membranes

Article
An Efficient Method to Determine Membrane
Molecular Weight Cut-Off Using Fluorescent
Silica Nanoparticles
Mariam Fadel, Yvan Wyart and Philippe Moulin *
Aix Marseille Univ, CNRS, Centrale Marseille, M2P2 UMR 7340, Equipe Procédés Membranaires (EPM),
Europôle de l’Arbois, BP80, Pavillon Laennec, Hall C, 13545 Aix en Provence CEDEX, France;
mariamfadel.mf@gmail.com (M.F.); yvan.wyart@univ-amu.fr (Y.W.)
* Correspondence: philippe.moulin@univ-amu.fr




Received: 31 August 2020; Accepted: 24 September 2020; Published: 1 October 2020


Abstract: Membrane processes have revolutionized many industries because they are more energy
and environmentally friendly than other separation techniques. This initial selection of the membrane
for any application is based on its Molecular Weight Cut-Off (MWCO). However, there is a lack
of a quantitative, liable, and rapid method to determine the MWCO of the membrane. In this study,
a methodology to determine the MWCO, based on the retention of fluorescent silica nanoparticles
(NPs), is presented. Optimized experimental conditions (Transmembrane pressure, filtration duration,
suspension concentration, etc.) have been performed on different membranes MWCO. Filtrations
with suspension of fluorescent NPs of different diameters 70, 100, 200 and 300 nm have been examined.
The NPs sizes were selected to cover a wide range in order to study NPs diameters larger, close
to, and smaller than the membrane pore size. A particle tracking analysis with a nanosight allows
us to calculate the retention curves at all times. The retention rate curves were shifted over the
filtration process at different times due to the fouling. The mechanism of fouling of the retained NPs
explains the determined value of the MWCO. The reliability of this methodology, which presents
a rapid quantitative way to determine the MWCO, is in good agreement with the value given by the
manufacturer. In addition, this methodology gives access to the retention curve and makes it possible
to determine the MWCO as a function of the desired retention rate.

Keywords: molecular weight cut-off; membrane; nanoparticles; retention rate; particle

tracking analysis

1. Introduction
Membrane technologies are showing an increased importance in separation techniques, because
they overcome many constraints linked to conventional processes. The use of these membrane processes
is spreading in many sectors of industrial activities, notably in dairy, food and beverages, and in the
manufacturing of chemicals, as well as in wastewater treatment, allowing for recovery and reuse
of already wasted materials. This action allows these industries to become more environmentally
friendly by decreasing the amount of waste generated, and also more profitable by the value components
being recovered and reused. The main key for such revolution is the proper selection and application
of these commercially produced and applied membranes [1–3]. The initial selection of the membrane for
any application is generally based on the “Molecular Weight Cut-Off” or “MWCO” [2]. A membrane’s
MWCO is defined as the minimum molecular weight of a solute that is 90% retained by the membrane,
according to the French Standard NF X 45-103 [4]. The MWCO assessment of a membrane, facilitates
theoretically the design processes of the engineering membrane systems and thus, ensures their

Membranes 2020, 10, 271; doi:10.3390/membranes10100271 www.mdpi.com/journal/membranes

Membranes 2020, 10, 271 2 of 14

predictable performance of retention [5]. There are a number of problems and limitations regarding
the application and the use of MWCO [6,7]. Despite this, the measurement and comparison of MWCO
curves and the determination of MWCO remain the most practical and universal means used to choose
and differentiate between membranes for different applications, before evaluating a specific membrane
in the real system of interest. Therefore, there is a need for accepted standards, leading to fast and
inexpensive methods for MWCO determination. Thus, a robust, cost-effective, and fast procedure
for the determination of MWCO is necessary for the current and future membrane industry, increasing
applications and uses of membranes for the development of the membrane research area. This therefore
implies a membrane retention study to be carried out in order to obtain a MWCO value of the chosen
membrane. Currently, for ultrafiltration (UF) and microfiltration (MF) membranes, an acceptable,
fast, and cost-effective universal standard method is still unavailable. In the methods currently
available in the literature, a range of compounds with different molecular weights, such as polyethylene
glycols (PEG) [5,8–11], oligostyrenes [12,13], alkanes [14,15], dextrans [8,11,16–20], pesticides [21],
hormones [22], heavy metals [23], dyes from the textile industries [24] and acids [10], were used as the
solute in the study of their membrane retention. Most of these tested methods are expensive and
laborious. For example, four or five separate filtration runs are needed to finally obtain the MWCO
using simple compound solutions. Few rapid, reliable, and relatively inexpensive chromatographic
techniques have been addressed that are capable of separating a relatively cheap mixture of the different
molecular weight compounds to allow a test of a single filtration to be used in the determination
of MWCO [1,5,16,25]. However, there are limitations concerning the application of these techniques.
For instance, the developed oligostyrene [13,26] and the protocols using PEGs [1] are for non-aqueous
solvents. Furthermore, this old method requires the use of relatively expensive reagents and the
precision and robustness of the latter, not yet well proven, in particular by benchmarking against
the membranes available on the market. Thus, It is important to choose the mixture of solute to
characterize the UF and MF membranes well, since blocking the pores of larger molecules on the
surface of the membrane [5,18,27], or membrane fouling in forms of retention of large foulants [28,29],
droplets deformability [30], and thermodynamics filtration resistance in gel/cake layer [30–33] can
potentially selectively impede the transport of different molecular weight solutes, and therefore deflect
the MWCO value from the real value [34]. In most cases, the major factor which does not allow
validation of tests, with a single filtration of several components, is the possibility of concentration
polarization affecting the membrane filtration [34]. Thus, it can result in denaturing the retention of
solutes, and giving apparent retention which is specific to the conditions only to this system of filtration.
This is the reason why continuous single-component filtrations to determine the MWCO are often
chosen [14,29]. Hermia proposed models that explain the main fouling mechanism (s) operating during
filtration [35]. The classical blocking filtrations are described as complete blocking, standard blocking,
intermediate blocking, and cake filtration, as schematized in Figure 1 [36]. The complete blocking
mechanism is described as the foulant particle arriving at the membrane being bigger than the pore wall;
thus, it entirely closes off the membrane pores. Standard blocking is described as the foulant particle
being smaller than the pore wall, which leads to constriction on the membrane pores. Intermediate
blocking is similar to complete blocking in that the pore is closed off at the membrane surface, but
different in that the blocking is caused by more than one particle. Cake filtration refers to the buildup
of foulant particles on the membrane surface due to complete blocking and intermediate blocking.
Membranes 2020, 10, 271 3 of 14
Membranes 2020, 10, x FOR PEER REVIEW 3 of 15

Complete Blocking Standard Blocking Intermediate Blocking Cake Filtration

Figure 1. Schematic representation of various fouling mechanisms by particulate foulants [36].

