Decomposition of formic acid over silica encapsulated and amine functionalised gold nanoparticles

Article abstract of DOI:10.1016/j.jcat.2016.11.020

Formic acid has recently attracted considerable attention as a safe and convenient source of hydrogen for sustainable chemical synthesis and renewable energy storage. Here, we show that silica encapsulated and amine functionalised gold nanoparticles are h

Full text of DOI:10.1016/j.jcat.2016.11.020

Journal of Catalysis 345 (2017) 149–156
Contents lists available at ScienceDirect
Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat
Decomposition of formic acid over silica encapsulated and amine
functionalised gold nanoparticles
⇑
Jerrik Mielby, Andreas Jonas Kunov-Kruse, Søren Kegnæs
DTU Chemistry, Technical University of Denmark, Kemitorvet 207, DK-2800 Kgs. Lyngby, Denmark
a r t i c l e i n f o
a b s t r a c t
Article history:
Formic acid has recently attracted considerable attention as a safe and convenient source of hydrogen for
sustainable chemical synthesis and renewable energy storage. Here, we show that silica encapsulated and
amine functionalised gold nanoparticles are highly active catalysts for the production of hydrogen by
vapour phase decomposition of formic acid. The core-shell catalysts are prepared in a reverse micelle sys-
tem that makes it possible to control the size of the Au nanoparticles and the thickness of the SiO2 shells,
which has a large impact on the catalytic activity. The smallest gold nanoparticles are 2.2 0.3 nm in
diameter and have a turnover frequency (TOF) of up to 958 hꢀ1 at a temperature of 130 °C. Based on
detailed in situ ATR-FTIR studies and results from kinetic isotope labelling experiments we propose that
the active site is a low-coordinated and amine functionalised Au atom, while H-assisted formate decom-
position into CO2 and H2 is the rate limiting step.
Received 6 October 2016
Revised 14 November 2016
Accepted 15 November 2016
Keywords:
Encapsulation
Gold nanoparticles
Heterogeneous catalysis
Silica
Ó 2016 Elsevier Inc. All rights reserved.
Formic acid
Hydrogen
ATR-FTIR
Kinetic isotope effect
1. Introduction
Supported gold nanoparticles have recently attracted much
attention because of their surprisingly high catalytic activity for a
Hydrogen produced from renewable resources holds great pro-
mise for sustainable production of chemicals and clean energy.
Unfortunately, the physical properties of hydrogen gas make trans-
portation, handling and refuelling difficult. Formic acid (HCOOH)
has great potential as a hydrogen storage material and as a means
to utilise CO2 [1]. Furthermore, formic acid may be produced
directly from biomass, for instance by the catalytic oxidation of cel-
lulose using heteropoly acids [2]. In general, the decomposition of
formic acid may occur via dehydrogenation (1) or dehydration (2)
number of reactions [4–6], in particular CO oxidation [7]. Although
the intimate mechanistic details are still not fully understood, it is
generally accepted that several factors contribute to the high activ-
ity [8]. In particular, the gold nanoparticles must typically be less
than 10 nm in diameter so that under-coordinated, reactive gold
atoms exist in large numbers at the edges and corners of the par-
ticles [9]. Supported gold nanoparticles have also previously been
used to decompose formic acid in both liquid [10,11] and vapour
phases [12–15]. Ojeda and Iglesia [12] showed that highly dis-
persed gold catalysts decomposed formic acid with higher metal-
time yields than similar Pt catalysts in vapour phase. The authors
suggested that the high catalytic activity was caused by disperse
Au species (undetectable by TEM) and not from visible metal
nanoparticles (3–4 nm), which were active for CO oxidation and
remained stable during thermal treatment. The decomposition of
formic acid has also been investigated by means of periodic density
functional theory calculations [16,17]. In particular, Yoo et al. [17]
studied the reaction on Ag, Cu, Pd and Pt. The study was extended
to other transition metals by scaling the adsorption energies of the
involved reaction species with two independent descriptors,
namely the CO and OH adsorption energies. Small Au clusters of
0.8 nm in diameter (13 Au atoms) were found to bind CO and OH
more strongly than bulk gold, which brought the activity of Au very
close to the top of the volcano plot. The small Au clusters were
HCOOH ! CO2 þ H2
ð1Þ
HCOOH ! CO þ H2O
ð2Þ
The two reactions pathways are linked by the well-known
water-gas shift reaction
CO þ H2O ¢ CO2 þ H2
ð3Þ
From the perspective of energy storage, the production of pure
H2 remains an important challenge because even small amount of
CO severely deactivates the platinum in fuel cell catalysts. Alterna-
tively, the efficient activation of formic acid may be exploited for
transfer hydrogenations in organic synthesis [3].
⇑
Corresponding author.
E-mail address: skk@kemi.dtu.dk (S. Kegnæs).
http://dx.doi.org/10.1016/j.jcat.2016.11.020
0021-9517/Ó 2016 Elsevier Inc. All rights reserved.

150
J. Mielby et al. / Journal of Catalysis 345 (2017) 149–156
even found to be more active and selective than Pt, which is in
good agreement with the results reported by Ojeda and Iglesia
[12].
mation of CO and CO2 as function of the reaction temperature as
measured inside the reactor. The formation of other important
products (in particular H2) was followed by an online Pfeiffer
Vacuum ThermoStar mass spectrometer (MS). All catalysts were
tested under the same reaction condition using a preprogrammed
heating profile going from 20 to 200 °C and then back to 20 °C by
2 °C/min. All kinetic data were collected after the temperature
had reached 200 °C. The kinetic isotope effect was investigated
by replacing HCOOH with HCOOD, DCOOH and DCOOD,
respectively.
