Novel Q-Carbon Anodes for Sodium-Ion Batteries

Featured Application

This work can be applied to sodium-ion batteries, especially for energy storage applications.

Abstract

The lack of a standard anode for sodium-ion batteries (SIBs) has greatly hindered their applications. Herein, we show that a novel phase of carbon, namely Q-carbon, is an effective anode material for sodium-ion batteries. The Q-carbon, which is a metastable phase of carbon consisting of about 80% sp³- and 20% sp²-bonded carbon, is synthesized by nonequilibrium pulsed laser annealing and arc-discharge methods. Two types of Q-carbons, Q1 and Q2, were evaluated as anode material for SIBs. Q1 had a slow quench and was used as the control, whereas Q2 was Q-carbon with a rapid quenching. Q1 exhibits a high initial columbic efficiency of 81% and a low-capacity retention of less than 60%, whereas Q2 has a low initial columbic efficiency of 58% and a high-capacity retention of 81%. Q2 exhibits a stable capacity of 168 mAh·g⁻¹ at a cycling rate of C/3 (124 mA·g⁻¹), which is comparable to other hard carbon anodes reported in the literature. This unique synthesis method opens a pathway for the further tuning of Q-carbon with higher trapping/charging of Na⁺ ions in improved SIBs.

1. Introduction

The generation of CO₂ due to human activities since the industrial revolution has led to an existential crisis for humanity, and transitioning away from fossil fuels is becoming the need of the hour [1,2,3,4]. Using renewable sources of energy can mitigate the issue of CO₂ generation. However, these intermittent sources of energy require efficient storage, which is a hot topic of research today [5]. The use of electric vehicles (EVs) is another attractive means of mitigating the climate crisis, as they represent zero-emission propulsion systems [6]. Lithium-ion batteries (LIBs) have become the prime choice as an energy storage means for EVs and electrical grid storage. LIBs are made of critical materials such as lithium, cobalt, nickel, and graphite, which are expensive and found in sensitive regions of the world [7]. Sodium-ion batteries (SIBs) have recently become prominent in the battery ecosystem due to their low cost and use of earth-abundant materials such as sodium, manganese, aluminum, iron, and Prussian blue (iron cyanide) [8]. SIBs have a few other advantages in terms of their wider range of operating temperature, faster charging rate, longer life span, and safety considerations. They can be also transported at 0 V due to use of copper-free current collectors and have similar form factors as LIBs [9]. Initially, SIBs were considered only for grid storage applications, but recently a few automotive and battery manufacturing companies have announced that they will be using SIBs in EVs. The performance of SIBs is appropriate for two-wheelers and other lower-range vehicles which are available in the market. For premium vehicles, the strategic use of SIBs along with LIBs has been proposed, where the energy density requirements will be met by LIBs and the high power requirements by SIBs with lower energy densities and longer life spans [10,11]. There are a multitude of hard carbons derived from the carbonization of organic biomass which have shown some potential for use as anodes for sodium-ion batteries; however, their varying microstructures exhibit diverse electrochemical behaviors, making practical applications very challenging. Soft carbons possess more graphitic crystallinity and higher electronic conductivity, which lead to superior rate performance. However, soft carbon still suffers from a lower specific capacity and poor low initial coulombic efficiency (ICE). Diamond-like carbon (DLC) films with a comparatively higher sp³ content have been used as anodes for lithium-ion batteries in the literature [12,13]. Therefore, the search for a low-cost alternative to current carbon anodes for SIBs is an ongoing topic of research. Sarkar et al. [14] and Shao et al. [15] have conducted detailed reviews on the performance of different hard and soft carbons.

Table 1. Comparison of discharge capacities of different carbon from the literature.

Carbon Type	Discharge Capacity [mAh·g−1]	Current Density [mA·g−1]	Ref.
Commercial hard carbon	270	20	[16]
Cellulose	300	37.2	[17]
Coal tar	270	100	[18]
Peanut shells	193	250	[19]
Hard–soft composite derived from biomass and oil waste	282	30	[20]
Sucrose	286	30	[21]
Coconut oil	277	100	[22]
Rice husk	372	25	[23]
Epoxy phenol novolac resin	480.3	50	[24]
Cotton	315	30	[25]

lists the performance of a few selected hard carbons.

