Tuning the surface properties of novel ternary iron(III) fluoride-based catalysts using the template effect of the matrix

Abstract

Sol–gel prepared ternary FeF₃–MgF₂ materials have become promising heterogeneous catalysts due to their porosity and surface Lewis/Brønsted acidity (bi-acidity). Despite the good catalytic performance, nanoscopic characterisations of this type of material are still missing and the key factors controlling the surface properties have not yet been identified, impeding both a better understanding and further development of ternary fluoride catalysts. In this study, we characterised the interaction between the bi-acidic component (FeF₃) and the matrix (MgF₂) on the nano-scale. For the first time, the formation pathway of FeF₃–MgF₂ was profiled and the template effect of MgF₂ during the synthesis process was discovered. Based on these new insights two novel materials, FeF₃–CaF₂ and FeF₃–SrF₂, were established, revealing that with decreasing the atomic numbers (from Sr to Mg), the ternary fluorides exhibited increasing surface acidity and surface area but decreasing pore size. These systematic changes gave rise to a panel of catalysts with tuneable surface and bulk properties either by changing the matrix alkaline earth metal fluoride or by adjusting their ratios to Fe or both. The template effect of the alkaline earth metal fluoride matrix was identified as the most probable key factor determining the surface properties and further influencing the catalytic performance in ternary fluoride based catalysts, and paves the way to targeted design of next-generation catalysts with tunable properties.

---

1. Introduction

Nanoscopic bi-acidic metal fluorides have been shown to play an active role in heterogeneous catalysis in recent years.^1–6 In this context, bi-acidic refers to the co-existence of Lewis and Brønsted acidity in the catalyst. Heterogeneous bi-acidic catalysts are easy to handle and environmentally friendly, and thus are frequently applied in industrial processes like alkylation, isomerisation, cracking, and esterification.^7,8 Commonly used heterogeneous bi-acidic catalysts are usually based on zeolites, silica, alumina, zirconia, and their combinations. Since metal fluorides are known to exhibit much more pronounced Lewis acidity compared with metal oxides,⁹ we wondered whether it is possible to prepare metal fluorides with bi-acidic properties. Indeed, partially hydroxylated MgF₂ and AlF₃ have already been successfully synthesised via a fluorolytic sol–gel process¹⁰ and were highly active as heterogeneous bi-acidic catalysts in the Friedel–Crafts alkylation (e.g. synthesis of vitamin E and vitamin K1)⁴ and the carbonyl–ene reaction (isomerisation of citronellal to isopulegol).² Further functionalisation of partially hydroxylated fluorides resulted in extended applications: fluorosulfonate grafted hydroxylated MgF₂ can catalyse the dehydration of xylose to furfural;^6,11 Au doped hydroxylated MgF₂ and Pd (or Pt) doped hydroxylated AlF₃ were active in the one-pot synthesis of menthol from citral due to their bi-acidity and hydrogenation ability;^12,13 Sn doped hydroxylated MgF₂ was reported as an efficient catalyst in the degradation of cellulose, offering a feedstock in bio-ethanol production.¹⁴

The previously reported bi-acidic fluoride catalysts focused on a narrow range of main group metal fluorides, i.e. AlF₃ or MgF₂. Naturally, questions arose as to whether it was possible to develop partially hydroxylated transition metal fluorides and whether these could provide comparable or even better catalytic performances than AlF₃ or MgF₂. In 2013 we reported a ternary FeF₃–MgF₂ catalyst that combined the advantages of the bi-acidic FeF₃ and the porous MgF₂ matrix and was highly active in the isomerisation of citronellal to isopulegol.⁵ The ternary FeF₃–MgF₂ catalyst was prepared via a one-step sol–gel route and the bi-acidity was introduced by a dehydration treatment of the Fe-precursor.⁵ In spite of the progress in the development of FeF₃–MgF₂, several issues needed to be clarified. First of all, the interaction between nanoscopic FeF₃ and MgF₂ required a detailed analysis since it would help understand the catalyst's structure and catalytic activity. More importantly, we devoted to investigate the formation pathway of the ternary FeF₃–MgF₂ catalyst with emphasis on revealing the function of the matrix material, MgF₂, during the synthesis process. Hence a template effect of MgF₂ was reported for the first time and was considered a key factor controlling the surface properties of FeF₃–MgF₂ catalysts. Inspired by the template effect of MgF₂, we established two novel ternary fluoride catalysts using other alkaline earth metal fluorides as matrices and discussed the role of the template effect in tuning the surface properties of this type of material.

Then the present work focuses on solving the following points. The interaction between nanoscopic FeF₃ and MgF₂ in the FeF₃–MgF₂ catalysts was characterised by scanning transmittance electron microscopy (STEM) coupled with energy-dispersive X-ray spectroscopy (EDX) and X-ray photoelectron spectroscopy (XPS). The formation pathway of the ternary FeF₃–MgF₂ was investigated by varying synthesis parameters in combination with characterisation methods such as elemental analysis (C, H, N, and F analysis) and powder X-ray diffraction (XRD). Finally, we synthesised FeF₃–CaF₂ and FeF₃–SrF₂ catalysts via the fluorolytic sol–gel process and characterised their physico-chemical properties with thermal analysis methods, STEM, N₂ physisorption, and chemisorption-IR. The isomerisation of citronellal to isopulegol was used as a model reaction to evaluate all the ternary fluoride catalysts studied in this work. We believe that our novel approach reveals the potential of ternary iron(III) fluoride-based catalysts and will promote their applications in heterogeneous catalysis. Moreover, such systems also offer potential applications as materials for making electrodes in fuel cell technology.