Here, we demonstrate that gold nanoparticles encapsulated in
silica are highly effective catalysts for the vapour phase decompo-
sition of formic acid under mild conditions. The catalysts are pre-
pared by means of the reverse micelle methodology recently
reported by Zhang et al. [18]. In this approach, the gold precursor
is first precipitated with an aqueous solution of ammonia and then
subjected to a controlled silica coating in a reverse micelle system
comprised of a commercial surfactant (Brij-10) in cyclohexane.
Furthermore, the method takes advantage of the strong interac-
tions between gold and (3-aminopropyl)trimethoxysilane
(APTMS), which gives a good control over the size and encapsula-
tion of the Au nanoparticles. The smallest and most active gold
nanoparticles were 2.2 0.3 nm in diameter and decomposed for-
mic acid with a turnover frequency (TOF) of up to 958 hꢀ1 at a tem-
perature of 130 °C. To the best of our knowledge, these are the first
detailed studies of the decomposition of pure formic acid over sil-
ica encapsulated Au nanoparticles in vapour phase, although Yadav
et al. previously reported good results for the decomposition of
sodium formate and formic acid mixtures in liquid phase using a
similar catalyst design [11]. Based on in situ ATR-FTIR spectroscopy
and results from kinetic isotope labelling experiments we propose
that the active site is a low-coordinated and amine functionalised
Au atom, while H-assisted formate decomposition into CO2 and H2
is the rate limiting step.
2.3. In situ ATR-FTIR spectroscopy
The reaction mechanism was investigated using a Thermo
Fisher iS50 Fourier transformation infrared (FTIR) instrument with
a Specac high-temperature Golden Gate diamond attenuated total
reflectance (ATR) accessory, which could heat the sample area up
to 300 °C. A custom build flow cell allowed the catalyst to be stud-
ied under reaction conditions by in situ ATR-FTIR. The catalyst was
gently pressed into a wafer and mounted in the ATR flow cell using
an inner piston, which pressed the catalyst against the ATR-
diamond to ensure intimate contact, while the formic acid vapour
was passed over it. A schematic drawing of the setup is given in the
supporting information (Fig. S10). Further details, descriptions and
drawings of the environmental ATR-FTIR device and the concept
behind it can be found elsewhere [19].
All in situ ATR-FTIR experiments were performed at 120 °C.
Before exposure to formic acid, the catalyst was first dried in a flow
of dry nitrogen (25 ml/min) for around 1 h at the same
temperature.
2. Experimental section
The procedure for the in situ experiments with DCOOD was sim-
ilar to the experiments described above. However, to exchange
mobile protons on the amine species, and thus make the interpre-
tation more straightforward, the catalyst was first stirred in D2O
for 48 h at 100 °C. The D2O treated catalyst was then dried in an
oven prior to the preparation for ATR-FTIR analysis. The enrich-
ment procedure did, however, not result in a complete exchange
of the mobile protons.
2.1. General synthesis
The Au@SiO2 core-shell catalysts were prepared according to a
modified literature procedure [18]. In a typical synthesis, 4.25 g
of Brij C10 was dissolved in 7.5 ml of cyclohexane at 50 °C. Then,
0.5 ml of HAuCl4 (0.25 M) was added dropwise under stirring, fol-
lowed by 0.8 ml of aqueous NH3 (25%), 150 ll APTMS and 2 ml of
tetraethyl orthosilicate (TEOS), which were added in order with
intervals of 2 min. The hydrolysis and condensation of TEOS were
allowed to proceed for 2 h at 50 °C followed by stirring overnight
with no heating. The product was collected by centrifugation,
washed with ethanol, dried at 80 °C and then reduced under
hydrogen gas (10% H2 in N2) for 2 h at 200 °C using a heating ramp
of 5 °C/min. In order to investigate the effect of the size of the Au
nanoparticles and the thickness of the SiO2 shells, we prepared
three other catalysts by increasing the amount of cyclohexane
(15 ml), decreasing the concentration of HAuCl4 (0.125 M) and
increasing the amount of TEOS (4 ml), respectively. Furthermore,
we prepared one catalysts comprised of pure SiO2.
2.4. Computational methods
Assignments of vibrational modes were carefully performed in a
combination of density functional theory (DFT) calculations and
reference experiments with deuterated formic acid. The DFT calcu-
lations were performed using the BP86 functional with the TZVP
basis set in Gaussian 09. No scaling factor was applied for the cal-
culated spectra. Silica surface was simulated with a small H3Si4O6-
AOH unit, which despite its simplicity, is known to give good
results for studies of vibrational modes [20]. Silica grafted propyl
amine was simulated using the analogous H3Si4O6A(CH2)3NH2.
Screenshots of the structures and selected vibrational modes are
shown in the supporting information Fig. S11. An overview of rel-
evant spectral assignments and approximate modes is also given in
supporting information Scheme S1.
