Potassium chloride (KCl), Expandable graphite (EG, 80 mesh), and potassium permanganate (KMnO₄) were obtained from Tianjin Damao Chemical Reagent Factory.

Thermally-exfoliated graphite (TE-G) was synthesized according to our previous work (Liu et al., 2021). Specifically, EG was heated at 500°C for 100 min under N₂ to obtain thermally-exfoliated graphene (TE-G).

In a typical synthesis, 1.0 g KMnO₄ was put into deionized water (80 ml) and stirred for 30 min to produce a uniform solution. Meanwhile, different amounts of TE-G powders were put into the KMnO₄ solution and stirred for 30 min, and then, the mixture was put into a stainless-steel autoclave. The hydrothermal reaction was conducted at 180°C for 15 h. The production was filtered, rinsed repeatedly by deionized water, and dried at 60°C for 12 h to obtain MnO₂/G composites. The redox reaction equation of C and KMnO₄ can be described as: 4MnO₄ ⁻ + 3C + H₂O → 4MnO₂ + CO₃ ²⁻ + 2HCO₃ ⁻. According to the equation, the theoretical mass ratio of KMnO4 and C can be calculated as 1/17.7. Therefore, to change the MnO₂ fraction in the MnO₂/G composite, the KMnO₄/TE-G mass ratio was set as 5, 10, and 20, and the resulting composites were coded as MnO₂/G-5, MnO₂/G-10, and MnO₂/G-20. In addition, the hydrothermal reaction of TE-G in deionized water and in KCl solution was carried out under the same condition, and the resulting samples were coded as TE-G-H₂O and TE-G-KCl, respectively. The rGO was used to prepare MnO_2/rGO composites. The ratio of KMnO₄/rGO was kept the same as the ratio of KMnO₄/TE-G composites, and the sample was named MnO₂/rGO-5, MnO₂/rGO-10, and MnO₂/rGO-20.

The field-emission scanning electron microscopy (FE-SEM; SU8010) and transmission electron microscopy (TEM; JEM-2100) were used to observe the morphologies of samples. The crystallographic feature was performed by X-ray diffraction (XRD; D8-Advance) with Cu Kα radiation source. X-ray photoelectron spectra (XPS) were recorded by using a Thermo Scientific K-Alpha XPS spectrometer. The working voltage was 12 kV, and the X-Ray source was Al Kα. Pore size distribution and the specific surface area were tested by using the SSA-7000 device, according to the BJH model and BET method.

A three-electrode system was used to test the electrochemical performance of samples in an electrolyte of 6 M KOH. The poly (vinylidene fluoride)/acetylene black/active materials were weighed at the ratio of 5:10:85, and dissolved in N-methyl-2-pyrrolidone (NMP) to prepare a slurry. Foam nickel (1 × 1 cm²) was used to support the slurry and served as the working electrode. The platinum sheet was acted as the counter electrode, and the saturated calomel electrode (SCE) was used as the reference electrode. A CHI 660E electrochemical workstation (Shanghai Chenhua Co. Ltd.) was used to test electrochemical impedance spectroscopy (EIS), galvanostatic charge/discharge curves (GCD), cycling stability, and cyclic voltammetry (CV) curves. Asymmetric supercapacitor (ASCs) devices were assembled by using TE-G as the negative electrode and MnO₂/G composite as a positive electrode with 6 M KOH electrolyte. The separator was glass fiber filter paper. In the ASCs device, Formula I (Wang et al., 2018): R = m ⁺/m⁻=(C⁻×ΔV⁻/(C⁺×ΔV⁺) can be applied to obtain the ratio of positive/negative electrode material. Formula II (Brousse et al., 2007; Hong et al., 2021): C=I×Δt/ΔV, was employed in calculating the specific capacitance (C) in a three-electrode system. Formula III (Brousse et al., 2007): Cs = 4C/M, Formula IV (Li et al., 2021): E = 0.5C (ΔV)²/3.6 and Formula V (Brousse et al., 2007): P = E/Δt can be applied to obtain specific capacitance (Cs), the energy density (E) and power density (P) of the ASCs, respectively.

