Lithium Batteries with Nearly Maximum Metal Storage - ACS Nano

May 16, 2017 - *E-mail: [email protected]. ... Tuo Wang , Rodrigo Villegas Salvatierra , Almaz S. Jalilov , Jian Tian , and James M. Tour. ACS Nano 2017 A...
1 downloads 0 Views 4MB Size
Download PDF
Lithium Batteries with Nearly Maximum Metal Storage Abdul-Rahman O. Raji,†,⊥ Rodrigo Villegas Salvatierra,†,⊥ Nam Dong Kim,† Xiujun Fan,‡ Yilun Li,† Gladys A. L. Silva,† Junwei Sha,† and James M. Tour*,†,‡,§ †

Department of Chemistry, ‡Smalley-Curl Institute and The NanoCarbon Center, and §Department of Materials Science and NanoEngineering, Rice University, 6100 Main Street, Houston, Texas 77005, United States S Supporting Information *

ABSTRACT: The drive for significant advancement in battery capacity and energy density inspired a revisit to the use of Li metal anodes. We report the use of a seamless graphene−carbon nanotube (GCNT) electrode to reversibly store Li metal with complete dendrite formation suppression. The GCNT-Li capacity of 3351 mAh g−1GCNT‑Li approaches that of bare Li metal (3861 mAh g−1Li), indicating the low contributing mass of GCNT, while yielding a practical areal capacity up to 4 mAh cm−2 and cycle stability. A full battery based on GCNT-Li/sulfurized carbon (SC) is demonstrated with high energy density (752 Wh kg−1 total electrodes, where total electrodes = GCNT-Li + SC + binder), high areal capacity (2 mAh cm−2), and cyclability (80% retention at >500 cycles) and is free of Li polysulfides and dendrites that would cause severe capacity fade. KEYWORDS: Li metal anodes, graphene, carbon nanotubes, full battery, Li dendrites (∼2.0 and 1, and a trace D band at ∼1360 cm−1. The Raman scattering signatures signify a high-quality multilayer graphene. The skewed baseline occurred because the spectrum is obtained atop Cu. (f) Raman spectrum of CNTs grown on the Cu-graphene substrate with the G band at 1587 cm−1, 2D band at 2652 cm−1, and D band at 1336 cm−1. (g) Raman RBMs of the CNTs in expanded format.

graphene. An average electronic conductivity of 1.45 × 103 S m−1 was measured from the I−V curves for the entire electrode from the bottom of the Cu current collector through the top of the GCNT (Figure S1). By adjusting the time of growth, different GCNT thicknesses from 10 to 50 μm can be obtained. The CNTs exist in bundles (Figure 1c,d), which are superlattices held together by van der Waals interactions.26 The CNT bundles are mainly vertical at their bottom but can bend at the top. In addition to an intertube spacing of ∼3.4 Å, the CNT bundles have 6 Å channels.27 The graphene and CNTs are confirmed in Figure 1e,f, and the radial breathing modes (RBMs) of the latter at 100−300 cm−1 indicate singleto few-walled CNTs (Figure 1g).28,29 The CNTs seamlessly connect to the graphene substrate and are able to store Li ions;27,30,31 however, due to low areal mass density, the capacity generated by the GCNTs is negligible (500 cycles), and the system is free of Li polysulfides and dendrites that would cause severe capacity fade.

