Sol-gel synthesis and characterisation of nanoscopic FeF3-MgF2 heterogeneous catalysts with bi-acidic properties

Article abstract of DOI:10.1002/cctc.201300115

Nanoscopic metal fluorides with surface hydroxy groups are of broad interest in heterogeneous catalysis. With both Lewis and Bronsted acid sites on the surface, these catalysts can be applied to a wide range of reactions. Having previously synthesised AlF₃- and MgF₂-based catalysts, we report a new transition metal fluoride with bi-acidity. A pre-dehydration procedure was developed to introduce hydroxy groups to a Lewis acid, FeF₃. Subsequently, ternary nanoscopic FeF₃-MgF₂ with enhanced porosity was prepared through a one-step fluorination. The interaction between MgF₂ and FeF₃ was elucidated. Surface characterisation revealed a remarkable increase in the surface area of FeF₃-MgF₂ compared with FeF₃. More importantly, medium-strong Lewis and Bronsted acid sites were detected on the FeF₃-MgF₂ surface. In line with its bi-acidity, FeF₃-MgF₂ was highly active in the model reaction, the isomerisation of citronellal to isopulegol. Finally, we discuss how the porosity and surface acidity jointly determine the activity of FeF₃-MgF₂. Our study demonstrates the feasibility of ternary FeF₃-MgF₂ and opens new possibilities to synthesise bi-acidic fluoride catalysts.

Full text of DOI:10.1002/cctc.201300115

CHEMCATCHEM
FULL PAPERS
DOI: 10.1002/cctc.201300115
Sol–Gel Synthesis and Characterisation of Nanoscopic
FeF3-MgF2 Heterogeneous Catalysts with Bi-Acidic
Properties
[a]
[b]
[b]
[a]
´
Ying Guo, Piotr Gaczynski, Klaus-Dieter Becker, and Erhard Kemnitz*
Nanoscopic metal fluorides with surface hydroxy groups are of
broad interest in heterogeneous catalysis. With both Lewis and
Brønsted acid sites on the surface, these catalysts can be ap-
plied to a wide range of reactions. Having previously synthes-
ised AlF3- and MgF2-based catalysts, we report a new transition
metal fluoride with bi-acidity. A pre-dehydration procedure
was developed to introduce hydroxy groups to a Lewis acid,
FeF3. Subsequently, ternary nanoscopic FeF3-MgF2 with en-
hanced porosity was prepared through a one-step fluorination.
The interaction between MgF2 and FeF3 was elucidated. Sur-
face characterisation revealed a remarkable increase in the sur-
face area of FeF3-MgF2 compared with FeF3. More importantly,
medium–strong Lewis and Brønsted acid sites were detected
on the FeF3-MgF2 surface. In line with its bi-acidity, FeF3-MgF2
was highly active in the model reaction, the isomerisation of
citronellal to isopulegol. Finally, we discuss how the porosity
and surface acidity jointly determine the activity of FeF3-MgF2.
Our study demonstrates the feasibility of ternary FeF3-MgF2
and opens new possibilities to synthesise bi-acidic fluoride
catalysts.
Introduction
Nanoscopic metal fluorides (nano-MFn) have become an impor-
tant class of heterogeneous catalysts because of their unique
properties.[1,2] Benefiting from the most electro-negative ele-
ment, fluorine, nano-MFn usually exhibit pronounced surface
acidity. Other common features of nano-MFn include low long-
range order, high surface areas, and well developed porosi-
ty.[2,3] These features make nano-MFn promising materials in
heterogeneous catalysis. Indeed, nano-MFn have shown high
activity in various reactions: AlF3 can act as a catalyst with ex-
tremely strong Lewis acidity in CCl2F2 dismutation and C3Br2F6
isomerisation,[3] and alkaline earth metal fluorides (MgF2, CaF2,
SrF2, and BaF2) are highly active catalysts with different chemo-
selectivity in the dehydrohalogenation of chlorofluoro-
butanes.[1]
Moreover, the fluorolytic sol–gel route is very flexible with tun-
able synthesis parameters.[4,5] The flexibility allows us to design
and to synthesise different nano-MFn derivatives, for example,
oxofluorides,[6] partially hydroxylated fluorides[5] and doped flu-
orides,[7] which thus expands the applications of nano-MFn in
heterogeneous catalysis.
Among different nano-MFn derivatives, partially hydroxylated
metal fluorides (also known as hydroxy fluorides) play an
active role in heterogeneous catalysis. Partially hydroxylated
nano-MgF2 and AlF3 are highly active in acid-catalysed reac-
tions such as the Friedel–Crafts alkylation (synthesis of vita-
min E and vitamin K1)[8] and the carbonyl–ene reaction (isomer-
isation of citronellal to isopulegol).[9] Further doping with
noble metals leads to multi-functional hydroxylated metal fluo-
rides,[10,11] which are able to catalyse the one-pot synthesis of
citronellal or citral to menthol. We noticed that established cat-
alysts mostly focused on main group elements such as Al and
Mg. Consequently, we wondered whether it is possible to de-
velop partially hydroxylated transition metal fluoride catalysts
with comparable or even better performances. In search of an
appropriate transition metal, we proposed three principles:
1) the corresponding fluoride should be a Lewis acid, 2) hy-
droxy groups should be introduced without changing the sta-
bility of the fluoride and 3) low-cost precursors would be pre-
ferred. Iron was therefore selected as a favourable candidate.