Figure 1 exhibits the preparation process of flower-like MnO₂/G microspheres. Firstly, under the presence of N₂, the expandable graphite (EG) was heated at 500°C to prepare TE-G. Under a high temperature, the intercalation agent in EG expands and violently decomposes, resulting in a large amount of gas spilling and forming micropores, mesopores, and macropores. As shown in Figure 1, TE-G shows an accordion structure with a thick lamella. Moreover, abundant cavity structures can be observed on TE-G. Secondly, under a hydrothermal process, KMnO₄ reacts with C to generate MnO₂, in which, each single-layer of graphene in TE-G reacts with KMnO₄ and is then wrapped by abundant MnO₂ nanosheets. The loading of MnO₂ thick nanosheets on graphene layers leads to the exfoliation of TE-G. From the inset SEM image, after the redox reaction, graphene nanosheets were wrapped by MnO₂ nanosheet arrays in different directions to produce flower-like microspheres. Compared with rGO, the consumption of graphene nanosheets and the generation of thick MnO₂ layers lead to the delamination of TE-G. In order to confirm the advantage of MnO₂/G composite, corresponding MnO₂/rGO composites were prepared, and the microstructure and electrochemical performance were investigated.

The morphologies of TE-G and different MnO₂/G composites samples were characterized by using TEM and SEM. From Figure 2A,B, pure TE-G presents an accordion structure with a large number of holes, and Figure 2C indicates the stacking structure of abundant graphene nanosheets. From these MnO₂/G samples, under a low ratio of KMnO₄/TE-G, a few graphene sheets in TE-G participate in the redox reaction with KMnO₄. Hence, a few MnO₂ nanosheets are generated on the graphene surface (Figure 2D). When the ratio of KMnO₄/TE-G increases to 10, dense MnO₂ nanosheet arrays are generated in all directions of graphene nanosheets, presenting a flower spherical structure (Figure 2F). From the high magnification SEM in Figure 2F, the resulting MnO₂ nanosheets arrays exhibit a honeycomb-like structure. The results show that the redox reaction between carbon and KMnO₄ produces MnO₂ nanosheets on graphene, which achieves the delamination of TE-G. From Figure 2G,H, there are dense MnO₂ nanosheets arrays dispersed on the graphene surface. Moreover, the TEM image also shows the connection of different MnO₂/G flower spheres, which may be resulted from the fracture of large graphene nanosheets during the high-temperature hydrothermal reaction process. As shown in Figure 2I, we can observe the diffraction fringes of MnO₂ on the graphene surface. The fringe spacing of ∼0.8 nm corresponds to the (001) facet of δ-MnO₂ (Wang J. et al., 2021). When the ratio of KMnO₄/TE-G reaches 20, excessive MnO₂ nanosheets are generated and piled up on the surface of the MnO₂/G composite (Figure 2E).

In order to prove that the exfoliating of TE-G is related to the KMnO₄-assisted hydrothermal reaction, two controls are designed by using only deionized water and KCl solution, respectively. In the absence of KMnO₄, the hydrothermal reaction cannot exfoliate the TE-G. As shown in Figure 3A, the accordion structure is kept the same as pure TE-G (Figure 2A). Under the presence of K⁺ derived from KCl, the resulting TE-G also keeps the same structure with pure TE-G in Figure 2A. Therefore, the exfoliation of TE-G is dependent on KMnO₄-assisted hydrothermal reaction, and the in-situ reaction between graphene nanosheet and KMnO₄ consumes carbon and introduces MnO₂ nanosheet arrays, which effectively exfoliate TE-G. To disclose the advantage of TE-G in preparing MnO₂/G composites, rGO was used as reductants, and resulting MnO₂/rGO composites were shown in Figure 3C,D. The MnO₂/rGO composite shows the same flower spheres as MnO₂/G composite (Figure 2D–F). The result indicates that the reaction between rGO and KMnO₄ is kept the same as the reaction between TE-G and KMnO₄, that is, the redox reaction of graphene nanosheets and KMnO₄. However, the major difference between the two reactions is the carbon precursors. TE-G has condensed graphene nanosheets with no oxygen-containing groups, while rGO is the exfoliated graphene containing oxygen-containing groups. Compared with rGO, TE-G has a low cost and high carbon content, which would consume more KMnO₄ and introduce much more MnO₂, while some rGO nanosheets are not wrapped by MnO₂ nanosheets arrays (Figure 3C), and much more MnO₂ nanosheets would enhance the electrochemical performance of MnO₂/G composites, which will be discussed further.

