Preparation and Characterization of PANI/MWCNT/RGO Ternary Composites as Electrode Materials for Supercapacitors

Electrode materials for supercapacitors are extensively studied at present; however, their structural and electrochemcial properties still require further improvement. Hence, a series of polyaniline/reduced graphene oxide binary composites (PANI/RGO, PG-i, i = 1–5) and polyaniline/multi-walled carbon nanotubes/reduced graphene oxide ternary composites (PANI/MWCNT/RGO, PCG-i, i = 1–6) were prepared by a facile hydrothermal method. The morphology, structure and electrochemical performance of these samples are systematically analyzed and discussed. Analysis shows that all the samples have abundant pore structure. By changing the ratio of components, PCG-5 shows superior comprehensive electrochemical performance, including a high specific capacitance of 478 F g−1 at 1 A g−1, a considerable capacitance retention (63.56%, from 1 A g−1 to 20 A g−1, and 55.33%, from 1 A g−1 to 50 A g−1) and an extraordinary cycling stability (64% after 3500 cycles, at 2 A g−1) in 1 M H2SO4. The results prove that the PANI/MWCNT/RGO ternary composites have great potential as supercapacitor electrodes.


Introduction
In order to meet the needs of practical applications, neotype energy storage devices and equipment are constantly being studied. Supercapacitors (SC), namely electrochemical capacitors, are large-capacity capacitors between conventional batteries and capacitors. SCs are favored by various researchers [1][2][3][4][5][6][7][8][9][10][11] because of their high power density, superior cycle stability, safety in use and fast charge-discharge characteristics. The energy storage mechanism of SCs can be classified into two categories, including an electric double layer capacitor (EDLC) and pseudo-capacitor. 1,[12][13][14][15][16] The former stores energy through charge accumulation in the double Helmholtz layer at the interface between the working electrode and electrolyte, while the latter stores energy through faradaic process on the surface of electrode material, including ion intercalation and de-intercalation and charge transfer among atoms on the surface. [15][16][17][18] Graphene is a particularly attractive carbon material in the field of SCs because it has a unique two-dimensional (2D) honeycomb structure, a high theoretical specific surface area up to 2630 m 2 g −1 , excellent electrical conductivity and superior mechanical properties. 19,20 Its preparation methods mainly include mechanical peeling, redox reaction and chemical vapor deposition (CVD). Among them, the method of oxidizing graphite powder to graphene oxide (GO) and then hydrothermally reducing the graphene oxide is more commonly used because of its simple operation and low cost. However, due to the van der Waals force between the layers, reduced graphene oxide (rGO) is easily stacked and agglomerated, which has only 100-200 F g −1 of specific capacitance that can be obtained in reality. 21,22 Multi-walled carbon nanotubes (MWCNT) possess high specific surface area (1600 m 2 g −1 ), good stability and high electrical conductivity (105 S cm −1 ). 23,24 According to previous reports, when MWCNT are used as a supercapacitor electrode material, their experimental intrinsic capacitance is relatively low (4-135 F g −1 ). 25 This is because of the limitations of micropores and internal resistance. 23,26 Taking the structure stability and the unique nanoscale three-dimensional winding structure into account, MWCNT can insert into graphene sheets and reduce graphene agglomeration effectively. For example, Kumar et al. prepared rGO/ MWCNT film (GP10C) successfully using a self-assembly plus reduction method, which showed a specific capacitance of 200 F g −1 at 0.25 A g −1 . 27 Polyaniline (PANI), a typical pseudo-capacitance electrode material, is also a conductive polymer (CP) that has been widely studied so far. It has the advantages of high conductivity, high specific capacitance, good intrinsic flexibility and simple preparation. 28,29 However, as continuous charge and discharge can cause the structural degradation of the PANI chains (e.g. volumetric expansion), 30,31 the rate performance and cycle life of pure PANI are severely limited. Therefore, to overcome these shortcomings, it is usually used in combination with other more stable materials, such as carbon materials and transition metal oxides. 