Operando Monitoring of the Solution-Mediated Discharge and Charge Processes in a Na–O 2 Battery Using Liquid-Electrochemical Transmission Electron Microscopy

HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Operando Monitoring of the Solution-Mediated Discharge and Charge Processes in a Na–O 2 Battery Using Liquid-Electrochemical Transmission Electron Microscopy Lukas Lutz, Walid Dachraoui, Arnaud Demortière, Lee Johnson, Peter Bruce, Alexis Grimaud, Jean-Marie Tarascon


Introduction 50
Compared to Li/Na-ion batteries, in which reversible energy storage relies on the use of redox 51 active transition metal oxides as positive electrodes, the metal-O 2 battery systems would 52 theoretically offer greater energy density owing to the use the redox of gaseous oxygen using 53 conductive and light carbon electrodes. 1, 2 The aprotic lithium-oxygen (Li-O 2 ) system has 54 been widely studied since the early demonstration of reversibility by K.M. Abraham. 3 55 Nevertheless, recent developments clearly pointed out towards drastic limitations in terms of 56 round trip efficiency as well as coulombic efficiency due to copious parasitic reactions of the 57 discharge product lithium peroxide (Li 2 O 2 ) with both the conductive electrode and the 58 electrolyte. 4-6 Following this conclusion, the sodium-oxygen (Na-O 2 ) system was then 59 proposed as a viable alternative due to its theoretical energy density of 1100 Wh/kg combined 60 with a better round trip efficiency and presumably limited parasitic reactions. 7 The Na-O 2 61 system is still, however, in its infancy, owing to several unresolved challenges, such as 62 limited capacities and low cyclability. 8,9 Hence, the initial excitement was quickly 63 counterpoised by the recent discoveries highlighting the unstable nature of the superoxide 64 discharge product sodium superoxide (NaO 2 ) that reacts with glyme-ethers solvent commonly 65 used in these systems. Despite these evident limitations, this system has been seen as an 66 interesting case study to better understand the complex redox reaction of oxygen in aprotic 67 solvent that involves a gas to solid phase transformation. Only mastering these complex 68 transformations will eventually trigger the development of rechargeable metal-O 2 batteries 69 and deliver the initial promises offered by the large energy density for these systems. 70 Further efforts are thus required to understand and master the formation and decomposition 71 processes of the micron-sized cubic NaO 2 product, which is at the core of the Na-O 2 72 electrochemistry and still under heavy debate. Contradictory results discussing either a 73 solution-mediated discharge and charge reaction, the need for phase transfer catalysts (e.g.: 74 cubes. Further, we visualize the formation of side products leading to the formation of 100 parasitic shell at the interface between NaO 2 crystals and the electrolyte, which remains as 101 solid residues on the electrode after charge. 102 Benchmarking the Na-O 2 micro-battery setup 103 Figure 1 shows a schematic of the micro-battery based on the electrochemical TEM cell 104 configuration (a-d) used throughout this work for the operando imaging of sub-micrometric 105 features during redox reactions at the positive carbon electrode (e). The operando cell was 106 assembled using an oxygen-saturated electrolyte made of 0.5M NaPF 6 dissolved into 107 monoglyme (DME), which contains < 20 ppm of water as determined by Karl-Fischer 108 titration. To establish its electrochemical performance, the operando cell was charged and 109 discharged in a cyclic-voltammetry mode, using a sweep rate of 10 mV/s between and Pt as 110 counter and pseudo-reference electrodes (Supplementary Figure S1). Such conditions were 111 used due to the extremely small size of the cell setup, restricting the volume of the electrolyte 112 as well as the available amount of dissolved O 2 . We first verified that these conditions provide 113 similar results as classical Swagelok cells, with namely the formation of discharge products 114 consisting of plentiful cubes (Figure 2a and 2b), which were identified as NaO 2 by combining 115 energy dispersive X-ray spectroscopy (EDX) and selected area electron diffraction (SAED) as 116 discussed later in greater detail. The growth of NaO 2 cubes during discharge was followed by means of fast TEM imagining 152 and high angle annular dark field STEM (HAADF-STEM) using the same