Dajian Li,Yuan Yuan,Jiawei Liu,Maximilian Fichtner,Fusheng Pan

aKarlsruhe Institute of Technology,Institute for Applied Materials-Applied Materials Physics(IAM-AWP),Hermann-von-Helmholtz-Platz 1,76344 Eggenstein-Leopoldshafen,Germany

b National Engineering Research Center for Magnesium Alloys(CCMG),College of Materials Science and Engineering,Chongqing University,400044 Chongqing,China

cKarlsruhe Institute of Technology,Institute of Nanotechnology(INT),Hermann-von-Helmholtz-Platz 1,76344 Eggenstein-Leopoldshafen,Germany

d Helmholtz-Institute Ulm(HIU),Helmholtzstr.11,89081 Ulm,Germany

Received 2 May 2020;received in revised form 19 September 2020;accepted 29 September 2020 Available online 12 November 2020

Abstract There is an increasing demand for rechargeable batteries in high-performance energy storage systems.The current dominating Li-ion batteries are limited by price,resource availability,as well as their theoretical capacities.So that the community has started to explore alternative battery chemistries.As a promising multivalent battery type,rechargeable magnesium batteries(RMBs)have attracted increasing attention because of high safety,high volumetric energy density,and low cost thanks to abundant resource of Mg.However,the development of high-performance anodes is still hampered by formation of passivating layers on the Mg surface.Additionally,dendrites can also grow under certain conditions with pure Mg anodes,which requires further studies for reliable operation window and substitutes.Therefore,this review specifically aims to provide an overview on the often overlooked yet very important anode materials of RMBs,with the hope to inspire more attention and research efforts for the achievement of over-all better performance of future RMBs.

Keywords:Rechargeable Mg batteries;Anode materials.

1.Introduction

Since the invention of lead-acid battery in the year 1859,rechargeable batteries have become an indispensable part for modern society,especially for the portable applications.Up to today,lithium-ion batteries(LIBs)are the most successful rechargeable battery system because of their high energy and power density compared to nickel–cadmium(NiCd)or nickel–metal hydride(NiMH)batteries[1,2].

Although LIBs have achieved great success,people are continuously looking for novel batteries to achieve higher safety,energy and power density[3].Especially,the energy capacity of LIBs is still unsatisfactory.Currently,the charge capacity of LIBs is restricted by the cathode materials,which can be mostly classified into three types according to their crystal structures as layered structure(LiCoO2,theoretical capacity 274 mA h g?1),spinel structure(LiMn2O4,theoretical capacity 148 mA h g?1),and olivine structured polyanionic compounds(LiFePO4,theoretical capacity 170 mA h g?1)[4].Due to limited potential lattice positions of removable Li ions in certain crystal structures,further increase of charge capacity of LIBs can be quite difficult.

Beside capacity limitation,emerging large-scale applications instead of portable electronics demand higher quantity of materials supply,and forces scientific and industrial societies[5,6]to look for more abundant alternatives compared to Li[7].Recently,sodium-ion batteries(SIBs)have stimulated extensive studies due to the natural abundance of Na and no need for expensive Cu and Co in the SIB[8].However,while the expected price of SIB is lower than that of LIB,SIBs have intrinsic low energy densities compare to LIBs.The operation voltages of SIBs are approximately 0.3 V lower than that of LIBs,which means energy density is also limited.The first commercial 18650 SIBs launched by CNRS and RS2E can provide 2000 cycles but only at 90 W h kg?1[9],while the LIB has reached 240 W h kg?1at cell level[5].

In order to develop systems with high gravimetric and volumetric energy densities,as well as improved safety,power density,lifetime,and reduced cost,intercalation batteries based multivalent(MV)chemistries such as Mg2+,Ca2+,Zn2+,Al3+,etc.,have attracted growing interest in the last decade[10].

Among the MV intercalation batteries,rechargeable magnesium batteries(RMBs)have attracted numerous attentions because Mg metal can provide high capacity(volumetric capacity,EV,Mg=3833 mA h cm?3),low reduction potential(-2.4 V vs.SHE)and dendrite-free deposition with up to 100%coulombic efficiency in the cells under certain operation conditions[11,12].Additionally,Mg is the 8thabundant element in the earth’s crust,and resource limitations are not in sight[13,14].Fig.1 shows number of publications with the topic of“magnesium battery”on web of science for two decades up to year 2019.

RMB has long been considered as a promising energy storage technique,but there have been delays in the development for several reasons.First developments of[13,15]have allowed fast charge RMBs with working potential window up to 2.1 V vs.Mg.A persisting challenge is the diffusion of Mg2+,which is sluggish in the cathode part,this will cause voltage hysteresis and insufficient magnesiation degree during cell operation.Due to the urgency for high performance cathodes for RMBs,most publications focus on this area[11,16–18].

Beside cathodes,proper electrolytes are also widely considered as the bottleneck for development of highperformance RMBs.A comprehensive review for the development of electrolytes and their compatibility with anode materials has been given in[19].Although the organohaloaluminate-based electrolytes can prevent passive layer formation on the Mg metal surface and have delighted the hope of pure Mg metal anode for RMBs[13],they are not stable for the high voltage cathodes and limited theoretical energy capacity.Recent developed electrolytes enable fast and reversible stripping and plating with high efficiencies in voltage windows>4V[20],and with no detrimental release of halide ions,which would corrode parts of the cell[21,22].

