Are cosmetics a signi ﬁ cant source of PFAS in Europe? product inventories, chemical characterization and emission estimates †

In this study, emission of per-and poly ﬂ uoroalkyl substances (PFAS) from the use of cosmetics in the European Economic Area (EEA; not including Lichtenstein and Iceland) was estimated for the ﬁ rst time. Using the European Commission database for information on cosmetic substances and ingredients (CosIng) (cid:1) 170 structures containing at least – CF 2 – or – CF 3 were identi ﬁ ed as ingredients in cosmetics on the European market. These structures were then cross referenced against the Cosmetic Database “ CosmEthics ” to identify PFAS-containing products. Among these products, polytetra ﬂ uoroethylene (PTFE) and C9-15 ﬂ uoroalcohol phosphate were the most frequently listed PFAS ingredients. Thereafter, a sample of 45 cosmetics spanning 5 product categories was purchased in Sweden and characterized for total ﬂ uorine (TF), extractable organo ﬂ uorine (EOF), and target PFAS. Using measured concentrations, the share of PFAS-containing products in each product category, sales data from Cosmetics Europe, as well as other parameters and assumptions, the annual emission of PFAS from cosmetics after use was estimated. Annual EEA-wide TF and EOF-based emissions ranged from (cid:1) 17 – 38 000 kg F per year and 37 – 5100 kg F per year, respectively, representing combined emission to wastewater and solid waste (low to high emission scenario). Sum per ﬂ uoroalkyl carboxylic acid (PFCA) emissions were considerably lower (21 kg P PFCAs per year; high scenario). While TF-and EOF-based emissions are signi ﬁ cant, they are considerably lower than estimates of TF emission from washing of PFAS-coated textiles in the EU. This work provides the ﬁ rst estimate of PFAS emissions from cosmetics and highlights the importance of using a multi-platform analytical approach for PFAS emission estimates. Environmental signi  cance Per-and poly  uoroalkyl substances (PFAS) are listed ingredients in European cosmetic products, but their contribution to PFAS contamination in liquid and solid waste remains unclear. In this work the prevalence of PFAS in cosmetic products from the European market was investigated using a combination of database mining and comprehensive analytical characterization. These data, together with information on consumer habits related to cosmetic removal, were applied to estimate PFAS emissions to wastewater and solid waste for the  rst time. Since waste treatment does not remove PFAS completely, the disposal of PFAS-containingcosmeticsmayleadtoenvironmentalcontamination.WhilePFASemissionsfromcosmeticsaresigni  cant,theyarelowerthanemissions from other sources, such as washing of PFAS-coated textiles.


Introduction
Per-and polyuoroalkyl substances (PFAS) are a diverse group of over 4700 chemicals, which have been widely used since the 1950s in industrial processes and consumer products. 1The OECD/UNEP Global PFC Group denes PFAS as substances that contain 1 or more uorinated methyl or methylene carbon atoms (i.e.-CF 3 or -C n F 2n -; n $ 1), with some noted exceptions. 2 The extraordinary strength of the C-F bond together with the combined lipophobic and hydrophobic properties of per-uoroalkyl chains make PFAS useful for a wide variety of applications.][5] Since the rst detection of PFAAs globally in humans and wildlife, 6 concerns surrounding the risks of PFAS have steadily increased.PFAS have been associated with a wide range of health effects, including cancer, thyroid and hormonal disturbance, premature birth, neuro-developmental deciencies and immunotoxicity. 7,80][11] However, some exemptions exist, and numerous PFAS remain unregulated and continue to be manufactured and used in consumer products globally.
Ingredients in cosmetics and personal care products are regulated by the European Parliament and Council of the European Union, 12 which stipulates that cosmetic manufacturers must ensure the safety of their products for human health.However, this regulation does not contain requirements on the use and release of substances that may impact the environment, such as PFAS.Instead, these substances are addressed by REACH, which stipulates that polymers and low molecular weight substances imported or manufactured in quantities of <1 tonne per year do not require hazard and risk assessments. 13PFOA and PFOA-related substances (i.e.substances that can degrade to PFOA) have been banned since the 4th of July 2020 in all products sold in the EU with limits of 25 and 1000 ng g À1 , respectively (REACH Annex XVII list of restricted substances (entry 68)).In addition, Sweden requires manufacturers and importers to register the occurrence of PFAS in products in the Swedish Chemicals Agency Products Register, regardless of the concentration used in the product.However, in other EU countries, such regulations are not in place and some PFAS may be completely overlooked in consumer products.
