Promoting solution phase discharge in Li-O 2 batteries containing weakly solvating electrolyte solutions

On discharge, the lithium-O 2 battery can form a Li 2 O 2 film on the cathode surface, leading to low capacities, low rates and early cell death, or it can form Li 2 O 2 particles in solution, leading to high capacities at relatively high rates and avoiding early cell death. Achieving discharge in solution is important and may be encouraged by the use of high donor or acceptor number solvents or salts that dissolve the LiO 2 intermediate involved in the formation of Li 2 O 2 . However, the characteristics that make high donor or acceptor number solvents good (e.g. high polarity) result in them being unstable towards LiO 2 or Li 2 O 2 . Here we demonstrate that introduction of the additive 2,5-Di-tert-butyl-1,4-benzoquinone (DBBQ) promotes solution phase formation of Li 2 O 2 in low polarity and weakly solvating electrolyte solutions. Importantly, it does so while simultaneously suppressing direct reduction to Li 2 O 2 on the cathode surface, which would otherwise lead to Li 2 O 2 film growth and premature cell death. It also halves the overpotential during discharge, increases the capacity 80-100 fold and enables rates > 1 mA cm -2areal for cathodes with capacities of > 4 mAh cm -2areal . The DBBQ additive operates by a new mechanism that avoids the reactive LiO 2 intermediate in solution. EC cat reaction, electrochemical

The high theoretical specific energy of the rechargeable Li-O2 battery has generated intense interest in the possibility of a practical device that could deliver energy storage significantly in excess of today's lithium-ion batteries [1][2][3][4][5][6][7][8][9] . However, major challenges hinder the development of such a technology [1][2][3][4][5][6][10][11][12][13][14] . Typically a Li-O2 battery is composed of a lithium metal anode separated by an aprotic electrolyte solution from a porous O2 cathode. The reaction at the cathode involves, on discharge, the reduction of O2 to form Li2O2, with oxidation of the latter on charge. Growth of Li2O2 on the cathode surface leads to low capacities, poor rates and early cell death [15][16][17] . In contrast, if Li2O2 can be induced to grow in the electrolyte solution then high discharge capacities at relatively high rates and avoiding early cell death is possible 15 . It is clearly important to operate a Li-O2 battery in which Li2O2 grows in solution.
A number of groups have elucidated the mechanism of O2 reduction to Li2O2 on discharge 15,16,[18][19][20][21] . The reduction proceeds through the following general steps: O2 + Li + + e -→ LiO2 (1) 2LiO2 → Li2O2 + O2 (2) LiO2 + Li + + e -→ Li2O2 (3) Whether Li2O2 grows in solution or as a film on the electrode surface depends on the solubility of the LiO2 intermediate; if LiO2 dissolves in the electrolyte solution then Li2O2 grows in solution. Solubility of LiO2 depends on the strength of the cation and anion solvation, i.e. on the solvent and salt donor and acceptor numbers 15,[22][23][24] . However, the very properties that make a good solvent for LiO2, (high polarity) often makes the solvent more susceptible to nucleophilic attack or proton abstraction by the reactive O2radical, leading to undesirable side-reactions 25,26 . The challenge is to form Li2O2 in solution on discharge in low donor number (weakly solvating) solvents.
Soluble catalysts or salts with high donor numbers (DN) can in principle promote solution phase growth of Li2O2 in low donor number solvents (e.g. ethers) 22,[27][28][29][30] . High DN salts have been shown to increase the capacity 4 fold and reduce the discharge overpotential by 30-50 mV over low DN salts 22 . Viologens 27,28 , phthalocyanines 29 and quinones 30 have been investigated as possible soluble reduction catalysts. While the studies of such catalysts are important, in most cases there is little or no direct evidence demonstrating that they promote formation of Li2O2 in solution and not on the electrode surface because they rely on electrochemical measurements alone. Yet past work on Li-O2 batteries has shown how essential it is to provide more than electrochemical evidence in this field 31 .
