Introduction

Per- and poly-fluoroalkyl substances (PFAS) find broad use in automotive, semiconductor, manufacturing, and energy production industries1. A problem with this use has been that the large number of C–F bonds make PFAS highly resistant to natural degradation after discard, leading to their accumulation in the environment over time2,3. As such, PFAS are often colloquially referred to as “forever chemicals”. Perfluorooctanoic acid, (PFOA), one of the most common PFAS, is of particular concern because of adverse human health effects4,5. It is present at significant concentrations in both industrial and consumer products, including in nonstick cookware, firefighting foams, and textiles6,7.

Concerns regarding this environmental stability have arisen as more extensive research has been conducted into PFAS toxicity. PFOA has been detected in up to 95% of the U.S. human population8. Studies have generally linked PFOA exposure to liver, kidney, and testicular cancer, thyroid disease and related issues, decreases in fertility, and increased adipose tissue in children4,5,9. Because of these potential health risks, and the massive global bioaccumulation occurring both within humans and the ambient environment, the development of ex situ PFOA degradation methods has significantly increased in recent years.

Harsh conditions are often necessary to destroy PFOA as a result of the strength of its many C-F bonds. Industrially, PFOA has been degraded using incineration at high temperature (650–1000 °C) in the presence of hydrocarbons, producing primarily HF and CO26,10. However, incineration produces hazardous and volatile shorter-chain PFAS byproducts11, and often does not even completely eliminate the initial PFOA. This can lead to environmental pollution through leachate and ash disposal12. Non-incineration PFOA degradation methods have been developed recently, but these often employ complex organometallic photocatalysts and/or electrochemistry7,13.

At laboratory scale, various combinations of electrochemistry7,14,15,16, advanced oxidation processes (including ozonation)17,18, and sonication19,20 have previously been employed to degrade PFOA, among other techniques3,6,21. Common pitfalls of such methods include high costs, high energy input requirements and high temperatures, use of toxic metals, and production of smaller-chain PFAS byproducts. Photochemical methods are being introduced as milder alternatives to these processes22,23,24.

It was previously demonstrated that Br2 exposed to intense focused sunlight is capable of attacking carbonyl bonds in CO2, leading to the formation of transient C–Br bonds25. We hypothesized that photoactivated Br2 might also be capable of attacking the C-F bonds present in PFAS. We have now confirmed this by demonstrating that either pure Br2, or the combination of MnO2, concentrated H2SO4, and NaBr, can degrade PFOA and even PTFE photolytically under focused solar irradiation. Substitution of less-hazardous NaBr for liquid Br2 is possible because MnO2 rapidly oxidizes Br− to Br2 in situ in concentrated H2SO4. Br2 is furthermore hypothesized to be reduced during perfluorocarbon degradation (oxidation), cyclically re-forming Br−.

Our best results have used MnO2 as oxidant, H2SO4 as solvent, and Br2 as photocatalyst, along with focused sunlight, to decompose PFOA to CO2 and inorganic fluoride. The only spectroscopically identified inorganic fluoride product species so far is HF, although there appears to be significantly more inorganic fluoride also present as manganese salts, insoluble in ethanol or water, that have not been amenable to characterization. The detected yields of CO2 and HF have so far reached 10–15% and 4%, respectively.

The reagents and apparatus to carry out this degradation are inexpensive and readily obtained, especially because NaBr can be substituted for pure Br2 as the sole source of bromine. Our results demonstrate the ability to degrade PFOA in a simple, controlled manner at lower temperatures than current industrial methods. The same products were obtained when polytetrafluoroethylene (PTFE) was substituted for PFOA as well, suggesting this process can potentially be used to degrade a wide range of PFAS.

Results

We have carried out test reactions on a scale of 25–70 mg for several different sets of PFAS and reagents, as well as on controls lacking one or more of the chemical ingredients, in sealed 1-mL borosilicate ampoules. The PFAS was either PFOA, PTFE, or a combination of both. These ampoules were positioned in direct sunlight at the focal point of an inexpensive 15-cm-dia paraboloid mirror (Supplementary Fig. S1) for 0.5–2 min periods to keep the temperature below ~ 100ºC, alternating with cooling in an ice bath. Total illumination times ranged from 1.5 to 12 h.

Visual detection of photoreaction

After sealing the prechilled ingredients in a glass ampoule and then allowing the ampoule to be heated by focused sunlight, the initial changes in visual appearance of the samples included the formation of orange Br2 vapor and melting of the powdery PFOA to form a clear oily layer on top of the H2SO4. These changes can be attributed solely to heating, because similar changes could be obtained simply by heating the samples to 100ºC.

Subsequent changes in appearance, although gradual, were eventually striking and could not be induced perceptibly simply by incubating the sample at 100ºC for many days. Figure 1 shows photographs of sample ampoules without illumination, and after 4 and 12 h of focused illumination.

