1 Introduction

Lithium-ion batteries (LIBs) are used in various electronic devices such as mobile phones and electric vehicles owing to their high energy density, high voltage, and long life. Because of technological innovations in these electrical devices, the demand for LIBs has been growing rapidly around the world. However, because the energy density of commercially available LIBs is already approaching the theoretical value, the development of new power storage devices that can replace LIBs crucial [1]. Therefore, secondary batteries utilizing multivalent cations (Mg2+ and Ca2+) instead of Li+ have been attracting attention with regard to the development of post LIBs [2,3,4,5,6,7]. Among these multivalent cation batteries, calcium-ion batteries (CIBs) utilizing Ca2+ as charge carriers have great potential as alternatives to LIBs [8,9,10]. Because CIBs use multivalent cations, they can achieve twice the capacity of LIBs. Additionally, because the standard electrode potential of Ca2+ is apporoximately 0.5 V lower than that of Mg2+ (− 2.87 and − 2.37 V with respect to the Standard hydrogen electrodes of Ca and Mg, respectively), CIBs can achieve high operating voltages [1]. Moreover, because Ca2+ is the fifth most abundant element in the Earth’s crust, it is advantageous in terms of sustainability [11]. Therefore, CIBs have more attractive features when compared with LIBs and other multivalent ion batteries. However, some problems exist concerning the development of CIBs for practical use. [12, 13]. Particularly, the use of Ca2+ metal anodes is not appropriate because the surface film formed on the Ca2+ metal prevents the reversible plating/stripping overpotentials of Ca2+ in conventional organic electrolytes. [12,13,14]. Furthermore, because multivalent cations, such as Ca2+, exhibit stronger Lewis acidity than Li+, they form strong solvation structures with the solvent molecules in electrolytes. Thus, high desolvation energy is required for intercalation into the host material, affording poor kinetics of charge intercalation [12, 15]. Therefore, optimization of Ca2+ host materials is desired to address these limitations.

To eliminate the instability of plating/stripping overpotentials of the Ca2+ metal electrode, a type of CIBs utilizing the Ca2+ intercalation/deintercalation reaction into the host material was proposed [16,17,18,19,20]. Graphite is one of the candidates as an anode material of CIBs because it can accommodate various guest ions into its layers and form a graphite intercalation compound (GIC). Graphite is generally used as an anode material for commercialized LIBs because of its inexpensiveness, safety, and capacity. In LIBs, Li+ is stored in the graphite layers to form a Li–GIC [21]. The intercalation of Ca2+ into graphite layers was previously limited to reports based on chemical methods only [22,23,24]. Recently, however, some reports suggested that the electrochemical intercalation of Ca2+ into graphite was possible [25]. These reports indicated that the potential carbonaceous materials could be used as anode materials for CIBs, and a charge/discharge system like that in LIBs could be constructed.

Therefore, instead of graphite, Marimo nano carbon (MNC) was investigated as a new anode material for CIBs. MNC is a spherical carbon material that comprises high-density carbon nanofilaments (CNFs) or carbon nanotubes arranged radially [26]. Furthermore, synthesized CNFs have remarkable properties: their internal structure (such as cup-stacked, coin-stacked, and hollow) can be controlled by the type of the catalyst metal [27]. Additionally, in cup-stacked and coin-stacked CNFs, the edge planes of the graphene sheets are exposed. Thus, they might be applied in a wide range of fields. As reported previously the use of coin-stacked CNFs with a graphite structure as an anode material for LIBs yielded higher capacity than that of a graphite anode [28]. Thus, the MNC anode may intercalate more Ca2+ between the graphene layers of CNF than the graphite anode. Consequently, higher charge/discharge capacities than graphite may be obtained. In this study, MNC with coin-stacked CNFs was synthesized as the anode material for CIBs. Additionally, heat-treated MNC (Ht-MNC) was prepared to increase the crystallinity of CNFs, and the performances of MNC and Ht-MNC as anode materials for CIBs were evaluated.

