Co9S8 nanoparticle-decorated carbon nanofibers as high-performance supercapacitor electrodes

This work reported Co9S8 nanoparticle-decorated carbon nanofibers (CNF) as a supercapacitor electrode. By using a mild ion-exchange method, the cobalt oxide-based precursor nanoparticles were transformed to Co9S8 nanoparticles in a microwave hydrothermal process, and these nanoparticles were decorated onto a carbon nanofiber backbone. The composition of the nanofibers can be readily tuned by varying the Co acetate content in the precursor. The porous carbon nanofibers offered a fast electron transfer pathway while the well dispersed Co9S8 nanoparticles acted as the redox center for energy storage. As a result, high specific capacitance of 718 F g−1 at 1 A g−1 can be achieved with optimized Co9S8 loading. The assembled asymmetric supercapacitor with Co9S8/CNF as the cathode showed a high energy density of 23.8 W h kg−1 at a power density of 0.75 kW kg−1 and good cycling stability (16.9% loss over 10 000 cycles).


Introduction
In recent years, metal suldes have attracted lots of interest for energy storage devices, 1-8 since they have a richer surface redox capability compared with their oxide counterparts. 9,10 This ensures metal suldes have higher activities in energy storage and conversion reactions. Among the various metal suldes, Co 9 S 8 has proved itself as a promising material for supercapacitor electrode applications due to its high theoretical capacitance. [11][12][13] Different types of cobalt sulde nanostructures have been explored as supercapacitor electrodes, including nanospheres, 14,15 nanocages, 16 nanoowers, [17][18][19] and so on. 20,21 Extraordinarily high specic capacitance of over 1600 F g À1 has also been reported in vapor grown Co 9 S 8 lms. 12 However, similar to metal oxides, metal suldes suffer from possible aggregation and poor electrical conductivity, which prevents full utilization of their capacitance. 22 It is then reasonable to incorporate Co 9 S 8 phases into carbon nanostructures to construct the composite, in which the carbon acted as the support to offer higher electronic conductivity and porous structure, while the better utilization of the Co 9 S 8 can be accomplished. 23,24 For instance, the CoS x /carbon core-shell nanospheres were decorated on electrospun carbon nanobers (CNF), and specic capacitance of 497 F g À1 at 0.5 A g À1 has been reported. 25 The enhanced performance was attributed to the synergistic effect from the enhanced conductivity and high specic capacitance of CoS x . Co 9 S 8 nanoakes/graphene composite was also prepared. Due to the better conductivity of graphene phase, high specic capacitance of 808 F g À1 at 5 mV s À1 was achieved, which was more than twice to that of pristine Co 9 S 8 . 13 In this work, ne Co 9 S 8 nanoparticles were decorated on electrospun carbon nanobers to form the Co 9 S 8 /CNF composite nanobers. Such Co 9 S 8 nanoparticles were obtained by mild ion-exchange synthesis route and were uniformly dispersed in CNF, which acted as the porous conductive matrix. This microstructure promoted both the electronic and ionic conductivity of the electrode and maintained the intrinsic activity from metal sulde. As a result, high specic capacitance (as high as 718 F g À1 for the composite) has been obtained. The assembled asymmetric capacitor also showed high energy density and good cycle stability (83.1% remained over 10 000 cycles).

Materials synthesis
The cobalt acetate/PAN (polyacrylonitrile) precursor composite nanobers were prepared by electrospinning method in the rst step. The electrospinning solution was prepared as following. Briey, 0.8 g of PAN (MW z 150 000) was dissolved in 10 mL of N,N-dimethylformamide (DMF) under stirring. 26 Aer that, different amounts of cobalt acetates were added into the above solution. The concentrations of Co 2+ were set to 0.125 M, 0.25 M and 0.5 M, respectively. The mixture was then stirred at 120 C for 1 h until a red and sticky solution was obtained. The above solution was then loaded into plastic syringe and electrospinning experiment was conducted. A high voltage of 18 kV was applied between the needle and the sample collector (distance 15 cm). The typical feeding rate was 1 mL h À1 .
