Journal of Catalysis 305 (2013) 277–289

Contents lists available at SciVerse ScienceDirect

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

Molybdenum-based catalysts for the decomposition of ammonia: In situ

X-ray diffraction studies, microstructure, and catalytic properties
Valeria Tagliazucca, Klaus Schlichte, Ferdi Schüth, Claudia Weidenthaler ⇑
Max-Planck-Institut für Kohlenforschung, Kaiser-Wilhelm-Platz 1, 45470 Mülheim an der Ruhr, Germany

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

Article history: The ammonia decomposition reaction over molybdenum-based catalysts is an example for the complex
Received 11 March 2013 influence of different factors, such as phase composition, size of crystalline domains, or defect concentra-
Revised 7 May 2013 tion, on the catalytic behavior of a material. In situ powder diffraction allows the direct analysis of how
Accepted 11 May 2013
catalysts change during a reaction with respect to the atomic structure or microstructure in terms of
Available online 29 June 2013
defects or size changes. In this article, the influence of catalyst treatment such as pre-reduction or ball
milling on the catalytic properties is discussed in detail.
Keywords:
Ó 2013 Elsevier Inc. All rights reserved.
In situ powder diffraction
Ammonia decomposition
Structure analysis
Molybdenum oxides
Molybdenum nitrides

1. Introduction catalysts with structure and/or phase changes monitored by in situ

Ammonia is an excellent hydrogen carrier, and therefore, the
decomposition of ammonia would be an elegant way to generate
hydrogen for fuel cell applications without formation of COx. On
the other hand, ammonia can be used directly as feed for solid
oxide fuel cells. Up to now, the benchmark catalyst is ruthenium
dispersed on carbon nanotubes [1,2]. The activity is significantly
influenced by the appropriate choice of promoters which are added
to the supported catalysts [3]. Due to economical aspects, the sub-
stitution of noble metal catalysts by metal oxides is under discus-
sion. As an example, iron oxide-based materials have been studied
as catalysts for ammonia decomposition [4]. Since several years,
compounds, such as molybdenum carbides and nitrides, are con-
sidered as potential catalysts for several catalytic reactions [5–8]
or as catalyst supports [9]. The formation of molybdenum nitrides
by nitridation of molybdenum oxides under ammonia flow was the
topic of a large number of publications. Heating MoO3 in NH3
atmosphere is one way to obtain molybdenum nitrides with high
surface areas [10–14]. The formation of different intermediates,
such as molybdenum bronze, molybdenum oxynitrides, and
molybdenum nitrides with different stoichiometries, makes this
system quite complicated. Even though many aspects of the nitrid-
ation of MoO3 have already been studied in detail, there are many
open questions, especially with respect to catalysis. Aim of our
work was to correlate the catalytic activity of molybdenum-based
powder diffraction. Another topic of interest was the influence of
microstructure properties on the catalytic performance of such cat-
alysts. For this purpose, also ball-milled catalyst precursors have
been investigated.
2. Experimental
2.1. Catalyst preparation
Commercial MoO3 (MoO3-C) (Merck, purity 99%) was used as
starting material for the synthesis of catalysts for the conversion
of ammonia. A second type of MoO3 samples was prepared by
impregnation and subsequent calcination of porous carbon with
an aqueous solution of molybdic acid (H2MoO4, min 83% of
MoO3, Riedel de Haën) according to the literature [15]. For the
preparation, either 7 mL of saturated aqueous molybdic acid aque-
ous solution or 7 mL of a 0.96 M phosphomolybdic acid hydrate
(puriss. p.a.) (PMA) aqueous solution was added dropwise under
stirring to 15 g of an as-received activated carbon (GPAK-500, Gel-
type purchased from CarboTex GmbH). The impregnated carbon
was then calcined for 2 h at 550 °C in air with a heating rate of
4 °C min 1. The samples obtained by this procedure are denoted
as MoO3-G.
2.2. Ball milling

⇑ Corresponding author. Ball milling of MoO3-C was performed in a Retsch MM2000

E-mail address: claudia.weidenthaler@mpi-mail.mpg.de (C. Weidenthaler). mixer mill equipped with tungsten carbide vials. One ball of the

0021-9517/$ - see front matter Ó 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jcat.2013.05.011
278 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289

Fig. 1. SEM images of sample (a) and (b) MoO3-C and of sample MoO3-G (c, d).

100.0 DIP h/h diffractometer (Bragg–Brentano geometry) using Cu Ka1,2

radiation. Divergence slit was set to 0.6 mm, receiving slit was
set to 0.6 mm, and the width of the horizontal mask was set to
80.0
Relative Intensity (%)

