Fluorspar to fluorochemicals upon low-temperature activation in water

Innovations with an impact on the large-scale production of chemicals are in demand considering the pressing global challenges faced by the manufacturing industry. These include the availability and management of raw materials, transportation disruptions, complex supply chains and climate change pressure¹⁰. This urgency applies to fluorochemicals, a class of molecules that remain in increasing demand owing to their critical role as pharmaceuticals to cure diseases, agrochemicals for food security and in the production of lithium-ion batteries, which is expected to soar over the coming decade^1,2,3,11. At present, the starting point of the entire fluorine industry is the treatment of fluorspar (fluorite, CaF₂) with concentrated sulfuric acid (H₂SO₄) at high temperature (>300 °C) to produce dangerous hydrogen fluoride (HF)^4,5. Subsequently, HF is used as is, or converted into various fluorinating reagents for the synthesis of fluorochemicals (Fig. 1a). For decades, other routes to HF have been studied, all featuring inorganic acids, such as hydrochloric acid (HCl), and requiring harsh reaction conditions¹². We recently disclosed an activation strategy whereby fluorochemicals as diverse as sulfonyl, benzyl, allyl and alkyl fluorides can be accessed with a fluorinating reagent directly prepared from acid-grade fluorspar (AGF; >97% CaF₂)¹³. This solid reagent obtained by ball-milling CaF₂ with dipotassium phosphate (K₂HPO₄) consists of two crystalline components identified as K₃(HPO₄)F and K_2−xCa_y(PO₃F)_a(PO₄)_b. This mechanochemical reaction stands out because it bypasses the production, storage and complex transport chain of HF. The reactivity of this fluorinating reagent for well-established transformations unavoidably requires independent investigation, and preliminary work has revealed some limitations. One of the challenges that we encountered is its broad application to the synthesis of fluoroarenes frequently used for the production of pharmaceuticals and agrochemicals. Also, mechanochemical reactions require specialized equipment that is available in only a few laboratories. This state of play encouraged the development of an alternative strategy to prepare fluorochemicals directly from fluorspar applying mild conditions for its activation. Departing from solid-state chemistry induced by mechanical energy, we surmise that activation of fluorspar in solution, and specifically in nature’s solvent water, would represent a highly attractive approach (Fig. 1b).

a, Overview of routes to fluorochemicals: (1) current industrial route, (2) preparation of fluorochemicals using phosphate-activated fluorspar, (3) synthesis of fluorinating reagents from fluorspar activated in water with oxalic acid, and B(OH)₃ or SiO₂. (Het)Ar, heteroaromatic. b, Cooperative activation of fluorspar using both a Brønsted acid and a fluorophilic Lewis acid (B(OH)₃ or SiO₂) to prepare fluorinating reagents for the synthesis of fluoroarenes (this work).

