Multilayer graphene coated vanadium(V) oxide as electrodes for intercalation based brackish water desalination

The sample preparation protocol, including hydrothermal cell and chemical vapor deposition (CVD) setup, is graphically represented in figure 1. Later the x-ray diffraction (XRD) spectra of the V₂O₅ powder, as-grown V₂O₅ honeycomb (V₂O₅-SS mesh), and graphene-coated (G-V₂O₅-SS mesh) structures are depicted in figure 2(a). Signature diffraction peaks corresponding to (110), (301), (310), (020) diffraction planes are omnipresent in all the three samples [27]. This observation implies the structural integrity of the V₂O₅ structure in all the three samples where the heat treatment during graphene coating hardly harms the honeycomb-like V₂O₅ structure. In addition to pristine V₂O₅ diffraction peaks, the graphene-coated V₂O₅ (G-V₂O₅)sample shows a feeble diffraction peak centered at 23° and a shoulder by strong peak at 43°, thus proving the presence of graphene coating [28] on V₂O₅ structure. Intense peaks at 45° and around 83° corresponds to the SS substrate.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Raman spectra of V₂O₅ powder, V₂O₅ grown on SS mesh, and G-V₂O₅ on SS mesh is presented in figure 2(b). Raman modes present in the V₂O₅ structures such as 149 cm⁻¹(B1 g), 310 cm⁻¹(Ag), 403 cm⁻¹(Ag), 532 cm⁻¹(Ag), 700 cm⁻¹(B1 g) and 983 cm⁻¹(B3 g) are observed in all the three samples [29]. The characteristic D-band and G-band of thin graphite-like structures are seen at 1348 and 1592 cm⁻¹ respectively, and two-dimensional band at 2650$\pm$2 cm⁻¹ that corresponds to a few layers' graphene (see figure S1 (available at stacks.iop.org/2DM/7/045025/mmedia)) is also visible for the G-V₂O₅ sample. In short, the XRD and Raman spectra validated the structure of grown V₂O₅ and G-V₂O_5. High-magnification SEM images of the V₂O₅ grown on SS mesh are shown in figure 2(c). The SEM images show a uniform honeycomb-like structure that grew vertically from the surface of each strand of the SS mesh. The SEM images in figure 2(d) show graphene covering the V₂O₅ structure grown on SS mesh. The coating of graphene improves the conductivity of the active material.

The transmission electron microscopy (TEM) results of the samples before (figure 3(b)) and after (figure 3(c)) desalination process is summarized in figure 3. The entire structure of V₂O₅ coated in graphene before desalination is shown in figure 3(b) (larger image). The following two images in figure 3(b) are the clear identification of the respective constituents: the upper image is V₂O_5, and the lower image is graphene. The particle embedded in graphene is a single crystal, and the respective d-spacing for both agrees with the phases. A similar set of images presented in figure 3(c) is taken after the adsorption desalination process. It is essential to elucidate that following the adsorption desalination process, the V₂O₅ is no longer a single crystal, but rather poly-crystals with a large number of defects. In the graphene, areas are sections with d-spacing of approximately 0.33 nm, whereas in other sections, the d-spacing can be as large as 0.42 nm. This is a potential indication that the Na atoms diffuse within the carbon and expand it more than 25%. Similar effects are observed in other regions, proving that it is not characteristic of this specific location. Similar effects are observed in the V₂O₅; yet, the most visible one is the poly-crystal frameworks. This combination of results demonstrates the presence of sodium and its detrimental effects on both graphene and V₂O₅. A complimentary demonstration of the presence of sodium is given in figure 3(a) (EDS results). This set of images shows the presence of all the constituents of the composite, but what is relevant is the presence of sodium that clearly shows its distribution. The EDS maps show a higher density in Vanadium(V), Carbon (C), Oxygen (O) when compared to Sodium (Na), implying a difference in concentration. Therefore, this proves that sodium is present in the composite after the adsorption desalination cycle.

