Soybean-derived hierarchical porous carbon with large sulfur loading and sulfur content for high-performance lithium–sulfur batteries

An hierarchical porous carbon nanostructure with intrinsic Oand N-dopants and an ultrahigh specific surface area of 1500 m g 1 is reported towards the goal of designing and achieving a better sulfur electrode for lithium–sulfur batteries (LSBs) that can provide both large sulfur loading and large sulfur content and are based on a facile fabrication process. This nanostructure was derived from crude soybeans in a facile pyrolysis process. Using it as a sulfur host, the S/C active composite with 80% sulfur content was made. Cells with different sulfur loadings were investigated and were found to demonstrate large capacity, high coulombic and energy efficiencies, and high cycling stability. In particular, for a sulfur loading of 5.5 mg cm 2 and a sulfur content of 80%, cells displayed a specific capacity of ca. 950 mA h g , which corresponds to an areal capacity of 5.2 mA h cm . Such a performance moves LSB technology closer to practical applications.


Introduction
To meet the insatiable demands for energy storage, new battery technologies are under intensive investigation. Considering their potential to provide an energy density of 2600 W h kg À1 , 3-5 times that of the conventional Li-ion batteries, and the low cost and environmental friendliness of sulfur, lithium-sulfur batteries (LSBs) are particularly attractive. [1][2][3][4][5][6][7] The insulating nature of sulfur and the reduction products (Li 2 S 2 and Li 2 S) prefer a catholyte chemistry, in which the reduction intermediates (lithium polysuldes) are soluble. This enables a solidliquid-solid conversion in the entire electrochemical process and therefore, overcomes the otherwise slow kinetics in a solidsolid reaction. 8 However, the soluble polysuldes may shuttle between and react on the cathode and the anode, causing a chemical shortcut in the cell 9,10 and thus giving rise to the irreversible loss of active materials, capacity fade, anode corrosion, low coulombic efficiency and self-discharge. 1,11 Such a shuttle effect, as well as other effects, prevent LSBs from achieving a large capacity with a reasonable cycling stability.
Steady progress has been made to advance LSB technology in recent years through cathode nanostructure engineering, [12][13][14][15][16] new electrolyte chemistry, 17,18 and lithium anode passivation, [19][20][21] among others. In terms of the cathode design aimed to diminish the adverse shuttle effects, the nanostructured conductive hosts, particularly carbonaceous nanomaterials, 13,22,23 have been well studied to physically prevent the soluble lithium polysuldes from migrating out of the cathode. 14, 16 More recently, chemical immobilization of poly-suldes through chemisorption 24,25 has been found to be more effective to achieve this goal. For the nonpolar carbonaceous surfaces, heteroatom nitrogen doping, 26,27 metal oxides, [28][29][30] or polymers 31 are commonly used for chemisorption of lithium polysuldes that have a polar nature. 32 Signicant progress has been made toward the practical application of LSBs. [33][34][35] As emphasized in ref. 8, a practical battery electrode must have an areal capacity of a few mA h cm À2 , and the electrode material must be cost-effective. However, a considerable number of the recently reported electrode structures is based on very complex fabrication processes; and hence, is not cost effective. Whereas, the structures based on facile processes simply cannot offer a reasonable performance. Furthermore, many reports show good performance, but use the structures with a small sulfur loading, and, in particular, with a small sulfur content in the active S/C composite. This results in a volumetric capacity, a very critical parameter for a battery, even lower for the LSBs than for the traditional Li-ion batteries. In this study, we endeavor to solve these problems by developing a facile process for fabrication of the cathode structure that offers impressive performance. Hierarchical porous carbon, intrinsically doped with nitrogen and oxygen, was prepared using crude soybeans, which inherently contain nitrogen and oxygen, as a precursor in a facile activation process involving mixing with KOH and pyrolysis in Ar. The resulting carbonaceous material displays a hierarchical porous structure with an ultrahigh specic surface area of ca. 1500 m 2 g À1 . Using it as sulfur host, the S/C composite with high sulfur content (80%) was made. Cells with different sulfur loadings were investigated, demonstrating good capacity and cycling stability. In particular, for a sulfur loading of 5.5 mg cm À2 and a sulfur content of 80%, cells displayed a specic capacity of ca. 950 mA h g À1 at 0.2C. This corresponds to an areal capacity as high as 5.2 mA h cm À2 , which is comparable to the reported performance of the state-of-the-art LSBs but prepared via a complex fabrication process. 24,36