4 mm. Data were collected in the 2h range of 20–85° for molybde-

60.0
40.0
(b)
20.0
(a)
0.0
12 16 20 24
2Theta/°
Fig. 2. XRD patterns of (a) MoO3-C and (b) MoO3-G.
same material with a diameter of 5 mm was used as grinding med-
ium. The ball-to-powder mass ratio was 8:1 and the rotational
speed was 900 rpm. Samples were milled for 5 or 10 h, and after
every hour, milling was suspended for 15 min.
2.3. Catalyst characterization
All samples were characterized by means of X-ray powder dif-
fraction (XRD) at ambient conditions, in situ XRD under reaction
conditions, scanning electron microscopy and transmission elec-
tron microscopy (SEM and TEM), nitrogen adsorption, and X-ray
photoelectron spectroscopy (XPS). Moreover, catalytic activity
was tested in a fixed-bed reactor. XRD experiments under ambient
conditions were carried out in reflection geometry on a Stoe STA-
num oxides and carbides and in the 2h range of 30–90° for the
molybdenum nitrides. In situ XRD experiments were performed
on a PANalytical X’Pert diffractometer equipped with an Anton
Paar XRK 900 high-temperature reaction chamber and with Cu
Ka1,2 radiation (40 kV, 40 mA) using the following slit configura-
tion: primary and secondary soller slits 0.04 rad, divergence slit
0.5°, and anti-scatter slits 1°. No monochromator but a secondary
Ni filter has been used. Data were collected in the 2h range of
20–85° for molybdenum oxides and carbides and the 2h range of
28 32 36 40 44 30–85° for the molybdenum nitrides. The Anton Paar reaction
chamber was equipped with a MarcorÒ sample holder (6–10 mm
diameter). Measurements were taken under pure ammonia flow
with a space velocity of 15 000 mL/gsample h and heating from room
temperature to 800 °C with a heating rate of 40 °C min 1. The sam-
ple was deposited on a sieve plate (10 mm diameter, 1 mm depth)
which allows the reaction gas to pass through the catalyst bed. The
product gases leave the reaction chamber through an exhaust pipe
under the sample.
For the analysis of the microstructure, XRD patterns were ana-
lyzed by the Whole Powder Pattern Modelling (WPPM) approach.
The method is based on a direct physical modeling of the micro-
structure from XRD data in terms of density of specific lattice de-
fects, shape, and size distribution of coherently diffracting
domains. Domain size, domain size distribution, and dislocation
density (q) were retrieved using the PM2K software based on the
WPPM approach [16]. The Cagliotti function [17] was used to
parameterize the instrumental effects and was obtained from the
analysis of diffraction data of a silicon reference material (NIST
SRM 640C). Domain size distribution of (spherical) domain diame-
 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289 279

Fig. 3. (a) NH3 conversion versus reaction temperature for MoO3-G, space velocity of 15,000 mL/gsample h, (b) stability test performed at 600 °C with space velocity of
15,000 mL/gsample h.

100
MoO2
90
MoxNy 80

MoN 70
60
wt.%

(h) 800°C
50
40
30
MoN /MoxNy

(g) 700°C
20
10 MoxNy Fm3m
MoN P-6m2
0
MoO2
(f) 650°C 300 350 HxMoO3
500 600
650 700
MoO3
750
Temp. [°C] 800
(e) 600°C
Fig. 5. Quantitative phase analysis from 300 to 800 °C starting from MoO3-G. The
MoO2

quantitative analysis has been obtained by Rietveld refinements.

(d) 500°C
powder). For a typical NH3 conversion experiment, the reactor
temperature was increased from 250 to 800 °C in 50 °C steps. At
(c) 350°C each temperature step, the reaction was allowed to equilibrate
HxMoO3 for 30 min to reach steady-state conditions. During each equilibra-
tion step, conversion data were recorded six times. Reaction prod-
300°C ucts were quantified using a micro-GC (Agilent). Stability tests of
(b) the catalysts were performed at defined temperatures for at least
16 h. The morphology of the MoO3 crystals was studied by scan-
MoO3

ning electron microscopy (Hitachi S-3500N for SEM and Hitachi

(a) 25°C
5500 for high-resolution SEM) and transmission electron micros-
copy (TEM, Hitachi HF 2000). N2 adsorption experiments at 77 K
were performed on a Nova 3200e instrument (Quantachrome
20 25 30 35 40 45 50 55
Instruments). Prior to analysis, the samples were outgassed in vac-
2Theta/° uum at 200 °C for 4 h. The specific surface areas were determined
from data in the 0.05–0.2 p/p0 range by applying the Brunauer–
Fig. 4. Selected powder diffraction patterns collected in situ during NH3 decom-
position on MoO3-G (space velocity of 15,000 mL/gsample).
Emmett–Teller (BET) method [18].

3. Results and discussion

ters was calculated refining l and r parameters, defining the log-
normal distribution of (spherical) domain diameters, from experi-
mental data.
Catalytic activity was tested in a fixed-bed reactor fed with pure
ammonia (SV 15,000 mL/gsample h, inner diameter of reactor: 6 mm,
50 mg catalyst, purity of ammonia 99.98%, catalyst was used as
Commercial MoO3-C was used as starting material for the first
series of experiments. It consists of micrometer-sized crystalline
particles with an inhomogeneous size range and platelet-shaped
crystals (Fig. 1a and b). A second series of experiments was per-
formed with MoO3-G crystals that had been obtained via the acti-
280 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289

12 wt% MoO3
67 wt% HxMoO3 4.5000 14.1500
21 wt% MoO2 a (NH3) c (NH3) b (NH3)

a and c lattice parameter [Å]

4.4000
a (air) c (air) b (air) 14.1000

b lattice parameter [Å]

400°C 4.3000
4.2000 14.0500
350°C 4.1000
4.0000 14.0000
300°C
3.9000
200°C 13.9500
3.8000

100°C 3.7000 13.9000

3.6000
25°C 3.5000 13.8500
0 50 100 150 200 250 300 350 400 450
22 23 24 25 26 27 28 29 30
temperature [°c]
2Theta/°
(a) (b)
Fig. 6. (a) Selected XRD patterns collected in situ under NH3 atmosphere in the range 25–400 °C, (b) Change in the lattice parameters with temperature during heating under
NH3 atmosphere (solid symbols) and air (open symbols).

Fig. 7. (a) Crystal structure of MoO3, unit cell, (b) part of the crystal structure of HxMoO3 according to the structure given by Adams et al. [26].