In some cases, soluble catalysts show an increase in discharge voltage (lower overpotential) as small as e.g. 40 mV 28,29 , which is very unlikely to be sufficient to shut-off the direct reduction of O2 to Li2O2, essential to stop detrimental Li2O2 film formation. Also, none of the previous studies in low donor number solvents exhibited a significant increase in capacity on discharge at a relatively high rate, which is important for a successful Li-O2 battery.
Here we demonstrate that addition of DBBQ (2,5-Di-tert-butyl-1,4-benzoquinone) to a weakly solvating (low DN) electrolyte solution, LiTFSI in ether 22 , promotes O2 reduction to Li2O2 in solution while halving the discharge overpotential (increasing the discharge potential), suppressing the growth of a Li2O2 film on the electrode surface thus postponing cell death, increasing the discharge capacity 80-100 fold and permitting discharge at relatively high rates > 1 mA cm -2 areal for an electrode capacity of > 4 mAh cm -2 areal. It operates by a new mechanism that does not involve the reactive LiO2 as an intermediate; the new mechanism also decouples the link between the nature of the electrolyte solution (solvating power) and the nature of the product (particles or surface film). The search for truly stable electrolyte solutions for Li-O2 batteries will focus on very low polarity and hence weakly solvating solvents. The significance of the present work is that if such stable solvents can be identified then DBBQ provides a route to solution growth of Li2O2 and hence potentially high rates, high capacities and sustained cycling, avoiding early cell death.

CV studies with DBBQ
The potential at which O2 is reduced to Li2O2 (the discharge plateau in a Li-O2 cell) is lower than the thermodynamic potential for O2/Li2O2, 2.96 V. A CV corresponding to this process is shown in Fig. 1.
To promote O2 reduction to Li2O2 in solution in low DN solvents while suppressing the direct reduction of O2 to form a Li2O2 film, which would otherwise passivate the electrode 15,16,21 , it is necessary to carry out the reduction of O2 to Li2O2 in solution at a higher potential than the surface reaction, which also has the advantage of increasing the cell discharge potential closer to its thermodynamic potential of 2.96 V (reducing the overpotential). To achieve this, molecules with a redox potential somewhat higher than the potential at which O2 is reduced (discharge plateau in a Li-O2 cell) are required. Quinones were selected as they are known to exhibit potentials in the relevant range 30,32 . Several quinones were investigated but most were found not to enhance O2 reduction, see Supplementary Fig. S1. Electrolyte preparation and cell assembly are described in the Supplementary Information. DBBQ, in contrast, showed promising electrochemistry, Fig. 1. The cyclic voltammograms for DBBQ obtained in 1 M LiTFSI in tetraethylene glycol dimethyl ether (TEGDME) and dimethoxyethane (DME) at a gold electrode under Ar exhibit quasi-reversible behavior, Fig. 1 and Supplementary Fig. S2. In the presence of O2, the reduction peak is enhanced significantly. Such a CV is similar to that of a catalyzed reduction 33 , where a redox active species, in this case DBBQ, is reduced and then takes part in a chemical reaction, here with O2 to form Li2O2, resulting in the rapid regeneration of more DBBQ, giving rise to the increased reduction current. The reduction potential is significantly higher than for the direct reduction of O2, Fig. 1, thus effectively suppressing the direct reduction of O2 to Li2O2 films on the electrode surface. The mechanism of O2 reduction by DBBQ is discussed further later; demonstration of the efficiency of DBBQ in promoting Li2O2 formation in solution and not on the electrode surface, as well as increasing the discharge potential of Li-O2 cells is presented below. CVs under Ar (blue) and O2 (red) and for direct O2 reduction without DBBQ (black). DBBQ concentration was 10 mM and CVs were carried out at planar Au electrodes, scan rate 100 mV s -1 .

