Unlocking the reducing power of metal nitrides by mechanochemical hydrodefluorination of fluorinated compounds

Abstract

The carbon-fluorine (C–F) bond, present in both aromatic C(sp²)–F and aliphatic C(sp³)–F moieties, possesses an exceptionally high bond energy. This inherent stability renders C–F bond activation highly challenging. In this work, we address this challenge and reclaim valuable hydrocarbons from fluorinated compounds, including aryl fluorides, alkyl fluorides, and perfluoroalkyl substances, by developing a mechanochemical hydrodefluorination strategy. This approach exploits metal nitrides, particularly magnesium nitride, as reductants, contrasting with previous perception of metal nitrides as merely ammonia surrogates in organic reactions. Remarkably, this mechanochemical protocol achieves the degradation of even persistent polytetrafluoroethylene (Teflon®), a prototypical “forever chemical”.

Environmental persistence of organofluorines, stemming from the high bond energy of C–F bonds and resultant chemical inertness, poses substantial environmental safety concerns. Here, the authors reclaim valuable hydrocarbons from fluorinated compounds, including aryl fluorides, alkyl fluorides, and perfluoroalkyl substances, by developing a mechanochemical hydrodefluorination strategy.

Introduction

Organofluorines are vital in pharmaceuticals, agrochemicals, and materials science due to their unique properties, including enhanced lipophilicity, high thermal stability, metabolic stability, and distinct biological activity^1–5. However, environmental persistence of organofluorines, stemming from the high bond energy of C–F bonds (Fig. 1a) and resultant chemical inertness, poses substantial environmental safety concerns^6,7. Owing to the high bond strength of C–F bonds, organofluorines exhibit highly negative reduction potentials^8,9; for example, the standard reduction potential of fluorobenzene approaches –3.0 V (versus saturated calomel electrode (SCE)). This makes direct reduction of organofluorines very difficult. To date, most methods for cleaving C–F bonds in perfluoroalkyl substances (PFASs), which include electrochemical^10,11, photochemical^12–14, mechanochemical^15–19, and base-mediated degradation^20,21, destruct the organic skeletons, precluding reclamation of valuable organic products. Moreover, hydrodefluorination of fluoroarenes remains underdeveloped, with its substrate scope largely confined to polyfluoroarenes or CF₃-containing molecules^22–27. Consequently, developing a general method for C–F bond activation applicable to alkyl and aryl fluorides, perfluoroalkyl substances, and fluoropolymers, while simultaneously yielding reusable hydrocarbon products, remains a daunting task¹⁴.

Fig. 1. Properties of organofluorines and Mg₃N₂.

a Bond dissociation energy of C-X bond. b The potential of Mg₃N₂. c Hydrodefluorination of fluorinated compounds using Mg₃N₂ via mechanochemistry.

Metal nitrides, particularly those of transition metals, find broad applications in academia and industry^28–31. For example, transition metal nitrides function as efficient electrocatalysts^32,33 and electrode materials for lithium-ion and sodium-ion batteries³⁴. Among main-group metal nitrides, lithium nitride (Li₃N) serves as a solid-state electrolyte³⁵, while magnesium nitride (Mg₃N₂) acts as a reactive additive³⁶ for all-solid-state lithium metal batteries. Mg₃N₂ (Fig. 1b), calcium nitride (Ca₃N₂), and aluminium nitride (AlN) are effective ammonia surrogates in few organic transformations in the presence of water and/or protic solvents^37–40. The N³⁻ anion in these metal nitrides exhibits high electron density and is intrinsically potent electron donor. Nevertheless, the use of these materials as reducing agents has yet to be reported (Fig. 1b).

