Selective Chemical Filters for VOF₃: Tailoring MgF₂ Filter Selectivity through Surface Chemistry

Abstract

In order to synthesize chemical filters for the selective removal of volatile fluorides, commercial magnesium fluoride MgF₂ with high specific surface area (HSA) was investigated. The amount of -OH groups substituting fluorine is not negligible, partly due to the high surface area, but also due to the synthesis route. These hydroxyl groups induce a Lewis basicity on the surface of metal fluorides. The amount of these Lewis basic sites has been tailored using fluorination with F₂ gas. The sorption of VOF₃, used as model gas, onto these fluorides was investigated. The versatility of surface chemistry as a function of a number of Lewis basic sites opens the way to filter selectivity mixture of volatile fluorides depending on their Lewis acidity. HSA MgF₂ acts as a stable matrix towards the gas to be purified, and the selectivity may be achieved by a higher Lewis acidity of the gaseous impurity.

1. Introduction

Volatile fluorides are involved in numerous industrial applications, either as reagents or pollutants. Most of them are Lewis acids and react with a wide variety of compounds; for example, in organic synthesis, gaseous VOF₃ is often used for oxidative coupling of phenolic rings [1]. Due to their ability to release atomic fluorine, volatile fluorides can also be used as fluorinating agents. Nanoparticles, thin films or atomic layers of tungsten and molybdenum metal are synthesized using WF₆ or MoF₆ as precursors [2,3]; metal deposition occurs through the reaction of MF₆ into M plus 3 F₂. A reactant such as Si₂H₆ is used together with MoF₆ to form Mo layers. Dense films of molybdenum oxide may be deposited by plasma-enhanced chemical vapor deposition using mixtures of MoF₆, H₂, and O₂ [4].

When the volatile fluorides are pollutants, however, chemical filters are needed to remove them. The most important application concerns UF₆, which plays a key role in the nuclear industry. UF₆, as a volatile uranium compound, allows uranium enrichment to ²³⁵U whatever the process, i.e., gaseous diffusion, centrifugation, or laser excitation. This particular case highlights the necessity for selectivity of the chemical filter. To provide high-purity nuclear fuel, international standards have been established and the quantity of pollutants allowed in nuclear pellets is restricted. Pollutants originate from both uranium ore, the chemical agents used in the conversion process, and/or the fission products of spent uranium. All these elements are fluorinated during the synthesis of UF₆ and their volatility and solubility in UF₆ vary depending on the element [5]. When the volatility of the impurity is close to that of UF₆, its removal is complicated. Selective chemical filters that do not react with UF₆ are thus highly needed. Ca(OH)₂, Mg(OH)₂, KOH, either separately or in combination, cannot be used because they react easily with UF₆ [6].

Our strategy involves metal fluorides which contain just a small part of basic -OH groups. Considering the reactivity of the volatile fluorides which must be trapped, the criteria for an effective chemical filter are a high surface area, high porosity to increase the interface with the target gas, chemical stability in order to avoid their decomposition, but also enough Lewis basic sites on the surface to react with volatile fluorides as Lewis acids. To reach sufficient selectivity, the surface chemistry or porosity of the filter must be able to match the chemisorption or physisorption of the gas onto the filter surface.

To fulfil these criteria, our choice was to go towards fluorides with high specific surface area (HSA). Usually, such fluorides are obtained via the sol–gel route [7,8,9,10,11] or by microwave-assisted solvothermal synthesis [12,13,14,15,16,17]. HSA metal fluorides are prepared using metal alkoxides in an organic medium or various solvents and metal precursors in an aqueous-HF medium. The synthesis conditions using microwave-assisted solvothermal routes, i.e., the choice of precursors, solvent, HF concentration and reaction temperature, strongly influence the formation of various networks with different chemical compositions. Magnesium difluoride MgF₂, better known for its catalytic [18,19] and optical properties [20,21,22], has been selected as a chemical filter based on previous data on the removal of impurities such as fluorides of technetium [23], molybdenum [24], ruthenium, neptunium or plutonium [25].

