INTRODUCTION the belief that these compounds will be
and polyfluoroalkyl substances (PFASs) are a group of diverse chemicals that are
used in or during the manufacture of many consumer and industrial products. Due
to their unique properties, such as thermal and chemical stability and water-
and oil-repelling abilities, PFASs are commonly used as surfactants in
firefighting foams, paints, non-stick cookware, carpets, upholstery, shampoos,
cosmetics, food contact material, etc.1,2 However, these unique properties lend PFASs to being of
human health and environmental concern, as they can bioaccumulate in humans and
wildlife and are extremely resistant to degradation.3,4
PFASs, containing eight or more carbon atoms 1, 5 (e.g., perfluorooctanoic acid (PFOA), perfluorooctanesulfonate
(PFOS)), have garnered significant attention in the scientific community, amongst
regulatory authorities, and within the public due to an increased understating of
the human health effects associated with exposure to these compounds and their
ecological persistence.3,4 In 2016, the U.S. Environmental Protection Agency
(USEPA) established a lifetime health advisory (HAL) of 70 ng/L for the sum of
PFOA and PFOS concentrations in drinking water.6,7
Also in 2016, PFOA was classified as possibly carcinogenic to humans by the
International Agency for Research on Cancer (IARC), 8 and the U.S. National Toxicology Program (NTP)
classified both PFOA and PFOS as immune hazards to humans.9 This accumulating evidence has led to the
development of short-chain PFASs and fluorinated alternatives (e.g., perfluoroether
substances) by fluorochemical manufacturers, with the belief that these
compounds will be less persistent and toxic.9,10 However, data from which these conclusions have been
made are limited and often not publically available.10
research has indicated the presence of legacy PFASs and a class of fluorinated
alternatives, perfluoroalkyl ether acids (PFEAs), in the Cape Fear River (CFR)
of North Carolina downstream of a PFAS manufacturing facility, as well as in
finished water from drinking water treatment plants (DWTPs) along the CFR
(Table S1 and Figure S1). 12,13 One of these PFEAs, the ammonium salt of
hexafluoropropylene oxide-dimer acid (HFPO-DA), a known PFOA alternative
commonly referred to as “GenX”, has also been identified in surface water in
Germany, the Netherlands, and China.14,15 A limited number of studies have evaluated the
toxicity, pharmacokinetic behavior, and environmental fate of GenX,15-22 but no
studies, to our knowledge, have evaluated the health impacts of the other PFEAs
identified in the CFR.
processes capable of removing PFASs include anion exchange,23,24 activated
carbon adsorption,23,24 nanofiltration,23
and reverse osmosis.25 Further,
preliminary data have demonstrated that PFEAs can be removed by anion exchange24 and activated carbon adsorption.13,24 No studies
have evaluated PFEA removal by high-pressure membranes.
2017 survey found that 72% of adults in the United States currently own at one
or more water filtration devices. 36% of those surveyed owned a refrigerator
filter, 11% owned an under-sink filter, and 11% owned a whole house filter.26 Refrigerator filters commonly contain an
activated carbon block (ACB), under sink filters commonly contain reverse
osmosis (RO) membranes and/or ACBs, and whole house filters commonly contain granular
activated carbon (GAC).27, 27.5 Data
from prior studies suggest that these point-of-use (POU; treat water where it
is consumed or used)28 and
point-of-entry (POE; treat water as it enters a residence)28 devices may be an effective PFAS barrier at
the household scale. NSF/ANSI Standard P473 has certified numerous POU-RO and
POU-ACB systems to remove PFOA and PFOS from a summed concentration of 1,500
ng/L to <70 ng/L.29,30 However, no testing has been conducted evaluating household treatment devices for removal of other legacy PFASs or PFEAs.
objective of this study was to (1) evaluate household water treatment systems
for the removal of legacy and emerging PFASs in the Cape Fear River basin to
provide (a) citizens information regarding effective PFAS exposure controls and
(b) DWTPs preliminary information regarding the effectiveness of activated
carbon and reverse osmosis for PFAS removal.
