Synchrotron X-ray tomography and scanning electron microscopy were applied to elucidate the spatial distribution of discharge product (NaO2) in the carbon cathode of sodium-oxygen batteries. Various batteries were discharged galvanostatically and their cathodes were analyzed. We observe a particle density gradient along the cathode that scales with the current density applied. Besides, we show that the particle size and shape of discharge product strongly depend on current density, and on whether the particles are deposited close to the oxygen reservoir or near the separator. We correlate our findings to transport limitations for the supplied oxygen and gain crucial information for optimal operation of sodium-oxygen batteries. Our findings imply that for low current densities pore clogging might occur, and that for elevated current densities small high surface area particles with limited electric conductivity form; both phenomena can decrease the available discharge and charge capacity significantly.
Sodium-oxygen batteries (Na/O2 batteries) are attractive candidates for future energy storage applications from the theoretical point of view, but fail so far to fulfill the optimistic expectations1. Up to now, their practical energy density and cycling stability are deficient2,3. Despite continuous progress4,5, fundamental understanding of the electrode reactions6 and the formation of reaction products is still limited7. Thus, deeper insight is needed to propel development of Na/O2 batteries.
The cell reactions during discharge of Na/O2 batteries are as follows2: Oxygen is reduced at an active site of the positive electrode (cathode) to form superoxide , peroxide , or oxide (O2−) species that, together with the cation (Na+) from the pure Na metal applied as anode, may form a solid metal oxide phase in the cathode structure4,8. So far both NaO23,4,5,7,8,9,10,11 and Na2O212,13,14,15 have been experimentally reported as discharge product in Na/O2 batteries. It appears that the nature of the electrolyte and the water concentration in the cell might determine whether NaO2 or Na2O2 is formed6,11,16.
While the nature of the discharge product in Na/O2 batteries is discussed intensively6,11,16,17, little is known about the spatial distribution of discharge product in the pores of the cathode layers. A parallel for the product distribution might be drawn from latest research on Li/O2 batteries: Model-based18 and X-ray photon spectroscopy19,20 investigations have shown that the primary discharge product, lithium peroxide (Li2O2), is deposited near the battery separator first, while unwanted lithium carbonate species (Li2CO3), formed from impurities or in electrolyte decomposition reactions, are deposited near the oxygen reservoir (i.e. at the top of the cathode). In addition, toroidal Li2O2 particles were visualized by various methods, such as scanning X-ray transmission microscopy21 and high-resolution transmission electron microscopy22.
On the one hand, Li2O2 can limit battery performance at elevated current densities due to its poor electric conductivity, and on the other hand Li2CO3 can cause pore clogging in the battery cathode. Both phenomena, limited electric conductivity of discharge products deposited and pore clogging, might appear for Na/O2 battery operation as well and could significantly affect battery performance and cycling stability.
While it was indicated that cubic particles of discharge product form on top of the cathode, as shown with scanning electron microscope (SEM) analysis4,5, the morphology of discharge product deep inside the cathode layers and the impact of current density on deposition behavior is not clear up to now. Recent studies have shown that X-ray tomography is highly suitable to monitor the discharge product in batteries in operando23,24,25,26. Moreover, imaging methods are suitable to resolve phases contributing to electrochemical processes, and degradation inside battery electrodes with high spatial resolution27,28.
In this work, we apply synchrotron X-ray tomography on cathodes of Na/O2 batteries with the aim to elucidate the spatial distribution of the discharge product within the cathode. We extend these measurements by SEM analysis of the formed particles on top of the cathodes of fully discharged batteries: Several Na/O2 batteries were discharged at a wide range of current density (50 to 1,000 μA · cm−2) to visualize the impact of current density on discharge product deposition. We will discuss the distribution of NaO2 particles with regard to diffusivity and solubility of Na+ and the reactant O2 in the non-aqueous electrolyte applied. This analysis is dedicated to gain further understanding of the recently explored, reversible cell chemistry of Na/O2 batteries.
Impact of current density on particle morphology
Figure 1 depicts in-plane cross sections of the carbon cathodes discharged at the respective current densities viewing on top of the battery from the O2 reservoir. For each cathode shown, the discharge product NaO2 (white particles) is deposited around the fibers of the carbon cathode (gray color). Pore space, i.e. filled with gas, is indicated with black color. At 100 μA · cm−2 large (above 10 to 40 μm maximal edge length) particles are visible, whereas at 200 μA · cm−2 small (≈10 μm maximal edge length) particles are present, and at 300 μA · cm−2 even smaller (below 10 μm maximal edge length) particles are formed.