2.2. Catalytic tests
The vapour phase decomposition of HCOOH was performed at
atmospheric pressure in a 3 mm quartz fixed-bed reactor. The for-
mic acid was introduced to the reactor by bubbling Ar (40 ml/min)
through pure formic acid kept in a thermostatic bath at 20 °C,
which resulted in gas composition of around 7% formic acid. The
reaction gas was preheated to the reaction temperature and then
passed through the reactor, which contained the fractionated cat-
3. Results and discussion
3.1. Characterisation
alyst (180–355
l
m). The amount of catalyst (50–100 mg) was
Fig. 1 shows a representative TEM image of the Au@SiO2 core-
shell catalyst as prepared by the general synthesis described in
Section 2.1. The image shows how the individual Au nanoparticles
were uniformly encapsulated in small spheres of SiO2 with a shell
thickness of 8–12 nm. TEM images of all the prepared catalysts are
shown in the supporting information.
adjusted to the Au loading in order to have a constant weight
hourly space velocity (WHSV) of around 138 g formic acid/
g Au hꢀ1. The reaction products, which were either CO2 and H2 or
CO and H2O, were quantified with an online Rosemount BINOS
100 non-dispersive infrared (NDIR) detector that followed the for-

J. Mielby et al. / Journal of Catalysis 345 (2017) 149–156
151
Fig. 3 shows the Au nanoparticle size distribution of four differ-
ent catalysts prepared using the reverse micelle method. The cata-
lysts are denoted by Au@SiO2 followed by the average size and
standard deviation in nm. The particle size distributions of the
Au nanoparticles were calculated from the measurement of around
300 nanoparticles by TEM. Fig. 3a shows the particle size distribu-
tion of the typical catalyst with a metal loading of 4.4 wt% Au and
an average size of 2.8 0.5 nm. The catalyst was prepared using
7.5 ml of cyclohexane, 0.5 ml 0.25 M HAuCl4 and 2.0 ml tetraethyl
orthosilicate (TEOS) according to Section 2.1. By increasing the
amount of cyclohexane to 15 ml we were able to decrease the aver-
age particle size from 2.8 0.5 nm to 2.6 0.3 nm, see Fig. 3b. By
decreasing the amount of gold to 0.5 ml 0.125 M HAuCl4 (2.2 wt%
Au) we were further able to decrease the average particle size to
2.2 0.3 nm, see Fig. 3c. In general, the diluted system resulted
in more narrow size distributions. We also tried to increase the
shell thickness by increasing the amount of TEOS to 4 ml. Fig. 3d
shows how the increased amount of TEOS decreased the average
size of the Au nanoparticles to 2.7 0.4 nm. The small decrease
in size and corresponding increase in the total number of particles
caused the shell thickness to increase slightly less than expected
from simple geometric calculations. While the general 4.4 wt%
Au@SiO2 2.8 0.5 catalyst (A) had an average shell thickness of
21.9 2.0 nm, the 2.2 wt% Au@SiO2 2.7 0.4 catalyst (D) had an
average shell thickness of 26.2 1.9 nm. Further attempts to
decrease the shell thickness by adding less TEOS without increas-
ing the size of the Au nanoparticles were not successful.
Fig. 1. TEM image of the Au@SiO2 2.8 0.5 core-shell catalyst.
The TEM image in Fig. 1 is in good agreement with the results
from X-ray diffraction (XRD) analysis shown in the supporting
information (Fig. S1). The diffraction pattern shows no clear peaks
from the core-shell catalyst since the SiO2 shells were amorphous
and the Au cores were relatively small (2.8 0.5 nm). The weak
diffraction peaks at 2h = 38, 44, 65 and 78 were assigned to Au
(111), Au(200), Au(220) and Au(311), respectively, in good
agreement with bulk gold. In general, the peaks are too weak to
estimate the size of the nanoparticles by line broadening analysis.
The surface area and pore volume were investigated by nitrogen
physisorption analysis performed at 77 K. The physisorption iso-
therm in Fig. 2 shows a significant hysteresis loop, which corre-
sponds to an average pore size distribution of around 11 nm as
determined by BJH-analysis of the isotherm desorption branch
(see Fig. S2 in the supporting information). The pore size distribu-
tion is in good agreement with the expected porosity related to the
inter-particle voids that exists between the individual spherical
particles. From the small amount of N2 that adsorbed at low partial
pressures, the isotherm also shows that the silica shells had very
limited porosity in the microporous range (<2 nm). A single point
read at p/p0 = 0.99 resulted in a total pore volume of 0.57 cm3/g,
while BET analysis resulted in a total surface area of 203 m2/g.
3.2. Catalytic activity
Fig. 4 shows the conversion of HCOOH as function of the tem-
perature. Since the total amount of Au was kept constant in all cat-
alytic experiments the catalytic activity can be directly compared
by the temperature at 50% conversion. While the core-shell cata-
lyst with the largest Au nanoparticles (2.8 0.5 nm) reached 50%
HCOOH conversion at a temperature of 132 °C, the catalyst with
slightly smaller nanoparticles (2.6 0.3 nm) reached 50% conver-
sion at 128 °C. The catalyst with the smallest Au nanoparticles
(2.2 0.3 nm) had a significantly higher activity and already
reached 50% conversion at 119 °C. The catalyst with increased
SiO2 shell thickness was less active than the other catalysts and
only resulted in 50% conversion at 140 °C.
Fig. 5 shows the selectivity towards H2 and CO2 as function of
the conversion in the same experiments (above 20% conversion
for clarity). The data show that the smallest gold nanoparticles
resulted in slightly better selectivity towards H2. We speculate that
the relatively small difference in selectivity may be explained by
the thermodynamic equilibrium, which favours the formation of
CO2 and H2 over CO and H2O at lower temperatures. In general,
all catalysts reached around 85–87% selectivity at full conversion,
which is lower than in liquid phase (up to 100%) [10,11], but signif-
icantly higher than previously reported for e.g. Au/TiO2 in vapour
phase (70%) [13]. Although Al2O3 previously has been shown to
favour dehydration [14], a recent study of different metal oxide
supports reports high selectivity for Au/Al2O3 (>99%) [21].
