Ps

Ba

b

a

ARRA

KSCNPP

1

srbraeigmmmpcr

bs

0d

J. of Supercritical Fluids 50 (2009) 82–90

Contents lists available at ScienceDirect

The Journal of Supercritical Fluids

journa l homepage: www.e lsev ier .com/ locate /supf lu

reparation of carbon black supported Pd, Pt and Pd–Pt nanoparticles usingupercritical CO2 deposition

. Cangüla, L.C. Zhangb, M. Aindowb, C. Erkeya,∗

Chemical & Biological Engineering Department, Material Science and Engineering Program, Koc University, 34450 Sarıyer, Istanbul, TurkeyInstitute of Material Science, Department of Chemical, Materials and Biomolecular Engineering, University of Connecticut, Storrs, CT 06269, USA

r t i c l e i n f o

rticle history:eceived 15 September 2008eceived in revised form 31 March 2009ccepted 1 April 2009

eywords:upercritical depositionarbonanoparticles

a b s t r a c t

The preparation of carbon black (Black Pearl 2000) supported single Pd, Pt and bimetallic Pd–Pt nanopar-ticles utilizing the supercritical CO2 deposition method was investigated. Palladium (II) acetylacetonate(Pd(acac)2) and dimethyl (cyclooctadiene) platinum (II) (Pt(cod)me2) were utilized as metallic precursors.The adsorption isotherms of Pd(acac)2 and Pt(cod)me2 on BP2000 in scCO2 at 20 MPa and 60 ◦C were deter-mined. A mass transfer model was found to represent the experimental data on the kinetics of adsorptionof Pd(acac)2 onto BP2000 in scCO2 fairly well with a fitting error of 5.6%. In order to prepare the supportednanoparticles, chemical reduction with H2 in scCO2 was utilized. Increasing reduction temperature andmetal loading caused an increase in Pd particle size. The Pd particles were irregularly distributed and the

latinumalladium

size range of the particles was 3–100 nm, whereas Pt nanoparticles were homogeneously distributed witha size range of 2–6 nm. Binary metal nanoparticles could also be produced by simultaneous adsorptionand subsequent reduction of the precursors. It was found that addition of Pt increased the homogeneityand reduced the particle size on the support compared to single Pd nanoparticles. From the X-ray spec-trometry data, it was seen that there was a heterogeneous mixture of the bimetallic nanoparticles. Pt-richnanoparticles had diameters of around 4 nm, whereas Pd-rich nanoparticles were larger with diameters

of around 10 nm.

. Introduction

Metal nanoparticles supported on high surface area carbonubstrates are used extensively as catalysts for a wide variety ofeactions. Among these catalysts, carbon-supported single Pd andimetallic Pd–Pt catalysts are commonly used for hydrogenationeactions in the fine chemical industry [1]. Bimetallic Pd–Pt cat-lysts show potential for the oxygen reduction reaction (ORR) inlectrochemical fuel cells [2], catalytic oxidation of sulfur dioxiden exhaust gas treatment, hydrogen peroxide production and hydro-enation of aromatics [3]. Carbon-supported single and binaryetal nanoparticles are prepared by a variety of methods includingolecular-capping-based colloidal synthesis [4], microemulsionethod [5], impregnation [6], sonochemical method [7], modified

olyol reduction [8] and deposition-precipitation method [9]. Theatalytic properties of such materials depend strongly on the prepa-

ation method.

The supercritical fluid (SCF) deposition technique has recentlyeen receiving increased attention for the preparation of carbon-upported catalysts. This process involves the dissolution of a

∗ Corresponding author. Tel.: +90 2123381866; fax: +90 2123381548.E-mail address: cerkey@ku.edu.tr (C. Erkey).

896-8446/$ – see front matter © 2009 Elsevier B.V. All rights reserved.oi:10.1016/j.supflu.2009.04.001

© 2009 Elsevier B.V. All rights reserved.

metallic precursor (MP) in a supercritical fluid and the exposure ofthe carbon support to the solution. After adsorption of the precursoronto the support, the metallic precursor is converted to its metalform by chemical or thermal reduction. This technique has beenused to prepare nanoparticles of a wide variety of metals includingPt, Pd, Ru and Rh supported on various types of carbon substrates[10–17] as well as other substrates [18].