Results and discussion
The schematic in Fig. 1a illustrates the entire process to produce the S/C composite. Crude soybeans, without being crushed into powder, were directly used as a doped carbon precursor. Soybeans mainly consist of protein (36%), carbohydrates (30%), and soybean oil (20%), with the remainder being water and ash. The protein-and carbohydrate-rich soybeans provide nitrogen and oxygen sources for the doped carbon material. Aer hydrothermally pretreating crude soybeans at 180 C for 20 hours, a black thick oil-like paste was obtained. The scanning electron microscope (SEM) image of the paste (Fig. S1 in the ESI †) reveals a morphology similar to a thick oil mixed with the irregularly-shaped larger particles. The crude soybeans were essentially disintegrated and liqueed in this process, resulting in a dispersion of carbonaceous intermediates. This pretreatment not only converts the solid soybeans into a dispersible paste for easy mixing with KOH in the subsequent step for activation, but is also critical for the creation of the C]C backbone for the nal activated carbon product. 37,38 The oil-like paste, aer being well mixed with KOH to form a slurry, was then activated at 800 C in the argon environment by a pyrolysis and corrosion process, to obtain the nal soybean-derived carbonaceous material. This material was then employed as a conductive matrix that was mixed with sulfur powder to obtain the S/C mixture with sulfur mass ratio as high as 80% in the active S/C composite.
The morphology and microstructure of the nal carbonaceous material was investigated by scanning electron microscopy (SEM) and transmission electron microscopy (TEM) to reveal its hierarchical porous structure. As shown in the SEM image in Fig. 1b, the strategy of hydrothermal pretreatment followed by pyrolysis in the presence of KOH activation agent transforms the soybean solid into a three dimensional (3D) bicontinuous porous structure. Such a bicontinuous pore morphology consists of the hollow, micrometer-or submicrometer-sized spheres that are embedded inside the structure, forming a nest, and are interconnected through the smaller pores that interpenetrate through the thin shells of spheres. This creates a hierarchical porous structure at the micro-scale. It is believed that the C]C backbone, created in the hydrothermal pretreatment process, has the tendency to fold into microspheres at higher temperatures in the pyrolysis process, while KOH-induced etching tends to disrupt the structure by forming a bicontinuous structure. 38 A similar hierarchical structure was also formed at the nanoscale, as revealed by the TEM images at different magnications in Fig. 1c and d. The hollow spheres have a diameter distribution ranging from hundreds of nanometers to tens of nanometers or even smaller sizes. The pores are interconnected in a 3D structure with the smaller pores nested inside the larger pores. The high-resolution TEM image in Fig. 1e suggests that the micropores with a diameter of less than 2 nm are present on the carbon surface.
To further quantify the obtained porous structure, the nitrogen adsorption isotherm was acquired using a micropore analyzer (Fig. 1f). The prominent nitrogen uptake at a relative pressure (P/P 0 ) below 0.01 is due to micropores, whereas the uptake at P/P 0 ¼ 0.05-0.3 is due to mesopores. The N 2 adsorption with a plateau at higher relative pressures followed by an increase at P/P 0 > 0.9 reveals the presence of large mesopores and macropores. 39 The pore size distribution was calculated from the nitrogen adsorption data (Fig. 1g), which clearly shows the presence of both micro-and meso-pores, and conrms the hierarchical porous structure of the obtained material. The cumulative pore volume plot (Fig. S2 †) shows that micropores, dened as pores with a diameter smaller than 2 nm, contribute most of the total pore volume of ca. 0.7 cm 3 g À1 . The Brunauer-Emmett-Teller (BET) specic surface area (S BET ) was found to be ca. 1500 m 2 g À1 . Such a hierarchical microporous/mesoporous structure is known to be effective for physically trapping the soluble polysuldes in the sulfur electrode, as well as for rapid Li + transportation.
