Tailoring Mg²⁺ Solvation Structure in a Facile All-Inorganic [Mg_xLi_yCl_2x+y·nTHF] Complex Electrolyte for High Rate and Long Cycle-Life Mg Battery

1 Introduction

Magnesium (Mg) as a metal anode can afford nearly twice the volumetric capacity than that of lithium (3833 vs 2062 mA h cm⁻³) and possesses several unique advantages such as low cost and safety.^{[1, 2]} In research of rechargeable Mg batteries, developing suitable electrolytes is pivotal. Over the past decades, intensive efforts have been invested in finding high-performance Mg²⁺ electrolytes.^[3-12] However, it is still challenging to develop an electrolyte that simultaneously possesses advantages of simple, low-cost, and high efficiency. Among the electrolytes that have been developed, the all-inorganic Mg²⁺ electrolyte is particularly attractive due to its simplicity and broad application prospects. The first successful all-inorganic electrolyte is synthesized by reacting MgCl₂ with a Lewis acid of AlCl₃, which is capable of stripping/plating Mg, non-nucleophilic (compatible with sulfur) and delivers an electrochemical window of nearly 3 V (vs Mg/Mg²⁺).^[13-15] Such electrolyte has been termed as magnesium aluminum chloride complex (MACC),^[13] although a later study by single-crystal X-ray diffraction indicates that it is actually composed of the dimer cations, [Mg₂Cl₃(THF)₆]⁺, and the AlCl₄⁻ anions.^[14] Despite of the advantages, a tedious conditioning process is required before a reasonable reversibility of Mg stripping/plating can be achieved in MACC, and its electrochemical activity may decay again with time.^{[16, 17]} Toxic CrCl₃,^[18] Mg powders,^[17] and high-cost MgTFSI₂^[19] have been added in MACC to relieve its onerous electrochemical activation process. Another problem of the MACC electrolyte is the inevitable co-deposition of Al during Mg plating, as a result of a higher reduction potential of Al (−1.66 V vs NHE) than that of Mg.^[13] To address this issue, AlCl₃ has been replaced by YCl₃ because Y has a lower reduction potential of −2.372 V (vs NHE).^[20] Despite the complicated preparation procedures that require 120 °C synthesizing temperature and PYR₁₄TFSI cosolvent, MgCl₂-YCl₃ complex solution delivers a relatively lower overpotential of ±110 mV (at 0.5 mA cm⁻²) for Mg stripping/plating without Y co-deposition and an electrochemical window of 3.0 V (vs Mg/Mg²⁺). Nevertheless, a conditioning process is still required for MgCl₂-YCl₃, and its solution structure is not clear yet. Therefore, the search for new high-performance all-inorganic Mg electrolyte is still critically needed.

Solvation chemistry plays a pivotal role in reactions on electrode surfaces and ultimately affects the performance of Mg batteries. Tailoring the solvation structure of an electrolyte is an effective way for improving the reversibility of metallic anodes. LiCl has long been considered as an efficient additive in organic synthetic chemistry, partially due to its high solubility in organic solvents, and the reduction potential of Li⁺ (−3.04 V vs SHE) is much lower than that of Mg²⁺, which eliminates the possibility of Mg-Li co-deposition. Previously, we reported that 1 M addition of LiCl to a MACC electrolyte ([Mg(THF)₆][AlCl₄]₂/PYR₁₄TFSI/THF) could dissolve the absorbed MgCl₂ on electrode/electrolyte interface and significantly reduce the polarization of Mg platting/stripping.^[9] However, the complicated preparation method and the use of high-cost ionic liquid make it less attractive for practical applications, and the detailed electrolyte solvation structure has not been studied. In this work, using a solvation structure tailoring strategy, a highly efficient all-inorganic magnesium electrolyte was synthesized by a simple room-temperature reaction of LiCl with MgCl₂ in THF solvent. Raman spectroscopy, nuclear magnetic resonance (NMR) spectroscopy, density functional theory (DFT) calculation, and molecular dynamics (MD) simulation revealed that a complex of [Mg_xLi_yCl_2x+y·nTHF] aggregate was formed in the solution owning to the interplay of Mg²⁺, Cl⁻, and Li⁺ in THF. We therefore refer this electrolyte as magnesium–lithium chloride complex (MLCC). Importantly, the coordination number (0.67) of THF in the first solvation sheath of Mg²⁺ in MLCC (0.3 M) is much lower than that (2.5) in MgCl₂/THF solution (0.3 M), which leads to a significantly enhanced de-solvation kinetics of Mg²⁺. This MLCC electrolyte possessed a stable electrochemical window up to 3.1 V (vs Mg/Mg²⁺) and enabled reversibly cycling of Mg stripping/plating at a high current density of 10 mA cm⁻² with an outstanding Coulombic efficiency (CE) up to 99%. Using this electrolyte, a Mg/Mo₆S₈ full cell can be cycled for over 10 000 cycles with a superior capacity retention of over 85 mA h g⁻¹ under an ultrahigh rate of 50 C. To our knowledge, this performance is the best among reported so far for all-inorganic Mg²⁺ electrolytes (Table S1).

