REVIEWS

Materials for next-generation

desalination and water purification
membranes
Jay R. Werber, Chinedum O. Osuji and Menachem Elimelech
Abstract | Membrane-based separations for water purification and desalination have been
increasingly applied to address the global challenges of water scarcity and the pollution of aquatic
environments. However, progress in water purification membranes has been constrained by the
inherent limitations of conventional membrane materials. Recent advances in methods for
controlling the structure and chemical functionality in polymer films can potentially lead to new
classes of membranes for water purification. In this Review, we first discuss the state of the art of
existing membrane technologies for water purification and desalination, highlight their inherent
limitations and establish the urgent requirements for next-generation membranes. We then describe
molecular-level design approaches towards fabricating highly selective membranes, focusing on
novel materials such as aquaporin, synthetic nanochannels, graphene and self-assembled block
copolymers and small molecules. Finally, we highlight promising membrane surface modification
approaches that minimize interfacial interactions and enhance fouling resistance.

Department of Chemical and
Environmental Engineering,
Yale University, New Haven,
Connecticut 06520–8286,
USA.
Correspondence to M.E. 
menachem.elimelech@yale.
edu
Article number: 16018
doi:10.1038/natrevmats.2016.18
Published online 5 Apr 2016
Addressing the lack of adequate and safe water is among
the most important challenges of our time1. Water is
Earth’s most precious resource for life and is becoming
perilously scarce and befouled1,2. Population growth,
industrialization and climate change constitute an
expanding list of stressors to water resources that exacer-
bate global water scarcity. To address this challenge, there
is an important need for the development of efficient and
sustainable technologies to tap unconventional sources
of water (for example, seawater, brackish groundwater
and wastewater) to augment water supply beyond what
is obtainable from the hydrologic water cycle2.
A global emerging problem is the contamination of
drinking water sources, such as rivers and lakes, with
micropollutants that originate from municipal, indus-
trial and agricultural wastewaters3. Micropollutants,
which include compounds such as hormones, pharma-
ceuticals, pesticides, personal care products and indus-
trial chemicals, can have detrimental effects even at low
concentrations. Furthermore, municipal wastewater
effluents in water-stressed regions are often recycled for
agriculture and even for indirect potable use, thereby
introducing a wide range of micropollutants to fresh
water resources4,5. The development of efficacious tech-
nologies for the removal of such micropollutants and
other contaminants, as listed in TABLE 1, is of paramount
importance to ensure clean and safe drinking water.
The management of complex industrial wastewaters,
such as those from the oil and gas industries and coal-
fired power plants, has become a major challenge in
recent years6–8. Such wastewaters are characterized by a
high level of total dissolved solids, up to six times that of
seawater, and a high fouling potential. These wastewaters
are therefore difficult to treat 8. Furthermore, potential
new regulations may require zero liquid discharge for
industrial wastewaters (for example, from coal-fired
power plants) or brines from inland desalination plants.
This will necessitate the development of low-cost, energy-
efficient technologies that effectively concentrate feed
waters up to a very high level of total dissolved solids
and, eventually, to a solid salt 9,10.
In addition to mining fresh water from waste­
water, there is a growing interest in recovering valuable
resources from municipal and industrial wastewaters11,12.
For example, municipal wastewater is now considered
to be a renewable resource from which water, nutrients
(for example, phosphate and nitrogen), energy and sub-
stances such as bioplastics can be extracted13,14. Resource
recovery will be an integral part of wastewater treatment
in the coming decades, and novel separation technologies
will be indispensable in this effort.
Membrane-based treatment technologies can have a
key role in water purification and desalination. These
technologies produce water of superior quality, are

NATURE REVIEWS | MATERIALS VOLUME 1 | MAY 2016 | 1

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

less sensitive to feed quality fluctuations and have a
much smaller footprint compared with conventional
‘Victorian’ water treatment technologies2. In addition,
membrane-based desalination technologies are inher-
ently more energy efficient than thermal approaches15.
For example, the energy required for seawater desali-
nation by current state-of-the-art reverse osmosis (RO)
technology is already within a factor of two from the
limit of a reversible thermodynamic process, whereas
thermal desalination technologies can consume as much
as five times the energy used in RO16.
Current materials and fabrication methods for mem-
branes are largely based on empirical approaches and
lack molecular-level design17,18. Consequently, control
of the structure in the selective layer of the membrane
is limited, thus hampering membrane performance and
increasing the cost of water purification and desalina-
tion. Recent improvements in water purification mem-
branes have been marginal because of inherent material
limitations, including the permeability–selectivity trade-
off, which limits the achievable water–solute selectivity,
and high fouling propensity 15. Novel materials and scal-
able, molecular-level design approaches for membrane
fabrication are imperative for overcoming these limita-
tions and for substantially advancing water purification
and desalination technologies.
Here, we review polymer-based membrane materi-
als for desalination and water purification, focusing on
molecular-level design approaches. We highlight the
main limitations and requirements of current membrane
materials, discuss how these limitations and needs can
be addressed using new materials, and identify where
new approaches are needed to significantly advance
water purification and desalination technologies.
State-of-the-art membranes
Membrane technologies for water purification and desal-
ination with separation largely based on size include the
established microfiltration (MF), ultrafiltration (UF),
nanofiltration (NF) and RO processes, as well as the
emerging forward osmosis (FO) process. BOX 1 pro-
vides an introduction to these processes. A wide range

Table 1 | Contaminants of concern in source waters and membrane technologies that can meet treatment objectives
Source water Key contaminants Treatment objectives Membrane technologies Problems and challenges
Natural waters
Seawater Boric acid (affects crop health), • Reduce salinity • Pre-treatment: • Energy of desalination
divalent cations (cause scaling microfiltration (MF) and • Ecological impact of
issues) ultrafiltration (UF) seawater intake and brine
• Desalination: reverse discharge
osmosis (RO)
Brackish (saline) Divalent cations • Reduce salinity • Pre-treatment: MF and • Inland discharge of saline
groundwater UF brine
• Desalination: RO and
electrodialysis*
Surface waters Natural organic matter (precursor • Remove particles and • MF, UF, nanofiltration (NF) • High fouling potential
for disinfection by-products), microbial pathogens and RO • Formation of toxic
microbial pathogens, • Reduce natural organic disinfection by-products
micropollutants and algal toxins matter during oxidative
disinfection
Fresh groundwater Natural arsenic, nitrates, iron and • Reduce nitrate, iron, • MF, UF, NF and RO • Seawater intrusion for
manganese (staining issues) and manganese and/or aquifers near ocean
divalent cations scale-forming ions • Aquifer over-exploitation
Wastewaters
Municipal Microbial pathogens, • Degrade organic matter • MF, UF, NF and RO • Requirements for potable
wastewater micropollutants, phosphates • Remove or inactivate reuse
(algal bloom concerns) and pathogens • Large footprint
ammonia • Remove nutrients (that is, • Odour of conventional
nitrogen and phosphate) treatment plants
• High membrane-fouling
potential
Shale-gas produced Drilling fluid additives (for • Drilling reuse: remove • Drilling reuse: MF and UF • Large water consumption
water example, surfactants, oxidants, suspended solids and • Desalination: NF, RO • High total dissolved solids
strong acids), oil and grease, scalants (for low-salinity waters), • Fluctuating water quantity
radium and divalent cations • Disposal: remove oily forward osmosis (FO) and and quality
compounds and reduce membrane distillation • Regulations over disposal
salinity (MD)‡ • High membrane-fouling
potential
Coal-fired power Toxic metals (for example, • Remove dissolved toxic • MF, UF, NF, RO, FO and • Potential zero liquid
plant flue gas arsenic, selenium and mercury), metals and reduce salinity MD discharge requirements
desulfurization ammonia and organic acids
*Electrodialysis is an electrically driven desalination method that uses ion-exchange membranes. ‡Membrane distillation is a thermally driven desalination process in
which a hydrophobic porous membrane separates a hot feed stream from a cold permeate. A vapour pressure gradient drives water vapour flux.

2 | MAY 2016 | VOLUME 1 www.nature.com/natrevmats

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

of membrane materials, both polymeric and inorganic,
have been used. Polymeric membranes are by far the
most widespread, largely owing to their high process-
ability and low cost. State-of-the-art polymeric mem-
branes and their corresponding fabrication methods are
discussed in this section.
Phase inversion membranes. Most commercially avail-
able porous membranes that are used for MF and UF,
as well as some dense membranes used for RO and FO,
are formed by phase inversion. This process involves the
controlled precipitation of a dissolved polymer in a thin
film to produce a porous membrane structure17. The main
phase inversion technique used is non-solvent-induced
phase separation (NIPS), in which a film of polymer
dissolved in solvent is immersed in a non-solvent bath,
typically a water bath, leading to solvent–non-solvent
exchange and phase separation into polymer-rich and
polymer-poor phases (FIG. 1a). These two phases ultimately
form the solid membrane matrix (from the polymer-rich
phase) and pores (from the polymer-poor phase). Other
phase inversion techniques include controlled solvent
evaporation, in which a volatile solvent is allowed to
evaporate from a liquid film that includes polymer and non-
solvent, and thermally induced phase separation, in
which a high-temperature solution of dissolved polymer
is slowly cooled until phase separation occurs.
During phase inversion, several key parameters
determine the resulting membrane morphology: for
example, the nature of the solvent and non-solvent, poly­
mer type, polymer concentration and fabrication tech-
nique. As shown in FIG. 1b, a wide range of morphologies
can result, including asymmetric membranes for UF that
have a thin nanoporous selective layer (which is often
called the active layer) and a microporous underlying
structure, and relatively symmetric microporous mem-
branes for MF. However, owing to the stochastic nature
of the phase inversion process, membranes formed by
traditional phase inversion have polydisperse pore size
distributions, which adversely affect the selectivity of the
resulting active layer, as discussed later.
In contrast to phase inversion, an alternative process
called track etching can form UF and MF membranes
with pores of uniform size17. Track etching involves two
steps: bombardment with charged particles to partially
degrade a track of polymer, followed by chemical etching

Box 1 | Membrane processes for water purification and desalination

0.1 nm 1 nm 10 nm 100 nm 1 μm 10 μm

Hydrated ions Viruses

Micropollutants Bacteria
Natural organic matter Protozoa (for example, Cryptosporidium, Giardia)
Algal toxins

Reverse osmosis,
Ultrafiltration
forward osmosis

Nanofiltration Microfiltration

Solution–diffusion Size exclusion

Several common membrane processes for water purification and desalination separate largely Nature Reviews
on the basis |ofMaterials
solute
size. Processes that are driven by hydraulic pressure include microfiltration (MF), ultrafiltration (UF), nanofiltration
(NF) and reverse osmosis (RO). Additionally, the process of forward osmosis (FO) is driven by an osmotic pressure
difference between the feed source water and a specialized solution of high osmotic pressure, which is referred to as
a draw solution17,124. The processes are classified on the basis of the size of the solutes that are retained, as shown in
the figure (in which the sizes refer to the diameter). The main function of MF membranes is to remove suspended
particles and microbial pathogens. UF membranes are designed to retain macromolecules, such as natural organic
matter, as well as smaller pathogens, such as viruses, which are only partially removed by MF. The molecular weight
cut-off (that is, the solute size at which 90% of species are rejected) of UF membranes is between approximately 5
and 500 kDa. MF and UF membranes are porous, and molecular separation is based on a sieving or size exclusion
mechanism. NF membranes can effectively remove scale-forming ions, such as calcium and magnesium, and can
partially reduce salinity. Typical molecular weight cut-offs of NF membranes are between 100 and 300 Da. RO and FO
membranes are designed for desalination. Current RO and FO membranes are non-porous and can remove nearly all
ions in addition to uncharged solutes of molecular weight greater than around 100 Da. Molecular transport in RO and
FO is governed by a solution–diffusion mechanism in which molecules partition into the selective layer of the
membrane, diffuse through the polymer network and desorb in the permeate17,24. Separation of species stems from
differences in their solubility and/or diffusivity in the selective layer. Separation in NF is based on a combination of
sieving and solution–diffusion mechanisms.