2 Experimental

Oxidized diamond (Diamond Innovations Co., Ltd.) was used as a catalyst support. It has oxygen-containing functional groups arranged regularly, therefore nano metal particles can be highly dispersed on it [29]. Therefore, an oxidized-diamond-supported Pd catalyst (Pd loading: 5 wt%) was prepared for CNF synthesis using the metal nanocolloidal method [30]. Oxidized diamond was added to deionized water, and the resultant dispersion was stirred for 30 min under ultrasonic treatment. PdCl2 as a precursor and citric acid as a dispersing agent were added, and the resultant solution was stirred for 30 min under ultrasonic treatment. Moreover, NaBH4 as a reducing agent was added to the solution. After stirring for 30 min under ultrasonic treatment, the supernatant was removed from the solution via centrifugation and dried at 60 °C. The supported catalysts were calcined at 450 °C for 5 h in N2 prior to the reaction.

MNC was synthesized via chemical vapor deposition (CVD) method using a horizontal rotating fluidized bed reactor [31]. We took 200 mg of the oxidized-diamond-supported Pd catalyst and uniformly distributed it on a quartz boat. After inserting the quartz boat into the center of the quartz tube, the reactor temperature was elevated to 600 °C under a N2 atmosphere. Subsequently, C2H4 was introduced at 30 mL min−1 and 600 °C. The reaction time was 1 h. The heat treatments of MNC were conducted at 2500 °C for 1 h under vacuum.

Field emission scanning electron microscopy (FE-SEM; S-4800, Hitachi, Ltd.) was conducted to observe the surface structure of MNC, and transmission electron microscopy (TEM; JEM-2010K, JEOL Ltd.) was conducted to observe the internal structure of the synthesized CNFs. Moreover, X-ray diffractometry (XRD; SmartLab, Rigaku) was conducted to calculate the interlayer distance before and after discharging from the peak position corresponding to the (002) plane of the graphite structure as follows:

$$d=\frac{\lambda }{2\mathrm{sin}{\theta }_{\mathrm{c}}},$$
(1)

where \(d\) denotes the phase-spacing, λ is the (Cu-Kα) X-ray wavelength, and \({\theta }_{\mathrm{c}}\) is the diffraction angle of graphite. The X-ray wavelength was 0.15406 nm, the voltage was 45 kV, and the current was 200 mA. The divergence slit width was 1/2°, receiving slit width was 0.15 mm, and scattering slit width was 1/2°. The range of the measurement was 5°—90°, and the scan speed was 50° min−1. The measurement was conducted after three cycles of discharging and charging to obtain representative results. Laser Raman spectroscopy (NRS-3100, JASCO) was conducted to analyze the degree of graphitization of the synthesized CNFs and the solvation structure. The excitation wavelength of the laser light source was 532 nm, and the laser power was 100 mW. The measurement was conducted with an exposure time of 5 s and comprised eight integrations. The wavenumber of the Raman shift was calibrated using Si and polypropylene standards. The surface area of the MNC was evaluated via the Brunauer–Emmett–Teller (BET) analysis of N2 adsorption isotherms at 77 K, and it was measured using a specific surface area analyzer (BELSORP-mini, Nippon Bel, Japan). Fourier transform infrared spectroscopy (FTIR; FT/IR-4200, JASCO) was conducted to analyze the solution structure. The measurement range was 500–4000 cm−1 at a resolution of 4 cm−1 and the integration number was 400. The solution structure was evaluated using the liquid-film method using a CaF2 window board.

MNC and Ht-MNC were used as working electrode materials, and spherical natural graphite (CGB-10; Nippon Graphite Industries, Ltd.) was used as a reference material. The MNC and graphite electrodes were mixtures of 95 wt% respective active materials and 5 wt% polyvinylidene fluoride (Sigma-Aldrich) Subsequently N-methylpyrrolidone was added to the mixtures to make a slurry. The slurry was coated onto Cu foil using the doctor blade method. After drying, the resulting electrode was punched to a diameter of 12 mm. Activated carbon (AC; YP-50F, Kurarey, Co., Ltd.) was used as the counter/reference electrode material instead of the Ca metal [25]. The AC electrodes comprised 80 wt% of YP-50F as the active material, 10 wt% of carbon black (TOKABLACK; TOKAI CARBON Co., Ltd.) as the conductivity promoter, and 10 wt% of Teflon resin as the binder. The mixture was rolled to a thickness of 0.1 mm and punched to a diameter of 12 mm. The MNC, Ht-MNC, graphite and AC electrodes were vacuum-dried at 80 °C overnight before being used. Cellulose fiber (TF4050) was used as the separator. An electrolytic solution was prepared by dissolving 0.5 M Ca(ClO4)2 and 0.5 M Ca(TFSI)2, in γ-butyrolactone (GBL). These salts were dehydrated via lyophilization overnight. These chemicals were purchased from FUJIFILM Wako Pure Chemical Ltd. In this GBL-based electrolyte, a reversible redox reaction has been reported, and Raman spectroscopy has revealed that the reaction is due to Ca2+ intercalation/de-intercalation into the graphite layer. Therefore, GBL-based electrolytes were also used in this study [32].

Reference-electrode validation tests were performed via cyclic voltammetry (CV) measurements of a three-electrode cell. A Pt electrode (1 cm × 1 cm) was used as the working electrode, and the AC electrodes were employed as the counter and reference electrodes. Ferrocene (Fc) was added to the electrolyte at a concentration of 0.05 M, and 300 μL of electrolytic solution was added to each cell. The CV measurements were conducted in a scanning potential range from − 0.6 to − 1.0 V at a scan rate of 1 mV s−1. To analyze the electrochemical behaviors of the MNC and Ht-MNC electrodes in organic electrolytes, charge/discharge measurements were conducted using an airtight flat cell (Fig. 1). The MNC, Ht-MNC, and graphite electrodes were used as the working electrodes, and the AC electrodes were employed as the counter electrodes. The charge/discharge measurements were conducted in a scanning potential range from − 2.4 to − 0 V and constant current of 0.1 C. All the cells were assembled in a N2-filled glove box, and all the electrochemical measurements were recorded using an electrochemical workstation (HZ-7000; Hokuto Denko, Japan).

Fig. 1
figure 1

Flat cell used during electrochemical measurement

Full size image

3 Results and discussion

Figure 2 shows the SEM images of MNC and Ht-MNC. CNFs are densely generated and many pores exist between the fibers (Fig. 1a, a′), Moreover many CNFs of Ht-MNC thicker than the CNFs of MNC are observed. (Fig. 1b, b′) The mean filament diameters of CNFs in the MNC and Ht-MNC were 24.1 and 70.8 nm, respectively.

Fig. 2
figure 2

Field emission scanning electron microscopy images of a and a′ Marimo nano carbon (MNC), b and b′ heat-treated MNC (Ht-MNC)

Full size image

Figure 3 shows the TEM images of CNFs of the MNC and Ht-MNC, As can be clearly seen from Fig. 3a, the synthesized CNFs have a coin-stacked structure, in which graphene sheets are stacked perpendicularly to the axis of the filaments. In Ht-MNC, the graphene sheets were stacked more regularly, and loops on the edge plane could be confirmed.

Fig. 3
figure 3

Transmission electron microscopy (TEM) image of a synthesized and b heat-treated carbon nanofilaments (CNFs)

Full size image

Figure 4 shows the model of MNC. MNC is a spherical carbon material that comprises (CNFs) or carbon nanotubes arranged radially. The CNFs synthesized in this study have a coin stacked structure.

Fig. 4
figure 4

Model of Marimo nano carbon (MNC)

Full size image

XRD was conducted to evaluate the structures of MNC, Ht-MNC, and CGB-10 (see Fig. 5). Compared with MNC, the peak intensity of the (002) plane of graphite considerably increased for Ht-MNC, and the FWHM of the peak decreased from 3.01° to 0.98° accordingly (see Fig. 5a). Additionally, the peak position of the (002) plane also shifted from 2θ = 25.81° for MNC to 2θ = 26.30° for Ht-MNC, and the interlayer distances were 0.345 nm and 0.339 nm, respectively, as calculated from Eq. (1). Because the CVD method synthesizes CNFs at reasonably low temperatures, the CNFs in MNC had low crystallinity. However, a graphite structure is formed when a carbon material is heat-treated at ≥ 2000 °C. Therefore, we hypothesize that the crystallinity of Ht-MNC was improved by reducing the defects of the carbon hexagonal network and developing the graphite structure via heat treatment. Additionally, the peaks derived from the Pd catalyst metal and oxidized diamond of the catalyst supports were confirmed for MNC, however, these peaks were not observed for Ht-MNC. Because the surface layer of diamond begins to graphitize at 1400 °C and diamond completely transforms to graphite at 1900 °C [33], the oxidized diamond transformed to a graphite structure via heat treatment at 2500 °C. The carbon source was absorbed by CNFs owing to the graphitization of the oxidized diamonds, thereby increasing the average fiber diameter. Figure 5b) compares Ht-MNC and natural graphite. For natural graphite, peaks derived from the (002), (100), (004), and (110) planes of graphite were confirmed. Similar peaks were confirmed for Ht-MNC, indicating the development of a graphite structure [34]. Because the synthesized CNFs have a coin-stacked internal structure (Fig. 3) with reported intercalation, the MNC anode may be able to intercalate Ca2+ or solvated Ca2+ between the graphene layers of the CNFs (Fig. 4).

Fig. 5
figure 5

X-ray diffractometry (XRD) spectrum comparisons of a Marimo nano carbon (MNC) and heat-treated MNC (Ht-MNC), and b Ht-MNC and natural graphite

Full size image

Table 1 presents the physical properties of each sample, as derived via Raman, XRD and BET analysis. From the Raman spectrum, the crystallinity was evaluated from the peak intensity ratio (IG/ID) of the G-band (1580 cm−1) due to the graphite structure and the D-band (1350 cm−1) due to the defect. For MNC, the D-band appeared strong, whereas for Ht-MNC, the peak intensity of the D-band decreased and the IG/ID value increased. However, the IG/ID value for Ht-MNC was not as large as that of natural graphite. Therefore, a correlation was observed between the evaluation results of crystallinity via Raman spectroscopy and XRD. Because the development of the graphite structure of Ht-MNC was confirmed via XRD, the exposure of the edge plane is considered to be the cause of a lower IG/ID value than that of natural graphite. Conversely, Ht-MNC had a more graphitic structure than MNC and more edge planes than natural graphite, therefore, it might be highly reactive with Ca2+. The interlayer distance also narrowed in the order of increasing graphitization degree. The BET analysis revealed that MNC has larger specific surface area and mesopores than natural graphite. However, Ht-MNC had smaller specific surface area and mesopores than MNC because of the large mean filament diameter.

Table 1 Characterization of each sample
Full size table

FTIR spectroscopy was conducted to evaluate the solvation structure in GBL-based electrolytes containing Ca(ClO4)2 and Ca(TFSI)2, (Fig. 6). As shown in Fig. 6, the GBL solvent exhibited the peak of C=O double bond at 1750 cm−1. For the electrolyte containing Ca (ClO4)2, a peak due to the perchlorate ion (ClO4−) at 1100 cm−1 was observed. For the electrolyte containing Ca (TFSI)2, peaks due to the TFSI− anion appeared at 1060, 1141, 1228, and 1335 cm−1 corresponding to the S–O stretching vibrations, SO2 symmetric stretching vibrations, CF3 stretching vibrations, and SO2 asymmetric stretching vibrations, respectively. In FTIR spectroscopy, when an interaction occurs with a material, the peak of the IR spectrum shifts or a new peak appears [35]. Each Ca salt was dissolved in the GBL solvent, but the C=O double bond peak did not change considerably. However, a new peak appeared at 1354 cm−1 for electrolytes containing Ca (TFSI)2; this peak could be derived from Ca-GBL.

Fig. 6
figure 6

Fourier transform infrared (FTIR) spectrum for each electrolyte

Full size image

Similarly, Raman spectroscopy was conducted to evaluate the strength of interaction between the solvent and each Ca salt (see Fig. 7). The GBL solvent exhibited the peak of C–O single bond at 930 cm−1 (Fig. 7a). Even when each Ca salt was dissolved, no new bands were observed, and the bandwidth displayed only approximately 2 cm−1 spreading. When each Ca salt was dissolved in EC- and PC-based electrolytes, new bands, derived from the solvated ions in which the solvent was coordinated to Ca2+, appeared, and the width of the bandshift was approximately in the range of 10–15 cm−1 [15]. Because the width of this bandshift represents the strength of interaction between the ion and solvent molecules [36], this result indicates indicated that the interaction between Ca2+ and the GBL solvent was very weak. This agrees with the order of the dielectric constants of these solvents [EC: 89.8, PC: 64.9, and GBL: 39.1], the higher the dielectric constant, the more readily the solvation proceeded. For the electrolyte containing Ca (TFSI)2, however, a new peak appeared at 693 cm−1. In Fig. 7b, such a peak cannot be confirmed even though the salt concentration increases; however, in Fig. 7c, the peak intensity increases as the concentration increases. Accordingly, for the electrolyte containing 0.5 M Ca (TFSI)2, the peak at 693 cm−1 is attributed to the solvated ions i.e., Ca-GBL [37]. Therefore, Ca (TFSI)2 is highly likely to interact with the solvent; thus, the order of the strength of interaction with Ca2+ is Ca (ClO4)2 < Ca(TFSI)2.

Fig. 7
figure 7

Raman spectrum for a each electrolyte and pure γ-butyrolactone (GBL) solvent, b electrolytes containing Ca (ClO4)2, and c electrolytes containing Ca (TFSI)2 at different concentrations

Full size image

Before testing the electrochemical performances of the anode materials, the AC quasi-reference electrode (AC QRE) was calibrated using a cyclic voltammogram of 0.05 M Fc in a GBL-based electrolyte containing 0.5 M Ca (ClO4)2 and 0.5 M Ca (TFSI)2, respectively, using a three-electrode cell. The Ca metal on the counter/reference electrodes for testing battery materials can only be used under limited conditions owing to irreversible plating/stripping due to the formation of a passivation layer. Thus, the use of the AC QRE, which facilitates stable evaluation in the potential window of the electrolyte, was proposed [1, 38, 39]. For non-aqueous systems, calibrating the AC QRE by measuring the redox potential of Fc as an internal standard is recommended. Accordingly, the reference potential of Fc/Fc+ for the AC QRE was calculated as follows:

$${E}_{{\mathrm{F}}_{\mathrm{c}}/{\mathrm{F}}_{\mathrm{c}}^{+}}{=E}_{1/2}=\frac{{E}_{\mathrm{p},\mathrm{ a}}+{E}_{\mathrm{p},\mathrm{c}}}{2}$$
(2)

where \({E}_{{\mathrm{F}}_{\mathrm{c}}/{\mathrm{F}}_{\mathrm{c}}^{+}}\) denotes the redox potential of Fc, and can be obtained from the half-wave potential \(({E}_{1/2})\). \({E}_{\mathrm{p},\mathrm{ a}}\) and \({E}_{\mathrm{p},\mathrm{c}}\) denote the anodic and cathodic peak potentials, respectively. Moreover, because \({E}_{\mathrm{p},\mathrm{ a}}\) was 0.130 V and \({E}_{\mathrm{p},\mathrm{c}}\) was − 0.085 V,\({E}_{{\mathrm{F}}_{\mathrm{c}}/{\mathrm{F}}_{\mathrm{c}}^{+}}\) was 0.023 V versus AC QRE. Similarly, because \({E}_{\mathrm{p},\mathrm{ a}}\) was 0.090 V and \({E}_{\mathrm{p},\mathrm{c}}\) was 0.020 V, \({E}_{{\mathrm{F}}_{\mathrm{c}}/{\mathrm{F}}_{\mathrm{c}}^{+}}\) was 0.055 V versus AC QRE. Therefore, the \({E}_{{\mathrm{F}}_{\mathrm{c}}/{\mathrm{F}}_{\mathrm{c}}^{+}}\) for the AC QRE obtained for each electrolyte is − 0.377 V and − 0.345 V lower than the \({E}_{{\mathrm{F}}_{\mathrm{c}}/{\mathrm{F}}_{\mathrm{c}}^{+}}\) for SHE. From this result, the potentials of Ca2+/Ca versus AC QRE were calculated to be − 3.247 and − 3.215 V. Hereinafter, all the electrochemical potentials will describe the potential of AC QRE and calibrated potential of Ca2+/Ca [20].

Figure 8 shows the charge/discharge curves of the MNC, Ht-MNC, and graphite electrodes in each electrolyte. In this study, charging is defined as reduction and Ca2+ or solvated Ca2+ intercalation, and discharging is defined as oxidation and Ca2+ deintercalation [25, 40, 41]. At present, it is not clear whether intercalation of Ca2+ or solvated Ca2+ into the graphite layer.

Fig. 8
figure 8

Charge/discharge curves of the a, b natural graphite electrode, c, d Marimo nano carbon (MNC) electrode and e, f heat-treated MNC (Ht-MNC) electrode in a, c, e 0.5 M Ca (ClO4)2/γ-butyrolactone (GBL) and b, d, f 0.5 M Ca(TFSI)2/GBL

Full size image

The electrolyte containing Ca (ClO4)2 showed higher charge/discharge capacity than the electrolyte containing Ca (TFSI)2. Moreover, the charge/discharge capacity of the electrolyte containing Ca (ClO4)2 increased as the cycle progressed. Hence, we hypothesize that Ca2+ or solvated Ca2+ intercalation/deintercalation occurs more in the electrolyte containing Ca (ClO4)2. Next, MNC and Ht-MNC showed higher charge/discharge capacities than natural graphite in the case of the electrolyte containing Ca (ClO4)2. Natural graphite is spheroidized to suppress the excessive reductive decomposition of the electrolytes, whereas MNC is highly reactive with the electrolytes owing to the exposed edge planes of the CNFs. However, Ht-MNC exhibited lower charge/discharge capacity than MNC. The charge/discharge capacity of MNC includes the capacity due to the amount of adsorption caused by the specific surface area. Consequently, the charge/discharge capacity of MNC is larger than Ht-MNC. In other words, Ht-MNC’s Ca2+ or solvated Ca2+ intercalation proceeded more than in MNC. Therefore, we consider that compared with MNC, Ht-MNC has a higher ratio of intercalation capacity to total capacity. Additionally, the Coulombic efficiency was correlated with the degree of graphitization of each material. Conversely, for the electrolyte containing Ca (TFSI)2, there was no considerable difference in charge/discharge capacity among the anode materials. This may be because Ca (TFSI)2 strongly interacts with the solvent molecule, thereby making Ca2+ or solvated Ca2+ intercalation unlikely to occur.

XRD was conducted to evaluate the intercalation of Ca2+ or solvated Ca2+ between the graphene layers of each anode material (Fig. 9). Before charging, MNC and Ht-MNC had peaks at 25.81° and 26.30° corresponding to the (002) plane (Fig. 5a); for the electrolytes containing Ca (ClO4)2, the (002) peaks of MNC and Ht-MNC shifted to 19.94° and 19.91°, respectively, after charging and the interlayer distances increased to 0.445 nm and 0.446 nm, respectively. This indicates that Ca2+ or solvated Ca2+ intercalates between graphene layers. Furthermore, for Ht-MNCs, the shifted peaks slightly reverted after discharge. This is thought to be due to development of the graphite structure, which facilitates reversible charging and discharging of Ca2+ or solvated Ca2+. However, the shifted peak of MNC did not return to the original 2θ value at all, indicating that it cannot be successfully discharged. This result is consistent with the Coulomb efficiency calculated from Fig. 8.

Fig. 9
figure 9

X-ray diffractometry (XRD) spectra of Marimo nano carbon (MNC) and heat-treated MNC (Ht-MNC) before charge and after charging in 0.5 M Ca (ClO4)2/GBL

Full size image

4 Conclusion

When MNCs were heat treated, their crystallinity was enhanced and graphite structure appeared. Moreover, when Ca(ClO4)2 and Ca(TFSI)2 were used as salts, the solution structure changed in GBL-based electrolytes, and the interaction strength with Ca2+ was in the order Ca(ClO4)2 < Ca(TFSI)2. For electrolytes containing Ca (ClO4)2 and Ca (TFSI)2, MNC and Ht-MNC were confirmed to exhibit higher charge/discharge capacity and reversible redox reactivity than natural graphite. Furthermore, Ht-MNC exhibited more advanced reversible Ca2+ or solvated Ca2+ intercalation between graphene layers than MNC.