High-Voltage Aqueous Magnesium Ion Batteries.

Nonaqueous rechargeable magnesium (Mg) batteries suffer from the complicated and moisture-sensitive electrolyte chemistry. Besides electrolytes, the practicality of a Mg battery is also confined by the absence of high-performance electrode materials due to the intrinsically slow Mg2+ diffusion in the solids. In this work, we demonstrated a rechargeable aqueous magnesium ion battery (AMIB) concept of high energy density, fast kinetics, and reversibility. Using a superconcentration approach we expanded the electrochemical stability window of the aqueous electrolyte to 2.0 V. More importantly, two new Mg ion host materials, Li superconcentration approach we expanded the electrochemical stability window of the aqueous electrolyte to 2.0 V. More importantly, two new Mg ion host materials, Li3V2(PO4)3 and poly pyromellitic dianhydride, were developed and employed as cathode and anode electrodes, respectively. Based on comparisons of the aqueous and nonaqueous systems, the role of water is identified to be critical in the Mg ion mobility in the intercalation host but remaining little detrimental to its non-diffusion controlled process. Compared with the previously reported Mg ion cell delivers an unprecedented high power density of 6400 W kg ion cell delivers an unprecedented high power density of 6400 W kg while retaining 92% of the initial capacity after 6000 cycles, pushing the Mg ion cell to a brand new stage.

Rechargeable magnesium (Mg) batteries (RMBs) have been considered as an attractive alternative to Li. [1−7] Challenges for developing a practical Mg full cell come from all components, but in particular from a Mg2+-conducting electrolyte that would remain electrochemically stable with both Mg anode and cathode.[5,6,8−10] Grignard reagents based on the complex between Mg and halocarbons dissolved in ethereal solvents were used most often as electrolytes because of their resistance against reduction by Mg anode, but they come with intrinsically low conductivity and poor electrochemical stability on the cathode surface.[8,11,12] Non-Grignard magnesium aluminum chloro complexes (MACC) are more conductive and oxidation-resistant, but their chemically corrosive nature induces both high cost and high hazard potential.[13,14] Of course, all these nonaqueous electrolyte systems are extremely moisture-sensitive and have to be prepared and handled following protocols even more strict than those for the electrolytes for LIBs. Besides electrolytes, the practicality of a Mg battery is also confined by the absence of high-performance cathode materials.[8,15−18] The most-frequently used chevrel phase molybdenum sulfide (Mo6S8) operates at about 1.1 V vs Mg during discharge, while other cathodes of higher voltage typically suffer from poor Mg ion diffusion in the host structures.[2,8,19−22] As a result, the Mg or Mg ion battery chemistry typically suffers from poor rate performance.[23]

Recent reports that a trace amount of H2O in nonaqueous electrolyte can significantly increase the kinetics of Mg2+ ion in the cathode host[24,25] have raised the once unlikely possibility of aqueous electrolyte for Mg batteries. Such aqueous Mg ion batteries (AMIBs) using appropriate cathode and anode intercalation hosts would be very attractive because of the low cost, noncorrosive, safe, and highly conductive aqueous electrolytes. In this line of thought, several cathode materials have already been evaluated in different aqueous electrolytes with decent electrochemical performances.[26−28] However, most reported aqueous Mg chemistries were only evaluated in half-cell configurations[24,26−30] due to the absence of proper electrochemical couples (cathode and anode). The problem comes from the limited candidate Mg2+-host materials available, while the narrow window (1.23 V) of the traditional aqueous electrolyte aggravates the restriction, which would consequently lead to low voltage and energy density of the AMIBs chemistry.[31,32]

In this work, we demonstrated a high voltage (1.9 V) AMIB full cell based on the expanded stability window (2.0 V) of Mg ion aqueous electrolyte and coupling of two novel Mg ion host materials. As shown in Figure a, lithium vanadium phosphate (Li3V2(PO4)3, LVP) was demonstrated as an ideal Mg ion host cathode material combining with high working voltage and fast kinetics. By coupling with a non-diffusion controlled polyimide anode (poly pyromellitic dianhydride, PPMDA), a 1.9 V AMIB full cell exhibits high specific energy density of 68 Wh/kg, high power density of 1440 W kg–1, and superior cycling stability of 6000 cycles. This highly reversible and fast aqueous Mg2+ ion battery chemistry confirms the feasibility of using aqueous electrolyte to address the once challenging intercalation kinetics of Mg ion cathode. This high voltage AMIB concept offers a safe and cost-wise energy storage solution to large-scale applications where power density, cost, and cycle-life far outweigh energy density, such as the grid-storage or renewable energy harvesting systems.