Fig. 1
figure 1

Visual appearance of the progressing photochemical reaction. A photographic comparison of the appearance of two samples in glass ampoules, both of which initially contained PFOA, MnO2, H2SO4, and NaBr as the sole initial source of bromine. One ampoule (left) was a control that had never been illuminated but rather remained refrigerated for 8 weeks. This limited the oxidation of the NaBr to form orange colored Br2, and also prevented the destruction of the PFOA, which is still visible as a clear/white residue on some areas on the inside glass surface. A different ampoule containing the same quantities of starting materials (center) had been illuminated cumulatively 4 h in the 3 days prior to taking this photograph, and was then refrigerated. It shows no evidence of remaining solid PFOA, but still shows a significant quantity of orange colored Br2. Another image of the same ampoule after a further 4 h of illumination (right) shows the formation of a distinctly blue photoproduct, particularly at the site of the most recent focusing of the sunlight for a 1-min interval. Additional blue-gray solid formed during prior illuminations is also clearly visible at the bottom of the H2SO4 layer. This sample was photographed immediately after a period of 1 min of focused solar illumination, and was too hot to touch. It was therefore left in the piece of Tygon tubing used to hold it.

Full size image

Spectroscopic detection of partially-degraded fluorocarbons and CO2

Figure 2 shows 19F NMR spectra, measured in deuterated chloroform (CDCl3), of unreacted PFOA (top) vs. the organic-soluble residue collected after several h of photoreaction (bottom). For the latter, the solid residue after photoreaction was first extracted with water, and then the remaining water-insoluble solid residue was extracted into ~ 1 mL CDCl3 and transferred into an NMR tube. The unreacted PFOA (Fig. 2, top) shows a number of peaks (a–g) with complex splitting patterns due to the many perfluorinated carbons in its chain, matching previously published PFOA spectra26. Additional peaks in the bottom spectrum (α, β, γ) were detected after several hours of solar illumination, clearly demonstrating the presence of partially decomposed fluorocarbon intermediates formed by the Br2-sensitized photoreaction.

Fig. 2
figure 2

19F NMR spectra measured at 400 MHz in CDCl3, showing PFOA before the photoreaction (top) and organic-soluble materials present after ~ 1 h photoreaction (bottom). Peak assignments a-g are based on previous work26. Neat PFOA (top): 19F NMR (CDCl3, 400 MHz) δ − 81.27 (tt, 3F, J = 10.4, 0.4 Hz), − 119.39 (tt, 2F, J = 13.1, 2.0 Hz), − 122.09 (s, 2F), − 122.50 (s, 2F), − 123.19 (m, 4F), − 126.61 (m, 2F). PFOA after photoreaction (bottom): 19F NMR (CDCl3, 400 MHz) δ − 64.12 (tt, 1F, J = 15.2, 2.7 Hz) − 81.27 (m, 3F), − 117.80 (tt, 1F, J = 16.4, 3.2 Hz), − 119.58 (tt, 2F, J = 13.4, 2.8 Hz), − 121.55 (m, 1F), − 122.14 (s, 2F), − 122.48 (s, 3F), − 123.24 (m, 5F), − 126.61 (m, 3F).

Full size image

When photoreaction ampoules were broken open after even ~ 12 h of illumination, room-temperature IR spectra of the gases released (Fig. 3, green trace) showed clear evidence of IR absorption due to C-F stretching vibrations near 1250 cm−1, in addition to somewhat larger CO2 bands near 2200–2400 cm−1. This spectrum corresponded to a nearly 50-fold dilution of the headspace gases in the pressurized ampoule, as there was a ~ fourfold expansion of the ~ 1-mL headspace upon breaking open the ampoule, then only ~ 3 mL of the resulting expanded gases were injected into an IR gas cell already containing ~ 50 mL argon.

Fig. 3
figure 3

FTIR spectrum of the gases released upon breaking open the photoreaction ampoule shown in Fig. 1 (green trace), along with spectra of 1, 2, and 4 mL injections of 100% CO2 (black traces). For the headspace sample (green trace), 3 mL out of the 4 mL total volume released upon breaking open the refrigerated photoreaction ampoule into a glass syringe, was injected and diluted into an Ar-filled gas cell. This gas cell had 100-mm path length, internal volume ~ 50 mL, was made of borosilicate glass, and was fitted with BaF2 windows and PTFE seals. The spectrum was measured using a Thermo Magna series FTIR instrument, (2 cm−1 resolution, 500 scans, 20 min), immediately following an otherwise-identical Ar-only background spectrum from the same gas cell. For the standard spectra of 100% CO2 shown in black, the indicated cumulative volumes of 100% CO2 had been injected after obtaining a background spectrum with 100% Ar. The green trace for the 3-mL injection of headspace gas coincidentally matched nearly perfectly with the black spectrum measured for the first (1-mL) injection of CO2. Plotted within is the absorbance value at 2360 cm−1 (A2360) vs injected volume of CO2, taken from the green 3-mL headspace spectrum and the spectra of 1, 2, and 4 mL injections of 100% CO2 (black points). Additional A2360 values are also plotted here for 8- and 16-mL injections of 100% CO2 (black), as well as for a 1-mL aliquot from a different 6-mL headspace volume, from a similar 12-h-photoreacted ampoule similar to that shown in Fig. 1 (green). Full spectra corresponding to all seven plotted A2360 values are available in the Supplementary Information.