In many review papers,the anode side of RMBs is neglected,with only one statement as Mg anode is dendritefree deposition with up to 100% coulombic efficiency[23].However,Whittingham and Banerjee have recently stressed that the safety of using Mg metal cannot be fully guaranteed as they are not totally free of dendrite formation[24].Additionally,incompatibility between Mg metal and conventional electrolytes has long been the major obstacle for the development of RMBs because Mg strongly tend to form passivating surface films with many solvents and subsequently block further electrochemical reactions.Regardless numerous efforts have been dedicated to the study of cathodes and electrolytes,no satisfactory battery performance together with pure Mg anode has been developed for RMBs up to now.Therefore,the development of alternative anode materials is still necessary to expand the potential candidates of suitable electrolytes and cathodes.We have to bear in mind that the compatibility between anodes,cathodes,electrolytes,as well as other battery components,such as separators and shells,must be fulfilled during development of novel RMBs.A compilation of reviews of Mg anode has been given in[25].During the preparation of the current manuscript,Niu,Zhang and Aurbach has made a review for the alloy anode materials for RMBs based on groups IIIA,IVA,VA elements[26].Nevertheless,the current review aims to include the latest updates and further bring attention to the importance of anode side of RMBs,their existing problem,and encourage the efforts to overcome them.

2.Pure Mg as anode

Mg was initially considered to be dendrite free when applied as anode of secondary batteries.However,this conclusion is highly dependent on the electrolyte and current density[27].Recent studies showed that some applications under fast charge or low-temperature conditions may cause severer dendrite growth than that Li anodes for LIBs[24,28].Additionally,interferences of pure Mg anodes and electrolytes can be problematic due to high polarizing ability of Mg2+cations.Lu et al.[29]found passive films,which can impede or block Mg ions in their migration towards the electrode surface.In fact,bulk Mg anode is not a functional reversible electrode in some polar organic electrolytes due to the formation of passivating barriers[30].Therefore,a series of efforts have been dedicated to the improvement not only of Mg electrolytes,but also of Mg anodes and their compatibility with the electrolyte.

Fig.2.Representative discharge curves of the Mg/air batteries made from Mg particles with different morphologies,at a constant current of 0.5 mA and a temperature of 25°C[36](reproduced with permission).

Recently,Son et al.[31]prepared slurry by mixing Mg powder with carbon black,polyacrylonitrile(PAN)and Mg(CF2SO3)2in dimethylformaide(DMF).The coated electrolyte was subsequently annealed at 300°C for 1 h under Ar to obtain an artificial interphase on the Mg metal.The coated Mg electrode showed similar deposition/stripping behaviour to that of pristine Mg electrode in APC electrolyte but over 1000 cycles in PC-based electrolyte solution,which can form impermeable passivation film with bear Mg.Full cell assembled using coated Mg anode and V2O5cathode,exhibited reversible capacities of 47 mA h g?1and 140 mA h g?1with PC and PC+3 M H2O electrolytes,respectively[31].However,Attias et al.[19]questioned if the insertion process with V2O5cathode was actually related to protons instead of Mg ions.

Lv et al.[32]used SnCl2-DEM solution to treat Mg foils and get an Sn-based artificial layer,which exhibited a long lifespan(over 4000 cycles,1400 h)in the subsequent stripping/platting measurements at current density of 6 mA cm?2.Zhang et al.[33]added GeCl4into Mg(TFSI)/DME electrolyte and allowed ion exchange of Ge4++Mg→Ge+Mg2+inside assembled cell to form a Ge based artificial layer,which showed a self-repair process in engraving-reassembling experiment.Full cells using Mg coupled with TiS2or Ti3C2in the modified electrolyte both delivered revisable capacity around 80 mA h g?1after 40 cycles.

Due to increased specific surface areas,nanomerization can make Mg more reactive[34,35].Li et al.[36]prepared controlled nano/mesostructures with spherical,platelike,rodlike,and sea-urchin-like shapes using a vapour-transport approach.Fig.2 shows the discharge curves of Mg-air battery using spherical and sea-urchin-like shape Mg particles with similar diameters.Higher discharge voltage,longer discharge time and higher energy density has been obtained using the seaurchin-like shape Mg particles.

Fig.3.Cell behaviours bulk(B)or graphene-like(G)MoS2 cathode and bulk(B)or nanosize(N)Mg anode with a discharge rate of 20 mA g?1[30](reproduced with permission).

Liang et al.[30]prepared ultrasmall Mg nanoparticles(mostly below 3 nm diameter),combined with graphenelike MoS2as cathode material.The cyclability of constructed RMBs was significantly improved compared to bulk-Mg anode,as shown in Fig.3.Liang et al.also proposed that the improved performance of nanostructured Mg anode can be attributed to increased surface area,which can lead to thinner passivating film and improved kinetic of Mg ion migration during cell operation.