RESULTS AND DISCUSSION Individual metallic CNTs have an electrical conductivity of 106 to 107 S m−1,22 on the order of in-plane conductivity of graphite and up to the conductivity level of copper, a high specific surface area (2600 m2 g−1),21,23 and when grown vertically from a substrate, a 3D structure suitable for nontortuous Li plating.24 Here, we first grow graphene via chemical vapor deposition (CVD) on a Cu substrate, followed by deposition of iron nanoparticles and aluminum oxide and subsequent CVD growth of CNTs at 750 °C using acetylene as the carbon source (Figure 1a).21,25 This method has been previously shown to produce CNTs that are covalently and seamlessly connected to the underlying graphene (Figure 1b),21,25 providing ohmic conductance between the underlying Cu onto which the graphene is grown and the CNTs on top of the 6363

DOI: 10.1021/acsnano.7b02731 ACS Nano 2017, 11, 6362−6369

Article

ACS Nano

contrast, deposition over flat substrates (graphene-covered copper foil, CuG) as shown in Figure 2l, using the same electrolyte, produces irregular deposits of Li (Figure 2m,n). Mossy structures are observed in less than 10 cycles. A previous report using the same electrolyte (4 M lithium bis(fluorosulfonyl)imide (LiFSI) in 1,2-dimethoxyethane (DME), see Methods)32 also has demonstrated dendrite suppression on flat Cu foil, although the tests were performed under low areal capacity (0.5 mAh cm2) conditions. Etherbased solvents have long been known for improved Li metal battery performance.33 The shape of the Li particles in this same work also demonstrates a change in morphology in which higher current densities (>1 mA cm−2) result in the formation of dendrite-shaped Li particles,32 as is observed under our conditions (at 2 mA cm−2) when testing pure CuG substrate. The 3D high surface area of the GCNT enables the dendrite suppression effect at high current densities and high capacity, in which the Li metal is infused inside the CNT bundle structure. The high surface area of GCNT contributes to reduce the local current density of Li deposition; the Li deposition occurs on a much larger area compared to a flat Cu foil. The low local current density also implies a low rate of Li-ion reduction to Li metal over the GCNT without exceeding the diffusion limit of Li ions from the electrolyte into the pores. Therefore, the GCNT structure enables an extension of the Sand’s time for dendrite-free Li plating.34 In addition, the seamless connection of the bottom of the vertically oriented CNTs with the graphene grown on Cu would remove interfacial resistance, thus improving the overall electrode electronic conductivity (Figure S1) and facilitating current distribution throughout the electrode to promote a homogeneous Li deposition in the GCNT carpet at high current density with low overpotential. This demonstrates that, in addition to the electrolyte, the GCNT is necessary to obtain the observed results. Figure 3 shows the electrochemical behavior of the GCNT-Li anodes versus CuG-Li. Figure 3a shows representative curves of the Li plating (discharge) and stripping (charge) from the sixth cycle, expressed in terms of gravimetric capacity. The measured capacity at 2 mAh cm−2 is 3120 mAh g−1GCNT‑Li, having as reference the masses of both GCNT and plated Li metal, which represent ∼80% of the capacity of the pure Li metal. As discussed previously, the GCNT has minimal contribution to the anode capacity at the same voltage range (Figure S2). The discharge and charge curves are characterized by remarkably flat voltages at −50 and +50 mV, respectively (Figure 3a), characteristic of Li metal plating and stripping.32,35 The charge/discharge curves are also similar in CuG-Li (Figure S3) at the first cycles. It is evident that the plated Li in the GCNT is metallic in contrast with Li-intercalated graphite (LiC6), where Li exists as an ion.35,36 The Li metal in GCNT-Li is further confirmed by X-ray diffraction (XRD) (Figure S4). Next, we compared the polarization evolution between charge and discharge over cycling. The polarization remains similar for GCNT-Li as it is observed over 200 h of continuous cycling (300 cycles) (partial cycling time shown in Figure 3b and Figure S5). In comparison, CuG-Li shows oscillating CE and increased polarization over time (partial cycling time shown in Figure 3c). This could be caused by the irregular Li particles formed on the bare CuG substrate (Figure 2l−n). Figure 3d shows the cycling stability and CE. Although the discharge (plating) capacity is fixed by the time-controlled galvanostatic deposition, the charge (stripping) capacity is regulated by voltage cutoff at 1 V; therefore, the ratio between these