Enhancing the discharge of Li-O2 cells with DBBQ
Li-O2 cells were constructed as described in the Supplementary Information (Methods Section). The cathode was a binder-free carbon-fiber gas diffusion layer (GDL, Freudenberg), similar to cathodes used widely for aprotic O2 cells 20,34 . Carbon electrodes are relatively stable on discharge 35 , which is our focus here. The anode consisted of LixFePO4, as used in previous Li-O2 studies instead of Li metal in order to avoid any oxidation of the anode by O2 36 . The LixFePO4 potential vs. Li + /Li, 3.45 V, was used to express all potentials in this work on the Li scale. The electrolyte solution was in all cases 1 M LiTFSI dissolved in the low donor and acceptor number ethers, TEGDME or DME.
Cells containing TEGDME and DME saturated with O2 (under 1 atm. of O2), were each discharged at several different areal current densities with and without DBBQ, Fig. 2. In the absence of DBBQ, the cells died rapidly, exhibiting very small capacities and poor rate capability, in accord with previous observations 3, 37 . The cells with DBBQ discharged under the same conditions exhibited a dramatic improvement, delivering up to 80 to 100 times higher discharge capacities before end of life. In TEGDME with DBBQ, a capacity of 10.6 mAh cm -2 areal (equivalent to 9.1 mg of Li2O2) was obtained at a current density of 0.2 mA cm -2 areal, while in DME with DBBQ, 7.3 mAh cm -2 areal (equivalent to 6.3 mg of Li2O2) was obtained at 0.5 mA cm -2 areal and 4 mAh cm -2 areal (equivalent to 3.4 mg of Li2O2) at 1 mA cm -2 areal. Moreover, areal current densities of 0.5 mA cm -2 (in TEGDME) and 2 mA cm -2 (in DME) were achieved, while halving the discharge overpotential, compared with the performance in the absence of DBBQ. To estimate the contribution of DBBQ reduction itself to the capacity, the cells were discharged under Ar, for DME and TEGDME, and at the same current densities as in Fig. 2. The discharge curves are given in Supplementary Fig. S3. A negligible capacity was observed. These values are all within the limits of the theoretical capacity for DBBQ reduction of 12.5 mAh m -2 BET.
It has been shown that the limit of Li2O2 film growth is 6 nm 17 , which equates to a maximum capacity of 15 mAh m -2 BET (0.4 mAh cm -2 areal). As is evident in Fig. 2, the cells without DBBQ exhibit end of life below this limit, indicating that Li2O2 formation is predominantly by the surface route. Whereas cells containing DBBQ are able to exceed the limit of film growth by an order of magnitude, signaling predominantly solution growth of Li2O2. To confirm that Li2O2 grows primarily in solution, away from the electrode surface, in the presence of DBBQ, despite the use of low donor/acceptor solvents, the discharged cathodes with and without DBBQ were extracted and examined by SEM. The results are shown in Fig. 3. In both TEGDME and DME, in the absence of DBBQ, the surfaces of the carbon fibers that constitute the GDL were covered with a film and there was no evidence of Li2O2 particles. In contrast, identical cells discharged under the same conditions, except for the presence of DBBQ, show substantial growth of particles in the pores of the electrodes and with the toroidal morphologies expected for Li2O2, Fig. 3. Equally important is that DBBQ suppresses film growth on the electrode surface. This is shown in Fig.  3 (c,g) where there is little evidence of film growth when DBBQ was present until close to cell death. There will always be some direct reduction to form Li2O2 on the surface, even at the higher potential where DBBQ is reduced, as the direct reduction to form a Li2O2 film is suppressed but not eliminated completely. It has been proposed recently that the presence of H2O can itself promote Li2O2 toroid formation in Li-O2 batteries 16,20 . Care was taken to rigorously dry the solvents, electrodes and all cell components used here. The H2O content at the beginning and end of discharge did not exceed 30 ppm, considerably smaller than the quantities required to promote toroid formation; at least 200500 ppm H2O is needed 16,20 . Overall, the SEM images demonstrate that DBBQ has successfully displaced the O2 reduction away from the electrode surface, promoting growth of large Li2O2 particles in the adjacent solution within the pores of the electrode. To demonstrate the particles observed in SEM are indeed Li2O2, powder X-ray diffraction (PXRD), infrared spectrometry (IR) and Raman spectroscopy were carried out on the porous electrodes extracted from the cells. The results are presented in Fig. 4. The PXRD pattern collected on the GDLs discharged in ethers exhibits only peaks associated with Li2O2. The results are confirmed by the IR and Raman spectra in Fig. 4, which also show Li2O2 as the primary product. Although ethers are one of the more stable solvents in Li-O2 batteries, it is known that they are not completely stable 38 . Small peaks associated with lithium acetate/formate and some Li2CO3 are evident as minor by-products in the IR, as identified previously for discharge in ethers 39 . There is little evidence of LiOH. To investigate the presence of any soluble by-products, NMR was carried out on the electrolyte solutions. The details are described in the Supplementary Information. In addition to the peaks associated with the electrolyte solutions, only a tiny peak assigned to lithium acetate was observed, Supplementary Fig. S4. In-situ differential electrochemical mass spectrometry (DEMS) was carried out to investigate the gas consumption on discharge. The procedure is described in the Supplementary information and the results are presented in Fig. 5. No gases were detected other than O2 and in particular there was no evidence of CO2, consistent with the degree of side-reactions in ethers being small. The total O2 consumed and total charge passed were measured and the integral gave a ratio of electrons to oxygen consumed of 2.03 e -/O2, consistent with the dominant reaction on discharge involving Li2O2 formation 38,40 . These results are in accord with charge/mass ratios seen previously for ethers 38 . Taken together, the PXRD, IR, Raman and DEMS indicate that the dominant product on discharge in the presence of DBBQ in ethers is Li2O2 and that it forms relatively large particles in the pores rather than on the surfaces of the porous electrode. The amount of Li2O2 present in the electrode was quantified by chemical analysis using TiOSO4 as described in the Supplementary information. The yield of Li2O2 (observed mass/mass predicted from charge passed) with DBBQ was 95% and 86% in DME and TEGDME, respectively. This compares with 91% and 81% reported previously for DME and TEGDME in the absence of DBBQ 38 . The slightly higher yields indicates that the relatively high surface area of the Li2O2 film that grows on the electrode in the absence of DBBQ leads to more decomposition of the electrolyte solution than is the case for the large particles in solution. It has also been suggested that LiO2 is responsible for solvent decomposition on discharge 26,41,42 and as discussed below, our analysis points to a mechanism that avoids this reactive intermediate.  Attempts to charge the cells after discharge proved fruitless, see Supplementary Fig. S5. This is to be expected since the Li2O2 is not well connected to the electrode surface and therefore direct electrochemical oxidation will be difficult. Therefore, especially in the presence of a reduction mediated discharge, it will be necessary to employ an oxidation mediator to charge the cell, as described previously 36,[43][44][45] .

The mechanism of O2 reduction in the presence of DBBQ
As mentioned above, DBBQ does not operate as an electrocatalyst like, for example, the phthalocyanines described previously 29,46 , for which O2 is bound to the electrocatalyst before, during and after reduction. Neither does it operate as a redox shuttle, transferring electrons from the electrode surface to reduce O2 in solution to LiO2 and further to Li2O2 by an outer sphere reaction. Instead, it operates by a different mechanism that changes the pathway of O2 reduction to Li2O2 avoiding the reactive LiO2 as an intermediate.