Elemental mapping and EDS were conducted to verify residual graphene in MnO₂/G composite. MnO₂ nanosheets (Figure 4A) keep the same with the SEM morphology (Figure 2F). The distribution of O is in accordance with the Mn (Figure 4C), which reflects the generation of MnO₂. In addition, the dispersed C signals verify the residual graphene in the MnO₂/G composite. From Figure 4E, the C content is at 35.31%, further demonstrating the residual carbon derived from graphene. To detect the precise carbon content in composite, TG curves of TE-G, MnO₂/G-5, MnO₂/G-10, and MnO₂/G-20 are given in Figure 4F. When the temperature is higher than 600°C, TE-G begins to decompose, and no residual carbon remains at 800°C. Compared with TE-G, the residual fractions of three composites are 72.5, 75.5, and 81.0% at 800°C in air. Based on the principle in Ref. (Wang J. et al., 2021), the final product of MnO₂/G composite is Mn₂O₃ at 800°C. According to the same Mn content, we can calculate the fraction of MnO₂, that is, 80.0, 83.2, and 89.3%, respectively. The residual carbon fractions in MnO₂/G-5, MnO₂/G-10, and MnO₂/G-20 are 20.0, 16.8, and 10.7%, respectively. Therefore, the TG result affirms the incomplete reaction of carbon (TE-G), and residual graphene nanosheet still exists in the MnO₂/G composite.

Fourier transform infra-red (FTIR) was provided in Supplementary Figure S1, the peak at ∼3,425 cm⁻¹ is attributed to the O-H vibration of GO or rGO. The peaks of ∼1,633 cm⁻¹ and 1,313 cm⁻¹ correspond to the stretching and bending vibration of C-O, respectively. Compared with GO or rGO, there are a few oxygen-containing groups in TE-G, which is conducive to the redox reaction between TE-G and KMnO₄, and a high fraction of carbon would consume a large amount of KMnO₄ and generate many more MnO₂ nanosheets. To verify the high fraction of MnO₂ in MnO₂/G composites, the TG curves of MnO₂/rGO-5 and MnO₂/rGO-20 were measured to obtain the content of MnO₂ in MnO₂/rGO composites. As given in Supplementary Figure S2, the residual fractions of MnO₂/rGO-5 and MnO₂/rGO-20 are 57.7 and 67.0%, respectively. Therefore, the fraction of MnO₂ can be calculated as 63.6 and 73.8%, respectively. The result shows that the MnO₂ content of MnO₂/rGO is much lower than that of the corresponding MnO₂/G composite. The reason can be ascribed to the low C fraction in rGO, resulting in fewer MnO₂ nanosheets.

