Preparation of Pd-coated anodic alumina membranes for gas separation media

Article abstract of DOI:10.1149/1.2432067

Different procedures of Pd electroless deposition onto anodic alumina membranes were investigated to form a dense metal layer covering pores. The main difficulty was related to the amorphous nature of anodic alumina membranes, determining low chemical stability in solutions at pH>9, where Pd plating works more efficiently. As a consequence, it was necessary to find the operative conditions allowing Pd deposition without damaging the membrane: to reduce alumina dissolution, the plating bath was buffered at pH 8.5 by addition of either NaHC O3 or Na2 B4 O7 H2 O. Acceptable conversion of Pd was found after a deposition time of 3 min. Single and multiple deposition steps (each lasting 3 min) were accomplished. After five deposition steps, pores remained open in the plating bath containing NaHC O3, while they were almost totally occluded in the bath containing Na2 B4 O7 H2 O. In this last bath, a dense layer of Pd, ～1 μm thick, was obtained after four deposition steps alternated with surface reactivation steps; the complete closure of pores was confirmed by gas permeability measurements. The different behavior of the two buffers can be attributed to a higher solubility of amorphous alumina in NaHC O3 solution.

Full text of DOI:10.1149/1.2432067

D188
Journal of The Electrochemical Society, 154 ͑3͒ D188-D194 ͑2007͒
0013-4651/2007/154͑3͒/D188/7/$20.00 © The Electrochemical Society
Preparation of Pd-Coated Anodic Alumina Membranes for Gas
Separation Media
Rosalinda Inguanta, Mariateresa Amodeo, Fabio D’Agostino, Maurizio Volpe,
,z
*
Salvatore Piazza, and Carmelo Sunseri
Dipartimento di Ingegneria Chimica dei Processi e dei Materiali, Università di Palermo, 90128 Palermo,
Italy
Different procedures of Pd electroless deposition onto anodic alumina membranes were investigated to form a dense metal layer
covering pores. The main difficulty was related to the amorphous nature of anodic alumina membranes, determining low chemical
stability in solutions at pH Ͼ 9, where Pd plating works more efficiently. As a consequence, it was necessary to find the operative
conditions allowing Pd deposition without damaging the membrane: to reduce alumina dissolution, the plating bath was buffered
at pH 8.5 by addition of either NaHCO3 or Na2B4O7·H2O. Acceptable conversion of Pd was found after a deposition time of
3 min. Single and multiple deposition steps ͑each lasting 3 min͒ were accomplished. After five deposition steps, pores remained
open in the plating bath containing NaHCO3, while they were almost totally occluded in the bath containing Na2B4O7·H2O. In this
last bath, a dense layer of Pd, ϳ1 ␮m thick, was obtained after four deposition steps alternated with surface reactivation steps; the
complete closure of pores was confirmed by gas permeability measurements. The different behavior of the two buffers can be
attributed to a higher solubility of amorphous alumina in NaHCO3 solution.
© 2007 The Electrochemical Society. ͓DOI: 10.1149/1.2432067͔ All rights reserved.
Manuscript submitted September 11, 2006; revised manuscript received November 8, 2006.
Available electronically February 2, 2007.
Porous alumina layers can be formed by anodic oxidation of
branes would be highly disadvantageous, so that composite mem-
branes, consisting of a thin film of palladium deposited onto a po-
rous support, are preferred. Different types of composite membranes
were proposed depending on the support and on the preparation
method of the palladium layer. For operations either at high tem-
perature or in aggressive environment, inorganic support, such as
porous stainless steel32 or ␥-alumina33 are used. The thin film of
palladium can be deposited by different procedures, like chemical
vapor deposition ͑CVD͒,34 electroless plating,32 and electro-
deposition.35 In particular, electroless deposition is an easy and fast
method to carry out compared to other ones.
In this work, we present a modified process of electroless depo-
sition of palladium onto AAM prepared in phosphoric acid solution.
AAMs are a suitable support for a palladium thin film because they
are easy to prepare and exhibits low resistance to mass transport due
to their ordered structure; unfortunately, they are not stable both in
acidic and in alkaline solutions owing to their amorphous nature.
The low chemical stability represents a severe limitation for Pd elec-
troless deposition, because the possible baths have pH external to
the stability interval of AAM. In fact, the optimum pH for palladium
deposition was reported to be about 10,36 but at this pH the mem-
brane is chemically unstable. For these reasons, a systematic study
was undertaken to detect conditions optimizing characteristics of
palladium deposit and membrane stability. In a previous work,37 we
proposed a procedure giving a deposit of palladium after 30 min of
immersion of the membrane in a bath having pH 8.4. However, with
this procedure Pd deposition was localized at the pore mouth, so that
the channels were not occluded and the composite membranes are
suitable only for catalysis, but not for gas separation. Another limit
of this procedure was the slight damage and the enhanced fragility
of membranes after 30 min deposition. Based on preliminary inves-
tigations showing that the conversion of palladium occurs predomi-
nantly in the first minutes, a novel procedure based on a shorter time
of deposition and on the use of an appropriate pH stabilizer was
developed. According to this procedure, complete occlusion of
membrane channels was achieved, as shown by scanning electron
microscopy and permeability measurements.
