Sunteți pe pagina 1din 12

Journal of Catalysis 320 (2014) 77–88

Contents lists available at ScienceDirect

Journal of Catalysis
journal homepage: www.elsevier.com/locate/jcat

Intermetallic compounds of Ni and Ga as catalysts for the synthesis


of methanol
Irek Sharafutdinov a, Christian Fink Elkjær a, Hudson Wallace Pereira de Carvalho b, Diego Gardini c,
Gian Luca Chiarello b, Christian Danvad Damsgaard c, Jakob Birkedal Wagner c, Jan-Dierk Grunwaldt b,
Søren Dahl a, Ib Chorkendorff a,⇑
a
Center for Individual Nanoparticle Functionality (CINF), Department of Physics, Technical University of Denmark, Fysikvej 312, DK-2800 Lyngby, Denmark
b
Institute for Chemical Technology and Polymer Chemistry, Karlsruhe Institute of Technology, Engesserstr. 20, D-76131 Karlsruhe, Germany
c
Center for Electron Nanoscopy (CEN), Technical University of Denmark, Fysikvej 307, DK-2800 Lyngby, Denmark

a r t i c l e i n f o a b s t r a c t

Article history: In this work, we present a detailed study of the formation of supported intermetallic Ni–Ga catalysts for
Received 8 April 2014 CO2 hydrogenation to methanol. The bimetallic phase is formed during a temperature-programmed
Revised 28 September 2014 reduction of the metal nitrates. By utilizing a combination of characterization techniques, in particular
Accepted 29 September 2014
in situ and ex situ X-ray diffraction, in situ X-ray absorption spectroscopy, transmission electron micros-
Available online 27 October 2014
copy combined with electron energy loss spectroscopy and X-ray fluorescence, we have studied the for-
mation of intermetallic Ni–Ga catalysts of two compositions: NiGa and Ni5Ga3. These methods
Keywords:
demonstrate that the catalysts with the desired intermetallic phase and composition are formed upon
Methanol synthesis
Intermetallics
reduction in hydrogen and enable us to propose a mechanism of the Ni–Ga nanoparticles formation.
Characterization By studying the effect of calcination prior to catalyst reduction, we show that the reactivity depends
XRD on particle size, which suggests that the reaction is structure sensitive.
TEM Ó 2014 Elsevier Inc. All rights reserved.
EXAFS

1. Introduction diluting palladium surface by silver atoms, i.e. breaking down the
surface into smaller Pd islands, leads to partial suppression of coke
Recently, increasing attention has been drawn towards alloys formation, and hence, improved selectivity and stability [12,13].
and intermetallic compounds for catalytic purposes: in a number Recently, a novel Pd–Ga intermetallic catalyst (IMC) has been
of applications, they might represent attractive alternatives to con- shown to have superior activity and selectivity for semi-hydroge-
ventional catalysts in terms of cost, activity, stability and selectiv- nation of acetylene compared to commercially used Pd–Ag alloy,
ity. The unique properties of such systems are attributed to their as well as being more resistant to coking [2]. The reason for
electronic and geometric structure, which might be very different improvement is believed to stem from the chemical nature of the
compared to their monometallic counterparts [1–7]. A good exam- Pd–Ga intermetallic compounds. In PdGa, for example, no direct
ple is direct synthesis of H2O2 from H2 and O2 over carbon-sup- Pd–Pd neighbours are present in the structure and segregation is
ported Au–Pd catalyst. The very high selectivity (greater than reduced due to strong covalent bonding between the two metals
95%) towards hydrogen peroxide is attributed to the formation of [2,14]. In other words, both the geometric and the electronic struc-
Pd-rich bimetallic nanoparticles, whereas the H2O2 selectivity of ture of the Pd–Ga IMC are much more well-defined than that of the
pure Pd or Au is significantly lower [8]. This has recently been Pd–Ag alloy.
shown to also work for electrocatalytic H2O2 synthesis where high Attempts have been made to utilize intermetallic compounds
selectivity is obtained by isolating Pt [9] and Pd [10] atoms by for methanol steam reforming (MSR) by using PdZn-based
surrounding them with Hg. [15,16] and PdGa-based [17] catalysts, since Cu/ZnO/Al2O3 is
Another example is selective hydrogenation of trace amounts of known to produce more than 1000 ppm CO as a by-product in this
acetylene to ethylene. Alumina-supported Pd–Ag alloy catalysts reaction [16].
are used in industry for this reaction [11]. It is believed that In general, utilization of intermetallic compounds offers a
number of benefits for catalytic applications: they are structurally
stable due to the covalent nature of the metal–metal interactions
⇑ Corresponding author.
[18], and the electronic structure of the active metal can be
E-mail address: ibchork@fysik.dtu.dk (I. Chorkendorff).

http://dx.doi.org/10.1016/j.jcat.2014.09.025
0021-9517/Ó 2014 Elsevier Inc. All rights reserved.
78 I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88

modified depending on the nature of neighbouring atoms. They are CO2 þ 3H2 ! CH3 OH þ H2 O
an excellent example of the concept of site isolation, which is
at atmospheric pressure with a stoichiometric mixture of CO2 (25%)
believed to be required in a number of heterogeneous catalysis
and H2 (75%) used as feedstock.
applications [19]. Despite the existence of a huge variety of inter-
Carbon monoxide, which is formed by the reverse water–gas
metallic phases, a major challenge is to prepare well-defined sup-
shift (rWGS) reaction
ported nanoparticles in a sustainable and reproducible way [3].
This has been reported for a few systems where intermetallic CO2 þ H2 ! CO þ H2 O
nanoparticles with controlled size and morphology were synthe-
sized following conventional chemical procedures [3,6], but there is the main by-product during methanol synthesis from a pure CO2/
seems to exist no generally applicable recipe to overcome this H2 feedstock.
problem. The outlet stream was analysed with an Agilent 7890A gas chro-
Recently, a computational screening study has shown that Ni– matograph (GC), equipped with thermal conductivity detector
Ga bimetallic catalysts could potentially catalyse the synthesis of (TCD) and flame ionization detector (FID) for the analysis of inor-
methanol from CO2 and H2. This has been experimentally confirmed ganic and organic compounds, respectively. The configuration of
by hydrogenation of CO2 to methanol over supported Ni–Ga cata- the GC was designed based on feed and expected reaction mixture
lysts. At pressures close to atmospheric, specific activity and selec- compositions, namely H2, CO2, CO, Ar, CH3OH, (CH3)2O, CH4 and
tivity towards methanol was found to be very close to that of higher hydrocarbons. Three columns in series are used in the
conventionally used Cu/ZnO/Al2O3 catalyst [20]. Formation of finely TCD line: 80/100 HATESEP Q packed column is responsible for mois-
dispersed Ni–Ga nanoparticles (average diameter ca. 5 nm) by ture removal, 19095P-QO4 capillary column for separation of CO2,
means of reduction of corresponding metal nitrates at 700 °C sug- and 19095P-MS6 molecular sieve column is dedicated to separation
gests that these catalysts should be very resistant to sintering under of H2, argon and CO. Organic products are separated with the aid of
methanol synthesis conditions. Conversely, Cu/ZnO/Al2O3 is well 19091J-413 capillary column in the FID line. A gas sample was
known to suffer from coarsening of copper nanoparticles [21]. In withdrawn every 15 min, and 5–7 measurements were taken at
this work, we demonstrate how the catalytic properties of Ni–Ga each temperature and for every gas composition to ensure stable
catalysts are closely linked to preparation conditions and Ni–Ga reading.
ratio, showing that the reactivity depends on particle size, which Before opening the reactor and transferring samples to further
indicates that the reaction is structure sensitive. Specifically, meth- characterization (X-ray diffraction (XRD), transmission electron
anol production and selectivity were investigated for catalysts in microscopy (TEM)), the catalysts were passivated in 1% O2/Ar flow
the nickel-rich side of the Ni–Ga phase diagram, confirming our ear- at 40 °C for 30 min to prevent severe re-oxidation of the catalysts
lier proposal of Ni5Ga3 as the most active composition. when exposed to air.
Finally, we demonstrate a model for the road to the formation of
well-dispersed intermetallic particles by a detailed study of the 2.3. Catalyst characterization
process using a variety of in situ and ex situ characterization tools.
2.3.1. Electron microscopy
For the purpose of statistical analysis of size distribution of the
2. Experimental Ni–Ga nanoparticles in the catalyst, transmission electron micros-
copy was performed using a FEI Tecnai T20 G2 microscope operat-
2.1. Catalyst preparation ing at 200 kV. Scanning TEM (STEM)–high-angle annular dark field
(STEM-HAADF) images of Ni–Ga nanoparticles synthesized on SiO2
A series of Ni–Ga catalysts with different compositions ranging nanospheres were acquired using a FEI Titan FEG TEM operated at
from Ni50Ga50 to Ni75Ga25 were prepared using incipient wetness 120 kV.
impregnation. A mixed aqueous solution of nickel and gallium Quantitative EELS analysis of individual Ni–Ga nanoparticles
nitrates (Sigma Aldrich) was impregnated on high surface area sil- was carried out in the FEI Tecnai T20 G2 TEM. Ni–Ga nanoparticles
ica (241 m2/g, Saint Gobain Norpro), dried and aged in air for 24 h on the surface of the 200 nm nanospheres allow an easy access to
at 100 °C and reduced in a flow of pure hydrogen for 2 h at 700 °C. the individual nanoparticles by the electron beam without probing
Alternatively, the catalysts were calcined in stagnant air for 4 h at the support, thus optimizing signal-to-background ratio for the Ni
400 °C or 700 °C prior to reduction. Reduction was always per- and Ga signals in the electron energy-loss spectra.
formed in the same reactor used for catalytic testing to prevent Single-particle spectra were acquired in STEM mode. Data treat-
contact of reduced material with air. The total loading of nickel ment and analysis were carried out using Gatan Digital Micrograph
and gallium in the reduced catalyst was 17 wt%. software. The background under the Ni ionization edges was fitted
For the purpose of quantitative EELS (electron energy loss spec- with a power-law model and subtracted. Plural scattering contri-
troscopy) analysis, we have prepared Ni–Ga catalysts supported on butions were removed using the Fourier-ratio method and quanti-
monodispersed SiO2 nanospheres (200 nm), according to the recipe fication was carried out using Hartree–Slater cross sections for Ni
described above. The method of SiO2 nanospheres preparation was and Ga L-edges. The choice of background and signal window off-
adopted from [22]. sets follows a preliminary study carried out on nickel (NiFe2O4)
and gallium (GaAs) containing samples, to ensure a reliable quan-
2.2. Catalytic measurements tification of various Ni–Ga compositions.

