A SYSTEMATIC EVALUATION OF FACTORS RESULTING IN INCOMPLETE RECOVERY WHEN IMPLEMENTING THE ADSORBABLE ORGANIC FLUORINE (AOF) AND EXTRACTABLE ORGANIC FLUORINE (EOF) METHODS A Thesis Presented to the Faculty of the Graduate School of Cornell University In Partial Fulfillment of the Requirements for the Degree of Master of Science by Yitong Pan August 2023 © 2023 Yitong Pan ABSTRACT Per- and polyfluoroalkyl substances (PFASs) are a global environmental contaminant of concern whose persistence in water systems is associated with negative effects on human health. The typical way for measuring PFAS concentrations in environmenta l samples is by means of liquid chromatography-mass spectrometry (LC-MS), but authentic analytical standards are only available for a few dozen out of thousands of individual PFAS species. Therefore, techniques that quantify total organic fluorine (TOF) are important for PFAS risk assessments. Adsorbable organic fluorine (AOF) and extractable organic fluorine (EOF) are techniques for measuring TOF. Both techniques rely on the loading of PFASs onto either activated carbon (AOF) or a WAX adsorbent (EOF) followed by a washing step to remove residual inorganic fluoride from the adsorbent. The EOF method also has an elution step that results in the transfer of captured PFASs to a small volume of an organic solvent. Both methods are completed with a combustion step to mineralize the PFASs and capture the evolved hydrogen fluoride for quantification; the AOF method performs the combustion on the washed activated carbon and the EOF method performs the combustion on the organic solvent eluate. The objective of this study was to systematically evaluate each step of the AOF and EOF methods for the recovery of a set of ten ultrashort-, short-, and long-chain perfluoroalkyl acids. We measured the overall recovery of fluoride for each method for each of the PFASs, and the recovery of fluoride and each PFAS around the loading, washing, and elution steps. We also measured the combustion efficiency of each PFAS by means of direct combustion without sample preparation. The overall AOF recovery ranged from 9.3% – 103.3% and the overall EOF recovery ranged from 21.0% – 108.1%, with higher recoveries generally measured for PFASs with increasing chain length in both methods. The three ultrashort-chain PFASs exhibited the lowest overall recoveries from 9.3 – 25.2% for AOF and 21.0 – 51.5% for EOF. Our systematic evaluation of each step in the AOF method revealed that decreases in the overall recovery are primarily the result of the loss of ultrashort- and short-chain PFASs during the washing step and the incomplete mineralization of perfluoroalkyl sulfonic acids (PFSAs) during combustion. Our evaluation of each step in the EOF method revealed that decreases in the overall recovery are primarily the result of the incomplete elution of short- and long- chain PFASs and the loss of ultrashort-chain PFASs during the washing step. Together, our experiments suggest that the EOF method is more appropriate than the AOF method for measuring TOF in samples containing significant fractions of ultrashort- and short- chain PFASs and that significant methodological improvement are possible with a focus on the washing, elution, and combustion steps. iii BIOGRAPHICAL SKETCH Yitong Pan is from Beijing, China. She received her Bachelor of Science with Honors degree in Marine Science from Xiamen University (XMU) in 2021. While studying at XMU, her research experiences include an eighteen-month research on persistent organic pollutants extraction from seawater and COD measurement influencing factor exploration in Cai research group and a two-year research on the biogeochemical cycle of phosphate at Jiulong River Estuary and optimization modeling on specific industr ia l coal-washing wastewater treatment in Wang research group. She also completed an eighteen-month internship as an AP physics instructor at Zhangmen 1 to 1 Education & Technology Group, a three-month internship as a seawater quality examiner at the State Key Laboratory of Marine Environmental Science, and a two-month internship as a conference coordinator at the Fifth Xiamen Symposium on Environmental Science. During the summer of 2019, she attended a four-week summer session at the Univers ity of Southern California with Prof. David A. Hutchins studying Our Future Changing Ocean. In the fall semester of 2019, she attended an international exchange program at the University of Amsterdam studying environmental chemistry. In August 2021, she joined the Environmental Processes focus area in the School of Civil and Environmenta l Engineering at Cornell University to pursue her Master of Science degree under the advisement of Prof. Damian E. Helbling. During the summer of 2022, she did a three- month internship as an associate scientist at Frontage Laboratories Inc. iv To the scientists who steadfastly devoted an unwavering commitment to the pursuit of knowledge and truth. v ACKNOWLEDGMENTS First, I would like to thank Professor Damian Eric Helbling for his help and support throughout my research project at Cornell University. Prof. Helbling is one of the most intelligent and hardworking people I have ever met. His expertise in the field of environmental engineering and insightful guidance on academic writing has provided me with a solid foundation for my future career path. I would also like to thank Professor Matthew Charles Reid for serving as the minor advisor on my committee and supporting the use of ion chromatography instruments in my research. Moreover, I am thankful to the members of the Helbling Research Group, past and present, for their kindness and support. Their encouragement has been instrumental in helping me overcome some of the toughest challenges I faced during my time here. I would also like to express my gratitude to all my friends and families, especially my parents, for their unconditional love and spiritual support. Their presence in my life has been a constant source of strength and inspiration, and I am forever indebted to them. Finally, I want to express gratitude to myself for my perseverance and dedication in the pursuit of science over the last few years. It has been a challenging journey, but I am grateful for the opportunities I have had to learn and grow. Though I may be just a tiny speck in the vast universe, I am always proud that I choose to be my own hero. vi 致谢 首先，我想感谢 Damian Eric Helbling 教授在我在康奈尔大学的研究项目中给予 我的帮助和支持。Helbling 教授是我所遇到的最聪明、最勤奋的人之一。他在环 境工程领域的专业知识和对学术写作的深入指导为我未来的职业道路打下了坚 实的基础。我还要感谢 Matthew Charles Reid 教授担任我毕业委员会的导师，并 支持我在研究中使用离子色谱仪器。 此外，我想感谢 Helbling研究小组的成员们，他们的理解和支持帮助我度过了最 困难的一段时期。我还要对所有的朋友和家人，特别是我的父母，表达感激之 情，感谢他们无条件的爱和精神上的支持。他们始终是我力量和灵感的源泉， 是上天给我最好的礼物。 最后，我要感谢自己在过去几年中在科研工作中的努力和坚持。这是一段充满 挑战的旅程，但我依旧感激于自己的成长和提升。即使我只是宇宙的尘埃，我 也无愧于选择成为自己的英雄。 vii TABLE OF CONTENTS Biographical Sketch .......................................................................................................... iii Acknowledgements ............................................................................................................v Table of Contents ............................................................................................................. vii List of Figures................................................................................................................... ix List of Abbreviations ..........................................................................................................x CHAPTER 1 - Introduction......................................................................................... 1 1.1 Per- and Polyfluoroalkyl Substances (PFASs) ........................................................... 1 1.2 Total PFAS Assessment........................................................................................... 5 1.2.1 Liquid Chromatography-Mass Spectrometry (LC-MS) .......................................... 5 1.2.2 Total Oxidizable Precursor (TOP) Assay .............................................................. 6 1.2.3 Total Organic Fluorine (TOF) ............................................................................. 7 1.3 AOF and EOF ........................................................................................................ 8 1.4 Research Objectives............................................................................................... 11 CHAPTER 2 - A systematic evaluation of factors resulting in incomplete recovery when implementing the adsorbable organic fluorine (AOF) and extractable organic fluorine (EOF) methods ............................................................................... 13 Abstract.......................................................................................................................13 2.1 Introduction ..........................................................................................................15 2.2 Materials and Methods ...........................................................................................18 2.2.1 Standards and Reagents ....................................................................................18 2.2.2 Adsorbable Organic Fluorine (AOF) Method .......................................................19 2.2.3 Extractable Organic Fluorine (EOF) Method .......................................................21 2.2.4 Evaluation of Each Step in the AOF and EOF Methods .........................................22 2.2.5 Evaluation of Combustion Efficiency .................................................................23 2.2.6 Fluoride Analys is by means of Ion Chromatography (IC) ......................................24 2.2.7 PFAS Analys is by means of Liquid Chromatography Mass Spectrometry (LC-MS) .25 viii 2.2.8 Data Analys is ..................................................................................................26 2.3 Results and Discussion...........................................................................................28 2.3.1 AOF and EOF Validation ..................................................................................28 2.3.2 Evaluation of AOF and EOF Methodological Steps ..............................................34 2.3.3 AOF and EOF Environmental Water Simulation ..................................................39 2.4 Conclusions and Implications .................................................................................44 CHAPTER 3 - Summary and future research ......................................................... 46 3.1 Summary ..............................................................................................................46 3.2 Future Research.....................................................................................................48 3.2.1 Supplement to the Evaluation ............................................................................48 3.2.2 Approaches for the Optimization .......................................................................51 3.3 Conclusion ............................................................................................................53 References .................................................................................................................... 54 Appendices ....................................................................................................................89 ix LIST OF FIGURES Figure 1-1. A general introduction of PFASs regarding their chemical structures, diverse applications, ubiquity in water systems, and harmful impacts on human health.......................1 Figure 1-2. PFAS concentration measurement with LC-MS..................................................6 Figure 1-3. PFAS TOF measurement with AOF-CIC.......................................................... 10 Figure 1-4. PFAS TOF measurement with EOF-CIC .......................................................... 