In addition to a variety of PFAS that are intentionally added to consumer products (mostly as emulsiers, antistatics, stabilizers, surfactants, lm formers, viscosity regulators and solvents), PFAS may occur unintentionally as impurities and/or degradation products. 14 Fujii et al. (2013) were the rst to measure PFAS in 24 cosmetics and personal care products, and found that 88% contained PFCAs, with sum PFCA concentrations up to 19 mg g À1 in sunscreens and up to 5.9 mg g À1 in other cosmetic products. 15Thereaer, Schultes et al. (2018) reported target PFAS in a wide range of cosmetic products from the Swedish market, showing high concentrations of PFCAs and PFCA-precursors in some samples, and even higher concentrations of total uorine (TF) and extractable organouorine (EOF). 16Most recently, a study of 231 cosmetics from Canada and the US reported that over 50% contained high levels of TF. 17 While the possible occurrence of inorganic uorine hampered determination of the exact proportion of PFAS-containing products, re-analysis of a subset of 29 products (i.e.$13%) using targeted methods revealed detectable PFAS in all samples, with concentrations ranging from 22-10 500 ng g À1 product weight. 17This is concerning, in particular since North American cosmetic producers are not necessarily required to list PFAS ingredients, 17 thereby hampering consumer efforts to avoid PFAS-containing products.
Given the large surface area of the skin ($22 m 2 ) 18 and the liberal application of some cosmetic products, dermal uptake may represent an important route of exposure to PFAS in cosmetics. 16However, many PFAS in cosmetics are ionic or polymeric, both of which are expected to be poorly absorbed by the skin. 19Dermal absorption of non-ionizable, low molecular weight PFAS is largely unknown.In addition to the potential for exposure from dermal uptake and ingestion of PFAS in cosmetics, PFAS used in these products may pose a risk to the environment.Environmental releases may occur during manufacture and use of cosmetics, or aer use, when the products are intentionally removed from the skin or hair.These residues ultimately enter landlls or wastewater treatment plants (WWTPs) where they are oen poorly retained and may subsequently enter the environment via landll leachate, 20 WWTP effluent, 21 or the use of sludge for agricultural fertilizer. 22owever, to date, the prevalence of PFAS in cosmetics on large scale has not been determined; nor have emissions to waste streams.
The overarching aim of this work was to assess the importance of cosmetics from the European market as a source of PFAS to waste streams.To achieve this goal, an inventory of PFAS in cosmetics was carried out using European cosmetic databases.Thereaer, a selection of decorative cosmetics, hair care, and skin care products were characterized using a combination of targeted PFAS analysis and uorine-specic measurements.Finally, using measured concentrations, the share of products containing PFAS, sales data, as well as other parameters and assumptions, the annual emission of PFAS into European wastewater and solid waste from cosmetics aer use was estimated for the rst time.

Sample collection
An inventory of PFAS and their frequency of occurrence in cosmetic products was determined through database searches (Fig. 1).The European Commission database for information on cosmetic substances and ingredients (CosIng) was searched (for "uoro") to extract unique PFAS ingredients potentially occurring in cosmetic products using International Nomenclature of Cosmetic Ingredients (INCI) identiers.All substances containing at least a -CF 2or -CF 3 (including polymers), were included in the search.Note that substances considered "exceptions" (i.e. a carbon atom with a H/Cl/Br/I atom attached to it 2 ) were also included.It is pertinent to note that one INCI name may comprise several PFAS (e.g."C9-C15 uoroalcohol phosphate", seven PFAS) but for practical reasons, each INCI name was considered as a single PFAS for the database analysis.
The $170 unique PFAS ingredients identied in CosIng were then cross referenced against the CosmEthics European database, in order to connect PFAS INCI names to specic cosmetic products within 11 product categories (Table S1 †).Each of the product categories contained a range of sub-categories.For example, anti-aging cream, masks, and eye gel were among the sub-categories within the category "Facial Care" (Table S1 †).In order to maximize sampling of sub-categories containing a high share of PFAS-containing products, the following approach was used: First, summary statistics on PFAS-containing products received from CosmEthics (EU/EEA barcode product database extract only) were reviewed and the share (%) of PFAScontaining products was calculated for each product sub-category.The most relevant sub-categories were identied as those with both the largest share of PFAS-containing products and the largest total number of products.Thereaer, a total of 50 anticipated samples was distributed among the most relevant product categories and sub-categories according to the number of PFAS-containing products in a given sub-category (i.e.percent distribution).Sampling aimed to collect the number of products identied from each of the targeted subcategories (Table S1 †).