NATURE REVIEWS | MATERIALS VOLUME 1 | MAY 2016 | 3

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

a b

Polymer, solvent
and additives
Thin film
Solid support
40 μm 20 μm

Non-solvent Solvent
Solid membrane
matrix
Pores

250 nm 2 μm

c m-Phenylene Trimesoyl chloride Crosslinked polyamide d

diamine
O O O O
H2N
Cl Cl N N N
H H H

NH2
O Cl O N N 500 nm
H H

Sulfur
Nitrogen

200 nm 200 nm

Figure 1 | State-of-the-art membranes for water treatment. a | A schematic diagram of phase inversion by non-solvent-
Nature Reviews | Materials
induced phase separation to form microfiltration (MF) and ultrafiltration (UF) membranes. A casting solution consisting of
polymer, solvent and desired additives is cast on a solid support as a thin film, which is followed by immersion in a non-solvent to
induce phase separation. b | Scanning electron microscopy (SEM) images of membranes formed from phase inversion. The upper
left panel is a cross-section of an asymmetric UF membrane with finger-like macrovoids cast from 9% polysulfone (PSf) in
dimethylformamide (DMF). The upper right panel is a cross-section of an asymmetric UF membrane with a sponge-like structure
cast from 12% PSf in DMF. The lower left panel is a top view of a hand-cast PSf UF membrane. The lower right panel is a top view of
a commercial 0.22-μm nominal pore size polyvinylidene fluoride MF membrane (EMD Millipore, Billerica, Massachusetts, USA).
c | A schematic representation of interfacial polymerization to form thin-film composite (TFC) reverse osmosis membranes.
Aqueous m‑phenylenediamine diffuses to the water/organic interface, reacting with trimesoyl chloride to form a dense
polyamide film with ridge-and-valley morphology. d | Microscopy images of a hand-cast TFC membrane. The upper panel
shows a top-view SEM image of rough polyamide film at a 45° viewing angle. The lower left panel is a transmission electron
microscopy (TEM) image of a thin cross-section showing the rough polyamide film formed on a PSf UF membrane. The
lower right panel shows a corresponding scanning TEM energy-dispersive X‑ray spectroscopy image; the presence of
sulfur and nitrogen delineates the PSf layer and polyamide film, respectively. The upper panels of part b are adapted with
permission from REF. 125, Elsevier. The upper panel of part d is adapted with permission from REF. 126, American Chemical
Society. The lower panels of part d are adapted with permission from REF. 20, American Chemical Society.

to form pores. The main limitation of track-etched
membranes is their low porosity (that is, low void vol-
ume within the membrane), which must be kept below
5% to prevent pore overlap. The low porosity results in
low water permeability, largely limiting track-etched
membranes to analytical separations.
Thin-film composite polyamide membranes. Thin-
film composite (TFC) polyamide membranes are the
gold standard for NF, RO and FO applications15. These
membranes comprise a non-porous, highly crosslinked
polyamide selective layer and an underlying porous
support layer, typically made of polysulfone. TFC
poly­amide membranes can achieve water permeability
and salt rejection far exceeding those for asymmetric
cellulose-acetate-based membranes, which were the
first-generation RO membranes developed more than
50 years ago17. This performance, combined with the
stability of TFC membranes over a wide pH range
(pH 2–11 for continuous operation), has led to their
widespread use in desalination applications.
The polyamide selective layer is formed through inter-
facial polymerization of a diamine — most commonly,
m‑phenylenediamine (MPD) for RO, FO and NF, and
piperazine (PIP) for NF — with a triacyl chloride, par-
ticularly trimesoyl chloride (TMC). During interfacial

4 | MAY 2016 | VOLUME 1 www.nature.com/natrevmats

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

polymerization (FIG. 1c), the polysulfone support is first
brought into contact with the aqueous diamine solution
to allow diamine penetration into the support. After
removing the excess solution, the support is immersed in
an organic phase containing TMC, which is not soluble in
water. As diamine monomers diffuse to the water/organic
interface, they react with TMC to form the polyamide
film, which adheres to the polysulfone support through
physical interactions. Subsequent hydrolysis of non-
reacted acyl chloride groups yields carboxyl groups,
which confer a negative charge to the membrane sur-
face. TFC RO membranes formed using aromatic MPD
and TMC monomers are fairly hydrophobic (with water
contact angles of 50–60°), whereas TFC NF membranes
formed using PIP and TMC are more hydrophilic (with
water contact angles of approximately 30°)19. Both MPD-
based and PIP-based TFC membranes are sensitive to
chemical degradation by strong oxidants, most notably
chlorine17. This sensitivity eliminates the usage of chlo-
rine as a membrane-cleaning agent and necessitates
removal of chlorine before the membrane step if it is
present in the feed stream.
The surface morphology of TFC membranes also
depends on the monomers used. PIP-based membranes
are smooth and have a root-mean-square roughness of
less than 10 nm. By contrast, MPD-based membranes
have a characteristic ‘ridge-and-valley’ structure, typically
having a root-mean-square roughness greater than 50 nm
and an overall end‑to‑end thickness of 100–500 nm. The
roughness is due to protruding polyamide nodules20, as
can be seen in FIG. 1d. The nodules are generally hollow;
transmission electron microscopy images suggest that the
nodule walls are roughly 10–30 nm thick20,21. Although
the rough nodular morphology increases fouling pro-
pensity (as discussed later), the increased surface area
stemming from this morphology may also increase water
permeability 21,22.
Inherent limitations and urgent requirements
The permeability–selectivity trade-off. Water flux
through porous (that is, MF and UF) membranes
and non-porous (that is, NF and RO) membranes is
described by equation (1)17:
Jw = A(ΔP –Δπ m) (1)
where Jw is the volumetric water flux, A is the water
permeability coefficient, ΔP is the applied hydraulic
transmembrane pressure and Δπm is the osmotic pres-
sure difference across the active layer. In typical water
treatment applications, Δπm is negligible for MF and UF,
whereas it is a crucial parameter for NF and RO.
For porous membranes, flow through the active layer
is modelled as laminar flow through an array of cylin-
drical pores (FIG. 2a, inset). Water permeability is related
by the Hagen–Poiseuille equation (equation (2)) 17 to the
solution viscosity (μ) and morphological characteristics
of the active layer — namely, the surface porosity (ε),
pore radius (rp) and active layer thickness (δm):
εrp2
A= (2)
8μδm
This simplified relationship best describes an idealized
membrane with uniformly sized cylindrical pores (that
is, an isoporous membrane). In reality, membranes that
are formed by phase inversion have non-cylindrical and
tortuous pores that vary substantially in size.
In addition to affecting water permeability, the
pore size distribution largely determines the rejection
(R = 1 − cpermeate/cfeed) of solutes present in the feed solution.
For a solute of radius a, rejection by a pore of radius rp
(with rp ≥ a) is a function of the ratio a/rp, as shown by
equation (3)23:
R = 1 – [2(1 – rap )2 – (1 – rap )4]exp[–0.7146( rap )2] (3)
For complete solute rejection (cpermeate = 0), the entire
pore size distribution must be smaller than the solute of
interest (rp ≤ a). Similarly, to completely reject a mixture of
solutes, the largest pore must be smaller than the smallest
solute that is to be retained.
For non-porous, dense membranes, water and solute
transport is governed by the solution–diffusion model.
Water and solute molecules partition into the active layer
of the membrane, diffuse through the polymer matrix
down their chemical potential gradients and desorb into
the permeate solution. The solubility and diffusivity
combine to define the diffusive water permeability (Pw)
and diffusive solute permeability (Ps), both of which are
intrinsic material properties — that is, they are independ-
ent of thickness. The water permeability coefficient of the
active layer is related to the diffusive water permeability
and layer thickness by equation (4)24:
PwVw
A= (4)
δmRgT
where Vw is the molar volume of water, Rg is the gas con-
stant and T is the absolute temperature. Similarly, solute
transport is modelled as Fickian diffusion (equation (5))24:
Ps
Js = Δcm = BΔcm (5)
δm
where B is the solute permeability coefficient and Δcm
is the solute concentration difference across the active
layer. Together, the coefficients A and B largely define the
selective layer performance of non-porous membranes.
The ideal membrane has both high water permeabil-
ity and high selectivity — a combination that is difficult
to attain. In recent studies, it has been proposed that
porous and non-porous water separation membranes
exhibit a permeability–selectivity trade-off that is sim-
ilar to the Robeson plot that has found extended use in
the field of polymeric gas separation membranes23,25,26.
These trade-offs all have a common implication: an
increase in permeability for a highly permeating species
(for example, water) comes with an even greater increase
in permeability for relatively highly retained species (for
example, macromolecules for UF or salt for RO).
For porous membranes, selectivity is intrinsically
linked with active layer pore size. This relationship is
shown in FIG. 2a for UF membranes, in which selectivity
for water over bovine serum albumin, given as the sepa-
ration factor 1/(1 − R), decreases sharply with increasing
water permeability at fixed surface porosity and active

NATURE REVIEWS | MATERIALS VOLUME 1 | MAY 2016 | 5

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

layer thickness23. This trade-off occurs for an ideal iso-
porous membrane when the pore size exceeds the solute
size and is even more pronounced for real phase inversion
membranes, which have a heterogeneous pore size dis-
tribution owing to the stochastic nature of selective layer
formation. For either case, decreasing the active layer
thickness or increasing surface porosity would increase
water permeability without sacrificing selectivity.
For non-porous membranes, a permeability–selectivity
trade-off was originally found by comparing the diffu-
sive water and salt permeabilities for several bulk poly­
mer materials25. A cubic relationship was proposed,
wherein Ps ∝ Pw 3. A similar trade-off relationship was
subsequently found to hold true for coefficients A and
B (that is, B ∝ A3) for a commercial TFC RO membrane
that was treated stepwise with chlorine and alkaline to
increase permeability27, as shown in FIG. 2b. Also shown
in FIG. 2b are the permselectivities (A/B) for current
commercial membranes, values for which were calcu-
lated on the basis of the technical specifications pro-
vided by the manufacturer 28. A comparison with the
proposed TFC trade-off relationship shows that many
of the membranes that are currently available outper-
form the previously proposed trade-off. In other words,
in contrast to the results for a single chemically treated
membrane, incremental gains in the water permeabil-
ity coefficient have been attained without experiencing
a loss in permselectivity for independently fabricated
membranes. These improvements are possibly due to
optimization of the interfacial polymerization reaction
conditions to yield slightly different film chemistries,
polymer network structures and film morphologies.
An updated trade-off relationship is shown in FIG. 2b,
reflecting currently achieved active layer performance.
Although the pattern of incremental performance gains
in commercial TFC membranes is likely to continue,
water–salt separation in these membranes is still fun-
damentally dependent on the separating abilities of
the aromatic polyamide material. Therefore, an upper
performance limit is still expected to exist for TFC
membranes, motivating research on novel materials to
achieve further performance gains.
Selectivity, not permeability, is key for desalination.
Research on novel active layers for desalination mem-
branes seeks to improve the water permeability or water–
solute selectivity, with permeability often being the focus.
However, for both seawater RO (SWRO) and FO desal-
ination, process modelling shows that increased water
permeability above levels that are currently achievable
(that is, 0.5–0.8 μm s−1 bar−1 or 2–3 l m−2 h−1 bar−1; FIG. 2b)
would negligibly decrease energy requirements and capi-
tal costs29,30. For example, increasing the water permeabil-
ity coefficient from 3 to 10 L m−2 h−1 bar−1 would decrease

a Water permeability coefficient, A (L m–2 h–1 bar–1) b Water permeability coefficient, A (L m–2 h–1 bar–1)
0 1000 2000 1 10
104

a 100
Water–salt permselectivity, A/B (bar–1)