The presence of weak basic sites was evidenced in MgF₂ prepared by sol–gel [26] and weak basic sites may prevent reaction with UF₆. Basic sites coexist with a large amount of Lewis acid sites, which explains the unique catalytic properties of MgF₂ prepared by sol–gel [8]. F⁻ and O²⁻ are both intrinsically Lewis bases; however, fluorine atoms reduce the basicity of oxygen atoms. Moreover, the basicity of MgF₂ is much lower than that of MgO. We propose to adapt the OH content in order to reach the selectivity, VOF₃ versus UF₆ for instance, but in a general way by using the Lewis acidity differences of the gases in the mixture. Moreover, the aim is to design a chemical filter that may be regenerated; metal fluoride acts as the support and the OH content changes during reaction with the impurity to be removed.

2. Materials and Methods

2.1. Materials

In order to test the tuning of the OH content, three MgF_2−x(OH)_x were used: a commercial MgF₂ received in pellet form (Nippon Puretec, Nagoya, Japan), which is used as either as received or after fluorination treatment, and a locally synthesized oxygen-free MgF₂. The commercial sample was chosen for its ease of use, in pellet rather than powder form, for all the filtering and regeneration operations as well as for the quantities available for future industrial uses. VOF₃ was synthesized locally and used as is.

2.2. Filtering Capacities

All experiments were performed in polytetrafluoroethylene (PTFE) bottles sealed in a nitrogen dry box due to the hygroscopic nature of fluorides. MgF_2−x(OH)_x pellets and VOF₃ powders were put in separate nickel baskets. The volatility of VOF₃ at 80 °C allowed the exposure of its gas on MgF_2−x(OH)_x samples. After exposure to VOF₃ for 24 h, the MgF_2−x(OH)_x pellets were crushed for characterization. The vanadium content on the filter surface was estimated by weight uptake and ICP analysis, both of which gave consistent data.

2.3. Fluorination

Pure molecular fluorine (Solvay, 99%+) was used. A chemical trap filled with soda lime scrubbed F₂ molecules from the exhaust in order to avoid their release into the atmosphere. The gas flow was set at 20 mL·min⁻¹. The treatment was performed at a fluorination temperature T_F = 300 °C. The temperature profile of the treatment is a heating ramp of 5 °C · min⁻¹, with a stabilization of temperature for 4 h. After this, the reactor was flushed with nitrogen in order to remove all reactive gases.

2.4. Characterization

2.4.1. X-ray Powder Diffraction

X-ray powder diffraction patterns were recorded with a Panalytical X’Pert powder diffractometer in Θ- Θ Bragg Bentano geometry. The samples were transferred to a sealed cell to avoid exposure to moisture, with an aluminum sample holder for some. All patterns were recorded between 5° and 70° in 2Θ with a step of 0.015° and a counting time of 60 min using a back graphite monochromated CuKα radiation (Kα1 = 1.54056 Å and Kα2 = 1.54439 Å). Profile matching refinements were performed using the FULLPROF software [27].

2.4.2. Adsorption/Desorption Isotherms

The nitrogen adsorption/desorption isotherms were measured using the Micromeritics ASAP 2020 instrument. Prior to each adsorption experiment, the samples were degassed at 473 K under primary vacuum and then under secondary vacuum. Pore volume, specific surface area, and pore size distribution were extracted from the N₂ adsorption/desorption isotherms at 77 K using the BET (Brunauer, Emmett and Teller) and BJH (Barrett, Joyner and Halenda) models for specific surface area (SSA) and pore size distribution for mesoporous materials, respectively.

2.4.3. Raman Spectroscopy

Raman spectra were collected at room temperature using a Bruker RFS 100/S apparatus with a Nd-YAG (aluminum-doped yttrium garnet) laser source at 1064 nm. A total of 500 scans were recorded between 4000 and 25 cm⁻¹ Raman shift. Samples were prepared in a sealed fluorinated ethylene-propylene tube (FEP, La Mothe-aux-Aulnais, Saint Gobain) that resulted in the presence of additional Raman bands (marked on the spectra). For the Raman analyses, the MgF₂ single-crystal (Sigma Aldrich, 99%) was used as the reference for oxygen-free magnesium difluoride.