In previous studies, iron fluoride has often been used as
a dopant and only its Lewis acidity was addressed.[12,13] Our
target of synthesising a hydroxylated nano-FeF3 is an impor-
tant milestone in the exploration of the potential of catalysts
based on iron fluoride. Moreover, high surface area and good
porosity are essential because heterogeneous catalysis is a sur-
face phenomenon. Thus we also intended to develop a high-
These examples of nano-MFn have all been prepared by the
fluorolytic sol–gel route,[3] which has become one of the most
useful methods to synthesise nano-MFn. It is an analogue to
the classic hydrolytic sol–gel route except that HF is used in-
stead of water. In the fluorolytic sol–gel route, the key step is
a fluorolysis reaction that happens usually at room tempera-
ture. The crystallisation of metal fluorides is therefore avoided,
and the nanoscopic features of the products are preserved.
[a] Y. Guo, Prof. Dr. E. Kemnitz
Institut fꢀr Chemie
Humboldt-Universitꢁt zu Berlin
Brook-Taylor-Straße 2, 12489, Berlin (Germany)
Fax: (+49)30-20937468
E-mail: erhard.kemnitz@chemie.hu-berlin.de
´
[b] Dr. P. Gaczynski, Prof. Dr. K.-D. Becker
Institut fꢀr Physikalische und Theoretische Chemie
Technische Universitꢁt Braunschweig
Hans-Sommer-Straße 10, 38106, Braunschweig (Germany)
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2223

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
surface-area nano-FeF3 material. This was achieved through
a one-step fluorination of mixed Fe and Mg precursors. Subse-
quently, we planned a panel of analytical measurements to
characterise the ternary FeF3-MgF2 catalysts. For brevity, FeF3
and FeF3-MgF2 refer to nanoscopic materials if no extra explan-
ation is added. The sol particle size distribution was deter-
mined by dynamic light scattering (DLS). The dried xerogel
products were characterised by powder XRD and Mçssbauer
spectroscopy. Importantly, the surface properties of the FeF3-
MgF2 samples were characterised by various methods. The sur-
face area and porosity were determined by N2 adsorption–de-
sorption measurements, and the surface acidity was studied by
NH3 temperature-programmed desorption (NH3-TPD) and
in situ FTIR spectroscopy. The in situ FTIR spectroscopy was de-
signed to determine both Lewis and Brønsted acid sites with
NH3 and CD3CN as probe molecules. The potential of FeF3-
MgF2 as a heterogeneous catalyst was evaluated by establish-
ing the isomerisation of citronellal as a model reaction. Despite
its importance in the fine chemical industry,[14,15] this reaction
has the advantage that both Lewis and Brønsted acid sites are
involved in the heterogeneous process.[16] Thus it can provide
strong evidence for the bi-acidity of FeF3-MgF2.
MXnþn HF ! MFnþn HX
FeðNO3Þ3 ꢀ 9 H2Oþ3 HF ! FeF3þ3 HNO3þ9 H2O
ð1Þ
ð2Þ
As we were interested in a catalyst with both Lewis and
Brønsted acidity, hydroxy groups that act as potential Brønsted
acid sites had to be introduced into FeF3. Depending on the
precursor, various methods are available for this purpose. For
metal alkoxides that were used in the synthesis of partially hy-
droxylated AlF3 or MgF2, an aqueous HF solution was sufficient
to initiate fluorolysis and hydrolysis in a single step, and thus
gave rise to M(OH)xFnꢁx xerogels. In contrast, for Fe(NO3)3·9H2O,
this route would have led to the formation of iron fluoride
monohydrate and hence is not an option. Pre-dehydration of
the Fe precursor, however, leads to the formation of iron hy-
droxide nitride. Hence, hydroxy groups could be introduced
before the fluorination step. The dehydration pathway of Fe-
(NO3)3·9H2O according to Wieczorek-Ciurowa and Kozak is
shown in Figure 2A.[17] Our investigations suggested an opti-
Results and Discussion
Nanoscopic FeF3
The fluorolytic sol–gel synthesis follows the general reaction
path described by Equation (1). Organic and inorganic com-
pounds can be used as the precursor, MXn. To synthesise FeF3,
we selected Fe(NO3)3·9H2O as the precursor because of its low
cost and high reactivity. The corresponding reaction path is de-
scribed by Equation (2). Through this route, we successfully
prepared a colourless, transparent FeF3 sol (Figure 1A). Drying
the sol resulted in the formation of an FeF3 xerogel. The broad
reflections in the powder XRD pattern of the dried FeF3 xero-
gel (Figure 1B) indicate its low degree of crystallisation, which
is a common feature of many nanoscopic metal fluorides.[1]
Figure 2. A) Schematic diagram of the thermal treatment of Fe(NO3)3·9H2O.
B) IR spectrum of FeF3 after activation at 2008C under vacuum.
mal pre-dehydration temperature of 658C. The partially dehy-
drated iron hydroxide nitrate was then used as the precursor
for the sol–gel synthesis.
To prove the existence of hydroxy groups in FeF3, we mea-
sured the IR spectrum of the pre-dehydrated FeF3 after remov-
ing surface impurities and weakly bonded water species by
heating the sample at 2008C under high vacuum (Figure 2B).
The weak vibration band at 1640 cmꢁ1 probably results from
a small quantity of strongly bonded water species. However,
its low intensity cannot explain the strong ꢁOH vibration
Figure 1. A) Images of iron nitrate nonahydrate precursor solution (left) and
FeF3 sol (right). B) Powder XRD pattern of an FeF3 xerogel (ꢁ refers to an
FeF3·H2O phase in PDF 26-783).