2. Results and discussion

2.1. Interaction between FeF₃ and MgF₂ in the ternary FeF₃–MgF₂ catalysts

In this work, we used STEM and XPS to characterise the morphology of FeF₃–MgF₂ catalysts and the nanoscopic interaction between FeF₃ and MgF₂ particles.

The STEM images (Fig. 1(a) and (b)) suggest that the particles in FeF₃–MgF₂ are spherical and are roughly homogeneous in size. The elemental mapping of Fe and Mg was generated by EDX in order to visualise the Mg and Fe elemental distribution in the material. The Fe and Mg distribution in a selected area that corresponds to the STEM image in Fig. 1(b) is shown in Fig. 1(c). The 2D hypermap reveals a homogeneous distribution of Fe and Mg. Two different topological situations may cause this distribution: (1) FeF₃–MgF₂ may consist of a layered structure, and the projection of Fe-rich layers and Mg-rich layers implies the pseudo-homogeneous Fe–Mg-distribution; (2) FeF₃–MgF₂ consists of a granular structure, and Fe and Mg are homogeneously distributed over the 3D space. Since the STEM images suggest a granular structure rather than a layered structure, we assume fully dispersed FeF₃ in MgF₂, and therefore, no isolated FeF₃-domains can be found embedded in MgF₂.

Fig. 1 (a) STEM image (scale bar: 10 nm) of FeF₃–MgF₂, (b) STEM image (scale bar: 5 nm) of FeF₃–MgF₂, and (c) 2D elemental hypermap (scale bar: 5 nm) of Fe and Mg generated from EDX results (red: Fe; green: Mg) in the region corresponding to (b). The contrast ratios of (a) and (b) were adjusted for better viewing.

XPS is a very useful technique for studying the electronic structure and chemical species in solid materials.¹⁵ In this work, XPS was applied to FeF₃–MgF₂ in order to clarify the effect of dispersing FeF₃ into the MgF₂ matrix based on the binding energy (BE) of Fe 2p_3/2 photoelectrons. The detailed Fe (2p_3/2) spectrum of a sol–gel prepared FeF₃ (synthesised as described by Y. Guo et al.⁵) is presented for comparison (Fig. 2(a)). The BE at 718 eV is assigned to the satellite peak to Fe 2p_3/2 (iron(III) fluoride), and the BE of the major component peak at about 715 eV corresponds to the Fe(III)–F bonds.¹⁶ The BE at 712 eV may refer to either minor Fe(III)–O or Fe(II)–F species. However, a previous study indicated that all the Fe atoms are in their +3 oxidation state.⁵ Thus the peak at the BE of 712 eV should be assigned to Fe(III)–O. The intensities of Fe (2p_3/2) components measured with FeF₃–MgF₂ (see Fig. 2(b)) are substantially different from the FeF₃ case. The major component here is at BE ∼712 eV which suggests a larger proportion of Fe–O bound species. Looking at the NIST XPS database¹⁷ reveals that Fe(III)OOH is to be expected at this BE^18,19 rather than Fe₂O₃ which is usually measured at a somewhat lower BE.²⁰ Since XPS is a surface sensitive technique,¹⁵ this result suggests that FeF₃–MgF₂ exhibits more surface –OH species compared with FeF₃. Because the surface –OH moieties are potential Brønsted acid sites,⁵ we expected FeF₃–MgF₂ to show more pronounced surface Brønsted acidity. Indeed, a previous study on FeF₃–MgF₂ confirmed its medium strong Brønsted acidity by chemisorption-IR and a catalysis test.⁵

Fig. 2 XPS spectra of Fe 2p of (a) sol–gel prepared FeF₃ and (b) FeF₃–MgF₂.

2.2. Formation pathway of ternary FeF₃–MgF₂ catalysts

The ternary FeF₃–MgF₂ catalysts were prepared by a sol–gel route in which a mixed Fe(OH)_x(NO₃)_3−x·yH₂O and Mg(CH₃COO)₂ suspension was fluorinated by HF and gave rise to a FeF₃–MgF₂ sol.⁵ To study the formation pathway of the ternary FeF₃–MgF₂ catalyst, we first examined whether there is any competition between Fe- and Mg-precursors during the fluorination process. It is known that the Mg(CH₃COO)₂ suspension, MgF₂ sol, and FeF₃ sol are white or colourless, whereas the Fe(OH)_x(NO₃)_3−x·yH₂O solution exhibits a reddish brown colour.⁵ Moreover, a significant colour change was observed during the synthesis of FeF₃–MgF₂ (the mixed precursor suspension was yellow; the FeF₃–MgF₂ sol was colourless).⁵ In light of this background, we designed two experiments to investigate the formation pathway of FeF₃–MgF₂ by using the colour change as an index.

In one attempt, a FeF₃ sol was prepared first, and a Mg(CH₃COO)₂ suspension was added afterwards to the FeF₃ sol.

