Large-Area Graphene-Based Nanofiltration Membranes by Shear
Transcript Of Large-Area Graphene-Based Nanofiltration Membranes by Shear
University of Kentucky
UKnowledge
Chemical and Materials Engineering Faculty Publications
Chemical and Materials Engineering
3-7-2016
Large-Area Graphene-Based Nanofiltration Membranes by Shear Alignment of Discotic Nematic Liquid Crystals of Graphene Oxide
Abozar Akbari Monash University, Australia
Phillip Sheath Monash University, Australia
Samuel T. Martin Monash University, Australia
Dhanraj B. Shinde Monash University, Australia
Mahdokht Shaibani Monash University, Australia
See next page for additional authors Follow this and additional works at: https://uknowledge.uky.edu/cme_facpub
Part of the Chemical Engineering Commons, and the Materials Science and Engineering Commons Right click to open a feedback form in a new tab to let us know how this document benefits you.
Repository Citation Akbari, Abozar; Sheath, Phillip; Martin, Samuel T.; Shinde, Dhanraj B.; Shaibani, Mahdokht; Banerjee, Parama Chakraborty; Tkacz, Rachel; Bhattacharyya, Dibakar; and Majumder, Mainak, "Large-Area Graphene-Based Nanofiltration Membranes by Shear Alignment of Discotic Nematic Liquid Crystals of Graphene Oxide" (2016). Chemical and Materials Engineering Faculty Publications. 19. https://uknowledge.uky.edu/cme_facpub/19
This Article is brought to you for free and open access by the Chemical and Materials Engineering at UKnowledge. It has been accepted for inclusion in Chemical and Materials Engineering Faculty Publications by an authorized administrator of UKnowledge. For more information, please contact [email protected]
Large-Area Graphene-Based Nanofiltration Membranes by Shear Alignment of Discotic Nematic Liquid Crystals of Graphene Oxide Digital Object Identifier (DOI)
https://doi.org/10.1038/ncomms10891 Notes/Citation Information Published in Nature Communications, v. 7, article no. 10891, p. 1-12. © The Author(s) 2017 This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ Authors Abozar Akbari, Phillip Sheath, Samuel T. Martin, Dhanraj B. Shinde, Mahdokht Shaibani, Parama Chakraborty Banerjee, Rachel Tkacz, Dibakar Bhattacharyya, and Mainak Majumder
This article is available at UKnowledge: https://uknowledge.uky.edu/cme_facpub/19
ARTICLE
Received 1 Sep 2015 | Accepted 29 Jan 2016 | Published 7 Mar 2016
DOI: 10.1038/ncomms10891
OPEN
Large-area graphene-based nanofiltration membranes by shear alignment of discotic nematic liquid crystals of graphene oxide
Abozar Akbari1, Phillip Sheath1, Samuel T. Martin1, Dhanraj B. Shinde1, Mahdokht Shaibani1, Parama Chakraborty Banerjee1, Rachel Tkacz1, Dibakar Bhattacharyya2 & Mainak Majumder1
Graphene-based membranes demonstrating ultrafast water transport, precise molecular sieving of gas and solvated molecules shows great promise as novel separation platforms; however, scale-up of these membranes to large-areas remains an unresolved problem. Here we demonstrate that the discotic nematic phase of graphene oxide (GO) can be shear aligned to form highly ordered, continuous, thin films of multi-layered GO on a support membrane by an industrially adaptable method to produce large-area membranes (13 Â 14 cm2) in o5 s. Pressure driven transport data demonstrate high retention (490%) for charged and uncharged organic probe molecules with a hydrated radius above 5 Å as well as modest (30–40%) retention of monovalent and divalent salts. The highly ordered graphene sheets in the plane of the membrane make organized channels and enhance the permeability (71±5 l m À 2 hr À 1 bar À 1 for 150±15 nm thick membranes).
1 Department of Mechanical and Aerospace Engineering, Nanoscale Science and Engineering Laboratory (NSEL), Monash University, Clayton, Victoria 3800, Australia. 2 Department of Chemical and Materials Engineering, University of Kentucky, Lexington, Kentucky 40506, USA. Correspondence and requests for
materials should be addressed to M.M. (email: [email protected]).
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
1
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
Advances in the design and synthesis of nanofiltration membranes with improved retention, flux and cost-effectiveness will have tremendous impact in several fields such as water treatment, selective chemical separations and drug delivery. Conventional polymeric nanofiltration membranes usually have limited chemical resistance, while ceramic membranes are not cost-efficient. Graphene is a one atom thick two-dimensional honeycomb sp2 carbon lattice, which is an exciting multifunctional material and possesses a combination of strong mechanical properties, chemical inertness and extremely large surface area1,2. Membranes prepared from graphene possess the best of both the worlds: they are chemically inert3 like ceramic membranes and can be made into films using graphene/graphene oxide (GO) fluid phase dispersions like polymers. Novel and exciting transport properties of graphene-based membranes such as high permeability and high selectivity for both liquids2,4–7 and gases8–11 have recently been reported. While these studies have unlocked potential applications, there is critical need to produce these membranes in large-areas using high throughput manufacturing routes, which may otherwise hinder their impact in membrane technologies. The ideal structure of a filtration membrane has a defect-free, thin, dense separation film that acts as a functional sieve, while the mechanical strength is provided by a porous and more permeable support. To achieve this asymmetric structure, researchers have grown continuous graphene films by chemical vapour deposition and transferred them to substrates followed by etching pores on the film, however, the transfer process limits the scalability of membrane production6,7. Another method to produce this structure is by restacking GO flakes by filtration of GO dispersions on a backing filter support2,12–14. However, producing a membrane by this approach requires large volumes of liquid, significant time and arguably has both alignment (of the GO sheets) and scalability issues. Other liquid phase processes such as dip-coating or layer-by-layer assembly similarly have potential issues with rapid productivity15. Therefore, a major challenge in this field is to define robust, scalable, liquid film processing approaches to produce large-area graphene-based membranes that will bridge laboratory curiosity to industrial productivity.
Here we introduce a scalable and industrially adaptable method to fabricate large-area graphene-based membranes by shear-induced alignment of liquid crystals of GO. The membranes have large in-plane stacking order of GO sheets and demonstrate outstanding water permeability while being able to sieve small organic molecules with performance metrics superior to well established and commercially available nanofiltration membrane.
Results Producing liquid crystalline phases of GO. Oxidation and exfoliation of graphite by the well-known Hummers’ method or its variations produces graphene nanosheets decorated with oxygenated functional groups also known as GO. The anisotropic GO nanosheets can be dispersed in liquids including water as stable colloidal suspensions with large volume fractions. As the concentration of the anisotropic particles increases, the orientation entropy of the suspensions starts to decrease only to be compensated by increase in the translation entropy leading to colloidal phase transitions from isotropic to nematic liquid crystalline phases—the onset of which has dependence on the thickness to diameter ratio of the disc-like mesogens of GO16. Liquid crystallinity defines a state between a crystal and a fluid, within which the constituent sheets become anisotropic but can still flow and respond to macroscopic force-fields such as shear17, and this state has been demonstrated in concentrated dispersions of GO1,18–20. Traditional means to produce concentrated GO dispersions, such as the application of heat21 or the use of vacuum equipment22, are time consuming and laborious. An innovative method was implemented in this work to quickly produce nematic GO dispersions (Fig. 1). We used superabsorbent polymer hydrogel beads (typically, cross-linked polyacrylate based copolymer), which are strongly hydrophilic. Concentration of a GO dispersion occurs because the hydrogel beads absorb and retain water23 without dissolving in water or absorbing GO sheets. This is demonstrated in Raman characterization in Supplementary Fig. 1 (Supplementary Note 1)—the characteristic peaks of GO were not observed
Add beads Remove
swollen beads
0.25 mg ml–1
Repeat until desired concentration
20 mg ml–1
Figure 1 | Procedure for concentrating GO. Photographs of stable dispersions of GO which have been concentrated by adding hydrogel beads.
2
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
ARTICLE
within the hydrogel beads swollen in a GO suspension. The time taken to concentrate a GO dispersion depends on the initial concentration, the desired concentration and the mass of beads used. Fig. 2a shows a GO dispersion with a concentration of 40 mg ml À 1.