In this study, carbon black supported single Pd, Pt and bimetal-lic Pd–Pt nanoparticle catalysts were prepared by scCO2 depositionusing Pd(acac)2 and Pt(cod)me2 as the MPs. Black Pearl 2000(BP2000) was used as the carbon support. BP2000 has a stan-dard BET surface area of 1450 m2/g, an external surface area of515 m2/g, a micropore volume of 0.454 cc/g and a total pore vol-ume of 1.75 cc/g [19]. The preparation of the Pd/BP2000 involvedadsorption of Pd(acac)2 from scCO2 solution, followed by chemi-cal reduction in a supercritical mixture of H2 and CO2. The effect ofreduction temperature and metal loading on the size of Pd particleswas also investigated. The preparation of the Pt/BP2000 involvedadsorption of Pt(cod)me2 from scCO2 solution, followed by chem-

ical reduction in a supercritical mixture of H2 and CO2. In order toprepare Pd–Pt/BP2000 catalysts, the carbon support was exposedto a solution of Pt(cod)me2 and Pd(acac)2 dissolved in scCO2 andboth of the precursors adsorbed on the surface of the supportsimultaneously. Subsequently, the precursors were reduced to their

rcritica

mHre

citsbaawilidwtw6

2

2

m(iwaA

2

c5StItitBsbtItiTfwa

atcsHspw

B. Cangül et al. / J. of Supe

etal form by chemical reduction in a supercritical mixture of2 and CO2. The prepared catalysts were characterized using X-

ay diffraction (XRD), transmission electron microscopy (TEM) andnergy-dispersive X-ray spectrometry (EDXS).

The thermodynamics and kinetics of adsorption of the pre-ursors from the scCO2 phase onto the carbon phase were alsonvestigated. The thermodynamics of adsorption is described byhe equilibrium adsorption isotherm of the MP–scCO2–substrateystem at the impregnation conditions and gives the correlationetween the concentration of the MP in the supercritical phasend the uptake amount of MP on the substrate. Knowledge of thedsorption isotherm is necessary to design large scale processeshich utilize SCF deposition to prepare supported particles. There

s limited data in the literature for adsorption equilibria of metal-ic precursor on porous supports. In this study, the adsorptionsotherms of Pt(cod)me2 and Pd(acac)2 on BP2000 in scCO2 wereetermined at 20 MPa and 60 ◦C. An adsorption kinetics model,hich was recently developed for describing the kinetics of adsorp-

ion of Ru(cod)(tmhd)2 into carbon aerogels from scCO2 solutions,as tested for the Pd(acac)2/BP2000-scCO2 system at 20 MPa and0 ◦C.

. Materials and methods

.1. Materials

Palladium (II) acetylacetonate (99%, molecular weight = 304.6 g,elting point = 205 ◦C) was purchased from Aldrich. Dimethyl

cyclooctadiene) platinum (II) (99%, molecular weight = 333 g, melt-ng point = 104 ◦C) was purchased from STREM. Black Pearl 2000as purchased from Cabot International. The chemicals were used

s-received. Carbon dioxide (99.998%) was purchased from Messerligaz.

.2. Experimental methods

The vessel used for adsorption and reduction experiments wasonstructed out of stainless steel and had an internal volume of4 ml. It was fitted with two sapphire windows (1 in. in diameter,apphire Engineering, Inc., Pocasset, MA) sealed on both sides ofhe vessel with polyetheretherketone O-rings (Valco Instruments,nc., Houston, TX), a thermocouple assembly, a vent line and a rup-ure disk assembly (Autoclave Engineers). The experimental set-ups given in Fig. 1. For the adsorption isotherm measurements, a cer-ain amount of MP (Pd(acac)2 or Pt(cod)me2), a certain amount ofP2000 in a pouch and a magnetic stir bar were placed in the ves-el for each run. The vessel was heated to the desired temperaturey a circulating heater/cooler (Cole Parmer, Model 12108-15) andhen it was pressurized with CO2 from a syringe pump (Teledynesco, Model 260 D) to the desired pressure. The system was kept athese conditions until equilibrium was reached. Then, the solutionn the vessel was drained slowly through the vent line into the hood.he vessel was cooled and the impregnated substrate was removedrom the vessel. The amount of MP adsorbed onto the substrateas determined by the weight change of the substrate using an

nalytical balance accurate to ±0.1 mg (AND GR-200).For the preparation of Pd/BP2000 catalysts, predetermined

mounts of Pd(acac)2 and BP2000 were placed in the vessel andhe vessel was heated to the desired temperature. The vessel washarged with CO2 to a pressure of 20 MPa at this temperature. The

ystem was kept overnight at these conditions. A 10 ml mixture of2 and CO2 (PH2 = 0.8 MPa, Ptotal = 27.6 MPa) was prepared in a ves-

el which was placed between the main vessel inlet and the syringeump. The 10 ml vessel was first filled with H2 at 0.8 MPa. After-ards, the vessel was filled with CO2 to a pressure of 27.6 MPa. Then,