MATERIALS AND METHODS
samples were taken on four sampling dates between June 30 and December 19,
2017. Locations sampled included eleven homes in Brunswick and New Hanover
counties, which are provided water from two DWTPs with raw water intakes along
the CFR downstream of the PFAS manufacturer. Water source for the homes was
determined from publically available distribution maps published by local water
utilities.31-33 One home, based on
our results, appears to have received a mixed supply of distributed water, derived
from both a PFAS-containing surface water source and a groundwater source with
non-detectable PFASs concentrations. Basic water quality parameters from
Consumer Confidence Reports (CCRs) for each DWTP are included in Table S2. Served
DWTP populations, maximum daily capacity, process configuration, and water
source are included in Table S3.
total, fourteen household drinking water treatment systems were tested for legacy
and emerging PFASs; this included six POU-RO systems, five POU-ACB systems, and
three POE-GAC systems. Manufacturer’s expected lifetime (MEL) for each system
was obtained by owner’s manuals, with this information and system age being
included in Table 1. It should be noted that all POU-RO systems tested
contained ACBs in pre- and post-filters, in addition to RO membranes. For these
systems, two MELs are included. In the case of POU-ACB 4, an in-line ACB filter
was placed before a refrigerator ACB filter. Two MELs are provided for this
setup as well.
Table 1. System information for each
household water treatment system. In the case where a manufacturer provided range
for an MEL, an average was used.
POU-RO system tested had a pressurized storage tank that was filled by DWTP
distribution system pressure. To obtain a temporally representative sample,
homeowners were asked to empty the POU-RO storage tank (i.e., open the faucet
until flow ceased) 12 to 6 hours prior to the sample time and 6 to 2 hours
prior to the sample time, with a minimum of 4 hours between each flush. The 4
hours ensured a full tank flush, according to manufacturer listed tank fill time
as included in Table S4.
POU-ACB and POE-GAC systems, filter flushing was done just prior to the sample
being taken. Flushing was done for activated carbon systems to ensure a
consistent empty bed contact time (EBCT) for each sample. During periods of low
water use (e.g., overnight, vacation), a water parcel may be in contact with
the activated carbon media for longer than during typical use, and therefore
bias results. Flush times varied based on the size of the system.
a conservative flow rate of 6 Lpm for kitchen faucets and 40 L bed volume
POE-GAC systems, a 6.67 minute EBCT was estimated. Three bed volumes of
flushing would require 20 minutes, and 1 minute of flow was estimated for home
piping between the system and the kitchen faucet, totaling 21 minutes of
assuming a flow rate of 6 Lpm for kitchen faucets and a 6 L bed volume for POU-ACB
under sink systems, a 1 minute EBCT was estimated. Three bed volumes of
flushing would require 3 minutes, and piping between the system and the kitchen
faucet was deemed negligible and not included in calculations.
a flow rate of 1 Lpm for refrigerator water outlets and a bed volume of 1 L for
POU-ACB refrigerator systems, an EBCT of 1 minute was estimated. Three bed volumes
of flushing would require 3 minutes, and piping between the system and the
refrigerator spout was deemed negligible and not included in calculations.
flushing and sample collection was done with the faucets completely open and
with cold water. Samples were collected in 1 L HDPE bottles and preserved with
5 mL of 1:1 nitric acid (15.8 N) and ultrapure water. Information about
analytical standards and liquid chromatography?tandem mass spectrometry (LC?MS/
MS) methods for PFAS quantification is provided in the Supporting Information.
removal for each household water treatment system is provided in Table 2.
Table 2. Household water treatment system
removal of PFASs occurring at the highest concentrations in water samples. Bolded
and italicized numbers indicate desorption. Two asterisks indicate desorption with
a non-detect initial peak area count; in this case, percent desorption cannot
be quantified. Calculations based on concentration are provided for GenX,
whereas all other calculations are based on chromatographic peak area counts.
Osmosis Systems. Multiple
studies have shown that nanofiltration and RO membranes are effective for the
removal of legacy PFASs, including short-chain PFASs, within DWTPs and in laboratory
experiments replicating DWTP conditions. This includes high rejection for perfluorobutanoic
acid (PFBA), with a formula weight of 214 Da.34-36 It has been suggested that for PFAAs, rejection is
governed by both size exclusion and electrostatic exclusion.35 The PFEAs detected in this study have formula
weights that range from 178 to 463 Da and all are predicted to be strong acids
and anions at typical DWTP pH values (Source?),
which indicates they would likely be effectively removed by a DWTP membrane
filtration. However, a direct comparison cannot be made between DWTP and POU-RO
systems due to operational differences. First, RO membranes in water treatment
plants are often operated at high pressures. For example, six common DWTP NF
and RO membranes are listed with normal operating pressures of ~75 to 225 psi when
operated for removal of organic contaminants.37
The POU-RO membranes tested in this study were pressured by distribution system
pressure, with listed operational pressures ranging from 35 to 120 psi. With a lower
pressure, NaCl rejection is expected to decrease, which is rejected by size
exclusion and electrostatic exclusion.37
Due to these similar rejection mechanisms, PFAA, and subsequently PFEAs
rejection would also be expected to decrease.
of PFASs by POU-RO systems is shown in Figure 1.