11 Figure 2-1. Recoveries of inorganic fluoride in experiments with individual PFASs spiked in Milli-Q water and the results comparison with published literature values for the AOF and EOF methods respectively ........................................................................................................ 29 Figure 2-2. The bar plots represent individual PFAS recoveries with or without the fluoride washing when the mixture of ten PFASs are spiked in Milli-Q water for AOF and EOF methods respectively ..................................................................................................................... 35 Figure 2-3. The AOF and EOF fluoride recovery results with a mixture of ten PFASs or a mixture of ten PFASs with groundwater-level fluoride spiked ............................................. 40 Figure 3-1. A: The quantile plot of data from Han et al., (2020) illustrating the impact of environmental water matrix factors on AOF recovery. B: The PCA biplot shows the relationship between AOF recovery in different water matrices and selected water characteristics ............ 50 x LIST OF ABBREVIATIONS AC activated carbon AFFFs aqueous film-forming foams AOF adsorbable organic fluorine CIC combustion ion chromatography DOM dissolved organic matter EOF extractable organic fluorine IC ion chromatography LC-MS liquid chromatography-mass spectrometry LOD limit of detection LOQ limit of quantification PCA principal component analysis PFAAs perfluoroalkyl acids PFASs per- and polyfluoroalkyl substances PFCAs perfluoroalkyl carboxylic acids PFSAs perfluoroalkyl sulfonic acids POPs persistent organic pollutants SPE solid-phase extraction TOF total organic fluorine TOP total oxidizable precursor WAX weak anion exchange 1 CHAPTER 1 INTRODUCTION 1.1 Per- and Polyfluoroalkyl Substances (PFASs) The occurrence of ubiquitous anthropogenic organic chemical contaminants in environmental water systems has been of increasing concern in recent years.1-3 As one of the most important classes of persistent organic pollutants (POPs), per- and polyfluoroalkyl substances (PFASs) garnered rapid public attention for their toxicity with regard to the ecosystem and human health (Figure 1-1).4,5 Figure 1-1. A general introduction of PFASs regarding their chemical structures, diverse applications, ubiquity in water systems, and harmful impacts on human health. PFASs are a group of synthetic substances containing multiple high-strength carbon- fluorine bonds (488 kJ/mol).6-8 Benefiting from the fully or partially fluorinated hydrophobic carbon chain and hydrophilic head group, many PFASs own unique 2 physiochemical properties, such as extraordinary thermal and chemical stability, water and grease repellency, and relatively high water solubility.4,6,9,10 Therefore, PFASs have been widely employed in numerous industrial applications and commercial products including paper making, food packaging, chromium plating, aqueous film-forming foams (AFFFs), aviation hydraulic fluids, stain-resistant fabrics, non-stick cookware, and personal care products since the 1950s.11-13 Two main groups of PFASs widely used in the industry include perfluoroalkyl carboxylic acids (PFCAs) and perfluoroalkyl sulfonic acids (PFSAs).13,14 These compounds, collectively known as perfluoroalkyl acids (PFAAs), are also the major transformation products of many polyfluoroalkyl substances over varying lengths of time. There are more than 9000 unique PFAS chemical structures identified so far and the number is still increasing.15-17 But only a limited number of long- and short-chain PFASs (containing three or more fluorinated carbon atoms), such as perfluorooctanoic acid (PFOA) and perfluorooctanesulfonic acid (PFOS), have been well-studied on its environmental sources, fate, transport, and toxicity.18-21 Ultrashort-chain PFASs (containing fewer than three fluorinated carbon atoms), on the other hand, though highly mobile and equally toxic as the long-chain homologs,13 have been largely overlooked due to limited quantification accuracy with conventional approaches which is challenging due to their high polarity.22-24 For example, trifluoroacetic acid (TFA, C2), the simplest substance among the PFCA group, is a widely applied mobile phase component for pharmaceutical bioanalysis and one of the terminal products of the degradation and transformation from PFAS precursors.22,25,26 It has also been detected 3 in various environmental media such as surface water, wastewater, groundwater, and seawater.27,28 Perfluoropropionic acid (PFPrA, C3), the predominant PFCA species identified in precipitation samples, is usually thought to be generated from fluoropolymer manufacturing industries and the hydrolysis of a major component from a fire suppression agent.29-32 Its occurrence in high concentrations has also been reported in wastewater, landfill leachates, and groundwater.22,33 Trifluoromethanesulfonic acid (TFMS, C1), a super acid material, is produced at more than 100 tons a year for its application in electronics and optical equipment manufacturers, semiconductor industries, and organic syntheses.25,34,35 TFMS was detected at high frequencies in surface water, groundwater, landfill leachates, and hazardous waste management facilities.22,36 These findings suggest that current research may underestimate the impact of a large proportion of poorly studied PFASs, which may be an important contributor to PFAS contamination in water samples.37 Typically, PFAS concentrations in natural water systems are relatively low, but elevated levels have been observed in groundwater and surface water near point sources like military training areas and fluorochemical manufacturing plants.21,38,39 The significant annual production of PFASs and their resistance to degradation in water treatment plants have made drinking water the primary pathway of PFAS exposure to humans.40-42 Numerous studies have demonstrated that persistent exposure to PFASs is associated with a range of adverse health effects, including high cholesterol, reduced kidney function, immune suppression, thyroid disorders, reduced fertility and fecundity, and increased cancer risks.42-47 In response to these findings, a cascade of regulations has 4 been implemented in recent years to limit PFAS levels in water matrices.48-52 Starting in 2006, eight major PFAS precursor producers joined the US Environmental Protection Agency (EPA) Stewardship Program towards the elimination of long-chain PFCAs.20 In 2009, the International Stockholm Convention listed PFOS and related substances under restricted or eliminated production or used chemicals.53-56 In 2013, long-chain PFCAs were identified as substances of high concern under the European chemicals regulation.20 In 2016, the EPA released a lifelong health advisory limit of 70 ng L-1 for individual or combined PFOA and PFOS in drinking water.57-59 This limit was significantly lowered in 2023 to 4 ng L-1 for both PFOA and PFOS.57,60-62 However, while these stringent legislations aim to mitigate the risk of PFAS to the environment and human health, they have inadvertently led to a shift from long-chain PFASs to shorter-chain analogs in chemical industries.63-68 In order to match the technical performance of compounds such as PFOA and PFOS, significantly larger quantities of short-chain PFASs are required, which raises concerns as many short-chain PFASs are not well understood in terms of their environmental occurrence and fate.54,69,70 Consequently, a growing discussion regarding grouping PFASs for risk regulation has been proposed.71-73 In 2020, the European Union (EU) proposed a limit value of 0.1 μg L-1 for the sum concentration of 20 short- and long-chain PFASs in drinking water.74-76 The revised EU Directive, set to take effect in 2023, includes an additional parameter for "total PFAS sum" with a limitation of 0.5 μg L-1.77 Member countries such as Norway, Sweden, Denmark, and Germany have also enacted related bans in support of the development of these regulations.78,79 These actions demonstrate that the global trend for water quality regulations is shifting towards regulating PFAS as a class. 5 1.2 Total PFAS Assessment The quantification of PFASs is mainly conducted through three different approaches: (i) liquid chromatography-mass spectrometry (LC-MS) measurement for the sensitive quantification of individual PFASs; (ii) the total oxidizable precursor (TOP) assay; and (iii) the total organic fluorine (TOF) analysis.80-84 The next paragraphs will summarize the advantages, disadvantages, and limitations of these three approaches. 1.2.1 Liquid Chromatography-Mass Spectrometry (LC-MS) LC-MS is the most conventional and widely-applied technique for quantifying individual PFASs in water matrices (Figure 1-2).85-87 The US EPA has established Method 537 and Method 1633 with this approach to quantify dozens of PFASs under the Unregulated Contaminant Monitoring Rule (UCMR5).88-90 The target analysis involves separating the sample components, ionizing the analyte (i.e., the individual PFAS), and detecting the molecular mass of the precursor ion and the product ions.91 LC-MS can provide accurate and precise results for quantifying and identifying individual PFASs at limits of detection (LOD) approaching 10 ng L-1 and lower.45 However, the analysis requires authentic analytical standards,92-94 which are only available for fewer than 100 of the more than 9,000 total PFASs identified.80 This indicates that a majority of PFAS precursors, intermediates, and terminal products cannot be accurately quantified, largely limiting the coverage of PFAS monitoring in water samples.84 Recent studies have shown that the sum of PFASs quantified by means of LC-MS is often less than 10% of the total organic fluorine in water samples.95 Additionally, sample preparation is critical for complex environmental water analysis 6 because PFASs are typically present at trace levels.45 The diversity of the media may lead to the loss of analyte due to insufficient extraction and retention, or signal suppression caused by matrix interferences, which can have a significant impact on the accuracy of PFAS determination.96 Moreover, LC-MS measurement can be time- consuming, expensive, and subjected to limited laboratories. These inconveniences also restrict its widespread application.97,98 As a result, LC-MS may not be an appropriate choice for accessing the total PFAS concentration for regulatory purposes. Figure 1-2. PFAS concentration measurement with LC-MS. 1.2.2 Total Oxidizable Precursor (TOP) Assay TOP is a complementary approach to the targeted LC-MS quantification.99-101 This assay estimates the total PFAS concentration by oxidizing unknown precursors to stable and measurable PFAAs under an alkaline environment with excessive hydroxyl radicals. The PFAA products of the TOP assay can be quantified by means of LC-MS to provide insight into the abundance and types of precursors present in the original sample.102-104 The TOP assay can provide more information about the unanalyzable PFAS precursors and dead-end products, which may contribute up to 80% of the total PFAS, giving it 7 better inclusivity of unexpected or unknown PFASs in various water matrices.105-109 Research has also shown that oxidative pre-treatment can increase the total PFCA concentration in urban runoff samples by 69%.104 Besides, the use of LC-MS for measurement allows for low LOD approaching 1-2 ng L-1.13 However, recent studies found that the oxidation process for some PFASs can be complex and nonspecific. A few compounds, such as hexafluoropropylene oxide-dimer acid (HFPO-DA) and dodecafluoro-3H-4,8-dioxanoate (ADONA), cannot be oxidized or be transformed into a PFAA for analysis.16 The molecular conversion rates for some substances can also be inadequate, for instance, 4:2 fluorotelomer sulfonic acid (4:2 FTSA) yields only 62% of the transformation efficiency in the TOP assay.107,110 Additionally, the oxidation of short-chain precursors may result in a significant amount of ultrashort-chain PFASs, for which LC-MS quantification methods are only now emerging.16,25,107 Therefore, the effectiveness of the TOP assay is highly dependent on the specific environmental PFAS circumstances and the presence of oxidizable precursors, so it also may not be the optimal choice for regulating total PFAS concentration in certain situations. 