In a second attempt, a mixed Fe(OH)_x(NO₃)_3−x·yH₂O and Mg(CH₃COO)₂ suspension was fluorinated in a single step by employing an under-stoichiometric amount of HF (two equivalents to Mg).

In the first attempt, addition of a stoichiometric amount of anhydrous HF to the Fe(OH)_x(NO₃)_3−x·yH₂O solution led to the formation of a colourless FeF₃ sol. Addition of the white Mg(CH₃COO)₂ suspension to this sol resulted first in the formation of an orange-red suspension that turned into a reddish brown clear sol (Fig. 3). This observation suggests that FeF₃ sol particles are highly active species that are able to react with Mg(CH₃COO)₂. It can be further deduced that Mg(CH₃OO)₂ should very probably be first fluorinated in the mixed Mg²⁺/Fe³⁺ precursor suspension, but fluorination of the iron precursor should proceed afterwards. The reddish brown sol shown in Fig. 3(h) was carefully dried at 30 °C to avoid any decomposition of the iron complex. The C, H, N and F elemental analysis results of the obtained brick-red powder as an intermediate substance suggested the co-existence of fluoride, nitrate, and acetate. The nitrate may result from the iron nitrate precursor in the synthesis of the FeF₃ sol. Thus the reaction between FeF₃ sol particles and Mg(CH₃COO)₂ probably gave rise to MgF₂ and an intermediate iron nitrate acetate complex.

Fig. 3 Images of (a) Mg(CH₃COO)₂ suspension in methanol, (b) FeF₃ sol in methanol, (c) adding the Mg(CH₃COO)₂ suspension to the FeF₃ sol, and image records of the reaction between Mg(CH₃COO)₂ and FeF₃ (Mg-to-Fe ratio = 1.5) at (d) t = 0 min, (e) t = 1 min, (f) t = 2 min, (g) t = 5 min, and (h) t = 14 min; (i) powder XRD patterns of an under-stoichiometrically fluorinated Fe–Mg sample after calcining at 600 °C (# refers to an Fe₂O₃ phase in PDF 33–664 and ° indicates an MgF₂ phase in PDF 1–1196); (j) schematic illustration of the template effect of MgF₂ sol particles.

In the second attempt, addition of under-stoichiometric HF to the Fe(OH)_x(NO₃)_3−x·yH₂O and Mg(CH₃COO)₂ suspension gave immediately a reddish brown sol that appeared to be similar to the one in Fig. 3(h). This sol was dried and the obtained powder was calcined at 600 °C before studying it by XRD (Fig. 3(i)). The reflexes can be assigned to an Fe₂O₃ phase and an MgF₂ phase, respectively. No evidence for FeF₃ or MgO was found. The XRD result confirmed the deduction that Mg becomes fluorinated first in the sol–gel synthesis, and the intermediate product is probably composed of MgF₂ and an iron nitrate acetate complex. It can be further speculated that the MgF₂ sol particles probably work as templates and induce the nucleation and govern the follow-up formation of FeF₃ (Fig. 3(j)). This template effect of MgF₂ would explain the uniform particle size in FeF₃–MgF₂ catalysts as revealed by STEM images. Moreover, the enhanced surface Brønsted acidity of FeF₃–MgF₂ as indicated by XPS results is probably introduced via the interaction between sol particles under in situ sol–gel synthesis conditions as the reaction pathway shows. Thus using MgF₂ as a matrix does not only template the textural structure but also tunes the surface acidity of FeF₃. Keeping in mind that both properties have a strong influence on the performances of a heterogeneous catalyst, we are convinced that the template effect of MgF₂ plays an important role in tuning the catalytic activities of FeF₃–MgF₂ catalysts. Consequently we were interested in whether other materials will also show this template effect. Hence we started the study on the template effects of heavier alkaline earth metal fluorides.

2.3. The template effects of heavier alkaline earth metal fluorides

To study the template effect of heavier alkaline earth metal fluorides, we prepared ternary FeF₃–CaF₂ and FeF₃–SrF₂ materials following the same route as that applied to FeF₃–MgF₂.⁵ Preparation of a ternary FeF₃–BaF₂ material was also considered. However, instead of the expected BaF₂, a Ba(NO₃)₂ phase was identified from the XRD pattern of the xerogel (Fig. 4(a)). The existence of Ba(NO₃)₂ indicated ion exchange between the precursors, iron nitrate and the barium acetate. In other FeF₃–alkaline earth metal fluoride systems, a similar ion exchange was evidenced by thermal analysis results (Fig. 4(b)). Release of NO⁺ (m/z = 30), a decomposition product of the nitrate, happens at different temperatures in different systems (FeF₃–MgF₂: ∼140 °C; FeF₃–CaF₂: ∼280 °C; FeF₃–SrF₂: ∼380 °C). The different decomposition temperatures clearly refer to different intermediate nitrate species. It is known that the thermal stability of alkaline earth metal nitrates increases with the atomic number.²¹ Thus these species can be assigned to Mg(NO₃)₂, Ca(NO₃)₂, and Sr(NO₃)₂, respectively. The ion exchange between iron nitrate and alkaline earth metal acetate probably results in these intermediate nitrate species. The general reaction pathway of this type of ternary fluorides is therefore summarised as follows: the mixed precursors undergo fast ion exchange followed by the two-step fluorination in which the alkaline earth metal fluoride works as a template, and finally give rise to the ternary fluorides (Fig. 5).