3.Metallic anode

3.1.Group XIV elements(Si,Ge,Sn,and Pb)

Several types of metallic anodes have been developed for applications of RMBs.Inspired by the application to LIBs,the group XIV metallic elements,i.e.Si,Ge,Sn,and Pb,have been studied for their suitability as anode materials for RMBs.

Si,Ge,and Sn share the same diamond structure(α),which can host up to 2 moles of Mg atoms with two kinds of possible lattice sites for 1 mole anode material,as shown in Fig.4.When combined with Mg element,Si,Ge and Sn also share very similar phase diagrams,which shows only one intermetallic compound,Mg2X(X=Si[37],Ge[38],Sn[39]),and negligible solubility at both sides of each phase diagram.

If the reaction between Mg and Si,Ge,or Sn follows the corresponding equilibrium phase diagram,only one voltage plateau should appear during each battery application[41].However,Wang et al.[42]found that Si,Ge and Sn can become amorphous during magnesiation process.The predicated[42]multi voltage plateaus for magnesiation of amorphous Si,Ge,and Sn are shown in Fig.5.According to Wang et al.[42],Mg diffusivity in Si(～10?17cm2/s)is much lower than that in Ge and in Sn(～10?14to 10?13cm2/s).Additionally,crystalline Si and Ge are not practical due to the required high overpotential.Therefore,regardless the highest theoretical capacity of Si(3817 mA h g?1)[40],most studies focused on Sn(theoretical capacity 911 mA h g?1)anode application.

Fig.4.Tetragonal(a)and hexagonal(b)insertion sites for Mg(orange)in Ge,Si,and Sn(blue)[40](reproduced with permission).

Fig.5.Electrode potentials vs Mg/Mg2+for amorphous(a-)MgxX(X=Si,Ge,and Sn)as a function of Mg concentration x predicated using ab inito calculation[42](reproduced with permission).

Besides diamond structure,Sn also can crystalize into tetragonal structure(β–Sn).But the formation ofβ–Sn leads to higher Mg diffusion barriers(0.435 eV)compare to that withα–Sn(0.395 eV)[43].Moreover,an already existing Mg ion near the diffusing Mg ion will increase its diffusion barrier to 0.77 eV for theβ–Sn[44].Therefore,β–Sn is not a suitable candidate for RMB applications.Jin and Wang[45,46]calculated the Mg diffusion behaviour through different crystal faces inα-Sn using DFT,and found the lowest diffusion energy barrier for Mg along the<100>direction.

Singh et al.[47]obtained a discharge capacity up to 903 mA h g?1for the first cycle using Sn electrode against pure Mg,as shown in Fig.6.The measured insertion/extraction voltage were 0.15/0.2 V against pure Mg under 0.002 C current rate.The formation/decomposition of the Mg2Sn phase was simultaneously confirmed using XRD technique.

Although Singh et al.[47]obtained near theoretical discharge capacity using Sn electrode for the first discharge process,only about 1/3 of the capacity was recovered during the following charge process.Nguyen et al.[48]assembled half cells using Sn powder electrode and obtained a reversible capacity up to 334 mA h g?1,which is comparable to the re-sults of Singh et al.[47].Nguyen et al.[48]also prepared 49 nm thick Sn film to perform electrochemical test.After 50 cycles,76% discharge capacity was retained with the Sn film,which also exhibited particle cracking and amorphization.Nacimiento et al.[49]used micro-and nano-particulateβ–Sn to assemble RMBs.Within half-cell setup,the microparticulate Sn exhibited low specific capacity less than 10 mA h g?1.The nanoparticulate Sn delivered initial capacity of 225 mA h g?1,but also fast capacity to less than 50 after 50 cycles at 0.05 C current.However,nanoparticulate Sn showed specific capacity up to 150 mA h g?1when combined MgMn2O4cathode[49].

Fig.6.1st cycle galvanostatic magnesiation/demagnesiation curves for Sn/Mg and Bi/Mg half-cells.Inset–XRD spectra for(1)as-fabricated,(2)magnesiated and(3)demagnesiated Sn anode(peak positions marked for Mg2Sn phase)[47](reproduced with permission).

Because most state-of-art RMB cathodes are Mg free compounds[16,23],anodes often serve as the Mg sources in RMBs.Nguyen and Song[50,51]synthesised Mg2Sn phase alloy using ball-milling technique starting from pure Mg and Sn powder.The obtained Mg2Sn phase was combined with V2O5cathode to obtain full cells,which exhibited good electrochemical performance[50],as shown in Fig.7.Asl et al.[52]also prepared Mg2Sn phase by melting weighted Mg and Sn.The obtained ingot was subsequently crushed,powered and ball-milled using high-energy mill.The obtained micrometer-sized Mg2Sn phase crushed into nano sized Sn particles during the initial magnesiation/demagnesiation cycles,which resulted better performance for the cycles afterwards.

Parent et al.[53]used nanosizedβ-SnSb particles as starting material and found that after several charge-discharge cycles,the particles irreversibly transform into Sn and Sbrich subparticles.In the following cycles,Sn-rich subparticles show highly reversible Mg-storage while Sb-rich subparticles contribute little to the performance.Therefore,Parent et al.[53]stated that good cyclability can be expected in RMBs only if the particle size of Sn anode is smaller than 40 nm in diameter.