103 rp
Separation factor, (1 – R)–1

102
10

101

100 1
0 1 2 3 4 5 6 7 8 1 10
Water permeability coefficient, A (10 m s bar )
–4 –1 –1
Water permeability coefficient, A (10–6 m s–1 bar–1)
Acrylic- and acrylonitrile-based polymers TFC RO membranes
Polysulfone-based polymers Chlorine- and alkaline-treated RO membrane
Cellulose-based polymers
Nature Reviews | Materials
Figure 2 | Permeability–selectivity trade-off relationships for ultrafiltration and reverse osmosis membranes.
a | Ultrafiltration (UF) membranes. Experimentally measured water permeabilities and rejection values of bovine serum
albumin are shown for membranes formed by phase inversion using polyacrylonitrile, polysulfone-based polymers and
cellulose-based polymers. The separation factor uses true rejection values (that is, rejection over just the active layer of the
membrane). The lines are modelled using equations (2) and (3) and a fixed surface porosity/active layer thickness ratio (ε/δm)
of 1 μm−1. The solid line assumes a log-normal distribution of pore sizes with a standard deviation of 0.2 relative to the mean
radius. The dashed line assumes a uniform pore size. The insets show how increasing the pore radius (rp) relative to the
solute radius (a) results in the observed trade-off. Data for panel a are from REF. 23. b | Reverse osmosis (RO) membranes.
The blue squares represent the permselectivity values for commercial thin-film composite (TFC) RO membranes. These
values are calculated from technical documentation of the manufacturer28. The green triangles represent permselectivity
values measured for a chlorine- and alkaline-treated commercial seawater RO membrane27. The dashed line is the
corresponding proposed trade-off relationship for TFC membranes. The solid line is an updated trade-off relationship that
better fits the performance of current membranes.

6 | MAY 2016 | VOLUME 1 www.nature.com/natrevmats

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

the SWRO energy requirements by less than 2%29.
This limited difference largely stems from the single-
stage operation of RO, which necessitates the use of a
hydraulic pressure greater than the osmotic pressure of
the brine exiting the RO stage, irrespective of the mem-
brane permeability 15. The use of hydraulic pressure is the
main determinant of the energy used by the RO stage.
Instead of attaining higher water permeability, a
more effective goal for desalination membrane materi-
als research would be to attain improved selectivity for
water over all dissolved solutes. For example, despite a
high salt rejection (>99.5% at standard test conditions)
for SWRO membranes, multiple RO passes are still
sometimes needed, such as in the production of high-
purity water for industrial use31. In addition, rejection
by current membranes of small neutral species is fairly
poor 32,33. Some of these species, such as boric acid in sea-
water and some micropollutants in wastewater, adversely
affect human and/or crop health. Incomplete rejection of
these species necessitates extensive post-treatment steps,
such as further RO passes, ion exchange and oxidative
chemical degradation2,34. These additional steps require
substantially higher chemical and energy usage, as well
as incurring higher capital costs, which could be avoided
if more selective RO membranes were available.
High fouling propensity. Membrane fouling — that is,
the accumulation of substances on the membrane sur-
face or within the membrane pores — is a major obstacle
for the efficient operation of membrane systems. There
are three types of fouling: organic fouling by adsorption
of dissolved organic matter; scaling by deposition of
precipitated salts or by surface nucleation and growth
of sparingly soluble salts; and biological fouling (bio-
fouling) by deposition and growth of microorganisms
to form strongly adherent biofilms17. Excessive fouling
deteriorates membrane performance (that is, water flux
and selectivity), necessitates chemical cleaning, which is
unsustainable and shortens membrane life, and increases
energy consumption and operating costs17. The sever-
ity of fouling can vary substantially and is affected not
only by feed water quality and process conditions (for
example, water flux and hydrodynamics), but also by the
fouling propensity of the membrane itself. The develop-
ment of membrane materials with low fouling propensity
remains highly important.
Many UF and MF applications require a high per-
formance despite feed streams that have a high fouling
potential: for example, the microorganism-filled sludge
that accumulates in membrane bioreactors for munici-
pal wastewater treatment 35. Phase inversion UF and MF
membranes commonly comprise synthetic polymers,
including polyvinylidene fluoride (PVDF), poly­sulfone,
poly­ethersulfone (PES) and polyacrylonitrile (PAN)17,36.
These poly­mers are easily processable, chemically,
thermally and mechanically robust, and inexpensive.
However, they are also fairly hydrophobic, which can
exacerbate organic and biological fouling through
increased adsorption of foulants to the membrane sur-
face and interior pores37. Cellulose-acetate-based mem-
branes are more hydrophilic and more fouling resistant,
but their application is limited by a narrow operating
pH range of 2–8 and an operating temperature of less
than 30 °C (REF. 36). In addition to material proper-
ties, the polydisperse pore size distribution of MF and
UF membranes that arises from phase inversion can
increase fouling propensity. Large pores tend to foul
rapidly owing to high local fluxes and can suffer from
irreversible fouling (that is, fouling by substances that
are not removed with rinsing steps) because of foulant
penetration into the membrane interior 38.
TFC polyamide membranes used in RO, FO and
NF are highly susceptible to all three types of fouling,
largely owing to a combination of surface morphology,
hydrophobicity and charge. The high surface roughness
of MPD-based TFC membranes enhances all types of
fouling as a result of increased surface area, thus provid-
ing greater foulant–membrane interaction and greater
opportunity for attachment 39. MPD-based TFC mem-
branes are also fairly hydrophobic, thereby increasing
rates of organic and biological fouling 40. Surface charge
exacerbates fouling for both MPD- and PIP-based TFC
polyamide membranes because of the surface carboxyl
functional groups. Through calcium bridging, these car-
boxyl groups increase the binding of organic foulants
such as natural organic matter (that is, partially aromatic
macromolecules with high levels of carboxylic and phe-
nolic functional groups)41. The enrichment of calcium
at the surface can also induce calcium sulfate scaling
through surface nucleation and growth42.
Designing highly selective membranes
Recent advances in chemical synthesis and directed
assembly offer the possibility of molecular-level design
of selective layers, potentially facilitating the fabrication
of highly selective membranes for both water purifi-
cation and desalination (that is, for UF, RO and FO).
For desalination membranes, specially designed water
channels could lead to a shift in how water itself moves
through the membrane: from partitioning into and dif-
fusing through polymeric thin films to single-file trans-
port through active-layer-spanning water channels.
Such sub-nanoporous water channels, if well designed,
could offer a significant improvement in desalination
performance.
Aquaporin as a model water channel. In the design of
sub-nanoporous water channels, materials research can
gain insight from the biological water channel aqua-
porin. These ubiquitous, integral membrane proteins
have homologues found in bacteria, archaea, plants
and animals. They exist as tetramers of four identical
monomers, and each monomer bears a water channel43.
Aquaporin is remarkable for its unique transport prop-
erties: some types of aquaporin are not only highly per-
meable to water, but also extremely selective, rejecting
ions, protons and even small neutral solutes, such as
urea, that readily permeate through current desalination
membranes44–47.
The high permeability of the aquaporin water
channel stems from its hourglass structure (FIG. 3a),
which features wide entrance vestibules and a narrow

NATURE REVIEWS | MATERIALS VOLUME 1 | MAY 2016 | 7

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

a Aquaporin Carbon Synthetically b

nanotubes designed nanochannels Block copolymer
(10–100 kDa)

5–50 nm

Polymerizable
surfactants
(~1 kDa) Graphene-based
frameworks

500 nm

Figure 3 | Selective membranes formed using molecular-level design. a | Sub-nanoporous materials Nature Reviews | Materials
with promise for
desalination. The top panel shows discrete channels spanning a thin, nearly impermeable film, including the membrane
protein aquaporin, single-walled carbon nanotubes with diameters ≤0.8 nm to exclude salt, self-assembled channel-forming
macrocycles and single-molecule synthetic water channels. The bottom panel shows contiguous media that fully form a
selective layer, including inverse hexagonal channels formed using wedge-shaped polymerizable surfactants, tortuous
channels formed in bicontinuous cubic phase using polymerizable surfactants and graphene-based frameworks, shown with
either graphene oxide or reduced graphene oxide as the grey sheets and covalent linkers in green. In all of the schematics,
water molecules are shown traversing the nanochannels. b | Controlled formation of the nearly isoporous ultrafiltration
membrane using block copolymers. Scanning electron microscopy image in panel b is from REF. 90, Nature Publishing Group.

hydrophobic centre that supports single-file transport of
water molecules46. Selectivity is size- and charge-based:
0.8 nm above the channel centre, the channel constricts
to just 0.28 nm in diameter (which is approximately the
size of a water molecule) owing to protruding phenyl­
alanine, histidine and arginine side chains. The arginine
side chain provides the constriction site with a fixed
positive charge. This tight, charged constriction site
provides a high energy barrier that prevents permeation
of charged and neutral solutes44–46,48.
Owing to its exceptional transport properties, aqua-
porin has recently been assessed as the selective agent in
desalination membranes49–51. Selectivity of a defect-free
aquaporin-based membrane should be very high and,
depending on the achievable packing density, water
permeability could match that of TFC membranes
(TABLE 2). Membrane fabrication has typically followed a
vesicle-based approach. The fairly stable bacterial aqua-
porin Z (AqpZ) is first incorporated into unilamellar
lipid or block copolymer vesicles, with AqpZ spanning
the vesicle bilayer. Block copolymers — that is, poly-
mers comprising two or more monomers divided into
chemically distinct sequences — with a hydrophobic
polydimethylsiloxane (PDMS) block have been most
commonly used50,51. Using physical or chemical interac-
tions, the vesicles are then ruptured on a porous support
to form a planar bilayer, which becomes the selective
layer of the membrane. Membranes that are formed
using this strategy have often exhibited poor salt rejec-
tion52, probably because of defects, such as unruptured
vesicles, although salt rejection of up to 98% has been
reported50. Membranes have also suffered from poor
stability, largely because of the non-crosslinked lipid
or block copolymer bilayer 49,50. Protein stability is an
additional concern. Although AqpZ reconstituted in
vesicles remains active after storage at 4 °C for several
months43, it is likely that it will not be stable enough for
continuous use.
Discrete synthetic nanochannels. Using aquaporin as a
model, it is theoretically possible to systematically design
discrete water channels that are suitable for desalination
and stable enough for easy processing and use. Such
channels should ideally exhibit the following attributes:
molecular-sieve-like constriction to exclude salt and
small neutral solutes; stable alignment within a poly-
meric thin film; intrinsic water permeability and achiev-
able packing density that are high enough to match the
water permeability coefficient of TFC membranes;
chemical and mechanical stability; and ease of fabrication
using consistent and scalable synthesis methods.
Despite their structural simplicity, carbon nanotubes
(CNTs) have the potential for remarkable performance
as nanochannels for water purification and desali-
nation. Much of their promise for desalination was
demonstrated by molecular dynamics simulations for a
single-walled (6,6)CNT (REF. 53) (FIG. 3a). The simulations
showed the formation of a single-file water wire with
ultra-fast water conduction (TABLE 2), which is largely
attributed to the atomically smooth, non-polar inte-
rior. Simulations have also indicated that (6,6)CNTs,
which have a diameter of 0.81 nm between opposing
carbon centres and an inner (pore) diameter of 0.47 nm,
can fully reject salt 54. Salt passage increases rapidly
with diameter: just 58% salt rejection is expected for
(8,8)CNTs with a 1.1-nm centre-to-centre diameter.