In September 2020, a total of 45 different cosmetics were purchased (43 listing PFAS as ingredients, 2 which did not) either online (1 product) or from local stores in Stockholm.For further consideration and emissions calculations, products were re-categorized into the following 5 product categories: "decorative cosmetics", "hair care", "skin care", "toiletries", and "perfumes and fragrances".This was necessary in order to align with the categories of the retail sales data provided in a report by Cosmetics Europe 23 which was later used for emissions calculations (see below).Among purchased products, a total of 24 were "decorative cosmetics", 6 were "hair care" and 15 were "skin care" (out of which 2 products were for males).An overview of the investigated samples, brand names and listed uorinated ingredients (i.e.INCI names) is available in Table S2 were purchased from Wellington Laboratories (Guelph, Canada).A list of isotopically-labelled internal standards is provided in Table S3.† Additional reagents used for sample preparation are provided in the ESI.† Sample handling and instrumental analysis.All samples were analyzed directly (i.e.no sample extraction) by combustion ion chromatography (CIC) for determination of total uorine (TF).In addition, a subset of 15 samples were selected in a manner which maximized the diversity of listed PFAS ingredients while covering the three main product categories and several subcategories.This subset was extracted with methanol and analyzed by CIC and liquid chromatography-tandem mass spectrometry (LC-MS/MS) for EOF and target PFAS, respectively.All extraction and instrumental analysis methods have been previously validated and reported (see Schultes et al. (2018) 16 and also tables S3, S4, and text in the ESI † for details).The results of ongoing QA/QC procedures as part of this work are provided in the following section.For comparison to TF and EOF data, targeted PFAS concentrations were converted to uorine equivalents (C F_PFAS ; ng F g À1 ) using eqn (1): where C PFAS and n F are the concentration and number of uorine atoms for a given target, A F is the atomic weight of uorine (19.0 g mol À1 ), and MW PFAS is the molecular weight of the target (g mol À1 ).Once the concentrations were converted to uorine equivalents (i.e.ng F g À1 ), they were summed to obtain SC F_PFAS concentrations, which were directly comparable to EOF and TF measurements.
Quality assurance and quality control.Interlaboratory comparison of TF measurements performed at Stockholm University have shown good agreement with other labs for paper matrices. 24In the present work, TF accuracy and precision were further conrmed through triplicate combustions of a certied reference material (CRM; BCR®-461, uorine in clay), which was measured at the same time as real samples.The average percent recovery of these measurements was 90% (2.5% relative standard deviation [RSD], n ¼ 3) and data were not corrected for losses.Boat blanks and a mid-level calibration standard were run intermittently to monitor potential carry-over and instrumental stability, respectively.All TF data were blanksubtracted and limits of detection (LODs) were calculated as 3 times the standard deviation of the blanks.Intra-sample variability of TF was assessed by analyzing 10% of samples in triplicate (i.e. 5 individual samples analyzed in triplicate).
Interlaboratory comparison of EOF measurements in water and sludge performed at Stockholm University have shown reasonable agreement with other labs. 25For cosmetics, accuracy and precision were assessed by replicate matrix spike/recovery experiments, consisting of a PFAS-free cosmetic product which was extracted and analysed with and without fortication of a suite of PFAS (277 ng F fortication level; n ¼ 3).The average percent recovery from these experiments was 61% (14% RSD) and data were not corrected for losses.Furthermore, an additional spiking experiment was performed using 250 ng sodium uoride (NaF; n ¼ 3), which was not recovered, con-rming that inorganic uorine was removed during the extraction procedure.In addition to analysing 10% of samples in triplicate (i.e. 2 individual samples analyzed in triplicate for EOF), three procedural blanks were handled in the same way as the samples to check for potential contamination introduced during the extraction procedure.A standard of PFOS (1 mg mL À1 ) analyzed routinely with samples revealed good comparability to theoretical concentrations (92-113%), conrming that the response measured on the CIC was F-specic.Finally, potential carry-over and instrumental stability were monitored using the same procedure as explained for TF analysis.All EOF concentrations were blank-corrected and the limits of detection were calculated based three times the standard deviation of the procedural blanks.
