Monitoring the Location of Cathode-Reactions in Li-O₂ Batteries

The oxygen electrode consisted of carbon black (C-Nergy Super C-65) and PTFE (Sigma Aldrich) as the binder in a weight ratio of 90:10, with a surface area of 61 m² g⁻¹, as previously reported.³⁶ Ethanol was used as the dispersing agent to form slurry that was kept under stirring overnight. The slurry was air-brushed over a stainless steel mesh (Alfa Aesar, 0.05 mm wire diameter), which works as current collector at the cathode side, and left drying at room temperature for a week in air. Finally, 10 mm diameter electrodes with carbon loadings between 0.20 and 1.32 mg were prepared and used in this study. Carbon loading was evaluated using a XP6 microbalance (Metler Toledo) with a resolution of ±1.0 μg.

Swagelok-type cells were used to perform the electrochemical measurements. Li-O₂ cells were assembled with a lithium foil anode (10 mm in diameter), a glass fiber separator (Whatman), electrolyte (0.1M LiClO₄ in DME, Sigma-Aldrich, 99.5%) and the oxygen electrode. A stainless steel mesh was used as the current collector. The cell assembly was carried out inside a glove box with Ar flux. For the electrochemical measurements, the cells were purged with dry O₂ and pressurized to 1 atm. DME was distilled then further dried for several days over freshly activated molecular sieves (type 4Å) resulting in a final water content of < 10 ppm (determined using a Mettler-Toledo Karl Fischer titration apparatus). Prior to use, LiClO₄ was dried with a Buchi oven by heating under vacuum at 120°C for 24 hours.

Electrochemical tests were conducted between 2 and 4.6 V at room temperature at 90 mA g_carbon⁻¹ using a Biologic-sas VMP control potenciostat. The current density was systematically varied for the cyclability studies in order to analyze its influence on the battery performance. The specific testing conditions are detailed later in the cyclability section.

Discharge products were characterized by means of X-ray powder diffraction (XRD) using a Phillips PW1710 diffractometer with Cu Kα (k = 1.5418 Å) radiation. XRD pattern was recorded using a gastight sample holder to prevent reaction of Li₂O₂ with moisture and CO₂. Diffractograms were recorded in the 20–60° 2θ range (0.02° step) and the obtained data were fitted using the FULLPROF program.³⁷ The morphology of the electrodes and the reaction products were examined by scanning electron microscopy (SEM) using a SEM Quanta 200 FEG operated in low vacuum mode at an accelerating voltage of 20.0 kV. X-ray photoelectron spectroscopy (XPS) was performed in a SPECS (Berlin, Germany) system equipped with a Phoibos 150 1D-DLD analyzer and monochromatic Al Kα radiation (1486.6 eV). An initial analysis of the present elements was conducted at wide scan mode (step energy 1 eV, dwell time 0.1 s, pass energy 40 eV) and after that, high-resolution spectra of the found elements were acquired (step energy 0.1 eV, dwell time 0.1 s, pass energy 20 eV) with an electron take-off angle of 90°. The binding energy of the adventitious carbon (C1s) was set at 284.6 eV to correct sample charging. The spectra were fitted with the CasaXPS 2.3.16 software which models the Gauss-Lorentzian contributions after background subtraction (Shirley). The specific surface area (SSA) of the materials was measured by N₂ adsorption-desorption isotherms at 77 K using a surface area analyzer (Micromeritics, ASAP 2020). Prior to the measurements, the C-65 commercial carbon and the C-65/PTFE (90:10) were degassed at 250°C for 3h and 130°C for 5h under vacuum (10⁻³ mbar), respectively. Brunauer-Emmett-Teller method (BET) was utilized to calculate the SSA, whereas mesopore size distributions were derived from the desorption branches of the isotherms based on the Barrett-Joyner-Halanda model (BJH).

