Room Temperature Defluorination of Poly(tetrafluoroethylene) by a Magnesium Reagent

Perfluoroalkyl substances (PFAS) are pervasive in the environment. The largest single use material within the PFAS compound class is poly(tetrafluoroethylene) (PTFE), a robust and chemically resistant polymer. Despite their widespread use and serious concerns about their role as pollutants, methods for repurposing PFAS are rare. Here we show that a nucleophilic magnesium reagent reacts with PTFE at room temperature, generating a molecular magnesium fluoride which is easily separated from the surface-modified polymer. The fluoride in turn can be used to transfer the fluorine atoms to a small array of compounds. This proof-of-concept study demonstrates that the atomic fluorine content of PTFE can be harvested and reused in chemical synthesis.


General Experimental
Standard Schlenk line and glovebox techniques were used for all manipulations under an inert atmosphere of dinitrogen or argon unless otherwise stated. NMR scale reactions were performed in J. Young NMR tubes equipped with internal standard capillaries of ferrocene ( 1 H NMR spectroscopy) and prepared in a glovebox. An MBraun Labmaster glovebox was used, operating at <0.1 ppm H2O and <0.1 ppm O2. 1 H, 13 C, 19 F and 11 B NMR spectra were recorded on Bruker 400 MHz or 500 MHz machines, and referenced against SiMe4 ( 1 H, 13 C), CFCl3 ( 19 F) and EtO•BF3 ( 11 B). NMR data were processed using the MestReNova software package.
Infrared spectra were obtained on a Cary630 spectrometer. XPS measurements were obtained on a Thermo Scientific K-Alpha+ X-ray Photoelectron Spectrometer. SEM images were collected using a JEOL 6010LA microscope. MAS-SS-NMR measurements were carried out by Dr. Nasima Kanwal at Queen Mary University London on a Bruker 400 MHz spectrometer. Single crystal X-ray diffraction data were collected using an Agilent Xcalibur PX Ultra A diffractometer. The CrysAlisPro software was used for data collection, as well as peak hunting, indexing reflections in reciprocal space, integration of the raw frames and application of corrections including interframe scaling, Lorentz, flood field and dark current corrections.
The structures were solved using the SHELXT program and least-square refined using the SHELXL program within the Olex2 system suite. 1,2 Solvents were dried over activated alumina from a solvent purification system (SPS) based upon the Grubbs design and de-gassed before use. Glassware was dried for 12 hours prior to use at 120 °C. Benzene-d 6 was de-gassed and stored over 3 Å molecular sieves before use. All reagents were acquired from Sigma Aldrich, Fluorochem or Alfa Aesar and used without further purification unless specified. Where liquids at 25 °C, reagents were dried over activated 3 Å molecular sieves and freeze-pump-thaw degassed prior to use. 1 was prepared following the literature procedure. 3 A range of PTFE products were purchased (Sigma Aldrich, Alfa Aesar) and tested, as per the manuscript, but the optimised reaction is with 1μm particle size PTFE purchased from Sigma Aldrich, used as purchased without any pre-treatment. PVDF was purchased from Sigma Aldrich.

NMR Scale:
In an N2 filled glovebox, 10 mg (0.014 mmol) of 1 and 0.028 mmol of DMAP (140 μL of 0.2 M C6D6 stock solution) was dissolved in 0.5 mL of C6D6, added to a J. Young NMR tube equipped with a ferrocene capillary internal standard, and a t=0 1 H NMR spectrum was recorded. The NMR tube was taken back into the glovebox and PTFE (10 mg, 0.1 mmol repeat unit) was added. The J. Young tube was inverted multiple times and left overnight. After 20 hours, the solution had gone from red to dark brown/red, while the PTFE powder had gone from white to a dark grey. 1 H and 19 F NMR spectra were recorded. The 85 % yield of the product 2 was determined in situ by integral comparison to the ferrocene internal standard in the 1 H NMR spectrum. This yield was cross-referenced by 19 F NMR spectroscopy, through addition of a fluorinated internal standard (1,2-difluorobenzene) at the end of the reaction, and comparison of the integral to that of 2.