8 | MAY 2016 | VOLUME 1 www.nature.com/natrevmats

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

In two early experimental works, aligned-CNT-based
membranes were fabricated using chemical vapour
deposition (CVD) methods55,56, yielding CNTs with
average inner diameters of 1.6 and 7 nm. Experiments
validated the proposed ultra-fast water permeation:
water permeability was found to be three to five orders
of magnitude higher than that predicted by continuum
flow models. However, the smallest particle rejected
was colloidal gold with a 2-nm diameter, indicating
that these membranes fit in the tight UF range. To
enhance CNT selectivity and potentially to allow desal-
ination, tip functionalization has been explored using
ligands such as biotin57, aliphatic groups58, charged
groups57–60 and zwitterions61. Although tip functionali-
zation has proven somewhat effective, the approach is
limited by a lack of control over ligand placement and
the permeability–selectivity properties of the chosen
flexible ligand.
A more ideal approach for ultra-selective CNT
membranes would be to directly use the rigid molec-
ular sieving properties of highly monodisperse, small
single-walled CNTs that intrinsically reject salt, as seen
in molecular dynamics simulations54. Such a membrane
has yet to be experimentally demonstrated, mainly
because of the heterogeneous size (that is, diameter)
distribution that typically arises during CNT synthe-
sis. Recent advances in monodisperse CNT synthesis
may overcome this limitation. For example, using a flat
platinum catalyst and a tailored precursor for CVD, sin-
gle-walled (6,6)CNTs with lengths greater than 100 nm
were selectively synthesized62. Scaling up this process, if
it is possible, would result in CNTs that are well suited
for desalination.
Synthetically designed nanochannels offer a custom-
izable alternative to CNTs. Although these nanochannels
lack the atomically smooth interior present in CNTs,
their properties can be individually tailored, allowing far
greater control over physical dimensions, performance,
processing and alignment. Synthetic nano­channels
can be divided into two classes: single-molecule and
supramolecular nanochannels (FIG. 3a).
A recent single-molecule example is the peptide-
appended pillar[5]arene channel63. Pillar[5]arenes are
macrocycles (that is, cyclic macromolecules) comprising
five chemically linked hydroquinone groups, in which the
benzene rings are oriented orthogonally to the macro­
cycle plane64. In peptide-appended pillar[5]arene chan-
nels, the macrocycle, which somewhat resembles a CNT,
defines the channel diameter, whereas the appended pep-
tide arms control the length of the channel. The channels
have been experimentally observed to readily insert into
and to self-align within lipid bilayers. They were also seen
to self-assemble into densely packed arrays. Although the
channels were highly permeable to water (TABLE 2), they
did not reject ions and had a molecular weight cut-off
of approximately 420 Da, making them unsuitable for
desalination. It may be possible to improve selectivity
through functionalization of the appended peptides but,
as discussed for CNTs, enhancing selectivity through the
use of smaller-diameter central constriction would be the
ideal solution.
Supramolecular synthetic nanochannels comprise
monomer units that self-assemble to form a water-
conducting channel. Two promising examples with sev-
eral similarities are modified cyclic peptide nanotubes65
and m‑phenylene ethynylene nanotubes66. Both types of

Table 2 | Transport properties for desalination-focused discrete water channels

Channel type Average single- Pore density Areal pore Measurement Demonstrated
channel water to match high- density method rejection
permeability water-flux TFC to match
coefficient (channels cm−2) TFC
at 25 °C
(molecules s−1 Pa−1)
Bacterial 5–36‡ (0.7–5.2) × 1012 7–47% Vesicle size All solutes
aquaporin Z43,127 change and
(~4 nm in length)* voltage clamp
(6,6)CNTs54,56 110§ 2.4 × 1011 0.25% Molecular NaCl (by simulation)
(1.4 nm in length) dynamics 2-nm colloidal gold
(by experiment)||
Peptide-appended 2.4‡ 1.1 × 1013 32% Vesicle size >420 Da
pillar[5]arenes63 change
(~4 nm in length)*
Modified cyclic 11¶ 2.4 × 1012 8% Molecular Not tested
peptide nanotubes128 dynamics
(2 nm in length)
m‑Phenylene 0.34# 7.9 × 1013 630% Vesicle size KCl and NaCl
ethynylene change
nanotubes66
(~4 nm in length)*
CNT, carbon nanotube; TFC, thin-film composite. *Approximate thickness of a lipid bilayer. ‡Permeabilities adjusted for
temperature using measured Arrhenius activation energies. §Simulated at 27 °C. ||For an aligned double-walled carbon nanotube
membrane with a 1.6-nm average pore diameter and with rejection measured using dead-end filtration. ¶For methylated cyclic
peptides to correspond to previous experimental work65. #Measured at 12 °C; value is unadjusted as Arrhenius activation energy
was not determined.

NATURE REVIEWS | MATERIALS VOLUME 1 | MAY 2016 | 9

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

nanotubes are formed using macrocyclic rings that have
chemically modifiable interiors to control selectivity. The
rings self-assemble by inter-ring hydrogen bonding and
π–π stacking to form solvent-stable, high-aspect-ratio
nanotubes. Through molecular-level design, the two
nanotube types were synthesized to have very small pore
widths (0.47 nm for modified cyclic peptide nanotubes
and 0.64 nm for m-phenylene ethynylene nanotubes).
Both types also have promising transport behaviour, as
shown in TABLE 2. In particular, m‑phenylene ethynylene
nanotubes have been shown experimentally to reject
sodium, potassium and chloride ions while maintain-
ing high water permeability in open (hydrated) chan-
nels. However, channels were open only 5% of the time,
resulting in low average water permeability compared
with TFC membranes (TABLE 2).
For all discrete nanochannels, the current challenge
is predominantly the synthesis of channels with the
desired water and solute transport properties. An equally
important technical challenge is developing scalable solu-
tion-phase methods for the assembly and alignment of
these channels in robust selective layers. Structural fea-
tures that aid alignment are likely to be important and
will be a major factor in nanochannel design. In addition,
much potential lies in using the self-assembling proper-
ties of amphiphilic surfactants and block copolymers.
Indeed, this approach has already been applied with some
success for CNTs67 and cyclic peptides68. Self-assembled
materials can also be used directly to form nanoporous
membranes, as described in the next subsection.
Self-assembled materials. Nanostructured polymer mem-
branes based on self-assembled materials have been a
topic of considerable interest for more than two decades.
The underlying motivation is the potential to derive highly
selective membranes because of the narrow pore size
distribution that results from equilibrium self-assembly.
This contrasts sharply with the broad size distributions in
conventional membranes that result from kinetically
dictated fabrication processes such as phase inversion.
Two broad classes of materials have been used.
The first class is reactive small molecules with a molar
mass generally less than 1 kDa that self-assemble into
lyotropic or thermotropic liquid crystalline meso-
phases69,70. Lyotropic liquid crystalline mesophases are
structurally ordered systems that form principally as a
function of small molecule concentration in a solvent
(typically water), whereas thermotropic liquid crystal-
line mesophases mainly form as a function of tempera-
ture in the absence of a solvent. Before they can be used
as membranes, liquid crystalline mesophases must be
poly­merized or crosslinked to form sufficiently robust
polymer films. The second class is block copolymers with
a molar mass generally greater than 10 kDa. Block copol-
ymers self-assemble into mesophases that are structurally
similar to those in small molecule systems71,72. Physical
entanglement of the polymer chains and/or the glassy
nature of the materials can provide mechanical integ-
rity without an explicit need for crosslinking. The ver-
satility of modern synthetic polymer chemistry enables
block copolymer membranes to be designed with a wide
range of physicochemical properties and functionalities.
For both small molecules and block copolymers, self-
assembly occurs because of partitioning of chemically
distinct segments of the molecules.
Although small molecules and block copolymers
can self-assemble into various structures (principally as
a function of composition), hexagonally packed cylin-
drical and bicontinuous cubic mesophases (FIG. 3) are
the structures of interest for membrane separations,
in which the cylindrical or bicontinuous cubic micro­
domains act as the pores for size-selective transport.
These mesophases form over limited ranges of composi-
tion that are thermodynamically defined. For both block
copolymers and small molecules, the characteristic pore
size is approximately set by the product of the size of the
molecule and the volume fraction of its pore-forming
segment. Currently accessible pore diameters range
from 5 to 50 nm (which lies in the UF range) for block
copolymers and 0.5 to 2 nm (which lies in the RO, FO
and NF range) for small molecules. In most cases, the
pores are formed by the selective removal (for example,
etching by base hydrolysis of polylactide73) or shrinkage
of material from the self-assembled structures, but for
small molecules, the porosity may be an intrinsic feature
of the system. Maximum porosity is usually 35–40%, as
it is defined by the small molecule or block copolymer
used and the composition range over which the desired
structure forms.
The ideal self-assembled selective layer has straight
pores that fully span the layer. For cylindrical meso-
phases, this requires the alignment of the cylindrical
pores perpendicular to the film surface as well as physical
continuity of the pores throughout the film. By contrast,
bicontinuous cubic mesophases intrinsically provide
three-dimensionally percolated pathways throughout
the system. However, these pathways are inherently tor-
tuous, which increases hydraulic resistance. In addition,
bicontinuous cubic phases are considerably less acces-
sible than cylinders owing to the limited composition
range in which they are stable — if they are present
at all — in most small molecule and block copolymer
systems. Efforts to develop high-performance mem-
branes based on self-assembled mesophases have there-
fore focused heavily on molecular design to ensure the
stable formation of bicontinuous cubic structures and
on the development of scalable directed self-assembly
methods to allow vertical orientation of cylindrical pores
in thin films.
Recent years have witnessed substantial progress
in molecular design for desalination-focused small
molecule membranes. Lyotropic liquid crystalline bi­­
continuous cubic mesophases based on ionic gemini
surfactants are particularly promising 74–77. For example, a
crosslinkable phosphonium system with pores of 0.75 nm
in diameter demonstrated 95% NaCl rejection and more
than 99% rejection of sucrose and MgCl2. Although water
permeability was low because of the 40‑μm thickness of
the selective layer, the thickness-normalized water per-
meability was comparable to that of commercial TFC RO
membranes, indicating that this material could be prac-
tically effective if thin selective layers of approximately

10 | MAY 2016 | VOLUME 1 www.nature.com/natrevmats

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

30 nm in thickness can be realized76. Several lyotropic
and thermotropic liquid crystalline phases (for example,
hexagonal and bicontinuous cubic) that use gallic acid
derivatives have also been developed78–83. For these mate-
rials, the pore size was readily tuneable over a range of
0.8–2 nm by varying the type of counterion present in salt
forms of the molecule84 and by using sacrificial species
that can be rinsed away after initially co‑assembling at
the centre of the cylindrical structures82,85,86. Gallic acid
derivatives have also been used to form thin films (that
is, selective layers) on porous supports. Spin-coating 83
and roll-casting 81 processes yielded film thicknesses of
50–100 nm and around 600 nm, respectively. In most of
the work carried out thus far on small molecule mem-
branes, separation properties of the resulting membranes
have fallen into the NF range. Further advances in molec-
ular design are crucial for fine-tuning both pore size and
permeability to address desalination applications more
effectively.
Progress has also been made in directing self-
assembly over large areas to achieve vertically aligned
cylinders. Magnetic fields provide a readily scalable
means of achieving such orientation control and have
been successfully used in both small molecule87 and
block copolymer88 systems. Recent work has shown an
alternative field-free approach based on so-called soft
confinement, in which symmetric physical confine-
ment of a small molecule mesophase during ordering
results in the spontaneous display of vertically oriented
microdomains89. The resulting films exhibit physically
continuous, vertically aligned 1-nm pores.
For block copolymer membranes, fabrication meth-
ods that combine self-assembly with a kinetic component
have proven highly effective. For example, self-assembly
with non-solvent-induced phase separation (SNIPS) is
a single, readily scalable process that can produce verti-
cally oriented pores in a thin selective layer at the mem-
brane surface with an integrally formed microporous
underlying support layer 90,91 (FIG. 3b). As SNIPS com-
bines self-assembly with kinetically limited phase inver-
sion, the resulting pore size distribution is broader than
that produced in equilibrium self-assembly processes.
Nevertheless, these membranes display sharper mole­
cular weight cut-off characteristics and thus improved
selectivity relative to NIPS-formed UF membranes in the
10–40-nm regime considered91–93. The combination of
block copolymer microphase separation and kinetically
limited structure formation also features in the formation
of membranes by polymerization-induced microphase
separation; this has recently been used to fabricate a
bicontinuous cubic phase with 4-nm pore size94.
Graphene. Owing to its atomically smooth and atom-
ically thin nature, graphene has become a widely
researched material for membrane separations, espe-
cially for desalination95–97. Graphene-based desalination
membranes have been proposed in two forms: nano­
porous graphene and graphene-based frameworks.
Both forms are envisioned to serve as selective layers
and to operate as molecular sieves with size-based
exclusion of undesired solutes.
To produce nanoporous graphene, defects are intro-
duced into pristine, single-layer graphene by plasma
etching or bombardment by ions or electrons97,98. Because
of the ultra-low thickness of the resulting pores, nano­
porous graphene was projected by molecular dynamics
simulations to have 102–103 times greater water permea-
bility coefficients than current TFC RO membranes (the
effect of thickness is expressed in equation (2)). Complete
salt rejection was projected to be possible for hydroxy-
lated pores that were 0.45 nm in diameter 99. Such perfor-
mance has recently been achieved for a 5‑μm‑diameter
sample of nanoporous graphene formed by oxygen
plasma etching of graphene grown through CVD97.
However, scaling up such a membrane is extremely chal-
lenging, because it requires the inexpensive formation
of a large area of defect-free, single-layer graphene, as
well as the scalable formation of uniformly sized nano-
pores100–102. In addition, even if achieved, the usefulness
of nanoporous graphene for desalination applications is
doubtful. As discussed earlier, increased water perme-
ability has a minimal effect above the current levels of
TFC desalination membranes, and unless the nanopores
can be even more finely tuned to also reject small neutral
solutes, nano­porous graphene will be at a disadvantage
compared with TFC membranes.
Graphene-based framework membranes differ in
morphology and water transport mechanism from
nanop­orous graphene and comprise a multi-layered
stack of finite-sized graphene sheets95,96,103 (FIG. 3a). They
are formed as laminates through vacuum filtration or
layer-by-layer deposition of graphene oxide, which is
inexpensively produced by oxidation of graphite and is
stable when suspended in water. Graphene oxide frame-
works allow ultra-fast permeation of water, which is pos-
tulated to be due to the slip flow along the atomically
smooth, unoxidized graphene channels; this is analogous
to the water flow within carbon nanotubes96. However,
the spacing between contiguous sheets increases from
0.9 ± 0.1 nm in humid air to 1.3 ± 0.1 nm when immersed
in water, resulting in rapid permeation of monovalent
and divalent cations and the corresponding counteri-
ons95. Graphene oxide frameworks still show a sharp
size-based cut-off, but the cut-off is too large for desali-
nation. In addition, graphene oxide laminates are gener-
ally unstable in water, unless they are stabilized by trace
multivalent cations104.
To potentially facilitate desalination, interlayer spac-
ing can be decreased and stability can be increased
through chemical or thermal reduction to remove oxy-
gen species. Full reduction decreases the interlayer spac-
ing to 0.36 nm, which is too small to allow water flow
between sheets and thus restricts water permeation to
defect-driven flow105. Although a reduced graphene oxide
membrane was recently shown to allow water permeation
with high salt retention when tested in FO, the defect-
driven permeation in this approach limits water perme-
ability 106. Another approach for decreasing interlayer
spacing and increasing stability is covalent crosslink-
ing 103,107,108. However, a covalent crosslinking method has
yet to be shown that facilitates salt rejection. Depending
on the success of these efforts, water-treatment-focused

NATURE REVIEWS | MATERIALS VOLUME 1 | MAY 2016 | 11

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

a b Surface modification that renders the membrane

Hydrophilic polymers Polymers with low surface energy surface highly hydrophilic substantially reduces organic
Poly(ethylene oxide) Poly(dimethylsiloxane) fouling and the adhesion of bacteria that eventually
n leads to biofouling. The most common hydrophilic
O
O O n
n Si materials for the preparation of fouling-resistant sur-
O
O O F n faces are polyethylene oxide (PEO)-based and zwitter-
O
O– F ionic polymers110,111 (FIG. 4a). PEO, which is also known
N+ S F as polyethylene glycol (PEG), binds water molecules by
O F F F
hydrogen bonding, whereas zwitterionic polymers —
Poly(sulfobetaine) Poly(hexafluoro-
butylmethacrylate) for example, poly(sulfobetaine) and poly(carboxybe-
Hydration layer taine) — bind water molecules more strongly through
electrostatic interactions111. The hydration layer of the
Hydrophilic side chain hydrophilic PEO and zwitterionic polymers provides a
Low-surface-energy block steric repulsive barrier that prevents the adsorption of
organic molecules and bacteria. Brush layers of these
Hydrophobic hydrophilic materials have been successful in prevent-
backbone
ing protein and bacteria adhesion in a wide range of
biomedical and industrial applications111,112.
Fouling resistance Fouling release Because of the adverse effects that fouling has
Water flow Water flow on membrane processes, the literature on develop-
ing fouling-resistant membranes is vast, and readers
are referred to recent comprehensive reviews113,114. A
few promising and widely applicable approaches for
designing fouling-resistant TFC and phase inversion
membranes are discussed below.

Anti-fouling TFC membranes by surface grafting.

Nature Reviews | Materials
Figure 4 | Polymeric materials and self-segregation approach for fouling
minimization in phase inversion membranes. In this approach, the phase inversion
casting solution contains a hydrophobic polymer and a comb copolymer with a
hydrophobic backbone and side chains that are at least partially hydrophilic. During
non-solvent-induced phase separation in water, the hydrophilic side chains
self-segregate at the membrane/water interface, whereas the backbone anchors itself in
the membrane matrix. a | Side chains are entirely hydrophilic, comprising poly(ethylene
oxide) or zwitterionic polymers. The resulting hydration layer inhibits irreversible
adsorption of organic and biological foulants. b | Side chains contain an additional
hydrophobic block with low surface energy. The low-surface-energy block promotes
foulant release in the presence of surface turbulence (for example, high cross-flows).
graphene oxide framework membranes may be effective
only for NF-range separations and not suitable for the
more urgently needed desalination applications.
Designing fouling-resistant membranes
As discussed earlier, phase inversion and TFC mem-
branes are prone to organic and biological fouling
because of their inherent surface chemical and physi-
cal properties. Proper tailoring of the surface chemical
properties of membranes can substantially enhance their
fouling resistance. Past work on the adsorption of pro-
teins to various surfaces provides useful insights into the
necessary surface properties that minimize adsorption of
organic foulants and microorganisms to membrane sur-
faces. Specifically, it was reported that surfaces with high
resistance to the adsorption of proteins are hydrophilic,
contain hydrogen bond acceptors, are electroneutral and
do not contain hydrogen bond donors109.
Surface grafting of hydrophilic polymers is one of the
most promising and versatile anti-fouling strategies
for TFC membranes. One approach involves binding
PEG-based or other anti-fouling materials to the free
surface carboxyl or primary amine groups of the poly­
amide layer 113,114. This method, however, often yields
limited fouling resistance owing to non-uniform and
incomplete surface coverage of anti-fouling materials
on the polyamide active layer, which is a direct result
of the limited number of accessible pendant carboxyl or
amine functional groups on the membrane surface115,116.
In addition, the method is limited even when using a
more ideal substrate, as steric repulsion of pre-formed
polymers limits the achievable brush density 117. A
more promising method for grafting dense and uni-
form anti-fouling polymer brushes on TFC membranes
involves the use of surface-initiated controlled radical
polymerization techniques. Of these techniques, atom
transfer radical polymerization (ATRP) is particularly
versatile and robust because it can precisely control
the density, length and architecture of the anti-fouling
brush layer 117. A recent study has successfully demon-
strated the use of ATRP to graft a dense brush layer of
poly(sulfobetaine methacrylate), a zwitterionic polymer,
to the polyamide surface of TFC RO membranes118. The
resulting membrane exhibited a very low water contact
angle (<10°), negligible adsorption of biomolecules (that
is, proteins), substantial reduction of foulant adhesion
forces to the membrane surface and a negligible change
in the permselectivity of the membrane.
Surface grafting often results in reduced water per-
meability and thus should be optimized to balance the
gain in fouling resistance and reduction in water perme-
ability 114. ATRP is advantageous in this regard because