1.2.3 Total Organic Fluorine (TOF) TOF is an analytical approach for measuring the total amount of organic fluorine in water samples.51,111-115 It has been developed by the EPA Office of Science and Technology (OW-OST) for the identification of PFASs in environmental samples.116 The analysis is typically conducted using combustion ion chromatography (CIC), a well-established technique for the assessment of total organic halogens, including total organic chlorine (TOCl), total organic bromine (TOBr), and total organic iodine 8 (TOI).117-119 The TOF assay involves PFAS pyrohydrolysis, hydrofluoric acid formation, and fluoride concentration measurement by means of ion chromatography (IC). 120-122 The decomposition of PFASs is conducted in a high-temperature and oxygen-rich furnace, resulting in the breakdown of carbon-fluorine single bonds.123,124 The mineralization product hydrofluoric acid is collected in an absorption solution, dissociated into fluoride, and measured using IC to determine the concentration.116,125 Unlike LC-MS and TOP measurements, the TOF assay does not differentiate between different PFAS structures, providing a comprehensive measure of total PFAS occurrence.111,126,127 Moreover, the assessment is relatively simple to conduct and less expensive, making it an ideal choice for PFAS regulations.16 However, the low concentration of organofluorine-containing chemicals and the presence of background inorganic fluorine from environmental samples can significantly affect the accuracy of the assessment.80 Therefore, several sample preparation techniques have been developed to address this challenge. The two most widely used approaches are adsorbable organic fluorine (AOF) and extractable organic fluorine (EOF).128,129 These techniques are explored more completely in Chapter 2 of this thesis and are summarized briefly in the following section. 1.3 AOF and EOF AOF is a commonly applied technique that integrates sample preparation with CIC for measuring TOF in water samples (Figure 1-3).124,130 In fact, the US EPA has recently established a draft protocol (EPA Method 1621) for the AOF assessment in aqueous matrices.131 Similarly, the German Institute for Standardization (Deutsches Institut für 9 Normung, DIN) has proposed Method 38409-59 for the determination of adsorbable organically bound fluorine in different water samples.16,132 The AOF method consists of the adsorption of organofluorine chemicals, removal of inorganic fluoride with a washing step, and measurement of fluoride using CIC.133,134 The adsorption process typically uses pre-packed activated carbon (AC) as adsorbents.16 The loaded AC is then washed with sodium nitrate solution to displace inorganic fluoride, followed by the TOF determination using CIC.49 Compared to direct combustion assays, the prior enrichment of PFAS with AOF can effectively increase the analyte concentration and minimize the background fluoride interference.135 The individual PFAS recoveries in Milli-Q water have been determined and reported in recent years for method cross-check confirmation purposes.128,133 However, the coverage is limited to long- and short-chain PFAAs, with minimal discussion on ultrashort-chain PFASs, indicating the need for a more comprehensive evaluation. Additionally, studies have shown that the AOF recovery for PFASs varies with chain length and is sensitive to minor experimental step changes.135,136 This highlights the need to detail the mechanisms leading to the recovery loss, which is typically attributed to incomplete adsorption efficiency but has not been verified.49 Furthermore, the AOF method has not been systematically validated for interference from environmental water matrices, such as fluoride and dissolved organic matter (DOM).49,134 Hence, a more comprehensive analysis is crucial to fully understand the performance of the AOF method and provide more information to optimize the method. 10 Figure 1-3. PFAS TOF measurement with AOF-CIC. EOF is another frequently applied approach in combination with CIC for the quantification of TOF content in various environmental matrices (Figure 1-4).137-141 For liquid sample assays, the EOF analysis involves solid-phase extraction (SPE) and CIC measurement.142 The SPE portion of the EOF method is detailed in International Organization for Standardization (ISO) Method 25101 and US EPA Method 533 for assessing PFAS in drinking water.143,144 Although SPE is a labor-intensive and time- consuming process, it enriches the PFAS concentration and helps decrease the interference of impurities from the original matrices, which enables more accurate instrumental detection.145-148 The procedure consists of the adsorption of organofluorine chemicals, removal of inorganic fluoride with a washing step, elution of the bound organofluorine-containing molecules, and measurement of fluoride using CIC.149,150 Commonly selected SPE materials include Oasis weak anion exchange (WAX), Oasis hydrophilic-lipophilic-balanced (HLB), and Phenomenex Strata-X, but only Oasis WAX is widely applied for EOF studies.146,151-153 Studies have shown that the choice of adsorbent can affect the selectivity and sensitivity of the extraction.154 In recent years, 11 there has been a focus on validating individual PFAS recoveries with EOF in Milli-Q water.145,155 The results revealed ideal recoveries for long- and short-chain PFAAs but poor recoveries for ultrashort-chain PFASs such as TFA.156 This loss of TFA has been attributed to the limited extraction efficiency of the specific adsorbent, but the detailed mechanisms were not thoroughly discussed.80 Furthermore, the current EOF method lacks an official protocol from regulatory agencies for total PFAS assessment. Therefore, it is important to have an integral evaluation with determined experimental steps. Further studies are needed to comprehensively evaluate the EOF method, optimize extraction efficiency, and validate its applicability in complex water samples. This will provide more information on the limitations of EOF and enable accurate quantification of PFAS in environmental water matrices. Figure 1-4. PFAS TOF measurement with EOF-CIC. 1.4 Research Objectives The motivation for this Master's thesis was to address knowledge gaps in the understanding of current AOF and EOF methods for the determination of TOF from 12 PFASs. The research involved a systematic evaluation of each step in the AOF and EOF methods, following conventional protocols, to investigate the causes of incomplete recoveries for individual PFAS species in Milli-Q water and simulated environmental water systems. CIC and LC-MS were used in tandem to measure fluoride and individual PFAS concentrations. The specific research questions addressed in this thesis include:  Question 1: How do current AOF and EOF methods work with 10 selected long-, short-, and ultrashort-chain PFASs? What relationships exist between individual PFAS recoveries and PFAS physiochemical properties?  Question 2: How does each methodological step affect the final AOF and EOF recovery? What are the primary reasons leading to incomplete recovery for selected PFASs?  Question 3: To what extent does the presence of inorganic fluoride influence the AOF and EOF PFAS assessment? Does the washing step effectively remove all fluoride? Chapter 2, entitled "A systematic evaluation of factors resulting in incomplete recovery when implementing the adsorbable organic fluorine (AOF) and extractable organic fluorine (EOF) methods," provides comprehensive answers to the research questions above, along with a discussion of the major findings. This study quantifies the limitations of current AOF and EOF methods and offers detailed insights for future research aimed at improving the accuracy and reliability of TOF determinations for PFASs in complex water matrices. 13 CHAPTER 2 A SYSTEMATIC EVALUATION OF FACTORS RESULTING IN INCOMPLETE RECOVERY WHEN IMPLEMENTING THE ADSORBABLE ORGANIC FLUORINE (AOF) AND EXTRACTABLE ORGANIC FLUORINE (EOF) METHODS Abstract Per- and polyfluoroalkyl substances (PFASs) are environmental contaminants of concern. Techniques that quantify total organic fluorine (TOF) such as the adsorbable organic fluorine (AOF) and extractable organic fluorine (EOF) methods are important for PFAS risk assessments. The objective of this study was to systematically evaluate each step of the AOF (loading, washing, combustion) and EOF (loading, washing, elution, combustion) methods for the recovery of ten ultrashort-, short-, and long-chain perfluoroalkyl acids. We measured the overall recovery of fluoride for each method for each PFAS, and the recovery of fluoride for each PFAS around the loading, washing, and elution steps. We also measured the combustion efficiency of each PFAS by direct combustion. The overall AOF and EOF recovery ranged from 9.3%-103.3% and 21.0%- 108.1%, respectively, with higher recoveries measured for PFASs with increasing chain length in both methods. The three ultrashort-chain PFASs exhibited the lowest overall recoveries from 9.3-25.2% for AOF and 21.0-51.5% for EOF. We found that decreases in the overall recovery are the result of losses of ultrashort- and short-chain PFASs during the washing step and the incomplete mineralization of perfluoroalkyl sulfonic 14 acids during combustion for AOF and incomplete elution of short- and long-chain PFASs and the loss of ultrashort-chain PFASs during the washing step for EOF. Together, our data suggest that the EOF method is more appropriate than the AOF method for measuring TOF in samples containing ultrashort- and short-chain PFASs and that methodological improvements are possible with a focus on the washing, elution, and combustion steps. 15 2.1 Introduction Per- and polyfluoroalkyl substances (PFASs) are a group of anthropogenic organofluorine-containing chemicals.48,50,157 The strong carbon-fluorine bonds provide PFASs with extraordinary thermal and chemical stability leading to their extensive use in industrial processes and consumer products.10,25 As a result of their widespread use, PFASs have an impact on water systems all over the world.158-160 This pervasive exposure can lead to human health issues including multi-organ failure, alteration of immune system functioning, and cancer.42,161,162 As a result, many countries recommend legislation to limit PFAS levels in drinking water.48-51 However, the number of individual PFAS species makes the regulation of individual species insufficient to reflect the hazard of the entire group.49,163 For example, the U.S. EPA proposed a national drinking water standard in March 2023 to establish maximum contaminant levels (MCLs) for perfluorooctanoic acid (PFOA) (4 ng/L) and perfluorooctanesulfonic acid (PFOS) (4 ng/L).62 However, a limit based on a hazard index approach was also proposed for mixtures of four PFASs including perfluorononanoic acid (PFNA), perfluorohexanesulfonic acid (PFHxS), perfluorobutanesulfonic acid (PFBS), and GenX.62 Likewise, the European Union has proposed quantifying the sum concentration of all quantifiable PFASs as a metric for risk assessment.35,75,164 Based on these actions, it appears that the trend for global water quality regulations is moving towards regulation as a class.54,73 Currently, the most conventional approach to measuring individual PFAS species in aqueous samples is by means of liquid chromatography-mass spectrometry (LC- 16 MS).75,145,165-167 LC-MS techniques can accurately quantify individual PFAS species with authentic analytical standards at limits of detection (LOD) approaching 10 ng L-1 and lower.45,168,169 However, because there are thousands of individual PFAS species and authentic analytical standards for only a few dozen, the scope of LC-MS measurements for assessing the occurrence of PFASs in any water