Preparative Scale:
In an N 2 filled glovebox, 500 mg (0.7 mmol) of 1 and 171 mg (1.4 mmol) of DMAP were added to a Schlenk flask (with no stirrer bar) and dissolved in 10 mL of C6H6. PTFE (500 mg, was then added. The flask was manually swirled for 15 seconds and left overnight. After 16 hours, the solution had gone from red to dark brown/red. The C 6 H 6 solution was separated from the solid by cannula filtration. The product 2 could be isolated by removal of the solvent in vacuo, in an 85 % yield (591 mg, 0.6 mmol). Single-crystals suitable for X-ray crystallography were grown by slow evaporation of a concentrated C6H6 solution of 2. The dark grey solid polymer product is isolated after cannula filtration, and washed with toluene (3x 5 mL) and hexane (3 Single Crystal X-Ray Crystallography: Figure S1: The X-ray crystal structure of 2 (50 % probability ellipsoids). Hydrogen atoms omitted for clarity.

Successive Defluorination Cycles
Successive defluorination cycles were carried out on the same sample of PTFE, as per the procedure described above. After each reaction, 2 was decanted away from the PTFE sample, which was then washed with toluene (3 x 1 mL) and dried under vacuum, before before being re-subjected to 1 + DMAP. We were able to observe the formation of 2 for 7 successive cycles of defluorination. The yield of 2 decreases throughout each successive defluorination, to a minimum of 5 % in the final cycle.

Equivalents of DMAP:
To investigate how the equiv. of DMAP affected the reaction, 3 parallel reactions with both 1,

Stoichiometry of PTFE
The number average molecular weight of PTFE is in the range 10 6 -10 7 g mol -1 . In an N2 filled glovebox, 10 mg (0.014 mmol) of 1 was dissolved in 0.    Ar F -C IV , para-Ar F -C IV and CF3 not observed.

Isolated reaction of octafluorotoluene and 1 and DMAP:
In an N 2 filled glovebox, 7 mg (0.01 mmol) of 1 was added to a J Young NMR tube and dissolved in 0.5 mL of C6D6, and a t=0 1 H NMR spectrum was recorded. The NMR tube was taken back into the glovebox and octafluorotoluene was added (2.08 μL, 0.014 mmol

Perfluorodecalin:
In an N2 filled glovebox, 50 mg (0.07 mmol) of 1 and 17 mg (0.14 mmol) of DMAP were added to a J Young NMR tube and dissolved in 0.5 mL of C6D6, and a t=0 1 H NMR spectrum was recorded. The NMR tube was taken back into the glovebox and perfluorodecalin was added (3.2 μL, 0.013 mmol). The NMR tube was inverted and left at room temperature for 3 hours.
Then, 1 H and 19 F NMR spectrum were recorded, revealing the formation of 2 in 90% yield, determined in situ by integral comparison to the ferrocene internal standard in the 1 H NMR spectrum.

Perfluoromethylcyclopentane:
In an N2 filled glovebox, 50 mg (0.07 mmol) of 1 and 17 mg (0.14 mmol) of DMAP were added to a J Young NMR tube and dissolved in 0.5 mL of C 6 D 6 , and a t=0 1 H NMR spectrum was recorded. The NMR tube was taken back into the glovebox and perfluoromethycyclopentane was added (5.8 μL, 0.033 mmol). The NMR tube was inverted and left at room temperature for 3 hours. Then, 1 H and 19 F NMR spectrum were recorded, revealing the formation of 2 in 90% yield, determined in situ by integral comparison to the ferrocene internal standard in the 1 H NMR spectrum.
In a N2 filled glovebox, 1 (50 mg, 0.07 mmol) was dissolved in toluene (3 mL) in a scintillation vial, and 2 equiv. of DMAP added (0.14 mmol, 17 mg). The reaction was stirred for 5 minutes before removal of the solvent in vacuo. Single-crystals were grown suitable for X-ray analysis by slow diffusion of n-hexane into a concentrated toluene solution of 1-(DMAP)2.
1-(DMAP)2 has been previously characterised by NMR and X-ray crystallography, and our data is in agreement. 7
In a N2 filled glovebox, 1 (50 mg, 0.07 mmol) was dissolved in toluene (3 mL) in a scintillation vial, and 1 equiv. of DMAP added (0.07 mmol, 8.6 mg). The reaction was stirred for 5 minutes before removal of the solvent in vacuo. Single-crystals were grown suitable for X-ray analysis by slow diffusion of n-hexane into a concentrated toluene solution of 1-(DMAP). Small brown-red plank crystals were successfully mounted but were found to be weakly diffracting, resulting in generally weak data, especially for the high angle diffraction peaks.