Fig. 4 (a) Powder XRD patterns of an “FeF₃–BaF₂” sample (+ refers to a Ba(NO₃)₂ phase in PDF 24–53); (b) MS profiles of ionic fragments NO⁺ (m/z = 30) of (i) FeF₃–MgF₂, (ii) FeF₃–CaF₂, and (iii) FeF₃–SrF₂ during thermal analysis.

Fig. 5 Schematic illustration of the general formation pathway of ternary iron(III) fluoride–alkaline earth metal fluoride.

The discovery of the ion exchange step not only provides important information on the formation pathway of this ternary fluoride system but also allows to predict the template effect of alkaline earth metal fluorides. Calculating the Gibbs free energy of the fluorination of alkaline earth metal nitrates with HF (Table 1) indicates that the formation of alkaline earth metal fluorides becomes less favoured with increasing the atomic number. As a consequence, the alkaline earth metal fluorides may lose their priority in the fluorination Step I (Fig. 5). Thus it can be predicted that the template effect may become less significant with increasing the atomic number of the alkaline earth metals in the ternary fluoride catalysts. To evaluate the template effect, we characterised the textural structure and surface acidity of FeF₃–CaF₂ and FeF₃–SrF₂ catalysts since these features are largely affected by the template effect.

Table 1 Gibbs free energy: the reaction of metal nitrate with HF to metal fluoride and nitric acid

	ΔrG (kJ mol^−1)
Mg(NO3)2	−92.3
Ca(NO3)2	−43.4
Sr(NO3)2	4.6
Ba(NO3)2	25.2

---

The textural structures of FeF₃–CaF₂ and FeF₃–SrF₂ catalysts were determined by STEM and N₂ adsorption–desorption measurements. One can roughly estimate the average particle size of different catalysts from the STEM images: FeF₃–SrF₂ > FeF₃–CaF₂ (Fig. 6). The differences in particle size may result in differences in the porous structure of the ternary fluoride materials, since the voids between particles contribute to the pore volume. Indeed, the average pore size of FeF₃–SrF₂ catalysts is almost two times as large as that of FeF₃–CaF₂ catalysts (Table 2, Column 3). Comparing all three ternary fluoride catalysts, their surface area and pore size decrease with the atomic number of the alkaline earth metal (Table 2, Columns 2 and 3). Taking the data of the corresponding alkaline earth metal fluorides as references, MgF₂ and CaF₂ showed a stronger influence in tuning the pore size of the ternary fluoride catalysts while SrF₂ hardly affects the pore size of FeF₃–SrF₂. The results confirmed that the template effect of alkaline earth metal fluorides in the aspect of controlling the porosity of the ternary catalysts declines from MgF₂ to SrF₂.

Fig. 6 STEM images (scale bar: 20 nm) of (a) FeF₃–CaF₂ and (b) FeF₃–SrF₂. The contrast ratios of the images were adjusted for better viewing.

Table 2 Summary of the surface properties of ternary FeF₃-based catalysts, FeF₃ and the corresponding alkaline earth metal fluorides

Sample	B.E.T. surface area (m^2 g^−1)	Average pore diameter (nm)	Surface Lewis acidity	Surface Brønsted acidity
FeF3–MgF2 (ref. 5)	400–500	2–5	Strong	Medium strong
FeF3–CaF2	ca. 200	5–15	Strong sites ≈ weak sites	Weak/medium strong
FeF3–SrF2	ca. 100	10–30	Strong sites ≪ weak sites	Weak/medium strong
FeF3	40–70	ca. 30
MgF2	240–260	2–3
CaF2	ca. 250	2–3
SrF2	ca. 130	2–3

---

Surface acidity of the FeF₃–CaF₂ and FeF₃–SrF₂ catalysts was studied by IR spectroscopy equipped with chemisorption devices. Different probe molecules (NH₃, pyridine, and CD₃CN) were used to determine the type and strength of surface acid sites (Lewis or Brønsted). Fig. 7 shows the chemisorption-IR spectra of samples after saturation with NH₃. Since the spectra are the subtractions of spectra after/before chemisorption, the negative bands at 1258–1260 cm⁻¹ are probably due to the decline of residual acetate band after NH₃ adsorption. The 1410 cm⁻¹ band in Fig. 7(a) can be assigned to protonated NH₄⁺ species which indicate Brønsted acidity.²² This band shifts towards the lower wavenumber range (Δδ = 26 cm⁻¹) compared with that in the spectra of FeF₃–MgF₂;⁵ it suggests that the Brønsted acidity of FeF₃–CaF₂ is probably weaker than that of FeF₃–MgF₂. The vibration band at 1222 cm⁻¹ and the shoulder at 1183 cm⁻¹ can be assigned to NH₃ species that coordinates to the surface Lewis acid sites.²² The broad peak at a lower wavenumber range (1134–1124 cm⁻¹) corresponds to H-bonding –OH⋯NH₃ species which typically refer to weak Brønsted acid sites.²²

Fig. 7 Chemisorption-IR spectra of (a) FeF₃–CaF₂ and (b) FeF₃–SrF₂ after saturating stepwise with NH₃.