In search for a suitable manifold to activate CaF₂ in water, we considered mild acidic conditions that provide direct access to frequently used fluorinating reagents other than HF. We propose that the synergistic use of a Brønsted acid and a fluorophilic Lewis acid would activate CaF₂, with Ca²⁺ capture as an insoluble salt and simultaneous HF capture in a form suitable for subsequent fluorination processes. A report by Lennox and Lloyd-Jones provided guidance by highlighting the beneficial influence of tartaric acid on the equilibrium in solution between aryl boronic acids, potassium fluoride and potassium trifluoroborate salts through precipitation of potassium bitartrate¹⁴. This knowledge prompted us to select a suitable Brønsted acid for AGF activation that enables concurrent Ca²⁺ sequestration as an insoluble salt. Oxalic acid (C₂H₂O₄, H₂Ox) stood out as a suitable candidate because calcium oxalate (CaOx) is poorly soluble in water (0.0061 mg ml^−1 (H₂O, 20 °C))^15,16. This choice was also driven by the appearance of economically viable and environmentally friendly methods for its synthesis, such as bio-based production in fast-growing microorganisms, or through carbon dioxide (CO₂) capture^6,7. The upsurge of interest in oxalic acid results from the ability of this solid organic acid to extract rare-earth elements from primary or waste ores and possibly replace existing recovery processes that depend on inorganic acids such as H₂SO₄ (ref. ¹⁷). Mechanistically, we surmise that the dissolution of AGF in water with oxalic acid in the presence of fluorophilic boric acid (B(OH)₃) may be driven by precipitation of CaOx and immediate capture of HF as a strong B−F bond (732 kJ mol^−1) with water as a by-product¹⁵. We prioritized B(OH)₃ as the Lewis acid prompted by the kinetic studies of Wamser on the reaction of hydrofluoric acid and B(OH)₃ in water¹⁸, and the target Balz–Schiemann reaction for the synthesis of fluoroarenes. Fluoride capture from AGF with B(OH)₃ was found to be highly effective in the presence of H₂Ox when the reaction was carried out in water for 15 h at 50 °C (Fig. 2). The [B−F] products were identified by ¹⁹F and ¹¹B nuclear magnetic resonance (NMR) spectroscopy in deuterium oxide (D₂O; Fig. 3a). By ¹⁹F NMR spectroscopy, HBF₄ (quartet, chemical shift (δ) = −150.3 parts per million (ppm), coupling constant ¹J_B–F = 1.1 Hz) and HBF₃OH (quartet, δ = −145.3 ppm, ¹J_B–F = 10.7 Hz) were identified as the major products¹⁹. In addition, a broad singlet (δ = −152.3 ppm) characteristic of difluoro(oxalate)borate species HOxBF₂ was observed in trace amount (<1%)²⁰. Notably, H₂Ox afforded a higher yield (96%) of [B−F] products than tartaric acid (2%), H₂SO₄ (69%) or HCl (60%) under these reaction conditions (Fig. 2). An investigation on fluorspar activation with a range of Brønsted acids gave insight on the interplay between acidity, denticity and solubility of the Ca²⁺ salt by-product. For monoacids (2 equiv. versus CaF₂), we noted a correlation between acidity and fluoride release in a pK_a (negative logarithm of the acid dissociation constant K_a) range between 5 and –0.5 (Supplementary Table 3). Multifunctional activators (1 equiv. versus CaF₂) were also investigated (Supplementary Table 4). Organic acids leading to five-membered Ca²⁺ chelates stood out with H₂Ox (pK_a1 = 1.3 and pK_a2 = 4.1 (where pK_a1 and pK_a2 are the pK_a values associated with the dissociation of the first and second proton of the diacid, respectively)) being the most effective activator followed by croconic acid (pK_a1 = 0.8 and pK_a2 = 2.2) and squaric acid (pK_a1 = 1.5 and pK_a2 = 3.4). Oxalic acid dihydrate (H₂Ox·2H₂O), which is more cost-effective than H₂Ox, gave the [B–F] products HBF₄ and HBF₃OH with efficacy similar to H₂Ox (total yield of 98%) upon treatment of AGF with B(OH)₃ at 50 °C for 15 h (Supplementary Table 12). Subsequent studies were therefore performed with H₂Ox·2H₂O.

Screening of Brønsted acids for cooperative activation of AGF with B(OH)₃ (2.0 mmol) at 50 °C in water. The yields of HBF₄ and HBF₃OH were determined by ¹⁹F NMR spectroscopy (details in Supplementary Tables 3–6). The first pK_a value (bold) refers to the dissociation of the first proton of the acid, and subsequent pK_a values (non-bold) refer to the dissociation of successive protons. pK_a, negative logarithm of the acid dissociation constant.