In this paper, we focus on quenched carbon (Q-carbon), which is a new allotrope of carbon which provides an alternative anode material for SIBs with higher initial coulombic efficiency (ICE), power, and stability. Q-carbon consists of randomly packed tetrahedra with about 80% sp³ and 20% sp² bonding. Q-carbon was synthesized originally by pulsed laser melting and quenching from an undercooled state. More recently, we have synthesized Q-carbon via plasma-enhanced chemical vapor deposition (PECVD) and arc discharge methods, which are suitable for wafer-scale integration of thin-film heterostructures and bulk scale-up processing, respectively. The Q-phases of carbon possess many extraordinary properties, including a higher hardness than diamond, room-temperature ferromagnetism, enhanced field emission, and record BCS superconductivity upon doping with boron [1]. In this paper, we compared two carbons, namely Q1 and Q2, where Q1, the control for this study, is an amorphous nanocarbon that has a slower cooling and Q2 is Q-carbon with rapid undercooling. Q-carbon has not previously been applied as an anode for sodium-ion batteries and its unique synthesis method and high-sp³ carbon content make it an interesting candidate. The diamond tetrahedra structure of the carbon may have a different storage mechanism for sodium. Various carbons with high sp³ content have been theoretically proposed as anode materials in the literature [26,27,28]; herein, we for the first time we test the hypothesis experimentally and report the findings.

2. Materials and Methods

Q1 and Q2 are mixtures of Q-carbon (nanoscale) and graphite (microscale) powders, one formed by pulsed laser deposition (PLD) and another by cathodic arc deposition. The pulsed laser deposition for Q1 uses a graphite sputtering target and vacuum arc discharge; Q2 uses a graphite electrode. In the PLD system, the target was a graphite target formed from fine graphite powders. There was a slow and constant flow of argon gas. A KrF excimer laser (248 nm, 25 ns pulse duration) was focused to a laser density of 3.5 J·cm⁻² on the carbon target. After ablation, carbon atoms were generated by ablation and were gathered on a metal collector at a water-cooled metal surface in the end of a tube. In the cathodic arc deposition system, the carbon plasma with carbon liquid droplets was generated by a vacuum arc discharge on a carbon cathode with an arc current of 200 A, arc duration of 5 ms, and arc repetition of 1 Hz. The ion current density on the substrate was 10–50 mA·cm⁻². The substrate was biased to negative 2.5 kV with a pulse duration of about one microsecond. The nanoscale liquid droplets resulted in the formation of Q-carbon powder of 10–20 nm, whereas microscale liquid droplets led to the formation of graphite on the substrate. This nanoscale Q-carbon on graphite sheets was ground to produce a mixture of nanoscale Q-carbon and microscale graphite. The composite powders were vacuum dried at 110 °C overnight before being used.