cycling conditions 153 as previously mentioned ( Figure 3, sequence a and e, Supplementary Video S1). Comparing 154 the electrochemical response in Figure 3 c with the image sequence in Figure 3 a shows that 155 the cube growth follows a solution-precipitation mechanism. Indeed, after an initial step 156 where the electrolyte is saturated by the electrochemically produced NaO 2 soluble species 157 (image at 5 s, cathodic current in CV of Figure S1), a point of super saturation is then reached 158 as characterized by the formation of small NaO 2 nuclei on the electrode surface (image at 10 159 s). This initial incubation period, where cathodic current corresponding to the electrochemical 160 formation of soluble NaO 2 is measured but no product is formed on the electrode, is 161 characteristic of a crystal growth following a solution-precipitation mechanism and therefore 162 rules out a surface-directed growth of NaO 2 cubes, for which NaO 2 would grow as a solid 163 following the cathodic current. Such nuclei subsequently grow in an isotropic manner (images 164 at 15 -60 s), by deposition of solvated NaO 2 on the surface of cubes. This growth ultimately 165 leads to the formation of NaO 2 cubes with a size of approx. 500 nm (image at 60 s). Hence, 166 three stages for the solution-mediated cube-growth precipitation process, similar to the first 167 description given by Janek and coworkers, 15 could be spotted. First, soluble NaO 2 is 168 electrochemically formed and quickly saturate the electrolyte (owing from the low solubility 169 of NaO 2 in organic solvents) 15,34 . Once supersaturation of the electrolyte is reached, small 170 aggregates of solvated (NaO 2 ) n species precipitate in the form of small NaO 2 nuclei on the 171 carbon electrode. Finally, upon discharge, soluble NaO 2 species are consistently produced and 172 deposit on the high surface energy nuclei, which ultimately grow into larger NaO 2 cubes. We 173 would like to emphasize here is that the electrode surface in Figure 3 sequence a and e, is hard 174 to visualize owing to the thick layer of electrolyte between the electrode and the Si 3 N 4 175 window. Hence, to facilitate its identification, a thin white line is used, as a guide to the 176 reader, to indicate the electrode border in TEM image in Figure 3 b. 177 178 Further exploiting the capability of the TEM setup, we visualize the sequential size evolution 179 of several cubes (Figure 3 b, supplementary Video S2). As NaO 2 is an insulator 15, 16 and 180 cannot grow by electrodeposition, it is evident that the gradual growth occurs by deposition of 181 NaO 2 from the solution at the outer crystal surface. For better quantification, the particle size 182 evolution during discharge as a function of the growth time was plotted, which revealed the 183 non-linear intermittent growth rate (Figure 3 c). Initially, the electrolyte is being saturated 184 with electrochemically generated NaO 2 and no significant deposit can be observed. Once the 185 saturation limit is reached, a rapid increase of the cube size is observed, which could be 186 associated with the large concentration of NaO 2(solv) in solution at the point of super 187 saturation. This initial burst is then followed by step-wise regime associated with domains of 188 low and high growth rates, dependent on the local concentration of NaO 2 in solution. Lastly, a 189 steady-state regime is reached towards the end of discharge where the growth rate diminishes parameters cannot be precisely measured from SAED patterns, their ratio can be estimated 246 with much higher precision. This estimate gives the a nano-cubes /a NaO2  1.015 value. Both, the 247 fluorite-type NaO 2 and antifluorite-type Na 2 O both possess the face-centered cubic unit cell 248 with the cell parameter ratio a Na2O /a NaO2 = 5.56Å/5.512Å = 1.009 that is reminiscent to the 249 experimentally measured ratio. Thus, one can tentatively identify the nano-cubes as defective 250 NaO 2 , with an increased Na:O atomic ratio (note that Na 2 O 2 would adopt an hexagonal 251 symmetry while Na 2 O 2 .2H 2 O would adopt a monoclinic symmetry). These high surface area 252 cubes may further favor the chemical reactivity towards electrolyte decomposition as seen by 253 the formation of the third shell, an amorphous layer at the outer surface, i.e. at the interface 254 between the cubes and the electrolyte, with reduced Na content as deduced by the small 