Investigation on the formation of Mg metal anode/ electrolyte interfaces in Mg/S batteries with electrolyte additives † Journal of Materials Chemistry A PAPER

The combination of sulfur (S) as cathode with a Mg anode is a promising approach for batteries because of the high theoretical capacity (1672 mA h g (cid:1) 1 ) as well as the abundance and relatively low cost of these elements. However, up to now Mg/S batteries face serious drawbacks like a large overpotential between charge/discharge cycles, rapid capacity fading, poor cycling e ﬃ ciency and slow kinetics, etc. Most of these problems are strongly correlated with polysul ﬁ de dissolution in the electrolyte and their subsequent shuttling to the Mg anode side. In this work, we investigate the Mg 2+ deposition/dissolution process on a Mg metal electrode using a non-nucleophilic magnesium electrolyte (0.4 M Mg[B(h ﬁ p) 4 ] 2 / DME) in the presence or absence of polysul ﬁ des. The large overpotential increase of the Mg metal anode in the presence of polysul ﬁ de species is reduced by the addition of optimum concentration of iodine additives to the Mg electrolyte, which help in the formation of a favorable, less corrosive, uniform and stable interfacial layer. In addition to the electrochemical studies, this work examines the chemical composition, thickness and morphology of the interphase layer formed on Mg metal by using X-ray photoelectron spectroscopy (XPS) with depth pro ﬁ ling and focused ion beam-scanning electron microscopy (FIB-SEM) techniques. Furthermore, the stability of di ﬀ erent MgX phases (X ¼ S, I 2 and F 2 ) formed between the Mg metal anode and di ﬀ erent additive containing electrolytes and Mg di ﬀ usion through these interphases are investigated by density functional theory (DFT) calculations.


Introduction
Although Li ion batteries are presently considered as the most promising technology for electric vehicle propulsion, the development of alternative and complementary battery chemistries and technologies is of great importance, especially aiming at large-scale applications, e.g., for stabilizing the power grid, where the cost (V per (kW h)) and sustainability are crucial indicators. Indeed, the implementation of Li based technologies on a large scale are increasingly questioned, since the controversial debates on lithium, cobalt, and nickel availability and moreover the cost cannot be ignored. 1 Amongst different battery chemistries, appealing alternatives involve the use of multivalent cations. Magnesium (Mg) is especially interesting due to the abundance of Mg raw materials, its low cost as compared to Li, high safety due to a lack of dendrite formation, non-toxicity, and stability. 2 Performance wise, the standard reduction potential for Mg is À2.37 V vs. SHE (standard hydrogen electrode) and the two electron reduction process results in a high theoretical energy density (Mg: 3837 A h L À1 as compared to Li: 2046 A h L À1 ). Even though most of the experimental and theoretical studies demonstrated a dendrite free deposition of Mg, 3,4 some recent reports showed the formation of uneven structures during Mg electrodeposition under certain circumstances. 5,6 This hints at an impact of the electrochemical environment (electrolyte concentration, coordinating ligands, etc.) and the applied current density on the growth morphology resulting during Mg electrodeposition.
The development of Mg batteries has been hindered for a long time by the lack of suitable electrolytes with high ionic conductivity and large voltage window, wherein Mg ions could reversibly deposit and dissolve. 7 Earlier studies showed that Mg could not be reversibly deposited from the solutions of simple magnesium salts such as Mg(ClO 4 ) 2 in conventional organic solvents. 7,8 The main reason for this is the formation of a dense blocking layer at the surface, due to the deposition of the reduction products formed by the reaction of Mg metal with various anionic electrolyte species, such as ClO 4 À , BF 4 À , etc. 8 The use of electrolytes with Grignard reagents (R-MgX, R ¼ alkyl; X ¼ Br, Cl), which are intrinsically strong reducing agents with limited oxidative stability and a high nucleophilicity, [9][10][11] requires cathode materials that are chemically inert, for example Mg x Mo 3 S 4 Chevrel phase cathodes. However, the high molecular weight of Chevrel phase cathode materials results in a low energy density ($60 W h kg À1 ) which makes it less attractive. 12 The well-known Mg(TFSI) 2 based electrolyte system is less interesting in rechargeable Mg batteries due to its large overpotential and unstable SEI formation resultant from the partial decomposition of the Mg(TFSI) 2 salt. 2,6,13,14 The combination of a Mg anode with conversion cathode materials such as sulfur (S) is promising due to their high theoretical capacity of 1671 mA h g À1 (3459 mA h cm À3 ). 