Report

Report no. 5/21

Performance of water
"

treatment systems for

PFAS removal
 report no. 5/21

Performance of water
treatment systems for
PFAS removal
This report was prepared by:

Kees Roest, Thomas L. ter Laak, Hans Huiting, Wolter Siegers, Nienke Meekel (KWR),
Coen de Jong, Martijn de Jong, Martijn van Houten, Tessa Pancras, Wim Plaisier
(Expertisecentrum PFAS), Joost Dalmijn (UvA)

Under the supervision of:

E. Vaiopoulou (Concawe Science Executive)

M. Hjort (Concawe Science Associate)

At the request of:

Concawe Special Task on Soil and Groundwater (WQ/STF-33)

Thanks for their contribution to:

Members of WQ/STF-33: M. Bonte, P. Eyraud, R. Gill, T. Greaves, T. Grosjean,

A. Medve, T. Nolte, J. Smith, C. Vautier

Concawe Legal Advisor: G. Crichlow

Members of Legal Issue Group (FuelsEurope)

Reproduction permitted with due acknowledgement

 Concawe
Brussels
June 2021

I
 report no. 5/21

ABSTRACT

Per- and polyfluoroalkyl substances (PFAS) are a group of widely used man-made
organic chemical substances. They contain alkyl groups on which all or many of the
hydrogen atoms have been replaced with fluorine. As such, they contain at least
one perfluoroalkyl moiety, –(CF2)n–. PFAS have been used because of their particular
physicochemical properties: most are stable at high temperatures, recalcitrant to
chemical oxidation and biological degradation (i.e., persistent), and act as a
surfactant. Although beyond the scope of this report, as reported by the Australian
Ministry of Defense 6, among others these properties mean there are a wide variety
of PFAS-containing materials (e.g stain- resistant fabrics, nonstick cookware,
polishes, personal care products, and fire-fighting foams), or materials where PFAS
is used in the process (e.g. Mist suppression in metal plating or photography). Many
such substances may also be bio-accumulative and toxic.

In this study several treatment technologies for PFAS removal were tested in the
laboratory on both groundwater containing PFAS, and firefighting wastewater
obtained from a firefighting training site where firefighting foam was applied. The
treatment technologies assessed were performance of sorbents,
coagulation/flocculation, nanofiltration, foam- and ozo fractionation technologies.
In all cases the PFAS removal effectiveness was evaluated.

This report provides:

 Criteria and background information to select potential treatment

technologies.
 Results of these performance tests of water treatment technologies for PFAS
removal.
 Practicalities such as availability of the technology and experimental feasibility
which are included in the evaluation.
 Recommendations for selection of treatment technologies for PFAS removal in
practice for impacted groundwater and firefighting wastewater.

Experiments showed that all tested sorbents were able to remove PFAS from
firefighting wastewater but the required sorbent dosages were in the g/L range. It
is therefore concluded that groundwater containing PFAS can be treated with one
of the tested sorbents directly, while for firefighting wastewater, which typically
has higher PFAS concentrations as well as other contaminants, a treatment train
approach is likely to be more efficient. An initial treatment, such as flocculation,
nanofiltration or foam- / ozo fractionation that removes bulk PFAS load including
co-contaminants followed by a polishing treatment (e.g. sorbents) that further
reduces PFAS concentrations to acceptable levels is advised, unless a relative small
fixed volume of firefighting wastewater needs to be treated.

This study provides a basis for readers of this publication to select, study and apply
the best available technologies to mitigate risks associated with PFAS
contamination.

II
 report no. 5/21

KEYWORDS

PFAS (per- and polyfluoroalkyl substances), PFCA (per- and polyfluoroalkyl

carboxylic acids), PFSA (per- and polyfluoroalkyl sulfonic acids), long-chain, short-
chain, precursors, treatment technologies, PFAS impacted groundwater,
firefighting wastewater, comparative evaluation, batch testing, pilot testing

INTERNET

This report is available as an Adobe pdf file on the Concawe website

([Link]).

NOTE
Considerable efforts have been made to assure the accuracy and reliability of the information
contained in this publication. However, neither Concawe nor any company participating in
Concawe can accept liability for any loss, damage or injury whatsoever resulting from the use
of this information.

This report does not necessarily represent the views of any company participating in Concawe.

III
 report no. 5/21

CONTENTS Page

SUMMARY VII

1. INTRODUCTION 1
1.1. PFAS 1
1.2. AIM 2

2. PART 1: TECHNOLOGY SELECTION 3

2.1. OVERVIEW OF TECHNOLOGY CONCEPTS 3
2.2. SELECTING TECHNOLOGIES FOR TREATING PFAS IN
GROUNDWATER AND FIREFIGHTING WASTEWATER 4
2.2.1. Preliminary criteria 4
2.2.2. Selection of technologies for experimental testing 4
[Link]. Sorption technologies 4
[Link]. Physical separation technologies 5
[Link]. Reactive degradation 5
[Link]. Selection of technologies 6

3. PART 2: TECHNOLOGY TESTING 7

3.1. PFAS IMPACTED WATER 7
3.2. EXPERIMENTAL SET UP OF REMOVAL TESTS 8
3.3. RESULTS AND DISCUSSION 8
3.3.1. Results PFAS composition 8
3.3.2. Results TOP assay 9
3.3.3. Discussion TOP assay results 12
3.3.4. Other constituents in PFAS impacted waters 12
3.3.5. Results batch sorption and flocculation tests - firefighting
wastewater 13
3.3.6. Results batch sorption and flocculation tests - impacted
groundwater 14
3.3.7. Results batch sorption and flocculation tests - Impacted
groundwater spiked with benzene 17
3.3.8. Discussion batch sorption and flocculant tests 20
3.3.9. Results foam fractionation with ozone and air 20
3.3.10. Discussion foam fractionation with ozone and air 22
3.3.11. Results nanofiltration tests 24
3.3.12. Discussion nanofiltration tests 25
3.3.13. Results small-scale column tests 26
3.3.14. Discussion of small–scale column tests 29

4. PART 3: TECHNOLOGY APPLICATION 31

4.1. ACTIVATED CARBON 33
4.1.1. Introduction 33
4.1.2. Performance 34
4.1.3. Key Design criteria 35
4.1.4. Operational aspects 37
4.1.5. Costs 38
4.2. REMBIND® 39
4.2.1. Introduction 39
4.2.2. Performance 39
4.2.3. Key Design criteria 40
4.2.4. Operational aspects 41
4.2.5. Costs 42
4.3. DEXSORB+® 42
4.3.1. Introduction 42

IV
 report no. 5/21

4.3.2. Performance 43
4.3.3. Key Design criteria 44
4.3.4. Operational aspects 45
4.3.5. Costs 46
4.4. POLYQA-OSORB® 46
4.4.1. Introduction 46
4.4.2. Performance 47
4.4.3. Key Design criteria 48
4.4.4. Operational aspects 49
4.4.5. Costs 50
4.5. FLUORO-SORB® 50
4.5.1. Introduction 50
4.5.2. Performance 50
4.5.3. Key Design criteria 51
4.5.4. Pre-treatment 51
4.5.5. Operational aspects 53
4.5.6. Costs 53
4.6. PERFLUORAD 53
4.6.1. Introduction 53
4.6.2. Performance 54
4.6.3. Key Design criteria 55
4.6.4. Operational aspects 55
4.6.5. Costs 56
4.7. NANOFILTRATION 57
4.7.1. Introduction 57
4.7.2. Performance 58
4.7.3. Key Design criteria 58
4.7.4. Operational aspects 59
4.7.5. Costs 60
4.8. FOAM- AND OZO FRACTIONATION 60
4.8.1. Introduction 60
4.8.2. Performance 61
4.8.3. Key Design criteria 61
4.8.4. Operational aspects 63
4.8.5. Costs 63

5. DISCUSSION AND CONCLUSIONS 65

5.1. DISCUSSION 65
5.1.1. Sorption 65
5.1.2. Flocculation 66
5.1.3. Separation by foam or membrane filtration 66
5.1.4. PFAS removal in relation to matrix composition and
physicochemical characteristics of the PFAS 67
5.1.5. Evaluating and combining techniques towards practical
application 67
5.1.6. Lookup table 68

V
 report no. 5/21

6. ACKNOWLEDGEMENTS 70

7. GLOSSARY 71

8. REFERENCES 73

APPENDIX 1: SAMPLE PREPARATION AND CHEMICAL ANALYSIS OF PFAS 75

APPENDIX 2: DETAILS OF TREATMENT CONCEPTS 83

APPENDIX 3: DETAILS OF EXPERIMENTAL SET UP BATCH SORPTION

EXPERIMENTS WITH ADSSORBENTS AND FLOCCULANT 85

APPENDIX 4: DETAILS OF EXPERIMENTAL SET UP FOAM FRACTIONATION WITH

FIREFIGHTING WASTEWATER 87

APPENDIX 5: DETAILS OF EXPERIMENTAL SET UP MEMBRANE SEPARATION

WITH FIREFIGHTING WASTEWATER 93

APPENDIX 6: DETAILS ON SET UP SMALL SCALE COLUMN EXPERIMENTS 95

APPENDIX 7: ANALYTICAL METHODS AND RESULTS OF OTHER CONSTITUENTS 98

APPENDIX 8: SORPTION ISOTHERMS OF BATCH SORPTION TESTS WITH

GROUNDWATER 101

APPENDIX 9: SORBENT LOADING RATES 102

APPENDIX 10: STRUCTURE OF ACTIVATED CARBON AND PFAS SORPTION 103

APPENDIX 11: OVERVIEW OF TREATMENT TECHNOLOGIES FOR GROUNDWATER

AND FIREFIGHTING WASTEWATER 1 104

VI
 report no. 5/21

SUMMARY

Per- and polyfluoroalkyl substances (PFAS) are a group of widely used man-made
organic chemicals. They contain one alkyl groups on which hydrogen atoms have
been replaced with fluorine. As such, they contain at least one perfluoroalkyl
moiety, –(CF2)n–. PFAS have been used because of their particular physicochemical
properties: most are stable at high temperatures, recalcitrant to chemical oxidation
and biological degradation (i.e. persistent), and act as a surfactant. Many of such
substances may also be bioaccumulative and toxic. PFAS were first synthesised in
the 1940’s and commercialised in the 1950’s in stain-resistant products. PFAS-
containing fire-fighting foams were developed for military applications in the mid-
1960’s but became available for commercial fire-fighting uses in the late 1960’s /
early 1970’s. The oil and gas industry is unlikely to have used PFAS-containing foams
prior to circa 1970. In 2018, the OECD defined over 4,000 different CAS numbers
representing chemicals that can be considered PFAS and this number is growing
each year. Release of PFAS-foams to the environment may have occurred during:

 Firefighting incidents
 Firefighting practice and training
 Vapour suppression during hot work
 Firefighting foam storage, dispensing and handling
 Firefighting system testing

Uses of foam for firefighting lead to contamination of groundwater. They were

commonly used at many industrial sites including petrochemical plants (and other
sites with bulk fuel storage, such as military and civil aviation facilities) and at
firefighting training sites co-located on petrochemical industry sites, or at third-
party training facilities operated commercially or by public fire brigades. This
report seeks to help Concawe members, and others, identify suitable technologies
to treat PFAS from firefighting foams in groundwater, and mitigate new emissions
by on-site treatment of firefighting wastewater residues. Within this study
treatment technologies for both types of impacted waters have been studied. In
addition the tested groundwater was spiked with benzene for selected experiments
to study the effects of petrochemical co-contamination on the treatment
performance.

This study contains two parts. First, a series of promising treatment technologies to
separate PFAS from aqueous phases were tested in a laboratory setting using
groundwater and wastewater obtained from a firefighting training site. Second,
these technologies were also evaluated in a desk based assessment considering their
practical application in the field, costs and potential production of waste. In
addition information was gathered on costs, generation of waste, reuse possibilities
and the physical footprint of such a technique, which can be a constraint when
insufficient space is available.

The selected technologies were: treatment by a variety of sorbents in batch (bottle)

and continuous (column) tests, coagulation/flocculation in batch tests and
nanofiltration, foam- and ozo-fractionation in a continuous tests. In batch tests all
tested sorbents are capable of removing over 85% total PFAS from both PFAS-
impacted groundwater (with or without added benzene) as well as firefighting
wastewater (FFWW). However, as concentrations of PFAS in the firefighting
wastewater were roughly three orders of magnitude higher than in the groundwater,
sorbent dosages in the g/L range were required to obtain PFAS removal. The tested

VII
 report no. 5/21

coagulant reduced PFAS concentrations at lower dosages as compared other

sorbents for both firefighting wastewater and PFAS-impacted groundwater.
However, PFAS levels did not fully drop below limits of detection at the highest
dosages.

The tested foam fractionation with and without ozone separated 60 to 90% of the
PFAS from the bulk water by concentrating it in a foam matrix. These removal rates
were considered insufficient as a stand-alone treatment for the tested waters as
PFAS concentrations were at µg/L level and mg/L level after treatment for
groundwater and firefighting wastewater, respectively. Therefore additional
treatment / polishing is advised. Similar results were observed for a nanofiltration
membrane that was solely tested with the firefighting wastewater. Although the
initial removal exceeded 95%, fouling occurred after several hours to a day of
operation. Thus, the firefighting wastewater might require pre-treatment to
prolong membrane service life. All in all this leads to the conclusion that all tested
sorbents were able to remove PFAS from firefighting wastewater but the required
sorbent dosages were very high. Therefore, a treatment train approach where an
initial treatment, such as flocculation, nanofiltration or foam- / ozo- fractionation
that removes bulk PFAS load including co-contaminants, and a second polishing
treatment (e.g. sorbents) that further reduces PFAS concentrations to acceptable
levels is likely to be more efficient for large volumes of firefighting wastewater.

Results were collected and summarized in the lookup table below. This table
provides a basis for readers of this publication to select, study and apply the best
available technologies to mitigate risks associated with PFAS present in groundwater
and firewater.

VIII
 report no. 5/21

Table I: Lookup table on advantages and disadvantages of treatment technologies. Colours ranging from red (-- very poor) via orange (-poor) and yellow
(fair) to green (+, good and ++, very good); n.a. (grey) is not assessed, or the data is unknown.

Performance Pre-treatment Main treatment Operational aspects Costs

Removal Hydraulic Effects of temperature and
Short Long Treatment strategy
mechanisms loading rate Contaminant pressure Energy Maintenance Chemical Waste
Chain chain Turbidity Fn/Mn TOC Contact time Bed lifetime CAPEX OPEX
/filtration loading rate use and personnel use production
PFAS PFAS
velocity

Sorption

Hydrophobic, (pi-
DESOTEC Granular
pi) electrostatic Fixed bed filtration,
Activated Carbon + ++ - - - 20-30 min 10- 15 m/h + + minimal 0 0 ++ + 0 --
and ionic lag/lead configuration
(GAC)
interactions

2 options:
Hydrophobic- ,
1: discontinuous batch 5-10 min,
electrostatic- 2-3
Rembind® (RB) + + with sedimentation - - - >80% unknown + + unknown, expected minimal 0 0 + + 0 unknown
and (an)ionic
2: mixed media filter sorption
interactions
with sand

Fixed bed filtration, unknown, 0 without

Hydrophobic-
Cyclopure DEXSORB+® lag/lead configuration, expected regeneration
and electrostatic - ++ - 0 ++ 3-10 min + 0 unknown, expected minimal 0 -- ++ + 0
(CP) with on-site around 10 - with
interactions
regeneration possible m/h regeneration

unknown, 0 without
Hydrophobic- not clear yet, possible unknown, expected low. Possible
expected regeneration
PolyQA-Osorb® (PQ) and ionic 0 + (a) fixed bed filtration - -- -- 1-5 min 0 n.a. pressure effects on swelling of 0 -- + + 0
around 10 -1 with
interactions with regeneration material
m/h regeneration
Flow-through filter or
Hydrophobic and
FLUORO-SORB® Des+ batch treatment with up to >15
electrostatic + ++ - - 0 2-10 min + + unknown, expected minimal 0 0 ++ + + 0
(FS) sedimentation or m/h
interactions
floatation

Flocculation

Hydrophobic- mixing tank, unknown, expected better

for filtration
PerfluorAd (PFAD) and electrostatic + (b) ++ (b) sedimentation, + 0 + up to 60 min ++ n.a. performance with increasing - - --- - - -
up to 15 m/h
interactions filtration temperature

Other separation
technologies

Increasing temperature has

positive effects on operational
Nanofiltration (c) Size exclusion - + Filter units 0 (d) 0 0 (d) n/a (short) n/a n.a. n.a. - - -- --- -- --
aspects, increased pre-pressure
reduces pump energy

multiple contact Unknown, but higher

Surface active
Foam fractionation tank(s) in series with temperatures negatively affect
characteristics of - + + + + 20-40 min n/a n.a. n.a. --- - - 0 + -
WITH OZONE compressed ozone gas transfer and saturation, this
PFAS
bubbles might have adverse effects

Discontinuous batch Unknown, but higher

Surface active
Foam fractionation contact tank with temperatures negatively affect
characteristics of - 0 + + + 20-40 min n/a n.a. n.a. 0 0 ++ ++ + +
WITHOUT OZONE compressed air gas transfer and saturation, this
PFAS
bubbles might have adverse effects

a) determination of sorption coefficients was complicated for GW

b) based on removal performance for firefighting wastewater
c) initial performance before fouling
d) suspended solids and TOC were higher than operational range stated by vendor

I
 report no. 5/21

1. INTRODUCTION

1.1. PFAS

Per- and polyfluoroalkyl substances (PFAS) are a group of widely used man-made
organic chemicals. They contain one alkyl groups on which hydrogen atoms have
been replaced with fluorine. As such, they contain at least one perfluoroalkyl
moiety, –(CF2)n–.4 PFAS have been used because of their particular physicochemical
properties: some are stable at high temperatures, recalcitrant to chemical
oxidation and biological degradation, and act as a surfactant. Many also have one
or more of the following properties: recalcitrant to degradation (i.e., persistent),
subject to bioaccumulation, and toxic. These polyfluorinated compounds may
transform into fully saturated perfluorinated compounds in the environment:
polyfluorinated compounds can act as precursors to the perfluorinated molecules.
In 2018, the OECD defined over 4,000 different CAS numbers representing chemicals
that can be considered PFAS.5

Figure 1: PFAS applications / sources (kindly provided by the Australian

Ministry of Defense)6

These substances act as surfactants and can reduce surface tension between
aqueous and hydrophobic phases, which makes them water and oil repellent. PFAS
are widely used, as illustrated in Figure 1 and described further by Gluge, et al. 7.
Their properties benefit applications but are a potential risk for the environment.
These properties potentially lead to persistence, bio-accumulation, and toxicity.
Furthermore, the properties also complicate removal during water treatment such
as remediation of impacted groundwater and treatment of residues from used
firefighting foams. In firefighting (including within the petrochemical industry),

1
 report no. 5/21

PFAS-containing fire-fighting foams have been used to suppress flammable liquid

fires or for firefighting training and spill control since about 1970. The composition
of these firefighting foams changed over the past decades as a result of the
development of new fluor-containing additives and regulations.

The potential persistence and toxicity of PFAS are reflected in water quality criteria
and other regulatory thresholds down to the ng/L range for some PFAS 8-9. This
challenges PFAS-producers, PFAS-containing product manufacturers, and users of
PFAS-containing products across society and industry, to reduce emissions to the
environment and study the impact of this diverse group of compounds.

As a result of firefighting activities in a training setting or real situation, PFAS can

be emitted. Firefighting initially results in large volumes of firefighting wastewater
(FFWW), which can contaminate surface waters and groundwater (GW) when
treatment and containment of the wastewater is not performed appropriately.
Conventional treatment systems such as rapid sand filtration are not effective in
removing PFAS from (waste) water. Activated carbon and ion exchange are
generally applied to remove PFAS from water, but alternative treatment options
are available. Therefore, Concawe seeks to understand the effectiveness of
multiple treatment options for PFAS in firefighting foam-impacted GW and FFWW
itself.

1.2. AIM

The aim of the report is to provide experimental results and criteria to support the
selection of appropriate treatment technologies to remove PFAS from impacted
water. This was done by experimental testing on lab scale to fill data gaps for these
treatment technologies. Thereby, this report builds upon the German Water Centre
(Technologiezentrum Wasser, TZW) review on treatment technologies for PFAS.1
Furthermore, practical criteria such as availability of the technology and
experimental feasibility are included in the evaluation. Thereby this report provides
a basis for the selection and application of the best available technologies to
mitigate PFAS contamination.

2
 report no. 5/21

2. PART 1: TECHNOLOGY SELECTION

In this chapter an overview is given of the different treatment technologies for the
removal of PFAS in water. These technologies are then evaluated against various
criteria, and technologies are selected for experiments.

2.1. OVERVIEW OF TECHNOLOGY CONCEPTS

There are multiple PFAS-treatment technologies available on the market. TZW1

(Technologiezentrum Wasser) evaluated a wide array of available technologies
marketed by numerous suppliers. Conceptually, there are three treatment
concepts. More details can be found in Appendix 2.

1. Sorption technologies: The PFAS will bind to sorption sites in or on the surface
of the sorbent. The compound will distribute between the aqueous phase and
the sorbent, to the energetically most favorable distribution. One can
distinguish adsorption from absorption (extraction), where adsorption is the
chemical interaction of a sorbate (e.g. PFAS) with a solid matrix (adsorbent)
that retains the sorbate on the surface of the matrix. In comparison, absorption
is the distribution of the sorbent into the porosity of a matrix where the
compound is physically retained. Within this report we use the generic term
sorption for the potential adsorption, ion exchange and complexation
interactions together.

2. Physical separation techniques: liquid separation techniques separate the

waste stream into two new streams. The first stream (concentrate) is typically
small in volume and contains the majority of the contamination, while the
second stream is much larger in volume and predominantly free of the
contamination and has nearly all of the volume of water. The separation can
be obtained by the use of a physical barrier such as a membrane. In addition,
this separation can also be obtained by the addition of a complexing agent that
interacts with PFAS by similar mechanisms as described in the sorption section
and where the complexing agent coagulates and flocculates and is filtered in
order to separate the water from the flocs containing the majority of the PFAS.
Finally, the separation can also be obtained by injecting air or ozone to
generate foam that pick up the PFAS from the water and concentrate them in
or on the air bubble interface of the foam layer which is skimmed from the
water surface.

3. Reactive degradation: these are (bio)chemical processes where the PFAS is

degraded into intermediate products or completely mineralized. The
effectiveness of the reactive degradation requires analytics that are able to
determine parent chemicals as well as all transformation products formed
during the reactions in order to define mass balances of the parents and the
transformation products versus total mineralization. If the complete
mineralization of the PFAS cannot be assessed, an additional polishing step is
recommended. Further treatment may be required to remove transformation
products. In case further treatment is necessary, sorption and/or separation
technologies come into play.

3
 report no. 5/21

2.2. SELECTING TECHNOLOGIES FOR TREATING PFAS IN GROUNDWATER AND

FIREFIGHTING WASTEWATER

2.2.1. Preliminary criteria

The treatment of impacted aqueous streams require fit for purpose treatment
technology. The suitability of any treatment technology is defined by many factors.
The following factors should be mentioned:

 The composition of the water stream and in particular the presence of

interfering components (the so-called matrix);
 The treatment efficiency for a wide array of PFAS (technology);
 Treatment costs (costs);
 Sustainability of the treatment (e.g., hazard-properties of the treatment
products);
 Practical implementation on site or at a central location and the technology
readiness level (practical application),
 Required reduction of contamination (target),
 Production of waste streams (waste management).

These criteria are interrelated, and as a result the following criteria have been used
in this study:

 Technology Readiness Level (TRL) and viability to apply on full scale (if not
applied on full scale yet);
 Single process or part of a treatment train;
 Removal efficiency;
 Inflow water quality requirements (pretreatment requirements);
 Waste produced;
 Formation of byproducts and post-treatment requirements;
 Use of chemicals;
 Energy consumption;
 Practical, full-scale applicability and robustness;
 Treatment of generated waste, regeneration or recovery of the sorbent;
 Costs, capital and operational expenditure (CAPEX, OPEX).

Furthermore, technologies can be combined which can improve efficiency or reduce

costs or provide other advantages (waste reduction, logistics) of the full treatment
train.

2.2.2. Selection of technologies for experimental testing

[Link]. Sorption technologies

The different technologies are sorted based on the type of sorbent:

 Granular Activated Carbon (GAC);

 Surface Modified Clay-based (e.g. Matcare®, Rembind®, FLUORO-SORB®);

4
 report no. 5/21

 Ion exchange resin based (Strong Base Anion resins);

 Biobased/Polymeric based (e.g., DEXSORB+®, PolyQA-Osorb®, CustoMem®/
Puraffinity CGM®).

From each of these type of sorbents, at least one product is tested except the ion
exchange material.

The selection of the sorption material is based on:

 The cooperation of manufacturers to supply materials;

 The suitability of the product for the performance of bench- and/or column
tests using relatively small volumes of GW or FFWW.

As stated, ion exchange material was not selected for testing. Both activated carbon
and ion exchange resin are considered as proven and commonly used technologies
by the Concawe STF33 experts. Activated carbon is however incorporated in the
testing program for it is selected as the bench mark technology.

[Link]. Physical separation technologies

Separation technologies are able to remove the bulk of the PFAS into a small amount
of water, like:

 Membranes (Nanofiltration / Reverse Osmosis)

 Foam separation (foam-/ ozo fractionation)
Membranes require an intensive pretreatment if fed with FFWW to prevent
suspended solids and organic material to foul or degrade membranes and thereby
affect treatment. Recently capillary NF membranes were commercialized with a
lower fouling tendency, likely to require less pretreatment. Therefore capillary NF
are preferably be tested for treatment of FFWW.

Foam- and ozo fractionation produce a very limited amount of wastewater

(compared to NF/RO). The IP for the technology has been developed by Evocra, a
company based in Australia. Evocra has not only approved the testing of the
technology, but also assisted KWR in executing the tests.

[Link]. Reactive degradation

Reactive degradation technologies are not considered for testing in this phase of
the project as they have a number of disadvantages:

 Testing them on lab scale is technically challenging due to the many different
test conditions (proprietary knowledge) and the complexity of the necessary
test equipment
 The formation of unknown transformation products
 The uncertainty to scale up to large installations for treating large volumes of
water
 The anticipated high costs (Capex as well as Opex)

5
 report no. 5/21

[Link]. Selection of technologies

In Table 1 an overview is given of the recommended technologies for experimental
testing of PFAS removal. This table is based on the tables as presented by TZW (see
Appendix 8), in which the TRL according to the EU Horizon 2020 programs10 is used
as a more quantifiable alternative for the “technical maturity” as used by TZW.

Table 1: Selection of technologies for experimental testing, “Yes” indicates where a

technology was selected, “No” that it was not.

Technology TRL Preliminary evaluation GW1 FFWW2 Example product/brand

(EU and selection for testing names
H2020)

Granular Activated 9 Yes Yes Yes e.g. Chemviron

Carbon F300/F400, Desotec
Organosorb 10AA, CABOT
Hydrodarco 4000
Surface modified clay- 5-6 Yes Yes Yes e.g. Matcare®, Rembind®,
based sorbents FLUORO-SORB®
IEX without 9 No, because of high TRL, No No SBA (strong base anion)
regeneration practical performance resins, e.g. Purolite®,
already largely known Lewatit®, Amberlite®,
Dowex®
Biobased/polymeric 5-6 Yes Yes Yes e.g. Customem®,
based sorbent DEXSORB+®, PolyQA-
Osorb®, PuraffinityCGM®
Flocculants 6-7 Yes Yes Yes e.g. PerfluorAd®, InSite®
Nanofiltration/reverse 9 Yes for FFWW, No for GW No Yes e.g. BWRO, DOW,
osmosis because presumably Hydranautics
concentrate volumes are
still large
Foam- and ozo 7-8 Yes Yes Yes e.g. OPEC systems, Evocra
fractionation

IEX with regeneration 7 No, because practical No No WBA (Weak base anion)
performance already resins, e.g. Lewatit®,
largely known, Amberlite®
regeneration require
research
Biobased/polymeric 4-6 No, first tests without No No e.g. Customem®,
based sorbent with regeneration DEXSORB+®, PolyQA-
regeneration Osorb®, PuraffinityCGM®
Distillation 7-9 No based on costs for No No n.a.
primary treatment

6
 report no. 5/21

3. PART 2: TECHNOLOGY TESTING

In this chapter the experimental evaluation and studies of the different

technologies are described and the results are discussed.

3.1. PFAS IMPACTED WATER

This study focuses on the treatment of PFAS impacted GW and FFWW. The FFWW
was collected from a firefighting training site in Hungary and the impacted GW was
obtained from a site in Germany. The collected FFWW contains C6 telomers, while
the impacted GW is impacted by historical contamination of PFAS origination from
firefighting activities decades ago. The composition of PFAS in the GW is a product
of the use of PFAS in the past, the hydrological conditions of the subsoil, the
mobility of the different PFAS in the soil and potential transformation of PFAS
(precursors) into other more stable PFAS during soil passage.

Table 2: Typical generalized characteristics of impacted groundwater (GW) and

firefighting wastewater (FFWW)

Impacted groundwater (GW)1 Firefighting wastewater

(FFWW)2

Volumes and duration Large volumes continuous or for Smaller volume generated
longer periods during emergency response
Concentration of PFAS Generally (sub) µg/L range Up to g/L range
Concentration of total organic carbon Generally mg/L range Variable mg/L – g/L range
Salt content Generally low, levels typically found Variable
in freshwater and pristine
groundwater
1
(historically) impacted GW with PFAS near petrochemical plants often contains residues of petrochemicals such as
BTEX.
2
FFWW contains PFAS present and transformed during firefighting activities and petrochemical residues, and organic
(soot) and mineral particles.