Zoom In Zoom Out Reset image size

V₂O₅ is considered as one of the best performing anode materials for Na ion battery owing to the interlayer spacing and crystallography that can accommodate Na ion as graphite does for Li-ions [24]. However, the Na intercalation redox reaction in Na-ion batteries is a sluggish process which may not occur at high rate voltage sweep cyclic voltammetry (CV) testing for supercapacitors. For this reason, no redox peaks are expected in CV for the pristine powder sample (figure 4(a)). Hence, to make a more electrochemically active surface area that can facilitate high rate adsorption-desorption kinetics, we employed a hydrothermal process to grow high surface area honeycomb like V₂O₅ structures on SS steel mesh (see figure 2(d)). The results show that even at high sweep rates, the adsorption associated oxidation peaks start to appear in the CV (discussed in the next section).

Zoom In Zoom Out Reset image size

Growth of V₂O₅ in a honeycomb structure over a substrate and coating with graphene to increase its conductivity allows us to investigate their electrochemical performance as supercapacitors. The capacitive behavior and reversibility of the electrode materials are studied using CV. A three-electrode system using a platinum spring as the counter electrode, Ag/AgCl as a reference electrode, and sodium chloride as the electrolyte were used for the electrochemical tests. For a comprehensive analysis, CV was performed for pure V₂O₅ powder (figure 4(a)), grown V₂O₅ on SS Mesh(figure 4(b)), and G-V₂O₅ (figure 4(c)). CV was performed on pure SS steel mesh as well to evaluate the contribution towards the total capacitance of the electrode, and it was found to be negligible (See SI figure 2). The CV's were performed at scan rates of 1 mV s⁻¹, 2 mV s⁻¹, 3 mV s⁻¹, and 5 mV s⁻¹ within the potential window of 0–0.8 V. The V₂O₅ powder showed capacitance of 67.80 F g⁻¹, while the grown V₂O₅ shows a capacitance of 161.37 F g⁻¹ at 5 mV s⁻¹. Furthermore, coating graphene increased the capacitance to 304.47 F g⁻¹ at 5 mV s⁻¹, which is a 55% increase from the as-grown V₂O₅ SS Mesh samples.

Both (figures 4(b) and (c)) cyclic voltammograms showed the oxidation and reduction peaks, although the reduction peak is very clearly visible in the G-V₂O₅ sample. The G-V₂O₅ sample shows two oxidation and reduction peaks. The peaks are due to adsorption occurring individually both at the carbon and V₂O_5, which in turn improves the capacitance. Both samples show the pseudocapacitance characteristics and the symmetric cycling curves indicative of the stability of the electrodes. The initial and final impedance curves (figure 4(d)) of both V₂O₅ and G-V₂O₅ showed very low impedance values, which is one of the primary criteria to act as a supercapacitor electrode. The impedance of the G-V₂O₅ electrode system is lower than that of the pristine V₂O₅ electrode, which is an indication of the improved electrical conductivity of the electrodes. The excess surface area offered by graphene improves the capacitance to the next level by incorporating the non-faradaic capacitance and the faradaic processes that happen at the V₂O₅ electrodes.

Density functional theory (DFT) calculations were performed to better understand the interaction between the V₂O₅ and graphene surfaces. All the simulations were completed using VASP [30–35]. Here we considered a 14-atom unit cell of V₂O₅ (four—vanadium atoms and ten—oxygen atoms) with lattice parameters (a = 3.61 Å, b = 11.55 Å, c = 4.79 Å) [36]. VASP simulations were performed using PAW-PBE potential with the following settings ISMEAR = 0 (Gaussian smearing), IDIPOL = 3, and ISPIN = 2 (spin-polarized calculation). Vander Waals contributions were included to account for the carbon atom interactions in the graphene. Vander Waals corrections were computed using Grimme D2 method (i.e. IVDW = 1). The cut-off energy (ENCUT) was set at 520 eV and EDIFF = 1.0 × 10⁵, EDIFFG = −0.02. Remaining input simulation parameters are listed for sake of reproducibility ALGO = Fast, ISMEAR = 0, MAGMOM = 48*5.0 120*0.6 192*1.0, PREC = N, SIGMA = 0.1.