2.4.4. NMR Spectroscopy

Multinuclear ¹⁹F, ¹H and ⁵¹V NMR measurements were carried out with a Bruker Advance Spectrometer with working frequency of 282.2, 300.0 and 78.8 MHz, respectively. A Magic Angle Spinning (MAS) probe operating with 2.5 mm rotors was used allowing a 30 kHz spinning rate. A sequence with a single π/2 pulse duration of 4.0 μs was used. The ¹⁹F, ¹H and ⁵¹V NMR chemical shifts were externally referenced to CFCl₃, tetramethylsilane (TMS) and solution of vanadium phosphate (1 M), respectively.

3. Results and Discussion

3.1. Fluorination to Tailor OH/F Ratio in MgF_2−x(OH)_x

In order to test the tuning of the OH content, three MgF_2−x(OH)_x were used. All the characterizations highlight the presence of -OH groups in the commercial sample, and synthesis via the sol–gel route is strongly suspected. The chemical composition may be written as MgF_2−x(OH)_x. The aim of post-fluorination of the commercial source is to tailor its OH/F ratio and surface chemistry, but to obtain pure oxygen-free MgF₂, the choice was made to fluorinate an oxygen-free source.

3.1.1. Conversion of MgB₂ into Oxygen-Free MgF₂

In order to select the precursor for the synthesis of oxygen-free MgF₂ via an etching during the fluorination of the element other than Mg, i.e., B, N; Si or P in MgB₂ (BF₃ evolution), Mg₃N₂ (NF₃), Mg₂Si (SiF₄) and Mg₃P₂, the following criteria were used:

• The expected pore size, in accordance with the size of the released molecules (0.243, 0.320, 0.377 and 0.377 nm) for MgB₂ (BF₃), Mg₃N₂ (NF₃), Mg₂Si (SiF₄) and Mg₃P₂ (PF₅), respectively.
• The toxicity of the gases released at the completion of the first reaction item.
• The presence of solid products other than MgF₂.

According to (3) MgC₂ cannot be retained because fluorocarbons may be formed. MgH₂ results after fluorination in the narrowest pores and has not been selected because HF is also undesirable according to (2). A toxic gaseous mixture S₂F₂/SF₆/F₂ is also formed from MgS, excluding this precursor according to (2). Mg₃N₂ is not considered because a thermal post-treatment is necessary to remove NH₄F from as-prepared MgF₂, that should lead to decrease the surface area. A narrow pore size being preferred, MgB₂ is selected rather than Mg₃P₂. The fluorination of this MgB₂ precursor at 300 K and 1 atm occurred in two steps: for the addition of F₂ in between 0 and 4 moles, the initial precursor is totally consumed to form solid boron and MgF₂; when F₂ is further added, the boron is then fluorinated as gaseous BF₃. From 4 moles of fluorine gas, the only solid product is MgF₂ and the gaseous mixture consists of BF₃ and F₂. A solid reaction yield slightly higher than 100% is explained by the presence of the intermediate product Mg(BF₄)₂ found at the completion of the reaction.

While the XRD pattern of the final product (Figure 1a) reveals only magnesium difluoride MgF₂ [28,29] with traces of MgB₂, the presence of the intermediate product Mg(BF₄)₂ is unambiguously revealed by FTIR spectroscopy (Figure 1b) considering the B-F bond vibration bands of at 1111, 1080, 461 cm⁻¹ [30].

