Thermal treatment of hexafluoropropylene oxide dimer acid (HFPO-DA) using a pilot-scale research combustor

Abstract

The thermal decomposition of single-component aqueous solutions of hexafluoropropylene oxide dimer acid (HFPO-DA), a “GenX” process chemical, was investigated in a pilot-scale research combustor. Two solutions containing target HFPO-DA concentrations of ~500 and 4000 mg/L were atomized at three post-flame temperatures of ~920, 860, and 750 °C, during which the stack gases were characterized for combustion products including fluorocarbon products of incomplete combustion (PICs). Analytical techniques included Other Test Method 50 (OTM-50) and real-time analysis using Fourier transform infrared spectroscopy (FTIR) and chemical ionization mass spectrometry (CIMS). Quantum chemical calculations were performed to identify pathways for the thermal decomposition of HFPO-DA. Identified PICs included 1H-perfluoroalkanes, fluoroether E-1 (heptafluoropropyl 1,2,2,2-tetrafluoroethyl ether), and two ultra short-chain perfluorocarboxylic acids (PFCAs): trifluoroacetic acid (TFA) and perfluoropropionic acid (PFPrA). Most PIC concentrations increased with decreasing peak temperatures with ~40–60% of the fluorine in the HFPO-DA converted into PICs at the lowest peak temperature examined (~750 °C). At higher peak temperatures, lower PIC concentrations were observed, suggesting that temperature is a critical parameter for mineralization. Modeling results proposed that the thermal decomposition of HFPO-DA can pass through 1,1,1,2,2,3,3-heptafluoro-3-[(1,2,2-trifluoroethenyl)oxy]propane (perfluoropropylvinyl ether, or PPVE) or fluoroether E-1 intermediate species with the PPVE route offering lower energy barriers.

Environmental implication

Thermal treatment is one of the few established techniques for the destruction of PFAS. However, concerns exist regarding the formation and emission of fluorocarbon products of incomplete combustion (PICs). This study examines the effect of peak temperature on fluorocarbon PIC formation in a pilot-scale combustion system during the thermal treatment of HFPO-DA, a PFAS commonly detected in the environment. By using new analytical techniques and quantum chemical modeling, this research enhances understanding of PFAS destruction, PIC formation, and thermal decomposition reaction pathways within combustion environments.

1.0. Introduction

Per- and polyfluoroalkyl substances (PFAS) are used in many industrial applications and commercial products and, as a result, are present in many waste streams and environmental media. PFAS are resistant to natural degradation and have potential health effects.^1–8 Some PFAS such as perfluorooctanoic acid (PFOA) are no longer used as a polymer processing aid in the manufacture of fluoropolymers due to a variety of contamination issues and health concerns.^1–6 In the U.S., manufacturers voluntarily agreed to phase out PFOA production by 2015.⁹ During this transition (2006–2015), PFOA was replaced by hexafluoropropylene oxide dimer acid (HFPO-DA) and its ammonium salt as the preferred polymer processing aid. HFPO-DA and its ammonium salt are the products of technology developed by The Chemours Company commonly referred to by the process trade name “GenX.”^{1, 10–14} Polymer processing aids reduce liquid surface tensions during polymerization reactions, allowing suspended polymer particles to grow larger. Polymer processing aids can then be separated from the fluoropolymer product and are often recovered for reuse.

HFPO-DA, a C₆ perfluoroalkyl ether carboxylic acid, was developed as a potentially less toxic and less stable alternative to PFOA. While changing the manufacturing process of fluoropolymers to the GenX process reduces the potential for further PFOA contamination, significant amounts of HFPO-DA have been found near manufacturing sites.^{1, 12, 15–18} HFPO-DA has its own potential health concerns and has recently been listed as one of six PFAS included in the U.S. EPA’s National Primary Drinking Water Regulation.¹⁹ At present, there is significant interest in methods to successfully treat HFPO-DA and other PFAS-containing wastes including contaminated soils and environmental media, wastewater residuals, and spent granular activated carbon (GAC) and ion exchange resins from drinking water pre-treatment technologies.^{14, 19–22}

One method to mitigate HFPO-DA contaminated material is thermal treatment that uses elevated temperatures and oxidative environments intending to decompose HFPO-DA into mineralization products of hydrogen fluoride (HF) and carbon dioxide (CO₂). For example, The Chemours Company, recently installed a thermal oxidizer (TO) at their Fayetteville Works facility to treat off-gases from the GenX process. The TO at Fayetteville Works is a natural gas-fired high temperature fast submerged-quench design based on the Sub-X thermal oxidizers developed by T-Thermal and Selas Linde North America specifically for process waste gases with high halogen concentrations.²³ This TO successfully demonstrated very high destruction efficiencies (>99.999%) for five GenX species including hexafluoropropylene oxide (HFPO), HFPO-DA, hexafluoropropylene dimer acid fluoride (HFPO-DAF), carbonyl difluoride (COF₂), and heptafluoropropyl-1,2,2,2-tetrafluoroethyl ether (fluoroether E-1).²⁴ Although products of incomplete combustion (PICs) measurements were not included in the study, this TO is designed to expose process waste gases to flame temperatures greater than 1000 °C under steady-state, well-mixed conditions.^{24, 25} However, the absence of knowledge of the PICs that may have formed under these conditions indicates further information is still needed. In addition, there are many other types of thermal treatment systems besides TOs, which often include mixed solid and liquid wastes of varying compositions, lower temperatures that vary with time, imperfect mixing, and waste introduction well away from standing flames.²⁵

Several studies have examined the thermal decomposition of perfluorocarboxylic acids (PFCAs) and perfluorosulfonic acids (PFSAs). Most focused on PFOA and perfluorooctanesulfonic acid (PFOS), two legacy C₈ PFAS no longer widely used but present in many wastes and environmental media. These studies have shown that thermal treatment can successfully decompose and mineralize PFCAs and PFSAs depending on the thermal conditions.^{13, 26–44} For example, in one experimental study, Shields et al.²⁶ concluded that temperatures >1100 °C successfully mineralized 10 PFAS present in a legacy aqueous film forming foam (AFFF) with below or near detection limit emissions of 30 fluorocarbon PICs. However, temperatures <1000 °C produced quantifiable fluorocarbon PICs that increased as temperatures decreased. Limited studies have focused on the thermal decomposition of HFPO-DA, with much of the existing research based on the use of quantum chemical modeling to examine the likely destruction pathways of HFPO-DA.^{13, 24, 39}

This study assessed the extent to which decomposition and mineralization of HFPO-DA occurred during thermal treatment in a pilot-scale combustor. Quantitative measurements of gas-phase species including fluorocarbon PICs present in the combustor exhaust were performed to better understand HFPO-DA destruction and mineralization as a function of temperature. Single-component solutions were used to relate fluorocarbon PICs to the original HFPO-DA and help inform mechanistic studies. Analytical methods employed included the newly developed Other Test Method 50 (OTM-50) providing targeted offline analysis of 30 non-polar volatile fluorinated PICs, plus real-time measurements of non-polar and polar volatile fluorinated products by Fourier transform infrared spectroscopy (FTIR) and chemical ionization mass spectrometry (CIMS). Additionally, quantum chemical simulations of HFPO-DA thermal decomposition were performed to gain mechanistic insights toward PICs formed during thermal treatment.