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2224

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
bands at 3597, 3333 and 3167 cmꢁ1. Similar features were seen
in the IR spectrum of ZnF2, and these ꢁOH vibration bands
have been taken as the evidence for the formation of ZnꢁOH
bonds.[17,18] Similarly, the IR spectrum shown in Figure 2B con-
firms that FeꢁOH bonds exist in FeF3.
and thereby exhibits a Zeeman-split six-line spectrum. In its
paramagnetic state, T>363 K, r-FeF3 exhibits a single-line spec-
trum, which indicates zero or negligibly small quadrupole split-
ting (Table 1).[19] In contrast, nanoscopic FeF3, also termed
amorphous FeF3 (a-FeF3) in the literature, exhibits a quadrupole
Ternary FeF3-MgF2
Table 1. Parameters from the 57Fe Mçssbauer spectra of FeF3, FM-2 and
FM-9 and literature values for a-FeF3 and r-FeF3.
So far, we proved that the pre-dehydration treatment of the Fe
precursor successfully introduced hydroxy groups into FeF3.
However, the obtained FeF3 showed a relatively low surface
area of 40–70 m2 gꢁ1. As we intended to use FeF3 as a heteroge-
neous catalyst, a high surface area is preferred because all the
active sites are on the surface. To prepare a high-surface-area
material, nanoscopic MgF2 was used as matrix. The one-step
fluorination of the mixed Fe and Mg precursors gave rise to
ternary FeF3-MgF2 materials (denoted as FM-n, in which n
refers to the Mg/Fe molar ratio). The synthetic route is shown
in Figure 3A. Nanoscopic MgF2 was also prepared as a -
reference.
1=2
Sample
IS
< QS >
< QS2
>
s
Tt
B
[mmsꢁ1
]
[mmsꢁ1
]
[mmsꢁ1
]
[K][a]
[T][b]
FeF3
FM-2
FM-9
a-FeF3
r-FeF3
0.45
0.43
0.43
0.58
0.73
0.96
0.50
0.56
0.60
–
0.30
0.47
0.75
–
–
–
–
–
–
–
[20]
0.45–0.46 0.55–0.58
0.49
30–40 53–56
363 61.8
[19]
0
0
0
[a] Magnetic transition temperature. [b] Local magnetic field at 4.2 K.
doublet with broad lines at room temperature,[20] which indi-
cates a lack of magnetic interactions and a lower-than-cubic
symmetry at the Fe site. The room temperature spectra of the
sol–gel-prepared FeF3 as well as FM-2 and FM-9 also exhibit
a broad doublet (Figure 4), and the parameters for FeF3 are in
excellent agreement with the literature values for a-FeF3
(Table 1). This agreement and the lack of magnetic hyperfine
lines confirm the nanoscopic nature of the sol–gel-prepared
FeF3. The broadening of the doublet indicates the existence of
a large number of slightly different Fe surroundings. The slight
asymmetry of the spectrum of FeF3 may be because of a weak
correlation between the isomer shift (IS) and quadrupole split-
ting (QS). The latter can be induced by local variation of chemi-
cal bonding as a result of the introduced OH groups, which co-
ordinate to the Fe atoms and thus result in different partially
fluorinated FeF6ꢁxOx units. Similar structural units have been
confirmed by 19F magic angle
Images of one representative sample, FM-2, before and after
fluorination are shown in Figure 3B. Similar to the synthesis of
FeF3, a clear FeF3-MgF2 sol was obtained. To determine the par-
ticle size distribution in the sol we used DLS. The hydrodynam-
ic diameter of the FM-2 sol particles follows an approximate
log-normal mono-distribution with the mean at approximately
6 nm (Figure 3C). This confirms the nanoscopic nature of
ternary FeF3-MgF2.
Structural characterisation of FeF3-MgF2
Mçssbauer spectroscopy is a powerful method to study the
local structure and magnetic properties of bulk and nanoscop-
ic FeF3. Bulk crystalline FeF3 of rhombohedral structure (r-FeF3)
possesses a long-range magnetic order at room temperature
spinning (MAS) NMR spectrosco-
py in partially hydroxylated
MgF2.[21] In addition, residual im-
purities, for example, strongly
bonded H2O (see Figure 2B),
may play a role.
The fitting parameters of the
57Fe Mçssbauer spectra of the
FeF3-MgF2 samples are summar-
ised in Table 1. The magnitude
of the IS indicates that Fe atoms
in all the samples are in the +3
formal oxidation state coordinat-
ed octahedrally by fluorine
atoms.[22] The non-zero QS and
the broadened line shape reflect
the random distortion of the oc-
tahedral coordination and struc-
tural disorder. The increasing
Figure 3. A) Schematic diagram of the one-step synthesis of ternary FeF3-MgF2. B) Images of the mixed suspension
of Fe-Mg precursors (left; Mg/Fe molar ratio=2:1) and the corresponding FM-2 sol (right). C) FM-2 sol particle size
distributions from DLS measurements.
magnitude of the mean value of
QS (< QS >) with increasing
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2225

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
showed properties similar to pure FeF3. These results can be
interpreted that the disturbance of MgF2 in FeF3 seems to be
introduced through the interaction between the sol particles
during the one-step sol–gel synthesis. This conclusion appears
to be in agreement with the deductions from the Mçssbauer
spectra.
Morphology of FeF3-MgF2
Ternary FeF3-MgF2 was further investigated by using TEM.
Nano-crystallites with clear crystalline planes can be distin-
guished in the high-resolution TEM (HRTEM) image of FM-2
(Figure 5). Assuming that FeF3-MgF2 consists of spherical parti-
cles, the radii of these crystallites are mostly below 2 nm. Thus
the HRTEM image provides direct proof of the nanoscopic
nature of FeF3-MgF2.
Figure 4. 57Fe Mçssbauer spectra of a) FeF3, b) FM-2 and c) FM-9.
Figure 5. HRTEM image of FM-2 (scale=5 nm).
MgF2 indicates the increase in the asymmetry around the Fe
atoms, whereas the increase in the root mean square
À
Á
Surface area and porosity of FeF3-MgF2
< QS2 >1=2 and standard deviation (s) reflects the increase
between the minimal and maximal asymmetry around the Fe
atoms. In conclusion, an increasing MgF2 content leads to in-
creasing disturbances of the Fe coordination polyhedra. This
indicates an at least partial mixing of Fe and Mg ions at the
atomic level.
