Size-dependent hydrogen sorption in ultrasmall Pd clusters embedded in a mesoporous carbon template

Article abstract of DOI:10.1021/ja101795g

Hydrogen sorption properties of ultrasmall Pd nanoparticles (2.5 nm) embedded in a mesoporous carbon template have been determined and compared to those of the bulk system. Downsizing the Pd particle size introduces significant modifications of the hydrogen sorption properties. The total amount of stored hydrogen is decreased compared to bulk Pd. The hydrogenation of Pd nanoparticles induces a phase transformation from fcc to icosahedral structure, as proven by in situ XRD and EXAFS measurements. This phase transition is not encountered in bulk because the 5-fold symmetry is nontranslational. The kinetics of desorption from hydrogenated Pd nanoparticles is faster than that of bulk, as demonstrated by TDS investigations. Moreover, the presence of Pd nanoparticles embedded in CT strongly affects the desorption from physisorbed hydrogen, which occurs at higher temperature in the hybrid material compared to the pristine carbon template.

Full text of DOI:10.1021/ja101795g

Published on Web 05/13/2010
Size-Dependent Hydrogen Sorption in Ultrasmall Pd Clusters
Embedded in a Mesoporous Carbon Template
Claudia Zlotea,*,† Fermin Cuevas,† Vale´rie Paul-Boncour,† Eric Leroy,†
Philippe Dibandjo,‡ Roger Gadiou,‡ Cathie Vix-Guterl,‡ and Michel Latroche†
Institut de Chimie et des Mate´riaux Paris Est (ICMPE), UMR 7182, CNRS, Thiais, France, and
Institut de Science des Mate´riaux de Mulhouse (IS2M), LRC 7227, CNRS, Mulhouse, France
Received March 3, 2010; E-mail: claudia.zlotea@icmpe.cnrs.fr
Abstract: Hydrogen sorption properties of ultrasmall Pd nanoparticles (2.5 nm) embedded in a mesoporous
carbon template have been determined and compared to those of the bulk system. Downsizing the Pd
particle size introduces significant modifications of the hydrogen sorption properties. The total amount of
stored hydrogen is decreased compared to bulk Pd. The hydrogenation of Pd nanoparticles induces a
phase transformation from fcc to icosahedral structure, as proven by in situ XRD and EXAFS measurements.
This phase transition is not encountered in bulk because the 5-fold symmetry is nontranslational. The kinetics
of desorption from hydrogenated Pd nanoparticles is faster than that of bulk, as demonstrated by TDS
investigations. Moreover, the presence of Pd nanoparticles embedded in CT strongly affects the desorption
from physisorbed hydrogen, which occurs at higher temperature in the hybrid material compared to the
pristine carbon template.
Introduction
A major problem for the study and the use of nanosized
particles is related to the instability at high temperature and the
tendency for coalescence into larger aggregates. Different types
of stabilizers (electrostatic or steric) are currently used to
overcome this drawback.7 Another possible solution is to embed
metal clusters into the porosity of an ordered porous material
with controlled pore size distribution. The well-defined pore
dimension controls the dispersion of the clusters size distribution.
The resulting material is a hybrid that combines the properties
of each component: the metal nanoparticles and the porous
structure. The advantage of this hybrid material as a hydrogen
store medium is the combined gas sorption: adsorption on the
surface of the pores of the porous material and hydrogenation
of the metal clusters by forming hydrides. Moreover, the
presence of metal clusters inside the pores is considered to
increase the hydrogen uptake of different porous materials via
the spillover effect (catalyzed formation of monatomic hydrogen
followed by migration to the surface of the support).8
In the present work, hydrogen sorption properties of a hybrid
material formed by Pd nanoparticles embedded into an ordered
mesoporous carbon material are reported and discussed. The
precursor and the hybrid material have been synthesized
following already published methods.9-14 The structural, mi-
crostructural, and textural properties of the carbon precursor and
the hybrid material have been determined. In situ X-ray, neutron
The downscaling and control of the metal particle sizes have
become an important issue for the design of new solid-state
hydrogen storage materials.1-4 Downsizing the metal particles
to few nanometers can introduce fundamental changes compared
to the corresponding bulk state and therefore may overcome
the drawbacks encountered in the latter materials. For small
particles the ratio of surface to bulk atoms becomes important
and surface effects may influence the total energy and modify
the fundamental properties such as thermodynamics and kinetics
of the hydrogen-metal interactions. Materials with small crystal
and particle size have improved kinetics of hydrogen sorption.3
Previously, theoretical calculations have proven that the ther-
modynamic properties of the Mg hydride can be destabilized
when the metal clusters become smaller than 1.3 nm.4
Another important aspect of downsizing the metal particle
size is related to their crystalline structure. Below a critical size
for fcc metal clusters, a complete rearrangement of atoms may
occur into new structures (icosahedral, dodecahedral) that
possess crystal symmetry properties completely different from
those of the bulklike systems.5,6 Both icosahedral and dodeca-
hedral structures have a 5-fold axis of rotation that is a
nontranslational symmetry.
† Institut de Chimie et des Mate´riaux Paris Est (ICMPE).
‡ Institut de Science des Mate´riaux de Mulhouse (IS2M).
(1) Dornheim, M.; Eigen, N.; Barkhordarian, G.; Klassen, T.; Bormann,
R. AdV. Eng. Mater. 2006, 8, 377.
(7) Ott, L. S.; Finke, R. G. Coord. Chem. ReV. 2007, 1075.
(8) Wang, L.; Yang, R. T. Energy EnViron. Sci. 2008, 1, 268.
(9) Ryoo, R.; Joo, S. H.; Jun, S. J. Phys. Chem. B 1999, 103, 7743.
(10) Ehrburger-Dolle, F.; Morfin, I.; Geissler, E.; Bley, F.; Livet, F.; Vix-
Guterl, C. Langmuir 2003, 19, 4303.
(2) Fichtner, M. Nanotechnology 2009, 20, 204009.
(3) Zaluska, A.; Zaluski, L.; Stro¨m-Olsen, J. O. J. Alloys Compd. 1999,
288, 217.
(11) Vix-Guterl, C.; Boulard, S.; Parmentier, J.; Werckmann, J.; Patarin,
J. Chem. Lett. 2002, 10, 1062.
(4) Wagemans, R. W. P.; van Lenthe, J. H.; de Jongh, P. E.; van Dillen,
A. J.; de Jong, K. P. J. Am. Chem. Soc. 2005, 127, 16675.
(5) Reinhard, D.; Hall, B. D.; Berthoud, P.; Valkealahti, S.; Monot, R.
Phys. ReV. Lett. 1997, 79, 1459.
(12) Zhao, D.; Huo, Q.; Stuky, G. D.; Feng, J.; Melosh, N.; Fredrickson,
G. H. Science 1998, 279, 548.
(13) Campesi, R.; Cuevas, F.; Gadiou, R.; Leroy, E.; Hirscher, M.; Vix-
Guterl, C.; Latroche, M. Carbon 2008, 46, 206.