Figure 2. Morphology of the GCNT-Li. (a) Schematic of GCNT-Li formation. (b) Voltage vs time of lithiation and delithiation processes of GCNT-Li. (c) Photograph of pristine GCNT, lithiated GCNT, and delithiated GCNT (scale bar = 1 cm). (d) SEM image of GCNT-Li (0.7 mAh cm−2 at 2 mA cm−2) by top view after 250 cycles. SEM images of GCNT-Li (4 mAh cm−2 at 2 mA cm−2) after 10 cycles by (e) side view, (f) expanded top view, and (g) expanded side view. SEM images of delithiated GCNT-Li (0.7 mAh cm−2 at 2 mA cm−2) after 250 cycles by (h) top view and (i) expanded top view. (j) TEM image of a CNT from GCNT-Li and (k) its higher magnification. (l) Schematic of Li deposited on graphene grown on Cu. (m) SEM image of Li deposited directly on graphene grown on Cu foil (0.7 mAh cm−2 at 2 mA cm−2) after 10 cycles, with no GCNT, showing the mossy and dendritic Li deposition and (n) its higher magnification.

from the GCNT are observed (Figure 2b). These are confirmed by a change in color of the GCNT from black to silver, indicating formation of Li metal (Figure 2c), and back to black upon Li stripping. SEM images of the GCNT (Figure 2d) and GCNT-Li (Figure 2e−g) show that the Li is not deposited atop the GCNT as a separate film but is rather plated homogeneously throughout the volume of the pillared CNT structure (Figure 2f), suggesting Li formation inside and outside the CNT bundles. The surface of the GCNT is homogeneously conductive based on its seamless connection with the basal graphene plane. Therefore, the CNT bundles are accessible to promote Li plating. The base view SEM image (Figure 2g) also indicates the presence of a deposited Li film, which underscores the significance of the micrometer-sized pores in Li-ion diffusion through the GCNT. No discernible exfoliation of the CNT bundles in the delithiated GCNT (Figure 2h,i) is observed. In Figure 2j,k, the transmission electron micrograph (TEM) images of the lithiated CNTs show deposition in the form of nanoparticles on the surface of the CNTs. The SEM images of the GCNT-Li presented in Figure 2d−i show no evidence of formation of dendritic, mossy, and related structures that have hindered application of Li metal anodes. The dendrite-free appearance even holds at 4 mAh cm−2. In 6364

DOI: 10.1021/acsnano.7b02731 ACS Nano 2017, 11, 6362−6369

Article

ACS Nano

Figure 3. Electrochemical characteristics of the GCNT anode. (a) Charge/discharge profile of GCNT-Li. Gravimetric capacity is based on the mass of GCNT-Li (see Methods). (b) Voltage profile of GCNT over time (inset: expanded charge/discharge cycle). (c) Voltage profile of CuG-Li (no GCNT) over time (inset: expanded charge/discharge cycle). (d) Cycle performance and CE of GCNTLi. The current density is 2 mA cm−2 (5.2 A g−1GCNT‑Li). The reference dotted black line is set to 99.9%.

Figure 4. Li storage and rate capabilities of the GCNT anode. (a) Li storage capacities of the GCNT from 0.4 to 4 mAh cm−2. (b) Cycling stability of different areal capacity anodes (1 to 4 mAh cm−2) expressed as areal capacity vs cycles. (c) Comparison of the gravimetric capacity of GCNT-Li (with different areal capacities) with other anode materials with respect to the mass of the anode at the fully lithiated state. The areal capacities of the GCNT-Li are from 0.4 to 4 mAh cm−2, represented by GCNT-Li-0.4 to GCNTLi-4. (d) Charge−discharge profiles measured at different current densities expressed in current density per area and per mass of electrode (GCNT-Li). The plating and stripping voltages are below and above 0 V, respectively, due to the overpotential, which increases with current density. (e) Voltage vs volumetric capacity.