15 But the high energy density sulfur cathode (electrophilic in nature) in rechargeable Mg batteries necessitates a non-nucleophilic electrolyte. 16 Previous electrochemical studies on rechargeable Mg/S batteries pointed out problems like large over-potentials, rapid capacity and voltage fading, slow kinetics, selfdischarge, 17 etc. where most of them are connected to the Mg polysulde (Mg PS) dissolution to the electrolyte. 18,19 Past efforts to make rechargeable Mg/S batteries mostly focused on the development of non-nucleophilic electrolytes, leaving it necessary to prevent the oxidation of electrolyte when getting into contact with the electrophilic sulfur. Recently our group developed a non-corrosive, chlorine-free Mg electrolyte with high ionic conductivity at room temperature ($11 mS cm À1 at $23 C), which is based on a uorinated alkoxyborate magnesium salt (Mg[B(hp) 4 ] 2 ). 20 With this electrolyte, an improved electrochemical performance was observed for a Mg/S cell with low cathode sulfur loading (<0.5 mg sulfur cm À2 ), 19 while higher sulfur loading on the cathode severely affected the cell performance and augmented the dissolution of polysulde to anode side. 21 Even though there have been some recent reports on Mg anode/electrolyte interfaces, most of these studies were performed with electrolytes suitable for Mg-ion batteries rather than Mg/S batteries. 22,23 Recently, we veried that the Mg metal anode/electrolyte interface has a higher contribution to the resistance of a Mg/S cell compared to that of the sulfur/carbon composite cathode, and this is also correlated with the poly-sulde dissolution from sulfur cathode. 21,24 Therefore, a thorough understanding of the interphase evolution and Mg ion dynamics at the Mg anode/electrolyte interfaces in the presence of sulfur species and different electrolyte additive species is important. Usually, the electro-deposition on metal anodes proceeds through a "nucleation and growth" mechanism, which consists of a series of consecutive steps: (i) diffusion/ mass transport of the complexed cation to the metal anode side through a liquid electrolyte, (ii) adsorption of the complexed cation combined with partial de-solvation and charge transfer, (iii) ad-atom surface diffusion on the electrode surface, (iv) formation of adatom clusters (nucleation) and (v) subsequent two-and three-dimensional crystal growth. 25 The activation energy barriers for each step in the metal deposition/ dissolution process determine the kinetics of the whole process and the morphology of the resulting lm. Compared with monovalent cations like Li + , the electro-deposition kinetics of multivalent ions like Mg 2+ are expected to be more hindered due to higher energy barriers for diffusion in the electrolyte and desolvation. 7 The electrolyte formulation can therefore play a major role in Mg batteries to achieve a fast cation mobility and high power density. Different ways to improve the Mg 2+ mobility in electrolytes include (i) the use of cation complexing and anion encapsulating agents, and (ii) the proper selection of solvents used in the electrolytes. 7 The tailoring of Mg metal/ electrolyte interfaces is possible by (i) the addition of additives to the electrolyte, which should help in the formation of fast Mg 2+ ion conductive and stable passivation layers, 26 4 ] 2 /DME electrolyte, can reduce the parasitic side reactions on a Mg metal anode during the prolonged cycling. 28 Li et al. illustrated the activation and stabilization of Mg anode by using a dual-salt Mg 2+ /Li + electrolyte. [29][30][31] Cao et al. improved the full cell performance of Mg ion batteries by creating a magnesium uoride (MgF 2 ) surface layer on a Mg anode. 32 In another study, Chen et al. investigated various coating materials for both anode and cathode electrodes in Mg batteries using density functional theory (DFT) based calculations and found that Mg-halides like MgBr 2 and MgI 2 or MgSe are promising candidates for anode coating. 33,34 In this work, we have initially investigated the Mg 2+ deposition/dissolution electrochemical process on a Mg metal anode in a non-nucleophilic magnesium electrolyte (Mg [B(hp) 4 ] 2 /DME) in the presence and absence of polysulde species. Later, the effect of the addition of small concentrations of iodine to this Mg electrolyte was investigated in the context of the large overpotential for Mg electrodeposition on Mg metal anodes in the presence of polysulde species, where the addition of iodine may help for the formation of a stable Mg ion conductive interfacial layer. Along with the electrochemical studies, the present work examines the chemical composition and morphology of the passivation layer formed at the electrolyte/electrode interfaces in the presence of various additives. Finally, we theoretically checked the thermodynamic stability of the interphases, which would form with various electrolyte additive species, and the Mg diffusion barrier through the MgI 2 interphase.