sample is relatively limited.11,16 In fact, recent studies have demonstrated that the sum of PFAS species in water and wastewater quantified by LC-MS represents only 1 – 10% of the total organic fluorine (TOF) in the sample.45,95 TOF is a metric of the total concentration of organofluorine-containing chemicals in a water sample.49,128,134 TOF is most frequently measured by means of combustion ion chromatography (CIC) where a sample is combusted in a furnace at high temperatures and the evolved hydrogen fluoride is captured and measured by means of ion chromatography (IC).51,170 CIC is a relatively simple and inexpensive technique that allows analysts to quantify TOF in a variety of sample matrices.171 However, low concentrations of organofluorine-containing chemicals and the presence of inorganic fluoride in most environmental water samples present challenges for CIC methods and necessitate some sample preparation prior to combustion.37 Adsorbable organic fluorine (AOF) and extractable organic fluorine (EOF) are two techniques that integrate sample preparation with CIC for measuring TOF in water samples.37,49,51,128,134,172 Both techniques rely on the loading of PFASs onto either activated carbon (AOF)49,131 or a WAX adsorbent (EOF)134,145,170,173 followed by a 17 washing step to remove residual inorganic fluoride from the adsorbent. The EOF method also has an elution step that results in the transfer of captured PFASs to a small volume of an organic solvent. Both methods are completed with a combustion step to mineralize the PFASs and capture the evolved hydrogen fluoride for quantification; the AOF method performs the combustion on the washed activated carbon and the EOF method performs the combustion on the organic solvent eluate. Numerous studies have demonstrated that both the AOF and EOF methods result in incomplete recoveries of fluoride for individual PFAS species, but no evidence is clearly presented to explain the mechanisms resulting in incomplete recoveries. Because the reported recoveries of fluoride for individual PFASs are often positively associated with chain length,75,174,175 incomplete adsorptions during the loading step are frequently cited as the primary factor limiting the recovery of fluoride for the AOF and EOF methods.4,49 However, a deeper understanding of the mechanisms resulting in incomplete recoveries is essential to improve the performance of both methods. This study presents a systematic evaluation of each step in the AOF and EOF methods to improve our understanding of the causes of incomplete recoveries of fluoride for individual PFAS species. The specific objectives of this research were to: (i) evaluate the performance of the AOF and EOF methods for a set of ten ultrashort-, short-, and long-chain perfluorocarboxylic acids (PFCAs) and perfluorosulfonic acids (PFSAs); (ii) identify the methodological steps for AOF and EOF that lead to incomplete recovery of the selected PFASs; and (iii) evaluate the extent to which environmentally relevant concentrations of inorganic fluoride influence recovery in the AOF and EOF methods. 18 The experiments were conducted using either spiked Milli-Q water or spiked simulated groundwater, following the conventional protocols for implementing the AOF and EOF methods with CIC. Uniquely, samples were collected around each step of the AOF and EOF methods (loading, washing, elution), and LC-MS measurements were coupled with CIC measurements to evaluate the recovery and loss of individual PFAS species around each methodological step. To the best of our knowledge, this is the first systematic and comprehensive evaluation of AOF and EOF methods to investigate the recovery of ultrashort-, short-, and long-chain PFCAs and PFSAs and identify causes of incomplete recovery. The findings from this study will motivate future optimization of the AOF and EOF methods and enable a more complete recovery of all PFASs. 2.2 Materials and Methods 2.2.1 Standards and Reagents We selected ten authentic PFAS standards including five PFCAs and five PFSAs with variable chain lengths. These included trifluoroacetic acid (TFA, two carbon atoms in chain denoted as C2), perfluoropropionic acid (PFPrA, C3), perfluorobutanoic acid (PFBA, C4), perfluorohexanoic acid (PFHxA, C6), PFOA (C8), trifluoromethanesulfonic acid (TFMS, C1), perfluoropropanesulfonic acid (PFPrS, C3), PFBS (C4), PFHxS (C6), and PFOS (C8). Stock solutions of each individual PFAS were prepared in methanol or water (depending on the solubility of each PFAS) at a concentration of 1 g L-1. Chemical structures, suppliers, concentrations, and solvents used for each stock solution are provided in Table A-1 of Appendix A. Spike solutions and spike mixtures were generated from the stock solutions by means of serial dilution 19 in water. Spike solutions and spike mixtures were prepared once at the beginning of the project and used for all subsequent experiments. We also acquired a 100 ± 1 mg L-1 sodium fluoride solution (Hach Company, USA) and a 50 μg mL-1 PFAS mixture (PFAC-MXA containing PFBA, PFHxA, PFOA, PFBS, PFHxS, and PFOS, Wellington Laboratories, USA) that we used for independent calibration of the ion chromatograph and the mass spectrometer, respectively. Finally, we acquired individual authentic analytical standards of PFBA, PFHxA, PFOA, PFPrS, PFBS, PFHxS, and PFOS dissolved in methanol at 50 μg mL-1 (Wellington Laboratories, USA) and prepared individual PFPrA (TCI, USA), TFMS (Spectrum Chemical, USA), and TFA (VWR, USA) standards at 50 μg mL-1 in methanol for the assessment of combustion efficiency. All PFAS stock solutions were stored at -20 ℃ in the dark. The spike solutions and spike mixtures were stored at 4 ℃ in the dark. 2.2.2 Adsorbable Organic Fluorine (AOF) Method We adapted an AOF method from the Draft EPA Method 1621131 and similar methods described in the literature that employ activated carbon adsorption and CIC.49,128 Briefly, the individual spike solutions or the spike PFAS mixture were used to generate triplicate 300 mL samples at a concentration of 10 µg L-1 of each PFAS in Milli-Q water or in a simulated groundwater spiked with inorganic fluoride to a concentration of 300 µg L-1. We used an AOX Adsorption Unit (Mitsubishi Chemical Analytech, Japan) to load two pre-packed activated carbon columns (Nittoseiko Analytech, Japan) connected in series. The activated carbon columns were first rinsed with Milli-Q water, then loaded with each 300 mL water sample at a flowrate of 200 mL h-1, and finally washed with 40 mL 20 of a 0.08 M potassium nitrate solution. This latter step is included in most AOF methods to remove residual inorganic fluoride that may be present in the water sample. The loaded and washed activated carbon columns were then cleaned with 2 mL of Milli-Q water and dried in the air. We used an Automatic Sample Combustion Unit AQF-2100H (Nittoseiko Analytech, Japan) to combust the loaded activated carbon according to the Draft EPA Method 1621.131 Ceramic boats were first washed with methanol and prebaked (Temperature: 900°C inlet and 1000°C outlet; Flow rate: 200 mL min-1 Argon and 400 mL min-1 Oxygen) for 10 minutes to remove impurities. After the prebake, the loaded activated carbon was transferred to the ceramic boats from the columns with a T-bar. We loaded the ceramic boats onto an ASC-240S Automatic Solid Sampler (Nittoseiko Analytech, Japan) and combusted each sample (Temperature: 900°C inlet and 1000°C outlet; Flow rate: 200 mL min-1 Argon and 400 mL min-1 Oxygen) for approximately 25 minutes. Detailed combustion parameters are provided in Table A-2. We added a ceramic boat blank after every two sample boats because one water sample was collected on two activated carbon columns. The HF gas that evolved during the combustion process was collected in an adsorption tube and transferred to a 50 mL polypropylene centrifuge tube (Corning, USA) containing 20 mL of MilliQ water in an AU-259 Gas Adsorption Unit (Nittoseiko Analytech, Japan). The volume in each centrifuge tube was adjusted to 25 mL prior to sample analysis by means of IC. 21 2.2.3 Extractable Organic Fluorine (EOF) Method We adapted an EOF method from those described in the literature that employ extraction by means of solid-phase extraction (SPE) cartridges and CIC.145,172 Briefly, the individual spike solutions or the spike PFAS mixture were used to generate triplicate 300 mL samples at a concentration of 10 µg L-1 of each PFAS in Milli-Q water or in a simulated groundwater spiked with inorganic fluoride to a concentration of 300 µg L-1. We loaded the prepared water samples (in triplicate) onto single 150 mg Oasis WAX 6 cc cartridges (Waters, USA) using a 12-port vacuum manifold as previously described.145 The cartridges were first washed with 4 mL 0.1% ammonia hydroxide in methanol, 4 mL methanol, and 4 mL Milli-Q water sequentially for conditioning, and then loaded with the 300 mL samples. The loaded cartridges were then washed with 20 mL of 0.01% ammonia hydroxide in water (to remove residual fluoride), dried in a vacuum, and eluted with 5 mL 0.1% ammonium hydroxide. The eluate was collected in 15 mL polypropylene centrifuge tubes (Corning, USA) and was evaporated under a gentle stream of nitrogen at room temperature. The dried sample was reconstituted with 600 μL of methanol and transferred to a ceramic boat for combustion. We used an Automatic Sample Combustion Unit AQF-2100H (Nittoseiko Analytech, Japan) to combust the reconstituted buffer according to a commonly used liquid combustion method.176 Ceramic boats were first washed with methanol and prebaked (Temperature: 900°C inlet and 1000°C outlet; Flow rate: 200 mL min-1 Argon and 400 mL min-1 Oxygen) for 10 minutes to remove background contamination. After the prebake, the elution buffer was transferred to the ceramic boats with a 200 mL pipette 22 in three consecutive transfers. We loaded the ceramic boats onto an ASC-240S Automatic Solid Sampler (Nittoseiko Analytech, Japan) and combusted each sample (Temperature: 900°C inlet and 1000°C outlet; Flow rate: 200 mL min-1 Argon and 400 mL min-1 Oxygen) for approximately 10 minutes. Detailed combustion processes are provided in Table A-2. We added a ceramic boat blank after every other sample boat to minimize interactions between sample burnings. The HF gas that evolved during the combustion process was collected in an adsorption tube and transferred to a 50 mL polypropylene centrifuge tube (Corning, USA) containing 20 mL of MilliQ water in an AU-259 Gas Adsorption Unit (Nittoseiko Analytech, Japan). The volume in each centrifuge tube was adjusted to 25 mL prior to sample analysis. 2.2.4 Evaluation of Each Step in the AOF and EOF Methods In addition to performing the complete AOF and EOF methods integrated with CIC, we also collected samples across each of the methodological steps to evaluate the concentrations of individual PFASs. For the AOF method, we collected 0.5 mL samples (AOF_start) from the original 300 mL prepared samples to confirm the starting concentration of each PFAS in the experiment. We also collected 0.5 mL samples from the 300 mL eluate from the loading step (AOF_load) and from the 40 mL eluate from the potassium nitrate washing step (AOF_wash). For the EOF method, we collected 0.5 mL samples (EOF_start) from the original 300 mL prepared samples to confirm the starting concentration of each PFAS in the experiment. We also collected 0.5 mL samples from the 300 mL eluate from the loading step (EOF_load), from the 20 mL eluate from the 0.01% ammonium hydroxide washing step (EOF_wash), and from the 23 5 mL elution step (EOF_elute). All samples were collected in polypropylene LC-MS vials (Thermo Scientific, SUN-SRi) and stored at 4°C until analysis. 