PFAS impacted GW was collected at a petrochemical site in Germany and

transported to the University of Amsterdam (UvA), The Netherlands, by using 20 L
jerry cans. The water was collected in July 2019 and January 2020, and was stored
in the dark at 4°C until use. A single jerry can was used to perform all batch sorption
studies except the complexant /coagulant PFAD that was supplied at a later stage.
PFAS impacted GW from firefighting activities often coincides with petrochemical
contamination, since PFAS-containing fire-fighting foams have been used to fight
Class B liquid (hydrocarbon) fires. In order to study the effect of residues of BTEX
on the treatment performance the collected GW sample spiked with 20 mg/L. The
benzene concentration is based on measured BTEX levels in petrochemical impacted
GW (personal communication with Concawe members). All batch experiments with
benzene spiked GW and the PFAD study with and without benzene spike were
performed with a second jerry can from the same batch.

FFWW was collected at a training site in Hungary and transported to the UvA in
Amsterdam, The Netherlands, by using 20 L jerry cans in July 2019. A single jerry
can was used to perform all batch sorption studies except the complexant
/coagulant PFAD that was supplied at a later stage. For the column tests GW was
collected at a petrochemical site in Germany and transported to KWR in
Nieuwegein, The Netherlands, by using fifty 20 L jerry cans on the 18th of June 2020.

7
 report no. 5/21

Here all water was mixed in a single container to be supplied to the columns. Details
on sample preparation and analysis can be found in Appendix 7.

3.2. EXPERIMENTAL SET UP OF REMOVAL TESTS

Different treatment techniques have been tested by batch and pilot testing. The
ultimate goal was to determine the performance of treatment methods to remove
PFAS from different types of water, and define the conditions that can affect the
performance. Table 3 shows the technologies selected for testing.

Table 3: PFAS remediation technologies tested in this research.

Sorption technologies

Granular activated carbon DESOTEC Granular Activated Carbon (GAC)

Remediation and PFAS sorbents Rembind® (RB), FLUORO-SORB® Des+ (FS)

Bio-based or polymeric based sorbent Cyclopure DEXSORB+® (CP), PolyQA-Osorb® (PQ)

Coagulation /Flocculation

Flocculant PerfluorAd (PFAD)

Other separation technologies

Nanofiltration1 NxFiltration (NF)

Evocra
Foam- and ozo fractionation
Foam ozo fractionation (FOF)
Evocra
Foam fractionation (FF)
Foam fractionation (FF)
1 Only applied to FFWW

Details on the experimental set up of the batch sorption experiments with FFWW
and GW for five sorbents and one flocculant can be found in Appendix 3. For foam-
and ozo- fractionation experiments with FFWW and GW a pilot test set up consisting
of a column with an air/ozone bubble supply at the bottom and an extraction
mechanism for the foam fraction at the top is presented in Appendix 4. The
nanofiltration treatment of FFWW is given in Appendix 5 and the small scale
columns experiments with GW can be found in Appendix 6.

3.3. RESULTS AND DISCUSSION

3.3.1. Results PFAS composition

This study tested treatment technologies for FFWW and impacted GW. The
concentrations and composition of PFAS as well as other constituents differ between
these two samples as the impacted GW contains historical PFAS contamination of
infiltrated FFWW from times when PFOS and PFOA were not banned (or AFFF
product specification changed to shorter chain PFAS compounds), and some
firefighting foam contained large quantities of especially PFOS 11.

8
 report no. 5/21

Figure 2: Relative composition of measured PFAS in impacted

groundwater (GW) and firefighting wastewater (FFWW).

composition of GW and FFWW

1.1

1.0 PFBA
PFPeA
0.9 PFBS
0.8
PFHxA
PFHpA
0.7 PFHxS
PFOA
fraction

0.6
L-PFOS
0.5 Br-PFOS
6:2 FTS
0.4
other PFAS <2%
0.3

0.2

0.1

0.0
W

W
G

W
FF

In Figure 2 it can be observed that the composition of the FFWW and GW differ.
FFWW is dominated by fluortelomer sulfonic acids, especially 6:2 FTS, while GW
contains a wider share of perfluorinated sulfonic and carboxylic acids. The
concentrations of the PFAS in the FFWW were three orders of magnitude higher than
in the impacted GW. The bars represent the average of triplicate samples. The
average total detected PFAS concentration in the impacted GW used for the sorbent
tests was 13.4 µg/L (SD 1.7 µg/L, n=3). The average total detected PFAS
concentration of the FFWW in the sorbent tests was: 9478 µg/L (SD 201 µg/L, n=3).

3.3.2. Results TOP assay

TOP Assay is a standardized pre-treatment of water samples or sample extracts

designed to expose underlying PFAS not amenable to standard analysis.
Perfluorinated carboxylates and sulfonates are stated to remain intact under the
conditions of the assay. A large series of samples was oxidized according to the TOP
assay as described in Appendix 7 and by Houtz and Sedlack12. Below the transition
of the composition of the samples are shown for all negative controls (NC)
(Figure 3).

9
 report no. 5/21

Figure 3: The ratio of total PFAS before and after TOP assay for
groundwater (GW n=11)), benzene spiked (BS n=6)) and
firefighting wastewater (FFWW n=6) in the negative controls
(without sorbent/flocculant or before treatment).

100

(NC/NC+TOPassay)
Molar ratio  PFAS

10

0.1
W

W
S
B

W
G

FF

For FFWW, the TOP analysis illustrates an average 41-fold increase based on Molar
concentrations, with a 95% Confidence limit of ranging from 31 to 52 of total PFAS.
The GW with and without benzene spike shows a small yet significant difference
between the TOP treatment and the untreated samples with a 1.45 (95% CL 1.21-
1.69) for GW and 1.40 (95% CL 1.07-1.73) for the benzene spiked GW.

10
 report no. 5/21

Figure 4: The ratio of selected individual PFAS before and after TOP assay for
groundwater (GW n=11)) benzene spiked (BS n=6)) and firefighting
wastewater (FFWW n=6) in the negative controls (without
sorbent/flocculant or before treatment).

PFBA PFPeA
10000 10000

1000 1000
(NC/NC+TOPassay)

(NC/NC+TOPassay)
ratio  PFAS
100 100
ratio  PFAS

10 10

1 1

0.1 0.1
W

W
S

W
W
B

B
W
G

W
G
FF

FF
PFHxA PFHpA
10000 10000

1000 1000
(NC/NC+TOPassay)

(NC/NC+TOPassay)
ratio  PFAS

ratio  PFAS

100 100

10 10

1 1

0.1 0.1
W

W
S

S
W

W
B

B
W

W
G

G
FF

FF

PFHxS 2:6 FTS

10000 10000

1000 1000
(NC/NC+TOPassay)

(NC/NC+TOPassay)
ratio  PFAS

ratio  PFAS

100 100

10 10

1 1

0.1 0.1
W

W
B

B
G

W
FF

FF

11
 report no. 5/21

3.3.3. Discussion TOP assay results

When looking at individual chemicals and their change in concentrations due to the
TOP assay, it can be observed that the TOP assay did not completely transform the
6:2 FTS, suggesting that the TOP assay did not result in full oxidation of this PFAS.
Despite the fact that the total PFAS concentration of (benzene spiked) GW was
hardly affected, minor shifts in the composition of PFAS occurred, where the five
and six carbon perfluorinated acids PFPeA and PFHeA, slightly increased, while the
6:2 FTS that was present at very low concentrations was reduced by the TOP
treatment. The lowest removal was observed for the (benzene spiked) GW. One
should note that the initial 6:2 FTS levels were very low in those samples, leading
to higher deviation in the analytical results.

For the FFWW the individual chemicals show clear patterns. For 6:2 FTS the median
removal was with 81% ranging from 71% to 100%. The FFWW is dominated by 6:2 FTS
(Figure 2). As seen in Figure 3 the TOP treatment resulted in a large increase in
total PFAS concentration. This can be mainly attributed to the increase of the 4-7
carbon chain perfluorinated carboxylic acids, PFBA, PFPeA, PFHxA and PFHpA of
which the median increases were a factor of 630, 1953, 49 and 852 respectively
with variations below one order of magnitude. This illustrates the presence of 6:2
FTS and many other precursors are transformed into other compounds by the TOP
treatment, resulting in the above mentioned 41 (31-52) fold increase in PFAS
detectable with the applied target method.

TOP assay was also performed on all samples that underwent various treatments
such as batch sorption studies, column studies, filtration, and foam fractionation.
These treatments affected the composition of the PFAS in a different way than the
TOP assay themselves. The impact of the treatment on the composition of the PFAS
mixture and the impact of the TOP oxidation on the composition of the PFAS mixture
could not be separated. Therefore the results of the negative controls and the
untreated source waters were used to illustrate the sole impact of the TOP assay
and thereby the composition of persistent and transformable (precursor) PFAS in
the tested waters. Furthermore, the treatment itself might lead to transformation
of precursors. For example the foam fractionation where the added ozone (or air)
might also act as oxidant for precursors during treatment. This might affect the
interpretation of the obtained results in the tests as oxidation processes of
especially the FFWW treatment efficiency might be biased for some compounds in
a positive manner (oxidizable PFAS like 6:2 FTS) and negatively by their products
(e.g. PFBA, PFPeA, PFHxA and PFHpA).

3.3.4. Other constituents in PFAS impacted waters

Analyses of other constituents of the studied water types (GW and FFWW) were
performed by Al-West of the Agrolab group in Deventer, the Netherlands. The
benzene spiked GW was not tested as the composition was identical to the GW with
the exception of the benzene addition. In Appendix 5 the analytical results are
given. FFWW showed three orders of magnitude higher content of organic carbon
measured as Total Organic Carbon (TOC) and Biological Oxygen Demand (BOD) and
Chemical Oxygen demand (COD) than the GW. Inorganic salts were present at
generally orders of magnitude higher levels in the FFWW. Only for inorganics that
are at least partially of natural origin such as Ca, Fe, Mn, SiO2 showed rather similar
levels in both waters.

12
 report no. 5/21

3.3.5. Results batch sorption and flocculation tests - firefighting wastewater

The total PFAS aqueous concentration of the FFWW was analysed for different
treatments with a variety of loads the sorbents Rembind® (RB), Granulated
Activated Carbon (GAC), PolyQA-Osorb® (PQ), Cyclopure (CP) and FLUORO-SORB®
(FS) and the flocculant PerfluorAd (PFAD).

Figure 5: PFAS decrease from the FFWW as a function of sorbent dosing

of the five selected sorbents and one flocculant, tested in
batch sorption tests with 7 doses and a negative control, each
in triplicate. The negative control is plotted as 0.0000001 mg/L
sorbent in order to include the data in the graph, using a
logarithmic X-axis. The tested sorbents were Rembind® (RB),
Granulated Activated Carbon (GAC), PolyQA-Osorb® (PQ),
Cyclopure (CP) and FLUORO-SORB® (FS) and the flocculant was
PerfluorAd (PFAD).

firefighting wastewater

RB
12000 GAC
PQ
10000
CP
FS
PFAD
sum PFAS g/L

8000

6000

4000

2000

0
.0

.5

.0

.5

.0

.5

.0

.5

0
0.

0.

1.

1.

2.
-4

-3

-3

-2

-2

-1

-1

-0

log (g/L sorbent)

In Figure 5 it can be observed that PFAD is effective in reducing the sum of PFAS
(dominated by 6:2 FTS) most rapidly, compared to the sorbents tested. This
observation can be explained by the competition of large amounts of PFAS, oil
residues and other organic contamination present in FFWW at levels up to mg/L
level (Table 23, Table 24, Table 25, Table 26 and Table 27, in Appendix 5). In
conclusion, all sorbents require dosages in the range of grams sorbent per liter of
treated water to render a relevant reduction of PFAS from the aqueous phase.

The fitted curves are logistic dose-response curves, the fit allows to determine the
loading of the sorbent when 50% of the initial total measurable PFAS concentration
in the treated water is reduced. The statistical evaluation also allows the
calculation of 95% confidence limits of the fitted values. The loading rate where
50% of the initial concentration that was around 10 mg/L (10.000 µg/L) is reached
and the rest of the detectable PFAS are sorbed to the sorbent in gram per kg sorbent
is 21.4 (95%CL = 18.0-25.5) for PFAD, 3.8 (95%CL = 3.2-4.4) for GAC, 3.8 (95%CL =
3.2-4.4) for CP, 1.7 (95%CL = 1.2-2.5) for RB, 1.4 (95%CL = 0.8-2.5)for FS. For PQ
data did not allow fitting of the confidence limits but the loading was estimated to
be 2.4 g/kg sorbent.

13
 report no. 5/21

As the FFWW contains much more undetected PFAS, the loading is potentially much
higher. The loading rate of non-detected PFAS cannot be determined. However the
TOP assay is indicative for part of the PFAS that were missed upon initial analysis,
and if we presume that the precursors of the PFAS detected after the TOP analyses
are sorbed to the sorbents in a similar manner as the PFAS that were detected,
these loading rates can be multiplied by a factor of ~41 leading to loading rates
ranging from 10 to over 500 (!) grams per kg sorbent. However, it should be noted
that we do not know the sorption behaviour of these precursors that are detected
as different compounds after oxidation. Nevertheless, this shows that the sorbents
are able to remove relevant fractions of PFAS from water.

3.3.6. Results batch sorption and flocculation tests - impacted groundwater

The total aqueous PFAS concentration of the impacted GW was analysed for
different treatments with a variety of loads the sorbents Rembind® (RB), Granulated
Activated Carbon (GAC), PolyQA-Osorb® (PQ), Cyclopure (CP) and FLUORO-SORB®
(FS) and the flocculant PerfluorAd (PFAD).

Figure 6: PFAS decrease from the impacted groundwater (GW) as a

function of sorbent dosing of the five selected sorbents and one
flocculant, tested in batch sorption tests with 7 doses and a
negative control, each in triplicate. The negative control is
plotted as 0.0000001 mg/L sorbent in order to include the data
in the graph, using a logarithmic X-axis. The tested sorbents
were Rembind® (RB), Granulated Activated Carbon (GAC),
PolyQA-Osorb® (PQ), Cyclopure (CP) and FLUORO-SORB® (FS)
and the flocculant was PerfluorAd (PFAD).

contaminated groundwater
14
RB
GAC
12 PQ
CP
10 FS
PFAD
sum PFAS g/L

0
-4

-3

-2

-1

log (g/L sorbent)

Note that the data for PFAD have a different offset as a different batch of GW was
used that contained 3.6 times more PFAS than the experiments with the sorbents.
In Figure 6 it can be observed that GAC, RB, and FS show the highest removal of
most PFAS, where can be observed that the GAC treatment results in a reduction
down to PFAS concentrations near zero, while the other two sorbents some of the
PFAS remains in the sampled aqueous phase, also at higher dosages of sorbent. PFAD
also shows a steep decline at low dosages, but the residual PFAS (~2 µg/L) at higher
dosages of PFAD. One has to note that the PFAD experiments were performed with

14
 report no. 5/21

a different batch of GW containing a higher initial PFAS concentration. It can be

concluded that all sorbents are rather effective in removing PFAS from GW while
the flocculant is very effective in reducing the PFAS concentrations but does not
lead to (near) full removal. Sorbent dosages below 100 mg/L or even 10 mg/L
appear to be sufficient to remove most of the PFAS from the aqueous phase
(Figure 7).

The fitted curves are logistic dose-response curves, the fit allows to determine the
loading of the sorbent when 50% of the initial total measurable PFAS concentration
in the treated water is reduced. The statistical evaluation also allows the
calculation of 95% confidence limits of the fitted values. The loading rate where
50% of the initial concentration is reached and the rest of the detectable PFAS are
sorbed to the sorbent in gram per kg sorbent is calculated. The results in gram per
kg sorbent are 61.5 (95%CL = not determinable) for PFAD, 10.5 (95%CL = 4.2-26.4)
for FS, 5.2 (95%CL = 1.2-22.0) for GAC, 1.5 (95%CL = 1.4-1.6) for RB, 1.2 (95%CL =
0.8-1.9) for CP and 0.9 (95%CL = 0.6-1.4) for PQ. It can be observed that the
confidence limits of these fitted values are often large. This is likely due to the fact
that the lowest dose of sorbent already leads to a significant decrease of the
dissolved concentration, which on its turn requires the extrapolation to obtain the
sorbent dose were 50% of the aqueous concentration was sorbed. This results in a
rather high uncertainty.

Potentially the GW contains additional PFAS that remained undetected with the
method applied. However, the applied TOP assay did not result in a significant
change of detectable PFAS, so the contribution of undetected precursor PFAS is
presumed limited. Generally the loading rate of the sorbents ranges from 0.9 gram
up to 61 gram of detectable PFAS per kg sorbent at aqueous concentrations of
roughly 7 µg/L for the sorbents and 25 µg/L for the flocculant PFAD.

15
 report no. 5/21

Figure 7: Removal of individual PFAS from the impacted and benzene spiked
groundwater (GW) by five selected sorbents and one flocculant, tested in
batch sorption tests with 7 doses and a negative control in triplicate. The
negative control is plotted as 0.0000001 mg/L sorbent in order to include
the data in the graph, using a logarithmic x-axis. The tested sorbents were
Rembind® (RB), Granulated Activated Carbon (GAC), PolyQA-Osorb® (PQ),
Cyclopure (CP) and FLUORO-SORB® (FS) and the flocculant was PerfluorAd
(PFAD).
PFBA in GW PFPeA in GW
120 120
RB RB
GAC GAC
100 PQ 100 PQ
CP CP
80 FS 80 FS
PFAD PFAD
% removal

% removal
60 60

40 40

20 20

0 0

-20 -20
-4

-3

-2

-1

-4

-3

-2

-1

2
log (g/L sorbent) log (g/L sorbent)

PFHxA in GW PFHpA in GW
120 120
RB RB
GAC GAC
100 PQ 100 PQ
CP CP
80 FS 80 FS
PFAD PFAD
% removal

% removal

60 60

40 40

20 20

0 0

-20 -20
-4

-3

-2

-1

-4

-3

-2

-1

log (g/L sorbent) log (g/L sorbent)

PFHxS in GW PFOA in GW
120 120
RB RB
GAC GAC
100 PQ 100 PQ
CP CP
80 FS 80 FS
PFAD PFAD
% removal

% removal

60 60

40 40

20 20

0 0

-20 -20
-4

-3

-2

-1

-4

-3

-2

-1

log (g/L sorbent) log (g/L sorbent)

L-PFOS in GW Br-PFOS in GW
120 120
RB RB
GAC GAC
100 PQ 100 PQ
CP CP
80 FS 80 FS
PFAD PFAD
% removal

% removal

60 60

40 40

20 20

0 0

-20 -20
-4

-3

-2

-1

-4

-3

-2

-1

log (g/L sorbent) log (g/L sorbent)

16
 report no. 5/21

3.3.7. Results batch sorption and flocculation tests - Impacted groundwater

spiked with benzene

PFAS impacted GW from firefighting activities can coincide with petrochemical

contamination. In order to study the effect of residues of petrochemicals on
treatment performance benzene was spiked to the impacted GW in a controlled
manner. The selected concentration (20 mg/L) is based on measured benzene levels
in petrochemical impacted GW (personal communication with Concawe members).
These are orders of magnitude lower than the petrochemical residues found in the
FFWW.

Figure 8: PFAS decrease from the impacted groundwater (GW) with 20

mg/L benzene spike as a function of sorbent dosing of the five
selected sorbents and one flocculant, tested in batch sorption
tests with 7 doses and a negative control, each in triplicate.
The negative control is plotted as 0.00001 mg/L sorbent in
order to include the data in the graph, using a logarithmic x-
axis. The tested sorbents were Rembind® (RB), Granulated
Activated Carbon (GAC), PolyQA-Osorb® (PQ), Cyclopure (CP)
and FLUORO-SORB® (FS) and the flocculant was PerfluorAd
(PFAD).

contaminated groundwater + benzene

15
RB
GAC
PQ
CP
FS
10 PFAD
sum PFAS g/L

0
-4

-3

-2

-1

log (g/L sorbent)

In Figure 8, it can be observed that GAC, FS show the highest removal of most PFAS.
Furthermore, it can be observed that the GAC treatment results in a reduction down
to PFAS concentrations near zero. PFAD also shows a steep decline at low dosages,
but the residual PFAS (~6 µg/L) at higher dosages of PFAD. One has to note that the
PFAD experiments were performed with a different batch of groundwater containing
a higher initial PFAS concentration. It can be concluded that all sorbents are rather
effective in removing PFAS from GW while the flocculant is very effective in
reducing the bulk PFAS concentrations but does not lead to (near) full removal.
Sorbent dosages below 100 mg/L or even 10 mg/L appear to be sufficient to remove
most of the PFAS from the aqueous phase. Dosages around 10-100 mg/L appear to
be sufficient to remove most of the PFAS from the aqueous phase for the best
performing sorbents.

17
 report no. 5/21

The fitted curves are logistic dose-response curves, the fit allows to determine the
loading of the sorbent when 50% of the initial total measurable PFAS concentration
in the treated water is reduced. The statistical evaluation also allows the
calculation of 95% confidence limits of the fitted values. The loading rate where
50% of the initial concentration is reached and the rest of the detectable PFAS are
sorbed to the sorbent in gram per kg sorbent is calculated. The results in gram per
kg sorbent are 61.0 (95%CL = not determinable) for PFAD, 22.8 (95%CL = 6.2-83.0)
for CP, 22.6 (95%CL = 2.3-226.0) for FS, 15.7 (95%CL = 6.9-35.6) for PQ, 5.5 (95%CL
= 2.1-14.3) for RB and 4.2 (95%CL = 3.2-5.5) for GAC. It can be observed that the
confidence limits of these fitted values are sometimes widely spread. This is likely
due to the fact that the lowest dose of sorbent already leads to a significant
decrease of the dissolved concentration, which requires the extrapolation to obtain
the sorbent dose were 50% of the aqueous concentration was sorbed. This results in
a rather high uncertainty.

Contrary to the experiment without benzene, the order of the most to least
efficient sorbent changed but the range of the sorbent loading was similar to the
treatment without benzene (4.2 gram up to 61 gram of detectable PFAS per kg
sorbent at aqueous concentrations of roughly 7 µg/L for the sorbents and 25 µg/L
for the flocculant PFAD).

When we compare these results of the FFWW and the (benzene spiked) impacted
GW the following can be observed. The loading rates of the impacted GW fall
roughly in the same range as the levels observed for the FFWW, when we take into
account that a major part of the PFAS in the FFWW are missed by the detection
method. As we cannot detect all PFAS we cannot compare the loading rates directly.
Nevertheless observed loading rates are in the range of what is observed in
literature (see chapter 4). Additional analysis is required to better cover the
spectrum of PFAS present in (especially the) FFWW.

18
 report no. 5/21

Figure 9: Removal of individual PFAS from benzene spiked groundwater (GW) by five
selected sorbents and one flocculant, tested in batch sorption tests with 7
doses and a negative control in triplicate. The negative control is plotted as
0.0000001 mg/L sorbent in order to include the data in the graph, using a
logarithmic x-axis. The tested sorbents were Rembind® (RB), Granulated
Activated Carbon (GAC), PolyQA-Osorb® (PQ), Cyclopure (CP) and FLUORO-
SORB® (FS) and the flocculant was PerfluorAd (PFAD).

PFPeA in GW
PFBA in GW
120
120 RB
RB
GAC
GAC 100
100
PQ
PQ
CP
CP
80 FS
80 FS
PFAD
PFAD

% removal
60
% removal

60

40
40

20
20

0
0

-20
-20

-4

-3

-2

-1

2
-4

-3

-2

-1

log (g/L sorbent)

log (g/L sorbent)

PFHxA in GW PFHpA in GW
120 120
RB RB
GAC GAC
100 PQ 100 PQ
CP CP
80 FS 80 FS
PFAD PFAD
% removal

% removal

60 60

40 40

20 20

0 0

-20 -20
-4

-3

-2

-1

-4

-3

-2

-1

2
log (g/L sorbent) log (g/L sorbent)

PFHxS in GW PFOA in GW
120 120
RB RB
GAC GAC
100 PQ 100 PQ
CP CP
80 FS 80 FS
PFAD PFAD
% removal

% removal

60 60

40 40

20 20

0 0

-20 -20
-4

-3

-2

-1

-4

-3

-2

-1

log (g/L sorbent) log (g/L sorbent)

L-PFOS in GW Br-PFOS in GW
120 120
RB RB
GAC GAC
100 PQ 100 PQ
CP CP
80 FS 80 FS
PFAD PFAD
% removal

% removal

60 60

40 40

20 20

0 0

-20 -20
-4

-3

-2

-1

-4

-3

-2

-1

log (g/L sorbent) log (g/L sorbent)

19
 report no. 5/21

3.3.8. Discussion batch sorption and flocculant tests

When the removal of individual PFAS present in the impacted waters is studied, it
can be observed that the longer the CF chain the more effective all sorbents become
in removing the PFAS. This is not surprising as the length of (fluorinated) chains
affects the aquatic solubility as well as the hydrophobicity of molecules. Similar
trends are observed for both ionic and non-ionic surfactants that are not
halogenated 13-15.

There are differences observed between the sorbents and flocculant and between
the different treated waters. Generally the GAC shows the most robust total
removal. PFAD seems to be most effective at the lowest dosages. This is especially
true for the FFWW. FS performs similarly well in a situation with benzene spike.
The performance of especially the flocculant PFAD but also to a lesser extent the
bio-based or polymeric based sorbents PQ and CP are not able to remove all PFAS
in the different treatments. This is hardly visible in the FFWW as initial
concentrations were very high, but is more pronounced for PFAS impacted GW
without and PFAS impacted GW spiked with benzene. Here, do not seem to reach a
zero concentration at high dosages but tend to drop to 2-20% of the initial PFAS
concentration even at extremely high dosages of sorbent / flocculant. This is most
pronounced in the impacted GW tests with and without benzene spike. This
observation is difficult to explain from a thermodynamic perspective of sorption
and complexation processes 16-17. Theoretically, the addition of sorbent leads to
more available sorption sites that consequently leads to lower freely dissolved
concentrations of the PFAS and presumably higher sorption coefficients when
typical nonlinear Langmuir or Freundlich-like sorption isotherm are assumed.
Therefore, the theory suggests an artefact such as incomplete separation of the
flocculant or polymers and the aqueous phase at high dosages, leading to
overestimation of the aqueous concentration and thereby underestimating the
removal at higher sorbent or flocculant dosage. Interestingly, the addition of the
benzene resulted in better performance of these polymer and bio-based sorbents
and lower residual fractions in the sampled aqueous phase (Figure 6 and Figure 8)
while the flocculant PFAD performed poorer.

Optimizing the sorbent water separation in pilot or full size applications will likely
result in larger removed fractions that can be brought close to 100%. Therefore, we
consider these residual concentrations an artefact of the experimental set up for
some of the sorbents and flocculant under certain test conditions.

3.3.9. Results foam fractionation with ozone and air

The foam fractionation experiments were executed with air and with ozone. Three
types of water were used, being FFWW, impacted GW and impacted GW spiked with
20 mg/L benzene (GW-BS). The benzene is a surrogate for BTEX residues that can
coincide with the PFAS contamination in GW underneath petrochemical plants and
other contaminated sites. The results are presented in Figure 10.

20
 report no. 5/21

Figure 10: Composition and concentrations of PFAS the treated water was sampled at
the time that no foam was formed anymore (0h) and 1 hour (1h) of
operation for the treatments with ambient air (Air) and ozone (O3).
Firefighting wastewater = FFWW, groundwater = GW and benzene spiked
groundwater = GW-BS, the feed water (negative control) = NC and the
Concentrate (foam) = Conc. All treatments were performed in triplicate for
GW-O3-Conc. and GW-BS-O3-0h one of the triplicates is excluded because
that particular sample strongly deviated for all chemicals. * FFWW-air-1h is
missing because the foam formation was too abundant and only one sample
was collected during the test (FFWW-air-oh).

30000
PFPeA
PFBS
PFHxA
20000
PFHpA
PFHxS
PFOA
L-PFOS
10000
g/L

Br-PFOS

PFUnA
FOSA
4:2 FTS
50 6:2 FTS
40 8:2 FTS
30
20
10
0 *
W W- 0 h

F F ir - C 1 h

-O .
W W - 0h

3- 1h

c.

G W - 0h
ir - 1 h

-O c.

c.

- B S- 0 h

h
G S - n c.

-B S- 0h

3- 1h

c.

C
G W- 0h

3- 1h
W nc

A r-1
-N
-N

N
on
on

G on

on
FF FW ir-

-A ir-

FF F W 3 -

-O 3-

G ir -

-A ir-

G 3-

-O 3 -

-
W o

-B o

S-
G W- O 3
S- O3
G W- Air
W

S- A i
C

C
W O

W O

C
-A

W A
-A

W A

-B
W

G ir -
G S-
W
W

W
FF

O
W B
-B

W B
G
W

G
W
W
FF

F
F

Figure 11 represents a picture of the untreated (left two bottles) and treated (right
two bottles) water.

Figure 11: Firefighting training wastewater, untreated vs. treated colour.

21
 report no. 5/21

Table 4 lists the volume of the treated water and the collected concentrate. For
the ambient air treatment the reduction after 1h operation could not be
determined, while the ozone treatment lead to a factor 5 reduction for FFWW. For
GW a factor 2.5 and 4 reduction was observed for air and ozone treatment,
respectively, while BS treatment resulted in reduction of a factor 5 and 8,
respectively. Together with the collected volumes of foam versus the volumes of
treated water a concentration factor between the foam and the water can be
determined for the sum of PFAS.

Table 4: Experimental conditions, PFAS fraction (%) remained in the treated water
and the concentration factor of summed PFAS between foam and water for
air and Ozo fractionation of groundwater (GW) benzene spiked (BS) and
firefighting wastewater (FFWW) and additional experimental conditions.

