Robust Expandable Carbon Nanotube Scaffold for Ultrahigh-Capacity Lithium-Metal Anodes

Transcription

1 Communication Lithium-Metal Anodes Robust Expandable Carbon Nanotube Scaffold for Ultrahigh-Capacity Lithium-Metal Anodes Zhaowei Sun, Song Jin, Hongchang Jin, Zhenzhen Du, Yanwu Zhu, Anyuan Cao, Hengxing Ji,* and Li-Jun Wan There has been a renewed interest in using lithium (Li) metal as an anode material for rechargeable batteries owing to its high theoretical capacity of 3860 ma h g 1. Despite extensive research, modifications to effectively inhibit Li dendrite growth still result in decreased Li loading and Li utilization. As a result, real capacities are often lower than values expected, if the total mass of the electrode is taken into consideration. Herein, a lightweight yet mechanically robust carbon nanotube (CNT) paper is demonstrated as a freestanding framework to accommodate Li metal with a Li mass fraction of 80.7 wt%. The highly conductive network made of sp2-hybridized carbon effectively inhibits formation of Li dendrites and affords a favorable coulombic efficiency of >97.5%. Moreover, the Li/CNT electrode retains practical areal and gravimetric capacities of 10 ma h cm 2 and 2830 ma h g 1 (vs the mass of electrode), respectively, with 90.9% Li utilization for 1000 cycles at a current density of 10 ma cm 2. It is demonstrated that the robust and expandable nature is a distinguishing feature of the CNT paper as compared to other 3D scaffolds, and is a key factor that leads to the improved electrochemical performance of the Li/CNT anodes. Lithium (Li)-based batteries have been considered to be among the most promising power sources for consumer electronics and electric vehicles, where a high energy density is a critical requirement. [1,2] However, as the capacities of anode/cathode Z. Sun, S. Jin, H. Jin, Z. Du, Prof. Y. Zhu, Prof. H. X. Ji, Prof. L.-J. Wan Hefei National Laboratory for Physical Sciences at the Microscale CAS Key Laboratory of Materials for Energy Conversion ichem (Collaborative Innovation Center of Chemistry for Energy Materials) School of Chemistry and Materials Science, University of Science and Technology of China Hefei , P. R. China jihengx@ustc.edu.cn Prof. A. Cao Department of Materials Science and Engineering College of Engineering Peking University Beijing , P. R. China Prof. L.-J. Wan Key Laboratory of Molecular Nanostructure and Nanotechnology Institute of Chemistry Chinese Academy of Sciences and Beijing National Laboratory for Molecular Sciences Beijing , P. R. China The ORCID identification number(s) for the author(s) of this article can be found under DOI: /adma materials of Li-ion batteries (LIBs) approach theoretical values, there has been only a marginal improvement in the energy density of LIBs in recent years. [3 5] In this context, because of their higher gravimetric capacity (3860 ma h g 1 ) and the lowest negative electrochemical potential ( V vs the standard hydrogen electrode) among the different electrode materials currently used in Li-ion technology, [6] Li-metal anodes have attracted renewed attention among the research community. [7,8] Despite impressive advantages and a half century of study, Li-metal anodes still suffer from serious issues that limit their practical application. These are: (i) uneven Li-plating on the anode surface leading to the formation of Li dendrites, which results in internal short circuiting of the battery and hence raises safety concerns; and (ii) the large change in dimension together with unstable Li/electrolyte interfacial chemistry leads to both the generation of dead Li and repeated buildup/breakdown of the solid state electrolyte interphase (SEI), resulting in irreversible Li consumption, which reduces Li utilization and coulombic efficiency (CE) of the battery. [6,7,9] Faced with the urgent need and high demand for high energy density batteries such as Li-S and Li-air batteries, several approaches have been proposed to overcome these issues from different perspectives. One method to avoid Li dendrites focuses