In order to analyze the crystal structure of samples, XRD testing was performed. As given in Figure 5A sharp diffraction peak at 26.4° is attributed to the (002) crystal plane of TE-G (Thommes and Cychosz, 2014). According to the Bragg equation: 2dsinθ = nλ, the layer spacing d is calculated as 0.34 nm. After reacted with KMnO₄, four peaks can be observed at 12.2°, 24.7°, 36.6°, and 65.6°, these peaks correspond to the (001), (002), (100), and (110) facets of δ-MnO₂ (JCPDS # 80–1098) (Wei et al., 2012; Zhu et al., 2017). When the amount of KMnO₄ increased, the (002) peak of carbon (TE-G) at 26.4° disappears, which is assigned to the loading of MnO₂ thick nanosheets on graphene layers leads to the exfoliation of TE-G. This phenomenon indicates the consumption of TE-G and results in a low fraction of carbon in MnO₂/G composites. Compared with MnO₂/G composite, hydrothermally-treated TE-G samples under deionized water or KCl both show a sharp diffraction peak at 26.4° (as shown in Supplementary Figure S3), which confirms that the TE-G cannot be exfoliated by H₂O or KCl under hydrothermal reaction. Therefore, the exfoliation of TE-G is dependent on KMnO₄. In addition, the chemical bonds and valance state of TE-G and MnO₂/G-10 samples were characterized by XPS. From the general spectra in Figure 5B. The TE-G spectrum shows the peaks of C and O elements. After reacting with KMnO₄, the peak of C weakens, and the peaks of O and Mn elements are stronger obviously, due to the generation of MnO₂ and the consumption of TE-G nanosheets. Figure 5C–E shows the magnified C, O, and Mn spectrum. From Figure 5C, the magnified C 1s spectrum can be divided into two peaks at 284.3 and 285.9 eV, which are attributed to the C-C/C=C bond and C=O bond (Yang et al., 2020), respectively. The high-resolution O 1s can be convoluted into two main peaks at 532.0 and 529.5 eV (Figure 5D) corresponding to the bond of C-O-Mn generated between graphene and MnO₂, and the bond of Mn-O-Mn in MnO₂ (Yang and Park, 2018; Yang et al., 2020). From the magnified Mn 2p spectrum (Figure 5E), the 2p orbital of Mn has two major peaks at 642.1 and 653.8 eV, corresponding to Mn 2p_3/2 and Mn 2p_1/2. The distance of the two peaks is around 11.7 eV, which reflects the +4 valence of the Mn element (Yang and Park, 2018; Li et al., 2020). Hence, the MnO₂ in the composite is further proved by XPS.

Figure 6 shows the N₂ adsorption-desorption isotherms curves and pore distribution of different samples according to the BJH model and BET method. TE-G exhibits a high adsorption capacity and a big specific surface area at low pressure. The specific surface area of TE-G is 1055.7 m² g⁻¹, and a big specific surface area facilitates the infiltration and stripping of TE-G. When increased the amount of KMnO₄, abundant MnO₂ nanosheets loading on graphene surface decreases the specific surface area. The specific surfaces of MnO₂/G-5, MnO₂/G-10, and MnO₂/G-20 are 252.3 m² g⁻¹, 76.1 m² g^−1, and 39.4 m² g⁻¹. From Figure 6B, TE-G has much more micropores and mesopores. The loading of MnO₂ on graphene decreases the fraction of micropores and mesopores. However, the macroporous structure of MnO₂/G composite would accelerate the charge transfer and ion diffusion, further improving the electrochemical performance.