255 sodium peak observed by EDX analysis (Figure 5 d,  To gain deeper understanding about this shell, XPS spectra of discharged GDL electrodes 273 were collected at various stages of discharge (Supplementary Figure S2). The C1s spectra 274 reveal the constant evolution of a parasitic carbonate-like species during discharge. To 275 quantify the amount of these carbonates generated upon discharge, we relied on the method 276 first described by Thotiyl et al. 35 that consist in the use of acid (H 3 PO 4 ) and Fentons`s reagent 277 are used to decompose inorganic and organic carbonates, with the CO 2 generated through 278 their decomposition being subsequently sampled by a mass spectrometer. 279 The released CO 2 concentration at various stages of discharge ( Figure 6) indirectly, 280 demonstrates the significant amount of inorganic Na 2 CO 3 and organic carboxylates formed on 281 the surface of carbon electrodes. Upon discharge, the concentration of inorganic carbonates 282 significantly increases (Figure 6 b), and this is in agreement with the growth of the shell 283 observed in Figure 4a. Moreover, when comparing the concentration of Na 2 CO 3 at the end of 284 discharge with the results obtained at the end of charge, a limited increase is found for GDL 285 electrodes, demonstrating that the parasitic products cannot be reoxidized and remain on the 286 electrode surface at the end of charge (Figure 6b). This result highlights the importance of 287 mastering this interface for decreasing the rate of parasitic product formation. 288 289 290 Figure 6: Discharge−charge profiles for GDL electrode at a rate of 25 A/cm 2 (a). Amount 291 of CO 2 evolved from the GDL electrode when removed from the cells at different states of 292 discharge and charge and treated with acid and Fenton's reagent to decompose Na 2 CO 3 and 293 organic carboxylates (b). CO 2 evolution originating from the instability of electrode and 294 electrolyte as deduced from the discharge of 13 C-carbon electrodes at various discharge rates 295 between 25 -250 uA/cm 2 (c). 12 CO 2 evolution indicates the electrolyte degradation leading to 296 inorganic carbonates (black) and organic carboxylates (red) whereas 13 CO 2 detection results 297 from the direct decomposition of the carbon electrodes. 298 299 Finally, to clarify the origin of the carbonate side product formation, which can result from 300 the electrolyte and/or from the electrode decomposition, discharge experiments using 13 C-301 labeled electrodes with the released CO 2 isotopes being detected by mass Spectrometry 302 analysis. These isotopic experiments revealed the presence of both 13 CO 2 and 12 CO 2 at the end 303 of discharge, which can only be explained by the decomposition of both the electrode surface 304 and the electrolyte, respectively (Figure 6c). However, the observed 12 C fraction was much 305 larger than the 13 C one, demonstrating that the majority of parasitic carbonates originates from 306 a) b) c) Na 2 CO 3 Carboxylates 12 C Carboxylates 12 C Na 2 CO 3 13 C Na 2 CO 3 the instability of the glyme-electrolyte in contact with the highly oxidizing NaO 2 discharge 307 product. Again, this result corroborates the shell formation we observed in Figure 4. 308 Additionally, we observed that upon elevated discharge currents, the amount of products 309 originating from the decomposition reactions increases, with a prominent contribution from 310 the electrode decomposition (Figure 6c), hence implying that an electrochemically-driven-311 electrode decomposition is also at play during discharge of Na-O 2 batteries. 312 Altogether, these experiments reveal the high reactivity of NaO 2 and further disproves, 313 together with previous literature reports, 8,[23][24][25]  carboxylate were identified to be the main parasitic products, with a variety of other side 325 products, such as formats and acetates also being found. 326 327 Overall, we believe that preventing the formation of this complex organic/inorganic shell will 328 be of prime importance to mitigate the drastic capacity loss observed upon cycling with 329 todays' Na-O 2 cells, which will be discussed in greater details below. 330

Resolving the mechanism in charge -the dissolution of NaO 2 331