To extend the aforementioned observations of the particle size for a wide range of current densities, SEM analysis was performed on cathodes from batteries that were galvanostatically discharged at 50, 200, 400, 600 and 1,000 μA · cm−2, respectively, until a cut-off voltage of 1.8 V was reached. The respective discharge curves with the discharge capacity indicated are shown in the supplementary information in Fig. SI2. The SEM images obtained are shown in Fig. 2(a–e). We observe crystallized discharge product in form of cubic or pyramidal particles deposited on the carbon fibers for all cathodes analyzed. The particle size of all cubic particles decreases almost linearly from approximately 30 μm to 5 μm maximal edge length with increasing current density, which is illustrated in Fig. 2(f).
It is to be noted that the morphology of the discharge product changes at 600 μA · cm−2: The cubic shape obtained for low current densities (Fig. 2(a–c)) changes to cuboctahedral shape (Fig. 2(d)) and then to pyramidal shape (Fig. 2(e)) at highest current densities.
Discharge product distribution
Figure 3 depicts the stack of cathodes that was analyzed with synchrotron X-ray tomography in a projected through-plane view through the cathode centers. This view is used to visualize the distribution of mean X-ray beam attenuation in each cathode along its thickness with mean gray values in arbitrary units: Black colors represent attenuation due to a gas phase, gray colors represent attenuation due to the carbon fibers of the gas diffusion layer, and white colors represent high attenuating regions, i.e. the mean value of attenuation due to discharge product (NaO2) together with carbon fibers. Assuming a one-electron transfer reaction for the NaO2 formation, the local charge distribution in Na/O2 cells can be derived from this figure.
It can be seen that on top of all cathodes (facing the O2 reservoir), predominantly the discharge product NaO2 (white color, 25–28 a.u.) is present. At the bottom of the cathodes (facing the separator), less discharge product is found (lower gray values, 24–25 a.u.). The distribution of gray values along the gas diffusion layer thickness at 100 μA · cm−2 appears to be more homogenous than for 200 than for 300 μA · cm−2. The local minimum of gray values in the middle of each battery cathode can be explained as follows: Each cathode comprises of two 200 μm thick gas diffusion layers that are stacked together in the battery housing (compare methods section). Consequently, less discharge product was deposited at the interface of both gas diffusion layers due to increased contact resistance.
Particle size distribution
The high spatial resolution and contrast of synchrotron X-ray tomography enables to separate carbon fibers of the cathode and the reaction product NaO2 in the images obtained, and thus to analyze the particle size distribution (PSD) in selected regions. The respective representative results of the PSD analysis for 300 μA · cm−2 and 100 μA · cm−2 are visualized in Fig. 4(a,b), and in Fig. SI3 in the supporting information.
Small particles (particle volume below 1,000 μm3, green and blue color) are present in the entire cathode from top to bottom of the battery that was discharged at 300 μA · cm−2, whereas much larger particles (particle volume above 1,000 μm3, orange/red and purple color) have formed at 100 μA · cm−2. For 300 μA · cm−2, particles with largest volume are located in the top of the cathode, whereas less particles are located in the bottom of the cathode. Thus, a slight gradient of particle volume from top to bottom of the cathode can be observed. All in all, at the top of the cathode, facing the O2 reservoir, more and larger particles are situated. For 100 μA · cm−2, a more pronounced, steeper, gradient from top to bottom of the cathode from largest particles (≈3,000 μm3) to smaller volume particles (≈1,000 μm3) is visible. As for the cathode discharged at 300 μA · cm−2, more large particles are present at the top of the cathode.
Furthermore, the PSD analysis of the entire sample region of the cathode (area of 2.5 × 2.5 mm2, compare Fig. SI4 in the supporting information) reveals that the number of particles is ~350,000 at 300 μA · cm−2, which is much larger than for 100 μA · cm−2 yielding ~80,000 particles, although both possess the same discharge capacity.