Owing to its high lattice energy, Mg₃N₂ is insoluble in organic solvents, fundamentally restricting its interaction with dissolved organic substrates. This inherent limitation makes mechanical milling a particularly effective approach for probing the reactivity of Mg₃N₂ under solvent-free conditions. Notably, mechanochemistry has enabled inorganic compounds like potassium phosphate and barium titanate to exhibit remarkable and often unexpected reactivity in organic transformations^18,41,42. Here, we harness ball-milling technique to unlock the reductive capability of Mg₃N₂ and related main-group metal nitrides, targeting reductive hydrogenation of C–F bonds in organofluorines (Fig. 1c) and degradation of polytetrafluoroethylene (PTFE, Teflon®).

Results and discussion

Reaction optimization

Using 1-fluoro-3,5-dimethoxybenzene (1a) as the model substrate, we optimized the reaction conditions (Fig. 2). To prevent potential deactivation of Mg₃N₂ by water, tetrahydrofuran (THF) was employed as a diluent, with a THF/H₂O mixture (v/v = 9/1, 40 μL) serving as the proton source. Initially, milling a mixture of 1a (0.20 mmol), Mg₃N₂ (3.0 equiv), and the THF/H₂O mixture in a stainless-steel jar (5 mL) with 4 stainless-steel balls (5 mm in diameter) at 35 Hz under ambient atmosphere for 3.5 h afforded the hydrodefluorinated product 2a in 32% yield, alongside with 37% of unconverted 1a. Adding LiOH (1.5 equiv) slightly increased the yield of 2a to 54%. Interestingly, replacing LiOH with NaOH or KOH significantly improved the yield to 89% and 91%, respectively. In contrast, other potassium-based alkaline salts proved less effective, yielding inferior results. To our satisfaction, 2a could still be obtained using pure H₂O as the proton source. While milling in THF alone gave 2a in 70% yield, the lack of deuterated product formation with THF-D₈ (Supplementary Fig. 11) indicated that the residual water in THF served as the operative proton source. Altering the THF/H₂O ratio did not improve yields, whereas omitting the THF/H₂O mixture drastically reduced the yield.

Fig. 2. Optimization of Hydrodefluorination of 1a.

Yields of 2a and unconverted (unconv.) 1a were determined by ¹H NMR spectra using CH₂Br₂ as internal standard.

Beyond Mg₃N₂, we evaluated several other commercially available main-group metal nitrides for their efficacy in the hydrodefluorination reaction. However, all proved significantly less effective than Mg₃N₂. To probe this disparity in performance, we assessed the stability of each metal nitride toward ambient air by monitoring color evolution (Supplementary Fig. 1). Ca₃N₂, also a Group IIA metal nitride, afforded only 9% yield of 2a with substantial recovery of unreacted 1a, attributed to rapid hydrolytic deactivation during milling. Although Li₃N exhibited air stability comparable to Mg₃N₂ and could degrade fluorinated compounds, it failed to produce 2a in high yield. In contrast, covalent aluminium nitride (AlN) showed insufficient reactivity due to its high stability, resulting in recovery of nearly all unreacted starting material. Replacing Mg₃N₂ with aqueous ammonia (25%) yielded no desired product, suggesting that NH₃ itself was not involved. Finally, a thermal control reaction conducted in THF (1.0 mL) at 65 °C for 24 h resulted in no conversion of starting material or no formation of 2a, underscoring the essential role of mechanical activation in this transformation.