a, Preparation of HBF₄ from AGF, B(OH)₃ and H₂Ox·2H₂O with ¹⁹F NMR (D₂O) spectra of reaction mixture containing [B–F] products HBF₄, HBF₃OH and HOxBF₂. b, Preparation of KF, NaF, CsF, Me₄NF and ⁿBu₄NF from AGF, SiO₂ and H₂Ox·2H₂O with ¹⁹F NMR (D₂O) spectra of reaction mixture containing [Si–F] products H₂SiF₆, H₂SiF₅OH and H₂OxSiF₄. The yields of H₂SiF₆, H₂SiF₅OH and H₂OxSiF₄ were determined by ¹⁹F NMR spectroscopy (further details provided in Supplementary Fig. 18). c, Monitoring the reaction of AGF (0.5 mmol) with anhydrous H₂Ox with and without B(OH)₃ by ¹⁹F NMR spectroscopy in D₂O (HF, HBF₄ and HBF₃OH yield quantified by ¹⁹F NMR spectroscopy; details in Supplementary Figs. 6 and 7).

It was also found that H₂Ox·2H₂O was a suitable activator for AGF when combined with silicon dioxide (SiO₂) in water at 50 °C for 15 h, with fluoride release as [Si–F] products (Si–F bond 577 kJ mol^−1) in 97% total yield¹⁵ (Fig. 3b). The formation of H₂SiF₆ (broad singlet, δ = −129.6 ppm), which in water exists in equilibrium with H₂SiF₅(OH) (broad singlet, δ = −128.8 ppm), was evidenced by ¹⁹F NMR spectroscopy of the reaction mixture²¹. Two triplets were also observed (δ = −124.5 ppm and −135.9 ppm, ²J_F–F = 8.9 Hz) and assigned to H₂OxSiF₄ (supported by ²⁹Si NMR spectroscopy; Supplementary Figs. 14 and 15)²². Subsequent studies focused on the synthesis of frequently used fluorinating reagents from these [Si–F] products (Fig. 3b). For this purpose, AGF (1.1 equiv.) was reacted with H₂Ox·2H₂O (1 equiv.) and SiO₂ (0.4 equiv.) in water at 50 °C for 15 h. The reaction mixture was filtered and treated with potassium hydroxide (KOH). The insoluble by-product of this filtration was unambiguously characterized as CaOx·H₂O by powder X-ray diffraction (Supplementary Fig. 17). Neutralization with KOH (2 equiv.) led to the formation of K₂SiF₆ (Supplementary Fig. 20), whereas treatment with excess KOH (6 equiv.) afforded KF^AGF (85% yield, calculated from AGF; purity analysis is included in Supplementary Information)²³. A similar protocol gave NaF^AGF (85% yield, 94% purity) and CsF^AGF (89% yield, 96% purity) from NaOH and CsOH·H₂O, respectively. Alternatively, treatment of the filtered reaction mixture with tetramethylammonium hydroxide (6 equiv.) afforded tetramethylammonium fluoride hydrate, which was converted to tetramethylammonium tert-amyl alcohol fluoride [Me₄NF·(^tAmOH)] ^AGF (88% yield), a fluorinating reagent well documented for nucleophilic aromatic fluorination (S_NAr)²⁴. This strategy also enabled the preparation of tetrabutylammonium fluoride hydrate, which was converted to the bench-stable reagent tetrabutylammonium tetra(tert-butyl alcohol) fluoride [Bu₄NF·(^tBuOH)₄]^AGF (71% yield calculated from AGF)²⁵. Mechanistic investigations by ¹⁹F NMR spectroscopy in D₂O established whether HF is formed upon dissolution of CaF₂ with H₂Ox (Supplementary Fig. 6). In the absence of a fluorophilic Lewis acid, HF is indeed observed (singlet, δ = −166.0 ppm), and an equilibrium is established with the amount of HF plateauing after 3 h at approximately 10% (ref. ²⁶; Fig. 3c). In the presence of either B(OH)₃ or SiO₂, the equilibrium is displaced via the precipitation of highly insoluble CaOx and immediate HF capture by the Lewis acid. Under these conditions, the singlet diagnostic of HF was not detected by ¹⁹F NMR spectroscopy during the entire course of the reaction. As anticipated, the reaction of AGF with fluorophilic Lewis acid in the absence of H₂Ox resulted in no fluoride release (Supplementary Information). These data highlight how Brønsted and Lewis acid cooperativity allows for fluorspar activation under mild conditions, prevents HF from building up, and enables access to HBF₄ (aq.), KF, NaF, CsF, Me₄NF·^tAmOH and ⁿBu₄NF·(^tBuOH)₄, directly from fluorspar.