The obtained cobalt acetate/PAN composite nanobers were then dried at 80 C for 12 h in oven, and then stabilized in air at 250 C for 1 h with a heating rate 2 C min À1 . The stabilized nanobers were then immersed into 0.1 M thioacetamide/H 2 O solution and sulfurized via a microwave hydrothermal method. The temperature and the duration were 120 C and 1 h. During the microwave hydrothermal treatment, the cobalt oxide-based phase was transferred to sulde phase. The sulfurized nano-bers were nally heated at 600 C for 3 h under nitrogen to give Co 9 S 8 /CNF samples. The composite nanobers were denoted as Co 9 S 8 /CNF-1, Co 9 S 8 /CNF-2, Co 9 S 8 /CNF-3 based on the concentration of Co 2+ (0.125, 0.25 and 0.5 M). In addition, control experiment was conducted to prepare Co 3 O 4 /CNF-3 composite without sulfurization process.

Characterization
The morphology of the samples was observed by scanning electron microscope (SEM, Hitachi-S4800) and transmission electron microscope (TEM, JEM1200EX). The crystalline phase analysis was performed by X-ray diffraction (XRD-6000, Shimadzu) using Cu Ka radiation (l ¼ 0.15 nm). Thermogravimetric analysis (TGA, TA-Q500) was carried out to estimate the content of Co 9 S 8 in the prepared composite. The pore size distribution and specic surface area were calculated according to N 2 adsorption-desorption method using Autosorb (Quantachrome Instrument Corp). The Fourier Transform infrared spectroscopy (FTIR) was performed on Nicolet 5700 spectrophotometer. The X-ray Photoelectron Spectroscopy (XPS) characterization was performed on Escalab250Xi instrument.
All the electrochemical measurements were carried out on an electrochemical workstation (CHI660E, ShangHai ChenHua Co., Ltd, China) at room temperature. The supercapacitor tests were performed in both three-electrode and two-electrode congurations in 6 M KOH electrolyte. The working electrode was prepared by mixing 80 wt% of active materials (Co 9 S 8 /CNF), 10 wt% of Ketjen black and 10 wt% polyvinylidene uoride (PVDF) dissolved in N-methyl pyrrolidone (NMP). The slurry was subsequently casted onto Ni foam and dried at 60 C in oven overnight. In the three-electrode system, Pt and Saturated Calomel Electrode (SCE) was used as the counter and reference electrode, respectively. In the two-electrode system, the asymmetric supercapacitors were built using the as-prepared working electrode as positive electrode and active carbon nanobers (ACNF) as negative electrode.
The electrochemical performances of the supercapacitors were evaluated by characterizing their cyclic voltammetry (CV), galvanostatic charge/discharge (GCD) and electrochemical impedance spectroscopy (EIS). The voltage window was from 0 to 0.5 V vs. SCE for the three-electrode system, and 0 to 1.5 V for the asymmetric Co 9 S 8 /CNF-ACNF cell. The GCD were tested at current density ranging from 0.5 A g À1 to 20 A g À1 and CV were tested at scanning rate of 2 mV s À1 to 200 mV s À1 . The EIS was measured at open circuit with 30 mV AC potential in the frequency ranged from 100 kHz to 0.01 Hz. Cycle stability measurements were tested at 10 A g À1 current density for 10 000 cycles.
The specic capacitance (C) of the device was calculated based on the integration of the discharge curve as following The energy density (E) and power density (P) were calculated according to the following equation: where Du is the potential, v is the potential scan rate (mV s À1 ), i is the discharge current (A), t is the discharge time (s), and m is the mass (g) of the active materials on the electrode.

Results and discussion
Fig . 1 shows the XRD patterns of the Co 9 S 8 /CNF series samples.