Sample pH ORP Treated Collected Sum PFAS not Sum PFAS

(mV) volume (L) concentrate removed (1h Concentration
(L) treatment / factor
negative control [concentrate]
(%)) / [treated
water]

FFWW-air fractionation n.a. n.a.2 6-7 0.1001 n.a.2 -

2

FFWW-Ozo fractionation 2.8 1090 6-7 n.a.2 18% -

GW-air fractionation 8.2 235 7-8 0.150 41% 67-77

GW-Ozo fractionation 4.43 1083 9-10 0.075 24% 380-422

BS-GW-air fractionation - - 7-8 0.100 21% 263-301

BS-GW-Ozo fractionation 4.23 1075 12-13 0.075 12% 1173-1271

1
Calculated with 0 hour water sample as the 1h treatment was corrupted by extensive foam formation
2
Not applicable as no sample of the 1h treatment could be obtained
3
adjusted pH

3.3.10. Discussion foam fractionation with ozone and air

Figure 10 and Table 4 show the results of the air and ozo fractionation. The setup
of foam fractionation is a delicate process of adjusting air (or ozone) flow,
potentially adding chemicals to allow or reduce foam formation to be sufficiently
stable for separation but also not too stable to enable collection of the foam.
Currently, foam fractionation is mainly applied to impacted GW. Treating FFWW is
currently very far out of the normal operation (personal communication Evocra).

The composition of impacted GW is different from water generated from firefighting

activities (Table 2). The excessive foam formation illustrated that the initial
settings and conditions of the pilot test system were not suitable for the FFWW.
Subsequently, conditions required adaptation and iron chloride was added and the
tested water was diluted in case of the air treatment to allow the collapsing of the
formed foam for separation. Even these adaptations were not sufficient to fully
solve the overproduction of foam. This was reflected in the omission of the samples
and determination of the volume of the foam and water generated one hour after
a stable foam layer was formed since the test was aborted to prevent the system
from getting completely contaminated with the foam including filters and tubing.
This illustrates the necessity to adapt and optimize the settings and conditions
related to the treated matrix, which apparently require extensive studies when
highly contaminated matrices such as FFWW are treated.

22
 report no. 5/21

It can be observed that the ozone treatment appeared to be more effective than
the air treatment for the GW and benzene spiked GW. However, within the current
experiments the air and ozone flow differed due to the capacity of the ozone
generator. Therefore, these differences between ozone and air treatment can be
either attributed to the effect of ozone vs. air or simply the dose of these gasses
that affects the formation and the structure of the foam. These two effects cannot
be fully distinguished in the applied experimental setting, but give opportunities
for further fine-tuning of the treatment.

The addition of benzene improved the removal of PFAS from impacted GW for both
the air and ozone treatment. Potentially the solvent affected the foam formation
or structure in such a way that PFAS removal was improved.

When the ozone treatment of FFWW was compared to GW with and without benzene
the residual PFAS in the treated water rate appeared similar. However the unbiased
comparison of the impacted GW and FFWW is hampered by multiple factors. First,
the comparison of the removal is biased by the composition of PFAS that differs
considerably between the impacted GW and the FFWW. The FFWW is dominated by
6:2 FTS while the impacted GW holds a wider spectrum of PFAS containing also more
polar (shorter carbon chain) constituents. This argues for the comparison of PFAS
removal for individual PFAS. However, the aeration and especially the ozone
treatment might induce the transformation of less persistent PFAS into more
persistent transformation product, as air and especially ozone will enhance
oxidation. This affects the composition of PFAS in the system during the experiment
and corrupts the analysis of removal of individual PFAS by the fractionation process.
A close look at individual PFAS in the ozone treated FFWW illustrates that this is
actually happening. PFPeA concentrations increased almost two-fold after
treatment (1h sample) and PFHxA, PFHpA were hardly removed during treatment
(between feed water and 1h sample) and appeared to increase 2 and 4 times
between the 0h and 1h sample.

So while quantitative interpretation of the results of both the summed PFAS levels
and individual PFAS levels for the FFWW is complicated, we can safely assume that
the comparatively mild oxidation by ozo fractionation is not leading to relevant
change in the summed PFAS in GW. Even the more aggressive TOP assay did not
result in a significant change in total PFAS for impacted GW (Figure 3).

Despite of suboptimal foam fractionation conditions and potential biases of the

oxidation of PFAS into other PFAS during treatment, the experiments show the
potential of foam fractionation as a (pretreatment) technology. The treatment
effectively lead to the concentration of PFAS in a small volume of foam and a
significant reduction of PFAS in the treated water. With additional testing and
optimizations applied, it is likely possible to improve the separation and removal of
the impacted GW further and to treat the FFWW in stable operation. This potentially
requires the following experimental steps:

 Optimizing gas inlet to minimize bubble formation at the start;

 Optimizing FeCl3 dosing;
 Change pump settings.

23
 report no. 5/21

3.3.11. Results nanofiltration tests

Membrane separation experiments

In the membrane filtration experiments, the pilot installation MexPlorer from
NXFiltration was used. Three different experiments were performed with this
installation using FFWW: transmembrane pressure, flow velocity and recovery tests
as described in Appendix 5.

In the tests according protocol 1 of the Standard Test Protocol Nanofiltration

Mexplorer (NXF-TR: 17-45-1C) of the technology provider the membrane was tested
at feed pressures of 3, 4 and 5 bar. At increasing pressure the water flux through
the membrane is increasing while the PFAS passage through the membrane is nearly
equal. This results in lower PFAS concentrations in the permeate at higher feed
pressures.

In the tests according to protocol 2 the membrane was tested at crossflow capacities
of 70, 100 and 150 L/h. When retaining a solute in membrane filtration, the
concentration of that solute significantly increases at the water/membrane
separation layer interface. This is called concentration polarization, a phenomenon
that happens almost immediately. High levels of concentration polarization will
reduce the overall rejection of the solute. The magnitude of this concentration
polarization can be reduced by actively transporting the higher solute
concentrations away from the water/membrane interface. In hollow fibre filtration
this is done by increasing the (tangential) crossflow velocity through the fibre.
Figure 12 illustrates the obtained results for the different treatments.

Figure 12: Composition and concentrations of PFAS in initial feed water

(NC) and treated water permeate obtained from protocol 1 and
protocol 2 (NXF-TR: 17-45-1C). Blank represents the results
from milliQ water. The treated water (permeate) was sampled
after 15 min of operation at a trans-membrane pressure (TMP)
of 3, 4, and 5 bar or a cross-flow velocity (CF) of 70, 100 and
150 l/h without oxidation.

15000
PFPeA
PFBS
10000
PFHxA
PFHpA
PFHxS
5000
PFOA
L-PFOS
g/L

Br-PFOS
1000
PFUnA
800 FOSA
95% removal 4:2 FTS
600 6:2 FTS
8:2 FTS
400

200

0
)

)
l/ h
r

l/ h

l/ h

(t 0
ba

ba

ba

(M
70

0
3

C
10

15

N
nk
P

F
TM

TM

TM

CF

F
C

la
C

24
 report no. 5/21

In addition an experiment was performed to obtain the removal efficiency for a

preset recovery. As the permeate flow dropped during operation, reaching the
recovered permeate volume required over 3 days. Figure 13 illustrates the results.
It can be observed that the long term operation leads to increasing amounts of PFAS
in the treated water (permeate) slowly reaching towards the level of the feed water
concentrations. This illustrates that the membrane is – with longer operation – not
capable of separating the PFAS from the aqueous phase.

Figure 13: Composition and concentrations of PFAS in initial feed water

(NC), the concentrate and treated water permeate obtained
from firefighting wastewater (FFWW). Blank represents the
results from miliQ water. The treated water (permeate) was
sampled after a series of run times up to 74h. The bars
indicated with two run times (10-29 and 34-51) are composite
samples of overnight collection of the permeate between the
two run times.

20000
PFPeA
PFBS
PFHxA
PFHpA
15000
PFHxS
PFOA
L-PFOS
g/L

Br-PFOS
10000
PFUnA
FOSA
4:2 FTS
6:2 FTS
5000
8:2 FTS

90% removal
95% removal
0
on NC )
nt )
h
h
10 0 h

h
h
34 4 h

h
h
h
h

te
Q
ce (t0
25

29

51
57
74
5

(M

ra
1
-2

3
-5
0.

nk
la
B

3.3.12. Discussion nanofiltration tests

The nanofiltration experiments were applied under different pressures and

transmembrane permeate flows (Figure 12). Furthermore the set up was operated
for 74 hr. to obtain a preset permeate recovery (Figure 13). With longer operation
of the membrane filtration the permeate flow dropped. This was likely due to the
fact that the FFWW matrix contained particles that blocked the pores of the
membrane hampering the water to pass the membrane. The nanofiltration
experiments show that the applied membrane is able to remove the PFAS effectively
during the first hours of operation (0.25 hr.-10 hr.), reaching removal rates of 90-
>95% (the membranes in Figure 12 were all operated for 0.25 hr.). However, when
the membrane was operated for a longer period trans membrane flows dropped and
the passage of PFAS slowly increase to levels approaching concentrations in the
feed-water (low removal). In addition, upon inspection, a penetrant “gasoline
smell” could be recognized in the filtrate after longer operation, illustrating the
permeability of the membrane for petrochemical residues as well. The
transmembrane pressure increased sharply with longer operation.

25
 report no. 5/21

3.3.13. Results small-scale column tests

The small-scale column pilot studies were designed to obtain in breakthrough curves
of the different PFAS present PFAS affected GW. These corves explores the
individual PFAS concentrations is as the fraction of the concentration in the feed
water. 20% breakthrough were defined and expressed in the bed volumes the
column has been exposed to. This is indicative for the treatment capacity of the
sorption material in the column for the different PFAS present in the water. More
than 40,000 bed volumes were passed. Based on the results of the batch sorption
experiments FLUORO-SORB® (FS) Cyclopure D+ (CP) and Desotec Granulated
Activated Carbon (GAC) were selected for the column studies because of their
observed performances in batch tests and the applicability of the material in
columns.

In Figure 14 the results of the small-scale column tests are shown, it has to be
noted that only results are shown of the compounds that showed relatively good
interpretable results: not all compounds were present in the influent water or
sometimes only at a concentration near the detection limit, which resulted in poorly
interpretable breakthrough curves.

Figure 14: Breakthrough curves of respectively PFBA, PFPeA, PFHxA,

PFHpA, PFOA, PFBS, PFHxS and PFOS tested obtained with the
small-scale column test.

26
report no. 5/21

27
 report no. 5/21

28
 report no. 5/21

In Table 5, estimations of breakthrough moments at 20 % breakthrough are given in

passed BV, to be able to compare the used sorbents. It has to be noted that the
choice for 20% breakthrough is debatable, but was selected as native PFAS
concentrations did not always allow a lower threshold due to limits of
quantification.

Table 5: Estimated breakthrough moments at 20 % breakthrough of the measured

compounds in passed Bed Volumes (BV), obtained with the small-scale
column test.

Amount of passed BV at 20% breakthrough

Influent
Compound concentration Remarks/observations
(µg/L) FLUORO-
Desotec GAC Cyclopure D+
SORB®

Desotec GAC shows

PFBA 1.17 2,000 7,000 Immediate desorption profile after
7,000 BV

Desotec GAC shows

PFPeA 2.17 10,000 13,000 1,000 desorption profile after
17,000 BV

PFHxA 4.98 > 40,000 19,000 7,000

All show direct 10-20 %

PFHpA 1.03 29,000 22,000 12,000
breakthrough

PFOA 1.5 > 40,000 40,000 > 40,000

PFBS 1.61 > 40,000 25,000 > 40,000

PFHxS 11.3 > 40,000 > 40,000 > 40,000

PFOS 35.5 > 40,000 > 40,000 > 40,000

With the small-scale column tests more than 40,000 BV were passed and the applied
method made it possible to compare the three chosen sorbents applied in a column.
It has to be noted however that the obtained breakthrough curves cannot be directly
converted to estimations of breakthrough of compounds in larger practice situations
using unground materials.

3.3.14. Discussion of small–scale column tests

All tested materials were capable in removing the longer chained PFAS molecules,
no break though was observed after > 40,000 bed volumes. No difference between
the sorbents was detected for the long chain PFAS molecules. They appear to be
removed quite efficiently. The activated carbon based material (Desotec GAC)
seemed to be a bit better capable of removing the shorter chained PFBA and PFPeA
in comparison with FLUORO-SORB® and Cyclopure D+. However, Desotec GAC
showed desorption of the shorter chained compounds PFBA and PFPeA after
breakthrough started. This might be caused by better sorbing compounds that
displace these shorter chained PFAS compounds. This indicates that these

29
 report no. 5/21

compounds are less strongly bound to the GAC material and can even be fully
desorbed after a longer time of operation. The other materials did not show this
phenomenon. For PFHxS Desotec GAC showed an earlier breakthrough compared
with FLUORO-SORB® and Cyclopure D+. However the concentration remained below
20% of the initial concentration at up to 43,000 bed volumes that have been tested.

30
 report no. 5/21

4. PART 3: TECHNOLOGY APPLICATION

This chapter focusses on application of the technologies, thus giving an overview

of the known operating ranges and constraints of the technologies, along with
identification of key design criteria for each treatment technology.

The basis for the chapter is the factual information from the previous report1, along
with the experimental results as described in previous chapters and additional
information from suppliers. Unless otherwise specified, information is derived from
this study.

Based on the information obtained from the technology-suppliers and the

experiences gained during the experimental phase, this overview of the
technologies provides input for the Concawe-members to evaluate the state of the
art and the most recent operational knowledge of the technologies. This chapter
gives insight into the potential cost-effectiveness of PFAS removal: high chemical-
or energy costs related to very limited PFAS removal could be, for example,
undesirable. The chapter finally identifies where knowledge and information is still
lacking regarding practical applications and costs.

The findings in the report are condensed in the form of a look-up table to help
Concawe members select the PFAS treatment technology conditions appropriate to
their site-specific characteristics. This table allow the Concawe-members to quickly
see the relevant technologies for specific treatment challenges and their available
information, the main text body gives additional (background) information.

This chapter briefly describes the basics of each technologies, followed by the
summary of the PFAS-removal performance based on Riegel, et al. 1 and the
experimental results. Furthermore, key design criteria are discussed that form the
basis for operational aspects and finally the costs for each technology. In order of
appearance, the following technologies are described:

 Activated Carbon;
 Rembind®;
 DEXSORB+®;
 PolyQA-Osorb®;
 FLUORO-SORB®;
 PerfluorAd;
 Nanofiltration;
 Foam- and ozo fractionation.

Disclaimer for further reading

It has to be stated that a comparison of the technologies for the purpose of site
specific application is far more complex than the intrinsic technology comparison
provided in this study. Several factors are at play when properly comparing
technologies and their applicability at specific locations and for specific purposes
under specific conditions in depth. Mainly very local conditions play a key role in
technology selection. The main influences are:

31
 report no. 5/21

Location characteristics
A GW contamination with PFAS will have completely different characteristics that
are related to the hydrology, geochemistry of the soil it is in, and the related
aqueous chemistry of the GW (salt content, pH, DOC etc.). For example, GW
contamination in peat or silty soils in deltas such as in the Rotterdam or Antwerp
Harbour areas will affect the contamination plume differently than upstream
locations along a river with higher flow velocity often coinciding with gravel / sandy
soils with extremely high permeability and lower sorption capacity such as is found
upstream in the Rhine and Meuse valley.

It can be expected that plume size, mobility and the rate of GW extraction in case
of a GW remediation will be far less in a delta such as the Rotterdam area than
upstream along the river Rhine. In addition, the location characteristics such as the
positioning and dimensions of a location determine the physical space available for
treatment as well as logistics and opportunities in the supply of materials used for
treatment and opportunities to remove or reuse waste. Finally, legal requirements
will also differ between locations depending on function (e.g. nature area,
farmland, industrial area, GW protection zone, or residential area’s) and between
countries.

The selection of treatment technology will be dependent on all of these

parameters:

In some cases, the available area on-site is limiting the footprint of a technology.
This gives the preference to short EBCT sorption filters, since these have an inherent
smaller footprint. For example in case the extraction rate needs to be very high, a
sorbent with a short Empty Bed Contact Time is preferable over the use of a sorbent
with a long EBCT. At extraction rates of approximately 100 m3/hr or more, the
volume of sorbent for the water treatment will become significant in case of long
EBCT’s.

In case of gravelly soils the use of recirculation might be a viable and very
economical option, particularly when the stand time of the sorbent could be
prolonged by it.

(Geo)chemical conditions of the water

The (geo)chemical composition of the water is another important factor. As we have
seen in the study, the presence of other organics and inorganics affect the
efficiency of the processes for all presented technologies. It is vital that these
effects are assessed for each case individually.

What we know from literature, this study and other projects is that some of the
sorbents are less affected by increased levels or organic or inorganic constituents1.
So dependent on the composition of the treated water, a selection for the sorbent
(or the necessary pretreatment) might differ.

Local regulations
The legal requirement for discharge of water or processing of waste could severely
influence the selection process for the treatment technology. The permitted
discharge concentrations (if defined in local jurisdiction) differ for PFAS from site
to site. This difference will affect the decision making process on whether a
treatment facility is needed or on how the actual treatment will take place.

32
 report no. 5/21

Equilibrium processes with regards to sorption

It is known that sorption is an equilibrium process, which means that the greater
the incoming concentration, the greater the load on sorbent, e.g. activated carbon
sorbs until a maximum loading is reached and the sorbent no longer is able to
remove the compound of interest. When the treated water contains a mixture of
contaminants, the affinity of the individual compounds will result in competition
where compounds with a stronger affinity will replace compounds with lower
affinity. Depending regulatory constraints for individual compounds, targets set by
the user and requirements for potential reuse, the application of the technique is
determined. This is typical for PFOS, a molecule with a significant sorption affinity
on activated carbon. A sorption isotherm can be found in Appendix 7. The steeper
the curve, the greater the sorption load. The greater the incoming concentration,
the greater the contaminant loading rate on activated carbon. This also shows that
the red curve (ORGANOSORB 10-AA) is much steeper than the green curve
(ORGANOSORB 10) and thus gives the macroporous activated carbon the greatest
load. Which we thus also observe in practice.

4.1. ACTIVATED CARBON

4.1.1. Introduction

Granular activated carbon, or activated carbon as a whole, is regarded as one of

the benchmark technologies for (ground)water remediation. The technology has
been widely applied many times from pilot to full-scale installations and has proven
its effectiveness. Activated carbon is marketed by a large number of specialized
vendors. Carbons specialized to remove PFAS are offered by various vendors such
as DESOTEC, CABOT-NORIT and Chemviron, and applied for GW remediation.

In general granular activated carbon sorbs all non-water soluble organic substances
present in water streams.

There are numerous types of granular activated carbon that differ in effectiveness
and in pricing, Appendix 7. The price difference between a high quality GAC and a
low quality GAC ranges a factor 5 to 10. Price is mainly dependent on the sorption
capacity of the activated carbon, which is dependent on the surface area, pore
volume and pore size distribution.

The removal mechanism is sorption in the pores of the activated carbon, based on
hydrophobic interactions between the (hydrophobic) fluorinated chain of the PFAS
and the activated carbon surface. Figure 15 below shows a flow diagram of a GAC
system including potential pre-treatment and disposal routes of the spent material1.

The working range for GAC for the treatment of impacted water is broad. Projects
in e.g. the drinking water industry have shown the effectiveness of the sorbent even
at very low concentrations (< 100 ng/L).

33
 report no. 5/21

Figure 15: PFAS flow diagram for continuous adsorption filtration with GAC. Note that
the values are indicative based on Riegel, et al. 1; removal efficiency and
target PFAS concentrations strongly depend on PFAS components to be
removed and site-specific conditions (e.g. influent concentration, presence
of competing substances, etc.).

• INCINERATION
INFLUENT or
100% PFAS
• THERMAL
reuse possible
REACTIVATION
< 10% mass loss
PRETREATMENT
Fe, Mn
(if required) GRANULAR SPENT ADSORBENT
sand filtration ACTIVATED up to 99% PFAS *
slightly PFAS removal CARBON

SLUDGE
DISPOSAL TREATED WATER
[approx. 0.1%
of influent vol.] 1% PFAS *
fixed-bed filter

4.1.2. Performance

Generally, the sulfonic acids (PFSAs, such as PFOS), are more effectively removed
by activated carbon than the carboxylic acids (PFCAs such as PFOA). The main
reason is the higher electrostatic interaction PFSA-compounds have with the carbon
surface due to the three oxygen atoms comparing to the two oxygen atoms at
PFCAs 1. Similarly, long-chain PFAS are more readily removed than short-chained
species due to greater hydrophobic interactions and increased confinement in the
pores. Actual removal efficiency thus differs based on matrix and specific PFAS.
Removal rates of >90% up to close to 100% are observed for some PFAS, whilst for
some short-chained PFCA’s removal of <50% is not uncommon in practical field
experiences with field conditions.

As discussed in Paragraph 3.3 batch experiments of the tested GAC shows higher
affinity for longer chain PFAS than for shorter chain PFAS. Below the sorption
coefficients (Ksw) that were derived from these data by means of fitting Freundlich
isotherms (Appendix 6), are listed in Table 6 below. Within this table the sorption
coefficients that are listed are the coefficients at an aqueous concentration of 1.0
µg/L. This normalisation at a concentration of 1.0 µg/L allows to compare
compounds independent of the native concentrations. The slope of the isotherm
indicates the linearity of the sorption with concentration. The sorption coefficient
is independent of concentration when 1/n is (near) 1 and becomes lower with
increasing concentrations when it is <1, values above 1 are not expected. The
sorption of the sulfonic acids appears to be stronger than that of the carboxylic
acids when compounds with the same fluorinated carbon chain length are
compared. These observations coincide with observations in literature summarized
in Riegel, et al. 1.

34
 report no. 5/21

Table 6: Freundlich sorption characteristics of different PFAS compounds on

activated carbon in impacted groundwater (GW).

Carboxylic acids Sulfonic acids

Length of 4 5 6 7 8 4 6 8 6
fluorinated
chain
solute PFBA PFPeA PFHxA PFHp PFOA PFBS PFHxS PFOS 6:2
A FTS

GAC KSW1 21038 51523 1013911 57016 430527 n.a.4 n.a.4 2500345 n.a4
(95% (13868 (32885 to (177011 (38726 (248886 to (2426610
Confidence to 80910) to to 744732) to
Limits) 31842) 5807644) 84140) 2576321)

GAC slope6 0.54 0.50 1.12 0.27 0.85 -0.06

(95% (0.38 to (0.38 to (0.69 to (0.17 (0.67 to (n.a.4)
Confidence 0.70) 0.61) 1.55) to 1.03)
Limits) 0.37)
1
Sorption coefficient ([PFAS] µg/kg sorbent / [PFAS] µg/L water) defined (extrapolated) at an aqueous concentration of
1 µg/L based on sorption isotherms listed in the Appendix 6). In most cases this concentration is at the high end of just
outside the concentration range tested, so it can be considered a worst case sorption coefficient, as sorption coefficients
generally increased with decreasing concentrations.
2
Number of bed volumes where a break-through of 20% of the initial concentration was observed.
4
Not applicable; data did not reveal clear sorption isotherms.
6
The slope of the sorption isotherm (log concentration on sorbent (µg/kg) vs. log concentration in aqueous phase (µg/L)

As discussed in paragraph 3.3.8 sorptive treatment is able to remove PFAS from both
the relatively clean matrix of impacted GW and the more heavily impacted matrices
such as FFWW. But the application to heavily contaminated matrices with a wide
variety of co-contaminants such as oil residues is not ideal as concentrations are
very high leading to relevant residues in the effluent. This argues that treatment of
FFWW requires pre-treatment in order to reduce the sorbent loading and thereby
allowing longer sorbent life higher bed volumes and lower concentrations PFAS in
the effluent.

4.1.3. Key Design criteria

Treatment strategy/treatment concept

Most commonly, a granulated activated carbon filter is used in a fixed-bed set up.
The impacted water will flow downward through the filter where sorption takes
place. Contact time can be easily managed this way. GAC-filters in a serial
configuration using lag/lead filters (pseudo moving bed) is very common: the first
filter is removing the bulk of the PFAS, the second is used for polishing. When the
first reaches breakthrough, the second filter becomes first in line, whilst the first
is replaced. This newly replaced filter starts out as polishing filter and thus the
cycle continues. This method reduces the overall activated carbon consumption.

Pre-treatment
There are a few reasons to pre-treat water before treatment with a granular
activated carbon filter:

Activated carbon possesses affinity for most hydrophobic organic compounds (e.g.
BTEX, aromatic hydrocarbons, oil/petroleum). These substances are in (direct)
competition with PFAS for sorption sites and thus have negative effects on the
lifetime of the filter bed, so depending on the concentration of these compounds,
pre-treatment might be cost effective. There are no general rule-of-thumb values

35
 report no. 5/21

available from which concentrations upward this becomes effective. Additionally,

low concentrations of multivalent ions, like iron and manganese, can cause filter
blocking/scaling. These compounds generally don’t compete with PFAS for sorption
sites, however they can pose a challenge for continued, reliable operation.
Similarly, particulate matter can cause filter blocking and pressure loss during
filtration resulting in frequent filter rinsing. Sand filtration is a common option to
remove suspended solids and a fraction of organic matter. For suspended particles,
a turbidity of 1.5 FTU/NTU/FAU or higher is a rule-of-thumb threshold for applying
pre-treatment.

Contact time
The main design parameter for an activated carbon filter is the Empty Bed Contact
Time (EBCT [min]), it defines the contact time of the water with the complete filter
that includes voids. Typically, this varies between 20 and 30 minutes of EBCT,
indiscriminate of the flow, but could be up to 60 minutes. A general rule-of-thumb
is 20 minutes. EBCT’s practically translate to total required filter volume: a flow of
150 m3/h with an EBCT of 20 minutes results in 50 m3 GAC* (wet and packed
volume), whilst an EBCT of 1 hour requires 150 m3. Generally, both contaminant
loading rate and thus the bed lifetime increase with increasing the EBCT.

Hydraulic loading rate/filter velocity

The filter velocity, or hydraulic loading rate, is expressed as the ‘speed’ of the
water through a filter. Typical filter hydraulic loading rates during operation vary
between 5 and 15 m3 per m2 filter surface per hour, expressed as filter velocity (vF,
[m/h]). A general rule-of-thumb is 10 m/h. According to literature18, taller
activated carbon filters might increase overall removal efficiency. The loading rate
is important because it influences the mass transfer from the bulk solution to the
carbon granules. A too low loading rate increased mass transfer resistance. The
filter velocity should be designed in tandem with the EBCT and thus the bed volume
to ensure enough contact time for sorption.

Contaminant loading rate

The contaminant loading rate is the amount of contaminant a certain amount of
carbon can sorb [mgPFAS/kgCarbon]. Through sorption isotherm experiments, the
loading capacity of a chosen type of GAC can be determined. Typical loading rates
range between 10 to 60 mg PFAS/kg GAC. Generally, both contaminant loading rate
and thus the bed lifetime increase with increasing the EBCT. These values vary
strongly depending on the presence of compounds competing for sorption in the
water matrix and the characteristics of the chosen GAC, so a generic rule-of-thumb
value is not given.

Within this study contaminant loading rates were orders of magnitude higher with
4.2 (95% Confidence Level = 3.2-5.5) and 1.5 (95%CL = 1.4-1.6) g PFAS/kg sorbent
for the GW without and with benzene spike respectively, and 3.8 (95%CL = 3.2-4.4)
for FFWW. However, these loading rates were obtained at rather high aqueous
concentrations of the detectable PFAS (~7 µg/L, for GW and around 5.000 µg/L for
FFWW).

The observed loading rates are not representative for loading rates that lead to high
removal percentages or acceptable effluent concentrations (sub µg/L-range) and
low sorbent dosages. For ~99% reduction of PFAS concentrations from GW probably
typically one to two orders of magnitude higher sorbent dosages are required, which
automatically results in lower dosing rates. Such loading rates fall in the same range
as observed in literature.

36
 report no. 5/21

Bed lifetime
A sorptive filter needs replacement when the target compound(s) are breaking
through. Normally this means that the treated water concentration is approaching
the treatment target concentration. Continuing filtration would mean that the
target concentration would be exceeded soon. The volume of water a filter can
treat before reaching this target concentration is called the bed lifetime. It is
expressed as the amount of Bed Volumes being the number of reactor volumes the
filter has treated (abbreviated to BV). Values range from several thousands to
several tens of thousands BV till breakthrough begins. These values vary strongly
depending on the properties of the PFAS, presence of compounds competing for
sorption in the water matrix, the desired removal rate and the characteristics of
the chosen GAC. A generic rule-of-thumb value cannot be given. Within the small
scale column studies, as shown in the table below, GAC showed > 40.000 bed
volumes without breakthrough for PFHxA, PFOA, PFHxS and PFOS, while PFBA,
PFPeA, PFHpA, and PFBS showed (20%) breakthrough at lower bed volumes
(Table 7). This correlates with the obtained sorption coefficients from the PFAS also
listed in the table. And with the generic trend that the shorter the fluorinated chain
is, the lower the affinity of the sorbent is.

In order to make best use of the potential of the sorbent to sorb contaminant, it is
best to make use of a lead and a lag vessel. The stand time of the lead vessel can
be prolonged because of the of a second security (lag) vessel.

Table 7: Bed volumes treated at 20% breakthrough in column testing with impacted
groundwater (GW).

Carboxylic acids Sulfonic acids

Length of 4 5 6 7 8 4 6 8 6
fluor chain
solute PFBA PFPeA PFHxA PFHpA PFOA PFBS PFHxS PFOS 6:2 FTS

GAC (20%)2 7.000 13.000 19.000 22.000 40.000 25.000 >40.000 >40.000 n.a.5
2
Number of bed volumes where a break-through of 20% of the initial concentration was observed.
5
Not applicable; concentrations were too low to assess sufficient reduction of 6:2 FTS in the treated water

Effect of temperature and pressure

Activated carbon filters can be constructed as gravity or pressurized filter units,
depending on the requirements of the client. Temperature has some effect, but
these effects are minimal compared to matrix effects and changes in contact time
of the carbon bed.

