Mechanistic Aspects and Side Reactions during Reversible Mg Deposition and Oxygen Reduction on a Pt Film Electrode in BMP-TFSI-Based Electrolytes: A DEMS Study

Reversible Mg deposition/stripping and O 2 reduction/evolution on a Pt film electrode in neat and O 2 -saturated 1-butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl) imide (BMP-TFSI) electrolytes, containing Mg(TFSI) 2 and/or Mg(BH 4 ) 2 as Mg source as well as Mg(BH 4 ) 2 and/or the crown ether 18-c-6 as additive, were investigated by online differential electrochemical mass spectrometry (DEMS) and by scanning electron microscopy/energy dispersed X-ray spectroscopy. Combined cyclic voltammetry and DEMS measurements reveal a complex network of partial reactions, including borohydride electro-oxidation by reaction with water or O 2 , chemical bulk reaction of these components, as well as electro-oxidation of H 2 , and electrolyte decomposition, in addition to the primary reactions Mg deposition/stripping and ORR/OER. They provide detailed insights into the potential dependent reactions occurring under these conditions, demonstrating that also the additive 18-c-6 undergoes decomposition upon reduction of Mg 2 + . Contributions from chemical bulk reactions are resolved by DEMS measurements in borohydride containing solution without a Pt electrode. Electrocatalytic borohydride oxidation, explored by similar measurements with a Pt electrode, can lead to H 2 or H + formation. Under open circuit potential conditions, charge compensation by the ORR results in the formation of a mixed potential. Consequences of these findings for applications in Mg-air batteries are discussed.


Introduction
Multivalent metal-ion and metal-air batteries have a high potential to fulfill the demands for an improved energy density and safe operation. [1]Among the various fundamental problems to be resolved, this requires a molecular-level understanding of the solvation/complexation of the multivalent metal ions, as a basis for a rational design and development of stable electrolytes for these next-generation post-Li batteries. [2,3]In spite of their larger charge density the multivalent metal ions are, however, likely to exhibit a higher tendency for ion pairing or ion clustering, which in turn leads to a lower ion mobility. [2,4]urthermore, they may also enhance the tendency for electrolyte decomposition and thus reduce the stability of the electrolyte, resulting in the formation of a non-conductive, passivating layer at the electrode surface. [2,3]Third, so far there are only a limited number of affordable intercalation materials. [5,6][9] Recently we started a systematic study on the deposition/ stripping of Mg and the reduction/evolution of O 2 from/in a room temperature ionic liquid (RTIL), specifically in 1-butyl-1methylpyrrolidinium bis(trifluoromethylsulfonyl) imide (BMP-TFSI), [10][11][12][13][14][15] which will be continued in the present study.21][22][23][24] However, they also show some disadvantages such as the above-mentioned reduced mobility due to ion pair formation, [25] which is more pronounced for more highly charged metal ions.Furthermore, the stability of these RTILs can be affected by the presence of the metal ions, due to complex formation with the metal ions.This may create a severe obstacle in the RTIL environment due to the high concentration of counter ions.Indeed, reductive decomposition of the TFSI À anion in the presence of Mg had been identified in several studies and proposed as a major reason for the observed passivation of the electrode for reversible Mg deposition/stripping or O 2 reduction/evolution. [10,11,15,[25][26][27] This was attributed to Mg 2 + complex formation, where TFSI anions are coordinated to Mg 2 + , [28,29] and decomposition of TFSI À upon reduction of the Mg 2 + central ion to Mg + .Another problem can arise from trace impurities of water in the RTIL, which lead to the formation of stable oxidic passivation layers. [30]To avoid such effects, additives are added to the RTIL such as BH 4  À , [16,31,32] or the borane dimethylamine complex NBH, [14] Figure 4 which may act as water scavenger, and/or complexing agents such as glymes, [29] ethers, [15,27,33] or also BH 4 À , [17,34,35] which should reduce or even exclude direct interaction between Mg and the RTIL components.Under model conditions, reversible Mg deposition/stripping was shown to be possible in the presence of both BH 4 À , introduced as Mg(BH 4 ) 2 , and the 18-c-6 crown ether as a complexing agent. [15]Alternatively, the RTIL anion can be modified to exclude Mg-induced decomposition, for example, by alkoxyfunctionalization, such that the functionalized anion can displace the TFSI À group from the coordination sphere of Mg 2 + . [36]his allowed reversible Mg deposition/dissolution with high coulombic efficiency, and it was suggested that the coordination sphere of the transient Mg + ions may play a key role in reversible Mg deposition/dissolution.In a combined experimental and theoretical study we could recently show that the crown ether 18-crown-6 (18-c-6) indeed binds more strongly to Mg than the TFSI À anion and can therefore displace TFSI from the inner coordination sphere of the Mg 2 + cations at room temperature. [15]On the other hand, the bond is not too strong, such that Mg deposition is still possible. [15]This led to the conclusion that the optimum additive should have an interaction energy with the central metal ion that is neither too weak, to facilitate displacement of the TFSI anion in this case, nor too strong, to still allow Mg deposition or formation of reoxidizable Mg oxy-species, along the lines predicted by the Sabatier principle. [37,38]Finally it should be noted that a recent study showed that in contrast to the positive effect of 18-c-6 on reversible Mg deposition/stripping the closely related crown ether 15-c-5 actually led to an inhibition of that reaction, which was attributed to a too strong complexation of the Mg 2 + ion. [39]n this work we present and discuss results of a systematic online differential electrochemical mass spectrometry (DEMS) study on the reversible Mg deposition/stripping (in O 2 -free electrolytes) and O 2 reduction (ORR)/evolution (OER) (in O 2saturated electrolyte) on a Pt film electrode during potential cycling in four different BMP-TFSI based electrolytes of different composition, containing different concentrations of Mg 2 + (as Mg(TFSI) 2 and Mg(BH 4 ) 2 ), of BH 4 À (as Mg(BH 4 ) 2 ) and of the crown ether 18-c-6 (see Table 1).This mainly differs from our previous studies [10][11][12][13][14][15] either by the different additive used as water scavenger, with BH 4 À in the present case and NBH in Ref. [14], or by the enforced presence of O 2 in the electrolyte, which was not the case in Ref. [15].Also, different from the purely electrochemical measurements mainly used in our previous studies, [10,11,15] DEMS measurements provide access to the potential dependent formation of gaseous products of the main reactions or of possible side reactions.Furthermore, the present measurements were performed at a fifty times slower potential scan rate than in our previous work with purely electrochemical detection, to be closer to realistic situations in most battery applications, especially those anticipated for Mg-air batteries, where the potential variation during discharge is rather slow.The much slower potential variation may affect the overall performance in a number of different ways.On the one hand, it should reduce mass transport limitations and the Ohmic losses caused by the rather low Mg 2 + concentrations, which result from the low solubility of the chemicals in the BMP-TFSI solvent, by the high viscosity and by the poor electrical conductivity of the latter.On the other hand, it may affect the impact of trace impurities such as residual water or oxygen because of the longer times available for the uptake of poisons.42] We start with characterizing Mg deposition/stripping and gas evolution in the O 2 -free electrolytes.We mainly present these DEMS data as temporal profiles of the potential, of the Faradaic current and of different mass spectrometric traces, as they allow a better identification of correlations between the different features.For better comparison with previous electrochemical studies, however, we also present conventional CVs and briefly summarize their main features in the Supporting Information (SI).The resulting electrodeposits were characterized by scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDS).Results on the gas evolution in a borohydride containing electrolyte under purely chemical conditions, without a Pt electrode, both in the presence and absence of O 2 , and the gas evolution/consumption upon interaction of the different BMP-TFSI based electrolytes with a Pt film electrode under open-circuit conditions, also in the presence and absence of O 2 are presented in the next section.Finally, these latter measurements were continued under potential control, monitoring the Faradaic current and the different mass signals during 3 complete cycles in the following section.With these latter measurements, which allow us to discriminate between electrocatalytic reactions and chemical bulk reactions, we aim at a better understanding of the role of the borohydride additive in the reversible Mg deposition/stripping reaction and ORR/OER.For similar reasons as above, we also in this case present conventional CVs and briefly summarize their main features in the SI, in addition to the temporal profiles.The main results and possible differences to previous findings obtained under slightly different conditions and with a different additive are combined in a comprehensive picture in the discussion section, including also a brief outlook on the consequences of these results for technical applications, followed by a brief summary.