aluminium in suitable electrolytes, where the oxide is sparingly
soluble. Typically, aqueous solutions of inorganic and organic acid,
like sulfuric, phosphoric, chromic, and oxalic acid, were used to
grow anodic porous layers with morphology dependent on the an-
odizing conditions. Recently, anodizing of aluminium in aqueous
solutions of organic acid electrolytes,1-3 and in mixed solutions of
sulfuric acid with addition of salicylic acid was also proposed. In
this last case, the use of salicylic acid allows the control of the
optical properties of anodic alumina membranes ͑AAMs͒.4
The characteristic feature of porous layers obtained by alu-
minium anodizing is the high order degree of the porous structure,
characterized by a close-packed array of columnar hexagonal cells,
each containing a central pore normal to the oxide layer surface.5 It
was assumed that parallelism of pore channels is due to the equilib-
rium between two processes: the field-enhanced oxide dissolution at
the oxide/electrolyte interface, and the oxide growth at the metal/
oxide interface.6-8 The growth of porous structures by aluminium
anodizing was extensively investigated in the past, particularly to
fabricate protective, corrosion-resistant coatings.9
The highly ordered porous structure makes anodic alumina an
ideal host or template for the fabrication of nanostructured materials
for applications in optoelectronics, sensors, magnetics, and elec-
tronic circuits.10-19
The recent ascendant interest in the fabrication of such porous
structures is also due to the possibility of obtaining membranes,
after removal of the residual metal and dissolution of the barrier
layer separating the porous oxide from the anodizing alum-
inium.20-27 These membranes are of great interest in separation pro-
cesses, such as micro- and ultrafiltration under severe conditions
where organic membranes cannot be used, as well as for fabrication
of catalytic membrane reactors ͑CMR͒, which exhibit higher con-
version, selectivity, and yield compared with conventional reactors.
Hydrogenation and dehydrogenation reactions are the most promis-
ing to be carried out in a CMR.28,29 Recently, attention was focused
on reforming reactions of light hydrocarbons and/or methanol to
produce hydrogen for energetics.30,31 In addition, pure hydrogen
production for metallurgical applications and semiconductor fabri-
cation are interesting. CMRs involving hydrogen, either as reagent
or product, use palladium metal as permselective element. Because
palladium is very expensive, employment of self-standing mem-
Experimental
Preparation of porous support.— The aluminium specimens,
used to grow anodic porous layers, were 100 ␮m thick foils, 99.99%
purity. High purity metal was preferred to minimize breakdown phe-
nomena and the formation of flaws or cracks in the porous matrix.
Prior to anodizing, samples were degreased with acetone and then
they were electropolished at 20 V for 5 min in a HClO4:C2H5OH
*
Electrochemical Society Active Member.
z E-mail: sunseri@dicpm.unipa.it
Downloaded on 2015-03-12 to IP 141.210.2.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 154 ͑3͒ D188-D194 ͑2007͒
D189
repeated ten times to obtain an optimal activated surface. The acti-
vated membrane was left drying in oven at 333 K overnight prior to
palladium deposition.
Table I. Composition of the sensitizing, activation and plating
baths for electroless deposition of Pd.
Sn Mössbauer spectroscopy was carried to study the membrane
surface after activation. Mössbauer spectra were recorded with a
conventional spectrometer ͑RITVERC GmbH͒ operating in the
transmission mode and having a Ca119SnO3 source. Calibration was
performed with spectra from Ca119SnO3 and 57Fe. The absorber
samples were placed on the sample holder of a liquid N2 cryostat
͑A.E.R.E. Harwell͒.
Palladium catalytic deposition was performed in a cylindrical
Pyrex flask filled with 10 mL of plating solution. The deposition
temperature was controlled at 333 K 1 K. After preheating the
solution for 5 min, the reducing agent was added to the system and
the solution was vigorously stirred before introducing the mem-
brane.
The composition of the plating bath, reported in Table I, was
specifically formulated to obtain a pH equal to 8.5. The amount of
ammonia necessary to stabilize palladium species in solution was
determined by a standard acid-base titration. The solution was buff-
ered by a suitable quantity of either sodium hydrogen carbonate or
borax to avoid dangerous changes of pH during plating. Hydrazine
was used as reducing agent rather than hypophosphite, because oxi-
dation of the latter produces hydrogen gas that could lead to a lower
quality of palladium deposit.41 For each deposition, the efficiency of
conversion was evaluated as ratio of the mass ͑mol͒ of palladium
deposited to the mass of palladium in the bath.
The metal coating was characterized by scanning electron mi-
croscopy ͑SEM͒, X-ray diffraction ͑XRD͒, and energy dispersive
spectroscopy ͑EDS͒. XRD patterns were obtained using a Philips
generator ͑model PW 1130͒ and a PW goniometry ͑model 1050͒.
The copper K␣ radiation and a scanning rate of 2 degree min−1
were used. Measured d-spacings were compared with the ASTM
index values. The distribution of Pd on the AAM surfaces and inside
the channels was determined by a Philips EDX microprobe ͑model
PV7760͒. The amount of palladium deposited was determined by
atomic absorption spectroscopy ͑AAS͒, using a Varian spectropho-
tometer ͑model 100 Plus͒. Before AAS, the composite membrane
was immersed in 5 mL of 1 M NaOH solution, where it was fully
dissolved. The palladium particles were oxidized to Pd2+ by addition
of 2 mL of 37% ͑w/w͒ HCl and 2 mL of 65% ͑w/w͒ HNO3. This
solution was diluted by addition of distilled water in order to reach
a palladium concentration detectable by AAS. To avoid depressing
of the atomic absorption signal due to aluminium ions, EDTA was
added to all samples prior to analyzing.