4.1.4. Operational aspects

Energy use
The dominant primary energy use in an activated carbon filter are the feed pumps.
Dependent on influent characteristics and treatment strategy, backwashing pumps
might be applied as well. Energy use is directly linked to the treated water flow
(m3/h), the hydraulic head of the filter and the filter configuration and should be
detailed on a case-by-case basis. Depending on the applied method for energy
calculation, one might take transportation energy as well as the energy requirement
for the off-site regeneration or disposal of the carbon into consideration. As a rule
of thumb, for large-scale activated carbon filters an energy use of maximum 0.1
kWh/m3 treated can be assumed based on van Nieuwenhuijzen, et al. 19. This
includes both feed- and backwash pumps with a hydraulic head of 6-8 m, and
excludes energy related to transport and regeneration/disposal.

37
 report no. 5/21

Maintenance and personnel

An activated carbon filter typically runs without intensive operation- and
maintenance efforts and can be fully automated. The average time spent should
typically not exceed 0.5 day per week assuming a plant of up to 25 m3/h, depending
on the exact filtration set up and water composition.

Chemical use
GAC-filtration does not apply any chemicals, besides the activated carbon itself.
The need to buy new- or regenerated carbon is discussed above in the sections on
bed lifetime and contaminant loading rate. Since the lifetime of a filter bed varies
widely depending on several factors, there is no rule-of-thumb value for GAC-use.

Waste production
The main waste produced from an activated carbon filter is the spent carbon.
Activated Carbon vendors typically offer management options for their spent
sorbent. The price of processing the spent carbon is generally included in the costs
for the carbon. Spent carbon can be either (thermally) reactivated or incinerated
off-site. Several providers are able to regenerate the activated carbon. In this
process the activated carbon is heated up to 900 degrees C, where a significant
portion of the PFAS become volatile, thus removing (practically) all PFAS
constituents from the GAC. The off-gas from this process needs to be treated further
to ensure complete removal. Regeneration is restricted to certain types of GAC and
to a certain level of contamination. The regenerated activated carbon is cheaper
to use. Its performance could be similar to virgin GAC. As with chemical use, waste
production is directly linked to several factors and thus a rule-of-thumb of waste
production is not available. A relevant part to mention is the CO2-footprint/emission
of this process. By (re)activating/regenerating the carbon the energy intensive
production of virgin activated carbon is avoided, resulting in a large CO2-footprint
reduction for reactivated carbon.

4.1.5. Costs

Capital costs
Capital costs are relatively low as these systems are relatively simple. Most vessels
operate under atmospheric conditions and the technology is mature. Depending on
the matrix some more extensive pre-treatment can be required, increasing the
capital costs. If, for example, a sand filter is required, the capital costs can more
or less double, so a general rule-of-thumb value for CAPEX is not available.

Operational costs
Operational costs are based on the abovementioned operational aspects. A major
factor is the costs of the carbon itself, often the costs for management of the spent
carbon is included in this price. Carbon is available in a wide price range, for PFAS-
specific carbons, a typical value of around EUR 3/kg is to be expected for high
quality virgin activated carbon, and around EUR 2/kg for high quality reactivated
carbon 1. Depending on water matrix, desired treatment efficiency etcetera, costs
are estimated in the range of EUR 0.04 to EUR 0.20 per m3 treated. The lower range
is based on long chain PFAS-removal with higher bed lifetimes, whereas the higher
range corresponds to short-chain PFAS removal with shorter bed lifetimes 1. Besides
these price estimates, there is no more accurate translation available to a cost of
EUR/m3 treated.

38
 report no. 5/21

4.2. REMBIND®

4.2.1. Introduction

Rembind® is a (proprietary) sorbent, based on a mixture of amongst others

amorphous aluminium hydroxide, activated carbon, organic matter and kaolinite.
The product is marketed by the RemBind Pty Ltd Company. Rembind® is originally
designed for soil treatment (immobilization), but a variation of the product can be
used for water treatment.

The removal relies on a combination of three mechanisms: electrostatic interaction

due to the surface charge of the mineral component, the binding of organic matter
to anionic functional groups of PFAS and the hydrophobic interactions from the
activated carbon component of Rembind®.

Figure 16: PFAS flow diagram for continuous sorption filtration with Rembind® sorbent
in a mixed fixed bed. Note that the values are indicative based on Riegel, et
al. 1; removal efficiency and target PFAS concentrations strongly depend on
PFAS components to be removed and site-specific conditions (e.g. influent
concentration, presence of competing substances, etc.).

• INCINERATION
INFLUENT or
PRETREATMENT 100% PFAS
(if required)
• LANDFILL
Fe, Mn DISPOSAL
sand filtration
slightly PFAS removal

SLUDGE
DISPOSAL SPENT ADSORBENT
[approx. 0.1% REMBIND up to 99% PFAS *
of influent vol.]

TREATED WATER
1% PFAS *
fixed-bed filter

4.2.2. Performance

The reported performance in Riegel, et al. 1 is mainly based on laboratory-, bench-

and pilot scale experiments. These experiments indicate removal of both short- and
long chain PFAS. High removal has been reported (>99%), however, this was based
on short contact times and Riegel, et al. 1expressed doubt about the validity of the
tests. Rembind® was only tested in Batch experiments in the current study, see
Table 8. As discussed in Paragraph 3.3 batch experiments of Rembind® showed
higher affinity for longer chain PFAS than for shorter chain PFAS. Below the sorption
coefficients that were derived from these data by means of fitting Freundlich
isotherms (Appendix 10). Within this table the sorption coefficients that are listed
are the coefficients at an aqueous concentration of 1.0 µg/L. This normalisation at
a concentration of 1.0 µg/L allows to compare compounds independent of the native
concentrations. The slope of the isotherm indicates the linearity of the sorption
with concentration. The sorption coefficient is independent of concentration when
1/n is (near) 1 and becomes lower with increasing concentrations when it is <1,
values above 1 are not expected. Especially for the sulfonic acids no sorption
coefficients could be accurately derived from the batch sorption tests.

39
 report no. 5/21

Table 8: Sorption characteristics of the sorbent in batch tests with impacted

groundwater (GW)

Carboxylic acids Sulfonic acids

Length of 4 5 6 7 8 4 6 8 6
fluor chain
sorbent PFBA PFPeA PFHxA PFHpA PFOA PFBS PFHxS PFOS 6:2
FTS

RB KSW1 13646 7447 33420 27990 63680 n.a.4 n.a.4 493174 n.a.4
(95% (7278 to (4267 to (25468 (3945 to (11858 (295801
Confidence 25586) 12972) to 198609) to to
Limits) 43853) 342768) 820352)

RB 1/n 0.83 0.17 0.14 0.30 0.65 0.22

((95% (0.42 to (0.02 to (0.01 (-0.17 to (0.15 to (-0.07 to
Confidence 1.24) 0.32) to 0.77) 1.14) 0.50)
Limits) 0.26)
1
Sorption coefficient ([PFAS] µg/kg sorbent / [PFAS] µg/L water) defined (extrapolated) at an aqueous concentration of
1 µg/L based on sorption isotherms listed in the Appendix 6). In most cases this concentration is at the high end of just
outside the concentration range tested, so it can be considered a worst case sorption coefficient, as sorption coefficients
generally increased with decreasing concentrations.
2
Number of bed volumes where a break-through of 20% of the initial concentration was observed.
4
Not applicable; data did not reveal clear sorption isotherms.
6
The slope of the sorption isotherm (log concentration on sorbent (µg/kg) vs. log concentration in aqueous phase (µg/L)

4.2.3. Key Design criteria

Treatment strategy/treatment concept

The Rembind® sorbent has two different treatment options: either a discontinuous
batch reactor followed by sedimentation of sorbent and decantation of treated
water or continuously in a mixed fixed-bed filter, applying sand to reduce pressure
loss. Below, both systems are described where applicable.

Pre-treatment
Rembind® exhibits similar sensitivities to organic compounds as activated carbon,
so similar pre-treatment options are available. There is currently no exact organics-
threshold value available. A total petroleum hydrocarbon content (TPH) of <15 mg/L
is considered acceptable, if the total petroleum exceeds this, additional treatment
is required. A recent study by Kabiri and McLaughlin 2 showed that the sorption of
PFAS by RemBind® is not influenced significantly by humic acids, competing ions or
pH values at levels expected to be found in the environment.

Contact time
The supplier indicates that recent experiments have shown that a contact time of
5-10 minutes is sufficient to sorb >80% of PFAS. Hydraulic loading rate/filter
velocity. For both the discontinuous- as well as the continuous system, there is no
information available regarding hydraulic loading rate/filter velocity.

Contaminant loading rate

Within this study loading rates could be derived at aqueous concentrations that
were half of the concentrations of the impacted GW (~7 µg/L) and FFWW (~5.000
µg/L). The contaminant loading rates where 1.5 (95%CL = 1.4-1.6) and 5.5 (95%CL
= 2.1-14.3) g PFAS/kg sorbent for the GW without and with benzene spike,
respectively, and 1.7 (95%CL = 1.2-2.5) for FFWW. These loading rates were
obtained at rather high aqueous concentrations of ~7 µg/L for impacted GW and

40
 report no. 5/21

5000 µg/L for FFWW. Interestingly, the loading rate of the benzene spiked GW
appeared to be higher than the one without benzene spike. Comparison of loading
rates of impacted GW and FFWW is not directly possible as the PFAS composition
strongly differs (Figure 2) and the fraction of undetected PFAS in especially FFWW
(Figure 3) is not accounted for.

The observed loading rates are not representative for loading rates that lead to high
removal percentages or acceptable effluent concentrations (sub µg/L-range) and
low sorbent dosages. For ~99% reduction of PFAS concentrations from GW probably
typically one to two orders of magnitude higher sorbent dosages are required, which
automatically results in lower dosing rates.

Bed lifetime
The supplier states that recent experiments by RemBind® show that the bed lifetime
is similar to high grade activated carbon.

Effect of temperature and pressure

There is no information available regarding the effects of temperature and pressure
of the continuous system. The effects are expected to be similar to activated
carbon, so one could assume that effects are negligible. A recent study by Kabiri
and McLaughlin 2 showed that a freeze/thaw cycle had no effect on PFAS binding to
Rembind®.

4.2.4. Operational aspects

Energy use
The dominant primary energy use in a continuous filter are the feed pumps.
Dependent on influent characteristics and treatment strategy, backwashing pumps
might be applied as well, however this option is not mentioned in Riegel, et al. 1.
Energy use is directly linked to the treated water flow (m3/h), the hydraulic head
of the filter and the filter configuration and should be detailed on a case-by-case
basis. Depending on the applied method for energy calculation, one might take
transportation energy as well as the energy requirement for the off-site disposal of
the material into consideration. As a rule of thumb, for large-scale filters,
comparable to activated carbon filters, an energy use of maximum 0.1 kWh/m3
treated can be assumed based on expert judgment. This includes both feed- and
backwash pumps with a hydraulic head of 6-8 m, and excludes energy related to
transport and disposal.

For the discontinuous powder application with a stirred contact tank and a
sedimentation tank, the major energy users are the feed pumps and the mixing
devices. Generally, these systems have a lower hydraulic head compared to
continuous packed bed filters and mixing devices as well as settlers require very
little energy compared to feed- and backwash pumps. The overall energy use can
thus be expected to be less than that of continuous systems.

Maintenance and personnel

Since the continuous system is comparable to an activated carbon filter (benchmark
system), the average time spent should not exceed 0.5 day per week, depending on
the exact filtration set up and water composition. For a discontinuous system, there
is no reason to assume it would be significantly higher or lower than the continuous
variety.

41
 report no. 5/21

Chemical use
The main chemical use is the sorbent. In continuous filtration, this is the only
chemical that is applied. Based on current available data, the exact chemical use
per m3 treated is unknown. In discontinuous systems, additional chemical dosing
can be implemented to increase the settling of the sorbent in the sedimentation
step. The supplier specified that no chemicals are required for the RemBind®
system. Spent material in a batch reactor is separated using filtration or
centrifugation.

Waste production
Spent sorbent is dealt with similarly to activated carbon and needs to be incinerated
or disposed. As it contains also inorganic material (aluminium oxide and kaolin),
there will also be some of ashes after incineration but according to a leading landfill
in Germany this ash content is not significant in terms of costs as stated by the
supplier. Overall, the cost of RemBind® including disposal for spent RemBind® is
similar to the cost of carbon including regeneration, according to the supplier. As
with chemical use, waste production is directly linked to several factors and thus a
rule-of-thumb of waste production is not available.

4.2.5. Costs

Capital costs
For continuous systems, the capital costs can be relatively low as these systems are
relatively simple, comparable to activated carbon. The vessels operate under
atmospheric condition. Depending on the matrix some more extensive pre-
treatment can be required, increasing the capital costs. If, for example, a sand
filter is required, the capital costs can more or less double. A major unknown factor
is the EBCT, which largely determines the required volume of the system. A general
rule-of-thumb value for CAPEX is not available.

Operational costs
Operational costs are based on the abovementioned operational aspects. A major
factor is the costs of the sorbent itself as well as the disposal of the material. There
is no accurate cost estimate available to determine EUR/m3 treated.

4.3. DEXSORB+®

4.3.1. Introduction

DEXSORB+® is a cyclodextrin-based sorbent derived from cornstarch. DEXSORB+® is

currently brought to market by CycloPure (USA). Cyclodextrin-based sorbents are
made by cross-linking of “cup-shaped” cyclodextrin molecules into porous
polymers. During this process, functional groups can be added to make the removal
more targeted towards for example PFAS. The sorbent is currently available as
powder or in granular form. Based on supplier contact, the granular form seems
most promising for water treatment application.

The removal mechanism relies on hydrophobic interactions between the inner part
of the cyclodextrin molecules and the carbon chains of PFAS as well as on
electrostatic interactions between the positively charged units and the anionic head
groups of the PFAS molecules.

42
 report no. 5/21

Figure 17: PFAS flow diagram for sorption filtration with cyclodextrin-based sorbent.
Note that the values are indicative based on Riegel, et al. 1; removal
efficiency and target PFAS concentrations strongly depend on PFAS
components to be removed and site-specific conditions (e.g. influent
concentration, presence of competing substances, etc.). According to the
supplier < 0.01% of influent volume needs further treatment of regenerate.

4.3.2. Performance

The reported performance in Riegel, et al. 1 is mainly based on laboratory- and

bench scale experiments. Pilot tests are currently being performed, but results are
not publicly available at the time of writing. The data suggest that both long-chain
and short-chain PFAS can be removed with faster sorption kinetics than activated
carbon. Removal efficiencies for FFWW up of 90-99% have been reported, and
Riegel, et al. 1 describe the performances as “showing potential” for these
applications. Breakthrough of different PFAS have been described, ranging from
~6.500 BV to >55.000 BV, depending on the matrix.

As discussed in Paragraph 3.3 batch experiments of the tested sorbent shows higher
affinity for longer chain PFAS than for shorter chain PFAS. Below some sorption
coefficients that were derived from these data by means of fitting Freundlich
isotherms (Appendix 10) and are listed in Table 9. Within this table the sorption
coefficients that are listed are the coefficients at an aqueous concentration of
1.0 µg/L. This normalisation at a concentration of 1.0 µg/L allows to compare
compounds independent of the native concentrations. The slope of the isotherm
indicates the linearity of the sorption with concentration. The sorption coefficient
is independent of concentration when 1/n is (near) 1 and becomes lower with
increasing concentrations when it is <1, values above 1 are not expected. The
sorption of the sulfonic acid PFOS is stronger than that of PFOA (with the same
carbon chain length) and any of the other carboxylic acids. The breakthrough curves
even more explicitly show the difference between the removal of the sulfonic and
carboxylic acids as the shorter chain carboxylic acids except PFOA all break through
within 40.000 bed volumes while even the shorter chain sulfonic acids don’t
(>40.000 BV). Results are in line with literature findings listed in the previous
paragraph, except for PFBA and PFPeA that showed almost immediately or direct
breakthrough.

43
 report no. 5/21

Table 9: Sorption and break through characteristics of the sorbent in batch and
column testing with impacted groundwater (GW).

Carboxylic acids Sulfonic acids

Length of 4 5 6 7 8 4 6 8 6
fluor chain
sorbent PFBA PFPeA PFHxA PFHpA PFOA PFBS PFHxS PFOS 6:2 FTS

CP KSW1 4613 16749 66681 116950 71614 n.a.4 n.a.4 314775 n.a.4
(95% (2472 to (13804 to (51286 (57544 to (23659 to (221820 to
Confidence 8610) 20324) to 237684) 216272) 446684)
Limits) 86497)

CP 1/n (95% 1.07 0.78 0.84 0.98 0.85 0.11

Confidence (0.51 to (0.71 to (0.73 to (0.80 to 1.16) (0.67 to (-0.13 to
Limits) 1.62) 0.86) 0.95) 1.03) 0.36)

CP (20%)2 03 1.000 7.000 12.000 >40.000 >40.000 >40.000 >40.000 n.a.5

1
Sorption coefficient ([PFAS] µg/kg sorbent / [PFAS] µg/L water) defined (extrapolated) at an aqueous concentration of
1 µg/L based on sorption isotherms listed in the Appendix 6. In most cases this concentration is at the high end of just
outside the concentration range tested, so it can be considered a worst case sorption coefficient, as sorption coefficients
generally increased with decreasing concentrations.
2
Number of bed volumes where a break-through of 20% of the initial concentration was observed.
3
Direct break through, no significant retention
4
Not applicable; data did not reveal clear sorption isotherms.
5
Not applicable; concentrations were too low to assess sufficient reduction of 6:2 FTS in the treated water
6
The slope of the sorption isotherm (log concentration on sorbent (µg/kg) vs. log concentration in aqueous phase (µg/L)

4.3.3. Key Design criteria

Treatment strategy/treatment concept

As mentioned above, DEXSORB+® is available as powder or in granular form.
Currently, the supplier is testing granular-based applications for water treatment.
The process flow diagram, as showed in Figure 17, accurately depicts the basic
principle, comparable to activated carbon, a granular media DEXSORB+® filter will
likely consists of either parallel or serial filters (lag-lead configuration).
Regeneration, which is unique to DEXSORB+® with respect to activated carbon, can
take place either in-line (on-site), or one of the filters can be taken out of the plant
and regenerated at a central facility off-site. This is dependent on the total
installation size and the economics that go along with transport and regeneration:
if multiple sites in a close proximity are applying this technology, a central
regeneration unit might be profitable. An exact tipping point for this application of
economy of scale is not known yet.

Pre-treatment
Similar to activated carbon, pre-treatment for iron and manganese might be
necessary. The benefit of DEXSORB+® is its resistance to fouling from complex
matrices: due to its physical characteristics only low-molecular weight organics
directly compete with PFAS, larger molecules such as NOM or DOC, pose less of a
hindrance to DEXSORB+®. No pH effects have been expected, so also pH corrections
are not required.

Contact time
Empty bed contact time in the DEXSORB+® filter vessel is substantially shorter than
activated carbon. Whereas activated carbon requires contact times ranging from 15
minutes up to several hours, the vendor indicates that contact times of several
minutes (3 to 5, up to a maximum of 10) should be sufficient for high removal.

44
 report no. 5/21

Hydraulic loading rate/filter velocity

Filtration velocity is not known, based on kinetics and required contact time, one
could assume that a faster velocity must be achieved to overcome mass transfer
limitations than e.g. activated carbon, i.e. >10 m/h.

Contaminant loading rate

This data is not available from Riegel, et al. 1. Experimental results from this study
show that loading rates were 1.2 (95%CL = 0.8-1.9) and 22.8 (95%CL = 6.2-83.0) g
PFAS/kg sorbent for the GW without and with benzene spike respectively, and 3.8
(95%CL = 3.2-4.4) for FFWW. However, these loading rates were obtained at rather
high aqueous concentrations of the detectable PFAS (~7 µg/L, for GW and around
5000 µg/L for FFWW). Interestingly, the loading rate of the benzene spiked GW
appeared to be remarkably higher than the one without benzene spike. Comparison
of loading rates of impacted GW and FFWW is not directly possible as the PFAS
composition strongly differs (Figure 2) and the fraction of undetected PFAS in
especially FFWW (Figure 4) is not accounted for.

The observed loading rates are not representative for loading rates that lead to high
removal percentages or acceptable effluent concentrations (sub µg/L-range) and
low sorbent dosages. For ~99% reduction of PFAS concentrations from GW probably
typically one to two orders of magnitude higher sorbent dosages are required, which
automatically results in lower dosing rates.

Bed lifetime
Bed lifetime varies strongly based on desired removal rate and the composition of
the wastewater. Literature reported bed lifetimes varying between 6.300 and
>55.000 1. The supplier has additionally reported bed lifetimes up to and over
150.000 BV. The small scale column studies showed > 40.000 bed volumes without
breakthrough for PFOA, and all sulfonic acids, while PFBA, PFpeA, PFHxA, PFHpA
showed (20%) breakthrough at lower bed volumes ranging from 0 to 12.000 BV
(Table 9) with the generic trend that the shorter the fluorinated chain is, the lower
the affinity of the sorbent is. This is in line with data reported by Riegel, et al. 1,
and lower than what was reported by the supplier. However this seeming
contradiction might also be a result of the selection of test chemicals or matrix
composition used in the various experiments.

Effect of temperature and pressure

There is no information available regarding the effects of temperature and
pressure. The system applies ambient pressures and ambient temperatures, tests
with elevated temperatures have not been conducted.

4.3.4. Operational aspects

Energy use
The main primary energy use for the DEXSORB+® system is the feed pumps. There
are no other significant energy consumers. For a system with on-site regeneration,
additional, smaller pumps for regeneration are required. For off-site regeneration,
transportation energy might be taken into account, as well as the energy
requirement for the off-site regeneration system. Exact energy use is thus
dependent on pump energy, and with that directly related to the hydraulic head
and the flow that is to be treated, and thus comparable to activated carbon
(<0.1 kWh/m3).

45
 report no. 5/21

Maintenance and personnel

The DEXSORB+® system should be comparable to an activated carbon filter, which
requires relatively little maintenance and upkeep. With an on-site automated
regeneration system, longer bed lifetimes should be achievable, thus lowering re-
filling frequencies of the sorbent media. Depending on the chosen regenerant fluid,
one might require some more regular inspection and maintenance, since e.g.
methanol is a flammable liquid. The supplier is currently looking into alternative
regenerants that are more user-friendly. Overall maximum staffing of
approximately 0.5-1 day per week should be expected.

Chemical use
Besides the sorbent itself, the DEXSORB+® system requires, if desired, the
regenerant. As mentioned above, regeneration can be done by flushing the system
with a methanol-salt solution, comparable to how some ion-exchange systems for
PFAS-removal are regenerated. Exact dosing ratios and chemical consumption is not
known.

Waste production
The DEXSORB+® system creates two waste streams: the spent sorbent and the
regenerant. The supplier claims that the sorbent can be infinitely regenerated to a
high degree. It is realistic to assume that a certain amount has to be replaced and
thus disposed of. This material can then be treated comparable to other spent
sorbent media, by for example incineration. The loaded regenerant fluid can be
concentrated further, re-used and/or disposed similar to residues from ion
exchange materials 1. At this moment, the amount of waste produced is not known.

4.3.5. Costs

Capital costs
Capital costs are relatively low, comparable to activated carbon if no regeneration
is applied and compared to an ion exchange system with regeneration. The vessels
can operate under atmospheric conditions. Depending on the chosen regenerant
fluid, one might have to take fire-safety into account, which will add to the costs
of the installation and is a potential drawback of this system. Exact cost calculations
for full-scale systems are not available.

Operational costs
Operational costs is based on the abovementioned operational aspects. A major
factor is the costs of the sorbent itself and the costs of the regenerant fluid. Current
estimates from the supplier is that the material will cost between EUR 15-20 per
kg. Currently EUR 20 is achievable on lab-scale, but once scaled-up to industrial
production, prices are expected to drop to <15 EUR. Regenerant can be a large cost
factor, methanol, for example, is commercially available for around EUR 300-500
EUR/ton. Besides these price estimates, there is no translation available to a cost
of EUR/m3 treated.

4.4. POLYQA-OSORB®

4.4.1. Introduction

PolyQA-Osorb® is a sorbent based on organosilica, also known as swellable

organically modified silica’s (SOMS). PolyQA-Osorb® is marketed by ABS Materials
(USA). PolyQA-Osorb® was not discussed in Riegel, et al. 1 since it was too new and
no (peer-reviewed) literature was available, however during the technology

46
 report no. 5/21

selection phase for this study it was deemed promising. This chapter thus diverts
from the rest of the previous chapters. Note that the available literature is generally
on small lab-scale tests, so the validity needs to be proven by peer-reviewed studies
and larger scale (pilot) tests.

The removal mechanism is a combination of hydrophobic and ionic interactions. The

SOMS are highly hydrophobic, thus attracting the PFAS carbon tail. The SOMS are
treated with a cationic quaternary amine (QA) polymer, creating the PolyQA-Osorb®
as described here. This QA-polymer attracts the hydrophilic, ionic head of the PFAS.
The sorbent swells as it sorbs more molecules, up to 2-3 times its original size,
allowing for higher sorption capacities. The supplier notes the regenerative capacity
as one of the most promising. By flushing the SOMS with solvent, PFAS can be pulled
out of the matrix. Doing so fully regenerates the sorbent, according to the supplier.

4.4.2. Performance

For demineralized water, PolyQA-Osorb® removed >80% for a group of PFAS, with
the majority being removed for >90%, up to 99%. Short chain PFAS are removed to
a lower extent (PFBA, PFBS), however also some longer chain PFAS perform
relatively poorly (PFDA, PFOSaAm). Overall, performance for short-chain PFAS is
significantly higher than that of activated carbon. The PFCS slightly outperform
their PFCA counterparts. In treatment tests of GW, >99% removal can be achieved
for several short- and long chain PFAS. Table 10 shows sorption coefficients derived
from batch sorption experiments by means of fitting Freundlich isotherms
(Appendix 8). Within this table the sorption coefficients that are listed are the
coefficients at an aqueous concentration of 1.0 µg/L. This normalisation at a
concentration of 1.0 µg/L allows to compare compounds independent of the native
concentrations. The slope of the isotherm indicates the linearity of the sorption
with concentration. The sorption coefficient is independent of concentration when
1/n is (near) 1 and becomes lower with increasing concentrations when it is <1,
values above 1 are not expected. The tested sorbent did not allow fitting of removal
curves for many of the native PFAS in GW. Therefore the relation with carbon chain-
length or sulfonic vs. carboxylic acid is difficult to derive. No column studies were
performed with PolyQA-Osorb®.

Table 10: Sorption characteristics of the sorbent in batch tests with impacted
groundwater (GW).

Carboxylic acids Sulfonic acids

Length of fluor 4 5 6 7 8 4 6 8 6
chain
sorbent PFBA PFPeA PFHxA PFHpA PFOA PFBS PFHxS PFOS 6:2
FTS
PQ KSW1 n.a.4 7762 29376 n.a.4 n.a.4 n.a.4 n.a.4 306196 n.a.4
(95% (4276 to (15276 to (260016 to
Confidence 12972) 56624) 361410)
Limits)
PQ slope (95% n.a.4 0.3569 1.333 n.a.4 n.a.4 n.a.4 n.a.4 0.5117 n.a.4
Confidence 0.1048 to 0.5891 to 0.4082 to
Limits) 0.6090 2.077 0.6151
1
Sorption coefficient ([PFAS] µg/kg sorbent / [PFAS] µg/L water) defined (extrapolated) at an aqueous concentration of
1 µg/L based on sorption isotherms listed in the Appendix 6. In most cases this concentration is at the high end of just
outside the concentration range tested, so it can be considered a worst case sorption coefficient, as sorption coefficients
generally increased with decreasing concentrations.
2
Number of bed volumes where a break-through of 20% of the initial concentration was observed.
4
Not applicable; data did not reveal clear sorption isotherms.
6
The slope of the sorption isotherm (log concentration on sorbent (µg/kg) vs. log concentration in aqueous phase (µg/L)

47
 report no. 5/21

4.4.3. Key Design criteria

Treatment strategy/treatment concept

Currently, the material application has not been clearly chosen by the supplier,
although it seems likely that it is applied in a similar manner to Ion exchange or
DEXSORB+®. PolyQA-Osorb®, like these two technologies, has the capacity to be
regenerated with high recovery. It is thus likely that the treatment concept will
consists of (several) filter vessels with regeneration possibility. How the swelling of
the material will affect this and how to overcome the operational drawbacks of the
swelling is not yet known.

Pre-treatment
There is no information available regarding pre-treatment, although based on its
characteristics, it is expected that PolyQA-Osorb® has interference with both
hydrophobic (organic) molecules as well as cationic interactions. Based on
experiences in the US, this interference seems to be worse compared to the
interferences experienced by activated carbon.

Contact time
The supplier states that an EBCT of 1 minute is sufficient for treatment.20 This has
not been tested during this lab research, so some caution is advised with this
contact time since it seems rather optimistic based on expert judgement from other
sorption systems.

Hydraulic loading rate/filter velocity

There is no exact information available regarding pre-treatment, although based on
the EBCT it can be expected that the filter velocity is higher than that of an
activated carbon filter, thus exceeding 10 m/h.

Contaminant loading rate

In literature, based on lab-scale tests contaminant loading rates of 5-50 mgPFAS/g
have been observed, depending on the target PFAS. This indicates significantly
higher loading rates than e.g. GAC, which has loading rates in the order of mg/kg
magnitude. Experimental results from this study show that loading rates were 0.9
(95%CL = 0.6-1.4) and 15.7 (95%CL = 6.9-35.6) g PFAS/kg sorbent for the GW without
and with benzene spike, respectively. However, these loading rates were obtained
at rather high aqueous concentrations of the detectable PFAS (~7 µg/L).
Interestingly, similar to DEXSORB+® the loading rate of the benzene spiked GW
appeared to be remarkably higher than the one without benzene spike.

Data of the experiment with FFWW showed 2.4 g PFAS/kg at very high aqueous
concentrations of the detectable PFAS (~5000 µg/L). The data did not allow to
calculate confidence limits. The observed loading rates are not representative for
loading rates that lead to high removal percentages or acceptable effluent
concentrations (sub µg/L-range) and low sorbent dosages. For ~99% reduction of
PFAS concentrations from GW probably typically one to two orders of magnitude
higher sorbent dosages are required, which automatically results in lower dosing
rates. Subsequently, loading rates from literature seem to be over one order of
magnitude higher than in our observations. Furthermore, the presumed difference
with GAC mentioned above was not observed within our study.