The PSD analysis and the gray value distribution yield that the discharge product in Na/O2 batteries is not homogeneously spread deep down into the cathode although all cathodes investigated are from batteries with 1.5 mAh discharge capacity. In detail, a gradient of discharge product can be observed along the cathode dimensions: NaO2 particles are predominantly deposited at the side of the O2 reservoir. By implication, the local charge density distribution in Na/O2 cells is higher near the O2 reservoir. This might be due to the limited solubility and low diffusivity of O2 in the electrolyte (see Hartmann et al.8). This hinders the transport of O2 to carbon fibers near the separator, where O2 should react with Na+ to form NaO2. The ionic conductivity of Na+ in the electrolyte is not the limiting factor for the deposition of NaO2, since particles are also formed at elevated current densities of 300 μA · cm−2 on top of the cathode (compare Fig. 1), and for 400 to 1,000 μA · cm−2 inside the carbon fibers on top of the cathode (compare Fig. 2), which is furthest away from the source of Na+, i.e. the metal anode.
It appears that the space occupied by discharge product strongly depends on the current density applied: The cathode is homogeneously utilized at low current densities, whereas more of the discharge product might be located near the O2 reservoir for elevated current densities (compare gradient of gray values in Fig. 3). By implication, structured cathodes with a porosity gradient (higher porosity facing the O2 reservoir, lower porosity facing the separator) might be necessary for Na/O2 batteries to obtain a homogenous discharge product distribution also at high current densities.
We furthermore deduce that the particle size of the discharge product directly scales with the current density applied: at 100 μA · cm−2 discharge product of cubic shape (above 10 μm edge length, compare Figs 1 and 2) forms, which is in line with other ex situ SEM image results3,29 and first-principle studies30.
Considering that the discharge product NaO2 is an insulator and that its formation and decomposition during cell cycling was identified as a solution-based process31, another conclusion is derived in the following from the tomography results obtained: At elevated current densities, very small NaO2 particles form around the carbon fibers in a film-like manner (compare Fig. 1). Even if sufficient Na+ conductivity is ensured, electrons might not reach the reaction zone at the interface of solid and electrolyte due to poor electrical conductivity of the reaction product deposited. Thus, it seems that only capacities below 0.5 mAh can be discharged above 200 μA · cm−2 (compare Fig. SI2 and Hartmann et al.29). Due to the poor electric conductivity of NaO2 particles, electron transport to the electrode/electrolyte interface might be limited, and cell potential decays rapidly. By implication, small particles of discharge product can decrease the available battery capacity significantly by surface blocking. Whereas on the other hand, large particles might cause pore clogging, and thus cause O2 shortage, as indicated by the three-dimensional representation of the cathode discharged at 100 μA · cm−2 shown in Fig. 4(b). To overcome this issue, either pumping of electrolyte (electrolyte flow battery) or active supply of oxygen might be a valid option for Na/O2 battery operation. Moreover, fluorinated ether might be added to the liquid electrolyte, for example as found for Li/O2 batteries32, to enhance oxygen activity and thus to achieve a homogenous distribution of NaO2 particles in the cathode. In addition, the use of redox mediators (dissolved in the liquid electrolyte33,34) might help to achieve a more homogenous distribution of discharge product and to assist the decomposition of NaO2 particles during charge.
All in all, our analysis yields comprehensive insight into the spatial distribution of discharge product in the battery cathode of Na/O2 batteries and will help to understand limitations for the operation at elevated current densities that arise thereof, the cell design chosen, and the operation mode selected. It might be for example beneficial to consider new operation strategies for Na/O2 batteries to achieve a more uniform distribution of discharge product. The analysis presented will help to systematically identify further limitations for Na/O2 batteries, which in the end might help to improve their capacity and cycle life. Other research fields (e.g. the model-based analysis of cathode processes and cell concepts35,36) might benefit from the here presented findings.
Battery preparation and discharge
Electrochemical measurements were performed using the ‘Giessen cell’3,29, a cell based on a Swagelok design with two electrodes that is illustrated in Fig. 5(a). Two binder-free gas diffusion layers type H23 with 200 μm thickness and diameter of 12 mm each (Freudenberg, Weinheim, Germany) were used as cathode, and pure sodium metal (BASF SE, Ludwigshafen, Germany) was used as anode (12 mm diameter). One glass microfiber filter (GF/A, Whatman, 12 mm diameter) separated the electrodes. Electrolyte comprising of diglyme (anhydrous, 99.5% Sigma Aldrich) as solvent and 0.5 M sodium triflate (NaSO3CF3, 98%, Aldrich) as conducting salt was applied on gas diffusion layer and separator. The cell design comprises an oxygen reservoir, which was flushed with oxygen (purity 5.0, Praxair) for 10 seconds at 105 Pa just before the electrochemical measurement. All cells were assembled in an Ar-filled glove box (GST4, Glovebox Systemtechnik) with water and oxygen contents below 5 ppm.