capacities (CE) indicates the degree of reversibility of Li metal after deposition. The first cycle CE is 96.2%, indicating a good reversibility after the first Li plating in the GCNT structure. After 225 cycles, there is no charge capacity fading and the CE is 99.6% (Figure 3d), indicating that during cycling, the Li could be continuously infused in the GCNT structure. Conversely, we found that plating Li metal onto a compact, pure single-walled CNT electrodes (not GCNT) produced a separate layer of Li metal atop the electrode (Figures S6 and S7), indicating that an open structure and nontortuous path enabled by the GCNT is necessary for the efficient Li incorporation and dendrite suppression. The GCNT-Li structure also enables a stable cycling operation at different areal densities (0.4 to 4 mAh cm−2) (Figure 4a,b). The voltage also shows areal capacity dependence. A small voltage gap of 100 mV between the Li plating and stripping curves is observed for 0.7 mAh cm−2 increasing to 200 mV at 4 mAh cm−2 due to the thicker plated Li (Figure 4a). By adjusting the amount of Li in the GCNT-Li, it is possible to optimize the specific capacity to reach maximum metal storage per carbon mass. The GCNT-Li at 4 mAh cm−2 has a capacity of 3351 mAh g−1GCNT‑Li, which is close (∼87%) to the theoretical capacity of Li (3861 mAh g−1Li). A correct comparison of Li metal with other anode materials would take into account their lithiated forms. Thus, the GCNT-Li (3351 mAh g−1GCNT‑Li) has 1.8 times higher gravimetric capacity than Li15Si4 (1857 mAh g−1Li15Si4) and 9.9 times higher than LiC6 (339 mAh g−1LiC6), and it is significantly higher than other lithiated anode materials (Figure 4c). The low density of GCNT-Li (∼0.05 g cm−3, with an average thickness of 50 μm) implies a large proportion of unoccupied space (porosity 97.8%; see Methods) within the electrode that can be used to store high amounts of Li metal as a thin coating

over the GCNT bundle (Figure 4c). In practical aspects, the specific capacity of electrodes must also include the current collectors. A detailed comparison is provided in Table S1, in which a high gravimetric capacity (∼750 mAh g−1) is also obtained when including the Cu current collector mass, significantly higher when compared to graphite anode materials. The prospects of even higher gravimetric capacities of the GCNT-Li, including the current collectors, involve replacing the dense Cu metal foils by conductive carbon films, as recently demonstrated.31 In Figure 4d, the GCNT is shown to plate and strip Li at a current density as high as 10 mA cm−2 (25 A g−1GCNT‑Li), producing a capacity of ∼0.7 mAh cm−2 (2100 mAh g−1GCNT‑Li), which is independent of the current density. The flatness of the curves is still maintained up to 4 mA cm−2 (10 A g−1GCNT‑Li). However, during the GCNT-Li cycling at 10 mA cm−2, a significant polarization is observed from the Li plating/stripping curves with loss of the characteristic flatness at lower current densities. The high current capability supersedes values reported for LIB electrodes.37,38 The excellent electrical conductivity of the GCNT monolith facilitates electron transport without the need for conductive additives. The vertical carpet nature of the CNTs would enhance Li-ion diffusion through nontortuous Li plating and stripping with flexible CNT movements.39 By controlling the time of GCNT growth (Figure S8), it is possible to tune the thickness of the GCNT carpets in order to optimize the volume occupied by the resulting GCNT-Li structure after Li plating. Figure 4e 6365

DOI: 10.1021/acsnano.7b02731 ACS Nano 2017, 11, 6362−6369

Article

ACS Nano shows that the volumetric capacity of the electrode could reach up to 1600 mAh cm−3 at an areal capacity of 4 mAh cm−2 based on the thickness of the GCNT-Li (∼25 μm). Upon Li plating and associated solid electrolyte interphase (SEI) formation, the thickness increased from ∼17 to ∼25 μm (Figure S8). Due to the low thickness of the GCNT, it is expected that the Li will be able to penetrate through the depth of the GCNT, thus utilizing the available volume very efficiently. The CEs of the first cycle and the first few cycles are important factors in predicting the performance of an electrode in a FB. Using the prelithiation method (see Methods), a sample cycled to 4 mAh cm−2 shows a cumulative capacity loss of 8.7% during the first five cycles (which can be seen in Figure S9a). The prelithiation involves a spontaneous reduction of the GCNT electrode by the Li foil with partial lithiation of the GCNT (forming Li+GCNT − not GCNT-Li) and formation of the SEI layer. Both processes can lead to an increase of the first cycle CE compared to the GCNT without prelithiation treatment (Figure S9b). Furthermore, we observed that the areal capacity influenced the first cycle capacity loss. The average CE of the first cycle can achieve >90% at 4 mAh cm−2 and >80% at the 1 mAh cm−2 anode (Figure S9b). Regarding the prelithiation process, the Li ions in the lithiated GCNT can drive the subsequent Li metal plating, which could explain the conformal production of Li metal films over the volume of the CNT bundles, as observed in our SEM and TEM images (Figure 2). In single-walled CNTs, it is known that Li ions are located preferentially in the interstitial channels between the individual CNTs, which expands the CNT bundle structure during Li+ intake.40 Therefore, the process to deposit Li metal could start from these expanded bundles or from the initial Li ions previously stored in the bundle. This is evident in the SEM images in which the Li metal is preferentially deposited over the CNT bundles, keeping its shape throughout the cycling. No Li metal is observed between the bundles or outside as separate films. This observation indicates a self-seeded mechanism of dendrite-free Li metal deposition using as starting material the lithiated GCNT. A similar reported structure was observed using metals as seeds for Li plating.41 This effect demonstrates a clear difference between our work and other reports that use a similar idea of having 3D scaffolds to minimize Li dendrites.12,13,15,17 We also demonstrated that a FB can be obtained by matching our dendrite-free GCNT-Li anode with a SC cathode with S content of ∼60 wt % (Figure S10). The S content in the cathode is reduced to 48 wt % with the addition of binder and carbon additives (see Methods). Cathodes based on SC have advantages over elemental sulfur (S8) cathodes, such as high compatibility with different electrolytes and absence of Li polysulfide diffusion (Figure S10);42 the latter generally leads to capacity fading over cycling in elemental sulfur cathodes.43 Figure 5a shows the cyclic voltammograms (CVs) of the GCNT-Li and the SC cathode (thirrd cycle) half-cells, each with total areal capacity of ∼2 mAh cm−2. The first cycle of the SC cathode half-cell has a CE of 83%, and the first cycle of the GCNT-Li anode half-cell has an average CE of 85% (Figure S10), both requiring a small excess of Li from the anode in the FB. The galvanostatic charge/discharge curve of the FB in Figure 5b shows that the discharge curve extends from 2.1 to 1.7 V. The specific capacity based on S mass is very close to that observed in the half-cells (Figure S10). A pouch FB based on GCNT-Li/SC is shown in Figure 5c. The FB can be cycled continuously at different rates (1 C = 1/discharge time, h) from

Figure 5. FB with a GCNT-Li anode and SC cathode. (a) CVs of GCNT-Li and SC cathode half-cells in 4 M LiFSI/DME at 0.5 mV s−1. (b) Galvanostatic charge/discharge curves of the FB at 0.1 C with areal capacity of 2 mAh cm−2. (c) Photograph of a FB prototype powering a LED. (d) Sequential rate performance test (0.2 to 9 C) and cycling stability of the FB. Inset: CE (%) of the rate and stability test. The reference blue dashed line is set to 99.9%. (e) Self-discharge (SD) tests of the FB after 8 h and 1 week showing charge curve followed by continuous discharge curve during and after the open circuit period. Inset: Voltage vs capacity of the SD tests. (f) Ragone plot of the GCNT-Li/SC FB, here considering the combined mass of the anode and cathode active materials (Li and S) and the full electrode mass (including binder, carbon additives, GCNT, excess of Li), excluding the current collector.

0.2 to 9 C. A cycle stability over 500 cycles is obtained at 1 C with ∼80% capacity retention (Figure 5d) and CE close to 99.9% (inset Figure 5d). The self-discharge (SD) was also tested in the FB, in which a stable voltage of 2.15 V can be achieved even after 1 week (Figure 5e) of SD. A capacity retention of 94 and 81% is measured after 8 h and 1 week of SD, respectively (inset, Figure 5e). Finally, the Ragone plot is calculated and presented in the Figure 5f for a range of energy and power densities. At the lowest power density, the energy density of the GCNT-Li/SC full-cell is 1423 Wh kg−1active materials (752 Wh kg−1total electrodes), where active materials = Li + S only and total electrodes = GCNT-Li + SC + carbon additives + binder. This is 3× higher energy density than that seen in Li−S full-cells with respect to the mass of active materials (Li−S).44 Moreover, the data appear attractive when compared to commercial LIB performances with 310 Wh kg−1active materials (220 Wh kg−1total electrodes),45 where active materials = graphite + LiCoO2; total electrodes = graphite + LiCoO2 + carbon additives + binder. However, a definitive comparison with a commercial cell is difficult at this stage because commercial cells are dual-sided and stacked, designed to minimize the contribution of current collectors and packaging materials. Further data on the gravimetric energy density calculation are shown in Tables S2 and S3. In a nonoptimized device, we achieved a volumetric energy density of 6366

DOI: 10.1021/acsnano.7b02731 ACS Nano 2017, 11, 6362−6369

Article

ACS Nano

for the electrochemical measurements (insertion/extraction and cycling) ranges from 1 to 10 mA cm−2, all performed at room temperature. For the Li plating (discharging process), a timecontrolled process with a constant current regime was applied with no cutoff voltage limit. The stripping process (charge process) was set to a constant current regime with a cutoff voltage of 1 V (vs Li+/Li). A control experiment was carried out using copper foil upon which graphene is grown by CVD. The gravimetric capacity was determined based on the combined mass of the GCNT, which ranges from 0.1 to 0.4 mg cm−2, and the Li metal, which was calculated from the amount of charge applied during the electrodeposition or plating (for instance, 4 mAh cm−2 is equivalent to 1.04 mgLi cm−2). Cathode Preparation and FB. SC cathodes were prepared by the decomposition of polyacrylonitrile (PAN) (Sigma-Aldrich, Mw 150k) in the presence of excess elemental sulfur. In a typical preparation, PAN, S, and graphene nanoribbons (GNRs) (EMD-Merck) in the mass ratio of 55:11:1 were ground together using a mortar and pestle. The GNRs improved the conductivity of the final material. The resulting powder is heated from room temperature to 450 °C at a rate of 5 °C min−1 in an argon atmosphere (1 atm). After 6 h, the SC powder was removed and used without purification. The SC powder has approximately 60 wt % S. The SC cathodes were prepared by mixing the SC powder with carbon black (Black Pearls 2000, Cabot Corp.) and polyvinylidene fluoride (PVDF, Sigma-Aldrich) in a mass proportion of 8:1:1, resulting in a total S content in the electrode of 48 wt %. Typical mass loading was 4−5 mg in 1 cm2 electrodes. Steel foils were used as current collectors. Electrolytes in the half-cell tests were either 4 M LiFSI/DME or 1 M LiPF6 in EC/DEC (1:1) in the voltage range of 1 to 3 V. The FB was assembled by previously producing GCNT-Li and then combining with a SC cathode using the 4 M LiFSI/DME electrolyte and Celgard K2045 as separator. The electrodes were ∼1 cm2. The areal capacity of the GCNT-Li is set to match the 30% irreversible capacity loss of the first cycle of the SC cathode. Materials Characterization. Coin cells were disassembled inside a glovebox to investigate the morphology of the GCNT electrodes after Li insertion/extraction. SEM images of the GCNT electrodes were obtained with an FE-SEM (JEOL-6500F) at an accelerating voltage of 20 kV. HRTEM images (JEOL FEG-2100F) were obtained after preparing the samples by sonicating the GCNT substrate in acetonitrile and dropping the dispersion over TEM grids. Electrical conductivity was measured with a 4155C Agilent semiconductor parameter analyzer. Measurements were conducted between the substrate (copper foil) and evaporated areas (30 μm thick Li, contact area 62500 μm2) over the top of the CNT.

234 Wh/Ltotal electrodes. There is no dendritic or mossy Li in the full-cell electrodes after 500 cycles (Figure S11). These results represent a promising achievement for a Li polysulfide- and dendrite-free battery.

CONCLUSIONS In conclusion, the concept of a 3D GCNT structure for reversible and stable Li storage can be extended to other materials with the combined features of a conductive high surface area, high porosity, and low mass density, yielding dendrite-free Li-based batteries. The high surface area electrode results in an effective decrease in the local current density of Li plating, promoting the homogeneous deposition of Li metal as dendrite-free thin metal coating over the bundles of GCNT. In addition to this effect, the high porosity (97.8%) and low gravimetric density of the GCNT contributes minimally to the weight of the GCNT-Li electrode. This holds promise for achieving superior energy density due to the near theoretical Li storage capacity and serves as the basis for the demonstrated SC∥GCNT-Li full-cell in a high concentration electrolyte to produce a safe, stable, and high-performance battery, thus becoming a harbinger of future systems. METHODS Graphene−Carbon Nanotube Preparation. The preparation of GCNT was similar to the previously reported methods.21,25 First, Bernal-stacked multilayer graphene was grown on copper foil (25 μm) using the CVD method as reported elsewhere.46 The catalysts for CNT growth are deposited by e-beam evaporation over the graphene/ Cu foil in the order graphene/Fe (1 nm)/Al2O3 (3 nm). The CNT growth was conducted under reduced pressure using a water-assisted CVD method at 750 °C. First, the catalyst is activated by using atomic hydrogen (H•) generated in situ by H2 decomposition on the surface of a hot filament (0.25 mm W wire, 10 A, 30 W) for 30 s, under 25 Torr (210 sccm H2, 2 sccm C2H2 and water vapor generated by bubbling 200 sccm of H2 through ultrapure water). After the activation of the catalyst for 30 s, the pressure is reduced to 8.3 Torr and the growth is carried out for 15 min. The GCNTs were weighed on a Cahn Instruments C31 microbalance with an accuracy of 0.0012% and ultimate precision of 0.1 μg at the fraction of load being 0.0001%. The average mass loading is 265 μg ± 100 μg per cm2. The porosity ϕ of the GCNT (97.8%) was calculated based on the equation ϕ = (1 − dGCNT/dgraphite) × 100% and the average density of the GCNT carpets (dGCNT = ∼0.05 mg cm−3) and the density of graphite (dgraphite = 2.266 g cm−3). The weight of Cu/graphene/Fe/AlOx after the graphene growth and metal deposition is recorded. Then the weight after the CNT growth is recorded. The weight difference is attributed to the carbon electrode, knowing that the CNTs (∼50 μm thick) contribute far more extensively to the weight than does the graphene (∼1 nm thick). Electrochemical Plating/Stripping of Li into/from GCNT. The electrochemical reaction was performed in 2032 coin-type cells using GCNT substrates and Li foil as both counter and reference electrodes. The GCNT substrates are circular with total area of ∼2 cm2. The electrolyte used was 4 M lithium bis(fluorosulfonyl)imide (LiFSI) (Oakwood Inc.) in 1,2-dimethoxyethane (DME). The LiFSI salt is vacuum-dried (