Figure 1. Removal of (a) legacy PFASs, GenX,
and PFMOBA and (b) PFEAs by point-of-use reverse osmosis systems. As authentic
standards were not available for PFEAs other than GenX and PFMOBA, chromatographic
peak area counts are shown in Figure 1b. GenX data are shown in both figures
for reference. Samples were collected in June and August 2017. PFASs with
concentrations below the QL (10 ng/L) were not plotted in Figure 1a. Nafion BP2
was not included in analysis for POU-RO 1, 2, and 3.
shown in Figure 1a, all legacy PFASs, GenX, and PFMOBA were brought below the
QL of 10 ng/L for the six POU-RO systems tested. However, as illustrated in
Figure 1b by the small GenX peaks for peak area counts relative to other PFEAs,
GenX, PFMOBA, and legacy PFASs only make up only a small portion of the total
PFAS signature. Due to this, system performance is better determined by looking
at total PFAS removal in terms of peak area counts, as provided in Table 2. In
the case of POU-RO devices, total PFAS removal ranged from 95 to 99%.
shown in Table 1, POU-RO systems 2 – 6 are relatively new. POU-RO 1, however,
was at 141% of the MEL for the RO membrane and 75% of the MEL for pre- and
post-filters at the time of sample. Due to only an influent and effluent system
sample taken, we cannot discern whether PFAS removal (97%) was governed by
activated carbon adsorption or membrane filtration. However, as the pre- and
post-filters were installed in October 2016… (Elaborate
once I have new ACB results). Therefore, this system’s performance
suggests that POU-RO systems are effective through their lifetime due to
effective membrane filtration. However, additional research is required to
further validate this claim.
Activated Carbon Block Devices. Granular
activated carbon (GAC) and powdered activated carbon (PAC) have been evaluated
for adsorption of legacy PFASs and PFEAs.
a thermally activated wood-based PAC in a laboratory study added at a high dose
to raw Cape Fear River water, short-chain PFEAs, such as PFMOPrA and PFO2HxA, were
essentially non-adsorbable and GenX was only slightly removed, while PFOA and
PFOS were effectively removed.
full-scale GAC column was tested at a DWTP on the CFR downstream of the PFAS
manufacturer with an EBCT of 14 minutes. At ~3,500 bed volumes, GenX had
reached 7% breakthrough, and after ~5,000 bed volumes, it had reached 74%
breakthrough. Nafion by-product 2, a larger PFEA at
464 Da, was effectively removed through ~5,000 bed volumes, but sampling from
the top sample port form the column showed 82% breakthrough at 10,000 bed
volumes. For reference, pilot-scale data showed PFOA breakthrough of >20%
after about 9,000 bed volumes and PFOS breakthrough of >20% after 13,500 bed
These results indicated that
NSF/ANSI Standard P473 testing for PFOA and PFOS removal certification is not
sufficient evidence to indicate that these filters will effectively remove
Similar to POU-RO systems, a direct
comparison cannot be made between POU-ACB systems and DWTP activated carbon
data. The use of an activated carbon block, rather than powdered or granular
activated carbon, alters
hydraulic flow patterns, pressure, and ____ that
have not yet been explored by research. Additionally, it is difficult to
determine EBCTs or bed volumes for POU-ACB filters due to unique system geometries,
and therefore it difficult to compare POU-ACB MELs to typical activated carbon
of PFASs by POU-ACB systems is shown in Figure 2.
Figure 2. Removal of (a) legacy PFASs, GenX,
and PFMOBA and (b) PFEAs by point-of-use activated carbon block systems. As
authentic standards were not available for PFEAs other than GenX and PFMOBA,
chromatographic peak area counts are shown in Figure 2b. GenX data are shown in
both figures for reference. Samples were collected in August 2017. PFASs with
concentrations below the QL (10 ng/L) were not plotted in Figure 2a.
POU-ACB samples were taken on August 4, 2017, which followed point-source
control measures taken by the fluorochemical manufacturer in late June. Prior
to these measures, total PFAS peak area counts were ~***** in the Cape Fear River, and likely
similar in the finished water of DWTP A and B as prior research has indicated.13
1, a countertop device that was installed more than two years prior to source
control measures taken by the fluorochemical manufacturer, demonstrated
desorption of GenX, as shown in Figure 2a, and PFMOAA and PFO2HxA, as shown in
Figure 2b, at these now lower influent concentrations. However, POU-ACB 1
continued to adsorb Nafion byproduct 2 (BP2), a larger PFEA. This may be
described by a shorter diffusion distance required to reach an adsorption site
within an activated carbon particle (Knappe’s textbook chapter; Randtke and
Snoeyiny). Total desorption, in terms of peak area counts, was 127%.
systems 2 – 5, including refrigerator and under sink all having been installed
in late June and early July, effectively removed PFASs, with total PFAS removal
ranging from 89 to 99%. However, all four of these systems were below 25% of
their MEL, and therefore performance throughout the manufacturer’s expected
lifetime is unknown.
