CNT Sheet Air Electrode for the Development of Ultra-High Cell Capacity in Lithium-Air Batteries

Lithium-air batteries (LABs) are expected to provide a cell with a much higher capacity than ever attained before, but their prototype cells present a limited areal cell capacity of no more than 10 mAh cm−2, mainly due to the limitation of their air electrodes. Here, we demonstrate the use of flexible carbon nanotube (CNT) sheets as a promising air electrode for developing ultra-high capacity in LAB cells, achieving areal cell capacities of up to 30 mAh cm−2, which is approximately 15 times higher than the capacity of cells with lithium-ion battery (LiB) technology (~2 mAh cm−2). During discharge, the CNT sheet electrode experienced enormous swelling to a thickness of a few millimeters because of the discharge product deposition of lithium peroxide (Li2O2), but the sheet was fully recovered after being fully charged. This behavior results from the CNT sheet characteristics of the flexible and fibrous conductive network and suggests that the CNT sheet is an effective air electrode material for developing a commercially available LAB cell with an ultra-high cell capacity.


Results and Discussion
Single-walled CNTs (SWCNTs) with metallic conductivity and a relatively small tube diameter of approximately 2 nm ( Figure S1) were chosen as the starting material in the fabrication of a CNT sheet air electrode to obtain a surface area that was as large as possible. Two types of CNT sheets were prepared through vacuum filtration of CNTs that had been ultrasonically dispersed in isopropanol (IPA) or N-methyl-2-pyrrolidone (NMP). CNT-IPA and CNT-NMP denote the corresponding CNT sheets prepared from the CNT slurries in IPA and NMP, respectively. CNT-IPA and CNT-NMP were composed of a non-woven textile of CNT bundles with ~50 nm and ~150 nm average widths, respectively, based on scanning electron microscopy (SEM) observations ( Figure S2) and had 65% and 60% porosity based on the sheet weight-thickness relationships ( Figure S3) for CNT-IPA and CNT-NMP, respectively. The LAB cells in the CR2032 configuration (2.0 cm 2 electrode area) using CNT-IPA or CNT-NMP as air electrodes had a constant discharge voltage of 2.65 V and a gradually increasing charge voltage of up to 4.5 V ( Figure S4). The CNT-IPA cell emitted an output current of up to 0.4 mA cm −2 ( Figure S5), and the cell repeated the discharge/charge cycle up to 73 times with a limited cycle capacity of 0.5 mAh cm −2 ( Figure S6); all of these cell performance parameters are comparable to those of LAB cells using solidified Ketjenblack ® 8 or graphene air electrodes 12 . Figure 1(a) shows the discharge capacities of cells possessing different CNT sheet thicknesses before reaching the cutoff voltage of 2.0 V. The capacities were plotted against the electrode weight, which is directly proportional to the sheet thickness. Although a negligible difference in battery performance was observed between the CNT-IPA and CNT-NMP cells in the limited cycle capacity operation ( Figure S4), the CNT-NMP cells had smaller discharge capacities not exceeding 1 mAh cm −2 regardless of their thickness, excluding a few outliers, owing to the difference in the CNT bundle width and hence the inhomogeneous occurrence of the LAB reaction inside the CNT-NMP, which will be discussed later. In contrast, most of the CNT-IPA cells had much higher discharge capacities. Though their capacities varied widely with the differences in each cell, the maximum discharge capacity at each electrode weight linearly increased with the sheet weight up to 5 mg (56 μ m thick), illustrating the efficiency of the sheet electrode at a discharge capacity of 5,000 mAh g CNT −1 (linear dotted line). As a result, the CNT-IPA cell achieved a maximum discharge capacity of 12 mAh cm −2 (24 mAh per cell), which was 6 times higher than the capacity of the conventional LiB cells of approximately 2 mAh cm −2 . Figure 1(b) shows the discharge profile. Even though the efficiency of the discharge capacity (5,000 mAh g CNT −1 ) is not outstanding compared to the efficiencies of all reported air electrode materials, the maximum cell capacity of 12 mAh cm −2 is quite competitive. The rechargeability of the cell was also confirmed, as shown by the dotted line in Fig. 1(b). However, the maximum capacity was not further improved by simply increasing the sheet thickness, which implies that the sheet electrode was becoming suffocated as the thickness was increased.