Encouraged by the mechanistic insight provided by operando TEM during the discharge of 332 Na-O 2 battery, we decided to explore the oxidation process following the same methodology. 333 From a sequence of images collected by HAADF-STEM (Supplementary Video S4), the 334 gradual dissolution of NaO 2 cubes during oxidation can be observed (Figure 7). More 335 importantly, this visualization shows that cubes dissolve concentrically from the outside 336 inwards. In detail, the 3D visualization of the processes at play during charge illustrate that 337 the dissolution of the cubes initially proceeds from the top surface, i.e. at the interface 338 between the cube and the electrolyte, leading to a steady decrease in size of the cube (Figure 7  339 a-d). This is in contrast to the previously reported electrode directed charge-transfer, i.e. the 340 direct oxidation of the cubes at the interface with the electrode. 12 To gain deeper insight into 341 this dissolution process, the height-profile evolution was followed for one cube throughout 342 the complete oxidation (Figure 7 e-h). From this profile, it can be observed that cubes, despite 343 being covered by the parasitic shell, start to dissolve from the top, i.e. the face exposed to the 344 electrolyte, hence demonstrating the porous nature of the organic shell. Upon further 345 charging, the overall height profile continuously decreases from the top of the cube, further 346 suggesting a collapse of the parasitic shell during charge. Ultimately, parts of the shell remain 347 at the end of charge spread on the surface of the electrode, visualized in Figure 7 i as a 348 patchwork and by the "walls" at 50 and 300 nm in the linear profile in Figure 7 h. This is in 349 good agreement with our ex situ observations where parasitic residues (organic and inorganic) 350 are found on the GDL carbon fibers after charge (Figure 6), as well as with previous reports. 8 351 Altogether, these new information shine light on a so-far poorly explained phenomenon, 352 namely the constant columbic losses measured upon cycling. Subsequent cycling will indeed 353 generate additional parasitic products that will accumulate at the electrode surface, ultimately 354 causing a rapid capacity loss and a drastic self-discharge. 355

356
In light of the solvation-desolvation equilibrium discussed above, our operando electron 357 microscopy measurements provide the definitive demonstration that the oxidation process in 358 Na-O 2 batteries follows a solution-mediated mechanism, as previously proposed 15 based on 359 the significant solubility of NaO 2 , 15, 34 its low dissolution energy 10, 13 as well as its insulating 360 nature that would prohibit direct oxidation at the electrode. 15, 16 During charge, solvated 361 NaO 2(solv) is oxidized at the electrode surface into Na + and O 2(g) , hence displacing the 362 equilibrium NaO 2(solid) = NaO 2(solv) to the right and forcing the dissolution of the cubes. 363 Through this process, the bottom edges of the non-conducting cubes, in direct contact with the 364 electrode surface, remain throughout the charge as evidenced by TEM. This clearly contrasts 365 with the previously proposed mechanism for which a direct charge transfer between the solid 366 and the electrode/current collector was a requirement. 12 367 368 Finally, we explored the consequences of the formation of parasitic products at the electrode 369 surface on subsequent cycles and NaO 2 formation. This revealed that NaO 2 nuclei were 370 exclusively formed during the second discharge on the uncovered, pristine electrode surface 371 (Supplementary Video S5). Hence, the parasitic products not only hamper the O 2(g) redox 372 reaction but also hinder NaO 2 nucleation on the carbon surface. In short, this study shows that 373 the formation of parasitic products has its origin in the high chemical reactivity of the NaO 2 374 cube surface, initially suspected to be less reactive than Li 2 O 2. Herein we have reported that fast imagining TEM and HAADF-STEM are powerful 389 analytical tools to understand the mechanistic pertaining to the charge/discharge processes in 390 DME based Na-O 2 batteries. We visualized the solution-mediated growth of NaO 2 in real-391 time and identified that the 3D growth process is governed by the equilibrium between 392 NaO 2(solv) <--> NaO 2(solid) and the mass transport of soluble product. By imaging the charge 393 process, we provide conclusive evidence that the same solvation-desolvation equilibrium is 394 responsible for the dissolution of the NaO 2 discharge product, which consumes the NaO 2 395 cubes from the NaO 2 -electrolyte interface towards the electrode and not from the cube-396 electrode interface. Therefore, we rule out the direct charge-transfer reaction as the major 397 oxidation path for NaO 2 cubes and clarify the mechanism of this widely discussed reaction. 