2.0. Experimental

Experiments were performed using the U.S. EPA Rainbow furnace, which is a small pilot-scale, refractory-lined, natural gas-fired research combustor. Detailed descriptions of the Rainbow furnace can be found elsewhere.^26–28 A schematic of the Rainbow furnace is presented in Figure 1, including the locations of HFPO-DA injections (ports 4, 6, and 10) and downstream extractive gas-phase sampling. In general, the six experiments described here followed procedures outlined in Shields et al.²⁶ with some exceptions described below.

Figure 1.

Schematic of U.S. EPA Rainbow furnace showing HFPO-DA injection and flue gas sampling locations. Exhaust is directed to a facility air pollution control system (APCS) including afterburner, baghouse, and NaOH wet scrubber.

For this study, the furnace was operated with a natural gas load and air/fuel stoichiometric ratio (SR) of 30.2 kW and 1.96, respectively. These settings established a temperature/residence time profile within the furnace similar to the 30-kW profile presented in Shields et al.²⁶ An updated temperature/residence time profile specific to these experiments is presented as Figure S1 in the supplementary material (SM). Two aqueous HFPO-DA solutions with target concentrations of 500 and 4,000 mg/L were prepared from liquid HFPO-DA (>97% pure, SynQuest Laboratories, Alachua, FL) and transferred to clean 19 L Cornelius kegs (Old Ale Wholesale Supply Co., Model KMS5G RBT, Taylor, MI). The aqueous solutions were then atomized along the centerline of the Rainbow furnace at a rate of ~15 mL/min using an atomization lance. Since the experiments described in Shields et al.²⁶, we have made improvements to the atomization lance to replace all air cooling with closed loop water cooling. Another improvement is the separate addition of PFAS liquids and atomization air through two concentric tubes within the atomization lance. The two fluids are then mixed at the injector tip where liquid atomization occurs.

Table 1 presents the experimental conditions for the six experiments in the order they were performed. Furnace load and stoichiometry were set, and the furnace was operated for several days to approach steady-state conditions. Ports 4, 6, and 10 (Figure 1) were chosen as injection locations for the HFPO-DA solutions to expose the HFPO-DA to approximate peak furnace temperatures of 920, 860, and 750 °C, based on ceramic shielded thermocouple measurements. Experiments using the low concentration HFPO-DA solution began at port 4 and proceeded to port 6 and port 10 over the course of one experimental day. At the end of the day, the furnace was left operating without any solution feed for over 12 h before a second experimental day repeated the process using the high concentration HFPO-DA solution.

Table 1.

Rainbow furnace neat HFPO-DA solution experimental conditions.^a,b

Natural Gas Flow Ratea (sL/min)	Air Flow Ratea (sL/min)	Injection Location	Peak Injection Temperature (°C)	Aqueous Soln HFPO-DA Concb (mg/L)	Aqueous Soln. Feed Rate (g/min)	HFPO-DA Feed Rate (mg/min)	Atomizing Air Flow Rate (sL/min)	Calculated HFPO-DA Flue Gas Conc (ppmv)
49.0	915.4	Port 4	918	441	15.25	6.72	10	0.492
49.0	915.4	Port 6	856	441	15.07	6.64	10	0.486
49.0	915.4	Port 10	747	441	15.19	6.69	10	0.490
49.0	915.4	Port 4	919	4230	15.34	64.82	10	4.75
49.0	915.4	Port 6	858	4230	15.11	63.85	10	4.68
49.0	915.4	Port 10	747	4230	15.39	65.03	10	4.76

^aNatural gas and air flows correspond to a thermal load of 30.2 kW and a stoichiometric ratio of 1.96.

^bAqueous HFPO-DA solutions were prepared with target concentrations of ~500 and ~4000 mg/L. Actual concentrations are listed in Table S1 from the average of pre- and post-injection samples collected from the atomization lance.

To verify solution concentrations, pre- and post-injection samples of the two aqueous HFPO-DA solutions were collected at the beginning and end of each experimental day. Prior to data collection, but after ~10 min of solution injection at port 4, the atomization air was stopped, the injector removed, and a small sample of the solution (~10 mL) was collected from the injection lance (pre-injection sample). This procedure was repeated at the end of the experimental day when the injector was removed from port 10 (post-injection sample). Both pre-injection and post-injection samples were analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS, Thermo Scientific Vanquish Horizon LC and TSQ Altis MS, Waltham, MA) and quantified using calibration procedures described in the SM. Actual HFPO-DA concentrations as well as trace concentrations of 24 other PFAS are presented in Table S1. Results indicate the 500 and 4000 mg/L target solutions produced actual HFPO-DA concentrations of 441 and 4230 mg/L, respectively, and that contamination from the other PFAS quantified was less than 2% of the HFPO-DA mass.

2.1. Analytical methods

During each of the six experimental conditions, extractive samples were collected following the procedures outlined in U.S. EPA OTM-50.⁴⁵ Each experimental condition yielded a ~4 L time-integrated gas-phase sample collected over 40 min using a ~100 mL/min critical flow orifice. In addition to the experimental samples, nitrogen (N₂) blank and combustion blank samples were collected. These samples were subsequently analyzed by GC-MS following procedures outlined in OTM-50. In addition to the OTM-50 samples, gas samples were extracted from the stack (see Figure 1) and through an inertly coated sintered filter to remove particulates before entering heated (150 °C) perfluoroalkoxy alkane (PFA) sample lines and directed to several real-time analytical instruments. These included two continuous emission monitors (CEMs, Models 702 LX and 703 LX, California Analytical, Orange, CA), two Fourier transform infrared spectrometers (FTIR, MultiGas Model 2030, MKS Instruments Inc., Andover, MA), and an online time of flight-chemical ionization mass spectrometer (TOF-CIMS, Tofwerk AG, Thun, Switzerland; Aerodyne Research Inc., Billerica, MA) operated in negative mode using iodide reagent ion, abbreviated as CIMS here for brevity.

One of the CEMs (Model 702 LX) was used to measure furnace exhaust concentrations of oxygen (O₂) and CO₂ and the other (Model 703 LX) to measure carbon monoxide (CO). These two sets of measurements were used in combination to verify combustion conditions and to quantify small amounts of air in-leakage caused by the facility’s induced draft blower and operation at a ~1.27 cm water draft. The CEM samples were preconditioned for water removal, and their O₂, CO₂, and CO measurements are presented on a dry basis.

The two FTIRs were used to quantify several volatile fluorocarbons, as well as HF, water vapor (H₂O_(g)) and CO₂. Both FTIRs have a 5.11 m heated cell set to 191 °C and set to measure wavelengths between 4500 and 600 cm⁻¹ at a resolution of 0.5 cm⁻¹ and a scan rate of 64 scans/min. One FTIR, operated normally with a heated sample, reported concentrations on a wet basis, and is henceforth referred to as “wet FTIR”. The second FTIR used two impingers placed in an ice bath to remove the bulk of water vapor followed by a Nafion permeation dryer (Model PD-200T, Perma Pure LLC, Lakewood, NJ) to reduce water vapor in the combustion gas from ~14% to ~0.2% (~140000 to ~2000 ppm_v) and reported concentrations on a dry basis, henceforth referred to as “dry FTIR.” Reducing the water vapor provided improved detection abilities for volatile fluorocarbons such as trifluoromethane (CHF₃), pentafluoroethane (C₂HF₅), and hexafluoroethane (C₂F₆) by reducing interference in the 1300 – 1100 cm⁻¹ region. The FTIR analysis was further improved by using CHF₃ (1.1 ppmv, ±5%) and C₂HF₅ (1.01 ppmv, ±5%) gas standards (Airgas USA LLC, Plumsteadville, PA) to improve the quantification within the MKS MultiGas FTIR software. A gas mixing system (Environics Inc, Series 4000, Tolland, CT) was used to create accurate diluted concentrations of both CHF₃ (100, 500, and 1000 pbb_v) and C₂HF₅ (25, 50 100, 500, and 1000 ppb_v). These spectra were then added to calibration process for the MKS software. The MKS MultiGas FTIR software quantifies gas concentration by fitting measured sample spectra to calibration spectra via classical least squares regression. Concentrations are determined from the scaling factors that minimize the difference between the measured sample spectrum and the scaled calibration spectra.