After the discharge experiment, the CNT sheet was observed to have expanded considerably by extruding into the 0.8 mm-diameter holes perforated on the stainless-steel case of the cathode for oxygen intake ( Fig. 2(a)). The cross-sectional SEM image of the sheet (Fig. 2(b)) revealed that the part of the CNT sheet that faced the open perforated holes (b,i) experienced the increase in sheet thickness, but the part harnessed by the cathode case (b,ii) did not expand from its initial thickness. Figure 2(c) shows the swelling ratio of the initial sheet thickness (~50 μ m) at the protruding parts (b,i) of CNT-IPA and CNT-NMP during the discharge/charge cycle. In the discharge process, CNT-IPA began to swell at 1 mAh cm −2 discharge and kept swelling to an approximately 3-fold thickness (~150 μ m) at 5 mAh cm −2 discharge. In the charge process, the sheet thickness of CNT-IPA monotonously shrank with increasing charge, returning to its initial thickness in the fully charged state. The CNT-NMP sheets abruptly distended by . A maximum discharge capacity of 12 mAh cm −2 (24 mAh per cell) was obtained using a cell with a 5 mg CNT-IPA electrode. (b) Discharge profile of the CNT-IPA cell that has a discharge capacity of 12 mAh cm −2 (solid line). A typical discharge-charge cycle profile of the cell at a fixed cycle capacity of 5 mAh cm −2 is shown by the dotted line.
4.8 times at 3 mAh cm −2 discharge with their extending error bars. The large error values for CNT-NMP resulted from the considerable roughness and layered voids created inside the sheet. The cross-sectional SEM observation and the corresponding energy-dispersive X-ray spectroscopy (EDS) analysis of CNT-IPA and CNT-NMP revealed that the battery reaction occurred inhomogeneously inside CNT-NMP ( Figure S7), which explains why the most of the CNT-NMP cells stopped discharging at low discharge capacities; the inhomogeneous reaction and hence the localized deposition of the discharge product suffocated the rest of the sheet, rendering it unusable for the discharge reaction and resulting in lower discharge capacities in the CNT-NMP cells. (d) SEM images of CNT-IPA after discharges of 1 mAh cm −2 (i), 2 mAh cm −2 (ii), and 5 mAh cm −2 (iii) and after being fully charged following a 5 mAh cm −2 discharge (iv). The dotted yellow squares show the magnified areas superimposed in each image. (e) Low-(i) and high-(ii) magnification TEM images of CNT-IPA after a 9 mAh cm −2 discharge. (f) XRD spectra of pristine CNT-IPA (i) and after discharges of 1 mAh cm −2 (ii), 2 mAh cm −2 (iii), and 5 mAh cm −2 (iv) and after being fully charged following a 5 mAh cm −2 discharge (v). The *symbols denote Li  The sheet was subjected to microscopic analysis after the discharge/charge experiment to investigate what triggered the swelling/shrinkage of the sheet. Figure 2(d) shows magnified SEM images of CNT-IPA after the discharge/charge experiment. The images obtained after discharge illustrate the deposition process of the discharge product; first, disk particles 50-150 nm in diameter were deposited on CNT bundles, and the number of particles then increased, with the particles evolving into a toroidal shape. Additionally, the bundles slightly thickened as the discharge progressed. The characteristic toroid shape suggests the formation of Li 2 O 2 as a discharge product, which has been reported in a wide range of LAB cells with non-carbonate electrolytes and carbonaceous air electrodes 7,11,12,15,[23][24][25][26][27] , and the detailed mechanism of the morphological evolution has been discussed by Yang and Thompson 21 . The full charge from a 5 mAh cm −2 discharged state completely decomposed the discharge product, restoring it to the initial bundle network structure. Transmission electron microscopy (TEM) observations of the CNT bundles after discharge (Fig. 2(e)) revealed the presence of a filmy product with a thickness of 5 nm or less covering the CNT bundles, in addition to the particle deposits of 50-150 nm observed in the SEM images. The deposition/decomposition of the discharge product in these toroidal or filmy morphologies caused the CNT sheet to reversibly swell/shrink, maintaining the conductive fibrous network of CNT bundles. X-ray diffraction (XRD) analysis of the CNT sheets after the discharge/charge experiment ( Fig. 2(f)) showed peaks that were identically attributed to Li 2 O 2 crystal diffraction after 2 mAh cm −2 discharge. As no other peaks aside from Li 2 O 2 were found, Li 2 O 2 was the exclusive discharge product of the discharge/charge reaction of the LAB. The peaks increased in intensity with further discharge to 5 mAh cm −2 but completely disappeared after reaching full charge, which is consistent with the toroid evolution observed by SEM.