Figure 1

Schematic illustration of the high voltage AMIB. (a) Schematic of the proposed AMIB components. (b) The expanded electrochemical stability window of 4 m Mg(TFSI)2 aqueous electrolytes measured with cyclic voltammetry (CV) on stainless steel working electrodes between −1.3 and 1.3 V vs Ag/AgCl at 10 mV/s. The potential has also been converted to Mg/Mg2+ reference (upper X-axis) for convenience. The O2/H2 evolution potential and Mg2+-intercalation potentials of various reported electrode materials are marked in the graph.

As reported in our previous work, superconcentration could suppress the water activity and expand the electrochemical stability window. Mg(TFSI)2 is thus employed as conductive salt at 4.0 m (molality, mol/kg) to prepare the electrolyte. Such an electrolyte is completely noncorrosive (pH ∼ 7), safe, and green compared with its nonaqueous counterparts,[10] or 10 times that of the Li+ electrolytes used in the state-of-the-art LIBs.[33] The electrolyte shows a 2.0 V window with a cathodic limit of ∼1.7 V and an anodic limit of ∼3.7 V vs Mg, which is wider than that of dilute MgSO4 electrolyte by 0.7 V (Figure b). The dilute MgSO4 electrolyte has been a typical aqueous electrolyte used, whose stability window is defined by the O2 and H2 evolution.[34] It should be noted that this 2.0 V window is even wider than that of most nonaqueous Mg ion electrolytes reported to date (Table S1).[35] When plotting the redox potentials of the reported Mg2+-intercalation electrode materials in the above electrochemical stability limits, it is apparent that all those materials either reside outside of the expanded stability window, thus being excluded as candidate electrode material, or fall right in the middle of it, an awkward position preventing the expanded window from being fully utilized. In particular, the most successful Mg2+-intercalation cathode, chevrel phase Mo6S8, operates at a potential of ∼1.1 V vs Mg and falls near the edge of this window.[36,37] Thus, appropriate Mg2+-host cathode and anode materials must be tailored for this electrolyte. As shown in Figure b, Mg-containing MgLiV2(PO4)3 and Mg-free PPMDA with Mg insertion/extraction potential at 3.7 and 1.7 V, respectively, are exactly at the edge of the upper and lower limits of the 4 m Mg(TFSI)2 electrolyte window and are perfect for cathode and anode of aqueous Mg ion batteries.

