Carbon nanotubes-bridged molybdenum trioxide nanosheets as high performance anode for lithium ion batteries

The search for novel nanomaterials driving the development of high-performance electrodes in lithium ion batteries (LIBs) is at the cutting edge of research in the field of energy storage. Here, we report on the synthesis of single wall carbon nanotube (SWNT)-bridged molybdenum trioxide (MoO3) nanosheets as anode material for LIBs. We exploit liquid phase exfoliation of layered MoO3 crystallites to produce multilayer MoO3 nanosheets dispersed in isopropanol, which are then mixed with solution processed SWNTs in the same solvent. The addition of SWNTs to the MoO3 nanosheets provides the conductive framework for electron transport, as well as a bridge structure, which buffers the volume expansion upon lithiation/de-lithiation. We demonstrate that the hybrid SWNT-bridged MoO3 structure is beneficial for both the mechanical stability and the electrochemical characteristics of the anodes leading to a specific capacity of 865 mAh g−1 at 100 mA g−1 after 100 cycles, with a columbic efficiency approaching 100% and a capacity fading of 0.02% per cycle. The low-cost, non-toxic, binder-free hybrid MoO3/SWNT here developed represents a step forward for the applicability of exfoliated MoO3 in LIB anodes, delivering high energy and power densities as well as long lifetime.