22,[32][33][34][35][36][37][38][39] For example, Dai et al. prepared a reduced graphene oxide/ polyaniline hollow sphere composite material (RGO/PANI-HS), which exhibited a specific capacitance of 529 F g −1 at 0.5 A g −1 . 37 Sardana et al. successfully synthesized PANI/ MWCNT composite hydrogels on carbon cloth by in situ oxidative polymerization, exhibiting a specific capacitance of 277.59 F g −1 at 0.25 A g −1 . 38 Han et al. prepared PANI/ CNT/graphene on a flexible ITO/PET substrate, and the sample showed a specific capacitance of 133.45 F g −1 at the scan rate of 100 mV s −1 . 39 In this work, we chose a simple one-step hydrothermal method for the preparation of PANI/MWCNT/RGO ternary composites. Theoretically, compared with pure graphene, the addition of PANI and MWCNT can reduce slightly the stacking of graphene sheets and contribute to specific capacitance, and graphene and MWCNT can slow down the structural collapse of PANI chains and extend the cycle life of the composite. As far as we know, there are few studies on the overall comparison between PANI/RGO and PANI/ MWCNT/RGO prepared by a hydrothermal method. In the present work, we intend to find the influence of PANI and MWCNT on the composite materials. Therefore, the morphology, distribution of mesopores and macropores, and electrochemical performance of these samples are mainly discussed and analyzed comprehensively. In addition, untreated carbon cloth (CC) was used as substrate to prepare binder-free working electrodes for characterizing the electrochemical performance of each sample, which can make full use of the hydrogel samples and characterize the original performance of these samples as accurately as possible.

Preparation
GO was prepared in advance by a modified Hummers' method. Then 0.05 g freeze-dried GO aerogel was weighed and dissolved in 50 ml deionized water (DI water). After that, 0.5 g thiourea (CS(NH 2 ) 2 ), 0.02 g MWCNT and X g dark green PANI powder (X = 0.01, 0.02, 0.02, 0.04, 0.05, 0.06) were added into the GO solution with continuous stirring. Subsequently, the mixture was transferred into hydrothermal kettles to react at 180°C for 12 h. After the reactor was cooled naturally, the obtained black hydrogels were filtered and washed several times with DI water. Lastly, these PANI/MWCNT/RGO composites were freeze-dried under vacuum for 24 h, and they were labeled as PCG-i (i = 1, 2, 3, 4, 5, 6). Furthermore, for comparison, MWCNT/RGO composite without PANI, graphene reduced only by CS(NH 2 ) 2 and PANI/RGO composites without MWCNT were prepared according to the above steps, and they were marked as MCG, RGO and PG-i (i = 1, 2, 3, 4, 5), respectively. Figure 1 is the schematic diagram of the material preparation process. To test the electrochemical performance of the samples, a series of working electrodes were made by a simple method. The active material was ultrasonically dispersed in ethanol (EtOH) directly. CC without treatment was immersed in acetone, alcohol, and DI water, and it was ultrasonically cleaned for 15 min each time. Then the dispersion was coated on cleaned CC and dried at 60°C several times.

Characterization
Scanning electron microscopy (SEM) (JSM-7500F, Japan) was used to characterize the surface morphology of these samples. X-ray diffraction (XRD) (DX-1000, China) with Cu K α radiation was used to characterize the crystalline structure. Raman spectroscopy (RS) (LabRAM HR, France) with a 532.17-nm laser was used to analyze the changes in the microstructure of these samples. Nitrogen adsorption-desorption (ASAP 2460, USA) and mercury intrusion porosimetry (MIP) (Autopore IV 9500, USA) were used to measure hydrogel specific surface area and pore-size distribution directly. Valence analyses were carried out using x-ray photoelectron spectrometry (XPS) (AXIS Supra (Kratos), USA) and charge calibration was carried out with a C 1s peak with binding energy of 284.6 eV. Electrochemical tests, including cyclic voltammetry curves (CV), galvanostatic charge/discharge curves (GCD) and electrochemical impedance spectroscopy curves (EIS) (10 6 -10 −2 Hz) were carried out by an electrochemical workstation (CHI660E, China). In a threeelectrode system, an Hg/Hg 2 SO 4 electrode and a platinum foil electrode served as the reference electrode and counter electrode, respectively. The electrode to be tested was used as the working electrode. All electrochemical tests were performed in 1 M H 2 SO 4 electrolyte.

