Sharp kinetic acceleration potentials during mediated redox catalysis of insulators


 Redox mediators could catalyse otherwise slow and energy-inefficient cycling of Li-S and Li-O2 batteries by shuttling electrons/holes between the electrode and the solid insulating storage materials. For mediators to work efficiently they need to oxidize the solid with fast kinetics yet the lowest possible overpotential. Here, we found that when the redox potentials of mediators are tuned via, e.g., Li+ concentration in the electrolyte, they exhibit distinct threshold potentials, where the kinetics accelerate several-fold within a range as small as 10 mV. This phenomenon is independent of types of mediators and electrolyte. The acceleration originates from the overpotentials required to activate fast Li+/e– extraction and the following chemical step at specific abundant surface facets. Efficient redox catalysis at insulating solids requires therefore carefully considering the surface conditions of the storage materials and electrolyte-dependent redox potentials, which may be tuned by salt concentrations or solvents.


Redox mediators could catalyse otherwise slow and energy-inefficient cycling of Li-S and Li-O2
14 batteries by shuttling electrons/holes between the electrode and the solid insulating storage 15 materials. For mediators to work efficiently they need to oxidize the solid with fast kinetics yet 16 the lowest possible overpotential. Here, we found that when the redox potentials of mediators 17 are tuned via, e.g., Li + concentration in the electrolyte, they exhibit distinct threshold potentials, 18 where the kinetics accelerate several-fold within a range as small as 10 mV. This phenomenon is 19 independent of types of mediators and electrolyte. The acceleration originates from the 20 overpotentials required to activate fast Li + /eextraction and the following chemical step at 21 specific abundant surface facets. Efficient redox catalysis at insulating solids requires therefore 22 carefully considering the surface conditions of the storage materials and electrolyte-dependent 23 redox potentials, which may be tuned by salt concentrations or solvents. 24 25 Electrochemistry with insulators is salient feature and central difficulty of topical future 26 battery chemistries such as Li-air (O2), Li-CO2, Li-Sulphur (Li-S) cells 1-10 . They differ in this respect 27 from current intercalation-type batteries, which rely on ion (de)insertion to balance charge upon 28 redox of the mixed-conducting solid host 1 . The interest in Li-O2, -CO2, and -S cells arises from high 29 theoretical energies, abundant elements, low cost and environmental friendliness. Li-O2/CO2 cells 30 interconvert O2 dissolved in the electrolyte into solid, insulating Li2O2 or Li2CO3 during 31 discharge/charge. Li-S batteries interconvert solid, insulating S8 and Li2S. Kinetic bottleneck during 32 these processes is charge transfer between electrode and the insulating, insoluble, solid storage 33 materials, causing high overpotentials and incomplete conversion even at low rates. 34 Redox catalysis using mediators can bypass those insulators, transporting charge through the 35 electrolyte phase where ion and electron/hole transport may be facile and may boost charge 36 transfer kinetics 3-5,11-14 . Equally important is to approach the cycling potential as close as possible 37 towards the formal potential of the storage material to maximize energy efficiency and to suppress 38 parasitic reactions 4,5,15-21 . Soluble redox mediators (RMs) are, therefore, now accepted to be key to 39 achieve these goals and have been studied in a wide variety for Li-O2 cells 3-5,11-13,18,21-24 . First 40 examples have been reported for S electrochemistry 3,25-27 . Redox mediation on, for example, 41 charging involves oxidizing the mediator RM red at the electrode surface to its oxidized form RM ox , 42 its diffusion to the surface of Li2O2 or Li2S, where RM ox extracts charge and reforms RM red . Main 43 requirements for successful redox catalysis include a suitable equilibrium potential of the redox 44 couple to drive the reaction and fast heterogeneous reaction rates between RM and both electrode 45 ( 0 ) and storage material. 0 is sufficiently fast 28 and well described by established theories of 46 electron transfer between redox molecule and metallic conductor 29 . However, for the rate limiting 47 electron transfer between RM ox and a redox active insulating solid, despite being essential, detailed 48 descriptors are missing. 49 Activating this most difficult electron transfer step is the primary goal of redox catalysis on 50 charging Li-S and Li-O2 batteries, which have important parallels in their charging reactions. The 51 insulating Li2S and Li2O2 undergo in a first step a one-electron oxidation to form Li polysulfides 52 (LiPSs) or Li superoxide (LiO2) intermediates. Further oxidation and/or disproportionation 53 eventually yields the most oxidized forms S8 and O2, respectively 10,20,30-32 . Reaction kinetics for RM ox 54 and Li2O2 were reported for a range of mediators, typically assuming faster kinetics with higher 55 mediator potential (driving force) 28