Figure 1. Schematic representation of various fouling mechanisms by particulate foulants [36].
Several techniques have been used to identify the structure and characterize the foulants
(tracers) in the
Several membrane,
techniques have suchbeenas, used
energy to dispersive
identify the X-ray analysis
structure and [37], high performance
characterize the foulants size
exclusion chromatography (HPSEC) [38,39], UV-absorbance and fluorescence spectroscopy [40,41],
(tracers) in the membrane, such as, energy dispersive X-ray analysis [37], high performance size
3D fluorescence
exclusion spectroscopy
chromatography [41,42], scanning
(HPSEC)[38,39], electron microscopy
UV-absorbance [38,42], and
and fluorescence Fourier transform
spectroscopy[40,41],
infrared spectroscopy (FTIR) [38]. Among all these techniques, fluorescence is increasingly
3D fluorescence spectroscopy [41,42], scanning electron microscopy [38,42], and Fourier used in
transform
the characterization of filtration membranes [43] because it makes it possible to target the used
infrared spectroscopy (FTIR) [38]. Among all these techniques, fluorescence is increasingly desired in
information.
the It acts asofafiltration
characterization contrast agent which allows
membranes the direct,
[43] because reliable
it makes it and simplified
possible identification
to target the desired of
the compounds of interest. The use of fluorescence therefore makes it possible to obtain qualitative
information. It acts as a contrast agent which allows the direct, reliable and simplified identification
information
of the compounds on theofmorphology
interest. The of usea of
membrane
fluorescence through: the makes
therefore differences in density
it possible that qualitative
to obtain it presents,
the determination
information of surface roughness,
on the morphology of a membrane and the size ofthe
through: thedifferences
pores, whether in densityor not
thatfollowed
it presents, by
microscopic analysis of the membrane. The use of defined tracers also makes it possible to consider
the determination of surface roughness, and the size of the pores, whether or not followed by
the effects ofanalysis
microscopic chargesofthat thecan take place
membrane. Thebetween the solutes
use of defined tracersandalso themakes
membrane and can
it possible serve as
to consider
a test of the integrity of the membranes. Fluorescent nanoparticles have many proven advantages:
the effects of charges that can take place between the solutes and the membrane and can serve as a
(1) they have dimensional stability which makes them non-deformable under pressure, (2) they (1)
test of the integrity of the membranes. Fluorescent nanoparticles have many proven advantages: do
not, under
they certain conditions,
have dimensional stabilityaggregate
which makes between themthemselves,
non-deformable (3) they can pressure,
under be manufactured
(2) they in dolarge
not,
quantities
under andconditions,
certain with preciseaggregate
sizes, andbetween
(4) they do not have preferential
themselves, (3) they canadsorption on the surface
be manufactured in large of
the membrane,
quantities and with as inprecise
the case of organic
sizes, and (4)or living
they compounds
do not [44]. Theadsorption
have preferential information oncollected
the surface eitherof
by microscopy
the membrane, as or inbythe
displacement
case of organic techniques
or livingdoes not give [44].
compounds a direct
The indication
information ofcollected
the membrane either
selectivity that one would expect under normal conditions of use of the membrane. However, for the
by microscopy or by displacement techniques does not give a direct indication of the membrane
user, one of
selectivity theone
that main objectives
would expectofunder
the characterization
normal conditions of the membranes
of use remains their
of the membrane. effectiveness
However, for the
for a selected separation. Measuring the retention rate of a solute for a porous membrane is one of the
user, one of the main objectives of the characterization of the membranes remains their effectiveness
means
for available
a selected to facilitate
separation. the selection
Measuring theof a membrane
retention rate offorathe separation
solute of a given
for a porous solute from
membrane a real
is one of
fluid
the (NF-X45-103
means available 1997) [4]. Thethe
to facilitate principle
selection consists of measuring
of a membrane retention ofofa aseries
for the separation givenofsolute
fluorescent
from
aNPs
real(tracers) of different sizes to obtain a selectivity curve as a function of the NP size. Numerous
fluid (NF-X45-103 1997) [4]. The principle consists of measuring the retention of a series of
studies triedNPs
fluorescent to find methods
(tracers) to determine
of different sizes the membrane
to obtain pore size,
a selectivity but until
curve today there
as a function hasNP
of the been no
size.
consistent method.
Numerous studies tried A recent
to findstudy, in 2020,
methods used a mixture
to determine of gold nanoparticles
the membrane pore size, butsolutes in a filtrate
until today there
membrane
has been no to determine
consistent its pore
method. A size
recentqualitatively,
study, in 2020, although
used athe retention
mixture of goldefficiency results solutes
nanoparticles are not
consistent
in a filtratewith 3 out of 5totested
membrane membranes
determine its pore andsizemany questions remain
qualitatively, althoughto be the
answered
retentionto understand
efficiency
results are not consistent
this phenomenon fully, likewith
why 3there out isofhigher
5 tested membranes
retention with smallandgold many questions remain
nanoparticles than with to the
be
answered to understand
larger particles [34]. this phenomenon fully, like why there is higher retention with small gold
nanoparticles
Therefore, thanthewith the larger
objective of thisparticles
work is [34].
to define an efficient method of measuring MWCO to
Therefore, the objective of
provide information for membrane users to assistthis work is to define an efficientselection.
in membrane method of To measuring
achieve thisMWCO objective, to
provide
the MWCO information
of ceramic formembranes
membranewere usersevaluated
to assist in bymembrane
the retention selection.
and countingTo achieve this objective,
of fluorescent silica
the MWCO of (NP)
nanoparticles ceramicusingmembranes were evaluated
particle tracking analysis. by the retention
Description of theand counting of
mechanism onfluorescent
NPs retention silicain
nanoparticles (NP) using particle tracking analysis. Description
the membrane is determined using both NPs in the feed and in the permeate which were collected of the mechanism on NPs retention
in the membrane
versus time usingistangential
determined using both NPs in the feed and in the permeate which were collected
filtration.
versus time using tangential filtration.

2. Materials and Methods

Membranes 2020, 10, 271 4 of 14

2. Materials and Methods

2.1. Membranes
Mono channel ceramic membranes, with tubular shaped geometry, made of oxides or silicon
carbide with high resistance mechanical materials, were used in this study. Membrane M1 and M2 present
0.1 and 0.2 µm molecular weight cut-off (MWCO) according to the manufacturer (Orelis, Salindres,
France). The inner diameter of a channel is 6 mm and the external one is 10 mm. Filtration
experiments were performed with membranes of 25 cm length. These membranes M1 and M2 have,
respectively, an average permeability of 1620 and 3840 L·h−1 ·m−2 ·bar−1 with a standard deviation of
50 L·h−1 ·m−2 ·bar−1 . Under filtration conditions, at neutral pH (7 ± 0.5), the membrane zeta potential is
about −22 mV (constructor data). It is, therefore, of the same sign and intensity as the NP suspensions.
This decreases the probability of NPs adsorption due to the repulsion forces between the NPs and
membrane materials [45].

2.2. Nanoparticles and Analyses

With respect to the membrane pore size, the three NPs sizes used center on the MWCO of
the membrane chosen. For M1- NPs 70, 100 and 200 nm were used and for M2-NPs 100, 200 and
300 nm. Spherical silica nanoparticles were used (Sicastar Red-F, Micromod Partikeltechnologie GmbH,
Rostock, Germany) with diameters of 70 nm (NPs-70), 100 nm (NPs-100), 200 nm (NPs-200) and 300 nm
(NPs-300). They are labeled with rhodamine, a fluorescent dye, covalently bound in the silica matrix.
The rhodamine labeling does not affect NPs adsorption or retention. It has been reported that there
is no variation in terms of flux and retention with addition of 50 mmol of NaCl in feed suspension
compared to filtration without salt [45]. Nanoparticles have a hydrophilic surface with Si-OH end
groups. Sizes were obtained with Zetasizer Nano S (Malvern, England) and with an individual
monitoring particle analyzer NanoSight NS300 (Malvern, England) based on Nanoparticle Tracking
Analysis (NTA). All measurements were performed at 25 ◦ C. Using the Nanosight NS300, the average
size and the size distribution for each NPs suspension were detected (Table 1). Four different NPs
size suspensions were prepared with a high concentration. These high concentrations were selected
to obtain a significant concentration in the permeate side, after filtration, to take into account the
apparatus measurement range (between 107 and 109 part mL−1 ). The pH of a solution can play a role
in the chemical form of fluorophores. A change in pH therefore leads to a distorted quantification
of the target species, so it is important to specify at which pH the measurements are taken. The pH
of the suspensions is equal to 7. The feed concentrations of these suspensions are: 7 × 1010 , 5 × 1010 ,
3 × 1010 and 9 × 109 part mL−1 for NP-70, NP-100, NP-200 and NP-300, respectively. For these high
concentrations, a dilution is necessary before analysis. The main characteristics of NPs suspensions
are summarized in Table 1. Samples were taken from each NPs feed solution and diluted before
determining their size distribution using Zetasizer Nano S. The distribution curves of the four different
NPs size suspensions are presented in Figure 2. The strength of these NPs used is their composition of
solid and structured materials (silica) which make them incompressible and keep their shape under
different operating conditions. The size of NPs will be the only reason why they pass through the
membrane. Considering the zeta potential of NP suspensions (close to or greater than −30 mV) [45]
and the fact that only NP are present in suspensions, the aggregation phenomenon is negligible and
validated by microscopy observations and by Nanosight measurements.
Membranes 2020, 10, 271 5 of 14

Table 1. Physicochemical characteristics of the nanoparticles (NPs) used in suspensions.