Rheological characterization of GO dispersions. Rheological properties of the GO dispersions are crucial to our fabrication method. We evaluated zero-shear viscosity by measuring the viscosity of the GO dispersions at a shear rate of 0.001 s À 1 (ref. 24). Fig. 2b demonstrates that the zero-shear viscosity increases with an increase in GO concentration. At low concentrations of GO, water molecules are attached to GO sheets via hydrogen bonds25, but similar to water-clay dispersions26, an increase in the number of GO sheets results in the assembly of graphene sheets and water molecules to form a three-dimensional network via hydrogen bonding, which decreases the fluidity of the dispersion. Furthermore, the large changes in the viscosity beyond 5 mg ml À 1 coincides with the onset of liquid crystalline nematic phase16. Fig. 2c presents apparent viscosity of the GO dispersion (Z) as a function of shear rate (g˚). The non-Newtonian
shear-thinning (pseudoplastic) behaviour was observed at different concentrations of the GO dispersion. Decreased viscosity of the GO dispersions with an increase in the shear rates is consistent with previous reports18,27. One can presume that the nematic phases in the GO dispersion are distributed randomly and do not align at low shear rates, which results in higher viscosity. At high shear rates, the randomly distributed nematic phases align in the direction of shear stress and produces less physical interaction with each other, resulting in decreased viscosity. Fig. 2c shows that the viscosity of the GO dispersions were in good agreement with the power law viscosity model. In the power law model, the exponent for ideal plastic material is À 1 and any deviation from this theoretical value shows a loss of plastic behaviour28. The exponents decrease from À 0.580 to À 0.867 by increasing GO concentration from 10 to 40 mg ml À 1, which affirms the increased plasticity arising from the nematic GO phases.
Interfacial properties of GO dispersion. Properties of the GO dispersion, typically at the solid–liquid interface, also assume importance in our membrane fabrication process. Several
Zero-shear viscosity (Pa s) (Pa s)
a
d
(y) H h0
Doctor blade
b
300
250 200 150 100
5 4 3 2 1 0
0 5 10 15 20
50
0 0
Nematic phase GO
10 20 30 40 50 60 GO concentration (mg ml–1)
e
f 10 mg ml–1
Process speed (U) L
40 mg ml–1
c 100
10 1
0.1 0.01 0.001
0
g
10 mg ml–1 20 mg ml–1 40 mg ml–1 ∝ γ° –0.868 ∝ γ°–0.666
∝ γ°–0.580
20
40
60
80 100
° (s–1)
Fixed substrate
(x)
h
i
832 nm
j 180
150
Height (nm)
150±15 nm
120
90
Profile 1
60
Profile 2
30
Profile 3
0 nm
0 0 200 400 600 800 1,000
Distance (nm)
Figure 2 | Fabrication of shear-aligned membrane from nematic GO. (a) Viscoelastic property of GO (B40 mg ml À 1). Scale bar, 1 cm. (b) Zero-shear viscosity of the dispersions increases with increasing GO concentration. Dashed line is a polynomial fit. (c) Rheology data for three different concentration
showing shear-thinning behavior. Solid curves are the fit of the experimental data with a power law model. (d) Schematic of shear-alignment processing
of a nematic GO to a film; L is the width of blade, h0 is the height of the channel, H is the height of the fluid in front of the blade and U is the processing speed. (e) Polarized light images of fully nematic GO at 40 mg ml À 1 (scale bar, 1 mm). (f) The red circle in the photograph identfies dewetting spots in the
SAMs, which is eliminated when processed from liquid crystalline GO (scale bars, 1 cm). (g) An SEM image demonstrates continuity and conformity of SAM over a porous Nylon substrate (scale bar, 1 mm). (h) Photograph of the gravure printing machine and (inset) images of 13 Â 14 cm2 GO membranes with different thicknesses. (i,j) AFM height map and corresponding height profiles of our membrane (scale bar, 1 mm).
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
3
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
criteria need to be satisfied to ensure accurate measurement of the surface tension and the contact angle29,30: the droplet has to be symmetric along the central vertical axis, the droplet should be shaped only by gravity and surface tension forces and no other forces such as viscosity should play a role in the motion or inertia of the droplet. Droplets formed by 40 and 60 mg ml À 1 GO dispersions did not satisfy these criteria due to high viscosity, so the surface tension and contact angle values were estimated by linear extrapolation (Supplementary Tables 1 and 2) for these two cases. The argument for using linear extrapolation for these dispersions is that the surface tension of GO dispersions decreases with increase in GO concentration given that GO possesses surfactant-like properties31. The contact angle between the GO dispersion and the Nylon substrate decreases with decreasing surface tension, which is consistent with inverse correlation between contact angle and surface tension in the Young’s equations26:
cos y / 1 ð1Þ gLA
where, y is contact angle between the GO dispersion and a Nylon substrate, and gLA is the interfacial surface tension of the GO dispersions.
Membrane fabrication. The primary goal of our work is to form large-area GO membranes by taking advantage of the discotic nematic phase of GO by a shear-induced, industrially adaptable liquid thin film process (Fig. 2d) referred to in this article as shear-aligned membrane (SAM). The discotic nematic colloidal phase (Fig. 2e) has a crucial role in enabling membrane formation, which goes beyond the requirements of high solid contents necessary to produce a continuous film. The GO colloidal dispersion used in our studies undergoes an isotropic to nematic phase transition at B5 mg ml À 1, remaining bi-phasic until B15 mg ml À 1, and fully nematic phases are formed at higher concentrations beyond 16 mg ml À 1 (ref. 16). Typical physical properties of the GO colloidal suspensions representing isotropic, bi-phasic and fully nematic phases are shown in Table 1.
The nematically ordered fluid phases of GO have non-Newtonian flow characteristics (Fig. 2b,c), which can be harnessed to produce large-area films by using shear forces as in doctor blading and dip coating32–34. Our choice of a rigid blade (known as doctor blade in industrial terminology) as the shear alignment method is dictated by factors including its use in large-scale, continuous, high-speed, liquid thin film processes as the metering, applicator gadget and wide-scale use in preparing polymeric films35. The size of a membrane that one can produce is limited only by the size of the shearing apparatus, thus large-area membranes can be produced with relative ease. We also hypothesize that the high shear stress will orientate the graphene sheets of nematic discotic phase17,36, packing them into a dense, continuous, uniform membrane over a porous support in
a rapid single step (Fig. 2d). To investigate this, we initially used a
lab-scale doctor blade that spreads the fluid under Couette
flow through a thin rectangular channel (Supplementary Fig. 2
and Note 2). The viscosity of the fluid is the dominant
material
parameter
in
the
imposed
shear
stress:
%
Z
U h
,
where
t
is
shear
stress,
Z
is
viscosity
of
the
GO
dispersion,
U
0
is
process
speed and h0 is the doctor blade gap size (Fig. 2d). GO fluids from 0.1 to 60 mg ml À 1 were studied with systematic variation of
viscosity. Photographs (Figs 2f and 3a) and scanning electron
microscopic (SEM) images (Figs 2g and 3b) revealed that
the uniformity of the cast film increased with increasing GO concentrations. The films made by 40 and 60 mg ml À 1 GO
suspensions have best uniformity and continuity. To demonstrate
the proficiency of our approach in membrane production, we made large-area GO membranes (40 mg ml À 1) by a gravure
printer (Supplementary Movie 1), with thicknesses ranging
from B65 to B360 nm, on porous Nylon substrates.
Although several methods using solution chemistry or energetic
radiation can be used to chemically reduce the membranes, as a
proof-of-concept to stabilize the membranes in aqueous
environment, we partially reduced the GO membranes by exposure (B5 min) to hydrazine vapour37,38.
Nanofiltration performance. Performances of the membranes as a function of membrane thickness were first evaluated by measuring the water permeability (using Reverse Osmosis water, known as RO water) and the retention for Methyl Red (an electroneutral probe molecule at pHB5.5 (ref. 39); Fig. 4a,b). Membranes with B150 nm thickness were found to exhibit the most promising trade-off between flux and retention. Consequently, the SAMs with B150 nm thickness was chosen for further characterizations. We have compared performance of the SAM with those prepared using the vacuum filtration technique2,12–14. Different thicknesses of GO membranes were prepared by changing the volume of the GO solution (10 mg l À 1) in the vacuum filtration process. These GO membranes were further reduced via hydrazine vapour following the same methodology used for SAM. We compared the water permeability and the retention of methyl red, a probe molecule that is electroneutral at the experimental pH (B5.5)39, for the SAM and the vacuum filtration membranes, with varying membrane thickness, measured here by AFM (Fig. 2i,j). While it is not surprising that the retention is enhanced, the water permeability is also improved as a result of the stacking order in the SAM. Water flux versus pressure measurements for three different varieties of membrane: SAM, vacuum filtered and a commercial membrane (NF270 membrane, Dow Chemical Company, USA) are shown in Fig. 5a. The SAM had a water permeability of 71±5 l m À 2 hr À 1 bar À 1, which is almost seven times better than vacuum filtration membranes (10±2 l m À 2 hr À 1 bar À 1) and almost nine times better than the NF270 membrane while demonstrating comparable or better retention for the electroneutral probe—methyl red (Fig. 4b).