2 Results and Discussion

To evaluate the Mg stripping/plating behavior in all-inorganic MLCC electrolytes, cyclic voltammetry (CV) and linear sweep voltammetry (LSV) measurements were performed using electrolyte samples prepared with different Mg²⁺:Li⁺ ratios (Figure 1a,b). For comparison, a control sample of pure MgCl₂ solution in THF was also tested, where its CV showed low redox peaks, indicating its poor Mg stripping/stripping capability (Figure 1a). In stark contrast, significantly pronounced redox peaks were obtained with increasing Li⁺ ratio. Considering the solubility issue, the highest Mg²⁺:Li⁺ molar ratio was limited to 2:4 in our experiments, which gave the lowest overpotential and the largest peak current density. The molar ratio of Mg²⁺ and Li⁺ also impact the oxidative decomposition voltage of the MLCC electrolytes, as shown in the LSV results (Figure 1b). The control sample of pure MgCl₂ solution displayed an electrochemical window of close to 3.5 V (vs Mg/Mg²⁺). With increasing Li⁺ ratio, the electrochemical window showed a slight decrease to 3.3, 3.1, and 2.8 V for Mg²⁺:Li⁺ = 2:2, 2:3, and 2:4, respectively.

Mg/Cu cells were used to further examine the overpotential and CE of the Mg plating/stripping processes in MLCC electrolytes (Figure 1c,d). The pure MgCl₂ control sample displayed a high Mg plating overpotential of 800 mV at the current density of 0.5 mA cm⁻² and a short-circuit event occurring at 0.2 mA h cm⁻² capacity (Figure 1c). In contrast, MLCC electrolytes significantly alleviated the polarization and avoided short circuit (Figure 1d). At a higher current density of 1 mA cm⁻², the Mg plating overpotential in MLCC electrolytes with Mg²⁺:Li⁺ = 2:1, 2:2, 2:3, and 2:4 were only 121, 85, 79, and 73 mV, respectively. Nearly 100% CE was achieved in MLCC electrolytes as shown in Figure S1. These showed a great promise of MLCC electrolytes for practical Mg batteries. Considering the overall electrochemical performance, we chose the MLCC electrolyte with Mg²⁺:Li⁺ = 2:3 and a Mg²⁺ concentration of 0.3 M (Figure S2) as the optimal sample for further characterization.