Parameter NPs-70 NPs-100 NPs-200 NPs-300

Diameter a
78.5 106.1 173.6 293.8
(nm)
Mode a
66.3 99.9 164.9 288.8
(nm)
Membranes 2020, 10, x FOR PEER REVIEW 5 of 15
Wavelength b
585 585 585 585
a Size average(nm)
and mode values measured by Nanosight NS300.
Feed concentration
7 × 1010 5 × 1010 3 × 1010 9 × 109
(part mL−1 )
ba Detection ranges of emission wavelengths given by manufacturer.
Size average and mode values measured by Nanosight NS300. b Detection ranges of emission wavelengths given
by manufacturer.

25

20
Intensity (%)

15

10

0
0 50 100 150 200 250 300 350 400 450 500 550 600
NP size (nm)
NPs 70 NPs 100 NPs 200 NPs 300

Figure 2. Nanoparticle size distribution curves (NPs) of the initial solutions (feed): NPs 70; 100;
200 and2.300
Figure nm, used insize
Nanoparticle the distribution
filtrations. curves (NPs) of the initial solutions (feed): NPs 70; 100; 200
and 300 nm, used in the filtrations.
2.3. Filtration Experiments
2.3. Filtration
To remove Experiments
the eventual residues before starting the filtration process, the membranes were
flushed with Milli-Q
To remove the eventual water (conductivity
residues before of 0.8 µS·cm−1
starting the) under 1 bar
filtration of transmembrane
process, the membranes pressure
were
(TMP). Then,
flushed a measurement
with Milli-Q of permeability
water (conductivity to ultra-pure
of 0.8 µS·cm−1) under (UP) water
1 bar was carried out. M1pressure
of transmembrane and M2
permeability
(TMP). Then,was found equal to
a measurement of1620 and 3840 L·h
permeability
−1 ·m−2 ·bar−1 , respectively. All filtration experiments
to ultra-pure (UP) water was carried out. M1 and M2
were performed
permeability wasin vertical
found cross
equal flow and
to 1620 filtration mode
3840 L·h −1·musing
−2·bar−1the lab-scale setup
, respectively. presentedexperiments
All filtration in Figure 3.
The module
were performed containing
in verticalthecross
membrane is placed
flow filtration vertically
mode using inthethe filtration
lab-scale system
setup (Figurein3),
presented so that
Figure 3.
filtration is carried out as homogeneously as possible, avoiding preferential
The module containing the membrane is placed vertically in the filtration system (Figure 3), so that deposits. This vertical
configuration
filtration is identical
is carried out astohomogeneously
that used for theasindustrial
possible, applications. Tangentialdeposits.
avoiding preferential filtrationThis
wasvertical
carried
out at constantisflowrate
configuration identical ofto
90that
L·h−1 andfor
used constant TMP of 0.5
the industrial bar ± 0.05 which
applications. was recorded
Tangential filtration(LEO Record,
was carried
KELLER,
out Winterthour,
at constant flowrate Switzerland)
of 90 L·h−1 versus time. The
and constant TMP experiment
of 0.5 bar was performed
± 0.05 which in was an recorded
air-conditioned
(LEO
laboratory
Record, (20 C Winterthour,
KELLER,
◦ ± 1 C) but the
◦ temperatureversus
Switzerland) was followed
time. Thethroughout
experimentthe wasexperiment
performedininorder to
an air-
correct the measured flux at 20 ◦ C in agreement with the variation of water viscosity. Pure pressurized
conditioned laboratory (20 °C ± 1 °C) but the temperature was followed throughout the experiment
airorder
in was connected
to correct the to the tank containing
measured flux at 20 °C theinNPs suspension,
agreement with which was connected
the variation to the module.
of water viscosity. Pure
The feed solution
pressurized air was is aconnected
suspension toof NPs
the tankin containing
Milli-Q water theultrafiltered
NPs suspension, by membrane
which was (Pore size 20 nm,
connected to
ALTEONTM I, Suez Aquasource ® , Péone, France). During the filtration process, permeates are
the module. The feed solution is a suspension of NPs in Milli-Q water ultrafiltered by membrane
collected
(Pore sizeover
20 nm, time, almost every
ALTEONTM 6 to 10
I, Suez s, and recorded
Aquasource ®, Péone, by France).
an electronic
Duringbalance (∆m = ±process,
the filtration 0.01 g)
(Mark Bell,are
permeates Berlin, Germany):
collected the permeate
over time, flux was
almost every 6 tocalculated. The permeate
10 s, and recorded by aniselectronic
collected continuously
balance (Δm
=but by ag)volume
± 0.01 (Mark of approximately
Bell, Berlin, Germany):5 mL which is the minimum
the permeate volume needed
flux was calculated. to make aisnanosight
The permeate collected
analysis. At the
continuously but end
byof each filtration,
a volume the membrane
of approximately 5 mLwas disconnected
which from thevolume
is the minimum moduleneededand a backwash
to make
was performed at 3 bars with a 4 L solution of Citric Acid (68%, Fisher
a nanosight analysis. At the end of each filtration, the membrane was disconnected from the module Scientific, Illkrich, France) of
and a backwash was performed at 3 bars with a 4 L solution of Citric Acid (68%, Fisher Scientific,
Illkrich, France) of pH ~ 1.8, heated at 50 °C to destroy/remove the NP and the membrane fouling.
Then, the membrane was rinsed with Milli-Q water at 3 bars. If the membrane permeability was not
recovered, a new acid backwash was performed. Each filtration experiment (one membrane – one NP
size) was repeated at least 2 times, to ensure the reproducibility of the results.
Membranes 2020, 10, 271 6 of 14

pH ~ 1.8, heated at 50 ◦ C to destroy/remove the NP and the membrane fouling. Then, the membrane
was rinsed with Milli-Q water at 3 bars. If the membrane permeability was not recovered, a new acid
backwash was performed. Each filtration experiment (one membrane – one NP size) was repeated at
least 2 times,
Membranes 2020, to
10, ensure the REVIEW
x FOR PEER reproducibility of the results. 6 of 15

PI
Retentate

FI

PI
Permeate

Pressurized air Balance

Nanoparticles
suspension
Pressurized tank

Figure 3. Schematic diagram of the vertical cross flow filtration set-up.

Figure 3. Schematic diagram of the vertical cross flow filtration set-up.
2.4. Nanoparticle Tracking Analysis Techniques—Nanosight (300)
2.4. Nanoparticle Tracking Analysis Techniques—Nanosight (300)
Permeates at various times, and feed and retentate at final filtration time were analyzed by
Permeates
Nanosight NS300 at“NTA”
various(Malvern,
times, and feed and
England). Thisretentate
technique at isfinal
basedfiltration
on NTAtime were analyzed
(Nanoparticle by
Tracking
Nanosight and
Analysis), NS300the “NTA” (Malvern,
identification of theEngland).
BrownianThis movementtechnique is based on makes
of nanoparticles NTA (Nanoparticle
it possible to
Tracking Analysis), and the identification of the Brownian movement
go back to the size of the latter. A fluorescence mode allows clear and precise identification of nanoparticles makes
of theit
possible to go back to the size of the latter. A fluorescence mode allows
nanoparticles used. This device was therefore used to obtain, in addition to the concentration of theclear and precise identification
of the nanoparticles
flows, used. of
the size distribution This
thedevice was therefore
suspensions used to of
of nanoparticles obtain, in addition
different sizes chosen.to theToconcentration
minimize the
of the flows, the size distribution of the suspensions of nanoparticles
fluorescence fading of NPs during the filtration runs, the permeate collection beaker was wrapped of different sizes chosen.withTo
minimize
aluminumthe foil.fluorescence
The optimal fading
measurementof NPsrange
during the filtration
of NTA is from 10 7
runs,
to 10 9
thepart
permeate collection beaker
mL ; suspensions
−1 were
was wrapped with aluminum foil. The optimal measurement range
diluted for analysis purposes when necessary (feed and retentate essentially). The aggregation, which of NTA is from 10 7 to 109