The bands related to the hydroxyl groups (3430 and 1647 cm⁻¹) are absent contrary to the other MgF₂ samples (main peak at 435 cm⁻¹ associated with Mg-F vibration mode) [31]. Two bands are identified in the ¹⁹F NMR spectrum (Figure 1c). The main band at −196 ppm is assigned to Mg-F bonds. A small shoulder centered at −145 ppm is due to BF₄⁻ present in very few amounts [32]. BF₄⁻ reveals the presence of an intermediate compound between MgB₂ and BF₃ following the mechanism:

$${\mathrm{M}\mathrm{g}\mathrm{B}}_{2\mathrm{}\left(\mathrm{s}\text{, black}\right)}\stackrel{\Delta ,+\mathrm{}4{\mathrm{F}}_2}{\to }{\mathrm{M}\mathrm{g}\left({\mathrm{B}\mathrm{F}}_4\right)}_{2\mathrm{}\left(\mathrm{s}\right)}\stackrel{\Delta ,\mathrm{}-\mathrm{}2\mathrm{B}{\mathrm{F}}_3}{\to }{\mathrm{M}\mathrm{g}\mathrm{F}}_{2\mathrm{}\left(\mathrm{s}\text{, white}\right)}$$

It is worthwhile to note that no shoulders indicating the presence of hydroxyl groups are observed in the final product. As a matter of fact, OH groups result in a change in electronic density around the ¹⁹F nuclei in their neighboring and consequently a small band should appear at higher chemical shift due to the decrease in the Mg-F bond ionicity (increase in the covalence of the Mg-F bond). A band or shoulder is then observed in the −160/−180 ppm range for other MgF₂ samples; its intensity is related to the synthesis and post-treatment but mainly to the nature and concentration of the hydroxyl groups substituting fluorine. The higher the intensity of this band, the higher the content of hydroxyl groups, (Figure 1c). The ¹H NMR spectrum of MgF₂ without oxygen (Figure 1d) confirms the absence of hydroxyl groups. The objective of preparing oxygen-free MgF₂ is thus reached. The sample obtained by fluorination of the magnesium diboride precursor exhibits a type II profile for the N₂ isotherm (Figure 1e), typical of non-porous or macroporous materials.

The absence of micropores indicates that the lattice is totally rebuilt during the chemical etching and BF₃ evolution. One should note the SSA (specific surface area) of 35 m² · g⁻¹ for the oxygen-free MgF₂. BF₃ gaseous molecules which are produced during the fluorination of MgB₂ allow a relatively high specific surface area to be maintained, that is unusual with gas/solid fluorination synthesis of fluorides.

3.1.2. Tailoring of OH/F Ratio in Conventional MgF₂

Considering the ¹⁹F NMR spectra (Figure 1c), the main band at −196 ppm is assigned to ¹⁹F nuclei in the F-Mg-F groups. This chemical shift is in accordance with the literature data [32]. A shoulder also appears for the raw compound and its intensity decreases after post-fluorination treatment. This shoulder is relative to the presence of hydroxyl groups (OH-Mg-F) in MgF₂. As mentioned previously, the position of the band gives information on the fluorine–oxygen environment of magnesium in MgF₂. By fitting the spectra using two Lorentzian lines, the amount of hydroxyl groups is obtained for each compound. The OH/F ratio is 0.14 for raw MgF₂ and 0.04 for the MgF₂ post-fluorinated at 240 °C. The composition of the two samples can be written MgF_1.75(OH)_0.25 and MgF_1.925(OH)_0.075. Regarding ¹H NMR spectra (Figure 1d), only one band is observed (the others being related to the rotor cap) corresponding to the hydroxyl groups in MgF₂. After fluorination, its intensity, i.e., the amount of OH groups, decreases, confirming the efficiency of the treatment in removing OH.

Conversely, oxygen-free MgF₂ exhibits the characteristics of a non-porous or macroporous compound, raw and post-fluorinated MgF₂ present a type IV hysteresis which is typical of a mesoporous material (Figure 1e). The BJH method indicates an average pore size of 8.3 and 20.3 nm for raw and post-fluorinated pellets of commercial MgF₂, respectively. The increase in pore diameter is due to the coalescence phenomenon induced both by the fluorination temperature and by the release of OH groups. It is not possible to extract an average pore diameter using the BJH method for the sample obtained from the boride precursor because the BJH method is only suitable for mesoporous materials. The BET surfaces are 100 and 72 m² · g⁻¹ for raw and post-fluorinated MgF₂, respectively. The post-fluorination treatment decreases the surface area by substituting fluorine atoms for hydroxyl groups.