Figure 2a exhibited the Mg plating/stripping cycling performance of this electrolyte in a Mg/Cu cell at an increased current density of 2 mA cm⁻². Note that the overpotential in the first cycle was slightly larger than that of the subsequent cycles, which may be caused by the electrochemically inactive substance absorbed on Mg anode before cycling. The voltage profiles were quite stable afterward with a low deposition overpotential of only 124 mV for over 250 cycles. Increasing the current density to 5, 8, or even as high as 10 mA cm⁻², despite of the increased polarization voltage (Figure 2b), Mg/Cu cells with the MLCC electrolyte could cycle stably for over 1000 cycles with a CE >99.0% (Figure 2c). The cycling stability of the MLCC electrolyte was further tested in a Mg/Mg symmetrical cell under different current densities from 0.2 to 10 mA cm⁻² (Figure 2d). A Mg/Mg cell with the MLCC electrolyte demonstrated an ultralong cycle life for over 1000 h at 2 mA cm⁻², with an average overpotential of <150 mV (Figure 2e), which outperform that of MACC electrolyte (Figure S3). Similar stable cycling results were achieved for higher current densities of 5 and 8 mA cm⁻² (Figure S4). Notice that the cycle life of 300 h at 8 mA cm⁻² corresponds to 1200 cycles (Figure S4b), which illustrates the excellent rate performance of the MLCC electrolyte.

The Mg deposit on Cu electrode obtained from an 1-h plating process in the optimal MLCC electrolyte at 0.5 mA cm⁻² showed relatively dense morphology without dendrite formation (Figure 3a,b). The elemental mapping results by energy-dispersive X-ray spectroscopy (EDX) indicated that the deposit was Mg with a small amount of oxygen and carbon as surface contaminants (Figure 3c). The much lower reduction potential of Li should exclude its co-deposition with Mg. To experimentally verify this, X-ray photoelectron spectroscopy (XPS) was performed and there was no Li detected in the Mg deposit (Figure 3d). The Mg 2p, O 1s, and C 1s XPS spectra (Figure 3e,f and Figure S5) showed that the O and C surface contaminants were mainly in the forms of MgO, Mg(OH)_2, and MgCO₃, which may be caused by the brief exposure of the sample to air before measurement, or from minor electrolyte decomposition products.

To investigate the solvation structure of MLCC electrolytes, MD simulations were performed (see SI for simulation details). MD results showed that the addition of Li⁺ in MLCC induced a complex [Mg_xLi_yCl_2x+y·nTHF] aggregate formation (Figure S6), causing a significantly change in the Mg²⁺ coordination environment (Figure 4a and Figures S7 and S8). Figure S9, Figure 4b–e and Figure S10 showed the radial distribution functions of Mg²⁺ in MgCl₂/THF and MLCC electrolytes with different Mg²⁺:Li⁺ ratios. The Mg-O_THF and Mg-Cl peaks in the Mg²⁺ first solvation sheath represented the direct interaction of Mg²⁺ with THF and Cl⁻, respectively. For pure MgCl₂/THF (Figure S9), there were 2.5 THF on average coordinated with Mg²⁺ in its first solvation shell. In MLCC electrolytes (Figure 4b–e), the coordination number of oxygen (in THF) in the first solvation shell of Mg²⁺ was significantly reduced to 1.09, 0.80, 0.67, and 0.54 for Mg²⁺:Li⁺ = 2:1, 2:2, 2:3, and 2:4, respectively. The lowered coordination number with THF assured a lower de-solvation energy of Mg²⁺ in the MLCC electrolytes, which would help its electrochemical kinetics in Mg plating/stripping. These results were consistent with the CV and overpotential measurements. On the contrary, the increased Cl⁻ coordination number in the first solvation sheath of Mg²⁺ may increase the binding sites for Mg²⁺ jumping motion in MLCC, thereby further facilitating the electromigration of Mg²⁺.^[21] Raman spectroscopy and ⁷Li NMR were performed to experimentally investigate the solvation structure of MLCC. In the Raman spectrum of MgCl₂/THF control sample (Figure 4f), the Raman bands at 280, 600, and 671 cm⁻¹ were from THF solvent, while the characteristic peak at 212 cm⁻¹ was attributed to the undissociated MgCl₂.^[22] In the spectra of MLCC electrolytes with increasing Li⁺ ratios, the intensity of this MgCl₂ peak decreased gradually, while a Raman peak appeared at 183 cm⁻¹ that corresponded to LiCl.^[9] Noticeably, a new weak Raman band at 230–258 cm⁻¹ also appeared. The emergence of the new Raman feature provided strong evidence that the MLCC electrolytes were not a simple mixture solution of MgCl₂ and LiCl; instead, there was a chemical reaction that resulted in a new structure in MLCC.