Fig.7.Electrochemical characteristics of Mg2Sn anode.(a)Cyclic voltammorgram at the scan rate of 50 mV s?1;(b)Discharge-charge voltage profiles;(c)cycling performance;(d)GITT curves[50](reproduced with permission).

Fig.8.Cell performance of a sputtered Pb film electrode vs.Mg[54].(a)Voltage curve;(b)Cycling performance(reproduced with permission).

Periyapperuma et al.[54]investigated magnesiation of Pb in film and powder form.The Pb film was sputtered on stainless steel foil with an average thickness of 0.24μm.A reversible capacity about 450 mA h g?1(see Fig.8)was obtained using ethylmagnesium chloride electrolyte and Mg foil counter electrode.Periyapperuma et al.[54]also built similar batteries using Pb powder(325 mesh),but only obtained about half of the reversible capacity compare to the ones using Pb foil.It should be noticed that the potential of Pb-Mg alloys is only about 125 mV against pure Mg,which is the lowest among known alloy anodes for RMBs.Therefore,it is possible to obtain higher voltage RMBs by using Pb as the anode material.

3.2.Bismuth

Besides Sn,Bi is also considered as the most promising alloy anode element for RMBs.Through the reaction 3Mg+2Bi?Mg3Bi2,Bi can provide a theoretical specific capacity of 385 mA h g?1.Although the specific capacity of Bi is not as high as Sn,theoretical volumetric capacity of Bi is 3783 mA h cm?3,which makes it a promising anode candidate for RMBs.Additionally,Bi anodes also exhibited good kinetic performance and relative low reduction/oxidation potential for application of RMBs.Ab-initio calculations[44,46]showed that diffusion energy barrier of Mg in the Bi+Mg3Bi2two-phase region locates in the range of 0.37-0.67 eV.However,Murgia et al.[55]reported that the two-phase region for Bi+Mg3Bi2is characterized with an open-circuit voltage at 230 mV against pure Mg anode using GITT technique.Meanwhile,the diffusion barrier for Mg ions to move from the surface to subsurface sites was calculated to be 1.27 eV[46],which indicates that surface modification could be essential for improving the battery performance.

Fig.9.Battery performances of Mg-Bi cells assembled with Bi nanotubes(NT)and microsized Bi powder[58].(a)Cyclic voltammograms of magnesiation/demagnesiation;(b)discharge/charge profile;(c)rate performance;(d)cycling stability and Coulombic efficiency(reproduced with permission).

Many groups have assembled RMBs using pure Bi against Mg electrodes and demonstrated promising battery performances.Arthur et al.[56]applied Bi as the anode of RMBs and demonstrated a reversible capacity around 240 mA h g?1,as shown in Fig.9.DiLeo et al.[57]deposited Bi on C nanotubes and obtained a reversible capacity over 180 mA h g?1.Shao et al.[58]used Bi nanotubes as anode for RMBs and obtained high reversible capacity of 350 mA h g?1with a current at 0.05 C which is 91% of the theoretical capacity of pure Bi.After 200 cycles with 0.02 C current,the RMBs assembled with Bi nanotubes[58]still maintained 303 mA h g?1reversible capacity.The excellent cell performance was attributed to the formation of even smaller nanoparticles from nanotubes during magnesiation/demagnesiation process.The newly formed nanoparticles remained connecting with each other to keep the charge capacities,while the formed nanopores served as fast ion transportation path.

Besides Shao et al.[58],several groups also reported that microstructures evolution of Bi electrode during operation of RMBs is essential for their performance.Murgia et al.[55]prepared half cells using micrometre particles of pure Bi and Mg3Bi2.During rate-capacity tests,the batteries built using pure Bi powder[55]exhibited even higher capacity than that from Bi nanotubes[58].Murgia et al.[55]confirmed the Bi+Mg3Bi2biphasic mechanism using Operando XRD characterization,and considered that the electrochemical grinding process could act as an in situ self-nanostructuration for Bi electrodes in RMBs.Liu et al.[59]also found the capacity increase when using Bi nanowires as active material for RMBs,and attributed the electrochemical grinding effect to the residual Mg3Bi2phase.A clear nanostructuration of the Bi anodes has been reported by Ramanathanet al.[60]by using TEM to investigate the morphology evolution of Bi anode,as shown in Fig.10.Benmayza et al.also investigated the thermal generation process for half cells with Bi anode during charge/discharge process,separated the total heat into irreversible and reversible part,plus the“other”part,which was attributed to pulverization/reformation of Bi material.A complex model to describe the porous electrode of Bi anode was later developed by the same group[61].