The structure, elemental composition and chemical bonding in the obtained porous structure were studied. The Raman spectrum shows two typical carbon peaks at ca. 1345 cm À1 and ca. 1590 cm À1 (Fig. S3a †). The intensity of the defect-induced D peak at ca. 1345 cm À1 is stronger than that of the G peak at ca. 1590 cm À1 . This indicates the porous nature of the carbonaceous material, and is consistent with the previous report. 37 The X-ray diffraction (XRD) pattern shows a broad peak at 26.5 2q (Fig. S3b †) and, together with the Raman spectrum, conrms the characteristics of the carbonaceous material. This is in agreement with the data for doped carbon materials. 26 The surface composition analysis based on the survey X-ray photoelectron spectroscopy (XPS) spectrum ( Fig. S4 †) found that the as-prepared carbonaceous material consists of 92.5 at% C, 5.2 at% O and 2.3 at% N. The high-resolution C 1s spectrum in O at 289.1 eV. 27,40 This conrms both O and N functional groups exist in the obtained carbonaceous material. The high-resolution N 1s spectrum in Fig. 1i was deconvoluted into four peaks, corresponding to pyridinic N at 398.5 eV, pyrrolic N at 400.3 eV, quaternary N at 401.4 eV, and oxidized N at 402.6 eV. 41,42 The high-resolution O 1s spectrum in Fig. 1j was deconvoluted into three peaks, assigned to C]O (carboxyl) groups at 530.5 eV, C]O (ester, amides) groups at 532.3 eV, and COH, N-OC groups at 534.2 eV. 43 This indicates the presence of various oxygen-containing functional groups with doubly and singly bound oxygen.
A polar carbonaceous surface, obtained through molecular functionalization or heteroatom doping, is essential for chemical bonding of the otherwise soluble polysuldes onto the carbon surface to minimize the shuttle effect. N-Doping has been extensively explored for this purpose. 24,25,28,36,44 However, the role of a small amount of oxygen heteroatom in carbonaceous materials converted from cellulose has not caught enough attention, although oxygen oen exists in such carbonaceous materials. Wang et al. 45 reported that the presence of oxygen impurities, particularly those bonded to vacancies and edges, may stabilize lithium polysuldes. Our recent rst principles-based calculations also suggest that O dopants, mainly through Li-O bonds, have signicantly larger binding energy than N dopants due to the more electronegative nature of O compared to that of N and C. Furthermore, O-and N-dopants can also work synergistically to enhance the bonding of polysuldes on the carbon surface. 46 Therefore, the small amount of oxygen, present in addition to nitrogen in the soybean-derived hierarchical porous carbon, will improve stabilization of lithium polysuldes that may otherwise diffuse into the electrolyte.
To summarize, both the electron microscopy imaging and nitrogen adsorption results reveal a 3D hierarchical porous structure of the as-prepared soybean-derived carbonaceous material, which is suitable as a conductive matrix for sulfur loading and polysuldes connement. In this hierarchical porous structure, sulfur species could be trapped in the micropores and result in enhanced electrochemical activity and reduced diffusion, whereas the mesopores serve as facile Li + transportation channels and provide buffer for volume expansion. 1 In addition to the physical connement by micro-and meso-pores, chemical binding of polysuldes onto the O-and N-doped matrix surface is more effective in alleviating their shuttle effect as well as enabling the uniform distribution of sulfur species on the matrix surface during cycling.