48
 report no. 5/21

Bed lifetime
Based on lab research, breakthrough of long-chain PFAS (PFOA, PFOS) occurs
between 20.000-30.000 BV, where breakthrough is defined here as C/C0 of 0.3.
Shorter chain PFAS see much faster breakthroughs, PFBA is the first to reach
breakthrough at 5.000 BV. The supplier states that for relevant scale application,
they’re expecting 500.000-750.000 BV, however this is not substantiated with
calculations or measurements.20 No column studies were performed, to either
support or reject the supplier’s claims, but the results of the sorption studies do
not give indications in that direction, but rather align with other lab scale results.

Effect of temperature and pressure

There is no information available regarding the effects of temperature and
pressure. It can be expected that elevated pressures have some adverse effect on
swelling sorbents, by for example causing blocking or reducing the swelling capacity
of the material, potentially leading to lower maximal contaminant loading rates on
the filter as a whole, because the porosity decreases. A lower porosity requires,
theoretically, longer EBCT’s for sufficient diffusion and thus removal. At similar
EBCT’s, with a swollen bed, lower removal will be achieved. The actual effect is
not known.

4.4.4. Operational aspects

Energy use
The main primary energy use for the PolyQA-Osorb® system is the feed pumps. There
are no other significant energy consumers expected. For a system with on-site
regeneration, additional, smaller pumps for regeneration are required. Exact
energy use is thus dependent on pump energy, and with that directly related to the
hydraulic head and the flow that is to be treated, and thus comparable to activated
carbon (<0.1 kWh/m3).

Maintenance and personnel

The PolyQA-Osorb® system should be comparable to an activated carbon filter,
which requires relatively little maintenance and upkeep. With an on-site automated
regeneration system, longer bed lifetimes should be achievable, thus lowering re-
filling frequencies of the sorbent media. Depending on the chosen regenerant fluid,
one might require some more regular inspection and maintenance, since e.g.
methanol is a flammable liquid. Overall maximum staffing of approximately 0.5-1
day per week should be expected.

Chemical use
Besides the sorbent itself, the PolyQA-Osorb® system requires the regenerant. As
mentioned above, regeneration can be done by flushing the system with methanol
solution, comparable to how some ion-exchange systems for PFAS-removal are
regenerated. Exact dosing ratios and chemical consumption is not known.

Waste production
The PolyQA-Osorb® system creates two waste streams: the spent sorbent and the
regenerant. The supplier claims that the sorbent can be indefinitely regenerated to
a high degree. It is realistic to assume that a certain amount has to be replaced and
thus disposed of. This material can then be treated comparable to other spent
sorbent media, by for example incineration. The loaded regenerant fluid can be
concentrated further and/or disposed similar to residues from ion exchange
materials1 . At this moment, the amount of waste produced is not known.

49
 report no. 5/21

4.4.5. Costs

Capital costs
Capital costs are relatively low, comparable to an ion exchange system. The vessels
can operate under atmospheric conditions. With the current regenerant (i.e.
methanol), fire-safety has to be taken into account, which adds to the costs of the
installation and is a potential drawback of this system. Exact cost calculations for
full-scale systems are not available.

Operational costs
Operational costs is based on the abovementioned operational aspects. A major
factor is the costs of the sorbent itself and the costs of the regenerant fluid.
Currently, exact material costs are unknown, although the supplier mentions that
the material is more expensive per kg than activated carbon or ion exchange resin.
How much more expensive is not known exactly, but currently the price for small
amounts of material far exceeds other sorbents. Regenerant can be a large cost
factor, methanol is commercially available for around EUR 300-500 EUR/ton. The
supplier notes that they expect the system to be overall cheaper than activated
carbon or ion exchange, but gives no actual values.

4.5. FLUORO-SORB®

4.5.1. Introduction

FLUORO-SORB® is a surface treated mineral (bentonite based). It is commercialized

by Cetco (USA). FLUORO-SORB® was not discussed in Riegel, et al. 1 since it was too
new and no (peer-reviewed) literature was available, but was deemed promising
during the technology selection phase of this study. The mineral itself has a very
high surface area >200 m2/gram. The surface treatment is hydrophobic and attracts
the PFAS molecules. The material is expandable and can accommodate a maximum
sorption capacity of > 15 mg/g for commonly regulated PFAS compounds.

This chapter diverts from the rest of the previous chapters. The removal mechanism
is not clearly described by the supplier, other than that it is a proprietary sorbent.

4.5.2. Performance

There is no prior information available regarding treatment performance. Table 11

shows sorption coefficients derived from batch sorption experiments by means of
fitting Freundlich isotherms (Appendix 6). Within this table the sorption
coefficients that are listed are the coefficients at an aqueous concentration of 1.0
µg/L. This normalisation at a concentration of 1.0 µg/L allows to compare
compounds independent of the native concentrations. The slope of the isotherm
indicates the linearity of the sorption with concentration. The sorption coefficient
is independent of concentration when 1/n is (near) 1 and becomes lower with
increasing concentrations when it is <1, values above 1 are not expected. The
sorption of the sulfonic acid PFOS is much stronger than that of PFOA (with the same
carbon chain length) or any of the other carboxylic acids. The breakthrough curves
even more explicitly show the difference between the removal of the sulfonic and
carboxylic acids as PFBA, PFPeA and PFHpA break through within 40.000 bed
volumes while even the shorter chain sulfonic acids don’t.

50
 report no. 5/21

Table 11: Sorption and break through characteristics of the sorbent in batch and
column testing with impacted groundwater (GW).

Carboxylic acids Sulfonic acids

Length of 4 5 6 7 8 4 6 8 6
fluor chain
sorbent PFBA PFPeA PFHxA PFHpA PFOA PFBS PFHxS PFOS 6:2 FTS

FS KSW1 15136 10186 23174 99770 66069 n.a.4 n.a.4 16865530 n.a.4
(95% (6471 to (6310 to (15849 (41687 to (10914 to (1782379
35481) 16444) to 399025) to
Confidence 238781)
33963) 159587915)
Limits)
FS slope 1/n 1.34 0.78 1.02 1.62 1.16 3.66
(95% (0.60 to (0.72 to (1.32 to 1.92) (0.50 to (1.80 to
(0.69 to
Confidence 1.99) 0.95) 1.32) 1.82) 5.50)
Limits)
FS (20%)2 2.000 10.000 >40.000 29.000 >40.000 >40.000 >40.000 >40.000 n.a.5
1
Sorption coefficient ([PFAS] µg/kg sorbent / [PFAS] µg/L water) defined (extrapolated) at an aqueous concentration of
1 µg/L based on sorption isotherms listed in the Appendix 6. In most cases this concentration is at the high end of just
outside the concentration range tested, so it can be considered a worst case sorption coefficient, as sorption coefficients
generally increased with decreasing concentrations.
2
Number of bed volumes where a break-through of 20% of the initial concentration was observed.
3
Direct break through, no significant retention
4
Not applicable; data did not reveal clear sorption isotherms.
5
Not applicable; concentrations were too low to assess sufficient reduction of 6:2 FTS in the treated water
6
The slope of the sorption isotherm (log concentration on sorbent (µg/kg) vs. log concentration in aqueous phase (µg/L)

4.5.3. Key Design criteria

Treatment strategy/treatment concept

FLUORO-SORB® is, according to the vendor, versatile in its deployment. For water
treatment, the most likely concept is in a flow-through filtration system, similar to
a fixed bed activated carbon filter. Due to short expected EBCT’s, a large filter can
be used for optimal removal.

4.5.4. Pre-treatment

There is not a substantial amount of information available regarding required pre-

treatment. Based on contact with the suppliers, a similar pre-treatment regime is
required as for GAC. As for all filter application, particle removal can always be an
option if too much solids are going into the system. According to the vendor,
FLUORO-SORB® is less affected by co-contaminants, showing high removal rates in
the presence of diesel, dioxane, BTEX and TCE. Natural organic matter seems to
have a negative effect on the removal of PFAS, although recent testing suggest
lower competition with total organic carbon compared to GAC.

Contact time
The vendor claims an EBCT of 2-5 minutes is sufficient for high removal, no more
details are given. Recent tests suggest an 8 minute EBCT is required when using a
200 grade material. A shorter EBCT thus means smaller filters.

51
 report no. 5/21

Hydraulic loading rate/filter velocity

There is no information available regarding the hydraulic loading rate. Recent tests
(using material Gen1 FS200/FS400) and Gen 2, suggest a hydraulic loading of 6
gal/minute per square foot, which translates to approximately 15 m/h hydraulic
loading rate.

Contaminant loading rate

There is no factual information available regarding the contaminant loading rate.
The supplier claims a loading rate between 0.5 and 1.75 mg/g, however this is not
further substantiated with data. This indicates significantly higher loading rates
than e.g. GAC, which has loading rates in the order of mg/kg magnitude. Contact
with the supplier suggests a loading rate of the Gen1 (FS200/FS400) material, which
is 4 times higher than bituminous GAC, depending on matrix effects and PFAS-
composition. Contact with the supplier suggests a maximum sorption capacity of
>15 mg/g, with breakthrough at an average sorption capacity of 0.2 mg/g with an
EBCT of 10 minutes. Currently the supplier is working on a Gen2 version, which is
supposed to have 4x higher loading rates compared to Gen1, with improved removal
for short chain and carboxylic acid PFAS. Experimental results from this study show
that loading rates were 10.5 (95%CL = 4.2-26.4) and 22.6 (95%CL = 2.3-226.0) g
PFAS/kg sorbent for the GW without and with benzene spike respectively while data
of the experiment with FFWW resulted in a loading rate of 1.4 (95%CL = 0.8-2.5).
These loading rates were obtained at rather high aqueous concentrations of the
detectable PFAS (~7 µg/L for impacted GW and even ~5.000 for FFWW).
Interestingly the loading rate obtained for impacted GW is remarkably similar to
the producers claim of maximum sorption capacity (>15 g/kg).

The observed loading rates are not representative for loading rates that lead to high
removal percentages or acceptable effluent concentrations (sub µg/L-range) and
low sorbent dosages. For ~99% reduction of PFAS concentrations from GW probably
typically one to two orders of magnitude higher sorbent dosages are required, which
automatically results in lower dosing rates. These levels are in line with the claimed
loading rates of 0.5-1.75 g/kg and break though loading rate of 0.2 g/kg by to the
supplier.

Bed lifetime
There is no information available regarding the bed lifetime. The column studies
illustrate that the small scale column studies showed > 40.000 bed volumes without
breakthrough for PFHxA, PFOA, and all sulfonic acids, while PFBA, PFpeA, and
PFHpA showed (20%) break though at lower bed volumes ranging from 2.000 to
29.000 BV (Table 11) with the generic trend that the shorter the fluorinated chain
is, the lower the affinity of the sorbent is. This is in line with data reported by
Riegel, et al. 1.

Effect of temperature and pressure

There is no information available regarding the effects of temperature and
pressure. Higher temperatures and pressures should not result in a degradation of
performance.

52
 report no. 5/21

4.5.5. Operational aspects

Energy use
The main primary energy use for the FLUORO-SORB® system is the feed pumps.
There are no other significant energy consumers expected. Exact energy use is thus
dependent on pump energy, and with that directly related to the hydraulic head
and the flow that is to be treated, and thus comparable to activated carbon (<0,1
kWh/m3).

Maintenance and personnel

The FLUORO-SORB® system should be comparable to an activated carbon filter,
which requires relatively little maintenance and upkeep and should thus not exceed
0.5-1 day per week.

Chemical use
Based on current available information, the FLUORO-SORB® system requires no
other chemicals then the sorbent itself.

Waste production
The FLUORO-SORB® system creates one waste stream: the spent sorbent. This
material can then be treated comparable to other spent sorbent media, by for
example incineration or landfilling. At this moment, the amount of waste produced
is not known. The supplier states that they are investigating the commercial
viability of regeneration process for FLUORO-SORB®.

4.5.6. Costs

Capital costs
Capital costs are expected to be relatively low, comparable to an activated carbon
filter. The vessels can operate under atmospheric conditions. Exact cost
calculations for full-scale systems are not available.

Operational costs
Operational costs are based on the abovementioned operational aspects. Contact
with the supplier suggests material costs between EUR 2,500 and 3,000 per ton of
material. The supplier notes that they expect the system to be overall cheaper than
activated carbon or ion exchange, but gives no actual values.

4.6. PERFLUORAD

4.6.1. Introduction

PerfluorAd is a flocculation technology that applies a proprietary liquid reactant to

remove the bulk PFAS from the water. PerfluorAd is marketed by Cornelsen
Umwelttechnologie GmbH (Germany).

The removal mechanism is based on both electrostatic and hydrophobic interactions

between the reactant and the PFAS. Generally, PerfluorAd is applied for higher PFAS
concentrations, from 0.3 µg/L onwards. The reactant is dosed in a stirred reactor,
where the coagulation/flocculation takes place, followed by the removal of the
PFAS-containing flocs, either by sedimentation, filtration or a combination thereof.
Finally, polishing is used to obtain desired PFAS-levels. Generally, this is done with
a sorptive technology like activated carbon or ion exchange resin.

53
 report no. 5/21

Figure 18: PFAS flow diagram for treatment with PerfluorAd. Note that the values are
indicative based on Riegel, et al. 1, applying 0.1 g/L flocculant dosage.
Removal efficiency and target PFAS concentrations strongly depend on PFAS
components to be removed and site-specific conditions (e.g. influent
concentration, presence of competing substances, etc.).

4.6.2. Performance

Based on Riegel, et al. 1, the performance of PerfluorAd varies depending on dosage

of the reactant and the target PFAS. Removal rates of up to 90% have been observed
for some PFAS, others were only removed up to 60% at higher dosages, so no
universally applied rule-of-thumb for removal can be given. It is noted that the goal
of PerfluorAd is not to achieve low effluent values, since it is designed as a bulk-
removal, followed by a polishing step that will reduce the PFAS to the desired
concentration. The batch experiments with PerfluorAd illustrated high loading
(complexation) rates. Loading rates were 61.5 (95%CL = not determinable) and 61.0
(95%CL = not determinable) g PFAS/kg sorbent for the GW without and with benzene
spike respectively. The lack of the confidence limit was not necessarily a problem
of poor data but rather of the result that even very low dosing of the flocculant
lead to significant reduction of the PFAS with the result that no accurate CL could
be determined. For FFWW the loading rate of determinable PFAS was 21.4 (95%CL
= 18.0-25.5). These loading rates were obtained at rather high aqueous
concentrations of the detectable PFAS (~7 µg/L for impacted GW and even 5.000
µg/L PFAS of the FFWW), such high dosages are exactly what this material is
designed for (removal of bulk PFAS, rather than polishing). These observed loading
rates are not representative for loading rates that lead to high removal percentages
or acceptable effluent concentrations (sub µg/L-range). Probably an additional
polishing step is required to reach such levels for this particular material.

54
 report no. 5/21

4.6.3. Key Design criteria

Treatment strategy/treatment concept

PerfluorAd is applied as a pre-treatment for bulk PFAS removal, followed by a
polishing step. The PerfluorAd system relies on mixing a fluid reagent with the PFAS-
containing water in a stirred tank, where flocs are formed. These flocs are removed
via either sedimentation, filtration or a combination of both. Generally, PerfluorAd
is applied for higher PFAS concentrations, from 0.3 µg/L onwards.

Pre-treatment
The PerfluorAd itself does not require significant pre-treatment, however, it may
be required to deal with some matrix effects if the following polishing step is
negatively impacted. PerfluorAd removes iron, colour and particles to some extent,
so less pre-treatment might be required. However there is no removal of manganese
or DOC expected with the PerfluorAd system.

Contact time
There is no information available regarding the contact time in the stirred tank.
Depending on the flocs and the chosen separation unit, sedimentation can take up
to 40 minutes, whilst a (sand)filter for floc-separation generally has a contact time
of 5-10 minutes. Depending on the chosen polishing step and its specific design
criteria, an additional contact time of 2-20 minutes has to be taken into account.

Hydraulic loading rate/filter velocity

There is no information available regarding the hydraulic loading rate of the stirred
tank. For flocculation, the characteristics of the flocs are not known, so loading
rate is not known. Generally, depending on floc integrity and strength, for filtering
through a sand filter, a filter velocity of 10-15 m/h can be assumed. Depending on
the chosen polishing step and its specific design criteria, the filter velocity for this
step is expected to be between 5-15 m/h.

Need for sedimentation vessel

The vendor mentions that sedimentation vessels are not always required. If a
filtration stage for separation iron and/or solids is already in place, only the stirred
tank is needed. It is not clear when a sedimentation vessel is required, but it stands
to reason that this step is added when the flocs cause rapid blocking of the filter
media, yielding a (too) high backwash frequency. At low suspended solids rates, a
direct in-line coagulation can be possible.

Effect of temperature and pressure

There is no information available regarding the effects of temperature and
pressure. However, with increasing temperature coagulation can be expected to
perform better and with lower mixing energy, due to the decreased viscosity of the
water. How relevant this effect is with regards to PFAS removal is unknown.

4.6.4. Operational aspects

Energy use
The primary energy use for the PerfluorAd system depends on the system
boundaries. If a filter is already present and only the mixing vessel is required,
energy use is expected to be very low. A stirring motor requires a low amount of
energy compared to feed- and backwash pumps of filters. In this case, the backwash

55
 report no. 5/21

pumps will have to be used more often, so there’s an increase in energy use. If
there’s no filter available, the feed- and backwash pumps of the filter are expected
to be the largest energy consumers. This is expected to be similar to a sand- or
activated carbon filter. Depending on the desire to dewater the PFAS-sludge, a
dewatering installation also needs to be installed and powered. The energy use
depends on the chosen dewatering technology and thus a generic rule of thumb is
not available. Finally, PerfluorAd requires a polishing step, which brings along its
own energy use. This is not described here.

Maintenance and personnel

There’s no explicit data available on maintenance and personnel, however, based
on the technology description, it is expected that it does not require significant
amounts of attention. Coagulation/flocculation followed by filtration and/or
sedimentation, combined with dewatering is always more demanding than just
filtration. The exact time needed is not known.

Chemical use
The vendor claims that the reactant can be tailor-made for precipitation of specific
PFAS. Dosing of the reagents lies between 25 to 2,000 mg/L, depending on the
matrix and the desired removal rate. A higher dosing yields a higher removal rate,
especially for short-chain PFAS. Exact use is thus highly dependent on matrix and
desired yield and therefore no rule-of-thumb values can be given. The batch
experiments are in line with the vendors dosing range, as FFWW as well as impacted
GW show significant removal at PFAD dosages as low as 10 mg/L for impacted GW
(Figure 6) and 1,000 mg/L for FFWW (Figure 5).

Waste production
The primary waste stream is the PFAS-containing sludge. The amount of waste
directly correlates to the dosing ratio, the amount of co-precipitated contaminants
(iron, colour, particles) and the ability to dewater the sludge. Exact waste
production is thus highly dependent on matrix and desired yield and therefor no
rule-of-thumb values can be given. The vendor claims that the active ingredients of
PerfluorAd are biodegradable, however PFAS-containing sludge will highly likely
have to be dewatered and subsequently dealt with according to local regulations,
for example incineration.

4.6.5. Costs

Capital costs
There are no cost calculations available, however, compared to a stand-alone
sorptive media filter, capital costs are expected to be at least double since two
filter units are required, along with the mixing vessel, storage for the reagent and
a dewatering step for the sludge.

Operational costs
There are no cost calculations available. The main goal of PerfluorAd, however, is
to decrease the operational costs of PFAS treatment by strongly concentrating the
bulk of the PFAS in the sludge, so that the (expensive) sorbent has a longer lifetime.

56
 report no. 5/21

4.7. NANOFILTRATION

4.7.1. Introduction

Nanofiltration is a pressure based separation technology that applies dense

membranes. This technology is not supplier-specific, several different technology
providers have their own variation of these membranes available. The removal
mechanism is the non-target removal via physical separation based on molecular
weight/size, along with chemical interactions with the membrane material. The
technology yields two water streams: the concentrate and the treated water, as
will be discussed below.

Figure 19: PFAS flow diagram for treatment with nanofiltration. Note that the values
are indicative based on Riegel, et al. 1. Removal efficiency and target PFAS
concentrations strongly depend on PFAS components to be removed and
site-specific conditions (e.g. influent concentration, presence of competing
substances, etc.).

antiscalant dosing
INFLUENT 0.002 to 0.004 g/L
100% PFAS
FURTHER TREATMENT
PRETREATMENT of CONCENTRATE
(if required)
Particle
sand filtration
UF [no PFAS removal] CONCENTRATE
Fe, Mn [approx. 20% of i nfluent vol.]
sand filtration
slightly PFAS removal
100% PFAS

SLUDGE TREATED WATER

DISPOSAL [approx. 80% of i nfluent vol.]
membrane
[approx. 0.1% 0 % PFAS
of influent vol.] filter

57
 report no. 5/21

4.7.2. Performance

NF membranes will reject solutes such as PFAS while water permeates through the
membrane. The rejections depends on the molecular weight cut-off of the
membrane and other membrane characteristics such as surface charge. Generally,
>90% of the PFAS can be removed, for many >99% is attainable. Within the
experiments initial removal was indeed 90%-95% but with longer operation, the
removal dropped to the point where the permeate nearly reached the concentration
in the feed water. While the membrane started fouling and the transmembrane flow
also dropped. Apparently the FFWW contained particulate material that corrupted
membrane functioning and separation.

4.7.3. Key Design criteria

Treatment strategy/treatment concept

Generally, the treatment concept is quite simple: a feed pump pre-pressurizes
water that is fed to the membrane units. Besides the membrane unit itself,
extensive pre-treatment is required (discussed below). Membrane cleaning is also
an important aspect. The concentrate stream, approximately 10-20% of the original
water flow, generally needs further treatment, since it’s still a relatively diluted,
watery volume.

Pre-treatment
Even though some membranes are more resistant to particulate matter, a general
rule of thumb is that the water needs to be free of particles, iron and manganese
to levels of <1 FTU turbidity, <0.05 mg/L iron and <0.02 mg/L manganese. Sand
filtration is a common-pretreatment, as well as micro- or ultrafiltration. These
steps do not retain PFAS, so the backwash water can be discharged safely. Acid
and/or antiscalant dosing might be required to prevent scaling of sparingly soluble
salts like CaCO3, CaSO4 or BaSO4.

Filter type
There are several options available for membrane units that can be applied
depending on the water matrix and the operating philosophy of the client. The most
well-known examples are tubular, spiral-wound and hollow fiber membranes.
Tubular membranes can handle viscous liquids with relatively high suspended solids
and can be cleaned both mechanically or chemically. Spiral wound membranes are
known for their space-efficiency and small footprint that can handle large
throughputs. Downside of this is that these membranes lose robustness and thus
require more careful pre-treatment. Hollow fiber membranes provide benefits of
both: relatively small footprint that can be cleaned relatively easily and that can
handle particles better than spiral wound. Downside of hollow fiber is that these
are maxed out at a TMP of 2 bar. Furthermore, different materials can be used.
Ceramic and polymer are the most common materials. Ceramic is more suitable for
hostile environments (solvents, wide pH ranges etc.), otherwise polymeric
membranes are the standard.

Pore size
Nanofiltration has pore sizes ranging from 0.5 to 2 nm. Increasing the pore-size
yields lower removal, since smaller molecules can pass through, but decreases
energy use. Pore size can, alternatively, be defined as the amount of Dalton to be
filtered. Dalton is a molecule weight measurement, corresponding to the molecular
weight [g/mol] and should be rounded down generously. Commonly occurring PFAS
are ranging from 200 to 700 g/mol. For example, PFOS and PFOA have molecular

58
 report no. 5/21

weights of 500 and 414 g/mol, respectively. If a water is impacted with both short
and long chain PFAS, one should consider a pore size of 200 Dalton. This lowest
molecular weight, thus the aimed size of the pores, is called the molecular cut off
weight.

Permeate flux
Another key design criterion is the permeate flux. This is a parameter defined as
the amount of liquid transported through one unit of membrane surface per unit of
time [L/[Link]]. The permeate flux and the ratio of the main treated water with
respect to the concentrate depends on the chosen membrane, the water matrix and
the transmembrane pressure. This ratio is also known as the recovery [%]. Maximum
recovery is 90%, usually 80% recovery is a safer rule of thumb for proper treatment.
Initial transmembrane flow and removal were acceptable, but with longer operation
in order to reach sufficient recovery (~80%) membrane started fouling and the
transmembrane flow also dropped resulting in very long operation (>3 days) in the
pilot experiment, while also the PFAS removal dropped. Apparently the FFWW
contained particulate material that corrupted membrane functioning and
separation, suggesting that membrane treatment require some pretreatment step
to remove the constituents that corrupt membrane functioning.

Trans-membrane pressure
Due to the small pores of the membranes, a high pressure of about 4 to 10 bars has
to be applied to treat drinking water with these technologies. Generally, the range
will be closer to 2-4 bar, 10 bar is on the higher end. The pressure drop (loss in
pressure at the end of the filter) is usually around 0.5 to 1.5 bar. When operating
in series, the pressure drop is the sum of all filters. When working in parallel, only
one pressure drop can be accounted for. Pumps and piping should be chosen
adequately considering these pressures.

Effect of temperature and pressure

There is no quantified information available regarding the effects of temperature
and pressure on PFAS removal. However, increasing temperatures decrease the
viscosity of the water which in turn has positive effects on the membrane flux and
requires less transmembrane pressure for treatment. This results in a lower pre-
pressure, resulting in lower energy use. Water that is already pressurized can,
depending on the pressure, reduce energy use of pre-pressuring the filters. If the
pressure of the water exceeds the maximum pressure of the chosen membrane a
pressure release or pressure reduction step has to be implemented.

4.7.4. Operational aspects

Energy use
The major factor for energy use is the pumps. This is also the major cost factor (see
below). Riegel, et al. 1 gives an energy use for NF of 0.4 kWh/m3. Additionally, pre-
treatment and concentrate treatment also require energy, depending on the
choices that are made there.

Maintenance and personnel

If set up correctly, membranes should be able to run relatively easily, although they
always require significantly more attention than media filtration. One should
account for 2-5 days per week (0.5-1 FTE) to keep a membrane unit up and running.

59
 report no. 5/21

Chemical use
Depending on the chosen membrane and the matrix, cleaning with water as
backwash could be sufficient, or more extensive cleaning programs are required
such as chemically enhanced backwashing and/or dosing of anti-scalant. The exact
amounts also depend on the matrix. For anti-scalant a dosing of 0.002-0.004 g/L is
a good starting point.

Waste production and concentrate treatment

There is no waste in the traditional sense of solid waste, however, the concentrate
does need to be treated further. This is a significant volume, up to 20% of the
original volume. There are several options available for this treatment, e.g. reverse
osmosis membranes that concentrate the PFAS into a smaller volume, or one could
use coagulant/flocculant based treatment on the concentrate stream. Eventually,
the PFAS has to be disposed of. Concentrate treatment is a significant part of NF
operations, but due to the varying options is not further discussed here.

4.7.5. Costs

Capital costs
Membrane installations are generally significantly more expensive with regards to
filtration units. As mentioned, membrane units already require media filtration as
pre-treatment. Exact costs are strongly dependent on the amount of pre-treatment
required and the chosen membrane units. Generally, it can be expected that a full
NF train has capital costs of 2 to 5 times that of a general (activated carbon) media
filter.

Operational costs
As mentioned, energy costs are a major factor in membrane filtration. For energy
alone this amounts to about EUR 0.05-0.10 per m3 treated, depending on energy use
and local energy prices. Furthermore the membranes have to be replaced.
Replacement frequency depends on operational regime and water matrix. The
waste/concentrate has to be handled, which can be a significant cost factor
depending on the chosen downstream processing. Based on this, there is no clear-
cut rule-of-thumb cost for treatment with NF.

4.8. FOAM- AND OZO FRACTIONATION

4.8.1. Introduction

Foam fractionation applies injection of compressed air (foam fractionation) or

ozone (ozo fractionation), which leads to the formation of bubbles into the water.
The PFAS surfactants adhere to the bubble walls and the PFAS-enriched foam is
collected at the aqueous surface for further (destruction-based) treatment. Foam
fractionation is marketed by OPEC systems (AUS) under tradename SAFF (Surface
Active Foam fractionation). Ozo fractionation is marketed by EVOCRA under trade-
name OCRA (ozofractionative catalyzed reagent addition). In both cases, the
removal mechanism relies on the foam-forming/surfactant qualities of the PFAS.
PFAS tend to favour air-water interfaces, due to their hydrophobic per- and
polyfluorinated carbon chains and the hydrophilic functional groups.

60
 report no. 5/21

Figure 20: PFAS flow diagram for treatment with foam fractionation. Note that this
scheme is generic for both SAFF and OCRA. The values are indicative based
on Riegel, et al. 1. Removal efficiency and target PFAS concentrations
strongly depend on PFAS components to be removed, chosen system and
site-specific conditions (e.g. influent concentration, presence of competing
substances, etc.).

4.8.2. Performance

Both SAFF and OCRA showed high removal rates in field-trials. Removal of 99-100%
removal were observed for both systems without polishing step. SAFF was able to
remove precursors and short-chain PFAS to some extent as well. OCRA has the added
effect of oxidizing some oxidizable precursors into persistent PFAS. Especially when
considering the addition of a polishing step, both systems can handle very high
influent concentrations and still yield >99.9% removal overall. Within the pilot
experiments such high removal rates were not obtained. Rather ~60-90% were
observed. The formation and stabilization of the foam seemed to be rather tricky,
especially with the heavily impacted FFWW, so potentially the experimental set up
was not optimal to fully enable PFAS removal.