Catalytic tests were performed under atmospheric pressure in a 2.3.2. X-ray diffraction
fixed-bed reactor made of quartz glass (internal diameter 6 mm). A X-ray diffraction experiments were conducted using a PANalyt-
sieve fraction of 0.212–0.354 mm was used. Total flow rate of gas ical X’pert Pro diffractometer using Ni filtered Cu Ka. For in situ
was kept at 100 nml/min, while the volume of the catalytic bed XRD measurements, an Anton Paar XRK-900 cell was used and
was 1.13 ml. Two thermocouples, before and after the catalyst the catalyst precursor was reduced at atmospheric pressure in
bed, were used to ensure the homogeneity of the temperature 90% H2/He mixture (total flow: 40 nml/min) in the heated sample
(DT 6 1.5 °C) in the reaction zone. The catalysts were tested in holder. The catalyst precursor was heated up to 700 °C in steps
the methanol synthesis reaction of 50 °C or 100 °C (5 °C/min), and the temperature was kept
I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88 79

constant for 2 h at each step. Short 15-min diffractograms (not pre- 12 100,0
sented here) were acquired before and after long 90-min diffracto-
99,5
grams, to confirm that no phase transformations occurred at 10

6
CH3OH production, TOF
(mol[CH3OH]*m *s )*10

CH3OH selectivity / %
isothermal conditions. The particle size was determined from 99,0

-1
Scherrer broadening of the main peak ((0 0 2) for NiO, (1 1 1) for Ni3- 8 98,5
Ga, (2 1 1) for Ni5Ga3 and (0 1 1) for NiGa). The instrumental broad-

-2
ening of the apparatus is 0.08°, which is negligible compared to the 6
205°C 80
broadening due to the small nanocrystals. 165°C

Rietveld refinement of the diffractograms was performed using 4 60


205°C
X’Pert HighScore Plus software. The pseudo-Voigt peak function 40
165°C
was employed, assuming flat plate sample geometry. Refinement 2
phase 20
was done by fitting various combinations of Ni–Ga phases, until
a satisfactory agreement between the experimental and fitted dif- 0
fractograms was achieved. 50 55 60 65 70 75
Nominal Ni content / wt%
2.3.3. Elemental analysis Fig. 1. Methanol production (solid lines) and methanol selectivity (dashed lines) as
The molar Ni/Ga ratio in the catalyst after the reduction/reac- a function of nominal Ni content in the NiaGab/SiO2 catalysts. Total pressure: 1 bar;
tion/passivation cycle was investigated by X-ray fluorescence anal- feed composition: 25% CO2 and 75% H2. Note a selectivity plateau between 62.5%
ysis (XRF). The XRF apparatus (MiniPal4, PANalytical) was and 67.5% nickel, reflecting the boundary concentrations for the d-phase.
calibrated using custom-made nickel and gallium solutions
(100 g/L in 5% nitric acid aqueous solution, deviation ±0.4%, SPEX
by-product was methane. The turnover frequency (TOF) values
CertiPrep Ltd.). The standard solutions were mixed in proportions
are based on the dispersion estimated from ex situ XRD data col-
corresponding to Ni/Ga ratio ranging from 40:60 to 80:20 and
lected after catalytic tests. The maximum in methanol production
impregnated on high surface area SiO2. The resulting calibration
is observed at 67.5% Ni, which falls into the d-phase region and is
line was used to determine the actual Ni/Ga ratio in the corre-
very close to the Ni5Ga3 composition. The selectivity is 100% at
sponding catalysts.
165 °C except for the catalyst containing 75% Ni, whereas at
205 °C, the selectivity decreases with increased Ni content. Inter-
2.3.4. X-ray absorption spectroscopy estingly, there is a selectivity plateau (approximately 98.5%)
The bimetallic particle formation was followed in situ by X-ray between 62.5% and 67.5% nickel, which is very close to the bound-
absorption spectroscopy (XAS). The XAS measurements were taken ary concentrations for the d-phase [27].
in transmission mode at the X1 beamline, at the DORIS III synchro- It is well known that metallic nickel is a good methanation cat-
tron facility, radiation source at HASYLAB, Germany, and at the alyst [28] with no selectivity to methanol. Our results show that
Swiss-Norwegian Beamlines at the ESRF, France. In both beamlines, with increasing Ni content in the bulk, the catalytic behaviour
a Si (1 1 1) double-crystal monochromator was used. The XAS spec- gradually approaches that of pure Ni with decreased methanol
tra were recorded at Ni and Ga K-edges. The catalyst precursor was and increased methane production. This likely suggests an
loaded in a quartz capillary (di = 1.0 mm, wall thickness 0.02 mm) increased number of Ni–Ni sites at the surface when moving
and subjected to temperature-programmed reduction in 90% H2/ towards the more Ni-rich part of the phase diagram. It was sug-
He mixture with a ramp rate of 5 °C/min. The temperature was gested that the active site for the production of methanol on this
measured by a thermocouple placed directly below the capillary class of catalysts is the Ni–Ga surface site [20], and the methanol
[23]. production trend thus points to a correlation between bulk compo-
The analysis of the X-ray absorption near edge structure sition and the number of active sites on the surface, which is max-
(XANES) and extended X-ray absorption fine structure (EXAFS) imized for Ni/Ga ratio of approximately 5:3.
data was carried out using Athena and Artemis software of IFEFFIT All the catalysts were analysed using XRF after the catalytic test.
package [24]. The spectra were energy calibrated from a reference The results of quantifications are summarized in Table 1: the com-
metal foil and then normalized. position of all the catalysts under investigation corresponds well to
The relative proportions between the starting and formed Ni the targeted Ni/Ga ratio. The slightly higher nickel content with
and Ga species were quantified by linear combinations (LC). The respect to the nominal could be due to inaccuracy during the deter-
LC fitting was performed in the spectral range of 20 and 90 eV mination of crystal water content of the metal nitrates. The actual
relative to the absorption edge. The EXAFS spectra were extracted composition of the synthesized nanoparticles deviated maximum
using the ‘‘autobkg’’ algorithm available at Athena. Ab initio FEFF6 1% from the expected value. These results also show that metallic
calculated phase shifts and amplitudes [25] were fitted to the nickel is not lost during the reaction due to the formation of nickel
experimental k3  v(k) EXAFS spectra in R-space (Fourier transfor- carbonyl.
mation, range of 3–12 Å1 multiplied by a Kaiser-Bessel window)
with amplitude and phase functions calculated with the FEFF6
3.2. Influence of preparation conditions
code [26].
In an attempt to optimize the performance of the Ni5Ga3/SiO2
3. Results and discussion catalyst and establish a correlation between particle size and reac-
tivity, we further investigated the influence of pre-calcination step
3.1. Effect of nickel content on nanoparticles dispersion, resulting phase and catalytic activity.
In our previous work, we have shown that different crystallo-
Fig. 1 shows methanol production and selectivity as a function graphic phases of Ni–Ga possess quite different catalytic properties
of nickel content for two different temperatures. Here, we only [20]. In order to better understand how the structure and catalytic
consider the selectivity towards organic compounds (CH3OH, performance are related, we investigate whether the methanol
(CH3)2O and CH4), i.e. not considering CO produced by the reverse synthesis on Ni5Ga3/SiO2 depends on the particle size. For this,
water–gas shift (rWGS) reaction. The only significant organic we tested and characterized the catalyst prepared in three
80 I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88