Table 1 | Physical properties of GO dispersion.
Concentration (mg ml À 1)
0.1 5 10 40
Volume fraction
0.005 0.27 0.55 2.22
Surface tension (mN m À 1)
71.9 68 66 49
Contact angle (°)
81 70 67 49
Zero-shear viscosity (at 10 À 3 s À 1) (Pa s)
0.00128 0.0094
0.8 66
Apparent viscosity (at 104 s À 1) (Pa s)
0.0016 0.002 0.0041 0.0164
Dewetting time (s)
0.012 0.151
15 4313
Drying time (s)
40 40 40 40
Physical properties of a typical isotropic (0.1 mg ml À 1), onset of bi-phasic (5 mg ml À 1), bi-phasic (10 mg ml À 1) and fully nematic (40 mg ml À 1) colloidal dispersions of graphene oxide (GO) demonstrating a decrease in surface tension and contact angle which promotes wetting of the fluid on the porous substrate and an increase in dewetting time compared with drying time.
4
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
a
1
2
3
ARTICLE
4
0.1 mg ml–1 Viscosity: 0.0012 Pa s Surface tension: 71.9 mN m–1
2.5 mg ml–1 Viscosity: 0.0016 Pa s Surface tension: 69.6 mN m–1
5 mg ml–1 Viscosity: 0.0094 Pa s Surface tension: 68 mN m–1
10 mg ml–1 Viscosity: 0.8 Pa s Surface tension: 66 mN m–1
5
6
7
8
b1
15 mg ml–1 Viscosity: 2.9 Pa s Surface tension: 65.1 mN m–1
20 mg ml–1 Viscosity: 5 Pa s Surface tension: 60 mN m–1
40 mg ml–1 Viscosity: 66 Pa s Surface tension: 38 mN m–1
60 mg ml–1 Viscosity: 311 Pa s Surface tension: 25 mN m–1
2
3
4
5
6
7
8
Figure 3 | Effect of GO concentration on uniformity and continuity of film. (a) Photographs (scale bars are 1 cm) and (b) SEM images (scale bars, 50 mm) of top surface of the shear-aligned membranes cast by progressively increasing concentration: (1) 0.1 mg ml À 1, (2) 2.5 mg ml À 1, (3) 5 mg ml À 1, (4) 10 mg ml À 1, (5) 15 mg ml À 1, (6) 20 mg ml À 1, (7) 40 mg ml À 1, (8) 60 mg ml À 1.
Water permeability (l m–2 h–1 bar–1) Methyl red retention (%)
a 160
140 120 100
80 60 40 20
0 0
Shear-aligned membrane Vacuum filtration membrane
100
200
300
400
Membrane thickness (nm)
b
100
80
60
40
20
0 0
– OO
CH
NN
+ CH3
N CH3
Shear-aligned membrane Vacuum filtration membrane
100
200
300
400
Membrane thickness (nm)
Figure 4 | Comparision of SAM and vacuum filtration membrane performance. (a) Water permeability versus thickness, and (b) Retention of methyl red, an electroneutral probe molecule. Inset of b is the structure of the methyl red. Error bars in these figures are from five measurements showing the maximum and minimum values.
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
5
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
Flux (l m–2 h–1)
a 300
250 200 150 100
50
Shear-aligned membrane Vacuum filtration membrane NF270
Retention (%)
b 105
100 95 90 85 80 75
RosB OG
MO
Ru
MR
RB
MnB
MB BB
MV
Positively charge probe molecule Negatively charge probe molecule Neutral probe molecule
0
0
1
2
3
4
5
Pressure (bar)
c
120
14
Feed Retentate Permeate Retention Adsorption
12 100
Concentration (mg l–1)
10
80
8
60
6 40
4
2
20
0
0
MV MR MO MnB OG RosB Ru RB MB BB
Probe molecule
Retention and adsorption (%) Retention (%)
70 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 Hydrated radius (Å)
d 45
40 35 30 25 20 15 10
5 0
MgCl2 Na2SO4 NaCl MgSO4 Salt solution
Figure 5 | Filtration performance. (a) Water flux versus applied pressure for three different membranes: SAM (red) with a thickness of 150±15 nm, vacuum filtration (blue) with a thickness of 170±20 nm, and NF270, a commercial nanofiltration membrane (green). SAM showed a retention of 90±2% for methyl red, while the vacuum filtration membrane and NF270 showed 50±5% and 90±1.5% retention, respectively. (b) Retention performance of the 150±15 nm thick shear-aligned membrane, as a function of hydrated radius, for probe molecules with different charges and sizes. (MV is methyl viologen,
MR is methyl red, MnB is methylene blue, MO is methyl orange, OG is orange G, Ru is Tris (bipyridine) runthenium (II) chloride, RB is Rhodamine B, RosB is
Rose Bengal, MB is methylene blue, BB is brilliant blue. The green, red and blue symbols represent electroneutral, negatively and positively charged probe molecules, respectively. (c) Retention details of the membrane for the probe molecules. (d) Salt retention by the 150±15 nm thick SAM, for four different
salt solutions. Error bars in these figures are from five measurements showing the maximum and minimum values.
The flux through the membrane increases linearly with increasing applied pressure (Fig. 5a). The modified HagenPoiseuille equation for slit-shaped pores13 (Flux % 12hL42DZPDx) gives an approximate explanation of fluid flow through these multilayered structures. Using this equation one can estimate the mass flow rate of a Newtonian fluid through porous materials per unit area (m3 s À 1 m À 2), where h (m) B0.95 Â 10 À 9 is the distance between neighbouring graphene sheets (estimated from X-ray diffraction, Supplementary Fig. 3), DP ¼ 0.5 Â 105 Pa is the pressure gradient, L ¼ 0.9 Â 10 À 6 m is the average lateral length of the graphene sheets, Z ¼ 0.001 Pa s is the viscosity of water at 20 °C, and Dx ¼ 150 Â 10 À 9 m is the thickness of the membrane. Comparison of the experimental results with estimated fluxes from the modified Hagen-Poiseuille equation reveal that the theoretical fluxes are four orders of magnitude smaller than the experimental results. This experimental enhancement is consistent with reports of water transport in nanotubes40 and the slit-pores of graphene2,13.
We evaluated the retention of this membrane for different probe molecules with different charges and hydrated radii;
Methyl Viologen (positive charge, at pH 6), Methyl Orange
(negative charge at pH 6), Methylene Blue (positive charge at pH
6.5), Orange G (negative charge at pH 6), Rhodamine B
(electroneutral at pH 6 (electroneutral at pH 6), Tris(bipyridine)
ruthenium(II) chloride (Ruthenium II) (positive charge at pH 6),
Methyl Blue (negative charge at pH 6), Brilliant Blue (negative
charge at pH 6.5) and Rose Bengal (negative charge at pH 6) (Fig. 5b). Before every experiment, the membranes are cleaned with ethanol, acetone and RO water followed by permeation of RO water until a stable permeability is observed. It is noteworthy that the cleaning process removed most of the probe molecules adhering to the membrane surface and almost 100% recovery of flux (Fig. 6a,b) is observed. Retention mechanism in membranes are reliant on size sieving, electrostatic repulsion and adsorption41,42, usually acting in tandem to affect separations. Sorption may dominate separations based on graphene-based materials43, so it is necessary to identify which of these mechanisms are crucial in our membrane. To calculate observation retention, R, (equation (2)) and the percentage of adsorption (equation (3)), we measured the concentration of each probe molecule in the feed (Cf), the permeate (Cp) and the retentate (Cr)44 evaluated by measuring the absorbance of the relevant peaks using a ultraviolet–visible spectrometer.