High-surface-area nanoscopic MgF2 (surface area=240–
260 m2 gꢁ1) was used as matrix to tune the surface area. To de-
termine the surface area and porosity of the FeF3-MgF2 materi-
als, N2 adsorption–desorption measurements were performed.
The isotherms shown in Figure 6A provide an overview of the
porous features of the FM-n samples. With an increasing con-
tent of MgF2, the isotherm changes gradually from an FeF3-
type (Figure 6Aa and b) to a MgF2-type (Figure 6Ac and d).
The FeF3-type isotherms show the main features of a type H3
hysteresis loop, which corresponds to a material that is either
non-porous or exhibits slit-shaped pores of non-uniform
sizes.[23] The MgF2-type isotherms follow a type H2 hysteresis
loop. Thus, the FM-n samples with a high Mg content (nꢂ2)
possess nearly cylindrical pores of non-uniform sizes, which are
probably ink-bottle shaped.[23]
The Mçssbauer spectra of the FM-n samples reflect the
structural interference between MgF2 and FeF3. However, the
origin of this interaction remains unclear. Thus two supplemen-
tary experiments were designed for further investigation. An-
other two FeF3-MgF2 samples were prepared with a Mg/Fe
ratio of 2:1. One sample was synthesised by mechano-milling
FeF3 and MgF2 xerogels, and the other was prepared by drying
the mixed FeF3 and MgF2 sols that were fluorinated separately.
These two samples were characterised by N2 adsorption–de-
sorption measurements and their catalytic activity was tested
with the model reaction (citronellal!isopulegol). The sample
from the mixed sols was almost identical to FM-2 in surface
area and catalytic activity, whereas the mechano-milled sample
The BET surface area and average pore diameter, both calcu-
lated from the isotherms,[24a] were plotted versus the Mg/Fe
molar ratio, respectively (Figure 6B and C). The surface area in-
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2226

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Moreover, not only the surface area but also the pore size
was tuned in ternary FeF3-MgF2. The average pore diameters
of the FM-n samples show a logarithmic relationship to the
Mg/Fe ratio (Figure 6C). The pore diameter of the FM-n (nꢂ2)
samples is in the range of 2–3 nm, which is at the border be-
tween micro- and meso-porosity.
Surface acidity of FeF3-MgF2
A study on the surface acidity of the FeF3-MgF2 system is vitally
important because it indicates its potential as a solid-acid cata-
lyst. Therefore, NH3 and CD3CN were employed as probe mole-
cules. TPD and in situ chemisorption–IR spectroscopy were
used as complementary methods to investigate the interaction
between the probe molecules and the sample surface.
To determine the strength of the surface acid sites, TPD
measurements were performed with NH3 as a probe molecule.
The NH3-TPD profile of FM-2 is shown in Figure 7. Reference
Figure 7. NH3-TPD profiles of a) FeF3, b) FM-2 and c) MgF2.
profiles of FeF3 and MgF2 are also presented. The profile of
FeF3 has its maximum at approximately 3008C, which indicates
medium–strong acid sites.[1] In the profile of MgF2, however,
the maximum at 1608C shows the existence of only weak acid
sites. Interestingly, with two maxima at 160 and 2808C, which
evidence the presence of both weak and medium–strong acid
sites, FM-2 shows hybridised features of the desorption profiles
of FeF3 and MgF2.
Figure 6. A) N2 adsorption–desorption isotherms of a) FeF3, b) FM-1, c) FM-2
and d) MgF2. B) Surface area of FM-n versus the Mg/Fe molar ratio. C) Pore
diameter of FM-n versus the Mg/Fe molar ratio.
The TPD method is useful to investigate the strength of acid
sites. However, it is not specific for Lewis or Brønsted acid
sites. Therefore, in situ chemisorption–IR spectra were recorded
with NH3 as a probe molecule. NH3 can interact with the acid
sites of a solid surface in different ways. We focused on three
of them: 1) protonation with surface hydroxy groups (Brønsted
acid sites), 2) coordination to electron-deficient metal atoms
(Lewis acid sites) and 3) hydrogen bonding through the N
atom of NH3 (Brønsted acid sites).[26] The differential spectra of
one representative sample, FM-2, before and after NH3 adsorp-
creases drastically with Mg content until it approaches the
maximum at a Mg/Fe ratio of 2:1. Higher Mg contents do not
result in a further increase, and the FM-2 sample exhibits the
highest surface area of 380–500 m2 gꢁ1. Interestingly, the sur-
face areas of most FM-n samples are higher than that of the
pure MgF2. This is probably caused by structural distortions of
MgF2 in the presence of FeF3. Additionally, HNO3 formed
during the fluorination [Eq. (2)] may also contribute to the
larger surface area, probably because of the stabilisation effect
of the acid on small sol particles, which prevents agglomera-
tion. Similar effects of acid additives have been discovered in
the synthesis of AlF3 and MgF2.[25]
tion are given in Figure 8A. The intense band at 1436 cmꢁ1 is
+
assigned to NH4
species stabilised by adjacent OH
groups,[24b,27] which confirms the existence of Brønsted acid
sites on the surface of FM-2. Additionally, this band shifts
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2227

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Figure 9. Differential IR spectra of CD3CN adsorbed onto FM-2 after activa-
tion at 808C.
