of active sites. The effect of metal loading on the total number of Ni surface sites is shown in Fig. 1. For a series of Ni/CZO catalysts prepared with varied Ni loading, following a brief calcination at 500 ◦ C for 30 min, it was found that the active Ni SA , as measured by H2 chemisorption, increased linearly with the logarithm of the loading over the range of 0.5–10 wt%. This linear trend holds with a correlation coefficient of 0.989, and solicits an underlying physical explanation that would predict such a simple relationship. A simple model might assume a fixed number of Ni nucleation sites on the blank support and the growth of a fixed number of cubo-octahedral shaped Ni particles at increasing weight loadings. Such geometry has been described previously for face-centered cubic metals [31]. This model predicts a decrease in dispersion as more Ni is added and an increasing number of atoms are present in the inner core or bulk of the particles. However, this predicted decrease in dispersion is smaller than the decrease depicted in Fig. 1. If the above described model had been

Fig. 1. Ni SA of prepared (non-aged) Ni/CZO catalysts measured by H2 chemisorption as a function of the nominal nickel loading (log scale). A linear trend line is displayed; this fits the data with a correlation coefficient of 0.989. Prepared Ni/CZO catalysts were calcined at 500 ◦ C for 30 min as part of the preparation procedure.

observed, then the Ni SA would have varied as a function of the Ni loading to the 2/3 power; but this simple model does not match the experimental observations. As the goal of this work was to focus on the effect of increasing Ni particle size, it was necessary to prepare a series of catalysts with different Ni loadings, resulting in different particle sizes, while attempting to correct for the corresponding increase in the total number of Ni surface sites. This correction was achieved by selectively sintering catalysts to an increasing degree as the Ni loading was increased. As a catalyst is sintered the average particle size will increase and the SA will decrease. This approach simultaneously accentuates the effect of particle size, while correcting for SA . A point of clarification should be made at this point regarding this experimental approach. The process of sintering is an ongoing process at ATR conditions. There exists for each Ni loading a most thermodynamically stable structure of Ni (corresponding to the value SA∞ , described below). Over the course of the ATR experiments it is expected that the SA should decrease with time-on-stream, as it trends towards SA∞ . Therefore, to correct for the increase in the number of active Ni surface sites, the goal was to prepare catalysts with the same initial SA . Sintering of Ni-based catalysts is a fairly well-explored issue as particle growth is a common problem affecting long-term catalytic performance. For example, Sehested et al. have described the growth of Ni particles under simulated steam reforming environments [24–26]. Their work suggested a power-law expression for decrease of SA with time (3) and an Arrhenius-type dependence on temperature (4). In these expressions, k refers to the rate constant for the decrease of SA , n is an integer generally greater than 7, SA∞ is the infinite-temperature limit of SA , and EA is the activation barrier for the process. dSA = kSAn dt

(3)

SA = SA∞ e−EA /RT

(4)

−

To systematically decrease the surface area to desired values, conditions similar to Sehested et al. were used. However, as opposed to the 30 bar used by Sehested et al., the pressures used in this process were less than 2 bar. Fig. 2 shows the effect of the hydrother-

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

207

Fig. 2. The effect of hydrothermal aging treatment temperature (a) and duration (b) on the measured SA for various prepared Ni/CZO catalysts.

mal reduction process on the SA of selected Ni/CZO catalysts. As expected, increased temperature and exposure time both caused decreases in Ni SA . These results did not follow expressions (3) and (4), perhaps due to the lower system pressure used in this study. However, the influences of treatment time and temperature were sufficiently predictable to develop a protocol for preparing catalysts with approximately equal SA . As shown in Table 1, the 3 wt%, 5 wt%, and 10 wt% Ni/CZO samples were treated at the indicated temperatures and durations in order to have nearly the same SA as untreated 0.7 wt% Ni/CZO. This approach resulted in four pre-reaction catalysts with 0.3 ± 0.03 m2 Ni SA /g catalyst. As these catalysts had different Ni loadings, they were expected to exhibit a range of particle sizes. Additional characterization of the pre-reaction catalysts was performed to demonstrate that the experimental catalysts indeed represented a range of particle sizes and had measurable differences in surface chemistry. 3.2. Characterization of pre-reaction catalysts XRD and STEM were each employed to independently verify that the catalysts prepared for this study exhibited significantly different particle sizes, while possessing comparable SA . An estimation of the surface-averaged mean particle diameter (dp ) can be derived from chemisorption data; however, accuracy is limited by Ni surfaces not accessible to probe adsorbate molecules. This is likely due to physical interaction of the metal with the support. Such calculations are expected to over-estimate particle sizes. XRD of the pre-reaction catalysts yielded a volume-averaged dp . Table 1 shows the results of the application of the Scherrer equation (1) to the Ni [1 1 1] and [2 0 0] reflections. XRD patterns showed lines for both fluorite-structured CeO2 and metallic Ni phases. A shift in CeO2 lines to slightly higher 2 values was consistent with the inclusion of some Zr in the fluorite lattice, however, asymmetric features, such as tailing on the higher 2 side of CeO2 lines, suggested inhomogeneity in the Ce–Zr composition [32,33]. No Ni lines could be resolved for the 0.7 wt% loaded catalyst. Additionally, the Ni [2 0 0] reflection for the 3 wt% Ni/CZO sample could not be resolved sufficiently to calculate a crystal size. Based upon the [1 1 1] reflection calculations, there is a clear trend in particle size across the pre-reaction catalysts such that 10 wt% Ni/CZO > 5 wt% Ni/CZO > 3 wt% Ni/CZO. This trend follows the dp trend approximated from H2 chemisorption, with XRD derived dp values consistently about 30% smaller than the values obtained by chemisorption. If assumed that this trend holds for the 0.7% Ni/CZO sample, an estimate of the average Ni particle size for that sample would be 4.9 nm. It is conceivable that the relatively low hydro-