Powder X-ray diffraction (XRD) patterns were collected on a Rigaku SmartLab X-ray diffractometer (Tokyo, Japan) with Cu-Kα radiation and a wavelength of 1.54 Å at an operation voltage of 40 kV and a current of 25 mA. The N₂ adsorption isotherms were measured at 77 K with a Micromeritics 3 Flex instrument (Norcross, GA, USA). The samples were degassed at 120 °C for 10 h before the measurements. The Brunauer–Emmett–Teller (BET) surface area, t-plot pore volume, and Barrett–Joyner–Halenda (BJH) desorption mesopore volume were calculated using MicroActive V6.00 software. Dynamic light scattering measurements were performed using a Zetasizer Ultra by Malvern Panalytical Ltd. (Malvern, United Kingdom) with a 633 nm He-Ne laser operating at 10 mW. Raman spectra were collected with a WITec confocal Raman microscope (Ulm, Germany) system using an excitation wavelength of 532 nm and an 1800 I·mm⁻¹ grating. A FEI Verios 460L scanning electron microscope (SEM) (Hillsboro, OR, USA) operated at 3 kV was used to characterize the surface morphologies of the samples. XPS data were collected with a Thermo-Fisher K-alpha XPS (Waltham, MA, USA) with a monochromatic Al-K_α, a 1486.6 eV source, 400 mm spot, and an argon ion flood gun. Precautions were taken to avoid any air and moisture exposure to the samples. An XPS vacuum chamber in an argon atmosphere was used to handle all samples. All samples were washed with anhydrous DMC bath and dried overnight under vacuum (in the antechamber of a glovebox) before being placed in the XPS vacuum chamber. Q-carbon powder was mixed with C45 (carbon black: CB) and PVdF in NMP in a weight ratio of 80:10:10 (Q-carbon: CB: PVdF) to form a homogenized slurry, followed by casting onto an Al foil using a doctor blade. The electrode sheet was first dried under UV light until the solvent (NMP) evaporated, followed by drying in a vacuum oven at 110 °C overnight. The electrode sheet was punched into circular discs with an area of 1.27 cm². The electrodes had an active material loading of approximately 2 mg·cm⁻². Half-cell coin cells were assembled in an argon-filled glovebox with Na metal foil as the negative electrode, a Q-carbon disc as the positive electrode, and a Celgard 2320 as the separator. The electrolyte used is 1.2 M NaPF6 dissolved in a 3:7 weight ratio of ethylene carbonate (EC) and ethyl methyl carbonate (EMC). The Q-carbon half cells were first cycled at a current rate of C/10 for three cycles, followed by cycling at C/3 between the voltage range of 0 and 3.0 V. Cyclic voltammetry (CV) was performed on a Biologic VSP (Seyssinet-Pariset, France) instrument in the voltage range of 0.01–3 V at a scan rate of 0.05 mV·s⁻¹. Electrochemical impedance spectrum (EIS) analysis was performed on all cells before and after cycling with the Biologic VSP instrument with an AC signal amplitude of 10 mV in a frequency range from 200 kHz to 10 mHz.

3. Results and Discussion

Figure S1a depicts the XRD patterns of the Q1 and Q2 carbon powders. The Q2 powder shows three broad 2θ peaks at ~10.5°, ~24.5°, and 44°, while the Q1 powder shows the latter two broad 2θ peaks at ~24.5° and 44°. The 10.5° peaks correspond to (001) graphene oxide, which is formed because of higher oxygen concentrations in the environment during the processing of the Q2 powder. The 24.5° peak corresponds to a graphitic (002) XRD peak and is shifted lower by 1.3° in the Q1 powder. Both peaks are much lower than the characteristic 26.7° of (002) graphite, suggesting a defective structure for both carbons, with Q1 having higher defect densities. The full width at half maximum (FWHM) values are related to crystallite sizes of graphite and were found to be 6.81 and 6.51 nm, respectively. The d₀₀₂ was calculated to be 3.629 Å and 3.801 Å for the Q1 and Q2 carbon, respectively. The interlayer spacing d₀₀₂ is an important measure of carbon’s microstructure and hence will significantly impact sodium-ion transport [29]. Figure S1b,c show the Raman spectra of the Q1 and Q2 carbon powders, respectively. Gaussian peak fitting is used for deconvolution of the peaks. In general, amorphous carbons have three peaks in the range 1000–1800 cm⁻¹; the first peak, D, at around 1320 cm⁻¹, the second peak, G, at around 1485 cm⁻¹, and the third peak, T, at around 1060 cm⁻¹. The D peak, G peak, and T peak correspond to the breathing mode of the sp² bonds due to the presence of defects in the structure, stretching of the sp² bonds present in the rings or chains, and vibrational mode of sp³-bonded carbon. The deconvoluted peaks ascribe I_d/I_g values of 4.37 and 2.067 for the Q1 and Q2 carbons, respectively. Generally, a higher I_d/I_g ratio means more sp² nature, indicating high disorder and a low graphitization degree. The current work shows the Raman results in Figure S1b,c, which are indicative of a higher sp³ fraction in the Q2 phase (I_d/I_g = 2.067) than the Q1 phase (I_d/I_g = 4.37) because of the lower I_d/I_g ratios in Q2 than those in Q1. The obtained Raman I_d/I_g of Q2 matches with the Raman spectra of 80% sp³-bonded Q-carbon in previous papers [1,2]. The Q-carbon phase of ~80% sp³ was confirmed in previous works by EELS, XPS, and Raman [1,2]. N2 adsorption–desorption isotherms, as shown in Figure S2, were collected to investigate the micropores and surface area for the Q-carbons. The Brunauer–Emmett–Teller (BET) surface area [30], Barrett–Joyner–Halenda (BJH) [31], and cumulative mesopore volume are obtained from this data. The BET surface area and micropore and mesopore volume for Q1 are 8.7 m²·g⁻¹, 0.002 cm³·g⁻¹, and 0.005 cm³·g⁻¹.