Although Bi anodes exhibit the self-nanostructuring property,electrodes prepared using nanosized Bi powder still exhibit better rate-capability as well as cycling stability in RMBs than that using bigger Bi particles,as shown in Fig.11[62].Nevertheless,Kravchyk et al.[62]observed that the Bi anode with 10 nm particles shows poorer capacity and cyclability than that with 18 nm ones,but did not give an explanation.Penki et al.[63]synthesized bismuth(Bi)/reduced graphene oxide(RGO)nanocomposites by using in situ reduction.The electrode made from composite of 60 % Bi and 40 % RGO delivered a specific discharge capacity of 413 mA h g?1,which is even higher than the theoretical capacity of Mg3Bi2(385 mA h g?1),in the first cycle.It also retained 372 mA h g?1in the 50thcycle.Penki et al.[63]attributed the good cyclability to RGO,which can minimize the mechanical stress during magnesiation/demagnesiation process.Wang et al.[64]prepared bismuth oxyfluoride(BiOF)using a solvothermal method.After the first discharge,during which metallic Bi was generated in situ,the BiOF delivered a reversible Mg storage capacity of 335 mA h g?1and 273 mA h g?1at 15 mA g?1and 300 mA g?1,respectively.

Fig.10.TEM images depicting changes in the structure of Bi during the processes of magnesiation/demagnesiation,(a)pristine Bi,(b)50% magnesiated Bi,(c)and(d)100% magnesiated Bi(i.e.Mg3Bi2),(e)50% demagnesiated.Mg3Bi2,(f)(g)and(h)100% demagnesiated Mg3Bi2(i.e.Bi)[60](reproduced with permission).

Fig.11.(a)Rate-capability and(b)cycling stability measurements of Mg-ion half-cells using nano(10 and 18 nm,“NPs”)and micro sized(0.5-2μm,“bulk”)Bi anodes[62](reproduced with permission).

Due to the promising battery performance,attempts have been dedicated to reducing cost for large scale production of Bi anodes for RMBs.Tan et al.[65]sintered Mg3Bi2alloys by heat treating Bi and Mg powders at 300,500,650 and 750°C.650°C was found suitable for obtaining pure Mg3Bi2phase alloy,which could deliver reversible capacity up to 350 mA h g?1.Lower temperatures were found not sufficient for the reaction and diffusion,while 750°C caused too much Mg evaporation.

3.3.Antimony

Sb and Bi share many similarities,such as(1)both Sb and Bi are rhombohedral in crystal structure;(2)Sb and Bi can form all proportional solid solution;and(3)both Sb-Mg or Bi-Mg phase diagrams contain similar intermetallic compounds(Mg3Sb2and Mg3Bi2).Therefore,different groups also investigated Sb with the hope of finding another good anode candidate for RMBs with theoretical capacity of 660 mA h g?1.

Arthur et al.[56]used Bi,Sb,and Bi-Sb alloys as RMB anodes to investigate their electrochemical performance.All alloys could get magnesiated(Fig.12(a))and reach close to their theoretical capacities at 0.01 C current(Fig.12(b)).However,the pure Sb anode showed poor cyclability at 1 C rate while Bi-containing anodes still retain most of their magnesiation/demagnesiation capacities(Fig.12(c)).Parent et al.[53]found that theβ–SnSb nanoparticles transformed into Sn and Sb-rich subparticles during the first several chargedischarge cycles.The Sn subparticles exhibit highly reversible Mg-storage.However,the Sb-rich subparticles showed little contribution to the charge-discharge capacity in the following cycles[53].

Fig.12.(a)Galvanostatic discharge profile of Bi,Sb,Bi0.88Sb0.12 anodes at a 0.01 C rate.(b)Observed capacity of the alloy anodes with a linear fit of these capacities;dashed line compared with the theoretical capacity;solid line.(c)Cycling capabilities of Bi,Sb,Bi0.88Sb0.12 anodes at a 1 C rate[56](reproduced with permission).

3.4.Others(In,Al,Li,Ga)

Murgia et al.[66]used micrometric In and MgIn powders as active material to study their battery performance against pure Mg.MgIn powders were prepared using high-energy ball milling technique;Mg-organohaloaluminate electrolyte was applied for the electrochemical investigations.Up to 425 mA h g?1revisable capacity was achieved at 0.02 C discharge rate for In powders.Good cyclability were also demonstrated with both In and MgIn initial active materials,as shown in Fig.13.However,the battery performance decreased drastically with increasing current rate.Further investigations are required for the application of In anodes in RMBs.

Sivashanmugan et al.[67]used Mg-13Li alloy as anode material and declared that Mg–Li/MgCl2/CuO cells could offer higher operating voltages and capacities than Mg–Al alloys[68].

Wang et al.[69]took advantage of low melting point of Ga(27.6°C)and introduced liquid phase during RMB operation process.Coin cells with initial setup of Mg2Ga5/APC+LiCl/Mg exhibited unprecedented long-cycle of 1000 cycles at high(dis)charge rate of 3 C at 40°C with a capacity above 200 mA h g?1,and coulombic efficiency above 95%.The reversible reaction between liquid Ga and Mg at 40°C is given in Eq.(1).Coin cells with the same setup showed almost no cyclability at 20°C,which is below the melting point of Ga.Obviously,the presence of liquid Ga during battery operation was essential for the high-rate and long cycle-life RMB.Phase diagram of Ga-Mg system and schematic explanation of the self-healing mechanism is given in Fig.14.In order to avoid all-sold reactions,the maximum Mg content in the Ga electrode was designed to not exceed Mg content of the Mg2Ga5phase.