part.mL −1; suspensions were diluted for analysis purposes when necessary (feed and retentate
could be the cause of a greater retention, was controlled by comparing NPs size in feed and retentate.
essentially). The aggregation,
The results validate which could
that no aggregation be the cause
occurred. Each sampleof a greater retention,consisted
measurement was controlled by
of 3 scans,
comparing NPs size in feed and retentate. The results validate that
and was performed under a calibrated constant continuous flow from a syringe of 1 mL. The flow to no aggregation occurred. Each
sample
be appliedmeasurement
must be strong consisted
enough of 3toscans, and the
increase wasnumber
performed under a counted
of particles calibrated (inconstant
comparison continuous
with a
flow from a syringe of 1 ml. The flow to be applied must be strong enough
static measurement where only the particles which are present in the analysis field are counted), but to increase the number of
particles counted (in comparison with a static measurement where
not too fast, otherwise the Brownian movement cannot be characterized for long enough (the particles only the particles which are
present in thethe
pass through analysis
fields,field
but doarenot
counted),
have timebuttonot
be too fast, otherwise
followed). the Brownian
The particles movement
must therefore cannot
take around
be characterized for long enough (the particles pass through the fields,
10 s to cross the screen, so the flow must be adjusted (the number to enter in the software) to obtain but do not have time to be
followed).
this speed. The particles must therefore take around 10 s to cross the screen, so the flow must be
adjusted (the number toquantitatively
The measurements enter in the software)
present the to average
obtain this speed.
concentration, means (defined as the average
The measurements quantitatively present the average
size of NPs), and mode (the most often value of NPs size) with their relative errors. concentration, means
These(defined
measurementsas the
average size of NPs), and mode (the most often value of NPs size) with their
were repeated at least 2 times to ensure their reproducibility. Importantly, the particle-tracking technique relative errors. These
measurements
employed herewere repeated
provides direct atevidence
least 2 times to ensure
of fouling their reproducibility.
mechanisms under specific Importantly,
operatingthe particle-
conditions.
tracking
This valuable information could benefit the design and optimization of the filtration processesspecific
technique employed here provides direct evidence of fouling mechanisms under [46].
operating conditions. This valuable information could benefit the design and optimization of the
filtration processes [46].

3. Results and Discussion

3.1. Retention Rate

Membranes 2020, 10, 271 7 of 14

3. Results and Discussion

3.1. Retention
Membranes 2020, Rate
10, x FOR PEER REVIEW 7 of 15

3.1.1.After
Retention of Nps
preparing the feed solution, the first step is to validate and optimize the parameters of the
Nanosight N300 analysis
After preparing technique.
the feed Figure
solution, 4 presents
the first the
step is to concentration
validate distribution
and optimize as a function
the parameters of
of the
the NPs size of the prepared feed solution of 5 × 10 10 of NPs 100 nm. The Nanosight measurements
Nanosight N300 analysis technique. Figure 4 presents the concentration distribution as a function of the
showsize
5.02
of×the
1010prepared
part.mLfeed solution of 5 × 1010 of NPs
−1 of average concentration, the 100
average sizeNanosight
of the NPsmeasurements
is equal to 106.1 +/-
NPs nm. The show
0.2 nm, 10
with a mode at 97.6 +/- 1.7 nm, which is compatible with the information
5.02 × 10 part.mL of average concentration, the average size of the NPs is equal to 106.1 ± 0.2 nm,
−1 given by the
supplier.
with a mode at 97.6 ± 1.7 nm, which is compatible with the information given by the supplier.

1,8E+9 9
1.8×10
97.6nm
1,5E+9 9
1.5×10
Concentration (part.mL-1)

9.0×10 9
1,2E+9

6.0×10 9
9,0E+8

6,0E+89
3.0×10

1.2×10 9
3,0E+8

0
0,0E+0
0 50 100 150 200
NP size (nm)

Figure 4. Concentration distribution for each NP size of suspension of NPs100 nm using the Nanosight
N300 analytical technique.
Figure 4. Concentration distribution for each NP size of suspension of NPs100 nm using the
Nanosight
Initial and N300
final analytical
retention technique.
rates are calculated from the concentrations of the first permeate and the
feed, and the last permeate and retentate collected, and analyzed at the end of the filtration, respectively.
Initial and final retention rates are calculated from the concentrations of the first permeate and
Under the same optimized detection parameters of Nanosight, the measurements of the other permeate
the feed, and the last permeate and retentate collected, and analyzed at the end of the filtration,
samples were performed. Figure 5A shows the concentration distribution for each NPs size of the feed
respectively. Under the same optimized detection parameters of Nanosight, the measurements of the
and permeate solutions after 30 min of the filtration of NPs 100 nm suspension with a membrane M2.
other permeate samples were performed. Figure 5A shows the concentration distribution for each
It is obvious that the smallest NPs are completely filtered by the membrane as they are shown in the
NPs size of the feed and permeate solutions after 30 min of the filtration of NPs 100 nm suspension
permeate curve. With the increase in the size of NPs, the concentration difference between the permeate
with a membrane M2. It is obvious that the smallest NPs are completely filtered by the membrane as
and the feed increases. The retention rate [17,22,29] was calculated by the Equation (1), according to
they are shown in the permeate curve. With the increase in the size of NPs, the concentration
the French Standard NF X 45-103 [4] and the variation of the retention rate versus Np size is given in
difference between the permeate and the feed increases. The retention rate [17,22,29] was calculated
Figure 5B.
by the Equation (1), according to the French Standard NF X] 45-103
[permeate
! [4] and the variation of the
t
retention rate versus Np size Retention
is givenRate t (%) =5B.
in Figure (1 − ) × 100 (1)
[Feed]t
where [permeate]t is the solute concentration in the permeate and [Feed]t is the solute concentration in
𝑝𝑒𝑟𝑚𝑒𝑎𝑡𝑒 −1
the feed at the same time 𝑅𝑒𝑡𝑒𝑛𝑡𝑖𝑜𝑛
t. The unit𝑅𝑎𝑡𝑒
of concentration
(%) = (1 − is in part mL ) ×
. With
100 this new methodology,
(1)
𝐹𝑒𝑒𝑑
the retention curve can be drawn for each time, and the effect of membrane fouling on the retention
rate
wherecan[permeate]
be estimated.
t is the solute concentration in the permeate and [Feed]t is the solute concentration

in the feed at the same time t. The unit of concentration is in part.mL−1. With this new methodology,
the retention curve can be drawn for each time, and the effect of membrane fouling on the retention
rate can be estimated.

A) B) M2-NPs100
M2-NPs100
1.8×10 9
2.5×107
2,5E+07 1,8E+09 1.8×10 9
1,8E+09
1.6×10 9
[Permeate] (part.mL-1)

1,6E+09 100 1,6E+09

1.6×10 9
2.0×107
[Feed] (part.mL-1)

2,0E+07
[Feed] (part.mL-1)

1.4×10 9
Retention rate (%)

1,4E+09 1,4E+09
1.4×10 9

1.2×10 9
1,2E+09 80 1.2×10 9
1,2E+09
1,5E+07
1.5×107 1.0×10 9
1,0E+09 9
60 1.0×10
1,0E+09
8.0×10 8
8,0E+08 8
1,0E+07
1.0×107 8.0×10
8,0E+08
6,0E+08
6.0×10 8 40 6.0×10 8
6,0E+08
5.0×106
5,0E+06 4,0E+08
4.0×10 8
4.0×10 8
4,0E+08
2,0E+08
2.0×10 8 20 8
2.0×10
2,0E+08
0,0E+000 00,0E+00 0 00,0E+00
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
NP size (nm) NP size (nm)
Permeate Feed Retention rate at 30 min Feed
 𝑅𝑒𝑡𝑒𝑛𝑡𝑖𝑜𝑛 𝑅𝑎𝑡𝑒 (%) = (1 − ) × 100 (1)
𝐹𝑒𝑒𝑑
where [permeate]t is the solute concentration in the permeate and [Feed]t is the solute concentration
in the feed at the same time t. The unit of concentration is in part.mL−1. With this new methodology,
the retention curve can be drawn for each time, and the effect of membrane fouling on the retention
Membranes 2020, 10, 271 8 of 14
rate can be estimated.