Oxygen electrodes with varying carbon mass were prepared in order to compare their discharge performances in a Swagelok based Li-O₂ cell. Table I summarizes the carbon loading for each sample, the applied current density and the first discharge specific capacity expressed in different units. The corresponding discharge profiles are shown in Figure 1. The onset potential is between 2.6 and 2.65 V and most of the discharge reactions take place between 2.68 and 2.5 V, similarly to those observed in the literature.^5,38,39

Zoom In Zoom Out Reset image size

When capacity is expressed in mA h g⁻¹ sample A presents a better performance. As shown in Table I, specific capacity (mA h g⁻¹) decreases as the mass of carbon increases. However, when current density is expressed in mA cm⁻² and capacity in either mA h cm⁻² or mA h units, the highest specific capacity is obtained for sample E, which corresponds to the highest carbon loading and current density.

Different specific capacities have been obtained for the electrodes. All electrodes were, however, cycled at the same intensity per carbon mass, suggesting that the entirety of the cathode material is not necessarily active. Furthermore, the fact that the thinner cathodes exhibit a higher specific capacity in mA h g⁻¹ while the thicker electrodes present higher specific capacities in mA h cm⁻² and mA h, evidences a difference on the ability toward ORR as the carbon amount increases. As previously mentioned, oxygen solubility is an important factor that influences the active electrode area in the discharge reaction. Figure 2a shows the schematic representation of the cathode before discharge. A significantly lower concentration of oxygen is observed at the electrode/separator interface due to limited oxygen mass transport to the reaction site in the liquid electrolyte present in the pores.²³ Thus, oxygen concentration is higher at the O₂/electrode interface and the concentration is reduced as the distance from this interface increases. Consequently, there are more oxygen molecules available for the ORR in the proximities of the O₂/electrode interface. Therefore, discharge products are favorable formed in the proximities of the surface exposed to the stainless steel mesh compared to the surface exposed to the separator. The accumulation of discharge products at the O₂/electrode interface leads to pore clogging (Figure 2b), resulting in the capacity fade observed in the final stages of the discharge profile shown in Figure 1.

Zoom In Zoom Out Reset image size

In order to corroborate this point, an additional discharge (sample F in Table I) was performed. The current rate in mA cm⁻² was the same as sample B (0.1223 mA cm⁻²), but with only half of the carbon mass, and consequently the current rate in mA g⁻¹ significantly differs from sample B (90 vs. 135 mA g⁻¹, respectively). A comparison of the discharge profiles is shown in Figure 3. The battery discharged at higher current density delivered lower specific capacity in mA h cm⁻² units (Figure 3a). This trend, however, was inverted when specific capacity was expressed in mA h g⁻¹. The reason lies in the fact that when the specific capacity of B is expressed in mA h g⁻¹ non-active carbon mass is included in the capacity calculations, leading to an underestimation of its real capacity.

Zoom In Zoom Out Reset image size

A post-mortem analysis of the E electrode was conducted in order to prove this hypothesis. The choice of this electrode is due to its highest carbon loading among all the studied samples and, therefore, the effect of the oxygen solubility on the electrode/separator interface, due to the longer pathway of oxygen molecules to reach to this interface, should be more noticeable for this sample.