Table 1 shows a summary of the prepared catalysts, their aver-
age particle size distribution and catalytic activity. The turnover
frequency (TOF) was determined from the H2 yield and the Au dis-
persion, and is given in mol H2 per mol of surface Au atoms per
hour at 130 °C. The Au dispersion D (i.e. the surface-to-volume
atomic ratio) was calculated from the average particle size (d) by
the relation D = 1.16/d, assuming equal proportions of the three
low-index planes (111), (100) and (110) on the surface of a spher-
ical face-centred cubic Au particle [22]. The calculated TOFs show
the remarkable effect that the size has on the intrinsic activity of
Au nanoparticles. Since it previously has been shown that the rel-
25
2
S
=203 m /g
3
BET
V
=0.572 cm /g
tot
20
15
10
5
0
0
0.2
0.4
0.6
0.8
1
p/p0
Fig. 2. N2 physisorption isotherm of the Au@SiO2 2.8 0.5 catalyst.

152
J. Mielby et al. / Journal of Catalysis 345 (2017) 149–156
30
25
20
15
10
5
30
Au@SiO2 2.8 0.5
Au@SiO2 2.6 0.3
A)
B)
25
20
15
10
5
0
0
0
1
2
3
4
5
6
7
8
9
10
0
1
2
3
4
5
6
7
8
9 10
Diameter (nm)
Diameter (nm)
30
25
20
15
10
5
30
25
20
15
10
5
Au@SiO2 2.2 0.3
Au@SiO2 2.7 0.4
C)
D)
0
0
0
1
2
3
4
5
6
7
8
9
10
0
1
2
3
4
5
6
7
8
9 10
Diameter (nm)
Diameter (nm)
Fig. 3. Au nanoparticle size distribution of the four core-shell catalysts.
100
80
60
40
20
0
100
80
60
40
20
Au@SiO 2.2 0.3
2
Au@SiO 2.6 0.3
2
Au@SiO 2.2 0.3
2
Au@SiO 2.8 0.5
2
Au@SiO 2.6 0.3
2
Au@SiO 2.7 0.4
2
Au@SiO 2.8 0.5
2
0
Au@SiO 2.7 0.4
0
20
40
60
80
100
2
80
100
120
140
160
Conversion (%)
Temperature (°C)
Fig. 5. Selectivity towards H2 and CO2 as function of the conversion of HCOOH.
Fig. 4. Light-off curve for the conversion of HCOOH as function of the reaction
temperature.
than for highly dispersed Au/Al2O3 as reported by Ojeda [12]. Fur-
thermore, the activation energy of the catalyst with increased shell
thickness was only 61 kJ/mol. The lower activation energy indi-
cates that the diffusion of formic acid through the silica shell
may limit the reaction rate, which shows that both the Au
nanoparticle size distribution and the SiO2 shell thickness have a
large impact on the catalytic activity.
ative number of low-coordinated Au corner atoms increases signif-
icantly around 2 nm [23,24], we speculate that these sites may play
an important role in the decomposition of formic acid.
Entry C in Table 1 shows that the smallest gold nanoparticles
resulted in a TOF of 958 hꢀ1 at 130 °C. For comparison, Gazsi
et al. [14] previously reached a TOF of around 692 hꢀ1 using Au/
SiO2 prepared by deposition-precipitation under similar reaction
conditions.
Fig. 6 shows the Arrhenius plots based on the four light-off
curves in Fig. 4 from around 15% to 30% conversion. The apparent
activation energies Ea (as obtained from the slope of the Arrhenius
plots) were around 68–71 kJ/mol, which are up to 18 kJ/mol higher
3.3. Kinetic isotope effect
The nature of the reaction was further investigated by means of
the H/D kinetic isotope effect (KIE). If the dissociation of the acidic

J. Mielby et al. / Journal of Catalysis 345 (2017) 149–156
153
Table 1
Outline of the prepared catalysts, their average particle size distributions, catalytic activity and apparent activation energies.
Entry
Catalyst
Au loading (wt%)
Au size (nm)
SiO2 size (nm)
Au dispersion (D = 1.16/d)
TOFa at 130 °C (hꢀ1
)
Ea (kJ/mol)
A
B
C
D
E
Au@SiO2
Au@SiO2
Au@SiO2
Au@SiO2
SiO2
4.4
4.4
2.2
2.2
n/a
2.8 0.5
2.6 0.3
2.2 0.3
2.7 0.4
n/a
21.9 2.0
20.2 1.3
21.1 1.6
26.2 1.9
34.2 2.2
0.41
0.45
0.53
0.43
n/a
488
627
958
421
70
68
71
61
n/a
No activity
a
The turnover frequencies are given in mol H2 per mol of surface Au atoms per hour at 130 °C.
hydrogen desorption or water-gas shift may scramble the hydro-
gen isotopomers under the given reaction conditions.
-0.8
-1
Au@SiO 2.2 0.3 (E =71 kJ/mol)
2
app
Au@SiO 2.6 0.3 (E =68 kJ/mol)
2
app
app
Au@SiO 2.8 0.5 (E =70 kJ/mol)
2
2
Au@SiO 2.7 0.4 (E =61 kJ/mol)
app
3.4. Mechanistic studies
Although all kinetic data were collected after the temperature
had reached 200 °C, we found that the catalytic activity was higher
during the initial heating ramp, while subsequent cooling and
heating cycles resulted in a similar light-off curve, see supporting
information Fig. S8. TEM analysis of the used catalyst resulted in
the same particle size distribution (2.2 0.3 nm), which indicated
that no significant sintering occurred under the given reaction con-
ditions. Moreover, the slope of the corresponding Arrhenius plot in
Fig. S9 did not indicate any changes in the diffusion or reaction
mechanism. This indicated that the decrease in activity was more
likely caused by a partial chemical deactivation such as e.g. sub-
strate poisoning.
