Mildly reduced graphene oxide membranes for water purification applications

Presently carbon allotropes namely graphene, graphene oxide (GO) and reduced graphene oxide (RGO) are being extensively utilized for water purification applications. The presence of myriad types of oxygen functional groups in the GO, however, makes this material very hydrophilic, allowing it to absorb water and to swell in moist or watery environments and to significantly damage its intended performance. In contrast, fully reduced graphene oxide membranes are not stable due to fewer oxide groups which are mainly responsible for GO flakes stacking. In the present work, the aforementioned problems are overcome by optimizing the oxygenated functional groups to develop mildly reduced graphene oxide (MRGO) membrane over PVDF (polyvinylidene fluoride) support. GO is reduced by L-Ascorbic Acid (LAA) with different amounts of wt.% and an optimized MRGO membrane is achieved at 10 wt.% of LAA, which is stable and showing comparatively lower swelling than GO membrane. All related structural and optical characterizations like XRD, SEM, EDAX, Raman, FTIR, and Contact angle have been done to evaluate the effect of mild reduction of GO. The studies are indicative of their potential application in water purification.


Introduction
Water purification is a method that eliminates harmful chemical compounds [1], organic/inorganic substances [2] including pharmaceutical wastes [3], cations [4] and biological pollutants [5] from water. It also involves distilling (turning a liquid in vapor and further condensing it in liquid form) and deionizing (ion removing by extracting dissolved salts) [6]. The purification of water is mainly aimed at providing clean water. Water purification also meets the needs of safe and drinking water applications in medical, pharmacological, chemical and industrial applications. The purification process reduces contaminant concentrations such as suspended particles, parasites, bacteria, algae, viruses, and fungi [7]. Many technologies are being used for water purification in which membrane-based separation is at high demand because it operates without heating and thus utilizes less electric energy than standard processes of heat separation like distillation, sublimation or crystallization [8]. Since the practical discovery of the first 2D material called graphene, much focus is paid to two dimensional structures in condensed matter physics, material science and chemistry [9,10]. Due to its unique atomic thickness and micrometer lateral dimensions, 2D materials have increasingly been explored as a fundamental medium to establish separation technologies [11][12][13][14]. Graphene oxide (GO), the oxidative form of graphene, is rated at a high level due to its unique permeation path, large surface area, outstanding anti-fouling properties, high chemical tolerance, high hydrophobicity [15,16]. It is a single layer of carbon monoatomic on the basal planes and sides, formed in a honeycomb with oxide groups (epoxide, carboxyl, and hydroxyl) [17,18]. Water permeance and dye rejection depend on various parameters. In a recent review article by Moghadam et al [19], membranes in form of atomically thin nanosheets or assembled laminated sheets using 2D materials like graphene , the filtration mechanism of gas and liquid species depend on size and density of pores, defects in the form of pores in the basal plane, interlayer spacing as well as surface functionality. Further in GO, the stacking arrangement of GO nanosheets, degree of reduction that determines the presence of functional oxygen groups, flake size and intrinsic defects [14] are the governing factors.
In another work by Du et al [15], they have discussed the dependence of permeation of liquids in GO membranes on the basis of complexity in layered, nano-porous structure as well as in the spatial distribution of oxygen-based functional groups on the basal plane. Permeation mechanism of water in GO membranes has been discussed in the light of Slip theory with slip length as a deciding parameter for level of interaction between water molecules and nano-channels of the GO membrane [20]. High adsorption of dye on the partially reduced graphene oxide sheets was previously reported by Alireza et al [21] and Kanika et al [22]. Various other mechanisms like expanded interlayer spacing, possible wrinkles, holes and inter-edge spaces for water permeation, solute-solvent membrane interaction with emphasis on solvent polarity, strong ion diploe forces have been discussed for organic molecules.
These oxide groups are responsible for GO flakes stacking [23] for making a stable (tough) membrane and also responsible for membrane swelling [24]. Membrane swelling in wet conditions is caused by the interaction of strongly hydrophilic GO nanosheets, which draw water molecules into the interlayer space of the GO membrane, hydrate the GO nanosheets and increase the d-spacing, which greatly hinders the performance of separation. [25,26].
If all oxide groups are removed then the developed membrane will not be stable in an aqueous medium. Hence, an optimum concentration of oxide groups on GO is essentially required so that the membrane swelling in wet condition could be first sufficient not to degrade. In this report, mildly reduced graphene oxide membranes are developed using L-Ascorbic Acid in different wt.% to optimize the concentration of oxide groups so that the MRGO membrane could be achieved which is stable and also show less swelling compared to other MRGO membranes.