Full size image

The green spectral trace in Fig. 3, obtained after injection of 3 mL of reaction headspace gas, shows CO2 absorption bands that nearly match those of a control injection of 1 mL of 100% CO2 (nearly superimposable black trace). The relative volume measurements are imprecise, especially for the 1-mL volume, because of dead space in the connection tubing between the glass syringe and the gas cell. Propagating this uncertainty, we can conclude that the headspace gas contained 25–35% CO2 by volume. In the 4 mL of total recovered volume, this corresponded to 1–1.5 mL (or 0.04–0.06 mmol/2–3 mg) CO2, equivalent to 0.5–0.7 mg C. This in turn corresponds to 8–12% conversion to CO2 of the original ~ 5 mg C in the 23-mg (0.055 mmol) PFOA sample. The sample that produced the other green data point in Fig. 3 generated 6 mL of headspace gas (Supplementary Fig. S2), corresponding to a carbon recovery of 12–18% from an original sample of 28 mg PFOA.

The relatively large magnitude of the strong C–F absorptions in this room-temperature spectrum (Fig. 3), and the absence of any H–F rotation-vibration absorption bands near 4000 cm−1, is most consistent with almost all of the ~ 70% of the gas in the reaction headspace that is not CO2 being low-molecular-weight fluorocarbon(s)27. Such a high fluorocarbon vapor pressure is inconsistent with unreacted starting material. PFOA has insufficient vapor pressure at room temperature to give rise to these IR bands, as confirmed by a control measurement with solid PFOA equilibrated inside the gas cell, for which the gas-phase C–F absorption bands at room temperature were below the level of detection (0.001 absorbance unit). The IR results in Fig. 3 thus indicate that the partially decomposed condensed-phase fluorocarbons detected after ~ 1 h of photoreaction (Fig. 2) likely give rise to further-decomposed small volatile fluorocarbons after 8 h additional illumination.

With the goal of identifying the decomposed fluorocarbons in the headspace giving rise to the C–F stretch vibrations in Fig. 3, we allowed 0.25 mL of the headspace gas to equilibrate with methanol, and then performed a liquid-chromatography-mass-spectrometry (LCMS) analysis. This resulted in only 1 identifiable solute peak from the liquid chromatogram, with one clearly identifiable m/z peak at 131 (see Supplementary Information). This mass matches most closely the value expected for a C3F5− ion.

When the glass photoreaction ampoules were broken open after 8 h of illumination, it was possible to obtain confirmation of the presence of CO2 in the released gases by trapping them with a 0.02 M Ca(OH)2 solution, collecting a powdery white precipitate, and measuring the powder X-ray diffraction spectrum (Fig. 4). The measured 1–2 mg mass of the recovered dry sample, containing 0.1–0.2 mg C, represented ~ 1% recovery of carbon from the original ~ 50 mg sample of PFOA, containing 12 mg C. The XRD indicated no significant presence of crystalline CaF2 or CaSO4 within the sample, as these would have shown significant additional diffraction peaks. From this we conclude that, while HF is formed during the photoreaction (see Fig. 5 below), it remains in condensed phase so that no significant amount of gaseous HF is released when the ampoules are broken open.

Fig. 4
figure 4

CaCO3 XRD measurement. Powder X-ray diffraction (XRD) measured from the solid precipitate formed when 5–10 mL of compressed gas from a photoreaction vial was allowed to mix with 10 mL of aqueous 0.02 M Ca(OH)2, and the white solid formed was centrifuged and dried.

Full size image
Fig. 5
figure 5

19F NMR spectra measured at 400 MHz in D2O. A, photoproducts from PFOA + PTFE/MnO2/NaBr; B, photoproducts from PFOA/MnO2/NaBr; C, photoproducts from PTFE/MnO2/NaBr; D, photoproducts from PFOA/MnO2/Br2; E, NaF standard (2 mg/mL) added to 10% H2SO4 in D2O. As an internal standard, TFA (− 76.55 ppm) was added to each sample to give a concentration of 0.2%.

Full size image

Detection of HF as a terminal photoproduct in the condensed phase

Figure 5 displays 19F NMR spectra for water (D2O) extracts of the condensed phases from closed-ampoule photoreactions. Trifluoroacetic acid (TFA) was used as a chemical shift standard that gives a sharp resonance at − 76.55 ppm. The peaks observable between − 120 and − 175 ppm for the photoproduct samples (Fig. 5A–D) are generally indicative of inorganic fluoride compounds, indicating C–F bonds in PFOA have been broken and the fluorine has been converted to inorganic compounds such as HF (more specifically, 2HF).

