Mixed Molybdenum Oxides with Superior Performances as an Advanced Anode Material for Lithium-Ion Batteries

A simple and effective carbon-free strategy is carried out to prepare mixed molybdenum oxides as an advanced anode material for lithium-ion batteries. The new material shows a high specific capacity up to 930.6 mAh·g−1, long cycle-life (>200 cycles) and high rate capability. 1D and 2D solid-state NMR, as well as XRD data on lithiated sample (after discharge) show that the material is associated with both insertion/extraction and conversion reaction mechanisms for lithium storage. The well mixed molybdenum oxides at the microscale and the involvement of both mechanisms are considered as the key to the better electrochemical properties. The strategy can be applied to other transition metal oxides to enhance their performance as electrode materials.

Materials Characterization. The XRD patterns were acquired with an X-ray diffractometer (Philips Analytical X'pert Pro) at a 2θ range of 10-70°. The diffractometer was equipped with a Cu Kα radiation (λ = 1.5418 Å) and operated at 40 kV and 40 mA. The air sensitive samples were protected by Kapton films in the XRD measurement. The morphology of the samples were investigated using a field-emission scanning electron microscope (Hitachi FE-SEM S4800). X-ray photoelectron spectroscopy (XPS) was used to survey the surface compositions and the chemical states of the samples (PHI5000 VersaProbe, PerkinElmer). The hydrocarbon C 1 s peak at 284.5 eV was chosen as the reference for the binding energies. The XPS spectra were analyzed with the software XPSPeak 4.1. The ratio of the peak area of the Mo 3d 5/2 peak to the Mo 3d 3/2 peak is fixed to 1.5 while the binding energy separation is set to 3.0 eV. A Bruker Advance III 400 MHz spectrometer with a 3.2 mm Magic Angle Spinning (MAS) probe was used for acquiring the 7 Li solid-state NMR spectra at a resonant frequency of 155.46 MHz. A spin-echo and a 2D exchanged pulse sequence were used in the data acquisition. In order to select reduce the intensity of the resonance due to the SEI, a T 2 filter was used in the spin-echo NMR data acquisition. After discharging to 5 × 10 −3 V, the self-supported electrode 16 was taken out, rinsed with dimethyl carbonate (DMC) to remove the residual electrolyte, before it was dried and packed into NMR rotors in an Ar glove box, and spun at 18 or 20 kHz for NMR data collection. A 1 M LiCl aqueous solution was used to set the spectral reference to 0 ppm. Electrochemical Characterization. Electrochemical measurements were performed on coin cells (model 2032) with lithium metal foils as the reference and counter electrodes at ambient temperature. 70 wt% of as prepared samples, 20 wt% acetylene black (conducting additive), and 10 wt% polyvinylidene fluoride (binder) were blended with N-Methyl-2-pyrrolidone (NMP) to make a slurry and then spread onto a copper foil current collector. After drying at 80 °C for 6 hours, the electrode was punched and then dried under vacuum at 120 °C for 12 hours. The electrodes were then transferred to an argon-filled glove box where both the moisture and oxygen concentrations were kept below 1 ppm. The electrolyte was prepared by dissolving 1 M LiPF 6 in a mixture of ethyl methyl carbonate (EMC), ethylene carbonate (EC) and dimethyl carbonate (DMC) (1:1:1 in weight) solution while a Celgard 2325 film was used as the separator. An electrochemical station (CHI660D) was used to obtain cyclic voltammetry (CV) at ambient temperature. CV measurements were performed over a potential range of 5 × 10 −3 -3.0 V (vs. Li + /Li) at a scan rate of 2 × 10 −4 V s −1 . Galvanostatic discharge-charge measurements were performed at different rate on a battery testing system (LAND CT2001A model, Wuhan Jinnuo Electronics Co. Ltd) with a voltage window range from 5 × 10 −3 to 3.0 V (vs. Li + /Li).

Results and Discussion
The XRD patterns of different molybdenum oxide nanoparticle samples are shown in Fig. 1. All of the XRD peaks for MoO 2 nanoparticles annealed with pure N 2 can be indexed to MoO 2 (Fig. 1a, ICSD:080830), while the sample heated with a high concentration of O 2 (1% O 2 in N 2 ) in the flow is completely oxidized to MoO 3 (Fig. 1c,  Figure 2 shows the SEM images of the obtained MoO 2 , MMO and MoO 3 samples. Small agglomerates with a diameter of about 50 nm composed of smaller and round shaped nanoparticles can be observed for the MoO 2 sample, while MoO 3 are much larger in size (100 to 500 nm). The SEM image of MMO shows this sample is composed of small particles with different sizes, resembling a mixture of the above two materials. The results indicate that MMO are not uniformly oxidized, in agreement with the XRD data.