4.8.3. Key Design criteria

Treatment strategy/treatment concept

SAFF applies a main primary contactor where water is treated and the foam is
removed. This primary contactor treats water in batches, thus for continuous
operation multiple parallel contactors are required. A second and (optional) third
step can be added to concentrate the foam further to reduce the amount of waste.
The primary water flow can be polished if desired/required. OCRA applies a serial
treatment with two or three consecutive treatment steps, followed by a polishing
step.

61
 report no. 5/21

Pre-treatment
For both foam- and ozo fractionation, significant effects of the matrix or co-
contaminants on PFAS removal are not to be expected. If a polishing filter is applied,
some more attention to pre-treatment can be beneficial, for example to remove
iron, manganese or suspended solids to a further extent. Foam fractionation showed
an effect of salinity increase, yielding higher amounts of non-PFAS surfactants in
the foam. Ozo fractionation has some impact from high levels of organic
compounds. The PFAS-removal is not affected according to Riegel, et al. 1, because
ozone depletion is not critical for the main treatment mechanism. The presence of
oxidisable compounds do increase the ozone use and thus the operational costs.
This oxidation yields a cleaner effluent, which is an added benefit of the OCRA
system.

Contact time
Exact contact times for both systems are not publicly available at the time of
writing. Based on contact with the vendors, a contact time of approximately 30-45
minutes seems a good rule-of-thumb. As mentioned the foam fractionation system
works as (sequencing) batch, whilst ozo fractionation is more geared towards a
continuous system.

Hydraulic loading rate

For both systems, hydraulic loading rate is not defined.

Bubble size
Bubble size is an important design parameter that can be influenced by amongst
others gas choice (ozone yields slightly smaller bubbles then air), type of aeration
system (spargers, recirculation with side stream injection) and the water matrix.
Generally speaking, the smaller the bubbles, the larger the air-water interface and
thus a better removal. Smaller bubbles tend to have lower upward velocity and with
that increase the required contact time. For both foam- and ozo fractionation,
bubble sizes are in the micro-/nanobubble range, varying from tens of nanometers
to tens of micrometers.

Concentration factor
Concentration factor is a dominant cost-determining parameter, since the PFAS-
containing concentrate has to be treated downstream. The smaller this flow, the
lower the waste costs are. Foam fractionation with the SAFF system with a three-
stage treatment/concentrator set up is designed to reach 0.0025% concentrate in
volume. The vendor claims that further concentration factors can be reached in the
(near) future. Ozo fractionation in the current two-stage process reaches 0.5 to 2%.

Effect of temperature and pressure

There is no information available regarding the effects of temperature and
pressure, however, elevated temperatures result in lower gas transfer rates and
lower saturation levels. This can have an adverse effect on the bubble size and
stability. Higher pressures could result in a higher solubility of gas, thus decreasing
the amount of bubbles with a negative effect.

62
 report no. 5/21

4.8.4. Operational aspects

Energy use
Energy use varies between the two systems. Generally, the air injection is
comparable per contactor step: pump energy is required to feed the contactors and
a recirculation pump or a compressor is required to feed the gas to the liquid phase.
Exact energy consumption figures for foam fractionation are not known, however
since it is only applying pumps and compressors, it is expected not to significantly
exceed an activated carbon filter, so energy use of approximately 0.1 kWh/m3
treated should be attainable. Ozo fractionation requires an additional energy
consumption due to the production of ozone either from ambient air or liquid
oxygen. This is a significant energy user. Overall, the vendor estimates energy
consumption at 3 to 7 kWh/m3 treated.

Maintenance and personnel

The foam fractionation system is relatively simple and mostly automated. The high
concentration factor means very little handling of waste streams. Exact
maintenance and personnel values are not known, but it is expected not to
significantly exceed an activated carbon filter, so this should not exceed 1 day per
week. Ozo fractionation is, as mentioned, very similar to foam fractionation with
the difference that the ozone system requires attention and maintenance and the
waste needs to be handled more often. These are not expected to be more than
double that of foam fractionation, so between 1-2 days per week.

Chemical use
Foam fractionation does not require the addition of chemicals, since it applies
ambient air. Ozo fractionation requires the use of ozone. Ozone has to be made on-
site, either from ambient air via e.g. pressure swing sorption or from liquid oxygen.
The ozone use is not known and thus the amount of liquid oxygen is not known.

Waste production
As mentioned in the subchapter on concentration factor, foam fractionation
produces 0.0025% of the influent water as concentrated waste. Ozo fractionation
produces 0.5-2%. Besides this, spent sorbent from the polishing step is a waste
production that both systems share but is not specified further, as well as pre-
treatment waste sludge from e.g. solids removal.

4.8.5. Costs

Capital costs
Capital costs for the contactor systems are relatively low as these are relatively
simple. The vessels operate under atmospheric conditions. However, a general rule-
of-thumb value for CAPEX is not available. For the ozo fractionation system, it
should be noted that the production- and application of ozone is capital intensive,
besides the ozone production step, additional safety measures have to be
implemented, increasing the CAPEX. CAPEX for the ozone system depends partially
on scale: small-scale ozone units generally apply pressure swing sorption, which
increases CAPEX but results in a lower OPEX because liquid oxygen is not required.
Liquid oxygen, however, increases ozone production efficiency, so has its own
advantages. The efficiency balance tends to favor ozone production from liquid
oxygen for larger scale systems. The ozo fractionation system is expected to be
significantly more expensive than the foam fractionation system, however how
much more expensive is difficult to predict and thus no general rule-of-thumb costs
are available.

63
 report no. 5/21

Operational costs
Energy use is a common cost factor for both foam fractionation and ozo
fractionation. For pumps and aeration, this is roughly similar, however ozo
fractionation requires significantly more energy due to the production of ozone.
Exact energy uses varies from 0.1 kWh/m3 for foam fractionation and 3-7 kWh/m3
for ozo fractionation. At energy prices of 0.1 EUR/kWh, this yields 0.01 EUR/m3 and
0.3-0.7 EUR/m3 respectively. Waste handling is another significant cost. Based on
the concentration factors and incineration costs of 1,000 EUR/m3, based on price
point 2019/2020 for incineration of PFAS-containing waste in the Netherlands, some
insight can be given into the OPEX. For foam fractionation with a concentration
factor of 0.0025%, this means that per m3 treated, 0.025 L of waste is produced,
costing EUR 0.025 to treat. Waste treatment costs thus come down to EUR 0.025/m3
treated. For ozo fractionation, with 0.5-2%, this comes down to EUR 5-20 per m3
treated. Further operational costs, like personnel and chemical use for the ozo
fractionation are not known, so a more detailed breakdown of operational costs is
not available.

64
 report no. 5/21

5. DISCUSSION AND CONCLUSIONS

5.1. DISCUSSION

This chapter is the integrated discussion of the experimental research and

evaluation of treatment technologies to remove PFAS from PFAS impacted GW and
FFWW.

Within this study two types of contaminated water were studied. One being PFAS
impacted GW and the other FFWW. Impacted GW and FFWW might be considered
at both ends of the spectrum of PFAS contamination. The concentration of
(detectable) PFAS differed almost by three orders of magnitude, with GW PFAS
concentration in the µg/L range and FFWW PFAS concentration in the mg/L range.
When samples were oxidized by a TOP analysis, the detectable PFAS concentration
in the FFWW increased with over an order of magnitude, while the concentration in
the impacted GW was not affected significantly. In addition to the PFAS
concentration, the PFAS composition also differed. The FFWW was dominated by
6:2 FTS, while the impacted GW contained various PFAS such as PFOS and PFOA.
Finally, the FFWW contained much more aromatic and mineral oil residues, metals
and organic and inorganic particles of undefined composition than the impacted
GW. The impacted GW and FFWW are therefore considered to provide the width of
the application range and are therefore very suitable to test treatment
technologies.

The differences in the two types of PFAS impacted waters lead to different
treatment results. By design and removal mechanism, physical separation
techniques such as foam fractionation appeared to be more suitable to (pre) treat
heavily contaminated while sorption techniques with the various tested sorbents
were more suitable to treat (polish) cleaner matrices. The table below give a very
generic overview of the applied treatment techniques and materials with some
notes on the applicability of the technique for the tested waters.

5.1.1. Sorption

The overview illustrates that the sorbents were capable of treating both types of
water, but to remove PFAS from heavily contaminated firefighting water clearly
requires much higher sorbent dosages. As initial concentrations are very high, such
reductions are not sufficient to reduce PFAS concentrations to acceptable levels 8.
Therefore, it is not advisable to apply the tested sorptive techniques to FFWW as a
single step treatment, while merely sorption techniques application could
effectively treat PFAS impacted GW. Therefore only impacted GW was used in the
column studies with GAC, FS and CP, while the FFWW was not treated in the column
tests.

Bench top flocculation and membrane studies are conceptually very similar to their
full scale treatment application while batch sorption tests the intrinsic sorption
capacity of the tested sorbents but the test set up is conceptually different from
their full scale treatment application Therefore batch sorption experiment cannot
directly predict the performance in column studies. In addition, one should be
careful in over-interpreting batch sorption and small-scale column studies. Small-
scale column studies are not directly translatable to larger columns or full scale
applications since the smaller grain size of the sorbent and shorter hydraulic
retention might affect sorption kinetics and surface characteristics. Nevertheless,
the results can be considered useful as determination of differences between
sorbents in a relative manner, and are therefore functional in selecting the sorbent
for pilot / full scale testing.

65
 report no. 5/21

5.1.2. Flocculation

The flocculant appeared to be more effective in (pre)treating FFWW than the

sorbents when compared on the basis of the sorbent/reagent mass per L water. The
observed doses that resulted in significant reduction of PFAS concentrations were
in the range of 10 mg/L for impacted GW and 1,000 mg/L for FFWW. In general, the
flocculation appears to be able to efficiently (pre)treat the heavily contaminated
FFWW, while the impacted GW with lower PFAS loads showed incomplete removal.
Potentially this is due to the incomplete flocculation or the incomplete
sedimentation and separation of the flocs, which leads to residues of PFAS in the
aqueous samples. Therefore, floc formation and separation of water and
precipitated material requires extra attention in future testing and full scale
application to optimize performance. Nevertheless the techniques seems to be
suitable to (pre)treat heavily contaminated water such as FFWW, but requires
optimization to reduce aqueous concentrations from less contaminated water to
levels near zero.

5.1.3. Separation by foam or membrane filtration

Two non-sorptive separation strategies were applied, being foam fractionation

(with air and ozone) and membrane separation. Both impacted GW, benzene spiked
GW and FFWW were applied in foam fractionation pilot set up, while only FFWW
was treated by membrane filtration. Similarities and differences were observed
between these techniques. Both techniques had difficulty in treating FFWW,
although due to different reasons. The membrane treatment resulted in membrane
fouling and low transmembrane flow and poorer removal after longer operation,
while the foam fractionation (initially) led to excessive foam formation, hampering
separation. Both these complications were not surprising as the heavily
contaminated FFWW contained high loads of particles and surfactants that lead to
these respective complications.

Separation techniques can be evaluated by two parameters, the removal efficiency

and the recovery. The membrane separation showed 90-95% good initial removal
before fouling but removal dropped significantly with longer operation. The foam
fractionation resulted in removal of 60-90%. When looking at the recovery as
permeate or as treated water, the foam fractionation enables to recover around
98-99% of the treated water while the recovery of the membrane treatment was
lower (~80%), took several days and resulted in dropping treatment performance
with increasing recovered volume.

The membrane treatment leads to a concentrate and treated water. Since the
experiments show that the concentration ratio of filtrate and feed water decreases
to marginal differences with longer operation to reach preset recoveries. The
membrane treatment might not be suitable to treat such dirty matrices such as the
FFWW and result in acceptable removal, recovered water fractions within a short
time. This suggests that some form of pretreatment is required to optimize the
performance.

The treatment performance of foam fractionation was not strongly affected by the
treated matrix, and even seemed to perform slightly better for benzene spiked
impacted GW when compared with the impacted GW without benzene, suggesting
that organic solvents might enhance the foam formation or constitution and thereby
enhancing removal. FFWW showed similar to higher removal rates than impacted
GW, but the difference in the PFAS between the water (and potential oxidation of
precursor PFAS during treatment of FFWW) does not allow direct comparison of the
two types of water. Nevertheless, the results of the benzene spiked GW, and

66
 report no. 5/21

performance of the FFWW suggest that the foam fractionation is able to cope with
heavily contaminated water sources, but requires optimization for the matrix it is
applied to.

For both separation techniques applied to FFWW, the removal is insufficient to

obtain PFAS concentrations that are acceptable for emission, as residues are still in
the (low) mg/L range. The separation treatments are suitable as a pre-treatment
step, but additional treatment is advisable.

5.1.4. PFAS removal in relation to matrix composition and physicochemical

characteristics of the PFAS

Impacted GW contained a series of PFAS. The detected PFAS have carboxylic acid
and sulfonic acid head groups and their carbon chain length varies between 4 and
8. Throughout the experiments it can be observed that (generally) the shorter the
PFAS chain, the lower the removal. In addition, the carboxylic acids often show
lower removal than their sulfonic acid counterparts with the same carbon chain
length. This was for example observed in the batch sorption tests and column
studies (Figure 7, Figure 9) and also, although a little less explicitly presented, in
the membrane study and the batch experiment with flocculant (Figure 10, Figure
12).

It is known that environmental waters contain even shorter chain PFAS, that were
not analysed in this study 21. These compounds will be challenging for all treatment
technologies tested, as they have lower sorption affinities and are better water
soluble. It is expected that they also permeate more easily through membranes due
to their smaller sizes and shorter hydrophobic tails. Furthermore, they are less likely
fully accumulating in foam layers and are more easily out competed by BTEX and
other petrochemicals. We therefore strongly recommend to study a wide spectrum
of PFAS when assessing treatment efficiency in field or in pilot tests in future, as
most likely not the long chain but the short chain PFAS will be the bottleneck for
the treatment efficiency.

5.1.5. Evaluating and combining techniques towards practical application

The results with the different sorption materials, the flocculation tests, the
membrane treatment and the foam separation indicate that their performance
varies under different conditions. We have seen that the sorbents are generally
more effective in rather clean matrices such as impacted GW, while the foam
fractionation and the flocculant appear to be suitable (as well) for more heavily
contaminated matrices. The membrane treatment was not tested for impacted GW,
but results of the FFWW indicate that membrane fouling is a risk, suggesting that
cleaner matrices are preferred (i.e. pre-treatment is preferred). We therefore can
conclude that impacted GW can be treated with one of the sorbents directly, while
the FFWW requires a multi-barrier approach, as the sorbents were able to remove
PFAS but required excessive sorbent dosages (g/L treated water). This is,
technically feasible, but often not desirable from a financial (OPEX) or
environmental (CO2-footprint) point of view, yielding high sorbent uses per year.

Therefore, likely a multi-barrier approach is interesting for some sites and specific
matrices. This means that a general solution is not available and a case-by-case
assessment must be made. An initial treatment removes bulk contamination
including co-contaminants and a polishing treatment that reduces PFAS loads to
acceptable concentrations. Among the techniques applied, presumably either a
combination of foam fractionation and sorbent or flocculant and sorbent appear to
be the most promising for further experimentation, and potential application.

67
 report no. 5/21

Furthermore, reactive treatment technologies were not tested in this study.

However in combined treatment systems these technologies may prove attractive
as well. The optimal (combination of) treatment technologies is therefore
dependent on the treated matrix and the wishes of the end user. In order to remove
PFAS from impacted aqueous phases both effectively (sufficient removal) and
efficiently (costs, practical issues) requires not only the selection of the most
suitable treatment technologies, but also the optimization of treatment conditions.

5.1.6. Lookup table

To get an idea of the advantages and drawbacks of each of the technologies, as

discussed in the previous paragraph, a lookup table has been constructed. This table
gives the advantages and disadvantages of each technology along with the marked
differences and key design parameters and can be used to get an overview of the
findings in this study (Table 12). The table provides a basis to select and apply the
best available treatment technologies to mitigate (further) PFAS contamination.
One should note that the interpretation is based on various experiments and
literature data, in some cases performance of techniques differed between the
tested contaminated waters, so performance was averaged. This table should
support the considerations of technologies, but additional information is required
to enable a balanced evaluation for each specific situation.

68
 report no. 5/21

Table 12: Lookup table on advantages and disadvantages of treatment technologies. Colours ranging from red (-- very poor) via orange (-poor) and yellow
(fair) to green (+, good and ++, very good); n.a. (grey) is not assessed, or the data is unknown.

Performance Pre-treatment Main treatment Operational aspects Costs

Removal Hydraulic Effects of temperature and
Short Long Treatment strategy
mechanisms loading rate Contaminant pressure Energy Maintenance Chemical Waste
Chain chain Turbidity Fn/Mn TOC Contact time Bed lifetime CAPEX OPEX
/filtration loading rate use and personnel use production
PFAS PFAS
velocity

Sorption

Hydrophobic, (pi-
DESOTEC Granular
pi) electrostatic Fixed bed filtration,
Activated Carbon + ++ - - - 20-30 min 10- 15 m/h + + minimal 0 0 ++ + 0 --
and ionic lag/lead configuration
(GAC)
interactions

2 options:
Hydrophobic- ,
1: discontinuous batch 5-10 min,
electrostatic- 2-3
Rembind® (RB) + + with sedimentation - - - >80% unknown + + unknown, expected minimal 0 0 + + 0 unknown
and (an)ionic
2: mixed media filter sorption
interactions
with sand

Fixed bed filtration, unknown, 0 without

Hydrophobic-
Cyclopure DEXSORB+® lag/lead configuration, expected regeneration
and electrostatic - ++ - 0 ++ 3-10 min + 0 unknown, expected minimal 0 -- ++ + 0
(CP) with on-site around 10 - with
interactions
regeneration possible m/h regeneration

unknown, 0 without
Hydrophobic- not clear yet, possible unknown, expected low. Possible
expected regeneration
PolyQA-Osorb® (PQ) and ionic 0 + (a) fixed bed filtration - -- -- 1-5 min 0 n.a. pressure effects on swelling of 0 -- + + 0
around 10 -1 with
interactions with regeneration material
m/h regeneration
Flow-through filter or
Hydrophobic and
FLUORO-SORB® Des+ batch treatment with up to >15
electrostatic + ++ - - 0 2-10 min + + unknown, expected minimal 0 0 ++ + + 0
(FS) sedimentation or m/h
interactions
floatation

Flocculation

Hydrophobic- mixing tank, unknown, expected better

for filtration
PerfluorAd (PFAD) and electrostatic + (b) ++ (b) sedimentation, + 0 + up to 60 min ++ n.a. performance with increasing - - --- - - -
up to 15 m/h
interactions filtration temperature

Other separation
technologies

Increasing temperature has

positive effects on operational
Nanofiltration (c) Size exclusion - + Filter units 0 (d) 0 0 (d) n/a (short) n/a n.a. n.a. - - -- --- -- --
aspects, increased pre-pressure
reduces pump energy

multiple contact Unknown, but higher

Surface active
Foam fractionation tank(s) in series with temperatures negatively affect
characteristics of - + + + + 20-40 min n/a n.a. n.a. --- - - 0 + -
WITH OZONE compressed ozone gas transfer and saturation, this
PFAS
bubbles might have adverse effects

Discontinuous batch Unknown, but higher

Surface active
Foam fractionation contact tank with temperatures negatively affect
characteristics of - 0 + + + 20-40 min n/a n.a. n.a. 0 0 ++ ++ + +
WITHOUT OZONE compressed air gas transfer and saturation, this
PFAS
bubbles might have adverse effects

n.a. not assessed

a) determination of adsorption coefficients was complicated for GW
b) based on removal performance for FFWW
c) initial performance before fouling
d) suspended solids and TOC were higher than operational range stated by vendor

69
 report no. 5/21

6. ACKNOWLEDGEMENTS

The authors would like to thank members of Concawe Special Task Force on Soil and
Groundwater (WQ/STF-33) for their valuable input, thorough and detailed
comments and support during this study. Also the suppliers of materials are
gratefully acknowledged. A lot of suppliers were willing to deliver their products
for the tests at no cost and some even provided technical support in
experimentation. The authors appreciate all the information and feedback that has
been received. A special thanks goes to Prof. dr. Annemarie van Wezel, Eugenie
Troia and previous colleague of them at the Institute for Biodiversity and Ecosystem
Dynamics (IBED) of the University of Amsterdam, Steven Droge, as well as to
colleagues at KWR including Jan Hofman and Frederic Béen for their support during
different stages of this PFAS removal performance tests project. The authors hope
that the results of the executed performance tests as presented in this report are
helpful to tackle the PFAS challenges.

70
 report no. 5/21

7. GLOSSARY

BOD Biological Oxygen Demand

BS Benzene Spiked
BV Bed Volumes
C Carbon
CAS CAS Registry Number
CF crossflows
CL Confidence Limit
COD Chemical Oxygen Demand
CP Cyclopure DEXSORB+®
EBCT Empty Bed Contact Time
F Fluor
FAU Formazin Attenuation Units
FeCl3 Iron(III)chloride
FF Foam fractionation
FFWW Firefighting Wastewater
FS FLUORO-SORB® Des+
FTS Fluorotelomere sulfonic acid
FTU Formazine turbidity unit
GAC Granular Activated Carbon
GW Groundwater
H Hydrogen
HDPE High Density Poly Ethylene
IEX Ion Exchange
Ksw Sorption coefficient
KWR KWR Water Research Institute
LC Liquid Chromatography
2
LC-MS Liquid Chromatography Tandem Mass Spectrometry
MeOH Methanol
MRM Multiple Reaction Monitoring
NC Negative Control
NF NxFiltration
NH4OH Ammonium hydroxide
NOM Natural Organic Matter
NTU Nephelometric Turbidity Units
OCRA Ozofractionative Catalyzed Reagent Addition

71
 report no. 5/21

OECD Organisation for Economic Co-operation and Development

ORP Oxidation-Reduction Potential
PFAD PerfluorAd
PFAS Poly/Per-Fluoroalkyl Substances
PFCAs Perfluorocarboxylic acids
PFOS Perfluorooctanesulfonic acid
PFSAs Perfluorosulfonic acids
PP Polypropylene
PQ PolyQA-Osorb®
PSU Pressure Swing Unit
PTFE Polytetrafluoroethylene
QA Quaternary Amine
RB Rembind®
RO Reverse Osmosis
SAFF Surface Active Foam fractionation
SOMS Swellable Organically Modified Silica's
SPE Solid-Phase Extraction
TCE Trichloroethene
TMP Transmembrane Pressure
TOC Total Organic Carbon
TOP assay Total Oxidizable Precursor assay
TPs Transformation Products
TRL Technology Readiness Level
TZW DVGW-Technologiezentrum Wasser
UvA University of Amsterdam

72
 report no. 5/21

8. REFERENCES

1. Riegel, M.; Egner, S.; Sacher, F. Review of water treatment systems for PFAS
removal; Concawe: Brussels, Belgium, August 2020, 2020.

2. Kabiri, S.; McLaughlin, M. J., Durability of sorption of per- and polyfluorinated alkyl
substances in soils immobilized using common adsorbents: 2. Effects of repeated
leaching, temperature extremes, ionic strength and competing ions. Sci Total
Environ 2021, 766, 144718.

3. Sorengard, M.; Kleja, D. B.; Ahrens, L., Stabilization and solidification remediation
of soil contaminated with poly- and perfluoroalkyl substances (PFASs). J Hazard
Mater 2019, 367, 639-646.

4. Buck, R. C.; Franklin, J.; Berger, U.; Conder, J. M.; Cousins, I. T.; de Voogt, P.;
Jensen, A. A.; Kannan, K.; Mabury, S. A.; van Leeuwen, S. P., Perfluoroalkyl and
polyfluoroalkyl substances in the environment: terminology, classification, and
origins. Integr Environ Assess Manag 2011, 7 (4), 513-41.

5. OECD Toward a new comprehensive global database of per- and polyfluoroalkyl

substances (PFASs): summary report on updating the OECD 2007 list of per- and
polyfluoroalkyl substances (PFASs). 2018; p 24.

6. Australian Government Department of Defence, What are PFAS? [Link],

Ed.

7. Gluge, J.; Scheringer, M.; Cousins, I. T.; DeWitt, J. C.; Goldenman, G.; Herzke, D.;
Lohmann, R.; Ng, C. A.; Trier, X.; Wang, Z., An overview of the uses of per- and
polyfluoroalkyl substances (PFAS). Environ Sci Process Impacts 2020, 22 (12), 2345-
2373.

8. Johnson, M. S.; Buck, R. C.; Cousins, I. T.; Weis, C. P.; Fenton, S. E., Estimating
Environmental Hazard and Risks from Exposure to Per- and Polyfluoroalkyl
Substances (PFASs): Outcome of a SETAC Focused Topic Meeting. Environmental
Toxicology and Chemistry 2020.

9. Cousins, I. T.; Ng, C. A.; Wang, Z.; Scheringer, M., Why is high persistence alone a
major cause of concern? Environmental Science: Processes and Impacts 2019, 21
(5), 781-792.

10. European Commission General Annexes - G. Technology readiness levels (TRL).

[Link]
xes/h2020-wp1415-annex-g-trl_en.pdf.

11. Naidu, R.; Nadebaum, P.; Fang, C.; Cousins, I.; Pennell, K.; Conder, J.; Newell, C.
J.; Longpré, D.; Warner, S.; Crosbie, N. D.; Surapaneni, A.; Bekele, D.; Spiese, R.;
Bradshaw, T.; Slee, D.; Liu, Y.; Qi, F.; Mallavarapu, M.; Duan, L.; McLeod, L.;
Bowman, M.; Richmond, B.; Srivastava, P.; Chadalavada, S.; Umeh, A.; Biswas, B.;
Barclay, A.; Simon, J.; Nathanail, P., Per- and poly-fluoroalkyl substances (PFAS):
Current status and research needs. Environ. Technol. Innov. 2020, 19.

12. Houtz, E. F.; Sedlak, D. L., Oxidative conversion as a means of detecting precursors
to perfluoroalkyl acids in urban runoff. Environmental Science and Technology
2012, 46 (17), 9342-9349.

73
 report no. 5/21

13. Droge, S. T. J.; Hermens, J. L. M., Nonlinear sorption of three alcohol ethoxylates
to marine sediment: A combined Langmuir and linear sorption process?
Environmental Science and Technology 2007, 41 (9), 3192-3198.

14. Droge, S. T. J. A closer look at the sorption behavior of nonionic surfactants in

marine sediment. Utrecht University, Utrecht, 2008.

15. Droge, S. T. J.; Goss, K. U., Development and evaluation of a new sorption model
for organic cations in soil: Contributions from organic matter and clay minerals.
Environmental Science and Technology 2013, 47 (24), 14233-14241.

16. Everett, D. H., Thermodynamics of adsorption from solution. Part 1. - Perfect

systems. Transactions of the Faraday Society 1964, 60, 1803-1813.

17. Everett, D. H., Thermodynamics of adsorption from solution: Part 2. - Imperfect

systems. Transactions of the Faraday Society 1965, 61, 2478-2495.

18. Eichelmann, C., Eignungsprufing PFC saniering. modellstandort Herbertingen:

Karslruher Altlastenseminar, 2017.

19. van Nieuwenhuijzen, A.; Bloks, B.; Essed, A.; de Jong, C. Verkenning technologische
mogelijkheden voor verwijdering van geneesmiddelen uit afvalwater; Stichting
Toegepast Onderzoek Waterbeheer (STOWA): Amersfoort, The Netherlands, 2017.

20. ABS Materials Inc PolyQA-Osorb®: A Superior Adsorbent for PFAS Contaminants.
[Link]
[Link].

21. Gagliano, E.; Sgroi, M.; Falciglia, P. P.; Vagliasindi, F. G. A.; Roccaro, P., Removal
of poly- and perfluoroalkyl substances (PFAS) from water by adsorption: Role of
PFAS chain length, effect of organic matter and challenges in adsorbent
regeneration. Water Research 2020, 171.

22. ALS PFAS Testing. [Link]

products/environmental/water-quality/pfas-testing (accessed 3 March 2021).

23. Hammer, J.; Haftka, J. J. H.; Scherpenisse, P.; Hermens, J. L. M.; de Voogt, P. W.
P., Fragment-based approach to calculate hydrophobicity of anionic and nonionic
surfactants derived from chromatographic retention on a C<inf>18</inf> stationary
phase. Environmental Toxicology and Chemistry 2017, 36 (2), 329-336.

24. Ateia, M.; Maroli, A.; Tharayil, N.; Karanfil, T., The overlooked short- and
ultrashort-chain poly- and perfluorinated substances: A review. Chemosphere 2019,
866-882.

74
 report no. 5/21

APPENDIX 1: SAMPLE PREPARATION AND CHEMICAL ANALYSIS OF PFAS

This appendix provides details of applied analytical methods and data analysis. Table 14 lists
PFAS included in various analysis protocols and regulatory lists. Currently, methods and
protocols are improved and regulation is evolving quickly due to the attention in society for
this group of chemicals.

Sample treatment
Firefighting wastewater (FFWW) and PFAS impacted groundwater (GW) were collected from the
20 L containers for analysis or use in experiments. The aqueous phase was sampled at least 1
day after delivery and storage at 4°C in the dark in order to let solid residues precipitate at the
bottom of the container and stabilize the floating oil layer on the surface of the water. Maximum
storage time was 4 months. The GW samples were used as received, the FFWW samples were
diluted 100-fold with solvent after extraction to enable analysis of high contaminant loads that
were otherwise outside (linear) calibration curves and reduce potential matrix effects.

For the continuously operated column studies, the impacted GW was filtered at KWR one day
after arrival (Friday 19th of June 2020) by a 1 µm cartridge filter to remove particles that could
block the small scale columns, as shown in Figure 21. The filtrated water was stored in a 550 L
stainless steel tank at 20 ̊C, connected to the small scale columns set up.

Figure 21: 1 µm cartridge filters used (top) and unused (bottom) to filtrate the water

Selected samples were oxidized according to the TOP assay according to Houtz and Sedlak 12.
TOP assay is a standardized pre-treatment of water samples or sample extracts designed to
expose underlying PFAS not amenable to standard analysis, see Appendix 7). This assay
illustrates the potential transformation products that are formed from the PFAS contamination
after long residence times in the environment (under oxic conditions) and is indicative for the
fractions of PFAS that are missed by measuring a selected set of PFAS. Perfluorinated
carboxylates and sulfonates are stated to remain intact under the conditions of the assay 22.