Mg deposition on a Pt film electrode from BMP-TFSI based electrolytes
First we recorded cyclic voltammograms together with online mass spectrometric signals on a Pt film electrode in the four different Mg-containing BMP-TFSI based electrolytes (see table 1).The Faradaic currents are presented as current-voltage plots in the Supporting Information (SI) in Figure S1 and discussed there in more detail.In summary, they show that reasonably stable cycling can be obtained for electrolytes III and IV, at least on the scale of the present experiments, but not for electrolytes I and II.The general trends in the CVs largely resemble those reported previously for a glassy carbon electrode in the same electrolytes [15] when considering the 50fold lower scan rate, which results in much lower current densities and in particular in a much more pronounced passivation in the present CVs per scan, due to the more pronounced formation of inhibiting species.
In Figure 1 we show the time dependence of the potential, of the current density and of selected mass spectrometric ion currents during the first three (four in Figure 1d) potential cycles in the different electrolytes.Before these measurements, the sample was kept under open circuit conditions in this electrolyte for several hours (OCPs see at the different electrolytes).
For electrolyte I (0.1 M Mg(TFSI) 2 + 0.1 M 18-c-6, Figure 1a, final OCP value: 0.95 V), where the Mg 2 + ions are coordinated to the TFSI À anion [28,29] and (predominantly) to the 18-c-6 ring, [27] the current is generally very low.When activating the potential control and stepping from the OCP to the positive potential limit, we find a small positive spike in the Faradaic current (inset in Figure 1a).Note that in the following we will describe this as a potential step to the positive potential limit.In contrast, the m/z = 2 (H 2 ) signal is essentially zero before, during and after the potential step in this borohydride-free electrolyte (Figure 1a).At potentials around the negative potential limit there is a small, but clearly resolved increase of the m/z = 2 ion current, indicating H 2 evolution.This occurs in parallel to the significant increase in Faradaic current density in this potential range.The asymmetric current peak shape (see also Figure S1a) indicates, however, that this peak is not only due to H 2 evolution, but also contains contributions from other irreversible processes.In the subsequent cycles, the H 2 signal remains at the noise level, which fits with the much smaller electrochemical current in these cycles.H 2 evolution can result from either reductive decomposition of the electrolyte and/or from reduction of trace impurities of water.Both are expected to lead to a passivation of the electrode, due to the formation of adsorbed electrolyte fragments and/or Mg (hydr)oxy species.The other ion currents depicted in Figure 1a are essentially featureless.
For electrolyte II (0.1 M Mg(BH 4 ) 2 + 0.1 M 18-c-6, Figure 1b, Figure S1b, final OCP value: ca.0.6 V), the Faradaic current densities are much higher.Furthermore, we find a small oxidation current with a symmetric peak centered at ca. 0.25 V, which is likely due to Mg stripping.The apparent Coulombic efficiency of this process (compared to the cathodic current in the Mg deposition regime) is, however, still rather poor, indicating either irreversible deposition processes and/or a passivation of the electrode as dominant reductive processes.Also in this case the H 2 signal was essentially zero, i. e., at the background level before the potential step, but increases when stepping from the OCP to the upper potential limit (see inset in Figure 1b).A similar increase is observed also for the Faradaic current, though on a lower scale than in electrolyte I, while the other signals do not change much.Both H 2 formation and Faradaic current decay rather quickly and almost reach the background level after 1000 s, at a potential slightly above 0.8 V.Here it is important to note that contributions to the Faradaic current resulting from capacitive charging are limited to the very initial phase, below 1 s. [12]bviously, an increase in the H 2 formation rate due to reductive H 2 formation fits neither with the initial increase in the oxidation current nor with the high anodic potential of around 0.8-1.0V.This apparent discrepancy can be resolved in two different ways.In the first case, H 2 formation results from the well-known chemical bulk oxidation of borohydride by reaction with trace impurities of water [43] according to In addition to the formation of (insoluble) Mg borates along reaction (1), also the formation of hydroxylated species such as borotetrahydroxylates is possible, as described by eq.(1a) For simplification, we will assume in the following that this reaction leads to borate formation, although small contributions from hydroxylated species cannot be ruled out, though it would require twice as much water per borohydride anion.Such kind of bulk reaction should, however, have started already upon preparing/resting the electrolyte, and not only when stepping the potential.Since we have no indication for measurable H 2 evolution before the potential step, we can exclude that the sudden increase in H 2 evolution stems from reactions (1) or (1a).It is also rather unlikely because of the low concentration of water trace impurities.
A potential dependent H 2 evolution could result from the above reactions, however, if at least part of the evolving H 2 is electro-oxidized at sufficiently anodic potentials along the reaction in eq. ( 2): which could lower the measured H 2 signal.In that case, the strong increase could be due to a potential dependent, sudden decrease of the H 2 oxidation rate, for example, due to Pt oxidation.[46] Considering, however, the very low concentrations of the H 2 O and/or O 2 trace impurities in these electrolytes and the rather small amount of H 2 formation by borohydride bulk oxidation detected in O 2 -saturated and borohydride-containing electrolyte in the absence of a Pt electrode (see Figure 3a and related discussion), the H 2 signal expected from the borohydride bulk reaction in the present electrolyte should be well below the detection limit of these DEMS measurements.There-fore, the pronounced increase in the H 2 signal upon stepping to the positive potential limit cannot be due to the combination of a potential independent borohydride bulk oxidation reaction and a potential dependent H 2 consumption due to H 2 oxidation.
From the same reason, Faradaic currents arising from the oxidation of H 2 created via the bulk reaction of borohydride with trace impurities of O 2 (eq.( 1)) should be negligible.Note that the situation is different for higher concentration of O 2 or H 2 O as, for example, in O 2 -saturated electrolyte, which will be discussed later.Alternatively, H 2 formation could be possible via the (partial) electrocatalytic oxidation of borohydride according to eq. ( 3) which for complete electrooxidation (x = 8) would result in the reaction (eq.4) In the present case, where the potential step is expected to lead to Pt oxidation, we would expect mainly reaction (3), as reaction ( 4) is only possible on a reduced metallic Pt electrode (for details see Ref. [47]).50] Subsequently, upon lowering the potential, PtO reduction sets in and borohydride oxidation can occur via reaction (4), leading to a fast decay of the H 2 signal.The measured oxidation current seems to decay even faster, which may be due to an increasing contribution from the reduction of trace impurities of O 2 according to the reaction in eq. ( 5) (see the Faradaic current in the inset in Figure 1b).At about 0.8 V, after about 1000 s, the situation experienced at the OCP (zero net Faradaic current and no measurable H 2 evolution, see insets in Figure 1b) is reached again.At about 0.25 V the reduction current (Mg 2 + reduction and possibly electrolyte decomposition) starts, becoming more pronounced at about À 0.25 V.
Overall, the data indicate that for the present electrolyte H 2 formation is dominated by electrooxidation of borohydride rather than by electrolyte decomposition.Note that small contributions to the Faradaic current from Pt surface oxidation and reduction were not considered in this discussion.
With further deceasing potential, we find a small peak in the m/z = 2 signal, possibly due to H 2 evolution via water reduction according to eq. ( 6) (6)   parallel to the distinct reduction peak in the range between À 0.5 V and the negative potential limit.Also the m/z = 15 ion current shows a small but distinct increase with the increasing reduction current (Figure 1b), smaller than in Figure 1a, which could tentatively be assigned to the reductive decomposition of electrolyte.
In the second cycle, there are no significant features visible any more in the Faradaic current and mass spectrometric signals, pointing to an efficient passivation of the electrode in the first cycle.In particular Mg stripping and Mg deposition peaks are almost completely absent in the second cycle.Nevertheless, the passivation must be less efficient in electrolyte II than in electrolyte I, considering the even lower Faradaic current signal in Figure 1a.The absence of a H 2 evolution peak in the second cycle, in contrast, is no direct proof for an efficient passivation, as this may result also from a depletion of H 2 O trace impurities in the first cycle.We tentatively explain the slight improvement in reversibility/passivation by the slower formation of inhibiting Mg oxy-species in electrolyte II, where the water scavenger BH 4 À species results in an efficient lowering of the level of H 2 O and O 2 trace impurities.