12 | MAY 2016 | VOLUME 1 www.nature.com/natrevmats

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
 REVIEWS

of the control it allows over the density and thickness of
the brush layer. However, methods for uniform immo-
bilization of the initiator molecules on the rough poly-
amide surface of TFC membranes need to be developed
to ensure dense and uniform brush layers. Long-term
anti-fouling performance and chemical stability of the
grafted brush layer, particularly under chemical clean-
ing operations, are also important considerations for TFC
surface modification. Furthermore, although model sur-
faces and small membrane coupons have been used in the
laboratory, upscaling to membrane modules and reduc-
ing the cost of membrane fabrication remain challenging.
Anti-fouling porous membranes by surface segrega-
tion. When applied to UF membranes, surface grafting
leaves the interior pore walls unmodified, resulting in
adsorption of organic foulants to the pore walls and
irreversible fouling 2,119. A more robust surface modifi-
cation approach for phase inversion membranes is in situ
surface segregation119,120. In this approach, an amphi-
philic copolymer comprising a hydrophobic backbone
and hydrophilic side chains is blended with the hydro­
phobic base polymer in the casting solution. During
precipitation of the base polymer in the coagulation
bath, the hydrophobic backbone of the amphiphilic co­­
polymers ensures strong anchoring within the polymer
bulk matrix, whereas the hydrophilic segments segregate
to all polymer/water interfaces, creating a surface with
an exposed hydrophilic brush layer that imparts fouling
resistance to the entire membrane (FIG. 4a). This one-step
membrane fabrication method has been successfully
demonstrated with the use of PAN-graft-PEO — which is
an amphiphilic comb copolymer with a water-insoluble
PAN backbone and hydrophilic PEO side chains — as
an additive in the fabrication of a PAN UF membrane
that exhibits complete resistance to irreversible fouling
by three types of organic foulants2,120.
The in situ surface segregation method for fabricating
highly hydrophilic UF membranes was recently extended
to the fabrication of membranes with surface proper-
ties that provide both resistance to irreversible foulant
adsorption and enhanced foulant release121. The method
involves the use of a ternary amphiphilic block co­­
polymer that comprises a hydrophobic anchoring block,
a hydrophilic fouling-resistant block and a nonpolar,
low-surface-energy fouling-release block as an additive
during the phase inversion process. If the amphiphilic
copolymer is properly designed and the correct solvents
are chosen, the hydrophilic and fouling-release segments
are forced to segregate to the polymer/water interface.
This method has been successfully demonstrated for the
fabrication of cellulose acetate121, PVDF122, and PES123
UF membranes with surfaces containing hydrophilic
PEO-based blocks for fouling resistance and PDMS
or fluorinated-methacrylate-based polymer blocks for
fouling release (FIG. 4b). The membranes exhibit excellent
performance in inhibiting the nonspecific adsorption of
organic foulants as well as in preventing the accumula-
tion and coalescence of organic and oily foulants on the
membrane surface — when surface turbulence is present
— owing to the fouling release property.
Conclusion and outlook
In this Review, we identified the lack of control over
water and solute permeabilities as the key limitation
hampering membrane performance. This crucial con-
straint is attributed to current membrane fabrication
methods that still remain more of an art than a science.
Several novel approaches for molecular-level design of
isoporous films that can function as selective layers have
been discussed. With these next-generation membranes,
solute rejection in both UF and desalination (that is, RO
and FO) membranes will be based on a sieving or size-
exclusion mechanism. Salt retention by size exclu-
sion will represent a paradigm shift for desalination
membranes, as solute and water transport in current
TFC membranes is governed by a solution–diffusion
mechanism.
Substantial progress has been made in the fabri­
cation of isoporous UF separation membranes by some
of the molecular-level design approaches discussed in
this Review. Fabrication of block copolymer UF mem-
branes with uniform, high-density pores is now possi-
ble using scalable approaches, such as SNIPS. However,
highly selective, isoporous UF membranes may offer
only marginal improvements over conventional UF
membranes for water purification applications, as high
selectivity is not an important objective in the current
applications of UF as pre-treatment before RO or as
a standalone water purification system. Isoporous UF
membranes are likely to have an important role in
future applications that involve recovery of substances
from various wastewaters, as well as in other industrial
separations.
Next-generation, highly selective desalination
mem­branes will represent an important advancement.
Enhanced membrane selectivity will have an important
role in improving water quality and eliminating the need
for additional separation stages — for example, for low-
ering boric acid concentration in seawater desalination
to acceptable levels for agriculture — thereby reducing
energy usage and the cost of desalination. The molecular-
level design approaches for desalination membranes
discussed in this Review represent an active area of
research. Despite these major efforts, however, it is still
not possible to consistently fabricate even small-area
membranes with greater water permeability and selec-
tivity than current state‑of‑the-art TFC membranes. A
particular challenge for several of these materials will
be the formation of thin, defect-free selective layers
supported by microporous layers. After the concepts of
these novel membranes are successfully demonstrated
at the laboratory scale, upscaling the small-area mem-
branes to low-cost, industrial-scale modules will be an
additional challenge.
Fouling control in next-generation membranes, in
addition to current membranes, will remain a major
challenge, because novel materials for high separation
performance are often not compatible with fouling
resistance. Surface grafting of fouling-resistant poly-
mer brushes — such as PEO and zwitterionic polymers
— will probably remain the best approach for impart-
ing fouling resistance for next-generation, isoporous

NATURE REVIEWS | MATERIALS VOLUME 1 | MAY 2016 | 13

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS

membranes. For ‘tight’ UF membranes, surface grafting
will not impart fouling resistance to the interior pore
walls; hence, proper choice of new polymer chemistries
for membrane fabrication with fouling and chemical
resistance will be crucial. Next-generation desalination
membranes are likely to overcome another limitation
of current TFC membranes: the high sensitivity of the
polyamide selective layer to oxidants such as chlorine
and ozone. This will allow better control of biofouling,
which is the Achilles heel of membrane desalination.

1. Elimelech, M. The global challenge for adequate and
safe water. J. Water Supply Res. Technol. 55, 3–10
(2006).
2. Shannon, M. A. et al. Science and technology for
water purification in the coming decades. Nature 452,
301–310 (2008).
3. Schwarzenbach, R. P. et al. The challenge of
micropollutants in aquatic systems. Science 313,
1072–1077 (2006).
4. Kolpin, D. W. et al. Pharmaceuticals, hormones, and
other organic wastewater contaminants in U.S.
streams, 1999–2000: a national reconnaissance.
Environ. Sci. Technol. 36, 1202–1211 (2002).
5. Tang, J. Y., Busetti, F., Charrois, J. W. & Escher, B. I.
Which chemicals drive biological effects in wastewater
and recycled water? Water Res. 60, 289–299
(2014).
6. Vidic, R. D., Brantley, S. L., Vandenbossche, J. M.,
Yoxtheimer, D. & Abad, J. D. Impact of shale gas
development on regional water quality. Science 340,
1235009 (2013).
7. Shaffer, D. L. et al. Desalination and reuse of high-
salinity shale gas produced water: drivers,
technologies, and future directions. Environ. Sci.
Technol. 47, 9569–9583 (2013).
8. Gregory, K. B., Vidic, R. D. & Dzombak, D. A. Water
management challenges associated with the
production of shale gas by hydraulic fracturing.
Elements 7, 181–186 (2011).
9. Bond, R. & Veerapaneni, S. Zero liquid discharge for
inland desalination (Awwa Research Foundation,
2007).
10. Mickley, M. Survey of high-recovery and zero liquid
discharge technologies for water utilities (WateReuse
Foundation, 2008).
11. van Loosdrecht, M. C. M. & Brdjanovic, D.
Anticipating the next century of wastewater
treatment. Science 344, 1452–1453 (2014).
12. Grant, S. B. et al. Taking the “waste” out of
“wastewater” for human water security and ecosystem
sustainability. Science 337, 681–686 (2012).
13. Sales, C. M. & Lee, P. K. Resource recovery from
wastewater: application of meta-omics to phosphorus
and carbon management. Curr. Opin. Biotechnol. 33,
260–267 (2015).
14. Wang, X. et al. Probabilistic evaluation of integrating
resource recovery into wastewater treatment to
improve environmental sustainability. Proc. Natl Acad.
Sci. USA 112, 1630–1635 (2015).
15. Elimelech, M. & Phillip, W. A. The future of seawater
desalination: energy, technology, and the environment.
Science 333, 712–717 (2011).
16. Semiat, R. Energy issues in desalination processes.
Environ. Sci. Technol. 42, 8193–8201 (2008).
17. Baker, R. W. Membrane Technology and Applications
(John Wiley & Sons, 2012).
18. Gin, D. L. & Noble, R. D. Designing the next
generation of chemical separation membranes.
Science 332, 674–676 (2011).
19. Tang, C. Y., Kwon, Y.‑N. & Leckie, J. O. Effect of
membrane chemistry and coating layer on
physiochemical properties of thin film composite
polyamide RO and NF membranes. Desalination 242,
168–182 (2009).
20. Lu, X. et al. Elements provide a clue: nanoscale
characterization of thin-film composite polyamide
membranes. ACS Appl. Mater. Interfaces 7,
16917–16922 (2015).
21. Karan, S., Jiang, Z. & Livingston, A. G. Sub‑10 nm
polyamide nanofilms with ultrafast solvent transport
for molecular separation. Science 348, 1347–1351
(2015).
22. Kwak, S. Y., Jung, S. G. & Kim, S. H. Structure-motion-
performance relationship of flux-enhanced reverse
osmosis (RO) membranes composed of aromatic
polyamide thin films. Environ. Sci. Technol. 35,
4334–4340 (2001).
23. Mehta, A. & Zydney, A. L. Permeability and selectivity
analysis for ultrafiltration membranes. J. Membr. Sci.
249, 245–249 (2005).
24. Geise, G. M., Paul, D. R. & Freeman, B. D.
Fundamental water and salt transport properties of
polymeric materials. Prog. Polym. Sci. 39, 1–42
(2014).
25. Geise, G. M., Park, H. B., Sagle, A. C., Freeman, B. D.
& McGrath, J. E. Water permeability and water/salt
selectivity tradeoff in polymers for desalination.
J. Membr. Sci. 369, 130–138 (2011).
A water–salt selectivity trade-off for desalination
using polymeric membranes is proposed for the
first time in this study.
26. Robeson, L. M. The upper bound revisited. J. Membr.
Sci. 320, 390–400 (2008).
27. Yip, N. Y. & Elimelech, M. Performance limiting effects
in power generation from salinity gradients by
pressure retarded osmosis. Environ. Sci. Technol. 45,
10273–10282 (2011).
28. Iwahashi, H. et al. Advanced RO system for high
temperature and high concentration seawater
desalination at the Arabian Gulf. IDA World Congress
(San Diego, 2015).
29. Cohen-Tanugi, D., McGovern, R. K., Dave, S. H.,
Lienhard, J. H. & Grossman, J. C. Quantifying the
potential of ultra-permeable membranes for water
desalination. Energy Environ. Sci. 7, 1134–1141 (2014).
30. Deshmukh, A., Yip, N. Y., Lin, S. & Elimelech, M.
Desalination by forward osmosis: identifying
performance limiting parameters through module-
scale modeling. J. Membr. Sci. 491, 159–167
(2015).
31. Singh, R. Production of high-purity water by membrane
processes. Desalin. Water Treat. 3, 99–110 (2012).
32. Miyashita, Y., Park, S.-H., Hyung, H., Huang, C.-H. &
Kim, J.-H. Removal of N‑nitrosamines and their
precursors by nanofiltration and reverse osmosis
membranes. J. Environ. Eng. 135, 788–795 (2009).
33. Ozaki, H. & Li, H. Rejection of organic compounds by
ultra-low pressure reverse osmosis membrane. Water
Res. 36, 123–130 (2002).
34. Fritzmann, C., Löwenberg, J., Wintgens, T. & Melin, T.
State‑of‑the-art of reverse osmosis desalination.
Desalination 216, 1–76 (2007).
35. Le‑Clech, P., Chen, V. & Fane, T. A. G. Fouling in
membrane bioreactors used in wastewater treatment.
J. Membr. Sci. 284, 17–53 (2006).
36. Cheryan, M. Ultrafiltration and Microfiltration
Handbook 31–70 (CRC Press, 1998).
37. Matthiasson, E. The role of macromolecular
adsorption in fouling of ultrafiltration membranes.
J. Membr. Sci. 16, 23–36 (1983).
38. Belfort, G., Davis, R. H. & Zydney, A. L. The behavior of
suspensions and macromolecular solutions in crossflow
microfiltration. J. Membr. Sci. 96, 1–58 (1994).
39. Elimelech, M., Zhu, X., Childress, A. E. & Hong, S. Role
of membrane surface morphology in colloidal fouling
of cellulose acetate and composite aromatic
polyamide reverse osmosis membranes. J. Membr.
Sci. 127, 101–109 (1997).
The roughness of polyamide thin-film composite
membrane surfaces is shown in this study to
exacerbate fouling.
40. Herzberg, M. & Elimelech, M. Biofouling of reverse
osmosis membranes: role of biofilm-enhanced osmotic
pressure. J. Membr. Sci. 295, 11–20 (2007).
41. Li, Q. & Elimelech, M. Organic fouling and chemical
cleaning of nanofiltration membranes: measurements
and mechanisms. Environ. Sci. Technol. 38,
4683–4693 (2004).
42. Mi, B. & Elimelech, M. Gypsum scaling and cleaning in
forward osmosis: measurements and mechanisms.
Environ. Sci. Technol. 44, 2022–2028 (2010).
43. Borgnia, M. J., Kozono, D., Calamita, G.,
Maloney, P. C. & Agre, P. Functional reconstitution and
characterization of AqpZ, the E. coli water channel
protein. J. Mol. Biol. 291, 1169–1179 (1999).
44. Beitz, E., Wu, B., Holm, L. M., Schultz, J. E. &
Zeuthen, T. Point mutations in the aromatic/arginine
region in aquaporin 1 allow passage of urea, glycerol,
ammonia, and protons. Proc. Natl Acad. Sci. USA
103, 269–274 (2006).
45. Hub, J. S. & de Groot, B. L. Mechanism of selectivity in
aquaporins and aquaglyceroporins. Proc. Natl Acad.
Sci. USA 105, 1198–1203 (2008).
46. Sui, H., Han, B. G., Lee, J. K., Walian, P. & Jap, B. K.
Structural basis of water-specific transport through the
AQP1 water channel. Nature 414, 872–878 (2001).
In this X‑ray crystallography study, a high-resolution
structure of aquaporin 1 was attained, which
elucidated the basis for its selectivity for water.
47. Murata, K. et al. Structural determinants of water
permeation through aquaporin‑1. Nature 407,
599–605 (2000).
48. Savage, D. F., O’Connell, J. D., Miercke, L. J. W., Finer-
Moore, J. & Stroud, R. M. Structural context shapes
the aquaporin selectivity filter. Proc. Natl Acad. Sci.
USA 107, 17164–17169 (2010).
49. Wang, M. et al. Layer‑by‑layer assembly of aquaporin
Z‑incorporated biomimetic membranes for water
purification. Environ. Sci. Technol. 49, 3761–3768
(2015).
50. Wang, H. et al. Highly permeable and selective pore-
spanning biomimetic membrane embedded with
aquaporin Z. Small 8, 1185–1190 (2012).
51. Kumar, M., Grzelakowski, M., Zilles, J., Clark, M. &
Meier, W. Highly permeable polymeric membranes
based on the incorporation of the functional water
channel protein aquaporin Z. Proc. Natl Acad. Sci.
USA 104, 20719–20724 (2007).
52. Zhong, P. S., Chung, T.‑S., Jeyaseelan, K. & Armugam, A.
Aquaporin-embedded biomimetic membranes for
nanofiltration. J. Membr. Sci. 407–408, 27–33 (2012).
53. Hummer, G., Rasaiah, J. C. & Noworyta, J. P. Water
conduction through the hydrophobic channel of a
carbon nanotube. Nature 414, 188–190 (2001).
This study provides the first demonstration, using
molecular dynamics, of ultra-fast water permeation
through the interior of carbon nanotubes.
54. Corry, B. Designing carbon nanotube membranes for
efficient water desalination. J. Phys. Chem. B 112,
1427–1434 (2008).
55. Majumder, M., Chopra, N., Andrews, R. & Hinds, B. J.
Nanoscale hydrodynamics: enhanced flow in carbon
nanotubes. Nature 438, 44 (2005).
56. Holt, J. K. et al. Fast mass transport through
sub‑2‑nanometer carbon nanotubes. Science 312,
1034–1037 (2006).
57. Hinds, B. J. et al. Aligned multiwalled carbon
nanotube membranes. Science 303, 62–65 (2004).
58. Majumder, M., Chopra, N. & Hinds, B. J. Effect of tip
functionalization on transport through vertically
oriented carbon nanotube membranes. J. Am. Chem.
Soc. 127, 9062–9070 (2005).
59. Fornasiero, F. et al. Ion exclusion by sub‑2‑nm carbon
nanotube pores. Proc. Natl Acad. Sci. USA 105,
17250–17255 (2008).
60. Corry, B. Water and ion transport through functionalised
carbon nanotubes: implications for desalination
technology. Energy Environ. Sci. 4, 751–759 (2011).
61. Chan, W. F. et al. Zwitterion functionalized carbon
nanotube/polyamide nanocomposite membranes for
water desalination. ACS Nano 7, 5308–5319 (2013).
62. Sanchez-Valencia, J. R. et al. Controlled synthesis of
single-chirality carbon nanotubes. Nature 512,
61–64 (2014).
63. Shen, Y. X. et al. Highly permeable artificial water
channels that can self-assemble into two-dimensional
arrays. Proc. Natl Acad. Sci. USA 112, 9810–9815
(2015).
64. Hu, X. B., Chen, Z., Tang, G., Hou, J. L. & Li, Z. T. Single-
molecular artificial transmembrane water channels.
J. Am. Chem. Soc. 134, 8384–8387 (2012).
65. Hourani, R. et al. Processable cyclic peptide
nanotubes with tunable interiors. J. Am. Chem. Soc.
133, 15296–15299 (2011).
66. Zhou, X. et al. Self-assembling subnanometer pores
with unusual mass-transport properties. Nat.
Commun. 3, 949 (2012).
Highly selective artificial water channels that
operate by rigid molecular sieving are reported in
this study.

14 | MAY 2016 | VOLUME 1 www.nature.com/natrevmats

©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.