-1.2
-1.4
-1.6
-1.8
-2
2.5
2.6
2.7
In order to investigate the partial deactivation in more detail,
we performed a series of in situ Fourier transform infrared (FTIR)
spectroscopy experiments using a custom build flow reactor that
allowed the active catalyst to be studied under reaction conditions
by attenuated total reflection (ATR). A detailed description of the
ATR-FTIR reactor is given in the supporting information.
After drying in a flow of N2, the most distinctive features of the
pure SiO2 sample (Table 1, entry E) could be assigned to the typical
modes of silica with well resolved bands at 795, 950 and
1045 cmꢀ1, together with a range of intense but overlapping bands
from 1120 to 1280 cmꢀ1 [20] (see supporting information S12). In
particular the band at around 1045 cmꢀ1 due to transverse optic
1000/T
Fig. 6. Arrhenius plot for the decomposition of formic acid over the prepared
catalysts.
H atom is rate limiting for the decomposition of HCOOH, the KIE
would be expected to give normal values (i.e. rH/rD > 1) for HCOOD,
but not for DCOOH. Similarly, rate limiting formate decomposition
would give normal values for DCOOH, but not for HCOOD, while
rate limiting formation of H2 would give normal values for both
HCOOD and DCOOH [12].
With the Au@SiO2 2.8 0.5 catalysts all isotopomers resulted in
normal KIE-values, see Table 2. The values were of apparent kinetic
origin since no significant HCOOD/DCOOH scrambling was
observed under the given reaction conditions. The values were,
however, slightly lower than previously reported in the literature
[12,10]. We suggest that the lower KIE-values may be explained
by internal diffusion limitations, which appeared to be rate limit-
ing. Further efforts to increase the diffusion by decreasing the shell
thickness or increasing the porosity are therefore currently under-
taken in our laboratory.
As expected, the decomposition of HCOOH and DCOOD resulted
in 100% H2 and D2, respectively. However, in contrast to previous
results from liquid phase experiments, we found that switching
the substrate to HCOOD and DCOOH resulted in a distribution of
both H2, HD and D2. It is important to underline that these results
do not question that the decomposition of formic acid is a uni-
molecular reaction [12], but simply indicate that recombinative
SiAO stretching modes, mSiAO, was very intense [20,25]. The pure
silica samples showed a rather weak but well-resolved mSiAOH
stretching band at around 3700 cmꢀ1. Even at 120 °C physisorped
water was present as evident by the broad mHOH band peaking at
around 3500–3400 cmꢀ1 as well as by the weak but well resolved
OH bending band at around 1600–1650 cmꢀ1. Weak traces of the
partly decomposed surfactant used in the synthesis were observed
in the form of very weak CAH bands (3000–2800 cmꢀ1) as well as a
weak band at around 1575 cmꢀ1. The amine-free SiO2 sample was
then exposed to HCOOH in N2 at 120 °C, which showed that the
formic acid was physisorped on the surface in its protonated form,
most probably by hydrogen bonding to the SiAOH groups (DFT
structures and full spectra are shown in the supporting informa-
tion Figs. S11(b) and S13). As expected, the physisorped species
were desorbed again when the sample was flushed with pure N2.
The amine containing Au@SiO2 2.2 0.3 core-shell catalyst
(entry C in Table 1) was first investigated at 120 °C in a flow of
N2 (supporting information S12) and then in a flow of HCOOH in
N2 (Fig. 7(a) and supporting information Figs. S13 and S14). In a
similar manner, the catalyst was also exposed to DCOOD, see
Fig. 7(b). Due to the larger mass of deuterium, modes involving
stretching or bending of deuterium normally shift downwards. In
some cases, such as e.g. partially exchange, new modes also appear
due to change of symmetry. Fig. S11 and Scheme S1 in the support-
ing information show the structure of the relevant deuterated spe-
cies as simulated by DFT.
Table 2
Kinetic isotope effect for the vapour phase decomposition of formic acid over
Au@SiO2 2.8 0.5 at 140 °C.
Isotopomer
KIE = rH/rD
1.0
1.7
1.9
4.1

154
J. Mielby et al. / Journal of Catalysis 345 (2017) 149–156
In general, the spectra of the Au@SiO2 2.2 0.3 catalyst showed
It has previously been shown that the rate of HCOOH decompo-
sition over Au nanoparticles may be significantly increased in the
presence of an amine [10]. The irreversible blockage of amine sites
due to amide formation therefore offers a good explanation for the
initial deactivation discussed earlier (see supporting information
S6 and S7). The key function of the amine is to facilitate OH bond
cleavage, which leads to the formation of Au-formate species.
The Au-formate species may then undergo further dehydrogena-
tion and react with the protonated amine to give H2 along with
CO2 via a b-elimination pathway [12]. We therefore propose that
the irreversible formation of formamide may decrease the reaction
rate by decreasing the number of silica grafted amine groups,
which functions as effective deprotonation sites under mild reac-
tion conditions. The formation of the proposed amide species is
shown in red in Fig. 8 along with a simplistic overview of the pro-
posed mechanism.
The kinetics of the relative amide formation was investigated by
extracting the absorbance of the amide mC@O using a Gaussian
deconvolution methodology described in our previous work [26].
The resulting kinetic profiles are shown in Fig. 9.
Fig. 9 shows how the rate of amide formation with HCOOH and
DCOOD was almost identical in the beginning of the experiment.
After 10 min, however, the rates started to deviate and a strong
kinetic isotope effect became evident. We believe that the kinetic
the same characteristic features as the pure SiO2 sample. Further-
more, the spectra also revealed the presence of the silane grafted
propylamine, which was added in the form of APTMS during the
synthesis. The most distinctive features of the propylamine species
were the NAH stretching band (mNH) peaking at around 3300–
3200 cmꢀ1 as well as an aliphatic mCAH and dCAH bands at around
3000–2900 cmꢀ1 and 1350–1450 cmꢀ1, respectively, and a dNH
band at around 1590–1650 cmꢀ1 slightly overlapping with the
dOH band from water (1640 cmꢀ1).