Results of similar experiments in the electrolytes III (0.1 M Mg(TFSI) 2 + 0.1 M Mg(BH 4 ) 2 + 0.1 M 18-c-6, final OCP value: 0.4 V) and IV (0.1 M Mg(TFSI) 2 + 0.1 M Mg(BH 4 ) 2 + 1.0 M 18-c-6, final OCP value: 0.6 V), which contain a higher Mg 2 + concentration as well as both additives, are plotted in Figure 1c and  1d.Most important, the current densities show an improved reversibility of Mg deposition and stripping (Figure 1c and 1d, Figure S1c and S1d).Nevertheless, the oxidation charge is still significantly lower than the reduction charge.As for electrolyte II (Figure 1b), both electrolytes result in a high initial H 2 formation rate at the upper potential limit, which decreases rapidly with decreasing potential, until reaching close to negligible values at ca. 0.5 V.The intensities are, however, somewhat lower than those in electrolyte II.Closer inspection of the Faradaic current (see insets in the two figures) reveals that similar to electrolyte II also in this case the potential step results in a small anodic current (about 10 μA cm À 2 ), which decays within 200-400 s, i. e., again on a much longer time scale than expected for capacitive charging.Also in this case the decay of the H 2 signal is slower than that of the Faradaic current, and there is still considerable H 2 evolution detected once the Faradaic current is negligible.As discussed above (Figure 1b), we relate the H 2 evolution under these conditions to a combination of partial electrochemical borohydride oxidation by reaction with trace impurities of water (reaction (3)), and PtO reduction.In subsequent cycles, H 2 formation is not observed any more.While the somewhat lower H 2 signals compared to electrolyte II and the absence of these signals in the second and third cycle may result also from lower levels and depletion of the H 2 O trace impurities rather than from surface passivation, the much more pronounced Faradaic current signals in the second and third cycle clearly indicate an improved reversibility/lower passivation.This is most likely related to the higher concentration of Mg 2 + in electrolytes III and IV compared to electrolyte II.Most simply, due to the limited amount of H 2 O trace impurities a smaller fraction of the Mg 2 + is converted into Mg oxy-species, increasing the probability for reversible Mg deposition/stripping and thus the reversible Mg deposition/stripping charge as compared to electrolyte I and II.Furthermore, we find a clear variation for the m/z = 15, m/z = 28 and even m/z = 27 signals with electrode potential in the mixed electrolytes (Mg(TFSI) 2 and Mg(BH 4 ) 2 ) in Figures 1c and 1d, in contrast to the very small increase of the m/z = 15 signal in Figure 1b and the essentially featureless other ion currents (except for H 2 ) in Figures 1a and 1b.These signals are typical for fragments resulting from the electron impact ionization of light hydrocarbons such as methane or ethene. [51]hey are likely to originate from the decomposition of 18-c-6 upon the reduction of its complex with Mg 2 + ions, and are detected at higher Mg 2 + concentrations.Borohydride can be ruled out as source for the m/z = 15 signal, since evaporation as BH 4 À anion from the electrolyte is impossible.Mg-induced decomposition of TFSI À , as proposed in references, [13,15] appears to be less likely under these conditions, since compared to electrolytes I and II their concentration hardly changed and they are present in large excess.Also, 18-c-6 is expected to more strongly complex Mg 2 + ions than TFSI À , [15] reducing the tendency for Mg-induced decomposition of TFSI À in electrolytes III and IV.Considering the structure of 18-c-6 one may expect ethylene oxide formation upon splitting the ring to fragments.However, the fragmentation pattern of ethylene oxide, where the m/z = 44 intensity should nearly be equal to that of m/z = 15, [51] does not fit to the relative intensities of the m/z = 15, m/z = 27, and m/z = 28 ion currents and to the rather featureless m/z = 44 signals in these figures.As another possibility, one might expect the formation of ethene (the featureless m/z = 30 signal (not shown) excludes ethane formation).Ethene formation had indeed been reported by Hegemann et al. upon stripping of Mg electrodeposited from a Magnesium Aluminum Chloride Complex (MAAC)-tetraglyme electrolyte, [52] which is rather similar to the crown ether used here as additive, since the 18-c-6 can be considered as a cyclic glyme.Those authors also proposed that decomposition of tetraglyme to ethene should be induced by electrons released upon Mg stripping.This, however, does not fit with the present observation of ethene formation upon Mg electrodeposition only at potentials well below Mg stripping, leaving questions for the underlying mechanism still open.Furthermore, ethene formation should result in a dominant m/z = 28 signal, a relatively high m/z = 27 signal and a rather low m/z = 15 intensity, [51] which differs from the ratio of these intensities in the experimental data.Therefore, there must be additional evolution of methane to reach a higher m/z = 15 signal, which requires CÀ C bond splitting of the 18-c-6 ring, in addition to CÀ O bond breaking.Apparently, reduction of Mg 2 + in the complex with 18-c-6 leads also to some reductive decomposition of 18-c-6, for example, via ring opening.The higher amount of 18-c-6 decomposition in electrolyte III and IV as compared to the other ones seems to be correlated with the higher Mg 2 + concentration in the electrolytes, which also leads to more efficient Mg 2 + reduction.Interestingly, after setting in at the lower potential limit, the gas evolution at m/z = 15, 27, and 28 continues and is tailing over the entire positive-going scan, which means that these fragments are most likely formed via immobilized intermediate species, which then decompose slowly with time in a chemical reaction.
3,35] It fits well, however, with other previous reports, where based on laser infrared multiple photon dissociation (IRMPD) spectroscopy in combination with DFT calculations the authors concluded that in the most stable conformers, where the 18-c-6 ring nearly completely surrounds the Mg 2 + ion, the crown ether ring is already opened. [53]While TFSI À decomposition during Mg 2 + reduction was demonstrated already in recent quantum chemical calculations, [15,25] the present data furthermore show that also complexation with 18c-6 is not fully stable, but can result in a decomposition of the crown ether upon reduction, which had not been considered before.Nevertheless, the significantly enhanced reversibility of Mg deposition/Mg stripping in electrolytes III and IV clearly indicates that the decomposition of the 18-c-6 crown ether has much less effect on the passivation process than the TFSI À decomposition and deposition of Mg oxy-species in the electrolytes I or II, respectively.The latter processes must be slowed down in electrolytes III and IV.
Overall, these measurements have shown that for an efficient Mg deposition from different BMP-TFSI based electrolytes both 18-c-6 and borohydride are necessary, in agreement with the data reported in Ref. [15].The mass spectrometric data furthermore indicated that in the absence of borohydride (electrolyte I), the small reduction current mainly arises from side reactions such as the evolution of H 2 from the electroreduction of trace amounts of water (reaction ( 6)) and the decomposition of the electrolyte components.This leads to a rapid passivation of the electrode, which was not evident from our previous, purely electrochemical measurements. [15]In borohydride containing electrolytes (II-IV), where the reduction currents are significantly higher, this is explained by the removal of trace impurities such as water by reaction with borohydride, by slow chemical bulk reaction in the electrolyte before the electrochemical measurements and/or by electrooxidation upon cycling, where the latter is indicated by potential dependent H 2 evolution.This improves the reversible Mg deposition/stripping, as confirmed by the appearance of Mg stripping peaks in the positive-going scan in these electrolytes (II-IV).There are also distinct Mg 2 + concentration effects, since in electrolytes III and IV with their twofold higher Mg 2 + concentration (Mg(TFSI) 2 + Mg(BH 4 ) 2 ), the long-term reversibility is much better than in electrolyte II with its lower Mg 2 + concentration (only Mg(BH 4 ) 2 ).While in electrolyte II a stripping peak is only observed in the first cycle, but not in subsequent ones, these increase in the other two electrolytes during subsequent cycles.As another major new finding, we find a measurable degradation of the 18-c-6 additive during Mg deposition, as indicated by ethene formation, which should go along also with an enhanced formation and deposition (see next section) of other decomposition products.While this is only visible for electrolytes III and IV with their higher Mg 2 + concentrations, we expect this to occur also in the other electrolytes, but at a lower rate.Obviously, this improves the sustained reversibility of Mg deposition/dissolution upon cycling, though a detailed understanding of the underlying effects is still missing.Finally, the observation of an improved reversible Mg deposition/stripping upon addition of borohydride as additive agrees also with the previous report of an improved Mg (deÀ )intercalation in a Chevrel-type electrode from a DME-based electrolyte. [54]

SEM/EDS characterization of Mg electrodeposits on a Pt film electrode from BMP-TFSI based electrolytes
Further information about the Mg deposition process and the nature of the deposits was obtained from SEM images and EDS maps, which were recorded after potential cycling as described in Figure 1, plus sample storage in a glove box. Figure 2 depicts the surface morphologies (upper two rows) and the elemental composition (lower rows) of the surface regions of the electrodeposits obtained on the Pt film electrode in the different electrolytes, respectively.Images of the element-specific distribution of Mg, O, F and C in the area presented in the upper row are shown in the bottom two rows of Figure 2. Quantitatively, the EDS results are summarized in Table 2.