For target PFAS measurements, accuracy and precision were assessed using the same replicate spike/recovery experiments as used for EOF, i.e. a PFAS-free cosmetic product which was extracted and analyzed with and without fortication of 10 ng of individual PFAS (Table S3 †).In addition to analyzing 10% of samples in triplicate (i.e. 2 individual samples analyzed in triplicate), three procedural blanks (i.e.no matrix) were handled in the same way as the samples to monitor for potential contamination introduced during the extraction procedure.Finally, during instrumental analysis, instrumental blanks and a mid-level calibration standard were run intermittently to monitor potential carry-over and instrumental stability, respectively.Procedural blanks did not show detectable contamination for any target PFAS.Therefore, limits of detection (LODs) were determined using the concentration obtained from the lowest calibration point with a well-shaped peak displaying an intensity >1000 and a signal-to-noise (S/N) ratio of >3 and converted to a weight-based concentration (e.g.ng g À1 ) using the average sample weight ($0.1 g).Replicate spike/ recovery experiments with target PFAS revealed an average percent recovery for all targets of 83% (RSD 13%; n ¼ 3).Most substances fell within the range of 85-95%, with the exception of PFHxDA and PFOcDA, which had lower recoveries (61 and 36%, respectively).Considering that internal standards were added aer extraction (necessary because a portion of the same extract was used for EOF analysis), these recoveries are reasonable, and reect some losses incurred during the extraction procedure and/or matrix effects arising from the absence of exactly matching, isotopically labelled internal standards.Consequently, no correction for losses was made.Detection limits ranged from 1 to 18 ng PFAS g À1 cosmetic product sample, which are higher than for environmental sample analysis due to the low quantity of sample used.
As a nal QC measure, a sample previously analyzed in Schultes et al. (2018) was re-analysed in the present study for TF, EOF and target PFAS concentrations.TF and EOF concentrations measured here were in good agreement with previous measurements (3100 vs. 2900 ng F g À1 and 1260 vs. 1380 ng F g À1 , respectively).For target analysis, sum PFCA ( P PFCA) concentrations were also in good agreement (6.27 ng F g À1 vs. 8.11 ng F g À1 , respectively), but higher diPAP concentrations were observed in the present work compared to Schultes et al. (2018) (487 ng F g À1 here vs. 72 ng F g À1 in Schultes et al. respectively).This is perhaps not surprising considering that a different method was used for diPAP determination in Schultes et al. (2018) and because challenges were noted in the previous study for determination of PAPs in these samples due to their extremely high concentrations.

Emission estimates
Annual PFAS or uorine emission (E PFAS ; kg PFAS per year or kg F per year) was estimated aer cosmetic use to wastewater and/ or solid waste for the European Economic Area (EEA; not including Lichtenstein and Iceland, population roughly 450 million people) according to eqn (2).
Each variable contains a number of calculations and assumptions which are explained in more detail in ESI (see Section S3, Tables S8-S11 †).C PFAS is the concentration of PFAS (or uorine) in a product obtained based on measurement of target PFAS (mg PFCAs per g product) or TF (mg F g À1 product), EOF (mg F g À1 product).The total amount of products sold annually (A products ; tons per year) was derived from the retail sales price from the European cosmetic products market (EEA, not including Lichtenstein and Iceland), the market share of the different product categories, 23 and the assumed average product price and size per product category.The share of products containing PFAS (f PFAS products ; unitless) was derived by dividing the total number of cosmetic products containing PFAS by the total number of products for each product category (numbers from the CosmEthics database, but product subcategories recategorized into cosmetics Europe categories to match the sales statics categories, see Section S3, Table S12 †).Finally, the fraction of the chemical released (f release ; %) from the product into a certain compartment (i.e.wastewater or solid waste; see Section S3 †) was obtained by considering consumer habits 26 and assumptions described in detail in the ESI (Section S3, Table S14 †).