A peak with identical shape and chemical shift of − 167.1 ppm is observed in all these photoproduct samples (Fig. 5A–D), as well as when a standard NaF sample is dissolved in 10% H2SO4 in D2O to give a similar acidity to our samples (Fig. 5E) and measured in an NMR tube with a PTFE insert. This − 167.1 ppm 19F chemical shift is within the range of values previously published for aqueous HF/DF28. These were noted to be strongly dependent on pH, and somewhat less dependent on level of deuteration and HF concentration.

Our observed chemical shift of − 167.1 ppm (Fig. 5A–E) is well upfield from the value of − 130 ppm previously observed for a pH 3 solution of Na2SiF629. The latter value, however, matches the value that we observed for a solution of Na2SiF6 in 10% H2SO4 in D2O (Fig. 6). We conclude that the chemical shift value of − 167.1 ppm that we observe for our photoproduct samples (Fig. 5A–D) is due to HF rather than either SiF6− or HSiF6, despite the presence of silica in both our reaction vessels and our NMR tubes.

Fig. 6
figure 6

19F NMR of HF and H2SiF6. Standard samples were prepared for comparison by first preparing a solution of 90:10 (w:w) D2O/H2SO4, then adding trifluoroacetic acid (TFA) to a concentration of 0.2%. A quantity of 1–2 mg of NaF or Na2SiF6 (resp.) was added to each of 2 separate freshly opened PTFE NMR tube inserts. Then, 0.5 mL of the 10% sulfuric acid solution was added to each, with gentle mixing, allowing 1 h for each sample to dissolve before making the NMR measurements.

Full size image

Somewhat surprisingly, we also saw no visual evidence of etching of the photoreaction ampoules, in the form of frosting or other optical perturbation, even after many days in contact with the HF-containing photoproducts in the presence of concentrated H2SO4. This could hypothetically be due to very strong HF–H2SO4 hydrogen-bonding interactions that render the HF unusually unreactive towards silica (see Discussion).

When we used borosilicate NMR tubes without PTFE inserts to measure NMR spectra of the D2O-diluted photoproduct samples, we did see peaks in the 140–160 ppm range that could be attributable to fluoroborate compounds (Fig. 5A–D). However, these fluoroborate peaks were not observed alongside the − 167.1 ppm HF peak, when we were careful to keep the HF photoproduct away from borosilicate glass as the H2SO4 was diluted (Supplementary Fig. S4).

By comparison to the known amounts of fluorine in the TFA standard as well as in the acidified NaF standard (Fig. 5E), we estimate that these inorganic 19F signals from the photolyzed samples (Fig. 5A–D) represent a 2–4% recovery of the 30–40 mg of fluorine present in the original 50–60 mg samples of PFOA, after 4–5 h worth of focused solar illumination with our 15-cm-dia mirror. The recovery from a sample made with PTFE (Fig. 5C) was a factor of ~ 10 lower than from the best PFOA sample.

Estimate of maximum likely mass of inorganic fluoride in the residual photoproduct solid

It is likely, but not proven, that inorganic fluoride is also present in the residual inorganic solid phase obtained after it was washed, first with ethanol (to remove residual fluorocarbons and excess H2SO4, and to chemically reduce residual MnO2 to MnSO4) and subsequently with water (to remove Na+, Br−, and SO42−). The final aqueous supernatant after these washes typically had a pH value near 4.5, suggesting that it might be buffered by fluoride. After drying the solid residue, up to ~ 5 mg of grey-black powdery solids was recovered. However, the color was too dark to be pure MnF2. This dark color is most consistent with partial reformation of MnO2 from the predominantly blue and/or white solids that were visible in the unbroken sample ampoule after extensive illumination (Fig. 1). However, the chemistry involved in such putative MnO2 re-formation is unclear, and IR and X-ray powder diffraction measurements have not yet given interpretable results for us to identify any components in the powdery solid.

Discussion

The main conclusion from our results is that at least some fluorocarbons, such as PFOA and PTFE, can be broken down with (i) MnO2 serving as an oxidant in (ii) H2SO4 solvent, and using (iii) photocatalytic Br2 and (iv) highly concentrated sunlight to initiate the reactions. These reactions did not proceed to any observable extent in control experiments lacking any one of these four crucial components.

One of the main reasons for investigating the reaction with PTFE was to test alternative hypotheses: whether our fluorocarbon degradation method requires an oxygen-containing functionality, or might also involve direct attack on C–F bonds by photoactivated Br2. The detection of HF as a photoproduct of PTFE/MnO2/Br2 (Fig. 5C), even in the absence of any added surfactant, suggests C–F bonds are indeed directly susceptible. However, the low yield of HF from PTFE degradation gives some pause, because fluorocarbon surfactants such as PFOA and GenX (hexafluoropropylene oxide dimer acid) have been used in the commercial synthesis of PTFE30. We have no way to assay whether contamination of such fluorocarbon surfactant(s) in our PTFE sample (although labeled as 100% PTFE) might account for the small apparent yield of HF (~ 0.2% of the original F from the PTFE sample). The hypothesis that photoreaction requires the terminal carboxyl group must therefore be considered as not yet ruled out. Nevertheless, the preliminary observation by LC–MS of a fluorocarbon product of m/z near 131 from PFOA starting material is difficult to reconcile with the hypothesis of reaction only at the carboxyl end group.

If C–F bonds are indeed being attacked, we hypothesize that it might be through a free-radical mechanism analogous to the classic mechanism for bromination of C–H bonds, which begins with dissociation of Br2 to Br atoms, then with the Br atoms abstracting H atoms from C–H bonds to form HBr31. Our hypothesis is that with sufficient energy, Br2 can be photoactivated to attack C–F bonds in a somewhat similar fashion. This would lead to formation of BrF and C–Br bonds in an analogous free-radical mechanism31. Then BrF, a known but unstable molecule, would presumably react quickly with Br− to form F− and Br2.

The bromine-sensitized photodegradation of PFAS molecules (which lack C–H bonds) does not proceed significantly even after many hours under direct but unfocused sunlight. The requirement for very high light intensity suggests that multiple photons are required for the initial stage of bromine-sensitized abstraction of fluorine from the stronger C–F bonds (relative to C–H). This might involve photodissociation of Br2 followed quickly by photoexcitation of a Br atom.

In addition to the hypothetical photochemical activation of Br2 to attack C–F bonds, many additional non-photochemical reactions take place subsequently. MnO2 is expected to be consumed indirectly by continually re-oxidizing HBr to Br2. However, the H2O generated by this redox reaction, as well as both MnO2 and Br2 themselves, all possibly participate directly in non-photochemical reactions with the initially brominated photoproducts; these are all classical reagents in organic chemistry oxidations. It is unlikely that H2SO4 is also participating as an oxidant, since reduction of sulfuric acid would be expected to yield some SO2. No evidence of the expected strong IR absorption near 1300–1400 cm−1 SO2 was observed in the headspace-gas spectrum in Fig. 3.32 Rather than speculate on the mechanisms taking place in experiments with sunlight focused intermittently by hand-held mirrors, deciphering the overall mechanism(s) will be most amenable to experimentation with steadier focused illumination from continuous wave and pulsed lasers, along with additional time-resolved spectroscopic techniques.

For PFOA, the photocatalytic processes form progressively smaller stable fluorocarbon products that can be detected in both condensed and gaseous phases with 19F NMR and IR spectroscopic measurements (Figs. 2 and 3). If these intermediates remain confined in a borosilicate reaction ampoule, they are ultimately broken down to form CO2 as detected by IR and XRD measurements, and inorganic fluoride (as HF) as detected by 19F NMR measurements (Figs. 3, 4 and 5).

As mentioned in the introduction, fluorocarbons can be thermally oxidized in incinerators. The high temperature of incinerators generally precludes confinement, and therefore leads to continual escape of incompletely-oxidized intermediates6. Furthermore, incineration requires the additional presence of hydrocarbons to yield HF. In the absence of such a source of hydrogen, thermodynamics predict that fluorocarbons are stable to reaction with O2.

For example, the PTFE combustion process e.g. O2(g) + ½(–C2F4–)(s) → F2(g) + CO2(g) is computed to have a positive ΔHº =  + 20.32 kJ mol−1, using an enthalpy of formation of PTFE, (–C2F4–)(s), from published work33. No thermodynamic help can be obtained by further oxidation of the F2 product by O2. In simplified terms, C–F bonds are more stable than C–O bonds, so it is not easy to replace the former with the latter.

Why then can the carbon in PFAS be converted into CO2 photocatalytically using the process described in the current work? The answer to this question becomes clearer if the overall reaction 1,

$${\text{MnO}}_{{{2}({\text{s}})}} + \, \raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} \left( { - {\text{C}}_{{2}} {\text{F}}_{{4}} - } \right)_{{({\text{s}})}} \to {\text{ MnF}}_{{{2}({\text{s}})}} + {\text{ CO}}_{{{2}({\text{g}})}}$$
(1)

is written as the sum of the combustion process 2 with additional reactions 3 and 4:

\({\text{O}}_{{{2}({\text{g}})}} + \, \raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} \left( { - {\text{C}}_{{2}} {\text{F}}_{{4}} - } \right)_{{{\text{(s}})}} \to {\text{ F}}_{{{2}({\text{g}})}} + {\text{ CO}}_{{{2}({\text{g}})}}\)

(2)

\(\Delta H = \, + {\text{20kJ mol}}^{{ - {1}}}\)

\({\text{MnO}}_{{{2}({\text{s}})}} \to {\text{ Mn}}_{{({\text{s}})}} + {\text{ O}}_{{{2}({\text{s}})}}\)

(3)

\(\Delta H = \, + {\text{521kJ mol}}^{{ - {1}}}\)

\({\text{Mn}}_{{({\text{s}})}} + {\text{ F}}_{{{2}({\text{g}})}} \, \to {\text{MnF}}_{{{2}({\text{s}})}}\)

(4)