Materials
Hydrogen peroxide (H 2 O 2 ) and graphite powder have been obtained by Sigma Aldrich, U.K. PVDF was imported from Millipore, USA. Other chemical components were obtained from Himedia such as KMnO 4 , HCl, H 2 SO 4 , LAA, NaNO 3 . Throughout the entire cycle of processing, MQ water (Millipore, United States, 18.2 MΩ cm) −1 was used.

Synthesis of MRGO
Graphene oxide (GO) synthesis was carried out using the well-known Hummers method [27] and is described in detail in our previous report [28]. For mildly reduced GO [29], a solution of 1 mg ml −1 in MQ water was sonicated for 30 min in an ultrasonic bath. The homogeneous GO solution was added with LAA of different wt. % using a magnetic stirrer for 5 min then the temperature was raised to 80°C for two hours. After that the solution was washed with MQ water using vacuum filtration until a pH value of 7 was attained , then the material was dried overnight in a glove box at 70°C. Since the focus of the work is to develop mildly reduced GO membrane, small wt.% of LAA was used. In the present work, 5, 8, 10, 20 and 40 wt.% of LAA has been used to reduce to GO.

MRGO Membrane Preparation
For MRGO membrane development, a 5ml MRGO solution (1 mg ml −1 in MQ water, sonicated for 5 min) was used over the PVDF membrane (0.22-micron pore size) with the help of pressure-assisted filtration assembly. Five membranes of different MRGO solutions namely MRGO5, MRGO8, MRGO10, MRGO20, and MRGO40 were developed with 5, 8, 10, 20 and 40 wt% of LAA respectively. All the five fabricated membranes were dried in a glove box at room temperature over two days. The measured thickness of all the developed membranes is 25 ± 3 μm (Surface Profiler, Dektak 150, Veeco Inc., USA). For comparison, a graphene oxide membrane is also developed with the same specifications and procedures. Figure 1 shows the digital images of all developed membranes. It can be seen that GO, MRGO5, MRGO8, and MRGO10 membranes are having stacked surface whereas MRGO20 and MRGO40 membranes are not showing the stacking surface as verified by a closer view of MRGO10 and MRGO20 membranes (figures 1(g) and (h)). The white color PVDF support can be seen in the closer view of MRGO20 membrane. It can be concluded that due to fewer oxide groups, stacking is almost zero in MRGO20 and MRGO40 membranes. Therefore, MRGO40 sample is discarded for further characterizations.

Characterization of GO and MRGO
Raman Spectroscopy (λ = 514.5 nm and E l = 2.41 eV, Renishaw Model no. Invia II, U.K.) was used to study the phase purity of GO and MRGO. The presence of functional groups in the prepared GO and MRGO samples were studied using Fourier transform infrared (FTIR, Perkin Elmer, Model no. Spectrum RXI, USA) spectroscopy.

Characterization of GO and MRGO Membranes
The interlayer spacing of GO and MRGO was measured using x-ray diffraction (XRD, Bruker D8 Discover, USA) with Cu Kα radiation (1.54059Å). The technique of contact angle measurement was used to test the hydrophilic nature of the developed membranes.

Results and discussion
3.1. Characterization of GO and MRGO Raman Spectroscopy, due to its non-destructive, rapid and high-resolute analysis, provides structural and electronic data, which is a simple characterization of carbon allotropes [30]. Literature studies indicate the presence of two major peaks for graphene oxide, D peak at 1345 cm −1 agreeing to structural defects (sp 3 hybridized carbon atoms) or partly disorderly graphitic domains, and G peak at 1600 cm −1 related with graphitic carbons (sp 2 hybridized carbon atoms) [31,32]. Figure 2 shows the Raman spectra of GO and different MRGO samples having D and G peaks as indicated in literature for GO samples [33]. It can be seen from table 1, there is a shift in the G band peak with changes in the FWHM. The calculated ratio of the D and G peak intensities (I D /I G ) of graphene oxide comes out to be 0.876 which increases to 1.18 after the mild reduction for MRGO20. These variations substantiate major changes in the surface structure [34] that are further confirmed by figure 1 and later in figure 5. The shifting of the G band toward the high-frequency region also substantiates defect enhancement of the graphene oxide structure.
This enhancement in I D /I G ratio is due to the change of the electronic conjugation state and stack defects of GO due to reduction [35,36]. As shown in figure 2, the I D /I G ratio improved significantly with the reduction and we can also see the appearance of 2D band. The shape, position, and intensity of the 2D band, which is a secondary D peak, is indicative of the quality of carbon rings in the graphene layers. The larger intensity is indicative of a single layer of graphene whereas it has a reduced intensity in case of multilayer graphene. As seen from figure 2(c), the 2D band is narrow and have low intensity whereas as we increase the reduction as represented in figure 2(e), the 2D band is broader with enhanced intensity. The transition is indicative reductive transformation of the graphene oxide structure with reduced number of layers and increase in defects [37,38]. This is very much in line with the GO's Raman spectrum reduced by hydrazine stated by Stankovich et al, implying that there was a reduction [36]. The calculated values of I D /I G ratio are summarized in table 1 also.  As reported in a previous work by Naoki et al [39] there was a clear inverse correlation between the amount of hydrazine i.e. increase in the degree of reduction and the oxygen content. Elemental analysis of the rGO showed that with increase in degree of reduction, the oxygen content of rGO proportionally decreased from 48 to 17 wt%. The minimum decrease in the oxygen content was 11% for maximum degree of reduction. Therefore, in our case of MRGO, it is expected that C/O ratio will increase and will be in between that of GO and rGO. It has been confirmed through the EDAX analysis results as shown later in figure 6.
FTIR spectroscopy has also confirmed the reduction of oxygen-containing groups in GO by L-AA as shown in figure 3. The intensities of the FTIR peaks referring to the presence of oxygen containing functions groups , such as the C=O and C=C stretching vibration peaks of carboxyl groups at 1723 cm −1 and 1620 cm −1 respectively, the vibration and distortion peak of O-H or COOH groups at 3395 cm −1 , the C-O (epoxy) stretching vibration peak at 1225 cm −1 , and the C-O (alkoxy) stretching peak at 1038 cm −1 [40,41] reduced slightly with the GO reduction. It is interesting to note from figure 3 that the oxygen containing functional groups have not totally disappeared indicating there is a minor reduction in the oxide functional groups in the MRGO samples compared to the GO sample due to the mild reduction process. It also confirms that MRGO has not been completely reduced into rGO [42]. Hence, both the Raman and FTIR studies confirm the reduction of the GO samples and with increasing the concentration of LAA, the reduction increases.