1-(DMAP
Nevertheless, connectivity data could be established with an acceptable R1(obs) value of

Reaction with Trimethylsilyl chloride:
In an N2 filled glovebox, 10 mg of 2 (0.01 mmol) was dissolved in C6D6 and added to a J Young NMR tube, 1,2-difluorobenzene was added as an internal standard (0.002 mmol), along with an external standard capillary tube of ferrocene in C6D6. 1 H and 19 F spectra were recorded. To this NMR tube, Me3Si-Cl (12.7 μL, 0.1 mmol) was added. After 1 hour, 1 H and 19 F spectra were recorded, and the product Me3SiF was formed in a quantitative yield, whose data matched those reported in the literature. 8

Reaction with BF3:
In an N2 filled glovebox, 10 mg of 2 (0.01 mmol) was dissolved in C6D6 and added to a J Young NMR tube, along with an standard capillary tube of fluorohexane in C6D6. 1

Reaction to form aluminium fluoride:
In an N2 filled glovebox, 10 mg of 2 (0.01 mmol) was dissolved in C6D6 and added to a J Young NMR tube, 1,2-difluorobenzene was added as an internal standard (0.002 mmol), along with an external standard capillary tube of ferrocene in C6D6. 1 H and 19 F spectra were recorded. To this NMR tube, Dipp BDIAlI2 ( Dipp BDI = (2,6-(i-Pr)-C6H3NCMe)2CH) (7 mg, 0.01 mmol) was added. After 1 hour, 1 H and 19 F spectra were recorded, and the product Dipp BDIAlF2 was formed in a quantitative yield, with data matching that as reported in the literature. 10

XPS:
Data is shown below for the C(1s) and F(1s) XPS scan of PTFE and PTFE-R.

Computational Methods
DFT calculations were run using Gaussian 09 (Revision D.01) 11 using the using the B3PW91 density functional, [12][13][14][15][16] and an ultrafine integration grid (keyword int=ultrafine). 17 Geometry optimisations and frequency calculations were carried out using BS1, while single point frequency calculations were then carried out at BS2 to obtain the final free energies. 18 BS1 was built as follows. Mg centres were described with Stuttgart SDDAll RECPs and associated basis sets, while a hybrid basis set was used for the other atoms: 6-31g**(C, H)/6-311+g*(N, F).
BS2 was built as follows. Mg centres were described with Stuttgart SDDAll RECPs and associated basis sets, while 6-311+g* was used for all other atoms.
BS3 was built as follows. Mg centres were described with Stuttgart SDDAll RECPs and associated basis sets, while 6-311++g* was used for all other atoms.
BS4 was built as follows. Mg centres were described with Stuttgart SDDAll RECPs and associated basis sets, while Αhlrichs triple-ξ basis set def2-TZVPP was used for all other atoms. 19 Geometry optimisation calculations were performed without symmetry constraints. The Gaussian 09 default optimisation criteria were tightened to 10 -9 on the density matrix and 10 -7 on the energy matrix. The default numerical integration grid was also improved using a pruned grid with 99 radial shells and 590 angular points per shell. Frequency analyses for all stationary points were performed using the enhanced criteria to confirm the nature of the structures as either minima (no imaginary frequency) or transition states (only one imaginary frequency).
Single point solvent corrections (benzene, ε = 2.2706) were applied using the polarizable continuum model (PCM) to free energies. 20 Single-point dispersion corrections using Grimme's D3 correction were applied to free energies, with Becke-Johnson damping applied for the B3PW91 functional. 21,22 Intrinsic reaction coordinate (IRC) calculations followed by full geometry optimisations on final points were used to connect transition states and minima located on the potential energy surface allowing a full energy profile (calculated at 298.15 K,