Unlike the nonaqueous battery where Mg ion comes from the Mg metal, Mg-containing electrode material should be used to fabricate an AMIB full cell. We synthesized the carbon-coated LVP containing alkali metal cations through a spray-drying method and exchanged Li+ for Mg2+in the first charge/discharge cycle, as illustrated in Figure a. The X-ray diffraction (XRD, Figure b) revealed that LVB has a monoclinic structure (P21/n) consisting of a three-dimensional framework with distorted VO6-octahedra and PO4-tetrahedra sharing a corner with each other[38−40] where three lithium atoms occupy three distinct crystallographic positions. Transmission electron microscopy (TEM) images (inset Figure b and Figure S1) show that LVP particles with an average size of ∼100 nm are well coated and connected by the carbon thin layer. Since the LVP is stable in the 4 m Mg(TFSI)2 electrolyte (Figure S2), the electrochemical performance of LVP was also evaluated in a three-electrode cell using 4 m Mg(TFSI)2 electrolyte with the activated carbon and Ag/AgCl as counter and reference electrodes. As shown in Figure c, LVP displayed two plateaus with a total capacity of 115.1 mAh/g during the initial charging (delithiation) process, indicating the extraction of nearly two Li from the structure.[39] In the following discharge step, Mg2+ rather than Li+ enters the framework structure simply because the concentration of the former in the electrolyte is much higher than that of the latter by several orders of magnitude. The initial magnesiation process demonstrated a different curve from the following cycles, which reflects that Li+ has not been totally replaced by Mg2+. The second charge process delivered 118.7 mAh/g with a small plateau at the same voltage of Li extraction, demonstrating that the residual Li was removed. The second discharge showed a much larger capacity and higher voltage than the first one. The third discharge was almost identical with the second, indicating the total replacement of Li+ by Mg2+. The structure revolution was also confirmed by the XRD patterns in Figures S3 and S4. The XRD peaks after the third discharge were quite similar to those of the pristine LVP. Comparison of magnified details demonstrated that the main peaks slightly shifted to higher theta, indicating slight lattice contraction, which may be attributed to the fact that Mg2+ (0.72 Å) has a slightly smaller ionic radius and a much stronger electrostatic attraction effect than Li+ (0.74 Å). It took several cycles to completely remove the two distinct Li+. In the subsequent cycles, the reversible Mg2+ insertion/extraction should dominate as the potential profile and capacity stabilize. This ion exchange process was also confirmed by the cyclic voltammogram (CV) in Figure S5. After 10 cycles, Mg was distributed uniformly in the electrode according to the element mapping (Figure S6).

Figure 2

Electrochemical performance of the LVP cathode. (a) The schematic of the working mechanism for the LVP cathode. (b) Powder X-ray diffraction patterns for the LVP cathode (inset: the TEM image). (c) The typical voltage profiles of LVP in 4 m Mg(TFSI)2 electrolyte at constant current of 1 C (100 mA/g as 1 C) with activated carbon and Ag/AgCl as counter and reference electrodes, respectively. (d) The cycling stability and Coulombic efficiencies of LVP cathode at 1 C rate. (e) The voltage profile of LVP cathode at various rates.

The cycle performance of LVP is presented in Figure d. In the initial 20 cycles, the capacity of LVP gradually reduced and then stabilized to 102 mAh/g. After 1000 cycles, a reversible capacity of 89.6 mAh/g was still retained with 100% Coulombic efficiency, corresponding to a low capacity decay rate of 0.012% and excellent cycle performance. A remarkable performance at high rate is also demonstrated in Figure e, where LVP delivered a capacity of 72.8 mAh/g even at 30 C, retaining 71.3% of the nominal capacity delivered at 1 C. Such an extraordinary rate capability is powerful evidence for the fast kinetics, which is rare for the intercalation chemistry of divalent Mg2+. The fast kinetics should be ascribed to the aqueous electrolyte because smaller capacity was obtained from the same LVP material in the organic electrolyte (Figure S7). In this regard, the role of H2O in generating favorable capacity and its responsibility for the electrochemical properties observed should not be overlooked.[41] The water molecule helps improve the Mg ion mobility in the LVP host, which may be ascribed to the shielding of multi charge in the presence of water. It is also possible the water props open the structures. The detailed mechanism still needs further exploring.

As for anode candidate, we explored polyimide family organic materials due to their faster kinetics than their inorganic counterparts as intercalation host for the divalent cation Mg2+.[42−46] Recently, PNTCDA (poly 1,4,5,8-naphthalenetetracarboxylic dianhydride), a member of the polyimide family, has been reported as anode host material with excellent electrochemical performances for both Li+ and Na+ intercalation while effectively suppressing hydrogen evolution in aqueous electrolytes.[47,48] As another kind of polyimide, PPMDA has a lower working voltage than PNTCDA, which benefits the aqueous full cell by contributing most of the stability window. In addition, PPMDA is also low-cost, green, and sustainable.[45] To achieve a high electronic conductivity, we designed and synthesized composites of PPMDA with 5% multiwall carbon nanotubes (MCNTs) through a simple in situ polymerization process. The morphology of PPMDA@MCNTs was characterized by transmission electron microscopy (TEM, Figure a), indicating that PPMDA@MCNTs exist in nanosheet structure with ∼10 nm layer thickness. The TEM image also shows that the MCNTs were uniformly distributed in the composite, forming an electronic conductive network. The resultant PPMDA@MCNTs was characterized using attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR, inset Figure a), where the bands at 1350 and 1580 cm–1 are assigned to the stretching vibration of imide C–N groups, while the bands at 1700 and 1670 cm–1 are attributed to the asymmetric and symmetric stretching vibrations of imide C=O bonds.