Further studies compared the reactivity of H₂Ox·2H₂O and H₂SO₄ with AGF. At 50 °C, the combination of AGF, concentrated H₂SO₄ and B(OH)₃ increased the yield of [B−F] products to 97% when the reaction time was extended to 24 h (versus 69% yield after 15 h). This reaction was also effective at 25 °C, offering [B−F] products in 94% yield (average of 2 runs) after a reaction time of 48 h. For comparison, H₂Ox·2H₂O afforded [B−F] products in 98% yield under these conditions (25 °C, 48 h). When SiO₂ served as the fluorophilic Lewis acid, activation of AGF with H₂Ox·2H₂O was effective at 25 °C, albeit requiring a prolonged reaction time of 72 h. After neutralization with KOH, KF was indeed isolated in 74% yield. Replacement of oxalic acid for H₂SO₄ did not give satisfactory results. At 50 °C for 24 h, the reaction of AGF, concentrated H₂SO₄ and SiO₂ gave only partial fluoride release as H₂SiF₆ and H₂SiF₅OH with a total [Si−F] product yield of 48% (quantified by ¹⁹F NMR spectroscopy). At 25 °C for 72 h, the total [Si−F] product yield was reduced to 29%. Notably, the treatment of the [Si−F] solutions with KOH afforded a solid material containing both KF and K₂SO₄ (46% KF (50 °C) and 36% KF (25 °C)) as evidenced by powder X-ray diffraction analysis, two salts that are challenging to separate (KF, 1,020 mg ml^−1; K₂SO₄ 120 mg ml^−1 (H₂O, 25 °C))¹⁵. The formation of K₂SO₄ is indicative of the incomplete reaction between AGF, H₂SO₄ and SiO₂. The overall superior reactivity of H₂Ox·2H₂O compared with H₂SO₄ correlates with the solubility of the calcium by-product formed upon AGF activation (CaOx.H₂O, 0.0061 mg ml^−1 versus CaSO₄.2H₂O, 2.1 mg ml^−1 (H₂O, 20 °C))¹⁵.

With an effective strategy to convert AGF into HBF₄ (aq.), KF and Me₄NF·^tAmOH, we had a duty to demonstrate that these AGF-derived reagents react as expected, with the synthesis of industrially valuable fluoroarenes and a focus on those not accessible via mechanochemical activation of AGF using a phosphate salt. For Balz–Schiemann chemistry, a two-step procedure followed the preparation of HBF₄ (aq.); addition of tert-butyl nitrite to a solution of aryl amine and aqueous HBF₄ led to the precipitation of the corresponding aryl diazonium tetrafluoroborate salt, which was isolated and subsequently heated to liberate the desired fluoroarene²⁷ (Fig. 4a). The protocol was validated first with 4-bromoaniline, AGF-derived HBF₄ (HBF₄^AGF; 1.1 equiv.) and tert-butyl nitrite (2 equiv.). Heating the resulting diazonium salt at 90 °C in chlorobenzene gave 4-bromofluorobenzene in 98% yield (as measured by ¹⁹F NMR spectroscopy), a key intermediate featured in the synthesis of the antidepressant citalopram²⁸. This chemistry was subsequently applied to prepare multiple fluoroarenes frequently used as building blocks in the synthesis of various organo-fluorine-containing drugs in up to 87% yield (dediazoniation yield). Examples include the precursors of lipitor (cholesterol lowering), norfloxacin (antibiotic), raltegravir (HIV), eravacycline (antibiotic), rosuvastatin (cardiovascular disease), flurbiprofen (anti-inflammatory), flunarizine (vertigo) and ezetimibe (cholesterol-lowering drug)^{29,30,31,32,33,34,35,36}. The methodology was also suitable for the preparation of fluoropyridines (4, 8 and 13) that are building blocks for drugs such as MK2 inhibitors (autoimmune diseases) and vericiguat (heart failure), and agrochemicals including the herbicide clodinafop^37,38,39. For diazonium salts prone to decomposition, we developed a one-pot protocol using tert-butyl nitrite and LiBF₄ enabling access to fluoroarenes (4 and 9) without the necessity to isolate the diazonium salt. For this purpose, LiBF₄ was prepared from fluorspar upon treatment of HBF₄^AGF with Li₂CO₃ (Supplementary Information).