Some general trends for the Mg electrodeposition can already be derived from the SEM images.Hardly any visible deposit can be identified after deposition from electrolyte I (0.1 M Mg(TFSI) 2 + 0.1 M 18-c-6, Figure 2a) at the present scale.Except for a guiding artifact of a damaged Pt film substrate in the middle part of the SEM image in Figure 2a, the surface is smooth and without resolved structures, similar to the pristine Pt film.At ca. tenfold higher magnification (Figure 2b) one can resolve a few aggregates and small structures, in addition to cracks in the Pt film.The small structures can be attributed to Mg particles whose further growth was inhibited by the deposition of a passivating layer of electrolyte fragments and Mg (hydr)oxides that were formed already during the first deposition scan. [40,41]There is little evidence for Mg deposition in this electrolyte, and also the passivating layer formed during deposition is not really resolved.These conclusions are supported by EDS elemental maps recorded at the same location as the SEM images (Figure 2c1-2c4), which show predominantly Pt (30.6 at.%) and especially C species (52.0 at.%), less O and F (7.8 and 6.8 at.%), while other elements are below 2 at.%. Carbon, oxygen, fluorine and trace amounts of sulfur may originate from BMP-TFSI traces on the surface or from BMP-TFSI decomposition products.Carbon and oxygen can also reflect residues from 18-c-6 decomposition and surface contamination picked up during transport through air, while Mg (hydr)oxide formation must be less important.These species are homogeneously distributed over the surface, except for a few defect structures, which are also visible in the SEM images.
For electrolyte II (0.1 M Mg(BH 4 ) 2 + 0.1 M 18-c-6, Figure 2d-2f) we find a closed deposit layer on the Pt film substrate  (Figure 2d), consisting of fused, sub-micrometer size hemispherical particles, as resolved at ca. 20-fold higher resolution (Figure 2e).The morphology of the Mg deposit differs from previous findings for magnesium deposition on a Pt electrode from a Grignard reagent based electrolyte, [55] where compact, hexagonally shaped particles with a uniform size of approximately 2-3 μm were observed, indicating a different Mg growth behavior in the present electrolyte.The observation of significant Mg deposition agrees fully with the much lower intensity of the Pt substrate signal (3.5 at.%) and the ca.tenfold higher Mg content (Table 2) compared to deposition from electrolyte I. Furthermore, the higher amounts of F and in particular O point to an enhanced deposition of Mg oxy-species and increased presence of TFSI species/fragments, resulting either from remaining electrolyte traces on the surface or from increased deposition of electrolyte fragments.Contributions from 18-c-6 decomposition products such as (oligoÀ )polymeric ether species, which may be expected, cannot be high based on the much lower carbon concentration.Finally, the observation of B-containing species, most likely borates, can be explained either by their formation during reaction in the experiments or by reaction of electrolyte traces on the emersed samples with O 2 or H 2 O during storage/transport to the SEM.In total, these data fully agree with the conclusions derived from the DEMS measurements that Mg deposition from BMP-TFSI based electrolyte is possible in the presence of borohydride (and 18-c-6), while in its absence (electrolyte I) this is inhibited.
The electrodeposit formed upon deposition from electrolyte III (0.1 M Mg(TFSI) 2 + 0.1 M Mg(BH 4 ) 2 + 0.1 M 18-c-6, Figure 2g-2i) results in a film of approximately similar total thickness as in electrolyte II, as indicated by the comparable EDS signal from the Pt substrate (see below).In contrast to the rather regular hemispheric structures formed in electrolyte II (Figure 2e), the electrodeposit obtained in electrolyte III exhibits relatively large, irregular agglomerates (Figure 2h).Some of them have cauliflower-like morphologies, similar to those reported for the Mg electrodeposits from BMP-TFSI based electrolyte in the presence of O 2 . [11,14]The structures resolved here differ from the mossy, non-dendritic morphology of the electrodeposits reported previously for deposition from a mixed Mg(TFSI) 2 , 18-c-6 and MPPp-TFSI electrolyte (molar ratio 0.32 : 0.32 : 1.6 M), [27] where the mossy shape was attributed to the inclusion of organic fragments and impurities in the plated Mg. [27] Overall, the RTIL based electrolytes result in a different growth behavior compared to the more dendritic growth in Grignard based electrolytes. [56]EDS measurements (Figure 2i) show a comparable amount of Mg deposit as for deposition from electrolyte II, further supporting our previous conclusion that Mg deposition in electrolyte III is similarly efficient as in electrolyte II.The O content, however, is only half of that in electrolyte II, while the C content is more than doubled (Table 2).Since the amounts of F and S originating from the residual BMP-TFSI/TFSI fragments are comparable to those obtained in electrolyte II (Table 2), the higher amount of carbon in the present deposit points to an enhanced formation and deposition of 18-c-6 decomposition products.This would correlate with the higher mass spectrometric signals for the 18c-6 decomposition (alkane/alkene fragments) (Figure 1c).
Finally, the electrodeposit obtained from the electrolyte IV with its higher 18-c-6 concentration (0.1 M Mg(TFSI) 2 , 0.1 M Mg(BH 4 ) 2 , 1.0 M 18-c-6, Figure 2j-2 l) shows again hemispherical particles (Figure 2j), which in higher resolution images (Figure 2k) reveal a distinct microstructure with fibers running in parallel to each other.Mg electrodeposits with comparable morphologies have been reported also for deposition from a so-called "hybrid electrolyte", consisting of a mixture of BMP-TFSI and tetraglyme at a molar ratio 1 : 2, with both Mg(TFSI) 2 and Mg(BH 4 ) 2 salts, which showed an extraordinary reversible Mg deposition and stripping on glassy carbon electrodes. [56]ince an XRD analysis of those deposits showed a dominant Mg growth along the (100) orientation, which was also supported by a theoretical study, [57] one may speculate that similar effects happen also in the present case.
The elemental composition of this electrodeposit shows a comparable Pt intensity and a significantly higher Mg intensity compared to deposition from electrolytes II and III, supporting our previous conclusion that the higher concentration of 18-c-6 improves Mg deposition.In addition, also the oxygen intensity increases significantly, while that of carbon decreases considerably, and the other signals do not change much.The presence of F, S, N and B again points to residues from the electrolyte and BMP-TFSI decomposition, whereas the higher concentration of O stands for the enhanced formation of Mg oxy-species and possibly 18-c-6 decomposition products such as (oligoÀ )polyethers, which are formed upon Mg 2 + reduction.
Overall, the SEM and EDS data support our conclusions from the CV/DEMS data.Significant Mg deposition is possible only in the presence of both borohydride and 18-c-6, while in the absence of one of these species it is negligible.Furthermore, they are compatible with the formation of Mg oxyspecies during Mg deposition, as indicated by the simultaneous presence of Mg and O. Finally, in all of these electrolytes Mg deposition results in the formation of organic overlayers, confirming our previous conclusions in Ref. [15]).These contain either TFSI À decomposition products and/or traces of electrolyte, and presumably 18-c-6 decomposition products.Traces of borates are also present, which may be formed either during reaction in the experiment or by reaction of borohydride in remaining electrolyte traces with the atmosphere during storage/transport of the emersed samples.

O 2 interaction with a Pt film electrode under open-circuit conditions in Mg 2 + -containing, O 2 -saturated BMP-TFSI based electrolytes
To explore the impact of O 2 in the electrolyte on the nature of the Mg deposit and the passivation of the electrode, we performed DEMS measurements on the interaction of O 2 with the Pt film electrode in the three BMP-TFSI based electrolytes I-III (Table 1) under open circuit conditions.