3.2. Sorption of VOF₃ in MgF_2−x(OH)_x

At this step, three different kinds of magnesium difluoride with various contents of OH groups were available for VOF₃ sorption tests: commercial (MgF_1.75(OH)_0.25), post-fluorinated (MgF_1.925(OH)_0.075) and oxygen-free MgF₂ (MgF₂). After exposure to VOF₃, the relative quantity of vanadium trapped was 3.6 w.% for the raw MgF₂, 2.1 w.% for the post-fluorinated sample and 1.3 w.% for the oxygen-free synthesized product. All weight uptakes were confirmed by ICP analysis (

Table 1. Weight uptake after VOF₃ sorption experiments.

MgF2 Treatment	Bulk Weight Uptake (%)	Vanadium Weight Uptake (%)	Vanadium ICP Analysis (%)
Raw	8.7	3.6	3.4
Post-fluorinated at 240 °C	5.1	2.1	2.0
Synthesized from MgB2	3.1	1.3	1.2

). The higher the OH/F ratio, the higher the sorption rate. The similitudes between the XRD patterns before and after exposure to VOF₃ indicate no new crystalline phase formed nor any change in crystallinity, the width of diffraction peak being the same as that of MgF₂ before the sorption of VOF₃.

The FTIR spectra of the samples (Figure 2a) reacting with vanadium oxyfluoride reveal the occurrence of magnesium—fluorine and vanadium—oxygen bonds. The V=O and V-F stretching bands of vanadium oxyfluoride are identified as raw and treated commercial MgF₂ after sorption at 1000 cm⁻¹ (V=O) and 820 cm⁻¹ (V-O) on the IR spectra (Figure 2a) [33,34]. The intensities of the vanadium–oxygen vibration bands are higher for raw MgF₂ than for treated MgF₂ in accordance with the amount of vanadium trapped (3.6 and 2.1 w.%, respectively); see Table 1. Only low bands are detected at 1000 cm⁻¹ (V=O) and 720 cm⁻¹ (V-F) in the case of oxygen-free samples.

Gas phase IR and pressure measurements were also performed to follow the nature of the gas generated or consumed during the sorption of VOF₃ on MgF₂. The pressure drops in a first sorption step as expected due to the trapping of VOF₃ molecules onto the surface of the filter, but increases rapidly after, indicating the release of other gases into the IR chamber. The gas-phase FTIR spectra pointed out a massive group of bands between 4500 and 3500 cm⁻¹ (Figure 2c). These bands are characteristic of gaseous HF. Their intensity increased during the sorption (Figure 2b).

HF is unambiguously the product of a chemical reaction occurring between MgF_2−x(OH)_x and VOF₃. This constitutes another proof of a chemisorption process. V=O and V-F vibration bands characteristic of vanadium oxyfluoride were also detected with Raman spectroscopy (Figure 3).

Data from the literature indicate that the first band at 1018 cm⁻¹ is assigned to the vibration of the V=O bond in VOF₃ whereas the second at 798 cm⁻¹ is related to the V-F bond [35]. As in the FTIR data of MgF₂ exposed to VOF₃, the peak intensity was higher for the raw MgF₂ than for the post-fluorinated sample, once again in accordance with the sorption rates. No bands were observed with the oxygen-free crystalline MgF₂ sample as the quantity of vanadium trapped (close to 1 w.%) was not sufficient to detect any sorption product.