gen uptake during chemisorption is due to a strong metal support interaction (SMSI) effect. It cannot be ruled out that the support becomes mobile during the hydrothermal treatment used to adjust nickel particle sizes. Based on the XRD results, the calculated dp in the [2 0 0] direction was smaller than that calculated in the [1 1 1] direction, which was surprising given that the [1 1 1] plane has a lower surface energy [34]. A simple Wulff-type construction predicts a direct relationship between surface energy of a given surface and the width of a particle in that direction [35]. Possible causes for this discrepancy between prediction and observation may be alternative surface energetics due to the chemical environment of the hydrothermal reduction conditions or a facet-specific interaction with the CZO support. Fig. 3 depicts representative STEM images for the pre-reaction catalysts. STEM coupled with EDS allows mapping of Ni atoms and provided a method to both visualize the shape of Ni particles and to calculate a number-averaged dp . Ni particles in 0.7% Ni/CZO were too small to be counted accurately for this analysis, but distinct Ni particles are visible with the CZO support material for the other samples. Particle size distributions fit to log-normal population distributions are shown in Fig. 4. This analysis confirmed the trend in particle size of the pre-reaction samples as measured by XRD and approximated from H2 chemisorption. The mean Ni particle diameter and the standard deviation of the mean particle diameter are both highest for the 10% Ni/CZO pre-reaction catalyst, followed by 5% Ni/CZO, and then 3% Ni/CZO. The dp values calculated from the [1 1 1] and [2 0 0] reflections were within 1.1 standard deviations of the particle sizes measured by STEM. The combination of the SA and dp measurements confirmed the successful preparation of a series of catalysts with equivalent prereaction Ni SA and different Ni particle sizes. In order to determine that the range in particle sizes were sufficient to affect a measurable change in reactivity, the pre-reaction catalysts were analyzed for their reducibility and oxygen reactivity, and also for the nature of their bonding with a chemisorbed probe molecule, CO. Fig. 5 shows the TPR profiles of the fully oxidized pre-reaction catalysts and the TPO profiles of the fully reduced pre-reaction catalysts. All four catalysts displayed two reduction features, a low-temperature feature between 270 and 290 ◦ C and a hightemperature feature between 630 and 640 ◦ C. The location and constant area of the high-temperature feature is consistent with the surface reduction of the CZO support [36]. The low-temperature feature, whose area increases with Ni loading, is assigned to reduction of NiO species. Results show no significant dependence of reducibility on Ni loading in pre-reaction materials once oxidized.

208

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

Fig. 3. Representative STEM images (left) and corresponding EDS Ni maps (right) for the pre-reaction samples 0.7 wt% Ni/CZO (a), 3 wt% Ni/CZO (b), 5 wt% Ni/CZO (c), and 10 wt% Ni/CZO (d). Scale bars in all images correspond to 100 nm.

By contrast, TPO data shown in Fig. 5b suggested that reduced pre-reaction samples had a significant oxygen reactivity dependence, which was based on Ni loading. All samples displayed a low-temperature oxidation feature at 164 ◦ C. There was an additional, broader oxidation feature displayed by 3, 5, and 10 wt% Ni/CZO, but not 0.7% Ni/CZO. This feature shifted to higher temperatures with loading, from 300 ◦ C for 3 wt% Ni/CZO to 400 ◦ C for 10 wt% Ni/CZO. Though it is not known what physical oxidation processes TPO features represent, the incidences and corresponding temperatures of the various features indicated that the resistance to oxidation of the Ni increased in the order: 0.7 wt% Ni/CZO < 3 wt% Ni/CZO < 5 wt% Ni/CZO < 10 wt% Ni/CZO. By comparing the TPO profiles of the samples at various heating rates, it is clear that these differences are not due merely to changes in diffusion rates. Infrared spectra of CO bound to pre-reaction catalysts provided a more direct probe of the chemical nature of the Ni surface. These experiments also offered some evidence as to the relative contribution of more reactive Ni sites on higher index planes versus less reactive Ni sites on low-index planes. The results, which are displayed in Fig. 6, demonstrated measurable differences in the CO bonding character of the four pre-reaction catalysts. All DRIFT spectra of Ni-loaded samples showed absorption frequencies consistent with CO adsorbed to Ni-metal [37–45]. A summary of the observed CO-stretching frequencies is given in Table 3. As each catalyst has a similar SA , as measured by H2