The CIMS was used to quantify C₂-C₈ perfluorocarboxylic acids (PFCAs) and HFPO-DA. Prior to the Rainbow furnace experiments, a response calibration for trifluoroacetic acid (TFA), perfluoroprionic acid (PFPrA), perfluorobutanoic acid (PFBA), perfluoropentanoic acid (PFPeA), perfluorohexanoic acid (PFHxA), perfluoroheptanoic acid (PFHpA), perfluorooctanoic acid (PFOA), and HFPO-DA was performed using a slightly modified version of the calibration apparatus for manual injection of liquid standards, or “CAMILS,” described in Davern et al.⁴⁶ A more detailed description of the calibration methods, as well as the resulting sensitivities and limits of detection for each compound, can be found in the SM and Table S2. Details of the CIMS instrument operation during the Rainbow furnace experiments as well as data analysis procedures have been described previously by Mattila et al.²⁸.

2.2. Computational methodology

The thermal decomposition of HFPO-DA was investigated by quantum chemical calculation using Gaussian 16.⁴⁷ Due to the large number of heavy atoms (20 in all), a high-level calculation such as G4 or G4MP2 is too computationally demanding.⁴⁸ The chosen methodology was ωB97XD with basis sets up to aug-cc-pVTZ which is capable of reproducing enthalpies of formation within 1.0 – 1.5 kcal/mol of the G4 methods.⁴⁹ For smaller molecules, G4MP2 was used.

Rate constants were calculated using the ChemRate program (Version 1.5.1, NIST., Gaithersburg, MD) using data obtained from quantum chemical simulations. Rate constants for bond fissions that did not have transition states were studied by canonical variational transition state (CVTST) methods.^50,51 Ansys Chemkin (2024 R2, Ansys Inc., Canonsburg, PA)⁵² was used to model the reactor kinetics. The Chemkin calculations were based on the combustion of methane and air (Table 1), the 30-kW temperature-residence time profile (see Figure S1)., and injection of the 4230 mg/L HFPO-DA solution at ports 4, 6, and 10 corresponding to peak injection temperatures of 920, 860, and 750 °C. The calculated HFPO-DA flue gas concentration at all three injection locations was 4.72 ppm_v and the combustion gasses were assumed to be well-mixed and plug flow.²⁶

3.0. Results and discussion

Table S3 presents average wet FTIR, dry FTIR, and CEM (dry) measurements for all six HFPO-DA experiments and four pre- and post-injection combustion blanks. Note that one set of wet FTIR measurements for port 10 are missing due to an operational error. The measured concentrations for O₂ and CO₂ were between 10.3 and 10.5% and 5.9 and 6.6%, respectively. Calculated exhaust O₂ and CO₂ concentrations based on natural gas and air flows (dry, assuming complete combustion, but not including injected HFPO-DA) are 10.9 and 5.7%, respectively. CO concentrations are <1 ppm_v for all combustion blanks and both HFPO-DA port 4 (~920 °C) injection experiments but increase slightly (~1–6 ppm_v) as the peak injection temperature decreases. This CO is related to incomplete HFPO-DA mineralization and not natural gas combustion as the pre- and post-injection CO concentrations are always <1 ppm_v and HFPO-DA injection is well downstream of the natural gas flame. Injection flow rates for the low (441 mg/L) and high (4230 mg/L) concentration solutions result in calculated HFPO-DA flue gas concentrations of ~0.49 and 4.72 ppm_v, respectively (see Table 1). Corresponding theoretical HF concentrations (assuming complete fluorine mineralization) are ~5.4 and 51.7 ppm_v, respectively. Table S3 indicates measured HF emissions between 0.70 and 26.1 ppm_v, which are lower than these theoretical values. HF is known to adsorb to and may react with the furnace’s internal alumino-silicate refractory and sampling system surfaces.^{26, 31, 53, 54} Desorbed HF is emitted hours and days later as evident in the pre- and post-injection measurements. This makes a total fluorine mass balance in many practical thermal destruction systems nearly impossible.

3.1. Thermal decomposition of HFPO-DA

Tables S4, and S5 present the OTM-50, and CIMS numerical results, respectively. These data are combined with the dry FTIR fluorocarbon data (Table S3) and presented graphically in Figure 2. Data related to the multiple fluorocarbon PFAS PICs quantified by real-time CIMS and dry FTIR are identified. All others not specifically designated were quantified by OTM-50. CIMS species include TFA and PFPrA and dry FTIR species include CHF₃ and C₂HF₅. The dry FTIR species can be compared directly to the same species measured by OTM-50 with good agreement. HF (see Table S3) was detected by the wet FTIR, confirming the decomposition of HFPO-DA. HFPO-DA destruction efficiency (DE) calculations based on measured feed rates and CIMS emission measurements (Table S5) indicate DE values >99.99% for the three lower solution concentration (441 mg/L) experiments and >99.999% for the three higher solution concentration (4230 mg/L) experiments. However, HFPO-DA was not detected above combustion blank levels in any of the six experiments and the different DEs (>99.99 vs. >99.999%) determined for the lower and higher HFPO-DA solution concentration experiments, respectively, are solely due to the 10 fold difference in the two HFPO-DA solution concentrations and the use of similar near combustion blank emission values used to calculate DEs for the six experiments.

Figure 2.

Fluorinated PICs identified by dry FTIR, OTM-50, and CIMS for the low concentration HFPO-DA solution (441 mg/L, upper panel) and high concentration HFPO-DA solution (4230 mg/L, lower panel) experiments. Dry FTIR and CIMS species are identified. All others were measured by OTM-50. Species quantified at concentrations >1 ppb_v from Tables S3, S4, and S5 are included. Histograms present species concentrations as a function of decreasing peak injection temperature (~920, ~860, and ~750 °C), left to 15 right. HFPO-DA solution concentrations and injection rates correspond to calculated flue gas concentrations of 0.49 (upper panel) and 4.7 ppm_v (lower panel).

OTM-50 includes targeted analysis of 30 C₁ to C₈ non-polar volatile fluorocarbons including all eight (C₁ to C₈) 1H-perfluoroalkanes. Figure 2 indicates that CHF₃ and C₂HF_5, are the dominant non-polar volatile fluorocarbon PICs formed, with much smaller (but detectible) concentrations of C₃HF₇ and C₄HF₉. Note that the C₅-C₈ 1H-perfluoroalkanes were not detected. The formation of both CHF₃ and C₂HF₅ is likely occurring from reaction between CF₃ and C₂F₅ radicals with hydrogen species. At 750 °C and well downstream of the natural gas flame, flame generated hydrogen (H) is unlikely. However, with ~14% water vapor present in the combustion gases, it is likely that H₂O_(g) and associated radicals such as OH or H, formed during second order reactions, account for the source of the hydrogen.^{33, 35, 55} Unlike H, OH radicals are more stable at these post-flame temperatures.^56–58 The same explanation may also be true for C₃HF₇, but a much more complicated mechanism is needed to account for the formation of C₄HF₉. HFPO-DA is comprised of a C₃ perfluoroalkane and a C₃ perfluorocarboxylic acid (similar to PFPrA) joined by an ether oxygen. The formation of C₄HF₉ would require carbon chain growth to be accomplished. This is possible via reaction of C₃F₇ and CF₂ or CF₃ radical, but at the low HFPO-DA concentrations used for these experiments, collision frequencies are likely very low.

