Four perfluorocarboxylic acid (PFCA) compounds including perfluorobutanoic acid (PFBA, 95%), perfluoropentanoic acid (PFPeA, 97%), perfluoheptanoic acid (PFHpA, 99%), perfluorononanoic acid (PFNA, 97%), and perfluorooctane sulfonate (PFOS, ∼40% in H₂O) were acquired from Sigma-Aldrich and they were used as reference materials. Additional descriptions of all chemical reagents and the sample preparation protocol were summarized in Supplementary Table S1.

The field samples were collected from a municipal WWTP located in Southern California. The MW-6 and MW-5 groundwater samples were collected from two groundwater monitoring wells installed around the WWTP, using dedicated Groundfos submersible pumps, stored in polypropylene sample bottles with Teflon®-free caps, and in compliance with the PFAS sampling guidance document published by the California State Water Resources Control Board (SWRCB). The PFASs of the field samples were determined to be at the ppt level using the EPA recommended Draft Method 1633 (EPA, 2022) by a commercial certified by the Department of Defense laboratory.

Several reference PFAS chemicals, including PFOS, and groundwater samples were prepared by simply drying the liquid mixtures on clean 1 × 1 cm² silicon (Si) wafer chips after depositing 25 µL of the liquid containing PFAS chemicals on the clean Si chip under ambient conditions (Wei et al., 2017; Fu et al., 2018; Sui et al., 2018). Samples were dried in a laminar flow and protected under Parafilm prior to analysis.

A ToF-SIMS V spectrometer (IONTOF GmbH, Münster, Germany) was used to analyze representative PFAS reference chemicals and PFAS-containing groundwater samples. The SIMS analysis was performed using a 25 keV pulsed bismuth (Bi₃ ⁺) primary beam ion under high vacuum of 10^–8 mbar during measurements. The Bi₃ ⁺ primary ion beam scanned over a 500 × 500 μm² area for field water samples and 200 × 200 μm² area for reference chemicals, respectively, with a resolution of 128 by 128 pixels. The pulsed current of Bi₃ ⁺ was set at 0.54 pA at a repeating frequency of 10 kHz. Each spectrum was acquired for 100 scans. The primary ion doses in all measurements were lower than the static limit, and the damage artifacts resulting from the Bi₃ ⁺ primary ion beam was negligible. Mass resolution was in the range of 3000–7000, varying from sample to sample depending on the sample roughness. At least five positive and five negative ion replicate spectra were collected at various locations randomly for each sample including groundwater samples.

ToF-SIMS spectral analysis and 2D image reconstruction were performed using the IONTOF SurfaceLab 7 software. Mass spectra were calibrated using CH⁺, CH₂ ⁺, CH₃ ⁺, C₂H₅ ⁺, C₃H₅ ⁺, Si₂C₅H₁₅O⁺, and Si₃C₇H₂₁O₂ ⁺ in the positive ion mode and CH⁻, C₂ ⁻, C₂H⁻, C₃H⁻, and SiO₂ ⁻ in the negative ion mode, respectively. Results were exported and plotted in Igor 8.0. Interference peaks such as Si were removed before running principal component analysis (PCA). Peaks were selected using spectral overlay. Selected peaks were used in PCA using Matlab (R2020a, Math Works, Inc., United States). SIMS spectral data were treated by normalization to the total ion intensity of selected peaks, square root transformation, and mean-centering prior to performing spectral PCA. More details were available in previous reports (Ding et al., 2016; Zhang et al., 2019a; Wei et al., 2020).