With NH₃ as a probe molecule, both Lewis and Brønsted acid sites were determined on the surface of FeF₃–CaF₂. However, the differential spectra of FeF₃–SrF₂ show only one positive band at 1138 cm⁻¹ indicating the H-bonding species that correspond to weak Brønsted sites (Fig. 7(b)).²² No vibration bands corresponding to protonated NH₄⁺ were recorded in the spectra. This is an indication that the Brønsted sites of FeF₃–SrF₂ are not strong enough to protonate NH₃. It cannot be ruled out that the protonated NH₄⁺ species require adjacent –OH for stabilisation,²³ although there is only a low coverage of hydroxyls (potential Brønsted sites) on the FeF₃–SrF₂ surface. Because pyridine is a stronger base than NH₃ in the gas phase (proton affinity in the gas phase: pyridine 922.2 kJ mol⁻¹; NH₃ 857.7 kJ mol⁻¹),²⁴ pyridine was used as a probe molecule to investigate FeF₃–SrF₂. The two bands at 1641 and 1614 cm⁻¹ in the differential spectra of FeF₃–SrF₂ can be assigned to protonated (PyH⁺) and coordinated (PyrL) species, respectively (Fig. 8).²⁵ These bands arise from the vibration mode 8a of pyridine.²⁶ The band at 1598 cm⁻¹ refers to H-bonding species on both Lewis and Brønsted sites.²⁵ Thus FeF₃–SrF₂ was confirmed to carry both Brønsted and Lewis acid sites on the surface, but its Brønsted acidity is weaker than that of FeF₃–CaF₂ and FeF₃–MgF₂.

Fig. 8 Chemisorption-IR spectra of FeF₃–SrF₂ after saturating stepwise with pyridine.

CD₃CN was used as a probe molecule to distinguish the strengths of Lewis acid sites on the surface of FeF₃–CaF₂ and FeF₃–SrF₂. The vibration band at 2313–2318 cm⁻¹ refers to strong Lewis acid sites, while the band in the lower wavenumber range (about 2260–2290 cm⁻¹) indicates weak Lewis sites.^25,27 Interestingly, a decreasing tendency in Lewis acidity can be seen in the order of Mg, Ca, and Sr. Strong Lewis acid sites are predominant on the surface of FeF₃–MgF₂.⁵ On the surface of FeF₃–CaF₂, the strong and weak Lewis sites co-exist and the intensities of the corresponding bands are comparable (Fig. 9(a)). In contrast, weak Lewis acid sites are the main species (major characteristic) of the surface of FeF₃–SrF₂ (Fig. 9(b)). The bands at 2114–2118 cm⁻¹ correspond to the symmetric CD₃ stretching,²⁷ and the weak bands at about 2220–2260 cm⁻¹ can be assigned to physisorbed species.²⁷ The so-called trio bands (three bands at 2890, 2400, and 1700 cm⁻¹)²⁵ that correspond to very strong Brønsted acidity are not observed in these spectra, probably because the surface Brønsted acid sites of FeF₃–CaF₂ or FeF₃–SrF₂ are not strong enough to protonate CD₃CN.

Fig. 9 Chemisorption-IR spectra of (a) FeF₃–CaF₂ and (b) FeF₃–SrF₂ after saturating stepwise with CD₃CN.

In short, with the help of chemisorption-IR, medium strong Brønsted acid sites were determined on the surface of all three ternary fluoride catalysts with the Brønsted acidity decreasing with the atomic number of alkaline earth metal. Since the template effect of the alkaline earth metal fluoride is believed to contribute to the Brønsted acidity, the results support the hypothesis that the template effect declines from Mg to Sr. Moreover, the Lewis acidity shows a similar trend: strong Lewis acid sites are predominant on the FeF₃–MgF₂ surface, while weak Lewis sites are the main species on the surface of FeF₃–SrF₂, the surface acidity of FeF₃–CaF₂ lies in between with both strong and weak Lewis acidity. Thus the strengths of Lewis/Brønsted sites can be easily tuned using different alkaline earth metal fluorides as matrices, which allows us to fine-tune catalysts with the desired surface acidity.

2.4. Catalytic tests

The isomerisation of citronellal to isopulegol was used as a model reaction to evaluate the catalytic performance of FeF₃–CaF₂ and FeF₃–SrF₂ samples. This reaction is one of the key steps in synthesising menthol,^22,28 which is widely used in drugs, cosmetics, and food additives.²⁹ It represents an intramolecular carbonyl–ene reaction, in which an allylic hydrogen (an “ene”) reacts with a carbonyl (an “enophile”) to form a new bond with the migration of the ene double bond and 1,5-hydrogen shift (Fig. 10).^30–32 Early studies of the reaction mechanism suggested that it is catalysed by Lewis acids, and later studies revealed that it actually requires the participation of both Lewis and Brønsted acid sites to achieve high yields towards isopulegol.^33,34 Thus the isomerisation reaction of citronellal can be used as an excellent model reaction to examine the surface acidity of especially bi-acidic catalysts.

Fig. 10 Schematic illustration of the isomerisation reaction of citronellal to isopulegol.