Table 1
X-ray fluorescence quantification of the composition of the supported Ni–Ga catalysts.

Nominal composition Ni50Ga50 Ni55Ga45 Ni60Ga40 Ni62.5Ga32.5 Ni65Ga35 Ni67.5Ga32.5 Ni70Ga30 Ni75Ga25
Nominal Ni content, mol% 50 55 60 62.5 65 67.5 70 75
Observed Ni content, mol % 50.31 55.46 60.65 63.14 66.08 68.12 71.08 76.04

different ways: (a) a catalyst calcined in stagnant air during 4 h at Table 2


400 °C prior to reduction, (b) a catalyst calcined in stagnant air dur- Average particle diameter as a function of calcination temperature determined by
statistical analysis of TEM images and ex situ XRD of used Ni5Ga3/SiO2 catalyst.
ing 4 h at 700 °C prior to reduction, and (c) a catalyst directly
reduced from nitrates. Calcination temperature (°C) Particle diameter (nm)
Fig. 2a and b summarize the results of activity tests and post- TEM XRD
reaction XRD profiles of the three differently prepared catalysts. N/A 4.7 ± 1.2 5.4
XRD data were acquired after the catalysts were passivated in 1% 400 5.2 ± 2.0 5.6
O2/Ar flow at room temperature; hence, partial oxidation of the 700 8.8 ± 2.5 8.5
intermetallic phase is expected. In this line, slight oxidation is evi-
dent from a very broad peak at 2h = 63° and 2h = 37°, which is a
footprint of small nickel oxide crystallites formation. far from equilibrium. The specific activity is defined as activity per
Very distinct Ni5Ga3 profiles are observed for the calcined sam- geometric area, derived from the value of a surface-weighted aver-
ples, whereas for the directly reduced catalyst, some of the reflec- age diameter, as found by statistical analysis of TEM images
tions (at 2h = 55° and 2h = 71°) are not seen. This could be due to according to the following formula (the particles were treated as
combination of at least two effects: very small particle size (below spheres) [30]:
5 nm) and a relatively weak overall signal from certain reflection
P 3
planes in the Ni5Ga3 crystal, as expected from reference diffracto- ni di
grams [29]. For such small crystals, a lack of long-range order in ds ¼ P 2
ni di
certain dimensions of the crystal could be expected, which might
lead to very weak (or no) signal in the diffraction pattern. The for- where ni is the number of particles with the diameter di.
mation of the Ni5Ga3 phase for these catalysts is additionally con- Fig. 4 shows that smaller particles are more active per unit area,
firmed by high-resolution TEM analysis, as shown in the in accordance with the trend in activity per unit mass of Ni5Ga3
Supplementary information (Fig. S3). intermetallic nanoparticles. In general, the fraction of low-coordi-
Pre-treatment conditions have a remarkable influence on the nated sites (corners, kinks and step sites) on the surface is
final state of the catalytic system (Fig. 2a): the catalyst calcined enhanced as the dimensions of the nanoparticles decrease, while
at 700 °C exhibits the lowest activity and the largest Ni5Ga3 crys- the fraction of flat terraces is higher for larger particles. This
tallites, as observed by the XRD analysis. Catalyst calcined at implies that the methanol synthesis reaction on Ni5Ga3 nanoparti-
400 °C has an intermediate activity and particle size, while non- cles could be structure sensitive [31,32]. Previously, we suggested
calcined Ni5Ga3 possesses the highest CH3OH activity and consists that the methanol production occurs at the Ni–Ga step sites, mean-
of the smallest nanoparticles. Average particle sizes were calcu- ing that the activity should increase when the particle size is scaled
lated by Scherrer broadening analysis of the main reflection down. Note that although the size distribution of the non-calcined
(2h = 43°), and the data are summarized in Table 2. and the 400 °C calcined catalysts is similar (Fig. 3), the surface-
Fig. 3 summarizes the size distribution results obtained by TEM weighted average diameter values are remarkably different (Fig. 4).
investigation of the catalysts and supports the results of XRD anal- The reason for detrimental effect of calcination on particle size
ysis discussed earlier: pre-treatment by calcination leads to bigger and, consequently, catalytic activity is likely due to high mobility
Ni5Ga3 crystallites with broader size distribution. In order to corre- of Ni during the calcination step. Previously, it was shown that cal-
late the particle size with the resulting methanol production, we cination of nickel in air leads to the formation of larger NiO crystal-
have calculated the specific activity of the catalyst. Activities were lites [33]. Assuming that NiO crystallites serve as nucleation
compared at 165 °C and 180 °C where the conversion is sufficiently centres for the formation of Ni–Ga intermetallic nanoparticles,

(a) (b)
CH3OH production / g[CH3OH]*(gcat)-1*h-1

[211]
[002]

0,24
non-calcined
Ni5Ga3
0,20
Intensity / a.u.
3
5G
a

[111]
Ni

o C
0,16 00
d

,4
ne

air-calcined
[022]

a3
lc i

G Ni5Ga3, 400oC
ca

Ni 5
n-

0,12 d
ine
no

C o

alc 700
a 3,
[002]

-c
0,08 air i
dN 5
G
air-calcined
ine
[040]

[422]
[402]

c Ni5Ga3, 700oC
cal
0,04 air-
[223]

[442]
[261]

165 180 195 210 225 240 40 50 60 70 80 90


Temperature / °C 2θ / °

Fig. 2. Activity towards CH3OH (a) and post-run in situ XRD profiles (b) of Ni5Ga3/SiO2 catalysts. Reference diffractograms: d-Ni5Ga3 in red [29], NiO in blue [51].
I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88 81

(a) (b)

10 nm 10 nm

(c) (d)

10 nm

Fig. 3. (a–c) Post-run TEM images and (d) size distributions of non-calcined, 400 °C – calcined and 700 °C – calcined Ni5Ga3 catalysts, respectively.