The structure considered is orthorhombic since, at low-to-moderate temperatures, this is the most prevalent crystal phase [37, 38]. The bulk structure was relaxed, and cut planes along directions (010, 001, 100) were generated (figure 5(a)). The surface energy of the 001 face was lowest of three calculated surfaces (E₀₁₀ = 1.38 J m⁻², E₁₀₀ = 0.41 J m⁻², E₀₀₁ = 0.033 J m⁻², table 1).

Zoom In Zoom Out Reset image size

The stability of (001) surface is also confirmed by preferential crystal growth seen along this direction, as inferred from XRD data in figure 2(a). This agrees with earlier reported literature observations as well [22, 23]. Further, the surface energies for (001) surface were calculated with additional layers. A bilayer (001) surface had E_surf = 0.031 J m⁻² (a difference of 0.2% compared to single-layer structure). The difference is small since V₂O₅ has a layered structure along [001] direction leading to weak interlayer interactions [39, 40]. Hence an interface with a single layer (001) face V₂O₅ as a substrate with a layer of graphene deposited on top was prepared, see figure 5(b). A supercell was constructed by minimizing the lattice vector mismatch of less than 1% between the V₂O₅ and graphene layers. Since V₂O₅ and graphene have different crystal symmetries, orthorhombic and hexagonal respectively, a fairly large supercell consisting of 360 atoms (48—vanadium, 120—oxygen, 192—carbon atoms with dimensions a = 7.24 Å, b = 69.16 Å, c = 23.04 Å) was needed to achieve the desired lattice mismatch threshold (See SI S3). The thickness of the vacuum layer is 15.98 Å and interlayer spacing of ∼3 Å. The thickness of the electrode is 7.06 Å.

The electronic density of states for both the V₂O₅ (001 surface) and the V₂O_5—graphene interface is plotted in figures 5(c) and (d). V₂O₅ surface exhibits a very clear bandgap with a value of about 1.91 eV, suggesting semiconducting behavior (figure 5(c)), while the V₂O₅–graphene interface has minor states populated near Fermi level E_F suggesting semi-metallic nature (figure 5(d)). We further analyzed the DOS by calculating the individual elemental contributions of V₂O₅, and C atoms. From figure 5(d), it can be seen that the conduction bands of bare V₂O₅ have shifted near to Fermi level states, and new valence states are introduced due to carbon atoms (i.e. graphene). Thus, the deposition of graphene on V₂O₅ leads to higher conductivity and an electroactive surface, which is in line with the electrochemical measurements of previous sections.

The surface charge distributions were studied by calculating charge density difference plots. It was obtained by subtracting the charge densities of the individual graphene and V₂O₅ components from the G-V2O5 composite. Figure 5(e) shows the iso-surface of the charge density difference at iso-level 0.00134 (yellow color—positive, blue color- negative). It can be seen that charge transfer happens from carbon to V₂O_5, particularly at the O sites near the graphene layer. Bader charge analysis was done to calculate the amount of charge accumulated at each atomic location. For the graphene, we observed a net charge transfer of 0.0118 e from graphene to the V₂O₅ surface. Bader charge of the graphene atoms in the surface can be seen varying between 3.8 and 4.2 e with a mean charge of 3.9882 e compared to the bare carbon valence charge of 4 e. Locally, we see Bader charge transfer of 0.0321–0.0327 e as shown in figure 5(f). This further confirms the charge transfer from graphene to V₂O₅, and similar behavior has been observed in the literature [41].

The performance parameters, salt adsorption capacity, charge efficiency, coulombic efficiency, and energy per ion removed, were calculated and plotted with respect to time/cycle numbers. The following equations were used to calculate the performance parameters. Salt adsorption capacity (SAC) represents the amount of the salt removed (mg) during the adsorption cycle per unit weight (g) combined with both the electrodes.

$$\begin{equation}SAC = {\rm{ }}\frac{{Q{\rm{*}}\int_{{t_0}}^{{t_{ads}}} {\left( {{C_o} - {C_t}} \right)} {\rm{ }}dt}}{{{W_{electrode}}}}\end{equation}$$