a, Fluoroarenes prepared via Balz–Schiemann reaction using AGF-derived HBF₄ [HBF₄]^AGF or metspar-derived HBF₄ [HBF₄]^M(I) (yield of diazotization/yield of dediazoniation). Diazotization reactions were carried out with 5.0 mmol of aryl amine, excluding compounds 3, 4, 6, 8, 9 and 11, which were carried out on 1.0 mmol of aryl amine based on differential scanning calorimetry analysis of the diazonium precursor (Supplementary Tables 29 and 30). Dediazoniation reactions were carried out on a 1.0-mmol scale (unless otherwise stated). All yields are for isolated products (unless otherwise stated). NSAID, nonsteroidal anti-inflammatory drug. b, Fluoroarenes prepared via halogen exchange (S_NAr) of chloroarenes using AGF-derived KF [KF]^AGF or metspar-derived KF [KF]^M(I), or fluorodenitration reactions of nitroarenes using AGF-derived Me₄NF·^tAmOH. Reactions were carried out on a 1.0-mmol scale (unless otherwise stated). All yields are for isolated products. c, Fluoroarenes prepared using AGF-derived HBF₄ for Balz–Schiemann reaction and AGF-derived Me₄NF·^tAmOH for S_NAr reactions. M, metspar; MK, MAP-activated protein kinase. ^a19F NMR yields using 4-fluoroanisole as internal standard. ^bDiazonium salt precursors of 4 and 9 are prepared at 10 °C (maximum process temperature <25 °C). Alternatively, 4 and 9 can be prepared via a one-pot protocol to bypass the isolation of the diazonium salt intermediate (Supplementary Information).

Next, AGF-derived KF (KF^AGF) (90% purity) was found to achieve high-yielding fluorination of chloroarene substrates using Me₄NCl (5 mol%) and DMSO as solvent to provide fluoroarenes 14 to 17 (Fig. 4b). We noted that the performance of KF^AGF was comparable to commercial KF (99% purity; Supplementary Table 22). Aromatic fluorodenitration using AGF-derived Me₄NF·^tAmOH proceeded in DMSO (30−80 °C) to access 2,6-difluorobenzonitrile (18), 4-fluoronitrobenzene (19) and 2-fluorobenzonitrile (20) in high yield.

The successful cooperative activation of AGF in water and its application to the synthesis of fluoroarenes encouraged an investigation on the reactivity of lower-grade metallurgical fluorspar (metspar). These studies were performed with materials sourced from China (Metspar^I CaF₂ (85%), SiO₂ (10%), CaCO₃ (<5%), S (0.12%), P (0.1%)), and Mexico (Metspar^II from CaF₂ (88.98%), SiO₂ (5.43%), CaCO₃ (4.02%), Al₂O₃ (0.41%), Fe₂O₃ (0.24%), S (0.011%), P (0.023%), Pb (<0.001%)). Both metspar^I and metspar^II yielded the [B–F] products HBF₄ and HBF₃OH with an overall yield of 83% each, upon activation with B(OH)₃ and H₂Ox·2H₂O at 50 °C for 15 h (Supplementary Information). The reaction of metspar with H₂Ox·2H₂O and SiO₂ also enabled the preparation of metspar-derived KF (KF^M) (53% (KF^M(I)) and 63% (KF^M(II)) yield, calculated from metspar I or II, respectively). Fluoroarenes 1, 14 and 16 were prepared in good yield using HBF₄^M(I) or KF^M(I) despite the reduced purity of metspar.