\(\Delta {\text{H}} = {-}{\text{855 kJ mol}}^{{ - {1}}}\)

\({\text{MnO}}_{{{2}({\text{s}})}} + \, \raise.5ex\hbox{$\scriptstyle 1$}\kern-.1em/ \kern-.15em\lower.25ex\hbox{$\scriptstyle 2$} ( - {\text{C}}_{{2}} {\text{F}}_{{4}} - )_{{({\text{s}})}} \to {\text{ MnF}}_{{{2}({\text{s}})}} + {\text{ CO}}_{{{2}({\text{g}})}}\)

(1)

\(\Delta H = {-}{\text{314kJ mol}}^{{ - {1}}}\)

Reactions 2 and 3 are both endothermic33,34. However, reaction 4 is highly exothermic35, with a negative ΔHº large enough to overcome the combined positive values of the other reactions. This is partly due to the enormous crystal energy stabilization of solid MnF2.

However, even when the MnF2 product is partially dissolved in H2SO4, it appears that the solution enthalpies of Mn2+ and HF are sufficiently favorable (exergonic; actual values do not appear in the literature) that coupling PFAS oxidation to their release into H2SO4 can still thermodynamically drive the otherwise unfavorable formation of C–O bonds from C–F bonds.

Unusually strong H-bonding interactions of HF with H2SO4 solvent would be consistent with several qualitative observations from our experiments. First, despite the measurable presence of up to ~ 1 mg of HF product in multiple solutions prepared in these borosilicate glass photoreaction ampoules, even over the course of many hours at intermittently elevated temperatures, we have seen no sign of etching of the glass, nor the formation of fluorosilicate products that would be expected to give rise to 19F NMR signals near − 130 ppm (see Figs. 5 and 6). Second, no odor of HF is detected from these opened ampoules, despite the boiling point of pure HF being just below room temperature. Third, when the pressure in these photoproduct vials is released and allowed to mix for many hours with Ca(OH)2, there is no sign of CaF2 formation. Fourth, there is no sign of HF absorption bands in the gas-phase IR spectrum (spectral region 3500–4000 cm−1 in Fig. 3). Together these observations support the idea that even at HF concentrations measurable with 19F NMR, strong H-bonding interactions with concentrated H2SO4 greatly suppress the HF chemical activity, probably in a similar way that they suppress the activity of H2O, whose activity coefficient approaches 10–9 when dissolved at low concentrations in H2SO436.

Reaction 1 between mixed-powder samples of perfluorocarbons (specifically PTFE) and MnO2 has been determined experimentally to initiate thermally at a temperature near 570 °C37. This is above the temperature at which PFAS generally depolymerize and/or evaporate to form gases. This determination therefore used milligram quantities of the reagents in a thermogravimetric differential scanning calorimeter designed to quickly remove evolved heat and gas. It would be impossible with any current technology to remove so quickly the evolved heat and gas from million-fold greater quantities of these reagents and products confined in an industrial-scale reactor. Using the known heat capacities of the products CO2 and MnF2, thermally triggering reaction 1 above 570 °C on a macroscopic quantity in a closed vessel could lead to a near-instantaneous runaway completion of the reaction, with a temperature increase of over 2000 K and a concomitant explosion potential from the evolved CO2.

The crucial advantage of our photocatalytic approach is thus that it allows the highly exothermic reaction 1 to be carried out at temperatures in the range 0–100 °C, in the presence of a thermally modulating solvent (H2SO4), thereby preventing runaway reaction. The reaction rate can be easily reduced to near zero at any time, simply by discontinuing illumination. Nevertheless, our spectroscopic results detecting HF and CO2 photoproducts from PFAS or PTFE using MnO2, NaBr, and sunlight, and the thermodynamic analysis above, suggest the bromine-photocatalyzed should go to completion with sufficient photoreaction time.

Results so far support the release of up to ~ 10–15% of the original carbon in the PFOA as CO2 after ~ 12 h total of focused solar illumination, and a likely conversion of most of the original remaining carbon to shorter-chain fluorocarbons. These fluorocarbon intermediates can be detected after various periods of photoreactions, either by extraction into organic phase and 19F NMR measurement (Fig. 2), or by collection of the gaseous photoproducts and measurement of their IR absorption spectrum (Fig. 3) and preliminary LC–MS analysis (see Supplementary Information). Such measurements have allowed only a general identification, but not a full speciation of the fluorocarbon intermediates and products from the photoreaction.

We tentatively conclude that the observed ion m/z value of 131 in mass spectral analysis of the headspace gas after ~ 12 h photoreaction corresponds to deprotonated C3HF5. This is a known molecular formula, some of whose isomers have been considered as substitutes for traditional fluorocarbon refrigerants because they are equally nontoxic, but degrade more easily in the atmosphere38. The absence of larger or smaller molecules in the headspace gas might be because photodegradation slows greatly when it reaches 3-carbon fluorocarbons, simply due to their greater volatility relative to longer-chain fluorocarbons. A much smaller fraction of these 3-carbon fluorocarbons would thus be expected to be present in condensed phase where the Br2-sensitized photoreaction is initiated. If this is the case, it might be possible to continue the degradation to completion, simply by using much longer periods of photolysis, and/or intermittently cooling the samples below 0 °C.