The spectra changed rapidly as soon as the catalyst was exposed
to HCOOH (t = 0 min indicates the time for introducing HCOOH).
Most notably, an intense mC@O band was formed at 1577 cmꢀ1
,
which clearly showed that the main part of the adsorbed formic
acid was deprotonated. This was in direct contrast to the sample
of pure SiO2, although a small amount of physisorped formic acid
was also evident on the Au@SiO2 2.2 0.3 catalyst (mOACAO at
1720 cmꢀ1). The intense carboxylate band at 1577 cmꢀ1 was
shouldered to both sides by various dNH3 and dNH2 modes from
the ammine/ammonium absorbing in the same region. Further-
more, two bands also appeared at 1380 and 1360 cmꢀ1, which
were assigned to the formate dCAH and symmetric
mCAO, respec-
tively. In general, the adsorption and formation of ammonium for-
mate species were relatively fast and approached steady state after
a few minutes with both HCOOH and DCOOD.
The carboxylate band appeared even sharper during DCOOD
absorption with no neighbouring amine bands or shoulders. We
believe that this is because the ND modes of the fully deuterated
amine is shifted below 1200 cmꢀ1. In the beginning, however,
some mixed NA(D/H) bending appeared in the 1500–1350 cmꢀ1
region as broad and poorly resolved peaks. The deuterium
exchange of the amines was further evident from the NH stretch-
ing modes at around 3400–3200 cmꢀ1, which was shifted down-
wards to around 2530–2450 cmꢀ1 (see full range spectrum in
supporting information Fig. S15).
After the bands associated with ammonium formate species
reached steady state, a new intense band slowly started to appear
at 1665 cmꢀ1 (1635 cmꢀ1 for the DCOOD experiment). After fur-
ther exposure, the intensity of the 1577 cmꢀ1 band seemed to
decrease slightly with the formation of the species associated with
the 1665 cmꢀ1 band. We assigned the latter band to amide species
formed between the grafted amines and formic acid in a side reac-
tion, see Fig. 8.
Fig. 8. Simplistic overview of the dual cycle amine-Au mechanism for formic acid
decomposition. The off-cycle deactivation of amine adsorption sites is shown in red.
Fig. 7. In situ ATR-FTIR spectra recorded at 120 °C on the Au@SiO2 2.2 0.3 core-shell catalyst. The spectra are coloured with a gradient indicating time of exposure. Initial
yellow spectrum, is indicated with a black line for clarity. (a) The pre-dried catalyst being exposed to formic acid vapours in N2. (b) Deuterium enriched (48 h in D2O at 100 °C)
and pre-dried Au@SiO2 2.2 0.3 catalyst being exposed to formic acid-d2 vapours in N2. Selected assignments are shown.

J. Mielby et al. / Journal of Catalysis 345 (2017) 149–156
155
isotope effect may be explained by the need for breaking two
amine NAD bonds, while the similar rates during the first 10 min
may be explained by the high H content in the beginning of the
experiment.
0.4
0.3
0.2
0.1
0
DCOOD
HCOOH
After around 40 min of exposure (80 min for the DCOOD exper-
iment) the flow was switched to pure N2 gas in order to study the
conversion of the absorbed formic acid species and discriminate
the reactive surface species from nonreactive, see Fig. 11. These
experiments showed that the species associated with absorbed
protonated formic acid (peak at 1725 cmꢀ1) disappeared faster
than the formate species (see supporting information).
In general, the formate species were converted by a first order
type
reaction
(see
peaks
at
1577/1575 cmꢀ1
and
1360/1325 cmꢀ1, respectively). We therefore used the deconvolu-
tion methodology, which allowed us to determine the kinetic pro-
files of this reaction. Fig. 10 shows a logarithmic plot of these
profiles.
0
10 20 30 40 50 60 70 80
Time / minutes
The data in Fig. 10 show that a large part of the formate species
(ꢁ85%) was converted by a first order reaction. It appears that the
adsorption and transfer to the Au site occur relatively fast, while
the subsequent decomposition of HCOOꢀ is rate-limiting. The
remaining part of the formate (ꢁ15%) was converted at a lower
rate, which indicates that the conversion becomes controlled by
internal diffusion. By estimation, the ratio of amine used in the
synthesis and the amount of surface gold atoms is roughly 13:1.
This indicates that the catalyst has a large number of amine
adsorptions sites, from where the HCOOꢀ species can diffuse to
an active gold site.
Fig. 9. Formation of the amide species during HCOOH and DCOOD exposure of on
the core-shell Au@SiO2 2.2 0.3 catalysts at 120 °C. The integrated absorbance is
derived from the deconvoluted difference spectra of the 1665 cmꢀ1 and 1635 cmꢀ1
bands, respectively.