Results and Discussion
Macroscopically, due to the self-assembly of GO in hydrothermal reduction, the prepared samples are hydrogels. 40,41 From Fig. 2 and Fig. S1, PANI (circled in red), MWCNT (circled in blue) and thin wrinkled graphene sheets can be observed, indicating that PG-i and PCG-i materials have been successfully prepared. The relatively small PANI particles (Fig. S2b) and MWCNT (Fig. S2c) are tightly wrapped or embedded in large graphene sheets. Because these composites are freeze-dried, they exhibit abundant pore structures, especially on the micrometer scale, which are beneficial to the diffusion of ions and the infiltration of the electrolyte to the working electrode. In addition, compared to PG-i, the three-dimensional network structure of MWCNT introduced in PCG-i may buffer and reduce the volume change and structural collapse of materials during the charging and discharging process, which can promote the transport of ions and electrons. However, due to the entanglement of MWCNTs, the uneven distribution of the components can also be observed, which may adversely affect the electrochemical performance of PCG-i. Figure S3 exhibits the XRD patterns of RGO, PANI, MWCNT, MCG, PG-4 and PCG-5. For intrinsic PANI, its characteristic peaks are located at 2θ = 14.77°, 2θ = 19.62°, 2θ = 25.52°, and their corresponding crystalline planes are (0 1 0), (1 0 0), (1 1 0), respectively. 42 The peak at 19.62° is attributed to the alternating polyaniline chain layers, the peak at 14.77° is related to periodicity parallel to polyaniline chains and the peak at 25.52° is related to periodicity perpendicular to polyaniline chains. 43 Some sharp peaks may be related to doping. Generally speaking, the characteristic peak positions of RGO and MWCNT in XRD patterns are about 2θ = 26° and 2θ = 43°, which correspond to (0 0 2) and (1 0 0) crystalline planes, respectively. As a result of the overlap of the peaks in the range of 2θ = 20-30°, the characteristic peak of these composites at 2θ = 25° is wider and more diffuse than that of pure RGO and MWCNT, indicating that these composites possess many lattice defects and a large degree of disorder.
Raman spectra of PG-4, MCG and PCG-5 are shown in Fig. 3. The Raman spectra of all samples are similar. They exhibit two conspicuous bands, namely the D band (−1350 cm −1 ) and the G band (−1592 cm −1 ). The D band is the disordered vibration peak of graphene, which represents the  defects in the graphene. The G band is caused by the inplane vibration of sp 2 carbon atoms, which represents the crystal phase structure of graphene. In addition, the 2D band (−2700 cm −1 ) is attributed to the double resonance transition of two phonons with opposite momentums in the graphene carbon atom. 5,44 A high Raman I D /I G value (where I D and I G are the intensity of the D band and G band, respectively) usually demonstrates a high degree of disorder, which means that the material can provide more electrochemically active sites and thus is beneficial to the combination with electrolyte ions and shows better electrochemical performance. Table I lists the detailed data of PG-i (i = 1-5), MCG and PCG-i (i = 1-6).