Where $Q$ is the flow rate in ml min⁻¹, ${t_0}$ and ${t_{ads}}$ are the cycle initiation and adsorption cycle completion times, ${C_t}$ and ${C_o}$ are the instantaneous effluent concentration and feed concentration respectively in mg l⁻¹ and ${W_{electrode}}$ is the combined electrode weight in grams. For an adsorption cycle, the ratio of the moles of the salt removed per unit mole of the electrical charge supplied during the adsorption cycle is termed as charge efficiency.

$$\begin{equation}Charge{\rm{ }}\;Efficiency = \frac{{{\rm{ }}\left( {Q{\rm{*}}\int_{{t_0}}^{{t_{ads}}} {\left( {{C_o} - {C_t}} \right)} {\rm{ }}dt/m} \right)}}{{\left( {\int_{{t_0}}^{{t_{ads}}} I *dt/F} \right)}}\end{equation}$$

Where, $m$ is the molecular weight of the salt (58.44 g mol⁻¹ NaCl), and F is Faraday's constant (96 485 C eq^–1). The coulombic efficiency represents the roundtrip charge efficiency. It is calculated as the ratio of electrical charge released during the desorption stage to the electrical charge supplied during the adsorption stage.

$$\begin{equation}Coulombic{\rm{ }}\;Efficiency = {\rm{ }}\frac{{\left( {\int_{{t_{ads}}}^{{t_{des}}} I *dt/F} \right)}}{{\left( {\int_{{t_0}}^{{t_{ads}}} I *dt/F} \right)}}\end{equation}$$

Energy consumption (J) during the adsorption stage was calculated by integrating the current supplied during the adsorption cycle and multiplying with the applied voltage. Furthermore, the calculated number was divided by the number of moles of the salt removed, a factor of two (to convert from moles to number of ions for NaCl) and a factor of 2480 J mol⁻¹ (RT = 2.48 kJ mol⁻¹) to convert the calculated value into the 'Energy per ion removed (KT)' units [2]

$$\begin{equation}Energy\;consumption = \frac{{{V_{ads}}*\int_{{t_0}}^{{t_{ads}}} I *dt}}{{{\displaystyle \left( {Q*\smallint_{{t_0}}^{{t_{ads}}} ( {{C_o} - {C_t}} ) dt/m}\right)} \atop {\displaystyle *2*2480}}}\end{equation}$$

Quantitatively, the salt removal performance was evaluated in terms of SAC, and the results are shown in figure 6(b). The average salt adsorption capacity for 5 mM feed concentration is around 6.7 mg g⁻¹. Further increasing the feed concentration leads to the saturation of SAC at 8.5 and 12.5 mg g⁻¹ for 10 and 15 mM NaCl feed solution respectively. Again, the fluctuation in the SAC values is probably because of the fluctuating baseline of the effluent concentration curves.

Charge efficiency (CE) is calculated to evaluate the extent of energy utilization towards the salt removal process. The results are shown in figure 6(c). The CE trend for 5 mM indicates a slightly lower charge efficiency (∼75%) than at the (∼85%) 10 and 15 mM feed concentration tests. Probably, at a lower concentration, the solution resistance in the spacer channel has high transport resistance leading to the increased solution resistive losses. These factors combined might have led to lesser charge efficiency at low feed concentration. This indicates that to evaluate the desalination performance selecting an appropriate feed concentration is very important.

The coulombic efficiency was calculated to evaluate the possibility and extent of any faradaic reactions. As shown in figure 6(c), the coulombic efficiency is approximately 86%, 80% and 85% for 5, 10 and 15 mM feed concentrations respectively. These high columbic efficiencies indicate that there is very little faradaic loss in this system. Furthermore, it should be noted here that the coulombic efficiency values are stable indicating that there was no intermittent noise within the desalination process.

Energy consumption per ion removed is calculated and presented in figure 6(d). As was seen in the charge efficiency case, the energy consumption in the case of 5 mM feed concentration (∼27 units) was higher than the other two cases (∼23 and ∼20 units for 10 and 15 mM feed concentration respectively). The important thing to be noted here is that the energy consumption for 10 and 15 mM feed concentration is very competitive as compared to the values reported in the various CDI literatures [2].