Rð% Þ ¼ Cf À Cp Â100ð% Þ
ð2Þ
Cf
Ads:ð% Þ ¼ Vf Cf À ðVrCr þ VpCpÞ Â100ð% Þ
ð3Þ
Vf Cf
The membrane showed high retention (490%) for the charged and uncharged solutes with a hydrated radius above 5 Å (Fig. 5b). Fig. 5c reports the analysis of feed, retentate and permeate concentration along with percentage retention and percentage
6
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
ARTICLE
a b 80 Pure water permeability
75
Water permeability in the presence of probe molecules
Permeability (l m–2 h–1 bar–1)
70
65
60
55
50
(1)
45
40 MV MR MO MnB OG RosB Ru RB MB BB
c 80
Probe molecules
Permeability (l m–2 h–1 bar–1)
75
70
65
(2)
60
55
50
45
40
35
30
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000
(3)
Time (min)
Figure 6 | Flux regeneration in SAM. (a) Permeability declined during the filtration tests with probe molecules. The results show that a maximum of 10% decline is observed, it is larger for small molecules (methyl viologen B10% reduction), and less for bigger probe molecules (B4.2% for methyl blue and brilliant blue) consistent with minimal sorption effects. Error bars are from five measurements showing the maximum and minimum values. (b) Photographs of the membrane before (1) and after filtration (2) of methyl red, and after the cleaning process (3) which shows regeneration of the parent-membrane surface (scale bar, 1 cm). (c) Demonstration of long-term viability, low-fouling behaviour of the membranes during filtration of BSA and flux recovery after chemical cleaning in five cycles. Each cycle commences with RO water permeation (Blue symbols), followed by permeation of BSA (Green symbols). Error bars in these figures are from five measurements showing the maximum and minimum values.
adsorption in all the experiments. The results show that the retentate concentration is always larger than the feed concentration, while the adsorption percentage is o10%, irrespective of the probe molecule species in consideration. The permeability during filtration of the probe molecule was usually 90–95% of their clean water permeability (Fig. 6a) further supporting minimum sorption. Based on these measurements, one can argue that SAMs primarily sieve molecules when the average interlayer space of the graphene sheets approaches the physical size of the probe molecules, reported here as hydrated radius. It is also worth noting that the negatively charged probe molecules have higher retention than the positively charged molecules suggesting that electrostatic effects are also important.
Salt retention. We evaluated the retention of monovalent and divalent salts (Na2SO4, MgSO4, MgCl2 and NaCl) at a concentration of 2 g l À 1. The membrane showed retention between 30–40% for all the salts (Fig. 5d). The salt retention capability of the membrane is not surprising as the interlayer spacing is small (B9.5 Å) and the membrane is abundant with various negatively charged oxygen functional groups, such as carboxyl, hydroxyl and epoxy, which persist even after the mild reduction used in stabilizing the membrane (Supplementary Fig. 4). These negatively charged groups particularly carboxylic acids, based on Donnan exclusion theory, will repel co-ions, and consequently retain counter ions to keep electroneutrality of the solution on each side of the membrane.
Long-term filtration tests. A key attribute of our membrane is the stability in aqueous environments and that the retention is
affected by sieving on the top surface of the membrane (Fig. 6b). This allows the membrane to be cleaned in polar and non-polar solvents for multiple reuse. Long-term filtration tests (over 24 h at 0.5 bar pressure) were carried out with BSA, a common laboratory model foulant in membrane-fouling studies45–47. The SAM showed fouling resistance and flux was recovered by a simple solvent cleaning (Fig. 6c). The fouling behaviour of membrane strongly depends on physical and chemical characteristics of the membrane surface such as pore size, porosity, pore morphology and most importantly the hydrophobicity48,49. Fortunately, our membrane retain hydrophilic groups (Supplementary Fig. 4) that decreases hydrophobic interaction with the organic probes and proteins47,48. As a result, a simple cleaning procedure using ethanol, acetone and RO water effectively recovered more than 90% of the flux after every cleaning cycle—this was true for the probe molecules and also for stronger foulant such as BSA (Fig. 6a–c).
Homogeneity of large-area membrane. To demonstrate the homogeneity of the large-area membrane, four pieces were incised from a single large-area membrane and their performance was evaluated. (Supplementary Fig. 5). RO water permeability and retention of each individual membrane is shown in Supplementary Table 3. It is seen that each of these membranes have almost similar performance in water permeability (mean—76.25 l m À 2 hr À 1 bar À 1 and s.d. of 5.7) and Methyl Red rejection (mean—91% and s.d. of 3) using methodology reported in the manuscript. These numbers are excellent evidence for the homogeneity of the large-area membrane.
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
7
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
Discussion
The underlying principles of fluid-physics necessary for fabricating uniform and continuous graphene-based membranes by shear-induced alignment of liquid crystals of GO and the role of stacking order in enhancing water permeability is now discussed. The uniformity and continuity of the membrane arises from a competition between two factors: the casting of a uniform liquid film and then maintaining the stability of the liquid film during drying. The GO dispersions are shear-thinning, pseudoplastic fluids27,50 especially in high volume fractions and are highly viscous in zero-shear and very thin at high shear rate (Fig. 2b,c)—this is instrumental in obtaining a uniform membrane by shear alignment. For example, a GO dispersion at 40 mg ml À 1 would have a zero-shear viscosity of 66 Pa s, but at a shear rate of 104 s À 1, relevant to our process, it will decrease to 0.0164 Pa s meaning that the nematic phase becomes fluid when forced under the micron scale outlet of the blade; membrane formation is also accentuated by the smaller surface tension and smaller contact angle of the nematic fluid (Table 1) to wet the underlying porous membrane. To obtain a uniform membrane it is critical to ensure that the processed liquid film from the GO dispersions remains uniform and continuous until it dries. If for any reason, the liquid film moves or migrates on the substrate, dewetting may ensue and the uniformity and continuity of the film degrades26,51,52 (Figs 2f and 3). Then to maintain stability of the liquid film during drying, the film needs to resist dewetting. In general, dewetting occurs on nonwettable substrates and can also be initiated by various film-thinning mechanisms, which persist until holes are produced and the film is ruptured. A large number of factors influence dewetting, such as solvent evaporation (especially in the case of low concentration dispersions), electrostatic repulsion (or attraction) forces between the dispersion and the substrate, dispersion migration due to gravity or capillary-driven flow, film thickness and viscosity and
surface tension gradients. Among all film-thinning mechanisms,
the predominant factor for dewetting is low viscosity and high
surface tension of the dispersion51,53. The dewetting time can be
estimated by tdewet ¼
mL ksy3
34,54,
where
tdewet
is
the
dewetting
time (s), s(N m À 1) and m (Pa s) are the surface tension and the
viscosity of the dispersion, respectively, and y (rad) is the contact
angle between the dispersion and the substrate. k is a constant related to the fluid property and is assumed to be 10 À 3 for water-based system34,54. L is the length scale, which is estimated as 10% of the substrate width34. Drying time is the time between
casting the liquid film and its solidification, which is defined by
tdry ¼ DJh, where Dh is a parameter estimated to 80% of the
0
thickness
of
the
liquid
film34
and
J0
is
the
solvent
evaporation
current (cm s À 1). To avoid rupture and obtain a continuous and
uniform film, the drying time must be lower than the dewetting
time. J0 was calculated by recording the mass loss of the liquid
film on drying. The volume of the liquid film was calculated by considering the density of GO (B1.8 g ml À 1)55, using the mass
and concentration of the GO dispersion. Dividing the volume
with the area of the liquid film, we obtained the thickness of the
liquid film during drying and subsequently calculated J0 (2 Â 10 À 6 cm s À 1). With increasing GO concentration, because
of the enhanced zero-shear viscosity, lower surface tension and
increase in the contact angle, the dewetting time-scale of the
nematic fluids can be easily increased by over six orders of
magnitude (Table 1). In fact, the dewetting time increased from
0.012 to 4,313 s with an increase in GO concentration from 0.1 to 40 mg ml À 1 (Table 1). Since the drying time (40 s) in case of 40 mg ml À 1 GO dispersion was significantly lower than the
dewetting time (4313 s), uniform films could be produced under
these conditions. Optical images and SEM in Fig. 3 confirm that
dewetting in the films formed from such high concentrations is
prohibited.
a
b
c
S~0.99
90
120
60
150
30
Slow axis 9o0r°ientation
135°
45°
180°
0°
d
e
180
0
f
S~0.30
90
120
60
150
30
Slow axis 9o0r°ientation
135°
45°
180
0
180°
0°
Figure 7 | Effect of stacking of GO sheets on performance of filtration membranes. Polarized light imaging results of (a–c) SAM and (d–f) vacuum filtration membrane. (a,d) Falsely coloured polarized light images, where the hue represents the azimuth as depicted by the legend (scale bars are 50 mm). Regions with the same hue represent the same azimuth angle, so the SAMs have higher in-plane stacking order while vacuum filtration membranes have lower stacking order. This is supported by their slow axis vector representations (b,e), which are expanded view of the boxed areas (scale bar, 10 mm), polar histograms of the azimuth angles and the in-plane order parameters (c,f). (c,f) Predicted organization of graphene sheets in membranes.