the redshift of the band at 1436 cmꢁ1 in the NH3 chemisorp-
tion–IR spectra (Figure 8A). The band at 2116 cmꢁ1 corre-
sponds to the symmetric CD3 vibration. The asymmetric modes
of the CD3 vibration are too weak to be detected. The band at
2260 cmꢁ1 is largely reduced after evacuation, which suggests
that it is caused by a physisorbed species. The band between
2317 and 2307 cmꢁ1 is characteristic of coordinated CD3CN
species on medium–strong and strong Lewis acid sites and its
frequency is shifted by approximately 13 cmꢁ1 to lower wave-
numbers compared with aluminium hydroxy fluoride.[31] This
implies that the Lewis acidity of FM-n is weaker than that of
aluminium hydroxy fluoride, which is in line with the order of
electronegativity of Fe (c=1.83) and Al (c=1.61).[32]
Figure 8. A) Differential IR spectra of NH3 adsorbed onto FM-2 after activa-
tion at 808C. B) Schematic diagram of the interaction between NH3 and the
solid surface (I: NH3 coordinated to a Lewis acid site; II: NH3 coordinated to
a Lewis acid site additionally stabilised by H-bonding; III: H-bonded NH3
species).
In conclusion, Lewis and Brønsted acid sites co-exist on the
surface of FM-n. The chemisorption–IR spectra of FeF3 used as
a reference are similar to those of the FM-n samples, whereas
the spectra of MgF2 with different probe molecules give only
evidence of some weak Lewis acid sites. Thus FeF3 dominates
the surface acidic properties of FM-n. Consequently, it is possi-
ble to tune the surface acidity of FM-n by varying the Mg/Fe
ratio.
slightly towards lower wavenumbers (Dd=14 cmꢁ1) compared
+
with the vibration frequency of NH4 on alumina surfaces (d=
1450 cmꢁ1).[28] Thus the Brønsted acidity of FeF3-MgF2 is proba-
bly weaker than that of alumina, which is commonly recog-
nised as a strong solid acid.
The assignments of the other three vibration bands are not
straightforward. The bands at 1191 and 1243 cmꢁ1 are in the
range (1280–1150 cmꢁ1 [26]
of vibration bands of coordinated
)
Catalytic potential of FeF3-MgF2
NH3 species at Lewis acid sites (Figure 8BI). The band at
1303 cmꢁ1 can only be assigned to coordinated NH3 species
with strong H bonding (Figure 8BII).[26,29] However, the refer-
ence is based on a study of [M(NH3)m]·halogenn coordination
compounds, whereas our results relate to solid surfaces. H-
bonded species (Figure 8BIII) were not detected on the sur-
face of FM-2. One possible explanation for this may be that
the coverage of hydroxy groups on the FM-2 surface is not
high enough for H bonding.[26]
To evaluate the catalytic potential of FM-n, the isomerisation of
citronellal to isopulegol was used as a model reaction. This re-
action is an intra-molecular carbonyl–ene reaction that requires
the participation of both Lewis and Brønsted acid sites.[33] It is
also a diastereoselective reaction with four isomer products,
which include isopulegol (molar ratio 70–75%), neo-isopulegol
(ꢃ20%), iso-isopulegol and neo-iso-isopulegol. This molar dis-
tribution of the four isomers is found for almost all acid-cata-
lysed reactions. The thermodynamic stability of the six-ring
transition state explains this distribution.[33] In our catalytic
tests, the four diastereoisomers follow the same distribution
with different catalysts. Thus, we concentrate our discussion
just on the conversion of citronellal and the chemical selectivi-
ty to isopulegol (including all diastereoisomers).
CD3CN is another widely used probe molecule because it
can interact with both Lewis and Brønsted acid sites.[24b,30] In
particular, the well-known trio bands are characteristic of
strong Brønsted acid sites. This refers to the simultaneous oc-
currence of three bands at approximately 2890, 2400 and
1700 cmꢁ1.[24b,30] In the differential spectra of FM-2 (Figure 9),
the trio bands are missing, which indicates that the Brønsted
acid sites on the FM-2 surface are not strong enough to form
an H-bonded complex with CD3CN. This is in agreement with
The catalytic performances of FeF3 and FM-n are presented
in Figure 10. As MgF2 is completely inactive in this reaction,
FeF3 must be the active component. The first series of tests
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2228

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Figure 10. A) Conversion and selectivity versus the Mg/Fe molar ratio with catalysts (20 mg). B) Conversion and selectivity to the product versus the Mg/Fe
molar ratio normalised to 10 mg FeF3. C) Conversion of citronellal versus the surface area of the catalysts. D) Selectivity to isopulegols versus the pore diame-
ter of the catalysts.
used equal amounts of catalysts with varying Fe contents. The
maximum conversion (as well as selectivity) was observed at
a Mg/Fe molar ratio of approximately 1.5 (Figure 10A). As the
FeF3 content varies in the samples, the results were normalised
to the same amount of FeF3 (10 mg) in all cases (Figure 10B).
Both the conversion and selectivity increase drastically for Mg/
Fe ratios between 0 and 2. For higher ratios, the conversion
and selectivity remain constant. As the amount of FeF3, the
active species, is constant, the higher conversion and selectivi-
ty are because of changes in other characteristics, for example,
the surface area and porosity.
gol formation because they suppress large by-products such
as dimers or trimers of isopulegol. Similar results were also
found for FM-n. The plot shown in Figure 10D proves that
high selectivity for isopulegols is only achieved in the low
nano-metre range of pores below 5 nm. Thus small meso-
pores or large micro-pores are favourable for this reaction.