398 Finally, we provide fundamental insights into the parasitic reactions occurring during cycling 399 of a Na-O 2 battery where time-resolved visualization revealed the chemical reactivity of NaO 2 400 at the interface with the electrolyte. As a result, parasitic products continuously accumulate on 401 the cube surface to form a thick shell surrounding the NaO 2 cubes, which passivates the 402 electrode surface as it cannot be reoxidized. This information is vital for optimization of the 403 battery, since this parasitic shell is responsible for the low efficiency during charge, as well as 404 for its poor cyclability by preventing crucial O 2 redox and further nucleation of NaO 2 . It must 405 therefore be recognized that the NaO 2 growth is solvent dependent, providing the possibility 406 of mediating the deposition process by controlling the solvation/desolvation event. Hence, 407 caution must be exercised prior to generalizing this finding. Through this first visualization of 408 the redox processes governing the Na-O 2 system, we further confirm the importance of 409 finding how the various components of the batteries locally interact with each other. We hope 410 these results will help in the development of new strategies to optimize cell components, such 411 as the electrolyte, in order to achieve high performing Na-O 2 batteries, and also serve to 412 motivate the development of operando electrochemical TEM cells. 413 414

Methods 415
Electrolyte preparation: 416 1,2-Dimethoxyethane (DME, 99.9%) was purchased from Sigma Aldrich and (NaPF 6 99.9 %) 417 was bought from Stella Chemifa. Solvents were dried by means of molecular sieves for 5 days 418 to remove excess water and Sodium salts were dried under vacuum at 80°C for 24 hours. The 419 0.5 M electrolyte solutions were prepared in an argon-filled glove box (0.1 ppm O 2 /0.1 ppm 420 H 2 O). The water content of the electrolyte solutions was analyzed by Karl Fischer titration 421 and was found to be below 20 ppm. The electrolyte was saturated with ultrapure O 2 , prior to 422 use in the in situ TEM cell. 423 424

Operando electrochemical (S)TEM experiments: 425
Operando TEM experiments were performed using a FEI-TECNAI G2 (S)TEM equipped 426 with a Schottky field-emission gun and an fast camera Oneview-Gatan (30 fps at 4k). For 427 these experiments the microscope was operated at 200 kV in both conventional TEM and 428 HAADF-STEM modes. In this study, we checked the effect of the electron beam used to 429 make the observations in TEM and STEM modes to be sure that the beam does not have any 430 effect on our results. During the observations, the dose was kept below 10 e -/nm 2 s in order to 431 limit beam damage effects. By this way, typical beam effects (bubble and precipitate 432 formations) due to the degradation of the electrolyte by radiolysis effect are avoided. As 433 shown in Supplementary Figure S3, the insignificant impact of the electron beam on the 434 liquid electrolyte was verified with the same dose of electron used during the fast imaging 435 acquisition, which shows a high stability of NaPF 6 /DME/O 2 electrolyte upon electron beam 436 irradiation for a relatively long period of time: 360s. The TEM holder used is a Protochips 437 Poseidon 510 owing both a microfluidic flow system and an electrochemical measurement 438 system with 3 electrodes. The micro-battery cell itself is localized in the holder tip and 439 consists of two silicon Echips sealed by Viton O-ring gasket: a top Echip (with 2 Pt electrodes 440 (reference and counter) and 1 glassy carbon electrodes (working), a 500 nm SU-8 polymer 441 spacer and a 50 nm thick Si 3 N 4 window) and bottom Echip (with a 500 nm spacer and a 50 442 nm thick Si 3 N 4 window). Mounted Echips are then compressed onto O-rings using screwed 443 lid of the holder inducing a good vacuum-sealing. The microfluidic system integrated in the 444 TEM holder allows to introduce and flow the electrolyte with a rate range from 0.5 to 5 445 µL/min. using a syringe pump system. The microfluidic system (cell and microtubes) is 446 flushed by argon gas to discard oxygen presence prior to start operando experiment. 447 448 Cyclic voltammetry 449