chemisorption, and the same mass fraction dilution in KBr, it was possible to interpret the relative quantity of individual COadsorption sites via the observed IR band intensity at a specific wavenumber. IR absorption bands near 2039 cm−1 were observed for each sample; these are generally assigned to linear carbonyl species bound to a single nickel site [37–40]. This absorption band can be used to infer the Ni dispersion [41]. The observed intensity of 2039 cm−1 IR band decreased with increasing nickel content, which suggests that the number of defect and low-coordinated Ni atoms decreased for the pre-reaction samples as the Ni loading increased. IR bands for bridged and multi-bonded Ni–CO (Nix≥2 –CO) were also observed between 2000 and 1800 cm−1 for each catalyst. It is generally accepted that the features between 2000 and 1930 cm−1 are associated with bridged species, Ni2 –CO, while features between 1900 and 1800 cm−1 are associated with multibonded species, Ni3,4 –CO [37–43]. The bridged-CO feature was very intense for 0.7 wt% Ni/CZO and much weaker for increasing Ni loadings. The multi-bonded region (1900–1800 cm−1 ) revealed that 0.7 wt% Ni/CZO contains multiple Nix –CO bonding features, while higher Ni loadings exhibited only one large Nix –CO feature. Such broad, low-frequency bands have been attributed to contributions of IR bands associated with triply- or quaternary-bonded CO on Ni [1 0 0] and [1 1 1] planes [43–45]. It is evident that the catalysts that experienced hydrothermal aging primarily exhibited such low-frequency CO absorptions, indicating a larger prevalence

Fig. 4. Particle size distributions obtained from STEM/EDS analysis for the pre-reaction catalysts: 3 wt% Ni/CZO (a), 5 wt% Ni/CZO (b) and 10 wt% Ni/CZO (c). Solid lines show the best-fit of log-normal distributions to each of the histogram data sets.

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

209

Table 3 Observed DRIFTS band positions and relative signal intensities for CO-adsorption on Ni (refer to Fig. 6). Absorption intensities are denoted in parenthesis. 2100–2000 cm−1

Pre-reaction sample

−1

2000–1900 cm−1 −1

(s)

0.7% Ni/CZO

2039 cm

3% Ni/CZO 5% Ni/CZO 10% Ni/CZO

2039 cm−1 (w) 2038 cm−1 (w) 2037 cm−1 (w)

1960 cm

(w)

1959 cm−1 (w) 1955 cm−1 (w) 1957 cm−1 (w)

1900–1800 cm−1 1895 cm−1 1863 cm−1 1826 cm−1 1829 cm−1 1829 cm−1 1834 cm−1

(s), (s), (s) (s) (s) (s)

of Ni [1 0 0] and [1 1 1] surfaces as compared to the catalyst with the lowest Ni content. 3.3. ATR of isooctane Characterization of the pre-reaction catalysts, described in Table 1, provided strong evidence that these samples exhibited a range of mean Ni particle sizes. This range of particle sizes was sufficient to result in catalysts with measurable differences in chemical behavior, while possessing equivalent SA . These catalysts were then tested for their ATR performance to determine the influence of particle size upon actual catalytic behavior. The experiments were performed under both sulfur-free and thiophene-containing feed conditions in order to specifically describe the influence of particle size on the poisoning of Ni by sulfur. Isooctane ATR was carried out at a H2 O/C ratio of 3.0, and O/C ratio of 0.75. As the desired product of this reaction is synthesis gas (H2 and CO), the idealized stoichiometry of reaction (5) may be used as a guide for describing this reaction. Based upon this stoichiometry, the maximum attainable value of YSG is 1.12. Previous analysis has shown that the equilibrium predicted YSG is 1.11 at these inlet conditions [27]. However, these past results have also demonstrated that actual YSG is limited by diminished isooctane conversion (X < 90%). Specifically, for a non-hydrothermally aged 10% Ni/CZO catalyst, YSG was 0.87. 24H O

2 11H2 + 8CO C8 H18 + 2H2 O + 3O2 −→

(5)

The reaction shown in (5) familiarizes the reader with the reactants and desired products of this reaction but masks the complex reaction mechanism that other researchers have endeavored to describe [9,22,46–49]. The chemistry occurring along an ATR catalyst is effectively divided into 3 sequential regions. Immediately upstream of the catalyst, the cracking of a fraction of isooctane into mainly olefin products such as isobutylene and propylene is driven by thermal energy. Highly exothermic partial and total oxidation reactions convert another fraction of isooctane near the front edge of the catalyst. Along the remainder of the catalyst bed, steam reforming and water–gas shift pathways convert the remaining

Fig. 6. DRIFT spectra from the adsorption of CO on the various pre-reaction Ni/CZO catalysts and the CZO support. Inset is zoomed in on the higher wavenumber region of the scan.