0,009 3.3. In situ XRD


Specific activity / [mol CH3OH*m *h ]
-1

In order to follow the dynamics of the formation of Ni–Ga inter-


-2

0,008
o
metallic nanoparticles starting from a nitrate precursor, we have
T = 180 C
conducted a series of in situ XRD measurements during tempera-
0,007 ture-programmed reduction (TPR) of impregnated Ni–Ga catalysts
prepared with Ni–Ga ratios corresponding to NiGa/SiO2 and Ni5-
0,006 Ga3/SiO2. The evolution of phases during TPR is presented in
Figs. 5a (NiGa) and 6a (Ni5Ga3), respectively. To assess the crystal
0,005
structures that develop during TPR, the XRD data were compared
o
to reference structures [35]. Crystal data for these are collected
T = 165 C
in Table 3. The Ni–Ga phase diagram and the relevant diffracto-
0,004 grams are presented in the Figs. S1 and S2, respectively.
For the Ni–Ga catalyst impregnated in a 1:1 ratio, alloying even-
tually leads to the formation of the b-NiGa phase. There are several
4 5 6 7 8 9 10 11 distinct temperature intervals to point out during the TPR as
Mean diameter (surface weighted) / nm shown by the in situ XRD data in Fig. 5a: at T > 200 °C, formation
of NiO crystallites starts. Between 300 °C and 350 °C, the intensi-
Fig. 4. Catalytic activity towards CO2 hydrogenation to methanol: influence of pre-
ties of the reflections from NiO are further enhanced, which is
treatment conditions. Feed composition: 25% CO2 and 75% H2; pressure: 1 bar.
indicative of sintering during this step. Increasing the temperature
further leads to alloying of Ni and Ga, i.e. formation of a0 -Ni3Ga, d-
avoiding calcination decreases particle size and increases activity Ni5Ga3 and finally b-NiGa phase, which corresponds to the Ni/Ga
as shown above. The size distribution obtained by direct reduction ratio in the initial impregnation mixture. Reflections corresponding
of the precursor is significantly more narrow compared to the Ni/ to Ga2O3 were not observed during in situ XRD, indicating that the
SiO2 catalyst prepared under similar conditions [34], meaning that oxide is amorphous in this case. However, XAS analysis revealed
gallium in the Ni5Ga3 reduces sintering. This confirms the hypoth- the presence of this intermediate (see Section 3.5).
esis that intermetallic nanoparticles are more stable towards Rietveld refinement was applied in order to quantify the rela-
agglomeration [2]. tive amounts of phases during TPR (Fig. 5b). Formation of the final
82 I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88

(a) (a)

(011)

(211)
(111)
(112)
(002)
700

650
700
600
(221)

(002)

650
Counts / a.u.

550

Temperature / oC

Temperature / oC
Counts / a.u.
(223)
(111)

500
600
(022)
(002)

450

(002)
550
400

(223)
(022)
(002)
(111)

350 500

(022)
300 400

(002)

(113)

(222)
(022)
250 300

(002)
(111)
200 200

40 50 60 70 80 40 50 60 70 80 90 100
2θ / o 2θ / o

(b) 100 (b) 100 δ−Ni5Ga3 formation


NiO β−NiGa analysed by
80 Rietveld Refinement
80
δ−Ni5Ga3 NiO
Phase / wt %

δ−Ni5Ga3
Phase / wt %

60
60 β−NiGa formation
analysed by
Retvield Refinement 40
40
α`−Ni3Ga
α−Ni3Ga
20 20

0 0
300 400 500 600 700 300 400 500 600 700
0 0
Temperature / C Temperature / C

Fig. 5. (a) Evolution of crystallographic phases during temperature-programmed Fig. 6. (a) Evolution of crystallographic phases during temperature-programmed
reduction (TPR) to form NiGa/SiO2 catalyst (Ni:Ga ratio 1:1 in the catalyst reduction (TPR) to form Ni5Ga3/SiO2 catalyst (Ni:Ga ratio 5:3 in the catalyst
precursor). Reflections are marked as follows: + (b-NiGa), ⁄ (d-Ni5Ga3), ¤ (NiO) precursor). Reflections are marked as follows: ⁄ (d-Ni5Ga3), ¤ (NiO) and o (a0 -Ni3Ga).
and o (a0 -Ni3Ga). Gas composition: 90% H2 in Helium. Total flow rate: 40 ml/min, Gas composition: 90% H2 in Helium. Total flow rate: 40 ml/min, and (b)
and (b) corresponding Rietveld refinement analysis of crystallographic phases corresponding Rietveld refinement analysis of crystallographic phases during TPR.
during TPR.

appears that, while the transformation of NiO and Ga2O3 into a


b-NiGa phase goes through more nickel-rich intermediate phases nickel-rich intermetallic state is a relatively easy process, the for-
(Ni3Ga and Ni5Ga3), with coexistence of two or more phases at mation of the d-Ni5Ga3 phase is much slower. The final diffraction
intermediate temperatures. pattern recorded after reduction at 700 °C still contains around 15%
For the catalyst impregnated with a mixture of nitrates corre- Ni3Ga in addition to the Ni5Ga3/SiO2 phase. Further details on the
sponding to Ni5Ga3/SiO2, the Rietveld refinement was more diffi- Rietveld refinement, including graphical and numerical data, are
cult due to a combination of several phenomena: more complex presented in Figs. S4 and S5, and Tables ST1 and ST2.
structure of the Ni5Ga3 phase, coexistence of phases with close This is in contrast to the identical catalyst prepared at the same
lying reflections and the absence of reflections from certain crystal- temperature and pressure in a fixed-bed reactor, as discussed in
lographic planes. In general, the onset of formation of intermetallic Section 3.2. The latter shows a much more pronounced d-Ni5Ga3
phases in the case of d-Ni5Ga3 synthesis is observed at lower tem- pattern (see ex situ XRD data, Fig. 2b). The reason could be the dif-
perature than for NiGa/SiO2 (Fig. 6b): NiO is formed at 250 °C, and ferent nature of the flow through the catalyst bed in the two cases:
already at 300 °C, further changes are seen. At this point, the cata- reduction in a fixed-bed reactor occurs under plug-flow conditions
lyst is a mixture of Ni3Ga and NiO, with nickel oxide as the domi- (height to diameter (h/d) ratio of 7–8). Flow conditions during
nant phase (70%). At 400 °C, a mixture of intermetallic a0 -Ni3Ga in situ XRD experiments are remarkably different: the catalyst pre-
and d-Ni5Ga3 phases is observed. With further temperature cursor is placed on a sample plate, where the h/d value is close to
increase, the Ni5Ga3 phase is formed to an extent of ca. 85%. It 0.3, which could result in preferential flow path formation and
I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88 83

Table 3
Reference data for some of the relevant Ni-containing crystal structures relevant in the present study.

Crystal structures International Crystal Structure Database (ICSD) reference


Phase Ni (at.%) Space group ICSD code Chemical structure Crystal structure a (Å) b (Å) c (Å)
b [49] 42–69.4 
Pm3m 103853 NiGa Cubic (bcc) 2.886 2.886 2.886
d [29] 62–67.5 Cmmm 103861 Ni5Ga3 Orthorhombic 7.53 6.72 3.77
a0 [49] 70–77 
Pm3m 103856 Ni3Ga Cubic (fcc) 3.585 3.585 3.585
Ni [50] 77–100 
Fm3m 52265 Ni Cubic (fcc) 3.524 3.524 3.524
NiO [51] – 
Fm3m 28910 NiO Cubic (fcc) 4.176 4.176 4.176