It is assumed that the instrument response counts (y) are linearly related to the standard concentration (x) for a limited range of concentration when there is a linear calibration curve (Armbruster and Pry, 2008). This model is used to compute the LOD. The LOD can be expressed as L O D = 3 S a / b . S a is the standard deviation of the response and b the slope of the calibration curve. The response can be estimated by the standard deviation of either y-residuals, or y intercept, of regression lines (Shrivastava and Gupta, 2011; Yu et al., 2020). The LODs were estimated based on the data and linear regressions fits (Supplementary Figures S1, S2), when using SIMS to quantify low concentration (≤1% usually) species (Médard et al., 2002; Coullerez et al., 2003). It is worth noting that six different concentrations of PFOS and PFPeA solutions were analyzed, and a linear relationship was obtained in the low concentration range (see Supplementary Figure S1). This result indicates that the assumption of a linear relationship between signal intensities and concentrations is reasonable. The LODs of PFPeA and PFOS were determined to be 28 and 5.6 ppm, respectively, using 25 μL sample deposition and the standard bunch mode spectral collection conditions.

The LOD can be calculated as L O D = 3 S a / b . Using this formula, the LOD is determined to be 27.97 mg/L for m/z ⁻ 168.994 and 5.59 mg/L for m/z ⁻ 268.980. The limit of quantification (which they call QL, the quantitation limit) LOQ can be calculated as L O Q = 10 S a / b . Using this formula, the LOQ is determined to be 93.23 mg/L for m/z ⁻ 168.994 and 18.63 mg/L for m/z ⁻ 268.980. The LODs of representative key peaks determined using fitting results (Supplementary Figures S1, S2) and the minimal concentrations used in SIMS analysis, aka LOQs were listed in Supplementary Table S2. Specifically, the LOQs are 2.50 mg/L for both m/z ⁻ 168.994 and m/z ⁻ 268.980 based on experimental values.

The LODs can be improved by increasing the secondary ion yields via either wider pulse width or longer analysis frames. Recent results have shown at least an order of magnitude increase in LODs using this approach (Coullerez et al., 2003; Klump et al., 2018). Sample properties, such as cohesive energy and density, could influence the LOD or surface sensitivity of the primary ion source due to the amount of surface erosion (Muramoto et al., 2012). Method development and optimization is needed to improve the LOD of the target PFAS analytes using ToF-SIMS in the future.

The estimated LODs does not seem to be low as what the more developed LC-MS/MS methods could offer. However, real field water sample analysis results, to be discussed below in more details, show that the SIMS LODs are equivalent to the ppt level LODs from LC-MS/MS. The surface sensitivity of ToF-SIMS has been widely used in forensic analysis (Lee et al., 2016; Cai et al., 2017; Terlier et al., 2020), such sensitivity is not easily translatable to an equivalent LODs in terms of the conventional definition of bulk samples. Regardless, because the forensic surface analysis capabilities inherent of ToF-SIMS, trace PFASs can be detected using simple sample preparation and a small amount (i.e., microliter) of water samples.

During SIMS spectral analysis, relative mass accuracy, defined as Δm = Abs (10⁶ × (m/z^— _obs − m/z^— _the)/m/z^— _the) in ppm, and measurement repeatability shown as standard deviation (S.D.) are key factors for obtaining reliable peak identifications. Peak identifications of analyzed PFAS compounds are summarized in Table 1 and Supplementary Table S2. The relative standard deviation (RSD%) is calculated as peak area S.D. divided by the mean peak area. The RSD% results of representative peaks based on the PFBA and PFOS samples are listed in Supplementary Tables S3–S5, respectively. The RSDs% are generally less than 2.5% for PFOS and PFBA, indicating good reproducibility.

The values of relative mass accuracy of most peaks are less than 30 ppm in the negative mode and less than 100 ppm in the positive mode, suggesting that the peak identification is dependable. The standard deviations of most peak areas are between 10% and 20% among all parallel samples. The standard deviations of peak height are larger than those of peak areas. Using peak area for peak identification would be more dependable in measurement evaluation because the intensity is spread over the mass scale due to imperfect energy compensation and topography effects for a specific ion. Therefore, the peak height consequently gives a smaller value with poorer repeatability. Thus, the peak area standard deviation is better to describe SIMS spectral repeatability. The PFAS measurement repeatability results shown in this work are satisfactory for static ToF-SIMS as a semi-quantitative analysis technique (Gilmore and Seah, 2000; Gilmore et al., 2005; Gilmore et al., 2007). Most importantly, SIMS offers sensitive and reproducible detection of characteristic ions and ion fragments of PFAS readily as shown in Supplementary Figures S3, S4.