Fig. 8. (a) and (b): SEM images and (c) HTEM image of HxMoO3.

vated carbon route. The mean particle sizes are about 20 lm. The
parallelepiped-type morphology is formed by superimposed layers
(island type, Fig. 1c and d). Both samples contain purely ortho-
rhombic MoO3 as crystalline phase (Fig. 2). Due to different pre-
ferred orientations, the intensity distributions of the diffraction
peaks are slightly different.
3.1. Catalytic performance and in situ XRD experiments
Starting from MoO3-G, a significant conversion of NH3 into
N2 and H2 is detectable at temperatures above 450 °C. The
activity increases considerably above 550 °C. Full conversion is
reached at 650 °C (Fig. 3a). Catalytic data were recorded at tem-
perature steps of 50 °C. After reaching the individual tempera-
tures, product gas composition was recorded six times every
5 min. Repetition of the measurement at one temperature
shows increases in conversion with time, underlining that the
reaction takes some time to reach steady state. After the six
analyses, typically constant conversion was reached and these
values are given in the figures. After cooling from 800 °C to
600 °C, stability tests prove constant activity of the material
for at least 16 h (Fig. 3b).
 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289 281

As will be shown later, this is caused by sintering at temperatures

Fig. 9. Typical XRD patterns for molybdenum-based catalysts (a) before (25 °C,
MoO3) and (b) after ammonia decomposition catalysis (800 °C, MoxNy).
However, while in the first cycle, the conversion at 600 °C is
about 80% (Fig. 3a), the high-temperature treatment at 800 °C
leads to a loss of activity. The subsequent stability test performed
after cooling from 800 to 600 °C shows only about 60% conversion.
above 650 °C.
In situ XRD measurements were performed to monitor phase
transformations, phase formation processes, and microstructure
changes taking place during the NH3 decomposition reaction. In or-
der to correlate catalytic data obtained from a fix-bed reactor with
in situ XRD data, the diffraction patterns were recorded under sim-
ilar catalytic reaction conditions.
Starting from MoO3, the orthorhombic phase is stable up to
300 °C (Fig. 4a). With increasing temperature under NH3 flow,
some reflections are significantly shifted toward lower diffraction
angles, and at 350 °C, molybdenum bronze, HxMoO3, and MoO2
form (Fig. 4c). Between 450 °C and 500 °C, the reduction of Mo6+
to Mo4+ proceeds and hexagonal molybdenum nitrides start to
crystallize. At temperatures above 650 °C, MoO2 is completely nit-
ridated. The stoichiometry of the nitrides changes with tempera-
ture, starting with a Mo/N ratio of about 1:1 at 600 °C and
ending in a stoichiometry of Mo/N of 1:0.6 at 800 °C (Fig. 4h).
Phase changes were followed quantitatively by Rietveld refine-
ments (Fig. 5). The precise compositions and therewith the crystal
structures of the nitride formed during ammonia decomposition
are difficult to assess due to the quality of the in situ XRD data.
Low-valent molybdenum nitrides (Mo2N1±x) are discussed to crys-
tallize in two different structure types: b-Mo2N1±x with a tetrago-

Fig. 10. (a) HTEM images of MoO3-C before catalysis and (b) of MoxNy obtained after catalysis at 800 °C.

Table 1
Nitrogen adsorption results obtained before and after catalysis at 800 °C.

Label Specific surface area BET (m2 g 1

) pre- Total pore volume Specific surface area BET (m2 g 1
) post- Total pore volume
catalysis cm3 g 1 catalysis cm3 g 1
MoO3-C 1 0.002 2 0.02
MoO3-G 1 0.005 22 0.10
MoO3-C- 13 0.09 2 0.03
b.m.10 h

Fig. 11. High-resolution SEM images with different magnifications of molybdenum nitride after catalytic conversion of NH3 up to 800 °C.
282 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289

17 MoO2

15 γ-MoxNy Fm3m

δ-MoN
13

MoN/Mox Ny
(h) 800°C
11
ln SV

9
(g) 700°C
7

5
(f) 600°C
3
1.1 1.15 1.2 1.25 1.3 1.35 1.4
1/T (*10-3) (e) 500°C

mono-MoO 2
Fig. 12. Arrhenius plot: change in the space velocity (SV) with temperature for a
fixed conversion of 12% measured during ammonia decomposition over MoO3-C.
(d) 400°C

(c) 350°C

(b) 550°C/H2

MoO3
(a) 25°C

Fig. 13. Lognormal distribution of crystallite domains obtained from WPPM of data
collected during the in situ measurements at 750 and 800 °C with MoO3-G as
starting material.
nal structure at lower temperatures and a cubic rock-salt-type
structure at higher temperatures [19].
The crystal structures of the molybdenum nitrides can be de-
duced from the molybdenum structure as follows: starting from
the bcc molybdenum structure, the tetragonal structure forms
by the occupation of the octahedral interstices [20]. The cubic c-
Mo2N1±x (Fm3m) structure can be generated by statistical filling
of the octahedral interstitials with nitrogen atoms [19,21]. Troits-
kaya and Pinsker observed an additional cubic primitive c-Mo2N1±x
(Pm3m) structure [22]. The Pm3m structure is interpreted as a
structure with ordered vacancies which can be distinguished from
the Fm3m structure by the presence of superstructure reflections.
The structures of these three compounds are closely related and
differ mostly in their nitrogen content. For the data collected ex
situ with a good counting statistics that are presented later in this
paper, distinction between the two cubic c-Mo2N1±x phases was
possible. Due to the structural complexity, the nitrogen-richer
phases formed at higher temperatures are assigned as MoxNy in
the following.
For the hexagonal molybdenum nitrides d-MoN, the crystal
structure family is even more complex. d-MoN is known to crystal-
lize at least in three crystal structures: d-MoN with WC structure
type (space group P-6m2) [23], d-MoN with NiAs structure type
(space group P63/mmc), and d-MoN with a distorted NiAs structure
(space group P63mc) [23,24]. The structure of Mo5N6 is described
in P63/m in the MoS2-type structure [25]. For the refinement of
all data containing hexagonal d-MoN, crystal structures in space
group P-6m2 and P63mc were considered [24].
20 25 30 35 40 45 50 55
2Theta/°
Fig. 14. Selected powder diffraction patterns collected at 25 °C (a), after reduction
in H2 for 30 min (b), cooling to 350 °C (c) and subsequent heating under NH3 flow
up to 800 °C (d–h).
The processes taking place up to 400–450 °C are very important
for the formation of high surface area compounds and therefore
deserve a more detailed discussion. Two main changes in the pow-
der patterns take place in the temperature range up to 400 °C: (a)
the systematic shift of MoO3 reflections and (b) the formation of a
molybdenum bronze (Fig. 6).
The shift of the MoO3 reflections between room temperature
and 350 °C (Fig. 6a) is caused by the anisotropic thermal expansion
of the orthorhombic crystal structure of MoO3. Since an identical
shift is also observed upon heat treatment under ambient atmo-
sphere (Fig. 6b), it is exclusively attributed to thermally induced
changes in the crystal structure. While a and c lattice parameters
are more or less constant, the b lattice parameter increases signif-
icantly (Fig. 6b). These results are in accordance with the observa-
tions by Adams et al. [26]. The anisotropic expansion of the unit
cell is based on the layered orthorhombic crystal structure of
MoO3. The structure is built of MoO6 octahedra which form layers
perpendicular to the crystallographic b-axis (Fig. 7a). The van der
Waals gap between the layers expands significantly with temper-
ature, resulting in an increase in the b lattice parameter, whereas
the other two parameters remain almost unchanged as depicted
in Fig. 6b.
The main phase present at 400 °C is assigned to hydrogen
molybdenum bronze HxMoO3. Cooling molybdenum bronze to
room temperature under NH3 leads to the reduction of the remain-
 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289 283