on improving the stability and uniformity of the SEI layer by optimizing the electrolyte concentration and composition. [10,11] Alternatively, high-modulus solid-state mechanical coatings have been proposed by Cui and co-workers. [12 15] Both these approaches are aimed at mechanically suppressing breakthrough of the Li dendrites and thereby improving the cycle life of the battery. Recently, Xu et al. developed a novel method where an electrostatic shield formed by additives in the electrolyte resulted in self-healing to enable the preferential deposition of Li + ions in valley areas instead of peak areas on the Li metal surface, [16,17] thereby suppressing dendrite growth. Current developments in nanomaterials and nanotechnology focus on Li dendrite inhibition by constructing conductive scaffolds and fillers to accommodate Li metal. [18 20] This strategy relies on Sand s law according to which Li dendrite nucleation and growth rates are slowed down with reduced local current density, [21,22] thereby suppressing the formation of Li dendrites (1 of 7)

2 Figure 1. a) Photograph of the CNT paper and the voltage profile as measured by DC current sweep, from which the electrical conductivity of the CNT paper is calculated (inset). b) SEM images of the top surface (left panel) and cross section (right panel) of a CNT paper. c) TEM image of a multiwall CNT. d) SEM images of the top surface (left panel) and cross section (right panel) of a Li/CNT electrode with Li loading of ma h cm 2. The enlarged part shows CNT bundles buried in Li matrix. From the point of view of practical applications of Li-metal anodes in advanced Li-based batteries, concurrent improvements in safety, rate capability, and specific capacity are required. For instance, an LIB with energy density of 300 Wh kg 1 requires an electrode areal capacity of 6 ma h cm 2. Plus, advanced Li metal batteries (Li-S and Li-air) operate preferentially with Li anodes with areal capacities of >10 ma h cm 2 and high Li utilization of >90% to meet the target of 500 Wh kg 1 energy density per cell. [5,23 25] Nevertheless, modifications to effectively inhibit Li dendrite growth still result in decreased reversible capacity and/or cycling life. For example, Li-metal-filled carbon fiber network demonstrated high areal and gravimetric capacities of 19 ma h cm 2 and 2000 ma h g 1, respectively, but for only one discharge, indicating a low reversibility. [26] The typical reversible areal capacities of <5 ma h cm 2 of the Li-metal anodes reported in the literatures are still lower than target values for Li-based batteries. [26 32] Notably, Li-plating/stripping cycle is always accompanied with large volume change of the Li metal even when surrounded by a conductive scaffold. For instance, 1 g of Li metal with a gravimetric capacity of 3860 ma h g 1 requires a minimum space of 1.87 cm 3 for plating, equivalent to the volume of 1 g Li metal. Such a huge volume change causes tensile stress to build up, leading to fracture of the Li anode structure and cell failure. Hence, effective strategies to concurrently achieve improved safety, reversible capacity, and cycling life must be explored, to ensure the availability of safe rechargeable Li-based batteries with high energy density. Here, a carbon nanotube paper with deposited Li metal (Li/CNT) is proposed as a stable ultrahigh-capacity Li-metal anode. The CNT paper employed here is made of a freestanding network of multiwall CNTs (MWCNTs) with moderate specific surface area (SSA, 118 m 2 g 1 ), tensile strength (σ = 27 MPa), and high electrical conductivity (σ = S m 1 ). We found the robust and expandable structure of CNT paper is a key factor, distinguish from other carbon-based 3D scaffolds, for stable Li-plating/stripping cycles, which can accommodate Li metal with a mass fraction of up to 80.7 wt%. Our experiments show that Li/CNT electrodes are capable of delivering reversible areal and gravimetric capacities of 10 ma h cm 2 and 2830 ma h g 1, respectively, with 90.9% Li utilization for 1000 cycles at a current density of 