Previous studies on chloride-based magnesium electrolytes normally attributed the Raman feature at 230–258 cm⁻¹ to the complex [Mg₂(μ-Cl)₃·6THF]⁺ species.^{[22, 23]} In MLCC electrolytes, however, the formation of [Mg₂(μ-Cl)₃·6THF]⁺ would require the first-step reaction of MgCl₂ and LiCl to form [MgCl·3THF]⁺ (Table S2).^[14] The calculated free energy increase for this reaction was too high (43.1 kcal/mol) to allow it to happen at room temperature, thus prohibiting the formation of [Mg₂(μ-Cl)₃·6THF]⁺ in MLCC. From our MD simulation results, we could disassemble the aggregated Mg-Cl-Li solvation structure in MLCC electrolytes into clusters of a ([Mg₂LiCl₅·4THF), b ([MgLi₂Cl₄·5THF]), and c ([Mg₂Li₃Cl₇·7THF]) as shown in Figure S11 and calculate their Raman modes using DFT method. As listed in Table S3, the Mg-Cl stretching modes for clusters a, b, and c were calculated at 234.19, 254.68, and 231.55 cm⁻¹, respectively, which were consistent with the experimental Raman features at 230–258 cm⁻¹. Therefore, the Raman experiment well supported the MD simulated solvation structure in MLCC.

Figure 4g showed the ⁷Li NMR spectra of LiCl/THF and MLCC electrolytes with different Mg²⁺:Li⁺ molar ratios. The singlet at 0.60 ppm in the spectrum of pure LiCl/THF represented the formation of dimer [Li₂(μ-Cl)₂·4THF], which was in accord with the previous reports.^{[24, 25]} Singlets with chemical shifts of 0.33, 0.40, 0.41, and 0.41 ppm were also observed in the spectra of MLCC electrolytes with Mg²⁺:Li⁺ = 2:1, 2:2, 2:3, and 2:4, respectively. Note that 1) the NMR peaks of all the MLCC samples were relatively shifted to higher field in comparison with that of the LiCl/THF sample; 2) the relative peak-shift decreased with increasing Li⁺ ratio in the MLCC sample; and 3) the NMR peaks of MLCC samples were slightly wider than that of the LiCl/THF sample. In order to correlate the NMR results with our MD simulated solvation structure, we used DFT method to calculate the ⁷Li NMR chemical shifts of clusters a, b, and c relative to that of the dimer [Li₂(μ-Cl)₂·4THF]. As shown in Table S3, the calculated relative shifts were all negative and <0.4 ppm, in agreement with the experimental NMR results of MLCC samples where the NMR peaks were in higher field in comparison with that of [Li₂(μ-Cl)₂·4THF]. MD simulation results showed that MLCC with increasing Li⁺ ratio contained less cluster a (displaying NMR at higher field) but more clusters b and c (displaying NMR at slightly lower field), also in agreement with the experimental observation in Figure 4g. Although the calculated adjacent peaks of clusters b and c seemed to be inconsistent with the experimental observation of singlets in MLCC samples, it could be explained by the fast conversion among multiple structures, where the exchange rate among different structures was faster than the difference in multiple peak centers, and hence resulted in a wider singlet peak (Figure 4g).^[26]

According to the experimental and calculation results, the MLCC solvation structure is in the form of [Mg_xLi_yCl_2x+y·nTHF] aggregates, and the formation of this local aggregation structure significantly decreases the coordinated number of THF in the first solvation sheath of Mg²⁺. Past reports showed that replacing the coordinated solvents of metal ions with anions could facilitate reduction in metal ions at metal surface under the influence of the electric field.^[27-32] In Figure S12 and Table S4, we give an estimation of the binding energies of Mg-Cl and Mg-O_THF in [Mg_xLi_yCl_2x+y·nTHF] (see Supporting Information for details). Under certain assumptions, the difference between Mg-Cl-Mg and Mg-O_THF binding energies is small, and the dissociation of the former could be assisted by electric field. In this situation, Mg²⁺ ions become surrounded with three Cl⁻ and one or even zero THF molecule (Figure S12a), making the Mg²⁺ easier to be reduced (Figure S12c). However, for MgCl₂/THF electrolyte, although Cl⁻ will be expelled away from the electrode/electrolyte interface (but still close to Mg²⁺), more coordinated THF molecules with large volume cause larger steric hindrance, making the Mg²⁺ more difficult to accept the electron and to be reduced (Figure S12b).