Valence analyses were carried out by XPS, and the fitted results are displayed in Fig. 4 and Fig. S4. Thiourea is used as both the nitrogen source and carbon source, and PANI is another nitrogen source during hydrothermal reactions, so the area of N 1s region of MCG in the survey spectra is obviously smaller than that of PG-4 and PCG-5. The high-resolution C 1s spectrum of PG-4 and PCG-5 can be well fitted to six parts, including C-C (284.   eV) and π-π conjugation (291.20 eV). The high-resolution N 1s spectrum can be fitted into three parts, namely pyridinic N (398.39 eV), pyrrolic N (399.82 eV) and graphitic N (400.72 eV). The presence of graphitic N proves that N has been doped into the hexagonal structure of graphene, which enhances conductivity and hydrophilicity of carbon composites. Pyridinic N and pyrrolic N provide π-electrons to the conjugated system in graphene layers. Therefore, they increase the specific capacitance through extra redox reactions. By comparing the area of each region, it can be seen that the relative content of pyrrolic N is the highest. The O 1s high-resolution spectrum can be fitted into four parts, namely O-S (531.14 eV), O-C (532.31 eV), O-C=O (533.58 eV) and O-C (534.89 eV). The S 2p high-resolution spectrum can be divided into three parts, namely -C-S-(164.23 eV), -C=S-(165.41 eV) and -C-SO x -C-, respectively. The first two S-containing groups (thiophene-like S) enhance conductivity, and the third S-containing group (oxidized S) participates in the Faradic reaction in liquid electrolyte to provide pseudo-capacitance. 45 It can be seen from the SEM images that composite materials have rich pore structures, especially mesopore (2-50 nm) and macropore (>50 nm) structures. Combined with Fig. S5d, the pores of the samples prepared in this work are mainly distributed in the range of mesopores and macropores. Many previous studies have shown that most graphene composites have a rich microporous structure. However, there are few specific reports on the analysis of their macropores and mesopores. Therefore, MIP was used to analyze the distribution characteristics of mesopores and macropores. Figure 5 shows the pore size distribution diagram of PG-4 and PCG-5. The interpolated table is the pore volume percentage in different pore ranges. It can be seen that both PG-4 and PCG-5 are rich in mesopores and macropores, which is conducive to the infiltration of the electrode sample in the electrolyte. The samples have such rich pore structures because the vacuum freeze-drying process can keep the microstructure of the hydrogel prepared by the hydrothermal method from being damaged. Figure 5b is a distribution diagram of the pore length ratio corresponding to the pore size, which can be used as a reference. It can be seen that in the range of <300 nm, the pore length of PG-4 and PCG-5 is long, especially when the pore diameter is less than 10 nm. In general, the smaller the pore, the deeper the pore. It can be inferred that there are three-dimensional microporous network structures inside these samples, which may be attributed to the incorporation of PANI chains and MWCNT. This plays a crucial role in improving the capacitance. MIP data analyses of these samples are listed in Table SI. From Table SI, the total pore area of each PCG-i sample is lower than that of the corresponding PG-i sample. This phenomenon is universal and reasonable. The reason is that MWCNT occupies some of the macropores and mesopores of the PG-i sample. The spacing between the layers of MWCNT is 0.34-0.42 nm, 46,47 which belongs to the pore size range of micropores, that is, the addition of MWCNT relatively reduces the macropores and mesopores and increases the micropores of these samples. Micropores generally have a longer pore length and contribute more to the specific surface area, while the instrument cannot measure the specific surface area of these pores in the range of the micropores' size. It can be seen that the porosities of PG-i (i = 1-5) and PCG-i (i = 1-6) are both above 84.79%, and both have a large total pore area (70-176 m 2 g −1 ). This means that there are many electrochemically active sites inside these materials. We can also see that within the range of pore distribution, which can be measured in this work, compared with RGO, the total pore area of PG-4 and PCG-5 has increased, and the porosity is only slightly reduced. This means that the addition of PANI and MWCNT has indeed contributed to the abundant pore structures. In addition, these materials are directly loaded on clean CC, which reduces the damage of grinding to the microstructures. Therefore, these samples may show superior electrochemical performance.