0.40
0.20
0.10
0.05
DCOOD
HCOOH
−1
−1.5
−2
y=−0.18x+−1.00
R2=0.95
The kinetic profile for the DCOOꢀ conversion deviated signifi-
cantly from the HCOOꢀ profile. The first half of the DCOOꢀ species
was converted with a first order rate constant, which was around
45% of that for HCOOꢀ. The remaining half of the DCOOꢀ species
were converted at a significantly lower rate. In a control experi-
ment we investigated the same effect on a fresh catalyst. By
switching between DCOOH/N2 and pure N2, we found that that
the difference between the DCOOꢀ and HCOOꢀ reaction rates
was less pronounced with a fresh catalyst with almost no amide
species (see details described in supporting information, including
Fig. S17 and Table S1).
y=−0.02x+−1.75
R2=0.94
y=−0.40x+−0.89
R2=0.998
−2.5
−3
y=−0.17x+−2.19
R2=0.98
−3.5
−4
0.02
0
5
10
15
20
25
30
35
The peak at 1665 cmꢀ1 was stable over time, which confirmed
our assignment as a nonreactive species. Fig. 11 shows more
clearly resolved bands became visible after the formate species
were removed. Based on DFT calculations, the bands at
Time / minutes
Fig. 10. Conversion of adsorbed formate species based on the integrated
absorbance of the deconvoluted spectra during N2 flushing of Au@SiO2 2.2 0.3
pre-saturated with HCOOH for 40 min and DCOOD for 80 min at 120 °C.
Au@SiO2 flushed with N2 @120 o
0 min
C
20
15
10
5
42 min
difference
46 min
52 min
Aliphatic(δCH
)
+
2
RNH , RNH ,water
2
(δNH+δO3H
)(a)
0
1800 1700 1600 1500 1400 1300
−1
Wavenumber / cm
Fig. 11. In situ ATR-FTIR spectra continuing experiments in Fig. 7, where the formic acid saturated Au@SiO2 2.2 0.3 catalyst was flushed with nitrogen at 120 °C. The spectra
are coloured with a gradient indicating time since experiment started (see Fig. 7). Initial yellow spectra are indicated by a black outline. (a) The flushing of the formic acid
saturated catalyst with N2, initial spectrum recorded prior to formic acid vapour exposure is show for comparison. The hatched area on (a) shows the difference between the
fresh and used catalyst, thus indicating the features on the amide species.

156
J. Mielby et al. / Journal of Catalysis 345 (2017) 149–156
1525 cmꢀ1 and 1387 cmꢀ1 further support the structure of the pro-
posed amide species.
[5] S. Kegnæs, J. Mielby, U.V. Mentzel, T. Jensen, P. Fristrup, A. Riisager, One-pot
synthesis of amides by aerobic oxidative coupling of alcohols or aldehydes
with amines using supported gold and base as catalysts, Chem. Commun. 48
(2012) 2427, http://dx.doi.org/10.1039/c2cc16768a.
[6] J. Mielby, J.O. Abildstrøm, F. Wang, T. Kasama, C. Weidenthaler, S. Kegnæs,
Oxidation of bioethanol using zeolite-encapsulated gold nanoparticles, Angew.
4. Conclusion
Chem.
anie.201406354.
- Int. Ed. 53 (2014) 12513–12516, http://dx.doi.org/10.1002/
Gold nanoparticles encapsulated in silica are highly active cata-
lysts for the production of hydrogen by vapour phase decomposi-
tion of formic acid. The Au@SiO2 core-shell catalysts were
prepared in a reverse micelle system that made it possible to con-
trol the particle size distribution and shell thickness, which have a
large impact on the catalytic activity. The smallest gold nanoparti-
cles were 2.2 0.3 nm in diameter and afforded a TOF of up to
958 hꢀ1 at a temperature of 130 °C. The remarkable particle size
effect on the TOFs and the results from in situ ATR-FTIR spec-
troscopy indicate that the reaction occurs by a dual site mechanism
involving an amine absorption site and a low-coordinated Au atom,
while results from kinetic isotope labelling experiments indicate
that H-assisted formate decomposition into CO2 and H2 is the rate
limiting step. The observed deactivation could be related to the for-
mation on inactive amide species in an off-cycle reaction between
the amine and formic acid. The study therefore suggests that the
design of a more stable amine absorption site could improve the
catalytic activity over time. In conclusion, the reverse micelle
method is simple and practical and results in a narrow size distri-
bution of small nanoparticles that are encapsulated by a thin shell
of silica that remains accessible to formic acid. By optimisation of
the amine adsorption site and the active metal composition to
improve the selectivity we therefore expect that silica encapsu-
lated metal nanoparticles may find use for activation of formic acid
for future renewable energy or chemical processes.
[7] M. Haruta, N. Yamada, T. Kobayashi, S. Iijima, Gold catalysts prepared by
coprecipitation for low-temperature oxidation of hydrogen and of carbon
monoxide, J. Catal. 115 (1989) 301–309.
[8] J. Mielby, S. Kegnæs, P. Fristrup, Gold nanoparticle-catalyzed formation of
nitrogen-containing compounds-from mechanistic understanding to synthetic
exploitation, ChemCatChem 4 (2012) 1037–1047, http://dx.doi.org/10.1002/
cctc.201200314.
[9] N. Lopez, T.V.W. Janssens, B.S. Clausen, Y. Xu, M. Mavrikakis, T. Bligaard, et al.,
On the origin of the catalytic activity of gold nanoparticles for low-
temperature CO oxidation, J. Catal. 223 (2004) 232–235, http://dx.doi.org/
10.1016/j.jcat.2004.01.001.
[10] Q. Bi, X. Du, L. He, Y. Liu, Y. Cao, H. He, et al., Efficient subnanomeric gold-
catalyzed hydrogen generation from formic acid decomposition under
ambient conditions, J. Am. Chem. Soc. 134 (2012) 8926–8933, http://dx.doi.
org/10.1002/anie.200.
[11] M. Yadav, T. Akita, N. Tsumori, Q. Xu, Strong metal – molecular support
interaction (SMMSI): amine-functionalized gold nanoparticles encapsulated in
silica nanospheres highly active for catalytic decomposition of formic acid, J.
Mater. Sci. 22 (2012) 12582–12586, http://dx.doi.org/10.1039/c2jm31309j.