Thresholds in the potential-dependent kinetics of RMs oxidizing Li2S and Li2O2 69
Decamethyl ferrocene (DFc) and lithium iodide (LiI) are commonly used RMs for the charging 70 process in Li-S batteries and Li-O2 batteries, respectively, and thus they are chosen as model RMs 71 in this work 14 . Their redox potentials, DFc/DFc + and I − /I 3 − , measured on the AgCl/Ag scale are 72 nearly independent of Li + concentration because of the species' large radii and weak solvation, 73 while Li/Li + does vary with Li + concentration following Nernst equation. Hence, DFc/DFc + and 74 I − /I 3 − vs Li/Li + vary with Li + concentration as shown in Supplementary Fig. 1. 75 Figure 1 shows the potential-dependent apparent rate constants app of DFc + and I3 -76 oxidizing Li2S and Li2O2, respectively. The rate constant for DFc + oxidizing Li2S ( DFc−Li 2 S app ) was 77 measured by following the DFc + concentration of a solution in contact with Li2S using UV-Vis 78 spectroscopy (see Methods and Supplementary Figure S2). The rate constant of I3 -oxidizing Li2O2 79 ( I 3 − −Li 2 O 2 app ) was measured using both scanning electrochemical microscopy (SECM) and differential 80 electrochemical mass spectrometry (DEMS) as detailed in Supplementary Note 1. Given the 81 complex mechanism with initial oxidation of Li2S or Li2O followed by further oxidation of the 82 intermediates or their disproportionation, apparent rate constants embrace all etransfer steps. In 83 either case, the rates followed first-order behaviour in RM ox concentration. They increase with 84 increasing mediator equilibrium potential. Surprisingly, however, is that in both cases kinetics 85 increased sharply by a factor of ~3 to 4.4 over a certain narrow range of equilibrium potentials, 86 whereas changes were gradual below and above these potentials. They represents a threshold, 87 where rather slow kinetics at lower potentials switches to much higher levels.   potential of the Li2S/Li2S2 redox couple, the relevant reaction for the first electron transfer step. As 100 a multi-step reaction, the reaction mechanism of Li2S oxidation is complicated and forms as a first 101 step partly soluble Li2S2 species as intermediate, which then over a series of 102 oxidation/disproportionation steps eventually forms S8. Therefore, the apparent kinetics could be 103 dominated by the oxidation of either solid Li2S or soluble polysulfides. To identify the rate-104 determining step, DFc + solutions in DME were separately added to two cuvettes with solid Li2S and 105 Li polysulfide dissolved in DME. The UV-vis spectra of both solutions were recorded after reacting 106 for 150 s. As shown in Supplementary Fig. 3, DFc + was completely consumed in the reaction with 107 polysulphides, but only partly with Li2S, which indicates that the reaction of DFc + oxidizing solid 108 Li2S is slower than oxidizing polysulfides and thus the former is the rate-determining step. 109 Therefore, the threshold of DFc/DFc + at 2.995 V in Fig. 1a is associated with the reaction of DFc + 110 oxidizing solid Li2S instead of oxidizing soluble polysulfides. 111 Turning to I3 -oxidizing Li2O2, a similar threshold was found around 3.56 V vs. Li + /Li (between 112 0.05 and 0.01 M Li + ), where the kinetics is accelerated 3-fold over only 17 mV. Our previous work 113 has shown that, again, the first electron extraction to form a superoxide is the rate determining 114 step 20 and, therefore, the threshold of ~3.56 V in Fig. 1b is associated with I3 -oxidizing solid Li2O2. 115