A panel of FeF₃–CaF₂ and FeF₃–SrF₂ catalysts with varying Ca (or Sr)-to-Fe ratios was prepared. The catalytic results of FeF₃–CaF₂ samples are presented in Fig. 11(a). Both conversion and selectivity increased drastically for the Ca-to-Fe ratio between 1 and 4. At higher ratios, the conversion reached nearly 100% and hence no further increase can be expected. The FeF₃–CaF₂ sample with a Ca-to-Fe ratio of 3 shows the highest yield towards isopulegol with a conversion of 97% and a selectivity of 84%, which is comparable to that of FeF₃–MgF₂ samples (conversion ca. 100% and selectivity ca. 85%).⁵ Evidently, the Ca-to-Fe ratio has a strong influence on the performance of FeF₃–CaF₂ catalysts. This can be rationalized by considering the template effect of CaF₂: adjusting the Ca-to-Fe ratio may affect the template effect of CaF₂ which further influences the surface properties of FeF₃–CaF₂; and the surface properties are crucial for the heterogeneous catalysis process.

Fig. 11 (a) Conversion of citronellal and selectivity and yield to isopulegol versus the molar Ca-to-Fe ratio; and (b) conversion of citronellal and selectivity and yield to isopulegol versus the molar Sr-to-Fe ratio. The catalyst amounts used in the catalysis tests were normalised to 10 mg FeF₃ per 1 mmol citronellal.

In comparison with FeF₃–MgF₂ and FeF₃–CaF₂, the FeF₃–SrF₂ samples are less selective in this reaction. The conversion of citronellal reached its maximum of nearly 100% at the Sr-to Fe ratio of 7 (Fig. 11(b)). The selectivity to isopulegol, however, was noticeably lower than that of the FeF₃–MgF₂ or FeF₃–CaF₂ catalysts. One possible reason is that FeF₃–SrF₂ exhibits weaker surface Brønsted acidity and therefore is less selective to isopulegol. Moreover, Chuah et al. studied the correlation between pore size and selectivity to isopulegol and found that micropores or small mesopores are favourable for higher selectivity, because they may exclude large by-products such as dimers or trimers of isopulegol.³⁴ Our previous work on FeF₃–MgF₂ confirmed this conclusion.⁵ Thus the larger average pore diameter of FeF₃–SrF₂ (10–30 nm) may make it less competitive for the formation of isopulegol.

3. Experimental

3.1. Preparation of FeF₃–MgF₂, FeF₃–CaF₂, and FeF₃–SrF₂ catalysts

The detailed synthesis route of FeF₃–MgF₂ has been described in our previous work.⁵

The syntheses of FeF₃–CaF₂ and FeF₃–SrF₂ catalysts followed a similar route to that of FeF₃–MgF₂. The alkaline earth metal acetate (Ca(CH₃COO)₂·H₂O or Sr(CH₃COO)₂·0.5H₂O, Sigma-Aldrich, 97%) was used as received. It was added to the pre-dehydrated iron nitrate precursor (Fe(NO₃)₃·9H₂O, Sigma-Aldrich, 98%; the pre-dehydration process has been described by Guo et al.⁵) at different Ca (or Sr)-to-Fe ratios. The mixed precursors were suspended in methanol (total concentration of metals: 0.2 mol L⁻¹) and fluorinated with stoichiometric amounts of HF in a methanolic solution. The obtained sol or suspension was aged for 14–16 h at room temperature, and then dried and calcined at 100 °C in vacuo for 2 h.

Caution: HF solution is highly hazardous and special protective wear is required!

CaF₂ and SrF₂ were also prepared via the sol–gel process. The syntheses followed the same route as that applied for FeF₃–CaF₂ or FeF₃–SrF₂ catalysts but without an Fe-precursor.

3.2. Elemental analysis (C, H, N, and F analyses)

The C, H, and N concentrations were determined with a Leco CHNS-932 analyser. The F concentration was determined with a fluoride-sensitive electrode after digesting the sample by fusion with soda-potash.

3.3. Powder X-ray diffraction (XRD)

The XRD patterns were obtained from an XRD-7 Seifert-FPM diffractometer using a CuKα beam.

3.4. Thermal analysis

Thermal analysis was performed by using Netzsch-STA-409C/CD coupled with Balzers-QMG-422-Quadrupole-Mass spectrometry. Both TG curves and the m/z ratio of the fragments were recorded during measurements.

3.5. Scanning transmittance electron microscope (STEM)

The ternary FeF₃–MgF₂ sample was re-suspended in methanol (2–3 mg mL⁻¹). A Lacey carbon film (Cu-networks, 300 meshes) was dipped in the sample suspension and dried in air. The STEM images were recorded with a Joel JEM-2200FS Transmission Electron Microscope using a 200 kV field emission gun (FEG) and a high angle annular dark field detector (HAADF).