For electrochemical tests, the selected voltage window is −0.6 V to 0.4 V. From Fig. 6, it can be seen that the capacitance of PCG-5 is the highest. According to Fig. 6d and e, the CV curves of RGO and MCG are similar to rectangles, while those of the other composites exhibit obvious oxidation peaks and reduction peaks, which are attributed to the doping and de-doping of counterions that occur in the PANI chains. 31 This shows that RGO and MCG mainly have electric double-layer capacitance while other composites exhibit obvious pseudo-capacitance due to the incorporation of PANI. As the scan rate increases, the oxidation peak (or reduction peak) shifts slightly to the left (or right). This is because at high scan rates, the transmission speed of electrolyte ions and the rapid transmission speed of electrons do not match the transmission speed of electrons inside or near the active material, and a large number of electrons are gathered together. Therefore, the charge potential increases and the discharge potential decreases, and the potential difference between the two redox peaks also becomes obvious. The peaks at 0.02 V and 0.07 V gradually overlap due to their close positions. PG-4 and PCG-5 are mainly compared and analyzed, because they have the highest capacitance values in their respective groups. It can be seen that the CV curves of PG-4 and PCG-5 are gradually deformed as the scan rate increases. Compared with the CV curves of PG-4, those of PCG-5 deforms more obviously, which might be due to the entanglement of MWCNT.
According to Eq. 1, the specific capacitance (C m , F g −1 ) of composites can be calculated from GCD curves.
where I, ∆t, ∆V and m are discharge current (A), discharge time (s), potential window (V), and the mass loading of active materials (g), respectively.
The calculated specific capacitance values of RGO, MCG, PG-i (i = 1-5) and PCG-i (i = 1-6) at 1 A g −1 are 71.6, 97.9, 249, 277, 313, 317, 235.3, 139, 336.5, 293, 165.1, 478 and 113.9 F g −1 , respectively, as displayed in Fig. 7a, and in Fig. 7b, c, the capacitance retention of RGO, MCG, PG-i (i = 1-5) and PCG-i (i = 1-6) from 1 A g −1 to 20 A g −1 are 65.50%, 73.59%, 85.14%, 69.24%, 70.86%, 58.86%, 61.03%, 75.53%, 61.10%, 57.74%, 70.14%, 63.56% and 55.13%, respectively. Even if the current density is increased to 50 A g −1 , all the samples except PG-5 and PCG-6 can maintain above 50% capacity. RGO does not have as good a rate performance as expected, which may be related to the destruction of the structural regularity of graphene by the incorporation of N (1) C m = I × Δt ΔV × m and S atoms. In some samples, the effect of this damage is partially compensated by PANI chains and MWCNT. The incorporation of PANI can increase the specific capacitance (comparison between PG-i and RGO), and the incorporation of MWCNT can also slightly increase the specific capacitance of the material (comparison between MCG and RGO), and PANI and MWCNT also have a competitive effect in composite materials (comparison between PCG-i and PG-i), which is specifically manifested in that PCG-i does not have an obvious law of specific capacitance change like PG-i. In summary, PCG-5 has a good capacitance retention (from 1 A g −1 to 20 A g −1 ) and the highest specific capacitance. The EIS curves are illustrated in Fig. 7d and e. It is obvious that charge transfer process and diffusion-controlled process control the total electrode process jointly. Through the intercept on the horizontal axis, the semicircle radius in the high-frequency region and the slope of the straight line in the low-frequency region, the equivalent series resistance (R s ), the charge transfer resistance (R ct ), and the Warburg impedance (W) can be obtained, respectively. It is known that the when straight line is more parallel to the vertical axis, it is more similar to an ideal supercapacitor. According to the fitting results (Table SII)  According to Eqs. 2 and 3, the energy density (E, Wh kg −1 ) and the corresponding power density (P, W kg −1 ) of PCG-5 can be obtained.