The schematic of spectral and 2D image analysis using ToF-SIMS is depicted in Figure 1. The volume of PFAS reference and groundwater samples is 25 μL and the sample can be easily prepared followed with analysis in ToF-SIMS without additional sample treatment. This method is simple in comparison with other techniques such as GC-MS/MS and LC-MS/MS. The latter requires extraction with organic solvents or pretreatment before analysis (Bach et al., 2016; Dauchy, 2019). ToF-SIMS is a mass spectral imaging technique, and both spectra and images are acquired during measurements. Characteristic peaks for each PFAS reference materials are observed in ToF-SIMS mass spectra (See Table 1 and Supplementary Table S2). Moreover, 2D images give direct visualization of the distribution of PFAS components, like the m/z ⁻ 268.980 C₅F₁₁ ⁻ and m/z ⁻ 218.986 C₄F₉ ⁻ in a mixture. This feature of spatial distribution of different components is especially appealing in studying complex PFAS mixtures.

To assure the precision of SIMS spectral measurements, at least five repetitions were acquired for every PFAS reference material and real-world samples in the positive and negative mode, respectively. Good repeatability is illustrated in Supplementary Figures S3, S4. Figure 2 depicts ToF-SIMS spectral comparison of long-chain and short-chain PFASs, including PFBA, PFPeA, PFHpA, PFNA, PFOS, two mixtures containing PFBA and PFOS as well as PFPeA and PFOS, and the Si wafer control in the mass range of m/z ⁻ 0–500 in the negative mode. Additionally, the positive spectral comparisons are depicted in Supplementary Figure S5.

Three main spectral results are shown in Figure 2. First, PFASs, including several representative PFCA compounds and PFOS, are fluorinated compounds. The fluoride ion peak, m/z ⁻ 18.999 F⁻, was observed with much higher intensities in all samples, indicating the detection of fluorine fragments. This finding shows the convenience of direct fluoride detection using SIMS compared to other bulk conversion methods (McDonough et al., 2019; Schultes et al., 2019). Second, typical pseudo-molecular ions [M-H]⁻ peaks were observed and identified for each PFAS reference compound in the negative mode (Table 1). Specifically, these ions are m/z ⁻ 212.968 C₄F₇O₂ ⁻ for PFBA, 262.894 C₅F₉O₂ ⁻ for PFPeA, 362.937 C₇F₁₃O₂ ⁻ for PFHpA, 462.942 C₉F₁₇O₂ ⁻ for PFNA, and 498.914 C₈F₁₇SO₃ ⁻ for PFOS (Figure 2), respectively. As to the two mixtures consisting of PFBA and PFOS and PFPeA and PFOS, the corresponding molecular ions peaks (i.e., m/z ⁻ 212.968 C₄F₇O₂ ⁻ and 498.914 C₈F₁₇SO₃ ⁻; 262.894 C₅F₉O₂ ⁻ and 498.914 C₈F₁₇SO₃ ⁻) are observed in Figure 2B, respectively.