2.2.5 Evaluation of Combustion Efficiency To evaluate the combustion efficiency for each of the selected PFASs, we performed direct combustion experiments with both solid and liquid samples. To approximate the conditions during combustion of samples prepared with the AOF method, we added unused activated carbon to a prebaked ceramic sample boat and spike 60 µL of each of the individual PFAS standards prepared in methanol at a concentration of 50 mg L-1. Samples were then combusted using the same conditions as described for the AOF method in the preceding. To approximate the conditions during combustion of samples prepared with the EOF method, we spiked 60 µL of each of the individual PFAS standards prepared in methanol at a concentration of 50 mg L-1 into a prebaked ceramic sample boat. Samples were then combusted using the same conditions as described for the EOF method in the preceding. For both analyses, boat blanks were added between every sample to avoid potential interferences. The HF gas that evolved during the combustion process was collected in an adsorption tube and transferred to a 50 mL polypropylene centrifuge tube (Corning, USA) containing 20 mL of MilliQ water in an AU-259 Gas Adsorption Unit (Nittoseiko Analytech, Japan). The volume in each centrifuge tube was adjusted to 25 mL prior to sample analysis. 24 2.2.6 Fluoride Analysis by means of Ion Chromatography (IC) Fluoride concentrations were measured on a Dionex ICS-2100 ion chromatograph (Thermo, USA) with a 4×250 mm Dionex Ion Pac AS18 analytical column. Potassium hydroxide was selected as an eluent during the measurement process with a flow rate of 1.0 mL min-1. We modified a previous multistep gradient program to allow a faster and more accurate measurement of fluoride as shown in Table A-2. 134 Under the baseline conditions, the target fluoride can be separated from matrix ions with a retention time of 3.7 minutes. The total duration for the chromatography run was 14 minutes. To determine the fluoride concentration in samples from the combustion experiments, an external fifteen-point calibration curve (concentration ranging from 0-1500 μg L-1) was established from the sodium fluoride stock solution. The calibration curves were run at the beginning of each analytical run. Instrumental blanks were added before and after the calibration, and after every four samples. The analytes were quantified through the calibration curve and the peak response from the IC based on ordinary least square regression. Because the AOF analysis involved the combustion of two activated carbon samples for each water sample, the peak responses from both samples were added together to estimate the fluoride concentration (The results are of no significant difference with the sum of individual sample fluoride concentrations, t-test, p = 0.73 > 0.05). The limit of quantification (LOQ) for each experiment was calculated with the calibration standard deviation approach177 and the results are provided in Table A-3. All data acquired for the AOF and EOF samples were at least 5 times higher than the method blank and above the LOQ. 25 2.2.7 PFAS Analysis by means of Liquid Chromatography Mass Spectrometry (LC-MS) We performed complementary HPLC-MS analyses to confirm the concentrations of each PFAS in the PFAS mixture experiments to assess the recovery of individual PFASs through the AOF and EOF experiments. Samples were measured by means of high- performance liquid chromatography (HPLC, Thermo, USA) coupled with a quadrupole- orbitrap mass spectrometer (MS, QExactive, Thermo, USA) in negative polarity mode with small volume injection. We adapted a recently described full-scan method for short- and ultrashort-chain PFASs to measure TFA, TFMS, PFPrA, and PFPrS25 and a previously described parallel reaction monitoring (PRM) method to measure PFBA, PFBS, PFHxA, PFHxS, PFOA, and PFOS.160 A detailed list of analytical information including the mass (m/z) of the parent and product ions along with the retention time for each PFAS is provided in Table A-4. Briefly, the mobile phase for the full-scan method for TFA, TFMS, PFPrA, and PFPrS consisted of (A) LC-MS grade water amended with 40 mM ammonium acetate and (B) LC-MS grade acetonitrile. Samples were injected at 20 µL onto an Acclaim mixed- mode WAX-1 3 μm 3.0×50 mm analytical column (Thermo, USA) for separation. The mobile phase gradient for the PRM method for PFBA, PFBS, PFHxA, PFHxS, PFOA, and PFOS consisted of (A) LC-MS grade water amended with 20 mM ammonium acetate and (B) LC-MS grade methanol. Samples were injected at 20 µL onto an Atlantis® dC18 5 µm 2.1×150 mm analytical column (Waters, USA) for separation. Detailed elution gradient information for both methods is provided in Table A-5. For 26 both methods, a ten-point external calibration curve with concentrations ranging from 0 to 200 μg L-1 was established to determine the PFAS concentrations in the samples. Calibration curves were measured at the beginning of every analytical run. Instrumental blanks were added before and after the calibration curves, and between triplicate samples. Quality control samples were run at the beginning and in the middle of measurement sequences to ensure the stability of the method. Analytes were quantified through the calibration curve and analyte peak area responses based on the best 1/x weighted linear least-squares regression. The LOQs for each PFAS were determined as the lowest point of the calibration curve that is significantly distinguishable from the sample blanks. 2.2.8 Data Analysis The overall recoveries for the AOF and EOF methods as well as the solid and liquid combustion efficiencies were calculated using Equation 2-1: Recovery = [F-] measured ×Vmeasured [F] PFAS ×msample ×100% (2-1) Where [F-]measured is the fluoride concentration (μg L-1) measured in each combusted sample; Vmeasured is the final volume of the centrifuge tubes (mL) that are used to collect the evolved HF gas during combustion; [F]PFAS is the mass fraction of fluorine in a certain individual PFAS (-); and msample refers to the mass of PFAS spiked to the water sample. The capture efficiency of the activated carbon was calculated using Equation 2-2: 27 AOF without washing = [AOF_start]-[AOF_load] [AOF_start] ×100% (2-2) where [AOF_start] is the initial amount of each PFAS in the prepared 300 mL solution [ng] and [AOF_load] is the amount of each PFAS in the effluent of the activated carbon column collected during loading [ng]. The AOF without washing is a metric of the percent of each PFAS that was captured by the active carbon during sample loading. The percentage of each PFAS remaining on the activated carbon column after the washing step was calculated using Equation 2-3: AOF with washing = [AOF_start]-[AOF_load]-[AOF_wash] [AOF_start] ×100% (2-3) where [AOF_wash] is the amount of each PFAS in the washing buffer effluent [ng]. The capture efficiency of the WAX adsorbent with and without washing was calculated using the EOF_start and EOF_elute samples collected in experiments conducted with and without the washing step. The general approach is described in Equation 2-4 and Equation 2-5. EOF without washing = [EOF_elute_without] [EOF_start] ×100% (2-4) where [EOF_start] is the initial amount of each PFAS in the prepared 300 mL solution [ng] and [EOF_elute_without] is the amount of each PFAS in the eluate of an EOF experiment that omitted the washing step [ng]. 28 EOF with washing = [EOF_elute_with] [EOF_start] ×100% (2-5) where [EOF_elute_with] is the amount of each PFAS in the eluate of a complete EOF experiment that includes the washing step [ng]. The percentage of fluoride that remains on the adsorbent after the washing step for both AOF and EOF methods were calculated using Equation 2-6: Fluoride remaining = [Fluoride_spike]-[Fluoride_rinse] [Fluoride_spike] ×100% (2-6) where [Fluoride_spike] is the initial amount of fluoride in the prepared 300 mL solution [mg] and [Fluoride_rinse] is the amount of fluoride lost during the washing step. 2.3 Results and Discussion 2.3.1 AOF and EOF Validation Our first objective was to evaluate the performance of the AOF and EOF methods for a set of ultrashort-, short-, and long-chain PFCAs and PFSAs. We define ultrashort-chain PFASs as PFAS having fewer than three fluorinated carbon atoms (TFA, PFPrA, TFMS), short-chain PFASs as those with three to six fluorinated carbon atoms (PFBA, PFHxA, PFPrS, PFBS, PFHxS), and long-chain PFASs as those with seven or more fluorinated carbon atoms (PFOA, PFOS). We performed triplicate experiments with 300 mL solutions of Milli-Q water spiked with each of the ten PFASs individually to a starting concentration of 10 μg L-1. The performance of each method was evaluated by calculating the recovery of fluoride according to Equation 2-1. We used LC-MS 29 measurements of diluted PFAS stocks to obtain the actual mass of each PFAS spiked into each solution (msample). The average and standard deviation of the recovery calculated for each PFAS with the AOF and EOF methods are provided in Figure 2-1A and Figure 2-1C, respectively. Figure 2-1. A and C: Recoveries of inorganic fluoride in experiments with individual PFASs spiked in Milli-Q water with the AOF (A) and EOF (C) methods, respectively (number = 3). The horizontal bar lengths represent the average recovery values; the error bars represent one standard deviation for triplicate measurements. The acronyms for PFASs are listed on the left side with the carbon number (n) in parentheses. Darker: perfluoroalkyl sulfonic acids (PFSAs); Lighter: perfluoroalkyl carboxylic acids (PFCAs). The box plot above shows the overall recovery patterns for selected PFASs with each method. B: AOF results comparison with published literature values. PFAS 30 acronyms are shown on the x-axis. The vertical bars represent the average recoveries with error bars. Black: this study (same as Figure 1A); grey: research from Forster et al., (2023) under lower adsorption pH and application of different washing buffers for fluoride removal16; light blue: research from Han et al. (2021) under similar conditions as this study49; dark blue: research from Abercron et al. at lower adsorption pH (2019)128. D: EOF results comparison with published literature values. PFAS acronyms are shown on the x-axis. The vertical bars represent the average recoveries with error bars. Black: this study (same as Figure 1B); grey: research from Forster et al., (2023)16 with Strata WAX as adsorbents; orange: research from Miyake et al. under similar conditions as this study (2007)145; red: research from Yeung et al. in Milli-Q water (2013)155. The average recoveries for the ten PFASs with the AOF method ranged between 9.3 – 103.3% (Figure 2-1A). It is interesting to note that seven of the PFASs had average recoveries between 69.5 – 103.3% whereas the three ultrashort-chain PFASs (TFMS, PFPrA, and TFA) had average recoveries between 9.3 – 25.2%. The large gap between the average recovery of PFBA and PFPrA is notable and demonstrates that there is an abrupt change in the performance of the AOF method between the C4 and C3 PFCA. This may be due to the changes in the adsorption mechanism for short-chain and ultrashort-chain PFCAs to the activated carbon.175,178,179 Previous studies demonstrate that hydrophobic interactions are the primary adsorption mechanism for PFASs with longer chains, whereas electrostatic interactions are dominant for ultrashort-chain PFASs.29,159,180-182 Our data suggest that hydrophobic interactions drive the recoveries of PFCAs and PFSAs with at least four carbon atoms, but weak electrostatic interactions between the anionic PFASs and the negatively charged activated carbon material limit the recovery of ultrashort-chain PFASs.183,184 This novel finding for ultrashort-chain 31 PFASs is important considering the emerging evidence that ultrashort-chain PFASs are nearly ubiquitous in industrial wastewater and environmental systems.13,25 With the exception of PFPrS for which we measured a higher average recovery than PFBS, there is also a trend of greater average recoveries for PFASs with longer perfluorinated chains, which aligns with the expected adsorption affinity of activated carbon for PFASs with varying chain length.16 PFOA and PFOS exhibit quantitatively complete recovery whereas PFHxA, PFHxS, and PFPrS