The sulfur cathodes were prepared by drop-casting the cathode slurry into a freestanding edge-oriented multilayer graphene foam (GF) that was used as a current collector. 47 It has a mass density of less than ca. 1.5 mg cm À2 , which is much lower than the conventional aluminum current collector. Cells with Li chips to be used as anodes were assembled to study their electrochemical property and test their performance. They were investigated at different charge-discharge (C-D) current densities, ranging from 0.1 to 1C, over a potential window of 1.8-2.6 V vs. Li/Li + . The specic capacity was calculated based on sulfur loading. The highly insulating sulfur particles may introduce dead volume, whereas dissolution and re-deposition of sulfur species in the C-D process will not only cause the initial "bursting release" of polysuldes away from the cathode, but also could help to redistribute the active species more uniformly on the conducting carbon matrix in the cathode. Therefore, the initial activation process was investigated by performing electrochemical impedance spectroscopy (EIS) in the rst few C-D cycles for a cell with a sulfur loading of 2.6 mg cm À2 at a 0.2C rate. The EIS measurement was conducted at the fully charged state (S 8 state) for each cycle. As shown in Fig. 2a, there was a relatively large discharge capacity loss aer the rst cycle, from 1130 to 1110 mA h g À1 . This was probably due to the initial "bursting release" of soluble polysuldes, and the formation of solid electrolyte interphase. Aer the rst cycle, the capacity became relatively stable. Fig. 2b presents the EIS results and the tting curves, obtained using the equivalent circuit shown in the inset. The intercept of impedance curve at the Z 0 axis represents the equivalent series resistance (R) that includes electronic resistance of electrode, ionic resistance of electrolyte, and contact resistance. The diameter of the semicircle on the impedance curve indicates the charge-transfer resistance (R ct ), which represents the resistance of the electrochemical reaction on the interface of the electrolyte and electrode. The inclined line at low frequency means Warburg impedance (W), which corresponds to the diffusion of lithium ions. The tted curves clearly indicate a dramatic decrease of R ct aer the rst cycle, from 6.7 U cm 2 to 1.8 U cm 2 . This suggests that the solid and insulating sulfur redistributed uniformly on the conductive matrix during electrochemical reaction that involves multistep solid-liquid-solid conversions. CV scans at 0.1 mV s À1 were conducted on another fresh cell for the rst few cycles, as shown in Fig. 2d. All peaks on the CV curves become sharper and are stable aer the rst cycle, whereas the anodic peaks shi to a higher voltage. This is consistent with the results obtained from the C-D curves. Fig. 3 represents the results from cells with a sulfur loading of 2.0 mg cm À2 . In Fig. 3a, the initial discharge capacity at 0.1C was as high as 1330 mA h g À1 . This indicates high sulfur utilization. Aerwards, the cells displayed stable specic capacities of 1300, 1120, 870, and 710 mA h g À1 at 0.1, 0.2, 0.5, and 1C, respectively. Aer cycling at different current densities, the specic capacity did not suffer any obvious loss when the current returned to 0.1C. The C-D proles at different C-D rates acquired for the h C-D cycle, i.e., aer the proles were stable at each rate, are plotted in Fig. 3b. The discharge proles exhibit two plateaus and two slopes. As proposed in ref. 48, the high plateau at ca. 2.38 V on the discharge curve at 0.1C corresponds to the rst reduction step of sulfur to long chain polysuldes such as S 8 À and S 6 À , whereas the slope that follows is the second reduction step with a gradual reduction of polysulde chain lengths, producing mainly S 4 À . The low plateau at ca.
2.10 V corresponds to a further chain reduction to short poly-sulde species such as S 2 2À and S 2À . The precipitation of these poorly soluble and highly insulating Li 2 S 2 and Li 2 S compounds leads to electrode passivation with a gradual ending of discharge. The charge prole also exhibits two plateaus, but these plateaus are not as prominent as the plateaus for the discharge prole. The separation between charge and discharge potentials increases with current density due to polarization loss. Similar observation can be made from the CV measurement (Fig. S5 †) that shows two well-separated cathodic peaks and two closely-spaced anodic peaks, with the cathodic and anodic peaks shi due to polarization at higher scan rates. Coulombic efficiency and energy efficiency from the h C-D curves at each rate are plotted in Fig. 3c. The cells exhibited a high coulombic efficiency of ca. 99% for all the measured rates. This suggests a weak shuttle effect. Energy efficiency is a more important parameter when considering energy storage and consumption, but it is unfortunately not reported in most publications. At 0.1 and 0.2C, our cells exhibited a high energy efficiency of ca. 92%. At 0.5 and 1C, the cells had energy efficiency of 89% and 82%, respectively. We believe that the excellent performance of the sulfur electrode is due to the synergistic effect of hierarchical porous structure and heteroatom doping in the soybean-derived carbonaceous material. The hierarchical porous structure