Figure 3

Electrochemical performance of the PPMDA@MCNTs anode. (a) TEM image for the PPMDA@MCNTs composite (inset: the FTIR spectral result). (b) The typical voltage profile of PPMDA@MCNTs in 4 m Mg(TFSI)2 electrolyte at constant current of 1 C (100 mA/g as 1 C) with activated carbon and Ag/AgCl as counter and reference electrodes, respectively. (c) The cycling stability and Coulombic efficiencies of the PPMDA@MCNTs cycled at 1 C rate. (d) The voltage profile of PPMDA@MCNTs cycled at various rates.

The electrochemical behavior of PPMDA@MCNTs was evaluated in 4 m Mg(TFSI)2 aqueous electrolyte in the three-electrode cells by galvanostatic charging/discharging at a current rate of 100 mA/g in the same potential range of 1.7–2.5 V vs Mg (or −0.9 V and −0.1 V vs Ag/AgCl). As shown in Figure b, the composite shows a sloping plateau between 2.0 and 1.7 V during discharge and between 1.7 and 2.2 V during the charge, delivering a reversible capacity of 110 mAh/g with an initial Coulombic efficiency of 84.3%. The latter quickly increases to 99% in the second cycle and reaches ∼100% after 20 cycles. The Mg2+ insertion into PPMDA during discharge was confirmed by comparing the elemental Mg mapping of the PPMDA@MCNTs electrode before and after electrochemical intercalation in anode half cells (Figures S8 and S9). No Mg can be observed in the fresh PPMDA after simple soaking in the electrolyte for 2 h followed by water-rinsing (Figure S9); however, a large amount and uniformly distributed Mg were observed in the fully magnesiated PPMDA after rinsing with water (Figure S8).

In the three-electrode cell, the PPMDA@MCNTs electrode also exhibits excellent cycling stability, retaining 87% of capacity after 500 cycles (Figure c), as well as superior rate capability (Figure d). Even at a very high rate of 20 C, 75.7% of the capacity achieved at 1 C can still be delivered. Such rate performance is among the best ever known for all reported magnesium electrodes.[15,17,18,22,28,49,50] It has been reported that the electrode kinetics of polyimide is not limited by the diffusion process,[31,51] which should account for the excellent rate. Thus, the above excellent performances of PPMDA@MCNTs in the potential range between 1.7 and 2.2 V make it a perfect anode candidate to couple with LVP in our superconcentrated Mg electrolyte. A similar capacity could be obtained from PPMDA using organic Mg ion electrolytes (Figure S10), which further proved that Mg2+ intercalation indeed occurred as in Figure , instead of possible proton intercalation into PPMDA. The different behaviors of LVP and PPMDA in organic electrolyte also suggest that the H2O acts as the critical solvent in improving the Mg ion mobility during the intercalation reactions with the host materials investigated here.