Before presenting and discussing these results, we will briefly discuss results of a test on possible contributions from a chemical bulk reaction between borohydride and O 2 , employing the same DEMS cell as used before, but a Pt-free FEP membrane inlet.Here we monitored changes in different mass signals when starting O 2 bubbling in a borohydride containing electrolyte (BMP-TFSI + 0.1 M Mg(BH 4 ) 2 ).O 2 bubbling leads to a distinct increase of the m/z = 32 (and m/z = 16) signal, but also to an increase of the m/z = 2 signal (Figure 3a).The latter signal, which is indicative of H 2 formation, starts even prior to the increase of the m/z = 32 signal.We attribute this to a chemical bulk reaction between borohydride and O 2 , as given, for example, by the equation ( 7) The significant delay between starting the O 2 bubbling (= onset of the increase in the H 2 signal) and the onset of the O 2 signal indicates that during this time the gaseous O 2 in the vicinity of the membrane is completely consumed by the chemical bulk reaction between O 2 and borohydride.There are no measurable changes in the signals related to CO 2 formation (m/z = 44 and 22) and alkene formation (m/z = 15, 27, 28), and the water related ion current (m/z = 18) shows only ill-defined bubbling-induced variations.Obviously, the chemical reaction between borohydride and O 2 is rather efficient under these conditions, as indicated, for example, by the observed H 2 evolution.It should be noted that in the presence of trace impurities of water also the formation of borohydroxides is for example, via reaction (8) To the best of our knowledge, these bulk reactions, specifically reaction (7), have not been considered so far for organic solvents or ionic liquids.Besides mitigation of O 2 formation and production of H 2 , such chemical reactions will also result in the formation of borates.Furthermore, in the presence of Mg 2 + ions, O 2 bubbling can also result in the formation of insoluble Mg(OH) 2 in the solution, via the reaction ( 9) The formation of insoluble species such as Mg(OH) 2 or Mg(BO 2 ) 2 is indicated also by the appearance of white colloidal aggregates in the formerly transparent solution of Mg(BH 4 ) 2 after O 2 bubbling.
Next, we monitored the response of the potential and the evolution of gaseous products upon O 2 bubbling under open circuit conditions, using a similar Pt film electrode as in the experiments to Figure 1. Figure 3b-3d shows chronopotentiometric transients (upper panels) and the corresponding chronoamperometric ion current transients before and after starting bubbling the electrolytes with O 2 (lower panels), where the onset of O 2 bubbling is indicated by the sudden increase in the O 2 signal (m/z = 32).
Starting with electrolyte I (0.1 M Mg(TFSI) 2 + 0.1 M 18-c-6, Figure 3b), where the open circuit potential (OCP) value of the Pt film electrode before introducing O 2 is about 1.0 V, the OCP slowly increases by ca.0.5 V upon admission of O 2 (at ca.400 s, see the small spike in the potential).The increase of the OCP upon O 2 bubbling is attributed to an oxidation of the electrode surface.More detailed information will be gained from the ion current transients recorded simultaneously (Figure 3b-3d, lower panels).Upon O 2 admission, the m/z = 32 ion current increases rapidly.Different from the situation in Figure 3a, there is no delay between O 2 admission and increase in O 2 signal, as there is no pathway for efficient O 2 consumption in this electrolyte.The additional structure in the initial phase of this signal in Figure 3b is due to an experimental artifact, caused by saturation of the signal at ca. 450 s, while from ~500 s the signal was followed on a less sensitive scale.Within about five minutes after the onset of O 2 purging the signal reached a nearly constant value.Hence, at this point the electrolyte is saturated with O 2 .This behavior fits to the increase of the OCP value discussed above.A slight increase of the m/z = 18 signal is again explained by an enhanced convective transport of water traces to the electrode upon O 2 bubbling, whereas the small increase of the m/z = 44 and m/z = 28 ion currents is likely due to the oxidation of organic residues adsorbed on the electrode at the increasing OCP value.The m/z = 2 and m/z = 15 ion currents finally remained at their initial level.
In the borohydride containing electrolytes II and III (Figure 3c, 3d), the OCP is about 0.5 V lower before the onset of O 2 bubbling than in electrolyte I (Figure 3b).This lower value is most easily explained by the formation of a mixed potential, resulting from the simultaneous oxidation of borohydride to borate (reactions (3) and ( 4)) and the reduction of trace impurities of O 2 (reaction ( 5)) and/or water (reaction ( 6)) after filling the electrolyte into the cell.Furthermore, reduction of an oxide layer on the Pt electrode by the strongly reducing borohydride may also play a role.Upon O 2 admission, the OCP response differs significantly from that in electrolyte I.In electrolyte II, the potential decreases by ca.0.25 V (Figure 3c), while in electrolyte III, where both Mg(TFSI) 2 and Mg(BH 4 ) 2 are present, it increases by ca.0.25 V (except for a brief excursion to lower values directly during the switch).Furthermore, these changes occur significantly faster than in electrolyte I, in particular in electrolyte III.Reasons for these differences in the OCP behavior will be discussed together with the changes in mass spectrometric signals upon O 2 admission, which are presented in Figure 3c and 3d.
For electrolyte II (Figure 3c), there is only a slow continuous increase of the m/z = 32 ion current, despite of O 2 bubbling, which contrasts the rapid increase in electrolyte I (Figure 3b).Also in this case there is no delay between H 2 formation (O 2 admission) and the increase in O 2 signal, despite the presence of BH 4 À .Obviously, different from the situation in Figure 3a, O 2 was not completely consumed by the bulk oxidation of borohydride (reaction (7)) in the initial phase, directly after O 2 admission, and therefore we see a small but measurable immediate increase in the O 2 signal at this point.Next, the m/z = 18 signal decreases abruptly as opposed to the increase in Figure 3b, and there is a pronounced, instantaneous increase of the m/z = 2 signal (H 2 evolution), followed by a slow decrease at later time.This contrasts the featureless m/z = 2 current trace in Figure 3b.For all three signals we see an additional transient structure in the time between about 500 and 1500 s after the onset of O 2 bubbling, which we will get back to when discussing the behavior in electrolyte III (Figure 3d).Finally, the ion currents m/z = 44, m/z = 28 and m/z = 15 remain unchanged.
We postulate that the admission of O 2 increases the tendency for the ORR (reaction (5)) and the electrocatalytic oxidation of borohydride via reactions (3) and ( 4), which in the absence of O 2 purging occurred at low rates.Under OCP conditions, these reactions must result in a zero net current.Upon admission of O 2 , the drastic increase in O 2 concentration will also lead to a corresponding increase of the borohydride bulk oxidation rate according to reaction (7), which in turn results in a significant rate of H 2 formation.Simultaneous H 2 electro-oxidation (reaction (2)) does not seem to occur at these potentials, as in that case the H 2 signal should be lower than for the Pt-free membrane (Figure 3a), which is not the case.Therefore, we expect that the main partial reactions contributing to the mixed potential formation are the BOR and the ORR. [58]In combination, these data indicate that the compensation of electron-generation and consumption due to electrooxidation and electro-reduction reactions before and after the admission of O 2 results in a shift of the OCP (mixed potential formation).The correlated sudden decay in the water signal, which is in contrast to the essentially constant signal on the Ptfree membrane, indicates that the traces of water in borohydride containing solution are consumed at the OCP only in the presence of the Pt film.This points to the participation of interfacial reactions in H 2 O removal, for example, via reactions (3) and ( 4) in the present case.Finally, contributions from the ORR (reaction (3)) should be small at these OCP values (0.5! 0.25 V, see Figure 3c).Another plausible cathodic partial reaction instead of the ORR in non-aqueous electrolyte (Eq.( 5)), which involves both the consumption of O 2 and H 2 O trace impurities, is given by reaction (9).This can lead to the formation of OH À and thus, in the presence of Mg 2 + in the solution, to Mg(OH) 2 .
In electrolyte III (Figure 3d), the general trends are rather similar to those in electrolyte II, with the main differences that the OCP increases rather than decreases upon O 2 bubbling, that all changes upon O 2 bubbling are smaller and that for longer times the consumption of O 2 and H 2 O, and the formation of H 2 do not co-decrease, but approach a saturation value.The different trend in OCP variation must be related to the higher Mg 2 + concentration, since in both electrolytes we have borohydride present and the twofold higher Mg 2 + concentration in electrolyte III is the only major difference.Most simply, this can be explained via deposition of Mg oxy-species, which affects the activity of the Pt surface such that H 2 electrooxidation is more affected than O 2 reduction.In that case, admission of O 2 will lead to an increase in OCP, opposite to the behavior in electrolyte II.Here it should be noted that these shifts in OCP require only minute differences in the oxidation and reduction rates, respectively, as the resulting charges are accumulated with time.
Similar to electrolyte II, we also see same transient changes, here in the period between 700 and 1500 s after the onset of O 2 bubbling (see the dotted vertical lines in Figure 3d).In the first part of this period, up to the time indicated by the dashed vertical line, the O 2 signal is higher for some time, indicative of lower O 2 consumption, while the H 2 O and H 2 signals remain constant and the OCP increases slightly (see the 10-fold magnified trace in this panel).Subsequently, the O 2 signal decreases again to the original level, and H 2 O and H 2 signals irreversibly decrease or increase, respectively.While the initial decrease in O 2 consumption is likely due to the ORR, using electrons provided by the incomplete oxidation of borohydride (reaction (3)), the increasing H 2 formation and H 2 O consumption starting at the dotted vertical line (at about 1200 s) point to an irreversible increase of borohydride (electroÀ )oxidation (reactions (3), ( 4) and possibly reaction ( 7)).For current neutrality (under OCP conditions), the above partial oxidation reactions are compensated by an increased O 2 reduction (ORR) via reactions ( 5) and ( 9).These variations are reflected by a small decay in the OCP (see inset in Figure 3d).The physical origin of these changes is not clear so far, but it should be of similar nature as that for the transient changes in Figure 3c in a similar time frame, supporting that these changes reflect reproducible changes in the surface chemistry, rather than artifacts that are induced, for example, by O 2 bubbling.