Leaching and recycling tests were performed to examine the
robustness of FM-n. The representative FM-2 sample survived
the leaching test, which suggests that the reaction undergoes
a heterogeneous process. In the recycling test, FM-2 was re-
generated following this process: the used catalysts were col-
lected by filtration, washed with acetone and then dried at
808C under an Ar flow for 12 h or under vacuum for 6 h. The
results are summarised in Table 2. The recovery rate was calcu-
lated by comparing the yields of isopulegol. In the first test,
three recycling runs were performed, and the reaction pro-
ceeded for 3 h in each case. The recovery rate remained
almost constant between 77–79%. The decline in catalytic ac-
tivity is probably because of the coordination of some organic
species (e.g., reactant, product or by-products) to the surface
acid sites as elemental analysis showed a significant increase in
Generally, high-surface-area materials are preferred in heter-
ogeneous catalysis. Therefore, we plotted the conversion of cit-
ronellal versus the surface area of FM-n (Figure 10C). Indeed,
the conversion rises rapidly in the surface area region below
300 m2 gꢁ1. In the high surface area region, however, the con-
version approaches nearly 100%, hence no further change can
be expected.
In previous studies, Chuah et al.[16] discussed the influence of
the catalyst pore size on the selectivity to isopulegol. It has
been concluded that small pores preferentially cause isopule-
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2229

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
the catalytic activity, we studied this by employing comple-
mentary surface characterisation with different probe mole-
cules. Weak and medium–strong acid sites were found on the
surface of FeF3-MgF2. More importantly, both Lewis and Brønst-
ed acid sites were identified. These results proved that the pre-
dehydration treatment of the iron(III) nitrate precursor generat-
ed hydroxy groups that are preserved after the fluorolytic sol–
gel synthesis, which explains the high activity of FeF3-MgF2 in
the model reaction.
Table 2. Summarised results of recycling tests.
Cycle Conv. Select. Recovery rate Carbon
[%] [%]
[%][a] content [wt%][b]
0
1
2
3
0
1
75 76 0.6
–
test 1
3 h reaction
63
50
44
98
86
71
69
61
73
79
79
77
77
–
2.5
3.6
4.0
0.6
2.7
test 2
6 h reaction
95
The FeF3-MgF2 catalysts exhibit great potential in heteroge-
neous catalysis. As a ternary fluoride, the combination of FeF3
and MgF2 creates an opportunity to tune the surface area, po-
rosity and surface acidity. We are convinced that this study on
the FeF3-MgF2 system evidences not only the potential of tran-
sition metal fluorides but also the possibility of targeted
design for new fluoride-based catalysts.
[a] Recovery rate=100%ꢁyield (cyclen)/yield (cyclenꢁ1);
gol)=conversionꢁselectivity. [b] Determined by elemental analysis with
an error of ꢄ0.3%.
yield (isopule-
the C content of the sample after each reaction. Only one recy-
cling run was performed in the second test. The reaction time,
however, was extended to 6 h as in the usual catalytic tests.
The recovery rate was approximately 95% after the first run.
Taking into account the results from the first recycling test, it
can be estimated that approximately 85% of the original yield
to isopulegol is to be expected after three recycling runs.
The catalytic results fully confirmed our expectations of
FeF3-MgF2. As robust bi-acidic solids, the FeF3-MgF2 catalysts
were highly active in the citronellal isomerisation reaction. The
surface Lewis and Brønsted acid sites jointly participated in the
reaction. Accordingly, a possible reaction mechanism is pro-
posed in Figure 11.
Experimental Section
Preparation of FeF3
Caution: HF is hazardous; protection required.
Fe(NO3)3·9H2O (Sigma–Aldrich, 99%, 20 mmol, 8.08 g) was dis-
solved in methanol (100 mL) and fluorinated with methanolic HF
solution (19.3 molLꢁ1, 3.1 mL). The obtained colourless, transparent
FeF3 sol was aged for 14–16 h before it was dried in vacuo. The
product, a powder-like xerogel, was calcined at 1008C under re-
duced pressure for 2 h.
The pre-dehydration treatment of Fe(NO3)3·9H2O was performed in
vacuo at 658C for 2 h. The obtained substance was then used as
a precursor for sol–gel synthesis following the same procedure as
described above.
Conclusions
The synthetic routes to both nanoscopic FeF3 and ternary FeF3-
MgF2 catalysts were explored. With the application of a pre-de-
hydration treatment to the iron(III) nitrate precursor, we suc-
cessfully introduced hydroxy groups, which are potential
Brønsted acid sites, onto the surface of FeF3. By dispersing
FeF3 in the inactive MgF2 matrix, we obtained bi-acidic cata-
lysts with high surface areas. The porosity as well as the acidity
of ternary FeF3-MgF2 can be tuned easily by changing the Mg/
Fe ratio. The FeF3-MgF2 catalysts were highly active in the iso-
merisation reaction of citronellal.
Preparation of FeF3-MgF2
Mg(CH3COO)2·4H2O (Sigma–Aldrich, >99%) was dehydrated at
2108C overnight in an oven. It was then added to the pre-dehy-
drated iron nitrate precursor at different Mg/Fe molar ratios. The
mixed precursors were suspended in methanol (total metal con-
centration=0.2 molLꢁ1
)
and fluorinated with stoichiometric
amounts of HF (in methanolic solution) in one step. The transpar-
ent FeF3-MgF2 sol was aged, dried and calcined under the same
conditions as in the synthesis of FeF3. The obtained xerogel prod-
uct was labelled as FM-n. F and M refer to FeF3 and MgF2 and n in-
dicates the Mg/Fe molar ratio.
Characterisation of the physico-chemical properties of the
ternary FeF3-MgF2 catalysts helped us to understand their cata-
lytic capabilities. Structural studies of the ternary FeF3-MgF2
samples indicated that the local structure around the Fe sites
was strongly disturbed as a result of their incorporation into
the MgF2 matrix. As the surface acidity has a direct impact on
MgF2 was also prepared as a reference. Dehydrated Mg(CH3COO)2
(20 mmol, 2.85 g) was suspended in methanol (100 mL). Methanol-
ic HF solution was added in stoichiometric amounts to the suspen-
sion. The drying procedure fol-
lowed the same steps as in the
synthesis of FeF3 and FM-n.