(6) Marks, L. D. Rep. Prog. Phys. 1994, 57, 603.
9
7720 J. AM. CHEM. SOC. 2010, 132, 7720–7729
10.1021/ja101795g  2010 American Chemical Society

Hydrogen Sorption in Pd Clusters
A R T I C L E S
volumetric device (Sievert’s method) equipped with calibrated and
thermostated volumes and pressure gauges. The samples were
enclosed in a stainless steel sample holder closed with a metal seal.
Before any sorption measurements, the samples were outgassed
under primary vacuum at 473 K for 10 h. The sample holder is
immersed in a liquid N2 Dewar at 77 K or in a thermostated water
bath maintained at 298 K, and high purity hydrogen (6 N) is
introduced step by step up to 9 MPa. The pressure variations due
to both gas cooling and hydrogen sorption are measured after
reaching thermodynamic equilibrium, usually in the range of
minutes. The real equation of state for hydrogen gas was used from
the program GASPAK V3.32.15 The PCI curves were measured
twice (i.e., two full adsorption-desorption cycles) to check the
hysteresis effect and the measurement repeatability. One additional
PCI curve at room temperature and H2 pressure up to 0.4 MPa has
been measured with the help of an automated volumetric instrument
(PCTPro) to confirm the manually recorded isotherm for CT/nPd.
Good agreement is noticed between the two types of measurements.
All capacities reported here in wt % are determined with respect
to the hydrogen loaded material after correction of the sample
weight loss by outgassing. Sample volume correction is derived
from density measurement obtained with a helium AccuPyc 1330
Micromeritics pycnometer. By use of this method, the measured
diffraction, and X-ray absorption experiments have been
performed to characterize the hydrogen sorption properties of
the hybrid material. A thorough comparison between the
hydrogen sorption properties of the hybrid material, the carbon
precursor, and bulk Pd will be addressed and will clarify the
size dependent properties of the Pd clusters.
Experimental Details
The carbon template (CT) was obtained by a replica method of
the SBA-15 silica, as described elsewhere.9-11 The SBA-15 silica
was prepared according to the synthesis procedure already presented
in the literature.12 It displays cylindrical pores with a diameter close
to 5 nm and a mean wall width of ∼4 nm. Carbon was deposited
into the SBA-15 porosity by chemical vapor deposition (CVD) of
a mixture of 3 vol % of propylene and argon at 1023 K. The total
amount of carbon introduced in the SBA-15 porosity is 37 wt %.
To remove the silica template, the mixed material (silica/carbon)
was stirred for 4 h in a HF solution (40% volume concentration)
before being filtered and washed with distilled water and finally
dried at 80 °C overnight. The as-synthesized CT displays a
hexagonal arrangement of carbon nanorods (4.5 nm) and a very
well organized mesoporosity. The total mesoporous volume is 1.1
m3/g with a narrow pore size distribution of around 5 nm.
densities of CT and CT/nPd are 1.8 ( 0.2 and 2.5 ( 0.1 g cm-3
,
respectively.
To generate in situ Pd nanoparticles (nPd) into the CT porosity,
a chemical wetting procedure was used, as described elsewhere.13,14
A solution of tetrachloropalladous acid (H2PdCl4) was prepared by
mixing 0.3353 g of PdCl2 (Alfa Caesar, purity 99.9%) into 20 mL
of 10% vol HCl aqueous solution under stirring at 300 K until
complete dissolution. The CT in powder form was then impregnated
twice with the tetrachloropalladinic acid solution for a final Pd
concentration of around 20 wt %. The mixture of powder and
The local structure of the nPd was investigated by X-ray
absorption spectroscopy (XAS) in transmission mode on SAMBA
beamline at the SOLEIL synchrotron. The X-ray absorption spectra
at the K edge of Pd were measured under vacuum at 77 and 298 K
and under hydrogen atmosphere (80 kPa) at 298 K. Bulk Pd powder
(<100 µm) was also measured under similar conditions and used
as reference. The powder samples were mixed with a polymer
(PTFE) to ensure the mechanical cohesion. The measurements were
performed in a special cell connected to a gas sorption device that
allows careful control of the hydrogen pressure.16 The XAS data
treatment and the EXAFS modeling and fit were performed by the
help of the MAX program package.17 The FEFF8 program was
used to obtain the theoretical phase shifts and amplitudes of the
Pd-Pd(O) shells.18 The photoelectron mean free path was deter-
mined empirically. The refined parameters are the coordination
number (N), the Debye-Waller factor (σ2), the distance between
the shells (R), and the energy shift (∆E0). A single Pd-Pd shell
model was successfully used for almost all spectra, excluding the
nPd under vacuum at 298 K, where a two-shell (Pd-O and Pd-Pd)
model has to be taken into account because of the presence of
oxygen. The energy shift ∆E0 was close to zero for all refinements.
2-
solution was stirred for 3 h and dried in air at 330 K. The PdCl4
ions were reduced by heating the impregnated CT in a Ar/H2 flow
(0.5 L min-1) at 573 K for 6 h. The sample was then outgassed
under secondary vacuum for 11 h at 573 K.
The chemical composition of the impregnated material has been
determined by an inductively coupled plasma optical emission
spectrometer (ICP-OES). The measured Pd content is 17.2 wt %,
close to the nominal composition. The hybrid material will further
be named CT/nPd.
Microstructural analyses were performed by transmission electron
microscopy (TEM-Tecnai F20 with a field emission gun of 200
kV, punctual resolution of 0.24 nm, and energy filtering GIF), and
the chemical composition was determined by energy dispersion
spectroscopy (EDS).
Neutron powder diffraction (NPD) experiments were performed
on D1B instrument at the Institute Laue Langevin. This instrument
can provide high neutron flux with a wavelength of 2.52 Å and is
suitable for in situ study under different gas pressure and temper-
ature conditions. In situ neutron diffraction was measured during
the thermal D2 desorption with the help of a special sample holder.
The silica sample holder can be used in a wide range of temperature
and pressure conditions. Deuterium isotope has been chosen instead
of hydrogen because of its lower incoherent scattering cross section.
Bulk Pd powder and pristine CT were also measured under
comparable gas pressure and temperature conditions and used as
references. The thermal desorption spectra (TDS) were recorded
simultaneously to neutron diffraction patterns as the total pressure
under dynamic high vacuum while applying a constant temperature
The textural properties were determined from nitrogen adsorp-
tion/desorption isotherms measured with a Micromeritics ASAP
2020 instrument. The specific surface area was obtained by the
Brunauer-Emmett-Teller (BET) method. The total pore volume
was computed from the amount of gas adsorbed at p/p0 ) 0.99,
and the micropores volume was calculated using the Dubinin-
Radushkevich equation in the relative pressure range 10-4-10-2
.