l Fluids 50 (2009) 82–90 83

the high pressure H2 and CO2 mixture was injected into the mainvessel containing the Pd(acac)2/BP2000 in scCO2 solution at 20 MPaand at the desired reduction temperature. The selected reductiontemperatures were 50, 60, 70 and 80 ◦C. The adsorption and reduc-tion pressures, and the amounts of substrate and MP placed intothe vessel were kept constant in each experiment to investigate theeffect of reduction temperature on Pd particle size. Reduction wasallowed to continue for 3 h. After the reduction process, the vesselwas depressurized slowly and allowed to cool to room temperature.The metal/substrate composite was then removed from the vessel,and the metal loading was measured from the weight change byusing an analytical balance. There was good agreement between themeasured Pd amount and the amount calculated from the adsorbedamount of Pd(acac)2 assuming that: the metallic precursor wascompletely reduced to its metal form, the organic ligands weretransformed to acetylacetone, and the acetylacetone was trans-ferred to the fluid phase. For the investigation of the effect of Pdloading (wt.% metal), the above procedure for the adsorption andreduction experiments were carried out, but this time the adsorp-tion and reduction temperature were kept constant at 80 ◦C and theamount of Pd(acac)2 placed in the main reactor vessel was variedfor each run.

In order to prepare Pt/BP2000 catalysts, a predeterminedamount of Pt(cod)me2 and BP2000 were put in the high pressuremain reactor vessel. The vessel was heated to 80 ◦C. Then, the vesselwas filled with CO2 up to 20 MPa by the syringe pump. Pt(cod)me2dissolved in scCO2 and adsorbed on the surface of BP2000. Afterthe system reached equilibrium, a 10 ml mixture of H2 and CO2(PH2 = 0.8 MPa, Ptotal = 27.6 MPa) was injected into the vessel usingthe syringe pump as described above. Reduction was continued for5 h. Then, the vessel was drained and the system was cooled toroom temperature. The metal loadings of the catalysts were calcu-lated from the weight change by using an analytical balance. It wasassumed that the metallic precursor was completely reduced to itsmetal form, the organic ligands were transformed to cyclooctaneand methane, and these products were removed from the systemduring depressurization. The preparation of Pd–Pt/BP2000 cata-lysts was carried out the same way as for the single metal catalysts.The only difference was that predetermined amounts of Pd(acac)2and Pt(cod)me2 precursors were placed into the vessel simultane-ously.

XRD continuous measurements were carried out by using a CuK� source Huber G 670 Imaging Plate in a 2� range of 5–86.915◦ witha scanning rate of 5.5◦ min−1. The morphology of the supportedcatalysts was characterized by high-resolution TEM. Specimens forTEM examination were prepared by carefully crushing the sam-ples with a mortar and pestle set. The resulting powders weresuspended in a volatile solvent and ultrasonicated to obtain auniform suspension. One or two drops of this suspension weredeposited onto a copper mesh grid coated with a holey carbonfilm (Quantifoil Micro Tools GmbH). The solvent was allowed toevaporate completely before examining the TEM specimens in aJEOL 2010 FasTEM operating at 200 kV. This instrument is equippedwith a high-resolution objective lens pole-piece (spherical aberra-tion coefficient Cs = 0.5 mm) giving a point-to-point resolution of<0.19 nm in phase contrast images. Chemical microanalysis was per-formed in situ using an EDAX Phoenix atmospheric thin windowEDXS.

3. Results and discussion

3.1. The adsorption isotherms and the kinetics model

The adsorption isotherms for Pd(acac)2 and Pt(cod)me2on BP2000 in scCO2 at 20 MPa and 60 ◦C are shown in

84 B. Cangül et al. / J. of Supercritical Fluids 50 (2009) 82–90

Fig. 1. Schematic diagram of th

F

FbPt

F6

ig. 2. Adsorption isotherm of Pd(acac)2/BP2000-scCO2 system at 20 MPa and 60 ◦C.

igs. 2 and 3, respectively. There are marked differencesetween the two isotherms. The adsorption isotherm for thed(acac)2/BP2000-scCO2 system is linear whereas the isotherm forhe Pt(cod)me2/BP2000-scCO2 system is non-linear and can be rep-

ig. 3. Adsorption isotherm of PtMe2COD/BP2000-scCO2 system at 20 MPa and0 ◦C.

e experimental system.

resented by the Langmuir model given by:

q = K1Q0C

1 + K1C(1)

where q is the uptake amount of MP on the substrate (mol/kg),C is the concentration of MP in the fluid phase (mol/m3), K1 isthe Langmuir adsorption constant (m3 scCO2/mol MP) and Q0 isthe adsorption capacity (mol MP/kg substrate). K1Q0 is the rela-tive affinity of the MP towards the surface of the adsorbent. Usingnon-linear regression analysis, K1 = 0.14 and Q0 = 4.5.