Substrate scope

With the optimized reaction conditions in hand, we explored the substrate scope of this mechanochemical hydrodefluorination of organofluorines (Fig. 3). Reaction times varied depending on the electronic properties of the substrates. Firstly, electron-rich dimethoxyfluorobenzenes (1a and 1b) afforded 1,3-dimethoxybenzene in 91% and 85% yields, respectively. Ortho-, meta-, and para-methoxyfluorobenzenes (1c–1e) delivered anisole in 81–85% yields, while 1-fluoro-4-phenoxybenzene (1f) gave diphenyl ether in an isolated yield of 81%. Amine-containing substrates performed efficiently. Ortho- and para-aminofluorobenzenes (1g and 1h) furnished N,N-dimethylaniline in 88% and 93% yields, respectively. When 4-fluoro-N,N-diphenylaniline (1i) and 1-(4-fluorophenyl)pyrrolidine (1j) were employed, triphenylamine and 1-phenylpyrrolidine were isolated in 84% and 65% yields, respectively. Alkyl-substituted fluorobenzenes were generally effective substrates. Ortho-, meta-, and para-fluorotoluenes (1k–1m) afforded toluene in moderate yields of 70–74% due to its volatility. In contrast, dimethylated (1n) and trimethylated (1o) analogs provided m-xylene (86%) and mesitylene (93%) in improved yields. Notably, isopropyl- and benzyl-substituted fluorobenzenes (1p and 1q) generated cumene and diphenylmethane in excellent yields of 92% and 99%. For fluorinated biphenyl scaffolds (1r–1t), biphenyl and 4-methylbiphenyl were isolated in 71–95% yields. The pyridine-containing fluorobenzene (1u) was tolerated and give 2-phenylpyridine with a slightly reduced yield of 78%. Both 1-fluoronaphthalene (1v) and 2-fluoronaphthalene (1w) worked successfully and gave naphthalene in 80% and 60% yields, respectively. Heterocyclic fluoroarenes, including both electron-rich fluoroindole (1x) and electron-poor fluoroquinoline (1y), reacted smoothly to generate N-benzylindole (82%) and 2-methylquinoline (56%). For electron-withdrawing substrates 1z–1ac bearing a ketal, amide, acid or ester group, the reaction rates were slower. To promote the conversion, the number of milling balls was increased. Under these conditions, the defluorinated products were obtained in yields ranging from 30 to 52%. The products derived from 1ab and 1ac were both obtained as ionic benzoate salts. In the case of 1ab, this resulted from direct neutralization of the carboxylic acid group with the added base, whereas for 1ac, it was due to hydrolysis of the ester followed by neutralization. We next examined difluoroarenes (1ad–1ak), requiring double the amount of Mg₃N₂ and H₂O to address the higher fluorine content. Consistent with earlier observations, organofluorines 1ad–1ah produced volatile anisole and toluene in 72–79% yields. To our delight, the yield of diphenylmethane from bis(4-fluorophenyl)methane (1ai) was high (95%). Moreover, para- and ortho-difluorobiphenyls (1aj and 1ak) could afford biphenyl in isolated yields of 71% and 89%, respectively.

Fig. 3. Substrate scope.

Conditions: A mixture of 1 (0.20 mmol), Mg₂N₃ (3.0 equiv, based on the amount of “F” atom), KOH (0.30 mmol), and THF/H₂O (v/v = 9/1, 40 μL) was milled in stainless-steel milling jar (5 mL) using four stainless-steel balls (5 mm) at 35 Hz, isolated yields. ^aYields were determined by ¹H NMR spectra. ^bEight milling balls. ^c4.0 equiv of Mg₂N₃. ^d3.0 equiv of KOH. ^eTHF/H₂O (v/v = 9/1, 80 μL). ^f1 (0.10 mmol) and THF/H₂O (v/v = 9/1, 100 μL).

Furthermore, our methodology could be extended to hydrodefluorination of alkyl fluorides. Difluorinated piperidine derivatives (1al and 1am) were converted to the corresponding piperidines in 78% and 69% yields, respectively. Arenes bearing trifluoromethyl group (1an–1aq) were compatible and afforded methylated arenes in 59–86% yields. Next, we applied the reaction to PFASs with C₃F₇, C₄F₉, and C₆F₁₃ chains. To mitigate volatility concerns, we selected several PFASs bearing methoxy groups (1ar–1au). To our delight, all examined PFASs underwent hydrodefluorination, providing defluorinated products in 27–45% yields, albeit accompanied by significant substrate degradation. These results are significant given the extreme rarity of PFAS hydrodefluorination^14,43. We also attempted the reduction of perfluorohexanoic acid (1av) and methyl perfluorooctanoate (1aw). However, no hydrodefluorinated products were detected, suggesting that even if C–F bond cleavage occurred, the resulting intermediates from PFASs without an aryl moiety underwent complete degradation. Substitution of Mg₃N₂ with Li₃N enabled aqueous-phase ¹⁹F NMR detection of free fluoride ions (Supplementary Fig. 9), confirming that C–F bond cleavage had indeed occurred. We further expanded the scope of this method to include the reduction of C–Cl (3a–3i) and C–Br (3j–3s) bonds in organochlorines and organobromines, achieving yields from 55% to 99% (Supplementary Fig. 3). Given the lower bond dissociation energies (BDEs) of C–Cl and C–Br bonds, the reaction conditions were adapted by modifying the base. Consequently, this protocol enabled rapid and selective removal of the weaker C–X bonds in substrates with multiple halogens on the same aromatic ring (3t–3v).