Finally, three different waste stream emission scenarios (i.e.low, average and high) were calculated for each of the three measurements (i.e.TF, EOF and targeted PFAS) based on: (i) concentrations (minimum, average and maximum, Tables S5-S7 and explanation in ESI †) in each product category; (ii) three release scenarios (i.e.f release % into wastewater and solid waste, Table S14 †) based on statistics on consumer removal habits 26 (Table S13 †) and (iii) three product price levels (Table S10 †), for the overall product categories.The difference between the total amount of PFAS in cosmetics and the amount that is emitted via wastewater was considered to end up in solid waste (i.e.disposal of cosmetic products removal pads/tissues etc. into waste, leover product amount in the package, disposal aer a product's lifetime).Further, PFAS emissions during the production of cosmetic products are not considered, neither are releases of volatile PFAS into the air.A detailed explanation of each input parameter for the different emission scenarios is provided in the ESI, † as well as an overview table on the ingoing parameters and their values applied in the emission calculation for the different scenarios (Tables S15-18 †).

Result and discussion
Product characterization TF and EOF analysis.TF was measured in a total of 45 cosmetics (43 PFAS-containing products and 2 not listing PFAS ingredients; Fig. 2 and Table S5 †).The 43 PFAS-containing samples comprised of 23 decorative cosmetics, 14 skin care and 5 hair care products.Among products listing uorinated ingredients, TF concentrations ranged from below LOD (i.e.<0.001 mg F g À1 in one hair spray sample and <0.035 mg F g À1 one mask) to 13.8 mg F g À1 sample (one skin care product).TF concentrations in skin care products were the most variable (<LOD -13.8 mg F g À1 ), followed by decorative cosmetics (0.02-6.01 mg F g À1 ) and hair care (<LOD -0.05 mg F g À1 ).Considerable variability was also observed within sub-categories, e.g.blusher/bronzer/contour (0.90-5.14 mg F g À1 ), foundation/BB cream (0.02-3.31 mg F g À1 ), anti-age cream (0.08-3.80 mg F g À1 ), and mask (<LOD -10.6 mg F g À1 ).Generally, the TF range of the cosmetic products compared well to prior studies on cosmetic products in Sweden 16 (<0.091-19.2mg F g À1 ), Denmark 27 (0.03 to 0.740 mg F g À1 ), and the US 17 (0.155-3.96 mg F g À1 ).
As was shown in Schultes et al. (2018), there was some evidence that inorganic uorine contributed to the TF signal in some samples. 16For instance, TF concentrations were above the LOD for eye liner pen 1, which did not list any PFAS ingredients.Assuming that the uorine does not occur from a residual or undeclared PFAS ingredient, it is likely attributable to inorganic substances (e.g.synthetic uorphlogopite/mica or uoride).Indeed 24 samples listed "mica" (not specied as synthetic or natural) and 7 samples listed "synthetic uorphlogopite" (Fig. 2 and 3) but the exact quantities remain unclear.While EOF and targeted PFAS analysis do not capture inorganic uorine and can be used to rule out its occurrence (as was the case here), these approaches may underestimate the total PFAS concentration (depending on the PFAS present), since some organo-uorine substances may not be extracted with the solvent and/ or measured in either case (i.e. during EOF or targeted PFAS analyses).We conclude that all three approaches may be susceptible to bias (TFover-reporting; EOF and targeted PFAS under-reporting); therefore, a combined approach, presenting upper and lower bound estimates, is the most transparent and comprehensive strategy for product characterization.
EOF measurements were performed in 15 of the 45 cosmetic products.Concentrations of EOF ranged from <LOD (<162 ng F g À1 for exfoliator, <325 ng F g À1 for concealer 2 and Eyeshadow 5) up to 4.93 mg F g À1 for foundation/BB cream 4 (Fig. 3a, Table S6 †).The second highest EOF concentration was detected in foundation/BB cream 3 (1.58mg F g À1 ).Similar to TF, EOF concentrations varied both within product categories and subcategories.While skin care samples appeared to have lower EOF concentrations than hair care and decorative cosmetic products, the small sample size (n ¼ 6) made it difficult to draw rm conclusions about trends of PFAS concentrations among product types.Overall these results compared well with prior work by Schultes et al. (2018) which reported EOF concentrations of <LOD up to 1.72 mg F g À1 , with the exception of one product in the current study which contained EOF concentrations up to 4.93 mg F g À1 (foundation/BB cream 4).