10 ma cm 2. CNT paper with length/width in tens of centimeters and thickness of 12.0 ± 0.2 µm (Figure 1a,b) was prepared by a floating catalyst chemical vapor deposition method. The electrical conductivity of the CNT paper, as measured by a four-point probe method, is S m 1 (inset of Figure 1a). Scanning electron microscopy (SEM) images acquired from both the top surface and cross section (Figure 1b) of the CNT paper show an assembly of MWCNTs in the form of bundles (Figure 1c). The calculated nominal mass density and areal density of the CNT paper are 0.57 ± 0.02 g cm 3 and 0.68 ± 0.03 mg cm 2, respectively (Table S1, Supporting Information), which corresponds to 73 vol% of void space in the CNT paper. The SSA (118 m 2 g 1 ) and pore diameter ( nm) are calculated from N 2 adsorption desorption isotherms (Figure S1, Supporting Information). The porosity data are in accordance with the morphology observed in the SEM images (Figure 1b). The porous structure of the CNT paper mainly arises from the voids between the CNT bundles. The porous and highly conductive network in the CNT paper allows reducing the local current density for Li-plating/stripping, thereby, according to the Sand s law, [9] inhibiting the formation of Li dendrites. The lightweight CNT paper is beneficial to yield high mass ratio of Li in the composite, and thus leads to improved specific capacity (2 of 7)

3 Figure 2. Voltage profiles of Li/CNT electrodes plating/stripping at a) current density of 2 ma cm 2 for 4.25 h and b) current densities of 2, 3, 4, 5, 6, 7.5, and 10 ma cm 2 for 1 h. The inset in panel (a) presents voltage profiles during the time windows of , , and h. c) Areal (C A ) and gravimetric capacities (C G ) of Li/CNT electrodes acquired from panel (b). Li metal was plated on the CNT paper by galvanic deposition at a current density of 0.1 ma cm 2 (the current density is calculated with respect to the projected top-surface area of the CNT paper). It is noted here that Li/CNT can also be fabricated by melt infusion of Li as shown in Figure S2 (Supporting Information). Due to the excellent conductivity of the CNT paper ( S m 1, inset of Figure 1a), the freestanding Li/CNT electrode circumvents the need for an additional current collector. The thickness of the Li/CNT paper is 65.0 ± 1.8 µm (Figure 1d) at the Li loading of ma h cm 2. SEM images of the Li/CNT top surface show a relatively smooth surface (Figure S3, Supporting Information), indicating that a dendrite-free and dense plating of Li is present. The CNT bundles appear to be buried in a continuous matrix of Li (right panel of Figure 1d), as further confirmed by the cross-sectional SEM image of the Li/CNT sample (left panel of Figure 1d). These results show clearly that Li metal can be densely filled into the voids in the CNT paper and that the CNT paper is both robust and expandable. It is interesting to note that this robust and expandable nature is a distinguishing feature of CNTs as compared to other carbon-based 3D scaffolds that have been used in Li-metal anodes. [20,25,27,31] We will show in the following section how this feature leads to ultrahigh gravimetric and areal capacities for the Li/CNT composite electrode, as measured by galvanic cycling. To demonstrate the capability of the CNT paper for reversible Li metal storage, Li/CNT electrodes were prepared with Li loading of 11 ma h cm 2, corresponding to a Li mass fraction of 80.7 wt%. Figure 2 shows the long-term cycling profiles of the symmetric coin cells. Our measurements show that the Li/CNT electrodes cycle at a current density of 2 ma cm 2 with Li-stripping/plating time of 4.25 h for more than 3000 h and have high areal and gravimetric capacities of 8.5 ma h cm 2 and 2410 ma h g 1, respectively (Figure 2a). Voltage profiles acquired at the cycling time windows of , , and h (inset of Figure 2a) show that there is very little voltage fluctuation during long-term cycling, indicating the excellent cycling stability of these electrodes. We can find a small voltage fluctuation ( 5 mv) at the end of the Li-plating/ stripping voltage plateaus of the first 20 cycles, which can be ascribed to the unstable Li/electrolyte interface at the initial stage. [19,27] However, such voltage fluctuation is not observed in the subsequent cycles. In sharp contrast, Li foil and Li/Cu electrodes (Figure S4, Supporting Information) show significant voltage fluctuation even under a much more moderate cycling condition with a current density of 1 ma cm 2 and Li-stripping/ plating time of 2 h, with short circuits occurring at the cycling times of 61.5 and 110 h, respectively, for the Li foil and Li/Cu electrodes. Such an abrupt voltage drop resulting from a short circuit is not observed for Li/CNT (Figure 2a and Figure S4, Supporting Information). The improved cycling stability of Li/ CNT as compared to those of Li foil and Li/Cu is attributed to the dendrite-free Li-plating on CNT. This conclusion is substantiated by the observation that in the absence of CNT, extensive formation of Li dendrites is observed in the SEM images of both Li foil and Li/Cu electrodes (Figure S5, Supporting Information) (3 of 7)

4 Figure 3. a) Thicknesses of Li/CNT and de-li/cnt at different Li loadings. The insets show SEM images obtained from cross sections of Li/CNT with Li loadings of 1.25, 10, and 13 ma h cm 2. b) Schematic of volume expansion Li/CNT and Li-metal foil as electrode during Li-stripping/plating. Figure 2b presents the voltage profiles of the Li/CNT electrodes measured at varying current densities in the range of 2 10 ma cm 2 at a constant Li-stripping/plating time of 1 h. The values of voltage hysteresis, given by the sum of the overpotentials for Li stripping and plating, are 190, 300, 400, 510, 600, and 720 mv at current densities of 2, 3, 4, 5, 6, and 7.5 ma cm 2, respectively. When the current density increases to 10 ma cm 2, the voltage hysteresis switches to 940 mv and decreases steadily to 900 mv in the subsequent cycles of Li-stripping/plating. Even after 2000 h (1000 cycles) of cycling at a high current density (10 ma cm 2 ) and an areal capacity (10 ma h cm 2 ), the cell shows no evidence of a dendriteinduced failure. The steady changes in voltage hysteresis with both current density switch and cycling time point to the good rate capability and stability of the Li/CNT electrode. To further demonstrate the cycling stability of the Li/CNT electrode, the reversible areal and gravimetric capacities are calculated from the voltage profiles in Figure 2a,b, and the values are plotted against the number of cycles (Figure 2c). The reversible areal capacity, which is a function of both current density and Listripping/plating time, reaches 10 ma h cm 2 at the current density of 10 ma cm 2 in the 41st cycle and remains steady for 1000 cycles. The mass fraction of Li in the Li/CNT electrodes used in the long-term cycling experiments in this study is 80.7 wt%, which corresponds to theoretical areal and gravimetric capacities of 11 ma h cm 2 and 3110 ma h g 1, respectively. Thus, the reversible stripping of 10 ma h cm 2 Li yields a practical gravimetric capacity of 2830 ma h g 1 and the Li utilization of 90.9%. The calculated reversible volumetric capacity is 1540 ma h cm 3. It has been noted that voids occupy 73.1 vol% of the space in the CNT paper. Thus, a 12 µm thick CNT paper can accommodate a Li film of thickness up to 8.8 µm, with a maximum areal capacity of 1.8 ma h cm 2 and a Li mass fraction of 40.7 wt%, assuming that the voids in the CNT paper are completely filled with Li and the CNT network is rigid. The cross-sectional SEM images (Figure 1d) of Li/CNT show, however, that the CNT paper expands and the CNTs are buried in a continuous matrix of Li, indicating nevertheless a uniform distribution and dense filling of Li in the CNT paper. This corroborates with the electrode performance of 1000 cycles of Li-stripping/plating with reversible areal and gravimetric capacities of 10 ma h cm 2 and 2830 ma h g 1, respectively, which are significantly higher than the calculated values. Therefore, it can be inferred that the continuous increase