Results and discussion
In order to illustrate the effect of various additives, we rst conducted cyclic voltammetry (CV) measurements using a 3electrode cell (PAT-Cell from EL-CELL GmbH) at a scan rate of 25 mV s À1 with a Pt working electrode, a Mg counter electrode and a Mg ring as the reference electrode. Fig. 1 illustrates the CV proles for (a) Mg electrolyte ((Mg[B(hp) 4 ] 2 ) in dimethoxyethane (DME) solvent), (b) Mg electrolyte containing iodine as additive, (c) Mg electrolyte containing polysulde species as additive, (d) Mg electrolyte containing both polysulde and iodine as additive. The CVs show an increase of the oxidation/ reduction peak current density from the initial to the 10 th cycle and this may be at least partly due to self-conditioning of the electrolyte and activation of the Mg metal surface by modication of the passivation layer. 35 Table 1 illustrates the onset potentials and peak currents (I p ) from CVs for the initial and selected subsequent cycles of Mg plating process on the Pt working electrode using electrolytes with different additives. The initial onset potential for the pure Mg electrolyte was reduced from À383 mV to À227 mV aer 10 cycles. The addition of iodine to the pure Mg electrolyte caused the onset potential to decrease to À280 mV in the initial cycle and further to À214 mV aer 10 cycles. Similarly, the peak currents for plating also increased aer the addition of iodine to the Mg electrolytes. The addition of polysulde species to the electrolyte increased the onset potentials and decreased the peak currents considerably. The onset potentials for polysulde containing electrolyte was À513 mV in the initial cycle and it further increased to À560 mV on the 10 th cycle. The very low values of the peak currents (À6.2 mA cm À2 and À0.70 mA cm À2 for the 1 st and 10 th cycles) aer  the addition of PS to electrolytes prove that the presence of polysulde species is highly detrimental for Mg stripping/ plating. The addition of iodine to the polysulde containing electrolyte reduced the onset potential to À480 mV for the initial cycle and further to À345 mV and À313 mV aer 5 th and 10 th cycles, respectively. Meanwhile, the peak currents increased (À20 mA cm À2 and À24.1 mA cm À2 for the 1 st and 10 th cycles) aer the addition of iodine to the PS containing electrolytes. This increase in peak current suggests that adding iodine to the PS containing electrolyte improves the Mg ionic conductivity of the interphase.
To investigate the long-term cycling performance of Mg electrolytes with additives, galvanostatic cycling was performed using symmetric MgkMg cells. A constant current density of 100 mA cm À2 was applied for 30 minutes during the charge/ discharge half cycle. These test conditions allow for an equal amount of charge shied from one electrode to the other and give a better comparison of the cycling stability of the metallic Mg anode in various electrolytes. Fig. 2 shows the galvanostatic cycling proles for (a and b) Mg electrolyte (0.4 M) and Mg electrolyte with iodine additive, (c and d) Mg electrolyte with a low amount of the polysulde additive, (e and f) Mg electrolyte containing both polysuldes and iodine as an additive. The plating potentials during the 10 th , 300 th , 700 th cycles for Mg electrolytes with various additives are given in Table 2.
The galvanostatic cycling proles (Fig. 2) for pure Mg electrolyte show slightly higher plating potential for the rst cycle, which subsequently decreases to À79 mV in the 10 th cycle. This decrease hints to the formation of an interphase between the electrolyte and the Mg metal anode. 14 Aer 300 cycles, the plating potential for the pure electrolyte was À45 mV, which further increased to À101 mV aer 700 cycles. A similar trend with an increasing voltage hysteresis during the long-term cycling (>300 h) for Mg[B(hp) 4 ] 2 /DME electrolytes with higher concentrations (>0.35 M) was reported also previously. 28,36 These studies concluded on a partial decomposition of the Mg [B(hp) 4 ] 2 electrolyte salt on the Mg metal anode in higher concentrated electrolytes and the formation of an unstable SEI layer during prolonged cycling. 28 Aer the addition of iodine additive to this 0.4 M Mg electrolyte, the plating potential was lower, even aer long cycling (700 h) of the cell, as shown in Fig.  2a, b, g and h. Aer 300 and 700 cycles, plating potentials of À43 mV and À50 mV, respectively, were reached. The lower and stable potential proves that the addition of iodine to Mg electrolyte is benecial and helps to form a stable interphase on the Mg metal electrode.