isooctane and cracking products to an equilibrium-limited mixture of primarily H2 O, H2 , CO, and CO2 , with a smaller amount of CH4 . YSG versus time-on-stream is plotted in Fig. 7 for all of the catalytic experiments in this study. In all cases the activity was stable throughout the experiment. Results were found to be reproducible when samples were taken from the same synthesis batch and also when the catalyst was prepared again. The standard errors from these separate sources (catalyst preparation and random variation in reactor performance) were not statistically different; therefore they were combined for all of the reported data. The dashed lines in Fig. 7 denote the time-averaged YSG , which may be taken as the steady-state value due to the minimal deviation over time-onstream. These steady-state values are plotted as a function of dp of each catalyst in Fig. 8a. These results demonstrated the effect of particle size on both sulfur-free and thiophene-doped isooctane ATR. Under sulfur-free reforming conditions decreasing Ni particle size had a negative impact on YSG . A prior study of methane stream reforming came to the conclusion that catalysts with smaller dp should be more active in reforming reactions [17]. However, the exact opposite

Fig. 5. TPR (a) and TPO (b) dTGA curves of the pre-reaction catalysts.

210

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

Fig. 7. Results of isooctane ATR experiments. Time-on-stream YSG for the four different Ni/CZO catalysts studied operated under sulfur free and thiophene-doped conditions. The results from the non-aged 0.7 wt% Ni/CZO are shown in (a), the results for the 3 wt% Ni/CZO, 5 wt% Ni/CZO, and 10 wt% Ni/CZO, aged as described in Table 1, are displayed in (b), (c), and (d), respectively. The reported values at each time point are averaged across the three experimental runs at each condition. The dashed lines represent the steady-state time-averaged YSG . (Reactor conditions: H2 O/C = 3, O/C = 0.75, GHSV = 200,000 h−1 , Tinlet = 500 ◦ C, [S]in = 0 or 7.9 ppmv.)

is observed in ATR of isooctane, and YSG decreased from 0.83 for the 50–60 nm particles in 10% Ni/CZO to 0.56 for the <5 nm Ni particles of the 0.7% Ni/CZO catalyst. This is not surprising, as isooctane steam reforming involves the cleaving of C–C bonds requiring larger ensembles of surface sites, while methane steam reforming involves only the cleaving of C–H bonds. Further evidence sup-

ports the notion that small Ni particles are actually less active in an ATR environment. Table 4 shows the breakdown of carbon exiting the reactor. As in the previously reported data [27], the 10 wt% Ni/CZO exhibited nearly 90% isooctane conversion with essentially no olefin products. The presence of isobutylene or prolylene would be associated with poor steam-reforming activity as they are pri-

Fig. 8. Steady-state yields of major species from the sulfur free (a) and thiophene-doped (b) isooctane ATR displayed as a function of pre-reaction Ni particle size.

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214 Table 4 Steady-state carbon balance from the sulfur free and thiophene-doped isooctane ATR experiments for the four catalysts investigated. ATR experiment

YCO

YCO2

YCH4

Y2–4

X

0.7% Ni no C4 H4 S 0.7% Ni w/ C4 H4 S 3% Ni no C4 H4 S 3% Ni w/ C4 H4 S 5% Ni no C4 H4 S 5% Ni w/ C4 H4 S 10% Ni no C4 H4 S 10% Ni w/ C4 H4 S

0.23 0.15 0.32 0.23 0.35 0.21 0.35 0.21

0.38 0.34 0.41 0.37 0.44 0.37 0.49 0.38

8.2e−3 1.7e−2 1.4e−2 1.7e−3 2.2e−2 7.4e−3 2.9e−2 6.9e−3

2.7e−2 8.5e−2 2.8e−3 3.1e−2 7.2e−4 3.2e−2 1.4e−4 2.9e−2

0.64 0.59 0.76 0.64 0.81 0.62 0.87 0.63

marily the product of cracking mechanisms, and these olefins are subsequently converted in endothermic steam reforming pathways to CO and H2 [48,49]. Compared to the 10 wt% Ni/CZO sample, the 5 wt% and 3 wt% samples had lower isooctane conversion, but gave a similar breakdown of carbon species. This observation suggests that olefins were still completely converted to reforming products but the unconverted isooctane was more difficult to reform as the Ni particle size decreased. One possible explanation for this phenomenon is that smaller Ni particles have smaller ensembles of Ni sites, thus making the adsorption of isooctane more difficult. The influence of ensemble size has been invoked previously to describe the interaction of sulfur with catalyst surfaces [50]. Compared to the other three samples, the 0.7 wt% Ni/CZO sample had lower isooctane conversion than the other samples, and additionally about 5% of carbon atoms exited the reactor as olefins, mostly isobutylene, indicating further decreased steam reforming activity of these small Ni particles. The temperatures measured at different axial positions throughout the reactor bed demonstrated the thermal consequences of the diminished reforming activity. Fig. 9a plots the maximum catalyst bed temperature and temperature at the downstream face of the catalyst bed with respect to the Ni particle size. These data demonstrated that as the reforming activity was diminished for smaller particle sizes, likewise, the temperature of the catalyst bed simultaneously increased. An elevated catalyst bed temperature indicates diminished enthalpic contributions of endothermic steam reforming reactions, as compared to exothermic partial oxidation reactions. The influence of particle size on the sulfur tolerance of the ATR reaction is found by examining the performance of the catalyst under conditions of thiophene exposure. As in sulfur-free conditions, the time-on-stream data showed stable performance over the course of all the thiophene-doped experiments (Fig. 7). However, the effect of particle size on YSG was much less than in the