Granular Activated Carbon Devices.
As detailed prior, GAC can adsorb legacy PFASs and PFEAs under certain performance
conditions, including an EBCT of 14 minutes. However, our calculations estimated
that POU-GAC systems would have an EBCT of 6.67 minutes or less, due to
conservative assumptions. One of these assumptions is that the entire POU-GAC
bed includes only activated carbon, when in actuality there is often other
media as well (e.g., sand, gravel, cation exchange resins, anti-bacterial
media, etc.). Limited manufacturer information about system size and GAC volume
limits quantitative analysis that would be typical of DWTP GAC beds, such as
EBCTs and bed volumes treated.
of PFASs by POE-GAC systems is shown in Figure 3.
Figure 3. Removal of (a) legacy PFASs, GenX,
and PFMOBA and (b) PFEAs by point-of-entry granular activated carbon systems.
As authentic standards were not available for PFEAs other than GenX and PFMOBA,
chromatographic peak area counts are shown in Figure 3b. GenX data are shown in
both figures for reference. Samples were collected in June and August 2017.
PFASs with concentrations below the QL (10 ng/L) were not plotted in Figure 3a.
1 is a system that appears to have received a mixed water supply of a PFAS-containing
surface water source (DWTP B) and a groundwater source with without PFASs (DWTP
C). On the day of sampling, we believe that the home was provided primarily groundwater
due to PFEAs being non-detectable in the untreated sample. It appears that the
system became loaded with PFEAs over time and then, as the influent
concentrations decreased due to source control measure and/or an altered water
source, released PFASs, such as GenX and PFMOAA.
systems 2 and 3 demonstrated varied performance. POE-GAC 2 performed poorly,
removing only 3% of all PFASs. Interestingly, GenX was removed by 45%, but
Nafion BP2 removal was essentially negligible. POE-GAC 3 also performed poorly
in terms of total removal, removing only 32% of all PFASs. However, for this
system GenX was removed by 56% and PFO2HxA, PFO3OA, and Nafion BP2 were
effectively removed; PFMOAA appears to have been displaced by these other,
larger PFEAs. PFMOAA is also hydrophilic compared to these other PFEAs, as
indicated by its predicted LogD value in Table S1, which typically indicates
paper provides preliminary information regarding household removal of legacy
and emerging PFASs.
systems, which also employ ACBs, were the most effective at removing PFASs. However,
the black box approach required when sampling previously installed POU systems
limits the conclusion that can be made whether removal is governed by membrane
rejection or activated carbon adsorption. POU-RO 1 does suggest that POU-RO
systems will be effective for their expected lifetime, but further research is
required to evaluate this claim.
installed under sink and refrigerator POU-ACB systems also performed well. **Discuss MEL and system performance over time once new
samples are analyzed, and discuss how MELs should be accepted with caution due
to surrogate testing not always being accurate** However, desorption is
a significant concern with activated carbon when PFAS concentrations vary (seasonal
manufacturing process, source control, etc.), as demonstrated by POU-GAC 1.
systems proved to be ineffective for PFAS removal. We suspect that this is
likely do to EBCTs that are insufficient for adsorption of recalcitrant
micropollutants, such as PFEAs.
1 and POE-GAC 3 provide interesting results that may be useful for DWTPs
impacted by PFEAs, particularly in the CFR watershed. POU-ACB 1 indicates the
stronger adsorption affinity for Nafion BP2 than GenX, PFMOAA, and PFO3OA. In
POE-GAC 3, PFMOAA is desorbed while all other PFASs present are adsorbed. These
results are consistent with past research indicating that long-chain PFASs are
more easily sorbed that short-chain PFASs due to an increase in compound
hydrophobicity associated with a longer fluorinated carbon chain.39 These results may also be explained by
the shorter distance larger PFASs need to travel within an activated carbon
pore to reach an adsorption site compared to their smaller PFASs, despite
larger PFASs tending to have slower diffusion kinetics and a limited number of
adsorption sites. Can we predict LogD of Nafion
to a recent survey, 24% of Americans own pitcher filters (e.g., Brita, PUR),
11% own faucet mounted filters, and 10% own portable water bottle filters.
Additional evaluation of these other devices would yield useful results for
those in PFAS impacted areas. Further, the methods of grab sampling used in
this study are basic, and laboratory tested should be conducted to better
understand POU-RO and POU-ACB removal of legacy PFASs and PFEAs.