A) B) M2-NPs100
M2-NPs100
1.8×10 9
2.5×107
2,5E+07 1,8E+09 1.8×10 9
1,8E+09
1.6×10 9
[Permeate] (part.mL-1)

1,6E+09 100 1,6E+09

1.6×10 9
2.0×107

[Feed] (part.mL-1)
2,0E+07

[Feed] (part.mL-1)
1.4×10 9

Retention rate (%)

1,4E+09 1,4E+09
1.4×10 9

1.2×10 9
1,2E+09 80 1.2×10 9
1,2E+09
1,5E+07
1.5×107 1.0×10 9
1,0E+09 9
60 1.0×10
1,0E+09
8.0×10 8
8,0E+08 8
1,0E+07
1.0×107 8.0×10
8,0E+08
6,0E+08
6.0×10 8 40 6.0×10 8
6,0E+08
5.0×106
5,0E+06 4,0E+08
4.0×10 8
4.0×10 8
4,0E+08
2,0E+08
2.0×10 8 20 8
2.0×10
2,0E+08
0,0E+000 00,0E+00 0 00,0E+00
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180 200
NP sizeREVIEW
Membranes 2020, 10, x FOR PEER (nm) NP size (nm) 8 of 15
Permeate Feed Retention rate at 30 min Feed

Figure 5. 5.
Figure (A)(A)
Particle size
Particle distribution
size distributionofof
thethe
feed and
feed andthethe
permeate
permeateofofNPs100
NPs100atat3030
min
minofof
thethe
filtration
filtration
process
processwith
withmembrane
membrane M2, and
M2, (B)
and (B)associated
associated retention
retention rate atat3030min
rate minofoffiltration
filtrationofofM2-NPs100
M2-NPs100
[Room
[Room temperature,TMP
temperature, TMP~0.5
~0.5 bar].
bar].

3.1.2.Effect
3.1.2. EffectofofFouling
Foulingononthe
theRetention
RetentionRate
Rate
Filtrationtime
Filtration timeshowed
showedananimpactimpactononthe theretention
retentionrateratecurves,
curves,although
althoughthe thecurves
curveshad hadthethesame
same
shape(Figure
shape (Figure6).6).The
The impact
impact ofofthe thefiltration
filtrationtime timeononthetheretention
retentionrateratewas
wasdetermined
determinedbybyhaving having
filtration last 30 min and collecting the permeates at different times
filtration last 30 min and collecting the permeates at different times during the filtration process.during the filtration process.
MWCOisisequivalent
MWCO equivalentfor for2 2and
and5 min,
5 min, around8080nm
around nmforfor a retentionrate
a retention rateofof 90%,
90%, butbut there
there is is a stronger
a stronger
shiftatat
shift 3030 min.
min. ForForthisthis
time,time,
the the
MWCO MWCO is estimated
is estimated around around
60 nm60fornma for a retention
retention rate ofrate of For
90%. 90%.
For M2-NPs100, the retention and accumulation of NPs-100 in/on
M2-NPs100, the retention and accumulation of NPs-100 in/on the membrane during the passage of the membrane during the passage
of first
the the first
mL mL of permeate
of permeate led led
to anto an increase
increase in membrane
in membrane fouling
fouling andand a modification
a modification of of selectivity.
selectivity.
Therefore, the filtration time was optimized for 2 min to determine the
Therefore, the filtration time was optimized for 2 min to determine the selectivity of the membrane. selectivity of the membrane.
Thisisiscompatible
This compatiblewithwiththetheduration
durationused usedininthe theprevious
previousmethods
methodswhichwhichwere werebased
basedonona ashort
shorttime
time
filtrationexperiment
filtration experimentfor forthethedetermination
determinationofofthe themembrane
membranepore poresize
size[34].
[34].This
This time(2(2min)
time min)is isa a
compromise between having a sufficient volume of permeate for
compromise between having a sufficient volume of permeate for Nanosight analysis and having Nanosight analysis and havinga a
minimum of fouling, in order to limit its impact on the determination
minimum of fouling, in order to limit its impact on the determination of the MWCO. of the MWCO.

M2-NPs100
120
time
100
Retention Rate (%)

80

60

40

20

0
0 20 40 60 80 100 120
NP size (nm)

t=30min t=5min t=2min

Figure 6. Retention rate of NPs100 in function of NPs size for the filtration with membrane M2 over
Figure 6. Retention rate of NPs100 in function of NPs size for the filtration with membrane M2 over
the time realized [TMP 0.5 bar].
the time realized [TMP 0.5 bar].
3.1.3. Influence of Nps Size on Retention Rate
3.1.3. Influence of NPs Size on Retention Rate
As shown in previous studies [45], the size of NPs has an impact on their retention by the
As shownSince
membrane. in previous studies
there is an [45],
interest in the
the size of NPsofhas
selectivity the an impact onthe
membrane, their retention
impact of NPsbysize
theof
membrane.
larger, closeSince there
to, and is anthan
smaller interest in the selectivity
the membrane of tested.
pore, was the membrane, the impact
With the two of NPs
membranes M1size
and of
M2,
larger, close to, and smaller than the membrane pore, was tested. With the two membranes M1
the retention rate curves plotted against the size of NPs decreased (Figure 7). For the membrane M1, and
M2, the retention rate curves plotted against the size of NPs decreased (Figure 7). For the membrane
M1, the retention of NPs presented a MWCO at 58.5–79.5 and 110 nm with NPs70-100-200,
respectively. In general, small particles penetrate deeper and faster into the membrane [28,29]. Thus,
for each NP size tested, the determination of MWCO is affected by the smallest NPs which are
blocked in the membrane. So we can deduce that the larger the chosen size, the closer we get to a
Membranes 2020, 10, 271 9 of 14

the retention of NPs presented a MWCO at 58.5–79.5 and 110 nm with NPs70-100-200, respectively.
In general, small particles penetrate deeper and faster into the membrane [28,29]. Thus, for each
NP size tested, the determination of MWCO is affected by the smallest NPs which are blocked in
the membrane. So we can deduce that the larger the chosen size, the closer we get to a satisfactory
determination of the MWCO as evidenced by the agreement with the manufacturer’s data and a study
in the literature [34]. The same conclusion is obtained for the membrane M2, where the MWCO of
the membrane was estimated at 85, 121.5 and 181.5 nm from the retention curves of NPs100-200-300,
respectively. This small difference between the value determined by this method and the value quoted
by the manufacturers
Membranes can REVIEW
2020, 10, x FOR PEER be addressed by the differences in methodology and test conditions 9 of[47].
15
Following the shape of the retention curves (Figure 7) within the different experiments performed,
they show
particle that theyof
distribution follow
NPs. the same
These path show
results with that of the increase
the consistency in the
of this size particle
method distribution
performed in two
of NPs. These results show the consistency of this method performed in
different membranes and each with three different sizes of NPs feed solutions. two different membranes and
each with three different sizes of NPs feed solutions.