The mechano-milled sample was
prepared by using
a planetary
micro-mill Pulverisette 7 premium
line (Fritsch). The FeF3 (2 mmol,
0.23 g) and MgF2 (4 mmol, 0.25 g)
xerogels were mixed and milled
for 2ꢁ60 min at 600 rpm with
a 15 min break in between.
Figure 11. Proposed isomerisation mechanism of citronellal to isopulegol at the active sites of FM-n catalysts.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2230

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
Two recycling tests were performed under the same reaction con-
ditions with the FM-2 catalyst. In the first test, the used catalyst
was collected by filtration, washed 4 times with acetone (10 min
each) and then dried at 808C under an Ar flow (20 mLminꢁ1). The
first test included three recycling runs, and each run lasted for 3 h.
In the second test, the catalyst was separated and washed with
acetone 3 times (5 min, 10 min and 12 h) and then dried at 808C in
vacuo for 6 h. One recycling run of 6 h was performed in the
second test.
Characterisation
DLS measurements were performed by using a Malvern Zetasizer
Nano ZS instrument in NIBS mode at 1738. The methanolic FM-2
sol (concentration of metal=0.2 molLꢁ1) was filtered through
a Nylon filter (ø=450 nm).
Powder XRD patterns were obtained by using an XRD-7 Seifert-
FPM diffractometer using a CuKa beam.
Mçssbauer spectra of FeF3, FM-2 and FM-9 at RT were recorded
with a standard transmission Mçssbauer spectrometer (Halder) in
the sinusoidal drive mode. The velocity scale was calibrated with
an a-Fe absorber and the IS value is given relative to a-Fe.
Acknowledgements
TEM images were recorded with a Joel JEM-2200FS Transmission
Electron Microscope using a 200 kV field emission gun (FEG). To
prepare the sample, FM-2 was re-suspended in methanol (2–
3 mgmLꢁ1). A Lacey carbon film (Cu-networks, 300 mesh) was
dipped in the sample suspension and dried in air.
We thank Holm Kirmse for recording the TEM images, Larissa
Schmidt for DLS measurements and discussion, Maik Dreger for
preparing the mechano-milled sample and Sigrid Bꢁßler for per-
forming the TPD measurement. We also thank Gudrun Scholz for
suggestions and discussion in structural characterisation and
Katharina Teinz for discussion and help in catalysis tests. Y.G. was
supported by a scholarship from the Chinese Scholarship Council
(CSC).
N2 adsorption–desorption measurements at 77 K were performed
by using a Micromeritics ASAP 2020 instrument. Samples were de-
gassed at 808C under high vacuum for 12 h.
NH3-TPD profiles were recorded by using a Perkin–Elmer FTIR-
System 2000 at 930 cmꢁ1. The pre-heated (1008C) samples were
loaded with NH3 at 608C. After purging with N2, the temperature
was raised at a rate of 108Cminꢁ1 up to 4508C while the amount
of desorbed NH3 was recorded.
Keywords: acidity
magnesium · sol–gel processes
·
heterogeneous catalysis
·
iron
·
For in situ FTIR spectroscopy, approximately 25–30 mg of the
sample was pressed at 1 ton into a self-supporting disc (2 cm2
area) for analysis. All samples were placed in a quartz cell equipped
with KBr windows. The cell was connected to a vacuum line and
a glass injection loop. A movable quartz sample holder permitted
the adjustment of the sample disc in the IR beam for spectra ac-
quisition and to place it into a furnace at the top of the cell for
thermal activation of the samples. After we activated the sample at
808C under vacuum, different probe molecules were added in
small doses. IR spectra were recorded by using a Thermo Scientific
Nicolet iS10 spectrometer. The IR spectrum of FeF3 after degassing
at 2008C under vacuum was recorded by using the same equip-
ment without the introduction of any probe molecules.
[1] K. Teinz, S. Wuttke, F. Bçrno, J. Eicher, E. Kemnitz, J. Catal. 2011, 282,
175–182.
[2] E. Kemnitz, S. Wuttke, S. M. Coman, Eur. J. Inorg. Chem. 2011, 4773–
4794.
[3] E. Kemnitz, U. Groß, S. Rꢂdiger, C. S. Shekar, Angew. Chem. 2003, 115,
4383–4386; Angew. Chem. Int. Ed. 2003, 42, 4251–4254.
[4] S. Wuttke, G. Scholz, S. Rꢂdiger, E. Kemnitz, J. Mater. Chem. 2007, 17,
4980–4988.
[5] S. Wuttke, S. M. Coman, J. Krçhnert, F. C. Jentoft, E. Kemnitz, Catal.
Today 2010, 152, 2–10.
[6] H. A. Prescott, Z.-J. Li, E. Kemnitz, J. Deutsch, H. Lieske, J. Mater. Chem.
2005, 15, 4616–4628.
[7] I. K. Murwani, K. Scheurell, E. Kemnitz, Catal. Commun. 2008, 10, 227–
231.
[8] N. Candu, S. Wuttke, E. Kemnitz, S. M. Coman, V. I. Parvulescu, Appl.
Catal. A 2011, 391, 169–174.
[9] S. M. Coman, P. Patil, S. Wuttke, E. Kemnitz, Chem. Commun. 2009, 460–
462.
Catalytic tests
[10] A. Negoi, S. Wuttke, E. Kemnitz, D. Macovei, V. I. Parvulescu, C. M. Teo-
dorescu, S. M. Coman, Angew. Chem. 2010, 122, 8311–8315; Angew.
Chem. Int. Ed. 2010, 49, 8134–8138.