Structural characterizations were performed by X-ray diffraction
(XRD) using a Bruker D8 Advance instrument (Cu KR, Bragg-
Brentano geometry). To characterize the crystallographic changes
induced by hydrogen absorption in CT/nPd, in situ XRD measure-
ments were performed by stepwise increase of hydrogen pressures
at ambient temperature. The sample was initially evacuated under
primary vacuum, and hydrogen pressure was incremented to 10
kPa. As reference, in situ XRD measurements have been performed
on bulk Pd (bPd) powder (<100 µm) under comparable conditions.
Hydrogen sorption properties were determined by measuring the
pressure-composition isotherms (PCI) at 77 and 298 K to 9 MPa
hydrogen pressure. The PCI curves were recorded using a manual
(15) Lemmon, E. W.; Peskin, A. P.; McLinden, M. O.; Friend, D. G.
NIST12 Thermodynamic and Transport Properties of Pure Fluids,
version 5.0; NIST: Gaithersburg, MD, 2000.
(16) Paul-Boncour, V.; Joubert, J.-M.; Latroche, M.; Percheron-Gue´gan,
A. J. Alloys Compd. 2002, 330-332, 246.
(17) Michalowicz, A.; Moscovici, J.; Muller-Bouvet, D.; Provost, K. J.
Phys.: Conf. Ser. 2009, 190, 012034.
(14) Campesi, R.; Cuevas, F.; Leroy, E.; Hirscher, M.; Gadiou, R.; Vix-
Guterl, C.; Latroche, M. Microporous Mesoporous Mater. 2009, 117,
511.
(18) Ankudinov, A. L.; Ravel, B.; Rehr, J. J.; Conradson, S. D. Phys. ReV.
B 1998, 58, 7565.
9
J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010 7721

A R T I C L E S
Zlotea et al.
Table 1. Textural and Hydrogen Sorption Properties of CT and
CT/nPda
properties
CT
CT/nPd
BET surface area (m2/g)
V0.99 (cm3/g)
960 ( 10
1.10 ( 0.1
0.2 ( 0.1
1.80 ( 0.1
660 ( 7 (795)
0.75 ( 0.1 (0.90)
0.3 ( 0.1 (0.3)
1.45 ( 0.1
Vµp (cm3/g)
excess hydrogen capacity,
4 MPa, 77 K (wt %)
excess hydrogen capacity,
4 MPa, 298 K (wt %)
0.13 ( 0.02
0.23 ( 0.02
a The values of BET surface area, total porous volume (V0.99), and
microporous volume (Vµp) are given by gram of sample and in
parentheses by gram of carbon.
rate of 0.42 K min-1 from 22 to 310 K. The samples were
previously loaded at room temperature with D2 gas with the help
of a Sieverts apparatus. The CT and CT/nPd were loaded with the
same amount of dosed D2. The bulk deuteride (bPdD0.6) was formed
at room temperature by deuterium uptake at 0.05 MPa pressure.
Afterward, all samples have been cooled under D2 atmosphere to
22 K. The remaining amount of gas that has been not adsorbed
(absorbed) by the sample was removed by pumping.
Results
1. Microstructural, Textural, and Hydrogen Sorption Pro-
perties. The structural and microstructural investigations of CT/
nPd have proven that the hybrid material is formed by Pd
nanoparticles (nPd) well dispersed in the porous structure of the
CT, as recently published by Campesi et al.13 The nPd adopts a
fcc structure with the lattice parameter a ) 3.93 ( 0.02 Å
comparable to the cubic crystal structure of bulk Pd (bPd).19 The
mean size of the nPd is 2.5 ( 0.5 nm, as determined by both XRD
(Scherrer’s formula) and TEM analysis. The textural properties of
CT and CT/nPd at 77K are listed in Table 1. The decrease of
specific surface area and total porous volume of CT/nPd hybrid
compared to the precursor CT can be explained by the increase of
the total weight (by 17.2%) together with the blocking of accessible
pores by insertion of nPd. The micropore volume does not change
significantely after the formation of the nPd, proving that nano-
particles are located in the mesopores of the CT material. The
change of textural properties has consequences on the hydrogen
storage properties, as discussed later.
Figure 1. PCI curves of CT (circle) and CT/nPd (square) recorded at 77
K (a) and 298 K (b). Full and empty symbols stand for ad/absorption and
desorption, respectively.
The hydrogen sorption properties have been determined by
measuring the PCI curves at 77 and 298 K (Figure 1). The Gibbs
excess hydrogen adsorption capacities of CT and CT/nPd at 77
and 298 K at 4 MPa are given in Table 1. At 77 K, the excess
sorption isotherms of both samples display a type I trend typical
for monolayer adsorption on porous materials (Figure 1a). The
hydrogen capacity of pristine CT is in good agreement with
previously reported values for porous materials with similar
surface area.20,21 At room temperature, the PCI curve of CT
(Figure 1b) displays a linear shape typical for diluted monolayer
adsorption (Henry type), as commonly reported for porous
materials at this temperature. The isotherm of CT/nPd shows a
steep rise at low pressure that can be attributed to the
hydrogenation of nPd clusters and formation of nPdHx hydride.
For higher pressure, a linear increase of the hydrogen uptake is
observed that is ascribed to H2 physisorption on the carbon
porosity. The desorption branch of the PCI curve of CT/nPd
shows an irreversible character at very low pressure.
Figure 2. Absorption PCI of CT/nPd at 298 K measured to 100 kPa (full
circles). For comparison, the absorption PCI of bPd is also plotted (line).
Hydrogenation of nPd has been measured at low pressure
and room temperature (where the hydrogen adsorption by CT
is negligible) and compared to bPd (Figure 2). Under hydrogen
pressure below 1 kPa, the nPd forms a solid solution of hydrogen
in Pd (R phase) followed by the formation of the hydride nPdHx
(ꢀ phase), as verified by the presence of a sloping plateau
pressure. The formation of the hydride nPdHx is complete at 5
(19) Bidwell, L. R.; Speiser, R. Acta Crystallogr. 1964, 17, 1473.
(20) Thomas, M. K. Catal. Today 2007, 120, 389.
(21) Stro¨bel, R.; Garche, J.; Moseley, P. T.; Jo¨rissen, L.; Wolf, G. J. Power
Sources 2006, 159, 781.
9
7722 J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010

Hydrogen Sorption in Pd Clusters
A R T I C L E S
Figure 3. In situ XRD patterns during hydrogenation of nPd at 290 K.
The applied H2 pressure is indicated. The diffraction peak at 74° (/) belongs
to the sample holder.
Figure 4. Comparison of the (220) diffraction peak for the metal and
hydride phases of nPd (top) and bPd (bottom). The diffraction patterns of
nPd correspond to those measured at 0 and 10 kPa hydrogen pressure in
Figure 3.
kPa hydrogen pressure. The plateau pressure of nPd appears in
the same pressure region as bPd, although it is sloped compared
to bPd. Moreover, the length of the plateau pressure of nPd is
reduced compared to bPd.
2. In Situ Hydrogen Sorption. To give a comprehensive
picture of the hydrogen sorption mechanism in CT/nPd hybrid,
in situ hydrogen sorption investigations have been performed
using XRD, XAS, and NPD. Systematic comparisons with bPd
and pristine CT have been performed under similar pressure
and temperature conditions.
Table 2. Structural Properties of nPd, nPdHx, bPd, bPdH0.6: Lattice
Parameter (a), Relative Lattice Expansion (∆a/a), Integrated Area
after Background Correction (I), and Full Width at Half Maximum
(fwhm) of the (220) Bragg Peaka
sample
a (Å)
∆a/a (%)
I (220) (a.u.)
fwhm (220) (deg)
nPd
nPdHx
bPd
3.93 ( 0.02
4.00 ( 0.02
3.884 ( 0.005
4.025 ( 0.005
1
1.07
1
2.85
2.19
0.53
0.70
1.8
3.6
bPdH0.6
1.17
The hydrogenation of the nPd of the hybrid material was
monitored by in situ XRD during stepwise increase of H2
pressure to 10 kPa at 290 K (Figure 3). The most intense (111)
and (200) diffraction peaks of nPd overlap with a broad
diffraction contribution from the CT.13 For the sake of the
clarity, the angular domain between 60° and 85° was chosen
for in situ measurements because it contains two distinct
diffraction peaks, indexed as (220) and (311) in the fcc structure,
and no carbon contribution. Moreover, in this angular domain
an extra diffraction peak belonging to the sample holder appears
at around 74° and serves as a reference. The diffraction peaks
of nPd shift to lower angle with increasing H2 pressure, in
agreement with the volume expansion due to the hydrogenation.
Below 1 kPa, a small shift of the Bragg peaks is noticed and
corresponds to the formation of the R phase (fcc). Higher H2
pressure (above 1.2 kPa) induces both a shift of the Bragg peak
and an increase of the peak height. This indicates the start of
the formation of the hydride nPdHx (the ꢀ phase). Because of
the strong overlap between the diffraction peaks of R and ꢀ
phases, a clear determination of the R-ꢀ miscibility limits
cannot be done by this method. For pressure higher than 5 kPa,
the formation of the nPdHx hydride is completed. These results
are in good agreement with the isotherm of nPd measured at
room temperature (Figure 2).
a The integrated area of the hydride phases are normalized to the
corresponding intensity of the metallic diffraction peak.
case of bPd than nPd, indicating different volume expansion.
The lattice parameters of nPd, nPdHx, bPd, and bPdH0.6 and
the relative lattice expansion due to hydrogenation are listed in
Table 2 together with the intensity (integrated area after
background correction) and full width at half-maximum (fwhm)
of the (220) peak (from Figure 4). The lattice expansion induced
by hydrogenation is larger in bPd compared to nPd. The peak
intensities of both hydride phases are slightly larger than the
corresponding metal states. However, the variation of the fwhm
is different; the nPdHx shows a narrowing of the fwhm relative
to nPd, while the fwhm increases from bPd to bPdH0.6. This
feature of nPd is completely reversible with hydrogenation/
dehydrogenation cycling.
In situ XAS investigations have been performed at the Pd K
edge to study the nPd and bPd in both metallic (under vacuum)
and hydride states (under 80 kPa hydrogen pressure) at 298 K.
The EXAFS spectra and the module of the Fourier transform
(FT) are shown in Figures 5 and 6 for nPd and bPd, respectively.
The FT peaks of the hydride phase are shifted to larger distances
compared to the metallic state, irrespective of the nature of Pd
(bulk and nanoparticles). A larger increase of the distance is
noticed for bPdH0.6 than for the nPdHx. The intensities of all
FT peaks decrease from metal to hydride state in the case of
bPd, whereas an opposite effect occurs for nPd. The FT of the
metallic nPd (Figure 5) shows that near the main peak due to
Pd-Pd backscattering, a smaller peak is present at lower
distance and can be accounted by the Pd-O backscattering, in
good agreement with previous EXAFS study of reduction and
Since the thermodynamics of hydrogenation of bPd is well
documented,22 a comparison between in situ XRD measurements
of hydrogenation of nPd and bPd (powder) was performed under
similar conditions. The (220) diffraction peak of nPd (at 0 kPa)
and nPdHx (at 10 kPa) from Figure 3 are replotted in Figure 4
and compared to those from bPd and its hydride bPdH0.6. The
peak shift due to the formation of the hydride is larger in the
(22) Frieske, H.; Wicke, E. Ber. Bunsen-Ges. Phys. Chem. 1973, 77, 48.
9
J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010 7723

A R T I C L E S
Zlotea et al.
Figure 5. Experimental EXAFS spectra (a) and corresponding module of
the Fourier transform (b) of nPd in metallic (line) and hydride (dashed line)
states at 298 K.
Figure 6. Experimental EXAFS spectra (a) and corresponding module of
the Fourier transform (b) of bPd in metallic (line) and hydride (dashed line)
states at 298 K.
Table 3. Refined Parameters of the EXAFS Spectra of nPd,
nPdHx, and bPd, bPdH0.6 Samples: Distance of the Nearest
Neighbors (R), Coordination Number (N), Debye-Waller Factor
(σ2), and Confidence Factor of the Fit (QF)
reoxidation of Pd nanoparticles.23,24 In contrast, the FT of the
corresponding hydride nPdHx has only one peak due to the
Pd-Pd backscattering with enlarged amplitude compared to
metal. The Pd-O backscattering peak disappears upon hydro-
genation, indicating that a reduction of the oxide layer occurs
while exposure to hydrogen. The oxidation state of nPd was
also checked by probing a completely different part of the same
sample at 77 K. The low temperature reduces the thermal
contribution of the Debye-Waller factor, and therefore, the
structural model can be undoubtedly determined. Surprisingly,
the FT at 77 K has only a single peak due to the first Pd-Pd
shell, indicating that this part of the sample is oxygen-free.
Moreover, a comparison of the related XANES spectra (from
the two different positions of the sample) clearly proves that
the part of the sample investigated at 298 K is slightly oxidized,
since the ratio of the first two oscillations after the absorption
edge is changed respective to the 77 K spectrum (Figure 1 of
the Supporting Information). This suggests that the as-prepared
sample is not homogeneous with respect to oxygen. Some parts
are slightly oxidized and subsequently reduced in hydrogen
atmosphere. This might indicate that the oxygen is chemisorbed
at the surface of nPd. This is not the case for bPd, which shows
a single peak in the FT for both metal and hydride (Figure 6).
sample
T (K)
shell
R (Å)
N
σ2 (Å2)
QF
nPd
77 Pd-Pd 2.75 ( 0.01 7.5 ( 0.6 0.004 ( 0.001 0.1
298 Pd-Pd 2.74 ( 0.01 6.4 ( 0.8 0.008 ( 0.001 0.1
298 Pd-O 1.97 ( 0.01 1.5 ( 0.3 0.004 ( 0.002
298 Pd-Pd 2.79 ( 0.01 9.0 ( 0.8 0.008 ( 0.001 0.1
77 Pd-Pd 2.74 ( 0.01 12 (fixed) 0.003 ( 0.001 0.1
298 Pd-Pd 2.74 ( 0.01 12 (fixed) 0.006 ( 0.002 0.5
nPdHx
bPd
bPdH0.6 298 Pd-Pd 2.84 ( 0.01 11.5 ( 0.9 0.008 ( 0.001 0.1
The first atomic shell was filtered between 1 and 3 Å and
refined (Figures 2 and 3 of the Supporting Information) with
one Pd-Pd shell for bPd, bPdH0.6, nPd (at 77 K), and nPdHx
and with two shells (RPd-Pd ) 2.74 Å and RPd-O ) 1.96 Å) in
the case of nPd (at 298 K). The refined structural parameters
are listed in Table 3, showing the distance of the nearest
neighbors (R), the coordination number (N), and the Debye-
Waller factor (σ2) together with the confidence factor of the
refinement (QF).
The distances of the nearest Pd-Pd neighbors, as determined
by EXAFS analysis, show good agreement with XRD data for
bPd and bPdH0.6 (a/ꢀ2 in Table 2), whereas it is 0.04 Å smaller
for nPd (2.74 Å vs 2.78 Å) and nPdHx (2.79 Å vs 2.83 Å). This
systematic difference between the nearest Pd-Pd distances
determined by EXAFS and XRD in the case of nanoparticles
might be explained by surface strain effects affecting this
(23) Caravati, M.; Meier, D. M.; Grundwaldt, J.-D.; Baiker, A. J. Catal.
2006, 240, 136.
(24) Grunwaldt, J.-D.; Caravati, M.; Baiker, A. J. Phys. Chem B 2006,
110, 25586.
9
7724 J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010

Hydrogen Sorption in Pd Clusters
A R T I C L E S
The in situ NPD of CT/nPd during D2 desorption (not shown
here) shows only broad diffraction peaks arising mainly from
the CT component, in agreement with its weak structural order,
but no diffraction peaks from nPd are observed. This is due to
the low amount of nPd (17 wt %), its small particle size, and
the very close scattering lengths of Pd and C (0.6 × 10-12 and
0.665 × 10-12 cm for C and Pd, respectively30). The neutron
diffraction patterns from CT are very similar to those of CT/
nPd, since mainly CT is contributing to diffraction. In both cases,
obvious differences between the diffraction patterns recorded
at full D2 loading and complete unloading (Figure 8a) are
noticed. The pattern difference clearly shows the existence of a
peak at around 2θ ) 45°, or Q ≈ 1.9 Å-1 (λ ) 2.52 Å).
Moreover, the difference between the diffraction patterns evolves
with temperature (Figure 8b). The position of the peak does
not change with temperature, but its area decreases with
temperature. The integrated area of this peak (after background
correction and normalization relative to the highest value) is
plotted as a function of temperature in Figure 8c together with
the thermal evolution of the D2 coverage, as extracted from the
TDS spectrum. A good agreement is noticed, which indicates
that the difference peak can be directly related to the surface
coverage of D2 molecules on the porous material. In a similar
way, the difference of diffraction patterns of CT/nPd shows the
presence of a peak at Q ≈ 1.9 Å-1, whose area and temperature
evolution correlate with the thermal progress of the D2 coverage,
as extracted from TDS signal (not shown here).
Figure 7. Comparison of the TDS signals from CT/nPd (line), CT (full
square), and bPd (empty circle) recorded between 22 and 310 K with a
constant temperature rate of 0.42 K min-1. For the sake of clarity, above
100 K, both CT/nPd curve (line) and bPd spectra (empty circle) were
normalized to the weight of Pd and then multiplied by 3.
parameter. The hydrogenation induces an expansion of RPd-Pd
distances in both hydrides; however, this is larger in bPd than
in nPd, in agreement with in situ XRD results (Table 2).
The Pd coordination number of the first shell in oxygen-free
samples is reduced from 12 in bPd (fcc structure) to 7.5 and
9.0 in the case of nPd and nPdHx, respectively, in agreement
with the small particle size.25 In the case of slightly oxidized
nPd (at 298 K), N decreases to 6.4 compared to 7.5 for oxygen-
free nPd (at 77 K). This feature is commonly observed during
oxidation and reduction of metallic particles.26
Discussion
The present hybrid material CT/nPd is formed by ultrasmall
Pd nanoparticles (2.5 nm) embedded in the mesopores of an
ordered carbon template. The insertion of nPd in the porosity
of the CT reduces the textural properties of hybrid material
(Table 1). This decrease supports the observed loss of excess
hydrogen sorption at 77 K in CT/nPd relative to CT (Figure
1a). The H2 sorption mechanism on CT/nPd at low temperature
is governed by molecular physisorption on the surface of the
porous material. At ambient temperature, the excess hydrogen
capacities of both CT and CT/nPd are small (below 0.5 wt %).
The CT/nPd shows higher capacity than CT (Figure 1b and
Table 1), and this can be explained by the hydrogenation of
nPd and the formation of nPdHx that is kinetically hindered at
77 K. For pressure up to 0.1 MPa, the adsorption of H2 on CT
is almost negligible (Figure 1b). Therefore, the hydrogen uptake
of CT/nPd measured at 0.1 MPa can be almost entirely attributed
to the formation of nPdHx. Under this assumption, the calculated
H/Pd ratio for nPdHx is approximately 0.6 (details of the
calculation are given in the Supporting Information), a value
higher than the commonly reported 0.3-0.4 for the hydride
formation of Pd nanoparticles under similar temperature and
pressure conditions.31-35 A cooperative effect between the metal
nanoparticles and the porous host might take place leading to
The Debye-Waller factor increases from 77 to 298 K because
of amplified thermal vibrations. It also rises from bPd to bPdH0.6,
in agreement with the introduction of defects by hydrogenation,
as already observed by in situ XRD and in previous EXAFS
studies.27,28 The Debye-Waller factor of the first Pd-Pd layer
remains unchanged for nPd and nPdHx at 298 K.
Several FT peaks are observed at distances larger than 3 Å
and can be attributed to simple and multiple-scattering on further
Pd atoms (Figures 5 and 6). These contributions were simulated
for bPd and bPdHx with the help of FEFF8 with additional
Debye-Waller factors those values increase with distance. The
simulation of the EXAFS spectra by taking into account the
presence of the interstitial H atoms in the metallic structure has
proven no additional multiple scattering effects compared to
Pd-Pd backscattering.
Thermal desorption of deuterium from deuterated CT/nPd,
CT, and bPd has been studied simultaneously with in situ
neutron powder diffraction. A comparison between the three
TDS spectra is shown in Figure 7. The desorption of physisorbed
D2 in CT/nPd occurs at higher temperature than that of CT.
The deuterium desorption from the fully deuterated nPd has a
double-peak shape like that of bPd, in good agreement with
the previous results from bPd.29 However, deuterium is desorbed
at lower temperature in nPd than in bPd.
(30) Bacon, G. E. Neutron Diffraction, 3rd ed.; Clarendon Press: Oxford,
U.K., 1975.
(31) Zu¨tell, A.; Nu¨tzenadel, Ch.; Schmid, G.; Emmenegger, Ch.; Sudan,
P.; Schlapbach, L. Appl. Surf. Sci. 2000, 162-163, 571.
(32) Narehood, D. G.; Kishore, S.; Goto, H.; Adair, J. H.; Nelson, J. A.;
Gutierrez, H. R.; Eklund, P. C. Int. J. Hydrogen Energy 2009, 34,
952.
(25) Frenkel, A. I.; Hills, C. W.; Nuzzo, R. G. J. Phys. Chem. B 2001,
105, 12689.
(26) Newton, M.; Belver-Coldeira, C.; Matrinez-Arias, A.; Fernandez-
Garcia, M. Nat. Mater. 2007, 6, 528.
(27) Rose, A.; Maniguet, S.; Mathew, R. J.; Slater, C.; Yao, J.; Russell,
A. E. Phys. Chem. Chem. Phys. 2003, 5, 3220.
(33) Pundt, A.; Kirchheim, R. Annu. ReV. Mater. Res. 2006, 36, 555.
(34) Zu¨tell, A.; Nu¨tzenadel, Ch.; Schmid, G.; Chartouni, D.; Schlapbach,
L. J. Alloys Compd. 1999, 293-295, 472.
(28) Davis, R. J.; Landry, S. M.; Horsley, J. A.; Boudart, M. Phys. ReV. B
1989, 39, 10580.
(35) Suleiman, M.; Faupel, J.; Borchers, C.; Krebs, H.-U.; Kirchheim, R.;
Pundt, A. J. Alloys Compd. 2005, 404-406, 523.
(29) Stern, A.; Resnik, A.; Shaltiel, D. J. Phys. F: Met. Phys. 1984, 14, 1625.
9
J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010 7725

A R T I C L E S
Zlotea et al.
compared to a previous report38 and does not lead to significant
improvement of the storage properties.
The hydrogenation isotherms recorded at low pressure and
298 K for nPd and bPd (Figure 2) show that thermodynamic
changes introduced by particle downsizing are not encountered.
The pressure plateaus of both nPd and bPd are in the same range,
although the plateau of nPd is slightly sloped. The origin of
the sloped pressure plateau is still an open question and has
been previously accounted for by a broad size distribution of
cluster sizes or mechanical stress induced by the stabilizing
matrix.33,35 In the present case, the size distribution of nPd is
narrow (1.0-3.5 nm) as determined by TEM, which might
indicate different origin of the sloping plateau. Moreover, the
maximum hydrogen solubility in the R phase increases and the
minimum solubility in the ꢀ phase decreases in nPd relative to
bPd. Subsequently, the R-ꢀ miscibility region in nPd (0.05-0.3
H/Pd) is lower than in bulk Pd (0.025-0.65 H/Pd), in good
agreement with previous results on small Pd clusters.33 There-
fore, the maximum amount of hydrogen absorbed in Pd
nanoparticles is lower than in bulk.
Results from in situ XRD investigations during hydrogenation
of nPd (Figure 3) are in good agreement with the isotherm
measurement at 298 K (Figure 2). The hydrogenation and
formation of the R and ꢀ phases in nPd occur at the same
pressure for both methods, confirming the validity of in situ
measurements. The advantage of in situ XRD is related to the
structural information that can be obtained on the R-ꢀ phase
transition in nPd.
In situ XRD reveals important differences between nPd and
bPd upon hydrogenation (Figure 4 and Table 2). The lattice
parameter of metallic nPd is slightly larger than that of bPd
and is in good agreement with previously published values.19
This dilation has been already reported for very small Pd
nanoparticles (2-3 nm).32,35 Upon hydrogenation, the relative
lattice expansion ∆a/a is larger in bPd compared to nPd (Table
2), pointing to a larger volume expansion introduced by
hydrogenation in bPd. This difference can be explained by a
lower hydrogen content than can be accommodated in the nPd
compared to bPd, in agreement with the low pressure isotherm
measurements at 298 K (Figure 2) and previous studies.31-34
Another important difference between nPd and bPd is related
to the change of the fwhm of the (220) Bragg peak from metal
to hydride (Figure 4 and Table 2). In the case of bPd, both the
peak intensity and the fwhm slightly increase from metal to
hydride. This is due to the formation of a large number of defects
by hydrogen insertion in the metal structure. In contrast, the
fwhm in the case of nPd narrows and its intensity rises when
forming the hydride phase. This feature is unusual and might
be related to an atomic rearrangement of clusters upon hydrogen
uptake.
Similar XRD results have been previously reported by
Kaszkur39 for 5.3 nm Pd nanoparticles. The increase of
diffracted intensity and the narrowing of diffraction peaks have
been attributed to an enhancement of crystal ordering in the Pd
cluster upon hydrogenation. In contrast with our results, Pundt
et al. have previously reported that hydrogenation on quasi-
free Pd clusters (5 nm) is accompanied by a drop of the
diffracted intensity and an increase of the width of the Bragg
peaks.40 This observation was performed by in situ synchrotron
Figure 8. Neutron diffraction patterns of the fully D2 loaded (22 K) and
completely unloaded CT (298 K) together with the difference pattern (a),
the thermal evolution of the difference pattern (b), and the thermal evolution
of the D2 coverage as measured by TDS (solid line) and the integral of the
difference neutron diffraction patterns (solid square) (c).
an enhancement in the hydrogen uptake, as already proposed
earlier.13,36-38 From the present data, it cannot be concluded
that a hydrogen spillover effect occurs, since no evidence for
hydrogen bonding in the CT material was proven. Nevertheless,
the observed enhancement of capacity is of minor importance
(36) Amorim, C.; Keane, M. A. J. Colloid Interface Sci. 2008, 322, 196.
(37) Contescu, C. I.; Brown, C. M.; Liu, Y.; Bhat, V. V.; Gallego, N. C.
J. Phys. Chem. C 2009, 113, 5886.
(39) Kaszkur, Z. J. Appl. Crystallogr. 2000, 33, 1262.
(40) Pundt, A.; Dornheim, M.; Guerdane, M.; Teichler, H.; Ehrenberg, H.;
Reetz, M. T.; Jisrawi, N. M. Eur. Phys. J. D 2002, 19, 333.
(38) Yang, R. T.; Wang, Y. J. Am. Chem. Soc. 2009, 131, 4224.
9
7726 J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010

Hydrogen Sorption in Pd Clusters
A R T I C L E S
XRD measurements. Using molecular dynamics simulations, the
authors claimed that the XRD results can be explained by a
cubic to icosahedral transition in the Pd clusters upon
hydrogenation.
The present and previous XRD studies on hydrogenation of
Pd nanoparticles show contradictory results. Nevertheless, it is
clear that short-range atomic reordering in Pd nanoparticles
occurs upon hydrogenation. In situ EXAFS measurements can
elucidate the nature of these structural modifications.
numbers: 13, 55, 147, 309, 561, 923, etc.44 For the present nPd
with an average size of 2.5 nm, the total number of atoms in a
cluster would be between 309 and 561. Mackay has demon-
strated that the cuboctahedral assembly can be transformed into
an icosahedron by geometrical rotations and vice versa.43
Because of the slight distortion of the icosahedron, the geo-
metrical packing density (0.688) is lower than for the close-
packing cuboctahedron (0.749). Lower packing density is in
agreement with hydrogen insertion and subsequent expansion
of lattice volume.
An analysis of EXAFS spectra recorded on oxygen-free
samples indicates a decrease of coordination number from bPd
(12) to nPd (7.5) (Table 3). This can be understood by an
increased number of dangling bonds due to the large fraction
of surface atoms in small clusters, in agreement with conclusions
from previous EXAFS studies on small metallic clusters.25-28,41
In a perfect geometrical arrangement (i.e., almost spherical
shaped), a fcc cluster can be built as a sequential filling of atomic
shells around a central atom. For a fcc cluster with a perfect
geometrical arrangement and an average diameter of 2.5 nm,
the number of atomic shells around the central atom is around
4 ((1) and the average coordination number is 8.9.42 The N
estimated from EXAFS (7.5) is slightly reduced relative to the
theoretical value of 8.9 that could be explained by a disorder
effect in real nanoparticles. The coordination number increases
from metal nPd (7.5) to hydride nPdHx (9.0). This feature is
striking and can be related to a rearrangement of clusters upon
hydrogenation, as suggested by the present in situ XRD results.
A similar change in coordination number has been observed
previously for Pd nanoparticles during CO/NO cycling (CO gas
flushing followed by NO).26 The authors have claimed that the
increase of N could be explained by either a resintering of
particles or a morphological variation from a flat toward a
hemispherical particle shape. In the present case, the increase
of N by sintering during hydrogenation can be refuted because
the nanoparticles are well dispersed within the mesopores of
CT, which serves as mechanical stabilizer against coalescence.
Moreover, the mobility of Pd atoms at room temperature is
limited. The second explanation that is based on a transition
from flat to hemispherical shape particles is also unlikely to
happen in the present case. Usually, metal nanoparticles have a
spherical shape to reduce the total surface energy, also confirmed
by the present and previous microscopic studies.13,14,32,36
On the basis of the present in situ XRD and EXAFS
measurements, we suggest that the phase transition that occurs
upon hydrogenation of nPd is accompanied by a structural
transformation from cuboctahedral to icosahedral symmetry. The
cuboctahedron is a close-packing fcc structure and can be seen
as a fcc polyhedron enclosed by (100) and (111) surfaces. The
icosahedral packing has 5-fold symmetry and can be described
as slightly distorted tetrahedrally shaped fcc units whose faces
are (111) planes sharing a common vertex, as initially proposed
by Mackay.43 The (100) and (111) fcc surfaces have eight and
nine nearest neighbors, respectively. The fcc cuboctahedral to
icosahedral transition increases the number of (111) surfaces,
and consequently, the average coordination number of the first
Pd-Pd neighbors increases.42 The number of atoms in a perfect
and complete geometric filling of both cuboctahedral and
icosahedral clusters is the same and follows the series of magical
The icosahedral structure of nPdHx is stabilized by the
hydrogen atoms, and the phase transition from cuboctahedron
to icosahedron is completely reversible by dehydrogenation, as
noticed in the XRD patterns during hydrogenation/dehydroge-
nation cycling (not shown here). The small size of nPd is
responsible for this phase transition, since it is completely absent
in bulk system. The 5-fold symmetry is forbidden in bulklike
systems because of the absence of translational periodicity.
The spectra of the thermal desorption of D2 from CT and
CT/nPd are shown in Figure 7. The desorption of physisorbed
D2 on the surfaces of both CT and CT/nPd takes place below
100 K (Figure 7). This feature is typical for physisorbed gas
molecules at low temperature, a process that is governed by
weak van der Waals forces. A comparison of the TDS spectra
from CT and CT/nPd below 100 K reveals noticeable differ-
ences: the maximum desorption peak from physisorbed D2 is
shifted to higher temperature in the case of CT/nPd compared
to the CT. This effect is important and implies higher energetic
bonds of D2 molecules physisorbed on the surface of the hybrid
material.
At higher temperature, the thermal desorption spectra of
absorbed deuterium in nPd and bPd have a two-peak shape
(Figure 7). The first desorption peak (at low temperature) points
to the dehydrogenation from the ꢀ phase, and the most intensive
peak corresponds to D2 desorption that accompanies the ꢀ to R
phase transformation.29 In the present case, the whole TDS
spectrum from nPd is shifted to lower temperature compared
to bPd. The two desorption peaks from nPd have broader
temperature domains (120-280 K) relative to bPd (190-300
K) and therefore are clearly separated. The maximum desorption
rate from deuterated nPd, corresponding to the maximum of
the main peak, occurs at lower temperature (248 K) relative to
bPd (291 K). Additionally, the onset temperature of desorption
from the ꢀ phase starts at lower temperature in nPdDx (120 K)
than in bPdD0.6 (180 K). This indicates that nPd has lower
thermal stability compared to bPd for both processes: the
dehydrogenation from the ꢀ phase and the ꢀ to R transformation.
Since this technique probes kinetic properties and since ther-
modynamic changes were not observed by downsizing the Pd
particle, this behavior could be related to a shorter diffusion
path in nanoparticles compared to bulk Pd.
In the entire temperature range (22-310 K), the D2 desorption
from the CT/nPd hybrid hints at a combination of the TDS
features of CT and bPd. However, both desorption processes
(from physisorbed and absorbed D2) have different thermal
stability compared to pristine CT and bPd. The presence of Pd
nanoparticles embedded in CT might increase the energy of
bonds between physisorbed D2 and carbon matrix, and subse-
quently the temperature of desorption rises compared to CT.
On the other hand, the nanometer size of particles kinetically
(41) Mierzwa, B.; Kaszkur, Z.; Moraweck, B.; Pielaszek, J. J. Alloys
Compds. 1999, 286, 93.
(42) Shandiz M. A. Phys. Lett.
A 2008, DOI: 10.1016/j.physleta.
2008.07.011.
(44) Handbook of Metal Physics, Metallic Nanoparticles; Blackman, J. A.,
(43) Mackay, A. L. Acta Crystallogr. 1961, 15, 916.
Ed.; Elsevier: Amsterdam, The Netherlands, 2009.
9
J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010 7727

A R T I C L E S
Zlotea et al.
destabilizes the D2 desorption from nPd deuteride relative to
bPd deuteride.
From the TDS spectrum of nPdDx deuteride the total amount
of desorbed deuterium can be determined as the integrated area
after background correction between 100 and 310 K. This result
can then be compared to the corresponding bPdD0.6 deuteride.
The latter value was measured experimentally and corresponds
to the fully loaded bulk deuteride at 0.05 MPa and room
temperature (D/Pd ) 0.6). As follows, a value of D/Pd ) 0.4
is obtained for nPdDx deuteride. The hydrogen/deuterium
capacity of nPd is lower than that of bPd, in good agreement
with the room temperature PCI (Figure 2) and in situ XRD
results.
The in situ NPD and thermal desorption measurements prove
that the difference between the diffraction patterns of fully
loaded and completely unloaded CT (Figure 8) and CT/nPd can
be related to the surface coverage of D2 molecules. A similar
conclusion has been proposed previously by Rousell et al.45 The
peaks obtained from the difference between fully loaded and
unloaded diffraction diagrams can be linked to the short-range
order of the D2 physisorbed molecules at the surface and to the
thermal evolution of the D2 surface coverage of the porous
carbon material.
Conclusions
The hydrogen sorption properties of the CT/nPd hybrid
material containing 17 wt % Pd nanoparticles (2.5 nm) have
been determined and compared to those of pristine CT and bulk
Pd. The CT and CT/nPd have very similar hydrogen sorption
mechanism at 77 K, which is governed by physisorption of the
hydrogen/deuterium molecules in the porosity of the material.
The desorption of adsorbed gas molecules occurs below 100 K
and implies more energetic reactions in CT/nPd than in CT, as
demonstrated by TDS.
At room temperature, nPd is hydrogenated below 0.1 MPa
followed by adsorption of hydrogen molecules on the porosity
of hybrid material for higher pressure. The thermodynamics of
hydrogenation at room temperature is not modified by down-
sizing the Pd particles relative to bPd. The plateau pressure of
the hydride formation in nPd arises in the same pressure range
as for bPd. In situ XRD and absorption isotherm at room
temperature together with thermal desorption measurements
have shown that the maximum hydrogen content stored in nPd
is lowered relative to bPd. The R-ꢀ miscibility region decreases
in nPd (0.05-0.3) compared to bPd (0.025-0.65).
The deuterium desorption from nPd occurs at lower temper-
ature than from bPd, as proven by TDS. Shorter diffusion path
in nanosized Pd can explain the improvement of desorption
kinetics compared to bulk. In the entire temperature range
(22-310 K), the D2 desorption from the CT/nPd hybrid hints
at a combination of the TDS features of pristine CT and bPd.
However, both desorption processes (from physisorbed and
absorbed D2) have different thermal stability compared to
pristine CT and bPd. The presence of nPd embedded in CT
increases the desorption temperature of physisorbed D2 com-
pared to CT. The nanometer size of Pd particles kinetically
destabilizes the D2 desorption from nPd deuteride relative to
bPd deuteride.
Figure 9. Comparison of the hydrogen storage capacities of the hybrid
material CT/nPd and the carbon precursor CT at 77 and 298 K (top).
Comparison of the maximum desorption temperatures (and related thermal
stability) during desorption of hydrogen for CT/nPd, CT, and bPd (bottom).
CT precursor (at 77 and 298 K) and the desorption temperatures
of the hybrid, CT, and bulk Pd. If the hydrogen capacity of
hybrid is reduced at 77 K compared to the pristine CT, this is
reversed at room temperature, where the hybrid has larger
capacity. Therefore, the thermal variation of the capacity in the
case of the hybrid is smaller than the CT. On the other hand,
the difference in the thermal stability between the physisorbed
and absorbed hydrogen is smaller for the hybrid relative to the
two precursors (CT and bulk Pd), as plotted in Figure 9. This
shows the double advantage of the presence of nPd embedded
in CT relative to both the hydrogen storage capacity and the
two desorption processes (from physisorbed and absorbed
hydrogen).
In situ XRD and XAS investigations during hydrogenation
of nPd have elucidated another size dependent effect on the
sorption properties. An atomic rearrangement of clusters during
hydrogenation was suggested by in situ XRD on nPd. This
behavior was clarified by in situ EXAFS measurements, which
indicate that nPd can transform from cuboctahedral to icosa-
hedral structure during hydrogenation. The icosahedral structure
is stabilized by hydrogen atoms, and the transition is completely
reversible by dehydrogenation. This peculiar transition is related
to the small size of Pd nanoparticles, given that it is forbidden
in bulk systems because of the nontranslational periodicity of
the 5-fold symmetry.
In conclusion, this study has clearly confirmed that down-
sizing the Pd particle dimensions dramatically changes the
interaction with hydrogen relative to the bulk system.
These results are gathered and compared in Figure 9, which
shows the hydrogen storage capacities of the hybrid and the
Acknowledgment. The present work has received funding from
the French ANR agency through the MATHYSSE project (Contract
ANR-07-PANH-007-01). The authors thank Diana Dragoe from
(45) Roussel, T.; Pellenq, R. J.-M.; Bienfait, M.; Vix-Guterl, C.; Gadiou,
R.; Beguin, F.; Johnson, M. Langmuir 2006, 22, 4614.
9
7728 J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010

Hydrogen Sorption in Pd Clusters
A R T I C L E S
ICMPE, Thiais, France, for ICP-OES measurements, J.-L. Soubey-
roux for neutron powder diffraction assistance as local contact on
D1B line at ILL (Grenoble, France), J.-P. Gonzales from the SANE
Department at ILL for the special neutron diffraction sample holder,
S. Belin for valuable help for X-ray absorption spectroscopy on
SAMBA beamline at SOLEIL synchrotron (Saint Aubin, France),
and Alain Michalowicz from ICMPE, Thiais, France, for precious
help with the EXAFS analysis and refinement programs.
CT/nPd hybrid material recorded at room temperature, XANES
spectra of Pd nanoparticles embedded in CT from two different
positions of the sample, and experimental filtered EXAFS
spectra and the module of the Fourier transforms together with
the corresponding refined curves for Pd nanoparticles and bulk.
This material is available free of charge via the Internet at http://
pubs.acs.org.
Supporting Information Available: Calculation of the H/Pd
for Pd nanoparticles from the hydrogen sorption isotherm of
JA101795G
9
J. AM. CHEM. SOC. VOL. 132, NO. 22, 2010 7729