3.6. X-ray photoelectron spectroscopy (XPS)

The XPS measurements were carried out with an ESCALab 200X photoelectron spectrometer (VG Scientific, U.K.) for all samples. XP survey spectra were recorded at an angle of emission of 0° using non-monochromatised Al Kα excitation (15 kV, 20 mA) using the analyser in the constant retard ratio mode with CRR = 10 for survey scans and in the constant analyzer energy mode with CAE = 20 eV for highly resolved spectra. The samples were analysed under high vacuum (<10⁻⁸ mbar). The binding energy scale of the instrument was calibrated following ISO 15472. The binding energy scale was corrected for static charging using BE = 285.0 eV for the C 1s component of aliphatic carbon originating from absorbed hydrocarbons. Quantification was performed using a routine implemented in the Thermo Advantage 4.43 software. The routine uses Scofield's photoionisation cross section data for the determination of relative sensitivity factors. This quantification routine uses the model of homogeneous and isotropic distribution of all constituents in the analyzed volume. The estimated IMFP was calculated with the Gries (G-1) equation using the NIST Electron IMFP Database Version 1.2.³⁵

3.7. N₂ physisorption

N₂ physisorption at 77 K was measured with a Micromeritics ASAP 2020 instrument. Samples were degassed at 80 °C under high vacuum for 12 h. The specific surface area was calculated using the B.E.T. equation, and the pore size distribution was determined by the B.J.H. method.

3.8. Chemisorption-IR

For the IR spectroscopy coupled with chemisorption of probe molecules (chemisorption-IR), about 25–30 mg sample was pressed at 1 ton into a self-supporting disc (2 cm² area) for IR analysis. The sample disc was placed on 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 allows adjusting the sample disc in the infrared beam for spectra acquisition and to displace it into a furnace at the top of the cell for thermal activation of the samples. The cell was connected to a vacuum line and a glass injection loop. After we activated the sample at 80 °C under vacuum, different probe molecules were added in steps by small doses. The amount of injected probe molecules was regulated by the valves. The IR spectra were recorded on a Thermo Scientific Nicolet iS10 spectrometer.

3.9. Catalysis

A citronellal isomerisation reaction was performed in a Thermomixer comfort (Eppendorf) equipped with glass tubes. The samples were heated at 80 °C in vacuo for 2 h before use to remove the surface impurities. Different amounts of fluoride samples were added to the reactant mixture (0.2 mL citronellal (Acros, 93%) solved in 2 mL toluene (Acros, 99.99%), and 0.1 mL undecane (Sigma-Aldrich, ≥99%) as the internal standard for GC evaluation). The glass tubes were sealed and maintained at 80 °C 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 m × 0.25 mm × 0.5 μm, temperature programming: starting temperature 120 °C, hold time 50 min, rate 15 °C min⁻¹, end temperature 270 °C, hold time 5 min).

4. Conclusion

The structural and surface characterisation of the FeF₃–MgF₂ catalysts revealed that MgF₂ strongly influences the porous structure and the chemical environment around the Fe atoms, and probably contributes to the enhanced Brønsted acidity of FeF₃–MgF₂. The investigation on the formation pathway of FeF₃–MgF₂ concludes for the first time a template effect of the MgF₂, which explains the role of the MgF₂ matrix in the FeF₃–MgF₂ catalysts. The template effect was introduced during the fluorination of mixed Mg²⁺ and Fe³⁺ precursors. During this step, Mg²⁺ becomes selectively fluorinated, forming dominantly MgF₂ particles in the sol, whereas fluorination of the Fe-precursor proceeds afterwards. The discovery of this template effect is essential for tuning the porosity and surface acidity of FeF₃–MgF₂ catalysts and inspired us to investigate other ternary FeF₃-based catalysts with heavier alkaline earth metal fluorides as matrices.

In light of this background, FeF₃–CaF₂ and FeF₃–SrF₂ catalysts were successfully prepared via the fluorolytic sol–gel route and for CaF₂ and SrF₂ the template effect was evidenced based on structural and surface characterisation as well. Using different alkaline earth metal fluorides as matrices significantly changed the porosity, and a systematic trend of increasing average pore size was observed whereas the surface area decreased with the atomic number of alkaline earth metals. Both Lewis and Brønsted acid sites were detected on the surface of FeF₃–CaF₂ and FeF₃–SrF₂ by chemisorption-IR. Notably, by comparing all three ternary fluoride catalysts, the FeF₃–SrF₂ showed weaker Brønsted acidity and much weaker Lewis acidity than FeF₃–CaF₂ and FeF₃–MgF₂. Thus the template effect decreases with increasing atomic number of the alkaline earth metal. The results confirmed the importance of the template effect of matrix materials in the ternary fluoride catalysts.

In short, the ternary iron(III) fluoride-based catalysts exhibit great potential in heterogeneous catalysis due to their pronounced and adjustable surface acidity and flexibility in preparation. This work highlights for the first time the template effect of the matrix materials in tuning the surface properties of the catalysts. It furthermore shows the feasibility of adjusting the surface properties of the catalysts by varying the matrix material or the alkaline earth metal to iron ratio or both. This fluorolytic synthesis approach opens access to next-generation catalysts based on ternary metal fluorides with adjustable surface properties.

Acknowledgements

The authors would like to thank Dr Holm Kirmse for taking the STEM images, Dr Micheal Feist for performing the thermal analysis, and Dr Katharina Teinz and Larisa Schimdt for the fruitful discussion. Y. Guo thanks the Chinese Scholarship Council (CSC) for a PhD scholarship.