Membrane:M1
A) 9
1.8×10
1,8E+09
100 1.6×10 9
1,6E+09
1.4×10 9

[NPs] (part.mL-1)
1,4E+09
Retention rate (%)

80 9
1.2×10
1,2E+09
1.0×10 9
60 1,0E+09
8.0×10 8
8,0E+08
40 6,0E+08
6.0×10 8

4,0E+08
4.0×10 8
20
2,0E+08
2.0×10 8

0 00,0E+00
0 50 100 150 200 250 300 350
NP size (nm)
NPs200-retention rate NPs100-retention rate
NPs70-retention rate NPs200
NPs100 NPs70

Membrane: M2 9
1.8×10
1,8E+09
B) 100 1.6×10 9
1,6E+09
1.4×10 9
1,4E+09
Retention Rate (%)

[NPs] (part.mL-1)

80 9
1.2×10
1,2E+09
1.0×10 9
60 1,0E+09
8.0×10 8
8,0E+08
40 6.0×10 8
6,0E+08
4.0×10 8
4,0E+08
20 8
2.0×10
2,0E+08
0 00,0E+00
0 50 100 150 200 250 300 350 400 450
NP size (nm)
NPs300 -retention rate NPs200-retention rate
NPs100-retention rate NPs300
NPs200 NPs100

Figure 7. Retention rates of the different size of NPs with two membranes: (A) Membrane M1 with the
Figure 7. Retention rates of the different size of NPs with two membranes: (A) Membrane M1 with
initial suspension solution of NPs 200, NPs 100 and NPs 70 nm, and (B) Membrane M2 with the initial
the initial suspension solution of NPs 200, NPs 100 and NPs 70 nm, and (B) Membrane M2 with the
suspension solution of NPs 300, NPs200 and NPs100 nm. [TMP ~ 0.5 bar, room temperature].
initial suspension solution of NPs 300, NPs200 and NPs100 nm. [TMP ~ 0.5 bar, room temperature].

To ensure the robustness of this methodology and thus the reproducibility of the results, the
experiments were repeated at the same operating conditions. Figure 8 represents the retention rate
of the NPs100 suspension filtrated with membrane M2 after 2min of the filtration process. The two
Membranes 2020, 10, 271 10 of 14

To ensure the robustness of this methodology and thus the reproducibility of the results, the
experiments were repeated at the same operating conditions. Figure 8 represents the retention rate
of the NPs100 suspension filtrated with membrane M2 after 2min of the filtration process. The two
curves of retention rate from the repeated experiments are almost identical, where the difference10ofofthe
Membranes 2020, 10, x FOR PEER REVIEW 15
MWCO value is ±0.6% between the two performed experiments.

M2-NPs100
120

100
Retention Rate (%)

80

60

40

20

0
50 60 70 80 90 100
NP size (nm)

1st experiment 2nd experiment

Figure 8. Retention rates of NPs 100 filtrated with membrane M2 of the two performed experiments:
1st and 2nd experiments carried out at same operating conditions.
Figure 8. Retention rates of NPs 100 filtrated with membrane M2 of the two performed experiments:
3.2. Fouling Mechanism
1st and 2nd and Mwco
experiments carriedMembrane
out at same operating conditions.
For each filtration run, the experimental data were plotted and compared to the linear forms of
3.2.
four foulingMechanism
Fouling models (cake andfiltration,
MWCO Membrane
intermediate blocking, standard blocking and complete blocking)
by calculating the correlation
For each filtration run, the coefficients
experimental (R2 )data
values were [23,29].
plottedIt was
and decided
compared to to
target relatively
the linear forms high
of
R 2 values (up to 0.95) in order to identify the most appropriate fouling model. Results showed that
four fouling models (cake filtration, intermediate blocking, standard blocking and complete blocking)
thecalculating
by permeate flux decreased coefficients
the correlation during the filtration
(R2) values with all NPs
[23,29]. It wassize, mainly
decided todue
target to relatively
cake filtrationhigh
fouling model 2
(R0.95)= 0.99 in most of the cases) as shown in (Figure 9). Whatever the size of the NPs
R 2 values (up to in order to identify the most appropriate fouling model. Results showed that
and the membrane molecular weight cutoff tested, at the end of filtration,
the permeate flux decreased during the filtration with all NPs size, mainly due to cake filtration the cake filtration appears.
This shows
fouling model very(R2clearly
= 0.99 in that theof
most determination
the cases) as shownof the molecular
in (Figure weight cut-offthe
9). Whatever is size
affected
of the byNPsthe
duration of the filtration and the quantity of NPs retained. When the size
and the membrane molecular weight cutoff tested, at the end of filtration, the cake filtration appears. of the NPs used is very small,
in theshows
This first stages
very of filtration
clearly that athe
standard blockingof
determination appears logicallyweight
the molecular with penetration
cut-off is affectedof the NPs byinto
the
the membrane
duration of the
Membranes and
2020, 10, then
filtration
271 a and
cakethe
filtration
quantityis formed. Conversely
of NPs retained. during
When thethe
sizefiltration
of the NPs of the largest
used
11 of is NP
14 very
sizes, ain
small, complete
Membranes
the first2020, blocking
10, x FOR
stages thenREVIEW
PEER a cake
of filtration filtrationblocking
a standard appearsappearsfrom thelogically
first minutes of filtration of
with penetration [46].
11 the
of 15NPs

into the membrane and then a cake filtration is formed. Conversely during the filtration of the largest
A A-i A-ii
M1-NPs200 M1-NPs200:Cake Filtration
NP sizes, a complete
1.8 blocking then a cake 18
filtration appears
M1-NPs200:Complete Blocking from 1.75
the first minutes of filtration [46].
1.7
15 1.7
1.6 R² = 0.9798 R² = 0.9921
1.65
t/V (s.mL-1 )

1.5 12
t/V (s.mL-1)
Q (L.h-1)

1.4 1.6
9
1.3 1.55
1.2 6 1.5
1.1 3 1.45
1 0 1.4
0 30 V(mL) 60 90 30 50 70 90
0 10 20 30 40
Experimental
MWCO100-NPs200 V(mL) V(mL)

Experimental
MWCO100-NPs200 Model Experimental
MWCO100-NPs200 Model

B B-i B-ii
M2 -NPs300 M2-NPs300:Complete Blocking M2-NPs300: Cake Filtration
0.75 30 0.72

0.7 25 0.7 R² = 0.9931

R² = 0.9772
t/V (s.mL-1)

0.68
t/V (s.mL-1 )

0.65 20
Q (L.h-1)

15 0.66
0.6
10 0.64
0.55
5 0.62
0.5 0 0.6
0 50 100 150 200 0 20 40 60 60 90 120 150 180
V(mL) V(mL) V(mL)
Experimental
MWCO 200-NPs300 MWCO 200-NPs300
Experimental Model Experimental
MWCO 200-NPs300 Model

Figure 9. Cont.
C C-i C-ii
M1-NPs100 M1-NPs100: Standard Blocking M1-NPs100: Cake Filtration
1.4 1.1 1.4
R² = 0.9898
R² = 0.9937
1.3 1.3
t/V (s.mL-1 )

t/V (s.mL-1)

1.05
t/V (s.mL-1)

1.2 1.2
1
1.1 1.1

0.95 1
1
 0.7 25 0.7 R² = 0.9931
R² = 0.9772

t/V (s.mL-1)
0.68

t/V (s.mL-1 )
0.65 20

Q (L.h-1)
15 0.66
0.6
10 0.64
0.55
5 0.62
0.5 0 0.6
0 50 100 150 200 0 20 40 60 60 90 120 150 180
V(mL) V(mL) V(mL)
Membranes 2020, 10, 271 11 of 14
Experimental
MWCO 200-NPs300 MWCO 200-NPs300
Experimental Model Experimental
MWCO 200-NPs300 Model

C C-i C-ii
M1-NPs100 M1-NPs100: Standard Blocking M1-NPs100: Cake Filtration
1.4 1.1 1.4
R² = 0.9898
R² = 0.9937
1.3 1.3

t/V (s.mL-1 )

t/V (s.mL-1)
1.05
t/V (s.mL-1)

1.2 1.2
1
1.1 1.1

0.95 1
1 0 10 20 30 40 50
0 20 40
60 80 100 120 t(s) 40 60 80 100 120
V(mL)
V(mL) MWCO100-NPs100
Experimental Model
MWCO100-NPs100
Experimental Model
MWCO100-NPs100
Experimental

D D-i D-ii
M2-NPs200: Standard Blocking
M2-NPs200 M2- NPs200: Cake Filtration
1.3 0.9
1.3
1.1
1.1
t/V (s.mL-1)

t/V (s.mL-1)
0.7

t/V (s.mL-1 )
0.9
0.9
0.7 R² = 0.9623
0.5 0.7 R² = 0.9879
0.5
0.5
0.3 0.3
0.3
0 20 40 60 80 100 120 0 5 1015 20 25 30
0 50 100 150
V(mL) t (s) V(mL)
MWCO200-NPs200
Experimental Model MWCO200-NPs200
Experimental Model
Experimental
MWCO200-NPs200