Introduction
The worldwide lithium-ion batteries (LIBs) market is increasing with a relevant growth rate of more than 20% per year [1], with an estimated global market reaching $130 billion in 2025 [2]. Today, LIBs find the main application in portable electronic devices [3], while the increasing demand in the transport sector, i.e. full electric vehicles (EVs) and hybrid electric vehicles (HEVs) [4], will drive the growth of LIBs market towards automotive and industrial applications [1]. Moreover, energy storage for grid balancing and micro-grids, e.g. residential storage battery in conjunction with photovoltaic systems, will also play a role for the growth of LIBs market [5]. Current LIBs [6], based on the capture and release of lithium ions, are usually composed by an intercalated lithium compound cathode (e.g. LiCoO 2 or LiFePO 4 ), a graphitic-based anode and an electrolyte [6,7], yielding a theoretical specific energy density of 387 Wh kg −1 (with LiCoO 2 as cathode) [7] and a measured energy density of 120-150 Wh kg −1 [7].
Crucial to the performance of these rechargeable batteries is the specific capacity (or gravimetric capacity, i.e. total ampere-hours -Ah-available when the battery is discharged at a certain current density, per unit weight) to evaluate the capability of lithium ions storage in the active materials [4]. In this regard, LIBs are still limited by the specific capacity of the commercial active materials, e.g. graphite anode with a theoretical specific capacity of 372 mAh g −1 [8]. For this reason, a current research effort focuses on the exploitation of alternative anodic materials with high theoretical specific capacity such as Si (3579 mAh g −1 ) [9], Sn (994 mAh g −1 ) [10], and SnO 2 (782 mAh g −1 ) [11]. However, the full development of such high-performance anode materials has been hindered by the significant capacity fading owing to the large volume contraction/expansion (100-300% with respect to the initial volume) [12] during chargedischarge cycling [13].
Carbon nanotubes-bridged molybdenum trioxide nanosheets as high performance anode for lithium ion batteries 2 H Sun et al Another promising strategy that is emerging in the last few years is to replace graphite with other layered anode materials, i.e. transition metal oxides [14][15][16][17][18] and graphene-based materials [19][20][21][22][23][24][25][26][27][28][29][30], which are expected to improve the energy storage of LIBs. In this context, graphene and its derivatives, i.e. reduced graphene oxide (RGO) [19,20] and graphene nanoplatelets (GNPs) [21], have been already largely explored for the realization of LIB anodes [22]. The aforementioned graphene-based materials display larger theoretical specific capacity (i.e. 744 mAh g −1 assuming lithium adsorbed on both sides of graphene to form Li 2 C 6 ) [23] when compared to graphite, electrochemical and thermal stability within the LIBs operational temperature range (−20/60 °C) [24], offering additional properties of flexibility and/or stretchability [25]. However, they also suffer severe limitations. In fact, in single-layer graphene (SLG), Li storage is not thermodynamically favoured, with low value of Li uptake [26][27][28]. So an hypothetical SLG anode will not be an efficient solution. Moreover, the morph ology (i.e. lateral size and thickness) of the flakes strongly influences the Li + storage capability of few-(FLG) and multi-layer graphene (MLG) flakes [29]. Indeed more capacity is delivered at high potentials upon flake size reduction, resulting in reduced and non-constant voltage output from the battery, a detrimental factor on the voltage efficiency [24]. Finally, RGO and GNPs experience critical issues for LIB applications. In fact, RGO show large irreversible capacity [19,20] and voltage hysteresis between lithiation and de-lithiation [30], while GNPs have not yet demonstrated considerable gain in maximum specific capacity with respect to graphite [21,25].
Recently, liquid phase exfoliation (LPE) [76,77] has attracted increasing attention to obtain nanoscale graphene and other 2D layered materials, which have strong covalent intra-layer bonding [78], but weak out-of-plane inter-layer interactions [79][80][81]. Henceforth, it is possible to peel off individual layers from the parent bulk crystal by supplying sufficient energy [82]. Thus, this process has become a widely used method for the production of a range of 2D flakes [74] (e.g. MoO 3 ) [83] from their parent bulk crystals [84]. During the LPE process, the inter-sheet forces are broken by the input of either shear or ultrasonic energy in the presence of a stabilizing liquid [73,85]. The resultant exfoliated flakes are generally defect-free and unfunctionalized [86,87].
In this paper, we demonstrate that multi-layer MoO 3 nanosheets produced by LPE of layered MoO 3 crystallites combined with single wall carbon nanotubes (SWNTs) can be used as an active binderfree material for LIBs. The hybrid anodes are produced by mixing multilayer MoO 3 nanosheets with solution processed SWNTs. Both the MoO 3 exfoliation process and the SWNTs de-bundling are carried out in isopropanol, allowing a simple deposition onto the copper substrate. The designed binder-free solution processed hybrid MoO 3 /SWNT anode displays a specific capacity of 865 mAh g −1 at 100 mA g −1 after 100 cycles, with a columbic efficiency of 99.7%. The SWNTs addition determines a network structure with the MoO 3 providing (1) long channels for electronic charge transport; (2) an active anode material, instead of polymeric binder, offering extra capacity for Li ions storage: (3) a buffer frame in the electrode, which reduce the capacity fading caused by the volume expansion of MoO 3 flakes during the lithiation process. To further confirm the essential roles of SWNTs, we also tested multilayers MoO 3 combined with carbon black (CB) nanoparti-cles, which are not able to create the network structure seen with SWNTs, yielding a significant capacity fading of the resulting battery. These results set the basis for the exploitation of exfoliated 2D MoO 3 sheets as anodic materials in Li-ion batteries.   and SWNTs are purchased from Carbon Solutions inc. All materials are used without any further purification.

LPE MoO 3 nanosheets preparation
Molybdenum trioxide powder (240 mg) is added to 2-propanol (IPA) (80 ml) in a 100 ml open top, flat bottomed beaker. The dispersion is ultrasonicated using a horn probe sonic tip (VibraCell CVX, 750 W, 25% amplitude) for 5 h. The sonic tip is pulsed for 6 s on and 2 s off to avoid damage to the processor and reduce any solvent heating. To minimize heating effects, an external cooling system circulated cooled water (5 °C) around the beaker during ultrasonication. To remove any un-exfoliated material the ultrasonicated dispersion is filled in glass vials (~30 ml) and ultracentrifuged at 1 krpm (~100 g) for 30 min. The supernatant is decanted (~20 ml) and further ultracentrifuged at 5 krpm for 105 min to remove small flakes. The supernatant is decanted (containing small flakes) and discarded while the sediment is redispersed in fresh IPA.

SWNTs dispersion preparation
P3 SWNTs are dispersed in IPA at a concentration of 0.1 gl −1 and ultra sonicated in both a horn sonic probe and a sonic bath to achieve a completely homogeneous dispersion. The procedure involves horn probe ultrasonication (30 min) followed by 1 h in a sonic bath and an additional 30 min in the horn probe tip.