Adding a low amount of PS to the pure Mg electrolyte and cycling under the same experimental conditions showed a signicant deterioration of the electrochemical responses of the Mg anode, as illustrated in the Fig. 2c, d, g and h. For the initial cycles, the plating potential considerably increased to 550 mV, which reduced in later cycles. The same trend was reported previously also on PS containing Mg electrolytes, 24 and a probable cause for this is that prior to the cycling, Mg metal is reacting with polysulde moieties to form a thin and fragile but highly resistive surface lm. 35 The thickness of the surface lm depends on the amount of polysulde species dissolved in the electrolyte. The formation of a PS containing surface lm on the Mg anode requires a high overpotential to remove this insulating layer and to initiate an electrochemical Mg oxidation/ reduction process at given constant current. 24,37 The breaking of this resistive surface lm during the plating/stripping process causes sudden uctuations and bumps in the voltage hysteresis prole. The addition of more polysuldes to the Mg electrolyte further increased the plating overpotential from À0.58 V to À1 V for the initial cycles (Fig. S2, ESI †).
The addition of iodine to the polysulde containing Mg electrolyte lowered the overpotential signicantly, as shown in Fig. 2e-h. It was À102 mV aer 10 cycles, increasing to À130 mV and À143 mV aer 300 and 700 cycles, respectively. Hence, the iodine additive increased the cycling stability of the polysulde containing electrolyte without any large increase of the overpotential . Furthermore, unlike the cells without iodine additive ( Fig. 2c and d), the iodine containing cells did not show any pronounced overpotential spikes during cycling.
Electrochemical impedance spectroscopy (EIS) was used to characterize the resistances that develop at the interphases of the symmetric MgkMg electrode cells in electrolytes containing different additives; the associated Nyquist plots are presented in Fig. 3. To reveal the underlying physical processes, the EIS data were tted with equivalent electrical circuits as shown in the inset of Fig. 3b. In the tted circuits, R1 indicates the electrolyte resistance and the electronic resistances of the current collectors. The high-frequency region semicircle, represented by a R SEI kQ2 circuit, stands for the bulk resistance of the interphase layer. The low-frequency region semi-circle, represented by a R CT kQ3 circuit, reects the Mg ion migration through the interphase. Since the cells are symmetric, the value of R SEI and R CT for an individual interface is simply calculated by dividing the tted resistances by a factor of 2. The resulting values are given in Table 3. Before any cycling at OCV, a more depressed semicircle was observed (Fig. 3a). The overall interfacial resistance (R SEI and R CT ) of the Mg anode is very high; it drops signicantly aer initial cycling. This indicates the formation of an interphase between electrode and electrolyte during the initial cycling as supported by galvanostatic and voltammetric results. The total interfacial resistances of the Mg anode increased signicantly when using a polysulde containing electrolyte, which points to the formation of a thick, insulating passivation layer. On the other hand, the addition of iodine to the pure Mg electrolyte or to polysulde containing electrolyte decreased the Mg metal anode resistances signicantly due to the formation of a favorable Mg ion conductive interphase. To gain information on the chemical surface composition of the Mg electrodes aer cycling in electrolyte with different additives, X-ray photoelectron spectroscopy (XPS) measurements were carried out. They give the composition and the chemical state (oxidation state) of the elements in the surface layer of Mg electrodes aer prolonged cycling under same test conditions. To compare with cycled state, XPS surface measurements were also conducted for fresh Mg metal electrodes aer dipping (30 minutes) in Mg electrolytes with different additives (Fig. S3, ESI †). Prior to the measurements, all the electrodes were shortly washed with DME solvent and dried under vacuum overnight. Since XPS is a surface sensitive method, which probes only $3-5 nm of the outermost surface layer, measurements were also recorded aer successive Ar + ion sputter treatment in order to get a depth prole of the elemental composition in the surface-near region. XP spectra were recorded and analyzed for all Mg electrodes before sputtering and aer sputter times of 3, 7, and 20 minutes, resulting in total sputter times of 0, 3, 10, and 30 min, respectively. Under the given conditions, the sputter rate is expected to be $1 nm min À1 .  (Fig. 4a, spectra shown in Fig. S6 †) reveals an increase of the carbon concentration upon sputtering, arriving at $70 at% aer 10 min sputtering (removal of $10 nm material) and staying constant aer that. At the same time, the atomic concentrations of oxygen and uorine decrease during the initial 10 minutes of sputtering and stay essentially stable aer that. The Mg concentration remains rather low at all stages of the experiment (e.g., $5 at% aer the last step).
