Catalytic Teflon AF-2400 membrane reactor with adsorbed ex situ synthesized Pd-based nanoparticles for nitrobenzene hydrogenation

Article abstract of DOI:10.1016/j.cattod.2020.03.062

Among the unconventional approaches of supporting catalyst nanoparticles, the layer-by-layer assembly of polyelectrolyte multilayers for nanoparticle adsorption represents an easy and convenient method. It enables the deposition of singularly adsorbed nanoparticles and prevents them from aggregating. In this work, polydopamine was grafted onto the internal surface of a Teflon AF-2400 tubular membrane, known for its excellent permeability to light gases and inertness to chemicals. Poly(acrylic acid) and poly(allylamine hydrochloride) were sequentially adsorbed onto the modified surface of the membrane. Ex situ synthesized spherical, cubical, truncated octahedral palladium or dendritic platinum-palladium nanoparticles were then incorporated. The catalytic membranes were assembled in a tube-in-tube configuration and tested over 6 h of continuous nitrobenzene hydrogenation with molecular hydrogen. Stable conversion was observed for the truncated octahedral and dendritic nanoparticles, while a progressive deactivation occurred for the other nanoparticles. Due to their small size, the 3.7 nm spherical nanoparticles exhibited the highest reaction rate, 629mol_reactant/(mol_catalyst?h), while the cubical nanoparticles showed the highest turnover frequency, ～3000 h^?1. The reactor concept developed in this work demonstrates how such a design can serve as a platform for conducting continuous multiphase catalytic reactions in flow using singularly adsorbed and finely tuned nanoparticles. The small volume of pressurized gas present in the tube-in-tube reactor offers improved process safety compared to a batch process, while the Teflon AF-2400 membrane provides control over the gas permeation during reaction.

Full text of DOI:10.1016/j.cattod.2020.03.062

Catalysis Today xxx (xxxx) xxx–xxx
Contents lists available at ScienceDirect
Catalysis Today
journal homepage: www.elsevier.com/locate/cattod
Catalytic Teflon AF-2400 membrane reactor with adsorbed ex situ
synthesized Pd-based nanoparticles for nitrobenzene hydrogenation
a
a
a
b
a
Baldassarre Venezia , Luca Panariello , Daniel Biri , Juhun Shin , Spyridon Damilos ,
a
b
a,
Anand N.P. Radhakrishnan , Chris Blackman , Asterios Gavriilidis *
a
b
Department of Chemical Engineering, University College London, Torrington Place, London, WC1E 7JE, UK
Department of Chemistry, University College London, 20 Gordon Street, London, WC1H 0AJ, UK
A R T I C L E I N F O
A B S T R A C T
Keywords:
Among the unconventional approaches of supporting catalyst nanoparticles, the layer-by-layer assembly of
polyelectrolyte multilayers for nanoparticle adsorption represents an easy and convenient method. It enables the
deposition of singularly adsorbed nanoparticles and prevents them from aggregating. In this work, poly-
dopamine was grafted onto the internal surface of a Teflon AF-2400 tubular membrane, known for its excellent
permeability to light gases and inertness to chemicals. Poly(acrylic acid) and poly(allylamine hydrochloride)
were sequentially adsorbed onto the modified surface of the membrane. Ex situ synthesized spherical, cubical,
truncated octahedral palladium or dendritic platinum-palladium nanoparticles were then incorporated. The
catalytic membranes were assembled in a tube-in-tube configuration and tested over 6 h of continuous ni-
trobenzene hydrogenation with molecular hydrogen. Stable conversion was observed for the truncated octa-
hedral and dendritic nanoparticles, while a progressive deactivation occurred for the other nanoparticles. Due to
Layer-by-layer assembly
Polydopamine
Tubular membrane
Flow reactor
Colloidal nanoparticles
Palladium catalyst
their small size, the 3.7 nm spherical nanoparticles exhibited the highest reaction rate, 629molreactant
/
−1
(
molcatalyst⋅h), while the cubical nanoparticles showed the highest turnover frequency, ∼3000 h . The reactor
concept developed in this work demonstrates how such a design can serve as a platform for conducting con-
tinuous multiphase catalytic reactions in flow using singularly adsorbed and finely tuned nanoparticles. The
small volume of pressurized gas present in the tube-in-tube reactor offers improved process safety compared to a
batch process, while the Teflon AF-2400 membrane provides control over the gas permeation during reaction.
1
. Introduction
Nanocatalysis has attracted considerable attention over recent years
polyelectrolyte assembly has been proven to be a versatile method to
create nanocomposite structures [20]. This approach consists of alter-
nating polyelectrolyte layers that interact electrostatically via their
functional building blocks, creating ultrathin organic films [21]. Decher
in 1991 developed the concept of the consecutive adsorption of poly-
electrolytes, after reporting a method on the successive physisorption of
anionic and cationic bipolar amphiphiles onto charged surfaces
[22,23]. The layer-by-layer assembly method of polyelectrolytes has
been employed in different fields, for both fundamental and applied
research, and among these is the adsorption of nanoparticles [20]. In
this regard, the work by Dotzauer et al. showed that the LbL method can
be applied to porous membranes in order to adsorb nanoparticles for
catalytic applications [24,25]. The internal surfaces of alumina and
polycarbonate membranes were modified by the adsorption of poly
(acrylic acid) (PAA) and poly(allylamine hydrochloride) (PAH), which
provided binding sites for gold nanoparticles [24]. These were em-
ployed in the hydrogenation of 4-nitrophenol into 4-aminophenol by
[
1–3]. The large surface-to-volume ratio of catalyst nanoparticles offers
a higher catalyst activity with respect to their bulk equivalent [4–6]. In
addition to a size effect, nanoparticle morphology can be of interest in
catalysis, due to the different reactivity of crystalline facets [2,6–8].
However, because of their thermodynamic tendency to aggregate, it is
of paramount importance to stabilize them onto supports, inorganic or
organic [9,10]. Inorganic porous supports for catalytic nanoparticles
have been extensively used [10–12], including carbon [13], silica/
mesoporous silica [14–16], or zeolites [17].
However, the synthesis of these supports generally includes a high
temperature calcination process of the inorganic precursors [18]. Re-
cent trends in supporting nanoparticles on polymeric structures have
attracted attention due to the low temperature and environmental
friendliness of the preparation process [19]. The layer-by-layer (LbL)
⁎
Corresponding author.