[12] M. Ojeda, E. Iglesia, Formic acid dehydrogenation on Au-based catalysts at
near- ambient temperatures, Angew. Chem. Int. Ed. Engl. 121 (2009) 4894–
4897, http://dx.doi.org/10.1002/ange.200805723.
[13] D.A. Bulushev, S. Beloshapkin, J.R.H. Ross, Hydrogen from formic acid
decomposition over Pd and Au catalysts, Catal. Today 154 (2010) 7–12,
http://dx.doi.org/10.1016/j.cattod.2010.03.050.
[14] A. Gazsi, T. Bánsági, F. Solymosi, Decomposition and reforming of formic acid
on supported Au catalysts: production of CO-Free H2, J. Phys. Chem. C 115
(2011) 15459–15466.
[15] A. Gallas-Hulin, J. Mielby, S. Kegnaes, Efficient production of hydrogen from
decomposition of formic acid over zeolite incorporated gold nanoparticles,
ChemistrySelect.
slct.201600831.
1
(2016)
3942–3945,
http://dx.doi.org/10.1002/
[16] A. Herron, J. Scaranto, P. Ferrin, S. Li, M. Mavrikakis, Trends in formic acid
decomposition on model transition metal surfaces: a density functional theory
study, ACS Catal. 4 (2014) 4434–4445.
Acknowledgments
[17] J.S. Yoo, F. Abild-pedersen, J.K. Nørskov, F. Studt, Theoretical analysis of
transition-metal catalysts for formic acid decomposition, ACS Catal. 4 (2014)
1226–1233.
[18] T. Zhang, H. Zhao, S. He, K. Liu, H. Liu, Y. Yin, et al., Unconventional route to
encapsulated ultrasmall gold nanoparticles for high-temperature catalysis,
ACS Nano 8 (2014) 7297–7304, http://dx.doi.org/10.1021/nn502349k.
[19] A.J. Kunov-Kruse, J. Thomsen, J. Piltoft, C. Petersen, Patent Application
EP12174778.6, 2012.
The authors gratefully acknowledge the support of the Danish
Council for Independent Research, Grant No. 12-127580, the sup-
port of the Lundbeck Foundation (Lundbeckfonden), Grant No.
R141-2013-13244 and the support from Villum Fonden research
grant (13158).
[20] A. Rimola, D. Costa, M. Sodupe, J.-F. Lambert, P. Ugliengo, Silica surface features
and their role in the adsorption of biomolecules: computational modeling and
experiments, Chem. Rev. 113 (2013) 4216–4313.
Appendix A. Supplementary material
[21] M. Zacharska, A.L. Chuvilin, V.V. Kriventsov, S. Beloshapkin, M. Estrada, A.
Simakov, et al., Support effect for nanosized Au catalysts in hydrogen
production from formic acid decomposition, Catal. Sci. Technol. 6 (2016)
6853–6860, http://dx.doi.org/10.1039/C6CY00552G.
Supplementary data associated with this article can be found, in
the online version, at http://dx.doi.org/10.1016/j.jcat.2016.11.020.
[22] G. Ertl, H. Knozinger, F. Schüth, J. Weitkamp (Eds.), Handbook of
Heterogeneous Catalysis, second ed., Wiley-VCH, Weinheim, Germany, 2008.
[23] N. Lopez, On the origin of the catalytic activity of gold nanoparticles for low-
temperature CO oxidation, J. Catal. 223 (2004) 232–235, http://dx.doi.org/
10.1016/j.jcat.2004.01.001.
[24] T.V.W. Janssens, B.S. Clausen, B. Hvolbæk, H. Falsig, C.H. Christensen, T.
Bligaard, et al., Insights into the reactivity of supported Au nanoparticles:
combining theory and experiments, Top. Catal. 44 (2007) 15–26, http://dx.doi.
org/10.1007/s11244-007-0335-3.
[25] K.T. Queeney, N. Herbots, J.M. Shaw, V. Atluri, Y.J. Chabal, Infrared
spectroscopic analysis of an ordered Si/SiO2 interface, Appl. Phys. Lett. 84
(2004) 493–495, http://dx.doi.org/10.1063/1.1644030.
[26] S. Saravanamurugan, A.J. Kunov-kruse, R. Fehrmann, A. Riisager, Amine-
functionalized amino acid-based ionic liquids as efficient and high-capacity
References
[1] S. Moret, P.J. Dyson, G. Laurenczy, Direct synthesis of formic acid from carbon
dioxide by hydrogenation in acidic media, Nat. Commun. 5 (2014) 1–7, http://
dx.doi.org/10.1038/ncomms5017.
[2] R. Wölfel, N. Taccardi, A. Bösmann, P. Wasserscheid, Green chemistry selective
catalytic conversion of biobased carbohydrates to formic acid using molecular
oxygen, Green Chem. 13 (2011) 2759–2763, http://dx.doi.org/10.1039/
c1gc15434f.
[3] M.J. Gilkey, B. Xu, Heterogeneous catalytic transfer hydrogenation as an
effective pathway in biomass upgrading, ACS Catal. 5 (2016) 1420–1436,
http://dx.doi.org/10.1021/acscatal.5b02171.
[4] S. Kegnæs, J. Mielby, U.V. Mentzel, C.H. Christensen, A. Riisager, Formation of
imines by selective gold-catalysed aerobic oxidative coupling of alcohols and
amines under ambient conditions, Green Chem. 12 (2010) 1437, http://dx.doi.
org/10.1039/c0gc00126k.
absorbents for CO2, ChemSusChem
10.1002/cssc.201300691.
7 (2014) 897–902, http://dx.doi.org/