CB dispersion preparation
Carbon black is dispersed in IPA at a known concentration (0.25 gl −1 ) and bath ultrasonicated for 2 h.

MoO 3 /SWNT/CB hybrid
The as produced MoO 3 , SWNTs and CB dispersions are then mixed, without centrifugation, to form hybrid structures of known wt%. Accurate weighing of an alumina membrane (pore size 25 nm) before and after filtration of MoO 3 dispersion allowed determining the concentration.

Characterization of SWNTs, CB, MoO 3 flakes and MoO 3 /SWNT hybrid structure
Transmission electron microscopy (TEM) images of MoO 3 , SWNT, CB and MoO 3 /SWNT hybrid are taken with a JOEL JEM 1011 transmission electron microscope, operated at 100 kV. The samples are then diluted 1:10 with pure IPA associated with 10 min ultra-sonication. 100 µl of the resulting dispersions are drop-cast at room temperature onto carbon coated copper TEM grids (300 mesh), and subsequently dried under vacuum overnight. Raman measurements are carried out with a Renishaw 1000 using a 50× objective, a laser with a wavelength of 532 nm and an incident power of ~1 mW. The samples are drop-cast onto a Si wafer (LDB Technologies Ltd), with 300 nm thermally grown SiO 2 . Electrochemical tests. The cyclic voltammetries (CVs) are performed at a scan rate of 50 µV s −1 between 1 V and 5 mV versus Li + /Li with a Biologic, MPG2 potentiostat/galvanostat. The electrochemical impedance spectroscopies are collected with a VMP3 (BioLogic). All the electrochemical measurements are performed at room temperature. Constant current charge/discharge galvanostatic cycles are performed for the as prepared binder-free anodes in half-cell and in full battery configuration at a different current density, using a battery analyzer (MTI, BST8-WA). The charge/discharge cycles are performed at different rates at room temperature.

Properties of MoO 3 , MoO 3 /CB and MoO 3 / SWNT dispersions
The synthesis of the MoO 3 /SWNT hybrid anode for LIBs starts with the solution processing of the two nanomaterials. The MoO 3 flakes in dispersions are obtained by LPE of pristine MoO 3 , while the SWNTs are prepared in IPA by tip ultrasonication, see Experimental for detailed information about their preparation, as well as for the processing of the CB nanoparticles.
The morphology of the as-produced MoO 3 and SWNTs are analysed by TEM. Figure 1(a) shows MoO 3 flakes with lateral sizes ranging from 50 to 300 nm, while figure 1(b) shows that the SWNTs are in bundles of ~10 nm in diameter, forming a spider web-like network onto the TEM grid. The TEM image of the hybrid MoO 3 /SWNTs sample (see figure 1(c)), clearly shows the bundles of SWNTs acting as bridges to connect isolated MoO 3 flakes, forming an interconnected network in the hybrid MoO 3 /SWNTs. The TEM of CB, reported in the supplementary information -S.I.-(see figure S3(b) (stacks.iop.org/TDM/5/015024/mmedia)), shows a particle size distribution of ~50 nm. We further carried out the characterization of the morphological properties of the dispersed mat erials by Raman spectroscopy, which is a fast and non-destructive technique commonly used to identify type of defects, doping, disorder and chemical modifications of nanomaterials [88]. In figure 1(d) we show the Raman spectra of MoO 3 , in blue, and SWNTs, in red. The full spectroscopy characterization, i.e. optical absorption and Raman, of the as-prepared MoO 3 , SWNTs and CB nanoparticles are discussed in the S.I.
The as-prepared samples are then exploited for the realization of electrodes, i.e. anodes, for LIBs. In particular, solution processed MoO 3 flakes and the MoO 3 /SWNT hybrid (9:1 ratio) is deposited onto Cu substrates. A reference sample, i.e. MoO 3 mixed with 10% CB, is also prepared by using the same process. The mass loading of MoO 3