To evaluate the reduction capability of metal nitrides toward fluorinated polymers, PTFE was ball-milled with Mg₃N₂ under optimal conditions (Method A). Fourier transform infrared (FT-IR) analysis of the resulting black powder (PTFE mix^MF) revealed the disappearance of characteristic PTFE stretching vibrations at 1150 and 1210 cm⁻¹ (Supplementary Fig. 6)¹³, indicating cleavage of C–F bonds. However, C–H vibrational peaks were very weak, suggesting that little, if any, polyethylene was formed. Owing to the poor solubility of magnesium fluoride (MgF₂), its yield was not quantified. Treatment of the black powder (PTFE mix^MF) with acid (1 M HCl) yielded an insoluble residue. Raman spectroscopy of this residue identified bands at 1314 cm⁻¹ (D-band) and 1577 cm⁻¹ (G-band), consistent with disordered graphitic carbon (Fig. 4a)¹⁸. Parallel experiments with Li₃N as reductant (Method B) produced a dark powder (PTFE mix^LF), which contained lithium fluoride (LiF). ¹⁹F NMR analysis of the water-soluble fraction showed a single fluoride signal at –122.3 ppm (78% yield, Fig. 4c)¹³. The mechanochemically generated LiF could also be reused as a nucleophilic fluorinating agent, like KF¹⁹. Raman analysis of the water-insoluble residue from PTFE mix^LF similarly displayed characteristic carbon bands at 1358 cm⁻¹ (D-band) and 1574 cm⁻¹ (G-band) (Fig. 4b).

Fig. 4. Degradation of PTFE using Mg₃N₂ and Li₃N.

a Raman spectroscopy (Method A). b Raman spectroscopy (Method B). c Aqueous-phase ¹⁹F NMR spectroscopy of fluoride ion (Method B). ^aDisordered (D) band and graphitic (G) band are labelled.

Mechanistic investigation

Given that Mg₃N₂ is synthesized from nitrogen gas and magnesium metal (Mg), and considering the strongly negative reduction potential (–2.61 V versus SCE) of Mg, we examined whether Mg could effectively achieve hydrodefluorination. As shown in Fig. 5a, reduction of the C(sp²)–F bond in 1a yielded only 46% of product 2a. When Mg was employed to reduce C(sp³)–F bonds in 1at (9 equivalents relative to fluorine atoms), no hydrodefluorinated product was detected, and only 13% of the starting material was recovered. These results clearly demonstrate that Mg has significantly lower efficiency compared to Mg₃N₂. Consequently, it can be concluded that Mg₃N₂ outperforms Mg in hydrodefluorination and that the reducing capability of Mg₃N₂ stems from nitride ions rather than from decomposition into metallic Mg during milling.

Fig. 5. Investigation of mechanism.

a Mg as reductant. b Radical-coupling experiments. c Control experiments. d Proposed reaction mechanism.