After sorption of VOF₃, the ¹⁹F NMR spectra show a single line due to Mg-F bonds (Figure 4a). The shoulder indicating the presence of hydroxyl groups disappears after the sorption. A chemical reaction occurs that removes OH⁻ groups from the surface of MgF₂ and involves the volatile compound. No further lines due to fluorinated vanadium compounds are observed in the final product. For the ¹H NMR spectra (Figure 4b), the peak intensity assigned to hydroxyl groups in MgF₂ decreases for another contribution at a chemical shift of +9 ppm. This band is assigned to the interaction between vanadium oxyfluoride and protons and confirms the chemical reaction between the OH⁻ groups and VOF₃. ⁵¹V NMR spectra recorded at different spinning rates in order to distinguish between isotropic and spinning bands (Figure 4c) reveal the presence of 3 isotropic bands for raw and post-fluorinated MgF₂ exposed to VOF₃. The isotropic bands at −561, −615 and −791 ppm refer to VO₂F₂⁻, VOF₃ and VOF₄⁻, respectively [36,37,38].

The presence of VO₂F₂⁻ ions evidences that VOF₃ reacts with the OH groups on the surface of the chemical filter to form MgF_2−x(VO₂F₂)_x. Assuming that the vanadium oxyfluoride anion coordinates Mg²⁺ via oxygen and occupies a distorted tetrahedral site, the steric hindrance to hydroxyl groups is radically different and VO₂F₂⁻ can substitute OH groups only at the surface and not in the rutile network. From an electronic point of view, V⁵⁺ is a second–order Jahn-Teller ion exhibiting a strong polyhedral distortion. This implies that this molecular species can accommodate a high distortion which permits the stabilization of MgF_2−x(VO₂F₂)_x compositions [39].

The generated HF increases the pressure in the reactor and can react with the VOF₃. The resulting product is HVOF₄ (trapped on the surface of the filter) in accordance with the isotropic band observed at −791 ppm. As expected, no isotropic band is detected for the oxygen-free MgF₂ because of the low quantity of vanadium trapped. VOF₃ sorption can be summarized as:

$$\mathrm{M}\mathrm{g}{\mathrm{F}}_{2-\mathrm{x}}{\left(\mathrm{O}\mathrm{H}\right)}_{\mathrm{x}\mathrm{}\left(\mathrm{s}\right)}+\mathrm{x}{\mathrm{}\mathrm{V}\mathrm{O}\mathrm{F}}_{3\mathrm{}\left(\mathrm{g}\right)}\to \mathrm{M}\mathrm{g}{\mathrm{F}}_{2-\mathrm{x}}{\left({\mathrm{V}\mathrm{O}}_2{\mathrm{F}}_2\right)}_{\mathrm{x}\mathrm{}\left(\mathrm{s}\right)}+\mathrm{x}\mathrm{}{\mathrm{H}\mathrm{F}}_{\left(\mathrm{g}\right)}$$

In addition to this expected reaction on the metal fluoride surface (chemisorption), a physisorption mechanism may also occur because some of the vanadium is trapped by oxygen-free MgF₂ (1.3 w.%).

Chemisorption involves the Lewis basicity of MgF₂ through OH groups. This basicity may be tailored both via the nature, i.e., the coordination number of OH/F anions, the content of OH groups and the cation associated with fluorine. To go further in the discussion, the combination of polarizable cations with low electronegativity (K⁺ in KMgF₃, Mg²⁺, Ca²⁺) and OH⁻ groups substituting F⁻ ions is another route to control the strength and number of Lewis basic sites keeping in mind the Lewis acidity of the gas that must be removed. The surface concentration and strength of the Lewis basic sites of the filter can be fitted to the Lewis acidity of the target gas in a gaseous mixture. Considering the structural features of the fluoride series: CaF₂ with a fluorite-type structure and fluorine atoms coordinated fourfold to Ca²⁺, MgF₂ with a rutile-type structure with fluorine atoms coordinated threefold to Mg²⁺ and KMgF₃ with a perovskite-type structure and fluorine atoms twice coordinated to Mg²⁺, a large variation in the concentration and strength of the Lewis basicity sites is expected. Furthermore, the concentration of OH⁻ groups can be adjusted by post-fluorination.