Additionally, characteristic PFASs fragment ion peaks were observed and identified (Table 1). Figure 2A shows that the main fragment ion peaks for PFBA are m/z ⁻ 68.999 CF₃ ⁻, 118.987 C₂F₅ ⁻, and 168.994 C₃F₇. Some higher intensity mass peaks (i.e., m/z ⁻ 81.020, 99.983, 169.034) belong to the fragments of PFBA according to the NIST WebBook reference mass spectra (Linstrom, 1997; NIST, 2023). However, they cannot be identified according to present literature. The pseudo-molecular peak m/z ⁻ 212.979 of PFBA is evident in the spectrum. PFPeA, PFHpA, and PFNA share common fragments peaks with PFBA, including m/z ⁻ 68.999 CF₃ ⁻, 118.987 C₂F₅ ⁻, and 168.994 C₃F₇ ⁻, because of similar molecular structures. In contrast, some peaks show relatively lower intensity possibly due to fragmentation difference among compounds. With the increase of molecular weight of reference PFAS chemicals, a series of fragments were observed, such as m/z ⁻ 218.986 C₄F₉ ⁻ (Figure 2B), 318.962 C₆F₁₃ ⁻ (Figure 2B), and 368.883 C₇F₁₅ ⁻ (Figure 2B). Similarly, some unidentified peaks (e.g., m/z ⁻ 81.020, 96.979, 120.952) are representative in PFPeA fragments. Peaks, such as m/z ⁻ 61.001, 76.974, 85.001, 112.992, 155.016, 220.933, and 242.943, come from PFNA fragments according to the NIST WebBook (Linstrom, 1997).

From the SIMS spectral comparison, representative fragment ion peaks from PFOS were observed and identified in Figure 2, such as m/z ⁻ 79.969 SO₃ ⁻, 98.956 FSO₃ ⁻, 129.954 CF₂SO₃ ⁻, 179.951 C₂F₄SO₃ ⁻, and 229.949 C₃F₆SO₃ ⁻. In the lab-prepared mixture samples, these peaks have significant occurrences with higher mass counts due to the presence of PFOS. Previous analyses using HPLC-MS/MS also report these characteristic peaks from PFOS (Berger et al., 2004). Higher intensity peaks, such as m/z ⁻ 310.954, 361.023, and 460.923, without identification might be related to PFOS, because these peaks appear in the spectra of PFOS and the two-component mixtures containing PFOS. The signal to noise ratios (SNRs) for the labeled ions with low relative abundance in the spectra, such as m/z ⁻ 118.987, 212.986, 268.980, 362.937, 368.833, and 419.984, are 3970, 124, 206, 71, 323 and 27, respectively, which indicate that these ions exist in the PFAS samples with reasonable signal intensities.

Spectral PCA was conducted to confirm the observation of spectral analysis of representative two-component mixtures, including PFBA and PFOS and PFPeA and PFOS, respectively, and to further elucidate characteristic PFAS peaks. Figs. S6a − S6b depict the scores plots of principal component one (PC1), PC2, and PC3; and Figs. S6c − S6d give the corresponding loadings plots in the negative ion mode. Representative pseudo−molecular ion peaks of PFBA, PFPeA, and PFOS have high loadings in the loadings plots. They act as key contributors separating selected PFASs as expected. For example, PFBA, PFHpA, and PFNA are situated in the PC1 positive score quadrant (Supplementary Figure S6A), suggesting that their molecular peaks should have positive PC1 loadings. The PCA results also demonstrate that molecular peaks, such as m/z ⁻ 212.968 C₄F₇O₂ ⁻, 362.937 C₇F₁₃O₂ ⁻ and 462.942 C₉F₁₇O₂ ⁻ corresponding to PFBA, PFHpA, and PFNA in positive PC1 loadings, respectively, are main contributors in the separation among different samples (Supplementary Figure S6C). PC1 cannot separate PFPeA from other samples, while PC2 and PC3 can with PFPeA residing in the PC2 positive scores plot (Supplementary Figure S6A) and PC3 negative scores plot (Supplementary Figure S6B). The molecular peak of PFPeA m/z ⁻ 262.894 C₅F₉O₂ ⁻ is situated in PC2 positive (Supplementary Figure S6D) and PC3 negative loading plots (Supplementary Figure S6E). The mixture substances and PFOS are only separated from other samples by PC1 due to the common component PFOS, and PC1 loadings plots (Supplementary Figure S6C), and the molecular peak of m/z ⁻ 498.914 C₈F₁₇SO₃ ⁻ is situated in PC1 negative.