hence longer time to reduce the entire volume of the sample. In the the surface composition of a number of IMCs has been shown to
case of b-NiGa, this does not hinder the formation of a pure inter- be dynamic under reaction conditions. In some cases, this has a
metallic compound, whereas synthesis of the structurally more detrimental effect on catalytic properties. For example, under
complex d-Ni5Ga3 likely requires longer time under in situ XRD methanol steam reforming conditions, partial decomposition of
conditions. However, despite the differences mentioned above, the surface of Pd2Ga nanoparticles to metallic palladium has been
we consider the in situ XRD a powerful tool to accurately show reported, resulting in unwanted methanol decomposition reaction
the catalyst development from nitrate precursors to intermetallic [39]. A similar picture is observed when Pd2Ga is used as a catalyst
compound. for liquid phase phenylacetylene hydrogenation: the surface is oxi-
In the case of b-NiGa, the composition of the final IMC corre- dized to form G2O3 in the presence of even remote amounts of H2O
sponds to the Ni/Ga ratio in the mixture of metal nitrates. For d- and O2 [40,41]. However, the structural changes in the surface
Ni5Ga3, some 15% a0 -Ni3Ga is present together with the targeted might also play a positive role. This has been demonstrated for
phase. This means that given high enough temperature, a desired PdZn/ZnO system, where PdZn is believed to be partially oxidized
phase is formed in a relatively short timescale compared to the to form ZnO, which increases the selectivity to CO2 under MSR con-
preparation of bulk Ni–Ga alloys, which requires much longer time ditions [42]. Likewise, the surface of the Ni5Ga3 nanoparticles could
due to slow diffusion rate [36]. However, the presence of the a0 - be altered as soon as the gas environment is switched from reduc-
Ni3Ga implies that some gallium is present in an unalloyed state ing (pure H2) to reaction (CO2/H2) mixture. This, coupled with the
(approximately 4% of the total gallium loading). formation of water and methanol, leads to significant increase in
It is noteworthy to mention that metallic nickel has not been the oxidation potential of the reaction mixture, which might be
observed in either case. This could either mean that reduction of enough to ignite the surface decomposition of the Ni5Ga3. As dis-
the NiO crystallites is immediately followed by incorporation of cussed above, the surface decomposition phenomenon could have
gallium into the crystal structure of nickel, or the dynamical both negative and positive impact on the catalysts performance.
changes during TPR are not fully resolved by in situ XRD. The for- Disintegration of the surface of the Ni5Ga3 phase to form a Ni3-
mation of the Ni3Ga phase during reduction is also confirmed by Ga–Ga2O3 or Ni–Ga2O3 mixture might be responsible for the loss
comparing the lattice parameters taking into account the thermal of active sites. This, along with carbon formation, could explain
expansion during TPR, Figs. S6 and S7. the slow activity drop within 20 h, observed in the previous work
Formation of Ni3Ga as a first intermediate compound at the [20]. On the other hand, a synergetic effect between Ni and
lowest temperature can be explained taking into account the fol- Ga2O3 formed under reaction conditions could be the key for
lowing considerations: both Ni and a0 -Ni3Ga have the same crystal understanding the high catalytic activity. It is well known that
structure, namely face-centred cubic (fcc). Hence, incorporation of Pd/Ga2O3 catalyst is active and selective for CO2 hydrogenation
gallium metal to form Ni3Ga occurs without changing the crystal to methanol [43–45]. In a similar fashion, the interplay between
structure of nickel and should be relatively easy process. According Ni and Ga2O3 could be responsible for methanol formation on the
to both theoretical [37] and experimental [38] studies, the exother- Ni–Ga catalysts. The viability of these hypotheses, however, needs
micity of the formation of nickel-rich Ni–Ga alloys from pure met- to be clarified by further experimental work.
als increases towards equimolar Ni/Ga ratio: approximately 30 to
50 kJ/mol with increasing heat effect towards equimolar Ni/Ga 3.4. STEM–EELS investigation of the model samples
ratio. However, there is apparently a significantly higher kinetic
barrier for the formation of the d-Ni5Ga3 and b-NiGa phases. This Quantitative EELS analysis was performed in order to assess the
could be due to re-arrangement of the crystal structure to ortho- elemental composition on a single-particle level to compare with
rhombic (d-Ni5Ga3) and body-centred cubic (b-NiGa), which the average composition given by XRD combined with Rietveld
requires rather high activation energy. refinement. For comparison, NiGa/SiO2, Ni5Ga3/SiO2 and Ni3Ga/
The presence of the Ni3Ga phase in the Ni5Ga3/SiO2 could be SiO2 were prepared and analysed in a similar way. In all cases,
responsible for the undesired side reactions, such as formation of the nanoparticles were synthesized on 200 nm SiO2 nanospheres.
CO and CH4, although such a correlation cannot be made unambig- The results of the quantification are presented in Fig. 7.
uously at this stage. Another source of the side reactions could be The single-particle EELS data do not allow for a statistical treat-
adjacent Ni–Ni sites, as shown by theoretical calculations [20]. ment to obtain the relative abundance of different phases, since the
Nevertheless, the state of the surface of the Ni–Ga catalysts under number of analyzed particles was too small. However, there are
reaction conditions requires further investigations. Suppression of clearly particles that fall substantially outside the targeted compo-
CO formation is the main challenge with the Ni5Ga3/SiO2 catalysts, sition. Even though the measured Ni to Ga ratio in the NiGa/SiO2
and further work is needed to understand the role of individual catalyst corresponds to the b-phase (Fig. 5a), a single outlier is
species and surface sites of this complex catalytic system. In-depth observed. For the Ni5Ga3/SiO2, the discrepancy is even more pro-
understanding of the surface composition might require that both nounced: in this sample, nanoparticles that fall into the a0 -region
supported and unsupported, well-defined Ni–Ga intermetallic are found together with those that fall into the intended d-phase
compounds (NiGa, Ni5Ga3, Ni3Ga) are studied individually, by uti- region. These results support the Rietveld analysis of the XRD data,
lization of a combination of surface sensitive techniques. Indeed, pointing towards the coexistence of the two phases after reduction.
84 I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88

80 allows for analysis of oxidation state and local environment of Ni


and Ga in the sample and can thus provide important information

phase
75 of the reduction and alloying processes. Importantly, XAS allows
Measured Ni content / at.%

for the detection of (amorphous) gallium oxide, not detectable by


70 `
XRD as discussed above. This detailed study was performed on Ni5-
Ga3/SiO2 as it possesses the highest activity among all Ni–Ga inter-

phase
65
metallics we have tested.
Linear combination fitting analysis of the XANES region of the
60
Quick XAS spectra recorded at Ni and Ga K-edges during TPR gives
55 insights into the oxidation state of the metals during reduction
(Fig. 9a and b). The signals from the mass spectrometer corre-

phase
50 sponding to water, NO and NO2 are also plotted as complementary
data.
45 The analysis of the isosbestic points observed at the Ni K-edge
XANES (near edge structure) suggests that Ni(NO3)2 is converted
1 2 3 4 5 6
to an intermediate phase prior to the reduction (Fig. S8 in the Sup-
Particle No plementary Information). Based on the XRD data (see Section 3.3),
this intermediate is identified as NiO. However, the fitting of the
Fig. 7. Measured compositions of the individual Ni–Ga nanoparticles. Targeted Ni
content is depicted in dashed lines for the NiGa (black), Ni5Ga3 (red) and Ni3Ga
Quick XAS spectra suggests the existence of other oxidized Ni
(blue) catalysts. The estimated experimental error of the quantification is 1.2% phase/phases along with NiO. It has been reported [34] that during
(NiGa/SiO2), 1.6% (Ni3Ga/SiO2) and 1.4% (Ni5Ga3/SiO2). The origin of the error is a the temperature-programmed reduction of nickel nitrate, a few
combination of uncertainty in the background determination and signal deconvo- intermediates could coexist at 250–300 °C, including Ni(NO3)2,
lution. (For interpretation of the references to colour in this figure legend, the
NiO and nickel phyllosilicates. This might explain the results of
reader is referred to the Web version of this article.)
the Quick XAS fitting. Therefore, the intermediate phase at 260 °C
is denoted as Ni2+. At higher temperatures, reduction of Ni2+ is acti-
XRD analysis of the NiGa and Ni5Ga3 synthesized on SiO2 nano- vated and the formation of metallic nickel completed at 500 °C.
spheres showed a similar diffractogram compared to the corre- The dynamics of gallium reduction is different: first, Ga(NO3)3 is
sponding catalyst prepared on high surface area SiO2. fully converted to Ga2O3 at 220 °C. This phase is stable up to 280 °C,
A representative HAADF-STEM image of the Ni5Ga3 particles where reduction of Ga2O3 starts. Complete reduction, according to
dispersed on the surface of the nanospheres is shown in Fig. 8a. XANES analysis, is achieved around 500 °C.
The average size of the nanoparticles, as well as the size distribu- The onset of the reduction of gallium oxide (280 °C) is slightly
tion, is significantly larger than for the catalysts used for activity higher than that of nickel oxide (260 °C). The rather low tempera-
tests, likely due to the low surface area of the nanospheres. An ture of gallium oxide reduction, compared to that expected for the
electron energy-loss spectrum of a single Ni5Ga3 particle is shown reduction of Ga2O3/SiO2 [46], appears to be triggered by the reduc-
in Fig. 8b, where both the Ni L-edge (855.0–895.1 eV) and the Ga L- tion of nickel oxide. Armbrüster et al. [47] investigated the forma-
edge (1225.1–1265.2 eV) are present. tion of Pd–Ga nanoparticles in solution and suggested that the
reduction of Ga takes place on the surface of already reduced pal-
3.5. In situ X-ray absorption spectroscopy ladium nanoclusters by atomically dissolved hydrogen therein.
Gallium oxide only starts to reduce after a certain amount of
In order to get a complementary understanding of the forma- metallic palladium is formed. A similar mechanism might take
tion of supported Ni–Ga nanoparticles, as well as to follow how place in the case of Ni–Ga nanoparticle formation, since metallic
the oxidation states of Ni and Ga are changing during the reduc- nickel readily chemisorbs and dissociate hydrogen at these tem-
tion, a series of in situ XAS experiments were carried out during peratures [48]. The results of in situ XRD experiments discussed
TPR of the nitrates impregnated on high surface area silica. XAS earlier also show that gallium is immediately incorporated into

(a) (b)

Deconvoluted spectrum

Background substracted
Ni L-edge
Intensity / a. u.