Factors governing the thresholds 116
These astonishing but unambiguous thresholds of RM red /RM ox at 2.995 V for Li2S and 3.56 117 V for Li2O2 could originate from many factors such as electrolytes, type of RM, or surface properties 118 of Li2S and Li2O2. We focus further on Li2O2 oxidation. Given that Li + is not involved in the I3 -/I -119 redox couple, I − /I 3 − relies on the Li + activity ( Li + ) in the electrolyte as detailed in Supplementary 120 Note 2. It can be manipulated either by directly changing the Li + concentration in a given solvent 121 or by changing the solvation ability of the electrolyte 35 , which changes the activity coefficient (γ) 122 and Li + . To prove this, a dimethyl sulfoxide (DMSO)/DME mixture electrolyte with various ratios 123 of DMSO/DME and constant 10 mM Li + was used to change the solvation of Li + and thus to 124 manipulate I − /I 3 − (Supplementary Fig. 4). Figure 2a compares the resulting apparent kinetics 125 versus I − /I 3 − with those obtained with varying Li + concentrations in pure DMSO. Although tuned 126 differently, an analogous step-change in kinetics at 3.56 V resulted. For example, 10 mM Li + in 127 DMSO yielded a potential beyond the threshold and fast kinetics while increasing DME raised Li + 128 activity and lowered the potential below the threshold. As the extreme, I3 -in contact with Li2O2 in 129 pure DME evolved almost no O2, Supplementary Fig. 5. Changing kinetics is, hence, not simply 130 arising from the solvent or Li + concentration, but rather from Li + and in turn the potential on 131 the Li/Li + scale. We conclude that the thresholds is genuinely linked to I − /I 3 − . 132

138
To further prove the threshold to be linked to redox potential rather than the particular RM, 139 the same experiments were carried out with 2,2,6,6-tetramethyl-1-piperidinyloxy (TEMPO) and 140 tetraethylene glycol dimethyl ether (tetraglyme) to substitute for LiI and DMSO. Both TEMPO and 141 tetraglyme have been extensively employed in the Li-O2 batteries 18,19 . O2 evolving from TEMPO + in 142 contact with Li2O2 is shown in Supplementary Fig. 6 and the apparent kinetics in Fig. 2b Together with a similar threshold for Li2S oxidation at a different overpotential, we conclude that 148 the thresholds are linked to the intrinsic surface properties of solid Li2O2 or Li2S such as crystal 149 facets. 150

The impact of facets 151
We hypothesize that the exposed facets of solid Li2O2 determine the charge transfer kinetics 152 given the reaction takes place at the surface where certain crystal facets are preferentially exposed. 153 To confirm the impact of facets, we measured the potential-dependent kinetics of I3 -oxidizing 154 amorphous Li2O2 that lacks dominant facets and therefore should likely not show thresholds.

155
Amorphous Li2O2 was synthesized as described earlier and its amorphous state confirmed by XRD, 156 Supplementary is compared with the data from crystalline 157 Li2O2 in Fig. 2c and shows no sudden acceleration at 3.56 V, confirming the threshold at 3.56 V to 158 be associated with specific abundant facets of crystalline Li2O2. 159 To identify the exposed facets, the crystalline Li2O2 was examined with selected area electron 160 diffraction (SAED) in the transmission electron microscope (TEM). The SAED pattern taken down 161 the [1120] zone axis, Supplementary Fig. 8b, is well indexed to Li2O2 (P63/mmc). The elongated 162 particle extends in [0001] direction with the (1120) facet dominating the surface followed (0001), 163 Supplementary Fig. 8a. Given that these facets dominate the surface of the Li2O2 crystallites, their 164 properties should predominantly govern the kinetics. 165