Fig.13.Cycling performance with In/MgIn electrolyde in half cell RMBs[66].(a)In powder,0.02 C;(b)MgIn powder,0.01 C(reproduced with permission).

Fig.14.(a)Equilibrium phase diagram of the Ca-Mg system.(b)Schematic of magnesiation/demagnesiation process during which liquid Ga and solid Mg2Ga5 reversibly converted into each other at constant temperature and pressure[69](reproduced with permission).

3.5.Multi-element alloys

One important hindrance for development of alloy anodes for LIBs is the volume expansion(280 % for Li15Si4and 244 % for Li17Sn4against Si and Sn,respectively)during lithiation process[70].Volume expansion can introduce internal stress to the electrodes and lead to cracks pulverisation of the anode materials during cycling.This will cause fast capacity decay to the total cell.By using multi-element alloys as anodes,more reaction steps will be involved during lithiation/delithiation process thus sudden change of lattice parameters/molar volumes can be avoid.Consequently,internal stress can be reduced and the lifetime of the complete cell can be prolonged.Similar strategy has been applied for the development of alloy anodes for RMBs.The combination of most investigated elements:Bi,Sn,and Sb,have become nature selections.

3.5.1.Bi-Sn alloys

Niu et al.[71]used chemical dealloying of rapid solidified Mg-Bi-Sn alloys to obtain nanoporous(NP)Bi,Bi6Sn4,Bi4Sn6,and Sn alloys starting form Mg90Bi10,Mg90Bi6Sn4,Mg92Bi8,and Mg90Sn10,respectively.NP Bi6Sn4and Bi4Sn6exhibited greatly improved battery performances for both rate capacity and cyclability compare to the NP Bi anode does.Bi4Sn6showed faster capacity decay compare to Bi6Sn4,and Niu et al.[71]attributed it to higher Sn content.Although Bi6Sn4exhibited prior performance than Bi4Sn6,the contribution of Sn to total capacity showed slower fading than that of Bi for the Bi-Sn alloys.After 200 cycles with Bi6Sn4electrode,the capacity fade related to Bi and Sn reactions are 37.9% and 22.7%,respectively.Based on their own results,Niu et al.[71]declared that the duo phase alloy likely“unlocks the kinetic limitation of Sn phase and grants Sn high reversibility to some extent”.By chemical deploying rapidly solidified Al-Bi-Sn with 95 at.% Al content,Niu et al.[72]prepared highly porous NP Bi-Sn alloys,among which the Bi6Sn4composition exhibited the highest reversible capacity,i.e.353 mA h g?1and 324 at mA h g?1at current of 100 mA g?1and 1000 mA g?1,respectively.Fig.15 shows the currently rate performance of porous alloy electrodes at different rate conditions[72].

Song et al.[73]prepared Bi-Sn alloy film at its eutectic composition using magnetron co-sputtering method.The obtained Bi-Sn film delivered high specific capacity(538 mA h g?1at 50 mA g?1current),good rate performance(417 mA h g?1at 1000 mA g?1current)and cycling stability(233 mA h g?1in the 200thcycle).Song et al.[74]also prepared Sn-Bi films with different compositions(Sn97Bi3,Sn82Bi18,Sn67Bi33,Sn33Bi4)by magnetron co-sputtering and rolling melt alloys and pure Sn.Sn-Bi electrodes demonstrated improved electrochemical performance than pure Sn electrode.Sn53Bi47electrode could deliver a specific capacity of 200 mA h g?1at a current of 1000 mA g?1.

Fig.15.Rate performance of porous Bi-Sn electrodes at different rate conditions[72](reproduced with permission).

3.5.2.Bi-Sb alloys

Although Sb shows little reversible Mg-storage capacity,Sb addition to Bi or Sn can improve their battery performance within RMBs.As reported by Arthur et al.[56](Fig.12(c)),BixSb1?x(x=1,0.55,0.88,0)alloy electrodes delivered 222,235,298,and 16 mA h g?1initial discharge capacity at 1 C rate,respectively.Especially,Bi0.88Sb0.12expressed higher initial capacity compare to pure Bi electrode(see Fig.12).However,after 100 cycles,the Bi0.88Sb0.12electrode faded to 215 mA h g?1while Bi anode exhibited 247 mA h g?1specific capacity.The increased fade progress was attributed to the increased Mg-Sb bond strength compared to Mg-Bi bond[54,75].However,if we look close to the trend with the Bi0.88Sb0.12and Bi in the Fig.12(c),the Bi-Sb alloy may show prior battery performance for prolonged battery tests.This effect could be explained with the phase evolution process of the Bi-Sb alloy proposed by Murgia et al.[76],with an example of SbBi composition given as Eq.(2):

The decomposition of Mg3SbBi was confirmed by operando XRD.If we look into the microstructure of an alloy after a eutectoid reaction,very fine even nano-structure can usually be observed.The elevated cyclability was attributed to combination of the eutectoid decompose of Mg3SbBi and nanostructuration of Bi.

