The interest in secondary Li/O₂ batteries has grown rapidly over the past two decades, as they exhibit a practically achieveable specific energy of about 1,700 Wh/kg, which equals that of gasoline and is well beyond those of conventional lithium ion, Ni metal hydride and Zn/air batteries.^1,2 The large energy density is due to the very light-weight cell components: a Li metal anode, a porous C cathode (gas diffusion electrode, GDE) and gaseous O₂ as active cathode material. During discharge LiO₂ and Li₂O₂ (as well as Li₂O) are formed via the oxygen reduction reaction (ORR) in the porous carbon matrix. These products are oxidized again upon recharge.

This seemingly simple system, however, has some major drawbacks due to the decomposition of the organic electrolytes during battery cycling – which is caused by the reaction of the respective electrolyte with either the Li anode, different components of the GDE and/or even the desired ORR products. The decomposition products are deposited in the GDE pores, leading to a minimized active electrode surface, decreasing capacities and ultimately cell failure upon continued cycling.³ There are two major reaction mechanisms leading to the formation if Li₂O₂, and the prevailing mechanism depends on the electrolyte. Li₂O₂ toroids are a common discharge product of the so-called solution growth pathway proceeding via an EC mechanism⁴: O₂ is reduced to O₂^-, which is dissolved and forms LiO₂. The solvated LiO₂ intermediates subsequently disproportionate to solid Li₂O₂ particles with different morphologies (which depend on the nature of the electrolyte)⁵. The ORR proceeds primarily via this mechanism when high DN electrolyte solvents are used and/or discharge parameters, such as low discharge current densities and low overpotentials, are applied⁶. With low DN solvents and/or high current densities and overpotentials, the discharge reaction proceeds via the surface growth pathway.

In this work, galvanostatic cycling and CV measurements were conducted with 1 M mixtures of LiTFSI/DMSO and LiTFSI/TEGDME, with and without additions of the redox mediator dimethylphenazine, DMPZ⁷, to analyze the electrolyte influence on the cycling behavior. At different states of charge, the GDE surface and pores were investigated by SEM and XRD to obtain information on the nature, the morphology and the distribution of the product deposits on and in the carbon structure.

Galvanostatic cycling measurements show a larger discharge capacity and lower discharge overpotential with LiTFSI/DMSO as well as a better recharge behavior in the first cycle, although both electrolytes exhibit very high recharge potentials. The TEGDME-based electrolyte on the other hand exhibits a higher stability over the first five cycles, whereas the discharge capacity of LiTFSI/DMSO decreases drastically already upon second discharge. CV measurements confirm these absolute and relative trends of the discharge capacities.

A SEM investigation of the GDE surfaces after first discharge reveals a considerable difference in discharge product morphology depending on the electrolyte. This is in agreement to previously published results: with DMSO as solvent the product consists of toroidal particles with diameters of approximately 200 nm, whereas discharge with the TEGDME-based electrolyte yields a product layer on the porous cathode surface.^5,8 Differential capacity analysis confirmed the difference in discharge product formation: with LiTFSI/TEGDME two reduction processes were detected to contribute to the discharge capacity, whereas discharge with LiTFSI/DMSO contains only one reduction process. X-ray diffraction after the first discharge with LiTFSI/DMSO shows the presence of crystalline Li₂O₂ in the GDE and the absence of crystalline LiOH. The discharge with the TEGDME-based electrolyte, on the other hand, did not yield any crystalline Li₂O₂ (or LiOH). FIB-SEM measurements after the first discharge with each electrolyte furthermore showed that the GDE pores were almost completely blocked by bulk product.

The addition of the DMPZ significantly reduced the charging potential, both for the DMSO and the TEGDME electrolyte. The cycling performance was, however, only improved for DMSO, and not for TEGDME.

1. Bruce et al., Nat. Mater. (2012) 11, 19-29.
2. Yao et al., Angew. Chem. Int. Ed. (2016) 55, 11344-11353.
3. Sharon et al., Isr. J. Chem. (2015) 55, 508-520.
4. Johnson, L. et al., Nat. Chem. 6, 1091–1099 (2014).
5. Mitchell, R. R., et al., J. Phys. Chem. Lett. 4, 1060–1064 (2013).
6. Aurbach, D., et al., Nat. Energy 1, 16128 (2016).
7. Lim et al., Nature Energy (2016) 1, 16066.
8. Aetukuri et al., Nat. Chem. (2014) 7, 50-56.