These microscopic analyses revealed that the large discharge capacity of the cell with the CNT sheet air electrode was a result of the flexible sheet characteristic with discharge-dependent swellability. The fixed position of the CNT sheets inside the cell cases raises concern that the attained discharge capacity might have been suppressed by the limited sheet swelling within the range of the windows of the cathode case holes. In an attempt to extend the swelling range of the CNT sheet during discharge, a gas diffusion layer (GDL), an array of solid carbon fibers ~10 μ m in diameter (TGP-H-060, 190 μ m thickness, Toray, Japan), was placed on both sides of the CNT sheet air electrode, which dramatically increased the discharge capacity relative to the cells without GDL insertion. Figure 3(a) shows a schematic illustration of the GDL-inserted air electrode. The red solid line in Fig. 3(b) shows a discharge profile of a cell with the air electrode of CNT-IPA sandwiched between two GDLs, recorded before reaching the cutoff voltage of 2.0 V. The cell achieved a 30 mAh cm −2 discharge capacity (60 mAh attained per cell), which is 15 times higher than that of conventional LiB cells. The authors believe that this ultra-high discharge capacity was a result of the inserted GDLs, which provided the CNT sheet with additional space for swelling during discharge. Importantly, the cell capacity purely resulted from the CNT sheet, as the GDL itself contributed little, less than 0.05 mAh cm −2 by a single GDL, to the discharge capacity. The cell with the CNT-GDL composite air electrode was also fully rechargeable, as shown by the dotted line in Fig. 3(b). Although the cyclic performance of this cell is being studied by our group, further studies on the anode and electrolyte are required to collect subsequent discharge/charge cycle profiles. The behaviors of the Li metal and the TEGDEM electrolyte during such long-term operation are currently unknown. The bumpy discharge voltage observed in the discharge profiles reflects the progress of the swelling of the CNT sheet inside the GDL. To the best of our knowledge, this cell capacity is the highest among the prototype LAB cells reported to date.
The cell with the CNT-GDL composite air electrode was disassembled after discharge, which revealed the enormous amount of CNT swelling over the GDL as the result of Li 2 O 2 deposition. Figure 4(a) shows photos of each electrode from the cell discharged at 20 mAh cm −2 . The lithium metal anode (200 μ m thickness) was partly depleted, revealing the stainless-steel anode case at the bottom (a, i), and this depletion could have stopped the discharge. Further increases in the discharge capacity to 40 mAh cm −2 (80 mAh per cell) can be obtained by using thicker lithium metal (1 mm), but these data will be reviewed in a subsequent report. The CNT sheet was turned into a crisp solid after 20 mAh cm −2 discharge, implying that the CNT bundle network was completely filled with the discharge product (a, ii). The CNT sheet thickness was approximately 150 μ m (swelling ratio ~3) at 15 mAh cm −2 discharge. The inserted GDL had a black mound on the surface (a, iii) in the area where the lithium anode had been depleted. The mound was 1-2 mm thick when viewed edge-on (a, iv), corresponding to ~10-fold swelling from the pristine GDL (initially 190 μ m thickness). Since the GDL itself contributes little to the discharge reaction, the black mound generated on the GDL surface is deemed to be the discharge product grown from the CNT bundles that had penetrated inside the GDL. The intruding CNT bundles into the GDL are discussed later. Additionally, the deposit mound was observed on the GDL that faced the separator, and little deposition appeared on the air side. XRD (Fig. 4(b)) and Raman spectroscopy (Fig. 4(c)) analyses of the CNT-GDL composite air electrode revealed that Li 2 O 2 was the exclusive discharge product on both the CNT and GDL parts, even in an ultra-high discharge state, because their spectra contained peaks that can be attributed only to Li 2 O 2 , except for the carbon graphite substrate. The SEM observation of the CNT part after discharge (Fig. 4(d)) displayed several grooves corresponding to traces of carbon fibers from the GDL (d, i), and the magnified image of the sheet surface showed heavily beaded particles of ~100 nm around the CNT bundles (d, ii), implying that no additional space existed to accommodate Li 2 O 2 particles on the bundle surface. The TEM image of the sheet (d, iii) also showed the beaded deposits on the CNT bundles observed the SEM image, and even from the densely packed state, the sheet was recovered to the non-woven bundle textile state at full charge (d, iv). On the other hand, the SEM observation of the black mound on the GDL (Fig. 4(e)) indicates a densely filled deposit between the gaps in the carbon fibers of the GDL (e, i). The magnified image of the deposit revealed granular particles of ~100 nm captured by thin fibers (e, ii), and the TEM observation revealed fine aggregates of the particles (e, iii). The presence of CNT bundles that had seeped into the GDL was clearly confirmed by the image of the GDL after being fully charged, depicting CNT bundles remaining inside the carbon fibers (e, iv).