This difference can most likely be attributed to substantiallyhigher concentrations of Pt(cod)me2 in the scCO2 phase. In theisotherms, the upper limit of the concentration in the fluid phasecorresponds to the solubility of the MP in the fluid phase. Sincethe solubility of Pt(cod)me2 in scCO2 (21.7 mol/m3) is much higherthan the solubility of Pd(acac)2 in scCO2 (0.66 mol/m3) at 20 MPaand 60 ◦C [20], the range of concentration of Pt(cod)me2 in the fluidphase is much larger. The area covered by Pt(cod)me2 or Pd(acac)2molecules on the BP2000 surface, S, can be calculated using:

S = Q0A�rp2 (2)

where Q0 is the maximum uptake, A is Avogadro’s numberand rp is the radius of a Pt(cod)me2 or Pd(acac)2 molecule. Thediameters of Pd(acac)2 and Pt(cod)me2 molecules were determinedusing the Marvin Beans software as 0.80 and 0.55 nm, respectively.The area covered by Pt(cod)me2 molecules is 620 m2/g whereasthe area covered by Pd(acac)2 molecules is 640 m2/g. This indi-cates that, at maximum uptake, both Pt(cod)me2 and Pd(acac)2molecules cover all the accessible surface of the BP2000. The uptakeamount of Pd(acac)2 on BP2000 is higher than the uptake amountof Pt(cod)me2 at the same fluid concentration in the supercriti-cal phase, indicative of a stronger precursor-support interaction forPd(acac)2/BP2000.

The kinetics of adsorption of Pd(acac)2 on BP2000 in the pres-ence of scCO2 was investigated by determining the uptake amountsat different times. Previously, we reported on a model for describ-

ing the kinetics of adsorption of a ruthenium precursor onto carbonaerogels from scCO2 solutions [21]. In this study, the applicabilityof such a model to predict the kinetics of adsorption of Pd(acac)2onto spherical BP2000 pellets was investigated. The batch adsorp-tion process can be represented by a differential material balance

rcritica

o

ε

qsrtorc(

BB

Si

vlg

wPFpAwasrg

B. Cangül et al. / J. of Supe

n a spherical adsorbent (BP2000) particle given by Eq. (3).

p∂CA

∂t+ �p

∂q

∂t= De

(∂2CA

∂r2+ 2

r

∂CA

∂r

)(3)

where CA is the concentration of Pd(acac)2 in scCO2 (mol/m3),is the uptake amount of Pd(acac)2 on BP2000 (mol/kg), r repre-

ents the radial coordinate (m), and t is time (s). The term ∂CA/∂tepresents the accumulation term, ∂q/∂t represents the adsorptionerm and ∂CA/∂r represents the flux term. The explanation of thether terms is given in Table 1. The accumulation term in Eq. (3) wasewritten in terms of the isotherm relationship and concentrationhange with time through the use of the chain rule resulting in Eq.4).

∂CA

∂t= De

εp + �p(∂q/∂CA)

(∂2CA

∂r2+ 2

r

∂CA

∂r

)(4)

The boundary conditions are;

oundary condition 1 ∂CA/∂t = 0 at r = 0.oundary condition 2 −V (∂CA/∂t) = mSDe (∂CA/∂r) at r = R.

where V is the volume of scCO2 in the vessel, m is the mass andis the external surface area of BP2000 particles, respectively. The

nitial conditions are:

Initial condition 1 : CA = 0 at t = 0 at 0 ≤ r < RInitial condition 1 : CA = CA0 at t = 0 at r = R

With the given boundary and initial conditions, Eq. (4) was con-erted to a set of ordinary differential equations using the method ofines and solved simultaneously using the values of the parametersiven in Table 1.

The values for most of the parameters used in the modelere obtained by independent measurements. The diffusivity of

d(acac)2 in scCO2 was calculated using the correlation given byunazukuri et al. [22]. The tortuosity factor was left as an adjustablearameter in the model and was regressed from experimental data.comparison of experimental data with model results obtainedith a tortuosity value of 1.3 is given in Fig. 4 and there is very good

greement between the two with an average fitting error of 5.6%. Aseen from Fig. 4, it takes around 4 h for the system to reach equilib-ium and the concentration in the fluid phase approaches the valueiven by the adsorption isotherm as expected.

Fig. 4. Comparison of experimental data with model results.

l Fluids 50 (2009) 82–90 85

3.2. Preparation of supported Pd nanoparticles

It was found that reduction of the Pd(acac)2 precursor couldbe carried out by subjecting the Pd(acac)2/BP2000 to a super-critical mixture of H2 and CO2. Before the injection of hydrogento the system, Pd(acac)2 is distributed between the scCO2 phaseand the BP2000 phase. The yellow-colored solution with dissolvedPd(acac)2 turned colorless a short time after H2 was injected asobserved from the sapphire window on the side of the vessel.When thermal reduction was carried out in scCO2 at 80 ◦C with-out any hydrogen, the color of the solution did not change andremained yellow for a prolonged period. The peaks for Pd in XRDspectra obtained from the catalysts indicated that chemical reduc-tion of adsorbed Pd(acac)2 was also occurring on the surface ofBP2000.