where C m , ∆V, ∆t are the specific capacitance of the device (F g −1 ), potential window (V) and discharge time (s), respectively. Accordingly, the energy density of PCG-5 is calculated as 66. 39    45.51 W h kg −1 , and 42.19 W h kg −1 at the corresponding power density of 0.5 kW kg −1 , 1 kW kg −1 , 2.5 kW kg −1 , 5 kW kg −1 , and 10 kW kg −1 , respectively. As illustrated in Fig. 8, the capacitance retention (at 2 A −1 ) of PG-4 is 60% and that of PCG-5 is 64% after 3500 cycles of continuous charging and discharging. This test result is reasonable. The cycle stability of electric double-layer capacitive materials is usually higher than that of pseudocapacitive materials. The addition of MWCNT increases the proportion of EDLC materials in PCG-5, which improves the cycle stability when compared to PG-4. And because of the entanglement between PANI and MWCNT, MWCNT slows down the swelling of PANI to some extent during the charging and discharging process. However, because the uneven distribution of the components in the ternary material might be more obvious, the effect of this improvement is limited. The capacitance contribution can be calculated roughly according to the CV curves .17,48-53 which can also be expressed as, where i (V), k 1 and k 2 are the current at a specific potential V (V) and constants which correspond to the scan rate v (mV s −1 ), respectively. k 1 v and k 2 v 1/2 represent the current parts contributed by the surface-limiting capacitive process and the diffusion-controlled process, respectively. 54 The values of k 1 and k 2 under different potentials can be obtained by the linear fitting results according to Eq. 5, where k 1 is the slope of the fitted straight line and k 2 is the line's intercept on the vertical axis. Theoretically, the more points are selected, the more accurate the curve will be. In this work, only a dozen points were selected as fitting points, and due to the influence of electrochemical polarization and the inherent limitations of Eq. 5, the fitting results only serve as a reference. Moreover, due to the limitation of Eq. 4, the fitting of k 1 values is inaccurate (Fig. 9b), so that the fitting area contributed by the capacitance (Fig. 9a) has overflow parts. Fig. 9c is a comparison diagram of the capacitance contribution ratio and the diffusion-controlled process ratio calculated according to the area ratio of each part in Fig. 9a. As scan rate increases from 5 mV s −1 to 100 mV s −1 , the capacitive process contributes 58.3%, 64.41%, 70.81%, 78.78% and 88.65% of the total capacitance while the ratio of diffusioncontrolled process is gradually decreased. This might be the mismatch between the transfer rate of electrons in the external circuit and the diffusion rate of ions in the electrolyte. In total, surface-limiting capacitance contributes the most, that is, the capacitive process in PCG-5 plays a dominant role.
In general, the specific capacitance of PANI/MWCNT/ RGO composites is slightly higher than that of the PANI/ RGO composites, and that of pure RGO is the lowest. In contrast, the rate performance of each composite is not much different. Compared with the materials reported in literature, 27,36-39,55-57 PCG-5 electrode still exhibits good electrochemical properties (Table II), showing its large research and application potentials.

Conclusions
In this work, RGO, PANI/RGO (PG-i, i = 1-5) binary composites, MCG and PANI/MWCNT/RGO (PCG-i, i = 1-6) ternary composites were prepared by a simple hydrothermal method. Analysis shows that the prepared samples have abundant pore structure and large specific surface area. The use of CC as a substrate can make full use of the hydrogel sample. MWCNT can slightly increase the specific capacitance of these materials, and PANI and MWCNT also have a competitive effect in the samples. The capacitive process of PCG-5 plays a dominant role. Among them, the as-prepared PCG-5 electrode possesses excellent comprehensive electrochemical performance. It shows a considerable specific capacitance (478 F g −1 at 1 A g −1 ), an excellent rate capability (63.56%, from 1 A g −1 to 20 A g −1 , and 55.33%, from 1 A g −1 to 50 A g −1 ) and an extraordinary cycling stability (64% after 3500 cycles, at 2 A g −1 ) at positive potential window of −0.6 V to 0.4 V. In short, PANI/MWCNT/RGO ternary composites have great research potential, and especially PCG-5 can be considered as a promising electrode material for supercapacitors.