211

sulfur-free case, as shown in Fig. 8b. The only significant variation was observed for the 0.7 wt% Ni catalyst, which exhibited a lower steady-state YSG . When comparing the isooctane conversions and distributions of carbon species for thiophene-doped isooctane ATR (Table 4), there was no clear difference between the performances of the three samples with the largest Ni dp . Moreover, these samples had similar temperature profiles, as shown in Fig. 9b. Interestingly, the isooctane conversion and distribution of carbon species in the thiophene-exposed 3, 5, and 10 wt% Ni/CZO experiments were nearly identical to those metrics in the sulfur-free 0.7% Ni/CZO experiment. As with YSG , 0.7 wt% Ni/CZO diverged from the other samples under thiophene-exposed conditions in having a higher catalyst bed temperature, a lower isooctane conversion, and a substantially increased proportion of isobutylene and propylene exiting the reactor. The yield of H2 S is displayed in Fig. 8b. It has been previously demonstrated [27] that under reforming conditions the complete or nearly complete conversion of thiophene to H2 S correlates with stable time-on-stream syngas yield and hydrocarbon conversion. It was proposed that sufficient hydrogen availability was required to clean off any sulfur bound to the Ni surface and generate H2 S. Thus, with decreasing rates of thiophene hydrodesulfurization, one would expect the sulfur concentration on the Ni sites to increase, and consequently, the reforming activity to decline. However, the results presented here demonstrate that this prediction does not hold for smaller particles of Ni. For example, as the average particle size decreased from 50–60 nm to 20 nm, the yield of H2 S dropped by a factor of two, leaving a significant fraction of thiophene unconverted. Surprisingly, there was no decline in activity over time-on-stream and there was not a measurable decrease in reforming yields. For the smallest particles, the values for YSG and YH2 S were even lower, but again the activity remained constant over the course of the experiment. These results suggest that thiophene had a diminished interaction with Ni as particle size was decreased. However, the sulfur-free activity demonstrates this was not due to elevated reforming activity, which was significantly diminished for smaller Ni dp . It was found that thiophene exposure increased the yield of olefins while decreasing isooctane conversion and YSG . These effects were more pronounced as the particle size increased. However, this is attributed simply to the higher activity of larger Ni particles under sulfur-free conditions. In essence, thiophene exposure effectively eliminated the disparities in performance which had been observed as a function of particle size under sulfur-free reforming conditions. While our experimental observations clearly point towards a dependence of thiophene-induced deactivation on particle size,

Fig. 9. Maximum catalyst bed temperature and exit bed temperature as a function of pre-reaction Ni particle size for sulfur free ATR (a), thiophene-doped ATR (b), and the influence of thiophene on the catalyst bed temperatures (c).

212

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

Table 5 Post-reaction SA , as measured by H2 chemisorption, and Ni particle size, as measured by XRD, of the Ni/CZO catalysts following ATR. Post-reaction sample

SA (10−1 m2 /g)

XRD Ni [1 1 1] dp (nm)

XRD Ni [2 0 0] dp (nm)

0.7% Ni no C4 H4 S 0.7% Ni w/ C4 H4 S 3% Ni no C4 H4 S 3% Ni w/ C4 H4 S 5% Ni no C4 H4 S 5% Ni w/ C4 H4 S 10% Ni no C4 H4 S 10% Ni w/ C4 H4 S