in the thickness of Li/CNT with Li loading should be an important factor leading to improved areal and gravimetric capacities. To confirm this, we have plotted the thicknesses of Li/CNTs with a series of Li loadings (red curve in Figure 3a) and have plotted the thicknesses of the scaffolds after complete Li stripping (blue curve in Figure 3a). For Li loadings of 0, 0.63, and 1.25 ma h cm 2, the corresponding thicknesses of Li/CNTs are 12.0 ± 0.2, 13.2 ± 0.7, and 14.0 ± 0.6 µm, respectively. If we assume that Li metal is densely plated on the top surface of the CNT paper, a Li loading of 1.25 ma h cm 2 corresponds to a Li metal foil of 6.1 µm in thickness. However, at low Li loading (0.63 and 1.25 ma h cm 2 ; region I in Figure 3a), we observe very little difference in thickness between the CNT paper, Li/CNT, and de-li/cnt, indicating that Li metal is primarily filled inside the CNT paper. The electrolyte wets CNT paper, and Li can be deposited at the CNT surface where it contacts with electrolyte. Therefore Li can be deposited inside the CNT paper. With further increase in Li loading to ma h cm 2, the thickness of Li/CNT increases almost linearly with Li loading (region II in Figure 3a). When the Li loading is >11 ma h cm 2 (region III in Figure 3a), a much faster thickness increase is observed. The experimentally measured electrode thicknesses with Li loadings of <11 ma h cm 2 (regions I and II) are in line with values calculated based on a model where we consider that a dense plating of Li metal fills the void space in the CNT paper (Figure 3a). For Li loading of <1.8 ma h cm 2 (region I in Figure 3a), Li metal is plated inside the free pores of the CNT paper until this space is entirely filled; thus, the thickness of Li/CNT remains nearly unchanged with Li loading. For Li loading of ma h cm 2 (region II in Figure 3a), the thickness of Li/CNT can be regarded as the sum of the thicknesses (4 of 7)

5 Figure 4. a) Raman spectra of CNT and AC/CNT paper. b) TEM image of an AC-coated CNT. c) Tensile stress strain curves of CNT and AC/CNT papers. d) CE values for Li/CNT and Li/AC/CNT electrodes. of a void-free CNT paper (3.24 µm; Table S1, Supporting Information) and a Li metal foil (4.85 µm per 1 ma h cm 2 of Li metal) since the voids are completely filled at this Li loading. Therefore, the CNT paper expands along the vector of the paper plane, and the thickness of Li/CNT increases linearly with Li loading. For Li loading of >11 ma h cm 2 (region III in Figure 3a), the experimentally measured Li/CNT thickness is obviously larger than the calculated value, which is due to the delamination of the CNT paper. As shown in the cross-sectional SEM image of Li/CNT with Li loading of 13 ma h cm 2 (inset of Figure 3a), the Li/CNT split into a few layers with gaps in micrometers. It is noteworthy that CNT bundles are observed both at the top and bottom surfaces of the Li/CNT disk even when the Li loading is as high as 13 ma h cm 2 (insets of Figure 3a), indicating that the CNT paper expands with Li loading and provides extra void space for Li accommodation. Since structure integrity is critical for the battery electrode, we suggest that a 12 µm thick CNT paper can accommodate Li of areal capacity up to 11 ma h cm 2 for stable Li-stripping/ plating cycle. When Li is completely stripped from Li/CNT (de-li/cnt; Figure 3a), the electrode thickness decreases. Such thickness change of the CNT paper between the lithiated and delithiated states was observed in the Li/CNT electrodes after 100 times of cycling at current density of 2 ma cm 2 (Figure S6, Supporting Information), and this phenomenon was also observed in the Li/CNT electrodes cycled at higher current densities of 5 and 10 ma cm 2 (Figures S7 and S8, Supporting Information). These results indicate a reversible thickness change of the CNT paper during the Li-stripping/plating cycle. The thickness difference between the Li/CNT and de-li/cnt is 27.6 µm at the Li loading of 10 ma h cm 2 (Figure 3a,b), which is significantly lower than the thickness change of 48.5 µm