exhibit average recoveries greater than 80%. Finally, our findings suggest that the recoveries for PFSAs are not significantly different than PFCAs with the same carbon number (t-test, p = 0.11 > 0.05).49 This is contrary to observations reported in the adsorption literature that demonstrate enhanced adsorption of PFSAs compared to PFCAs with the same carbon number due to the presence of an additional CF2 group in the PFSA and a greater electrostatic potential of the sulfonic acid functional group.15,182,185,186 Together, these findings suggest that adsorption during sample loading might not be the factor limiting the recovery of fluoride in these AOF experiments. We compare our AOF experimental data with data from other studies reported in the literature in Figure 2-1B. Because no previous study has reported recoveries for ultrashort-chain PFASs with the AOF method, we only compare six of the short- and long-chain PFASs here. As a means to compare the results across the four studies, we performed a Pearson correlation analysis on the average fluoride recoveries reported for each of the six PFASs. The Pearson correlation matrix is provided as Figure A-1A in Appendix A. Analysis of the Pearson correlation coefficients presented in Figure A- 32 1A suggests that the results of the present study are relatively similar to those reported in Forster et al. (2023) and Han et al. (2021). The minor differences in recoveries are attributed to small changes in experimental conditions such as the selection of activated carbon materials and washing solutions along with combustion parameters and analytical uncertainties. Notably, the results of the present study are quite different from the data reported by Abercron et al. (2019). In the Abercron et al. (2019) study, the authors adjusted the pH of their samples to approximately 2 which can have an effect on the protonation state of the surface of the activated carbon and of the short-chain PFASs.159,187 The consequence of this adjustment is a lower recovery of fluoride for long-chain PFASs and an improved recovery of fluoride for short-chain PFASs.128 Overall, this analysis demonstrates how relatively small changes in experimental conditions can lead to variable AOF recoveries, and it is important to standardize the method for the purposes of future regulations. The average recoveries of the ten PFASs using the EOF method ranged from 21.0% to 108.1% (Figure 2-1C). Compared to AOF, the average EOF recoveries are significantly higher (paired two-sample t-test, p = 0.001 < 0.05). As noted for the AOF method, seven of the short- and long-chain PFASs had average recoveries ranging between 79.8 – 108.1%, whereas the three ultrashort-chain PFASs had average recoveries ranging between 21.0 – 51.5%. Although the recovery of TFMS was relatively high (51.5%), there was a large gap in the recovery of PFBA (89.8%) and PFPrA (32.3%) that suggests a shift in the mechanism that determines fluoride recovery for short-chain PFCAs and ultrashort-chain PFCAs in the EOF method. One explanation for this observation is the 33 presence of strong hydrophobic interactions between short- and long-chain PFASs that facilitate aggregate formation when binding to anion exchange resins.186,188 This effect might be limited for the C2 and C3 PFCAs thereby limiting their recovery in the EOF method. Nevertheless, the measured recoveries of greater than 80% for all of the short- and long-chain PFASs demonstrate the effectiveness of the EOF method for these PFASs.16 The short- and long-chain PFSAs do not exhibit significantly different average recoveries compared to the PFCAs with the same carbon number (t-test, p = 0.35 > 0.05), whereas the ultrashort-chain PFSAs exhibit greater recoveries when compared to the PFCA with the same carbon number (t-test, p = 0.0004 < 0.05). This finding differs from the study by Zaggia et al. (2016) that reported PFSAs yield a better sorption capacity than PFCAs for anion exchange resins. However, to our knowledge, no previous studies have focused on accounting for this difference. We compare our EOF experimental data with data from other studies reported in the literature in Figure 2-1D. Because no previous study has reported recoveries for ultrashort-chain PFASs with the AOF method, we only compare available data for six of the short- and long-chain PFASs here. As a means to compare the results across the four studies, we performed a Pearson correlation analysis on the average fluoride recoveries reported for each of the six PFASs. The Pearson correlation matrix is provided as Figure A-1B in Appendix A. Analysis of the Pearson correlation coefficients presented in Figure A-1B suggest that the results of the present study are relatively similar to those reported elsewhere, but it must be noted that there is less variability in average recoveries among the six PFASs studied and therefore some 34 comparisons result in Pearson correlation coefficients near zero despite good agreement overall. Nevertheless, we attribute the minor deviations among the studies to small changes in experimental conditions. For example, Forster et al. (2023) selected a Strata WAX (Phenomenex) adsorbent and adjusted the pH to approximately 5 during sample loading. These changes can result in different specific surface areas, porosities, and surface activities on the adsorbent, which can affect the capture efficiency and lead to different recoveries.188-191 Other methodological differences of note include varied drying times of the WAX adsorbents prior to elution and varied combustion parameters. However, because there has been no previous systematic evaluation of experimental procedures, it is difficult to identify the specific reasons for these differences. 2.3.2 Evaluation of AOF and EOF Methodological Steps Our second objective was to identify the methodological steps for AOF and EOF that lead to incomplete recovery of the selected PFASs. To address this objective, we first developed an experiment that integrated LC-MS measurements with CIC analyses to perform mass balances around each step of the AOF and EOF methods. We employed triplicate experiments with 300 mL solutions of Milli-Q water spiked with a mixture of all ten PFASs to a starting concentration of approximately 10 μg L-1 each. Measurements for the AOF evaluation involve LC-MS analysis for PFAS concentrations from the influent to the activated carbon columns (AOF_start), the effluent from the activated carbon columns (AOF_load), and the effluent of the activated carbon columns during the washing step (AOF_wash). The AOF capture efficiency (without washing) and the AOF capture efficiency (with washing) were 35 calculated with Equation 2-2 and Equation 2-3, respectively. Measurements for the EOF evaluation involve LC-MS analysis for PFAS concentrations from the influent to the WAX adsorbents (EOF_start) and the effluent of the WAX adsorbents during the elution step (EOF_elute) in experiments conducted without and with the washing step. The EOF capture efficiency (without washing) and the EOF capture efficiency (with washing) were calculated with Equation 2-4 and Equation 2-5, respectively. We also evaluated the combustion efficiency of each PFAS by spiking authentic standards directly into the combustion boat. The results were calculated with Equation 2-1. Figure 2-2. The bar plots represent individual PFAS recoveries with or without the fluoride washing when the mixture of ten PFASs are spiked in Milli-Q water for AOF (A, left) and EOF (B, right) methods, respectively. PFAS acronyms are shown on the x- axis with separated classes. The vertical bars represent the average recovery of 3 replicates. The lines represent the recoveries of individual PFASs with different combustion procedures for solid (A, left) and liquid (B, right). The height of each point represents the combustion efficiency for that PFAS. The line connecting the points is provided to assist in visual interpretation. The results represent a single batch experiment corrected with positive control burning. 36 The AOF method involves three separate steps: loading of the activated carbon, washing of the activated carbon, and combustion of the activated carbon. In previous literature, adsorption during loading of the activated carbon has been discussed as the factor limiting the recovery of fluoride in AOF experiments.49,134 However, our findings reported in Figure 2-2A suggest that all of the selected PFASs were completely captured onto the activated carbon adsorbent during the loading step, as the PFAS recoveries during AOF experiments without washing were 100% for all ten PFASs. This result may seem unexpected in the context of previous adsorption literature that reports incomplete capture or ultrashort- and short-chain PFASs on activated carbon,192,193 but can be explained by the excess amount of activated carbon used in the AOF method (80 mg for 300 mL = 267 mg L-1) relative to the amount used in most adsorption experiments (10 mg L-1).192 In contrast, we see incomplete recoveries of PFBA and the three ultrashort-chain PFASs when we add the washing step to the AOF method (Figure 2-2A). The washing step is intended to remove inorganic fluoride which can bind to the activated carbon adsorbent through electrostatic interactions.16,194 Our data demonstrate that this washing step can also remove PFBA and ultrashort-chain PFASs to a significant extent as a function of chain length, which will have an important and negative effect on fluoride recoveries. We attribute this phenomenon to the greater relative role of electrostatic interactions versus hydrophobic interactions in the binding of these PFASs to activated carbon. Finally, it is interesting to note that we observed incomplete recovery of PFASs during the combustion step. The average recoveries of fluoride for the ten PFASs ranged between 78.9 – 102.0% with no clear trend observed with carbon number (line in Figure 2-2A). However, a significant difference was found 37 when comparing the two PFAS groups with the same carbon number: PFSAs exhibited an average recovery of 88.0% whereas PFCAs had an average recovery of 98.5% (t-test, p = 0.02 < 0.05). We attribute this difference to the incomplete mineralization of organofluorine for PFSAs in the presence of activated carbon due to the formation of other volatile intermediates. Overall and in contrast to previous findings that attribute limited recovery of fluoride in AOF experiments to incomplete adsorption during the loading step,49 our systematic evaluation of the methodological steps suggests that limited recovery of fluoride in AOF experiments is primarily the result of the loss of ultrashort- and short-chain PFASs during the washing step and the incomplete mineralization of perfluoroalkyl sulfonic acids (PFSAs) during combustion. The EOF method involves four separate steps: loading of the WAX adsorbent, washing of the WAX adsorbent, solvent-based elution of the WAX adsorbent, and combustion of the elution solvent. Although incomplete recoveries have also been reported for the EOF method in the literature,16 the impact of each step on the recovery has not been thoroughly explored in previous literature. We first note that the capture efficiency of the WAX adsorbent is 100% for all ten PFASs as measured using the EOF_start and EOF_load samples and Equation 2-2. Therefore, losses of PFASs during the EOF method must occur during the washing step or the elution step. Previous studies utilizing WAX adsorbents for SPE have reported extraction and elution efficiencies for six short- and long-chain PFASs (PFOS, PFOA, PFHxS, PFHxA, PFBS, PFBA) ranging between 74.1 – 100.9%,151,195 which is comparable to our results reporting EOF without washing recoveries ranging between 74.8 – 97.5% (Figure 2-2B). The losses here must be 38 attributed to the elution step during which some of the PFASs may not completely elute from the WAX adsorbent. It is also worth noting that the ultrashort-chain PFASs exhibited perfect EOF recoveries in the