To further demonstrate the merit of fast de-solvation kinetics of the MLCC electrolyte, high-rate performance of Mg/Mo₆S₈ full cells was tested. The cell was assembled using a Mo₆S₈ cathode (areal density, 0.65–0.7 mg cm⁻²), a Mg foil anode, and a Whatman glass fiber separator. In this full cell, it should be noted that the anode undergoes Mg stripping and plating during cycling, while the Mo₆S₈ cathode may mainly involve Li⁺ intercalation as previously reported for mixed cation system.^[7] As shown in Figure S13, the Mg/Mo₆S₈ cells could be cycled for over 1000 cycles at the high current density of 5, 10, and 20 C (1 C = 128.8 mA g⁻¹) with a high CE (>99%). The slight increase in discharge capacity might due to a slow wetting process of the electrolyte and/or an activation process of the passivation layer on the Mg anode surface.^[7] Figure 5a showed the galvanostatic charge/discharge profiles of the 1st, 100th, 300th, and 500th cycles at 5 C. The Mg/Mo₆S₈ cell delivered an initial discharge specific capacity of 120.7 mA h g⁻¹ and retained 110 mA h g⁻¹ after 300 cycles, indicating a high electrochemical reversibility. Figure 5b showed the galvanostatic charge/discharge curves of the 200th cycles at different rates. Although the discharge plateau was lower as the current density increased, the small voltage hysteresis reflected the good rate performance of the Mg/Mo₆S₈ cells with the MLCC electrolyte. In Figure 5c, an ultralong cycle life of a Mg/Mo₆S₈ cell was demonstrated at an ultrahigh current density of 50 C. After 10 000 cycles, the specific discharge capacity was stably maintained above 85 mA h g⁻¹ with a CE close to 100%. As far as we know, this is the best rate performance among Mg/Mo₆S₈ cells reported so far (Figure S14, including mixed Mg⁺/Li⁺ systems).^[7] It should be noted that our key conclusion on the role of Mg²⁺ solvation structure in MLCC for efficient Mg plating and striping is strongly supported by the results of Mg/Mg symmetric cells, which do not involve Li intercalation reaction or Li plating; and the excellent performance of the full cell is just another support of MLCC's advantages.

3 Conclusions

In conclusion, an all-inorganic MLCC electrolyte was synthesized by a reaction of MgCl₂ and LiCl salts in THF solvent at room temperature. Using this electrolyte, highly reversible Mg deposition/stripping process, with a CE as high as 99% for over 1000 cycles, was realized in a Mg/Cu cell at a high current density of 10 mA cm⁻². A Mg/Mg symmetric cell also demonstrated an ultralong cycle life of over 1000 h at 2 mA cm⁻². For the Mg/Mo₆S₈ full cells using the MLCC electrolyte, a superior capacity retention of 85 mA h g⁻¹ after 10 000 cycles was achieved at an ultrahigh rate of 50 C. Raman spectroscopy and NMR characterizations, together with MD simulation and DFT calculation, revealed that the solvation structure of MLCC consisted of [Mg_xLi_yCl_2x+y·nTHF] aggregates owing to the interplay of Mg²⁺, Cl⁻, and Li⁺ in THF. This aggregated structure significantly reduced the number of coordinated THF molecules in the first solvation sheath of Mg²⁺ and thus reduced its de-solvation energy and boosted its electrochemical kinetics. This well explained the excellent rate performance of the MLCC electrolyte.

4 Experimental Section

Conflict of Interest

The authors declare no conflict of interest.