For most samples, EOF concentrations were $10-10 000-fold lower than TF concentrations, indicating that the listed PFAS ingredient(s) were poorly extracted and/or that the products contained high concentrations of inorganic uorine (Table S16 †).The few instances of EOF exceeding TF concentrations (e.g.foundation/BB creams 3 and 4, pressed powder 2) are attributable to the variability in TF and EOF analyses, and have been documented previously. 28In these three cases the EOF and TF are considered equivalent, indicating that the organo-uorine ingredient is efficiently extracted and that inorganic uorine is negligible (Table S16 †).However, the observation of EOF concentrations at similar levels to TF in the case of pressed powder 2, was surprising, considering that the listed ingredient (polyperuoroethoxymethoxy diuoroethyl PEG phosphate) View Article Online appears to be polymeric and was not expected to be extracted efficiently with methanol.In comparison, all other polymercontaining products (e.g.Eyeshadow 5 and lip liner, pen 1, eye liner, pen 2 and aershave) displayed EOF levels that were much lower than TF (range 0.01-6.8%).It can be speculated that the polymer in pressed powder 2 is much smaller (i.e.shorter chain length) and more soluble in methanol than the two other polymers in the aforementioned products.Targeted analysis.Target PFAS measurements were performed for the same samples subjected to EOF analysis (i.e. 15 of the 45 cosmetic products); and the results of these analyses are summarized in Fig. 3b.In total, 7 out of 15 samples contained measurable concentrations of at least one PFAS.P PFCA concentrations ranged from <LOD to 9560 ng g À1 with the highest concentrations in mask 2. The second highest P PFCA concentration (425 ng g À1 ) was observed in another mask product (mask 1, Table S7 †).Samples of foundation/BB creams, masks, concealer 1, eye shadow 5, lip liner pen 1 contained up to 7 out of the 21 monitored PFAS above detection limits, whereas samples of anti-age cream 2, exfoliator, lip liner pen 2, pressed powder 2, concealer 2, shampoo, styling cream and aershave did not contain detectable concentrations of targeted PFAS.None of the PFSAs from the target list were detected above the LOD, which is in accordance with previous studies from Europe, 14,16 but contrasts recent work on North American cosmetic products, which detected low (i.e.< 30 ng g À1 ) concentrations of PFSAs in most Canadian products and 1 US product.The most frequently detected PFCA was PFBA in seven of the products, followed by PFHxA in three of the products and PFHpA in two of the products.PFPeA, PFOA, PFNA, PFDA, PFDoDA and PFTeDA were each detected in one sample (PFPeA in foundation/BB cream 4, the other PFCAs in foundation/BB cream 3, Fig. 3b).PFOA was measured in one sample and occurred at concentrations above the EU limit of 25 ng g À1 (112 ng g À1 in foundation/BB cream 3). 30Notably, analysis of 29 North American cosmetic products 17 revealed sum PFAS concentrations ranging between 22 and 10 500 ng g À1 , with 6 : 2 uorotelomer alcohols (FTOHs) and methacrylates (FTMAs) among the dominant PFAS.While FTOHs and FTMAs were not included in the present study, it was expected that both classes of PFAS (if present) would be captured by EOF measurements.Overall, the diversity of PFAS ingredients and residuals in consumer products which are not captured by targeted approaches highlights the importance of including TF-and EOF-based measurements for a comprehensive picture of PFAS contamination in consumer products.
Our targeted PFAS analysis also detected intentionally added PFAS ingredients in two products.Foundation/BB cream 3 and 4 Fig. 2 TF concentrations (mg F g À1 ) of different cosmetics, grouped according to product category.Bar colors represent samples containing polymers (red), perfluorinated cyclic/aliphatic species lacking a functional group (yellow) and PFAS species with functional group(s) (grey) that appeared on the ingredient list.Samples of mask 1, treatment 2 and hairspray 2 are < LOD.Structures T and Y are adopted from literature, 29  contained "peruoroalkylethyl phosphate" and "uoroalcohol phosphate", respectively, which assumably refers to poly-uoroalkyl phosphate esters (PAPs).Indeed, diPAPs were measured in both products, at concentrations up to 2.3 mg g À1 (6 : 2 diPAP; Fig. 3b).6 : 2/8 : 2 and 8 : 2 diPAP were detected at concentrations up to 0.67 mg g À1 (Fig. 3b), which far exceeds the EU limit of 1000 ng g À1 for PFOA-related substances. 30These diPAP concentrations exceeded concentrations of all other PFAS by several orders of magnitude in some cases, demonstrating that targeting listed ingredients will greatly increase P PFAS concentrations, compared to if exclusively impurities are measured (e.g.PFCAs, FTOHs, etc.).This underlines the need for more analytical standards and methods in order to quantify the diversity of PFAS which are intentionally added to products.