Electrochemical properties of MoO 3 , MoO 3 /CB and MoO 3 /SWNT electrodes
In order to fully understand the optimal SWNTs/MoO 3 weight ratio on the electrochemical performances of the MoO 3 /SWNT hybrid anodes, we prepared other two samples with different SWNTs/MoO 3 weight ratio of 2:8 and 3:7, respectively, see Experimental. The Raman and SEM characterization of the electrodes are presented and in depth discussed in the S.I. As shown in figure S4, the three MoO 3 /SWNT hybrid anodes show a homogenous coverage of SWNTs and MoO 3 nanoflakes onto the Cu substrates.
The CVs ( figure 3(a)) of the three samples, i.e. MoO 3 , MoO 3 /CB and MoO 3 /SWNT are collected at a scan rate of 50 µV s −1 starting from 5 mV versus Li + / Li potential, to cover the lithiation processes in both MoO 3 and SWNT [89]. In the first reduction sweep, the MoO 3 exhibits two peaks at 2.3 and 2.7 V, which can be linked with the insertion of Li ions into the interlayers of the MoO 3 structure to form Li x MoO 3 , and another peak at 0.4 V, which corresponds to the conversion reaction of Li x MoO 3 into Mo and Li 2 O [42,54,93].
The two processes determine, the accommodation of six Li ions in each MoO 3 , reaching theoretical  specific capacity of 1117 mAh g −1 [56,57], as summarized by equations (1) and (2) [54,93], (2) In the reverse oxidation process, metallic molybdenum is converted into amorphous MoO 2 , in the 1.0 V-2.2 V range [90]. From the 2nd cycle onward, a clear shift is observed in the conversion reaction peak at 0.4 V, which is consistent with the formation, upon oxidation, of MoO 2 with lower reactivity with respect to MoO 3 [97]. Furthermore, in the second cycle a new peak, at 1.5 V, appears which can be assigned to the lithium insertion into amorphous MoO 2 [91,97]. For the MoO 3 and the hybrid MoO 3 /CB samples, the reduction peaks at 1.5 V and 0.4 V rapidly disappear during the following 8 cycles. On the contrary, in the case of MoO 3 /SWNT sample the intensity of these two peaks is maintained from cycle 2 to cycle 10, leading to a remarkable improvement on its electrochemical stability with respect to the other two samples. Figure 3(b) shows the charge-discharge voltage profiles of bulk MoO 3 at 100 mA g −1 , in order to get a complete electrochemical response for the Li ion transfer [92], during the lithiation/de-lithiation process at the anode. In the 1st charge (lithiation) process, two plateaus at 2.3 V and 0.4 V are observed in all the three samples. These two plateaus have already been attributed to the formation of Li x MoO 3 and its following conversion reaction into metallic Mo and LiO 2 , respectively [42,54,93]. These reactions have also contrib ution on the large initial specific capacity obtained in the three samples, i.e. 864 mAh g −1 for MoO 3 , 1332 mAh g −1 for MoO 3 /CB, and 1357 mAh g −1 for MoO 3 /SWNT. The higher initial specific capacity shown by both the MoO 3 /SWNT and MoO 3 /CB samples, compared with the MoO 3 sample, can be associated to the enhanced electrical conductivity in the hybrid electrodes due to the presence of the carbon nanomat erials [65,93]. On the contrary, significant differences of the three samples are shown by the specific capacities obtained at the 1st discharge (de-lithiation) process. In fact, initial specific capacities of 481, 625 and 962 mAh g −1 are achieved for the MoO 3 , MoO 3 /CB, and MoO 3 /SWNT samples, respectively. For all the 3 samples, the capacity drop between the 1st charge and discharge processes (see figure 4(a)), is caused by the combination of several irreversible processes, including: (1) the solid electrolyte interphase (SEI) formation [62]; (2) the structural modulation during Li + insertion/extraction into the inter-layers and intra-layers of MoO 3 [94]; (3) the conductivity loss caused by the electrode pulverization upon lithiation/de-lithiation [95].