Chemical analysis
Within this study, 20 PFAS were selected based on their use and related potential environmental
relevance, their diversity in their physicochemical properties and the available sensitive and
robust analytical techniques at UvA. Table 13 lists the test substances.

75
 report no. 5/21

Table 13: The 20 selected PFAS test substances

Substance Abbreviation Internal standard1

PFSAs
18
Perfluorobutane sulfonate L-PFBS O2 PFHxS
Perfluorohexane sulfonate L-PFHxS
13
Perfluorooctane sulfonate (linear) L-PFOS C8 PFOS
13
Perfluorooctane sulfonate (branched) Br-PFOS C8 PFOS
13
Perfluorodecane sulfonate L-PFDS C8 PFOS
PFCAs
13
Perfluorobutanoic acid PFBA C4 PFBA
13
Perfluoropentanoic acid PFPeA C2 PFPeA
13
Perfluorohexanoic acid PFHxA C2 PFHxA
13
Perfluoroheptanoic acid PFHpA C4 PFHpA
13
Perfluorooctanoic acid PFOA C8 PFOA
13
Perfluorononanoic acid PFNA C9 PFNA
13
Perfluorodecanoic acid PFDA C6 PFDA
13
Perfluoroundacanoic acid PFUnA C7 PFUnA
13
Perfluorododecanoic acid PFDoA C2 PFDoA
13
Perfluorotridecanoic acid PFTrA C2 PFTrA
13
Perfluorotetradecanoic acid PFTeA C2 PFTeA
PF Sulfonamide
13
Perfluorooctane sulfonamide PFOSA C2 PFOSA
Fluorotelomer sulfonates
13
4:2 Fluorotelomer sulfonate 4:2 FTS C2 4:2 FTS
13
6:2 Fluorotelomer sulfonate 6:2 FTS C2 6:2 FTS
13
8:2 Fluorotelomer sulfonate 8:2 FTS C2 8:2 FTS
1 Recoveries were automatically corrected with the appointed internal standard

Preventing cross contamination

Because of the ubiquitous presence of PFAS, care is taken to work in clean (surfactant-free) lab
conditions. Polytetrafluoroethylene (PTFE) materials are not being used in the lab materials and
instruments. Even clothes of the researcher in the lab are selected in order to prevent cross
contamination by, for example, Gore, Gore-Tex fabrics and dirt repelling sprays used to shield
clothing and shoes. Furthermore, due to the tendency of PFAS to sorb to glass surfaces, high
density poly ethylene (HDPE) bottles and tubes are used to handle the PFAS containing samples
and minimize losses by sorption to lab materials.

76
 report no. 5/21

Details of sample preparation and analysis procedure

50 mL of samples stemming from GW and FFWW (tests) were extracted in triplicate using solid-
phase extraction (SPE) using weak anion-exchange (WAX) SPE cartridges containing 60 mg of
sorbent. The cartridges were eluted with 1.2 ml MeOH (0.1% NH4OH). In every batch a procedural
blank, ultrapure water spiked with ~1 ng mass labelled internal standards, was added. Before
SPE elution, deuterated internal standards were added to the eluent to account for matrix
effects during analysis. A volume of 0.5 mL of each extract was diluted with 0.5 mL 0.1% acetic
acid and transferred to polypropylene vials for liquid chromatography analysis (PP-LC vials).
Analysis was performed with high performance liquid chromatography coupled to tandem mass
spectrometry (HPLC-MS/MS) (Shimadzu Prominence 20 XR-AB Sciex 4000 QTrap), with Kinetex
C18 Evo column (100x2.1mm x 2.6 µm) + C18 Evo guard column to protect the separation column.
The chromatographic separation was performed using H2O (2 mM ammonium acetate) and MeOH
(2 mM ammonium acetate) as mobile phases. Analytical blanks and standards were injected in-
between (roughly every 12 samples) and at the end of the sequence. The method was validated
for all analysed PFAS with the exception of the perfluorotelomers that were later added to the
analytical method according to the standard NEN 7777+C1 of the Dutch Royal Normalisation
institute. Further. Quantifications were performed with 15 point calibration lines, standards
with accuracies outside 75-125% were removed from the calibration lines to obtain a minimal
linearity R2 of ≥0.99. The first mass transition (multiple reaction monitoring (MRM)) were used
for quantification, the second mass range transition for confirmation of the identity of the
chemical.

77
 report no. 5/21

Table 14: PFAS target compounds of the UvA method and compounds listed by other
Concawe analyses (STF-33, Analytik PFC & Lastfire) as well as potentially
relevant regulatory lists such as the proposed Drinking Water PFAS list and
the Groundwater Watch list.

type Abbrev. UvA 1 STF-33 Analytik Lastfire 4 DWD 5 GW-

2 PFC 3 Watch 6

Sulfonic acids
C4 PFBS X X x X X
C5 PFPeS
C6 PFHxS X X x X x X
C7 PFHpS X X X
C8 PFOS X X x X x X
C9 PFNS X X
C10 PFDS X X X x
Carboxylic acids
C4 PFBA X X x X X
C5 PFPeA X X x X X
C6 PFHxA X X x X X X
C7 PFHpA X X X X X
C8 PFOA X X x X x X
C9 PFNA X X x X x X
C10 PFDA X X X X X
C11 PFUnA X X X
C12 PFDoA X X X
C13 PFTrA X X
C14 PFTeA X X
Fluorotelomers
C>> 4:2 FTS X X
C>> 6:2 FTS X X
C>> 8:2 FTS X X
Perfluoroalkane
sulfonamides
C>> PFOSA X X X
C>> N-EtFOSA X
C>> N-MeFOSA X
C>> N-EtFOSAA X
C>> N-MeFOSAA X
C>> N-EtFOSE X
C>> N-MeFOSE X
This method is comparable to the EPA 537 method, but has an extended list of target PFAS.
5
In draft version of revised Drinking Water Directive as “list of Sum of PFAS” (most recent version we are aware of as of 3rd July 2019).
6
Groundwater Watch List – “List Facilitating” ([Link]
312a4359676f/First%20List%20facilitating%20Annex%20I%20and%20II%20review%20process%20of%20the%20Groundwater%20Di
rective%20(Endorsed%20V2.1%20-June%202019).pdf)
In gray, PFAS are presented that are not part of the validated method, but can be determined semi-quantitative.

78
 report no. 5/21

Dilution
The FFWW samples are suspected to contain high levels of PFAS residuals and furthermore are
probably heavily contaminated with soot, hydrocarbons and other AFFF constituents and contain
a substantial fraction of sand. Thus, these samples constitute a complex, multiple-phase matrix.
Therefore, the aqueous phase in between particle residues on the bottom of the vessel and
potential floating layers of petrochemical substances will be used for testing, to reflect the
water treated after sedimentation and removal of floating layers. Because of the potential high
concentrations of perfluorinated compounds, as observed during preliminary analysis, a dilution
step is proposed in order to reduce matrix effects, improve the elution rate during extraction
and to prevent the analytes from saturating the detector of the MS2 system. The amount of
dilution will depend on the concentrations of PFAS found in FFWW samples by screening analyses
and Total Oxidizable Precursor (TOP)-assays. A most probable dilution range will be up to 3
orders of magnitude.

Extraction
In order to analyse the PFAS concentrations in the FFWW and GW samples using LC-MS2 an
extraction and purification procedure has to be followed. This procedure makes use of a SPE
column with a weak ion exchange sorbent (WAX) (see Table 15). A blank sample containing only
ultrapure water and a control sample with a reference material (local tap water) is also included
in the procedure. Triplicates will be used throughout the extraction method. Before the SPE
each tested sample is spiked with a 13C internal standard and left overnight to equilibrate at
room temperature. In addition, the control sample is spiked with a 12C spike mix. Subsequently,
the SPE manifold is loaded with the SPE cartridges and conditioned with 4 mL 0.1% NH4OH in
methanol without vacuum. Then the cartridges will be equilibrated using 4 mL ultrapure water
(Milli-Q, Merck) after which the sample is loaded into the cartridge and vacuum (<20 In Hg) will
be applied. After elution of the sample the vacuum pressure will be relieved. Subsequently, the
cartridge will be washed with 2 mL 25 mM acetate buffer with a pH of 4 and 2 mL of methanol.
Then, the cartridge will be dried under vacuum for at least 5 minutes, after which a sample rack
with a 1.5 mL HPLC vial is placed in the manifold and a 0.2 µm filter is connected between the
cartridge and manifold. Extraction takes place using 2 x 600 µL 0.1% NH4OH in methanol. The
captured extract will be diluted 1:1 with 0.1% acetic acid in water after which it is stored at
4 ⁰C until analysis.

Table 15: Used SPE materials

Cartridge Oasis WAX (Waters), 60mg, 3cc

Manifold 20 cartridge positions (Waters)
Sample 50 mL FF water
Conditioning 4 mL 0.1% NH4OH in MeOH
Equilibrating 4 mL ultrapure water (Milli-Q)
Wash 2 mL 25 mM acetate buffer (NH4CH3CO2/CH3COOH) pH 4 & 2 mL MeOH
Elution 2 x 500 µL 0.1% NH4OH in MeOH
Dilution of extract 0.1 % CH3COOH in water

TOP-Assay
The PFAS impacted water is suspected to contain various precursors to PFCAs and PFSAs.
Therefore, the concentrations of various perfluorinated precursors will be analysed following
the total oxidizable precursor (TOP) assay. For this assay, the protocol of Houtz & Sedlak, 2012
will be followed, where the PFAS containing water is subjected to oxidation by adding K2S2O8
and heating in a basic pH environment. This assay relies on hydroxyl radicals (OH), which are
formed with the thermolysis of persulfate in basic pH and which transform the precursors to
PFCAs and PFSAs of related chain lengths. PFCAs and PFSAs themselves remain stable in these
conditions. Thus, the amount of oxidizable precursors could be established by assessing the

79
 report no. 5/21

difference in PFCA and PFSA concentrations between persulfate amended and un-amended
samples. Furthermore, the transformation of FOSA in a spiked blank sample could be used in
order to verify the conversion of PFAS by the TOP-Assay. In order to complete the TOP-Assay and
to obtain reliable results, it is important that the samples are not contaminated with high loads
of organic matter or contain high concentrations of PFAS, that both can hamper the oxidation.
For that reason, the FFWWs are diluted ~2500x to reduce such matrix effects and prevent
incomplete of the substances of interest.

For the TOP-assay, 0.8 g of K2S2O8 and 760 µL of 10N NaOH is added to the 50 mL diluted sample
(60 mM & 150 mM respectively). Subsequently, the samples are placed in a water bath under
constant shaking at 85 ⁰C for six hours, after which they are neutralized using HCl. After the
TOP-assay the samples will be subjected to the aforementioned SPE protocol.

LC-MS2 method
The quantification and detection of the target PFAS substances will be achieved by performing
liquid chromatography tandem mass spectrometry (LC-MS2). Characteristics of the LC-MS2 system
and method can be found in Table 16. A previously validated (following the Dutch NEN7777+C1
guideline) method which has been used in an investigation on PFAS in surface water for the
Dutch Department of Waterways and Public Works (Rijkswaterstaat) will be used to detect and
quantify the PFAS in the samples. Furthermore, quality control of this method has been
performed following the NEN 6603 norm. The mobile phase gradient curve and the analytical
characteristics of the target substances can be found in Figure 22 and Table 17 respectively.

Table 16: LC-MS2 set up.

HPLC Shimadzu LC20-AD

2
MS AB Sciex 4000 QTRAP
Column Phenomenex Kinetex C18 Evo (2.6 µm, 100 x 2.1 mm)
Guard column Phenomenex C18 Evo
Ion source Turbo spray
Eluent A Ultrapure water (Milli-Q), 2% NH4CH3CO2
Eluent B Methanol, 2% NH4CH3CO2
Acquisition MRM mode, negative ESI
Injection volume 5-50 µL
Flow rate 0.3 mL/min
Column Tx 35 ⁰C

Figure 22: Liquid chromatography mobile phase gradient.

80
 report no. 5/21

Table 17: Analytical characteristics PFAS, quantifier (Q3) in bold. Q1 is the parent
molecule mass and Q3 gives the masses of fragments that are used for
identification and quantification.

PFAS Retention time Q1 Q3 (tr1/tr2) Reporting

limit (pg/L)
PFBA 2.6 213 169 145
PFPeA 3.5 313 219 47.6
PFHxA 3.9 315 269/119 58.7
PFHpA 4.3 363 319/169 27.1
PFOA 4.7 413 369/169 18.3
PFNA 5.3 463 419/219 11.1
PFDA 5.7 513 469/219 21.3
PFUnA 6.3 563 519/269 20.2
PFDoA 6.8 613 569/319 23.7
PFTrA 7.3 663 619/369 24.4
PFTeA 7.7 713 669/369 40.1
PFOSA 6.4 498 78/169 25.0
L-PFBS 3.6 299 80/99 12.0
L-PFHxS 4.3 399 80/99 23.3
L-PFOS 5.2 499 80/99 21.2
Br-PFOS 5.1 499 80/99 20.7
L-PFDS 6.2 599 80/99 9.51

QA/QC
 All samples will be extracted and injected once.
 After each 18 injections a Method Blank sample is injected to evaluate carry over and
potential drift in response and separation.
 Method Blanks should show no contamination for all compounds analyzed.
 In every batch a Procedure Control Sample will be included that is subjected to the same
extraction procedures and sample treatment in order to obtain the impact of sample
preparation.
 The results of the Procedure Control Samples will be monitored in a Shewart chart.
 Quantification of all measurements will be performed with a linear ten point calibration
line (with r2 > 0.98 for all analytes).
 Samples will all be quantified within the linear dynamic range (0.07 to 140 pg absolute
injected) of the calibration line.
 Analyte concentrations will be corrected for total procedural recovery of the mass
labeled internal standards.
 The (method) LOQ is defined as the lowest validated spike level meeting the method
performance acceptability criteria (mean recoveries for each representative commodity
in the range 50-130% for the mass labeled internal standards (and 60-130% for the native
compounds) and with a RSD ≤ 20%).

81
 report no. 5/21

 Calibration levels will be used only when complying with the following criterion:
o (i) The nominal concentration of the different levels and the corresponding
concentrations calculated with the linear calibration line (by the HPLC software:
Analyst v1.5.1) of the levels does not deviate more than 40%. Analytes will be
identified and quantified using the additional criteria:
o (ii) Analyte peak area of the sample should be higher than that of the lowest
calibration level (defined by criterion (i)).
o (iii) Analyte peak area of the sample should be lower than that of the highest
calibration level.
o (iv) The peak area ratio of mass transitions 1 and 2 of the analyte in the sample
does not deviate more than ±30% of the average ratio of the same mass
transitions determined for the calibration levels.
 Branched PFOS is quantified using branched PFOS levels with linear PFOS as internal
standard due to the limited availability of a 13C labeled standard for branched PFOS.
 13
C spike injections for each 10 sample injections will be used to check if sensitivity of
the analysis changed during the course of measuring.

Data analysis of sorption and flocculation tests

Assuming equal volumes, temperature, pressure and sufficient contact time, an equilibrium
exists between the amount of sorbent/flocculant and the concentration of PFAS in the liquid.
The main output variable of batch sorption test will therefore be the concentration of PFAS per
volume of liquid in the sample. These equilibrium values (Ce) could be used to calculate the
equilibrium sorption capacity (qe) of the sorbent by the following calculation:

In which V is the volume of liquid, C0 the original concentration in water and m the mass of
sorbent/flocculant. Sorption isotherms are given by the relation between the equilibrium water
concentration (Ce) and the equilibrium sorption capacity (qe). Therefore, the tested parameter
will be m, the mass of used sorbent. The original concentrations (C0) of PFAS in the matrices will
be established by sampling and analysis of the FFWW, and GW. The PFAS mass sorbed in the
sorbent will not be measured. Therefore, the measurement of the sorbent mass must be very
accurate. In case of PFAS that are not part of the standard series, C0-Ce can be replaced with
the responses of the chemical before and after treatment for a semi quantitative analysis of
the qe.

Foam fractionation
For foam fractionation the concentration of PFAS in the liquid before and foam and liquid after
fractionation will be analyzed. There are multiple variables which could be tested for this
technique which still depend on the development of a lab-scale test set up. These variables
could include the operating time or the total volume of treated water.

Nanofiltration
In the nanofiltration set up, the concentration of PFAS before (similar to C0) and after passing
the membrane will be analyzed. In nanofiltration systems, the main parameter is flux (J), which
in the proposed set up will be given by the total volume of treated water (membrane pore area
will be constant). As in the following equation:

Therefore, the test variable in the experiment will be volume of treated water, from which the
flux can be calculated. A relationship between flux and solute concentration after passing the
membrane will be established.

82
 report no. 5/21

APPENDIX 2: DETAILS OF TREATMENT CONCEPTS

Sorption technologies

In sorption technologies the PFAS will bind to sorption sites on the surface of the sorbent. The
compound will distribute between the aqueous phase and the sorbent, to the energetically most
favorable distribution. One can distinguish adsorption and extraction (absorption), where the
first is based on a solid matrix that adsorbs compounds on a surface and the second is based on
a non-aqueous liquid matrix where the compound dissolves in.

Examples of sorbents where the interaction of the sorbent and sorbate is on the material surface
are:
 Activated carbon, a carbon based material with a very large surface area, capable of
adsorbing contaminants. Activated carbon is commonly applied to treat water. It is not
a very specific adsorbent and can be considered as the benchmark technology.
 Ion exchange material, in most cases a chemical based sorbent (resin) containing
functional groups. These functional groups can specifically bind negatively or positively
charged contaminants.
 Complexing agents are sorbents (resins) containing functional groups or pockets that
bind chemicals with specific size, structure and composition.

All these sorbents are limited in the number of sites on the surface. Upon use the material
becomes saturated and at the end of the life-time of the sorbent, PFAS can be desorbed by
competition with other sorbates (e.g. ions, other organic molecules, other PFAS).

The rate of saturation is dependent on multiple parameters. Below some of these influences:
 Flow and contact time;
 The number and concentrations of the PFAS present in the water stream;
 The presence of competing substances (amongst others other contaminants, natural
organic material, iron and manganese);
 Biological activity.

Depending on the structure of the sorption material (surface-volume ratio and pore size),
sorption processes require a certain contact time which determines the dimensions of the
treatment system. Generally, sorption materials are most effective in rather clean matrices,
where competition and fouling are limited. The surfactant properties of the PFAS lead to
concentration on the interface of water and sorbent of absorbents, instead of dissolving in the
sorbent 14, 23. The smaller (short chain) PFAS are more aqueous soluble and have a lower tendency
to sorb and are easily out-competed on sorbents 24.

Saturated (spent) sorption materials either have to be disposed as waste or need regeneration
procedures to be reused.

Physical separation technologies

Liquid separation technologies

Liquid separation techniques separate the waste stream in two new streams. The first stream
(concentrate) is preferably very small in volume and contains all (or most) of the contamination
while the second stream is much larger in volume (almost) free of the contamination and has
nearly all of the volume of water. The separation can be obtained by the use of a physical barrier
such as a membrane or by another treatment concept that provides two phases that differ in
contaminant load.

83
 report no. 5/21

Most common technologies are the membrane based technologies like reversed osmosis and
ultrafiltration. Less known are the ozo fractionation and air fractionation. The latter
technologies are based on the formation of microbubbles that pick up the PFAS from the water
stream which are then concentrated in the produced foam layer. As mentioned above
complexation, flocculation and subsequent coagulation and filtration might also considered
phase separation techniques.

The separated and concentrated waste stream needs further treatment by means incineration
or reactive degradation.

Flocculation technologies

Flocculation is a technology where flocs are formed by agglomeration of suspended solids and
precipitation of coagulants. The primary process is a sorption/complex forming process that
might be classified as a form of sorption while the secondary process the complexes coagulate
and are separated from the bulk aqueous phase by filtration or precipitation. The generated
waste (the separated flocs) needs further treatment by means incineration or reactive
degradation.

Reactive degradation technologies

The most common degradation technology for the degradation of PFAS is high temperature
incineration. PFAS containing waste streams (spent resin, separated concentrates) are in most
cases sent to certified processors. Saturated activated carbon, suited for regeneration is heat
treated at 900 degrees C, the off-gases (containing the PFAS) are treated at higher
temperatures. Several new reactive degradation technologies are under development for the
water based streams. Electrochemical oxidation and sonochemical oxidation are good examples
of these developments, whereas electrochemical oxidation is based on the degradation of PFAS
using specific electrodes. Sonochemical oxidation is based on ultrasound degradation
technology. So far these technologies cannot be considered as ready to market because upscaling
and commercial issues but should be considered as promising.

84
 report no. 5/21

APPENDIX 3: DETAILS OF EXPERIMENTAL SET UP BATCH SORPTION

EXPERIMENTS WITH ADSSORBENTS AND FLOCCULANT

Sorption behaviour of native PFAS present in FFWW and GW was studied measuring sorption
isotherms for five sorbents and removal with one flocculant. The sorbents and flocculants were
added at a wide range of dosages, the dosage window for the flocculant was smaller than for
the sorbents. The range was selected around the recommended dose by the supplier (Table 18).

Table 18: Sorbent and flocculant dosage in batch sorption studies

Sorbent/flocculant Groundwater Groundwater Firefighting

(abbreviation) (GW) benzene spike wastewater (FFWW)
(GW-BS) (>>mg/L)
Dosage sorbent (mg/L)
Desotec Granular Activated
carbon (GAC)
Rembind® (RB) 0, 2, 20, 60, 200, 600, 2000, 20000
FLUORO-SORB® (FS)
PolyQA-Osorb® (PQ)
Cyclopure D+ (CP)
PerfluorAd (PFAD) 0, 1, 10, 20, 30, 70, 150, 1000 0, 100, 250, 500,
1000, 1500, 3000,
10000

For all tested sorption and flocculation technologies batch experiments were performed using
250 mL HDPE bottles containing 200 mL of (benzene spiked) GW or FFWW sample. For the
sorption technologies, a dosing range of seven points covering four orders of magnitude around
the 200 mg/L dosing was set up. All seven doses were tested in triplicate. For the flocculation
experiments, also seven-point dosing ranges was set up, with a triplicate for each dose. For all
sorbents and flocculant negative controls containing the water sample with no sorbent and
blanks containing the sorbent in ultrapure water were tested.
Here, a well-mixed quantity of flocculant was pipetted in 200 mL of the treated water sample.
The flocculation dosing was adapted to the PFAS concentrations of the tested matrix. This was
necessary because increasing the dosage of flocculants affected coagulation and further
separation of water and the coagulated flocs.
For the sorbents the samples were shaken at 90 rpm for seven days on a heavy-duty shaker in a
dark incubation room with a stable temperature of 21 ˚C. For the flocculant the samples were
shaken for at least 30 minutes, after which a small dose (10µL) of ferrous chloride solution (40%
FeCl3) was added in order to further stimulate flocculation and coagulation. After treatment,
the flocs were separated from the matrix by filtration with Whatman 595 1/2 filter paper
(4-7 µm). The filtrate was captured and stored in a refrigerator at 4˚C until SPE and analysis.
Subsequently, for all FFWW samples with high PFAS concentrations aliquots of 1.5 mL were
pipetted into Eppendorf tubes. These were centrifuged in an Eppendorf centrifuge for 10 minutes
at 12000 rpm, after which 500 µL of supernatant was pipetted into 50 mL of ultrapure water.
The resulting samples were stored in a refrigerator at 4˚C until SPE and analysis.

85
 report no. 5/21

For the GW and benzene spiked GW samples, aliquots of 50 mL were taken in 50 mL centrifuge
tubes. Subsequently, these aliquots were centrifuged for 30 minutes at 4000 rpm in a Hettich
benchtop centrifuge. After this, the supernatant was transferred to another 50 mL centrifuge
tube and stored in a refrigerator at 4 ˚C until SPE and analysis.
For the determination of the sorption capacity of the sorbents and flocculant, treatments
(dosages) were selected that resulted in a significant reduction concentrations as compared to
the control (no sorbent).

86
 report no. 5/21

APPENDIX 4: DETAILS OF EXPERIMENTAL SET UP FOAM FRACTIONATION

WITH FIREFIGHTING WASTEWATER

For foam- and ozo- fractionation, a pilot test set up consisting of a column with an air/ozone
bubble supply at the bottom and an extraction mechanism for the foam fraction at the top was
obtained from Evocra and set up at KWR. The water to be treated was recirculated through the
bottom part of the column (volume approx. 16 L), while air or ozone was introduced to the
recirculating water through a venturi tube, in line with the recirculating water. The foam was
collected in the top part of the column, where it collapsed and the concentrate was collected
in a separate sample collector. Foam- and ozo fractionation is a process that requires some
experimentation of operational conditions (recirculation flow speed, air pressure, setting height
of the foam collection tower, total liquid hold up (height of foam/liquid interface)) in order to
obtain stable operation.
Feed water was fed into the top of the column via a peristaltic feed pump. Liquid flow rate was
determined by estimating volume per time run at certain settings, and estimated drop in raw
feed water container volume. Recirculation of water within the column is achieved by a
centrifugal pump, allowing for the measurement of pH and ORP (Oxidation-Reduction Potential)
parameters, as well as the production of vacuum to inject air or ozone into the stream, and
subsequently the column.
Figure 23: Evocra OCRA Foam-and ozo fractionation column assembly with ozone
generator

Due to the inherent restrictions of overall gas flow of the provided ozone generator, there
appeared to be a restriction in overall gas injection when compared to air trials. This may have
contributed to the larger amount of liquid able to be fed into the column for the GW tests, as
the foam layer within the column was much smaller in comparison to the air-only test. This may
have potentially reduced contaminant removal by having a coarser foam (lower surface area)
for contaminant removal.

87
 report no. 5/21

The ozone generator was set at 80% PSU, as there was no significant increase in ozone
concentration above this point. Typical concentration of ozone produced varied from 20 – 25
g/Nm3, with concentrations usually hovering around 22 g/Nm3. ORP target levels for all tests
were as high as possible as this typically allows for the highest removal efficiency.
For the experiments, pH was not adjusted during air only trials as prior experience indicated no
improvement in performance. Ozone trials pH was adjusted by the addition of 90% w/w acetic
acid, intended to create smaller bubbles to increase the contact surface, down to a target pH
of 4.5, however as the reagent dose pump had to be manually controlled this was not always
possible.
The column was flushed in between trials with demineralized water until no foam was generated
and observed in rinse water by visual inspection during recirculation and air injection.
The aqueous phase was sampled as follows. All samples were analysed in triplicate:
 The feed water before treatment (negative control);
 The treated water at the time that no foam was formed anymore (t = 0 = no foam
formation);
 The treated water at time + 1 hour that no foam was formed anymore (t = 1 = no foam
+ 1 hr);
 The concentrated foam (concentrate) was sampled once at the end of the tests.

Note that the subsequent trials with the column are generally considered the first step in
optimizing the equipment for scale up. Further optimization for continuous operation is required
and there is the potential for increased efficiency in both removal and waste volume reduction.

Groundwater without benzene spike

At the start of the foam fractionation test (air only) the column was fed with approximately 7 -
8 L of GW. The frequency drive of the recirculation pump was set at 50 Hz, resulting in a
recirculation flow of approximately 2.5 m3/h. The air was injected freely through the venturi.
At the start of the ozo fractionation test (Ozone) the column was fed with approximately 9 - 10
L of GW. Here also the frequency drive of the recirculation pump was set at 50 Hz, resulting in
a recirculation flow of approximately 2.5 m3/h. The ozone injection was restricted by the
capacity of the ozone destruction device. Therefore, air treatment produced a larger foam layer
in the column when compared to ozone fed trials. During both trials the t = no foam sample, and
t = no foam + 1 hr sample could be obtained.

Groundwater with benzene spike

At the start of the foam fractionation test (air only) the column was fed with approximately 7 -
8 L of GW. The frequency drive of the recirculation pump was set at 50 Hz, resulting in a
recirculation flow of approximately 2.5 m3/h. The air was injected freely through the venturi.
At the start of the ozo fractionation test (Ozone) the column was fed with approximately 12 -
13 L of GW. Here also the frequency drive of the recirculation pump was set at 50 Hz, resulting
in a recirculation flow of approximately 2.5 m3/h. The ozone injection was restricted by the
capacity of the ozone destruction device. Therefore, air treatment produced a larger foam layer
in the column when compared to ozone fed trials. During both trials the t = no foam sample, and
t = no foam + 1 hr sample could be obtained.

Firefighting wastewater
Foam fractionation of FFWW was complicated with air as well as ozone as the matrix resulted in
excessive foam formation. Details on the development of the method can be found in
Appendix 3.

This lead to the following adapted procedure for the foam fractionation test (air only). At the
start of the ozo fractionation test the column was fed with approximately 6 - 7 L of FFWW.

88
 report no. 5/21

To control foam formation the FFWW was dosed with FeCl3 to a concentration of 250 mg/L and
diluted a factor 2 with tap water. Additionally, the frequency drive of the recirculation pump
was reduced to 25 Hz. Despite these adaptations only one sample could be obtained (t = no foam
sample) as longer operation resulted in excessive foam formation hampering further sampling.

With the lessons learned of the air only test, the ozo fractionation (Ozone) test was performed
as follows. At the start of the ozo fractionation test the column was fed with approximately 6 -
7 L of FFWW. To control foam formation the FFWW was dosed with 250 mg/L FeCl3. Additionally,
the frequency drive of the recirculation pump was reduced to 20 Hz. The inherent restriction on
the ozone gas inflow into the column reduced the amount of foaming within the column and
potentially aided the trial. After an additional hour, the t = no foam + 1 hr sample could be
obtained, but the volume of formed foam could not be determined accurately.