The effects of reduction temperature and metal loading on theaverage Pd particle size were also investigated using XRD. Fourdifferent reduction temperatures of 50, 60, 70 and 80 ◦C wereselected. The XRD spectra obtained from the Pd/BP2000 catalystswith 20 wt.% Pd loading prepared at these four different reduc-tion temperatures are given in Fig. 5a. The size of the Pd particleswas obtained from the spectra using the Scherrer formula. Sincethe characteristic {1 1 1} and {2 0 0} peaks overlap with the broadC peaks, the size of the Pd particles was calculated from the fullwidth at half maximum of the {2 2 0} peak. As shown in Fig. 5b, theaverage Pd particle size increased as the chemical reduction tem-perature in scCO2 was increased. The average size doubled from 7to 14 nm as the temperature was increased from 50 to 80 ◦C. Themost likely explanation for this effect is an increase in the mobilityof the reduced palladium atoms with increasing reduction temper-ature, which results in more extensive particle coarsening duringthe reduction process. In a study by Kim et al. [23] where Pd(hfac)2was deposited on SiOx in liquid CO2 and reduced with H2 at atmo-spheric pressure, it was found that the average Pd particle diameterincreased from 5.9 to 7.9 nm when the reduction temperature wasincreased from 75 to 150 ◦C. The smaller Pd particle size obtainedon SiOx can be attributed to the lower mobility of the Pd atoms onSiOx.

Examples of TEM images obtained from the catalyst with 20 wt.%Pd loading prepared by chemical reduction in H2/scCO2 at 80 ◦C areshown in Fig. 6. Images such as Fig. 6a, which were obtained at lowermagnifications, reveal clearly the structure of the BP2000 supportand the sizes and distribution of the Pd nanoparticles. The lighterfeatures in such images correspond to the support, which consistsof inter-linked homogeneous C particles 20–50 nm in diameter asexpected for BP2000. The darker features superimposed upon thesupport are the Pd nanoparticles which are distributed rather inho-mogeneously. These Pd nanoparticles are present in a wide rangeof different sizes ranging from 3 nm up to around 100 nm. Highermagnification images such as Fig. 6b show that the C particlesthat comprise the support consist of disordered carbon. More-over, such images reveal that the larger Pd nanoparticles (>20 nm)have an irregular morphology suggesting that these may arise bythe agglomeration of smaller nanoparticles migrating across thesupport surface instead of by coarsening due to the atomic diffu-sion of metallic Pd. Further evidence for this is shown in particlesize distributions obtained by measuring the projected diameterof over 100 particles in images such as Fig. 6b and binning thesemeasurements to give histograms of the percentage of particlesin each size range. One such histogram obtained from the cata-lyst with 20 wt.% Pd loading is presented in Fig. 6c. In this case

the particles measured had a mean size of 37.6 nm with a standarddeviation of 28.2 nm. This mean particle size is substantially largerthan the value obtained by XRD, but since the use of the Scher-rer formula to analyze XRD data gives the crystal or grain size, thediscrepancy between the XRD and TEM values is consistent with

86 B. Cangül et al. / J. of Supercritical Fluids 50 (2009) 82–90

Table 1Parameters used in the solution of adsorption kinetics model.

Parameters Definiton Unit Value

BP2000 �P Density of the porous carbon particles kg/m3 444R Radius of the spherical carbon particles m 6.5 × 10−4

εp Porosity – 0.78m Mass of carbon used in experiments kg 10−4

S External surface area per mass of carbon particles m2/kg 10.38� Tortuosity—the adjustable parameter of the system – 1.3

Bulk scCO2 properties V Volume of scCO2 (volume of high pressure vessel) m3 54 × 10−6

CA0 Initial concentration of Pd(acac)2 in scCO2 at t = 0 mol/m3 1.22

Slope of the isotherm – q/CA m3/kg 3.22

Diffusivity DAB Binary diffusion coefficient m2/s 1.72 × 10−8 [22]ds Pd(acac)2 diameter m 8 × 10−10

of caty

tp

actea3dFot

nhcPwocwPbstts

Ft

dp Average pore sizeDe Effective diffusivi

he largest particles being polycrystalline agglomerates of smallerarticles.

The effect of metal loading on the growth of Pd particle size waslso investigated. In these experiments, the chemical reduction wasarried out at 80 ◦C in H2/scCO2 (PH2 = 0.8 MPa, Ptotal = 27.6 MPa) andhe amount of Pd(acac)2 put into the vessel before the adsorptionxperiments was varied. The amounts were selected based on thedsorption isotherm. The Pd loadings of the prepared catalysts were, 5, 12 and 20 wt.%. XRD spectra obtained from the catalysts withifferent Pd loadings are presented in Fig. 7a. As can be seen inig. 7b, the average Pd particle size increased with the metal loadingf the Pd/BP2000 composites: the average size increased from 7.8o 14 nm as the loading was increased from 3 to 20 wt.% Pd.

Ye et al. [16] decorated functionalized MWCNTs with Pdanoparticles by supercritical deposition using Pd(hfa)2·xH2O (hfa:exafluoroacetylacetonate) as the precursor. The reduction wasarried out in a H2/scCO2 mixture at 80 ◦C. TEM images of thed/MWCNTs revealed well dispersed and spherical particles thatere anchored onto the external walls of MWCNTs and the size

f these particles were about 5–10 nm at a 10 wt.% Pd loading. Inomparison, the TEM image of a commercial Pd/activated carbonith the same metal loading showed the existence of very large

d particles that were irregularly distributed on the activated car-on surface. The surface of the BP2000 support used in this study

eems to be similar in nature to the surface of activated carbon forhe formation of Pd nanoparticles by supercritical deposition sincehe Pd particles are not homogeneously distributed and there areome very large particles on the BP2000 support.

ig. 5. (a) XRD spectra of the 20 wt.% Pd loaded samples prepared by chemical reductemperature on the growth of Pd particles with XRD results shown in (a).

rbon particle m 2 × 10−9

m2/s 1.77 × 10−9

3.3. Preparation of Pt/BP2000 catalysts

12 wt.% metal loaded Pt/BP2000 catalysts were prepared first byadsorption of a certain amount of Pt(cod)me2 onto BP2000 fromscCO2 at 20 MPa and 80 ◦C. Subsequently, the reduction of the pre-pared Pt(cod)me2/BP2000 composites was carried out in H2/scCO2.The XRD data obtained from the Pt/BP2000 catalysts revealed amuch finer distribution of metal than that for the Pd/BP2000 cat-alysts, with a mean particle size of 2.8 nm. This is consistent withthe TEM data and representative images are shown in Fig. 8. Theparticles are too fine to be observed in images obtained at lowermagnifications (such as that used in Fig. 6a) and high magnifica-tion images (e.g. Fig. 8a) reveal a very uniform dispersion of metalnanoparticles with a narrow size distribution. The size and mor-phology of the Pt nanoparticles are revealed more clearly in imagesobtained at even higher magnifications (e.g. Fig. 8b) and a particlesize distribution obtained from such images is presented in Fig. 8c.These nanoparticles are equiaxed and most range from 2 to 6 nm indiameter with a mean of 3.7 nm and a standard deviation of 1.8 nm.While there is some suggestion of agglomeration in these images,tilting experiments in the TEM have shown that this is mainly aprojection artifact and that the vast majority of the Pt nanoparti-cles are well separated. We note that these data are comparablewith those obtained in our previous work on Pt/C-aerogel catalysts

prepared under similar conditions [10,11,24]. This indicates that themobilities of the Pt atoms and nanoparticles on the BP2000 sur-face are significantly lower than the corresponding mobilities forPd.

ion in H2/scCO2 at different reduction temperatures. (b) The effect of reduction

B. Cangül et al. / J. of Supercritical Fluids 50 (2009) 82–90 87

F paredf

3

ca8

Fs

ig. 6. (a) and (b) TEM images of the 20 wt.% metal loaded Pd/BP2000 catalysts prerom images such as (b).

.4. Preparation of Pd–Pt/BP2000 catalysts

16 wt.% metal loaded Pd–Pt/BP2000 binary metallic nanoparti-les were obtained first by simultaneous adsorption of a certainmount of Pd(acac)2 and Pt(cod)me2 from scCO2 at 20 MPa and0 ◦C onto the surface of BP2000. Subsequently, the precursors

ig. 7. (a) XRD spectra of Pd/BP2000 catalysts prepared by chemical reduction in H2/scCO2

hown in (a).

by chemical reduction in H2/scCO2 at 80 ◦C. (c) Particle size distribution measured

were reduced concurrently in a H2/scCO2 mixture at 27.6 MPa◦

(PH2 = 0.8 MPa) and 80 C. The molar ratio of Pt:Pd was 0.92 in

the prepared catalysts and these showed much better control ofnanoparticle size than in the Pd/BP2000 catalysts. The XRD spec-trum obtained from the Pd–Pt/BP2000 sample is presented in Fig. 9along with the corresponding XRD spectra from the Pd/BP2000 and

at 80 ◦C. (b) The effect of metal loading on the size of the particles from XRD results

88 B. Cangül et al. / J. of Supercritical Fluids 50 (2009) 82–90

Fig. 8. (a) and (b) TEM images of the Pt/BP2000. (c) Particle

Fig. 9. XRD spectrum for Pd–Pt/BP2000 catalyst.

size distribution measured from images such as (b).

Pt/BP2000 samples for comparison. The average metal particle sizeof in the Pd–Pt/BP2000 catalyst calculated from the XRD spectrumusing the Scherrer formula is 9 nm.