A three-electrode system was used to investigate the electrochemical performance of different samples, by using 6 M KOH electrolyte. Figure 7A presents the CV curves of TE-G and different MnO₂/G composites at 20 mV s⁻¹. The CV curve of TE-G displays a quasi-rectangular shape, reflecting the EDLC characteristic of TE-G. When introducing MnO₂, the resulting MnO₂/G composites show two pseudocapacitive peaks of MnO₂ corresponding to the faradic redox reaction of MnO₂. The faradic redox reaction mechanism of MnO₂ is verified as the valence shift between Mn⁴⁺/Mn³⁺ and Mn³⁺/Mn²⁺ (Zhou et al., 2015; Xie et al., 2019). The redox peaks centered at ∼0.1 and ∼0.4 V (vs Hg/HgO) can be assigned to the reversible redox reaction: Mn⁴⁺ ↔ Mn³⁺ + e⁻, while the other pair of redox peaks around ∼0.3 and ∼0.6 V (vs Hg/HgO) originate from the faradic redox reactions related to Mn³⁺↔ Mn²⁺ + e⁻ (Toupin et al., 2004; Zhou et al., 2015; Xie et al., 2019), corresponding to the two faradic redox peaks in CV curve further reflecting the pseudocapacitance characteristics from MnO_2. In addition, the CV curve of the MnO₂/G-10 sample has the largest area among these samples, revealing the maximum specific capacitance. Figure 7B exhibits the GCD curves of TE-G and different MnO₂/G samples. TE-G shows a linear symmetrical triangle, reflecting a typical EDLCs feature related to the adsorption and desorption of ions. When increased the amount of KMnO₄, the pseudocapacitive feature can be verified by the shape of GCD curves. The MnO₂/G-10 composite exhibits the longest discharge time of 250.0 s, much longer than that of pure TE-G (53.8s). On the basis of the equation of SC = I·Δt/(mV), the specific capacitance would be obtained. From Figure 7C, TE-G has a specific capacitance of 107.6 F g⁻¹ at 1 A g⁻¹. When hybridizing with MnO₂, MnO₂/G composites show high specific capacitances. Among these composites, the MnO₂/G-10 sample has the maximum specific capacitance of 500 F g⁻¹ at 1 A g⁻¹. Even operated at 10 A g⁻¹, the capacitance is 314 F g⁻¹, which is assigned to a large number of MnO₂ nanosheets with high capacitance content loaded to the surface of graphene. In comparison with MnO₂/G-10, the MnO₂/G-20 composite has a capacitance of 158 F g⁻¹, because of the stacked MnO₂ aggregation on graphene (Figure 2E). The MnO₂ aggregations obstruct the fast transfer of charges/ions, further decreasing the capacitance. Therefore, the MnO₂/G-20 composite exhibit a lower specific capacitance and poor electrochemical performance.

Figure.7D shows the EIS plots of different samples. Each EIS curve consists of an oblique line in the low-frequency range and a hemisphere in the high-frequency range. The Warburg impedance (W1) can be reflected by an oblique line, which reflects the diffusive resistance of the electrode in the electrolyte. The intercept in the X-axis and the diameter of the hemisphere reflect the internal resistance (R1) and charge transfer resistance (R2), respectively. ZView software was used to obtain the fitting curves (solid line) in Figure.7D. The fitting data were listed in Supplementary Table S1. The R1 values of TE-G, MnO₂/G-5, MnO₂/G-10, and MnO₂/G-20 are 0.484, 0.480, 0.213, and 0.217 Ω, respectively. In addition, the R2 values are 0.341, 0.669, 0.332 and 0.379 Ω, respectively. Therefore, the MnO₂/G-10 composite exhibits the minimum value of R1 and R2 among these samples, which indicates the minimum internal resistance and charge transfer resistance. The reason can be explained as the residual graphene in composite enhances the electronic conductivity. Moreover, hierarchical porous flower spheres of MnO₂ promote the fast transfer of charges/ions, which facilitate the pseudocapacitive reaction of MnO₂ in the electrolyte. Unfortunately, abundant MnO₂ aggregated clusters impede the rapid transfer of charges/ions, increase the internal resistance, which leads to the poor electrochemical performance of KMnO₄/G-20 composite.

To further verify the performance advantage of MnO₂/G composite, the electrochemical performance of MnO₂/rGO composites are given in Supplementary Figure S4. Both CV curves and GCD curves of different MnO₂/rGO composites show the pseudocapacitive feature of MnO_2, the area of MnO₂/rGO composites enclosed by the CV curve is much smaller than that of the MnO₂/G-10 composite. In addition, the maximum discharge time of MnO₂/rGO-20 is 64.7 s, the specific capacitance can be calculated as 129.4 F g⁻¹, much lower than that of the MnO₂/G composite. The reason is the less carbon fraction in rGO limits the redox reaction with KMnO₄, resulting in less MnO₂ nanosheets loading on rGO (in Supplementary Figure S2). Therefore, TE-G shows an obvious performance advantage to high-cost rGO.