absence of a washing step (101.4 – 117.2%), which can be explained by their relatively weak binding and relatively high elution potential. The washing step employed in this experiment relies on 20 mL 0.01% NH4OH/H2O washing buffer, which has been previously shown to effectively remove fluoride from WAX adsorbents while maintaining high PFAS recoveries.145 However, as shown in Figure 2-2B, the recoveries for ultrashort-chain PFASs were significantly reduced when compared to the results obtained without the washing step (t-test, p = 0.009 < 0.05). This significant difference was not observed for other PFASs (t-test, p = 0.77 > 0.05), again suggesting that the relatively weak binding of the ultrashort-chain PFASs might limit their recovery. Finally, it is interesting to note that we again observed incomplete recovery of PFASs during the combustion step. Previous studies have reported combustion efficiencies of PFASs in a neat solution. It was found that the average efficiencies were around 90%, with a negative correlation observed between the increasing chain length of PFCAs and combustion efficiency, but no clear trend for PFSAs.172,196 In our experiment, we used a stronger air flow and a longer combustion time to achieve a more complete PFAS decomposition. The combustion efficiencies obtained for all selected PFASs ranged from 90.5% to 106.9%, with an average of 100.5% (line in Figure 2-2B). In contrast to solid combustion efficiency, there was no significant difference between PFSAs and PFCAs with the same carbon number (t-test, p = 0.051 > 0.05). This could be attributed to more direct mineralization and transformation for PFASs without the presence of activated carbon. Overall, our 39 findings demonstrate complete capture of all ten PFASs by the WAX adsorbent used in the EOF method, but losses of ultrashort-chain PFASs during the washing step and incomplete elution of short- and long-chain PFASs during the elution step. This contradicts previous studies that attributed low recoveries of ultrashort-chain PFASs to limited extraction efficiency.156 However, it is important to note that these factors alone may not fully explain the low recoveries of ultrashort-chain PFASs observed in Figure 2-1B. We surmise that additional losses of ultrashort-chain PFASs may occur by means of volatilization during the evaporation step after elution which may further lower the recoveries of ultrashort-chain PFASs in EOF experiments; evaluation of volatilization losses was outside the scope of this thesis. 2.3.3 AOF and EOF Environmental Water Simulation Our third objective was to evaluate the extent to which environmentally relevant concentrations of inorganic fluoride influence the fluoride and PFAS recoveries when implementing the AOF and EOF methods. This is an essential objective considering the impact of the washing step on the overall recovery of fluoride in the AOF and EOF methods. To address this objective, we performed experiments with the mixture of ten PFASs spiked to a concentration of 10 µg L-1 each into Milli-Q water and into simulated groundwater containing 300 µg L-1 of fluoride. We evaluated the fluoride and PFAS recoveries for the AOF and EOF methods by conducting triplicate experiments with and without the washing step. The results were evaluated by calculating the recoveries of fluoride according to Equation 2-1. The average and standard deviation of the fluoride 40 recovery calculated for the PFAS mixture under different conditions with the AOF and EOF methods are provided in Figure 2-3A and Figure 2-3B, respectively. Figure 2-3. A: AOF fluoride recovery results with a mixture of ten PFASs or a mixture of ten PFASs with fluoride spiked to 300 µg L-1. B: EOF fluoride recovery results with a mixture of ten PFASs or a mixture of ten PFASs with fluoride spiked to 300 µg L-1. The two bars in the same section indicate whether or not the washing step was included. Results represent an average of triplicates. Asterisks were used to indicate the significance level compared between groups through conducting Welch Two Sample t- tests, the significant codes for p-value: 0 ‘****’ 0.0001 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05. The bars at the bottom of each figure represent the percentage contribution of the PFASs and the unremoved fluoride to the overall PFAS recovery with fluoride spiked in the presence of washing steps. The AOF analysis revealed an average fluoride recovery of 60.2% for the PFAS mixture in Milli-Q water with the washing step and an average fluoride recovery of 82.6% for the PFAS mixture in Milli-Q water without the washing step (the upper part of Figure 2-3A). This significant difference (t-test, p = 0.005 < 0.05) reinforces our previous findings that the washing step results in a substantial loss of PFASs during the AOF method leading to a lower fluoride recovery. However, the washing step is not the sole 41 reason for the observed incomplete recovery because the average recovery without washing is also less than 100%. Our systematic evaluation of the methodological steps suggests that this deviation is attributable to the relatively low combustion efficiencies for PFSAs, as we have previously demonstrated that all PFASs were successfully captured onto the activated carbon during the loading step. These results strengthened our previous conclusions. The washing step proved essential in minimizing the interference of fluoride when the PFAS mixture was spiked into the simulated groundwater matrix containing fluoride. The average fluoride recovery with the washing step in the simulated groundwater was 56.7% (the lower part of Figure 2-3A), which is similar to the fluoride recovery in Milli- Q water. The slight difference may be due to the competition of adsorbent surface sites with fluoride in the simulated groundwater matrix. Without the washing step, the recovery reached 446.2% of the theoretical value that would be expected based on the organofluorine in the PFAS alone. This extreme value indicates that fluoride can be captured by activated carbon through electrostatic interactions during the loading step and re-transferred into hydrogen fluoride via combustion, as observed in previous studies.16,49 However, this recovery is lower than the recovery based on the theoretical value that would be expected (505.6%) based on the organofluorine in the PFAS and the fluoride in the simulated groundwater matrix. This can be explained by the limited adsorption and solid combustion efficiency of fluoride, which were found to be 76.5 ± 2.6 % and 90.1 ± 5.8 %, respectively (95% confidence interval), in control experiments conducted for this study. These results highlight the importance of including a washing 42 step in the analysis of AOF in fluoride-containing water matrices. We also assessed the performance of the washing step by calculating the remaining percentage of fluoride using Equation 2-6 (Fluoride_spike, Fluoride_rinse). The results are shown in the bottom bar of Figure 2-3A. Our findings indicate that most fluoride can be removed during the washing step, and only 0.83% of the inorganic fluoride contributes to the final fluoride recovery values (the bottom bar of Figure 2-3A). This interference is negligible and does not affect the AOF determination. The average EOF recoveries for the PFAS mixture in Milli-Q water with or without the washing step are 74.4% and 84.6%, respectively (the upper part of Figure 2-3B). The performance of EOF with washing is significantly better than AOF (paired t-test, p = 0.02 < 0.05), which is consistent with our previous findings for individual PFAS recoveries. With the washing step, the EOF recovery dropped, but not as significantly as AOF. We attribute that to the tighter binding between ultrashort-chain PFASs and the WAX adsorbent compared to activated carbon. The recovery for EOF without washing was not 100% in Milli-Q water, even though it was demonstrated that all PFAS were captured by the WAX adsorbent. We attribute this to the incomplete elution observed in the evaluation of each methodological step. As the binding becomes too tight, the elution buffer may not successfully desorb all of the short- and long-chain PFASs. These results corroborate our previous conclusions. In the case of mixing PFASs in simulated groundwater, the average EOF recovery with washing was 76.1% (the lower part of Figure 2-3B). This result is not statistically 43 different from the 74.4% recovery from the Milli-Q water washing experiment (t-test, p = 0.65 > 0.05). This indicates that the 0.01% NH4OH buffer can completely remove bound fluoride as intended. Without the washing step, the average EOF recovery increased to 120.4%, indicating that fluoride can sorb to the WAX adsorbent and be eluted into the organic solvent. However, we found that the extraction and liquid combustion efficiency of fluoride were poor, at 6.8 ± 1.4 % and 4.2 ± 0.2 %, respectively (95% confidence interval), in control experiments conducted for this study. This explains the relatively minor interference of fluoride in the EOF method compared to AOF, but as fluoride concentration becomes high, its influence may still be considerable. Thus, our results demonstrate that the EOF washing step is still necessary to be applied in the analysis of real water matrices, even though washing can lead to a loss in fluoride recovery from PFAS, which is consistent with the findings from other literature.16 The washing step performance was also evaluated using Equation 2-6 as described previously for AOF. We found that only 0.27% of the PFAS mixture recovery resulted from unremoved fluoride (the bottom bar of Figure 2-3B). This is an improvement over the 0.83% observed in the AOF experiment, indicating that the EOF method with washing is can be reliably applied to complex water matrices with fluoride interference. Together, these results reveal that that the EOF method is more appropriate than the AOF method for measuring TOF in samples containing significant fractions of ultrashort- and short-chain PFASs. 44 2.4 Conclusions and Implications There is a need to quantify TOF in water samples and a variety of other matrices as a means to accurately and comprehensively assess exposure risk to PFASs and other organofluorine-containing chemicals. In this study, we systematically evaluated the performance of the AOF and EOF methods for the recovery of ten PFCAs and PFSAs. We intentionally included three ultrashort-chain PFASs in the study because no previous work has assessed the performance of the AOF and EOF methods for ultrashort-chain PFASs. We also uniquely integrated our TOF analyses with LC-MS measurements to assess the performance of each step of the AOF and EOF methods in bulk and for individual PFASs. Our study revealed that the overall recovery ranged between 9.3 – 103.3% and between 21.0 – 108.1% for the ten selected PFASs in Milli-Q water for the AOF and EOF methods, respectively. In both methods, the recovery of the ultrashort-chain PFASs was significantly lower than the recovery of short- and long-chain PFASs. In comparing our data to other studies in the literature, we found that even small changes in the methodology can lead to significant changes in the measured recovery. Our systematic evaluation of each step in the AOF method revealed that decreases in overall recovery are primarily the result of the loss of ultrashort- and short-chain PFASs during the washing step and relatively low combustion efficiencies for PFSAs. Our evaluation of each step in the EOF method revealed that decreases in overall recovery are primarily the result of incomplete elution of short- and long-chain PFASs and the loss of ultrashort-chain PFASs during the washing step. Experiments in simulated groundwater 45 spiked with inorganic fluoride demonstrated the essential need for the washing step which nearly completely eliminated inorganic fluoride from the sample prior to combustion. Together, our experiments suggest that the EOF method is more appropriate than the AOF method for measuring TOF in samples containing significant fractions of ultrashort- and short-chain PFASs in pure water or in simulated groundwater spiked with inorganic fluoride. We expect that our findings will stimulate continued optimization of AOF and EOF methods, particularly for ultrashort- and short-chain PFASs. Because ultrashort- and short-chain PFASs were lost during the washing step for the AOF and EOF methods, identifying techniques to more selectively remove inorganic fluoride would be a significant development. This could be done as a sample pretreatment step (e.g., fluoride ion-selective membranes) or by identifying adsorbents and washing solutions that selectively remove inorganic fluoride without interfering with ultrashort-chain PFASs. Careful evaluation of activated carbon combustion parameters or EOF elution parameters could also improve the overall recovery of the AOF and EOF methods, respectively. 