For Mg electrodes cycled in pure Mg electrolyte and Mg electrolyte with iodine additive (Fig. S4, S5 and S7 †), we nd an initial increase of the F concentration as well as a steady increase of the Mg and a concurrent reduction of the C  concentration with increasing sputtering time. Finally, for the electrolyte containing polysulde and iodine additive in combination (Fig. 4b, spectra shown in Fig. 5), we observe a continuous growth of the Mg concentration and simultaneous decrease of the C, O, and F concentrations with increasing sputtering time, i.e., from the surface to deeper layers. Further information is obtained from the binding energies of the different elements recorded aer prolonged cycling periods aer different sputtering times. Here we will focus on the electrode cycled in Mg electrolyte containing polysulde and iodine in combination (spectra in Fig. 5). The spectra recorded on the electrodes cycled in the other electrolytes, in (i) pure Mg electrolyte, (ii) in Mg electrolyte with iodine additive, and (iii) in polysulde containing Mg electrolyte (Fig. S4-S6 †), show qualitatively similar features, but with differences in the quantitative composition and in the trends during sputtering.  Starting with the initial, non-sputtered surface, the C1s spectra show a dominant peak at 284.8 eV, which we attribute to C-C/C-H moieties. Furthermore, peaks at higher binding energies (287.4 eV and 289.0 eV) are assigned to C-O and C]O groups. CF x species from the hp (hexauoroisopropanolate) ligands of Mg[B(hp) 4 ] 2 and its decomposition products lead to the signals at 293.0 eV and 294.2 eV, respectively. Peaks related to the C-O and C]O groups were also observed in the O1s spectra at 532.0 eV (C-O) and 533.1 eV (C]O), respectively. The spectra in the F1s region are initially dominated by a peak at 688.5 eV, which is due to the CF x groups of the hp ligand. In addition, there is a small peak at 685.5 eV, which is attributed to MgF 2 . The sulfur concentration is comparatively low on the unsputtered surface. The position of the S2p peak doublet (162.5/163.7 eV), which is better visible aer sputtering, indicates the presence of small amounts of disuldes (MgS x¼2 ). Finally, in the I3d region we nd a low-intensity peak at $619.3 eV, which according to the literature is typical for iodide species. 38 From these XPS results, we believe that the topmost part of the interphase layer may contain some electrolyte species, with only small amounts of decomposed electrolyte salts.
In Fig. S8 (ESI †), we compare the Mg2s spectra recorded on Mg electrodes cycled in pure Mg electrolyte and in electrolytes with various additives, aer different sputtering times. Since some of the cycled Mg electrodes contained iodine species, we chose to concentrate on the Mg2s peak region to avoid a possible interference in the Mg2p region with the I4p peak. For pure Mg electrolyte, the spectra in the Mg2s (Mg2p also) regions show initially the exclusive presence of oxidized Mg species (peaks at 89.3 eV and 50.6 eV). Aer the second sputter step, i.e., aer removal of $10 nm material, additional peaks appear at lower binding energy (87.3 eV and 48.6 eV), which are characteristic of metallic Mg. Aer removal of $30 nm of material, metallic Mg amounts to $20%, while the rest still consists of oxidized species. We note that the peaks of metallic Mg appear at $1 eV lower binding energy in our measurements compared with what is usually reported in literature. 39 We attribute this shi to a differential charging effect, which shis the peaks of the non-conducting phase and thus also the binding energy reference, but does not affect the conducting Mg phase. For Mg electrodes cycled in electrolyte with iodine additive, the Mg2s spectra show the presence of metallic Mg even aer the rst sputter step (i.e., removal of $3 nm material) and the percentage of metallic Mg is $25% aer removal of $30 nm of the surface layer. For the Mg electrode cycled in polysulde containing electrolyte, the peak of metallic Mg appears only aer 30 minutes sputtering (with $10% of the total Mg signal intensity). These results indicate the formation of a relatively thick passivation layer on polysulde species containing Mg electrolyte. Finally, for the Mg electrode cycled in polysulde and iodine containing electrolyte, the Mg2s spectra show the peak of metallic Mg even aer the rst sputter step (i.e., removal of $3 nm material), its intensity amounts to $45% aer removal of $30 nm of the surface layer. It is interesting to note that the S2p peak doublet recorded on this electrode is shied to lower binding energy (161.6/162.8 eV) compared with the sample without iodine, which could indicate the formation of mono-sulde (MgS) instead of disulde (MgS 2 ) species in the presence of iodine. 39 All in all, these XPS data indicate that the deeper regions of the interphase layer mainly consist of inorganic species, in particular of MgF 2 , MgI 2 , MgO and MgS x , etc.