E E-i E-ii
M1-NPs70 M1-NPs70: standard filtration M1-NPs70: Cake Filtration
1.4 1.35
1.3
1.3
1.2
1.2 t/V (s.mL-1 ) 1.3
t/V (s.mL-1)
t/V (s.mL-1 )

1.1 1.1 R² = 0.9593

1.25 R² = 0.9802
1
1
0.9
0.9 1.2
0.8 0 10 20 30 40 50 30 60 90 120
0 30 60 90 120 V(mL)
t(s)
V(mL)
MWCO100 -NPs70 MWCO100
Experimental
-NPs70 Model Experimental
MWCO100 -NPs70 Model
Experimental

F F-i M2-NPs100: Standard blocking F-ii

M2-NPs100 M2-NPs100: Cake Filtration
1.1 1 1.05
1.05 1.04
0.95
1 R² = 0.9983 1.03 R² = 0.9888
t/V(s.mL-1 )

t/V(s.mL-1)
t/V(s.mL-1)

0.95 0.9 1.02

0.9 1.01
0.85 0.85
1
0.8 0.99
0 50 100 150 0.8
V(mL) 40 60 80 100 120 140
0 10 20 30 40 50 V(mL)
t(s)
Experimental
MWCO200-NPs100 MWCO200-NPs100
Experimental Model
Experimental
MWCO200-NPs100 Model

Figure 9. Fouling models found for NPs filtrations of (A) NPs 200 with membrane M1, (B) NPs 300 nm
Figure 9. Fouling models found for NPs filtrations of (A) NPs 200 with membrane M1, (B) NPs 300 nm
with membrane M2, (C) NPs 100 with membrane M1, (D) NPs 200 nm with membrane M2, (E) NPs
with membrane M2, (C) NPs 100 with membrane M1, (D) NPs 200 nm with membrane M2, (E) NPs
70 with membrane M1, and (F) NPs100 with membrane M2. The experiments are performed at same
70 with membrane M1, and (F) NPs100 with membrane M2. The experiments are performed at same
experimental conditions [TMP ~ 0.5 bar, room temperature].
experimental conditions [TMP ~ 0.5 bar, room temperature].
4. Conclusions
In this feasibility study, a new and fast method is presented to determine the molecular weight
cut off (MWCO) of a membrane based on the retention of fluorescent NPs. Two Membranes—M1
and M2—of different pore sizes have been investigated. The reproducibility of the experiments was
validated. The retention rates of NPs with sizes larger, closer to, and smaller than the membrane
pore size given by the manufacturer at optimized experimental conditions were studied. It was
shown that the retention of NPs changed over the filtration process due to the membrane fouling: as
expected, a shift of the MWCO value is observed whatever the experiment after a certain duration of
the filtration process. The membrane fouling was studied as a function of the NP size with the Hermia
model. From a standard to a cake fouling modelling, the results are in agreement with the NP size.
So, the selectivity of the membrane should be detected at the first few minutes of the filtration process.
Membranes 2020, 10, 271 12 of 14

The membranes used presented good retention rates curves with different sizes of NPs, increasing
gradually until they formed a plateau at 100% of retention. This allows determining the MWCO at a
90% retention rate according to the French Standard NF X 45-103 although another could be selected.
A MWCO of 181.5 nm has been determined for Membrane M2 with NPs of 300 nm, and 110 nm for the
Membrane M1 with the NPs of 200 nm, close to the values given by the supplier (200 and 100 nm,
respectively). With this new methodology, it is possible to obtain a continuum of information of the
retention rates versus the NP sizes, and as a function of the membrane application it is possible to
obtain the retention rate for a specific molecule. These NPs can later be destroyed with an acid wash,
to recover the permeability of the membrane. The robustness of this methodology is in progress and
will be the subject of a forthcoming paper.

Author Contributions: Conceptualization and methodology (P.M., Y.W.), formal analysis (P.M., Y.W., M.F.);
investigation (M.F.), original draft preparation (M.F.); writing—review and editing, (P.M., Y.W., M.F.); supervision
(P.M.); project administration (P.M.); funding acquisition (P.M.) All authors have read and agreed to the published
version of the manuscript.
Funding: The project leading to this publication has received funding from Excellence Initiative of Aix-Marseille
University-A*MIDEX, a French Investissements d’Avenir programme. It has been carried out in the framework of
the Labex MEC.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Li, X.; Monsuur, F.; Denoulet, B.; Dobrak, A.; Vandezande, P.; Vankelecom, I.F.J. Evaporative light scattering
detector: Toward a general molecular weight cutoff characterization of nanofiltration membranes. Anal. Chem.
2009, 81, 1801–1809. [CrossRef] [PubMed]
2. Razdan, U.; Joshi, S.V.; Shah, V.J. Novel membrane processes for separation of organics. Curr. Sci.
2003, 85, 761–771.
3. Alam, J.; Dass, L.A.; Alhoshan, M.S.; Ghasemi, M.; Mohammad, A.W. Development of polyaniline-modified
polysulfone nanocomposite membrane. Appl. Water Sci. 2012, 2, 37–46. [CrossRef]
4. AFNOR. Norme Française NFX45-103; French Association of Normalization: Paris La Défense, France, 1997.
5. Tam, C.M.; Tremblay, A.Y. Membrane pore characterization—Comparison between single and
multicomponent solute probe techniques. J. Membr. Sci. 1991, 57, 271–287. [CrossRef]
6. Meireles, M.; Bessieres, A.; Rogissart, I.; Aimar, P.; Sanchez, V. An appropriate molecular size parameter for
porous membranes calibration. J. Membr. Sci. 1995, 103, 105–115. [CrossRef]
7. Patterson, D.A.; Yen Lau, L.; Roengpithya, C.; Gibbins, E.J.; Livingston, A.G. Membrane selectivity in the
organic solvent nanofiltration of trialkylamine bases. Desalination 2008, 218, 248–256. [CrossRef]
8. Sarbolouki, M.N. A General Diagram for Estimating Pore Size of Ultrafiltration and Reverse Osmosis
Membranes. Sep. Sci. Technol. 1982, 17, 381–386. [CrossRef]
9. Escott, R.E.A.; Mortimer, N. Analysis of polyethylene glycol and derivatives by high-performance liquid
chromatography using elevated temperatures and low-wavelength ultraviolet detection, and supercritical
fluid chromatography. J. Chromatogr. A 1991, 553, 423–432. [CrossRef]
10. Hilal, N.; Al-Abri, M.; Al-Hinai, H.; Abu-Arabi, M. Characterization and retention of NF membranes using
PEG, HS and polyelectrolytes. Desalination 2008, 221, 284–293. [CrossRef]
11. Morão, A.; Pessoa de Amorim, M.T.; Lopes, A.; Escobar, I.; Queiroz, J.A. Characterisation of ultrafiltration
and nanofiltration membranes from rejections of neutral reference solutes using a model of asymmetric
pores. J. Membr. Sci. 2008, 319, 64–75. [CrossRef]
12. Sairam, M.; Nataraj, S.K.; Aminabhavi, T.M.; Roy, S.; Madhusoodana, C.D. Polyaniline membranes for
separation and purification of gases, liquids, and electrolyte solutions. Sep. Purif. Rev. 2006, 35, 249–283.
[CrossRef]
13. See Toh, Y.H.; Loh, X.X.; Li, K.; Bismarck, A.; Livingston, A.G. In search of a standard method for the
characterisation of organic solvent nanofiltration membranes. J. Membr. Sci. 2007, 291, 120–125. [CrossRef]
Membranes 2020, 10, 271 13 of 14