X-ray photoelectron spectroscopy (XPS) measurement was performed in order to investigate the surface valence states of Mo in MMO, and the results are shown in Fig. 3. The peaks of Mo 3d and O 1 s can be observed in the survey spectra (Fig. 3a), and the former was further examined. The two peaks at about 232.0 and 235.0 eV are characteristic of Mo 6+ , and can be attributed to Mo 6+ (3d 5/2 ) and Mo 6+ (3d 3/2 ), respectively 18 . The lower energy peak centered at 229.3 eV can be attributed to Mo 4+ (3d 5/2 ), suggesting that the Mo ions on the surface of MMO are in mixed valence states 19 . The spectrum was fitted into two doublets ascribed to Mo 6+ (3d 5/2 )/Mo 6+ (3d 3/2 ) and Mo 4+ (3d 5/2 )/Mo 4+ (3d 3/2 ) ( Table 1). The Mo 6+ /Mo 4+ ratio is determined to be 4.0 according to the line fitting results, hence the valence state of molybdenum on the surface of MMO can be determined as 5.6 18,20 .
To further investigate the overall valence state of the MMO, TGA-DSC test was carried out to monitor the weight changes as a function of temperature in the air atmosphere and the results are shown in Fig. 4. The small weight loss (about 0.3%) at the temperature lower than 150 °C can be attributed to the loss of absorbed water. The   weight increase at 300-500 °C corresponds to the oxidization of MMO (MoO x ) to MoO 3 in the air and a wide exothermic peak is associated with this process. The total weight increase is 4.30% (Fig. 4) Cyclic voltammetry (CV) was used to study the Li insertion/extraction behaviors in different samples and the results are presented in Fig. 5a. The redox peaks centered at 1.25/1.53 V and 1.48/1.77 V are attributed to the reversible monoclinic-orthorhombic-monoclinic phase transitions in the partially lithiated Li x MoO 2 according to the CV data of MoO 2 nanoparticles ( Figure s1a) and previous reports 6,19,21 , consistent with the presence of MoO 2 in MMO which was illustrated in the XRD pattern and the TG curve 6,21 . Besides these two redox pairs, the whole CV curve of MMO is similar to that of our MoO 3 nanoparticles ( Figure s1b) and literature [22][23][24] due to the high concentration of MoO 3 , as implied by XRD and XPS data 22 . There are also intense peaks below 0.5 V, which can be associated with a conversion reaction with lithium, where Mo oxides are reduced completely to metallic  Mo along with the formation of Li 2 O. The curve changes very little in the following cycles, suggesting that the partially lithiated Li x MoO 2 structure is very stable during the charge-discharge progress, thus MMO is expected to have a good cycling performance 25 . Meanwhile the difference between the first and following cycles can be attributed to irreversible reactions such as the irreversible phase transition, the decomposition of the electrolyte and the formation of a solid electrolyte interface (SEI) layer 22,26 .
To further investigate the electrochemical performance of the samples, galvanostatic experiments were performed on the coin cells. Figure 5b shows the charge/discharge plots of MMO at a current of 200 mA·g −1 . The initial discharge capacity is 1068.2 mAh·g −1 , slightly higher than the theoretical capacity of MMO (1021 mAh·g −1 ), which can be assigned to some irreversible reactions, e.g. the generation of an SEI layer 23,27 . In agreement with the CV results, two short discharge plateaus at about 1.55 and 1.26 V can be observed 13,21 . The expected charge plateaus at around 1.4 and 1.7 V are not clearly visible, which is similar to previously reported MoO 3 nanostructures 26 . The charge/discharge curves are very stable with continued cycling, suggesting good cycling performances, except that the voltage of discharge plateaus decreases and eventually the plateau can no longer be observed. Since the plateaus are usually attributed to the phase transition in the partially lithiated Li x MoO 2 , the decrease in voltage can be ascribed to the transformation of crystalline phase to amorphous-like structure 13,21 .