67. Mauter, M. S., Elimelech, M. & Osuji, C. O.
Nanocomposites of vertically aligned single-walled
carbon nanotubes by magnetic alignment and
polymerization of a lyotropic precursor. ACS Nano 4,
6651–6658 (2010).
68. Xu, T. et al. Subnanometer porous thin films by the
co‑assembly of nanotube subunits and block
copolymers. ACS Nano 5, 1376–1384 (2011).
69. Gin, D. L., Gu, W., Pindzola, B. A. & Zhou, W. J.
Polymerized lyotropic liquid crystal assemblies for
materials applications. Acc. Chem. Res. 34, 973–980
(2001).
70. Gin, D. L., Bara, J. E., Noble, R. D. & Elliott, B. J.
Polymerized lyotropic liquid crystal assemblies for
membrane applications. Macromol. Rapid. Commun.
29, 367–389 (2008).
71. Zhang, Y., Sargent, J. L., Boudouris, B. W. &
Phillip, W. A. Nanoporous membranes generated from
self-assembled block polymer precursors: quo vadis?
J. Appl. Polym. Sci. 132, 41683 (2015).
72. Jackson, E. A. & Hillmyer, M. A. Nanoporous
membranes derived from block copolymers: from drug
delivery to water filtration. ACS Nano 4, 3548–3553
(2010).
73. Zalusky, A. S., Olayo-Valles, R., Wolf, J. H. &
Hillmyer, M. A. Ordered nanoporous polymers from
polystyrene–polylactide block copolymers. J. Am.
Chem. Soc. 124, 12761–12773 (2002).
74. Sorenson, G. P., Coppage, K. L. & Mahanthappa, M. K.
Unusually stable aqueous lyotropic gyroid phases from
gemini dicarboxylate surfactants. J. Am. Chem. Soc.
133, 14928–14931 (2011).
75. Hatakeyama, E. S., Wiesenauer, B. R., Gabriel, C. J.,
Noble, R. D. & Gin, D. L. Nanoporous, bicontinuous
cubic lyotropic liquid crystal networks via
polymerizable gemini ammonium surfactants. Chem.
Mater. 22, 4525–4527 (2010).
76. Zhou, M. et al. New type of membrane material for
water desalination based on a cross-linked
bicontinuous cubic lyotropic liquid crystal assembly.
J. Am. Chem. Soc. 129, 9574–9575 (2007).
77. Pindzola, B. A., Jin, J. & Gin, D. L. Cross-linked normal
hexagonal and bicontinuous cubic assemblies via
polymerizable gemini amphiphiles. J. Am. Chem. Soc.
125, 2940–2949 (2003).
78. Soberats, B. et al. 3D Anhydrous proton-transporting
nanochannels formed by self-assembly of liquid
crystals composed of a sulfobetaine and a sulfonic
acid. J. Am. Chem. Soc. 135, 15286–15289 (2013).
79. Kerr, R. L., Miller, S. A., Shoemaker, R. K., Elliott, B. J.
& Gin, D. L. New type of Li ion conductor with 3D
interconnected nanopores via polymerization of a
liquid organic electrolyte-filled lyotropic liquid-crystal
assembly. J. Am. Chem. Soc. 131, 15972–15973
(2009).
80. Smith, R. C., Fischer, W. M. & Gin, D. L. Ordered
poly(p‑phenylenevinylene) matrix nanocomposites via
lyotropic liquid-crystalline monomers. J. Am. Chem.
Soc. 119, 4092–4093 (1997).
This is an early demonstration of the formation of
a nanoporous polymeric structure using surfactants
based on polymerizable gallic acid.
81. Zhou, M., Kidd, T. J., Noble, R. D. & Gin, D. L.
Supported lyotropic liquid-crystal polymer
membranes: promising materials for molecular-size-
selective aqueous nanofiltration. Adv. Mater. 17,
1850–1853 (2005).
82. Broer, D. J., Bastiaansen, C. M., Debije, M. G. &
Schenning, A. P. Functional organic materials based
on polymerized liquid-crystal monomers:
supramolecular hydrogen-bonded systems. Angew.
Chem. Int. Ed. Engl. 51, 7102–7109 (2012).
83. Henmi, M. et al. Self-organized liquid-crystalline
nanostructured membranes for water treatment:
selective permeation of ions. Adv. Mater. 24,
2238–2241 (2012).
84. Deng, H., Gin, D. L. & Smith, R. C. Polymerizable
lyotropic liquid crystals containing transition-metal and
lanthanide ions: architectural control and introduction
of new properties into nanostructured polymers. J. Am.
Chem. Soc. 120, 3522–3523 (1998).
85. Lee, H.‑K. et al. Synthesis of a nanoporous polymer
with hexagonal channels from supramolecular discotic
liquid crystals. Angew. Chem. Int. Ed. Engl. 40,
2669–2671 (2001).
86. Ishida, Y. et al. Guest-responsive covalent frameworks
by the cross-linking of liquid-crystalline salts: tuning of
lattice flexibility by the design of polymerizable units.
Chemistry 17, 14752–14762 (2011).
87. Feng, X. et al. Scalable fabrication of polymer
membranes with vertically aligned 1 nm pores by
NATURE REVIEWS | MATERIALS VOLUME 1 | MAY 2016 | 15
magnetic field directed self-assembly. ACS Nano 8,
11977–11986 (2014).
88. Gopinadhan, M. et al. Thermally switchable aligned
nanopores by magnetic-field directed self-assembly of
block copolymers. Adv. Mater. 26, 5148–5154 (2014).
89. Feng, X. et al. Thin polymer films with continuous
vertically aligned 1 nm pores fabricated by soft
confinement. ACS Nano 10, 150–158 (2015).
90. Peinemann, K. V., Abetz, V. & Simon, P. F. Asymmetric
superstructure formed in a block copolymer via phase
separation. Nat. Mater. 6, 992–996 (2007).
In this study, block copolymer self-assembly and
phase inversion are combined for the first time to
readily form porous membranes with a low pore
size polydispersity.
91. Phillip, W. A. et al. Tuning structure and properties of
graded triblock terpolymer-based mesoporous and
hybrid films. Nano Lett. 11, 2892–2900 (2011).
92. Gu, Y. & Wiesner, U. Tailoring pore size of graded
mesoporous block copolymer membranes: moving
from ultrafiltration toward nanofiltration.
Macromolecules 48, 6153–6159 (2015).
93. Clodt, J. I. et al. Performance study of isoporous
membranes with tailored pore sizes. J. Membr. Sci.
495, 334–340 (2015).
94. Seo, M. & Hillmyer, M. A. Reticulated nanoporous
polymers by controlled polymerization-induced
microphase separation. Science 336, 1422–1425
(2012).
95. Joshi, R. K. et al. Precise and ultrafast molecular
sieving through graphene oxide membranes. Science
343, 752–754 (2014).
This is an experimental demonstration of molecular
sieving through graphene oxide laminates, albeit
with a size cut-off that is too large for desalination.
96. Nair, R. R., Wu, H. A., Jayaram, P. N., Grigorieva, I. V.
& Geim, A. K. Unimpeded permeation of water
through helium-leak-tight graphene-based
membranes. Science 335, 442–444 (2012).
97. Surwade, S. P. et al. Water desalination using
nanoporous single-layer graphene. Nat. Nanotechnol.
10, 459–464 (2015).
98. Russo, C. J. & Golovchenko, J. A. Atom‑by‑atom
nucleation and growth of graphene nanopores. Proc.
Natl Acad. Sci. USA 109, 5953–5957 (2012).
99. Cohen-Tanugi, D. & Grossman, J. C. Water
desalination across nanoporous graphene. Nano Lett.
12, 3602–3608 (2012).
100. O’Hern, S. C. et al. Selective ionic transport through
tunable subnanometer pores in single-layer graphene
membranes. Nano Lett. 14, 1234–1241 (2014).
101. O’Hern, S. C. et al. Nanofiltration across defect-sealed
nanoporous monolayer graphene. Nano Lett. 15,
3254–3260 (2015).
102. O’Hern, S. C. et al. Selective molecular transport
through intrinsic defects in a single layer of CVD
graphene. ACS Nano 6, 10130–10138 (2012).
103. Mi, B. Graphene oxide membranes for ionic and
molecular sieving. Science 343, 740–742 (2014).
104. Yeh, C. N., Raidongia, K., Shao, J., Yang, Q. H. &
Huang, J. On the origin of the stability of graphene
oxide membranes in water. Nat. Chem. 7, 166–170
(2014).
105. Su, Y. et al. Impermeable barrier films and protective
coatings based on reduced graphene oxide. Nat.
Commun. 5, 4843 (2014).
106. Liu, H., Wang, H. & Zhang, X. Facile fabrication of
freestanding ultrathin reduced graphene oxide
membranes for water purification. Adv. Mater. 27,
249–254 (2015).
107. Hung, W.‑S. et al. Cross-linking with diamine
monomers to prepare composite graphene oxide-
framework membranes with varying d‑spacing. Chem.
Mater. 26, 2983–2990 (2014).
108. Zhang, Y., Zhang, S. & Chung, T. S. Nanometric
graphene oxide framework membranes with enhanced
heavy metal removal via nanofiltration. Environ. Sci.
Technol. 49, 10235–10242 (2015).
109. Ostuni, E., Chapman, R. G., Holmlin, R. E., Takayama, S.
& Whitesides, G. M. A survey of structure–property
relationships of surfaces that resist the adsorption of
protein. Langmuir 17, 5605–5620 (2001).
Through adsorption measurements of proteins on
model surfaces with defined functional groups, this
study describes general characteristics of
non-fouling surfaces.
110. Banerjee, I., Pangule, R. C. & Kane, R. S. Antifouling
coatings: recent developments in the design of
surfaces that prevent fouling by proteins, bacteria,
and marine organisms. Adv. Mater. 23, 690–718
(2011).
©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.
REVIEWS
111. Jiang, S. & Cao, Z. Ultralow-fouling, functionalizable,
and hydrolyzable zwitterionic materials and their
derivatives for biological applications. Adv. Mater. 22,
920–932 (2010).
112. Callow, J. A. & Callow, M. E. Trends in the development
of environmentally friendly fouling-resistant marine
coatings. Nat. Commun. 2, 244 (2011).
113. Kang, G. D. & Cao, Y. M. Development of antifouling
reverse osmosis membranes for water treatment: a
review. Water Res. 46, 584–600 (2012).
114. Rana, D. & Matsuura, T. Surface modifications for
antifouling membranes. Chem. Rev. 110, 2448–2471
(2010).
115. Shaffer, D. L., Jaramillo, H., Romero-Vargas Castrillón, S.,
Lu, X. & Elimelech, M. Post-fabrication modification of
forward osmosis membranes with a poly(ethylene glycol)
block copolymer for improved organic fouling resistance.
J. Membr. Sci. 490, 209–219 (2015).
116. Van Wagner, E. M., Sagle, A. C., Sharma, M. M.,
La, Y.‑H. & Freeman, B. D. Surface modification of
commercial polyamide desalination membranes using
poly(ethylene glycol) diglycidyl ether to enhance
membrane fouling resistance. J. Membr. Sci. 367,
273–287 (2011).
117. Barbey, R. et al. Polymer brushes via surface-initiated
controlled radical polymerization: synthesis,
characterization, properties, and applications. Chem.
Rev. 109, 5437–5527 (2009).
118. Ye, G., Lee, J., Perreault, F. & Elimelech, M. Controlled
architecture of dual-functional block copolymer brushes
on thin-film composite membranes for integrated
“defending” and “attacking” strategies against biofouling.
ACS Appl. Mater. Interfaces 7, 23069–23079 (2015).
119. Kang, S., Asatekin, A., Mayes, A. M. & Elimelech, M.
Protein antifouling mechanisms of PAN UF
membranes incorporating PAN‑g‑PEO additive.
J. Membr. Sci. 296, 42–50 (2007).
120. Asatekin, A., Kang, S., Elimelech, M. & Mayes, A. M.
Anti-fouling ultrafiltration membranes containing
polyacrylonitrile-graft-poly(ethylene oxide) comb
copolymer additives. J. Membr. Sci. 298, 136–146
(2007).
The in situ surface segregation approach is used in
this study to readily fabricate fouling-resistant
polyacrylonitrile UF membranes.
121. Chen, W. et al. Engineering a robust, versatile
amphiphilic membrane surface through forced surface
segregation for ultralow flux-decline. Adv. Funct.
Mater. 21, 191–198 (2011).
122. Chen, W. et al. Efficient wastewater treatment by
membranes through constructing tunable antifouling
membrane surfaces. Environ. Sci. Technol. 45,
6545–6552 (2011).
123. Zhao, X. et al. Engineering amphiphilic membrane
surfaces based on PEO and PDMS segments for
improved antifouling performances. J. Membr. Sci.
450, 111–123 (2014).
124. Shaffer, D. L., Werber, J. R., Jaramillo, H., Lin, S. &
Elimelech, M. Forward osmosis: where are we now?
Desalination 356, 271–284 (2015).
125. Tiraferri, A., Yip, N. Y., Phillip, W. A., Schiffman, J. D.
& Elimelech, M. Relating performance of thin-film
composite forward osmosis membranes to support
layer formation and structure. J. Membr. Sci. 367,
340–352 (2011).
126. Lu, X., Arias Chavez, L. H., Romero-Vargas
Castrillón, S., Ma, J. & Elimelech, M. Influence of
active layer and support layer surface structures on
organic fouling propensity of thin-film composite
forward osmosis membranes. Environ. Sci. Technol.
49, 1436–1444 (2015).
127. Pohl, P., Saparov, S. M., Borgnia, M. J. & Agre, P.
Highly selective water channel activity measured by
voltage clamp: analysis of planar lipid bilayers
reconstituted with purified AqpZ. Proc. Natl Acad.
Sci. USA 98, 9624–9629 (2001).
128. Ruiz, L., Wu, Y. & Keten, S. Tailoring the water
structure and transport in nanotubes with tunable
interiors. Nanoscale 7, 121–132 (2015).
Acknowledgements
The authors acknowledge support received from the US
National Science Foundation (NSF) under award numbers
CBET-1437630 and CMMI-1246804 and through the NSF
Nanosystems Engineering Research Center for
Nanotechnology-Enabled Water Treatment (ERC‑1449500).
The authors also acknowledge the NSF Graduate Research
Fellowship awarded to J.R.W.
Competing interests statement
The authors declare no competing interests.
 O N L I N E O N LY

Figure permissions
Figure 1
The upper panels of part b are adapted with permission from REF. 125,
Elsevier. The upper panel of part d is adapted with permission from
REF. 126, American Chemical Society. The lower panels of part d are
adapted with permission from REF. 20, American Chemical

Figure 3
Scanning electron microscopy image in panel b is from REF. 90, Nature
Publishing Group.

Subject Terms
Physical sciences / Materials science / Nanoscale materials [URI
/639/301/357]
Physical sciences / Engineering / Chemical engineering
[URI /639/166/898]
Physical sciences / Chemistry / Materials chemistry / Soft materials
/ Polymers
[URI /639/638/298/923/1028]
Physical sciences / Chemistry / Materials chemistry / Graphene
[URI /639/638/298/918]
Physical sciences / Materials science / Nanoscale materials / Carbon
nanotubes and fullerenes
[URI /639/301/357/73]
Physical sciences / Chemistry / Environmental chemistry / Pollution
remediation
[URI /639/638/169/896]
Earth and environmental sciences / Environmental sciences /
Environmental chemistry / Pollution remediation
[URI /704/172/169/896]
Physical sciences / Chemistry / Environmental chemistry
[URI /639/638/169]

Web summary

000 Materials for next-generation desalination

and water purification membranes
Jay R. Werber, Chinedum O. Osuji and Menachem
Elimelech
Membranes have an increasingly important role in
alleviating water scarcity and the pollution of aquatic
environments. Promising molecular-level design
approaches are reviewed for membrane materials,
focusing on how these materials address the urgent
requirements of water treatment applications.
©
2
0
1
6
M
a
c
m
i
l
l
a
n
P
u
b
l
i
s
h
e
r
s
L
i
m
i
t
e
d
.
A
l
l
r
i
g
h
t
s
r
e
s
e
r
v
e
d
.