3.3. Regeneration of the Chemical Filter

The hydroxyl groups were removed upon exposure to VOF₃ for MgF₂. With the aim to regenerate the chemical filter a two-step process was investigated: the first step consisted of the removal of vanadium with a solvent for V species (cleaning), whereas the second involved a rinsing of the surface. The investigations were performed on a sample with V content of 11,000 ppm after exposure of VOF₃ (ICP data). The extraction rate is given as a function of the cleaning/rinsing agent pair (Figure 5a). A high yield extraction (95%) was achieved with HNO₃ as both cleaning and rinsing agent. It is to note that this removal occurred without significant losses of MgF₂ (only 4 w.%, Figure 5b). After removal of the vanadium, the number of basic sites (OH) must be recovered via the regeneration process. To reach this goal, MgF₂ may be treated with NaOH.

¹⁹F MAS NMR is once again a powerful tool to evidence and quantify the presence of OH groups (Figure 6b) in addition to XRD that proves the unchanged presence of the MgF₂ structure without significant change in Mg(OH)₂ (Figure 6a).

The chemical compositions extracted from the fit of the NMR spectra according to the method described before (Figure 6c) are: MgF_1.79(OH)_0.21, MgF_1.72(OH)_0.28, MgF_1.65(OH)_0.35 as a function of the NaOH concentration, i.e., 0.25, 0.5 and 1 M, respectively. The number of basic sites can then be tailored. In order to nearly recover the initial O/F ratio of 0.25 (MgF_1.6(OH)_0.4), the concentration must be 1 M (O/F = 0.21); the duration of the treatment is 1 h at 60 °C. Two VOF₃ filtering/regeneration cycles were carried out and the change in specific surface area was studied.

Table 2. Change in the specific surface area during two filtering/regeneration sequences.

	Process Step	Starting Material	After Filtering	After NaOH Regeneration
SSABET (m²·g−1)	First run	72	37	64
	Second run	64	38	51

shows the slight decrease in the BET surface after a full filtering/regeneration cycle.

Whereas the specific surface area decreased after the exposure to VOF₃, the initial value was nearly recovered after regeneration with NaOH (1 M concentration, 1 h at 60 °C). Such characteristics prove the possibility of regeneration of the selective filter for both reuse and recovery of vanadium. When the selective filter is used for the purification of UF₆, some uranium species will be present on the surface of the filter; the regeneration aims to remove these species too, underlining its primary importance.

4. Conclusions

Metal fluorides have been investigated as selective chemical filters for the removal of VOF₃, a model gas for volatile fluorides. HSA MgF₂ containing different contents of OH⁻ groups and an oxygen-free MgF₂ with a rather high specific surface area (35 m² · g⁻¹) were tested. It is worth noting that such fluorination synthesis using MgB₂ precursor is reported for the first time. The sorption mechanism identified for MgF₂ consists of a chemical reaction between VOF₃ and Lewis basic sites, i.e., OH⁻ groups. The higher the amount of OH⁻ groups, the higher the quantity of vanadium trapped. Without hydroxyl groups (free-oxygen MgF₂), physisorption is possible but the amount of vanadium is lower than that of a raw and post-fluorinated commercial MgF₂. It was expected that hydroxyl groups are involved in the reaction with VOF₃ but our data evidence physisorption too. Moreover, understanding the sorption mechanism using complementary techniques allowed us to select the most promising selective and regenerable filter. Both the amounts and strength of the Lewis basic sites may be tailored using a post-fluorination treatment with F₂ gas. This versatility opens the route for the selectivity of filtering for mixtures of volatile fluorides according to the Lewis acidity of the target gas, and the present materials can be considered for active materials of gas sensors [40,41,42]. Since the reactions with the impurity to be removed occur at the surface of the HSA MgF₂, the metal fluoride matrix is maintained and controlled regeneration of the hydroxyl groups with treatment in NaOH solution is possible.