Owing to the high ionic conductivity ($10 À5 S cm À1 ) and low electronic conductivity of MgI 2 ($10 À9 S cm À1 ), its presence in the Mg interphase layer is highly benecial for less resistive Mg stripping and plating. 38 XPS studies already conrmed the formation of MgF 2 , MgS and MgI 2 species in the interphase of cycled Mg metal electrodes. To better understand the Mg ion transport through these MgX (X ¼ S, I 2 and F 2 ) species, DFT calculations (for more details please see the experimental section) were performed. 40 The formation energy for the insertion (DE MgX,interstitial ) of an Mg atom into the considered MgX phases (X ¼ S, I 2 and F 2 ), as well as the energy required to create a vacancy (DE MgX,vacancy ) that is necessary for the substitutional diffusion were calculated. For MgF 2 , MgS and MgI 2 species, the DE MgX,interstitial values are 6.84 eV, 4.7 eV, and 1.78 eV, respectively. The related DE MgX,vacancy energies are 13.28 eV, 5.29 eV and 4.80 eV, respectively. Hence, the MgI 2 species have signicantly lower energies for the occupation of an interstitial site and the formation of vacancies, both of which are relevant for Mg ion transport, as compared to MgS and MgF 2 . In the interlayer region between the iodine layers, there are two adjacent local minima for the interstitial Mg atom corresponding to a tetrahedral and an octahedral site. The octahedral site is the energetically more favorable one; hence the diffusion path corresponds to the migration from the octahedral via the tetrahedral to the next symmetrically equivalent octahedral site. Hence there is only one possible pathway, as for example in many other materials such as spinel materials. 41 Along this pathway we have performed a nudged elastic band calculation yielding a migration barrier of about 0.25 AE 0.05 eV between the octahedral and the tetrahedral site which thus yields the diffusion barrier along the minimum energy path, as shown in Fig. 6a. Earlier studies reported higher diffusion barriers for the transport of Mg ions in MgS (0.943 eV), MgF 2 (1.123 eV) and MgO (1.851 eV) phases. 34 It also needs to be mentioned that the calculated formation energies for the MgI 2 species are still rather high, indicating that other magnesium iodide compounds might be present in the interphase layer contributing to the facile Mg transport; whose theoretical identication, however, is beyond the scope of the present paper.
To gain information on the morphology of the surface and interphase of the cycled Mg electrodes, we performed electron microscopy (SEM) and focused ion beam (FIB)-SEM experiments. Fig. 6b-d shows electron microscopy images of the pristine polished Mg metal electrode, and of Mg metal electrodes aer extensive cycling in polysulde or polysulde-plusiodine containing electrolytes. The addition of polysulde to the Mg electrolyte completely changed the surface morphology of the deposited Mg. The corrosion of the Mg surface was very obvious in the presence of polysulde species and the extent of corrosion was found to strongly depend on the amount of pol-ysulde species in the electrolyte. Cycling in the electrolyte containing small amounts of polysulde gave a surface with large and irregular deposits with thicker interphase (>500 nm) as shown in Fig. 6c and 7a. Increasing the polysulde content in the Mg electrolyte led to the formation of a highly corroded Mg surface with rectangular nodules and pit-holes as shown in Fig. S9 (ESI †). The addition of iodine to the polysulde containing electrolyte resulted in a much smoother Mg surface (see Fig. 6d, e and 7b), even aer long cycling, indicative of a uniform Mg deposition/dissolution behavior. The FIB-SEM image (Fig. 6e) of the Mg metal electrode cycled in electrolyte with polysulde and iodine additives illustrates the formation of a uniform passivation layer with an approximate thickness of $165 nm. Insets in Fig. 6e shows the energy dispersive X-ray analysis (EDXA) spectrum from the interphase region of Mg metal electrode cycled in iodine and polysulde containing electrolyte. The spectrum clearly illustrates the presence of the chemical species like uorine, iodine, sulfur, oxygen, etc. and these results are in accordance with the XPS results (Fig. 5).