The reduction of quinones, such as DBBQ, in Li + electrolyte solutions under Ar is known to form Liquinone complexes, in this case LiDBBQ, Fig. 1 [47][48][49] . In the presence of O2 the reduction potential for DBBQ/LiDBBQ does not change, Fig. 1, indicating that the same reduction reaction (DBBQ to LiDBBQ) occurs, i.e. there is no binding of O2 to DBBQ prior to the initial electron transfer, unlike the phthalocyanines 29 , the first step is as shown in equation (4). However, the reduction current is enhanced significantly, Fig. 1. The observed CV is similar to that of an ECcat reaction, electrochemical reduction followed by a chemical step, in which the reduced form of the redox couple takes part in a chemical reaction that re-generates the oxidized form of the couple to feed the reduction 33 . Here DBBQ is re-generated from LiDBBQ by the latter reducing O2 in a chemical step, which goes on to form Li2O2.
In the absence of DBBQ, reduction of O2 to Li2O2 proceeds via the LiO2 intermediate 15,16,[18][19][20][21] , and it is the need to reach the potential for formation of LiO2 that pins the O2 reduction at a potential (discharge plateau in a Li-O2 cell) significantly negative of the standard potential for Li2O2 formation, 2.96 V, Fig. 1. Where the energetics of an intermediate dictates the potential required to carry out an electrochemical reaction this is referred to as a "thermodynamic overpotential" 50 . In the presence of DBBQ O2 reduction effectively takes place at the potential for DBBQ reduction, Fig. 1 (Fig. 6) and the potential correspondingly raised, as seen in the higher voltage for the discharge plateau in galvanostatic discharge of Li-O2 cells, Fig. 2.
The sequence of proposed reaction steps at the cathode on discharging a Li-O2 cell containing DBBQ is summarized in equations (4) to (6). Equation (4)   These reactions can be summarized by the schematic shown in Fig. 6, and the consequences of this scheme are relatively simple electron transfer and dominate solution phase product formation that translate into high rates and capacities during cell discharge.
As noted above, DBBQ does not act as a conventional catalyst, it does not bind O2 and facilitate LiO2 formation by stabilizing the superoxide intermediate. Instead DBBQ is reduced to LiDBBQ that binds O2 to form LiDBBQ, (equation 5). The characteristics that make DBBQ suitable for this function are, a reduction potential positive of the potential for formation of LiO2 formation thus avoiding direct formation of LiO2, a reduction potential negative of the overall reduction potential to Li2O2 such that a driving force remains to push the reaction towards peroxide formation and the ability to bind O2 when in the reduced form (LiDBBQ).

Outlook
O2 reduction to Li2O2 by the DBBQ mediated route brings a number of benefits. The electrochemistry at the electrode surface is now DBBQ reduction rather than direct formation of Li2O2, in an electrolyte solution that does not dissolve LiO2 (weakly solvating electrolyte solution). As a result, Li2O2 formation is moved into solution without the need for high donor/acceptor number solvents or salts. DBBQ shuts down the direct formation of a Li2O2 film on the cathode, thus postponing cell death, increasing capacity 80-100 fold and facilitates discharge rates of > 1 mA cm -2 areal for cathodes with capacities of > 4 mAh cm -2 areal. The discharge potential is also increased (overpotential is halved). O2 reduction to Li2O2 in the presence of DBBQ follows a new route that avoids the reactive LiO2 in solution. The search for truly stable electrolyte solutions for Li-O2 batteries will focus on very low polarity and hence weakly solvating solvents. The significance of the present work is that if such stable solvents can be identified then DBBQ provides a route to solution growth of Li2O2 and hence potentially high rates, high capacities and sustained cycling, avoiding early cell death. These results demonstrate the importance of moving to a mediated reaction on reduction and imply that the future of the lithium-air battery involves the mediated formation and decomposition of lithium peroxide, where the latter fulfills the role of storage medium only.

Methods
Methods and any associated references are available in the online version of the paper. The authors declare no competing financial interests.