8
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
UKnowledge
Chemical and Materials Engineering Faculty Publications
Chemical and Materials Engineering
3-7-2016
Large-Area Graphene-Based Nanofiltration Membranes by Shear Alignment of Discotic Nematic Liquid Crystals of Graphene Oxide
Abozar Akbari Monash University, Australia
Phillip Sheath Monash University, Australia
Samuel T. Martin Monash University, Australia
Dhanraj B. Shinde Monash University, Australia
Mahdokht Shaibani Monash University, Australia
See next page for additional authors Follow this and additional works at: https://uknowledge.uky.edu/cme_facpub
Part of the Chemical Engineering Commons, and the Materials Science and Engineering Commons Right click to open a feedback form in a new tab to let us know how this document benefits you.
Repository Citation Akbari, Abozar; Sheath, Phillip; Martin, Samuel T.; Shinde, Dhanraj B.; Shaibani, Mahdokht; Banerjee, Parama Chakraborty; Tkacz, Rachel; Bhattacharyya, Dibakar; and Majumder, Mainak, "Large-Area Graphene-Based Nanofiltration Membranes by Shear Alignment of Discotic Nematic Liquid Crystals of Graphene Oxide" (2016). Chemical and Materials Engineering Faculty Publications. 19. https://uknowledge.uky.edu/cme_facpub/19
This Article is brought to you for free and open access by the Chemical and Materials Engineering at UKnowledge. It has been accepted for inclusion in Chemical and Materials Engineering Faculty Publications by an authorized administrator of UKnowledge. For more information, please contact [email protected]
Large-Area Graphene-Based Nanofiltration Membranes by Shear Alignment of Discotic Nematic Liquid Crystals of Graphene Oxide Digital Object Identifier (DOI)
https://doi.org/10.1038/ncomms10891 Notes/Citation Information Published in Nature Communications, v. 7, article no. 10891, p. 1-12. © The Author(s) 2017 This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ Authors Abozar Akbari, Phillip Sheath, Samuel T. Martin, Dhanraj B. Shinde, Mahdokht Shaibani, Parama Chakraborty Banerjee, Rachel Tkacz, Dibakar Bhattacharyya, and Mainak Majumder
This article is available at UKnowledge: https://uknowledge.uky.edu/cme_facpub/19
ARTICLE
Received 1 Sep 2015 | Accepted 29 Jan 2016 | Published 7 Mar 2016
DOI: 10.1038/ncomms10891
OPEN
Large-area graphene-based nanofiltration membranes by shear alignment of discotic nematic liquid crystals of graphene oxide
Abozar Akbari1, Phillip Sheath1, Samuel T. Martin1, Dhanraj B. Shinde1, Mahdokht Shaibani1, Parama Chakraborty Banerjee1, Rachel Tkacz1, Dibakar Bhattacharyya2 & Mainak Majumder1
Graphene-based membranes demonstrating ultrafast water transport, precise molecular sieving of gas and solvated molecules shows great promise as novel separation platforms; however, scale-up of these membranes to large-areas remains an unresolved problem. Here we demonstrate that the discotic nematic phase of graphene oxide (GO) can be shear aligned to form highly ordered, continuous, thin films of multi-layered GO on a support membrane by an industrially adaptable method to produce large-area membranes (13 Â 14 cm2) in o5 s. Pressure driven transport data demonstrate high retention (490%) for charged and uncharged organic probe molecules with a hydrated radius above 5 Å as well as modest (30–40%) retention of monovalent and divalent salts. The highly ordered graphene sheets in the plane of the membrane make organized channels and enhance the permeability (71±5 l m À 2 hr À 1 bar À 1 for 150±15 nm thick membranes).
1 Department of Mechanical and Aerospace Engineering, Nanoscale Science and Engineering Laboratory (NSEL), Monash University, Clayton, Victoria 3800, Australia. 2 Department of Chemical and Materials Engineering, University of Kentucky, Lexington, Kentucky 40506, USA. Correspondence and requests for
materials should be addressed to M.M. (email: [email protected]).
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
1
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
Advances in the design and synthesis of nanofiltration membranes with improved retention, flux and cost-effectiveness will have tremendous impact in several fields such as water treatment, selective chemical separations and drug delivery. Conventional polymeric nanofiltration membranes usually have limited chemical resistance, while ceramic membranes are not cost-efficient. Graphene is a one atom thick two-dimensional honeycomb sp2 carbon lattice, which is an exciting multifunctional material and possesses a combination of strong mechanical properties, chemical inertness and extremely large surface area1,2. Membranes prepared from graphene possess the best of both the worlds: they are chemically inert3 like ceramic membranes and can be made into films using graphene/graphene oxide (GO) fluid phase dispersions like polymers. Novel and exciting transport properties of graphene-based membranes such as high permeability and high selectivity for both liquids2,4–7 and gases8–11 have recently been reported. While these studies have unlocked potential applications, there is critical need to produce these membranes in large-areas using high throughput manufacturing routes, which may otherwise hinder their impact in membrane technologies. The ideal structure of a filtration membrane has a defect-free, thin, dense separation film that acts as a functional sieve, while the mechanical strength is provided by a porous and more permeable support. To achieve this asymmetric structure, researchers have grown continuous graphene films by chemical vapour deposition and transferred them to substrates followed by etching pores on the film, however, the transfer process limits the scalability of membrane production6,7. Another method to produce this structure is by restacking GO flakes by filtration of GO dispersions on a backing filter support2,12–14. However, producing a membrane by this approach requires large volumes of liquid, significant time and arguably has both alignment (of the GO sheets) and scalability issues. Other liquid phase processes such as dip-coating or layer-by-layer assembly similarly have potential issues with rapid productivity15. Therefore, a major challenge in this field is to define robust, scalable, liquid film processing approaches to produce large-area graphene-based membranes that will bridge laboratory curiosity to industrial productivity.
Here we introduce a scalable and industrially adaptable method to fabricate large-area graphene-based membranes by shear-induced alignment of liquid crystals of GO. The membranes have large in-plane stacking order of GO sheets and demonstrate outstanding water permeability while being able to sieve small organic molecules with performance metrics superior to well established and commercially available nanofiltration membrane.
Results Producing liquid crystalline phases of GO. Oxidation and exfoliation of graphite by the well-known Hummers’ method or its variations produces graphene nanosheets decorated with oxygenated functional groups also known as GO. The anisotropic GO nanosheets can be dispersed in liquids including water as stable colloidal suspensions with large volume fractions. As the concentration of the anisotropic particles increases, the orientation entropy of the suspensions starts to decrease only to be compensated by increase in the translation entropy leading to colloidal phase transitions from isotropic to nematic liquid crystalline phases—the onset of which has dependence on the thickness to diameter ratio of the disc-like mesogens of GO16. Liquid crystallinity defines a state between a crystal and a fluid, within which the constituent sheets become anisotropic but can still flow and respond to macroscopic force-fields such as shear17, and this state has been demonstrated in concentrated dispersions of GO1,18–20. Traditional means to produce concentrated GO dispersions, such as the application of heat21 or the use of vacuum equipment22, are time consuming and laborious. An innovative method was implemented in this work to quickly produce nematic GO dispersions (Fig. 1). We used superabsorbent polymer hydrogel beads (typically, cross-linked polyacrylate based copolymer), which are strongly hydrophilic. Concentration of a GO dispersion occurs because the hydrogel beads absorb and retain water23 without dissolving in water or absorbing GO sheets. This is demonstrated in Raman characterization in Supplementary Fig. 1 (Supplementary Note 1)—the characteristic peaks of GO were not observed
Add beads Remove
swollen beads
0.25 mg ml–1
Repeat until desired concentration
20 mg ml–1
Figure 1 | Procedure for concentrating GO. Photographs of stable dispersions of GO which have been concentrated by adding hydrogel beads.
2
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
ARTICLE
within the hydrogel beads swollen in a GO suspension. The time taken to concentrate a GO dispersion depends on the initial concentration, the desired concentration and the mass of beads used. Fig. 2a shows a GO dispersion with a concentration of 40 mg ml À 1.