Figure 4a shows the XRD patterns (2θ = 20–60°) of the cathode after discharge and the protective polymer film used for the air-tight oxygen system. Strong reflections from the polymer film on the discharged electrode at 2θ = 21.7° and 26.7° make the characterization of the discharge products difficult. Consequently, Rietveld refinement was performed in a 2θ range between 30° and 60° in order to identify the discharge product structure (Figure 4b). Within this range, six distinct reflections (2θ = 32.92°, 34.99°, 47.31°, 48.83°, 58.78°, 58.83°) are found which correspond to the formation of hexagonal Li₂O₂ (JCPDS 74-0115). Additional reflections at 2θ ∼ 36.4°, 40.3° and 54.3° can be attributed to the polymer foil used in the air-tight system and, due to the absence of crystallographic information, have been excluded from the refinement. Furthermore, two reflections (2θ = 43.58°, 50.76°) from the stainless steel mesh were identified and taken into consideration. From the residual plot, a mismatch in the relative intensities of the Li₂O₂ reflections at lower diffraction angles becomes obvious, i.e. the first reflection at 2θ = 32.92° shows considerably lower intensity than expected. We attribute this observation to the electrode configuration in which the stainless steel mesh is located on top of the C-based electrode. As the measurement was conducted in Bragg-Brentano geometry, diffraction at lower angles was most likely influenced by the relative height of the materials. Thus, the stainless steel mesh acted as a physical barrier for part of the radiation diffracted by the Li₂O₂-phase leading to lower absolute intensities. At higher diffraction angles, this effect dissapears and a normal behavior of the intensities is observed. Hence, it can be concluded that Li₂O₂ is the main discharge product, and further evidence is provided in the following sections. It is also worth noting that no Li₂CO₃ side-reaction product was observed although it is widely accepted that the Li₂O₂ formed during the discharge process can subsequently react with the electrolyte leading to the formation of Li₂CO₃.^39–41 In addition, it must be noted that the carbonate formation can be also associated to the instability of the carbon electrode.⁴² SEM was used to locate the discharge products.

Zoom In Zoom Out Reset image size

Discharge products accumulated in the O₂-electrode interface are shown in Figure 5. Toroidal-shaped characteristic Li₂O₂ particles with a diameter of 2–3 μm are found. This type of discharge product has been widely described by several authors^38,43–45 at low discharge rate currents. Recently Aetukuri et al.⁴⁶ attributed this discharge product morphology to the presence of water in the electrolyte when DME solvent was used. Nevertheless, water content in the electrolyte was controlled to be less than 10 ppm by Karl Fisher measurements. In addition, toroidal-shaped discharge products observed by Aekuturi et al.⁴⁶ can be described as aggregates of flat discs forming toroid-like structures which differ from those in Figure 5. No discharge products were observed on the electrode-separator interface which supports our hypothesis. As previously discussed, Lu et al.³⁰ reported that an electrode with a 20 μm thickness was fully active, using DME as electrolyte solvent. Our results suggest that this does not happen when thicker electrodes are used. 3D computed tomography mapping of the lithium discharge products and air electrode reported by Nanda et al.³¹ show, however, that the discharge products appear at both the O₂/electrode and electrode/separator interface. The authors attribute this phenomenon to the low discharge current density (5 μA cm⁻²). In our study, discharge products do not appear at the electrode/separator interface and the concentration is reduced as the distance from the O₂/electrode interface increases, in good agreement with the simulations reported by Andrei et al.²⁸ In order to definitely compare the activity toward ORR of these two surfaces by the accumulation of the discharge products XPS measurements were carried out. The results are shown in Figure 6.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 6a shows the XPS spectrum at the O₂/electrode interface. The presence of the Li₂O₂ discharge product detected by XRD was confirmed at the O₂/electrode interface at ∼56 eV in the Li 1s spectra. The existence of Li₂CO₃ (at ∼290 eV in the C 1s spectra) in the interface was also found. On the other hand, in Figure 6b at the electrode/separator interface no discharge product (Li₂O₂) or side reaction (Li₂CO₃) products were detected, proving the inactivity of this surface. In summary, the XPS measurements confirm that the positive electrode shows different activity toward ORR in each surface, consistent with the SEM analysis. Thus, it can be concluded that the activity of the cathode toward ORR is reduced as the distance from the oxygen gas-electrolyte interface is increased.