Materials and methods
TEGDME was distilled under vacuum and DME was distilled under Ar. All solvents were further dried for several days over freshly activated molecular sieves (type 4Å, Aldrich) before use. The final water content was < 10 ppm (determined by Karl Fischer titration). Lithium bis(trifluoromethane)sulfonimide (LiTFSI, Aldrich) was dried at 70 o C under vacuum over several days.
Cyclic voltammetry (CV) was performed using a VMP3 electrochemical workstation (Biologic) and a multi-necked, air-tight glass cell within a glove box. The measurements were carried out at room temperature and IR correction was used. 2 mm diameter polycrystalline Au disks (BAS Inc.) were employed as the working electrodes. A platinum wire served as the counter electrode and a partially oxidized LiFePO4 composite electrode behind a Vycor frit served as the reference electrode, as described previously 15 . Swagelok Li-O2 cells were constructed as described previously 51 . Binder-free gas diffusion layers (GDL, H2315, Quintech) served as the O2 electrode. The porosity of the GDLs is 80 %, roughness factor (total surface area/ areal area) is 90 and the Brunauer-Emmett-Teller surface area is below 1 m 2 g -134 . Three pieces of GDLs (4 mm x 4 mm) were stacked to form the cathode giving a final roughness factor of 270 (3 x 90), a glass fibre filter (Waterman) was used as the separator and a partially oxidized LiFePO4 electrode was used as the anode. The two-phase LixFePO4 has a fixed potential of 3.45 V vs. Li + /Li. 200 µl of electrolyte solution was used, consisting of either TEGDME or DME containing 1 M LiTFSI with DBBQ as indicated in the main article. All cell components were dried at 90 o C under vacuum prior to use. Assembled cells were placed in glass tubes, which were filled with dried O2 inside the glove box. Cells were discharged inside an Ar-filled glove box.

Characterisations of discharged electrodes
For post-cycling characterisation, the cells were dissembled in a glovebox and the cathode and separators were rinsed with a small amount of TEGDME or DME, the resulting solutions were subjected to Karl Fischer titration to determine the water content after discharge. The electrodes were rinsed again with DME and dried prior to further characterisation. The morphology of discharge electrodes were observed by FE-SEM using a Zeiss-Merlin. PXRD was carried out with a Rigaku X-ray diffractometer in an air-sensitive holder. FTIR spectra were measured with a Thermo IR spectrometer (Nicolet 6700) in a N2-filled glove box. Raman spectra were measured with a Rinishaw Invia spectrometer (10 mW laser power at 785 nm) with an air-sensitive sample holder. For NMR analysis, 100 µl of electrolyte was extracted from the discharge electrodes and separators then diluted with 0.7 ml of CDCl3, measurements were recorded on a Bruker spectrometer (400 MHz). A DEMS cell was constructed as described previously 40 . A GDL served as working electrode and a partially oxidised LiFePO4 composite electrode served as anode. The electrolyte solution was 10 mM DBBQ in 1 M LiTFSI in TEGDME. A continuous 95% O2 / 5% Ar gas flow was purged through the cell as a carrier gas at a flow rate of 0.3 ml min -1 .
The quantity of Li2O2 formed was determined by UV-vis spectrometry (Thermo Evolution 200) using a UV-vis titration method reported previously 20,52 . The unwashed discharged electrode and separators were added to a vial containing a known amount of water; Li2O2 reacts with water to produce H2O2 in solution. 1 ml of this solution was mixed with 2 ml of 2 % TiOSO4 dissolved in 1 M H2SO4 solution and a yellowish complex [Ti(O2)] 2+ (λmax= 405 nm) was formed. The UV-vis absorption spectrum of the solution was measured and compared to a calibration curve, which was obtained by measuring solutions with known amounts of commercial Li2O2 (Aldrich). The purity of commercial Li2O2 was determined by titration using KMnO4 and this was taken into account when constructing the calibration curve.