0.47 ± 0.21

N/A

N/A

0.11 ± 0.10

N/A

N/A

0.07 0.27 0.03 0.17 0.30

85.7 ± 8.4 85.0 ± 5.7 83.1 ± 6.7 88.8 ± 7.4 104.9 ± 12.3

77.5 ± 13.3 74.2 ± 3.1 72.9 ± 4.1 83.7 ± 14.6 87.4 ± 7.8

0.20 ± 0.07

114.6 ± 10.1

90.7 ± 6.5

0.49 0.27 1.00 0.34 1.48

± ± ± ± ±

with larger particles being more prone to deactivation, the overall effect of thiophene is rather complex and we cannot, at this point, offer a clear mechanistic explanation from this data set alone. 3.4. Characterization of post-reaction catalysts The novelty in the experimental design of this study was that the initial catalytic materials had equivalent SA with a sufficiently broad range of mean dp . However, the above description of particle size trends should be tempered by the fact that under typical ATR reaction conditions it should be expected that particle sizes will change beyond the control of the experiment design. H2 chemisorption was used to measure the SA of the post-reaction catalysts. The results of this analysis are reported in Table 5, and it can be seen that the active SA decreased significantly over the course of ATR runs. This effect was more pronounced for smaller values of pre-reaction dp , and also for thiophene-doped isooctane ATR. The post-reaction SA of certain samples decreased during reaction enough that H2 adsorption was near the limit of detection, which resulted in the large standard errors reported in Table 5. XRD-calculated Ni dp for the post-reaction samples (reported in Table 5) corroborated the H2 chemisorption results, showing that Ni particle sizes of the catalyst samples increased significantly over the course of the ATR experiments. This increase was most dramatic for the 3 wt% Ni catalyst and least dramatic for the 10 wt% Ni sample. However, unlike the chemisorption results, XRD data suggest the increase in post-reaction dp are not statistically different between sulfur-free and thiophene-exposed samples. It is plausible, that this discrepancy may be explained by the formation of surface sulfides in the presence of thiophene, which would cause decreased gas uptake during chemisorption. The results of Table 5 may be taken to imply that the SA decreased significantly for catalysts that underwent less extensive hydrothermal reduction, and therefore the micro-structures of small Ni particles were simply not as stable as the structures of the larger Ni particles. However, it seems that the rate of sintering during reaction was most strongly dependent upon the catalyst bed temperature. For example, the SA of the 3 wt%, 5 wt%, and 10 wt% Ni catalysts following thiophene-exposed ATR were reasonably similar despite their large difference in initial dp , explained by the similar in situ temperature profiles of these samples. As the TPR and TPO analyses of the pre-reaction catalysts demonstrated, the oxidation of smaller particles occurred at lower temperatures than the oxidation of larger particles. As no major differences in the reduction feature locations in the TPR profiles were observed, it would follow that under ATR conditions the in situ oxidation state of Ni would be closely related to particle size. TPR profiles of post-reaction samples clearly indicate that the oxidation state of Ni is related to particle size. Fig. 10 shows that the average

Fig. 10. Post-reaction average Ni oxidation state as determined by TPR given as a function of pre-reaction Ni particle size.

oxidation state of Ni was very high for the catalyst that initially had the smallest Ni particle size. Conversely, the average oxidation state of the catalyst with the largest particles was below +1. While it appears that the presence of thiophene may have, on average, lowered the oxidation state of Ni for the intermediate sized particles, the large error associated with this analysis makes this distinction difficult to infer from a statistical perspective. XPS was used to further characterize the surface of the postreaction catalyst. The Ni 2p region is shown in Fig. 11 for postreaction samples. The total intensity of the Ni 2p region compared to other peaks, such as the Ce 3d line, was found to increase as a function of Ni loading with a correlation coefficient of over 0.95. The peaks in the Ni spectra at 872, 861, and 855 eV can be attributed to Ni 2p1/2 satellite and 2p3/2 lines, respectively [51,52]. Previous researchers have demonstrated that the position of the Ni 2p3/2 line can be related to the oxidation state of the surface Ni species, with a BE of around 855 eV attributed to Ni2+ /3+ species and a BE of around 852 eV attributed to metallic Ni [53]. Fig. 11 clearly shows two distinct 2p3/2 features suggesting the presence of both Ni0 and Ni2+/3+ in all samples except for the 3 wt% Ni/CZO catalyst used in the thiophene-doped isooctane ATR experiment. Differences between the main 2p3/2 and the satellite line positions have been previously correlated to changes in the oxidation state of Ni and the electronegativity of Ni-bound ligands [54,55]. The values of this difference, (LS)3 , measured for the spectra in Fig. 11 are listed in Table 6. The value of satellite displacement was found to generally increase as a function of the particle size of the initial catalyst material. This increase corroborates the findings of

Table 6 Satellite displacement measurements from the XPS Ni 2p3/2 line in the spectra shown in Fig. 11. Post-reaction sample

(LS)3 (eV)

0.7% Ni no C4 H4 S 0.7% Ni w/ C4 H4 S 3% Ni no C4 H4 S 3% Ni w/ C4 H4 S 5% Ni no C4 H4 S 5% Ni w/ C4 H4 S 10% Ni no C4 H4 S 10% Ni w/ C4 H4 S

5.15 4.93 5.31 5.79 5.71 5.26 5.98 5.26

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

213

Fig. 11. Representative post-reaction Ni 2p core X-ray photoelectron spectra. The backgrounds of these spectra have been subtracted and the peaks have been deconvoluted.

the TPR analyses, suggesting that the oxidation state of Ni increased for smaller Ni particles during the ATR experiments. For purposes of comparison, the expected satellite displacements for metallic Ni and NiO are 6 eV and 3.5 eV, respectively [55]. Comparing the influence of thiophene on the XPS results, the satellite displacements were lower for three of the samples (0.7, 5, and 10 wt% Ni) that had been exposed to thiophene. However, the presence of thiophene coincided with a positive shift of the satellite for the 3 wt% Ni/CZO catalyst. A shift in the satellite position to lower binding energies may suggest the binding of a more electropositive ligand than oxygen, such as sulfur. However, no peaks were seen in the S 2p region of the spectrum (165 eV) to corroborate the presence of surface bound sulfur species. Unfortunately, it is difficult to make a definitive determination as to the presence or absence of sulfur because of the poor XPS signal-to-noise ratio of this region. Evidence of the apparent correlation between pre-reaction Ni dp and post-reaction average Ni oxidation state supports an explanation for poor reforming activity of small Ni particles. The active phase for reforming reactions, Ni0 , is less stable in catalysts with smaller particle sizes, under the ATR reaction environment. It is possible that the formation of surface or bulk nickel oxides can be just as detrimental on the reforming activity of hydrocarbons, such as isooctane and isobutylene, as the interaction of sulfur with active sites. 4. Conclusions This study investigated the influence of Ni dp upon isooctane ATR. Particle size was effectively controlled by altering the loading of the metal over a CZO support material. The catalyst design corrected for the additional influence of Ni loading upon active metallic SA . This was achieved by selectively aging the catalyst material under hydrothermal reducing conditions such that the pre-reaction catalysts exhibited identical SA , as measured by H2 chemisorption. This approach allowed for the specific effects of Ni particle size to be determined. Characterization of the pre-reaction materials convincingly demonstrated the disparity in particle size distributions between the different catalyst samples and proved that these differences resulted in measurable changes in the surface reactivity of the Ni sites. The ATR activity of these catalysts was then compared under sulfur free and thiophene-doped reaction conditions in order to determine the effect of particle size on these separate reaction conditions. It was found that under sulfur free conditions, Ni particle size has a significant influence on the activity of the catalyst