Background information and rationale for experimental set up

Foam fractionation for the FFWW was complicated. The first experiment showed to be a very
strong foaming water by a benchtop test. This test involved 10 mL of FFWW and an air stone
placed in a tall 100 mL cylinder and observing the foam produced when air is introduced via the
air stone. The air stone was able to convert approximately 10 mL of FFWW into a structural foam
that does not readily collapse.

Figure 24: Firefighting wastewater before air stone bubble test (1) and after air stone
bubble test for 40 sec air flow (2)

(1) (2)

To control foam formation to acceptable structure and volumes, the FFWW was treated with 250
mg/L FeCl3. As foaming was still very strong the FFWW was diluted with 1:1 ratio of tap water
(rendering FeCl3 concentration 125 mg/L). This dilution was also to expand the available volume
of water for the ozone trial as there was potential to exhaust all available water.

89
 report no. 5/21

Figure 25: FFWW with FeCl3 dosed, continuous air feed by air stone, limited structural
foam accumulation

At the start of the second foam fractionation test the column was fed with approximately 6 - 7
L of FFWW. The frequency drive of the recirculation pump was set at 25 Hz, resulting in a
recirculation flow unable to be detected by the rotameter. Between the second and third
attempt, the head was rinsed to remove the foam within and to allow for more foam to be
generated, this foam was not collected as it would be diluted with water. Before the third
attempt an additional dose of FeCl3 was added to bring the concentration of FeCl3 to 250 mg/L
the FFWW again. The conditions were the same as for the second attempt. At the end of this
attempt, the recirculation pump was started for a brief amount of time to collect a single treated
water sample. At time of sample, ORP and pH were not observed as the recirculation pump was
unable to run long enough for probes to stabilize, however are expected to not deviate
significantly from untreated/raw water values. pH was not adjusted, as further stabilization of
foam (if possible) was not desired. During the last trial only one sample could be obtained due
to excessive foam formation.

Figure 26: Firefighting training wastewater, air foam fractionation, column full of
structural foam

90
 report no. 5/21

Ozo fractionation of firefighting wastewater

Using the lessons from the air foam fractionation tests, the pump frequency was lowered to 20
Hz, and the test was started with a 250 mg/L FeCl3 concentration in the feed water. In addition
the inherent restriction on the ozone gas inflow into the column reduced the amount of foaming
within the column and potentially aided the trial.

At the start of the ozo fractionation test the column was fed with approximately 6 - 7 L of FFWW.
The frequency drive of the recirculation pump was set at 20 Hz, resulting in a recirculation flow
unable to be detected by the rotameter. The rotameter and recirculation line was also full of
foam, and as such the gas-liquid mixture would not be an accurate representation of flow. After
starting the recirculation pump, it was noticed that the foam formed was less stable. This might
be due to the higher FeCl3 concentration (250 vs. 125 mg/L) and/or the use of ozone instead of
air as the ozone might oxidize part of the organic chemicals (carbohydrate surfactants) that
would have given the foam its strength and stability. Similar structural foam entered the
concentrate collection head, the first attempt had to be stopped to protect the ozone destruct
catalyst.

Figure 27: Firefighting training wastewater, ozo fractionation, structural foam present
in foam collection head

Foam destabilization was much faster than previous air test and 30 - 60 minutes was waited in
between before starting the second attempt.

Figure 28: Firefighting wastewater, ozo fractionation, foam section breaking away
from liquid interface

91
 report no. 5/21

Foam within the column collapsed considerably, however foam within the foam collection head
was very stable, but not cleaned out, so this was done manually. Additional foam collected into
the foam collection head, which did not collapse into readily collectible water, and large
sections of foam rose and fell throughout the column and were unable to enter into the foam
collection head. Approximately 1 L of additional water was fed into the column during this test.
After 1 hr, and a large section of foam sticking to the top of the column, and another moving up
and down, a t = no foam sample was collected and a 1 hr timer started. After an additional hour,
the t = no foam + 1 hr sample could be obtained, but the volume of formed foam could not be
determined.

92
 report no. 5/21

APPENDIX 5: DETAILS OF EXPERIMENTAL SET UP MEMBRANE SEPARATION

WITH FIREFIGHTING WASTEWATER

The nanofiltration treatment of FFWW was tested using a Mexplorer test kit of NXFiltration with
an MP pilot module type Mexfil MP025 dNF40. The Mexfil MP025 dNF40 test module is a small
module that contains approximately 100 membrane fibres. Each fibre has an effective length of
approximately 23 cm and an internal diameter of 0.7 mm. The total membrane area of the test
module is 0.05 m2.
Figure 29: Mexplorer Portable test unit

The Standard Test Protocol Nanofiltration Mexplorer (NXF-TR: 17-45-1C) of the technology
provider was followed for the operation of the pilot. The impacted FFWW was fed into the
membrane in crossflow mode. The system was tested using different protocols in order to see
the performance under different membrane setting:

 Protocol 1 - dNF Membrane Evaluation: measurement of the retention of the PFAS at

three different permeate fluxes. This has been done with short lasting tests at a
transmembrane pressure (TMP) of 3, 4 and 5 bar with a crossflow of 80 L/h. The produced
permeate and concentrate were recirculated to the feed tank to represent longer
operation periods;
 Protocol 2 – Influence of crossflow velocity: this has been done with short lasting tests
at crossflows (CF) of 70, 100 and 150 l/h with a TMP of 3 bar. The produced permeate
and concentrate were recirculated to the feed tank to represent longer operation
periods;
 Protocol 3 – Influence of Recovery: A single long term test was done at a TMP of initially
3 bar and a CF of 80 L/h. As the flux decreased due to increasing osmotic pressure the
TMP was increased after 34 hours of operation to 5 bar. The total test ran for 74 hours.
The produced concentrate was recirculated to the feed tank and the produced permeate
was discharged. This resulted in a decrease in the volume of the feed water simulating
the increase in recovery.

93
 report no. 5/21

To obtain a stable operation after each change in test conditions the pilot ran at least 15 minutes
at the specified setting before sampling. Samples of 40 mL were taken for PFAS analyses.
The test conditions and measured parameters during the protocols are given in Table 19, Table
20 and Table 21.
Table 19: Operating conditions during protocol 1 tests

Feed Pressure Crossflow Crossflow velocity Permeate Flux Permeate

(bar) (L/h) (m/s) (L/m2.h) conductivity
(S/cm)
3 80 0.58 6.9 3.11
4 80 0.58 9.6 3.62
5 80 0.58 16.0 3.4

Table 20: Operating conditions during protocol 2 tests

Crossflow Crossflow velocity Feed Pressure Permeate Flux Permeate

(L/h) (m/s) (bar) (l/m2.h) conductivity
(S/cm)
70 0.51 3 6.9 3.39
100 0.72 3 6.9 3.76
150 1.08 3 6.9 3.89

Table 21: Operating conditions during protocol 3 test

Time Crossflow Feed Pressure Permeate Flux Volume feed Simulated

(h) velocity (bar) l/m2.h tank Recovery (%)
(m/s) (L)
0 0.58 3 6.9 10 0
0.25 0.58 3 6.9 10 0
5 0.58 3 6.9 8 20
10 0.58 3 4.8 7 30
29 0.58 3 2.2 4 60
34 0.58 5 n.d. 3.7 63
51 0.58 5 n.d. 2.5 75
57 0.58 5 n.d. 2.2 78
74 0.58 5 n.d. 1.8 82
n.d. = not determined

94
 report no. 5/21

APPENDIX 6: DETAILS ON SET UP SMALL SCALE COLUMN EXPERIMENTS

Three small scale columns with an internal diameter of 6.5 mm were used to simulate a contact
time of 25.2 minutes using the RSSCTs particle size independent diffusion model. To fill the
small scale columns, the 3 different types of sorbents were collected and rinsed over the 90 µm
sieve with ultrapure water (Veolia Elga Corus Analytic) with most of the dust being washed away.
The sieved fraction with particle size 90-106 μm was then transferred to a 1 litre glass beaker
and rinsed several times with ultrapure water to remove the remaining part of the dust.
The following sorbents were used for the column experiments with GW containing PFAS
compounds:

 Desotec Granulated Activated Carbon (GAC);

 FLUORO-SORB® Des2 (FS);
 Cyclopure DEXSORB+® (CP).

All three sorbents were delivered grinded and sieved to a fraction of 90-180 µm. Because of a
possible dry sieving by the supplier, the materials possessed a high amount of dust defined as
particles < 90 µm. After wet sieving and thorough washing a usable amount of 90-180 µm fraction
was collected. For the GAC however, an extra milling and sieving was necessary to obtain enough
material to fill the mini column. Stainless steel tubes were used for the small scale columns
(Figure 30) with an internal diameter 6.5 mm and length up to 20 cm. The sorbent was kept in
the column by means of a stainless steel screen on both sides of the column. By means of vacuum
suction on the bottom of the column, the small scale columns were filled with the 90-106 µm
sorbent, which was poured into the column as a slurry. The small scale column was filled with
sorbent to 15.8 cm column height (5.24 mL). To prevent the sorbents from settling at the column
during the experiment, the remaining part (4.8 cm) was filled with inert sand. After this, the
columns were placed in the set-up and rinsed upwards with demineralized water at 2 L/h for
approximately 2 h. The effluent was visually checked to see e.g. wash out of the sorbent.
Figure 30: Set up with the small scale columns and flow control

The column studies result in breakthrough curves of the different PFAS present in the tested
water. In these curves the concentration is expressed as the fraction of the concentration in the
feed water. The 20% breakthroughs were defined and expressed as the bed volumes the column
has been exposed to. This is indicative for the treatment capacity of the sorption material in
the column for the different PFAS present in the water.

95
 report no. 5/21

Background information

Table 22 lists the modelling data of the small scale columns after conversion from large column
to a small scale column, EBCT stands for Empty Bed Contact Time.
Table 22: Conversion of 25.2 min contact time to the small scale column.

Parameter Large column Small column

Diameter particles (mm) 1.24 0.098 (0.090-0.106)
Height column (mm) 2,000 158
Diameter column (mm) 500 6.5
Water flow (L/h) 935 2.0
Filtration rate (m/h) 2.0 60.3
EBCT (min) 25.2 0.16
Bed volume (L) 393 0.00524

The experiment started the 22nd of June at 14:45h and was stopped the 29th of June at 8:00h,
2020. The small scale columns were fed with the desired flow by means of constant flow
regulators. A constant pre-pressure of approximately 3.5 bar was used for the regulators. This
pressure was necessary to guarantee a constant flow even with a pressure build up in time. In
front of the columns, interchangeable stainless steel 2 µm filters were used to prevent fouling
of the small scale columns. These filters were replaced once during the experiment. The flow
rate of all columns was manually checked daily and adjusted when necessary, but was rather
constant in time. In Figure 31 an overview of the setup is given, with in front a part of the jerry
cans used to transport the water. In Figure 32 a close-up of the sample collector is given.
Figure 31: Overview photos of the mini column set up with the jerry cans, stainless
steel tank, pump and small scale columns

The samples were collected every 4 h by using a sample collector (type KWR v1), see Figure 32.
On a daily base, the samples were removed and stored at 4 ̊C in a cooling chamber. 1 L HDPE
bottles were used to collect and store the water.

96
 report no. 5/21

Figure 32: The sample collector, manufactured by KWR

Influent samples were taken before and after the filtration step, at the beginning and the end
of the sorption experiment and every day after passing a blank (empty) small scale column. The
experiment lasted for approximately 7 days to reach > 40.000 BV (Bed Volumes). In total 29
samples of 1 L per column were taken, with the first one after 1 h, and for a part analyzed at
the Institute for Biodiversity and Ecosystem Dynamics (IBED) laboratory of the University of
Amsterdam. From the 1L delivered samples, 50 mL was collected and spiked with ~1 ng
(absolute) mass labelled internal standards. From every series of samples from the three
sorbents the odd numbered samples were chosen for analysis and all the blanks. The samples
were extracted and analysed as described in Appendix 7.

97
 report no. 5/21

APPENDIX 7: ANALYTICAL METHODS AND RESULTS OF OTHER CONSTITUENTS

Methods used:

· GW: Heavy metals (NEN 5740), incl. Barium and filtration, (AS3000) (volatile) mineral oil,
volatile chlorinated hydrocarbons, (NEN-EN-ISO 17294-2 (2004)) Ca, Fe, Mg, Mn, Na, potassium,
(NEN-ISO 15923-1) Cl, NH4, NO2, (AS3100) NO3, SO4, (NEN-EN-ISO 17294-2 (2004)) dH, (AS 3100)
EC, pH, (NEN-ISO 15923-1) phosphate, (NEN 6578) fluoride, (NEN-EN-ISO 9963-1) bicarbonate,
(NEN 6541) silicate, (NEN-EN 1484 ) TOC, (NEN-EN 872) TSS, (NEN-EN 1899-1) BOD, (NEN 6633+A1
(2006)) COD.
· FFWW: Heavy metals (NEN6953 (ont.cfNEN6961, with ISO17294-2(2004)) incl. filtration, (NEN-
EN-ISO 10301) mineral oil, volatile chlorinated hydrocarbons, (cf NEN6953(ont.cfNEN6961, [Link]
ISO17294-2 (2004))) Ca, Fe, Mg, Mn, Na, potassium, (AS 3100) Cl, NO3, SO4, (NEN-ISO 15923-1)
phosphate, (NEN 6578) fluoride, (NEN-EN-ISO 9963-1) bicarbonate, (NEN 6541) silicate, (NEN-EN
1484) TOC, (NEN 6621) TSS, BOD, (NEN 6633+A1 (2006)) COD.

Results of constituents of impacted GW and FFWW:

In Table 23 the analytical results of generic water characteristics are summarized. Some
compounds of the FFWW were not measured (given as n.m.). Only the liquid phase was analysed,
so any solids, when present in the water, were not analysed.

Table 23: Analytical results of common parameters for the studied groundwater and
firefighting wastewater.

Parameter common Groundwater (GW) Firefighting

wastewater (FFWW)
Conductivity (25°C) (µS/cm) 652 n.m.
pH 7.8 n.m.
Temperature (°C) 19 n.m.
Carbonate (mg/L) <6 <6
Hydrogen carbonate (mg/L) 230 58
Fluoride [F] (mg/L) 0.14 14
Ammonia-N (mg/L) <0.1 n.m.
Chloride (Cl) (mg/L) 62 390
Nitrate-N (mg/L) 0.5 < 3.0
Nitrite-N (mg/L) 0.01 n.m.
Ortho-phosphate (P) (mg/L) 0.05 510
Silicate (SiO2) (mg/L) 10 5.2
Sulphate (mg/L) 63 2,900
Total alkalinity (mmol/L) 2.6 n.m.
Total alkalinity (°dH) 15 n.m.
BOD 5 days (mg/L) <1 370
COD (mg/L) <5 4,500
Suspended solids (NEN-EN 872) (mg/L) 2.3 25
TOC (mg C/L) 1.6 1,500
Calcium (Ca) (µg/l) 86,000 77,000
Ferric (Fe) (µg/L) < 5.0 8,000
Potassium (K) (µg/L) 4,700 6,300
Magnesium (Mg) (µg/L) 12,000 55,000
Manganese (Mn) (µg/L) 2,100 870
Sodium (Na) (µg/L) 32.000 260.000

98
 report no. 5/21

In Table 24 the analytical results of heavy metals are given of the studied water types. It is
shown that the FFWW contained up to several orders of magnitude higher concentrations of most
heavy metals, the only exception is barium (Ba), which is of natural origin. In most cases levels
in GW were below detection limits so accurate ratios between the two water types could not be
determined.

Table 24: Analytical results of heavy metals for the studied groundwater and
firefighting wastewater.

Parameter aromatics Groundwater (GW) Firefighting

wastewater (FFWW)
Barium (Ba) (µg/L) 120 57
Cadmium (Cd) (µg/L) < 0.20 1.3
Cobalt (Co) (µg/L) < 2.0 6.8
Copper (Cu) (µg/L) < 2.0 64
Mercury (Hg) (µg/L) < 0.05 < 0.05
Lead (Pb) (µg/L) < 2.0 12
Molybdenum (Mo) (µg/L) < 2.0 4,2
Nickel (Ni) (µg/L) < 3.0 46
Zinc (Zn) (µg/L) < 10 6.700

In Table 25 the analytical results of aromatics in the studied types of water are given. FFWW
contained at least two orders of magnitude higher amounts of aromatic compounds in
comparison to GW. In most cases levels in GW were below detection limits so accurate ratios
between the two water types could not be determined.

Table 25: Analytical results of aromatics for the studied groundwater and firefighting
wastewater.

Parameter aromatics Groundwater (GW) Firefighting

wastewater (FFWW)
Benzene (µg/L) < 0.20 89
Toluene (µg/L) < 0.20 110
Ethylbenzene (µg/L) < 0.20 12
m,p-Xylene (µg/L) < 0.20 31
ortho-Xylene (µg/L) < 0.10 96
Sum Xylenes (µg/L) 0.21 130
Naphthalene (µg/L) < 0.020 < 1.0
Styrene (µg/L) < 0.20 89

99
 report no. 5/21

In Table 26 the analytical results of chlorinated hydrocarbons and volatile compounds are given.
As can be seen, both water types contained residues of chlorinated hydrocarbons. Levels in
FFWW were 1 to two orders of magnitude higher than in GW. For volatile compounds the
difference was even larger with a difference of 3 or more orders of magnitude between FFWW
and GW. In case of volatile compounds, levels in GW were below detection limits so an accurate
ratio between the two water types could not be determined.

Table 26: Analytical results of several organic contaminants for the studied
groundwater and firefighting wastewater.

Parameter organic contaminants Groundwater (GW) Firefighting

wastewater (FFWW)
Dichloromethane (µg/L) 0.4 < 2.0
Sum cis/trans-1,2-Dichloroethene (µg/L) 0.14 1.4
Sum Dichloroethene (µg/L) 0.21 2.1
Sum Dichloropropanes (µg/L) 0.42 45
Volatile compounds C6-C10 (µg/L) < 10 5.800

Finally, in Table 27 the analytical results are given of mineral oil fractions present in the studied
water types. As can be seen, FFWW contained all measured mineral oil fractions while GW did
not contain any of these fractions above limits of quantification. Since levels in GW were below
detection limits no accurate ratio between the two water types could not be determined.

Table 27: Analytical results of mineral oil fractions for the studied groundwater and
firefighting wastewater.

Parameter mineral oil fractions Groundwater (GW) Firefighting

wastewater (FFWW)
Hydrocarbon fraction C10-C40 (µg/L) < 50 640
Hydrocarbon fraction C10-C12 (µg/L) < 10 35
Hydrocarbon fraction C12-C16 (µg/L) < 10 67
Hydrocarbon fraction C16-C20 (µg/L) < 5.0 78
Hydrocarbon fraction C20-C24 (µg/L) < 5.0 120
Hydrocarbon fraction C24-C28 (µg/L) < 5.0 110
Hydrocarbon fraction C28-C32 (µg/L) < 5.0 120
Hydrocarbon fraction C32-C36 (µg/L) < 5.0 81
Hydrocarbon fraction C36-C40 (µg/L) < 5.0 34

100
 report no. 5/21

APPENDIX 8: SORPTION ISOTHERMS OF BATCH SORPTION TESTS WITH

GROUNDWATER
This appendix provides figures of the sorption isotherms of six PFAS derived from the sorption
experiments of the five sorbents exposed to impacted groundwater (GW). No isotherms for
other constituents are presented as the removal was too high to obtain trustworthy aqueous
concentrations and derived concentrations on the sorbent.

Figure 33: Sorption isotherms of batch sorption studies performed with impacted
groundwater. The presented sorbents were Rembind® (RB), Granulated
Activated Carbon (GAC), PolyQA-Osorb® (PQ), Cyclopure (CP) and FLUORO-
SORB® (FS)

PFBA in GW PFPeA in GW
5 5
RB RB
GAC GAC
PQ PQ
CP CP
FS FS
log (g/kg sorbent)

log (g/kg sorbent)

4 4

3 3

2 2
-3

-2

-1

-3

-2

-1

0
log (g/L) log (g/L)

PFHxA in GW PFHpA in GW
6 6
RB RB
GAC GAC
PQ PQ
CP CP
5 FS 5 FS
log (g/kg sorbent)

log (g/kg sorbent)

4 4

3 3

2 2
-3

-2

-1

-3

-2

-1

log (g/L) log (g/L)

PFOA in GW L-PFOS in GW
7 7
RB RB
GAC GAC
PQ PQ
6 CP 6 CP
FS FS
log (g/kg sorbent)

log (g/kg sorbent)

5 5

4 4

3 3

2 2
-3

-2

-1

-2

-1

log (g/L) log (g/L)

101
 report no. 5/21

APPENDIX 9: SORBENT LOADING RATES

In the batch sorption experiments the sorbents reach high loads of PFAS in the tests. The levels
of detectable PFAS for GW and benzene spiked GW at a 50% reduction of the initial aqueous
concentration fall in the range of 1-60 g/kg sorbent. For FFWW the loading rate is more difficult
to determine since a large fraction of the PFAS in the FFWW is not detectable with the applied
analytical technique. Nevertheless the detectable PFAS sorbed fall within the same range, and
if we assume that a relevant fraction of the PFAS is missed, the loading rate is presumably even
higher. This can be expected since the concentration in the FFWW is, even when non-detected
PFAS are excluded, almost two orders of magnitude higher. The presence of the artificially
spiked benzene does not affect the loading rate of the sorbent by a large extent. The effect of
the complex mixture of other contaminants in the FFWW cannot be quantified since the PFAS
concentrations and the co-contaminants are at a different order of magnitude than for GW.
When the effect of the addition of benzene is studied more carefully, it can be observed that
the effect on the sorption coefficients of the large PFAS is marginal while for the smaller PFAS
(especially PFBA and PFPeA) the sorption seems to be slightly reduced and the relation between
the sorbent dose and the removal seems to flatten (PFBA). Although this was not clearly
reflected in the figures for the total PFAS removal, the reduction of removal and flattening of
curve suggests the following. Benzene competes for sorption sites with the smaller PFAS while
larger PFAS are presumably better sorbed and experience less competition from a small organic
chemical such as benzene.

102
 report no. 5/21

APPENDIX 10: STRUCTURE OF ACTIVATED CARBON AND PFAS SORPTION

All types of activated carbon can be characterized in three categories (or combinations of these
categories). These three categories are defined by pore size. The IUPAC (International Union of
Pure and Applied Chemistry) defined three types of pores present in activated carbon:
Macropores: > 50 nm, Mesopores: 2-50 nm and Micropores: < 2 nm. The final pore distribution of
the granular activated carbon is determined by the product (e.g. coal, coconut) used for the
production of the granular activated carbon and the activation conditions. The pore distribution
with sorption sites determines the sorption capacity, where smaller pores have more sorption
surface area and more capacity. The accessibility of the sorption surface is higher when there
are more macropores that enable diffusion of molecules inside the material. So the optimal
structure is a mixture of pore sizes with sufficient sorption sites. For PFAS constituents like PFOS,
PFOA or Gen-X, the use of macro porous granular activated carbon is recommended. This
phenomena is clearly visible in Figure 34 (provided by Desotec) where the sorption isotherm for
a macropore granular activated carbon Organosorb 10AA) is compared with a micropore granular
activated carbon (Organosorb 10).

Figure 34: Examples of sorption isotherms for PFOS on activated carbon.

103
 report no. 5/21

APPENDIX 11: OVERVIEW OF TREATMENT TECHNOLOGIES FOR GROUNDWATER AND FIREFIGHTING WASTEWATER 1

Evaluation matrix for treatment of groundwater

Technical Reliability of Technical Reaching No need for No need for Impact of Tolerance Impact of Mass Mass No use of No by- Key treatment Treatment
maturity data complexity threshold pre- pre- high organic of high salt high removal removal chemicals products parameter costs
values for treatment treatment matter content Petroleum effectivene effectivene
long-chain for for turbidity content substances ss for Short- ss for Long-
PFAS Fe / Mn content chain PFAS chain PFAS

+ positive - negative o neutral

Operating time
Sorption technologies:
(PFOS)
Many peer
Activated Carbon Full-scale
reviewed data
+ + - + - + - - + + + 40,000 BV +

Pilot- Single peer

MatCare
scale reviewed data
+ + - + - + - No data + + + No data No data

Pilot- Single peer

Rembind® scale reviewed data
+ + - + - + - No data + + + No data No data

IEX without Many peer

regeneration
Full-scale
reviewed data
+ + - + - - - + + + + 150,000 BV +

CustoMem with- Vendor-

out regeneration
Lab-scale
Information
+ No data - + - + - No data + + + No data No data

Cyclodextrin- Pilot- Vendor-

based sorbent scale Information
+ + - + - + - + + + + 6,300 – >55,000 BV No data

Flocculent dosage /
Flocculation technologies:
removal efficiency
Single peer 100 mg/L /
PerfluorAd Full-scale
reviewed data
O +/- +/- + - + - + + - + 90%
No data

Pilot- Vendor- 2 g/L /

InSite
scale Information
O - +/- + - + - No data + - + 99%
No data

104
 report no. 5/21

Technical Reliability of Technical Reaching No Fe / Mn No pre- High High salt Petroleum Short-chain Long-chain No use of No by- Key treatment Treatment
maturity data complexity threshold pre- treatment organic content substances PFAS PFAS chemicals products parameter costs
values treatment for matter
turbidity

+ positive - negative o neutral

Energy demand /
Liquid-liquid separation technologies:
separation factor
Nanofiltration, Full-scale Many peer - + - - + + + + + - + 0.4 kWh/m³ / o
reverse osmosis reviewed data 20%
Foam Pilot-scale Vendor- - + - + + + + + + + +/- 3 – 7 kWh/m³ / o
fractionation Information < 2%
IEX with Pilot-scale Single peer O + - + - - - + + - + 0.01 kWh/m³ / +
regeneration reviewed 0.01 – 0.1 %
CustoMem with data
Vendor- 0.01 kWh/m³ /
Lab-scale O No data - + - + - No data + - + +
regeneration Information 0.01 %
Cyclodextrin- Pilot-scale Vendor- O + - + - + - + + - + 0.01 kWh/m³ / No data
based sorbent Information No data

Distillation Pilot-scale Vendor- - No data - + + + + + + + + 50 kWh/m³ -

Information 20%
Energy demand for
Destruction technologies:
PFAS destruction
Electrochemical Lab-scale Single peer + + - + O + + + + + - 50 kWh/m³ +
degradation reviewed
data
Single peer
Oxidation Lab-scale + - - + - + - + + - - No data No data
reviewed
data
Single peer
Sonochemistry Lab-scale + + - + + + - + + + + 100 kWh/m³ o
reviewed
data
Single peer
UV radiation Lab-scale + - - - - + No data + + + + No data No data
reviewed
data
Distillation & Pilot-scale Vendor- - + - + + + + + + + No data 4,500 kWh/m³ -
Plasma Information
treatment
Nanoscale Lab-scale Single peer - - - + No data No data No data No data + + + No data No data
zerovalent iron reviewed
Water data
Single peer
Full-scale - + + + + + + + + + + 700-2000 €/m³ -
Incineration reviewed
data

105
 report no. 5/21

Evaluation matrix for treatment of firefighting wastewater (+: positive, -: negative, o: neutral)

Technical Reliability of data Technical Reaching No pre- Petroleum Short-chain Long-chain No by-products Key treatment Treatment
maturity complexity threshold treatment for substances PFAS PFAS parameter costs
values turbidity

+ positive - negative o neutral

Sorption technologies: Throughput

Single peer
Activated Carbon Full-scale + + + - - + + No data No data
reviewed data

Single peer
MatCare Lab-scale + No data + - No data + + No data No data
reviewed data

Single peer
Rembind® Pilot-scale + + + - + + + No data No data
reviewed data

IEX without
No data No data + No data + - o - + No data No data
regeneration

CustoMem without
No data No data + No data + - No data + + No data No data
regeneration

Cyclodextrin-based Vendor-
Pilot-scale + + + - + + + No data No data
sorbent Information

Flocculent dosage /
Flocculation technologies:
removal efficiency

Vendor-
PerfluorAd Pilot-scale O +/- + - + + + No data No data
Information

Vendor- 2 g/L /
InSite Pilot-scale O - + - No data + + No data
Information 99%

106
 report no. 5/21

Technical Reliability of Technical Reaching No pre- Petroleum Short-chain Long-chain No by- Key treatment Treatment
maturity data complexity threshold treatment for substances PFAS PFAS products parameter costs
values turbidity

+ positive - negative o neutral

Liquid-liquid separation technologies Energy demand /

separation factor
Nanofiltration, Pilot-scale Single peer - No data - + + + + No data No data
reverse osmosis reviewed data

Foam fractionation Pilot-scale Vendor- - + + + + + +/- 3 – 7 kWh/m³ / O

Information < 2%
IEX with No data No data O No data + - o - + 0.01 kWh/m³ / No data
regeneration No data
CustoMem with No data No data O No data + - No data + + 0.01 kWh/m³ / No data
regeneration No data
Single peer
Cyclodextrin-based reviewed data 0.01 kWh/m³ /
Lab-scale O + + - + + + No data
sorbent No data

Distillation Pilot-scale Vendor- - No data + + + + + 50 kWh/m³ -

Information 20%

Destruction technologies: Energy demand for

PFAS destruction
Electrochemical Lab-scale Single peer + + + + + + - 250 kWh/m³ +
degradation reviewed data

Oxidation Lab-scale Single peer + - + - + + - No data No data

reviewed data

Sonochemistry Lab-scale Single peer + + + - + + + 300 kWh/m³ +

reviewed data

UV radiation Lab-scale Single peer + - - No data + + + No data No data

reviewed data

Distillation & Vendor-

Pilot-scale Information - + + + + + No data 4,500 kWh/m³ -
Plasma treatment

Nanoscale Lab-scale Single peer - - + No data No data + + No data No data

zerovalent iron reviewed data

Water Incineration Full-scale Single peer - + + + + + + 700-2,000 €/m³ -

reviewed data

107
 "

Concawe
Boulevard du Souverain 165
B-1160 Brussels
Belgium

Tel: +32-2-566 91 60
Fax: +32-2-566 91 81
e-mail: info@[Link]
[Link]