Characterization of GO and MRGO membranes
The XRD technique is used to calculate the inter-layer distance of the developed membranes as well as degree of reduction. The characteristic graphite x-ray diffractogram contains the two distinguishing peaks of this material, which are at 2θ = 26°(d 002 = 0.34 nm), and 2θ = 55°(d 004 = 0.17 nm), which matches with the crystallographic planes (002) and (004) respectively. Such peaks are moved to lesser angles after oxidation. In GO x-ray diffractogram, there is a (002) peak at 2θ = 10.6°(d 002 = 0.834 nm) which confirms oxidation efficiency, as the graphene nano-platelets are increasing inter-planar distance because oxide groups are inserted [43,44]. After the mild reduction, the 002 peaks shifted towards a higher angle because of the reduction of oxide groups. It can be seen from figure 4 that as the percentage of reduction increasing, the 2θ peak is also increasing which contributes to the reduction in the inter-layer distance of the membranes. It can be clearly seen from the XRD results that there is an increase in the FWHM as the degree of reduction increases. The results are in conformity to the exfoliation and reduction processes of GO where the functional oxide groups are reduced. Also, no peak is observed at higher angles between 2θ = 23°−26°. In an earlier work by Naoki et al [39], it was observed in the XRD spectra, that as the reduction proceeded, the high angle peak disappeared when the oxygen content was less than 33 wt%. Further, no peak appeared at 2θ = 25°-27°, indicating that the layered graphitic structure was not recovered. Thus, the XRD results in our case are indicative of partial reduction of GO into MRGO. The calculated values of interlayer spacing are also summarized in table 2.
In order to further ascertain partial reduction, SEM and EDAX analysis was carried out. Figures 5(a)-(d) shows the surface morphology of GO, MRGO5, MRGO8 and MRGO10 membranes using SEM micrographs. The SEM morphology in figure 5(a) indicates that the conglomerates in graphene oxide have flakes which are tightly grouped . As the reduction is increased (figures 5(b)-(d)), these become much looser. It can be interpreted that the flakes are affected by strong Van der Waals bonds during aggregation as increase in the reduction level  leads to a decrease in the oxygen-containing functional groups. It can be seen that the GO membrane has the smother surface due to the higher concentration of oxide groups that are responsible for the compaction between the GO sheets. As we are reducing the oxide groups using chemical reduction, the compaction is decreasing with increase in the reduction level. Figure 6 shows the EDAX analysis of the developed membranes. It indicates that as we are increasing the reduction degree, the C/O ratio is increasing from 1.55 for GO to 1.93 for MRGO20 membrane. All the characterization results confirm partial reduction.  Water contact angle measurements were carried out to characterize the hydrophilic properties of the membranes in which a low contact angle suggested generally a hydrophilic existence [45]. Figure 7 shows that the GO membrane has a water contact angle of approximately 45°. Nevertheless, the MRGO5, MRGO8, and MRGO10 membranes are still hydrophilic and the water contact angles are 48°, 55°, and 60°respectively. As the LAA wt% level gradually increases, the membrane surface wetting decreases. This is because of eliminating a small number of functional groups containing oxygen on the surface of the membrane which is responsible for membrane surface wetting [46].