with a large specic surface area facilitates the uniform distribution of sulfur species in the carbon matrix during electrochemical reactions, favors electrolyte penetration into electrode matrix, and physically connes soluble lithium polysuldes, whereas O-and N-dopants render carbon surface strongly adsorbing the otherwise soluble lithium polysuldes and reduce the charge transfer barrier in the electrochemical reactions. The synergetic physical and chemical effects from this unique carbon material collaboratively contribute to the high performance of the sulfur electrode that contains this material in an LSB cell. The excellent stability was further investigated by the long-term cycling test. Three cells with a sulfur loading of 2.0 mg cm À2 were cycled at 0.1C for 150 cycles, 0.2C for 700 cycles, 0.5C for 800 cycles ( Fig. 3d and S6 †). These cells had the nal capacity of 1020, 640, and 460 mA h g À1 and the capacity retention of 78%, 57%, and 53%, respectively. To further demonstrate the practical potential of the asprepared soybean-derived carbonaceous material, cells with a high sulfur loading of 5.5 mg cm À2 were assembled to achieve a high areal capacity. This is vital for commercializing LSBs since the conventional lithium-ion batteries typically have an areal capacity (C A ) of 4-5 mA h cm À2 . For most cathode designs, the specic capacity will signicantly decrease at a higher sulfur loading because a larger portion of the polysuldes will diffuse away; most publications have therefore reported a low sulfur loading, typically approximately 2 mg cm À2 or less. However, our cells with a sulfur loading as high as 5.5 mg cm À2 still displayed high specic capacities of 1200, 1070, 970 and 710 mA h g À1 at 0.05, 0.1, 0.2 and 0.5C, respectively. These correspond to areal capacities of 6.6, 5.9, 5.3 and 3.9 mA h cm À2 (Fig. 4a and b), and are comparable with the areal capacity of practical lithium-ion batteries. The cell was cycled at 0.2C and 81% of its initial capacity was maintained aer 100 cycles (Fig. 4c and d). Since the stability of LSB cells depends on both the sulfur cathode and the lithium anode, the failed cell (Fig. S7 †) was disassembled for inspection to reveal the degradation and failure mechanisms of the high-loading cells. From a photograph and SEM images (Fig. S8 †), no evident change was found in the cathode; however, the lithium metal anode was severely eroded and a thick solid electrolyte interphase (SEI) layer had formed on the surface. Thus, the capacity fading was more likely caused by lithium corrosion and electrolyte decomposition and depletion, 49 rather than a considerable degradation of the sulfur cathode. An improved Li anode design would further enhance the electrochemical stability of the cells with a high sulfur loading. Although further progress is needed, we emphasize that the cathode design based on the soybean-derived hierarchical porous carbon with intrinsic heteroatom dopants shows a better overall performance than many reported nitrogen-doped carbon-based designs. 24,[50][51][52] Considering its large areal capacity, high sulfur loading and utilization, and good stability, this material represents signicant progress toward the practical application of LSBs. The excellent performance of the battery described herein, prepared using cheap materials and easy processes moves the LSBs a step closer to practical application.

Conclusion
In this study, a simple pyrolysis/activation process was employed to derive a hierarchical porous carbon nanostructure from crude soybeans. Its ultrahigh specic surface area of ca. 1500 m 2 g À1 with dominant micropores can enhance the connement of soluble polysuldes, whereas the intrinsic Oand N-dopants synergistically contribute to the chemical binding of polysuldes as well as reducing the electron transfer resistance during electrochemical conversion. As a result, sulfur electrodes with a high sulfur content, large sulfur mass loading, and excellent performance can be fabricated using this new material. Cells with such sulfur cathodes demonstrated large capacity, high coulombic and energy efficiencies, and high cycling stability. For a sulfur loading of 5.5 mg cm À2 , cells displayed a specic capacity of ca. 950 mA h g À1 at 0.2C, which corresponds to an areal capacity of 5.2 mA h cm À2 . It is envisioned that with a better lithium anode design, application of the cathode developed in this study will move the LSB technology a step closer to being practical.

Experiments
Preparation of soybean-derived activated carbon 30 g of original soybeans, without being crushed into powder, was rst sealed into a 100 mL Teon-lined autoclave, and then heated at 180 C for 20 h. To produce activated carbon, 20 g of the obtained soybean black paste was mixed with KOH in a mass ratio of 1 : 1, and then 20 mL deionized water was added to make a uniform slurry by stirring. The slurry was dried at 120 C in vacuum to obtain the uniformly mixed soybeans and KOH. This mixture was pyrolyzed at 800 C in an argon environment for 2 h with a heating rate of 2.5 K min À1 . The pyrolyzed product was washed in dilute HCl a few times, and then in deionized water repeatedly to remove ion impurities. The nal soybean-derived activated carbon product was obtained by drying the washed product at 80 C in vacuum overnight.