Three cells of aforementioned composition were galvanostatically discharged at 100, 200 and 300 μA · cm−2 in a temperature chamber (Binder) at 298 K with a battery cycling system 4300 (Maccor) until a discharge capacity of 1.5 mAh was reached, to achieve comparable loading with discharge product. The corresponding discharge curves are shown in Fig. SI1 in the supporting information. After discharge, these batteries were disassembled in the Ar-filled glove box. 2.5 mm diameter samples of the respective cathodes (compare Fig. SI4 in the supporting information) were cut out and put all together in a PEEK sample holder, sealed with hot glue, and then analyzed with synchrotron X-ray tomography as visualized in Fig. 5(b). The stacking of the cathodes analyzed is illustrated in Fig. 3.
For SEM imaging analysis, batteries of aforementioned composition were discharged with the ‘Giessen cell’ setup with three electrodes with the aforementioned battery cycling system at 50, 200, 400, 600 and 1,000 μA · cm−2, respectively (see supplementary information, Fig. SI2) and their cathodes were analyzed ex situ with SEM, viewing on top of the cathode (i.e. from the O2 reservoir).
Synchrotron X-ray tomography imaging was performed at the synchrotron tomography station of the Helmholtz-Zentrum Berlin (BAMline at Bessy II). The respective setup is illustrated in Fig. 5(b). To ensure sufficient transmission through the samples, a monochromatic X-ray beam with an energy of 15 keV was chosen. A 4,008 × 2,672 pixel2 CCD camera (PCO 4000 with a CdWO4 scintillator screen) was used to capture local radiograms over 360 degrees. Stitching always two 180 degree separated local radiograms to one radiogram together resulted in a 7,900 × 7,900 × 2,672 pixel reconstruction with a voxel size of 0.4383 μm3. For the whole sample, two local measurements had to be performed and were stitched together with the software Fiji37.
A conventional X-ray tube was used for full view tomographic measurements of cathodes for the results in Fig. SI4 in the supporting information. The accelerating voltage was tuned to 50 kV while the current at the tungsten anode was adjusted to 200 μA. No filter was applied. 1,500 projections were taken for a full tomography (full range over 360 degrees). Each angle step was exposed for 2.2 s three times to increase the signal-to-noise-ratio. The projections were taken using a Hamamatsu flat panel detector with 2,316 × 2,316 pixel with a pixel size of 50 μm (6.25 μm pixel size in the reconstructed two-dimensional images).
SEM measurements were performed with a Merlin high-resolution Schottky field emission electron microscope (Zeiss SMT) equipped with an X-Max EDS detector (Oxford Instruments). All analyzed cathodes were washed prior each SEM measurement to remove excess liquid electrolyte.
Particle size distribution analysis algorithm
A sub stack that contained two slices of every of the cathodes analyzed was created in order to train a machine learning data-mining algorithm. For that purpose the trainable Weka Segmentation38 plug-in for Fiji37 including a fast random forest classifier was used. The classifier was trained by means of the sub stack. Subsequently, the segmentation of the full data set was performed by adapting the classifier to this data set. Finally the segmented data set was labeled using the software Mavi 1.5.139 and was analyzed with the software Avizo Fire 8.040. The three-dimensional visualization in Fig. 4 of the segmented data was created using VGStudio MAX 2.2.641.
How to cite this article: Schröder, D. et al. Visualizing Current-Dependent Morphology and Distribution of Discharge Products in Sodium-Oxygen Battery Cathodes. Sci. Rep. 6, 24288; doi: 10.1038/srep24288 (2016).
C.L.B. thanks the ‘Fonds der Chemischen Industrie’ (FCI) for a Ph.D. scholarship. The project was supported by the BASF International Scientific Network for Electrochemistry and Batteries.
This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/