Ga L-edge

100 nm

800 900 1000 1100 1200 1300


Energy loss / eV

Fig. 8. (a) STEM-HAADF image of Ni5Ga3 nanoparticles on a group of 200 nm SiO2 spheres. Microscope: FEI Titan ATEM operated at 120 kV and (b) in black – deconvoluted
EELS spectrum of a single Ni5Ga3 nanoparticle on silica nanospheres, in red – the signal after background subtraction in the quantification region (E = 855–895 eV for nickel
and 1040–1265 eV for gallium). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88 85

Fig. 9. H2O (m/z 18), NO (m/z 30) and NO2 (m/z 46) evolution monitored by mass spectrometer (dashed lines) and linear combination results of the in situ XANES spectra
(solid lines) at (a) Ni K-edge and (b) Ga K-edge during temperature-programmed reduction of metal nitrates to form Ni5Ga3/SiO2.

the structure of the nickel-containing phase as soon as the NiO is new series of peaks at 2.1 Å and 2.5 Å emerge. The system is very
reduced, since no pure metallic nickel was detected during TPR. dynamic between 230 °C and 360 °C, implying that main variations
The results of the linear combination fitting of Quick XAS spec- in the local structure occur in this temperature range. At T > 360 °C,
tra can be well interpreted in the light of the Rietveld analysis data. the structure is stabilized, and only strong backscattering contribu-
Analysis of the XANES region shows that the ratio between oxi- tion at 2.1 Å is seen. At the Ga K-edge, a slightly different picture is
dized and reduced nickel at 300 °C (Fig. 9a) is approximately 70/ observed: there are four distinct temperature intervals corre-
30 wt%, while at 400 °C, nickel is primarily present in reduced sponding to the states of the system. Below 160 °C, gallium is
state. This is in a good agreement with the Rietveld refinement mainly present in the initial nitrate form (Ga–O distance repre-
data (Fig. 6b), which suggests that the amount of NiO at 300 °C is sented by a peak at 1.5 Å). As the temperature is increased, the
70 wt%, while almost no nickel oxide is present at 400 °C. The backscattering contribution at 2.8 Å vanishes (highlighted in blue),
results show that, despite the differences between sample geome- indicative of gallium oxide formation as concluded from linear
try, temperature profiles and flow conditions, the two methods combination fitting analysis. Between 250 °C and 420 °C, the local
yield comparable results. environment of gallium is changing significantly. The local envi-
Although the XANES is a powerful tool that allows following the ronment of Ni is also changing in the same temperature interval.
reduction and alloying of Ni and Ga, it is difficult to derive the com- At higher temperatures, only one strong peak at around 2.1 Å is
position or coexistence of different intermetallic species due to the observed. The small peak observed at low R-values (1.0–1.5 Å) at
similarities of the density of states and chemical environment of temperatures P500 °C is due to noise probably stemming from
these intermetallic compounds. signal attenuation due to high temperatures and the fast recording
The Fourier-transformed EXAFS spectra recorded at the Ni and mode, see Section 3.2 in the Supplementary Information for further
Ga K-edges during TPR are represented in Fig. 10a and b, respec- discussion. The contribution cannot be attributed to oxygen back-
tively. This method tracks the changes in local environment around scattering since it is not observed after the sample was cooled
the absorbing atoms. At the Ni K-edge, main changes in the local (Fig. 11) as discussed below.
environment of Ni appear when the temperature is increased It is important to note that the main backscattering contribu-
above 230 °C (highlighted in red): the intensity of the main peak tion to the final spectrum is observed at R(Å) = 2.1 Å in the R-space
at R(Å) = 1.6 Å, which is due to Ni–O contribution, vanishes and a representation of both Ni and Ga edges (not corrected for phase

(a) (b)

700oC
500oC
χ(k) function) / a.u.

360oC
FT ( k3-weighted

420oC
o
230 C
250oC
160oC

25oC 25oC
0 1 2 3 4 0 1 2 3 4
R/Å R/Å

Fig. 10. Fourier-transformed EXAFS profiles at the (a) Ni K-edge and (b) Ga K-edge during temperature-programmed reduction of metal nitrates to form Ni5Ga3/SiO2.
86 I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88

(a) (b)
FT ( k3 -weighted χ(k) function) / a.u. Ni edge Ga edge
o
Ni5 Ga3 , 20o C

FT ( k3 -weighted χ(k) function) / a.u.


Ni5 Ga3 , 20 C
10
fit Ni5 Ga3 structure 10 fit Ni5 Ga3 structure

5 5

0
0

-5

-5
-10 Ni5 Ga3 , 20o C
Ni5 Ga3 , 20o C
fit Ni5 Ga3 structure fit Ni5 Ga3 structure

0 1 2 3 4 5 0 1 2 3 4 5 6
R/Å R/Å

Fig. 11. Comparison of the experimental (o and dashed lines) and calculated (solid lines) k3 weighted Fourier-transformed EXAFS spectra at (a) Ni and (b) Ga K-edges of the
Ni5Ga3/SiO2 catalyst recorded at 20 °C after TPR up to 700 °C. Magnitude of the FT–v(R) function – in black, imaginary part of the v(R) function – in red. (For interpretation of
the references to colour in this figure legend, the reader is referred to the Web version of this article.).

shift). The observation that the Ni–Ga and Ga–Ni distances are values of mean square deviation of the interatomic distances
identical suggests that the two metals have formed a bimetallic (r2), see Table 4.
phase, further corroborated by data fitting. For this purpose, the Both in situ XRD and in situ XAS experiments provide solid evi-
formation of the targeted Ni5Ga3 phase was fitted with the refer- dence that the desired Ni–Ga phase can be achieved given that the
ence crystal structure model. In the d-Ni5Ga3, there are three differ- reduction temperature is high enough. Comparative analysis of the
ent Ni atoms and two different Ga atoms with respect to their local results obtained from both techniques gives valid grounds for pro-
environment. The structural parameters obtained by EXAFS fitting posing the formation mechanism of the intermetallic Ni–Ga phases
correspond to the average of those sites [29]. Fig. 11 shows the during reduction from nitrates. The mechanism of the d-Ni5Ga3
experimental and calculated Fourier transformation of the EXAFS phase formation is summarized in Fig. 12. We propose that at
spectra of the catalyst at room temperature at the Ni K-edge (a) the low temperatures (below 260 °C), Ni-containing nucleation
and Ga K-edge (b). The structural parameters obtained from the fit- sites are being formed through decomposition of the Ni(NO3)2. This
ting procedure are presented in Table 4. The experimental and fit- is supported by the in situ XRD data (see Section 3.3). However,
ted data are in good agreement, as suggested by the small analysis of the XAS data suggests the existence of other oxidized
disagreement factor q. Accordingly, the first (and the highest) peak Ni-containing phase/phases along with NiO, which could be nickel
on both Fourier-transformed spectra corresponds to the nearest phyllosilicates and/or the remains of Ni(NO3)2 [34]. This intermedi-
coordination shells, which has a quite complex structure. Both Ni ate is therefore represented as Ni2+. Ga(NO3)3, in contrast, is
and Ga are characterized by complex chemical environments: the decomposed to form a single compound, Ga2O3. As the tempera-
first peak on the Fourier-transformed Ni EXAFS spectrum accom- ture is increased above 260 °C, the reduction of nickel followed
modates the contribution from three close coordination shells, by the reduction of Ga2O3 is initiated. Already at 300 °C, a notable
while in the case of Ga, it is formed by two close shells. All mea- fraction of reduced species is observed, which are assigned to a
surements were taken at room temperature; however, the EXAFS combination of Ni2+ species, Ni3Ga and Ga2O3. As the temperature
refinement unravelled high mean square deviation of the inter- is increased further, the proportion of Ni3Ga increases in the
atomic distances (r2), which in this case can be explained by high expense of Ni2+ and Ga2O3. Finally, at T > 500 °C, the Ni3Ga phase
static disorder. To some extent, it can be attributed to the small is gradually transformed into the Ni5Ga3. Alloying always proceeds
particle size. In other words, the deviation of the interatomic dis- through a more Ni-rich phase until stoichiometric composition is
tances from the expected values can be explained by a significant achieved upon long enough reduction time and high enough tem-
fraction of surface atoms. perature. Here, we propose that metallic nickel is not formed dur-
The high static disorder is also revealed by the XRD: the ing TPR, based on the in situ XRD data. However, the time
refinement of the Debye–Waller factor of the Ni5Ga3 after the tem- resolution of the diffractograms might be not high enough to
perature-programmed reduction is 2.037 Å2, which corresponds to detect all the dynamic changes in the catalyst precursor structure:
the mean square displacement of the atoms of 0.0258 A2. This is in after the temperature is stabilized at a certain value, the data
a good agreement with the EXAFS fitting, which resulted in high acquisitions take 90 min. Although Quick EXAFS has a significantly