Determining whether C3HF5 is indeed the principal species in the residual headspace gas after 12 h photolysis will probably require mass spectrometry with high-resolution and the ability to analyze ionization fragments with m/z < 50. However, further speciation by IR and NMR spectroscopy will likely also be required, because C3HF5 has multiple isomers. Gas-phase IR spectra of only two of them (cis- and trans-1,2,3,3,3-pentafluoropropene) have been published38. Each of these spectra shows distinctive strong absorptions due to C–F stretches in the range 1030–1260 cm−1. Vibrations in this range are seen in our headspace IR spectrum (Fig. 3), but these frequencies are not perfect matches for published data38.

After 12 h photoreaction, the combined volume of gas released as CO2 and 3-carbon fluorocarbons can possibly account for 100% of the original carbon in the PFOA. That is, as stated above, 10% of that carbon is accounted for as 0.04–0.06 mmol CO2 making up 25% of the volume of the headspace gas. To account for the other 75% of the volume of released headspace gas, there must be 0.12–0.18 mmol of such non-CO2 gas. If all such non-CO2 gases have molecular formulas with 3 carbons, as suggested by the mass spectrometry (see Supplementary Information), then these other headspace gases contain 0.4–0.6 mmol C. The total amount of C in the headspace gases can therefore easily be as high as 0.5–0.75 mmol, which is enough to account for essentially all the original 0.055 mmol C in the 23-mg PFOA starting material.

Therefore, there need not be significant amounts of fluorocarbons remaining in the condensed-phase (liquid and solid) residue in order to account for all of the original C from the PFOA sample.

Further spectroscopic and/or XRD analysis will also be required to identify the transient bluish photoproduct seen after long periods of illumination (Fig. 1). Although its several-h lifetime at room temperature was too brief to analyze spectroscopically yet, this was the clearest direct visual evidence of a photochemical process occurring during ~ 12 h of illumination. Few Mn compounds with the six other elements present in our samples (C, O, S, Na, Br, H) are blue. We speculate that it could be the known stable aquamarine-colored compound MnF4, or possibly one of its unreported analogs, MnF2Br2 or MnF3Br.

Not only is the sunlight-focusing apparatus required for further laboratory-scale optimization and analysis of this photoreaction extremely inexpensive, but so too are the reagents. These reagents are also abundant in the natural environment, easily transported, and even available at reagent-grade purity as consumer products: MnO2 as a ceramic glaze colorant, concentrated H2SO4 as a drain cleaner, and NaBr as a hot tub disinfectant. These reagents, as well as the products that have been identified so far (CO2, HF) carry environmental hazards that are familiar and routinely dealt with industrially, e.g. by recycling and neutralization. By contrast, other commercially available compounds being investigated for PFOA photocatalytic destruction, e.g. boron nitride22 and indium oxide,2 are at least 5–10 times more expensive and not as easily disposed or recycled.

To summarize, both PFOA and PTFE have been degraded using a combination of MnO2, H2SO4, either liquid Br2 or NaBr, and focused sunlight, at temperatures of 0–100 °C. Compared to most other PFAS degradation methods being developed, this process is less expensive, depends exclusively on the sun as an energy source, and does not require dangerously high temperatures or persistently toxic metals. Future work will be needed to scale up the reaction, to extend its scope to other PFAS, to determine if the reaction can be carried to completion yielding only inorganic fluoride and CO2, and to separate Mn2+ and inorganic fluoride products from other reagents and H2SO4 solvent.

Methods

Sodium fluoride (NaF) and sodium hexafluorosilicate (Na2SiF6) were purchased from AK Scientific. Polytetrafluoroethylene (PTFE) 100% ultrafine powder (Materialix™ brand bicycle-chain lubricant) was purchased online. All other chemicals were purchased from Sigma-Aldrich and used as received without further purification. PTFE NMR tube inserts were obtained from New Era Enterprises, Inc. (USA).

Sample preparation

Samples were prepared by adding to 1 mL glass ampoules: 25–50 mg PFAS (PFOA, PTFE, or a 50/50 mixture of the two); 60–90 mg MnO2, 10–100 mg NaBr or Br2, and 0.5–1 g concentrated H2SO4.When Br2 was used, it was added to the ampoule last, after the ampoule body and other ingredients had been precooled to dry ice temperature. This was done to minimize Br2 evaporation during handling, weighing, and flame-sealing. When NaBr was used, H2SO4 was similarly added last after the ampoule body and other ingredients had been precooled to dry ice temperature. In these cases, this precooling served to delay the oxidation of NaBr by MnO2/H2SO4 to form Br2, a crucial step in the eventual photocatalysis process, that leads to rapid evaporative loss of gaseous Br2 from unsealed containers. Once all the unreacted ingredients had been added, and were again brought to dry ice temperature, the ampoule was quickly flame sealed at the neck. Samples were kept at 4ºC until suitable cloud-free solar illumination period(s) became available.

Solar illumination

A handheld mirror apparatus (see Supplementary Information for details) containing a 15-cm diameter paraboloid mirror (Mirascope™) was constructed to focus sunlight onto samples. Solar illuminations were conducted in 0.5–2-min intervals over periods of 3–12 h by holding the ampoule at the focal point of the paraboloid concave mirror apparatus under cloud-free sunlight. Cooling was performed in an ice bath for 1–2 min between illumination intervals.

The PFOA inside the ampoules was initially visible as a white solid or clear liquid phase, depending on temperature, floating on top of the H2SO4 phase. Over the initial 4–8 h of illumination, this floating PFOA disappeared and the black MnO2 solid turned lighter grey or brown. After an additional 2–4 h of illumination, a distinctly blue solid began to replace the brown-grey solid, particularly at any location on the ampoule’s interior surface where sunlight was focused tightly onto the sample. See Fig. 1 for photographic images.

The blue color of the solid photoproduct faded partially over hours, once the sample was removed from solar illumination and cooled. The blue color disappeared very rapidly and completely once the glass sample ampoule was broken open, turning initially to brown and then to a dark grey, i.e. a much darker color than commercially obtained MnF2.

Following solar illumination, photoreaction ampoules (see Fig. 1) were refrigerated until they were ready to be broken open.

Gas trapping

A gas trapping apparatus (see Supplementary Fig. S2 for details) was constructed to allow any CO2 produced during solar illumination to be captured upon cracking open the top of the photoreaction ampoule. The ampoule tip was broken while the sample was still a bit below room temperature, leading to an initial explosive release of 5–10 mL. A slower expansion of < 0.5 mL was observable as the gases came to room temperature. The gases were directed through a tight Luer-lock connection, either into a 30 mL glass syringe in order to measure the volume of released gases, or through a rubber septum sealing a 25 mL Erlenmeyer flask containing 10.0 mL of 0.02 M Ca(OH)2. The latter flask was kept sealed to the ampoule for 3 days, in order to allow as much gaseous CO2 photoproduct as possible to diffuse into the solution to form CaCO3. The Erlenmeyer flask was then swirled, and its contents were transferred to a 15-mL polypropylene centrifuge tube and centrifuged to collect the fine white precipitate. Alternatively, a 1-mL aliquot of the gases collected in a syringe was injected into a 50-mL gas cell that had just previously been purged with Ar and its FTIR background spectrum measured. The same cell containing 2% added gas from the photochemical reaction then had its FTIR spectrum re-measured and ratioed to the background spectrum.

Inorganic fluoride isolation

We first ascertained that NaF and MnF2 standard samples were soluble, both individually and when combined, to at least 5 mg/mL in 10% H2SO4/water. Therefore, we assumed we could use this solvent to extract inorganic fluoride salts of Mn2+ and Na+ from the photoreaction-product solids. After removal of the Tygon tubing seal used for gas trapping, the scored neck of the photoreaction ampoule was broken, and the sample remaining in and on all pieces of the broken ampoule was transferred into a 1 × 13 cm borosilicate culture tube by using 3 × 0.5 mL aliquots of ice-cold D2O. This culture tube was then centrifuged for 5 min at 7000 RPM, and the dark red/purple liquid supernatant was separated from the remaining solids. Those solids were then re-extracted with ~ 1 mL CDCl3 for 19F NMR measurements of the organic residues. For the NMR measurements on the aqueous phase, it was necessary to remove Mn2+, which is strongly paramagnetic and prevented even locking onto the deuterium NMR signal as a necessary precursor to the 19F NMR measurement. To this end, concentrated KMnO4 in D2O was added in aliquots to the supernatant liquid phase and the solid black MnO2 product was removed by successive centrifugations, until the intense characteristic purple color of permanganate persisted in the supernatant for several hours.

With some NMR measurements, the HF-containing sample was kept away from borosilicate glass as the H2SO4 was diluted, by removing the supernatant from the freshly-broken-open photoreaction ampoule to a 15-mL polypropylene centrifuge tube (instead of a 1 × 13 cm borosilicate culture tube). Such polypropylene tubes were then also used for dilution with D2O, treatment with KMnO4, and removal of MnO2 precipitate. The NMR sample in 90:10 D2O/H2SO4 was then measured using a PTFE NMR tube insert (New Era NE3505).

NMR measurements were performed on a Bruker Avance III HD 400 MHz NMR instrument.

Powder X-ray diffraction (XRD) measurements were performed using a Bruker D2 Phaser with a LYKXEYE 1D silicon strip detector using Cu Kα radiation (λ = 1.5406 Å).

FTIR measurements were performed on a Thermo Magna i550 instrument, using a borosilicate gas cell (Pike Technologies, 100-mm pathlength) equipped with 32-mm-dia BaF2 windows sealed against the glass with PTFE gaskets.