TEM images such as Fig. 10a–c obtained from the Pd–Pt/BP2000catalyst samples show metal nanoparticles that are distributedmore homogeneously than those in the Pd/BP2000 catalyst butless homogeneously than those in the Pt/BP2000 catalyst. Simi-larly, the particle size distributions are narrower than those for thePd/BP2000 catalyst but broader than those for the Pt/BP2000 cat-alyst, with the diameters for most of the particles lying between 2and 10 nm (e.g. Fig. 10d). In this case, the mean particle size mea-sured from the TEM images (4.3 ± 2.7 nm) is significantly less thanthat obtained from the XRD data because the particle size distribu-tion is bimodal; the number-averaged size measured from TEM datawill thus be influenced more strongly by the large number of smallparticles than the volume-averaged XRD value. Attempts to mea-sure the chemistry of these particles by EDXS were complicated bythe high particle density in these bimetallic catalysts, which meant

that spectra acquired from a particular particle included contribu-tions from adjacent particles. Despite this, a comparison of manysuch spectra from regions that contain predominantly larger orsmaller particles (e.g. areas such as those shown in Fig. 10b andc, respectively) revealed that the smaller particles are less Pd-rich

B. Cangül et al. / J. of Supercritical Fluids 50 (2009) 82–90 89

io of 0

tti

4

suiatwaPstcplutscfisBi

Fig. 10. (a–c) TEM images of the Pt–Pd/BP2000 with Pt:Pd molar rat

han the larger particles. On this basis it is tempting to speculatehat Pt and Pd particles may form separately and then alloy bynterdiffusion but further work is needed to test this hypothesis.

. Conclusion

The scCO2 deposition technique was used to prepare BP2000upported single Pd, Pt and bimetallic Pd–Pt nanoparticles. The MPssed for the study were Pd(acac)2 and Pt(cod)me2. The adsorption

sotherms were measured at 20 MPa and 60 ◦C for the Pd(acac)2nd Pt(cod)me2 on BP2000 in scCO2. It was shown that the adsorp-ion isotherm for the Pd(acac)2/BP2000-scCO2 system was linearhich was attributed to the low solubility of Pd(acac)2 in scCO2

t the experimental conditions. The adsorption isotherm of thet(cod)me2 on BP2000 in scCO2 at 20 MPa and 60 ◦C was repre-ented by the Langmuir model. The uptake amount of Pd(acac)2 onhe support was higher than that of Pt(cod)me2 at the same fluidoncentration in the supercritical phase indicative of a strongerrecursor-support interaction for Pd(acac)2/BP2000. The metal

oadings of the supports could be thermodynamically controlledsing the adsorption isotherms. A mass transfer model was foundo describe the adsorption kinetics for the Pd(acac)2/BP2000-scCO2ystem at 20 MPa and 60 ◦C. It was found that the Pd precursorould be reduced in a H2/scCO2 mixture. From the XRD data, it was

ound that increasing reduction temperature and Pd metal load-ng increased the Pd particle size. From the TEM images it waseen that Pd particles were not homogeneously distributed on theP2000 support with numerous large particles with sizes rang-

ng from 3 to 100 nm. In contrast to Pd nanoparticles, Pt particles

.92. (d) Particle size distribution measured from images such as (c).

were found to be homogeneously distributed with small particlesizes between 2 and 6 nm on BP2000. Binary metal nanoparticlesof Pd–Pt on BP2000 could also be formed. It was found that addi-tion of Pt increased the homogeneity and reduced the particle sizeon the support compared to single Pd nanoparticles. At the sameexperimental conditions, it was found that supported bimetallicPt–Pd nanoparticle size was 9 nm whereas single Pd size was 14 nmfrom XRD data. From the EDXS results, it was seen that the smallerparticles were less Pd-rich than the larger particles.

References

[1] A.L. Dantas, P.D. Alves, D. Arando, M. Schmal, Characterization of carbon sup-ported Pd catalysts: influence of electronic and particle size effects usingreaction probes, Appl. Catal. 277 (2004) 71–81.

[2] H. Li, G. Sun, N. Li, S. Sun, D. Su, Q. Xin, Design and preparation of highly activePt–Pd/C catalyst for the oxygen reduction reaction, J. Phys. Chem. 111 (2007)5605–5617.

[3] M. Markus, M. Kumar, K. Eklund, H. Sjöholm, S. Munzin, Hydrogenolysis ofhydroxymatairesol over carbon supported palladium catalysts, Catal. Lett. 1103(2005) 125–131.

[4] J. Luo, P.N. Njoki, Y. Lin, D. Mott, L. Wang, C. Zhong, Characterization of carbon-supported AuPt nanoparticles for electrocatalytic methanol oxidation reaction,Langmuir 22 (2006) 2892–2898.

[5] L. Xiong, A. Manthiram, Nanostructured Pt-M/C (M = Fe and Co) catalysts pre-pared by a microemulsion method for oxygen reduction in proton exchangemembrane fuel cells, Electrochim. Acta 50 (2005) 2323.