Fig.16.Charge-discharge profiles of a Mg-SnSb cell tested with the voltage of 0～0.8V(vs.Mg)[77](reproduced with permission).

3.5.3.Sb-Sn

Parent et al.[53]used nanosizedβ-SnSb particles as starting material to build RMBs.After several charge-discharge cycles,the particles irreversibly transformed into Sn and Sbrich subparticles.In the subsequent following cycles,Sn-rich subparticles showed highly reversible Mg-storage,while Sbrich subparticles contribute little to the performance.The difficult demagnesiation of Sb was attributed to strong Mg-Sb bond[75].Neverthless,the SnSb alloy prepared by Cheng et al.[77]show obvious“activation”process during the magnesiation/demagnesiation process,as shown in Fig.16.The activation effect was attributed to the formation of porous networks of Sn-rich and Sb-rich subparticles,whose microstructure evolution is given in Fig.17.

3.5.4.Other alloying elements(In,Cu,Pb)

Murgia et al.[78]prepared InBi compound using high energy ball milling.The obtained InBi compound showed reversible phase conversion reactions during magnesiation/demagnesiation process.For the simply mixed In and Bi powders,only pure elements were observed during demagnetization process without formation of InBi compound.Therefore,the alloying process using high energy ball milling process was critical for the activation of In together with Bi.Total magnesiation/demagnesiation reaction for InBi compound is given as Eq.(3):

The previous results[53,56]suggested that Sb is not a promising anode candidate for RMBs.However,Blondeau et al.[79]recently reported that In can unlock the reversibility of Sb.Blondeauet al.prepared InSb compound using ball milling technique.During the first magnesiation process using the InSb compound,only one plateau around 90 mV appeared.However,the characteristic voltage plateau of magnesiation/demagnesiation of Mg3Sb2around 400 mV was observed in the following cycles.This indicates that phase transformations appeared in the electrode.However,XRD results indicated the reformation of InSb compound during the demagnesiation process.Additionally,slow capacity increment was observed in the following cycles,which may be related to ongoing morphology evolution.

Fig.17.morphology evolution for SnSb particles in half cell RMB:(a)pristine(b)magnesiated;(c)demagnesiated[77].

Fig.18.CV profiles for(a)Sn,(b)Pb0.4Sn0.4,(c)In0.5Sn0.5 and(d)the 10th cycle of electrode[80](reproduced with permission).

Kitada et al.[80]investigated Cu-Sn,Pb-Sn,and In-Sn alloys as possible alloy anodes for RMBs.Pb-Sn and In-Sn alloys showed improved battery capacity and kinetic performance compare to pure Sn electrode,as shown in Fig.18.Contradictory,Cu-Sn compounds stayed relative inert for the Mg intercalation/deintercalation.

4.Insertion anodes

Beside alloy anodes,which usually include complex phase transformations during magnesiation/demagnesiation process,several intercalation type anodes with which no phase transaction occurs during the battery operations were also proposed.

Several groups have studied carbon-based anode materials for RMBs.Maeda and Touzain[81,82]used highly oriented pyrolytic graphite(HOPG)combined with MgCl2dissolved in dimethylsulfoxide(DMSO),and found reversible electrochemical insertion of Mg2+into the HOPG.Kim et al.[83]investigated cointercalation of Mg2+ions into natural graphite with linear ether solvents,diethylene glycol dimethyl ether(DEGDME)and 1,2-dimethoxyethane(DME)experimentally and performed ab initio calculation to show DEGDME is the most advantageous for cointercalation due to its high binding energy with Mg2+ions(10.05 eV).Er et al.[84]studied the adsorption of divalent Mg ions on defective graphene and graphene allotropes by first-principles density functional theory.A theoretical capacity up to 1042 mA h g?1was proposed with graphene including 25% divacancy defects.Lim et al.[85]prepared three-dimensional macroporous graphitic carbon nanosubstrates which demonstrated improved rate capacity as well as cyclabilities.

Table 1Summary of experimentally measured specific capacities of alloy anode materials for RMBs.NP–nanopowder;NW–nanowire;NT–nanotube;P–porous.;F-film.

(continued on next page)

Table 1(continued)

(continued on next page)

Table 1(continued)

Although the usage of carbon-based material facilitate the electrolyte adoption,low specific charge and short cycle could happen due to the co-insertion process of solvent molecules(solv)as shown in Eq.(4)[86].Therefore,following investigations are required,especially on the cycle life aspect.

Chen et al.[87]applied Na2Ti3O7as anode material RMB.During the first magnesiation process,Mg substituted half of the Na positions in Na2Ti3O7following reaction(5).Afterward,half amount of the Mg integrations could be reversibly inserted & extracted following reaction(6).Na2Ti3O7delivered reversible capacity of 78 mA h g?1(theoretical capacity 88 mA h g?1)for over 100 cycles under 20 mA g?1current.It also demonstrated good rate performance up to 1000 mA g?1current,under which the reversible capacity was 40 mA h g?1.Furthermore,full Mg-ion batteries(MgNaTi3O7/Mg-Mg(ClO4)2-diglyme/V2O5)delivered an energy density of 53 Wh kg?1.