From the patterns, one can conclude that Co 9 S 8 phase can be successfully formed aer the microwave-assisted sulfurization and carbonization process, since the diffraction peaks can be identied as standard Co 9 S 8 (JCPDF65-6801) in the patterns. The peak intensity in Co 9 S 8 /CNF-1 is rather weak, which may be due to that the content of the sulde phase was not high. However, in Co 9 S 8 /CNF-2 and Co 9 S 8 /CNF-3 samples, clear diffraction peaks can be observed, which indicates the Co 9 S 8 in the composite nanobers had good crystallinity. In contrast, the XRD pattern of Co 3 O 4 /CNF-3 sample matched with Co 3 O 4 phase as shown in Fig. S1. † From the above comparison, one may propose the formation mechanism of Co 9 S 8 phase in the composite nanobers. During the stabilization process, the cobalt acetate was decomposed into cobalt oxide by heating. The aerward microwave hydrothermal step initiated the ionexchange reaction to transform Co 3 O 4 into Co 9 S 8 , since TAA can decompose and release sulfur contained reactants. 3,[27][28][29][30][31] It is also known that the material with lower solubility (K sp ) is more thermodynamically stable than that with higher K sp value. The ionic radius of S 2À (0.18 nm) anions is also close to the oxygen (0.14 nm) anions. Thus, Co 9 S 8 nanoparticles were formed. The high temperatures in the carbonization process helped to improve the crystallinity of the cobalt sulde and Co 9 S 8 nanoparticles can form. More details about the elemental composition of the Co 9 S 8 / CNF samples were investigated by XPS. Fig. 2 gives the core level scans of Co 2p and S 2p of the Co 9 S 8 /CNF-3 sample. The high resolution Co 2p XPS spectrum ( Fig. 2(a)) shows the presence of Co 2+ and Co 3+ . The peaks at 778.2 eV and 792.9 eV were belonged to Co 3+ 2p 3/2 and 2p 1/2 , and the two peaks at 782.0 eV and 797.5 eV were assigned to Co 2+ 2p 3/2 and 2p 1/2 , respectively. 32,33 The S 2p spectrum can be divided into two main peaks and one satellite peak, as shown in Fig. 2(b). The two peaks at 161.4 eV and 162.6 eV can be attributed to the metal-sulfur bonds. 34,35 These XPS results are consistent with the XRD analysis mentioned above and suggest that Co 9 S 8 is successfully formed. The surface composition of the composite nanobers was also estimated from the XPS full spectra and the data was shown in Table S1. † It can be seen that the content of the Co 9 S 8 in the samples increased as the Co 2+ ions concentration increased in the precursor.
The formation of Co 9 S 8 was further conrmed by FTIR spectroscopy. Fig. S2 † gave the FTIR spectra taken from the Co 9 S 8 /CNF samples. The bands located at 606 and 470 cm À1 were assigned to Co 9 S 8 phase, which has been discovered in metal sulde materials. 3,36 The content of Co 9 S 8 in the composite nanobers were examined by TGA method, and the result was shown in Fig. S3. † The TGA of the pristine Co 9 S 8 was also presented in Fig. S4 † as the reference. The samples were heated from 50 C to 800 C in air. There were three stages of weight loss in the Co 9 S 8 /CNF samples during the heating process. At rst (before 250 C), there was a slight weight loss which probably was attributed to the evaporation of water remained in the sample. When the temperature raised from 250 C to 550 C, two competition process co-existed. On one side, because of the partial oxidation of Co 9 S 8 , 37 the weight of the sample will be slightly increased which can be seen from the pristine Co 9 S 8 in Fig. S4(b). † On the other side, the decomposition of carbon will cause a huge weight loss. Considering these two process, the composite nanobers showed weight loss. At the last step, the cobalt sulde converted into cobalt oxide when the temperature was above 700 C and the weight became constant. The nal remained weight of the three samples was 18.9%, 28.5% and 38.8% as shown in Fig. S3, † and from Fig. S4(b) † it is known that the weight loss was about 11.8% when the pristine Co 9 S 8 converted to Co 3 O 4 . 38 Thus, the Co 9 S 8 contents in Co 9 S 8 /CNF-1, Co 9 S 8 /CNF-2, and Co 9 S 8 /CNF-3 samples can be estimated to be $21.4%, 32.3% and 44.0%, respectively.