Permeation and swelling performance of membranes
The rejection of dye, methylene blue (MB) with an active membrane area of 12.56 cm 2 was tested for all the developed membranes. The experiment was conducted with a vacuum filtering system (Dead end filtration) for 5 h with reliability tested over fifteen days. 50ppm MB dye solution was applied to verify the permeation and percentage rejection of all the membranes, as summarized in table 2. The values are measured at room temperature at the pressure of 540 mm Hg and are based on constant values obtained by repeated tests. The water flux, F (Lm −2 h −1 ), is determined by equation (1) at 540 mm Hg applied pressure [47].
where V represents the maximum (50 ppm) MB filtration solution that flows through the membrane in the litre (L), filtration time is t in an hour (h), and the active filtration area is A in m 2 . It can be seen from figure 8, that the water permeation decreases and dye rejection increases as we increase the degree of reduction of GO. This is attributed to the fact that the spacing between the layers depends on the oxide groups present on the membrane surface that decrease with the degree of reduction. Here it can be seen that the GO membrane is giving maximum flux but lowest dye rejection whereas MRGO10 membrane showing lowest flux but 100 percent dye rejection. Figure 9 shows the 50 ppm MB dye rejection using UV-vis scan of the permeate solution for all the developed membranes. It can be clearly seen that MRGO10 has completely rejected the MB solution.
The swelling performance of all membranes is shown in figure 10. It can be seen that in the starting all the five membranes are showing the 100% dye rejection but as the time increases the rejection decreases for GO, MRGO5, and MRGO8 membranes.
Zheng et al [24] has shown that there is a significant increase in the interlayer d-spacing of a GO membrane in wet condition as compared to dry condition. They measured the d-spacing of a dry layer-stacked GO film using XRD as 0.8 ± 0.1 nm and values obtained in the present investigation are in good agreement. The reason for increase in the interlayer 'd' spacing was attributed to the hydrophilic nature of the GO nanosheets. They attracted the water molecules and thereby hydrated the GO sheets. The number and distribution of functional oxygen groups has a major role in deciding the hydrophilic nature of GO nanosheets through interlayer d spacing. Thus, the dye rejection is directly dependent upon interlayer spacing of developed membrane. As we can see from the swelling graph due to swelling of the membranes, the dye rejection performance is poor for membranes having higher amount of oxides. It is in consonance to our XRD, SEM and EDAX results. In another work by Joshi et al [48], they have shown a sharp molecular cutoff of 9 ± 1 A°for GO laminates in dry state. The interlayer 'd' spacing increased to 13 A°when they were soaked in water. Therefore, taking into consideration the thickness of graphene to be near 3.4 A°, GO membranes after swelling has a poor dye rejection of methylene blue molecule, which has size of is 0.84 nm. However, the MRGO membranes after swelling have an increasing dye rejection because of the reduced interlayer d spacing with 100% rejection ration for MRGO10.
Thus, it can be related to the interaction of highly hydrophilic GO nanosheets with water which attract water molecules to GO's interlayer space, hydrating the GO nanosheet, thus increasing the d-spacing and significantly decreasing the dye rejection. Since the GO membrane has a large percentage of oxide groups, therefore it gets more hydrated compared to those containing lesser percentage of oxide groups such as MRGO membranes. GO, MRGO5, and MRGO8 membranes show swelling as indicated by a decrease in dye rejection with time whereas MRGO10 membrane provide 100% MB dye rejection and no swelling due to less percentage of oxide groups present on its surface for long time filtration. The results are encouraging for the realization of efficient membranes for the water purification systems.
The stability and reproducibility in dye rejection of the MRGO10 membrane was measured for more than three months as shown in figure 11. It has been found that there is no swelling of the membrane showing a dye rejection variation of less than 1%. The variation with error bars is as shown below-

Conclusion
Mildly reduced GO membranes were developed over PVDF support to check the MB dye rejection performance. It is concluded that MRGO20 and MRGO40 membranes were not stable due to fewer oxide groups which are responsible for GO-GO flakes stacking as we can from figure 6 that the C/O ratio is increasing which means that the oxide groups are decreasing. Here we optimize the reduction degree for membrane stability and 100% MB dye rejection. MRGO10 membrane showing the 100% MB dye rejection and water flux of 46 lm −2 h −1 at 540 mm Hg pressure. It can therefore be concluded from this study that the percentage of oxide groups plays a key role in the stacking of Graphene oxide membrane and they can be suitably optimized for desirable usage in water purification.

Acknowledgments
The authors are grateful to the University Grant Commission (UGC), Department of Electronic Science, South Campus, Delhi University and Principal, Acharya Narendra Dev College, University of Delhi for financial and infrastructure support.