An aqueous Mg ion full cell consisting of PPMDA anode and LVP cathode was thus fabricated. The anode/cathode mass ratio was set to 1:1 to balance the capacity of anode and cathode, and the upper cutoff voltage was set as 1.9 V based on the electrochemical stability window of the aqueous electrolyte. As shown in Figure a, the AMIB full cell delivers a discharge capacity of 52 mAh/g and an energy density of 62.4 Wh/kg based on the total mass of active materials in both electrodes at 1 C rate (100 mA/g). As expected, excellent rate performance is also exhibited (Figure b), with 74.9% of the capacity at 1 C retained even at the high rate of 60 C (6000 mA/g). This high rate capability results in a specific power density of ∼6400 W kg–1 at 60 C, which unprecedented for Mg2+-intercalation chemistries to the best of our knowledge. The self-discharge rate was rigorously evaluated by leaving the fully charged cell at open-circuit voltage while monitoring the cell voltage fading. At the end of the 24 h storage the ratio of self-discharged capacity to charge capacity was calculated (Figure S10). As shown in Figure c, about 85.3% of the original capacity was retained, demonstrating a highly stable electrolyte at the surfaces of the fully magnesiated anode and demagnesiated cathode. The negligible self-discharge rate presents a significant advantage over the electrochemical capacitors, which offer high power densities but are plagued by rather high self-discharge.[52]Figure d displays the cycling stability and Coulombic efficiency of the aqueous Mg ion full cells at both 2 and 20 C rates, respectively. Excellent cycling stability with capacity retention of 86.8% for 1000 cycles at 2 C and capacity retention of 92% for 6000 cycles at 20 C was demonstrated. Such excellent cycling stability is already comparable to that of electrochemical capacitors, but with much higher energy densities. A high Coulombic efficiency of 100% is achieved after the first several cycles at both of these rates.

Figure 4

Electrochemical performances of new aqueous Mg ion full cell. (a) The typical voltage profiles of the AMIB full cell employing PPMDA anode and LVP cathode in 4 m Mg(TFSI)2 electrolyte at constant current of 1 C (100 mA/g). (b) The rate cycle performance of the AMIB full cell. (c) The residual discharge capacity after 24 h storage at fully charged state. (d) The cycling stability and Coulombic efficiencies of the cell at the rates of 20 and 2 C (inset). (e) Performance comparison of electrode materials for Mg ion batteries. (f) The power comparison for Mg ion batteries and bivalent Zn batteries.

For comparison, we plotted the capacity and potential of both Mg ion electrodes of this work, LVP and PDMA, in Figure e against other representative Mg ion host materials previously investigated,[24,25,27,53−56] where the cycling stability is color-coded with red, blue, and green representing <100 cycles, 100–500 cycles, and >500 cycles, respectively. The highly concentrated electrolyte enables a 1.9 V Mg ion battery using LVP cathode and PDMA anode, representing a milestone advance for aqueous Mg ion batteries and even comparable to some nonaqueous Mg batteries, such as the well-known Mo6S8/Mg battery. LVP has the highest operating potential in all Mg ion battery cathodes, while PMDA has the most proper potential as the anode for 4m Mg(TFSI)2 electrolyte. The highly soluble MnO2 and V2O5 in aqueous electrolytes reduce the cycling stability, and their high potential also reduces the cell voltage. Most importantly, the AMIB full cell exhibits a much superior rate capability. As shown in the Figure f, the AMIB can provide an impressive power density of 6400 W/kg, which is higher by 30 times than the highest record achieved by all the organic electrolyte Mg batteries. The AMIB is even competitive to other multivalent-ion batteries[57−60] (Zn ion battery, where the cathode/anode mass ratio was set as 1) and other well-known aqueous battery systems (Figure S11). Considering the high safety, the low cost and green aqueous electrolyte, the environmental friendliness of electrode material, the high specific power, the low self-discharge rate, and the long cycle life, this aqueous Mg ion full cell apparently makes a promising candidate for large-scale applications where energy density is outweighed by the above merits, such as the energy storage units in grid storage or renewable energy systems.[61−63]

In summary, we demonstrated a unique high voltage AMIB in the concentrated Mg(TFSI)2 electrolyte. The resultant full Mg ion cell based on LVP and PPMDA exhibited superior electrochemical performances including excellent rate capability, high power density of 6400 W kg–1, and high-capacity retention of 92% after 6000 cycles. Detailed comparisons of both the LVP and PPMDA in both aqueous and nonaqueous media were conducted. The sluggish Mg ion diffusion within the LVP crystal structure in the nonaqueous electrolyte was accelerated in the aqueous system, while the polyimide material behaved similarly. The success of such a highly reversible and safe full Mg cell could find application in the large-scale energy storage market. More sophisticated material designs, such as tuning organic functional groups and screening more viable cathodes, as well as electrolyte and interphase tailoring, could lead to cell chemistries of higher energy density.