The morphology of the Co 9 S 8 /CNF was characterized by SEM. Fig. 3 show the corresponding images for the samples. From Fig. 3(a) (Co 9 S 8 /CNF-1, with the lowest Co 9 S 8 loading), it can be seen that the sample was composed of stacked nanobers, which is typically seen in the electrospun samples. The diameter of an individual ber was $300 nm and the length can reach to tens of microns. The stacked nanobers formed a porous matrix, which is benecial for continuous electron transfer and ion penetration. 39 When the Co 9 S 8 content increased, the porosity of the ber network can be well retained as shown in Fig. 3(b). In Fig. 3(c) and (d) (which is the magni-cation of Fig. 3(c)), when the Co 9 S 8 content was the highest, numerous nanoparticles were found to decorated on the ber surface, and the size was estimated to be 100 nm. A comparison of the Co 3 O 4 /CNF-3 can be found in Fig. S5(a), † which revealed the similar porous structure. In Fig. S5(b), † it can be seen that the surface of the ber was coarse, and small Co 3 O 4 can be identied. The above morphology analysis shows that the Co 9 S 8 nanoparticles prepared by ion-exchange method possessed the similar morphology as their oxide counterpart.
The microstructure of the Co 9 S 8 /CNF samples was further characterized by TEM and showed the similar trends. Fig. 4(a) shows the TEM image taken from Co 9 S 8 /CNF-1. Fibers with coarse surface can be observed, and the Co 9 S 8 phase can be barely identied since its content was low. In Fig. 4(b) (Co 9 S 8 / CNF-2), ne nanoparticles with darker contrast can be clearly seen. In Fig. 4(c) and (d) (Co 9 S 8 /CNF-3), one may see a large number of Co 9 S 8 nanoparticles with the size of $100 nm appeared on the surface of the carbon nanobers, which agreed well with the SEM characterization result. These nanoparticles were rmly attached on the CNF surface, and such structure prevented the aggregation of the nanoparticles. The above morphology evolution indicated that the existence of the Co 9 S 8 nanoparticles was more evident and the size became larger when the Co 9 S 8 phase increased. Fig. 5(a) shows the N 2 adsorption/desorption isothermal curves for the samples. The isotherm showed typical type IV behavior, indicating the presence of the mesopores. In addition, large N 2 uptake can be found when the relative pressure was larger than 0.9, which is the sign of the macropores. The  existence of both mesopores and the macropores are benecial for high performance supercapacitor electrode, since they can act as the ion transfer pathway and ion reservoir, respectively. 23 The calculated surface area for Co 9 S 8 /CNF-1, Co 9 S 8 /CNF-2, and Co 9 S 8 /CNF-3 samples were 13.61 m 2 g À1 , 12.20 m 2 g À1 and 12.02 m 2 g À1 , respectively. Such values are comparable to the reported CoS x /C nanosphere decorated carbon nanobers samples. 25 The corresponding pore size distribution calculated from the adsorption branch in the mesopore range can be found in Fig. 5(b). A peak at 4 nm can be found in all samples, which may be due to the decomposition of the polymer in the carbonization process, and the Co 9 S 8 /CNF-3 possessed the narrow pore size distribution.