[6] E. Antolini, J.R.C. Salgado, L.G.R.A. Santos, G. Garcia, E.A. Ticianelli, E. Pastor,E.R. Gonzalez, Carbon supported Pt–Cr alloys as oxygen-reduction catalysts fordirect methanol fuel cells, J. Appl. Electrochem. 36 (2005) 355–362.

[7] C.A. Angelucci, M. Silva, F.C. Nart, Preparation of platinum–ruthenium alloyssupported on carbon by a sonochemical method, Electochemica 52 (2007)7293–7299.

9 rcritica

[

[

[

[

[

[

[

[

[

[

[

0 B. Cangül et al. / J. of Supe

[8] W. Wang, D. Zheng, C. Du, Z. Zou, X. Zhang, B. Xia, H. Yang, D. Akins, Carbonsupported Pd–Co bimetallic nanoparticles as electrocatalysts for the oxygenreduction reaction, J. Power Sources 167 (2007) 243–249.

[9] G. Ren, H. Shi, Y. Xing, Synthesis and composition evolution of bimetallic Pd–Ptalloy nanoparticles, IOP Electron. J.-Nanotechnol. 18 (2007).

10] C.D. Saquing, T.T. Cheng, M. Aindow, C. Erkey, Preparation of platinum/carbonaerogel nanocomposites using a supercritical deposition method, J. Phys. Chem.B 108 (2004) 7716–7722.

11] C.D. Saquing, D. Kang, M. Aindow, C. Erkey, Investigation of the supercriticaldeposition of platinum nanoparticles into carbon aerogels, Micropor. Mesopor.Mater. 80 (2005) 11–23.

12] A. Bayrakceken, U. Kitkamthorn, M. Aindow, C. Erkey, Decoration of multi-walled carbon nanotubes with platinum nanoparticles using supercriticaldeposition with thermodynamic control of metal loading, Scripta Mater. 56(2007) 101–103.

13] A. Bayrakceken, A. Smirnova, U. Kitkamthorn, M. Aindow, L. Türker, I. Eroglu,C. Erkey, Pt-based electrocatalysts for polymer electrolyte membrane fuel cellsprepared by supercritical deposition technique, J. Power Sources 179 (2008)

532–540.

14] Y. Zhang, D. Kang, M. Aindow, C. Erkey, Preparation and characterization ofruthenium/carbon aerogel nanocomposites via a supercritical fluid route, J.Phys. Chem. B 109 (2005) 2617–2624.

15] Y. Zhang, C. Erkey, Preparation of supported metallic nanoparticles using super-critical fluids: a review, J. Supercrit. Fluids 38 (2006) 252–267.

[

[

l Fluids 50 (2009) 82–90

16] X.R. Ye, Y. Lin, C.M. Wai, Decorating catalytic palladium nanoparticles on carbonnanotubes in supercritical carbon dioxide, Chem. Commun. 5 (2003) 642.

[17] X.R. Ye, Y. Lin, C. Wang, M.H. Engelhard, Y. Wang, C.M. Wai, Supercritical fluidsynthesis and characterization of catalytic metal nanoparticles on carbon nan-otubes, J. Mater. Chem. 14 (2004) 908–913.

[18] G.I. Garrido, F.C. Patcas, G. Upper, M. Tuerk, S. Yilmaz, B. Kraushaar-Czarnezki,Supercritical deposition of Pt on SnO2-coated Al2O3 foams: phase behaviourand catalytic performance, Appl. Catal. A Gen. 338 (2008) 58–65.

19] M. Kruk, M. Jaroniec, Y. Bereznitski, Adsorption study of porous structure devel-opment in carbon blacks, J. Colloid Interface Sci. 182 (1996) 282–288.

20] O. Aschenbrenner, S. Kemper, N. Dahmen, K. Schaber, E. Dinjus, Solubilityof �-diketonates, cyclopentadienyls, and cyclooctadiene complexes with var-ious metals in supercritical carbon dioxide, J. Supercrit. Fluids 41 (2007)179–186.

21] Y. Zhang, B. Cangul, Y. Garrabos, C. Erkey, Thermodynamics and kinet-ics of adsorption of bis(2,2,6,6-tetramethyl-3,5-heptanedionato) (1,5-cyclooctadiene) ruthenium (II), J. Supercrit. Fluids 44 (2008) 77.

22] T. Funazukuri, C.Y. Kong, S. Kagei, Impulse response techniques to measure

binary diffusion coefficients under supercritical conditions, J. Chromatogr. A1037 (2004) 411–429.

23] J. Kim, G. Roberts, D.J. Kiserow, Supported Pd catalyst preparation using liquidcarbon dioxide, Chem. Mater. 18 (2006) 4710–4712.

24] Y. Zhang, D. Kang, C. Saquing, M. Aindow, C. Erkey, Supported platinum nanopar-ticles by supercritical deposition, Ind. Eng. Chem. Res. 44 (2005) 4161–4164.