Overall, these findings, in particular the differences between the OCP and mass spectrometric responses to O 2 bubbling in the absence and presence of borohydride in the electrolyte, indicate that in borohydride-free electrolyte I, O 2 exposure under OCP conditions leads to Pt surface oxidation, which in turn results in an increasing OCP.In the borohydride containing O 2 -saturated electrolytes II and III, O 2 admission results in a complex network of reactions.These include the bulk chemical oxidation of borohydride to borates, by reaction with O 2 , as well as a number of electrocatalytic reactions, such as electrooxidation borohydride and of the H 2 produced that reaction and the electro-reduction of O 2 .Under OCP conditions, these reactions lead to the formation of a mixed potential.This situation is similar to that experienced in 'catalytic' borohydride oxidation in aqueous electrolytes, which is also characterized by simultaneous oxidation of borohydride, reduction of water to OH À and H 2 formation. [47]As a practical consequence, one has to consider that the use of borohydride as additive will result in undesired H 2 evolution and a gradual mitigation of borohydride at the open circuit potential in realistic Mg batteries.

O 2 reduction and evolution on a Pt film electrode from Mg 2 +containing, O 2 -saturated BMP-TFSI based electrolytes
After saturation of the Mg 2 + containing BMP-TFSI based electrolytes with O 2 at the OCP (see Figure 3), we performed similar cyclic voltammetry measurements as described in Figure 1 under continuing O 2 bubbling.Again, we followed the Faradaic current density as well as different mass spectrometric signals simultaneously during potential cycling in the O 2 -saturated electrolytes I-III.Experimentally, the procedure was similar to that described before, with the electrode potential stepped from the OCP to the upper potential limit.Also in these measurements, effects from O 2 reduction/evolution, Mg deposition/stripping and the formation of insoluble deposits due to electrolyte decomposition, Mg (hydr)oxide formation or borate formation and finally H 2 oxidation have to be considered.Again, small contributions to the Faradaic current from Pt surface oxidation/reduction are not considered in the discussion.For better comparison with literature data we again plotted the Faradaic current densities as current-voltage plots in the Supporting Information in Figure S2, they are discussed there in more detail.In summary, they show that reasonably stable reversible cycling performance can be obtained for electrolytes III and IV, at least on the scale of the present experiments, but not for electrolytes I and II.Furthermore, the general trends in the CVs largely resemble those reported previously for a glassy carbon electrode in similar electrolytes, but using dimethyl aminoborane as a water scavenger. [14]This underlines the importance of a reducing additive in combination with a sufficiently high Mg 2 + concentration for preventing electrode passivation by oxy-species that are stable against re-oxidation, a precondition for sustained reversibility in the oxygensaturated electrolytes.
In Figure 4 we show plots of the Faradaic current densities and of different ion currents as a function of time during the first three cycles.A presentation of the initial phase on an expanded time scale is given in Figure S3.These latter plots demonstrate that also on that time scale there are no significant changes in the signals directly after stepping to the upper potential limit, independent of the electrolyte (for details see Figure S3).Hence, for all electrolytes the potential before the potential step the OCP was close to that reached upon the potential step.
In O 2 -saturated electrolyte I (0.1 M Mg(TFSI) 2 + 0.1 M 18-c-6, Figure 4a), we find an initially slow increase of the Faradaic reduction current at > 0.1 V, which is accompanied by a small increase of the m/z = 32 current signal.At present we can only speculate that this subtle increase in the O 2 signal is related to the reductive removal of some deposits that were formed under OCP conditions and which affect the permeability of the membrane.The first reduction peak in the negative-going scan with its onset at ca. 0.0 V (see also Figure S2a) goes along with increasing O 2 consumption, as evidenced by the steep decay of the m/z = 32 signal.This reaches its minimum at ca. 10000 s (~À 0.6 V), and remains about constant until reaching ca.0.0 V in the backward scan.The sharp peak in the m/z = 2 ion current, which coincides with the equally sharp reduction current peak, indicates that both result from a reduction reaction associated with H 2 formation such as reduction of H 2 O trace impurities.We cannot exclude, however, contributions from a change in the ORR selectivity, from a 1-electron process to a 2-electron process, which would perfectly agree with our previous data. [12,14,59]Interestingly, a comparable H 2 evolution was not observed in the O 2 -free electrolyte in this potential region (Figure 1a), indicating that the presence of O 2 activates the reductive hydrogen evolution from traces of water and possibly also from electrolyte components.This interpretation is supported also by the fact that both the Faradaic current peak and the H 2 signal were very small in later cycles, after depletion of trace impurities or passivation of the surface against electrolyte decomposition.Finally, there is a reproducible increase of the m/z = 44 signal when approaching the upper potential limit of 1.5 V (Figure 4a), which parallels the small increase of the oxidation current at potentials positive of ca.1.25 V (Figure 4a).These features are most likely due to the oxidation of adsorbed organic residues from electrolyte components.Due to the low intensities of these signals we cannot decide upon the origin of these species, since we do not have any indication of BMP-TFSI decomposition or 18-c-6 decomposition under these conditions from the other DEMS signals.
Similar traces recorded in analogous DEMS measurements in electrolyte II (0.1 M Mg(BH 4 ) 2 + 0.1 M 18-c-6) are plotted in Figure 4b (see also Figure S2b).Different from the measurements in the O 2 -saturated electrolyte I (Figure 4a), the m/z = 32 current is initially close to zero, although the electrolyte is O 2saturated, indicative of an efficient reactive O 2 removal at high potentials.Assuming that the O 2 signal at the upper potential limit reflects essentially the level of bulk oxidation of borohydride by reaction with O 2 (eq.( 7)), the subsequent decay of this signal (at about 5000 s) can be associated with the increasing consumption of borohydride (reaction (7)) and, at potentials below 0 V, the onset of the ORR (reaction ( 5)).Furthermore, we find efficient H 2 evolution under these conditions.The H 2 signal decreases about exponentially with time and is essentially insensitive to the potential changes, which points to a potential independent bulk reaction such as that described by reaction (7).This will be discussed in more detail below.(Note also that this decay is much slower than the H 2 decay in Figure 1b and  1c.)With further decreasing potential and its increase in the backward-going scan, the O 2 signal is close to the zero level, indicating that bulk chemical O 2 consumption and ORR together result in essentially complete consumption of O 2 .The ongoing ORR will contribute part of the cathodic current to the Faradaic current signal, while the remaining current fraction must come from electrolyte decomposition and/or H 2 evolution (see below).At the end of the reduction peak in the Faradaic current, the O 2 mass signal increases again, reflecting a decreasing O 2 consumption.Most simply, this can be explained by a complete decay of the electrochemical ORR rate, such that only O 2 removal by the bulk reaction ( 7) is left.The small positive peak in the Faradaic current and in the O 2 mass signal at about 0.1 V, which is better visible in the second cycle, is likely to reflect electrocatalytic O 2 evolution.Apparently, a small part of the oxy-species formed upon the oxygen reduction and immobilized at the electrode surface can be re-oxidized to O 2 in the positive-going scan.In the subsequent scans, this general scheme repeats, with the only difference that from scan to scan the O 2 signal in the positive potential range becomes higher, as if O 2 removal via the bulk chemical reaction with borohydride becomes less efficient with time.This explanation is also supported by the negligible Faradaic current in these potential ranges.The reactions proposed and discussed above also imply that the nature of the deposits formed in electrolyte II is different from that in electrolyte I, being mostly Mg oxy-species in the present case and mostly electrolyte decomposition products in electrolyte I.This also seems to agree well with our EDS observations in O 2 -free electrolyte.