The cycling stability of TE-G, MnO₂/G-10, and MnO₂/rGO-20 composite was tested at a current density of 5 A g⁻¹. As given in Figure 8, the specific capacitance of TE-G increases and then decreases during the first 500 cycles, which is assigned to poor wettability between TE-G and the electrolyte. As TE-G only contains a small amount of oxygen-containing groups, the wettability between TE-G and the electrolyte is poor. With the progress of the charge-discharge cycle, the wettability between TE-G and the electrolyte is improved, and the specific capacitance gradually increases. However, due to the limitation of the material itself, the specific capacitance content of TE-G decreases gradually with the increase of the cycle numbers. The specific capacitance of TE-G declines from the original 87.0 F g⁻¹ to 78.6 F g⁻¹ after 5,000 cycles. The capacitance retention rate is 90.4%, for the EDLC feature. The capacitance of MnO₂/rGO composites decreases from an initial 108.6 F g⁻¹–94.1 F g⁻¹, and the capacitance retention rate is 86.6%, which was attributed to the lower content of MnO₂ and more graphene lamellar residues (as shown in Supplementary Figure S2). In comparison with MnO₂/rGO, the MnO₂/G-10 composite has a low capacitance retention rate of 85.3%, and the capacitance decreases to 307.0 F g⁻¹ from 340.1 F g⁻¹. The low capacitance retention of the MnO₂/G-10 sample is attributed to a high fraction of MnO₂ nanosheets in the composite. However, considering the high specific capacitance, MnO₂/G-10 composite still presents outstanding cycling stability, which is attributed to the residual graphene layer remaining in the flower spherical structure.

To testify the outstanding performance of MnO₂/G composite with a flower spherical structure, we list the capacitance and long-term cycling performance of reported MnO₂/graphene in Table 1. Considering the difference in testing conditions, the MnO₂/C electrode material has the largest specific capacitance of 480.3 F g⁻¹ (Jeong et al., 2021). The maximum capacitance retention is 99.4% for 5,000 cycles (Vimuna et al., 2020). Although our MnO₂/G composite (MnO₂/G-10) has a low capacitance retention rate of 85.3%. Particularly, the specific capacitance of our MnO₂/G composite is 500 F g⁻¹, much higher than reported samples. Therefore, the MnO₂/G composite with flower spheres structure displays an excellent electrochemical performance, which can be ascribed to two aspects. First of all, a large amount of MnO₂ nanosheets arrays loading on graphene constructs a homogeneous hierarchical porous structure, which promotes the transport of electrons and ions, and reduces the charge transfer resistance. Moreover, the special microstructure facilitates the interface contact between MnO₂ nanosheets and electrolyte and releases a high specific capacitance. Secondly, TE-G is composed of stacking graphene layers, which facilitates the redox reaction between KMnO₄ and C, resulting in a high fraction of MnO₂ in composite, which increases the pseudocapacitance. Moreover, the residual graphene layer in composite improves the conductivity of electrode material and decreases the internal resistance, which enables an outstanding rate capability and cycling performance.

In addition, we assembled an asymmetric supercapacitor (ASC) with TE-G negative electrode and MnO₂/G-10 positive electrode. From Figure 9A, the CV curves show quasi-rectangular shapes. With an increase of scan rate, the area of the CV curve increases, with a shape of quasi-rectangular, further indicating the EDLC feature. The ASC can be operated stably under a broad voltage window of 0–1.0 V (Figure 9B). From Figure 9C, the longest discharge time reaches 201.6 s, corresponding to the maximum specific capacitance of 100.8 F g⁻¹ at 0.5 A g⁻¹. The capacitance reduces to 78.8 F g⁻¹ at a large current density of 5 A g⁻¹, indicating an excellent rate capability. Figure 9E provides the energy density (E) and power density (P) at different current densities. The energy density is 14.0 Wh kg⁻¹ at the power density of 250.0 W kg⁻¹. With an increase of power density, the energy density drops to 10.94 Wh kg⁻¹ (2500 W kg⁻¹), further reflecting the excellent power/energy combination. Figure 9F shows the cycling stability of ASC. When measured at 5 A g⁻¹, the capacitance retention rate is 98.4% after 5,000 cycles, reflecting the superior cycling stability. Therefore, flower-like MnO₂/G microspheres exhibit outstanding performance in ASC.