Finally, the effect of carbon loading and current rate on the cyclability was evaluated. Electrodes loaded with 0.2 and 0.4 mg of carbon were cycled in the voltage window of 2–4.5 V with a fixed discharge capacity of 0.235 mA h cm⁻² at a current rate of 0.045 mA cm⁻² (Figure 7a). The battery loaded with higher carbon mass was able to perform 21 cycles before capacity fade while the battery loaded with 0.2 mg of carbon was only able to perform 11 cycles. This could be attributed to the blocking of the active area with increasing cycle numbers due to the non-reversibility of the side products. As previously discussed, the entire electrode is not active for ORR during the first discharge as the Li₂O₂ mainly forms at the oxygen/electrode interface where the oxygen concentration is higher,⁴ filling the pores of the carbon particles. The formed toroid-shaped products gradually block the oxygen diffusion paths into the interior of the cathode leading to incomplete utilization of the cathode pore volume. However, during charge the formation of insulator products (e.g. Li₂CO₃) could result in a partial pore blocking avoiding the restriction of the ORR to the top of the oxygen side of the electrode.⁴⁷ The formation of Li₂O₂ even at the cathode/separator interface of the thicker electrodes can therefore occur (see Figure 8). Finally, the progressive accumulation of side products results in the capacity fade where thicker electrodes are able to deliver the selected specific capacity during higher number of cycles. In addition, batteries with similar carbon loadings (0.4 mg) were cycled at different current rates (0.045, 0.090 and 0.225 mA g⁻¹) with a fixed discharge capacity of 0.235 mA h cm⁻² (Figure 7b). It was confirmed that low current rates enhance the lifetime of the battery, as lifetime in the battery cycled at 0.045 mA cm⁻² was over 4 times the battery cycled at 0.225 mA cm⁻². This fact can be ascribed to two factors. First, higher discharge rates lead to larger gradient of the oxygen concentration across the carbon cathode, which restricts the reaction to the region close to the gas phase/cathode interface.⁴ Second, the difficulties for removing discharge products at higher current densities, as has been demonstrated by Adams et al.³⁸ and Wang et al.,⁴⁸ are expected to reinforce this decrease on the performance.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It has been demonstrated by means of SEM and XPS that discharge reaction occurs mainly at the oxygen gas-electrolyte interface. The percentage of active area in the cathode material for ORR is reduced as the distance from the O₂/electrode interface increases due to limited oxygen mass transport to the reaction site in the liquid electrolyte present in the pores. The specific capacity expressed in mA h g⁻¹ is directly related to the mass loaded onto the cathode. In addition, ORR and discharge product accumulation has been demonstrated not to occur in the entire C-65 electrode area in a DME-based electrolyte at 90 mA g⁻¹. Consequently, if expressed in mA h g⁻¹, the capacity will be underestimated for thicker cathodes, as the reaction occurs to less extent in the rest of the cathode. Thus, reducing the cathode thickness to the most active area results in higher specific capacities expressed in mA h g⁻¹. However, this cannot be considered as a better performance and comparison of our results with those reported in literature is not possible due to the lack of electrochemical and electrode data. ORR takes place throughout the cathode but to different extents as shown by the increase in capacity (mA h cm⁻² or mA h) with carbon loading. Finally, it has been confirmed that the increase of the current rate and decrease of the carbon loading lead to earlier battery death.

This work demonstrates that even though specific capacity is an important parameter for real-world batteries that should not be the goal of fundamental materials research due to the wide variations observed in the same material with small differences in carbon loading and measurement conditions.

This work has been partially financed by the Ministerio de Educación y Ciencia under project MAT2010-19442 and MAT2013-41128-R and by the Eusko Jaurlaritza/Gobierno Vasco under projects IT-570-13, SAIOTEK (S-PE11UN064 and S-PE12UN140) and ETORTEK (ENERGIGUNE'12 - IE12–335). I. Landa-Medrano thanks the Universidad del País Vasco (UPV/EHU) for his predoctoral fellowship. R. Pinedo thanks the Universidad del País Vasco (UPV/EHU) for funding his research activities under the "Especialización de Personal Investigador del Vicerrectorado de Investigación de la UPV/EHU" programme. N. Ortiz-Vitoriano acknowledges a Marie Curie International Outgoing Fellowship within the EU Seventh Framework Programme for Research and Technological Development (2007–2013). The authors gratefully acknowledge Dr. Karthik Mani and Edurne Redondo for the BET measurements and Begoña Acebedo and Egoitz Martin for the SEM and XRD measurements, all of them from the CIC Energigune. In addition, authors thank Dr. María Belén Sánchez from SGIker of UPV/EHU for the XPS measurements.