For the MoO 3 and MoO 3 /CB samples, we observed a coulombic efficiency (the ratio between discharge and charge capacities) at the 1st cycle as low as 55.7% and 46.9%, respectively, see figure 4(a). Moreover, both samples show capacity fading upon cycling, which is the main drawback of MoO 3 anodes due to the pulverization of the electrode [65], with capacity loss of 84% and 64%, respectively, after 60 cycles. Alternatively, the MoO 3 /SWNT sample shows a coulombic efficiency as high as 70.9% at the 1st cycle, a value which is significantly enhanced compared to the MoO 3 and MoO 3 / CB samples. Additionally, the MoO 3 /SWNT sample shows a tangible improvement on the stability of the electrochemical performance, delivering reversible capacity of 950 mAh g −1 at 50th cycle, with only a 1.2% capacity loss from the 1st cycle.
In order to further understand the different electrochemical performance of the three MoO 3 -based electrodes, we carried out electrochemistry impedance spectroscopy (EIS) for all the three samples at charged state, after 60 cycles. The Nyquist plots of the electrodes are presenting a semi-circle at high-to-medium frequency [60], demonstrating the different interface resistances in the three samples. The interface resistance occurring at high frequency is associated with phenomena such as Li + ion diffusion through the SEI film and/or in the active material, and the contact layer between the electrode and current collector [96][97][98].
As obtained from figure 4(b), the interface resistance of the MoO 3 sample is ~160 Ω, which is significantly reduced to ~80 Ω for the MoO 3 /CB sample. The MoO 3 /SWNT hybrid structure gives the lowest value of interface resistance, i.e. ~40 Ω, which is one fourth and one half with respect to the ones shown by the MoO 3 and MoO 3 /CB-based electrodes, respectively. The reduction of the interface resistance upon the addition of carbon additives, especially SWNTs, compared with the MoO 3 , might be attributed to the different structural morphology of the electrodes after lithiation/de-lithiation processes [60,61].
Thus, in order to understand the relation between the electrochemistry performance of the three samples and their structural morphology after chargedischarge cycles, we carried out post-mortem SEM measurements on MoO 3 , MoO 3 /CB and MoO 3 / SWNT electrodes after 60 charge/discharge cycles. As shown in figure 5(a), the MoO 3 electrode clearly presents cracks and fractures with width of 200-400 nm, likely caused by the volume change during the charge/ discharge cycles. These cracks determine a drop in the electrical conductivity, with consequent capacity fading [99,100], as clearly presented in figure 4. As shown in figure 5(b), large cracks over 1 µm are observed in the MoO 3 /CB electrode as well. Even if, compared to free MoO 3 , the presence of CB seems able to furnish better electrical conductivity during the first cycles MoO 3 /CB electrodes still suffer a remarkable capacity fading upon cycling. This is likely due to the inability of CB to keep the anode material in continuous contact with the current collector [102,103]. Although the MoO 3 /SWNT sample shows cracks after 60 cycles, the cracks are much narrower with respect to the ones presented by the MoO 3 and MoO 3 /CB electrodes.
Moreover, the carbon network of nanotubes ensures high electrical conductivity upon the expansion/contraction processes of MoO 3 . This conductive framework is therefore beneficial for both mechanical stability [101] and the specific capacity of the anodes.