towards production of synthesis gas. Surprisingly, it was the highly dispersed small Ni particles (<5 nm) which exhibited unfavorable reaction characteristics. High catalyst bed temperatures and increased olefin output suggested that catalysts with small Ni particle sizes exhibited poor steam reforming activity. The effects of particle size were somewhat counterbalanced with the presence of thiophene in the feed stream. There was more pronounced deactivation of the larger Ni particles due to thiophene addition. Based upon the analysis of the products of the reforming reaction and post-reaction characterization of the catalysts, two possible explanations for the role of Ni particle size can be presented. The first explanation is that small ensemble sizes of Ni, found on the smallest Ni particles, make it difficult for large molecules such as isooctane to adsorb and react. The second explanation is based upon differences in the apparent oxygen reactivity of the catalysts. As smaller Ni particles were more easily oxidized, it is possible that these Ni sites were oxidized under ATR reaction conditions and the resulting NiO surfaces had diminished hydrocarbon reforming activity. Acknowledgements Financial support for this study was provided by the U.S. Army Tank-Automotive Research, Development & Engineering Center under Cooperative Agreement Number W56HZV-05-2-0001, and by the U.S. Department of Energy under Contract Number DE-FC2606NT42813. The authors would like to acknowledge the help of Dr. Kai Sun at the Electron Microbeam Analysis Laboratories (EMAL) of the University of Michigan for his assistance with the electron microscopy. The STEM equipment is supported at the EMAL facility under NSF grant #DMR-9871177, and the XPS is funded through NSF grant #DMR-0420785. The authors also wish to thank Dr. Galen Fisher for his help and guidance. References [1] G. Somorjai, Introduction to Surface Chemistry and Catalysis, John Wiley & Sons, New York, 1994, pp. 5–10. [2] C. Henry, in: C. Bréchignac, P. Houdy, M. Lahmani (Eds.), Nanomaterials and Nanochemistry, Springer, Berlin, 2007, pp. 3–34. [3] R. Van Hardeveld, F. Hartog, Surf. Sci. 15 (1969) 189–230. [4] M. Flytzani-Stephanopoulos, G.E. Voecks, Int. J. Hydrogen Energy 8 (1983) 539–548. [5] A. Lutz, R. Bradshaw, L. Broomberg, A. Rabinovich, Int. J. Hydrogen Energy 29 (2004) 809–816. [6] A. Docter, A. Lamm, J. Power Sources 84 (1999) 194–200. [7] J. Schwank, A. Tadd, Catalysis 22 (2010) 56–85.

214

J.M. Mayne et al. / Applied Catalysis A: General 400 (2011) 203–214

[8] M. Krumpelt, T. Krause, J. Carter, J. Kopasz, S. Ahmed, Catal. Today 77 (2002) 3–16. [9] D. Shekhawat, T. Gardner, D. Berry, M. Salazar, S. Haynes, J. Spivey, Appl. Catal. A 311 (2006) 8–16. [10] J. Rodriguez, J. Hrbek, Acc. Chem. Res. 32 (1999) 719–728. [11] J. Sehested, Catal. Today 111 (2006) 103–110. [12] J.R. Rostrup-Nielsen, in: J.R. Anderson, M. Boudart (Eds.), Catalysis Science and Technology, Springer-Verlag, Berlin, 1984, pp. 1–117. [13] T.P. Beebe Jr., D.W. Goodman, B.D. Kay, J.T. Yates Jr., J. Chem. Phys. 87 (1987) 2305–2315. [14] J.H. Larsen, I. Chorkendorff, Surf. Sci. Rep. 35 (1999) 163–222. [15] F. Abild-Pederson, O. Lytken, J. Engbæk, G. Nielsen, I. Chorkendorff, J.K. Nørskov, Surf. Sci. 590 (2005) 127–137. [16] J. Wei, E. Iglesia, J. Catal. 224 (2004) 370–383. [17] G. Jones, J.G. Jakobsen, S.S. Shim, J. Kleis, M.P. Andersson, J. Rossmeisl, F. Abild-Pederson, T. Bligaards, S. Helveg, B. Hinnemann, J.R. RostrupNielsen, I. Chorkendorff, J. Sehested, J.K. Nørskov, J. Catal. 259 (2008) 147–160. [18] H.S. Bengaard, J.K. Nørskov, J. Sehested, B.S. Clausen, L.P. Nielsen, A.M. Molenbroek, J.R. Rostrup-Nielsen, J. Catal. 209 (2002) 365–384. [19] J.R. Rostrup-Nielsen, Steam Reforming Catalysts, Danish Technical Press, Copenhagen, 1975. [20] T. Borowiecki, Appl. Catal. 4 (1982) 223–231. [21] X. Chen, A. Tadd, J. Schwank, J. Catal. 251 (2007) 374–387. [22] B. Gould, X. Chen, J. Schwank, J. Catal. 250 (2007) 209–221. [23] B. Gould, X. Chen, J. Schwank, Appl. Catal. A 334 (2008) 277–290. [24] J. Sehested, J. Catal. 217 (2003) 417–426. [25] J. Sehested, J.A.P. Gelten, S. Helveg, Appl. Catal. A 309 (2006) 237–246. [26] J. Sehested, J.A.P. Gelton, I.N. Remediakis, H. Bengaard, J.K. Nørskov, J. Catal. 223 (2004) 432–443. [27] J. Mayne, A. Tadd, J. Schwank, J. Catal. 271 (2010) 140–152. [28] ASAP 2020, Chemi Operator’s Manual, Micromeritics Instrument Corporation, 2005. [29] D. Rafaja, Mater. Struct. 7 (2000) 43–50. [30] K. Ahmed, L.S. Kershenbaum, D. Chadwick, Ind. Eng. Chem. Res. 29 (1990) 150–156.