14. See-Toh, Y.H.; Ferreira, F.C.; Livingston, A.G. The influence of membrane formation parameters on the
functional performance of organic solvent nanofiltration membranes. J. Membr. Sci. 2007, 299, 236–250.
[CrossRef]
15. See-Toh, Y.H.; Ferreira, F.C.; Livingston, A.G. The influence of membrane formation on functional performance
of organic solvent nanofiltration membranes. Desalination 2006, 199, 242–244. [CrossRef]
16. Nobrega, R. Transfer of dextran through ultrafiltration membranes: A study of rejection data analysed by gel
permeation chromotography. J. Mech. Appl. Math. 1989, 45, 17–36. [CrossRef]
17. Aimar, P.; Meireles, M.; Sanchez, V. A contribution to the translation of retention curves into pore size
distributions for sieving membranes. J. Membr. Sci. 1990, 54, 321–338. [CrossRef]
18. Tkacik, G.; Michaels, S. A rejection profile test for ultrafiltration membranes & devices. Bio/Technology
1991, 9, 941–946. [CrossRef]
19. Causserand, C.; Aimar, P.; Vilani, C.; Zambelli, T. Study of the effects of defects in ultrafiltration membranes
on the water flux and the molecular weight cut-off. Desalination 2002, 149, 485–491. [CrossRef]
20. Mulherkar, P.; Van Reis, R. Flex test: A fluorescent dextran test for UF membrane characterization. J. Membr. Sci.
2004, 236, 171–182. [CrossRef]
21. Chen, S.S.; Taylor, J.S.; Mulford, L.A.; Norris, C.D. Influences of molecular weight, molecular size, flux,
and recovery for aromatic pesticide removal by nanofiltration membranes. Desalination 2004, 160, 103–111.
[CrossRef]
22. Nghiem, L.D.; Schäfer, A.I.; Elimelech, M. Pharmaceutical retention mechanisms by nanofiltration membranes.
Environ. Sci. Technol. 2005, 39, 7698–7705. [CrossRef] [PubMed]
23. Hilal, N.; Busca, G.; Hankins, N.; Mohammad, A.W. The use of ultrafiltration and nanofiltration membranes
in the treatment of metal-working fluids. Desalination 2004, 167, 227–238. [CrossRef]
24. Yu, S.; Gao, C.; Su, H.; Liu, M. Nanofiltration used for desalination and concentration in dye production.
Desalination 2001, 140, 97–100. [CrossRef]
25. Weber, R.; Chmiel, H.; Mavrov, V. Characteristics and application of new ceramic nanofiltration membranes.
Desalination 2003, 157, 113–125. [CrossRef]
26. Loh, X.X.; Sairam, M.; Steinke, J.H.G.; Livingston, A.G.; Bismarck, A.; Li, K. Polyaniline hollow fibres for
organic solvent nanofiltration. Chem. Commun. 2008, 47, 6324–6326. [CrossRef]
27. Tragardh, G. Characterization methods for ultrafiltration membranes. Desalination 1985, 53, 25–35. [CrossRef]
28. Li, R.; Fan, H.; Shen, L.; Rao, L.; Tang, J.; Hu, S.; Lin, H. Chemosphere Inkjet printing assisted fabrication of
polyphenol-based coating membranes for oil/water separation. Chemosphere 2020, 250, 126236. [CrossRef]
29. Wu, M.; Chen, Y.; Lin, H.; Zhao, L.; Shen, L.; Li, R.; Xu, Y.; Hong, H.; He, Y. Membrane fouling caused by
biological foams in a submerged membrane bioreactor: Mechanism insights. Water Res. 2020, 181, 115932.
[CrossRef]
30. Tummons, E.; Han, Q.; Tanudjaja, H.J.; Hejase, C.A.; Chew, J.W. Membrane fouling by emulsified oil:
A review. Sep. Purif. Technol. 2020, 248, 116919. [CrossRef]
31. Teng, J.; Wu, M.; Chen, J.; Lin, H.; He, Y. Chemosphere Different fouling propensities of loosely and
tightly bound extracellular polymeric substances (EPSs) and the related fouling mechanisms in a membrane
bioreactor. Chemosphere 2020, 255, 126953. [CrossRef]
32. Teng, J.; Shen, L.; Xu, Y.; Chen, Y.; Wu, X.; He, Y.; Chen, J.; Lin, H. Chemosphere Effects of molecular weight
distribution of soluble microbial products ( SMPs ) on membrane fouling in a membrane bioreactor ( MBR ):
Novel mechanistic insights. Chemosphere 2020, 248, 126013. [CrossRef] [PubMed]
33. Gha, N.; Qamar, A. Membrane fouling quantification by specific cake resistance and flux enhancement using
helical cleaners. Sep. Purif. Technol. 2020, 239, 116587. [CrossRef]
34. Chan, Q.; Entezarian, M.; Zhou, J.; Osterloh, R.; Huang, Q.; Ellefson, M.; Mader, B.; Liu, Y.; Swierczek, M. Gold
nanoparticle mixture retention test with single particle detection: A fast and sensitive probe for functional
pore sizes of ultrafiltration membranes. J. Membr. Sci. 2020, 599, 117822. [CrossRef]
35. Hermia, J. Constant pressure blocking filtration laws-application to power-law non-Newtonian fluids.
Chem. Eng. 1982, 60, 111–120.
36. Wang, J.; Cahyadi, A.; Wu, B.; Pee, W.; Fane, A.G.; Chew, J.W. The roles of particles in enhancing membrane
filtration: A review. J. Membr. Sci. 2020, 595, 117570. [CrossRef]
37. Butt, F.H.; Rahman, F.; Baduruthamal, U. Characterization of foulants by autopsy of RO desalination
membranes. Desalination 1997, 114, 51–64. [CrossRef]
Membranes 2020, 10, 271 14 of 14

38. Lee, N.H.; Amy, G.; Croué, J.P.; Buisson, H. Identification and understanding of fouling in low-pressure
membrane (MF/UF) filtration by natural organic matter (NOM). Water Res. 2004, 38, 4511–4523. [CrossRef]
39. Pelekani, C.; Newcombe, G.; Snoeyink, V.L.; Hepplewhite, C.; Assemi, S.; Beckett, R. Characterization
of natural organic matter using high performance size exclusion chromatography. Environ. Sci. Technol.
1999, 33, 2807–2813. [CrossRef]
40. Filloux, E.; Gallard, H.; Croue, J.P. Identification of effluent organic matter fractions responsible for
low-pressure membrane fouling. Water Res. 2012, 46, 5531–5540. [CrossRef]
41. Peiris, R.H.; Budman, H.; Moresoli, C.; Legge, R.L. Understanding fouling behaviour of ultrafiltration
membrane processes and natural water using principal component analysis of fluorescence
excitation-emission matrices. J. Membr. Sci. 2010, 357, 62–72. [CrossRef]
42. Xiao, P.; Xiao, F.; Wang, D.S.; Qin, T.; He, S.P. Investigation of organic foulants behavior on hollow-fiber UF
membranes in a drinking water treatment plant. Sep. Purif. Technol. 2012, 95, 109–117. [CrossRef]
43. Meng, F.; Liao, B.; Liang, S.; Yang, F.; Zhang, H.; Song, L. Morphological visualization, componential
characterization and microbiological identification of membrane fouling in membrane bioreactors (MBRs).
J. Membr. Sci. 2010, 361, 1–14. [CrossRef]
44. Hir, M. Le Caractérisation du Colmatage des Membranes D’ultrafiltration de Production d’eau Potable par
des Nanoparticules Fluorescentes. Ph.D. Thesis, Aix Marseille University, Marseille, France, 2017.
45. Le Hir, M.; Wyart, Y.; Georges, G.; Siozade, L.; Moulin, P. Nanoparticles Retention Potential of Multichannel
Hollow Fiber Drinking Water Production Membrane. J. Membr. Sci. Res. 2018, 4, 74–84. [CrossRef]
46. Cai, Y.; Schwartz, D.K. Single-nanoparticle tracking reveals mechanisms of membrane fouling. J. Membr. Sci.
2018, 563, 888–895. [CrossRef]
47. Causserand, C.; Rouaix, S.; Akbari, A.; Aimar, P. Improvement of a method for the characterization of
ultrafiltration membranes by measurements of tracers retention. J. Memb. Sci. 2004, 238, 177–190. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).