PFAS emissions
Low, high, and average emission scenarios for PFAS from the use of cosmetics were estimated using TF, EOF, and P PFCA (C 4 -C 18 )-concentrations, resulting in a total of 9 estimates for combined emission to wastewater and solid waste (Table 1).These estimates should not be confused with environmental emissions, since the fraction of PFAS retained and/or destroyed during treatment (i.e. by wastewater treatment plants, landlls, or incineration plants) has not been included in the calculation, nor were emissions occuring during cosmetic production.For TF-based estimates, emissions ranged from 17 kg F (low emission scenario) to 38 000 kg F (high emission scenario) per year.Since TF concentrations capture all uorinated substances (both organic and inorganic) these estimates will err on the side of overestimation.Moreover, caution is warranted when converting these uorine concentrations to PFAA-equivalents (for example to estimate surface water concentrations), since many of the uorinated ingredients captured by TF measurements are not PFAA-precursors and may not be water soluble.Rather, they are polymers which may contribute to particulate (i.e.nano-or microplastic) pollution, or possibly retained in the wastewater treatment plant.By comparison, EOF-based emission estimates, which ranged from 37 kg F (low emission scenario) to 5100 kg F (high emission scenario) per year do not include contributions from inorganic uorine, but are limited to (semi-) polar PFAS, typically PFAAs and/or PFAA-precursors.Polymeric To investigate the importance of PFAS emission from cosmetics relative to other sources, emission estimates calculated here were compared to literature data, expressed on a uorine equivalent basis (Fig. 4).While onward into the environment 32 (Fig. 4).Considering PFCA emissions alone, the contribution from cosmetics appears low.However, it is pertinent to note that our emission estimates for cosmetics only include release from the products themselves, whereas other studies included the emission from production as well.Considering the high emission scenario using TF data, emissions from cosmetics are only 3.5-fold lower than emissions from outdoor textile washing (i.e.38 tonnes F per year vs. 133 tonnes F per year) 33 and TF data are in the same range as estimates for the global emissions of uorine during the life cycle of PFHxS-and PFDS-based products (0.4-11.1 tonnes F per year compared to the cosmetic range 0.02-38 tonnes F per year). 34

Implications
The present work demonstrates that both regulated and unregulated PFAS exist in cosmetic products from the EEA, sometimes at concentrations exceeding EU limits for PFOA (25 ng g À1 ) and PFOA-related substances (1000 ng g À1 ) since the 4th of July 2020 for products sold in the EU. 30 Considering that regulation of PFOA and PFOA-related substances under REACH has existed since 2019 and are now included under Annex A of the Stockholm Convention, new products or product version ought to exclude these compounds.While the proportion of products listing PFAS ingredients is small relative to the total number of products on the market, the emissions to wastewater and solid waste may be signicant, albeit lower than other sources (e.g.outdoor textiles).Nevertheless, questions remain as to why these products are needed in cosmetics or why some regulated PFAS still appear in these products and how to exclude them in case of impurities.

Study limitations, uncertainties and future improvements
This study was subject to a number of limitations and uncertainties related to inventory development, product sales estimates, analytical characterization and emission estimates, which are described in detail in the ESI.† A brief overview of the most important sources of error from each parameter is described here: Inventory development.The total number of PFAS occurring in cosmetic products and/or existing as INCI names is likely an underestimate, either because products were missed during the database search, were not named correctly in the database, or because one INCI name oen includes several different PFAS.
Analytical characterization.A weakness of the study was that the total number of samples analyzed represents a small fraction of the total number and variety of products available.Moreover, the focus of the study was primarily on products with listed PFAS ingredients; the extent of PFAS contamination in supposedly "PFAS-free" (according to listed ingredients) products in the EEA remains unclear.While high PFAS concentrations were previously observed in North American cosmetics 17 which did not disclose PFAS ingredients, a similar result in European products would be surprising, given the requirements for disclosure of ingredients.However, PFAS as impurities from the raw materials or subsidiary technical materials cannot be ruled out in the current study either, as they do not count as ingredients (Regulation (EC) No 1223/2009 of the European Parliament and of the Council of 30 November 2009 on cosmetic products).Neither can PFAS be ruled out in products that do not follow the existing legislations, which we discovered in a few cases. 12Further, emissions based on TF, EOF, or target PFAS each have advantages and limitations: while TF captures all PFAS it may be subject to overestimation due to the cooccurrence of inorganic uorine.On the other hand, while EOF is not affected by inorganic uorine (assuming an appropriate extraction procedure is used to remove it), some PFAS are clearly not extractable using the polar solvents employed here.Thus, EOF may underestimate the PFAS content in some cases.Finally, target PFAS analysis is useful for sensitive and quantitative determination of individual substances, but undoubtedly underestimates total PFAS content, since listed PFAS ingredients are seldom included in targeted analytical methods and since the PFAS analyzed for was in this study limited to PFCAs, PFSAs and PAPs.