References

1. S. M. Coman, S. Wuttke, A. Vimont, M. Daturi and E. Kemnitz, Adv. Synth. Catal., 2008, 350, 2517–2524.
2. S. M. Coman, P. Patil, S. Wuttke and E. Kemnitz, Chem. Commun., 2009, 460–462.
3. S. Wuttke, S. M. Coman, J. Kröhnert, F. C. Jentoft and E. Kemnitz, Catal. Today, 2010, 152, 2–10.
4. N. Candu, S. Wuttke, E. Kemnitz, S. M. Coman and V. I. Parvulescu, Appl. Catal., A, 2011, 391, 169–174.
5. Y. Guo, P. Gaczyński, K.-D. Becker and E. Kemnitz, ChemCatChem, 2013, 5, 2223–2232.
6. I. Agirrezabal-Telleria, F. Hemmann, C. Jäger, P. L. Arias and E. Kemnitz, J. Catal., 2013, 305, 81–91.
7. K. Tanabe and W. F. Hölderich, Appl. Catal., A, 1999, 181, 399–434.
8. K. Wilson and J. H. Clark, Pure Appl. Chem., 2000, 72, 1313–1319.
9. Y. Guo, S. Wuttke, A. Vimont, M. Daturi, J.-C. Lavalley, K. Teinz and E. Kemnitz, J. Mater. Chem., 2012, 22, 14587–14593.
10. S. Wuttke, S. M. Coman, G. Scholz, H. Kirmse, A. Vimont, M. Daturi, S. L. M. Schroeder and E. Kemnitz, Chem. – Eur. J., 2008, 14, 11488–11499.
11. I. Agirrezabal-Telleria, Y. Guo, F. Hemmann, P. L. Arias and E. Kemnitz, Catal. Sci. Technol., 2014, 4, 1357–1368.
12. A. Negoi, S. Wuttke, E. Kemnitz, D. Macovei, V. I. Parvulescu, C. M. Teodorescu and S. M. Coman, Angew. Chem., Int. Ed., 2010, 49, 8134–8138.
13. A. Negoi, K. Teinz, E. Kemnitz, S. Wuttke, V. Parvulescu and S. Coman, Top. Catal., 2012, 55, 680–687.
14. S. Wuttke, A. Negoi, N. Gheorghe, V. Kuncser, E. Kemnitz, V. Parvulescu and S. M. Coman, ChemSusChem, 2012, 5, 1708–1711.
15. K. Siegbahn and K. Edvarson, Nucl. Phys., 1956, 1, 137–159.
16. M. Kasrai and D. S. Urch, J. Chem. Soc., Faraday Trans. 2, 1979, 75, 1522–1531.
17. A. V. Naumkin, A. Kraut-Vass, S. W. Gaarenstroom and C. J. Powell, NIST X-ray Photoelectron Spectroscopy Database, NIST Standard Reference Database 20, Version 4.1, National Institute of Standards and Technology, Gaithersburg, MD., U.S.A., 2012.
18. B. J. Tan, K. J. Klabunde and P. M. A. Sherwood, Chem. Mater., 1990, 2, 186–191.
19. N. S. McIntyre and D. G. Zetaruk, Anal. Chem., 1977, 49, 1521–1529.
20. E. Paparazzo, J. Phys. D: Appl. Phys., 1987, 20, 1091.
21. A. F. Holleman and E. Wiberg, Lehrbuch der Anorganischen Chemie, Walter de Gruyter, Berlin, New York, 2007.
22. A. A. Tsyganenko, D. V. Pozdnyakov and V. N. Filimonov, J. Mol. Struct., 1975, 29, 299–318.
23. E. H. Teunissen, R. A. Van Santen, A. P. J. Jansen and F. B. Van Duijneveldt, J. Phys. Chem., 1993, 97, 203–210.
24. A. Auroux, Acidity and Basicity, Springer-Verlag, Heidelberg, 2008.
25. G. Ertl, H. Knözinger and J. Weitkamp, Handbook of Heterogeneous Catalysis, Wiley-VCH, Weinheim, 2008.
26. D. Cook, Can. J. Chem., 1961, 39, 2009–2024.
27. A. G. Pelmenschikov, R. A. van Santen, J. Janchen and E. Meijer, J. Phys. Chem., 1993, 97, 11071–11074.
28. J. C. Leffingwell and R. E. Shackelford, Cosmetics Perfumery, 1974, 89, 69–89.
29. D. H. Pybus and C. S. Sell, The Chemistry of Fragrances, RSC Paperbacks, Cambridge, 1999.
30. B. B. Snider, M. Karras, R. T. Price and D. J. Rodini, J. Org. Chem., 1982, 47, 4538–4545.
31. W. Oppolzer and V. Snieckus, Angew. Chem., Int. Ed. Engl., 1978, 17, 476–486.
32. H. M. R. Hoffmann, Angew. Chem., Int. Ed. Engl., 1969, 8, 556–577.
33. Z. Yongzhong, N. Yuntong, S. Jaenicke and G.-K. Chuah, J. Catal., 2005, 229, 404–413.
34. G. K. Chuah, S. H. Liu, S. Jaenicke and L. J. Harrison, J. Catal., 2001, 200, 352–359.
35. C. J. Powell and A. Jablonski, NIST Electron Inelastic-Mean-Free-Path Database, Version 1.2, SRD71, National Institute of Standards and Technology, Caithersburg, MD, 2010.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c4dt03229b

---