Thresholds for ( ) and (0001) facets 166
We further explored the chemistry underpinning the threshold potential for Li2O2 oxidation 167 using density functional theory (DFT) calculations. Particularly, we determined the overpotentials 168 needed to oxidize the dominating facets, in turn rationalizing the threshold potential to activate 169 fast oxidation pathways. We go beyond previous DFT work modelling Li2O2 oxidation, which only 170 allowed for full removals of the stoichiometric formula via electrochemical steps 37-40 . I.e., two Li + 171 and one O2 via either −Li + , −O2, −Li + or −Li + , −Li + , −O2. However, recent experimental work 172 highlighted the dominance of superoxide disproportionation as the O2 evolving step in general and 173 for the formation of the highly reactive singlet oxygen ( 1 O2) in particular 17,20,31,41,42 . We therefore 174 explicitly allow for disproportionation as well. To do so, we did not limit the charging process to a 175 stoichiometric formula (i.e. two Li + per O2), but allow for more than two Li + ions to be removed 176 before O2 evolves. 177 Using the computational procedure detailed in the Methods, we calculated the reaction 178 energy for each intermediate reaction step. Steps are either electrochemical to desorb Li (one 179 electron and one Li + ) or chemical to desorb O2. After each Li removal, the system relaxed to 180 equilibrium with the remaining atoms reorganizing, releasing the reorganization energy ∆ reorg 181 in the j th step and the entire slab assuming a new total energy ∆ . Li is removed from the Li2O2 182 surface one after another with the assistance of an overpotential . The energy ∆ required to 183 desorb O2 chemically after removing j Li indicates the ease of the overall process to desorb j Li and 184 one O2. At least two Li need to desorb before any O-O moiety could become superoxide-like. Hence, 185 O2 desorbing after two Li would refer to direct oxidation of a peroxide moiety to O2. O2 desorbing 186 after removing four or more Li would refer to disproportionation, leaving behind a Li-deficient Li2-187 xO2 surface. 188 We examined the dominant (1120) and (0001) facets whose structures are shown in Fig. 3a,d. 189 The structural unit with the O-O dimer surrounded by six Li atoms is shown in Supplementary Fig.  190 9   adjacent superoxides exist at the surface as also seen in the Bader charge, Supplementary Fig. 10b.

204
Since the O2 desorption energy is with 0.6 eV still significant, spontaneous O2 desorption appears 205 unlikely. However, after removing six to seven Li, the relaxed O-O bond length is close to the 1.23 Å 206 of molecular O2, Fig. 3c, which is no longer strongly chemically bonded, Fig. 3f. Importantly, this 207 process can be interpreted as disproportionation. As indicated by the Bader charge after removing 208 beyond for the central moiety, the neighbouring ones remain close to the lengths of initial peroxide, 212 Supplementary Fig. 10. This surface disproportionation leaves behind a Li-deficient Li2O2 surface 213 and an easily released O2 molecule. 214 Figure 3b shows the corresponding reaction energy profiles for the electrochemical steps at 215 various overpotentials. A minimum overpotential of 0.54 V is required to make the process all the 216 way to seven Li removals energy-downhill, where O2 is released most easily. Lower overpotentials 217 mean higher energy barriers for O2 release, which is associated with low rates. Consequently, the 218 0.54 V are the overpotential required to activate an overall fast oxidation/O2 release pathway at 219 the (1120)  remaining close to the initial value, Supplementary Fig. 11. An overpotential of 0.78 V is required 229 to make the energy profile downhill for the entire process to take place spontaneously (Fig. 3d).

230
This predicts a second threshold at 0.78 V (or 3.74 V vs Li/Li + ) where RM ox oxidizing Li2O2 is expected 231 to accelerate further. 232 To confirm this hypothesis and to identify the second threshold experimentally, TEMPO was 233 used as the RM in TEGDME where we could manipulate TEMPO/TEMPO + to above 3.7 V. Figure 4  234 shows the measured rate constant over the full voltage range. The first threshold is followed by a 235 gradual increase up to ~3.7 V, where another steep acceleration followed with kinetics doubling. 236 This increase is centred around 3.74 V or an overpotential of 0.78 V and hence matches perfectly 237 the DFT prediction.

244
Correctly predicting the two thresholds strongly supports the facet-dependent reaction 245 pathways during mediated oxidation of Li2O2. To better understand the difference between these 246 two facets, it is helpful to consider the reorganization energy ∆ reorg shown in Supplementary 247 herein is the total reorganization energy between initial and product states. Next to the 259 reorganization energy of the Li2O2 slab as discussed above ( Supplementary Fig. 10), it also accounts 260 for the reorganization of the RM and the solvation shell of both reaction partners. Given the 261 complicated multi-step delithiation process until eventual O2 release, rigorous treatment following 262 Marcus theory is beyond the scope of the work, but we suggest that the underlying ideas explain 263 the decreasing kinetics observed here. Overall, the two thresholds and the observed maximum 264 establish target potentials for maximum rates. 265 266