Discussion of Computational Model
DMAP binding to 1 is assumed to be fast and reversible. This assumption is reinforced by the 1 H NMR spectroscopic data at 25 °C of 1-(DMAP) which reveals a broadened set of signals corresponding to 1 symmetrical ligand environment, demonstrating the fluxional behaviour of DMAP, where it is rapidly moving between Mg centres. Previous work by the Jones group came to the same conclusion. 27 In the reaction of 1 with PTFE, 2 equiv. of DMAP are used, as this drives the reaction to the thermodynamic product of 2. However, the reaction of 1 and PTFE also proceeds at room temperature when only 1 equiv. of DMAP is added. This reaction formed a mixture of products, which was resolved to form the singular thermodynamic product of 2 by the addition of a second equiv. of DMAP.
Hence in the computational model, we consider 1-(DMAP) as the active species and zeroenergy point for the reaction with C2F6. The assumption is that DMAP can transfer to a different magnesium species with an insignificant energy penalty generating 1-(DMAP) in situ.
Calculations on a series of isodesmic reactions that suggest that DMAP can exchange between Mg centres in 1 and 2 with only a small energy penalty (Scheme S11). This model is more appropriate than considering the formal association and dissociation of DMAP to Mg atoms, as this method assumes DMAP can be free in solution, which is highly unlikely in practise, and occurs a significant and unrealistic energy penalty (~ 10 kcal mol -1 ).
Scheme S11: Equilibria demonstrating the ease of DMAP transfer between different magnesium species present in the reaction mixture.

Alternative Mechanisms
After establishing the lowest-energy transition state for C-F cleavage of C2F6 with 1-(DMAP), we calculated similar transition states for the same bond breaking by 1 and 1-(DMAP)2 ( Figure   S16). Attack of 1 at C2F6 was calculated via TS-1A (∆G ‡ 298K = 30.2 kcal mol -1 ), and by 1-    We could also support this result through experiment, where addition of the radical trapping reagent 9,10-dihydroanthracene to the optimised reaction for 1 + DMAP + PTFE led to no change, giving the same product 2 in 85 % yield (Scheme S12). 6 Scheme S12: Addition of a radical trapping reagent 9,10-dihydroanthracene to 1 + DMAP did not alter the outcome of the reaction

NBO Data for TS-1, TS-1A and TS-1B
A full NBO analysis was carried out and the relevant NPA charges are tabulated below.

Assessment of the Functional
An assessment of the computational methodology was carried out by a series of functional benchmarking calculations (Table S6). The functionals tested were the Minnesota hybrid-meta functional M06-2X, 17 the long-range corrected functional (with Grimme's D2 dispersion correction) ω-B97XD. 28,29 The same basis set and pseudopotential combination were maintained throughout. Single point solvent corrections (benzene, ε = 2.2706) were applied using the polarizable continuum model (PCM) to free energies. 20 Single point dispersion corrections using Grimme's D3 correction were applied to free energies in the cases of M06-2X and ω-B97X (noting that ω-B97X-D has Grimme's D2 dispersion correction built into the functional). 21 Consistent results were found across the different functionals. The chosen methodology, with B3PW91 as the functional, follows previous work in our group for C-F bond cleavage using Mg-Mg nucleophiles. 30 It is noted that the functional benchmarking was carried out using the basis-set package BS4.  Free-energies in kcal mol -1 . BS3 was used for this functional assessment.

Assessment of Basis Sets
An assessment of the computational methodology was carried out by a series of basis set benchmarking calculations (Table S7)