To elucidate the reaction mechanism, deuteration experiments were performed using THF-D₈ as the deuterium source (Supplementary Fig. 11). ¹H NMR analysis revealed negligible deuteration (<1%) with THF-D₈, confirming that residual water served as the hydrogen origin under open-air conditions. We therefore propose that any aryl radicals generated in situ would be rapidly reduced to aryl anions by excess Mg₃N₂. To test this hypothesis, 1a was milled with Mg₃N₂ alone (Fig. 5b). Notably, ¹H NMR analysis detected a small quantity of the homocoupled product 5a. Furthermore, when pyrrole was introduced as a radical scavenger in the absence of proton source, the coupling product 6a was also observed by ¹H NMR. Together, these findings confirm the formation of aryl radicals during the milling process. To better understand the reaction mechanism, control experiments with free radical scavengers were performed. Under the optimal reaction conditions, the addition of 3.0 equiv of either 2,2,6,6-tetramethylpiperidine-1-oxy (TEMPO) or butylated hydroxytoluene (BHT) completely inhibited the reaction, leaving 90% and 93% of the starting material unreacted, respectively (Fig. 5c).

Proposed mechanism

Based on the experimental results and literature precedents^14,44,45, a plausible reaction mechanism is outlined in Fig. 5d, using aryl fluoride as an example. The process begins with KOH-mediated activation of Mg₃N₂ under ball-milling conditions. This is followed by single electron transfer (SET) reduction of the aryl fluoride, forming an anion radical intermediate I. Subsequent C–F bond cleavage generates aryl radical II and fluoride ions (F⁻), where the released F⁻ combines with Mg²⁺ to form insoluble MgF₂. Under reductive conditions, aryl radical II undergoes rapid reduction to aryl anion III, which is then protonated by H₂O to yield the hydrodefluorination product. Notably, this protonation step regenerates KOH in situ, explaining why the amount of KOH remains constant, even when the quantities of the C–F bonds in organofluorines and Mg₃N₂ are increased.

In summary, we have developed a mechanochemical strategy for the hydrodefluorination of fluorinated compounds, offering a sustainable route to recover valuable hydrocarbons. Departing from the conventional view of metal nitrides as merely ammonia surrogates, this approach leverages their reductive capability under ambient air and at room temperature. Our methodology demonstrates broad applicability, being equally effective for both aryl and alkyl fluorides and extendable to PFASs. Significantly, our protocol achieves degradation of even persistent PTFE using Mg₃N₂ or Li₃N. Furthermore, this work addresses longstanding challenges related to the inherent stability and solubility of Mg₃N₂, representing its first successful application as a reducing agent. We intend to further explore the properties of these long-neglected metal nitrides to unlock their potential in modern synthetic chemistry.

Methods

General procedure for the hydrodefluorination of organofluorines

Organofluorine 1 (0.20 mmol), Mg₃N₂ (3.0 equiv based on the amount of “F” atom), KOH (0.30 mmol), and THF/H₂O (v/v = 9/1, 40.0‒100.0 μL) were added quickly in a dried 5 mL stainless-steel milling jar with four or eight stainless-steel balls (5 mm, diameter) under air. The reaction mixture was milled at 35 Hz for 2‒5 h on a HMD400 mixer mill at room temperature. After the reaction was complete, the resulting mixture was combined by washing with ethyl acetate. Then, the reaction mixture was filtered through a silica gel plug with ethyl acetate as the eluent, and evaporated in vacuo. The residue was separated by flash column chromatography on silica gel with ethyl acetate/petroleum ether as the eluent to afford product 2.

For low boiling-point products, the reaction mixture was dissolved in CDCl₃, followed by the addition of CH₂Br₂ (7.0 μL, 0.10 mmol) as the internal standard after the reaction was completed. Yields of the desired products were then calculated using crude NMR spectra.

Author contributions

G.-W.W. supervised the project. G.-W.W. and J.-S.C. proposed the concept and wrote the manuscript. J.-S.C., L.-F.G., and H.P. performed the experiments and collected the data.

Peer review

Peer review information

Nature Communications thanks Hua-Jian Xu, and the other, anonymous, reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data from this study are provided in the published article and its Supplementary Information file. Detailed experimental procedures, NMR, and other analysis data can be found in the Supplementary Information. For further inquiries, data can be obtained by contacting the corresponding author.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-70813-5.