In combination, the XPS and microscopy results indicate that the presence of polysuldes is highly detrimental and creates a thick non-uniform passivation layer that blocks the smooth deposition/dissolution of Mg ions and results in large overpotentials. In contrast, the addition of optimum concentration of iodine additives (MgI 2 formation on the Mg surface) leads to a less corrosive, stable and Mg ion conducting interfacial layer with low thickness (cf. schematic depiction Fig. 7a and b). This interphase can support the uniform, smooth deposition/dissolution of Mg during extensive cycling with low overpotentials even in the presence of different chemical species such as sulfur, oxygen and uorine species. In addition, the XPS sputter proles demonstrate , i.e., the cathode electrode retained 95%, 78%, 49% and 37% of its initial capacity aer the 2 nd , 5 th , 50 th , 100 th cycles, respectively (Fig. 7d). For the initial 15 cycles, the cell gave a Coulombic efficiency of # 97% and then aer it increased to 98.5% for the following 85 cycles. Electrochemical performances of Mg/S full cell can be further improved by optimizing the sulfur cathodes. 43,44

Conclusions
In summary, we have investigated the electrochemical Mg 2+ deposition/dissolution on a Mg metal anode (a) in 0.4 M uorinated alkoxyborate magnesium electrolyte, (b) in Mg electrolytes with either polysulde or iodine additives, and (c) in Mg electrolyte contained both polysulde and iodine additives. The presence of polysulde species leads to a large overpotential at the Mg anode during cycling. Furthermore, microscopic and spectroscopic characterizations reveal signicant corrosion of the Mg metal electrode in the presence of polysulde species and the formation of a highly resistive thicker interphase layer, which increasingly blocks the Mg stripping/plating process. The large overpotential on Mg anodes observed in the presence of polysulde species was considerably reduced upon the addition of optimum concentration of iodine to Mg electrolyte and this also eluded the time consuming electrolytic conditioning behavior. The formation of a MgI 2 layer on the Mg anode helps (i) to reduce the detrimental side reactions of the polysulde species on Mg metal anode and (ii) to form an electrochemically stable, thinner, and uniform Mg ion conductive interphase between the Mg metal and uorinated alkoxyborate magnesium electrolyte. DFT calculations obtained lower vacancy/interstitial formation energies and a very low diffusion barrier of interstitial Mg in MgI 2 compared to other MgX (X ¼ F 2 , S, O) phases, which supports the high mobility of Mg in MgI 2 phases in the interphase. This approach, the use of iodine additives, renders an easy way for the creation of a stable and ion conductive interphase in practical room temperature Mg/S batteries. The present studies furthermore underline that the proper choice of additives can strongly modify and even determine the composition of the solid electrolyte interphase and this way inuence the performance on the Mg anode side. It also opens an avenue for the development of new electrolytes for Mg/S batteries operated with a Mg metal anode.

Experimental section
Synthesis of materials and electrochemical measurements 0.4 M uorinated alkoxyborate magnesium salt (Mg[B(hp) 4 ] 2 ) in dimethoxyethane (DME) solvent was used as Mg electrolyte, it was synthesized according to our previous report. 20 Since ether based solvents are more suitable for Mg battery electrolytes/ electrodes as compared to organic carbonate solvents, due to their higher reduction stability, the electrolyte and all additives for the current study were prepared in DME solvent. 45,46 Mg polysulde (Mg PS) was used as sulfur additive, and synthesized according to our previous report. 19 The concentration of sulfur species in the Mg PS solution was 68 mmol L À1 and determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES). S2p X-ray photoelectron spectra (XPS) of the Mg PS (adsorbed on carbon matrix) showed the S2p peak doublets of bridging and terminal S atoms (163.4/164.6 eV & 161.5/162.7 eV), respectively, as compared to bulk sulfur (164.0/165.2 eV) (Fig. S1, ESI †). For sulfur cathode preparation, nitrogen doped (graphene-multiwall carbon nanotube) hybrid structure (NC) was used as the carbon host matrix. The sulfur was impregnated within the carbon host matrices (S/NC) by melt inltration technique at 155 C under an argon gas atmosphere. A detailed procedure for the synthesis of sulfur cathode was given in our previous report. 21,42 Cyclic voltammetry measurements (CVs) were performed using a three-electrode cell (PAT-Cell, EL-CELL GmbH) at a scan rate of 25 mV s À1 with a Pt disc as the working electrode, a Mg disc as counter electrode and a Mg ring as reference electrode. Galvanostatic cycling experiments for symmetric MgkMg cells were carried out at a current density of AE100 mA cm À2 for 30 minutes with each charge/discharge half cycle. Electrochemical impedance spectroscopy (EIS) for symmetric MgkMg cells was performed with an applied sinusoidal excitation voltage of 10 mV in the frequency range from 1 MHz to 50 mHz. All the cells were assembled inside an argon lled glove box (MBraun GmbH) and electrochemically tested using a Biologic VMP-3 multichannel potentiostats/galvanostats. Initial electrochemical experiments for Mg electrolyte were carried out without the presence of any polysulde species, followed by measurements with the addition of different concentrations of polysulde species, from low ($14 mM) to high ($50 mM) values. In a second series of experiments, we added optimum concentration of iodine-based additives ($45 mM) to Mg electrolyte and performed the same electrochemical experiments in the absence and presence of polysulde species.