Rheological characterization of GO dispersions. Rheological properties of the GO dispersions are crucial to our fabrication method. We evaluated zero-shear viscosity by measuring the viscosity of the GO dispersions at a shear rate of 0.001 s À 1 (ref. 24). Fig. 2b demonstrates that the zero-shear viscosity increases with an increase in GO concentration. At low concentrations of GO, water molecules are attached to GO sheets via hydrogen bonds25, but similar to water-clay dispersions26, an increase in the number of GO sheets results in the assembly of graphene sheets and water molecules to form a three-dimensional network via hydrogen bonding, which decreases the fluidity of the dispersion. Furthermore, the large changes in the viscosity beyond 5 mg ml À 1 coincides with the onset of liquid crystalline nematic phase16. Fig. 2c presents apparent viscosity of the GO dispersion (Z) as a function of shear rate (g˚). The non-Newtonian
shear-thinning (pseudoplastic) behaviour was observed at different concentrations of the GO dispersion. Decreased viscosity of the GO dispersions with an increase in the shear rates is consistent with previous reports18,27. One can presume that the nematic phases in the GO dispersion are distributed randomly and do not align at low shear rates, which results in higher viscosity. At high shear rates, the randomly distributed nematic phases align in the direction of shear stress and produces less physical interaction with each other, resulting in decreased viscosity. Fig. 2c shows that the viscosity of the GO dispersions were in good agreement with the power law viscosity model. In the power law model, the exponent for ideal plastic material is À 1 and any deviation from this theoretical value shows a loss of plastic behaviour28. The exponents decrease from À 0.580 to À 0.867 by increasing GO concentration from 10 to 40 mg ml À 1, which affirms the increased plasticity arising from the nematic GO phases.
Interfacial properties of GO dispersion. Properties of the GO dispersion, typically at the solid–liquid interface, also assume importance in our membrane fabrication process. Several
Zero-shear viscosity (Pa s) (Pa s)
a
d
(y) H h0
Doctor blade
b
300
250 200 150 100
5 4 3 2 1 0
0 5 10 15 20
50
0 0
Nematic phase GO
10 20 30 40 50 60 GO concentration (mg ml–1)
e
f 10 mg ml–1
Process speed (U) L
40 mg ml–1
c 100
10 1
0.1 0.01 0.001
0
g
10 mg ml–1 20 mg ml–1 40 mg ml–1 ∝ γ° –0.868 ∝ γ°–0.666
∝ γ°–0.580
20
40
60
80 100
° (s–1)
Fixed substrate
(x)
h
i
832 nm
j 180
150
Height (nm)
150±15 nm
120
90
Profile 1
60
Profile 2
30
Profile 3
0 nm
0 0 200 400 600 800 1,000
Distance (nm)
Figure 2 | Fabrication of shear-aligned membrane from nematic GO. (a) Viscoelastic property of GO (B40 mg ml À 1). Scale bar, 1 cm. (b) Zero-shear viscosity of the dispersions increases with increasing GO concentration. Dashed line is a polynomial fit. (c) Rheology data for three different concentration
showing shear-thinning behavior. Solid curves are the fit of the experimental data with a power law model. (d) Schematic of shear-alignment processing
of a nematic GO to a film; L is the width of blade, h0 is the height of the channel, H is the height of the fluid in front of the blade and U is the processing speed. (e) Polarized light images of fully nematic GO at 40 mg ml À 1 (scale bar, 1 mm). (f) The red circle in the photograph identfies dewetting spots in the
SAMs, which is eliminated when processed from liquid crystalline GO (scale bars, 1 cm). (g) An SEM image demonstrates continuity and conformity of SAM over a porous Nylon substrate (scale bar, 1 mm). (h) Photograph of the gravure printing machine and (inset) images of 13 Â 14 cm2 GO membranes with different thicknesses. (i,j) AFM height map and corresponding height profiles of our membrane (scale bar, 1 mm).
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
3
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
criteria need to be satisfied to ensure accurate measurement of the surface tension and the contact angle29,30: the droplet has to be symmetric along the central vertical axis, the droplet should be shaped only by gravity and surface tension forces and no other forces such as viscosity should play a role in the motion or inertia of the droplet. Droplets formed by 40 and 60 mg ml À 1 GO dispersions did not satisfy these criteria due to high viscosity, so the surface tension and contact angle values were estimated by linear extrapolation (Supplementary Tables 1 and 2) for these two cases. The argument for using linear extrapolation for these dispersions is that the surface tension of GO dispersions decreases with increase in GO concentration given that GO possesses surfactant-like properties31. The contact angle between the GO dispersion and the Nylon substrate decreases with decreasing surface tension, which is consistent with inverse correlation between contact angle and surface tension in the Young’s equations26:
cos y / 1 ð1Þ gLA
where, y is contact angle between the GO dispersion and a Nylon substrate, and gLA is the interfacial surface tension of the GO dispersions.
Membrane fabrication. The primary goal of our work is to form large-area GO membranes by taking advantage of the discotic nematic phase of GO by a shear-induced, industrially adaptable liquid thin film process (Fig. 2d) referred to in this article as shear-aligned membrane (SAM). The discotic nematic colloidal phase (Fig. 2e) has a crucial role in enabling membrane formation, which goes beyond the requirements of high solid contents necessary to produce a continuous film. The GO colloidal dispersion used in our studies undergoes an isotropic to nematic phase transition at B5 mg ml À 1, remaining bi-phasic until B15 mg ml À 1, and fully nematic phases are formed at higher concentrations beyond 16 mg ml À 1 (ref. 16). Typical physical properties of the GO colloidal suspensions representing isotropic, bi-phasic and fully nematic phases are shown in Table 1.
The nematically ordered fluid phases of GO have non-Newtonian flow characteristics (Fig. 2b,c), which can be harnessed to produce large-area films by using shear forces as in doctor blading and dip coating32–34. Our choice of a rigid blade (known as doctor blade in industrial terminology) as the shear alignment method is dictated by factors including its use in large-scale, continuous, high-speed, liquid thin film processes as the metering, applicator gadget and wide-scale use in preparing polymeric films35. The size of a membrane that one can produce is limited only by the size of the shearing apparatus, thus large-area membranes can be produced with relative ease. We also hypothesize that the high shear stress will orientate the graphene sheets of nematic discotic phase17,36, packing them into a dense, continuous, uniform membrane over a porous support in
a rapid single step (Fig. 2d). To investigate this, we initially used a
lab-scale doctor blade that spreads the fluid under Couette
flow through a thin rectangular channel (Supplementary Fig. 2
and Note 2). The viscosity of the fluid is the dominant
material
parameter
in
the
imposed
shear
stress:
%
Z
U h
,
where
t
is
shear
stress,
Z
is
viscosity
of
the
GO
dispersion,
U
0
is
process
speed and h0 is the doctor blade gap size (Fig. 2d). GO fluids from 0.1 to 60 mg ml À 1 were studied with systematic variation of
viscosity. Photographs (Figs 2f and 3a) and scanning electron
microscopic (SEM) images (Figs 2g and 3b) revealed that
the uniformity of the cast film increased with increasing GO concentrations. The films made by 40 and 60 mg ml À 1 GO
suspensions have best uniformity and continuity. To demonstrate
the proficiency of our approach in membrane production, we made large-area GO membranes (40 mg ml À 1) by a gravure
printer (Supplementary Movie 1), with thicknesses ranging
from B65 to B360 nm, on porous Nylon substrates.
Although several methods using solution chemistry or energetic
radiation can be used to chemically reduce the membranes, as a
proof-of-concept to stabilize the membranes in aqueous
environment, we partially reduced the GO membranes by exposure (B5 min) to hydrazine vapour37,38.
Nanofiltration performance. Performances of the membranes as a function of membrane thickness were first evaluated by measuring the water permeability (using Reverse Osmosis water, known as RO water) and the retention for Methyl Red (an electroneutral probe molecule at pHB5.5 (ref. 39); Fig. 4a,b). Membranes with B150 nm thickness were found to exhibit the most promising trade-off between flux and retention. Consequently, the SAMs with B150 nm thickness was chosen for further characterizations. We have compared performance of the SAM with those prepared using the vacuum filtration technique2,12–14. Different thicknesses of GO membranes were prepared by changing the volume of the GO solution (10 mg l À 1) in the vacuum filtration process. These GO membranes were further reduced via hydrazine vapour following the same methodology used for SAM. We compared the water permeability and the retention of methyl red, a probe molecule that is electroneutral at the experimental pH (B5.5)39, for the SAM and the vacuum filtration membranes, with varying membrane thickness, measured here by AFM (Fig. 2i,j). While it is not surprising that the retention is enhanced, the water permeability is also improved as a result of the stacking order in the SAM. Water flux versus pressure measurements for three different varieties of membrane: SAM, vacuum filtered and a commercial membrane (NF270 membrane, Dow Chemical Company, USA) are shown in Fig. 5a. The SAM had a water permeability of 71±5 l m À 2 hr À 1 bar À 1, which is almost seven times better than vacuum filtration membranes (10±2 l m À 2 hr À 1 bar À 1) and almost nine times better than the NF270 membrane while demonstrating comparable or better retention for the electroneutral probe—methyl red (Fig. 4b).