[31] G. Agostini, R. Pellegrini, G. Leofanti, L. Bertinetti, S. Bertarione, E. Groppo, A. Zecchina, C. Lamberti, J. Phys. Chem. C 113 (2009) 10485–10492. [32] T. Montini, A. Speghini, L. De Rogatis, B. Lorenzut, M. Bettinelli, M. Graziani, P. Formasiero, J. Am. Chem. Soc. 131 (2009) 13155–13160. [33] E. Kümmerle, G. Heger, J. Solid State Chem. 147 (1999) 485–500. [34] L. Vitos, A. Ruban, H. Shriver, J. Kollar, Surf. Sci. 411 (1998) 186–202. [35] I. Chorkendorff, J.W. Niemantsverdriet, Concepts of Modern Catalysis and Kinetics, John Wiley & Sons, Weinheim, 2003, p. 180. [36] A. Tadd, Ph.D. Dissertation, University of Michigan, 2006, p. 33. [37] G. Blyholder, J. Phys. Chem. 68 (1964) 2772. [38] K. Hadjiivanov, M. Mihaylov, D. Klissurski, P. Stefanov, N. Abadjieva, E. Vassileva, L. Mintchev, J. Catal. 185 (1999) 314–323. [39] M. Agnelli, H.M. Swaan, C. Marquez-Alvarez, G.A. Martin, C. Mirodatos, J. Catal. 175 (1998) 117–128. [40] E.B. Pereira, G.A. Martin, Appl. Catal. A 103 (1993) 291–309. [41] C. Resini, T. Venkov, K. Hadjiivanov, S. Presto, P. Riani, R. Marazza, G. Ramis, G. Busca, Appl. Catal. A 353 (2009) 137–143. [42] K. Hadjiivanov, M. Mihaylov, N. Abadjieva, D. Klissurski, J. Chem. Soc. Faraday Trans. 94 (1998) 3711–3716. [43] D.G. Blackmond, E.I. Ko, J. Catal. 96 (1985) 210–221. [44] J.C. Campuzano, R.G. Greenler, Surf. Sci. 83 (1979) 301–312. [45] J.C. Bertolini, B. Tardy, Surf. Sci. 102 (1981) 131–150. [46] R. Horn, K. Williams, N. Degenstein, L. Schmidt, J. Catal. 242 (2006) 92–102. [47] R. Horn, K. Williams, N. Degenstein, A. Bitsch-Larsen, D. Dalle Nogare, S. Tupy, L. Schmidt, J. Catal. 249 (2007) 380–393. [48] A. Qi, S. Wang, G. Fu, D. Wu, Appl. Catal. A 293 (2005) 71–82. [49] S. Yoon, I. King, J. Bae, Int. J. Hydrogen Energy 33 (2008) 4780–4788. [50] J. Rodriguez, Prog. Surf. Sci. 81 (2006) 141–189. [51] N.S. McIntyre, M.G. Cook, Anal. Chem. 47 (1975) 2208–2213. [52] D. Srinivas, C.V.V. Satyanarayana, H.S. Potdar, P. Ratnasamy, Appl. Catal. A 246 (2003) 323–334. [53] K.M. Shaju, G.V. Subba Rao, B.V.R. Chowdari, Electrochim. Acta 48 (2003) 1505–1514. [54] G.C. Allen, S.J. Harris, J.A. Jutson, J.M. Dyke, Appl. Surf. Sci. 37 (1989) 111–134. [55] A.P. Grosvenor, M.C. Biesinger, R.C. St. Smart, N.S. McIntyre, Surf. Sci. 600 (2006) 1771–1779.