Electrochemical properties of MoO 3 /SWNT electrodes with different mix ratio
From the obtained results, it is clear that the SWNTs addition (10% with respect to the MoO 3 flakes) is beneficial for the electrochemical properties of the electrodes. Thus, in order to further investigate the contribution of the SWNTs addition in the MoO 3 / SWNT hybrid anode, we designed other two electrodes, adding SWNTs to the MoO 3 flakes at a weight ratio of 20% and 30%, with mass loading of 0.74 mg and 0.76 mg, respectively. As shown in figure 6(a), while the percentage of SWNTs in the MoO 3 /SWNT electrodes rises from 10% to 30%, the initial capacities of the three samples reach 1357, 1161 and 1044 mAh g −1 , with corresponding discharge capacity at the first cycle of 927, 675, and 566 mAh g −1 , respectively. The specific capacity and coulombic efficiency (see figure 6(a)) demonstrate that all the hybrid electrodes with different mixed ratios show remarkable stable cyclability up to 50 cycles, if compared with the MoO 3 anode.
The post-mortem SEM images of the three MoO 3 / SWNT electrodes shown in figures S5(b)-(d) clearly demonstrate that the SWNTs in the MoO 3 /SWNT electrode create a network between the cracked 'islands' following the MoO 3 volume change during charge/discharge cycles [65]. Notably, the width of the cracks is reducing with the percentage increase of SWNT in the MoO 3 /SWNT hybrids. A possible explanation could be linked with the fact that the increasing amount of SWNTs, as buffer between the MoO 3 flakes, can efficiently attenuate the volume change during charge/discharge cycles, reducing the mechanical degradation of the electrodes.
The EIS results of the 3 samples shown in figure 6(b), demonstrates how the higher is the percent age of SWNTs in the hybrid MoO 3 /SWNT electrodes, the lower is their charge transfer resistance. In fact, a charge transfer resistances of ~40 Ω, ~30 Ω and ~17 Ω have been obtained for the sample with 10%, 20% and 30% of SWNTs with respect to the MoO 3 flakes, respectively. However, although higher percentage of SWNTs (i.e. 20-30%) in the hybrid structure can provide better electrical conductivity (i.e. charge transfer resistances of 30 Ω and 17 Ω for the electrodes containing 20% and 30% of SWNTs with respect to the MoO 3 flakes) this is not directly associated to an increase of the electrode specific capacity. In fact, the increasing percentage of SWNT has determined a tangible decrease of the specific capacity with respect to the total loading of MoO 3 /SWNT hybrid electrodes. This could be linked with the high irreversible capacity that affects CNTs-based anode for LIBs [102]. In fact, the irreversible capacity increases from 32% for the 10% MoO 3 /SWNTs sample to 46% in the case of 30% MoO 3 /SWNTs one. Moreover, the 10% MoO 3 / SWNTs sample shows the highest capacity retention (71.6% after 50 cycles) over charge/discharge cycles, obtained by dividing the charge-capacity to the initial capacity (see figure 4(a)), amongst the electrodes, i.e. the hybrids MoO 3 /SWNTs and the MoO 3 one.
Moreover, we have also calculated the specific capacity of each electrode, as shown in figure 6(c), labeled by different SWNTs content from 0 to 30%. The specific capacities are calculated using the mass loading of MoO 3 and MoO 3 /SWNT, respectively. In both cases, the 10% SWNT sample reaches the highest specific capacity of 1028 mAhg −1 MoO3 (926 mAhg −1 MoO3/SWNT ), see figure 6(d), which represent the 92% of the theoretical specific capacity of MoO 3 [55,56]. As summarized in table S1 (see S.I.), the performance of our MoO 3 /SWNT binder-free anode favorably compare with state of the art MoO 3 -based LIBs [39,42,44,49,56,60,65,72,103,104]. The reported electrochemical analysis indicates that the addition of 10% SWNT in the hybrid structure with MoO 3 flakes represents the best compromise in term of mechanical and electrochemical properties of the as-produced anodes.

Conclusion
By exploiting the combination of multilayer MoO 3 nanosheets, produced by LPE of layered MoO 3 crystallites, with solution processed SWNTs we demonstrated a high performance binder-free MoO 3 /SWNTs hybrid anode for LIBs. Contrary to CB nanoparticles, the SWNTs addition determines a network structure with the MoO 3 , which is beneficial for the mechanical and (electro)chemical performances of the as-produced anode by providing (1) long channels for electronic charge transport; (2) an active anode material, instead of polymeric binder, offering extra capacity for Li ions storage: (3) a buffer frame in the electrode, which reduce the capacity fading caused by the volume expansion of MoO 3 flakes during the lithiation process. The designed binder-free solution processed hybrid MoO 3 /SWNT (90:10) anode has demonstrated a specific capacity of 865 mAh g −1 at 100 mA g −1 after 100 cycles, with a columbic efficiency of 99.7% and a capacity fading of 0.02% per cycle. We believe that the low-cost, nontoxic, binder-free hybrid MoO 3 /SWNT can boost the development of high-performance anodes for LIBs.