Table 1 | Physical properties of GO dispersion.
Concentration (mg ml À 1)
0.1 5 10 40
Volume fraction
0.005 0.27 0.55 2.22
Surface tension (mN m À 1)
71.9 68 66 49
Contact angle (°)
81 70 67 49
Zero-shear viscosity (at 10 À 3 s À 1) (Pa s)
0.00128 0.0094
0.8 66
Apparent viscosity (at 104 s À 1) (Pa s)
0.0016 0.002 0.0041 0.0164
Dewetting time (s)
0.012 0.151
15 4313
Drying time (s)
40 40 40 40
Physical properties of a typical isotropic (0.1 mg ml À 1), onset of bi-phasic (5 mg ml À 1), bi-phasic (10 mg ml À 1) and fully nematic (40 mg ml À 1) colloidal dispersions of graphene oxide (GO) demonstrating a decrease in surface tension and contact angle which promotes wetting of the fluid on the porous substrate and an increase in dewetting time compared with drying time.
4
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
a
1
2
3
ARTICLE
4
0.1 mg ml–1 Viscosity: 0.0012 Pa s Surface tension: 71.9 mN m–1
2.5 mg ml–1 Viscosity: 0.0016 Pa s Surface tension: 69.6 mN m–1
5 mg ml–1 Viscosity: 0.0094 Pa s Surface tension: 68 mN m–1
10 mg ml–1 Viscosity: 0.8 Pa s Surface tension: 66 mN m–1
5
6
7
8
b1
15 mg ml–1 Viscosity: 2.9 Pa s Surface tension: 65.1 mN m–1
20 mg ml–1 Viscosity: 5 Pa s Surface tension: 60 mN m–1
40 mg ml–1 Viscosity: 66 Pa s Surface tension: 38 mN m–1
60 mg ml–1 Viscosity: 311 Pa s Surface tension: 25 mN m–1
2
3
4
5
6
7
8
Figure 3 | Effect of GO concentration on uniformity and continuity of film. (a) Photographs (scale bars are 1 cm) and (b) SEM images (scale bars, 50 mm) of top surface of the shear-aligned membranes cast by progressively increasing concentration: (1) 0.1 mg ml À 1, (2) 2.5 mg ml À 1, (3) 5 mg ml À 1, (4) 10 mg ml À 1, (5) 15 mg ml À 1, (6) 20 mg ml À 1, (7) 40 mg ml À 1, (8) 60 mg ml À 1.
Water permeability (l m–2 h–1 bar–1) Methyl red retention (%)
a 160
140 120 100
80 60 40 20
0 0
Shear-aligned membrane Vacuum filtration membrane
100
200
300
400
Membrane thickness (nm)
b
100
80
60
40
20
0 0
– OO
CH
NN
+ CH3
N CH3
Shear-aligned membrane Vacuum filtration membrane
100
200
300
400
Membrane thickness (nm)
Figure 4 | Comparision of SAM and vacuum filtration membrane performance. (a) Water permeability versus thickness, and (b) Retention of methyl red, an electroneutral probe molecule. Inset of b is the structure of the methyl red. Error bars in these figures are from five measurements showing the maximum and minimum values.
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
5
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
Flux (l m–2 h–1)
a 300
250 200 150 100
50
Shear-aligned membrane Vacuum filtration membrane NF270
Retention (%)
b 105
100 95 90 85 80 75
RosB OG
MO
Ru
MR
RB
MnB
MB BB
MV
Positively charge probe molecule Negatively charge probe molecule Neutral probe molecule
0
0
1
2
3
4
5
Pressure (bar)
c
120
14
Feed Retentate Permeate Retention Adsorption
12 100
Concentration (mg l–1)
10
80
8
60
6 40
4
2
20
0
0
MV MR MO MnB OG RosB Ru RB MB BB
Probe molecule
Retention and adsorption (%) Retention (%)
70 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 Hydrated radius (Å)
d 45
40 35 30 25 20 15 10
5 0
MgCl2 Na2SO4 NaCl MgSO4 Salt solution
Figure 5 | Filtration performance. (a) Water flux versus applied pressure for three different membranes: SAM (red) with a thickness of 150±15 nm, vacuum filtration (blue) with a thickness of 170±20 nm, and NF270, a commercial nanofiltration membrane (green). SAM showed a retention of 90±2% for methyl red, while the vacuum filtration membrane and NF270 showed 50±5% and 90±1.5% retention, respectively. (b) Retention performance of the 150±15 nm thick shear-aligned membrane, as a function of hydrated radius, for probe molecules with different charges and sizes. (MV is methyl viologen,
MR is methyl red, MnB is methylene blue, MO is methyl orange, OG is orange G, Ru is Tris (bipyridine) runthenium (II) chloride, RB is Rhodamine B, RosB is
Rose Bengal, MB is methylene blue, BB is brilliant blue. The green, red and blue symbols represent electroneutral, negatively and positively charged probe molecules, respectively. (c) Retention details of the membrane for the probe molecules. (d) Salt retention by the 150±15 nm thick SAM, for four different
salt solutions. Error bars in these figures are from five measurements showing the maximum and minimum values.
The flux through the membrane increases linearly with increasing applied pressure (Fig. 5a). The modified HagenPoiseuille equation for slit-shaped pores13 (Flux % 12hL42DZPDx) gives an approximate explanation of fluid flow through these multilayered structures. Using this equation one can estimate the mass flow rate of a Newtonian fluid through porous materials per unit area (m3 s À 1 m À 2), where h (m) B0.95 Â 10 À 9 is the distance between neighbouring graphene sheets (estimated from X-ray diffraction, Supplementary Fig. 3), DP ¼ 0.5 Â 105 Pa is the pressure gradient, L ¼ 0.9 Â 10 À 6 m is the average lateral length of the graphene sheets, Z ¼ 0.001 Pa s is the viscosity of water at 20 °C, and Dx ¼ 150 Â 10 À 9 m is the thickness of the membrane. Comparison of the experimental results with estimated fluxes from the modified Hagen-Poiseuille equation reveal that the theoretical fluxes are four orders of magnitude smaller than the experimental results. This experimental enhancement is consistent with reports of water transport in nanotubes40 and the slit-pores of graphene2,13.
We evaluated the retention of this membrane for different probe molecules with different charges and hydrated radii;
Methyl Viologen (positive charge, at pH 6), Methyl Orange
(negative charge at pH 6), Methylene Blue (positive charge at pH
6.5), Orange G (negative charge at pH 6), Rhodamine B
(electroneutral at pH 6 (electroneutral at pH 6), Tris(bipyridine)
ruthenium(II) chloride (Ruthenium II) (positive charge at pH 6),
Methyl Blue (negative charge at pH 6), Brilliant Blue (negative
charge at pH 6.5) and Rose Bengal (negative charge at pH 6) (Fig. 5b). Before every experiment, the membranes are cleaned with ethanol, acetone and RO water followed by permeation of RO water until a stable permeability is observed. It is noteworthy that the cleaning process removed most of the probe molecules adhering to the membrane surface and almost 100% recovery of flux (Fig. 6a,b) is observed. Retention mechanism in membranes are reliant on size sieving, electrostatic repulsion and adsorption41,42, usually acting in tandem to affect separations. Sorption may dominate separations based on graphene-based materials43, so it is necessary to identify which of these mechanisms are crucial in our membrane. To calculate observation retention, R, (equation (2)) and the percentage of adsorption (equation (3)), we measured the concentration of each probe molecule in the feed (Cf), the permeate (Cp) and the retentate (Cr)44 evaluated by measuring the absorbance of the relevant peaks using a ultraviolet–visible spectrometer.