γ-MoxNy
(a) δ-MoN
(b)

(F)
(C)

(E)
(B)

(D)
(A)
12 16 20 24 28 32 36 40 44 12 16 20 24 28 32 36 40 44
2Theta/° 2Theta/°

Fig. 15. (a) Powder pattern of MoO3 after pre-reduction at 550 °C in H2 and (b) Powder pattern of pristine MoO3 after (A, D) subsequent NH3 decomposition reaction up to
800 °C, (B, E) followed by stability test for 16 h at 600 °C, and (C, F) two NH3 decomposition cycles up to 800 °C and stability test for 16 h at 600 °C. All diffraction patterns have
been collected ex situ after cooling to room temperature and removal from the reactor.

Fig. 16. Quantitative phase composition of samples heated up to 800 °C in the catalytic reactor under NH3 atmosphere (a) pre-reduced with H2 at 550 °C prior to NH3
decomposition and (b) pristine MoO3 before the catalytic experiment.

Fig. 17. NH3 conversions for pre-reduced MoO3 and pristine MoO3 during the first
and the second cycle.
ing MoO3 to MoO2, while the bronze disappears. The HxMoO3
structure can be preserved by changing the gas atmosphere at
400 °C to N2 and cooling under N2 atmosphere. The crystal struc-
ture of the bronze is formed by insertion of hydrogen into the
van der Waals gap between the loosely bound layers of the MoO3
structure [13]. This changes the symmetry from Pbnm with
a = 3.972 Å, b = 14.106 Å, and c = 3.659 Å to Cmcm with
a = 3.882 Å, b = 14.390 Å, and c = 3.720 Å [26]. With the thermal
expansion, hydrogen is intercalated (Fig. 7b). The formation of
hydrogen-intercalated molybdenum bronze has been further
Fig. 18. Lognormal domain size and domain size distribution for pre-reduced and
pristine samples after the first and the second cycle evaluated for the cubic nitride
phases obtained after reaction at 800 °C.
investigated by a combined theoretical and experimental approach
by Adams et al. [27]. The expansion of the crystal structure along
the b-direction of the original MoO3 structure introduces extra
strain to the structure, and as a consequence, the large crystalline
domains decompose into small crystalline domains. The gas
adsorption analysis of the bronze sample measured after heating
to 400 °C and cooling to room temperature reveals a significant
284 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289

Table 2
MoO2
Microstructure data for samples obtained after ammonia decomposition at 800 °C
starting from pre-reduced and pristine samples. With Dave.: mean domain size in nm, γ-MoxNy
sd: size distribution, q: dislocation density.

Dave. (nm) sd q (1016 m 2) δ-MoN

δ-MoN/γMoxNy
Pristine, first cycle 5.2 3.8 23 (h) 800°C
Pre-reduced, first cycle 10.6 6.9 1.3
Pristine, second cycle 7.6 4.2 17
Pre-reduced, second cycle 12.3 13.5 38
(g) 750°C

change in the specific surface area. The specific surface area in-
creased from 1 m2 g 1 for MoO3 to 10 m2 g 1 for HxMoO3, and a to- (f)
650°C
tal pore volume of 0.062 cm3 g 1. HTEM/SEM images (Fig. 8) show
that starting from smooth single crystals of MoO3, the surface of
the bronze crystals became fragmented and the large single-crystal 500°C
(e)
domains were decomposed into smaller pieces of HxMoO3.

mono-MoO2
The final nitrides formed during ammonia decomposition con-
sist of very small crystalline domains. This is supported by the
450°C
XRD patterns (Fig. 9) of pristine MoO3 measured at room temper- (d)
ature and of MoxNy collected during reaction at 800 °C. From these
data, it is possible to deduce the change in the domain sizes from
the widths and the shape of the reflections.
The formation of smaller crystalline domains is also supported (c) 400°C
by electron microscopy images taken before and after catalysis
(Fig. 10). The images show the topotactic nature of the breakup
of the pristine crystals. The overall morphology of the pristine 350°C
(b)
MoO3 crystals is maintained, while the large single crystals are
decomposed into small MoxNy domains.