As shown in Fig. 5c, the electrochemical performance of MMO is much better than MoO 2 or MoO 3 . For MoO 2 , the first discharge capacity is only 670.5 mAh·g −1 while the capacity gradually increases upon cycling and attains 781.7 mAh·g −1 after 15 cycles. This activation process has been reported by other groups and may be attributed to the partial loss of crystallinity of the material during the cycling 13,28,29 , accompanied with a larger fraction of the material undergoing further conversion reaction instead of stopping at Li x MoO 2 phase 30 . Nonetheless, a rapid capacity deterioration decay can be observed once the capacity reaches the maximum value and only a capacity of 325.7 mAh·g −1 remains after 50 cycles. The initial capacity of MoO 3 is 1017.6 mAh·g −1 , however, it drops to 593.4 mAh·g −1 after only 30 cycles due to the big volume changes during cycling. Such decrease can also be observed in MMO, which has a large fraction of MoO 3 . However, the drop in capacity is much less. The capacity decreases from the first discharge capacity of 1068.2 mAh·g −1 to 842.3 mAh·g −1 after 20 cycles, corresponding to a capacity loss of 12% (calculated based on the capacity of 2 nd and 20 th cycles). After that, the capacity increases slowly and reaches 976.3 mAh·g −1 after 90 cycles. The discharge capacity of the MMO only decreases slightly to 930.6 mAh·g −1 after 200 cycles, which is equivalent to a capacity retention of 95% of the maximum value of 977 mAh·g −1 , demonstrating high capacity and excellent stability. Such high reversible capacity of more than 930.6 mAh·g −1 at 200 mA·g −1 is almost three-fold that of commercial graphite (~330 mAh·g −1 ). Although the overall profile of MMO's cycling curve is similar to that of MoO 3 , the key difference is the capacity decay is much smaller for MMO than MoO 3 . This decay, which can be seen in many reports of MoO 3 , is generally assigned to the huge volume and/or structure change during the conversion reaction 23,27 . Since the volume and structure changes are very limited in the insertion reaction 31 , the presence of stable partially lithiated Li x MoO 2 in MMO is expected to help stabilize the whole electrode and avoid the capacity loss. To further investigate the hypothesis above, the electrochemical performance of the mixed MoO 2 and MoO 3 sample was prepared by ball milling (MMM) ( Figure S2b). Although MMM is also composed of MoO 2 and MoO 3 with similar proportion (Figure S2a), the electrochemical performance is apparently worse than that of MMO. Although an initial capacity of 1050 mAh·g −1 can be obtained, which is similar to that of MMO, only a capacity of approximately 100 mAh·g −1 can retain after 100 cycles at a current rate of 200 mA·g −1 . The very rapid decay can be ascribed to the phase isolation which is expected to accelerate the pulverization process. Owing to the poor electrochemical performances of Mo 4 O 11 32 , which is the intermediate phase in the transformation from MoO 2 to MoO 3 33 , the small fraction of Mo 4 O 11 in MMO is also unlikely to be associated with the improved electrochemical properties.
The rate capability of MMO was further investigated and the results are shown in Fig. 5d. The material was tested with the potential range of 5 × 10 −3 -3.0 V at 0.1 A·g −1 , 0.2 A·g −1 , 0.5 A·g −1 , 1 A·g −1 , 2 A·g −1 , 5 A·g −1 and 0.1 A·g −1 for 10 cycles each. The specific capacity reaches 994 mAh·g −1 at 0.1 A·g −1 , 883 mAh·g −1 at 0.2 A·g −1 , 735 mAh·g −1 at 0.5 A·g −1 , 644 mAh·g −1 at 1 A·g −1 , 538 mAh·g −1 at 2 A·g −1 , 397 mAh·g −1 at 5 A·g −1 , respectively. A specific capacity of 397 mAh·g −1 at 5 A·g −1 for this material is as high as 40% of its capacity at 0.1 A·g −1 (994 mAh·g −1 ). The results show that MMO has a better rate capability compared to previous reports 26,34 It is noteworthy that this high performance is achieved without coating the active material with carbon based conductors such as graphene. Table 2 summarizes some recent works on molybdenum oxide based materials as anodes for lithium ion batteries. The specific capacity (930.6 mAh g −1 , 0.2 A·g −1 ) and cycling performance of MMO in this work excels most of the molybdenum oxide based materials previously reported in the literature, and even those with carbon coating [35][36][37][38][39][40][41] .
To further understand lithium storage mechanisms of MMO, Solid-State NMR and XRD experiments on lithiated samples were performed. 7 Li NMR spectroscopy proves to be a powerful method to investigate the relationship between local structures and electrochemical performances of materials for lithium ion batteries and to reveal the electrochemical reaction mechanisms 42 . 7 Li MAS NMR spectra of the MMO and MoO 3 lithiated to 5 × 10 −3 V are given in Fig. 6. A broad resonance can be seen at 2.6 ppm in the spectrum of lithiated MoO 3 , which can be attributed to Li 2 O which is the product of the conversion reaction. The resonance of the lithiated oxide is broader compared to the spectrum of commercial Li 2 O standard sample, indicating more complex local structures of the lithium ions after the lithiation process, and/or the magnetic fields generated by the Mo particles 22 . Similarly, the broad peak at 2.6 ppm in the spectrum of lithiated MMO can be assigned to Li 2 O formed in the conversion reaction. A sharp peak at − 7.0 ppm can also be observed for MMO, indicating that there is another local environment of lithium ions.