For full cell measurements, the cathode slurry was prepared by mixing S/NC powders and the polyvinylidene uoride binder (PVDF, received from Kynar) in the 90 : 10 mass ratio using Nmethyl-2-pyrrolidone (NMP) as solvent. This obtained slurry was cast onto aluminum foil by doctor blade techniques and thereaer dried at 60 C for 24 h. The sulfur loading at the cathode electrode was $0.4 mg sulfur cm À2 . 80 mL optimized Mg electrolyte (0.4 M Mg[B(hp) 4 ] 2 /DME with iodine additive) was used in each assembled cell with two electrode congurations (Swagelok). All the cells were assembled inside an argon-lled glove-box. The electrochemical performance of the full cells was tested by galvanostatic cycling with 0.05C (1C ¼ 1672 mg sulfur cm À2 ) using an Arbin BT-2000 battery tester at room temperature (25 C). The capacities were calculated based on the mass of sulfur in the cathode electrode.
X-ray photoelectron spectroscopy. The chemical state of the sample surfaces was determined by XPS measurements using monochromatized Al K a (1486.6 eV) radiation (PHI 5800 Mul-tiTechnique ESCA System, Physical Electronics). The measurements were done with a detection angle of 45 , using pass energies at the analyzer of 93.9 and 29.35 eV for survey and detail spectra, respectively. To avoid surface contamination, the samples were transferred in inert gas atmosphere to the sample load lock of the XPS system. To compensate for surface charging effects the samples were neutralized with electrons from a ood gun (current 3 mA). For binding energy calibration, the C1s peak of adventitious C was set to 284.8 eV in the measurements before sputtering, while for the measurements on the sputtered samples we used the F1s peak of MgF 2 at 685.5 eV for calibration. Peak tting was done with the CasaXPS program package using a Shirley-type background and Gaussian-Lorentzian peak proles. For the S2p peaks, doublets with a ratio of the peak areas of 2 : 1 and a spin-orbit splitting 1.2 eV were used.
Electron microscopy. Scanning electron microscopy (SEM) was carried out using the instrument LEO GEMINI 1550 VP equipped with a Silicon Dri Detector (OXFORD Instruments). To carry out focused ion beam-scanning electron microscopy (FIB-SEM), the samples were transferred under argon with a Quorum vacuum transfer system, which has been attached to a Zeiss Auriga 60 dual-beam FIB. Cross sections were cut using a 16 nA Ga-ion beam for rough trenches, followed by a polishing of the cross section surface with 600 pA. Imaging was performed using 5 kV e-beam with a secondary electron (SE) detector (Everhart-Thornley-detector) under pixel-averaging mode. EDAX single-spot measurements were carried out with an EDAX-Octane-Super-A detector (60 mm 2 ) using 5 kV hightension for 120 seconds each.
Computational methods used in the DFT calculation. Periodic density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP) 47,48 . Exchange-correlation effects were accounted for within the generalized gradient approximation (GGA) employing the functional of Perdew, Burke, and Ernzerhof, 49 which is well-suited to describe salt properties. [50][51][52] The core electrons were represented by projector augmented wave (PAW) pseudopotentials, 53 as supplied in VASP, 54 with a cutoff energy of 300 eV for the MgI 2 based systems, 400 eV for the MgF 2 based systems, and 300 eV for the MgS based systems. This cutoff is sufficient to reproduce the main properties of the considered materials such as bulk cohesive energy and lattice constants.
The crystalline systems were allowed to fully relax for the system energy calculations. In the case of the interstitial magnesium atom in MgI 2 , only the iodine and magnesium layers directly above and below the interstitial atom were allowed to relax. The system energies were determined within a 4 by 4 by 4 geometry for the MgI 2 systems, a 3 by 3 by 3 geometry for the MgF 2 system, and a 4 by 4 by 4 geometry for the MgS system. The diffusion barrier for interstitial diffusion in the MgI 2 system was calculated in the same geometry as the system energies of this system. All systems were calculated using a 5 by 5 by 5 gamma-centered k-point grid. The diffusion path in the case of the interstitial diffusion in MgI 2 was calculated employing the nudged elastic band (NEB) method, [55][56][57] with nine images along the reaction path. The relaxation in the