MoO3
Upon nitridation, the specific surface areas and the total pore
volumes have increased significantly (Table 1). This is related to (a) 25°C
the formation of a porous structure with pores ranging also inside
the particles (Fig. 11). 20 24 28 32 36 40 44 48
Some of the pores are accessible from the surface of the parti-
2Theta
cles, but not all pores might be completely accessible. Correlating
the phase composition of the sample with the catalytic conversion Fig. 20. Selection of powder patterns collected in situ during the decomposition of
allows the conclusion that the activity is increased with the ammonia over MoO3 ball milled for 5 h.
appearance of nanosized MoxNy phases. This is supported by the
Arrhenius-type plot shown in Fig. 12. Since dependence of rate
on the partial pressures of reagents and product is complex, the appearance of the oxide and the increase in the content of nano-
space velocity was adjusted to result in equal conversion at differ- sized molybdenum nitrides.
ent temperatures by changing catalyst amount and flow rate. Un- Microstructure analysis by Whole Powder Pattern Modelling
der these circumstances, the space velocity is a direct measure (WPPM) allows a more detailed view into the processes during
for reaction rate. The plot of the ln of space velocity for a fixed con- the reactions at high temperatures. The size distribution is as-
version against inverse temperature does not display a linear sumed to be lognormal according to Langford et al. [28]. The anal-
dependency, but a distinct kink at about 550 °C, with decreased ysis of the peak profile reveals that at 800 °C, the domains are
activation energy at temperatures exceeding this value. This slightly bigger (Dave. 4.5 nm, sd: 2.1) compared to the sizes of those
change in the apparent activation energy is associated with the dis- determined at 750 °C (Dave. 3.1 nm, sd: 4.5) (Fig. 13). Nevertheless,

100
(b)
pristine MoO3
Relative Intensity (%)

80

60

40
MoO3
5h ball milling
20

0
20 22 24 26 28 30 32
2Theta/°

Fig. 19. (a) SEM after ball milling for 5 h, (b) XRD of MoO3 before and after ball milling.
 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289 285

Fig. 21. Conversion of catalysts pre-treated under different conditions (a) commercial MoO3 (MoO3-C) as received and ball-milled for 5 and 10 h. (b) Stability tests run at
600 °C after performing the first cycle either at 650 °C or at 800 °C, space velocity of 15,000 mL/gsample h.

Fig. 22. SEM micrographs of MoO3-C-b.m.5 h. (a) Before catalysis and (b) after catalysis at 800 °C.