46 CHAPTER 3 SUMMARY AND FUTURE RESEARCH 3.1 Summary This thesis provides a systematic evaluation of each step in the AOF and EOF methods to investigate the causes of incomplete recoveries for individual PFAS species in Milli- Q water and simulated environmental water systems. The research questions outlined in Chapter 1 have been addressed, and the following is a summary of the major findings of each question:  Question 1: How do current AOF and EOF methods work with 10 selected long-, short-, and ultrashort-chain PFASs? What relationships exist between individual PFAS recoveries and PFAS physiochemical properties?  Answer 1: The overall recovery for the ten selected PFASs in Milli-Q water ranged between 9.3 – 103.3% and 21.0 – 108.1% for the AOF and EOF methods, respectively. This result is generally in accordance with most studies in the literature, but small changes in the methodology can have a significant impact on the measured recovery. In both methods, the recovery of the ultrashort-chain PFASs was significantly lower than the recovery of short- and long-chain PFASs. Specifically, the three ultrashort-chain PFASs exhibited recoveries ranging from 9.3 – 25.2% for AOF and 21.0 – 51.5% for EOF. These results suggest that higher 47 recoveries are typically associated with increasing chain length for both methods.  Question 2: How does each methodological step affect the final AOF and EOF recovery? What are the primary reasons leading to incomplete recovery for selected PFASs?  Answer 2: The AOF method involves three separate steps: loading of the activated carbon, washing of the activated carbon, and combustion of the activated carbon. Our systematic evaluation of each step in the AOF method revealed that decreases in overall recovery are primarily the result of the loss of ultrashort- and short-chain PFASs during the washing step and the incomplete mineralization of PFSAs during combustion. On the other hand, the EOF method involves four separate steps: loading of the WAX adsorbent, washing of the WAX adsorbent, solvent-based elution of the WAX adsorbent, and combustion of the elution solvent. Our evaluation of each step revealed that decreases in overall recovery are primarily the result of incomplete elution of short- and long-chain PFASs and the loss of ultrashort-chain PFASs during the washing step.  Question 3: To what extent does the presence of inorganic fluoride influence the AOF and EOF PFAS assessment? Does the washing step effectively remove all fluoride?  Answer 3: Experiments in simulated groundwater spiked with inorganic fluoride 48 demonstrated the essential need for the washing step as fluoride can be captured by both adsorbents, eluted into the organic solvent (in the EOF method), and transferred into hydrogen fluoride via combustion. The washing step can effectively eliminate inorganic fluoride from the sample prior to combustion. In the presence of the washing steps, the contribution of fluoride from external media was found to be 0.83% for AOF and 0.27% for EOF. 3.2 Future Research Future studies in this field may benefit from an even more comprehensive evaluation of the current AOF and EOF methods and an optimization of these measurements to enable the complete recovery of all PFASs in the presence of complex matrix constituents. Such efforts could provide valuable insights and lead to the development of more accurate and effective methods for analyzing TOF in environmental water samples. The detailed rationale and suggested approaches are described below. 3.2.1 Supplement to the Evaluation This study provides a comprehensive evaluation of each step in the AOF and EOF methods to interpret incomplete fluoride recoveries for specific PFAS species. However, it should be noted that the ten PFASs selected in this study, while covering various chain lengths and highlighting the incomplete quantification of ultrashort-chain PFASs, are all anionic PFASs that share similar chemical properties. The performance of the AOF and EOF methods for PFASs with different electrical properties, such as cationic, nonionic, and zwitterionic PFASs, remains largely uncharacterized.197 Previous studies 49 have demonstrated that the activated carbon used in the AOF method exhibits comparable adsorption capacity among PFASs with different charge states.198 However, recent research by Han et al. (2021) suggests that zwitterionic PFASs exhibit lower AOF recoveries compared to PFAAs with the same fluorinated carbon number, and the functional groups as well as branching also affect AOF recoveries.49 For the EOF method, the WAX resin used in the loading step is a selective adsorbent for anionic PFASs.168,199 The adsorption efficiency of cationic and nonionic PFASs might be limited in the EOF method, as their binding would primarily depend on weaker non- electrostatic interactions (e.g., van der Waals, dipole-dipole, hydrophobic interactions.).146,200 The recovery loss for these PFASs might be caused by limited adsorption efficiency rather than incomplete elution and removal during the washing step. Therefore, it might be essential to include PFASs with different types in future research to achieve a more comprehensive understanding of the AOF and EOF methods. In addition, it should be noted that the validation of the methods in this study did not fully account for the potential impact of environmental matrix constituents. While inorganic fluoride is the primary concern in CIC measurements, other factors such as DOM, conductivity, and pH can also significantly affect the accuracy of the analytical results. For example, a deeper analysis of data presented elsewhere suggests that matrix constituents have a significant effect on fluoride recoveries in the AOF method. A quantile plot (Q-Q plot) analysis on data from Han et al., (2020) shows that the AOF recovery of the same analyte in Milli-Q water spiked with external constituents does not follow a multivariate normal (MVN) distribution (Figure 3-1A), which is typically 50 assumed when the experimental error is the only influencing factor. Our principal component analysis (PCA) of AOF recoveries for PFOS, PFOA, and PFHxS spiked in different types of water also indicates that water characteristics account for approximately 30% of the total variance (Figure 3-1B). Additionally, the efficiency of the SPE step in the EOF method can be impacted by the presence of other constituents in the water matrix.200,201 Studies have demonstrated significant variations in the recoveries of PFASs spiked in ultrapure water, drinking water, tap water, and effluent from a municipal wastewater treatment plant when implementing SPE.202-204 These findings emphasize the need for further investigation into external environmental interference when evaluating the accuracy of the AOF and EOF methods. Figure 3-1. A: The quantile plot of data from Han et al., (2020) illustrating the impact of environmental water matrix factors (fluoride, DOC, and other constituents) on AOF recovery. The 1:1 blue line represents the ideal MVN distribution of the data. The red points within the blue ellipse on the upper side indicate outliers and suggest that the assumption of normally distributed recoveries under different environmental conditions is invalid. B: The PCA biplot shows the relationship between AOF recovery in different water matrices (groundwater, natural water, pure water, and wastewater) and selected water characteristics (DOM, conductivity, pH, and fluoride concentration). Two 51 principal components (PC) were selected according to the North’s rule, PC1 and PC2 explain 85.1% of the total variance. The position of the points represents the correlation of the recoveries corresponding to the two PCs after dimensional reduction, with colors representing the background water matrix. The direction and length of the arrows represent the correlation and strength between different impact factors, respectively. 3.2.2 Approaches for the Optimization We expect our findings to stimulate further refinement of the AOF and EOF methods, especially for ultrashort- and short-chain PFASs. Because our experiments suggest that the recovery loss is at least partially due to the washing step in both methods, identifying techniques that can more selectively remove inorganic fluoride would be a significant improvement. One potential approach for addressing this issue is to incorporate a sample pretreatment step to decrease the inorganic fluoride concentration in water samples. Fluoride ion-selective membrane filtration is a promising method for this purpose,205,206 but choosing the suitable membrane can be challenging because we want to maintain the PFAS concentration. Commonly applied nanofiltration (NF) or reverse osmosis (RO) membranes may not be ideal selections as they can also remove PFASs.207,208 Future research focuses on developing specific chemical properties to remove fluoride without impacting the PFAS concentration could be a useful pathway to explore. Another potential strategy is to select different washing solutions that can selectively remove inorganic fluoride while preserving ultrashort- and short-chain PFASs. Recent research by Forster et al. (2023) suggests that the washing solution applied in the EOF method, ammonium hydroxide, may also be an appropriate choice for fluoride removal in AOF.16 However, because the desorption mechanisms for 52 fluoride and short-chain PFASs are both based on ion exchange, achieving a balance in the experimental effect by adjusting the washing buffer concentration and volume might be difficult.209 Therefore, future research may focus on developing alternative approaches for fluoride replacement based on different principles to minimize the removal impact on ultrashort- and short-chain PFASs. There are several other pathways that could potentially enhance the accuracy of the AOF and EOF measurements. Firstly, it may be worthwhile to consider the application of PFAS selective adsorbents, such as molecularly imprinted polymers (MIPs) and β- cyclodextrin (β-CD) polymers, instead of traditional activated carbon or WAX adsorbents for the loading step.48,166,210 MIPs and β-CD polymers rely on polymer- structural interactions as their primary mechanisms for PFAS capture,160,211 which may eliminate the need for a washing step to remove inorganic fluoride and thus minimize the loss of ultrashort-chain PFASs. Secondly, the elution efficiency of the EOF method could be improved by selecting alternative eluent buffers. Recent studies have demonstrated that the incorporation of NH4OH, H2O, and acetonitrile (ACN) could facilitate a more thorough elution compared to conventional methanol elution.212-216 Therefore, adjusting the organic solvent ratio may help elute short- and long-chain PFASs and ultimately improve the overall EOF recovery. Lastly, the combustion parameters could be scrutinized to regulate the PFAS-fluoride transfer efficiencies. For the AOF method, a more robust airflow and longer combustion time could be implemented to ensure the complete mineralization of PFSAs. For the EOF method, modifying the liquid combu