Table 4
Structural parameters obtained from EXAFS data analysis at Ni and Ga K-edges.

Edge Shell Atom N R (Å) r2 (103 Å2) q (%)


s f
Ni K 1st Ga 4.0 ± 0.4 2.47 ± 0.02 11.0 ± 1.0f 5.2
2nd Ni 4.0 ± 0.4s 2.42 ± 0.07f 12.2 ± 4.0f
3rd Ni 4.0 ± 0.4s 2.82 ± 0.05f 10.0 ± 6.0f
Ga K 1st Ni 8.0 ± 0.3f 2.48 ± 0.01f 11.2 ± 1.6c 0.8
2nd Ga 2.0 ± 0.7f 2.86 ± 0.02f 12.2 ± 4.0c

s – set (fixed) value; f – fitted parameter; c – constrained to be same as obtained by fitting at Ni edge; N – number of atoms in the shell; R (Å) – interatomic distance; r2 –
mean square deviation of the interatomic distances; q – disagreement between experimental and fitted data.
I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88 87

There is, however, still very little known about the state of the
Ni–Ga catalysts surface under working conditions. The assumption
on the nature of the active sites is based on the density functional
theory calculations and needs to be further validated. For this,
experimental data (in situ XPS and in situ FTIR) collected under
reaction gas atmosphere are required.
The utilization of the complementary characterization tools
enables establishing the road to fine tuning the size and crystallo-
graphic phase of Ni–Ga nanoparticles and consequently to opti-
mize the catalyst in terms of methanol production.

Acknowledgments

Fig. 12. Proposed mechanism of the formation of Ni5Ga3 phase during tempera- This work was supported by the Danish National Research
ture-programmed reduction from nitrates. Foundation’s Centre for Individual Nanoparticle Functionality
(DNRF54) and by the Catalysis for Sustainable Energy initiative,
which is funded by the Danish Ministry of Science, Technology
better time resolution, it cannot be used for unambiguous discrim-
and Innovation. XAS experiments were performed on the X1 beam-
ination of Ni3Ga and Ni. Time-resolved in situ XRD experiments are
line, at the DORIS III Synchrotron facility, radiation source at HASY-
required for more detailed studies of the phase transformations
LAB, Germany, and at the Swiss Norwegian Beamlines at the ESRF,
during TPR.
France. We are thankful to the local staff, in particular Dr. H. Eme-
rich of SNBL of ESRF, for support during the in situ XAS and XRD
4. Conclusions experiments. Helge Rasmussen is acknowledged for assistance
using the XRD equipment. The A.P. Møller and Chastine Mc-Kinney
In this work, we have demonstrated that by employing a variety Møller Foundation is gratefully acknowledged for the contribution
of complementary in situ and ex situ characterization methods the towards the establishment of the Center for Electron Nanoscopy in
mechanism of formation of supported Ni–Ga nanoparticles can be the Technical University of Denmark.
studied. A combined analysis of in situ XRD, Rietveld refinement
and in situ XAS data has been performed in order to reveal the evo-
Appendix A. Supplementary material
lution of crystallographic structure of Ni–Ga nanoparticles as a
function of temperature. It was demonstrated that supported
Supplementary data associated with this article can be found, in
nanocrystalline intermetallic Ni–Ga particles of the desired phase
the online version, at http://dx.doi.org/10.1016/j.jcat.2014.09.025.
and composition can be produced by direct reduction of the corre-
sponding nitrates in a flow of hydrogen, which are active and selec-
References
tive towards hydrogenation of CO2 to methanol. Here, the
dynamics of reduction was thoroughly investigated and a mecha- [1] F. Besenbacher, I. Chorkendorff, B.S. Clausen, B. Hammer, A.M. Molenbroek, J.K.
nism of Ni–Ga nanoparticle formation was proposed. Upon decom- Norskov, I. Stensgaard, Science 279 (1998) 1913–1915.
position of nitrates, highly dispersed Ni(II) nanocrystallites serve as [2] K. Kovnir, M. Armbrüster, D. Teschner, T.V. Venkov, F.C. Jentoft, A. Knop-
Gericke, Y. Grin, R. Schlögl, Sci. Technol. Adv. Mater. 8 (2007) 420–427.
centres for further reduction of the nickel and gallium oxides and [3] J.C. Bauer, X. Chen, Q. Liu, T.-H. Phan, R.E. Schaak, J. Mater. Chem. 18 (2008)
metallic gallium is incorporated into the nickel crystal lattice to 275–282.
form an intermetallic phase (Fig. 12). Formation of the stoichiom- [4] S. Park, Y. Xie, M.J. Weaver, Langmuir 18 (2002) 5792–5798.
[5] C. Roychowdhury, F. Matsumoto, V.B. Zeldovich, S.C. Warren, P.F. Mutolo, M.
etric Ni–Ga compound always proceeds through one or more Ni- Ballesteros, U. Wiesner, H.D. Abruna, F.J. DiSalvo, Chem. Mater. 18 (2006)
rich Ni–Ga phases. 3365–3372.
Unravelling the mechanism of the intermetallic phase forma- [6] X. Ji, K.T. Lee, R. Holden, L. Zhang, J. Zhang, G.A. Botton, M. Couillard, L.F. Nazar,
Nat. Chem. 2 (2010) 286–293.
tion could only be achieved by application of various complemen-
[7] M. Armbrüster, K. Kovnir, M. Friedrich, D. Teschner, G. Wowsnick, M. Hahne, P.
tary X-ray and electron microscopy methods. It has been shown Gille, L. Szentmiklosi, M. Feuerbacher, M. Heggen, F. Girgsdies, D. Rosenthal, R.
that the resulting catalyst activity depends intimately on both Schlögl, Y. Grin, Nat. Mater. 11 (2012) 690–693.
composition and preparation conditions. In terms of Ni/Ga ratio, [8] J.K. Edwards, B. Solsona, E.N. Ntainjua, A.F. Carley, A.A. Herzing, C.J. Kiely, G.J.
Hutchings, Science 323 (2009) 1037–1041.
the maximum methanol turnover frequency has been observed [9] S. Siahrostami, A. Verdaguer-Casadevall, M. Karamad, D. Deiana, P. Malacrida,
for bimetallic Ni–Ga nanoparticles containing 67.5 wt% of Ni, as B. Wickman, M. Escudero-Escribano, E.A. Paoli, R. Frydendal, T.W. Hansen, I.
demonstrated by a combined study of catalytic activity and ele- Chorkendorff, I.E.L. Stephens, J. Rossmeisl, Nat. Mater. 12 (2013) 1137–1143.
[10] A. Verdaguer-Casadevall, D. Deiana, M. Karamad, S. Siahrostami, P. Malacrida,
mental composition (by means of XRF). Second, the results of T.W. Hansen, J. Rossmeisl, I. Chorkendorff, I. Stephens, Nano Lett. 14 (2014)
in situ XRD and Rietveld refinement suggest that dispersion could 1603–1608.
be enhanced by reducing the alloying temperature. Third, forma- [11] D.B. Thanh C. N., Sarrazin P., Cameron C., US Patent, 2000.
[12] D. Mei, M. Neurock, C.M. Smith, J. Catal. 268 (2009) 181–195.
tion of a more phase pure Ni5Ga3/SiO2 catalyst could be required [13] Y.M. Jin, A.K. Datye, E. Rightor, R. Gulotty, W. Waterman, M. Smith, M.
to suppress the methane formation further. In fact, it is still to be Holbrook, J. Maj, J. Blackson, J. Catal. 203 (2001) 292–306.
clarified whether CH4 stems from the presence of the Ni-rich Ni3Ga [14] M. Armbrüster, H. Borrmann, M. Wedel, Y. Prots, R. Giedigkeit, P. Gille, Z. Fur
Kristallographie-New Cryst. Struct. 225 (2010) 617–618.
phase, small pure Ni particles or it is produced on the adjacent Ni– [15] M. Armbrüster, M. Behrens, K. Fottinger, M. Friedrich, E. Gaudry, S.K. Matam,
Ni sites on the surface of Ni5Ga3. H.R. Sharma, Catal. Rev. – Sci. Eng. 55 (2013) 289–367.
With the aid of TEM and XRD, the influence of calcination on [16] M. Friedrich, D. Teschner, A. Knop-Gericke, M. Armbrüster, J. Catal. 285 (2012)
41–47.
catalysts activity was studied: it was demonstrated that when cal-
[17] L.D. Li, B.S. Zhang, E. Kunkes, K. Fottinger, M. Armbrüster, D.S. Su, W. Wei, R.
cination in stagnant air is used for pre-treatment of the catalyst Schlögl, M. Behrens, ChemCatChem 4 (2012) 1764–1775.
precursor, final size of the Ni–Ga nanoparticles is negatively [18] K. Kovnir, J. Osswald, M. Armbrüster, R. Giedigkeit, T. Ressler, Y. Grin, R.
affected. As the temperature of calcination is increased, the particle Schlögl, in: E.M. Gaigneaux, M. Devillers, D.E. De Vos, S. Hermans, P.A. Jacobs,
J.A. Martens, P. Ruiz (Eds.) Scientific Bases for the Preparation of
dimensions are scaled up, and the overall activity per unit mass Heterogeneous Catalysts, Proceedings of the 9th International Symposium,
and per unit surface area of the catalyst is decreased. 2006, pp. 481–488.
88 I. Sharafutdinov et al. / Journal of Catalysis 320 (2014) 77–88