Material characterization
The morphology and the hierarchical porous structure of the soybean-derived activated carbon were revealed by a LEO eld emission scanning electron microscope (SEM) and a Hitachi H-9500 transmission electron microscope (TEM). The surface elemental composition and chemical bonding were analyzed using a Physical Electronics PHI 5000 VersaProbe X-ray photoelectron spectrometer (XPS) with photoelectrons excited by monochromatic Al K a radiation (1486.6 eV). The survey scans in the 0-1400 eV binding energy (BE) range were collected with a pass energy of 187.85 eV and a step of 0.8 eV. For the highresolution scans, the pass energy was 23.5 eV with a step of 0.1 eV. The C 1s, O 1s and N 1s core level spectra were tted using Physical Electronics PHI MultiPak soware with a Gaussian sum function, a Shirley background subtraction and an iterative leastsquares optimization algorithm. To correct for sample charging, the BE of the spectra was referenced to the C-C peak C 1s spectrum with BE at 284.6 eV. 24 Before the XPS measurement, the sample was dried at 60 C in vacuum overnight to remove the adsorbed water. The Raman spectrum was recorded on a Bruker Optics Senterra dispersive Raman microscope spectrometer with a spectral resolution of $3-5 cm À1 using 532 nm laser excitation. The X-ray powder diffraction (XRD) pattern was obtained using a Philips X'Pert system. The nitrogen adsorption measurement was conducted at 77 K using a Quantachrome Autosorb iQ (ASiQ) micropore analyzer. The sample was outgassed at 150 C for 18 h under high vacuum. The multi-point specic surface area (S BET ) was determined from the adsorption isotherm using the Brunauer-Emmett-Teller (BET) method. The pore size distribution was calculated from the nitrogen adsorption data using a nonlocal density functional theory N 2 /carbon equilibrium transition kernel at 77 K based on a slit-pore model.

Electrode fabrication and electrochemical measurements
The active S/C composite with sulfur loading of 80 wt% was made by mixing sulfur and the soybean-derived activated carbon. These components were mixed in a mass ratio of 4 : 1 (S : C) by manually grinding for 1 h in mortar using a pestle. The mixture was then sealed in a Teon-lined autoclave and heated at 155 C for 12 h, so that sulfur and the activated carbon could be well mixed. The cathode paste was then made by adding 20 wt% carbon black, and 10 wt% polyvinylideneuoride (PVDF) binder. The S/C composite was mixed with carbon black by grinding together, while PVDF was dissolved in N-methylpyrrolidone (NMP) solvent using magnetic stirring until the solution was uniform. Then, the ground mixture was transferred into the solution, and stirred continuously for a few hours until a uniform slurry was obtained. To fabricate the cathode, the slurry was drop-casted onto freestanding edge-oriented graphene (EOG) foam and dried at 60 C for 20 h. The EOG foam used was obtained by EOG growth inside a sacricial Ni foam. 53,54 It has a mass density of less than ca. 1.5 mg cm À2 , much lower than the commonly used Al current collector, e.g., 14.9 mm thick, 4.02 mg cm À2 . The as-prepared cathodes were assembled into coin cells in an Ar-lled Vigor glovebox where oxygen and moisture are both held below 1 ppm. Lithium chips were used as the anode. Separators were soaked with electrolyte that consisted of 1.0 M bis(triuoromethane)sulfonimide lithium salt (LiTFSI) in a mixed solvent of 1,2-dimethoxyethane (DME) and 1,3-dioxolane (DOL) (1 : 1, v/v) with 2 wt% lithium nitrate (LiNO 3 ) as an additive. The ratio of the volume (mL) of the electrolyte to the mass (mg) of sulfur was controlled at 15 : 1. The newly-assembled cells were stored for 12 h before testing, allowing electrolyte diffusion into the cathode. Cell characterization was carried out using cyclic voltammetry (CV), charge-discharge (C-D), and electrochemical impedance spectroscopy (EIS). C-D tests were conducted with a Land CT2001A battery test system at different current densities over a voltage window of 1.8-2.6 V. CV tests were executed using a Biologic SP-150 electrochemical workstation at different scan rates from 0.05 mV s À1 to 0.2 mV s À1 . EIS was measured in the frequency range from 1 MHz to 0.1 Hz with a sinusoidal signal of 10 mV at an open circuit potential of 2.6 V.