Figure 3 depicts the normalized 2D image comparison of the spatial distribution of molecular ion peaks among the single components and mixtures in the negative ion mode. The dark sub-regions in Figures 3B,C indicate low ion counts. Detection of PFAS mixture compounds without further sample treatment shows selectivity of ToF-SIMS as a technique to analyze PFASs. Figures 3A−C represents the 2D normalized images and distributions of key peaks of the single components and the two-component mixture consisting of PFOS and PFBA. Representative molecular ion peaks of m/z ⁻ 212.968 PFBA are in red and m/z ⁻ 498.914 PFOS in green. PFOS shows a higher molecular ion peak intensity than PFBA. It is not surprising that the PFOS is predominant in the mixture, in agreement with findings in the spectral analysis. Similarly, the second mixture of PFPeA and PFOS (Figures 3D−F) shows consistent results as the other mixture in Figures 3A–C. The normalized intensity of the molecular ion peak m/z ⁻ 498.914 PFOS is higher than that of m/z ⁻ 262.894 PFPeA. 2D SIMS images give direct visualization of main components as an attractive feature in mass spectral imaging, showing long-chain and short-chain PFASs spatial distribution. This is a unique SIMS feature that bulk MS analysis could not provide.

Figure 4A−b show the spectral comparison plots of real groundwater samples. Fluorinated compounds, such as m/z ⁻ 168.994 C₃F⁻, 268.980 C₅F₁₁ ⁻, and 368.833 C₇F₁₅ ⁻, can be detected in two groundwater samples named MW-5 and MW-6, respectively. In addition, characteristic pseudo-molecular ions [M-H]⁻ peaks were observed and identified for the groundwater samples in the negative mode, such as m/z ⁻ 262.894 C₅F₉O₂ ⁻ and m/z ⁻ 362.937 C₇F₁₃O₂ ⁻. Furthermore, fragment ion peaks with relatively higher masses from PFOS, like m/z ⁻ 179.951 C₂F₄SO₃ ⁻, were observed in groundwater samples. This finding indicates that ToF-SIMS is an extremely sensitive technique for the PFPeA and PFOS detection from the environmental water sample. Interestingly, the representative normalized 2D images of the PFASs related peaks, including m/z ⁻ 218.986 C₄F₉ ⁻, m/z ⁻ 268.980 C₅F₁₁ ⁻, m/z ⁻ 318.962 C₆F₁₃ ⁻, and m/z ⁻ 368.833 C₇F₁₅ ⁻, were observed (Figure 4C), giving direct evidence of PFAS detection. The polluted ground water containing PFAS was in the form of a slurry. Dilution was used to dissipate the particles more evenly on the Si substrate. The mass ion spatial distribution depicted in Figure 4 gives a representation of ions of interest in the complex mixture and their relative abundance to each other in a small volume, namely, several microliters were used to prepare the sample. The relative abundances of the selected ions are different between MW5 and MW6, which were collected from different wells in a polluted site.

Supplementary Figure S8 depicts the comparisons of SIMS 2D images of m/z ⁻ 219, 269, and 363 between the Si substrate (a–c) and the ground water sample MW-6 (d–f), respectively. Unlike the 2D normalized images in Figure 3, 4 in the main text, these results are shown in the measurement counts. The counts of the real-world sample MW-6 are on the order of 10⁴ for peaks of interest. Such intensity indicates that the detected peaks are real and not noise. Comparable results of MW-5 are depicted in Supplementary Figure S9. Spatial distribution of PFAS is important because there is a huge interest to understand the PFAS laden materials to address grand challenges in understanding the fate PFAS degradation and environmental restoration. First, having the PFAS distribution will help answer the question of where PFAS compounds reside in the PFAS-laden materials, e.g., clay or resin as amendment. Liquid extraction can tell you the amount of PFAS but not the location. Also, the ability to offer chemical maps of PFAS and its PFAS degradation products in the PFAS loaded amendments would be attractive to decipher the reaction pathways. Again, liquid extraction and bulk LC-MS or GC-MS analyses could tell you how much not where and how relative to the original location of the PFAS.