for a Li metal foil. Stiff frameworks made of both carbon and metal have been reported for Li anodes with the aim to reduce volume expansion of the composite electrode, which, in turn, is expected to prevent stress building inside the cell. However, Li-stripping invariably leaves voids in the anode either inside a stiff framework or in the space between the anode surface and the separator, like, for instance, in a Li metal foil (Figure 3b). [6,7] These voids need to be filled with electrolyte for subsequent Li-plating. In our case, the unique structure of Li/CNT enables a) I=1mAcm Capacity of Li/CNT LFP CE of Li/CNT LFP 80 1 Capacity of Li foil LFP CE of Li foil LFP Cyclic Number (n) b) 4 c) 600 I=1mAcm -2 C area (ma hcm -2 ) Voltage (V) 3 2 Li/CNT LFP Li foil LFP C area (ma hcm -2 ) Hysteresis (mv) Li/CNT LFP Li foil LFP Cyclic Number (n) Figure 5. a) Cycling performance of Li/CNT and Li-foil anodes in full cells with LiFePO 4 cathode at current density of 1 ma cm 2. b) Charge/discharge curves of the full cells at the 150th cycle. c) Voltage hysteresis between the plateaus of the charge and discharge curves of the full cells. CE (%) (5 of 7)

6 steady Li-stripping/plating for 1000 cycles at ultrahigh areal and gravimetric capacities (10 ma h cm 2 and 2830 ma h g 1, respectively), with 90.9% Li utilization at the current density of 10 ma cm 2. The results shown in Figure 3 indicate that the robust and expandable network of the CNT paper is responsible for the excellent electrochemical performance of the Li/CNT. To further study the role of the mechanical property of the CNT paper, we prepared a CNT paper with a coating of amorphous carbon (AC content of 5 wt%, named as AC/CNT paper) using glucose as the precursor (see the Experimental Section in the Supporting Information for more details). According to previous reports, the CNT should slide over each other during the expansion of CNT paper, especially at the junction of crossstacked CNTs because of the weak Van der Waals interaction, and amorphous carbon coating can lock such junction, thus limiting the slide of CNT and the expansion of CNT paper. [33,34] The successful coating of AC on the CNT paper is supported by the broad D and 2D bands, as well as the presence of additional shoulder peaks on both sides of the D band in the Raman spectrum of the AC/CNT paper (blue line, Figure 4a). [35,36] Moreover, the transmission electron microscopy (TEM) image in Figure 4b shows an AC layer of 2 nm thickness on the MWCNT surface, which is not observed in the un-coated CNT (Figure 1d). The series and charge transfer resistances obtained by fitting the electrochemical impedance spectroscopy data of the AC/CNT paper (Figure S9, Supporting Information) increase by 31% and 38%, respectively, comparing to those of the CNT paper. The tensile stress strain curves of CNT and AC/CNT papers along the plane of the paper were recorded to compare the mechanical properties (Figure 4c). The CNT paper presents a Young s modulus of 520 MPa and a tensile strength of 27 MPa, which are typical values of CNT paper. [37] Whereas, the Young s modulus (210 MPa) and tensile strength (13 MPa) of AC/CNT paper are much lower. Especially, the tensile stress strain curve of the CNT paper shows a long elastic deformation zone as marked in Figure 4c, which is absent in the curve of the AC/CNT paper. These results indicate that the AC/CNT paper is relatively fragile than CNT paper. We compared the CE of Li/ CNT and Li/AC/CNT in Figure 4d. The CE of the electrodes was measured at a current density of 1 ma cm 2 with Li loading of 5 ma h cm 2 in the range of Li-stripping potential V. The Li/CNT paper has a CE of >97.5% for 100 cycles. However, intense fluctuation of the CE is observed after 50 cycles for Li/AC/CNT. A decrease in CE in the presence of non-sp 2 carbon has been observed in previous studies on AC-coated graphite anodes for LIB applications and has been attributed to the increased rate of electrolyte decomposition during SEI formation. [38] However, in Li-plating, the SEI is primarily