The (m/z = 2) signal, which is characterized by a slow, about exponential decrease, reaches zero after about 40 000 s (~11 h).This decay is overlaid by some broad, weak features around the negative potential limit (Figure 4b).Most important, the distinct peak observed in electrolyte I at the first negative potential limit is essentially absent here.Also, there is no evidence for H 2 electro-oxidation at more positive potentials, as would be expected for a Pt electrode.In that case, the H 2 signal should be much lower in the potential regime where this reaction is active.The weak features in the H 2 signal we tentatively attribute to either reductive electrolyte decomposition or to water reduction.The fact that they are much smaller than in electrolyte I supports our interpretation that in electrolyte I the deposit consists mainly of electrolyte decomposition products, while in electrolyte II these are largely Mg oxy-species.The slow decrease of the H 2 signal, which is definitely related to the simultaneous presence of borohydride and O 2 , is likely due to a gradual consumption of borohydride by bulk reaction with O 2 (Eq.( 7)).Here it is interesting to note that H 2 electrooxidation is apparently inhibited at more positive potentials, while O 2 reduction via the ORR is active at potentials below 0.0 V. Most simply, this can be explained assuming that the Pt surface was covered by Mg oxy-species already during the OCP phase in O 2saturated electrolyte and that this cover layer is present also during potential cycling, at least at more positive potentials, and blocks the H 2 oxidation reaction.For the ORR this means that this reaction can proceed also on a surface covered by Mg oxy-species.
The other signals, including the m/z = 44 and the m/z = 15 ion currents (as well as the m/z = 27 and m/z = 28 signals, not shown) are essentially featureless in this electrolyte.We expect that the absence of CO 2 formation is mainly due to the lower value of the upper potential limit in this electrolyte (1.0 V rather than the 1.5 V in electrolyte I), which impedes oxidation of electrolyte decomposition products.Furthermore, the constant value of these signals is different from our observations in the absence of O 2 , where these signals showed a distinct potential dependent structure, indicating that the crown ether is stable in the presence of O 2 and thus during O 2 reduction, but not in the absence of O 2 .Most simply, this can be explained by the formation of insoluble Mg (per)oxy-species that remove Mg 2 + from the complexes with 18-c-6, and thus lower 18-c-6 decomposition upon Mg 2 + reduction.
Similar time-resolved profiles measured in the O 2 -saturated electrolyte III (0.1 M Mg(TFSI) 2 + 0.1 M Mg(BH 4 ) 2 + 0.1 M 18-c-6) are plotted in Figure 4c (see also Figure S2c).Interestingly, the Faradaic current density is about four times larger than those obtained in electrolytes I and II (Figure 4a, 4b), which we relate to the higher Mg concentration.The general features are, however, rather similar.For the m/z = 2 and m/z = 18 signals, respectively, we find a rapid increase/decrease during the initial stage of the present experiment (at about 6000 s), which is caused by a re-adjustment of the O 2 bubbling rate.The general shape of the respective signal traces largely resembles those in Figure 4b.Therefore, they shall not be discussed again in detail.Main differences compared to electrolyte II are the more pronounced oxidative current peaks and peaks in the m/z = 32 signal in the positive-going scans at about 0.15 V. (Note that also the small Faradaic current reduction peak at 5000 s (À 0.1 V) in the first negative-going scan, which was not seen in Figure 4b, is attributed to the re-adjustment of the O 2 bubbling rate.)Hence, similar as in electrolyte II the potential controlled electrochemical reactions are accompanied by the bulk chemical reaction between O 2 and borohydride, which can continue until the latter is essentially depleted in the electrolyte.The more pronounced peaks at about 0.15 V point to a more efficient accumulation of oxidizable Mg oxy-species close to the electrode at more cathodic potentials, which is indicated also by the significantly higher ORR currents at more cathodic potentials (note the different scales in Figure 4b and 4c).This fully agrees with our conclusion of a more reversible ORR/OER and hence less pronounced passivation in this electrolyte as compared to electrolyte II (see the discussion of Figure S2c).It also agrees with previous findings by Eckardt et al. for the ORR/ OER on a glassy carbon electrode in the same electrolyte, though with a different additive. [14]In that work the ORR products in Mg 2 + -containing BMP-TFSI were concluded to be MgO and MgO 2 , where only the latter ones can be re-oxidized to form O 2 . [14]][61] Therefore, we assume that these species are formed also in the present case, where the latter can be reoxidized to O 2 at 0.15 V.
To address the effects of O 2 bubbling and borohydride depletion upon O 2 bubbling in more detail, we performed a comparative experiment in electrolyte III where we used a longer time resting at the OCP with active O 2 bubbling (Figure S4a), followed by the ORR measurement (Figure S4b).As expected, we again find continuous H 2 formation during O 2 bubbling at the OCP, which is accompanied by only a slight increase in O 2 signal at the onset of O 2 bubbling (Figure S4a).Hence, there is a continuous consumption of O 2 during this time by bulk reaction with borohydride.During the subsequent potentiodynamic run, however, the H 2 formation rate decays much more rapid than in Figure 4c, concomitant with an increase of the O 2 signal.This fully agrees with expectations for a situation where most of the borohydride was already consumed during the waiting time at the OCP, and therefore cannot contribute to the overall reaction via reaction (7) any more.Except for the H 2 signal, the other signals show essentially similar trends during potential cycling as observed in Figure 4c.Comparable results were obtained also for test measurements performed with different O 2 bubbling rates, supporting the above conclusions.
Overall, the DEMS measurements in the different O 2saturated Mg-containing electrolytes revealed that the reduction and oxidation currents contributing to the Faradaic current signal are a result of a complex combination of different reduction and oxidation processes that are going on at a given potential and that can (partly) compensate each other in the Faradaic current signal.The increasing reduction current at low electrode potentials, at about constant O 2 consumption, indicates a change in the ORR selectivity, from a 1-electron process to a 2-electron process.This would perfectly agree with our previous findings obtained in the BMP-TFSI based electrolyte in the absence of any additive, [12,59] or in the presence of the NBH additive, [14] which were all performed under welldefined mass transport condition in a thin-layer flow cell.The m/z = 32 OER signals confirmed our previous observation that passivation of the Pt electrode is significantly lowered or even largely removed in the combined presence of both borohydride and 18-c-6 additives.This effect of a lower passivation is even more pronounced for the higher relative Mg 2 + concentrations in electrolyte III, as indicated by the higher OER currents/O 2 signals compared to electrolyte II.For reaction in O 2 -saturated borohydride containing solution we find reactions involving O 2 , borohydride, 18-c-6, and water trace impurities as reactants.Furthermore, some of these reactions occur both as chemical bulk reaction and as electrochemical reaction, depending on the potential.The different contributions can be identified by comparison of Faradaic current, O 2 consumption/generation, water consumption and H 2 formation.Our conclusion of a considerable bulk reaction between borohydride and O 2 , which results in the continuous formation of H 2 , was confirmed by DEMS measurements performed in borohydride containing electrolyte under purely chemical conditions, in the absence of a Pt electrocatalyst, which also show considerable H 2 evolution.Depending on the potential, H 2 can also be generated by electrocatalytic oxidation of borohydride on the (Mg-free) Pt electrode, and possibly consumed by electrocatalytic H 2 oxidation.These reactions can proceed without a measurable Faradaic current when coupled to a simultaneous reduction reaction on the same electrode ('mixed potential' formation).Finally, in Mg 2 + containing electrolyte, the ORR via reaction (9) can result in OH À , which leads to the generation of insoluble Mg(OH) 2 products that appear as flakes in the borohydride containing electrolyte.

Discussion
Combining the results presented and discussed in the previous sections and previously reported data we arrive at the following main insights: 1.The initial efficiency for Mg deposition (in the first scan) in the O 2 -free electrolytes is significantly enhanced in the presence of both additives, BH 4 À (Mg(BH 4 ) 2 ) and the crown ether 18-c-6, as compared to electrolyte I with only 18-c-6 as additive (see Figure S1 and Table 2).The same is true also for the reversibility in the first cycle.The fact that both additives are required for improved deposition/dissolution indicates a synergistic behavior with different roles of both additives, where borohydride acts as water scavenger and 18-c-6 as complexing agent, hindering the reductive decomposition of TFSI À in Mg-TFSI complexes.Sustained reversibility (also in subsequent cycles), however, is achieved only in electrolytes III and IV, which compared to electrolyte II contain double amount of Mg 2 + , as well as identical amounts of BH 4 À and 18-c-6 (in electrolyte III) or an excess of 18-c-6 (in electrolyte IV).Apparently, the Mg 2 + concentration is important as well, which may reflect a decreasing sensitivity towards trace impurities with increasing Mg 2 + concentration.Compared to previous findings for Mg deposition from the same electrolytes, but 50 times faster scan rate, the deactivation of reversible Mg deposition/ stripping (per cycle) is much faster, reflecting the more efficient decomposition of the electrolyte due to the much longer interaction times. [15]. In electrolytes III and IV with their higher Mg 2 + concentration DEMS results reveal an enhanced reductive decomposition of 18-c-6, as indicated by the potential dependent formation of ethene and methane (m/z = 15, 27, 28 fragments).This is proposed to happen upon reduction of its complex with Mg ) and of a complexing agent (18-c-6), but depend sensitively also on the relative concentrations of Mg 2 + and the two additives.Furthermore, while in the absence of O 2 the interaction between Mg 2 + and 18c-6 will cause slow decomposition of the Mg 2 + -[18-c-6] additive, most likely by destructive reduction, this is not observed in the presence of O 2 .We suggest that in this case the non-destructive removal of Mg 2 + is facilitated via the reversible formation and deposition of MgO 2 .