The electrochemical performance of the composite nano-bers was rstly evaluated in 3-electrode congurations. Fig. 6(a) shows the corresponding CV curves of the electrodes at the scan rate of 20 mV s À1 . For comparison, the CV curves of Co 3 O 4 /CNF-3 sample was also demonstrated. From the curves for the Co 9 S 8 /CNF samples, pairs of broad redox peaks can be observed (two oxidation peaks between 0.35 to 0.45 V vs. SCE but only one broad reduction peaks between 0.25 to 0.32 V vs. SCE), 40,41 which are attributed to the multiple oxidation steps of Co 9 S 8 in alkaline electrolyte by the following equations: 41,42 Co 9 S 8 + OH À ¼ Co 9 S 8 (OH) + e À (4) Co 9 S 8 (OH) + OH À ¼ Co 9 S 8 OH + e À Specically, two well dened reduction peaks can be found in Co 9 S 8 /CNF-3 sample, while one reduction peak was found in other samples. Such phenomenon has been reported in the previous Co 9 S 8 supercapacitor electrode. This may be attributed to that the Co 9 S 8 /CNF-3 sample possessed the highest Co 9 S 8 content. The above observation conrms the charge storage in Co 9 S 8 /CNF electrodes is mainly accomplished by the fast-redox reactions in Co 9 S8 phase. The peak current density can be used to evaluate the specic capacitance of the electrode, which followed the order of Co 3 O 4 /CNF-3 < Co 9 S 8 /CNF-1 < Co 9 S 8 /CNF-2 < Co 9 S 8 /CNF-3, and the detail CV curves measured at different scan rates for Co 9 S 8 /CNF-3 can be found in the ESI Fig. S6(a). † This indicated Co 9 S 8 /CNF-3 had the largest specic capacitance, which was proved by the galvanotactic method. Fig. 6(b) demonstrated the constant current charge-discharge curves of the corresponding electrodes at 1 A g À1 . All the curves had the shape of quasi-triangle with plateaus, which matched with their CV curves in which clear redox peaks were found. The longest discharge time was found in Co 9 S 8 /CNF-3 electrode and decreased in the same order as the CV results. The specic capacitance at 1 A g À1 was 718 F g À1 , 568 F g À1 and 382 F g À1 for Co 9 S 8 /CNF-3, Co 9 S 8 /CNF-2, and Co 9 S 8 /CNF-1 electrodes, respectively. Such values are much larger than conventional electrical double layer capacitance obtained in pristine porous carbon nanobers (typically less than 200 F g À1 ) and are superior to the reported CoS x -CNF composite nanostructures. For examples, CoS x @carbon core-shell nanospheres decorated carbon nanobers electrode showed 497 F g À1 capacitance at 0.5 A g À1 current density. 25 In addition, the performance of Co 9 S 8 /CNF-3 electrodes was also better than most of the reported pristine cobalt sulde nanostructures, such as hollow Co 9 S 8 nanospheres 14 (306 F g À1 at 0.1 A g À1 ), cobalt sulde nanotubes 43 (285 F g À1 at 0.5 A g À1 ), ower-like cobalt sulde hierarchitectures 17 (674 F g À1 at 3 A g À1 ), and 3D ower-like Co 9 S 8 hierarchical architectures 19 (522 F g À1 at 0.5 A g À1 ). Considering the Co 9 S 8 content was 44% in the composite, the utilization of the active site was much higher. This proves that the Co 9 S 8 nanoparticles decorated CNF matrix can effectively store the charges. It should be noted here that Co 3 O 4 /CNF-3 electrode had the lowest specic capacitance of 154 F g À1 at 1 A g À1 . This highlights the importance of the ion-exchange process, which turned the oxide to the more electrochemically   active sulde. Fig. 6(c) shows the specic capacitance of the electrodes measured at different current densities, which reects the rate capability. Even at 20 A g À1 , Co 9 S 8 /CNF-3 electrode can still maintain specic capacitance of 292 F g À1 . Electrothermal impedance spectra were further used to explain the origin of the difference in the electrochemical performance. From Fig. 6(d), all the curves showed straight line in low frequency range, which is the characteristic of capacitive behavior. The curve of Co 9 S 8 /CNF-3 electrodes showed the largest slope, suggesting a lower diffusion resistance. Such results agreed the discussion above that it has the highest capacitance. Overall, the high specic capacitance of Co 9 S 8 / CNF-3 electrode can be attributed to (1) the high loading of the active Co 9 S 8 phase in the composite, which contributed higher pseudocapacitance and (2) the small size of the Co 9 S 8 nanoparticles and the highly porous nature of the CNF backbone, which facilitated both the electron and ion transferring in the electrode.