Sensitizing solution ͑1 L͒
Activation solution ͑1 L͒
Pd͑NH3͒ ͑NO3͒ ͑10%
SnCl2·2H2O
1 g
1.5 mL
1 mL
4
2
solution͒
HCl ͑37%͒
1 mL
HCl ͑37%͒
Palladium plating bath composition ͑1 L͒
Palladium precursor
Complexing agent
Buffer
Pb͑NH3͒ ͑NO3͒ ͑10% solution͒
25 mL
70 g
20 mL
2.6 g
4
2
Na2EDTA
NH3·H2O ͑25%͒
NaHCO3
or
NH3·H2O ͑25%͒
0.025 M borax, 1 M HCl
N2H4 ͑1 M͒
20 mL
25 mL
17.8 mL
Reducing agent
solution ͑1:6 by volume͒ vigorously stirred. Electrodes were
mounted onto a holder to have a flat circular surface ͑39 mm diam͒
exposed to the solution, whose temperature was fixed at −1 1°C
by means of a Lauda refrigerator ͑model RE 106͒. Electrolyte tem-
perature was chosen to attain “hard anodizing” conditions, where a
hard and abrasion resistant anodic layer is formed.9 Anodizing was
conducted in 0.4 M H3PO4 solution in two steps: in the first, cell
voltage was raised using a linear potential scan at 0.2 V s−1 up to
160 V, while in the second one, the final voltage was held constant
allowing to pass a potentiostatic charge of 160 C cm−2. A Glassman
High Tension power source ͑series ER͒ was used. The detachment
of anodic oxide layer from the remaining metal was performed by
chemical dissolution in a 0.1 M CuCl + 20% w/w HCl solution at
5°C. Pore bottom opening was achieved by immersion in 1 M
NaOH solution at room temperature for 20 min. This treatment was
made only on the metal-side surface to avoid enlargement of pore on
the solution-side surface.
The morphological features of membranes were determined by
SEM analysis ͑Philips model X30 ESEM͒.
Deposition of palladium.— To achieve a good metal film adhe-
sion and to control electroless plating, surface cleaning and activa-
tion were essential. Surface cleaning was performed by a four-step
procedure. AAMs were sonicated for 1 min at room temperature in
an alkaline aqueous solution ͑1 g/L NaOH͒, then they were soni-
cated for 2 min in distilled water. After the ultrasonic treatment,
membranes were rinsed with isopropyl alcohol for 10 min and with
distilled water for a further 10 min. The cleaned membrane was
dried in oven at 333 K for at least 4 h. After the cleaning procedure,
the porous support surface was seeded with Pd nuclei, which ini-
tiates the autocatalytic reduction of the palladium complex in the
subsequent plating step. Several methods for surface activation are
reported in the literature.38,39 All of these methods are based on the
use of solutions containing Sn2+ ions, therefore lowering the purity
of the palladium layer. In this work, the two-step immersion proce-
dure was preferred because the amount of tin in the final deposit was
less.40 This procedure consists in an alternate sequence of sensitiz-
ing steps, performed by immersion in SnCl2/HCl aqueous solution,
followed by dipping in the activation bath, this last being an acidic
aqueous solution containing a Pd salt. Compositions of sensitizing
and activation baths are reported in Table I.
Permeabilities of gases ͑H2, He, CH4, and N2͒, through both the
as-grown AAM and composite Al2O3-Pd membrane, were measured
with a homemade permeation cell at room temperature. The gas flow
rate was measured with a soap-bubble flow-meter at different trans-
membrane pressures while the permeated side was kept at atmo-
spheric pressure.
Results and Discussion
Cleaning and sensitizing/activation.— After anodizing, mem-
branes exhibited the morphological features shown by the SEM pic-
tures of Fig. 1, where Fig. 1a and b are relative to the solution-side
surface and to the metal-side surface, respectively. This figure shows
an asymmetrical pore structure on the two surfaces, with higher
interpore wall thickness for the membrane surface formed in contact
with the electrolyte. The different morphology can be attributed to
the fabrication procedure of AAM: the formation of fresh oxide
occurs only at the metal/oxide interface, owing to the reaction of
Al3+ ions formed with O2− ions migrating from the oxide/electrolyte
A fresh stannous chloride solution was used after five sensitizing
steps, because coalescence and precipitation of the suspended col-
loidal particles of SnCl2 occurred after a few hours. On the contrary,
the palladium solution was left to stabilize overnight before using
and it could be stored for several days. A typical sensitizing-
activation cycle was 2 min in sensitizing solution, 30 s in distilled
water, 2 min in activation solution, and 30 s in distilled water.
The procedure was conducted at room temperature and it was
interface, where they are injected into the oxide simultaneously with
7,8
the ejection in solution of Al3+
.
As a consequence, pores on the
solution-side surface are due to the anodizing process and they are
formed in the initial stages of anodizing, while pores on the internal
surface are formed by chemical dissolution of the barrier oxide layer
Downloaded on 2015-03-12 to IP 141.210.2.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D190
Journal of The Electrochemical Society, 154 ͑3͒ D188-D194 ͑2007͒
Figure 2. Mössbauer spectrum of an activated membrane showing the pres-
ence of tetravalent tin sites only.
Electroless deposition.— Different parameters, like pH, time,
and number of deposition steps were investigated to evaluate their
influence on the final characteristics of the deposited film. pH was
fixed at the value of either 10.0 or 8.5. In the last case, two different
buffers, NaHCO3/NH3 and Na2B4O7·H2O/HCl, were used. Because
pH plays a fundamental role on the electroless deposition process, in
the following we present separately the results obtained according to
both pH values.
Figure 1. SEM images of anodic alumina membranes prepared in 0.4 M
H3PO4 at 160 V − 1 °C and 160 C cm−2: ͑a͒ solution side; and ͑b͒ metal
side after dissolution of barrier layer in 1 M NaOH solution.
NaHCO3/NH3, pH 10.— Initially, deposition was carried out at
pH 10 to increase the conversion of palladium. From a thermody-
namic point of view, an increase of pH favors the conversion, be-
cause hydrazine is a better reducing agent in an alkaline rather than
in acidic medium. In fact, redox potentials are43
and they are formed after the anodizing process. Of course, the
average pore size on this last surface depends on the dissolution
conditions: for example, the morphology of Fig. 1b was obtained
after chemical etching of the metal-side surface in 1 M NaOH at
room temperature for 20 min. Prior to electroless deposition, mem-
brane was cleaned in 1 g L−1 NaOH aqueous solution, which is
aggressive toward amorphous alumina. As a consequence, an en-
largement of about 20% of pore size and a thickness reduction ͑from
60 to about 54 ␮m͒ were observed.
The successive sensitizing/activation treatment plays a funda-
mental role in the electroless deposition of a metal on a nonconduc-
tive surface, because it leads to the formation of catalytic seeds. In
this case, clusters of Sn-Pd, uniformly distributed onto the surface
were formed. In particular, during the sensitizing step, a colloidal
form of Sn2+ is deposited,39,40 while during the successive activation
step the Pd2+ species is reduced to Pd0 according to the reaction
N2 + 4H2O + 4e = N2H4 + 4 OH− Eؠ = − 1.16 V
͓2͔
͓3͔
H2 + 5H+ + 4e = N2H+5 Eؠ = − 0.23 V
in basic or acidic solution, respectively.
pH also affects the chelating action of EDTA and thus the con-
version of palladium. At low pH, the degree of ionization of EDTA
is low; therefore the concentration of free palladium ions in the
solution is high because the chelating action is weak. As a result, the
deposition bath is unstable, because the reduction of palladium ions
can occur not only on the membrane surface but also on Pd nuclei
formed in the bulk of solution. The solution turns to black and no Pd
ions are available for plating. On the contrary, as pH increases, the
chelating action of EDTA is enhanced because it is highly dissoci-
ated. As a result, the concentration of free palladium ions in solution
is low, and the reduction is confined to the catalytic seeds on the
membrane surface. At very high pH the fully ionized chelate anion
forms the strongest metal chelate ͑Pd-EDTA͒ complex. In this case,
the deposition of Pd is hindered because the concentration of Pd
ions in solution is very low. Thus pH values between 9 and 11
enhance Pd conversion for the combined effect of both the redox
potential of the reducing agent and the chelating action of EDTA.36
The X-ray pattern of Fig. 3, relative to a membrane held for
3 min in the deposition bath at pH 10, shows the presence of palla-
dium onto the membrane surface, while the broad band at about
Sn2+ + Pd2+ = Sn4+ + Pd0
͓1͔
The complete oxidation of Sn2+ to Sn4+ was confirmed by the
Mössbauer analysis ͑Fig. 2͒, showing the presence of only Sn4+ on
the membrane surface soon after the sensitizing/activation treatment.
In fact, Fig. 2 shows only a large band at an isomer shift of
0 mm s−1, typical of Sn4+, while the doublet typical of Sn2+ at iso-
mer shift of 2.55 mm s−1 is absent.42
Atomic absorption analysis revealed that the total amount of Pd
deposited during the activation step was about 2.93 ϫ 10−2
mg cm−2, while EDS analysis revealed different concentrations of
Pd and Sn on the two surfaces, as shown in Table II. This difference
can be attributed to the different morphology of the two surfaces,
and in particular to the different membrane specific surface ͑ratio of
the Al2O3 mass to the apparent surface͒ exposed to the sensitizing/
activation solution. This interpretation is supported by data in Table
II, showing that the highest amount of Sn and Pd was deposited on
the solution-side surface, which exhibits a higher specific surface.
However, from Table II a nearly equal Pd/Sn ratio in the catalytic
particles is derived for both surfaces.
Table II. Surface atomic ratio of anodic alumina membranes af-
ter sensitizing-activation treatment.
Solution side
Metal side
Sn/Al atom %
Pd/Al atom %
0.264
0.091
0.094
0.033
Downloaded on 2015-03-12 to IP 141.210.2.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 154 ͑3͒ D188-D194 ͑2007͒
D191
usually lower. Interestingly, Fig. 4 evidences that membrane surface
was damaged with pore walls collapsed in many points ͑see Fig.
4b͒. Dissolution of AAM, at this pH, was confirmed by EDS analy-
sis revealing that white particles consist of Pd, Al, and O, and by
weight measurements, carried out before and after deposition, re-
vealing a weight loss of about 9% for the membrane. Considering
the weight gain due to metal deposition, the true weight loss of
alumina during the immersion in the deposition bath was still higher
after only 3 min of immersion. The formation of the white clusters
could originate from the precipitation of alumina from the solution
caused by a local pH decrease associated with the oxidation of hy-
drazine ͑Reaction 2͒. The white clusters are not evident on the
metal-side surface ͑Fig. 4b͒ because, during the electroless deposi-
tion, AAM was immersed horizontally into the solution with this
surface toward the bottom of the cell. In conclusion, these deposi-
tion conditions caused membrane partial dissolution, enhancing its
fragility, while pores remained almost completely open.
Figure 3. ͑Color online͒ XRD pattern of Pd-Al2O3 composite membranes
obtained after 3 min in a deposition bath buffered at pH 10 in the presence of
NaHCO3.
NaHCO3 buffer, pH 8.5.— On the basis of previous result, depo-
sition of palladium was investigated in a solution buffered at pH 8.5
by addition of NaHCO3. Membranes were held for 3 min in this
solution. At this pH, dissolution of alumina will be very slow, even
if a reduction in the conversion of palladium is expected. The mor-
phology of a membrane prepared in these conditions is similar to
that of Fig. 4: pores were still open, but their size was reduced.
Membrane was also slightly damaged in these conditions, as con-
firmed by the presence of the white clusters. To explain the dissolu-
tion process of alumina at this pH, we have to focus our attention on
the electroless deposition reaction
2␪ϭ28° reveals the amorphous nature of the AAM. By using the
Scherrer formula,44 we calculated a grain size of about 11.5 nm.
AAS revealed that the total amount of palladium deposited in these
conditions was 0.205 mg cm−2. The conversion was 53%, and the
rate of deposition was 2.7 ϫ 10−2 mg s−1
.
EDS analysis revealed surface concentrations of Pd on the two
surfaces of about 53 and 47 atom %. This difference can be attrib-
uted to the different surface concentrations of catalytic sites formed
during the sensitizing/activation treatment ͑see Table II͒, consider-
ing that the highest Pd concentration was detected on the surface
having a higher concentration of catalytic seeds. The SEM pictures
reported in Fig. 4 show metal deposits consisting of spherical par-
ticles, about 40 nm large, deposited around pore mouth and white
particles of different size distributed onto the surface. The single
grain size of 11.5 nm was not detected because SEM resolution is
2+
Pd͑NH3͒ + 4H2O + 2e = Pd + 4NH+4 + 4 OH−
͓4͔
4
Reaction 4 causes a local increase of pH, which could determine
the dissolution of alumina, while the oxidation of hydrazine, causes
a local decrease of pH. Because the cathodic area, where palladium
is reduced, and anodic area, where hydrazine is oxidized, are in
contact, it is highly probable that alumina in contact with higher pH
locally dissolves and it precipitates from the solution zones at lower
pH, forming clusters embedded of palladium. XRD analysis re-
vealed the presence of Pd on both membrane surfaces, while AAS
revealed a total amount of deposited metal of 0.104 mg cm−2. Con-
version and deposition rate of Pd were 27% and 1.28
ϫ 10−2 mg s−1, respectively, while surface Pd concentrations,
evaluated by EDS analysis, were equal to 26 and 23.6 atom %.
Comparison of these findings with those at pH 10 indicates that pH
reduction has a detrimental effect on the palladium deposition pro-
cess, but improves the quality of the composite membrane because it
appears less damaged ͑weight loss was about 0.9%͒ and with nar-
rower pores.
To improve both the quality of the deposit and the efficiency of
deposition, the time of deposition was increased from 3 to 9 min,
but the improvement was poor. The total amount of Pd deposited
increased to 0.221 mg cm−2, while the deposition efficiency and rate
decreased to 19% and 9.8 ϫ 10−3 mg s−1, respectively. The weight
loss increased up to about 13% while pores were still partially open.
The membrane section-view, reported in Fig. 5, shows spherical
particles deposited on the surface tending to form pillars perpen-
dicular to the surface, so that shutting of the pore mouth was par-
tially hindered. Figure 5 shows also the presence of Pd nanoparticles
deposited on the lateral surface of pore channels. The distribution of
Pd nanoparticles along the channels length was detected by EDS
analysis. Figure 6 shows that Pd particles were deposited essentially
within a thickness interval from 5 to 10 ␮m from pores mouth, and
not in the central channel part. Figure 6 shows also that the Pd/Al
atomic ratio was different in the proximity of the two surfaces, con-
firming the asymmetry of the membranes.
To increase the amount of Pd deposited and to decrease dissolu-
tion, multiple deposition steps ͑up to five͒, each lasting 3 min, were
conducted. For each deposition step, a fresh solution was used. The
amount of Pd deposited increased from 0.104 mg cm−2, after the
Figure 4. SEM micrographs of Pd-Al2O3 composite membranes obtained
after 3 min in a deposition bath buffered at pH 10 in the presence of
NaHCO3: ͑a͒ solution side; and ͑b͒ metal side.
Downloaded on 2015-03-12 to IP 141.210.2.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D192
Journal of The Electrochemical Society, 154 ͑3͒ D188-D194 ͑2007͒
Figure 5. Section view of a Pd-Al2O3 composite membranes obtained after
9 min in a deposition bath buffered at pH 8.5 in the presence of NaHCO3.
Figure 7. Conversion of Pd vs deposition number in a deposition bath buff-
ered at pH 8.5 in the presence of NaHCO3.
first deposition step, to 0.383 mg cm−2, after 5 deposition steps,
while the efficiency of deposition decreased from about 27% down
to about 20% after the fifth deposition step, as shown in Fig. 7. This
behavior can be explained taking into account that, as the number of
the deposition steps increases, surface concentration of free catalytic
sites decreases, because they are covered by deposited Pd. Conse-
This conclusion is confirmed by the section view of Fig. 9 showing
a dense and continuous palladium film, about 1 ␮m thick, with good
adhesion to the substrate. The amount of Pd deposited was about
0.458 mg cm−2. This finding indicates a specific influence of the
buffer agent on the features of the deposit.
To verify the effectiveness of the procedure adopted, permeabil-
ity measurements were carried out. Figure 10 shows the fluxes of
different gases vs the transmembrane pressure for a composite mem-
brane after three deposition steps. The linear dependence indicates a
Knudsen mechanism of gas permeation, described by45,46
quently,
deposition
rate
decreases
from
1.38 to 1.02
ϫ 10−2 mg s−1. SEM analysis revealed that after 5 steps pores were
still open on both surfaces, which were quite damaged. Note that
damage was higher than after one deposition step, indicating that the
total immersion time in the plating bath plays a fundamental role in
altering the surface morphology of the AAM.
To reduce damage and to restore the concentration of catalytic
sites, successive deposition steps ͑up to four͒ were alternated with
reactivation of the surface. Fresh activation and plating solutions
were employed for each step. After four deposition steps a conver-
sion efficiency of about 31% was achieved, but the deposit did not
uniformly cover the surface and some pores were still open.
2 ␧r
v
Pk =
͓5͔
3 ␶LRT
In Eq. 5, Pk is the permeability coefficient ͑in mol s−1m−2Pa−1͒,
Na2B4O7·H2O buffer, pH 8.5.— Because the previous deposition
procedures did not lead to a dense palladium layer occluding pores,
bath composition was modified by substituting NaHCO3 with a bo-
rax buffer, keeping pH at the same value of 8.5. The deposition was
performed in 3, 4, or 5 steps, each 3 min long. Analogous to the bath
buffered with NaHCO3, even after 5 deposition steps surface cover-
age by the metal was not complete. As shown in Fig. 8a, several
pores were still open, even if surface coverage is better than in
solutions containing NaHCO3. Following the previous procedure,
the successive deposition steps were alternated with reactivation of
the surface to restore the concentration of catalytic sites. Figure 8b
shows that after the fourth deposition step channels were occluded.
Figure 8. SEM pictures of Pd-Al2O3 composite membrane after multiple
depositions, each 3 min long, in bath buffered at pH 8.5 in the presence of
Na2B4O7·H2O: ͑a͒ after 5 consecutive depositions; and ͑b͒ after 4 deposi-
tions with intermediate reactivation.
Figure 6. ͑Color online͒ EDS analysis along the channel of a Pd-Al2O3
composite membranes obtained after 9 min in a deposition bath buffered at
pH 8.5 in the presence of NaHCO3.
Downloaded on 2015-03-12 to IP 141.210.2.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 154 ͑3͒ D188-D194 ͑2007͒
D193
Figure 9. Section view of a Pd-Al2O3 composite membrane obtained after 4
deposition steps, each lasting 3 min, with intermediate reactivation, in the
bath buffered at pH 8.5 using Na2B4O7·H2O.
Figure 11. Permeability as function of the inverse square root of molecular
weight for the composite membrane of Fig. 10.
␧ the porosity, r the mean pore radius of the membrane, ␶ the tor-
tuosity factor, equal to unity for uniform straight pores normal to the
H2. Although metal film appears continuous, He permeation sug-
gests the existence of microholes caused by small impurities and/or
surface defects; consequently, H2 flux should consist of two parts,
one coming from the microholes in the palladium layer and the other
from the dense part.47,48
The separation factor, defined as the ratio between H2 and He
flux, was calculated at the transmembrane pressure of 0.1 MPa: for
composite membranes obtained with reactivation steps this separa-
tion factor exceeds the theoretical Knudsen diffusion value, and in-
creases from 2.74 to 6.55 with increasing the number of deposition
steps. This improvement of the H2 separation performance through
the Pd-Al2O3 composite membranes must be ascribed to the selec-
tive surface diffusion transport mechanism.45
membrane surface, L the thickness of the membrane, and
average velocity for an ideal gas given by
the
v
0.5
8RT
=
͓6͔
v
ͩ ͪ
␲M
where M is the gas molecular weight.
Assuming a Knudsen flow in cylindrical channels, from Eq. 5
and 6, the pore radius can be estimated by the slope of the gas
permeability vs ͑1/M͒0.5, once known pore population and mem-
brane thickness. This is displayed in Fig. 11 for a membrane ob-
tained after three deposition steps: for this membrane a porosity of
19.6% and a thickness of 57.3 ␮m were evaluated by SEM analysis,
giving a mean pore radius of 54.7 nm. Figure 12 shows the variation
of the average pore radius with the number of deposition steps: a
relevant decrease occurs after four deposition steps, while the varia-
tion between the fourth and the fifth deposition step is negligible.
After five deposition steps, the mean pore size of the composite
membrane was about 19 nm.
The higher effectiveness of the bath plating containing
Na2B4O7·H2O in determining pore closure, compared to the identi-
cal bath containing NaHCO3 buffer, was specifically investigated, as
it was not reported previously. AAS measurements, carried out on
activated membranes held for 3 min in both solutions, revealed a
loss of catalytic seeds of about 16% in the presence of NaHCO3 and
9% in the presence of Na2B4O7·H2O. This finding indicates that
AAMs are slightly more soluble in the plating bath containing the
NaHCO3 buffer, with the consequence that also the catalytic seeds
formed in the activation step are partially lost during the electroless
deposition. For this reason, the amount of Pd deposited after four
steps with reactivation was lower in the presence of NaHCO3
Permeability measurements were also performed on membranes
functionalized with deposition steps alternated with reactivation
steps, and they confirmed the shutting of the channels shown in Fig.
8b. In fact, nitrogen and methane fluxes were negligible at all trans-
membrane pressures, while small H2 and He fluxes were measured
through the composite membrane, and they increase sensibly at high
trans-membrane pressures. Figure 13 shows the gas fluxes of H2 and
He for a membrane functionalized with four deposition steps alter-
nated with surface reactivation: He flux is much lower than that of
͑0.365 mg cm−2
͑0.462 mg cm−2͒.
͒
than in the presence of Na2B4O7·H2O
Figure 10. ͑Color online͒ Gas flux vs pressure for Pd-Al2O3 composite
membrane after 3 electroless deposition steps of Pd, each 3 min long, in the
bath buffered at pH 8.5 with Na2B4O7·H2O, T = 25 °C.
Figure 12. ͑Color online͒ Variation of pore radius with the deposition num-
ber for Pd-Al2O3 composite membranes obtained after multiple depositions,
each 3 min long, in bath buffered at pH 8.5 in the presence of Na2B4O7·H2O.
Downloaded on 2015-03-12 to IP 141.210.2.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

D194
Journal of The Electrochemical Society, 154 ͑3͒ D188-D194 ͑2007͒
References
1. S. Ono, M. Saito, M. Ishiguro, and H. Asoh, J. Electrochem. Soc., 151, B473
͑2004͒.
2. S. Ono, M. Saito, and H. Asoh, Electrochem. Solid-State Lett., 7, B21 ͑2004͒.
3. S. Ono, M. Saito, and H. Asoh, Electrochim. Acta, 51, 827 ͑2005͒.
4. Y. Yang and Q. Gao, Phys. Lett. A, 333, 328 ͑2004͒.
5. J. W. Diggle, T. C. Downie, and C. W. Goulding, Chem. Rev. (Washington, D.C.),
69, 370 ͑1969͒.
6. J. P. O’Sullivan and G. C. Wood, Proc. R. Soc. London, Ser. A, 317, 511 ͑1970͒.
7. J. Siejka, and C. J. Ortega, J. Electrochem. Soc., 124, 883 ͑1977͒.
8. K. Shimizu, K. Kobayashi, G. E. Thompson, and G. C. Wood, Philos. Mag. A, 66,
643 ͑1992͒.
9. V. F. Henley, Anodic Oxidation of Aluminium and Its Alloys p. 6, Pergamon Press,
Oxford ͑1982͒.
10. K. Nielsch, F. Muller, A.-P. Li, and U. Gosele, Adv. Mater. (Weinheim, Ger.), 12,
582 ͑2000͒.
11. W. G. Yelton, K. B. Pfeifer, and A. W. Staton, J. Electrochem. Soc., 149, H1
͑2002͒.
12. H. Chirac, A. E. Moga, M. Urse, and T.-A. Ovari, Sens. Actuators, A, 106, 348
͑2003͒.
13. S. W. Lin, S. C. Chang, R. S. Liu, S. F. Hu, and N. T. Jan, J. Magn. Magn. Mater.,
282, 28 ͑2004͒.
Figure 13. ͑Color online͒ Gas flux vs pressure Pd-Al2O3 composite mem-
brane after 4 depositions, each 3 min long, with intermediate reactivation in
bath buffered at pH 8.5 in the presence of Na2B4O7·H2O, T = 25 °C.
14. P. Ciambelli, D. Sannino, M. Sarno, and J. B. Nagy, Adv. Eng. Mater., 6, 804
͑2004͒.
15. T. Wu, I. C. Leu, J. H. Yen, and M. H. Hon, J. Phys. Chem. B, 109, 9575 ͑2005͒.
16. I. Z. Rahman, A. Boboc, K. M. Razeeb, and M. A. Rahman, J. Magn. Magn.
Mater., 290–291, 246 ͑2005͒.
17. Y. Piao, H. Lim, J. Y. Chang, W.-Y. Lee, and H. Kim, Electrochim. Acta, 50, 2997
͑2005͒.
18. X. Zhang, Y. Hao, G. Meng, and L. Zhang, J. Electrochem. Soc., 152, C664
͑2005͒.
19. J. Lee, Y. Yun, J. Oh, and Y. Tak, Electrochim. Acta, 51, 1 ͑2005͒.
20. R. C. Furneaux, W. R. Rigby, and A. P. Davidson, Nature (London), 337, 147
͑1989͒.
Conclusions
Electroless deposition of Pd onto AAM is limited by the amor-
phous nature of the support causing a significant dissolution in the
pH interval 9 ÷ 11, where the rate of deposition is maximum. Ex-
periments at pH 10 showed that after 3 min of immersion in the
plating bath an efficiency of Pd deposition of about 53% was
achieved, but pores were open and the surface of the membrane was
heavily damaged. As a result, different procedures were investigated
to form a dense metal layer and to find the best compromise between
bath utilization and membrane stability. The pH value was lowered
to 8.5 and it was controlled by two different buffers: NaHCO3 or
Na2B4O7·H2O. Despite the lower pH, dissolution of alumina oc-
curred at some extent, as evidenced by the presence on the surface
of white alumina/Pd clusters precipitated from the solution. This
behavior can be attributable to a local increase of pH due to Pd2+
reduction coupled with a local decrease of pH due to hydrazine
oxidation. Because the cathodic and anodic areas are in close prox-
imity, alumina dissolved from the cathodic areas precipitated on the
anodic ones. The local change of pH did not interfere with the pH of
the bulk solution which remained constant during the electroless
deposition of Pd.
Single and multiple depositions were carried out to form a dense
layer of Pd, and the best results were obtained by using the
Na2B4O7·H2O buffer. Multiple deposition steps were performed us-
ing a fresh plating bath after each step, and they were alternated
with reactivation steps to restore the surface concentration of cata-
lytic sites.
Permeability measurements showed the progressive closure of
the pores with the number of deposition steps. After four deposition
steps alternated with surface reactivation, a complete shutting of
pores was achieved, owing to the formation of a dense Pd layer
permeable only to H2 and He. The higher effectiveness of the plat-
ing bath containing Na2B4O7·H2O could be due to a lower solubility
of amorphous alumina in this solution. Further work is in progress to
clarify the transport mechanism of gases and to optimize the plating
process for a possible extensive use of these composite membranes
for gas separation processes.
21. P. P. Mardilovich, A. N. Govyadinov, N. I. Mukhurov, A. M. Rzhevskii, and R.
Paterson, J. Membr. Sci., 98, 131 ͑1995͒.
22. P. P. Mardilovich, A. N. Govyadinov, N. I. Mazurenko, and R. Paterson J. Membr.
Sci., 98, 143 ͑1995͒.
23. P. P. Mardilovich, A. N. Govyadinov, N. I. Mukhurov, A. M. Rzhevskii, and R.
Paterson J. Membr. Sci., 98, 131 ͑1995͒.
24. J. Randon, P. P. Mardilovich, A. N. Govyadinov, and R. Paterson J. Colloid Inter-
face Sci., 169, 335 ͑1995͒.
25. P. Bocchetta, C. Sunseri, A. Bottino, G. Capannelli, G. Chiavarotti, S. Piazza, and
F. Di Quarto, J. Appl. Electrochem., 32, 977 ͑2002͒.
26. P. Bocchetta, C. Sunseri, G. Chiavarotti, and F. Di Quarto, Electrochim. Acta, 48,
3175 ͑2003͒.
27. P. Bocchetta, C. Sunseri, R. Masi, S. Piazza, and F. Di Quarto, Mater. Sci. Eng., C,
23, 1021 ͑2003͒.
28. R. Dittmeyer, V. Hollein, and K. Daub, J. Mol. Catal. A: Chem., 173, 135 ͑2001͒.
29. E. Drioli and M. Romano, Ind. Eng. Chem. Res., 40, 1277 ͑2001͒.
30. T. P. Tiemersma, C. S. Patil, M. van Sint Annaland, and J. A. M. Kuipers, Chem.
Eng. Sci., 61, 1602 ͑2006͒.
31. S. Wieland, T. Melin, and A. Lamm, Chem. Eng. Sci., 57, 1571 ͑2002͒.
32. J. Tong, H. Suda, K. Haraya, and Y. Matsumura, J. Membr. Sci., 260, 10 ͑2005͒.
33. G. Xomeritakis and Y.-S. Lin, AIChE J., 44, 174 ͑1998͒.
34. S.-E. Nam, S.-H. Lee, and K.-H. Lee, J. Membr. Sci., 153, 163 ͑1999͒.
35. N. Itoh, N. Tomura, T. Tsuji, M. Hongo, Microporous Mesoporous Mater., 39, 103
͑2000͒.
36. J. N. Keuler, L. Lorenzen, R. N. Sanderson, and V. Linkov, Plat. Surf. Finish., 84,
34 ͑1997͒.
37. M. Volpe, R. Inguanta, S. Piazza, and C. Sunseri, Surf. Coat. Technol., 200, 5800
͑2006͒.
38. N. Feldstain, J. Electrochem. Soc., 121, 738 ͑1974͒.
39. A. M. T Van Der Putten., J. De Bakker, and L. G. J. Fakkink, J. Electrochem. Soc.,
139, 3475 ͑1992͒.
40. J. Horkans, J. Electrochem. Soc., 130, 311 ͑1983͒.
41. Y. S. Cheng and K. L. Yeung, J. Membr. Sci., 182, 195 ͑2001͒.
42. J. C. Serrano-Ruiz, G. W. Huber, M. A. Sanchez-Castillo, J. A. Dumesic, F.
Rodriguez-Reinoso, A. Sepulveda-Escribano, J. Catal., 241, 378 ͑2006͒.
43. G. Charlot, Les Reactions Chimiques En Solution, p. 402, Masson et Cie, Paris
͑1969͒.
44. A. R. West, Solid State Chemistry and Its Applications, p. 173 John Wiley & Sons
Ltd, Chichester ͑1985͒.
Acknowledgments
45. A. J. Burggraaf, in Fundamentals of Inorganic Membrane Science and Technology,
A. J. Burggraaf and L. Cot, Editors, p. 331, Elsevier Science B. V., Amsterdam
͑1996͒.
46. H. de L. Lyra and R. Paterson, J. Membr. Sci., 206, 375 ͑2002͒.
47. H. B. Zhao, K. Pflanz, J. H. Gu, A. W. Li, N. Stroh, H. Brunner, and G. X. Xiong,
J. Membr. Sci., 142, 147 ͑1998͒.
This work was supported financially by APQ Ricerca della Re-
gione Siciliana Delibera CIPE no. 17/2003, “Laboratorio
dell’Innovazione nel Settore dei Beni Culturali.”
The Universita di Palermo assisted in meeting the publication costs of
this article.
48. H. B. Zhao, G. X. Xiong, and G. V. Baron, Catal. Today, 56, 89 ͑2000͒.
Downloaded on 2015-03-12 to IP 141.210.2.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).