Emission estimates.PFAS emissions estimated in the present work were only calculated for the product aer use and release to wastewater and solid waste.The releases during production and potential losses while use (e.g. to air) was not considered, nor the fraction released to the environment following treatment of solid or liquid waste.Finally, information on consumer habits pertaining to cosmetic removal were not available for all product categories, thus assumptions were required.Since the quantity of cosmetic products sold per year (on a mass basis) was unavailable, several assumptions were made in order to estimate these values, which could lead to large differences in emission estimates.The two parameters with the biggest inuence on the cosmetic product amount sold per year are likely the price per product and the size of a product.A 10% change in either of these parameters would each result in a corresponding 10% change in the total emission estimates.

Fig. 1
Fig.1Schematic of cosmetic products database screening and selecting samples, analysis and emission estimation workflow."PFAS-containing" and "PFAS-free" refer to the ingredient list on the product.
Fig. 2 TF concentrations (mg F g À1 ) of different cosmetics, grouped according to product category.Bar colors represent samples containing polymers (red), perfluorinated cyclic/aliphatic species lacking a functional group (yellow) and PFAS species with functional group(s) (grey) that appeared on the ingredient list.Samples of mask 1, treatment 2 and hairspray 2 are < LOD.Structures T and Y are adopted from literature, 29 but are uncertain based on the INCI name provided.Samples labelled PFAS-FREE did not list any PFAS ingredients.*Mica listed among ingredients; #synthetic fluorphlogopite listed among ingredients.

Fig. 3
Fig. 3 (a) Fluorine mass balance showing TF, EOF and targeted PFAS concentrations (mg F g À1 sample).The targeted PFAS ( P PFAS) concentrations were converted to fluorine concentrations for comparative purposes.(LOD TF ¼ 35.45 mg F g À1 , LOD EOF ¼ 0.32 mg F g À1 , LOD P PFAS ¼ 0.0015 mg F g À1 ).*Mica listed among ingredients; #synthetic fluorphlogopite listed among ingredients.(b) Concentrations (ng g À1 sample) of targeted PFAS in 5 decorative cosmetics and 2 skin care samples (masks).Samples not shown (8 out of 15) were <LOD.Samples are plotted on a logarithmic scale (base 10) to enhance readability.The letters on top of each bar correspond to structures provided in Fig. 2.
cosmetics aer use release up to 0.015 tonnes F from C 4 -C 18 PFCAs annually within the EEA, annual emission of C 4 -C 14 PFCAs during the entire life-cycle of PFOA and PFNA-based products between 1951 and 2015 were much higher (25-240 tonnes F per year). 31In comparison, Li et al. (2017) estimated that degradation of sidechain uorotelomer-based polymers in waste stocks would release $34-635 tonnes of C 4 -C 12 PFCAs per year from 2015

Fig. 4
Fig. 4 Comparison of PFAS emission from other consumer products.Values in the brackets show the fluorine equivalent emission calculated for each study; ranges represent low to high emission scenarios.Note that cosmetics emissions only consider PFAS release after the use of cosmetics, while emissions from the life cycle of PFDS-and PFHxS-based products considers production as well.

Table 1
Annual estimates for total (emissions to wastewater and solid waste), each in a low, average and high emission scenario for the different cosmetic product categories based on the TF (kg F per year), EOF (kg F per year) and P PFCA (kg P PFCA/per year) measurements.All values are rounded to two significant figures.Emission estimates for Perfumes and Fragrances category were negligible and are not shown.The higher emissions for EOF compared to TF (low emission scenario) is because the lowest measured EOF concentration among 15 samples characterized for EOF was higher than the lowest measured TF concentration among the 45 samples characterized for TF.A value of "0" indicates negligible emission for a given emission scenario based on the input parameters and assumptions