Development of a standardized adsorbable organofluorine screening method for wastewaters with detection by combustion ion chromatography

Abstract

Per- and polyfluoroalkyl substances (PFAS) are man-made organofluorine chemicals that can contaminate environmental waters and have gained worldwide attention over the past two decades. PFAS are most frequently detected by mass spectrometric targeted analysis methods which may not detect all the PFAS in samples. This report describes the investigation of adsorbable organofluorine (AOF) with detection by combustion ion chromatography (CIC) for detection of PFAS in surface waters and wastewaters that adsorb to granular activated carbon (GAC) with the recognition that this technique measures more than just PFAS. Overall mean recoveries of 77–120% were obtained in 17 of the 18 tested surface water and wastewater matrices spiked with perfluoropentane sulfonate (PFPeS) and 55–119% mean recoveries were obtained in 11 of the 12 surface water and wastewater matrices spiked with a PFAS mixture. Poor method performance (34–39% mean recoveries) was observed in landfill leachate wastewater. Method detection limits of 1.4–2.2 μg L⁻¹ were achieved using 100 mL sample volumes adsorbed onto commercially available GAC. This report demonstrates that this AOF technique can be a useful screening tool for estimating organofluorine concentrations when PFAS contamination is suspected.

Introduction

Organofluorine chemicals, such as per- and polyfluoroalkyl substances (PFAS), fluorinated pesticides, and fluorinated pharmaceuticals, are man-made chemicals with the potential to be contaminants in environmental waters across the globe. Of this group of organofluorine chemicals, PFAS are of particular concern due to their persistence and prevalence in the atmosphere, the environment, aquatic and terrestrial biota, the human body, household products, food, food packaging, and drinking water. Many analytical methods have been developed for the targeted analysis of individual PFAS chemicals in a variety of matrices, but the search for analytical methods that capture all PFAS has expanded significantly in the last decade.

The U.S. Environmental Protection Agency’s (EPA) 2019 PFAS Action Plan identifies the need for additional analytical methods for PFAS.¹ According to the National Institute of Environmental Health Sciences, there are over 4000 PFAS currently in existence² and this group of chemicals is ever-expanding. Currently, the most frequently employed targeted analysis methods use liquid chromatography tandem mass spectrometry (LC/MS/MS) for the detection and quantitation of specific PFAS. Although LC/MS/MS methods are very sensitive with typical detection limits of <10 ng L⁻¹, commercially available standards do not exist for all PFAS, which makes analysis with targeted methods extremely limiting. One analytical approach gathering interest within EPA and the scientific community is the estimation of total PFAS by non-targeted analysis (NTA; detection of both known and unknown chemicals). NTA is a critical step in the effort to comprehensively address PFAS as a group.

There are a number of non-targeted and aggregate chemical class methods being investigated for the analysis of a greater number of PFAS than targeted analyses can determine such as high-resolution mass spectrometry (HRMS),^3,4 total oxidizable precursor (TOP) assay,^4–6 particle-induced gamma ray emission (PIGE) spectroscopy,^5,6 fluorine-19 nuclear magnetic resonance (¹⁹F-NMR) spectroscopy,⁷ extractable organofluorine-combustion ion chromatography (EOF-CIC),⁴ and adsorbable organofluorine (AOF)-CIC.^5,6 Although these techniques have strengths and weaknesses (some briefly mentioned below), all are valuable tools in the assessment of PFAS. Generally, NTA methods are hampered by poor sensitivity unless concentration techniques, such as solid phase extraction (SPE), are used. Unfortunately, PFAS methods that rely on sample preparation techniques to achieve suitable sensitivity will not capture all PFAS or all organofluorine due to losses during sample preparation.

HRMS is a useful tool to identify novel/unknown PFAS in samples but is quantitative only if well-characterized standards are available. The TOP assay estimates PFAS precursors that can be converted by oxidation to perfluoroalkyl acids (PFAAs) and the PFAAs are quantitated by LC/MS/MS. Thus, the TOP assay, even though it targets specific PFAS compounds, provides an estimation of total precursors in a sample but may not identify specific precursor compounds or unknown PFAS that are not degraded to measurable PFAAs. PIGE is not widely available and is primarily a surface analysis technique that measures total fluorine. Water samples could be dried and the residue measured however an extraction technique would need to be employed for removing inorganic fluoride.^{8 19}F NMR has been used to identify and characterize organofluorine compounds⁹ and can also be used to quantify total organic fluorine in a sample by integrating multiple peaks associated with organofluorines. ¹⁹F NMR has been developed for measuring total concentrations of PFAS by monitoring the chemical shift associated with the terminal CF₃ peak.^9,10 One of the advantages of ¹⁹F NMR is that most interferences from common classes of organofluorine pesticides or pharmaceuticals, as well as inorganic fluorine, are eliminated. In the case of CIC, total fluoride is detected by IC after formation of hydrogen fluoride during hydropyrolysis of liquid or solid samples containing organofluorine compounds. Like other methods, CIC methods require sample preparation techniques to reach environmentally relevant detection limits. Additionally, EOF-CIC and AOF-CIC methods are not reliant on the availability of specific individual PFAS standards so will capture more PFAS than targeted methods, but the measured totals may also include other fluorinated organic chemicals such as fluorinated pesticides and pharmaceuticals. EOF-CIC uses conventional SPE techniques like those used in EPA Methods 537.1 ¹¹ and 533 ¹² to create a liquid extract that is analyzed by CIC and quantitated as fluoride ion. One limitation of EOF is that only specific PFAS that are retained on the SPE sorbent will be detected. For example, the SPE sorbent used in Method 537.1 does not retain PFAS with less than a six-carbon chain and the sorbent in Method 533 will not retain neutral PFAS. Alternatively, AOF methods adsorb PFAS onto granular activated carbon (GAC), combust the GAC and quantitate fluoride ion. AOF will adsorb both ionic and neutral PFAS, but does not quantitatively retain all small carbon chain (≤4) PFAS.¹³