formed at the Li/electrolyte interface, which means that the greater electrolyte decomposition at the AC/electrolyte interface is not likely to be the only reason for the observed decrease in CE. Since the AC/CNT is fragile and does not efficiently resist tensile deformation (Figure 4c), the Li/AC/CNT peels off at moderate Li loading (Figure S10, Supporting Information), leading to dead Li, which causes significant fluctuation of CE. Since the Li-plating/stripping cycle inevitably results in a large volume change of the Li metal and builds up tensile stress inside the conductive framework of the Li/scaffold composite, a rigid or flexible framework that is strong enough to withstand the tensile stress induced by Li metal volume change is required for Li anodes. In this regard, we note that an accurate measurement of the Young s modulus and tensile strength along the vector in the plane of the CNT paper is still challenging and the mechanical properties of the CNT paper are anisotropic. The cycle life of Li-metal anodes paired with LiFePO 4 (LFP) can be a good indicator of the possible practical application of the Li/CNT anode. As shown in Figure 5a, the full cell assembled with Li/CNT anode and LFP cathode (Li/CNT LFP) delivers an areal capacity of 2.44 ma h cm 2 with a CE of 97.7% at the first cycle, and maintained the areal capacity of 2.26 ma h cm 2 at the 150th cycle, indicating a low capacity decay rate of 0.05% per cycle. The initial areal capacity of the Li-foil LFP cell (2.52 ma h cm 2 ) is slightly higher than that of the Li/CNT LFP; however, it decays fast to 1.79 ma h cm 2 at the 150th cycle, corresponding to a capacity decay rate of 0.19% per cycle, which is about four times of the cell with Li/ CNT as the anode. We note that the initial capacity of the Li/ CNT LFP cell is 3% lower than that of the Li-foil LFP cell, which is due to the variation of the areal capacity of the LFP disks. Moreover, the Li/CNT LFP cell shows a lower voltage hysteresis (Figure 5c) in the charge/discharge profiles than Li-foil LFP, which can be attributed to the highly conductive CNT paper as the current collector. In addition, the voltage hysteresis of the Li/CNT LFP cell increases by 28% from 0.20 V, yet the Li-foil LFP cell shows a voltage hysteresis increase of 46% from 0.30 V after 150 cycles. The much smaller change of the voltage hysteresis of the Li/CNT LFP cell is consistent with the better cycling stability of the Li/CNT LFP cell. The cycling performance was carried out in ether-based electrolyte, which is known to be a better electrolyte for Li-metal anodes owing to the more flexible SEI formed at the Li metal/electrolyte interface to accommodate the surface fluctuation during charging/ discharging. [39] Nevertheless, the Li/CNT outperformed Li-foil anode in the full cell test. In summary, we have demonstrated the application of CNT paper, an assembly of multiwall CNT, to be a promising scaffold to accommodate Li metal for anode of Li-based batteries. Ultrahigh areal and gravimetric capacities of 10 ma h cm 2 and 2830 ma h g 1, respectively, which can be maintained for 1000 cycles at 10 ma cm 2, were obtained and Li dendriteinduced failure was avoided. Owing to the moderate SSA and high electrical conductivity of the CNT paper, the local current density was decreased, as a result of which the Li/CNT electrode presented low voltage hysteresis with inhibited Li dendrite formation and a long cycling life. The CNT paper, which is lightweight, yet robust and expandable, was able to withstand the huge volume change and consequently retained the microcircuit inside the electrode intact during Li stripping/plating cycles. Thus, the Li/CNT electrode presented stable voltage hysteresis at high areal and gravimetric capacities with high Li utilization. Based on these observations, we recommend that optimized scaffolds for Li-metal anodes, either rigid or flexible, have to be strong enough to withstand the unavoidable volume expansion of Li metal during Li-plating. We expect that our work will provide a new vision to the development of practical rechargeable Li metal (6 of 7)