The BET surface area and micropore and mesopore volume for Q2 are 82.24 m²·g⁻¹, 0.0228 cm³·g⁻¹, and 0.0225 cm³·g⁻¹. Q1 has a low surface area and low mesopore volume, whereas Q2 has a higher surface area and mesopore volume. Particle size distribution was also measured using a laser diffractometer, as shown in Figure S3. The particle size for Q1 was 74.3 nm, whereas Q2 showed a wider bimodal distribution of particle sizes, concentrated at 146.1 nm and 488.7 nm. A summary of all structural properties of the Q1 and Q2 carbon are given in

Table 2. Structural parameters of Q1 and Q2 carbon. Pore structure characterization and BET surface area for Q1 and Q2 carbon.

Sample	d002 (Å)	Lc002 (nm)	Id/Ig	Micropore Volume (cm3·g−1)	Mesopore Volume (cm3·g−1)	BET Surface Area (m2·g−1)	d (nm)
Q1	3.629	6.81	4.37	0.002	0.005	8.7	74.3
Q2	3.801	6.51	2.067	0.0228	0.0225	82.24	146.1	488.7

. The electrochemical performance of Q-carbons was evaluated using sodium half cells. Figure 1 shows the discharge (Na intercalation) and charge (Na deintercalation) capacities and coulombic efficiencies of the Q1 and Q2 carbons cycled at a current density of 124 mA·g⁻¹. Q1 exhibits initial discharge and charge capacities of 344.5 mAh·g⁻¹ and 279.8 mAh·g⁻¹, respectively, with a columbic efficiency of 81%. After the 4th cycle, the coulombic efficiency increases to above 99%, with the reversible capacity gradually decreasing. The capacity retention is 63% (6 cells; 152.07 ± 14 mAh·g⁻¹) at the end of 200 cycles. Q2 provides an initial columbic efficiency of 58% with an initial discharge capacity of 427.5 mAh·g⁻¹ and charge capacity of 247.98 mAh·g⁻¹. After eight cycles, the coulombic efficiency increases to above 99% while the capacity remains stable. The capacity retention was found to be 74% (6 cells: 166.22 ± 4 mAh·g⁻¹) at the end of 200 cycles.