Later the SAC and specific capacitance values and other relevant parameters are tabulated and compared with the existing literature. The as developed G-V₂O₅ samples prevail most of the reported materials on feed conductivity, second-best in SAC, energy-efficient in terms of adsorption voltage and second highest in specific capacitance. More details are provided in the supplementary information, see SI S4 & S5. All these studies used a composite electrode based on graphene or graphene oxide materials. One major difference is that we used slightly higher feed water salinity, but this is due to our previous experience that low salinities (<200 us cm⁻¹) are quite far from practical use. More importantly, solution resistance becomes very high leading to higher specific energy consumption and side reactions. We reported the charge efficiency and specific energy consumption values for several cycles which is not the case for previous studies which in turn rendering them difficult to implement.

Vanadium pentoxide was prepared using a hydrothermal setup, and a layer of graphene was coated on the surface. A low purity V₂O₅ as a raw material was used to make this process more cost-effective. The primary solution was prepared by using a 0.1 g of V₂O₅ dissolved in 30 ml of double-distilled water followed by 1 ml of hydrogen peroxide. The mixture was stirred well for 4 h before adding it into an autoclave containing 5 × 1 cm vertical strips of SS mesh. The SS mesh was sonicated in an isopropanol medium and air-dried for 10 min. The mesh was weighed and transferred into a hydrothermal setup. The mixture was heated to 205 °C for 16 h to obtain a uniform growth. Later the SS mesh was removed from the autoclave and repeatedly washed with water and ethanol to remove impurities and weakly grown particles. The mesh was vacuum dried at 90 °C for 12 h before using a CVD technique to coat a few layers of graphene on the surface of the as-grown V₂O₅ honeycomb-like structure. The coating of graphene is expected to improve electrical conductivity. Ultra-pure acetylene (99.99%) was used at a flow rate of 100 sccm to grow a few layers of graphene at 625 °C for 3 min. The V₂O₅ grown on SS mesh and the V₂O₅ grown on SS mesh and coated with graphene were taken for several characterizations.

A field emission JEOL JSM 7500 F FE-SEM operating at 5 kV was used to image the surface morphology of the sample at micrometer length scales. TEM images were recorded in a Jeol 2010 F system operated at 200 kV. The TEM is equipped with an Energy-dispersive x-ray for chemical analysis. A Renishaw in Via Raman microscope with an excitation wavelength of 532 nm was used to record the Raman spectra. An autolab potentiostat under three-electrode configurations was employed to run the CV. Here the G-V₂O₅ acts as a working electrode, Ag/AgCl as reference electrode platinum spring as counter electrode and 1 M Na₂SO₄ as the electrolyte. The voltage window with which the experiments were done is 0–0.8 Volts.

A 'flow by' type CDI cell configuration (as shown in figure 7) was used to conduct desalination tests on a 1 cm × 5 cm electrode. The same bench-scale test set up was used as in the previous work [13]. In brief, the setup included a feed and effluent container, a peristaltic pump, a CDI reactor unit, an electrochemical analyzer unit for voltage supply, and a pH and conductivity probe at the exit of the CDI reactor. Additionally, a computer system interfacing with the electrochemical analyzer, pH, and conductivity probe to control the applied voltage, desalination cycle time, record current, pH, and conductivity data. The CDI unit assembly mentioned above consisted of a cathode and anode electrodes of the same material and a 1 mm thick (61% porosity) spacer. Desalination tests were conducted using NaCl (5, 10, 15 mM) feed solutions. Desalination cycles consisted of 1000 s each adsorption and desorption, with 1.2 V applied during adsorption and 0.0 V during the desorption cycle. Additionally, before starting the desalination cycles, the samples were equilibrated with the feed solution for one hour. Conductivity values were recorded every 1 s and converted to the concentration (mg l⁻¹ NaCl) using an initial concentration-conductivity relationship (10 mM, 1.13 ms cm⁻¹).

Zoom In Zoom Out Reset image size