The presence of the other lithium compound is further supported by XRD data of lithiated MMO (Fig. 7). The Kapton film was used to protect the air-sensitive fully lithiated sample and its broad background peak can be   Thus, the value of x can be determined as 0.058, implying 16% of the MoO 2 material underwent the insertion/ extraction reaction in the MMO material. The capacity of MMO increased after 50 cycles, indicating that more MoO 2 underwent conversion reaction 20,30 . This activation process, which has been reported many times, can be ascribed to the transformation from insertion reaction to conversion reaction, corresponding to the slowly decreasing of discharge plateaus at about 1.5 V 21 .
Two-dimensional exchange (EXSY) 7 Li spectra provides more detailed information on the structure of MMO and additional evidences for its better electrochemical properties 43,44 . With a short mixing time of 1 μ s, only two peaks on the diagonal at (2.6, 2.6) and (− 7.0, − 7.0) corresponding to Li 2 O and Li 0.98 MoO 2 can be observed in the 2D EXSY spectrum (Fig. 8a). However, small cross peaks at (2.6, − 7.0) and (− 7.0, 2.6) can be observed at a much longer mixing time of 100 ms for MMO (Fig. 8b) while no peak is present at the same mixing time for MMM ( Figure S3). It suggests that the two species in MMO are close in proximity and this mixing time is long enough for chemical exchange processes to occur in which lithium ions move across the interfaces between the two sites. The conclusion that the insertion and conversion reaction products Li 0.98 MoO 2 and Li 2 O are close in space implies that MoO 2 and MoO 3 regions are well mixed at the microscale in MMO, while this is not the situation in the sample MMM. The much lower capacity in MMM is presumably associated with the large volume change of MoO 3 (104% volume change) during cycling 27,45 and the resulting aggregation and/or sluggish kinetics of lithium ion insertion for the MoO 2 nanoparticles 28,29 . Since the MoO 2 and MoO 3 species are not well mixed at the microscale for MMM, inhomogeneous distribution causes phase isolation during cycling, which will lead to pulverization and rapid capacity decay. It has been demonstrated that superior stability can be achieved for the sample with a closer combination between mixed valence lithium vanadates formed in an in-situ transformation than other common coating methods 14,46 . The boost of the electrical performances may be explained as the results of synergistic interactions 34,[47][48][49][50] . Since transformation from MoO 2 to Li 0.98 MoO 2 in insertion reaction only involves small volume/structure change in the very initial cycles, the well dispersed MoO 2 helps to keep the integrity of the electrode and prevent the loss of active materials.
Although it is common that only insertion reaction occurs for bulk MoO 2 , nanostructured MoO 2 often undergoes conversion reaction in further lithiation, presumably due to faster kinetics 30 . According to the XRD and NMR results of lithiation samples of MoO 2 (Figure S4), the intermediate phase Li 0.98 MoO 2 still exist after 5 cycles, which helps stabilize the structure 44 . The much more stable Li 0.98 MoO 2 in the lithiated MMO nanoparticles may be attributed to the high concentraion of MoO 3 around MoO 2 in MMO. MoO 3 is associated with a higher equilibrium potential and a lower activation barrier compared to MoO 2 , thus conversion reaction will be easier to occur during lithiation 26 . The generated amorphous Mo and Li 2 O have poor electronic conductivity and ionic conductivity, therefore slowing down the kinetics and making further lithiation of Li 0.98 MoO 2 more difficult 26,30 .

Conclusion
In conclusion, MMO is developed from the moderately oxidization reaction of molybdenum dioxide nanoparticles. The nominal composition of the material can be calculated as MoO 2.63 , according to the TGA results. MMO shows a high cyclability of 930.6 mAh·g −1 after 200 cycles at a current of 200 mA·g −1 as the anode material for lithium ion batteries. The formation of stable Li 0.98 MoO 2 in the lithiation of MMO is confirmed with NMR and XRD on lithiated sample. 2D NMR results confirm that MoO 2 and MoO 3 are well mixed at the microscale. Both insertion/extraction and conversion reaction mechanisms are involved and the fraction of MoO 2 which undergoes the former reactions is estimated to be 16%. The insertion/extraction process stabilizes the electrode material and decrease the volume change generated from conversion reaction of MoO 3 . This synergistic interaction is believed to be the key to the excellent electrochemical performances and this approach can be applied in many other transition metal oxides to provide advanced electrode materials.