[19] W.M.H. Sachtler, Catal. Rev. – Sci. Eng. 14 (1976) 193–210. [37] W.X. Yuan, Z.Y. Qiao, H. Ipser, G. Eriksson, J. Phase Equil. Diffus. 25 (2004) 68–
[20] F. Studt, I. Sharafutdinov, F. Abild-Pedersen, C.F. Elkjær, J.S. Hummelshøj, S. 74.
Dahl, I. Chorkendorff, J.K. Nørskov, Nat. Chem. 6 (2014) 320–324. [38] S. Martosudirjo, J.N. Pratt, Thermochim. Acta 17 (1976) 183–194.
[21] M.V. Twigg, M.S. Spencer, Appl. Catal. A – General 212 (2001) 161–174. [39] A. Haghofer, K. Föttinger, F. Girgsdies, D. Teschner, A. Knop-Gericke, R. Schlögl,
[22] R. Lindberg, J. Sjoblom, G. Sundholm, Colloid Surf. A – Physicochem. Eng. G. Rupprechter, J. Catal. 286 (2012) 13–21.
Aspect 99 (1995) 79–88. [40] G. Wowsnick, D. Teschner, I. Kasatkin, F. Girgsdies, M. Armbrüster, A. Zhang, Y.
[23] J.-D. Grunwaldt, N. van Vegten, A. Baiker, Chem. Commun. (2007) 4635–4637. Grin, R. Schlögl, M. Behrens, J. Catal. 309 (2014) 209–220.
[24] B. Ravel, M. Newville, J. Synchr. Rad. 12 (2005) 537–541. [41] G. Wowsnick, D. Teschner, M. Armbrüster, I. Kasatkin, F. Girgsdies, Y. Grin, R.
[25] S.I. Zabinsky, J.J. Rehr, A. Ankudinov, R.C. Albers, M.J. Eller, Phys. Rev. B 52 Schlögl, M. Behrens, J. Catal. 309 (2014) 221–230.
(1995) 2995–3009. [42] M. Friedrich, S. Penner, M. Heggen, M. Armbrüster, Angew. Chem. – Int. Edit.
[26] J.J. Rehr, R.C. Albers, Rev. Mod. Phys. 72 (2000) 621–654. 52 (2013) 4389–4392.
[27] H. Okamoto, J. Phase Equil. Diffus. 31 (2010) 575–576. [43] D.L. Chiavassa, S.E. Collins, A.L. Bonivardi, M.A. Baltanás, Chem. Eng. J. 150
[28] P. Sabatier, J.B. Senderens, Comptes Rendus Hebdomadaires Des Seances De L (2009) 204–212.
Academie Des Sciences 134 (1902) 514–516. [44] L. Li, B. Zhang, E. Kunkes, K. Föttinger, M. Armbrüster, D.S. Su, W. Wei, R.
[29] S. Bhan, K. Schubert, J. Less-Common Met. 17 (1969) 73–90. Schlögl, M. Behrens, ChemCatChem 4 (2012) 1764–1775.
[30] G.P. Bergeret, Particle size and dispersion measurements, in: Handbook of [45] X.W. Zhou, J. Qu, F. Xu, J.P. Hu, J.S. Foord, Z.Y. Zeng, X.L. Hong, S.C.E. Tsang,
Heterogeneous Catalysis, Wiley-VCH, 2008, pp. 738–765. Chem. Commun. 49 (2013) 1747–1749.
[31] M.P. Andersson, E. Abild-Pedersen, I.N. Remediakis, T. Bligaard, G. Jones, J. [46] S.E. Collins, M.A. Baltanás, J.L. Garcia Fierro, A.L. Bonivardi, J. Catal. 211 (2002)
Engbwk, O. Lytken, S. Horch, J.H. Nielsen, J. Sehested, J.R. Rostrup-Nielsen, J.K. 252–264.
Norskov, I. Chorkendorff, J. Catal. 255 (2008) 6–19. [47] M. Armbrüster, G. Wowsnick, M. Friedrich, M. Heggen, R. Cardoso-Gil, J. Am.
[32] J.K. Norskov, T. Bligaard, B. Hvolbaek, F. Abild-Pedersen, I. Chorkendorff, C.H. Chem. Soc. 133 (2011) 9112–9118.
Christensen, Chem. Soc. Rev. 37 (2008) 2163–2171. [48] Y. Yamamoto, N. Nawa, S. Nishimoto, Y. Kameshima, M. Matsuda, M. Miyake,
[33] T. Toupance, M. Kermarec, C. Louis, J. Phys. Chem. B 104 (2000) 965–972. Int. J. Hydrog. Energy 36 (2011) 5739–5743.
[34] C. Louis, Z.X. Cheng, M. Che, J. Phys. Chem. 97 (1993) 5703–5712. [49] R. Guerin, A. Guivarch, J. Appl. Phys. 66 (1989) 2122–2128.
[35] http://www.fiz-karlsruhe.de/icsd.html. [50] E.R. Jette, F. Foote, J. Chem. Phys. 3 (1935) 605–616.
[36] R. Ducher, R. Kainuma, K. Ishida, Intermetallics 15 (2007) 148–153. [51] H. Kedesdy, A. Drukalsky, J. Am. Chem. Soc. 76 (1954) 5941–5946.

S-ar putea să vă placă și