The Clean Water Act prohibits the discharge of unauthorized pollutants from a point source to U.S. waters.¹⁴ Wastewaters entering wastewater treatment plants (WWTP) can contain organofluorine chemicals from residential, industrial, agricultural, and business contributions to the WWTP influent. The exact composition of organofluorine chemicals in these diverse sources is difficult to obtain, and the true load of organofluorines may be underestimated by applying only targeted methods to their analysis. Thus, standardized screening methods are needed for the aggregate measurement of man-made organofluorine chemicals in wastewaters. These standardized screening methods can be used to monitor organofluorine concentrations in WWTP influents and effluents to aid in optimizing treatment processes and minimize fluorine containing chemicals from being discharged to the receiving waters.

While there are no, widely available, inexpensive analytical tools to measure total organofluorine or total PFAS currently, AOF-CIC was chosen to investigate in the development of a wastewater screening method for adsorbable organofluorine. AOF-CIC has been applied to surface waters, groundwaters,^15,16 and wastewaters.^13,16,17 This manuscript discusses the development of an EPA standardized screening method for quantification of AOF in wastewaters with detection by CIC. This research addresses laboratory reagent blank (LRB) levels of AOF, and the impact of inorganic fluoride removal efficiency on PFAS recoveries. Recovery data are reported for 35 PFAS, two PFAS mixtures, two fluorinated pharmaceuticals, and two fluorinated herbicides in laboratory fortified blanks (LFBs; fortification of deionized water). AOF measurements in five wastewater treatment plant effluents are reported as well as PFAS laboratory fortified sample matrix (LFSM) recovery data in 13 challenging wastewater matrices and two surface water matrices. Chemical names for PFAS acronyms used in this report can be found in ESI Table S1.^†

Materials and methods

Reagents/supplies

Optima LC/MS grade methanol was purchased from Thermo Fisher Scientific (Waltham, MA). Ultrex II Ultrapure Nitric acid was purchased from Mallinckrodt Baker (Phillipsburg, NJ). Potassium nitrate (KNO₃; ≥99.0%), sodium thiosulfate (Na₂S₂O₃; 99%) and the 1000 mg L⁻¹ primary fluoride standard were purchased from Sigma-Aldrich (St. Louis, MO). The 1000 mg L⁻¹ second source fluoride standard was purchased from Reagents (Belmont, NC). Individual stock PFAS solutions were purchased from Wellington (Ontario, Canada) at 50 ng μL⁻¹ in methanol except for PFPeS (50 ng μL⁻¹), PFOS (100 ng μL⁻¹), PFOA (100 ng μL⁻¹), PFDA (100 ng μL⁻¹), PFHxS (50 ng μL⁻¹), and PFBS (50 ng μL⁻¹) which were purchased from Accustandard (New Haven, CT) in methanol. Oxyfluorfen and trifluralin were purchased from Accustandard at 1000 ng μL⁻¹, whereas fluconazole and atorvastatin were purchased from Absolute Standards (Hamden, CT) at 1000 ng μL⁻¹. The PFAS mixture used in the study was purchased from Wellington Laboratories (P/N: EPA-537PDS-R1) and contained 18 PFAS. Quartz Pallflex 85 mm type 2500QAT-UP filters were purchased from Pall (Port Washington, NY). GAC columns were purchased from Mandel Scientific (Houston, TX). Deionized water was obtained from a MilliporeSigma (Burlington, MA) Milli-Q IQ7000 system.

Sample collection

Surface waters #1 and #2 were collected in August of 2021, wastewater effluents #1 and #11–13 in May of 2021 and wastewaters #2–10 were collected in November of 2018. All waters were stored refrigerated until adsorption and analysis by AOF-CIC. Wastewaters #2 and #4 were manually vacuum filtered using quartz filters prior to GAC adsorption due to high total suspended solids.

AOF

Two adsorption units were used for these studies: a Nittoseiko Analytech (formerly Mitsubishi Chemical Analytech); (Yamato, Kanagawa, Japan) TXA-04 and an Analytik Jena (Jena, Germany) APU sim. Two 40 mg Mitsubishi GAC columns were used in each Mitsubishi column holder apparatus on both units to ensure comparable data between the two adsorption units. The 100 mL samples were loaded onto the GAC columns at a flowrate of 3 mL min⁻¹ with one of the adsorption units. The GAC columns (80 mg total GAC) were washed with 20 mL of 8.2 g L⁻¹ KNO₃ to remove inorganic fluoride. Using a 10 mL polypropylene syringe, 3 mL of reagent water followed by 3 mL of air was slowly pushed through each GAC column holder. This rinse of the GAC removes residual KNO₃ and prolongs the life of the pyrolysis tube by decreasing the potential for devitrification. To study breakthrough of organofluorine chemicals, the top GAC and bottom GAC from each column holder apparatus were pushed into separate ceramic boats using a T-shaped GAC column rod and analyzed by CIC.

CIC

A Nittoseiko Analytech combustion unit was coupled to a Thermo Scientific Dionex Integrion IC and comprised of a solids autosampler (ASC 270LS), a furnace (AQF-2100H), an absorption module (GA 211) and an external solution selector (ES-211). The halides produced by the combustion unit were analyzed by a Thermo Scientific Dionex Integrion HPIC system.

The CIC parameters used for analyses are shown in Table 1. Initially, chromatographic separation of fluoride was performed on a Dionex IonPac AS20 2 mm × 250 mm column and a Dionex IonPac AG20 2 mm × 50 mm guard column. All data produced using the AS20 column are noted in the tables. Otherwise, the Dionex IonPac AS24 2 mm × 250 mm separator column and a Dionex IonPac AG24 2 mm × 50 mm guard column were used to increase the linear dynamic range. The total run-time was 20 min on both columns. Control of the CIC system (combustion, absorption, and ion chromatography) and analysis of the results were performed by means of the software NSX-2100 version 10.2.5 and Chromeleon version 7.2.10. All AOF-CIC results are background subtracted by the LRB in the case of LFBs, or by the native AOF concentration in the case of LFSMs, using the following equation

$$\text{AOF}=(C_{\text{S}}\frac{V_{\text{abs}}}{V_{\text{S}}}-C_{\text{(LRB or native)}})$$

where AOF is total concentration of adsorbable organic fluorine in μg L⁻¹, C_S is sample concentration (sum concentration of top and bottom GAC) in μg L⁻¹ reported without blank subtraction and dilution factors, V_abs is total volume of absorption solution in mL of the sample prior to IC injection, C_{LRB or native} is either the LRB or native concentration (sum concentration of top and bottom GAC) in μg L⁻¹, and V_S is sample volume in mL.

Table 1.

CIC conditions

IC conditions for Dionex IonPac AS20 2 × 250 mm column			IC conditions for Dionex IonPac AS24 2 column × 250 mm column
Time (min)	mM KOH	Curve	Time (min)	mM KOH	Curve
Initial	4.0	5	Initial	8.0	5
3.500	4.0	5	3.500	8.0	5
9.000	75	8	9.000	75	9
10.250	75	5	10.250	75	5
10.251	4.0	5	10.251	8.0	5
20.000	4.0	5	20.000	8.0	5
IC parameters
Flow rate		0.3 mL min−1
Injection volume		100 μL
Column temperature		30 °C
Suppressor		Dionex ADRS 600 2 mm in constant current mode
Detector cell temperature		35 °C
Combustion parameters
Furnace inlet temperature		900 °C
Furnace outlet temperature		1000 °C (minimum)
Pyrolysis tube		Quartz tube with ceramic insert and quartz wool
Argon carrier flow		200 mL min−1
Oxygen flow		400 mL min−1
Humidified argon flow		100 mL min−1
Water supply scale		2
Sample boat		Ceramic
Absorption solution		Water
Final absorption solution volume		10.99 mL
Autosampler combustion program
Position (mm)		Wait time (s)	Speed (mm s−1)
90		60	10
End		600	10
Cool		60	40
Home		120	20

Calibration

External IC calibration was performed using sodium fluoride standards injected directly into the IC (no combustion). A 10 mg L⁻¹ primary stock solution was prepared in deionized water from the purchased 1000 mg L⁻¹ primary fluoride standard. Calibration standards were prepared in deionized water from the 10 mg L⁻¹ primary fluoride stock solution at 2, 5, 10, 50, 100, 250 and 500 μg L⁻¹. The calibration curve was verified with a second source 1000 mg L⁻¹ fluoride standard. A 10 mg L⁻¹ secondary stock solution was prepared in deionized water from the purchased 1000 mg L⁻¹ secondary fluoride standard. A second source quality check standard (QCS) was prepared in deionized water from the 10 mg L⁻¹ secondary fluoride stock solution at 50 μg L⁻¹ and quantitated at least quarterly using the primary standard calibration curve. The QCS recovery must meet quality control criterion of 85–115% to ensure accurateness of the primary standard calibration curve. Fluoride was quantitated using a quadratic 1/x weighted calibration curve.

MDL procedure

The sensitivity limits of this AOF-CIC method were determined using the EPA’s Office of Water method detection limit (MDL) procedure.¹⁸ In summary, nine LRBs and nine LFBs fortified with PFPeS at a fluoride concentration of 5.0 μg L⁻¹ were analyzed over four days by AOF-CIC using the TXA-04 adsorption unit. For the APU sim adsorption unit, seven LRBs and seven LFBs fortified with PFPeS at a fluoride concentration of 5.0 μg L⁻¹ were analyzed over three days by AOF-CIC. The MDL_b (MDL based on method blanks) of the LRBs and the MDL_s (MDL based on spiked samples) of the LFBs for each adsorption unit were calculated by multiplication of the Student’s t-value (99% confidence level with n – 1 degrees of freedom) by the standard deviation of the replicate analyses. The reported MDL for each adsorption unit was the greater of MDL_b or MDL_s.

Results/discussion

Background AOF levels

To assess blank AOF levels (Table 2), six LRBs were analyzed using six different lots of GAC. The mean AOF blank contamination was measured at 0.65 μg L⁻¹ in 100 mL of deionized water. The combustion of six empty ceramic boats produced a mean fluoride concentration of 0.032 μg L⁻¹ (calculated based on 100 mL samples) eliminating the ceramic boats and the combustion process as the source of the fluoride blank contamination in the LRBs (Table 2). Unused dry GAC from 12 glass columns (two of each lot) was pushed directly into empty ceramic boats and combusted. To mimic LRBs, the two dry GACs with the same lot number were summed for total of six replicates with a mean AOF concentration of 0.68 μg L⁻¹ (Table 2). Similarly, 12 GAC columns (two of each lot) were KNO₃ washed, combusted, and the two GACs with the same lot number were summed for a total of six replicates with a mean AOF concentration of 0.67 μg L⁻¹ (Table 2). There is no significant difference in the mean AOF measured for the LRBs, dry GAC, or KNO₃ washed GAC indicating the AOF baseline contamination is originating from the carbon sorbent used in the adsorption columns and that it cannot be removed with a KNO₃ pre-wash. Ultimately, the only way to lower the LRB concentration, and thus the MDL, using this commercially available GAC sorbent is to adsorb a larger sample volume.

Table 2.

Baseline contamination of AOF LRBs, empty ceramic boats, dry combusted GAC, and KNO₃ washed GAC^a

Boat #	Measured AOF concentration, μg L−1							% RSD
	1	2	3	4	5	6	Mean
Ceramic boats	0.041	0.031	0.030	0.029	0.030	0.033	0.032	14
GAC lot #	1	2	3	4	5	6
LRBs	0.77	0.48	0.66	0.48	0.76	0.75	0.65	21
Dry GAC	0.72	0.53	0.76	0.51	0.76	0.80	0.68	19
KNO3 washed GAC	0.51	0.51	1.04	0.44	0.75	0.77	0.67	34

^aAll calculations based on 100 mL sample volume and 10.99 mL absorption solution volume. TXA-04 adsorption unit used.

Contamination of the GAC can also lead to inconsistencies if the contamination is not reproducible and below the MDL. Over the course of method development, 95 LRBs, representing seven lots of GAC, were analyzed by CIC resulting in a mean concentration of 0.76 μg L⁻¹ with 31% RSD and a range of 0.43–2.0 μg L⁻¹. High LRB results (defined here as >1.0 μg L⁻¹) were never observed in an entire batch of GAC but rather encountered in single GAC columns of lots that were shown to have typical background levels. When LRBs and LFBs fail QC criteria (to be determined after multi-laboratory validation), data have to be discarded which could mean loss of samples and revenue for commercial laboratories running AOF-CIC. Additionally, this potential for single-column GAC contamination could occur in sample analyses unknowingly unless duplicate sample analyses are performed. Over the course of our studies, 7.4% of the LRBs were above 1.0 μg L⁻¹.

Inorganic fluoride removal efficiency

To measure AOF, the inorganic fluoride present in many environmental samples must be removed prior to CIC analysis. The accuracy of the AOF-CIC measurement will depend on the removal efficiency of inorganic fluoride from the samples. Two common methods of removing inorganic fluoride were found in the literature. Adsorbable organic halide methods, such as EPA Methods 9020B¹⁹ and 1650 ²⁰ (neither measures AOF), remove inorganic halides by washing the GAC with 8.2–8.5 g L⁻¹ NaNO₃ or KNO₃ after sample loading. Others have added 1.0 g L⁻¹ NaNO₃ to the sample as well as in the wash step to aid in the removal of inorganic fluoride from the GAC.¹³ KNO₃ was chosen in our studies over NaNO₃ to reduce devitrification of the pyrolysis tube.

A study was conducted to evaluate the efficiency of inorganic fluoride removal in LRBs, with and without the presence of 2.0 mg L⁻¹ inorganic fluoride, containing (a) 1.0 g L⁻¹ KNO₃ in the sample and 1.0 g L⁻¹ KNO₃ in the wash step, (b) 8.2 g L⁻¹ KNO₃ only in the wash step and (c) 1.0 g L⁻¹ KNO₃ only in the wash step. Triplicate 100 mL LRBs were prepared for each of the above conditions and the AOF measured. A mean AOF value of 0.56 μg L⁻¹ was obtained for LRBs (no added inorganic fluoride) with 1.0 g L⁻¹ KNO₃ in the sample and the wash step and 0.87 μg L⁻¹ if the GAC is only washed with 8.2 g L⁻¹ KNO₃ (Table 3). The addition of 2 mg L⁻¹ inorganic fluoride (source was sodium fluoride) to the LRBs in (a), (b) and (c) produced mean AOF values of 0.83 μg L⁻¹, 1.6 μg L⁻¹ and 4.9 μg L⁻¹, respectively. These data support the addition of KNO₃ to the sample in addition to the GAC wash step as in previous reports^13,15,17,21 for more efficient removal of inorganic fluoride from the GAC.

Table 3.

Inorganic fluoride removal efficiency – AOF concentrations and relative standard deviations (RSDs) in acidic and neutral LRBs (n = 3) containing inorganic fluoride using KNO₃ in the sample + GAC wash step versus in only the GAC wash step^a

pH of LRBs	KNO3 conc. in LRBs g L−1	KNO3 wash conc. g L−1	Inorganic F conc. mg L−1	Mean AOF conc. μg L−1, n = 3	RSD%
Neutral	1.0	1.0	0.0	0.56	2.3
Neutral	0.0	8.2	0.0	0.87	32
Neutral	1.0	1.0	2.0	0.83	5.5
Neutral	0.0	8.2	2.0	1.6	6.3
Neutral	0.0	1.0	2.0	4.9	14
<1.0	0.0	1.0	2.0	9.3	9.8

^aTXA-04 adsorption unit and AS20 separator column used.

Although adding KNO₃ to samples as well as the GAC wash step more efficiently removed the inorganic fluoride in LRBs, adverse effects were observed for some PFAS when KNO₃ was added to the sample. Two studies were conducted comparing AOF-CIC measurements of LFBs (n = 3) individually fortified with 35 PFAS in 100 mL of deionized water containing (a) 1.0 g L⁻¹ KNO₃ in the LFBs and in the GAC wash step and (b) 8.2 g L⁻¹ in the GAC wash step only. The 35 individual PFAS were selected to cover a range of chain lengths, from C4 to C14, and functional groups. Using 1 g L⁻¹ KNO₃ only in the wash step with these LFBs, instead of 8.2 g L⁻¹, was not pursued due to the poor inorganic fluoride removal demonstrated in Table 3. Fourteen PFAS shown in Fig. 1 had a change [(b – a)/a × 100] of greater than 20% (all recovery data can be found in ESI Table S2^†). Five PFAS had higher recoveries (22–34%) with the KNO₃ added to the sample while nine PFAS had higher recoveries (27–145%) without the KNO₃ in the sample. PFAS are surfactants and the more hydrophobic, longer chains are known to adsorb to surfaces.^22–24 Adding KNO₃ to the sample may introduce a salting out effect that increases the adsorption of the more hydrophobic, longer chains to the sample bottles, tubing and syringes used in the sample loading process onto the GAC. Except for PFBA, all the PFAS in Fig. 1 contain eight or more carbons. Recovery of PFBA was higher with the addition of KNO₃ to the sample probably because the KNO₃ decreased the water solubility of PFBA thereby decreasing its breakthrough on the GAC columns. Why N-AP-FHxSA, N-MeFOSE, FDET and PFTeA had better performance with KNO₃ in the sample is unknown, however it should be noted these four PFAS are not well recovered, and surface adsorption can be irreproducible. Although no one condition is optimum for all PFAS, the 8.2 g L⁻¹ KNO₃ only in the wash step resulted in significantly higher recoveries for more of the PFAS analyzed in this study. Therefore, 8.2 g L⁻¹ KNO₃ in the wash step only was chosen as optimum for this screening method.

Fig. 1.

Effect of KNO₃ on AOF-CIC measurements (n = 3) using the TXA-04 of LFBs individually fortified with selected PFAS in 100 mL of deionized water containing (a) 1.0 g L⁻¹ KNO₃ in the LFBs and in the GAC wash step and (b) 8.2 g L⁻¹ in the GAC wash step only. Percent change (y-axis) is calculated as [(b – a)/a × 100] using experimental recoveries obtained for a and b. (AS20 separator column).

The effectiveness of the 8.2 g L⁻¹ KNO₃ wash step at removing inorganic fluoride (0, 1, 2, 4, 6, and 8 mg L⁻¹) in LFBs and LFSMs, fortified with PFPeS at a fluoride concentration of 5.1 μg L⁻¹, is shown in Fig. 2. AOF recoveries of PFPeS LFBs increased with increasing inorganic F concentration ≥2 mg L⁻¹. In contrast, the 8.2 g L⁻¹ KNO₃ wash effectively removed the inorganic fluoride up to 8 mg L⁻¹ in the surface water and wastewater PFPeS LFSMs. These results are consistent with Wagner and colleagues who proposed that natural organic matter in tap water samples may compete with inorganic fluoride adsorption on GAC, resulting in less retention of the inorganic fluoride on the GAC.¹⁵ It is not known if this matrix effect will be universal across all wastewater matrices. The highest inorganic fluoride measured in the surface water and wastewater samples evaluated in this study was 1.3 mg L⁻¹, thus the 8.2 g L⁻¹ KNO₃ wash used in these studies was sufficient to remove native inorganic fluoride present in these surface and wastewaters.

Fig. 2.

Effect of inorganic fluoride concentration on recovery of PFPeS (spiked at 5.1 μg L⁻¹ fluoride) in LFBs, surface water and wastewater effluent using the TXA-04. (AS20 separator column) * n = 4 for surface water and wastewater LFSMs with 0 ppm inorganic fluoride. Wastewater LFSMs were not conducted at 1 ppm to conserve the collected water for the higher inorganic F concentrations where adverse results may be encountered.

Sample preservation

Typically, EPA methods stipulate sample preservation agents that are used to prevent microbial degradation, to dechlorinate samples (if residual chlorine is present), to stabilize the pH, or prevent chemical degradation of the analytes. Given that this screening method measures total AOF, it is not possible to study the degradation of all organofluorine compounds. However, in the case of PFAS, if degradation occurs and the carbon chain length is >4, the degradation product likely will still contain fluorine, due to the strength of the carbon–fluorine bond and will be detected by AOF-CIC. Acidification of the sample with nitric acid has been used to prevent microbial growth;^13,15,17 however, in our studies, acidification of LRBs to pH <1 with nitric acid in the presence of 2.0 mg L⁻¹ inorganic fluoride and 8.2 g L⁻¹ KNO₃ used in the GAC wash step resulted in carryover of 9.3 μg L⁻¹ inorganic fluoride into the AOF-CIC measurement (Table 3) and 306% recovery (data not shown) of LFBs acidified with nitric acid to pH <1 and fortified with PFPeS at a fluoride concentration of 5.1 μg L⁻¹. The pK_a of hydrogen fluoride (HF) is 3.19,²⁵ so at a sample pH of less than 3.19, half or more of the HF is not dissociated resulting in increased adsorption of inorganic fluoride to the GAC. Based on these results, this screening method requires the sample pH to be ≥5.0 and refrigerated at or below 6 °C, but not frozen, until sample adsorption and analysis. None of the samples tested in this report contained free chlorine but sodium thiosulfate (8 mg/100 mL sample) was tested in LFBs (n = 3) spiked with PFPeS at a fluoride concentration of 7.0 μg L⁻¹ with no adverse effects (111% ± 1.9% recovery; data not shown). Thus, sodium thiosulfate could be used to dechlorinate wastewater or surface water samples containing free chlorine.

LFB recoveries on two adsorption units

Mean recoveries of 35 PFAS, two fluorinated pharmaceuticals and two fluorinated herbicides were evaluated on two adsorption units. Using the final method conditions of 8.2 g L⁻¹ KNO₃ only in the GAC wash step, the mean recoveries (n = 3) of the 35 PFAS ranged from 21–116%, the pesticides were 48–50%, and the pharmaceuticals were 87–101% using the TXA-04 adsorption unit (ESI Table S3^†). The APU sim produced mean recoveries (n = 3) of 38–100% for the PFAS, 50–51% for the pesticides, and 87–98 for the pharmaceuticals (ESI Table S3^†). Comparing the two adsorption units, 11 PFAS, shown in Fig. 3, have a % change [defined as (b – a)/a × 100 where b = the APU sim adsorption unit and a = the TXA-04 adsorption unit] greater than 20% using the APU sim unit and one PFAS, PFDPA, with 23% greater recovery on the TXA-04. The Mitsubishi GAC columns and column holders were used on both adsorption units to ensure comparable data if the units were used concurrently to gather data. This also served to eliminate the GAC as the source of differences observed between the two units. These 11 PFAS have significantly increased recoveries with the APU sim adsorption unit. This may be due to the APU sim having less surface area (no transfer lines from the sample bottle to the syringes; samples directly poured into 100 mL syringes) for the PFAS to adsorb to prior to adsorption onto the GAC.

Fig. 3.

AOF mean recovery variations (n = 3) in LFBs analyzed on two adsorption units: (a) TXA-04 and (b) APU sim. Percent change (y-axis) is calculated as [(b − a)/a × 100] using experimental recoveries obtained for a and b.

It has been suggested that some low PFAS recoveries may be due to incomplete combustion.²⁶ To evaluate the possibility of incomplete combustion, 10 analytes with mean recoveries ≤70% (FDEA, FDUEA, 10 : 2 FTS, PFTrA, FDET, 11Cl-PF3OUdS, N-Me-FOSE, N-EtFOSE, trifluralin, and oxyfluorfen) on the APU sim were spiked directly from the individual stocks onto the top GAC and dried in a 55 °C oven for 10 minutes to evaporate the methanol. The top GAC column was then placed in a column holder along with a bottom GAC column and washed with 8.2 g L⁻¹ KNO₃. The design of this incomplete combustion experiment ensures any losses observed are not due to surface adsorption to sample bottles or adsorption unit tubing. The KNO₃-washed spiked GAC mean recoveries (n = 3) for the 10 analytes with low AOF recoveries were 91–116% compared to mean AOF recoveries of 38–71% using the APU sim. This study eliminates incomplete combustion as the source of the low recoveries of the longer chain PFAS, trifluralin, and oxyfluorfen in our studies. Instead, as noted previously, surface adsorption is the more likely reason for the ≤70% recoveries in the case of the eight PFAS. Trifluralin and oxyfluorfen are both zwitterions which may play a role in their adsorption to GAC. Ultimately, these LFB data on two adsorption units suggest that not all AOF data in the literature may be comparable depending on the design of the adsorption unit. Additionally, the inability to mitigate surface adsorption during AOF measurements results in a potentially low bias for some PFAS, depending on the short versus long carbon chain composition of the sample.

These LFB recovery data were also evaluated for selection of a PFAS used for fortification of LFB and LFSM quality control samples in the AOF-CIC wastewater method. Longer chain PFAS were eliminated due to surface adsorption resulting in poor recoveries. PFAS with six or seven carbons are well recovered in LFBs but would not mimic breakthrough of the shorter chain PFAS on the GAC in the presence of matrix interferences. Thus, the five-carbon chain PFPeS was chosen for the LFB and LFSM samples as it was consistently well recovered in LFBs and yet not so strongly adsorbed to the GAC that breakthrough in LFSMs could not be measured. Mean recovery (n = 61) of 100% with 11% RSD was obtained for LFBs fortified with PFPeS at fluoride concentrations of 5.1–40 μg L⁻¹ indicating PFPeS performed well as a QC analyte for LFBs. Mean recovery (n = 60) of 93% with 27% RSD was obtained for LFSMs fortified with PFPeS at fluoride concentrations of 5.1–40 μg L⁻¹ indicating PFPeS performed well as a QC analyte for LFSMs also.

Breakthrough

In addition to total AOF recoveries, the percentage of the total AOF (top + bottom GAC) measured on the top GAC and bottom GAC was evaluated for each of the 39 organofluorine chemicals in this study. Four PFAS had more than 5% of the total measured AOF quantified on the bottom GAC in LFBs, indicating breakthrough of the top GAC for these PFAS in deionized water (Fig. 4). This breakthrough is not surprising for PFBA and PF4OPeA as they are short-chain, hydrophilic PFAS. The six-carbon chain PFHxPA may be exhibiting breakthrough because it is estimated to be more water soluble (515 g L⁻¹) than analogous six-carbon chain PFAS such as PFHxA (21.7 g L⁻¹) and PFHxS (2.3 g L⁻¹).²⁷ FDET is a 10-carbon chain and not expected to breakthrough to the second GAC especially since FBET, a six-carbon chain, did not exhibit breakthrough. No reason has yet been determined for the breakthrough of 25% of FDET onto the bottom GAC.

Fig. 4.

Breakthrough onto bottom GAC of four PFAS in LFBs fortified at a fluoride concentration of 9.9–10.1 μg L⁻¹ using the APU sim. Breakthrough was calculated as the % of the AOF measured on the bottom GAC divided by the total AOF measured (n = 3).

MDL study

An MDL study was performed for each adsorption unit according to the EPA Office of Water’s 2017 MDL procedure¹⁸ and the MDL was based on the total of the top and bottom GAC for each sample. For the Mitsubishi TXA-04, nine LFBs and nine LRBs were analyzed producing an MDL of 2.2 μg L⁻¹. Similarly, seven LFBs and seven LRBs were analyzed on the Analytik Jena, resulting in an MDL of 1.4 μg L⁻¹. These MDLs are comparable to detection limits reported by von Abercron, et al.¹³ Others have reported sub μg L⁻¹ LODs with homemade sorbents or sorbents no longer commercially available.^15–17 Han and colleagues reported a limit of detection (LOD) of 300 ng L⁻¹ using activated carbon, however, 300 mL sample volumes were used to reach those detection limits.²⁸ While increasing our sample size would lower our MDL, adsorption of larger sample volumes will likely decrease the recovery of the short-chain organofluorine compounds due to break through on the GAC especially in sample matrices other than deionized water where natural organic matter will compete for adsorption onto the GAC. In addition, loading more than 100 mL of wastewaters containing significant total suspended solids is unlikely due to clogging of the GAC columns.

WWTP effluents

Five WWTP effluents and two surface waters (wastewaters #1, 11–14 and surface waters #1 and 2 in Table 4) were tested using this screening method. The AOF measured for the wastewater effluents was 2.5 to 3.3 μg L⁻¹, except for one wastewater that was less than the MDL of 2.2 μg L⁻¹, using the TXA-04 adsorption unit. The AOF measured for the two surface waters was 1.9 and 5.4 μg L⁻¹ using the APU sim adsorption unit. Native AOF concentrations for wastewaters #2–10 are not reported due to the age of the samples.

Table 4.

AOF recovery^a,b and precision of LFBs (n = 9), surface waters (n = 3), and wastewaters (n = 3) fortified with PFPeS

	Source	Fluoride fortification conc., μg L−1	Total mean AOF conc., μg L−1	Mean recovery%	RSD, %	% of total AOF on bottom GAC
LFBs	Deionized water	10.1	10	101	11	2.7
Surface water #1	Retaining pond	10.1	12	120	2.3	14
Surface water #2	Lake	10.1	9.3	91.9	8.1	13
Surface water #2	Lake	40.2	42	104	1.8	15
Wastewater #1b	WWTP effluent	7.02	7.6	109	7.5	20
Wastewater #2c	Landfill leachate	19.9	7.7	38.7	1.1	46
Wastewater #3d	Metal finisher	19.9	18	91.8	12	26
Wastewater #4c	WWTP effluent	10.1	8.5	84.1	1.9	30
Wastewater #5	Hospital	10.1	9.6	94.8	1.7	13
Wastewater #6d	WWTP influent	10.1	7.8	77.4	29	10
Wastewater #7	Bus washing station	10.1	8.1	80.2	11	38
Wastewater #8	Power plant	10.1	8.9	8	1.6	2.3
Wastewater #9	Pulp & paper effluent	10.1	9.1	90.3	4.9	28
Wastewater #9	Pulp & paper effluent	40.2	36	89.4	3.3	32
Wastewater #10	WWTP effluent	10.1	8.7	85.9	7.1	14
Wastewater #11b	WWTP effluent	7.02	7.8	111	0.94	9.0
Wastewater #12b	WWTP effluent	7.02	6.3	89.1	20	15
Wastewater #13b	WWTP effluent	7.02	7.7	109	4.6	27
Wastewater #14b	WWTP effluent	5.05	4.8	95.7	3.5	21

^aAll measurements corrected for native AOF concentration. APU sim adsorption unit used except where otherwise noted.

^bAS20 separator column used and TXA-04 adsorption unit.

^cWastewater vacuum filtered manually prior to spiking due to high total dissolved solids that clogged the GAC and adsorption unit tubing.

^dQuartz wool prefilter was used on top of the GAC to prevent clogging of the GAC with particulates. The quartz wool prefilter was also combusted and added to the total AOF measured.

LFSMs were prepared from the 16 matrices in Table 4 by fortification with PFPeS at organic fluoride concentrations of 5.1–40.2 μg L⁻¹ and analyzed by AOF-CIC. Mean LFSM recoveries of 77–112% with ≤30% RSD were obtained in the seven WWTP effluents and 92–120% with ≤20% RSD for the two surface waters as shown in Table 4. More challenging wastewaters #2, 3, and 5–9 (influents to WWTPs with no or minimal treatment) were also tested with LFSMs fortified with PFPeS at 10.1–40.2 μg L⁻¹ organic fluoride concentrations. Wastewaters #2 and #3 had to be spiked at 20 μg L⁻¹ due to high native AOF concentrations of 24 and 34 μg L⁻¹, respectively. Wastewaters #2 and #4 were manually vacuum filtered prior to fortification due to high total suspended solids (168–244 mg L⁻¹) that clogged the GAC and tubing of the adsorption unit. Mean LFSM recoveries were 77–94% with ≤29% RSD except for the poor LFSM recovery of 39% in landfill leachate (wastewater #2) which was by far the most challenging matrix of the ones studied. Except for wastewater #2, the LFSM recoveries were within ±23% of the PFPeS LFB mean recovery of 99%.

Table 5 contains similar LFSM data to Table 4 for the same two surface waters and nine of the wastewaters except these LFSMs were fortified with a mixture of the 18 PFAS included in EPA Method 537.1¹¹ at organic fluoride concentrations of 10.0–20.0 μg L⁻¹. Mean LFSM recoveries were 55–88% with ≤34% RSD except for wastewaters #2 (34%) and #7 (119%). Thus, all the surface waters and seven of the nine wastewaters were comparable to the LFB mean recovery of 74% considering the 16% RSD. Given the inability to mitigate PFAS surface adsorption to the sample bottles or adsorption unit sample path, it is not surprising the PFAS mixture recoveries in LFBs, surface waters and wastewaters are lower than the PFPeS spiked samples in Table 4 (excluding wastewater #7). Wastewater #7 (bus washing station) is the only matrix in Table 5 that resulted in recoveries above 80%. Non-homogeneity of the sample is suspected resulting in the possibility of more particulates with adsorbed organofluorine chemicals combusted with the GAC. The landfill leachate (wastewater #2) was by far the most challenging matrix of the ones studied.

Table 5.

AOF recovery^a and precision of LFBs (n = 9), surface waters (n = 3), and wastewaters (n = 3) fortified with a PFAS mixture^b

	Source	Fluoride fortification conc., μg L−1	Total mean AOF conc., μg L−1	Mean recovery%	RSD, %	% of total AOF on bottom GAC
LFBs	Deionized water	10.0	7.6	75.9	15	1.3
Surface water #1	Retaining pond	10.0	7.7	80.4	4.0	19
Surface water #2	Lake	10.0	7.6	76.4	10	14
Wastewater #2c	Landfill leachate	20.0	6.8	34.2	1.3	60
Wastewater #3d	Metal finisher	20.0	12.8	64.1	5.3	23
Wastewater #4c	POTW effluent	10.0	5.5	55.1	2.8	34
Wastewater #5	Hospital	10.0	6.2	62.0	9.0	12
Wastewater #6d	POTW influent	10.0	5.5	55.3	34	15
Wastewater #7	Bus washing station	10.0	11.9	119	15	35
Wastewater #8	Power plant	10.0	8.8	88.2	3.8	3.9
Wastewater #9	Pulp & paper effluent	10.0	7.2	71.9	13	28
Wastewater #10	POTW effluent	10.0	7.4	73.8	5.6	19

^aAll measurements corrected for native AOF concentration. APU sim adsorption unit used.

^bPFAS mixture (Wellington Laboratories P/N: EPA-537PDS-R1) containing 18 PFAS.

^cWastewater vacuum filtered manually prior to spiking due to high total dissolved solids that clogged the GAC and adsorption unit tubing.

^dQuartz wool prefilter was used on top of the GAC to prevent clogging of the GAC with particulates. The quartz wool prefilter was also combusted and added to the total AOF measured.

Surrogates, typically chemicals similar in nature to the method analytes, are frequently used in EPA methods to monitor the sample preparation and extraction process. Organofluorine compounds cannot be used as surrogates and frequent QC failures may be encountered if other ionic species (e.g., nitrate, chloride, sulfate, phosphate) are used as surrogates due to the prevalence of ionic compounds in environmental samples. Since surrogates are not feasible in this AOF technique, the amount of fluoride on the bottom GAC can be used as a measure of organofluorine breakthrough of the GAC columns. GAC breakthrough would require either dilution of the sample or a smaller sample size be adsorbed. Although more wastewaters will need to be tested in a multi-laboratory validation study before setting a breakthrough threshold, this study indicates a typical breakthrough range of 4–38% in these wastewaters with two exceptions at 46 and 60% in the landfill leachate LFSMs. Contaminants in the leachate wastewater appear to be overwhelming the capacity of the GAC causing breakthrough and low recoveries of the PFAS.

Chromatograms in Fig. 5 depict the separation of fluoride from nearby eluting organic acids and chloride on the AS24 separator column in a (a) 50 μg L⁻¹ calibration standard, (b) top GAC of LRB, (c) top GAC of wastewater #5, and (d) top GAC of wastewater #5 spiked with PFPeS at 10.1 μg L⁻¹ fluoride. Two peaks at retention times 8.69 and 11.17 min were determined to be acetate and formate ions, respectively. These organic acids are not retained on the GAC in AOF, thus peaks for these ions are not observed in the AOF chromatograms. However, if adequate separation is not achieved between fluoride and the organic acids, the fluoride calibration curve will be biased. Under the IC conditions presented in this report, fluoride and chloride are separated by 6.35 min, thus high levels of chloride in samples will not adversely affect the quantitation of fluoride.

Fig. 5.

Chromatograms of a (a) 50 μg L⁻¹ calibration standard, (b) top GAC of LRB, (c) top GAC of a wastewater effluent sample, and (d) top GAC of a wastewater effluent spiked with PFPeS at 10.1 μg L⁻¹ fluoride. Chromatograms stacked with a 5% offset.

Conclusions

An AOF-CIC screening method has been developed that demonstrates overall recoveries of 77–120% in 17 of the 18 tested matrices spiked with PFPeS and 55–119% in 11 of the 12 matrices spiked with a PFAS mixture. Poor method performance was obtained in the landfill leachate wastewater due to breakthrough. MDLs of 1.4–2.2 μg L⁻¹ were achieved using 100 mL sample volumes. These data reported by the authors can be used to set the quality control measures and criteria necessary to ensure comparable performance among laboratories using this wastewater screening method and the specific GAC used in this report.

This report demonstrates that this AOF technique can be a useful PFAS analytical tool in the balancing act between selectivity, sensitivity, and inclusivity. However, as with most analytical techniques, limitations must be recognized. While the GAC used in AOF is likely to capture more PFAS than SPE techniques used in EOF, surface adsorption of PFAS were not mitigated in this AOF procedure. Thus, losses of longer carbon chain PFAS were observed and the extent of the losses will be dependent on the surface area of the adsorption unit used. Although MDLs can be lowered using larger sample volumes, aqueous samples high in particulates are not amenable to larger sample volumes due to clogging of the GAC and the adsorption unit. Consequently, there is a need for commercially available GAC with lower AOF background levels or approaches to remove the background AOF prior to using the GAC.