The cycling was performed on five cells of each type and all cells showed similar performance. The discharge capacity for Q1 and Q2 was 152.078 ± 15 mAh·g⁻¹ and 164.18 ± 9 mAh·g⁻¹. CV and rate experiments were also performed two times to confirm performance. The ICE and final capacity of Q1 and Q2 cells are reported in Table S1. Figure 1a,b show the charge/discharge profile of the first three cycles for Q1 and Q2 carbon, respectively. Notice the larger leftward shift between the 1st and 2nd; the 3rd discharge curve is larger for Q2 than Q1 carbon, which is due to the higher surface area of Q2, which leads to extensive SEI formation and lower initial columbic efficiency. Figure 1c,d show the charge/discharge profile of the 1st, 50th, 100th, 150th, and 200th for Q1 and Q2 carbon. It is noticed that the discharge curves for the 1st, 50th, 100th, 150th, and 200th cycle of Q1 carbon keeps shifting to the left, signifying its faster capacity fade over Q2. As a comparison, the cycling profiles for Q2 are close to each other, indicating superior cycling stability. For both Q1 and Q2, the curve can be divided into three sloping regions, with the first region being 0 to 0.22 V, the next being 0.22 to 1 V, and the last being the 1 to 3 V region. The absence of a flat voltage plateau at low voltage shows the absence of a graphitic region or nano-pores in both Q1 and Q2 [30]. As noted from the Raman spectra, the slower quench from higher temperature for Q1 leads to the formation of larger particles with low surface area. In addition, slower undercooling might also provide a longer time for closure of the micropores, resulting in lower porosity. Both the lower porosity and surface area suppress irreversible capacity, leading to the formation of thinner solid-electrolyte interface (SEI) layers on the electrode surface and a higher ICE [30].

As a comparison, the rapid quench of Q2 leads to a higher porosity, high surface area, large irreversible capacity loss and a low ICE. The difference of cycling stability of Q1 and Q2 can be ascribed to their difference in I_d/I_g ratio, which is 4.37 for Q1 and 2.067 for Q2. Compared with a graphite-like structure, the higher ratio of the disordered carbon in Q1 is much easier for sodium intercalation/deintercalation, resulting in a higher capacity, although with a lower surface area [17,31,32]. It seems that although Q1, with a higher amount of disordered carbon, is much easier for sodium intercalation/deintercalation with a high ICE due to its low surface area, it has a high capacity but is less stable with cycling, resulting in a continual loss of capacity with cycling, probably due to the continual built up of the SEI layer. On the other hand, Q2, with a higher amount of graphite and a higher surface area, has a low ICE and low capacity, but its cycling stability is much better, probably due to the formation of a better SEI layer. To investigate the cause of the cycling difference between Q1 and Q2, cyclic voltammetry (CV) was performed at a scan rate of 0.05 mV·s⁻¹ in the range of 0 to 3 V. Figure 2a,b shows the first three CV curves of the Q1 and Q2 carbon. For both Q1 and Q2, there is an irreversible reduction peak between 1.0 V and ~0.2 V during the first cathodic scan, resulting from the irreversible reaction of the electrolyte and the formation of the SEI on the electrode surface. The following two cycles overlap well, leading to improved CEs. The CV curves show a pair of strong redox peaks near 0 V which are associated with the intercalation/de-intercalation of sodium-ion into/out of the graphitic layers or nanopores. The overlapping wide peaks between 0.25 and 1.75 V are associated with the surface adsorption/desorption of sodium ions. It is noted that Q1 has a better symmetric redox peak than Q2 and the current densities of Q1 are much higher those of Q2, which is consistent with the higher capacity of Q1 shown in Figure 1. As mentioned earlier, the higher ratio of disordered carbon in Q1 compared to Q2 is beneficial for sodium intercalation/deintercalation, which is further validated by the rate capability test shown in Figure 2c,d. Q1 has the highest capacity at the lowest current density of C/10 (37.2 mA·g⁻¹). Both Q1 and Q2 show a consistent capacity drop with an increasing current rate, as depicted in

Table 3. Average capacity of Q1 and Q2 carbon at different current densities. Both Q1 and Q2 show good rate performance but Q1 outperforms Q2.

C Rate	Current Density (mA·g−1)	Q1: Average Capacity (mAh·g−1)	Q2: Average Capacity (mAh·g−1)
C/10	37.2	293.25	215.72
C/5	74.4	277.93	201.15
C/3	124	247.84	171.85
1C	372	198.46	120.08
2C	744	137.66	60.54
5C	1860	55.18	20.46
10C	3720	5.21	4.38
C/10	37.2	282.59	216.97

, with Q1 and Q2 decreasing from 293.25 mAh·g⁻¹ and 215.72 mAh·g⁻¹ at 37.2 mA·g⁻¹ to 5.21 mAh·g⁻¹ and 4.38 mAh·g⁻¹ at 3720 mA·g⁻¹, respectively. From the data in both Figure 1 and Figure 2, it is confirmed that Q1 indeed outperforms Q2 under all current densities.

Figure 3a,b shows the C1 spectra with four signature peaks corresponding to moieties present in the SEI: (a) carbonate (CO₃) moieties with a binding energy (BE) of 289.5 eV, O-C=O moieties at BE of 288 eV, C-O species at BE of 286.2 eV, and C species at BE of 284.0 eV, which is usually associated with hydrocarbon surface contamination in XPS spectra [33] and the presence of defects in the graphitic structure; Estrade-Szwarckopf et al. [34] claim this could be sp³ sites. The higher carbonate peak in Q1 carbon (7.1%) than in Q2 (5.3%) indicates that the SEI layer is thicker in Q1 than in Q2, which is consistent with the continual decrease of capacity with cycling shown in Figure 1. Figure 3c,d show the F1s spectra for Q1 and Q2 carbon; it has two signature peaks for trace salt NaPF_x at BE of 686.3 eV and NaF at BE of 683.2 eV. Q1 shows a higher NaPF_x salt content of 6.5% on the surface of the SEI compared to Q2 with only 2.8%, whereas Q2 is the opposite, showing a high NaF (5.3%) content on the surface compared to the 3.2% in Q1. The combined composition of NaF and NaPFx on Q1 (9.7%) is higher than that in Q2 (8.1%), suggesting that the SEI is thicker in Q1 than in Q2, consistent with the carbonate data. Other spectra such as O 1s, P 2p, and Na 2s have been fitted and quantified and are given in Table S2. The full XPS spectra are shown in Figure S4. The SEM images of both Q1 and Q2 under two different magnifications are shown in Figure S5.

NaF is formed by the degradation of NaPF₆ salt onto the surface of the SEI. Li et al. [35] claims that a SEI with a homogenous morphology contributes to the long term cycling stability of the electrode. A thin and stable SEI leads to better long cycling stability of the batteries. Based on the cycling data and XPS results, we can conclude that Q1 has a thicker SEI than Q2, which is supported by the following electrochemical impedance spectroscopy (EIS) data. EIS was performed on the Q1 and Q2 carbon before and after 200 cycles at a frequency ranging from 10⁶–10⁻² Hz. As shown in Figure S6, the Nyquist plot for the pristine Q1 and Q2 carbon collected at OCV shows a semicircle at mid/high frequencies, which is mainly due to the surface film formed at the sodium anode electrode. The Nyquist plot after cycling for the Q1 and Q2 carbon changes drastically, with an extra semicircle at mid frequencies due to the charge-transfer resistance of the electrode−electrolyte interfaces (R_ct) at the carbon electrode. The R_ct for Q1 is much higher than that for Q2, confirming the formation of a thicker SEI in Q1, which is consistent with the cycling data.

4. Conclusions

The use of Q-carbon is an interesting pathway for making sodium-ion battery anodes with a high ICE and high capacity and stability. The higher amount of disordered carbon in Q1 with a low surface area is beneficial for fast sodium intercalation/deintercalation but lacks the necessary cycling stability, whereas the lower amount of disordered carbon in Q2 with a high surface area leads to a low ICE and low capacity but with much improved cycling stability, which all can be traced back to the thinner and stable SEI layers. Further tuning of process parameters can be performed to find the optimum particle size and distribution, surface area, and porosity. Our initial hypothesis was that we would observe a different storage mechanism due to the diamond tetrahedra structure of the carbon. However, the disordered structure and micro- and meso-pores in the carbon dominate the sodium storage mechanism. This could be due to the high thermodynamic barrier for sodium bonding with the diamond tetrahedra structure and needs to be explored further.