Overall, these DEMS measurements provided detailed insights into the complex reaction network active during potential cycling in O 2 -free/O 2 -saturated, Mg-and borohydride-containing BMP-TFSI electrolyte, extending our previous knowledge from CV measurements on the O 2 -free electrolytes [15] and from DEMS measurements using NBH as water scavenger. [14]Particularly relevant for technical application in Mg-air batteries are the results that there is a highly efficient electrochemical pathway for reaction of the water scavenger BH 4 À with water, which depending on the potential and state of the electrode leads to H 2 evolution or H + formation and can proceed also under OCP conditions.In O 2 -saturated borohydride-containing electrolyte the bulk chemical reaction between borohydride and O 2 is strongly enhanced and results in the generation of insoluble products, which appear as flakes in the electrolyte.This essentially excludes the use of BH 4 À as water scavenger in these electrolytes in Mg-air batteries.

Summary
As part of an extensive series of model studies on mechanistic aspects of the reactions in magnesium-air batteries we have performed systematic DEMS measurements on i) the deposition/stripping of Mg on a Pt film electrode in four different BMP-TFSI based electrolytes, containing Mg(TFSI) 2 and/or Mg-(BH 4 ) 2 as Mg source, BH 4 À as water scavenger, and the crown ether 18-c-6 as complexing additive, and on ii) the reduction/ evolution of O 2 in the same Mg-containing electrolytes.From the potential dependent appearance of different gaseous reaction products and electrolyte decomposition products and their correlation with the Faradaic current, we could identify a number of different reactions that occur in the above processes during potential cycling.These include deposition/stripping of Mg in O 2 -free electrolyte and of Mg oxy-species in O 2 -saturated electrolyte (ORR/OER), reductive (Mg-assisted) 18-c-6 decomposition, electrochemical borohydride oxidation by reaction with water traces, which at certain potentials is much more active than the well-known bulk chemical reaction, and the bulk chemical reaction between borohydride and O 2 or H 2 O. Reaction with O 2 occurs at considerable rates in O 2 -saturated electrolytes, which precludes the use of this water scavenger in Mg-air batteries.Overall, this work underlines the potential of DEMS measurements for a detailed understanding also of complex battery chemistries.

Experimental Section
The online DEMS experiments were performed in a beaker-type small-volume (about 0.5 cm 3 ) DEMS cell, using a Pt film electrode sputtered onto a gas permeable membrane, which on the backside was directly interfaced to the vacuum chamber with the quadrupole mass spectrometer.A detailed description of the employed DEMS instrument was provided in that Ref. [62].In short, the cell consists of a polyether ether ketone (PEEK) cylinder and a PEEK Ucup, which connects to the analysis chamber with the mass spectrometer (Pfeiffer Vacuum QMA 410) via a fluorinated ethylene propylene (FEP) membrane (Bola, purchased from Bolender, thickness 50 μm, exposed area 0.25 cm 2 ).The cell was located in an Arfilled glove box (MBraun LabMaster Pro, O 2 < 0.1 ppm; H 2 O < 0.5 ppm) equipped with an O 2 supply line (Alphagaz, 99.9995 %), and connected to the external analysis chamber via a stainless steel bellow.For dismounting the cell, the connection to the analysis chamber could be closed with an open/close valve in the glove box.The Pt film working electrode (thickness ca.70 nm) was sputtered onto the FEP membrane, using an Ar plasma sputter coater (Leica EM ACE600).Working electrodes of 13 mm diameter were punched from the Pt-sputtered membrane and stored in another Ar-filled glove box (MBraun LabStar, O 2 < 0.5 ppm; H 2 O < 1 ppm).Before use they were dried in the glove box on a hot plate at 100 °C for ca.30 min.Two fresh cuts from a Mg foil (99.9 %, Goodfellow, 0.25 mm thick), which were scratch-cleaned in a glove box environment, were used as counter and quasi-reference (À 1.0 V vs. Fc/Fc + ) electrodes, respectively.In contrast to a Ag/AgCl reference, the Mg/MgO reference was found to be stable in these electrolytes also in the presence of O 2 . [12,14]All potentials in this paper are given relative to that of the Mg/MgO reference.The DEMS cell was stored, assembled and operated in the glove box.For the electrochemical measurements we used a computercontrolled potentiostat (PAR 263 A), and the selected ion currents were acquired simultaneously.The mass spectrometric signals presented below are background corrected, such that the lowest value was set to zero.Scanning Electron Microscopy (SEM) and Energy Dispersive Spectroscopy (EDS) characterization of the deposits was performed in a Zeiss Crossbeam 340 field-emission electron microscope, after extensive rinsing with acetone followed by drying under vacuum.
The electrolytes were prepared in the glove box, by dissolving the appropriate amounts of Mg(BH 4 ) 2 (Sigma Aldrich, 95.0 %), Mg(TFSI) 2 (Solvionic, 99.5 %, < 250 ppm H 2 O) and 18-crown-6 ether (Alfa Aesar, 99.0 %, < 0.29 % H 2 O) in 1-butyl-1-methylpyrrolidinium bis(trifluoromethylsulfonyl) imide (BMP-TFSI, Solvionic, 99.9 %, < 20 ppm H 2 O) under stirring (composition see Table 1).Karl Fischer titration of the resulting electrolytes yielded water contents in the range of ca. 30 ppm. [15] 0.5 mL of the respective electrolyte was filled into the DEMS cell, counter and reference electrode were introduced from the top.Together with the working electrode they were connected to the potentiostat to rest at the open circuit potential (OCP) until stable values of the OCP were obtained and the background signals of the corresponding ion currents had stabilized.For O 2 saturation the electrolytes were purged with O 2 by bubbling O 2 via a capillary immersed from the top, while the O 2sensor of the MBraun LabMaster Pro glove box was deactivated.After each experiment with O 2 purging the glove box was flushed for 30 min to 1 h with Ar to reduce the O 2 content to < 0.1 ppm.

Table 1 .
Composition of the different electrolytes used in this study.

Table 2 .
Elemental surface composition of the electrodeposits formed on the Pt film electrode in the corresponding BMP-TFSI electrolytes I-IV, based on the EDS measurements in Figure2.Data are averaged over the upper row images, relative deviations due to local variations are estimated to be � 3 %.
c-6 are present, reflecting a reversible deposition/ oxidation of Mg oxy-species.Furthermore, we could demonstrate that the reaction in borohydride containing solution results in a complex combination of different partial reactions between O 2 , borohydride, 18-c-6 and water trace impurities, and also H 2 electro-oxidation, which occur depending on the potential and can be identified by comparison of Faradaic current, O 2 consumption/generation, water consumption/formation and H 2 formation.In addition, in O 2 -saturated borohydride containing solutions also bulk chemical oxidation plays an important role, for example, by the formation of H 2. This is true at least in the initial phase of the reaction, before the depletion of borohydride in the electrolyte.Finally, different from O 2 -free electrolytes, decomposition of18-c-6 does not seem to play an important role in O 2 -saturated electrolyte, as concluded from the absence of potential dependent decomposition signals such as m/z = 15.7. The data indicate that the reversible Mg deposition/stripping and the OER/ORR in neat and O 2 -saturated BMP-TFSI based electrolytes, respectively, require not only the presence of a reducing water scavenger (BH 4 À 2 + ions.Possible deposits formed during this process such as (oligoÀ )polyethers on the electrode surface, however, do not (fully) inhibit reversible Mg deposition/stripping. Instead, the sustainable reversibility is enhanced.2 -saturated, Mg-containing electrolyte are at least partly due to the ORR and OER, as identified by the consumption and release of O 2 in online DEMS measurements in these electrolytes.This underlines the reversibility of the ORR/OER by deposition and re-oxidation of Mg peroxo-species, as the formation of MgO would inhibit this.Passivation of the Pt electrode is lower if both borohydride and 18-