[11] A. Negoi, K. Teinz, E. Kemnitz, S. Wuttke, V. Parvulescu, S. Coman, Top.
Catal. 2012, 55, 680–687.
[12] J. Krishna Murthy, U. Gross, S. Rꢂdiger, E. Kemnitz, Appl. Catal. A 2004,
278, 133–138.
[13] J. Krishna Murthy, U. Groß, S. Rꢂdiger, E. ꢃnveren, E. Kemnitz, J. Fluorine
Chem. 2004, 125, 937–949.
[14] J. C. Leffingwell, R. E. Shackelford, Cosmet. Perfum. 1974, 89, 69–89.
[15] Y. Nakatani, K. Kawashima, Synthesis 1978, 147–148.
[16] G. K. Chuah, S. H. Liu, S. Jaenicke, L. J. Harrison, J. Catal. 2001, 200, 352–
359.
[17] K. Wieczorek-Ciurowa, A. J. Kozak, J. Therm. Anal. Calorim. 1999, 58,
647–651.
[18] Y. Guo, S. Wuttke, A. Vimont, M. Daturi, J.-C. Lavalley, K. Teinz, E. Kem-
nitz, J. Mater. Chem. 2012, 22, 14587–14593.
[19] G. K. Wertheim, H. J. Guggenheim, D. N. E. Buchanan, Phys. Rev. 1968,
169, 465–470.
The citronellal isomerisation reaction was performed by using
a Thermomixer comfort (Eppendorf) equipped with glass tubes.
The samples were heated at 808C in vacuo for 2 h before use to
remove the surface impurities. Different amounts of fluoride sam-
ples were added to the reactant mixture of citronellal (0.2 mL,
ꢃ1 mmol; Acros, 93%), undecane (0.1 mL; Sigma–Aldrich, ꢂ99%)
as an internal standard for GC evaluation and toluene (2 mL; Acros,
for HPLC, 99.99%) as the solvent. The glass tubes were maintained
at 808C for 6 h. After the reaction, the catalysts were separated by
filtration, and the products were analysed by GC (Shimadzu GC-
2010 with an HP-1 column of 100 mꢁ0.25 mmꢁ0.5 mm, tempera-
ture programming: starting temperature=1208C, hold time=
50 min, rate=158Cminꢁ1, end temperature=2708C, hold time=
5 min).
The leaching test was executed by removing the catalyst from the
reaction mixture after 3 h from the beginning of the reaction. The
rest of the solution was kept under the reaction conditions for an-
other 3 h. At the end, the reaction mixtures before and after the
break were both analysed by using GC.
[20] a) J. M. Grenꢄche, Hyperfine Interact. 1999, 122, 9–19; b) M. Eibschꢂtz,
M. E. Lines, L. G. Van Ultert, H. J. Guggenheim, G. J. Zydzik, Phys. Rev. B
1984, 29, 3843–3851; c) M. Eibschꢂtz, M. E. Lines, L. G. Van Uitert, H. J.
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2231

CHEMCATCHEM
FULL PAPERS
www.chemcatchem.org
´
Guggenheim, G. J. Zydzik, Phys. Rev. B 1981, 24, 2343–2348; d) G. Ferey,
F. Varret, J. M. D. Coey, J. Phys. C 1979, 12, L531.
[28] K. Gꢅra-Marek, M. Derewinski, P. Sarv, J. Datka, Catal. Today 2005, 101,
131–138.
[21] S. Wuttke, S. M. Coman, G. Scholz, H. Kirmse, A. Vimont, M. Daturi,
S. L. M. Schroeder, E. Kemnitz, Chem. Eur. J. 2008, 14, 11488–11499.
[22] F. Menil, J. Phys. Chem. Solids 1985, 46, 763–789.
[23] G. Leofanti, M. Padovan, G. Tozzola, B. Venturelli, Catal. Today 1998, 41,
207–219.
[24] a) K. S. W. Sing, J. Rouquerol in Handbook of Heterogeneous Catalysis,
Vol. 2 (Eds.: G. Ertl, H. Knçzinger, J. Weitkamp), Wiley-VCH, Weinheim,
2008, pp. 429–431; b) H. Knçzinger in Handbook of Heterogeneous Cat-
alysis, Vol. 2 (Eds.: G. Ertl, H. Knçzinger, J. Weitkamp), Wiley-VCH, Wein-
heim, 2008, pp. 709–711.
[29] K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordination
Compounds, Vol. 2, Wiley, Hoboken, 2009, pp. 1–3.
[30] A. G. Pelmenschikov, R. A. van Santen, J. Janchen, E. Meijer, J. Phys.
Chem. 1993, 97, 11071–11074.
[31] K. Teinz, Edelmetallbeladene Aluminiumhydroxidfluoride als Neue Hetero-
gene Katalysatoren, Dissertation, Humboldt-Universitꢆt zu Berlin, Berlin,
2012, pp. 38–40.
[32] Handbook of Chemistry and Physics, 84th ed., Section 9 (Ed.: D. R. Lide)
CRC, Boca Raton, 2003–2004, pp. 74.
[33] W. Oppolzer, V. Snieckus, Angew. Chem. 1978, 90, 506–516; Angew.
Chem. Int. Ed. Engl. 1978, 17, 476–486.
[25] K. Teinz, unpublished results.
[26] A. A. Tsyganenko, D. V. Pozdnyakov, V. N. Filimonov, J. Mol. Struct. 1975,
29, 299–318.
Received: February 13, 2013
[27] E. H. Teunissen, R. A. Van Santen, A. P. J. Jansen, F. B. Van Duijneveldt, J.
Phys. Chem. 1993, 97, 203–210.
Published online on May 31, 2013
ꢀ 2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ChemCatChem 2013, 5, 2223 – 2232 2232