E-mail address: a.gavriilidis@ucl.ac.uk (A. Gavriilidis).
https://doi.org/10.1016/j.cattod.2020.03.062
Received 7 December 2019; Received in revised form 8 March 2020; Accepted 27 March 2020
0920-5861/ © 2020 Elsevier B.V. All rights reserved.
Please cite this article as: Baldassarre Venezia, et al., Catalysis Today, https://doi.org/10.1016/j.cattod.2020.03.062

B. Venezia, et al.
Catalysis Today xxx (xxxx) xxx–xxx
sodium borohydride, which flowed through the membrane pores. Re-
sults showed that the LbL method was effective in creating catalytic
sites inside the membrane and allowing for continuous operation.
Liu et al. applied the LbL method to deposit polyelectrolyte layers onto
the outer surface of a polypropylene hollow fibre membrane. Catalyst
particles were then immobilized onto the modified surface by reducing the
adsorbed precursor species [26]. Nitrobenzene hydrogenation to aniline
was studied by pumping it outside the tubular membrane, while hydrogen
was pressurized inside, in a membrane reactor configuration. The reactor
showed stable conversion over hours and process conditions were varied
to prove the flexibility of this catalytic membrane reactor. However, the
synthesis of nanoparticles performed in situ, showed a consequent loss of
control over the catalytic particle size and morphology. Ex situ synthesis
can provide better tuning of the nanoparticles [27,28]. The adsorption of
ex situ 12 nm Au nanoparticles inside LbL-modified polyethersulfone
hollow fibre membranes was demonstrated by Ouyang et al. [29]. This
study showed that the fibre surface modification led to a high density of
unaggregated nanoparticles that were highly active in the continuous li-
quid phase hydrogenation of 4-nitrophenol.
water. Separately, 200 mg of PEG-PPG-PEG were dissolved in 10 mL of
DI water. After the complete dissolution of the polymer, 0.1 mL of the
Pd precursor solution were injected in the polymer solution and stirred
for 24 h at room temperature.
2.1.2. Pd truncated nano-octahedra synthesis
Pd nano-octahedra were prepared according to Lim et al. [44]. In
detail, 68.4 mg of Na
2
PdCl were dissolved in 3.6 mL of DI water. Se-
4
parately, 105 mg of PVP, 60 mg L-ascorbic acid and 60 mg of citric acid
were dissolved in 8 mL of DI water and heated to 100 °C under reflux in
a stirred three-neck flask placed in an oil bath. A volume of 3 mL of the
Pd precursor was then injected in the heated flask. After 3 h, the solu-
tion was removed from the oil bath and allowed to cool at room tem-
perature in open air.
2.1.3. Pd nanocubes synthesis
Pd nanocubes were prepared according to Lim et al. [45]. In detail,
67.1 mg of Na
2
PdCl were dissolved in 3.6 mL of a 1.68 M KBr water
4
solution the day before the synthesis. On the day of the synthesis,
105 mg of PVP, 60 mg L-ascorbic acid and 60 mg of citric acid were
dissolved in 8 mL of DI water and heated to 80 °C in a stirred three-neck
flask placed in an oil bath. Then, 3 mL of the prepared Pd precursor
stock were injected in the heated solution. After 3 h, the solution was
removed from the bath and allowed to cool at room temperature in
open air.
Dense, hydrophobic, gas-permeable membranes that show high li-
quid pressure breakthrough, like the Teflon AF-2400, have recently at-
tracted considerable interest, due to their high permeability to light gases
and chemical inertness [30,31]. Teflon AF-2400 has been employed in a
variety of gas-liquid and gas-liquid-solid reactions [32–37]. First in-
troduced by the group of Ley [38], Teflon AF-2400 has been employed in
a tube-in-tube configuration for both homogeneous and heterogeneous
continuous hydrogenations. Hydrogen was pressurized outside, while the
reacting liquid was directed inside the membrane [33]. In the case of
homogeneous catalysis, the catalyst was premixed with the substrate and
flowed in the tubular reactor, while when a heterogeneous catalyst was
employed, the catalyst was packed in a column placed downstream of the
tube-in-tube, which was used as pre-saturator.
2.1.4. Pt-Pd nanodendrites synthesis
Pt-Pd nanodendrites were prepared according to Lim et al. [44]. In
detail, Pd nano-octahedra were prepared as previously described. Then,
1 mL of this solution was added to 6 mL of DI water. In this solution,
35 mg of PVP and 60 mg L-ascorbic acid were dissolved. This solution
was heated to 90 °C in a stirred three-neck flask placed in an oil bath.
In order to be able to support the catalyst on the surface of a tubular
Teflon AF-2400 membrane using an LbL approach, an initial surface
modification is required. The discovery in 2007 of the adhesive properties
of polydopamine (PDA) onto different materials by the group of
Messersmith [39], paved the way for applications of PDA coatings in ma-
terial science, chemistry and engineering [40]. PDA contains aminoethyl
and catechol functional groups, which respectively display a positive and
negative charge, making it suitable in the LbL assembly method [41].
In this work, the inner surface of a tubular Teflon AF-2400 mem-
brane was modified, via a PDA coating followed by the sequential de-
position of a PAA and PAH layers. Nitrobenzene hydrogenation with
molecular hydrogen was used as a model reaction and was carried out
over size- and shape-tuned ex situ synthesized palladium-based nano-
particles that were adsorbed onto the inner surface of the modified
tubular membrane. Palladium was chosen as the catalyst due to its high
selectivity in the hydrogenation of the nitro group [42]. Characteriza-
tion analyses and reaction tests were performed on different adsorbed
nanoparticles, demonstrating this reactor concept as an attractive
platform for nanoparticle catalyst testing.
Separately, 27 mg of K
2
PtCl were dissolved in 3 mL of DI water. This
4
solution was then injected in the heated solution containing the Pd
nano-octahedra, PVP and ascorbic acid. After 3 h, the solution was re-
moved from the heating bath and allowed to cool at room temperature
in open air.
After the synthesis, each colloidal solution was transferred into a
dialysis membrane (Mw cut-off: 12−14 kDa, Medicell Membranes Ltd)
placed in a DI water bath for 24 h. This was performed in order to re-
move unreacted precursors and excess of reducing agent.
2.2. Membrane modification and nanoparticle adsorption
The tubular Teflon AF-2400 membrane (internal diameter, ID:
0.8 mm, outer diameter, OD: 1.0 mm, length: 30 cm, Biogeneral) was
functionalized with PDA. The polymerization of dopamine was per-
formed in a basic environment and triggered by oxygen dissolved in the
solution. The procedure was taken from Messersmith and co-workers
[39,40]. All polymeric solutions were pumped using plastic syringes
(20 mL, HSW) and a syringe pump (PHD ULTRA, Harvard Apparatus).
Polydopamine was prepared mixing 16 mg dopamine hydrochloride
(Sigma Aldrich) and dissolving it in a 8 mL buffer solution at a pH of 8.5
2
. Materials and methods
(
Tris-HCl 0.010 M, 2BScientific). The solution was pumped at 17 μL/
2.1. Nanoparticle synthesis
min through the membrane for 8 h. Afterwards, the buffer solution was
pumped for 1 h at a flowrate of 20 μL/min. The membrane was then
dried in an oven (Lenton) at 60 °C overnight. The procedure for the
adsorption of the following polyelectrolytes was similar to the one
adopted by Dotzauer et al. and Liu et al. [24,46]. The anionic poly-
electrolyte solution was prepared by mixing 0.67 mL of poly(acrylic
acid) (PAA, Mw ∼5,000, Sigma Aldrich) with 3.33 mL of DI water and
117 mg of NaCl. The pH was increased by dropwise addition of 2 M
NaOH (Sigma Aldrich) until a pH value of 3 was reached. The poly-
electrolyte solution was then pumped at a flowrate of 20 μL/min for 2 h
through the PDA-modified membrane. Afterwards, DI water was
pumped at 40 μL/min for 1 h and subsequently the membrane was
Sodium tetrachloropalladate(II) (Na PdCl4, 99.99 %), potassium
2
tetrachloroplatinate(II) (K
2
PtCl , 99.99 %), polyvinylpyrrolidone (PVP,
4
Mn ∼55,000), potassium bromide (FT-IR grade, ≥99 %), L-ascorbic
acid (reagent grade), citric acid monohydrate (ACS reagent, ≥99.0 %)
and Pluronic® P123 (PEG-PPG-PEG, Mn ∼5,800) were purchased from
Sigma Aldrich.
2
.1.1. Pd nanospheres synthesis
Pd nanospheres were synthesized according to Piao et al. [43]. In
detail, 29.4 mg of Na
2
PdCl were dissolved in 1 mL of deionized (DI)
4
2

B. Venezia, et al.
Catalysis Today xxx (xxxx) xxx–xxx
Fig. 1. Graphical representation of the PDA/PAA/PAH layer-by-layer modification of a Teflon AF-2400 tubular membrane followed by adsorption of palladium
nanoparticles (PDA: polydopamine, PAA: poly(acrylic acid), PAH: poly(allylamine hydrochloride)).
placed in the oven at 60 °C for 2 h. The preparation of the cationic
polyelectrolyte was performed by dissolving 700 mg of poly(allylamine
hydrochloride) (PAH, Mw ∼17,500, Sigma Aldrich) in 2 mL DI water
and 58 mg NaCl. The mixture was stirred until a clear and homogeneous
solution was obtained. The pH was raised to 6.5 by gradual dropwise
addition of 2 M NaOH. The final polymeric mixture was pumped at a
flowrate of 20 μL/min through the membrane for 2 h. DI water was then
pumped at 40 μL/min for 2 h, followed by drying in the oven at 60 °C
for 2 h. Nanoparticle adsorption was carried out by pumping the na-
noparticle solution at 10 μL/min through the LbL-modified Teflon AF-
detect the surface elemental composition and the palladium oxidation
state on the modified membrane. The XPS data were calibrated to an
ionization attributed to adventitious carbon at 285.0 eV. To determine
the metal loading of the catalytic membranes before and after reaction,
inductively coupled plasma mass spectroscopy (ICP-MS, 820, Varian)
was used. For this characterization, the membranes were cut into pieces
and aqua regia (3 volumes HCl, 1 volume HNO ) was used to dissolve
3
the metal nanoparticles to obtain a leachate, which was used for ana-
lysis.
2
400 tubular membrane for 12 h. Fig. 1 is a graphical representation of
2.4. Nitrobenzene hydrogenation
the steps for making the Teflon AF-2400 membrane catalytic, involving
an LbL-modification of the internal surface followed by nanoparticle
adsorption.
Nitrobenzene hydrogenation was employed as model reaction to
perform a comparative study of the catalytic performance of the pal-
ladium-based nanoparticles. Scheme 1 shows the generally accepted
reaction mechanism [47,48]. There are two relevant reaction pathways.
The first is the direct three-step reduction of the nitro group into the
nitroso, hydroxylamine and amine group. The second pathway involves
a condensation reaction of nitrosobenzene and N-phenylhydroxylamine
to produce azoxybenzene, which can be further reduced into aniline in
a three-step hydrogenation involving the consecutive production of
azobenzene and hydrazobenzene. However, the mechanism for ni-
trobenzene hydrogenation still represents a topic of research and a clear
reaction mechanism has not been established yet [47,49].
2.3. Materials characterization
Prior to adsorption, the nanoparticles were characterized by trans-
mission electron microscopy (TEM, 2100 EXii, 120 kV acceleration
voltage, JEOL) in order to determine their average size and mor-
phology. Surface characterization analyses were performed before re-
action. Pieces from the membranes were cut along the axial dimension
and flattened. Atomic force microscopy (AFM, Dimension Icon, Bruker)
was used for the visualization of the Teflon AF-2400 membrane before
and after the LbL modification, and after nanoparticle deposition. The
AFM was used in Peak Force Tapping Mode at 2 kHz Peak Force
Frequency and 0.5–1 Hz line rate at 256 scans per line resolution. Post-
processing of the raw pictures was performed by a second degree
polynomial flattening. X-ray photoelectron spectroscopy (XPS, K-alpha,
Al source 1486.6 eV, Thermo Scientific) was employed in order to
2.5. Reactor design and operation
The LbL-modified catalytic membranes were cut to a length of
2
20 cm, with an inner surface area of 5 cm covered with nanoparticles.
They were assembled inside a 15 cm long polytetrafluoroethylene
Scheme 1. Reaction pathways for the reduction of nitrobenzene to aniline.
3

B. Venezia, et al.
Catalysis Today xxx (xxxx) xxx–xxx
(
PTFE) tube (ID: 0.063″, OD: 1/8”, Thames Restek), in a tube-in-tube
define product selectivity for aniline, SAN, and nitrosobenzene, SNS,
configuration, using two polyether ether ketone (PEEK) T-junctions
respectively. Here CAN,out and CNS,out are the concentrations of aniline
(
Upchurch) that sealed the membrane to the inlet and outlet tubes. Pure
and nitrosobenzene at the outlet of the reactor.
hydrogen from a hydrogen generator (PH200, Peak Scientific) was
pressurized in a dead-end configuration outside the membrane (the gas
outlet T-junction was sealed). Hydrogen was delivered from the other
T-junction, permeated through the membrane and reacted on the pal-
ladium-based nanoparticles with nitrobenzene which flowed inside the
membrane. A solution of nitrobenzene in the concentration range of
CNB,in
CNB,out
X =
CNB,in
(1)
(2)
(3)
CAN,out
CNB,in CNB,out
SAN =
3
0−100 mM in ethanol (both Sigma Aldrich) was placed inside an 8 mL
C
NS,out
S
NS
=
stainless-steel syringe (Harvard Apparatus) and was pumped by means
of a syringe pump (PHD ULTRA, Harvard Apparatus) at a constant
flowrate of 15 μL/min. This corresponded to a residence time of ∼5 min
inside the reactor. The syringe was connected to the inlet of the tube-in-
tube reactor, while the outlet was connected to a pressure sensor
CNB,in
CNB,out
The average nitrobenzene consumption rate, r, was calculated ac-
cording to Eq. (4), where FNB,in is the inlet molar flowrate of ni-
trobenzene and nmetal the molar amount of palladium metal in the re-
actor. For the Pt-Pd nanodendrites the total amount of palladium and
platinum metal was taken into account, despite the latter being much
higher than the former. To estimate the activity of surface atoms, a
turnover frequency, TOF, was defined according to Eq. (5), where R is
the ratio between the number of bulk atoms below the first external
atomic layer to the total amount of metal atoms in the nanoparticle.
Hence, 1-R is the nanoparticle dispersion and its estimation is reported
in the Supplementary Information.
(
P X 309, Omega) and to an adjustable back-pressure regulator (BPR-01,
Zaiput). These were used to keep the liquid pressure inside the mem-
brane at 6 bar, while the hydrogen was stagnant and pressurized at
5
bar outside the membrane. The lower gas pressure was used to pre-
vent the formation of bubbles into the liquid. Fig. 2 shows the sche-
matic of the set-up. The tube-in-tube reactor was kept inside a bath
filled with water and placed on a hotplate (US152, Stuart). A constant
and uniform water temperature of 30 °C was maintained and measured
by a thermocouple immersed inside the bath and connected to the
hotplate.
X FNB,in
nmetal
r =
(4)
(5)
Samples were analyzed by a gas chromatograph (7820A, Agilent
Technologies) using a flame ionization detector (FID), a HP-INNOWAX
r
TOF =
1
R
(
19091−133) capillary column and a liquid auto-sampler. The two
main products detected were aniline and nitrosobenzene. Taking into
account the nitrobenzene, aniline and nitrosobenzene species at the
outlet, the nitrogen balance was calculated relative to the concentration
of nitrobenzene at the inlet. This value ranged between 98 % and 102 %
for all the experiments. Nitrobenzene conversion, X, was calculated
using Eq. (1), where CNB,in and CNB,out are the concentration of ni-
trobenzene at the inlet and the outlet of the reactor. Eqs. (2) and (3)
3
. Results and discussion
.1. TEM characterization of nanoparticles
Nanoparticle size and morphology were analyzed via TEM and size
3
distribution was obtained using Pebbles [50]. At least 120 particles per
Fig. 2. Experimental set-up schematic of the tube-in-tube LbL-modified Teflon AF-2400 membrane reactor for the continuous hydrogenation of nitrobenzene on
palladium-based nanoparticles.
4

B. Venezia, et al.
Catalysis Today xxx (xxxx) xxx–xxx
Table 1
nanoparticles. The main Pd 3d5/2 peak observed for the cubical nano-
particles was at 335.0 eV, attributed to metallic palladium. This sug-
gests that the Pd cubical nanoparticles displayed less oxidation com-
pared to the Pd spherical, truncated octahedral and dendritic
nanoparticles.
Average particle size along with standard deviation of palladium-based nano-
particles, obtained by TEM.
Nanoparticle
Average particle size, nm
Pd nanospheres
3.7 ± 0.6
3.9 ± 0.6
24.9 ± 3.7
6.2 ± 1.2
Pd truncated nano-octahedra
Pd nanocubes
3.4. Hydrogenation of nitrobenzene in flow
Pt-Pd nanodendrites
3
.4.1. Effect of substrate concentration
Continuous reaction experiments with the LbL-modified catalytic
sample were analyzed. Average nanoparticle sizes are presented in
Table 1 along with their standard deviation, while TEM pictures are
reported in the Supplementary Information.
tubular membrane were performed over more than 6 h. The effect of
two nitrobenzene concentrations, namely 30 mM and 100 mM, was
studied using palladium nanospheres. Conversion and product se-
lectivities are reported in Fig. 6. Nitrobenzene conversion for the lower
concentration reached completion over the whole reaction period,
along with a selectivity to aniline in the range of 95–100 %. Conversely,
for the case of 100 mM, conversion dropped from 36 % to 26 % after 6 h
of operation. Selectivity to aniline ranged between 85–88 % and ni-
trosobenzene selectivity increased to 15 % in the last hours of reaction.
This difference can be ascribed to the fact that at a higher concentration
of nitrobenzene, the nitrosobenzene produced in the first reaction step
could not be further hydrogenated under the same catalyst contact
time. However, at a lower substrate concentration the nitrosobenzene
produced could be fully hydrogenated into aniline. Moreover, using
around 3-fold higher substrate concentration, conversion dropped by
ca. a factor of three, from 95 to 100 % to 30–40 %. The maximum
possible hydrogen supply rate through the membrane was estimated to
be 2.9 mL/min, assuming full consumption of hydrogen at the catalyst
location (see Supplementary Information). On the other hand, con-
sidering a conversion of 100 % of nitrobenzene to aniline at a flowrate
of 15 μL/min and an inlet concentration of 100 mM, the hydrogen
consumption rate was assessed to be only 0.1 mL/min. It is therefore
reasonable to assume that there was no limitation of hydrogen supply.
3.2. Atomic force microscopy analysis of membranes
Atomic force microscopy micrographs of the membrane surface
before adsorption of the nanoparticles are presented in Fig. 3. The
pristine Teflon membrane surface presents a relatively smooth surface
with an elevation difference of around 10 nm between the lowest and
the highest point. Fig. 3 also displays the modified surface of the
membrane after the LbL sequential adsorption of the polyelectrolyte
multilayers. It is evident that the surface lost its smoothness after the
modification, presenting bulges with an average size of 100 nm and
elevation of ca. 25 nm. Fig. 4 shows the micrographs of the inner sur-
face of the membrane with the adsorbed palladium-based nano-
particles. The surface covered with spherical nanoparticles presented
scattered bulges, which can be ascribed to the polyelectrolyte assembly,
features with a characteristic size in the order of 10 nm and singularly
resolved nanoparticles. The surface covered with truncated octahedral
nanoparticles displayed homogeneously adsorbed singular nano-
particles. It is possible to observe that some bulges created during the
LbL modification provided higher adsorption surface and are covered
by the nanoparticles. For the case of the dendritic Pt-Pd nanoparticles,
these could be observed in the form of clusters of nanoparticles with
dimensions larger than 40 nm. The cubes, which were approx. 25 nm
were singularly resolved by AFM analysis. They were scattered over a
wavy surface and some of them were found buried inside the polymeric
support.
3.4.2. Comparison of catalytic performance of different types of
nanoparticles
Reaction experiments involving the use of palladium-based nano-
particles with different morphology and size were carried out over a
period of 6−7 h of continuous reaction. The nitrobenzene inlet con-
centration was kept constant at 100 mM, along with a temperature of
30 °C and hydrogen pressure of 5 bar and results are presented in Fig. 7.
As seen in Fig. 7, the nanocubes showed 96 % conversion in the first
hour of operation, but this value steadily dropped to 54 % after 6.5 h.
Over the whole reaction period, selectivity to nitrosobenzene increased
from 5 % to 12 % while aniline selectivity dropped from 95 % to 88 %,
possibly because of the simultaneous drop in conversion. The palladium
truncated nano-octahedra exhibited much lower but steady conversion
(7–8 %) after 2 h reaction time and higher selectivity to nitrosobenzene,
3.3. X-ray photoelectron spectroscopy analysis of membranes
XPS revealed palladium and platinum on the inner surface of the
catalytic membranes (see Supplementary Information). Fig. 5 shows the
high resolution Pd 3d overlay spectra of all samples. Despite the low
signal-to-noise ratio, it is possible to observe that the peaks of the
spherical, truncated octahedral and dendritic nanoparticles are shifted
to approx. 0.3 eV higher energy as compared to the cubical
Fig. 3. AFM micrographs of (a) pristine Teflon AF-2400 membrane and (b) LbL-modified PDA/PAA/PAH membrane inner surface before adsorption of nanoparticles.
5

B. Venezia, et al.
Catalysis Today xxx (xxxx) xxx–xxx
Fig. 4. AFM micrographs of the LbL-modified PDA/PAA/PAH tubular Teflon AF-2400 membrane inner surface after the adsorption of (a) Pd spherical, (b) Pd
truncated octahedral, (c) Pd cubical and (d) Pt-Pd dendritic nanoparticles.
ICP-MS analysis revealed a palladium and platinum metal content
adsorbed in the modified tubular membranes in the range of
2
0
.5−12 μg/cm . These values were determined both before and after
the reaction and were used to estimate the consumption rate and the
turnover frequency of nitrobenzene reacting on the catalytic nano-
particles. Table 2 shows the results.
Palladium nanospheres showed a conversion of 36 % in the first hours
2
of reaction, with a loading of 1.09 μg/cm , which resulted in a con-
sumption rate of 629 molreactant/(molcatalyst⋅h). However, after reaction the
nanospheres-loaded membrane presented leaching of around 12 % of the
initial loading and a nitrobenzene conversion decrease of 28 %. These
values suggest that deactivation cannot be only explained by a mere
leaching effect of the catalyst from the reactor, but to other parallel effects,
such as poisoning or particle restructuring. Similarly to the nanospheres,
but with half the catalyst loading, the truncated nano-octahedra showed
an initial rof similar value, 538 molreactant/(molcatalyst⋅h). This can be as-
cribed to their similar average particle size. However, unlike the nano-
spheres, the truncated nano-octahedra exhibited almost no leaching
during reaction and their deactivation should be attributed to other effects.
The reaction rate of the Pt-Pd nanodendrites was assessed taking
into account both the platinum and the palladium loading. It decreased
Fig. 5. X-ray photoelectron spectra of the Pd 3d region of the spherical, trun-
cated octahedral, cubical and dendritic nanoparticles adsorbed over the LbL-
modified PDA/PAA/PAH Teflon AF-2400 tubular membrane inner surface.
−1
from ∼340 to ∼260 h , from the start to the end of the reaction.
However, a leaching of 29 % from the original platinum loading of
with an aniline selectivity of 75 %. An even lower selectivity to aniline,
at higher conversion values, was recorded for the platinum-palladium
nanodendrites. Despite a first drop from 35 % to 21 % in the first 2 h of
operation, conversion was stable for more than 4 h, with a selectivity to
nitrosobenzene and aniline equal to 40 % and 60 % respectively. In the
work of Kataoka et al. a similar selectivity of 40 % to nitrosobenzene
was achieved using platinum nanoparticles immobilized inside a cata-
lyst-coated capillary microreactor [48]. An inlet concentration of
2
3.38 μg/cm was observed. The same relative drop was observed for the
2
2
palladium loading which decreased from 0.17 μg/cm to 0.12 μg/cm .
This effect might be among the possible causes for the drop in con-
version from 36 % to 19 % (-47 %). However, based on the external
layer of atoms and assuming the nanoparticle as a sphere, the nano-
−1
dendrites exhibited an initial TOF of 1414 h , which was of the same
order of magnitude as the initial TOF of the palladium nanospheres
−
1
−1
5
0 mM nitrobenzene at a flowrate of 10 μL/min was employed along
(1628 h ) and truncated nano-octahedra (1457 h ).
with 100 μL/min hydrogen flowrate.
Results from batch hydrogenations of nitrobenzene using palladium
6

B. Venezia, et al.
Catalysis Today xxx (xxxx) xxx–xxx
Fig. 6. Nitrobenzene conversion, X, aniline selectivity, SAN and nitrosobenzene selectivity, SNS, during the hydrogenation of nitrobenzene in the membrane reactor
with the Pd nanospheres. Nitrobenzene inlet concentration, (a) 30 mM, (b) 100 mM. Liquid flowrate, 15 μL/min; liquid pressure, 6 bar; hydrogen pressure, 5 bar;
temperature, 30 °C.
nanoparticles resulted in reaction rates in the same order of magnitude as
our work. In the work of Huang et al., 2.6 nm Pd nanoparticles stabilized
inside P123 micelles in water were tested at 45 °C and 30 bar hydrogen
pressure [42]. The reaction rate was 245 molreactant/(molcatalyst⋅h) after
such a study was performed for 2-methyl-3-butyn-2-ol hydrogenation
over unsupported PVP-stabilized Pd nanocrystals [6]. In that work, it
was concluded that the activity of nanocubes and nano-octahedra had a
similar trend, while nanocuboctahedra had the lowest activity because
of the large presence of edge atoms, which disfavoured the 2-methyl-3-
butyn-2-ol hydrogenation.
2
h reaction time. Harraz et al. synthesized 4 nm Pd nanoparticles sta-
bilized by polyethylene glycol in ethanol and tested them at room tem-
perature and 1 bar H pressure. A reaction rate of 126 molreactant
molcatalyst⋅h) after 180 min reaction time was observed [51].
2
/
In our work, estimating the turnover frequency based on the surface
atoms of the nanoparticle, we found that the cubical nanoparticles
(
−1
To the best of our knowledge, comparative studies on the effect of
demonstrated the highest TOF, in the order of 3000 h . This was two
times higher than the nanospheres and truncated nano-octahedra.
However, the reactor with adsorbed nanocubes exhibited a much larger
palladium nanoparticle morphology in the batch or continuous hydro-
genation of nitrobenzene are not reported in the literature. However,
Fig. 7. Nitrobenzene conversion, X, aniline selectivity, SAN and nitrosobenzene selectivity, SNS, during the hydrogenation of nitrobenzene in the membrane reactor.
The reactor contained (a) Pd nanospheres, (b) Pd truncated nano-octahedra, (c) Pd nanocubes and (d) Pt-Pd nanodendrites. Nitrobenzene inlet concentration,
1
00 mM; liquid flowrate, 15 μL/min; liquid pressure, 6 bar; hydrogen pressure, 5 bar; temperature, 30 °C.
7

B. Venezia, et al.
Catalysis Today xxx (xxxx) xxx–xxx
Table 2
Palladium and platinum metal loading per membrane inner surface, nitrobenzene conversion, X, consumption rate, r, and turnover frequency, TOF, before and after
reaction for the Pd spherical, Pd truncated octahedral, Pd cubical and Pt-Pd nanoparticles.
Nanoparticle
2
TOF, h−1
Metal loading on the membrane, μg/cm
X, %
r, molreactant/(molcatalyst⋅h)
Start
End
Start
End
Start
End
Start
End
Pd nanospheres
1.09
0.56
12.18
3.38
0.17
0.96
0.53
4.08
2.37
0.12
36
15
96
36
36
26
7
629
538
150
341
515
239
252
257
1628
1457
2293
1414
1332
647
Pd truncated nano-octahedra
Pd nanocubes
54
19
19
3854
1067
Pt-Pd nanodendrites (Pt)
Pt-Pd nanodendrites (Pd)
conversion drop (−44 %), which can be attributed to Pd leaching of
4. Conclusions
2
6
6 % with respect to the initial loading of 12.18 μg/cm . An interesting
aspect of the nanocubes adsorption was the 12-fold higher loading that
could be achieved compared to the nanospheres, possibly arising from
their larger average particle size. The ratio between the average volume
of nanocubes (25 nm) and nanospheres (4 nm) can be estimated to be
A Teflon AF-2400 tubular membrane was modified via adsorption of
polyelectrolyte multilayers onto which palladium-based nanoparticles
with different size and shapes were adsorbed. The use of a Teflon AF-
2400 membrane, allowed operating at high liquid pressure without gas
bubble formation in the liquid. AFM showed the presence of adsorbed
nanoparticles on the inner surface of the membrane, along with a wavy
topography of the membrane surface after modification with the
polyelectrolyte assembly. X-ray photoelectron spectroscopy highlighted
a shift to high binding energies for the palladium in the spherical,
truncated octahedral and dendritic nanoparticles with respect to the
cubical. This indicated a higher oxidation state for the these nano-
particles compared to the nanocubes. The catalytic membranes were
assembled in a tube-in-tube reactor configuration and tested for ni-
trobenzene hydrogenation in flow. The palladium nanospheres of 4 nm
size showed a stable conversion close to 100 % for over 6 h of con-
tinuous reaction with almost no by-product formation, when using
30 mM nitrobenzene inlet concentration; however, deactivation was
observed with 100 mM nitrobenzene concentration. A comparative
study of the performance of different palladium-based nanoparticles
2
44. However, the palladium mass ratio between the adsorbed nano-
cubes and the adsorbed nanospheres was only ∼12, which indicates
that that a lower number of nanocubes were adsorbed on the membrane
surface with respect to nanospheres.
The enhanced TOF of the nanocubes can be connected to the XPS
results. Their higher reduction degree, as evidenced by their 0.3 eV shift
to lower energy, might be the cause for the enhanced activity.
Palladium metal has in fact the important role of dissociating and
providing surface-reactive hydrogen atoms in the nitrobenzene reaction
[
52,53]. The higher catalytic activity of the nanocubes can also be re-
lated to their larger size compared to the other nanoparticles. Carturan
et al. showed an increased TOF as the degree of metal dispersion de-
creased, in the batch hydrogenation of nitrobenzene [54]. The authors
attributed this to the fact that the larger palladium particles favoured
hydrogen dissociation and reduction of the metallic surface, oxidized by
the adsorption of nitrobenzene.
revealed
a high consumption rate of nitrobenzene of around
Palladium nanoparticles supported on the inner surface of tubular
membranes can represent a suitable alternative to packed-bed reactors
where internal mass transfer resistances occur inside catalyst particles
600 molreactant/(molcatalyst⋅h) in the first hours of operation using the Pd
nanospheres, attributed to their small size. However, the highest turn-
over frequency based on the nanoparticle surface atoms was observed
−1
and the wetting liquid. Experiments with Au-Pd/TiO
2
catalyst packed
for the nanocubes, which was in the order of 3000 h . This was pos-
sibly due to the lower oxidation state of cubical nanoparticles. Leaching
of metals from the reactor was found to be the highest for the one
loaded with nanocubes, while almost no losses were recorded in the
case of the truncated nano-octahedra.
inside a Teflon AF-2400 tubular membrane were performed in our
group and indicated the presence of internal mass transfer resistance
[
36]. Use of immobilized palladium nanoparticles in flow reactors for
the hydrogenation of nitrobenzene can be found in the work of Liu et al.
26]. In that work, a polypropylene hollow fibre membrane was mod-
[
ified on its outer surface with PDA followed by poly(sodium 4-styr-
enesulfonate) and PAH. This was assembled in a tube-in-tube mem-
brane configuration and nitrobenzene hydrogenation was performed in
the annulus, while hydrogen was pressurized inside the tubular mem-
brane. The nanoparticles were synthesized in situ after adsorption and
reduction of the palladium precursor on the external surface of the
modified hollow fibre membrane. After 10 h continuous reaction, the
catalyst loading was ca. 47 μg, and the reactor could achieve a ni-
trobenzene conversion of 48 %. The inlet substrate concentration was
CRediT authorship contribution statement
Baldassarre Venezia: Conceptualization, Methodology, Writing -
original draft, Writing - review & editing, Visualization, Investigation.
Luca Panariello: Conceptualization, Investigation. Daniel Biri:
Methodology, Investigation. Juhun Shin: Data curation, Investigation.
Spyridon Damilos: Investigation, Visualization. Anand N.P.
Radhakrishnan: Data curation, Investigation. Chris Blackman:
Writing
- review & editing, Resources. Asterios Gavriilidis:
3
0 mM with a liquid flowrate of 10 μL/min resulting in an average
Supervision, Writing - review & editing, Resources.
consumption rate of around 19 molreactant/(molcatalyst⋅h), which was
significantly smaller than the one achieved by the nanospheres in our
work. This can possibly be due to the much larger average nanoparticle
size of 100 nm. In a similar work, using the same inlet nitrobenzene
concentration and liquid flowrate, nitrobenzene hydrogenation was
Declaration of Competing Interest
No competing interests to declare.
performed over a Pd/γ-Al
2
O catalyst coated on a poly(dimethylsi-
3
loxane) flat membrane with hydrogen flowing on the other side of the
membrane [55]. This showed 52 % conversion at 110 mM nitrobenzene
inlet concentration, 10 μL/min liquid flowrate and using 1.9 mg catalyst
coated on the membrane. The catalytic layer produced in that work was
homogeneously distributed over the membrane with a depth of 1 μm.
Acknowledgements
The authors thank EPSRC for funding (grant EP/L003279/1 and for
BV’s DTP studentship), Dr Richard Thorogate for the AFM analysis and
Prof John McArthur for the ICP-MS analysis.
8

B. Venezia, et al.
Catalysis Today xxx (xxxx) xxx–xxx
Appendix A. Supplementary data
[28] A.R. Tao, S. Habas, P. Yang, Small 4 (2008) 310–325.
[
[
29] L. Ouyang, D.M. Dotzauer, S.R. Hogg, J. Macanás, J.-F. Lahitte, M.L. Bruening,
Catal. Today 156 (2010) 100–106.
Supplementary material related to this article can be found, in the
online version, at doi:https://doi.org/10.1016/j.cattod.2020.03.062.
30] T.C. Merkel, I. Pinnau, R. Prabhakar, B.D. Freeman, Gas and vapor transport
properties of perfluoropolymers, in: Y. Yampolskii, I. Pinnau, B.D. Freeman (Eds.),
Materials Science of Membranes for Gas and Vapor Separation, John Wiley & Sons,
Chichester, 2006, pp. 251–270.
References
[
[
31] I. Pinnau, L.G. Toy, J. Memb. Sci. 109 (1996) 125–133.
32] M. Brzozowski, M. O’Brien, S.V. Ley, A. Polyzos, Acc. Chem. Res. 48 (2015)
349–362.
[
1] D. Astruc, Transition-metal nanoparticles in catalysis: from historical background to
the state-of-the art, in: D. Astruc (Ed.), Nanoparticles and Catalysis, 2008, pp. 1–48
Weinheim.
[
33] M. O’Brien, N. Taylor, A. Polyzos, I.R. Baxendale, S.V. Ley, Chem. Sci. 2 (2011)
1250–1257.
[
[
[
[
[
2] B. Roldan Cuenya, Acc. Chem. Res. 46 (2013) 1682–1691.
3] F. Zaera, Catal. Lett. 142 (2012) 501–516.
[34] P. Koos, U. Gross, A. Polyzos, M. O’Brien, I. Baxendale, S.V. Ley, Org. Biomol. Chem.
9 (2011) 6903–6908.
4] H.-J. Freund, Chem. Eur. J. 16 (2010) 9384–9397.
[35] A. Polyzos, M. O’Brien, T.P. Petersen, I.R. Baxendale, S.V. Ley, Angew. Chem. Int.
Ed. 50 (2011) 1190–1193.
5] S. Cheong, J.D. Watt, R.D. Tilley, Nanoscale 2 (2010) 2045–2053.
6] M. Crespo-Quesada, A. Yarulin, M. Jin, Y. Xia, L. Kiwi-Minsker, J. Am. Chem. Soc.
[36] G. Wu, A. Constantinou, E. Cao, S. Kuhn, M. Morad, M. Sankar, D. Bethell,
G.J. Hutchings, A. Gavriilidis, Ind. Eng. Chem. Res. 54 (2015) 4183–4189.
[37] B. Venezia, M. Douthwaite, G. Wu, M. Sankar, P. Ellis, G.J. Hutchings,
A. Gavriilidis, Chem. Eng. J. 378 (2019) 122250.
1
33 (2011) 12787–12794.
[
[
7] K. An, G.A. Somorjai, ChemCatChem 4 (2012) 1512–1524.
8] S. Mostafa, F. Behafarid, J.R. Croy, L.K. Ono, L. Li, J.C. Yang, A.I. Frenkel,
B.R. Cuenya, J. Am. Chem. Soc. 132 (2010) 15714–15719.
[
38] M. O’Brien, I.R. Baxendale, S.V. Ley, Org. Lett. 12 (2010) 1596–1598.
[
9] G. Prieto, J. Zečević, H. Friedrich, K.P. de Jong, P.E. de Jongh, Nat. Mater. 12
[39] H. Lee, S.M. Dellatore, W.M. Miller, P.B. Messersmith, Science 318 (2007) 426–430.
[
40] J.H. Ryu, P.B. Messersmith, H. Lee, ACS Appl. Mater. Inerfaces 10 (2018)
7523–7540.
(
2012) 34.
[
[
10] D. Astruc, F. Lu, J.R. Aranzaes, Angew. Chem. Int. Ed. 44 (2005) 7852–7872.
11] H.-U. Blaser, A. Indolese, A. Schnyder, H. Steiner, M. Studer, J. Mol. Catal. A Chem.
[41] J. Liebscher, R. Mrówczyński, H.A. Scheidt, C. Filip, N.D. Hădade, R. Turcu,
A. Bende, S. Beck, Langmuir 29 (2013) 10539–10548.
1
73 (2001) 3–18.
[
[
12] M.L. Toebes, J.A. van Dillen, K.P. de Jong, J. Mol. Catal. A Chem. 173 (2001)
5–98.
13] A.L. Dantas Ramos, Pd.S. Alves, D.A.G. Aranda, M. Schmal, Appl. Catal. A Gen. 277
[42] C. Huang, X. Wang, F. Yu, B. Yuan, C. Xie, S. Yu, Res. Chem. Intermed. 44 (2018)
1
3–26.
7
[43] Y. Piao, Y. Jang, M. Shokouhimehr, I.S. Lee, T. Hyeon, Small 3 (2007) 255–260.
[
44] B. Lim, M. Jiang, P.H.C. Camargo, E.C. Cho, J. Tao, X. Lu, Y. Zhu, Y. Xia, Science
324 (2009) 1302–1305.
(
2004) 71–81.
[
[
[
[
[
[
[
[
14] M.K. Richmond, S.L. Scott, H. Alper, J. Am. Chem. Soc. 123 (2001) 10521–10525.
15] C.P. Mehnert, D.W. Weaver, J.Y. Ying, J. Am. Chem. Soc. 120 (1998) 12289–12296.
16] C.M. Crudden, M. Sateesh, R. Lewis, J. Am. Chem. Soc. 127 (2005) 10045–10050.
17] K. Okumura, K. Nota, K. Yoshida, M. Niwa, J. Catal. 231 (2005) 245–253.
18] H.H. Kung, E.I. Ko, Chem. Eng. J. Biochem. Eng. J. 64 (1996) 203–214.
19] H. Wu, L. Zhuo, Q. He, X. Liao, B. Shi, Appl. Catal. A Gen. 366 (2009) 44–56.
20] C. Jiang, V.V. Tsukruk, Adv. Mater. 18 (2006) 829–840.
[45] B. Lim, M. Jiang, J. Tao, P.H.C. Camargo, Y. Zhu, Y. Xia, Adv. Funct. Mater. 19
(
2009) 189–200.
[46] J. Liu, X. Zhu, Q. Liao, R. Chen, D. Ye, H. Feng, M. Liu, G. Chen, Chem. Eng. J. 332
(
2018) 174–182.
[47] A. Corma, P. Concepción, P. Serna, Angew. Chem. Int. Ed. 46 (2007) 7266–7269.
[48] S. Kataoka, Y. Takeuchi, A. Harada, T. Takagi, Y. Takenaka, N. Fukaya, H. Yasuda,
T. Ohmori, A. Endo, Appl. Catal. A Gen. 427-428 (2012) 119–124.
21] V.V. Tsukruk, V.N. Bliznyuk, D. Visser, A.L. Campbell, T.J. Bunning, W.W. Adams,
Macromolecules 30 (1997) 6615–6625.
[49] E.A. Gelder, S.D. Jackson, C.M. Lok, Chem. Commun. (2005) 522–524.
[
50] S. Mondini, A.M. Ferretti, A. Puglisi, A. Ponti, Nanoscale 4 (2012) 5356–5372.
[
[
[
[
22] G. Decher, J.D. Hong, Ber. Bunsenge. Phys. Chem. 95 (1991) 1430–1434.
23] G. Decher, Science 277 (1997) 1232.
[51] F.A. Harraz, S.E. El-Hout, H.M. Killa, I.A. Ibrahim, J. Catal. 286 (2012) 184–192.
[52] B.D. Adams, A. Chen, Mater. Today 14 (2011) 282–289.
24] D.M. Dotzauer, J. Dai, L. Sun, M.L. Bruening, Nano Lett. 6 (2006) 2268–2272.
25] D.M. Dotzauer, A. Abusaloua, S. Miachon, J.-A. Dalmon, M.L. Bruening, Appl. Catal.
B 91 (2009) 180–188.
[53] M. Guo, H. Li, Y. Ren, X. Ren, Q. Yang, C. Li, ACS Catal. 8 (2018) 6476–6485.
[54] G. Carturan, G. Facchin, G. Cocco, G. Navazio, G. Gubitosa, J. Catal. 82 (1983)
5
6–65.
[
[
26] M. Liu, X. Zhu, R. Chen, Q. Liao, D. Ye, B. Zhang, J. Liu, G. Chen, K. Wang, Chem.
Eng. J. 354 (2018) 35–41.
[55] M. Liu, X. Zhu, R. Chen, Q. Liao, H. Feng, L. Li, Chem. Eng. J. 301 (2016) 35–41.
27] P. Sonström, M. Bäumer, Phys. Chem. Chem. Phys. 13 (2011) 19270–19284.
9