In total, for all three peak HFPO-DA injection temperatures, over 90% of the fluorinated PICs identified by OTM-50 are accounted for by two 1H-perfluoroalkanes. However, the formation of both CHF₃ and C₂HF₅ from the thermal decomposition of PFAS has not been well studied because most experimental bench-scale studies using externally heated tube furnaces do not often well represent the very high water vapor concentrations in combustion environments.^{33, 35} However, if the two PFCAs from the CIMS measurements are included, the contribution of the 1H-perfluoroalkanes drop to ~20–30% of the total fluorinated PICs and TFA and PFPrA comprise over 60% of the total. Again, several studies suggest that PFCAs can be formed via hydrolysis reactions, which again highlights the importance of water vapor in PFAS thermal destruction chemistry. ^{59, 60}

When comparing the formation of CHF₃ and C₂HF₅ over the three peak injection temperatures, the data indicate that the concentration of C₂HF₅ decreases as the injection temperature increases, especially between port 10 (750 °C) and 6 (860 °C). This is likely caused by either the direct decomposition of C₂HF₅ or by the C₂F₅ radical readily decomposing into CF₂ and CF₃ at higher temperatures, while at the lower temperature (750 °C), the fission of C₂F₅ may be sufficiently slow for a hydrogen species to collide forming C₂HF₅.^{34, 61–63} A similar trend is also seen for CHF₃ where the CHF₃ concentration also decreases as temperature increases. However, unlike C₂HF₅, CHF₃ is prominent at all temperatures. The presence of CHF₃ at higher temperatures is probably caused by a few different factors. The first is that CHF₃ requires temperatures above 700 °C to decompose.^64–67 Even though these temperatures are achieved at the HFPO-DA solution injection ports, CHF₃ formation may be delayed along the decreasing temperature profile such that insufficient temperatures are available when needed.^64–67 Additionally, CHF₃ will likely be forming from CF₃ radicals, which tend to have slower second order reaction rates when compared to CH₃ or even CF₂.^{29, 44, 68, 69} This implies that CF₃ reactions to form mineral species such as HF and CO₂ may be slow, while the formation of CHF₃ or C₂F₆ may be favored.^{70, 71} Another observation from the OTM-50 results is noted for C₂F₆ in which the concentration increased with increasing temperature suggesting that some CF₃ radicals are left unreacted and recombined to form C₂F₆. A similar observation is seen for C₃F₈. The unreacted CF₃ radicals are likely from the high temperature decomposition of intermediate PICs such as the 1H-perfluorolkanes (e.g. C₂HF₅) or PFCAs which is occurring at higher temperatures (ports 4 and 6). This would result in increased availability of CF₃ radicals for the formation of C₂F₆ and C₃F₈.

Other PICs such as C₂F₄ and E-1 are observed predominantly during experiments at the lowest peak temperature (~750 °C). This is not surprising, for C₂F₄ as the fluorocarbon backbone of HFPO-DA, is either directly forming C₂F₄ or recombining two CF₂ singlets to form C₂F₄.¹³ At 750 °C, C₂F₄ is starting to decompose into CF₂ singlets, which are being consumed by H₂O, O₂, O, OH or H.^{30–35, 55} However, the formation of E-1, another 1H-perfluoroalkane, was surprising as Adi and Altarawneh¹³ and Blotevogel et al.³⁹ did not predict the formation of E-1. Our mechanistic study with a potential pathway to E-1 is discussed in a later section. To our knowledge, measurement of the PIC E-1 (Figure 2) has not been previously reported in the PFAS thermal treatment literature. Furthermore, except for the two perfluoroalkanes (C₂F₆ and C₃F₈) discussed above, as peak temperature increased, concentrations for all other OTM-50 PICs decreased, implying that peak temperature is a critical parameter affecting their destruction.

When comparing results from the low and high concentration HFPO-DA experiment, the overall trends and distribution of fluorocarbon PICs are very similar. Although perhaps obvious, an important observation is that PIC concentrations are significantly higher in samples collected using the higher concentration HFPO-DA solution. This is partially due to an increased amount of F present but is also related to the increased importance of second order reactions that are dependent on concentration and collision frequency. Further, higher initial concentrations help increase PIC concentrations above method and instrument detection limits offering improved confidence, precision, and a more diverse set of PICs to be detected.

Figure 3 presents the real-time concentration traces for the FTIR and CIMS measurements for the lower and higher HFPO-DA solution concentrations (upper and lower panels), respectively. Measurement periods corresponding to the OTM-50 sampling periods are shaded (gray) and average concentrations presented in Tables S3, S4, S5, and Figure 2 were calculated over these time periods. Dry FTIR data for CHF₃ and C₂HF₅ show robust responses for both HFPO-DA solution concentrations. FTIR results can be difficult to quantify and are limited to the database of spectra available, which are usually in the ppm_v range and are affected by interferences caused by species such as CO₂ and H₂O_(g). Even with significant reduction of H₂O_(g) in the dry FTIR measurements, attempted here to increase sensitivity, the improvements are limited, as significant drifts in H₂O_(g) were observed as the experiments continued. Additionally, the database included with the MKS MultiGas FTIR lacked the ppb_v range for CHF₃ and C₂HF₅, which can cause the calibration curve applying Beer’s law to have variable accuracy. Thus, in the post-analysis phase, we improved the quantification process by adding lower concentration spectra of 100, 500 and 1000 ppb_v for CHF₃ and 25, 50, 100, 500 and 1000 ppb_v for C₂HF₅ to the MKS software. This resulted in the dry FTIR measurements to have an increased sensitivity, and an improved accuracy compared to the wet FTIR measurements, but not to the levels of the OTM-50 measurements. In general, reducing H₂O_(g) concentrations in this manner improved the dry FTIR observable concentration limits for CHF₃ and C₂HF₅ to ~10 ppb_v compared to ~100 ppb_v for the wet FTIR measurements. However, even with these limitations, the dry FTIR data are valuable as they offer a potential means to measure and evaluate the concentrations of several common PFAS PICs in real-time at relatively low cost.

Figure 3.

Fluorinated PICs identified by real-time dry FTIR (64 scans/min) and CIMS (1 Hz) for the low concentration HFPO-DA solution (441 mg/L, upper panel) and high concentration HFPO-DA solution (4230 mg/L, lower panel) experiments. Shaded areas represent the time periods over which canister samples were collected for OTM-50 analysis, except in the case of the shaded area labeled “CB” (upper panel) for which an OTM-50 combustion blank was not collected. Shaded areas also represent the time periods over which data were averaged and reported in Tables S3, S5, and Figure 2. CIMS mixing ratios have been corrected for sample dilution. Port 4, 6, and 10 designations correspond to peak injection temperatures of ~920, ~860, and ~750 °C. HFPO-DA solution concentrations and injection rates correspond to calculated flue gas concentrations of 0.49 (upper panel) and 4.7 ppm_v (lower panel).

Comparing absolute concentrations, there is a ~30% and ~50% difference between the OTM-50 and the dry FTIR measurements for CHF₃ and C₂HF₅, respectively, with the dry FTIR indicating higher concentrations (Figure 2). The potential for over estimation of concentrations by the FTIR is not surprising due to limited signal-to-noise ratios, drifting baselines, lower quality calibration spectra, and most importantly, interference from other combustions gases such as H₂O, nitrogen oxides (NO_x) and CO₂. All of these can cause variation in the concentration, especially at the low concentrations examined here, and cause detectable concentrations of C₂HF₅ in the combustion blanks. Interference from overlapping fluorocarbons can also cause major issues at higher concentrations because as the C-F peaks get larger in the 1000 – 1300 cm⁻¹ region there is more overlapping. An example of this is the C₂HF₅ and CHF₃ spectra, which can overlap at 1150 cm⁻¹. We are still developing the dry FTIR analysis method with the objective to also remove both NO_x and CO₂ from the FTIR spectrum to further improve the data quality.

The wet FTIR was also used to monitor HF and showed that the highest temperatures resulted in the highest HF concentrations. However, the nature of HF and its transport properties through such a large volume makes it very difficult to determine the concentration of HF accurately for the system.^72–74 HF and other acid gases are known to react with silicate-based refractory materials, which complicate accurate determination of HF concentrations in the exhaust.^{31, 44, 72, 75, 76}

The CEM and both FTIRs detected increasing CO concentrations as the HFPO-DA peak injection temperatures decreased. This was most notable for the higher HFPO-DA solution concentration experiments, with CO concentration of 3.9 and 5.6 ppm_v measured during HFPO-DA injection at port 10. As discussed, this is likely related to the HFPO-DA and not the natural gas and is probably directly correlated to PICs that formed under conditions that lack sufficient OH radicals to convert CO, CF₂, and CF₃ to HF and CO₂.^{33, 44, 55} This results in increased concentrations of C₂F₄ and CHF₃.

While the FTIR and OTM-50 methods measure select volatile non-polar fluorocarbon PICs, there is evidence that volatile and semi-volatile polar PFAS PICs may also be forming within combustion systems. Measurements of volatile ultra-short chain PFCAs using CIMS provides insight into a portion of that “missing” chemical space. As seen in Table S5 and Figures 2 and 3, the CIMS measurements indicate that substantial amounts of TFA and PFPrA (C₂ and C₃ perfluorocarboxylic acids) are being generated within the Rainbow furnace. Concentrations of both PFCAs increase with decreasing peak HFPO-DA injection temperature with TFA concentrations of ~0.45 and 2.20 ppm_v for the 441 and 4230 mg/L HFPO-DA solution experiments injected at the lowest (~750 °C) peak temperature (port 10) location, respectively. These emission concentrations can be compared to the calculated HFPO-DA combustion gas concentrations (assuming no destruction) of ~0.49 and 4.72 ppm_v. Of course, volume concentrations cannot be compared directly, but it is significant that both sets of concentrations are of the same order of magnitude. Also noteworthy, is the absence of any PFCA larger than C₃, consistent with the molecular structure of HFPO-DA discussed above. PFCAs, including larger PFCAs, have been measured in previous studies examining individual compounds and mixtures that included longer chain PFAS^{26, 28} and have been predicted via a mechanism involving hydrolysis of acyl fluoride intermediate species.^{39, 59, 60} However, the relatively large amounts of TFA and PFPrA identified here, the relatively low kinetic rates for hydrolysis reactions with acyl fluorides to form PFCAs from the literature,^{39, 59, 60} and the limited potential for HFPO-DA to form acyl fluorides, would imply that the formation of TFA and PFPrA also occurs via other pathways and needs further study. Formation of short-chain PFCAs was also used by Shields et al.²⁶ as an apparent reason for the lower DEs observed for several PFCAs but not for a similar set of perfluorosulfonic acids (PFSAs). The development of additional kinetic mechanisms involving the formation of PFCAs is outside the scope of this study.

3.1.1. Real-time measurements

When visually comparing the real-time PIC concentrations measured by CIMS and dry FTIR (Figures 2 and 3), it appears that the generation of TFA and PFPrA may be correlated with the generation of CHF₃ and C₂HF₅. To investigate this potential relationship, a correlation analysis was conducted by plotting the TFA and the PFPrA time series concentrations each against both the CHF₃ and C₂HF₅ time series concentrations. For the low HFPO-DA solution concentration data (Figure 3), the resulting correlation coefficient, R², for both TFA and PFPrA versus CHF₃ were 0.77 and 0.70, respectively, suggesting only a mild correlation (Figure S3). However, it should be noted that the data used to calculate these correlations were limited by the detection limit of the FTIR for CHF₃ and C₂HF₅ and the comparatively low levels of CHF₃ produced under this condition. For the same low concentration condition, however, the TFA and PFPrA versus C₂HF₅ correlations produced coefficients of 0.96 and 0.92, respectively, suggesting a strong correlation. Additionally, for the high solution concentration data the R² values were all greater than 0.91, indicating a strong correlation between all the real-time PICs that were measured (Figure S3). This strong correlation between the polar volatile ultra-short chain PFCAs and the non-polar volatile 1H-perfluoroalkanes implies that the mechanisms involved in the formation of these four PICs are correlated. Furthermore, it emphasizes the importance of sufficiently high thermal treatment temperatures to reduce the formation of PICs, as all four real-time measured PICs decreased at similar rates with higher temperatures, supporting the idea that higher temperatures create less PICs.

3.2. Mechanistic understanding of the thermal decomposition of HFPO-DA

As previously stated, the thermal decomposition of HFPO-DA has been studied by quantum chemical simulations.^{13, 39} However, from the experimental results of our work we can now expand on the mechanisms that might be involved. Figure 4 is a global potential energy diagram of the potential pathways through which HFPO-DA may decompose.

Figure 4.

Global potential energy diagram for the thermal decomposition of HFPO-DA. Numbers in bold are barrier heights and identified as transition states (TS). Numbers in italics are enthalpies of reaction. Both sets of values are reported in kcal/mol at 298 K. A simplified version without structures can be found in in the supplementary material (Figure S4).

Similarly to Adi and Altarawneh¹³ and Blotevogel et al.³⁹, we found that HFPO-DA predominantly decomposes into perfluoropropylvinyl ether (PPVE). However, this product may not arise directly from HFPO-DA but from an HFPO-DA rotational isomer (HFPO-DArot) seen in Figure 5.

Figure 5.

Structures of HFPO-DA (left) and HFPO-DArot (right).

HPFO-DArot has a rotated H atom directed towards an F atom on the adjacent C atom. This rotamer lies a mere 1.6 kcal/mol above HFPO-DA and requires a modest barrier of 10.2 kcal/mol (Figure 4). For the reversible reaction HFPO-DA = HFPO-DArot, the forward and reverse rate constants are determined to be k_f = 10^13.16 exp(−11.97 kcal mol⁻¹/RT) s⁻¹ and k_r = 10^13.05exp(−9.4 kcal mol⁻¹/RT) s⁻¹, respectively, and the equilibrium constant, K = [HFPO-DArot]/[HFPO-DA] is the ratio of k_f/k_r. At 298 K (25 °C), K = 0.017 meaning that the rotamer comprises just 1.67% of the total HFPO-DA. At 973 K (700 °C), K = 0.346 and the rotamer comprises 25.7% of the total HFPO-DA. Based on this information, the time required (relaxation time) to reach 99.9% of the equilibrium rotamer concentrations is calculated to be 6.0 × 10⁻¹¹ s. Hence, equilibration is extremely rapid at 700 °C and above.

Previous studies did not take this rotamer into account, but as shown in Figure 4, most reaction flux passes through HFPO-DArot. Two major routes from the rotamer have been identified. The first reaction forms PPVE and is the fission of HF involving the H atom and the F atom on the adjacent C atom. This enables the simultaneous spontaneous fission of CO₂ to take place. This reaction is summarized as reaction (1):

$$\mathrm{H}\mathrm{F}\mathrm{P}\mathrm{O}-\mathrm{D}\mathrm{A}\mathrm{r}\mathrm{o}\mathrm{t}\to \mathrm{H}\mathrm{F}+{\mathrm{C}\mathrm{O}}_2+{\mathrm{C}\mathrm{F}}_2{\mathrm{C}\mathrm{F}\mathrm{O}\mathrm{C}}_3{\mathrm{F}}_7\left(\mathrm{P}\mathrm{P}\mathrm{V}\mathrm{E}\right)$$ (1)

The formation of the observed PIC E-1 (Figure 2) represents a new reaction pathway that has not previously been discussed in the thermal treatment literature. The initiation involves a hydrogen atom transfer to the adjacent C atom leading to the spontaneous fission of CO₂ according to reaction (2):

$$\mathrm{H}\mathrm{F}\mathrm{P}\mathrm{O}-\mathrm{D}\mathrm{A}\mathrm{r}\mathrm{o}\mathrm{t}\to {\mathrm{C}\mathrm{O}}_2+{\mathrm{C}\mathrm{F}}_3{\mathrm{C}\mathrm{H}\mathrm{F}\mathrm{O}\mathrm{C}}_3{\mathrm{F}}_7(\mathrm{E}-1)$$ (2)

It is notable that the OTM-50 results (Table S4 and Figure 2) report significant concentrations of E-1. Both rate constants have been calculated as k₁ = 10^14.275 exp(−59.30 kcal mol⁻¹/RT) s⁻¹ and k₂ = 10^14.431 exp(−64.40 kcal mol⁻¹/RT) s⁻¹.

At elevated temperatures, HFPO-DA itself can scission its −COOH group. This has an enthalpy of reaction of 84.2 kcal/mol and has a canonical variational transition state theory (CVTST) rate constant of 10^19.33 T^−0.92 exp(−83.06 kcal mol⁻¹/RT) s⁻¹, producing C₅F₁₁O and cis-HOCO according to reaction (3):

$$\mathrm{H}\mathrm{F}\mathrm{P}\mathrm{O}-\mathrm{D}\mathrm{A}=\mathrm{c}\mathrm{i}\mathrm{s}-\mathrm{H}\mathrm{O}\mathrm{C}\mathrm{O}+{\mathrm{C}}_5{\mathrm{F}}_{11}\mathrm{O}$$ (3)

To determine these rates, we also calculated the enthalpy of formation (Δ_fH^°₂₉₈) for HFPO-DA through three separate calculations. The first used isogyric (atomization) method at ωB97xD/aug-cc-pVTZ//wB97xD/6–311+G(d,p) to get −686.8 kcal/mol. Additionally, two isodesmic calculations using equation (4):

$$\mathrm{H}\mathrm{F}\mathrm{P}\mathrm{O}-\mathrm{D}\mathrm{A}+2{\mathrm{C}\mathrm{F}}_4={\mathrm{C}\mathrm{F}}_3\mathrm{C}\mathrm{O}\mathrm{O}\mathrm{H}+{\mathrm{C}\mathrm{F}}_3{\mathrm{O}\mathrm{C}\mathrm{F}}_3$$ (4)

The first isodesmic calculation used ωB97xD/6–311+G(d,p) and the second ωB97xD/aug-cc-pVTZ giving Δ_fH^°₂₉₈ of −689.2 and −687.4 kcal/mol respectively.

All rates for the initial decomposition of HFPO-DA can be found in Chemkin format in Table S6. A comparison between the relative formation of PPVE, E-1, and C₅F₁₁O as a function of initial peak HFPO-DA injection temperature is shown in Figure S5. For simplicity, only the first 0.5 s of residence time is shown, and these calculations only include reactions involving the formation of HFPO-DArot, PPVE, E-1, and C₅F₁₁O. Evident from Figure S5 is that HFPO-DA and HFPO-DArot are consumed very quickly at the two higher injection temperatures, but decay slowly to ~80 ppb_v concentrations for the lowest injection temperature. In terms of the relative importance of PPVE, E-1, and C₅F₁₁O routes, all three injection temperatures clearly indicate that the formation of the PPVE is ~5 times faster than E-1 and ~100 times faster than C₅F₁₁O.

Even though the mechanism (Figures 4 and S5) predicts E-1 and PPVE as two key initial intermediate products of HFPO-DA decomposition, PPVE will likely be a short lived intermediate at the experimental temperatures due to the relatively small C-O bond enthalpy of 43.2 kcal/mol to form C₃F₇ + CF₂CFO. Previous studies^{13, 39} assumed that PPVE would result from fission of an F atom from the terminal CF₃ group which would be expected to have an enthalpy of reaction of 68.6 kcal/mol at 298 K. However, potential energy scans reveal that this fission does not take place, and this is indicated by the cross (X) in Figure 4. Instead, the F atom transfers to the adjacent radical-bearing -CF- with a discrete barrier of 43.4 kcal/mol. Thus, the barrier found for C-C fission is lower, and the reaction flux is to C₃F₇ + CF₃COF.

Unfortunately, none of the analytical methods used here are well suited for PPVE measurement. OTM-50 uses a set of cooled impingers to remove moisture and HF and PPVE has been reported to be water soluble (1.8 mg/L at 22.3 °C).⁷⁷ PPVE was not detected in the tentative identified compound (TIC) analysis of the OTM-50 samples. PPVE measurement by FTIR is possible. However, the lack of library spectra and interference from water vapor and CO₂ limited our ability to identify spectral peaks at the ppb_v levels likely to be expected. Similarly, CIMS measurements for PPVE were not included in the high-resolution peak fitting as PPVE would likely result in limited signal. PPVE measurement remains a topic for future research.

As seen in Figure 4, the major route for thermal decomposition of PPVE was to form C₃F₇ and CF₂-CF=O (CF₂CFO) with a low enthalpy of reaction of 43.34 kcal/mol at 298 K. CF₂CFO can also directly scission into CF₂ + FCO. This does not have a discrete barrier and evaluation of its rate constant requires CVTST methods. This has been carried out using trial transition states between C-C bond lengths of 2.227 and 3.027 Å. This relatively small molecule will be in the fall-off regime at 1 atm. Using ChemRate, with a value of ΔE_down of 200 cm⁻¹ gave a fall-off value of 10^52.9 T^−12.04 exp(−73.3kcal mol⁻¹/RT) s⁻¹.

CF₂CFO can also undergo an F transfer to form CF₃CO which, in turn, can scission into CF₃ + CO. Both reactions exhibit discrete barriers. The potential energy surface (PES) for this process is shown in Figure 6.

Figure 6.

Potential energy diagram of CF2CFO. All values are in kcal/mol at 298 K.

Because of the shallow well of the CF₃CO intermediate and the very small barrier for its decomposition, this will be a “well-hopping” reaction which will not be pressure-dependent at 1 atm. The overall reaction to CF₃ + CO will have the same rate constant as the initial rate constant. The rate constant was calculated to be 10^12.94 T^0.188 exp(−64.3 kcal mol⁻¹/RT) s⁻¹. The latter rate constant is significantly faster than the 1 atm direct fission rate constant, hence the principal route for decomposition of CF₂CFO will be to form CF₃ and CO.

As seen in Figure 4 the enthalpy of reaction of all minor routes have been calculated. A simple Chemkin model has been created using the Rainbow furnace parameters (discussed in the experimental section). Table S6 presents the calculated Arrhenius rate parameters in Chemkin format for the reduced 14 reaction kinetic set describing the initial thermal decomposition of HFPO-DA. NASA coefficients (for 30 species) are also provided in the SM. The first seven reactions in Table S6 describe HDPO-DA transition to HFPO-DArot, formation of PPVE, E-1, and C₅F₁₁O, and further decomposition to C₁ and C₂ fluorocarbon species based on the present work. Seven additional reactions using estimated rate constants describe additional decomposition pathways for PPVE, E-1, and C₅F₁₁O.

The Chemkin model results (Figures S5 and S6) show that at the highest two injection temperatures (~920 and 860 °C), the decomposition of HFPO-DA occurs quickly within the first 100 ms. However, at the lowest injection temperature (~750 °C), HFPO-DA persists at concentrations ~80 ppb_v. This is not surprising as temperature rather than residence time has the larger effect on the thermal decomposition of PFAS.^{26, 29, 31–35, 44}

The model results also reveal that the primary pathway of HFPO-DA decomposition is through PPVE + HF and CO₂ (Figure S5). However, Figure S6 indicates no PPVE to be present as a product for any of the simulations. With additional decomposition reactions included, PPVE is thermally decomposing nearly at the same rate as it is forming into C₃F₇ and CF₂CFO. Even if analytic methods were available, the kinetic calculations indicate no measurable PPVE. Contrary to PPVE, the model predicts measurable E-1 with highest concentrations for the lowest (~750 °C) injection temperature. This is consistent with our OTM-50 measurements. Using the full 14 reaction reduced mechanism (Table S6), other stable species predicted include C₃F₆O, C₂HF₅ and CO. E-1 and C₂HF₅ were the only fluorocarbon PICs measured that were included as reaction products in the reduced mechanism. It is important to note that 14 reactions are insufficient to model the complete decomposition of HFPO-DA to final reaction products. The purpose of these calculations was to identify initial pathways and identify potential intermediate species. Some variation between the model and experimental is expected due to potential surface reactions that are known to occur with HFPO-DA, additional second order reactions from OH radicals, or the 1.5 – 2 kcal error in the density functional theory (DFT) calculations.^78–82

Unstable species such as FCO, CF₃, CF₂ C₃F₇ CF₂CFO, HOCO, and CF₃COF are very likely to either directly react with the excess H₂O_(g), O₂, or fission into other species. FCO, for example, is well known to undergo reaction (5):⁸³

$$\mathrm{F}\mathrm{C}\mathrm{O}(+\mathrm{M})=\mathrm{C}\mathrm{O}+\mathrm{F}(+\mathrm{M})$$ (5)

Important reactions are also noted between H₂O_(g) and fluorine radicals such as F and CF₃, suggesting that it is crucial to maintain large amounts of excess of water vapor. These include HF formation via reaction (6) with F and H₂O_(g):^{33, 35}

$$\mathrm{F}+{\mathrm{H}}_2{\mathrm{O}}_{\left(\mathrm{g}\right)}=\mathrm{O}\mathrm{H}+\mathrm{H}\mathrm{F}$$ (6)

and CHF₃ formation via reaction (7) with CF₃ and H₂O_(g):^{29, 44}

$${\mathrm{C}\mathrm{F}}_3+{\mathrm{H}}_2{\mathrm{O}}_{\left(\mathrm{g}\right)}={\mathrm{C}\mathrm{H}\mathrm{F}}_3+\mathrm{O}\mathrm{H}$$ (7)

These chain propagating reactions also result in the formation of additional OH radicals useful elsewhere such as the oxidation of CO to CO₂.^{33, 35} We have limited our model to initial first order dissociation reactions and exclude all second order reactions as additional work in this area is needed, especially regarding products such as TFA and PFPrA whose formation mechanisms are not well understood.^{29, 40, 41, 44} As previously mentioned OTM-50 results did suggest that some CF₃ radicals do not react with hydrogen species but combined into C₂F₆ via reaction (8):

$${\mathrm{C}\mathrm{F}}_3+{\mathrm{C}\mathrm{F}}_3(+\mathrm{M})={\mathrm{C}}_2{\mathrm{F}}_6(+\mathrm{M})$$ (8)

It is also possible that the formation of C₂F₆ is occurring at temperatures that are too low for further decomposition.

Nevertheless, it is evident from all pathways that the formation of both CHF₃ from CF₃ radicals and the direct formation of C₂HF₅ or reactions with C₂F₅ are feasible. The formation of TFA could occur from CF₂CFO or CF₃CO.^84–87 However, the mechanisms leading to TFA and PFPrA are not well known and require further study. In the case of PFPrA it is possible that it will decompose to TFA.^{29, 37, 44, 88}

3.3. HFPO-DA destruction and PIC formation

3.3.1. Destruction efficiency

Based on CIMS measurements, all six experimental conditions (three temperatures and two HFPO-DA concentrations) report high DEs >99.99%, and this is supported by the modeling results that indicate HFPO-DA destruction (>99.9%) within 100 ms for the two highest peak injection temperatures (920 and 860 °C). At the lowest peak injection temperature (750 °C) the model predicted DE (97.7%) is less complete and HFPO-DA is seen to persist at flue gas concentration of ~80 ppb_v indefinitely. This result is interesting, as it again highlights the importance and sensitivity of peak temperature on HFPO-DA destruction and the difference that an approximate 110 °C increased initial temperature can have. For HFPO-DA, and likely for other PFAS, initial decomposition reactions and molecular destruction are controlled by first order unimolecular reactions that may be very fast if sufficient energy is available. High temperatures and not necessarily large residence times or turbulent mixing are required for this initial step. The temperature-residence time profile presented in Figure S1 indicates the three HFPO-DA injection locations and the location of port 21 near the end of the horizontal furnace section immediately prior to the vertical turn and the first set of sampling locations (CEMs and CIMS, see Figure 1). Residence times from ports 4, 6, and 10 to port 21 are ~6.0, 5.5, and 5.0 s, respectively. The difference in residence times between ports 4 and 10 is ~1 s (~20%). Residence times to the FTIR and OTM-50 sampling locations are likely >10–20 s. Residence time has an important role to ensure intimate mixing of reactants especially for second order reactions. However, for PFAS whose initial decomposition reactions are unimolecular with high activation energies, temperature is the most important variable. This is supported by both the DE measurements and modeling results as discussed above. Based on the calculated kinetics, the effect of temperature on HFPO-DA destruction is highly nonlinear and reaction rates increase quickly to time scales on the order of milliseconds as temperatures increase above 860 °C. However, as these intermediate species decompose further, second order reactions with O₂, H₂O and associated radical species becomes more important to promote the formation of HF and CO₂, and these require intimate mixing of reactants and longer residence times to achieve this mixing. With respect to a declining temperature-time profile typical of many combustion systems, high initial temperatures provide the energies needed to quickly accomplish the initial unimolecular decomposition reactions and sufficient energies at a later point in time along the temperature profile to maximize conversion to HF and CO₂.

3.3.2. PIC formation

We have previously discussed the difficulties of attempting to close a total fluorine mass balance due to the propensity of HF to adsorb on and react with internal surfaces and reactor materials (to form SiF₄, for example) and desorb over long times later. However, although properly accounting for mineral fluorine is often problematic, it may be possible to perform an organic fluorine mass balance especially when the amounts of organic fluorine fed is known with reasonable precision. In this case, we know the concentration and purity of the two HFPO-DA solutions, their feed rates, and the feed rates of air and natural gas. From these, we can calculate the theoretical HFPO-DA concentration in the flue gas at the points of injection. Table S7 presents results from a set of calculations comparing fluorine fed in the form of HFPO-DA and fluorine recovered in the form of fluorocarbon PICs measured by CIMS and OTM-50. We did not include the FTIR measurements of CHF₃ and C₂HF₅ as these species are also measured by OTM-50. Available fluorine from HFPO-DA was determined based on 11 F atoms per HFPO-DA molecule and measured F in the form of fluorocarbon PICs was determined based on the sum of measured concentrations of quantified fluorocarbon PICs multiplied by their respective F atom count. Comparison of F available and F measured as fluorocarbon PICs indicates that at the lowest peak HFPO-DA injection temperature (~750 °C), 58 and 37% of the available F was recovered as fluorocarbon PICs for the low and high solution concentration experiments, respectively. These values fell to 15 and 17% and 3 and 5% for the ~860 and 920 °C experiments, respectively. Together, TFA and PFPrA account for between 40 to >60% of the PICs measured. This suggests there is potential for release of both TFA and PFPrA via the stack gases especially at the lower peak injection temperature (750 °C). While the biological toxicity of TFA and PFPrA are presumably lower than longer PFCAs,^{89, 90} their persistence as environmental contaminants warrants attention.^{91, 92} However, it is important to note that many thermal treatment systems such as hazardous waste incinerators are operated at much higher temperatures (>1000 °C) and many include primary and secondary combustion chambers. Further, TFA, PFPrA, and other PFCAs are likely to be at least partially soluble in aqueous solutions used by wet scrubbing air pollution control devices for acid gas control and contribute to a liquid discharge.

Four species (TFA, PFPrA, CHF₃ and C₂HF₅) accounted for the largest portion (96–98%) of the total fluorocarbon PICs measured with six others measured at concentrations >1 ppb_v. To put this into perspective, we tried to quantify 45 fluorocarbon species (30 by OTM-50, 10 by CIMS, and 5 by FTIR). These methods are limited to a select set of polar and non-polar volatile fluorocarbon species for which these methods are sensitive and for which we have calibration standards and available infrared spectra. This list is not comprehensive and does not include all classes of fluorocarbon PICs possible. An important observation is the persistence of the CF₃ group among all major fluorocarbon PICs. Considering that two CF₃ groups are available in each HFPO-DA molecule, the concentration of HFPO-DA in the gas phase for the low solution concentration HFPO-DA experiments (0.49 ppm_v) correspond to a total of 0.98 ppm_v of available CF₃ groups. Summing the CF₃ groups found in the measured fluorocarbon PICs yielded a minimum of 83, 22, and 4.8% of unmineralized CF₃ groups remaining within fluorocarbon PICs at the peak injection temperatures of ~750, 860, and 920 °C, respectively. Similarly, for the high solution concentration HFPO-DA experiments (9.46 ppm_v of available CF₃), the percentage of unmineralized CF₃ groups within measured PICs were 51% (750 °C), 25% (860 °C) and 8.7% (920 °C). These results suggest that the CF₃ group may be uniquely resistant to mineralization (to HF and CO₂) and that it plays a dominant role in the formation of fluorocarbon PICs.

4.0. Conclusion

The thermal treatment of two aqueous HFPO-DA solutions in a pilot-scale combustor was successfully conducted at three peak injection temperatures of ~750, 860, and 920 °C, with near combustion blank concentrations of HFPO-DA measured and DEs >99.99% determined for all the experimental conditions examined. However, even with satisfactory DEs, significant amounts of both polar and non-polar volatile fluorocarbon PICs were observed. Compared to a calculated HFPO-DA initial flue gas concentration of 4.72 ppm_v (4230 mg/L solution), real-time CIMS measurements quantified two polar volatile ultra-short PFCAs (TFA and PFPrA) at a total flue gas concentration >3.1 ppm_v, while OTM-50 and real-time FTIR measurements quantified several non-polar volatile fluorocarbon PICs (including CHF₃ and C₂HF₅) at a total flue gas concentration >1.4 ppm_v. When considering the total fluorine introduced with the HFPO-DA, for the lowest HFPO-DA peak temperature examined (~750 C), 58 and 37% of the HFPO-DA fluorine converted into fluorocarbon PICs for the low and high solution concentration experiments, respectively. Fluorocarbon PICs (amounts and fractions decreased) with increasing peak injection temperature.

Along with the increasing concentrations of fluorocarbon PICs, CO, not likely associated with the natural gas flame, also increased with decreasing peak injection temperatures, suggesting an association. This is likely directly correlated to the OH radical concentration needed to convert CO into CO₂ and known to be a key step in the mineralization of PFAS. Further research is warranted to fully understand the correlation of fluorocarbon PICs and CO under a range of combustion conditions. However, for the higher HFPO-DA solution concentration examined here (4230 mg/L), the CO increase was small (from <1 to 6 ppm_v) over the range of decreasing peak injection temperatures (from 920 to 750 °C). Interestingly, C₂F₆ and C₃F₈ were found to increase in concentration with increasing peak injection temperatures. While counterintuitive, this may be due to temperature-related increased localized concentrations of CF₃ and C₂F₅ near the point of injection as HFPO-DA quickly decomposes at higher temperatures and increased collision frequencies of these radicals to form stable C₂F₆ and C₃F₈. Correlation graphs between concentrations of four fluorocarbons measured using the real-time CIMS and FTIR methods showed strong correlations between the polar and non-polar volatile PICs of TFA, PFPrA and CHF₃ and C₂HF₅ suggesting that the formation of all four species may be linked.

A mechanistic study, completed using quantum chemical simulation, discovered an improved pathway to PPVE and a new route to E-1. Further analysis of these two major pathways indicated that PPVE will be the dominant decomposition product for the initial decomposition of HFPO-DA. However, improved understanding of PPVE decomposition revealed that, at the temperatures studied, PPVE should rapidly break down into unstable species of C₃F₇ and CF₂CFO, which will likely lead to the formation of CHF₃, and C₂HF₅ via reaction with hydrogen species such as water vapor, while E-1 could persist at the studied temperatures and was found to be present in the OTM-50 samples demonstrating its relative greater stabilty compared to PPVE. An important finding from the modeling was that HFPO-DA rapidly decomposes via multiple unimolecular reactions within milliseconds in the reactor and that initial peak temperature and not residence time controls these initial decomposition reactions. However, as decomposition proceeds, second order reactions become critical and may require sufficient temperature and mixing to minimize emissions of fluorocarbon PICs.

Experimental conditions were selected to examine the effect of peak temperature on the destruction of HFPO-DA and more importantly, the formation of fluorocarbon PICs. PICs could be quantified at all three peak temperatures examined. However, at the highest temperature examined (~920 °C) the total fluorocarbon PIC concentrations from HFPO-DA decomposition were much reduced suggesting that high peak exposure temperatures (not necessarily high residence times) are most related to minimizing fluorocarbon PICs.