the mean sizes of the crystalline domains are still very small. Addi-
tionally to the slight increase in size, the size distribution becomes
broader, probably due to sintering of the particles.
All experiments have also been performed with commercial
MoO3 as starting material. Since the phase changes are absolutely
the same as for MoO3-G, we do not show the in situ diffraction pat-
terns here.
3.2. MoO3 with pre-reduction in H2 at 550 °C, in situ diffraction studies
In order to investigate the influence of the pre-reduction on the
state and the activity of the catalyst, MoO3 was first reduced in the
XRD reaction chamber under 100% H2 atmosphere at 550 °C and
then kept for 30 min before cooling to 350 °C. After cooling, the
gas atmosphere was switched to 100% NH3 and the catalytic
decomposition was started (Fig. 14). In the following, the samples
used after pre-reduction in H2 are named pre-reduced and the
samples directly used for ammonia decomposition without any
conditioning are called pristine.
The reduction of MoO3 leads to the formation of monoclinic
MoO2 which is stable up to 700 °C. In difference to MoO2 formed
during the direct reaction of pristine MoO3 with ammonia, the
crystallites of MoO2 formed by the pre-reduction are larger than
100 nm (Fig. 14b). Without pre-reduction, the formation of nano-
sized molybdenum nitrides starts at about 500 °C, while the large
crystalline MoO2 domains are nitridated only at very high temper-
atures of about 700 °C (Fig. 14).
Additional experiments were performed to analyze changes
taking place during cycling of the catalytic reaction and/or stability
test for 16 h. Two series of experiments were performed in the con-
ventional reactor: one series with a pre-reduced sample (Fig. 15a)
and a second series starting from pristine MoO3 (Fig. 15b). Starting
from pre-reduced MoO3 (MoO2), cubic MoxNy forms during the first
cycle at 800 °C (Fig. 15A). This is in line with the results obtained
by the in situ X-ray diffraction experiment. However, an additional
stability test performed at 600 °C for 16 h leads to the formation of
hexagonal d-MoN (Fig. 15B). The phase mixture is maintained also
after two cycles and subsequent stability test (Fig. 15C). Compared
to the previously reported refinements, the crystalline phases
could not unambiguously be assigned to one hexagonal and one
cubic structure. The refinements indicate that the cubic phases
crystallize in Pm3m, whereas the presence of two hexagonal d-
MoN phases, one with the distorted NiAs structure in space group
P63mc and the other one with the WC structure type in P-6m2,
cannot be excluded. For further investigation, either high-resolu-
tion diffraction data or electron diffraction data are required.
If no pre-reduction step is performed before ammonia decom-
position, the dominant crystalline phase is cubic MoxNy. The hex-
agonal d-MoN phase forms only if the reaction is cycled and a
stability test for 16 h is included (Fig. 15F). The quantitative phase
composition obtained by Rietveld refinements for both sample ser-
ies is summarized in Fig. 16. Pre-reduction leads to the formation
of large amounts of hexagonal MoN during cycling and/or long-
term stability tests, which could be one reason for the reduced
activity during the second cycle (Fig. 17).
Ammonia conversion curves of the two samples series are
shown in Fig. 17. Starting from pristine MoO3 leads to a higher
activity of the catalyst compared with the pre-reduced sample
which could be due to the ease of nitride formation in the pristine
sample. However, both materials loose activity in the second cycle.
One reason for the deactivation might be sintering of the cata-
lysts. In order to analyze the changes in the domain size and do-
main size distribution, the cubic molybdenum nitride phase was
analyzed by line profile analysis (Fig. 18).
After one cycle up to 800 °C, the pristine sample consists of
smaller nitride domains with a narrow size distribution. The ni-
286 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289
Fig. 23. (a) Conversions for pristine and ball-milled MoO3-C in the first and second
cycles. (b) Stability tests at 600 °C for pristine and ball-milled MoO3-C performed
after the first and the second cycles. (c) Conversion during the fifth cycle for pristine
and ball-milled samples.
tride domains formed from the pre-reduced sample are much lar-
ger and have a broader size distribution (Table 2).
For both samples, the crystalline domains increase in size dur-
ing the second cycle, but the domains from the pristine sample re-
main much smaller. There also exist differences in the
concentration of dislocations. After reaction, the dislocation den-
sity, q, for the pristine material is one magnitude higher than in
the pre-reduced sample. However, already after the second cycle,
the dislocation density of the pre-reduced sample has increased
and the concentrations of dislocations for both samples become
similar. This indicates that after the first cycle, the size of the crys-
talline domains and the size distribution are predominantly
responsible for the better activity of the catalysts used without
any pre-treatment. From these results, it becomes clear that reduc-
tive pre-treatment of the MoO3 starting material with H2 is not
supportive for a high catalytic activity of the system. It rather
has a negative effect on the catalytic properties.
The experimental results demonstrate that not only the nitrid-
ation is important for activation but also that the crystal structures
and the composition of the phases as well as the crystallite sizes
seem to play an important role for the catalytic activity.
3.3. Changing the microstructure: Ball-milling experiments
The influence of the microstructure in terms of crystallite size
and specific surface area on the catalytic properties was studied
in more detail. For this purpose, commercial MoO3 was ball-milled
for 5 and for 10 h. As expected, ball milling decreases the domain
sizes of the MoO3 crystallites and thus causes a significant broad-
ening of reflections (Fig. 19). The reduction in crystallite sizes goes
along with an increase in the specific surface area from below
1 m2 g 1 to 13 m2 g 1 (Table 1).
Ball-milled (b.m.) MoO3 is almost completely reduced to MoO2
during ammonia decomposition already at about 300 °C, and
molybdenum nitrides already form at 400 °C instead of above
500 °C (Fig. 20). The catalytic activity of ball-milled MoO3 com-
pared to the pristine one is significantly higher (Fig. 21). At this
point, it has to be mentioned that the formation of molybdenum
bronze was not observed during the in situ XRD experiments. It
is possible that the bronze formation takes place so rapidly that
it is simply missed by the experiment.
Improved catalytic activity obviously results from the smaller
size of the crystalline MoO3 domains and an increased specific sur-
face area (Fig. 21a). However, prolonging ball milling to 10 h does
not improve activity further. Another important factor influencing
the activity is the maximum reaction temperature the catalyst has
been exposed to. Comparison of two ball-milled samples that were
used either at 650 °C or 800 °C in a first catalytic run showed sig-
nificantly different conversions in a second run. In long-time sta-
bility tests with these samples, 90% conversion for at least 14 h
was reached for the sample that had been used at 650 °C in the first
run, whereas the sample that had been used at 800 °C first only
converted about 70% ammonia in the second run at 600 °C
(Fig. 21b).
More pronounced sintering of the catalyst particles at 800 °C, as
can be easily confirmed by SEM (Fig. 22), is one of the main reasons
for the reduced catalytic activity.
Compared to pristine MoO3, the ball-milled sample has higher
activity in the first cycle, but the activity for the pristine MoO3 is
also significantly increased during the second cycle (Fig. 23). The
catalytic activity of the different samples operated at 600 °C is
maintained at least for 16 h (Fig. 23b). Interesting are the activity
changes observed during cycling of both sample series (Fig. 23c).
The differences in activity are discussed for conversion at 550 °C.
For pristine MoO3, the conversion at 550 °C is about 20% in the first
cycle and it increases to 40% during the second cycle and reaches
about 50% conversion during the fifth cycle. The conversion for
the ball-milled sample is about 45% in the first cycle, and it further
increases to 55% during the second cycle and decreases during the
fifth cycle to 45%: After 4–5 cycles, the activities for both series be-
come more or less comparable.
In order to understand this behavior, the crystalline phases
were determined quantitatively and the domain size distribution
and dislocation densities were evaluated from the powder diffrac-
tion data. Diffraction data were collected ex situ after running the
reaction in the reactor up to 650 °C and/or several repetitions of
the cycling between 350 °C and 650 °C. The phase analyses by Riet-
 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289 287

Fig. 24. Quantitative phase composition for (a) pristine material after different cycles and (b) ball-milled material after different cycles. The treatment includes stability test
for 10 h.