The evidence supporting Li 2 O 2 as the exclusive discharge product indicates that the ideal LAB reaction nominally written as 2Li + + O 2 + 2e − → Li 2 O 2 is occurring even during the discharge to ultra-high capacity. The 30 mAh cm −2 discharge capacity achieved by the cell with the CNT-GDL composite air electrode consumed 150 μ m of the thickness of the Li metal anode to form 120 μ m of bulk thickness of Li 2 O 2 on the cathode. This thickness of Li 2 O 2 could barely be held by the rigid carbon cathode but was successfully stored in the flexible CNT sheet by the swelling of the sheet to a thickness of a few millimeters. The CNT sheet (CNT-IPA, ~50 nm bundle width) had a calculated surface area of 60 m 2 g −1 . Although this surface area is much less than that of the rigid carbon cathode, the swellable nature of the sheet provides distinctive advantage to LAB cells for extending their areal cell capacity. The 120 μ m-thick Li 2 O 2 deposit on CNT-IPA numerically leads to ~80 nm-thick Li 2 O 2 on the CNT bundle surface, which is close to the particle size observed by SEM and TEM in Figs 2 and 4. However, the detailed reaction mechanism by which the Li 2 O 2 discharge product grows is not clear at this time. "Plating" and "precipitation" processes have been discussed for Li 2 O 2 growth in LABs 28 . In the "plating" process, which is considered a two-electron reaction, the LAB reaction occurs only on the cathode surface to form Li 2 O 2 , but the insulative nature of Li 2 O 2 limits its growth to only a few nm in thickness. This process indicates that the filmy deposit on the CNT bundles observed in Fig. 2(e, ii) was Li 2 O 2 but would hardly be able to produce particles with thicknesses of ~100 nm. On the other hand, the "precipitation" process involves a one-electron reaction in an electrolyte to form solvated LiO 2 (Li + + O 2 + e − → LiO 2 ), which is subsequently precipitated (disproportionated) on the cathode surface as Li 2 O 2 while producing O 2 (2LiO 2 → Li 2 O 2 + O 2 ). Contaminated protons promote the solvation of LiO 2 , and micron-sized toroidal Li 2 O 2 shapes have been observed on the cathode in an electrolyte containing a few thousand ppm H 2 O 29,30 . The presence of small particles with ~100 nm thickness in Figs 2 and 4 suggests that Li 2 O 2 was generated through the precipitation process in this study. However, the experiments were performed in a dry state (H 2 O < 30 ppm), and it is uncertain whether this process can provide the mechanical stress needed to swell the CNT sheet to a thickness of a few millimeters. Since the mechanism of Li 2 O 2 growth is directly related to the design of the air electrode for deriving cell performance, further study is in progress to determine the Li 2 O 2 deposition process.
In summary, LAB cells with CNT sheet air electrodes were prepared to extend the actual cell capacity of the LAB cell, exploiting the sheet characteristics of the flexible and fibrous network. The CNT sheet air electrode swelled along with the discharge and resulted in LAB cells that can discharge up to 30 mAh cm −2 . This discharge capacity is 15 times higher than that of current LiB cells and is the highest among the prototype LAB cells reported to date, to the best of our knowledge. As a result, the sheet electrode experienced enormous expansion owing to deposition of the discharge product on the CNT bundles and grew into an expanded cathode with a thickness of a few millimeters. The discharge product deposit was solely identified as Li 2 O 2 based on XRD and Raman analyses, which implies that the ideal LAB reaction was occurring inside the cathode during the discharge. Therefore, the CNT sheet is a promising air electrode material for achieving an ultra-high cell capacity in LAB cells. In addition, the invention of long-lasting air electrode provide paths for studying appropriate designs of Li metal anodes and electrolytes for LAB. These push new frontier for constructing practically available LAB cells in the future.

Methods
Preparation of the CNT Sheet Air Electrode. Seventy-five milligrams of SWCNTs with an average tube diameter of 2 nm (EC2.0, Meijo Nano Carbon) was dispersed in 150 g of IPA or NMP with an ultrasonic homogenizer (450D, Branson) equipped with a 1/8" micro-tip for 3 h in an ice bath. The prepared slurry of SWCNTs was filtered under vacuum through a PTFE membrane (Omnipore TM , 1 μ m pore size, EMD Millipore) to obtain a free-standing and binder-free CNT sheet. After drying in a vacuum oven at 60 °C overnight, the CNT sheet was deliberately peeled off from the membrane and cut into a 16 mm-diameter cathode (air electrode, 2 cm 2 effective area). Air electrodes of various thickness (10-250 μ m thick) were prepared by adjusting the filtered amount of the slurry.
Battery Assembly and Testing. Batteries were assembled in a CR2032 coin cell with perforated stainless-steel cases (Hohsen Corp.) in a dry room with a supply-air dew point of < − 90 °C (< 0.1 ppm of H 2 O) and an environmental dew point of approximately − 60 to − 50 °C. The cell comprised layers of a Li metal foil anode (16 mm in diameter, 200 μ m thick, Honjo Metal), glass fiber separator (GF/A, Whatman), and CNT sheet air electrode facing the perforated side of the cell. Two films of a PP/PE/PP trilayer porous membrane (20 μ m thick) were placed on both sides of the GF/A separator to prevent the glass fibers from sticking to the anode and air electrode. An ether-based electrolyte, composed of 1 M lithium bis(trifluoromethanesulfonyl)imide (LiTFSI, Kishida Chemical Co., Ltd.) dissolved in anhydrous tetraethylene glycol dimethyl ether (TEGDME, < 10 ppm H 2 O, Japan Advanced Chemicals) was prepared in an Ar-filled glovebox (Glovebox Japan Inc.) and immersed in the GF/A glass fiber separator and the CNT sheet air electrode. The prepared electrolyte had a water content of less than 30 ppm, determined by Karl Fisher titration before the immersion. Discharge and charge profiles were collected under oxygen gas flow at a constant current density of 0.1 mA per cell (0.05 mA cm −2 ) using a battery discharge-charge system (HJ1001SD8, Hokuto Denko). The lower limit of the discharge voltage was set at 2.0 V, where the LAB normally reaches its fully discharged state.
Characterization. After the discharge/charge experiment, the batteries were disassembled in the dry room to collect the CNT sheet air electrode for analysis. The collected CNT sheets were copiously rinsed with super dehydrated tetrahydrofuran (< 10 ppm H 2 O, Wako Pure Chemicals) and then dried under vacuum. The completely dried CNT sheets were transported using air-isolated sample holders to prevent possible reactions with the humidity in ambient air. The morphologies and structures of the CNT sheet air electrode were analyzed using a JSM-7800F field-emission scanning electron microscope (FE-SEM, JEOL) with a 5 keV accelerating voltage. EDS was performed using an X-Max N analyzer (Oxford Instruments) attached to the SEM. TEM images of the CNTs and discharge product were obtained from a JEOL JEM-ARM 200 F at 200 or 80 keV accelerating voltage. XRD measurements were carried out on a D8 ADVANCE (Bruker) powder X-ray diffractometer with a CuKα source (λ CuKα = 1.542 Å).