7 batteries and can be extended to other similar metal-based energy-storage systems. Supporting Information Supporting Information is available from the Wiley Online Library or from the author. Acknowledgements The authors appreciate funding support from the National Natural Science Foundation of China (Grant Nos , , and ) and the Fundamental Research Funds for the Central Universities (Grant No. WK ). The authors also acknowledge support from the 100 Talents Program of the Chinese Academy of Sciences. Conflict of Interest The authors declare no conflict of interest. Keywords carbon nanotube paper, lithium-metal anodes, robust structure, ultrahigh capacity Received: February 7, 2018 Revised: May 2, 2018 Published online: June 19, 2018 [1] J. B. Goodenough, Energy Environ. Sci. 2014, 7, 14. [2] D. Larcher, J.-M. Tarascon, Nat. Chem. 2015, 7, 19. [3] O. Schmidt, A. Hawkes, A. Gambhir, I. Staffell, Nat. Energy 2017, 2, [4] J. W. Choi, D. Aurbach, Nat. Rev. Mater. 2016, 1, [5] H. Kim, G. Jeong, Y.-U. Kim, J.-H. Kim, C.-M. Park, H.-J. Sohn, Chem. Soc. Rev. 2013, 42, [6] P. Albertus, S. Babinec, S. Litzelman, A. Newman, Nat. Energy 2018, 3, 16. [7] Y. Sun, N. Liu, Y. Cui, Nat. Energy 2016, 1, [8] D. Lin, Y. Liu, Y. Cui, Nat. Nanotechnol. 2017, 12, 194. [9] X.-B. Cheng, R. Zhang, C.-Z. Zhao, Q. Zhang, Chem. Rev. 2017, 117, [10] B. Zhou, L. Guo, Y. Zhang, J. Wang, L. Ma, W. H. Zhang, Z. Fu, Z. Peng, Adv. Mater. 2017, 29, [11] J. Zheng, M. H. Engelhard, D. Mei, S. Jiao, B. J. Polzin, J.-G. Zhang, W. Xu, Nat. Energy 2017, 2, [12] G. Zheng, S. W. Lee, Z. Liang, H.-W. Lee, K. Yan, H. Yao, H. Wang, W. Li, S. Chu, Y. Cui, Nat. Nanotechnol. 2014, 9, 618. [13] N. W. Li, Y. X. Yin, C. P. Yang, Y. G. Guo, Adv. Mater. 2016, 28, [14] J. Luo, C. C. Fang, N. L. Wu, Adv. Energy Mater. 2018, 8, [15] Y. Liu, Q. Liu, L. Xin, Y. Liu, F. Yang, E. A. Stach, J. Xie, Nat. Energy 2017, 2, [16] Q. Xu, Y. Yang, H. Shao, Electrochim. Acta 2016, 212, 758. [17] F. Ding, W. Xu, G. L. Graff, J. Zhang, M. L. Sushko, X. Chen, Y. Shao, M. H. Engelhard, Z. Nie, J. Xiao, J. Am. Chem. Soc. 2013, 135, [18] Q. Li, S. Zhu, Y. Lu, Adv. Funct. Mater. 2017, 27, [19] C.-P. Yang, Y.-X. Yin, S.-F. Zhang, N.-W. Li, Y.-G. Guo, Nat. Commun. 2015, 6, [20] T. T. Zuo, X. W. Wu, C. P. Yang, Y. X. Yin, H. Ye, N. W. Li, Y. G. Guo, Adv. Mater. 2017, 29, [21] J.-N. Chazalviel, Phys. Rev. A 1990, 42, [22] R. Miao, J. Yang, X. Feng, H. Jia, J. Wang, Y. Nuli, J. Power Source 2014, 271, 291. [23] J. Hassoun, B. Scrosati, Adv. Mater. 2010, 22, [24] C.-X. Zu, H. Li, Energy Environ. Sci. 2011, 4, [25] S. Jin, Z. Sun, Y. Guo, Z. Qi, C. Guo, X. Kong, Y. Zhu, H. Ji, Adv. Mater. 2017, 29, [26] Z. Liang, D. Lin, J. Zhao, Z. Lu, Y. Liu, C. Liu, Y. Lu, H. Wang, K. Yan, X. Tao, Proc. Natl. Acad. Sci. USA 2016, 113, [27] D. Lin, Y. Liu, Z. Liang, H.-W. Lee, J. Sun, H. Wang, K. Yan, J. Xie, Y. Cui, Nat. Nanotechnol. 2016, 11, 626. [28] C. Jin, O. Sheng, J. Luo, H. Yuan, C. Fang, W. Zhang, H. Huang, Y. Gan, Y. Xia, C. Liang, Nano Energy 2017, 37, 177. [29] Q. Yun, Y. B. He, W. Lv, Y. Zhao, B. Li, F. Kang, Q. H. Yang, Adv. Mater. 2016, 28, [30] Y. Gao, R. Yi, Y. C. Li, J. Song, S. Chen, Q. Huang, T. E. Mallouk, D. Wang, J. Am. Chem. Soc. 2017, 139, [31] Y. Zhang, W. Luo, C. Wang, Y. Li, C. Chen, J. Song, J. Dai, E. M. Hitz, S. Xu, C. Yang, Proc. Natl. Acad. Sci. USA 2017, 114, [32] Y. Liu, D. Lin, Z. Liang, J. Zhao, K. Yan, Y. Cui, Nat. Commun. 2016, 7, [33] W. Zhao, Y. Li, S. Wang, X. He, Y. Shang, Q. Peng, C. Wang, S. Du, X. Gui, Y. Yang, Q. Yuan, E. Shi, S. Wu, W. Xu, A. Cao, Carbon 2014, 76, 19. [34] D. McKenzie, D. Muller, B. Pailthorpe, Phys. Rev. Lett. 1991, 67, 773. [35] A. C. Ferrari, J. Robertson, Phys. Rev. B 2000, 61, [36] M. Tamor, W. Vassell, J. Appl. Phys. 1994, 76, [37] G. Xu, Q. Zhang, W. Zhou, J. Huang, F. Wei, Appl. Phys. A 2008, 92, 531. [38] V. Sharova, A. Moretti, G. A. Giffin, D. V. Carvalho, S. Passerini, C 2017, 3, 22. [39] D. Aurbach, O. Youngman, Y. Gofer, A. Meitav, Electrochim. Acta 1990, 35, (7 of 7)