Selected peak spectral PCA was conducted to confirm the observation of spectral analysis of laboratory prepared mixtures and real-world groundwater samples. As shown in Figure 5A, PC1 and PC2 can explain more than 66% of all data. PFPeA, PFBA, PFOS, mixture of PFOS and PFBA, and mixture of PFOS with PFPeA are situated in the PC1 positive score quadrant, suggesting the molecular peaks have high positive PC1 loadings, including m/z ⁻ 68.999 CF₃ ⁻, m/z ⁻ 98.956 FSO₃ ⁻, m/z ⁻ 129.954 CF₂SO₃ ⁻, m/z ⁻ 229.949 C₃F₆SO₃ ⁻, and m/z ⁻ 498.914 C₈F₁₇SO₃ ⁻. PC2 separates the PFHpA, PFNA, and PFOS from the two groundwater samples (i.e., MW-5 and MW-6), PFPeA, PFBA, PFOS + PFBA, and PFOS + PFPeA. The characteristic peaks in the PC2 negative loadings are m/z ⁻ 268.980 C₅F₁₁ ⁻, m/z ⁻ 368.833 C₇F₁₅ ⁻, and m/z ⁻ 429.937 C₇F₁₄SO₃ ⁻. This finding is consistent with the spectral analysis results as discussed before. In addition, Figure 5B shows the PCA results of PC2 vs. PC5. PC5 separates the two groundwater samples containing PFASs, and the relevant characteristic peaks are shown in the loadings plots, for example, the peak m/z ⁻ 262.894 C₅F₉O₂ ⁻ has a higher loading in PC5 positive, and m/z ⁻ 362.937 C₇F₁₃O₂ ⁻ has a higher loading in PC5 negative. Loadings plots of PC1, PC2, and PC5 are shown in Figures 5C–E respectively.

The ppt level concentrations of approximately 20 ppb of PFASs in the field groundwater samples were verified based on the commercial laboratory analysis using LC-MS/MS. Additional comparison and quantification will be investigated in the next step. Thus, our results demonstrate that ToF-SIMS can detect PFAS at concentrations in ppt level using micrometer of real water samples, i.e., significantly lower than the estimated LODs. Furthermore, our finding show that ToF-SIMS has the potential to tackle with the challenge of determining PFAS contamination in drinking water and groundwater using the forensic analysis (Zhou et al., 2016; Terlier et al., 2020) and source tracking capabilities (Kempson et al., 2003; Héberger, 2008; Kind and Fiehn, 2010). The latter is a topic that is worth of additional investigation.

PFAS contamination in groundwater and soil is a major concern in the environment (Nzeribe et al., 2019; Anderson et al., 2021; Kurwadkar et al., 2022). Concerns over investigation-derived waste (IDW) continues to grow (Singh et al., 2019; Lenka et al., 2021; Longendyke et al., 2022). IDW refers to water, soil and drill cuttings produced during well installations and sampling activities performed during contaminated site investigations. Recently, the memorandum of Temporary Prohibition on Incineration of Materials Containing PFASs calls for a better understanding of PFAS-laden materials (DOD, 2022). Our results show that ToF-SIMS can provide mass spectra in one-dimension and 2D maps of PFAS as well as PFAS dissociation products. Therefore, ToF-SIMS, as an imaging technique, offers a unique and much needed solution to analyzing and imaging PFAS compounds directly on the surface or substrate of the PFAS laden materials, unlike the bulk LC-MS/MS or GC-MS/MS approaches. The latter methods require sample preparation and extraction, which destroys the PFAS-laden materials.