Fig. 25. XRD patterns of the ball-milled starting material after (a) 1 cycle, (b) 2
cycles, and (c) 5 cycles.
veld refinements reveal clear differences for both sample series
(Fig. 24). In case of the non-ball-milled samples, the main phase
is cubic c-MoxNy (Fm3m) accompanied by not completely reduced
and nitridated MoO3. Even after 5 cycles, 10–12 wt.% MoO2 remain
unreacted. Additionally, small amounts of hexagonal d-MoN (dis-
torted NiAs type) are formed (Fig. 24a).
Compared to this, the amount of MoO2 in the ball-milled sam-
ple after one cycle is much lower (4 wt.%). This is assigned to a
faster nitridation of the smaller MoO3 crystallites produced by ball
milling. Significantly different is the formation of about 55 wt.%
hexagonal d-MoN after the fifth cycle (Figs. 24b and 25) for the
ball-milled sample. The formation of high amounts of hexagonal
MoN could be one reason for the reduced activity of the ball-milled
MoO3 after 5 cycles.
Another factor which may influence the catalytic activity is the
change in the specific surface area of the pristine sample which in-
creases from 1 m2 g 1 before catalysis to 98 m2 g 1 after the first
cycle, then to 122 m2 g 1 after the second cycle, and it finally de-
creases to 110 m2 g 1 after the fifth cycle (Table 3). This is related
to the changes in the domain size and size distribution as deter-
mined from line profile analysis of the diffraction profiles
(Fig. 26) After one reaction cycle, the mean domain size (Dave.)
for the hexagonal d-MoN is 2.8 nm (size distribution (sd) of 0.12)
which is similar to the size of 3.1 nm (sd 1.4) after two cycles. After
5 cycles, the domains size is significantly increased to 10.3 nm (sd
0.2). For the main cubic phase, the trend is slightly different: Start-
ing from 9.3 nm (sd 0.85) after the first cycle, the mean size de-
creases to 3.3 nm but the distribution becomes broader (sd 2.8)
after 2 cycles. This could be associated with further nitridation of
MoO2 during the second cycle which may lead to additional forma-
tion of smaller molybdenum nitride domains and an increase in
specific surface area. After 5 cycles, the cubic phase shows sinter-
ing, resulting in a mean domain size of 8.2 nm and a significantly
broader size distribution (sd 19.6). This is also reflected in the low-
er specific surface area as listed in Table 3. It was also possible to
analyze the changes in the dislocation densities for the cubic
phase, while for the hexagonal phase, not all physical parameters
required for calculation were accessible (Table 3). The dislocation
density for the cubic phase increases from 8.4  1015 m 2 after
the first cycle to 21.0  1015 m 2 after the fifth cycle. The increase
in catalytic activity observed for the non-ball-milled catalysts dur-
ing the first reaction cycles is most likely influenced by more than
one factor: for both crystalline phases, the mean domain size de-
creases after the first cycle which is going along not only with an
increase in the specific surface area but also with an increase in
the total pore volume from 0.087 cm3 g 1 to 0.104 cm3 g 1. In
addition, the concentration of defects in the cubic phase increases

Table 3
Microstructure data determined from line profile analysis.

Dave. (nm) Cubic sd q (1015 m 2) Dave. (nm) hexagonal sd BET m2 g 1

TPV cm3 g 1

Non-ball-milled
1 Cycle 9.3 0.9 8.4(±1.6) 2.8 0.1 98 0.088
2 Cycles 3.3 2.8 9.3(±2.5) 3.1 1.4 122 0.104
5 Cycles 8.2 19.6 21(±1.9) 10.3 0.2 109 0.112
Ball-milled
1 Cycle 4.3 5.0 12(±0.8) 3.3 2.8 69 0.107
2 Cycles 5.4 11.6 15 3.4 3.1 53 0.118
4 Cycles 10.9 5.8 9.8(±0.3) 8.3 17 41 0.109
288 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289
Fig. 26. Domain size distribution determined from line profile analysis for the hexagonal and cubic molybdenum nitride phases for both series; left: for the series without ball
milling, right: for the series with ball milling of pristine MoO3 before the first cycle.
with the number of cycles, an effect which is presently not fully
understood.
The ball-milled sample starts from smaller MoO3 crystallites
with a higher surface area which may lead to a faster nitrida-
tion and therefore a higher activity in the first cycle. After
the first cycle, the accessible specific surface area of the ball-
milled sample is lower (69 m2 g 1) compared to the pristine
MoO3 material even though the domain sizes of both com-
pounds are smaller. The mean domain sizes and the size distri-
butions are similar for both molybdenum nitride phases
(Table 3). The mean domain size of 4.3 nm for the cubic phase
is much smaller than observed for the non-ball-milled sample.
With 12  1015 m 2, the dislocation density determined after
the first cycle is slightly higher than for the non-ball-milled
sample, but the concentration decreases during cycling (Table 3).
Already after the second cycle, sintering starts, as reflected by a
broadened size distribution. After 4 cycles, the mean domain
size is about 10.9 nm for the cubic phase which is larger than
observed for the non-ball-milled sample. This causes also a de-
creased surface area and in total a slight decrease in the overall
activity compared to the first cycle. Ball milling is advantageous
if the reaction runs continuously at 650 °C. Cycling of the reac-
tions seems to lead to pronounced and fast sintering of the
molybdenum nitride domains. Scanning electron microscopy
images display the formation of needle-like crystallites with
sizes of several tens of nanometers. In addition, the hexagonal
phase became the predominant crystalline phase present. All
these factors contribute to the decrease in catalytic activity dur-
ing cycling.
4. Conclusion
In situ XRD allows monitoring of phase formation and transfor-
mation pathways during nitridation of MoO3 with ammonia. These
results in combination with the information obtained by micro-
structure analysis provide a better understanding of structure–
activity relationship during ammonia decomposition starting from
molybdenum oxides.
In summary, it has been shown that ammonia decomposition
over catalysts starting from molybdenum oxides is an example
for a self-optimizing system. High activity of the catalysts is corre-
lated with the appearance of molybdenum nitride phases, and
activity is enhanced by the breakup of the initially large MoO3 do-
mains into nanosized MoN domains. The catalytically active nano-
sized molybdenum nitrides are formed directly during
ammonolysis. Starting the reaction from large-sized MoO3 crystal-
lites requires an activation phase during the first cycle. This is asso-
ciated with substantial changes in the atomic structure and of the
microstructure with respect to defects and domain sizes. The for-
mation of the molybdenum bronze is an important step for the
generation of high surface area molybdenum nitrides which takes
place during the decomposition reaction itself. The formation of
the molybdenum bronze leads to structural strain which induces
the decomposition of the originally large crystalline MoO3 domain
into smaller MoN domains.
Pre-reduction of the oxide with hydrogen has been shown to low-
er the catalytic activity rather than improving it. This is correlated
with the formation of large MoO2 crystallite domains without any
formation of small-sized molybdenum bronze. The large oxide par-
 V. Tagliazucca et al. / Journal of Catalysis 305 (2013) 277–289 289