Furthermore, an asymmetric supercapacitor device was assembled by employing Co 9 S 8 /CNF-3 as the positive electrode and activated carbon nanobers (ACNF) as the negative electrode. Such activated carbon nanobers were prepared according to previous literature. 26 The corresponding CV curves at different scan rate can be found in Fig. 7(a). The curve had a near rectangular shape at all scan rates, in the meantime small redox peaks can be also observed. This indicates that both the electrical double layer capacitance (EDLC) and the pseudocapacitance contributed to the charge storage. From CV one can also see that the device can operate at 1.5 V potential window, since no obvious water decomposition current was observed. The galvanotactic charge-discharge curves can be found in Fig. 7(b). The curve had a symmetric quasi-triangle shape, which is the sign of high coulomb efficiency. The specic capacitance of the device was 82 F g À1 at 0.5 A g À1 , and gradually dropped to 48 F g À1 at 5 A g À1 . Even at 10 A g À1 current density, the specic capacitance of the device can still reach to 34 F g À1 , accounting the whole mass of both positive and the negative electrode. The Ragone plot, which reects the energy and the power density of the device, can be found in Fig. 7(c). The specic energy density was 23.8 W h kg À1 when the power density was 0.75 kW kg À1 . The energy density retained at 10.7 W h kg À1 when the power density increased to 7.5 kW kg À1 . At the highest power density of 15 kW kg À1 , the energy density can maintain 5.5 W h kg À1 . These data are superior to some of the cobalt sulde-based devices, including the hierarchical Co 9 S 8 nanostructures grown on nickel foam//AC capacitor (21.03 W h kg À1 at 0.13 kW kg À1 and 13.33 W h kg À1 at 1.3 kW kg À1 ) and MOF-derived Co 9 S 8 -carbon//AC capacitor (14.85 W h kg À1 at 0.68 kW kg À1 and 6.63 W h kg À1 at 6.82 kW kg À1 ). 44,45 The stability of the electrodes was also evaluated by charging the electrode with 10 A g À1 current density for more than 10 000 cycles, and the capacitance change was demonstrated in Fig. 7(d). The capacitance retention was 83.1% aer the testing, suggesting much better stability compared to those in literature (55-65% retaining aer 500 cycles for selfassembled Co 9 S 8 and 83.3% retaining aer 5000 cycles for networked ultralong CoS 1.097 nanotube). 46,47

Conclusions
In summary, Co 9 S 8 nanoparticles decorated carbon nanobers composite electrodes were prepared. The porous CNF matrix offers electron and ion transfer pathway, while the Co 9 S 8 nanoparticles prepared by ion-exchange method owned small size and were uniformly distributed on the CNF surface. Such structure ensured the better utilization of the active Co 9 S 8 phase to delivery high pseudocapacitance. The optimized Co 9 S 8 /CNF-3 electrode demonstrated high specic capacitance of 718 F g À1 . Furthermore, high energy density and power density (energy density of 23.8 W h kg À1 at the power density of 0.75 kW kg À1 , energy density of 5.5 W h kg À1 at power density of 15 kW kg À1 ) as well as good cycling stability (83.1% performance fading over 10 000 cycles) were demonstrated in the asymmetric capacitor based on such Co 9 S 8 /CNF-3.

Conflicts of interest
There are no conicts to declare.