Rð% Þ ¼ Cf À Cp Â100ð% Þ
ð2Þ
Cf
Ads:ð% Þ ¼ Vf Cf À ðVrCr þ VpCpÞ Â100ð% Þ
ð3Þ
Vf Cf
The membrane showed high retention (490%) for the charged and uncharged solutes with a hydrated radius above 5 Å (Fig. 5b). Fig. 5c reports the analysis of feed, retentate and permeate concentration along with percentage retention and percentage
6
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
ARTICLE
a b 80 Pure water permeability
75
Water permeability in the presence of probe molecules
Permeability (l m–2 h–1 bar–1)
70
65
60
55
50
(1)
45
40 MV MR MO MnB OG RosB Ru RB MB BB
c 80
Probe molecules
Permeability (l m–2 h–1 bar–1)
75
70
65
(2)
60
55
50
45
40
35
30
0 200 400 600 800 1,000 1,200 1,400 1,600 1,800 2,000
(3)
Time (min)
Figure 6 | Flux regeneration in SAM. (a) Permeability declined during the filtration tests with probe molecules. The results show that a maximum of 10% decline is observed, it is larger for small molecules (methyl viologen B10% reduction), and less for bigger probe molecules (B4.2% for methyl blue and brilliant blue) consistent with minimal sorption effects. Error bars are from five measurements showing the maximum and minimum values. (b) Photographs of the membrane before (1) and after filtration (2) of methyl red, and after the cleaning process (3) which shows regeneration of the parent-membrane surface (scale bar, 1 cm). (c) Demonstration of long-term viability, low-fouling behaviour of the membranes during filtration of BSA and flux recovery after chemical cleaning in five cycles. Each cycle commences with RO water permeation (Blue symbols), followed by permeation of BSA (Green symbols). Error bars in these figures are from five measurements showing the maximum and minimum values.
adsorption in all the experiments. The results show that the retentate concentration is always larger than the feed concentration, while the adsorption percentage is o10%, irrespective of the probe molecule species in consideration. The permeability during filtration of the probe molecule was usually 90–95% of their clean water permeability (Fig. 6a) further supporting minimum sorption. Based on these measurements, one can argue that SAMs primarily sieve molecules when the average interlayer space of the graphene sheets approaches the physical size of the probe molecules, reported here as hydrated radius. It is also worth noting that the negatively charged probe molecules have higher retention than the positively charged molecules suggesting that electrostatic effects are also important.
Salt retention. We evaluated the retention of monovalent and divalent salts (Na2SO4, MgSO4, MgCl2 and NaCl) at a concentration of 2 g l À 1. The membrane showed retention between 30–40% for all the salts (Fig. 5d). The salt retention capability of the membrane is not surprising as the interlayer spacing is small (B9.5 Å) and the membrane is abundant with various negatively charged oxygen functional groups, such as carboxyl, hydroxyl and epoxy, which persist even after the mild reduction used in stabilizing the membrane (Supplementary Fig. 4). These negatively charged groups particularly carboxylic acids, based on Donnan exclusion theory, will repel co-ions, and consequently retain counter ions to keep electroneutrality of the solution on each side of the membrane.
Long-term filtration tests. A key attribute of our membrane is the stability in aqueous environments and that the retention is
affected by sieving on the top surface of the membrane (Fig. 6b). This allows the membrane to be cleaned in polar and non-polar solvents for multiple reuse. Long-term filtration tests (over 24 h at 0.5 bar pressure) were carried out with BSA, a common laboratory model foulant in membrane-fouling studies45–47. The SAM showed fouling resistance and flux was recovered by a simple solvent cleaning (Fig. 6c). The fouling behaviour of membrane strongly depends on physical and chemical characteristics of the membrane surface such as pore size, porosity, pore morphology and most importantly the hydrophobicity48,49. Fortunately, our membrane retain hydrophilic groups (Supplementary Fig. 4) that decreases hydrophobic interaction with the organic probes and proteins47,48. As a result, a simple cleaning procedure using ethanol, acetone and RO water effectively recovered more than 90% of the flux after every cleaning cycle—this was true for the probe molecules and also for stronger foulant such as BSA (Fig. 6a–c).
Homogeneity of large-area membrane. To demonstrate the homogeneity of the large-area membrane, four pieces were incised from a single large-area membrane and their performance was evaluated. (Supplementary Fig. 5). RO water permeability and retention of each individual membrane is shown in Supplementary Table 3. It is seen that each of these membranes have almost similar performance in water permeability (mean—76.25 l m À 2 hr À 1 bar À 1 and s.d. of 5.7) and Methyl Red rejection (mean—91% and s.d. of 3) using methodology reported in the manuscript. These numbers are excellent evidence for the homogeneity of the large-area membrane.
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications
7
ARTICLE
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms10891
Discussion
The underlying principles of fluid-physics necessary for fabricating uniform and continuous graphene-based membranes by shear-induced alignment of liquid crystals of GO and the role of stacking order in enhancing water permeability is now discussed. The uniformity and continuity of the membrane arises from a competition between two factors: the casting of a uniform liquid film and then maintaining the stability of the liquid film during drying. The GO dispersions are shear-thinning, pseudoplastic fluids27,50 especially in high volume fractions and are highly viscous in zero-shear and very thin at high shear rate (Fig. 2b,c)—this is instrumental in obtaining a uniform membrane by shear alignment. For example, a GO dispersion at 40 mg ml À 1 would have a zero-shear viscosity of 66 Pa s, but at a shear rate of 104 s À 1, relevant to our process, it will decrease to 0.0164 Pa s meaning that the nematic phase becomes fluid when forced under the micron scale outlet of the blade; membrane formation is also accentuated by the smaller surface tension and smaller contact angle of the nematic fluid (Table 1) to wet the underlying porous membrane. To obtain a uniform membrane it is critical to ensure that the processed liquid film from the GO dispersions remains uniform and continuous until it dries. If for any reason, the liquid film moves or migrates on the substrate, dewetting may ensue and the uniformity and continuity of the film degrades26,51,52 (Figs 2f and 3). Then to maintain stability of the liquid film during drying, the film needs to resist dewetting. In general, dewetting occurs on nonwettable substrates and can also be initiated by various film-thinning mechanisms, which persist until holes are produced and the film is ruptured. A large number of factors influence dewetting, such as solvent evaporation (especially in the case of low concentration dispersions), electrostatic repulsion (or attraction) forces between the dispersion and the substrate, dispersion migration due to gravity or capillary-driven flow, film thickness and viscosity and
surface tension gradients. Among all film-thinning mechanisms,
the predominant factor for dewetting is low viscosity and high
surface tension of the dispersion51,53. The dewetting time can be
estimated by tdewet ¼
mL ksy3
34,54,
where
tdewet
is
the
dewetting
time (s), s(N m À 1) and m (Pa s) are the surface tension and the
viscosity of the dispersion, respectively, and y (rad) is the contact
angle between the dispersion and the substrate. k is a constant related to the fluid property and is assumed to be 10 À 3 for water-based system34,54. L is the length scale, which is estimated as 10% of the substrate width34. Drying time is the time between
casting the liquid film and its solidification, which is defined by
tdry ¼ DJh, where Dh is a parameter estimated to 80% of the
0
thickness
of
the
liquid
film34
and
J0
is
the
solvent
evaporation
current (cm s À 1). To avoid rupture and obtain a continuous and
uniform film, the drying time must be lower than the dewetting
time. J0 was calculated by recording the mass loss of the liquid
film on drying. The volume of the liquid film was calculated by considering the density of GO (B1.8 g ml À 1)55, using the mass
and concentration of the GO dispersion. Dividing the volume
with the area of the liquid film, we obtained the thickness of the
liquid film during drying and subsequently calculated J0 (2 Â 10 À 6 cm s À 1). With increasing GO concentration, because
of the enhanced zero-shear viscosity, lower surface tension and
increase in the contact angle, the dewetting time-scale of the
nematic fluids can be easily increased by over six orders of
magnitude (Table 1). In fact, the dewetting time increased from
0.012 to 4,313 s with an increase in GO concentration from 0.1 to 40 mg ml À 1 (Table 1). Since the drying time (40 s) in case of 40 mg ml À 1 GO dispersion was significantly lower than the
dewetting time (4313 s), uniform films could be produced under
these conditions. Optical images and SEM in Fig. 3 confirm that
dewetting in the films formed from such high concentrations is
prohibited.
a
b
c
S~0.99
90
120
60
150
30
Slow axis 9o0r°ientation
135°
45°
180°
0°
d
e
180
0
f
S~0.30
90
120
60
150
30
Slow axis 9o0r°ientation
135°
45°
180
0
180°
0°
Figure 7 | Effect of stacking of GO sheets on performance of filtration membranes. Polarized light imaging results of (a–c) SAM and (d–f) vacuum filtration membrane. (a,d) Falsely coloured polarized light images, where the hue represents the azimuth as depicted by the legend (scale bars are 50 mm). Regions with the same hue represent the same azimuth angle, so the SAMs have higher in-plane stacking order while vacuum filtration membranes have lower stacking order. This is supported by their slow axis vector representations (b,e), which are expanded view of the boxed areas (scale bar, 10 mm), polar histograms of the azimuth angles and the in-plane order parameters (c,f). (c,f) Predicted organization of graphene sheets in membranes.
8
NATURE COMMUNICATIONS | 7:10891 | DOI: 10.1038/ncomms10891 | www.nature.com/naturecommunications