ticles with low surface areas need much longer time for the complete sity Trento) is gratefully acknowledged for discussion of micro-
nitridation, and several conditioning cycles are required before the structure data. We thank Ulrich Holle (MPI Mülheim) for
activity is comparable to non-pre-reduced material. Catalytic exper- technical support with XRD.
iments at temperatures above 650 °C lead to a loss of activity, clearly
related to sintering of the catalyst particles. Operating ammonia References
decomposition only up to 650 °C produces high surface area molyb-
denum nitrides. The molybdenum oxides are not completely nitri- [1] S.F. Yin, B.Q. Xu, X.P. Zhou, C.T. Au, Appl. Catal. A 121 (2004) 1.
[2] S.F. Yin, Q.H. Zhang, B.Q. Xu, W.X. Zhu, C.F. Ng, C.T. Au, J. Catal. 224 (2004) 384.
dated during a first cycle, but progressively during successive [3] W. Pyrz, R. Vijay, J. Binz, J. Lauterbach, D. Buttrey, Top. Catal. 50 (2008) 180.
cycles. At the same time, growth of the crystalline domains starts. [4] M. Feyen, C. Weidenthaler, R. Güttel, K. Schlichte, U. Holle, A.H. Lu, F. Schüth,
However, the domain sizes are still small, and the specific surface Chem. Eur. J. 17 (2011) 598.
[5] J.S.J. Hargreaves, D. Mckay, J. Mol. Catal. A-Chem. 305 (2009) 125.
areas remain high. As consequence, it seems that particle growth [6] H.J. Lee, J.G. Choi, C.W. Colling, M.S. Mudholkar, L.T. Thompson, Appl. Surf. Sci.
does not much contribute to deactivation. 89 (1995) 121.
The microstructure properties can be additionally influenced by [7] S.S. Pansare, J.G. Goodwin Jr., Ind. Eng. Chem. Res. 47 (2008) 4063.
[8] R.S. Wise, R.J. Markel, J. Catal. 145 (1994) 335.
ball milling. Ball milling has been shown to be an appropriate way [9] K.J. Blackmore, L. Elbaz, E. Bauer, E.L. Brosha, K. More, T.M. McCleskey, A.K.
to generate highly active molybdenum-based materials for ammo- Burrella, J. Electrochem. Soc. 158 (2011) B1255.
nia decomposition which remain highly active if the catalyst is oper- [10] H.K. Park, J.K. Lee, J.K. Yoo, E.S. Ko, D.S. Kim, K.L. Kim, Appl. Catal. A 150 (1997)
21.
ated at a maximum temperature of 650 °C. Nitridation occurs faster,
[11] R. Kojima, K. Ayka, Appl. Catal. A 219 (2001) 141.
and at lower temperatures, the size of the crystalline domains is [12] C.H. Jaggers, J.N. Michaels, A.M. Stacy, Chem. Mater. 2 (1990) 150.
smaller and ball milling induces structure defects. However, after [13] J.-G. Choi, R.L. Curl, L.T. Thompson, J. Catal. 14 (1994) 218.
[14] C.W. Colling, J.G. Choi, L.T. Thompson, J. Catal. 160 (1996) 35.
performing several catalytic cycles, the activity decreases slightly,
[15] M. Schwickardi, T. Johann, W. Schmidt, F. Schüth, Chem. Mater. 14 (2002)
probably as a result of crystallite growth, sintering and the reduction 3913.
in defects. It can be concluded that even by starting from materials [16] M. Leoni, T. Confente, P. Scardi, Z. Kristallogr. Suppl. 23 (2006) 249.
with different microstructural properties, the catalytic properties [17] G. Cagliotti, A. Paoletti, F.P. Ricci, Nucl. Instrum. Methods 3 (1959) 223.
[18] S. Brunauer, P.H. Emmett, E. Teller, J. Am. Chem. Soc. 60 (1938) 309.
of the different materials approach each other after several cycles. [19] D.A. Evans, K.H. Jack, Acta Crystallogr. 10 (1957) 833.
For example, after cycling the reaction, the performance of non- [20] D. Isheim, D.N. Seidmann, Metal. Mater. Trans. A 33A (2002) 2317.
ball-milled and ball-milled samples turn out to be very similar. [21] P. Ettmayer, Monatsh. Chem. 101 (1970) 127.
[22] N.V. Troitskaya, Z.G. Pinsker, Kristallografiya 4 (1959) 33.
The molybdenum nitride system is an example that the cata- [23] C.L. Bull, P.F. McMillan, E. Soignard, K. Leineweber, J. Solid State Chem. 177
lytic properties of a material are governed by various factors. This (2004) 1488.
can be structure changes of the crystalline phases, changes in do- [24] E. Zhao, J. Wang, Z. Wu, Phys. Status. Solidi B 247 (2010) 1207.
[25] A.Y. Ganin, L. Kienle, G.V. Vajenine, J. Solid State Chem. 179 (2006) 2339.
main sizes, and with this of specific surface areas, as well as varia- [26] S. Adams, K.H. Ehses, J. Spilker, Acta Crystallogr. B49 (1993) 958.
tions in defect concentrations. [27] B. Braida, S. Adams, E. Canadell, Chem. Mater. 17 (2005) 5957.
[28] J.I. Langford, D. Louër, P. Scardi, J. Appl. Cryst. 33 (2000) 964.
Acknowledgments

We thank Hans Bongard, Silvia Palm and Bernd Spliethoff (MPI

Mülheim) for electron microscopy studies. Matteo Leoni (Univer-