Wu et al.[88]assembled battery using Mg and Li4Ti5O12electrodes with Mg(AlCl2BuEt2)2/THF electrolyte.A revisable capacity around 175 mA h g?1was obtained under 15 mA g?1current.During the charge process,Li atoms were extracted from 8a sites of Li4Ti5O12.During the following discharge process Mg atoms entered empty 8a sites which was originally occupied by Li atoms.Some Mg atoms also entered previously empty 16c sites.By adding 0.25 M LiCl into the Mg(AlCl2BuEt2)2/THF electrode,Li4Ti5O12delivered reversible capacity around 175 mA h g?1with small capacity decay of 0.01% per cycle for over 500 cycles.With the addition of LiCl to the system,Li4Ti5O12could deliver a reversible capacity around 120 mA h g?1under current rate of 300 mA g?1.While without LiCl addition,only around 50 mA h g?1could be obtained under the same conditions.The improved battery performance was attributed to co-intercalation of Li and Mg into Li4Ti5O12electrode.[89].

Zeng et al.[90]firstly reported the possibility of using Li3VO4as the insertion type anode for RMBs.However,the Mg insertion kinetic is sluggish with Li3VO4material.To conquer this problem,Zenet al.applied mesoporous Li3VO4/C hollow spheres,obtained 318 mA h g?1for the first cycle,and maintained 57 % at the 15thcycle.Black phosphorus was also predicated to be able to deliver a high capacity as well as high rate capacity due to the covalent Mg-P bonding[91].

5.Summary and perspectives

RMBs have attracted constantly increasing attention as one of the most promising post lithium battery systems.Because of the possibility of using pure Mg as the anode material,most studies focus on the development of advanced cathode materials.However,a good cathode material must combine with compatible anode and electrolyte for its full usability.The widely studied Grignard-based electrolyte solutions are compatible with pure Mg anodes without formation of thick passive layers.However,the Grignard-based electrolyte solutions are limited by narrow voltage window as well as toxicity.The newly developed fluorinated alkoxyborate-based electrolytes exhibited a high anodic stability,high ionic conductivity and high coulombic efficiency of magnesium deposition.However,high-performance anodes beside pure Mg are always demanded to futher broad the choice of usable electrolytes,cathodes,as well as other cell components.

The most investigated alternative anodes for RMBs are alloy anodes based on Bi,Sn,Sb,In,Pb etc.,where Bi and Sn seem to be more promising than the others.These alloy anodes allow usage of conventional electrolytes,but also have shortages such as slow Mg2+transport and significant volume change during magnesiation/demagnesiation process.The slow kinetics can be overcome by reducing particle size or by modifying the morphology.The volume change is mostly combined with phase transformation,which means both crystal structure and molar volume of formed intermetallic phases change drastically at certain compositions during the battery operations.The drastic volume change usually leads to pulverization of the anode materials and battery capacity fading for LIBs.In contradictory,the volume change often cause formation of continuous nanopores instead of pulverization for RMB alloy anodes.This nanomerization process can improve Mg2+transportation inside the anode part as well as cyclability of the entire cell.Additionally,it can be enhanced by combining different alloy elements,e.g.by adding Sb or In into Bi or Sn,resulted from variation of phase transformation sequences.Initial morphology of the anode materials can also greatly influence their subsequent performance.Therefore,multi-element alloy anodes with carefully designed initial morphology and reaction routine may further enhance the performance of alloy anodes for RMBs.Additionally,the novel Ga anode,which involves liquid phase during cycling process,demonstrated the possibility of self-healing for RMB alloy anodes.The reported specific capacity,rate capacity,and cyclability of alloy anodes for RMBs are summarized in Table 1.

Although fewer studies than alloy anodes were reported,insertion anodes already showed promising potential for RMBs with various mechanisms.Mg2+can be inserted/extracted with Li3VO4anode material,which may be enhanced by doping in future studies.Besides development of complete new materials,doping with elements offers further potential for improving the already found anode materials.

During the first several“demagnesiation”process,Na in Na2Ti3O7or Li in Li4Ti5O12can be partially extracted from the initial materials.Given proper Mg sources,Mg ions will refill the left vacancies during the next magnesiation process.The implanted Mg ions stay active in the following cycles to provide reversible cyclability for the entire RMBs.The Mg-containing anodes after the substitution process are most likely in metastable states,which can hardly be obtained through normal synthesis process.Therefore,this substitution process provides a unique direction for extending the selection range of RMBs anodes.However,long-term electrochemical tests are required to validate the stability of the substituted anodes for RMBs.

In summary,although anodes are less reported than cathodes or electrolytes for the RMBs,this doesn’t mean that the corresponding studies should be abandoned.In contrast,development of high-performance anodes is essential for the future of novel RMBs.More studies on looking for potential host materials as well as proper combinations with other battery components are highly demanded.

Acknowledgments

This work is financially supported by the German Research Foundation DFG project(LI 2839/1-1)and National Natural Science Foundation of China(51971044).MF acknowledges funding from EU research and innovation framework programme via“E-MAGIC”project(ID:824066).This work contributes to the research performed at CELEST(Center for Electrochemical Energy Storage Ulm-Karlsruhe).