Accelerated kinetics in operation 267
To test the impact of the potential thresholds on batteries, we charged electrodes preloaded 268 with commercial Li2O2 using 10 mM TBAI in DMSO containing 1 M or 0.05 M LiTFSI, where the 269 I − /I 3 − couple operates below/above the threshold potential. These Li + concentrations provide in 270 either case sufficient conductivity. If anything, the somewhat lower conductivity of the 0.05 M 271 (higher potential) electrolyte would lessen the effect of accelerated kinetics. Cells were charged 272 using linear sweep voltammetry and O2 evolution followed by DEMS, Fig. 5. Cells without RM 273 served as base case for direct electrooxidization of Li2O2, Supplementary Fig. 12. Given that above 274 3.6 V I3 − is further oxidized to I2, only the O2 evolution below 3.6 V (indicated by the shaded region) 275 is taken to judge kinetics. Iin 1 M Li + electrolyte roughly doubled the O2 yield compared to absence 276 of the mediator (Fig. 5b, Supplementary Fig. 12b). Lifting I − /I 3 − above the threshold with 0.05 M 277 Li + raised the O2 yield by as much as 5-fold (Fig. 5a), confirming strongly boosted mediated kinetics 278 above the identified threshold.

285
Electrochemically formed Li2O2 may expose dominant facets to different extend than 286 chemically formed (commercial) Li2O2. We therefore did the same experiments except for forming 287 the Li2O2 by discharging the electrodes in DMSO electrolyte, Fig. 5c,d and Supplementary Fig. 12c,d.

288
At low mediator potential (1 M Li + ), the O2 yield doubled against the control without RM while it 289 was boosted more than 5-fold at a high mediator potential (0.05 M Li + ). Analogous results in cells 290 using chemically and electrochemically formed Li2O2 are all in accord with boosted kinetics beyond 291 the threshold that is related with the dominant (1120) facet. 292 The effect is further confirmed using galvanostatic cycling of cells with the same Icontaining 293 electrolytes, Supplementary Fig. 13. In line with above results, the charging overpotential with 294 lower Li + concentration (higher I − /I 3 − ) is lower than that with the high Li + concentration. The 295 charging plateau is with ~3.6 V only slightly above the threshold of 3.56 V. The higher oxidation 296 rate constant allows a smaller overpotential being sufficient to produce a RM ox concentration 297 capable of oxidizing Li2O2 at the applied current. This threshold or switch-on effect with Ionly 298 takes I − /I 3 − to grow by 10 mV, which we have shown can arise from factors such as Li + 299 concentration and type of solvents. Therefore, unintentionally positioning I − /I 3 − below or above 300 the threshold may explain some contradictory conclusions and debates about the capability of I3 -301 oxidizing Li2O2 during the charging process in literature 35,45 , which span from highly active to nearly 302 inactive.

303
Conclusions 304 In summary, we have shown that the kinetics of mediators oxidizing insulating solids such as 305 Li2S, and Li2O2 show distinct potential thresholds, where reaction kinetics accelerate several-fold. 306 The step in kinetics happens over a potential change of as little as 10 mV. and tend to be oxidized at lower overpotential. 317 For mediated oxidation to be fastest, the mediator should exceed the threshold potentials of 318 dominant facets. Adjusting the potential and boosting rate capability may be as simple as reducing 319 the Li + concertation as long as ionic conductivity remains sufficient. The results resolve 320 contradictory conclusions in the literature about the ability of the I3 -/Iredox couple to oxidize Li2O2.

321
We give a rational for the most effective use of RMs to oxidize insulating active materials such as 322 those in metal-sulfur, metal-air, or metal-CO2 batteries. The properties and abundance of individual 323 facets of the solid product determine required RM potentials for maximum charging rates.       Supplementary Fig. 10. The details will be discussed below.

441
The reaction free energy of the above reaction is

462
After obtaining the bulk redox potential, the next step is to explore the intrinsic barrier of the 463 surface decomposition. It is widely accepted that the Li2O2 decomposition includes both the 464 electrochemical step (the desorption of Li + ) and the chemical step (the desorption of O2), respectively.

465
The reaction energy of the j th (j ranges from 1 to 7) electrochemical step ( ) is defined as

469
The chemical step for the j th step is defined as: