Mechanism of gelation in high nickel content cathode slurries for sodium-ion batteries

Sodium-ion batteries are a prospective sustainable alternative to the ubiquitous lithium-ion batteries due to the abundancy of sodium, and their cobalt free cathodes. The high nickel O3-type oxides show promis-ing energy densities, however, a time dependency in the rheological properties of the composite elec- trode slurries is observed, which leads to inhomogeneous coatings being produced. A combination of electron microscopy and infra-red spectroscopy were used to monitor the O3-oxide surface changes upon exposure to air, and the effect upon the rheology and stability of the inks was investigated. Upon exposure to air, NaOH rather than Na 2 CO 3 was observed on the surfaces of the powder through FTIR and EDS. The subsequent gelation of the slurry was initiated by the NaOH and dehydroﬂuorination with crosslinking of PVDF was observed through the reaction product, NaF. Approximately 15% of the CF bonds in PVDF undergo this dehydroﬂuorination to form NaF. As observed in the relaxation time of ﬁtted rheological data, the gelation undergoes a three-stage process: a dehydroﬂuorination stage, creating saturated struc- tures, a crosslinking stage, resulting in the highest rate of gelation, and a ﬁnal crosslinking stage. This work shows the mechanism for instability of high nickel containing powders and electrode slurries, and presents a new time dependent oscillatory rheology test that can be used to determine the process window for these unstable slurry systems. (cid:1) 2022 The Authors. Published by Elsevier Inc. ThisisanopenaccessarticleundertheCCBYlicense(http:// creativecommons.org/licenses/by/4.0/).


Active material synthesis and characterisation
To investigate the gelation effects upon the electrode formulation, 100 g batches of material was synthesised to ensure a consistency in the testing. NaNi 1/2 Mn 1/4 Sn 1/8 Ti 1/8 O 2 (Na-NMST) was synthesised by a solid-state method. Stoichiometric quantities (to produce 100 g of product) of Na 2 CO 3 , Ni(NO 3 ) 2 Á6H 2 O, MnCO 3 , SnO 2 , and TiO 2 , (Alfa Aesar, !99.5% purity) were ball milled using a Glen Creston roller mill in a 1000 mL HDPE Azlon roller pot. Due to the volatility of sodium during the solid-state synthesis, an extra 5 wt% of Na 2 CO 3 was added. A charge of 375 g of 5 mm and 375 g of 10 mm zirconia balls were used for milling. 250 mL of isopropanol (IPA) was initially added, and the pot was milled for more than 48 h to ensure homogeneity. The pressure inside the pot was released every $30-60 min and, over the course of milling, an extra 200 mL of IPA was added to account for its volatility.
The IPA was removed and condensed under vacuum at approximately 70°C and the precursor was fired at 900°C for 24 h. The resulting as-synthesised Na-NMST was transferred hot into a glove box filled with argon to prevent any reaction with air.
Scanning electron microscope (SEM) images of the powder were taken using a Zeiss SUPRA 55-VP with elemental analysis performed using an Oxford Instruments energy-dispersive X-ray (EDS) spectrometer attached to the SEM. SEM images and qualitative elemental analysis were obtained for the as-synthesised material and material placed on SEM stubs that had been exposed to lab air for ca. 72 h.

Electrode slurry formation and characterisation
The active material was milled in a Retsch PM100 Planetary Ball Mill with one third of the total being conductive carbon black (IMERYS, Super C65) to coat the particles and aid in the conductive properties of the electrode. The powders were milled at 300 rpm for 45 min, with a reversal of direction every 15 min. The zirconia balls used had diameters of 5 and 10 mm, with a loading of 45 g and 15 g for the 5-and 10-mm balls, respectively. Electrode slurries were formed by mixing the carbon coated Na-NMST (NaNi 1/2 Mn 1/4 Sn 1/8 Ti 1/8 O 2 ) with the remaining conductive carbon black, and an 8 wt% PVDF (polyvinylidene fluoride) (Solvey 5130) binder in NMP (n-methyl-2-pyrrolidone) solution in a ratio of 89:6:5 by weight. The cathode ink was mixed in sealed pots in a Thinky ARE-250 planetary mixer, situated in a Munters dry room with a maximum dew point of À40°C. Materials were measured out under a recirculator hood (also situated in a dry room) or in an argon filled glove box. There were three stages of mixing: initially the active material and the extra NMP, measured to give a final solids content of 40 wt%, followed by the addition of carbon black, and finally, addition of the PVDF solution. Each mixing stage lasted 5 min at a revolution speed of 1300 rpm.
The electrode slurry was chemically characterised using a Bruker Vertex 70v Fourier Transform Infrared spectrometer (FTIR) with a diamond attenuated total reflectance accessory. Data were recorded from 4500 to 100 cm À1 , and all measurements were made using 16 scans at 4 cm À1 resolution, with appropriate background measurements taken for each spectrum. Measurements were taken every 10 min whilst the slurry was situated on the diamond and exposed to air. The slurry samples were transported to the FTIR spectrometer in sealed vessels following their mixing in the dry room.
Physical measurements of the slurry rheology were obtained using a Malvern Panalytical Pro + rheometer with 40 mm diameter roughened parallel plates and a 1 mm gap. Frequency sweeps were performed, from 0.1 Hz to 100 Hz at two different amplitudes, 0.02% and 0.1%, both of which reside in the linear viscoelastic region (LVER). A constant pre-shear of 10 s À1 for 30 s was performed at the start of measurements to eliminate any time dependence in the samples and ensure consistency between results. The frequency sweeps were performed every 5 min on the electrode slurry, between roughened parallel plates. Three repeats were performed.

Electrode formation and characterisation
Electrodes were formed by coating the electrode slurry onto an aluminium substrate using a doctor blade set at 150 lm, aiming for a target loading of 80 gsm. They were then dried on a hotplate at 80°C for 40 min, followed by overnight in a vacuum oven at 120°C. For additional characterisation of the slurry exposure, three different coatings were made from the electrode slurries, one made from the just-mixed slurry (i.e., immediately after mixing), one was coated 20 min after mixing and the final one 40 min after mixing (at which point the slurry gelation was noticeable). All coating was performed inside a recirculator hood situated in a dry room.
X-ray photoelectron spectroscopy (XPS) data on the asmanufactured electrodes was acquired using a Kratos Axis Ultra DLD spectrometer. Samples were transferred to the spectrometer using an airless transfer unit with an argon atmosphere. The surfaces of the samples were found to charge slightly under the Xray beam during the experiments and to compensate for this, the surfaces were flooded with a beam of low-energy electrons during data acquisition. This in turn required subsequent re-referencing of the data, using the graphitic component of the C1s region at 284.3 eV as the reference point. The data were analysed with the CasaXPS software package, using Shirley backgrounds and mixed Gaussian-Lorentzian (Voigt) line shapes and asymmetry parameters where appropriate.
Cross-section Scanning electron microscopy (SEM) was performed on a Philips XL30 FEG instrument within a holder where the cross-section of the sample facing up, the images are taken under an acceleration voltage of 10 kV. The elemental distributions were measured by energy-dispersive X-ray spectroscopy (EDX, Oxford Inca 300).
Half-cells were assembled in an argon filled glovebox, using sodium metal as the counter (and reference) electrodes, Celgard 2325 and glass fibre (GF/A, Whatman) discs, both cut to 16 mm as dual separators, and the cathode electrodes cut to 14.8 mm.
Electrodes had an average thickness of 50.3 lm (±4.3 lm) and an average loading of 77.2 gsm (±7.95 %). They were not calendared, and the porosity of the electrodes was estimated to be 15 % (±2.5 %). The GF/A separator was placed next to the electrode, and the Celgard 2325 next to the metallic sodium to reduce dendrite growth. 70 lL of 1 M NaPF 6 in EC:DEC (1:1 V/V) with FEC additive (5.3 wt% FEC) (Flurochem) was used as the electrolyte. Cells were made into CR2032 coin cells. Cells were subjected to an initial 10 mA/g formation cycle, and 15 mA/g for all subsequent cycles. EIS measurements were performed at 2 V, scanning between 10 kHz and 100 mHz. A repeat of at least three cells were made for each test.

Results and discussion
To understand the surface and chemical evolution of the active material in air the active material powder was examined by measuring a FTIR spectra every 30 min for 360 min in air, as show in Fig. 1a. The very broad peak at the high frequencies, between $2500 cm À1 and $3500 cm À1 , is indicative of H-bonds, from OH stretching in both hydroxyl groups and absorbed surface water molecules [42]. At $ 2350 cm À1 an asymmetrical double peak is observed, typical of CO 2 [43]. Interestingly, the intensity of this peak does not change over time, suggesting that the absorbed CO 2 is at saturation on the surface. This absorbed CO 2 can react to form carbonates with subsequent reabsorption of. The peaks between 1423 cm À1 and 669 cm À1 relate to carbonates and hydroxides. Carbonates show a stretching at $1425 cm À1 , typically asymmetrical [44] and hydroxides can show an OH bending at around 1420 cm À1 [45]. A comparison of these peaks with pure carbonate (Na 2 CO 3 [46] and NiCO 3 [47]) and hydroxide compounds (NaOH [47]) is shown in Fig. 1b. The three highlighted regions in Fig. 1b demonstrate the shared peaks, with the bands at $1423 cm À1 and $879 cm À1 being the most observable. It is clear that NiCO 3 is not being formed as there is no overlap with the peak at 820 cm À1 , despite sharing the peak at $1434 cm À1 . From the broad region of OH stretching between $2500 cm À1 and $3500 cm À1 , as well as peaks at 1422 cm À1 , 879 cm À1 and 701 cm À1 , it is likely that NaOH and Na 2 CO 3 is present (Fig. 1b). The spectrum for NaOH shares the very broad region of OH stretching between $2500 cm À1 and $3500 cm À1 , as well as peaks at 1422 cm À1 , 879 cm À1 , and 701 cm À1 . Na 2 CO 3 similarly shares peaks at 1429 cm À1 and at 878 cm À1 .
SEM and EDX ( Fig. 2) of the powder shows platelet-like particles existing both singly and as agglomerates. Primary particles are approximately 750 nm Â 350 nm Â 150 nm in size (as seen in  Fig. 2b. Approximately 500 mg of the active material powder was mounted onto a SEM stub and exposed to the air in an open vial for 3 days, after which acicular particles are observed emanating from the particle agglomerates (Fig. 2b).
From EDS it is observed that the acicular particles contain sodium and oxygen in a 1:1 ratio, with lower levels of the transition metals and carbon ( Fig. 2d and e). This indicates that these acicular particles are NaOH, supporting the results of the FTIR analysis.
When the NMST is processed into a slurry, initially a pourable ink is formed, however over time this transforms to a jelly like material as shown in Fig. 3. Initially, just after the electrode slurry has been mixed, it can flow very easily and freely (Fig. 3a). Once the electrode slurry has reached a gel-like state ( Fig. 3b) it becomes impossible to coat homogeneously and reliably. A physically stabilised gel can be broken up with increased shear or additional solvent, however if a chemical gel forms this is typically irreversible. One challenge with these time dependent inks is to determine the point at which the ink or slurry becomes uncoatable, often this is determined qualitatively in the laboratory through observation, and in manufacturing through defect observations in the coatings and adhesion properties after or during coating, this leads to potential waste and increased cost [48][49][50][51]. For simple slurry cast coatings there is a wide viscosity process window, however with increasing viscosity a change in the viscoelastic properties is also observed. There are several challenges in the measurement of these changes particularly with high solids fraction electrode slurries (>40 wt%) where there is an increase the risk of colloidal jamming.
Computational calculations and experimental measurements have shown that this can occur at a solids volume fraction of 54.5% [52].
In this work, gelation has a significant effect upon rheological properties, and the properties are elucidated with time to monitor the onset of different physical and chemical changes in the slurry. Oscillatory rheology was performed to determine the nature of the gelation and to analyse the degree of gelation of the electrode material. Fig. 4a shows the storage, G 0 and loss G 00 , moduli for the aging electrode slurry. It can be seen that both G 0 and G 00 increase over the course of the aging measurements, with a larger increase observed for G 0 , consistent with the notion of a gelling slurry.
G 0 and G 00 both demonstrate a convergence of measurements obtained at different frequencies towards the end of the rheological measurements. The convergence is a strong indication of a gel point, GP [53]. As this point of convergence occurs where the viscoelasticity no longer depends on observation time, it is clear that this is not the formation of a physical gel (such as a colloidal glass) but rather a reaction is occurring to form a chemical gel. Rheological indicators of a gel, include the convergence of G 0 and G 00 in time at different frequencies and the crossover point of G 0 and G 00 , which occurs almost immediately [54]. It should be noted that '0 min' refers to the start of the rheological measurements not the point at which the slurry is mixed (and it does not take into account the transfer time of less than 5 min), it therefore seems logical that there is gelation occurring during the mixing of the electrode slurry. The changes in the phase angle, Fig. 4b, illustrate the balance between the viscous and elastic forces. Initially, the phase angle is at $40°with slight differences depending on the frequency (or observation time). This indicates a system with a slight dominance towards the storage modulus and the elastic behaviour. This phase angle decreases as the electrode slurry aging progresses, until 65-70 min into the experiment where the convergence occurs, and the phase angle reaches $1-2°. At this point the gel is realised, and elastic behaviour is almost fully dominant. By the end of the rheological measurements, the physical manifestation of the gel had reached a point where it was completely uncoatable, with properties as shown in the photograph in Fig. 4b.  To determine the coatability of the gel, the data was fitted to a two mode Maxwell model using the Reptate software package [55]. Using a Maxwell Model [55,56] the oscillatory data can be fitted with single or multiple modes, each comprised of an elastic, Hookean spring element (E) and a viscous, Newtonian Fluid, dashpot element (g) in series. Multiple modes can be used in parallel to increase the parameters of fitting and improve the quality of fit. In this work one mode per decade of data is used, as shown in Fig. 5 [55].
Relaxation time, s, is the ratio between the viscous element and the elastic element, as shown in Eq. (1). This is a theoretical property where times much shorter than the relaxation time will behave as a solid (spring), and times longer than the relaxation time will behave as a fluid (dashpot). As stress is equal across a mode and strain is the summed, a rearrangement and differentiation of Hooke's Law and a rearrangement of the Newtonian law of viscosity allows the following equations (Eqs. (2) and (3)) to be written, where a denotes the mode and x is the angular frequency of the oscillation.
From Fig. 6, it is observed that the primary relaxation time slowly increases for the first 30 min. During this time, the secondary relaxation time is relatively unchanged and upon deformation the electrode slurry will still revert to viscous behaviour relatively quickly. Then, the primary relaxation time rapidly increases indicating that the gelation is rapidly progressing. After    45 min, which is suggested as the maximum processing window of this electrode slurry, the rate of primary relaxation time increase drops. The gel structure is formed, with a macromolecule of connected chains throughout the electrode slurry. After this point, the interconnections between chains will strengthen slowly and the gel strength will increase. In terms of the secondary relaxation time, values are much lower, and only change slowly over the period of gelation. The secondary relaxation times denotes the behaviour of the electrode slurry over very short times scales. Below approximately 0.2 s, the secondary relaxation time signifies that the electrode slurries exhibit a strong elastic response, similar to that of a yield stress in a Herschel-Bulkley fluid. Above this secondary relaxation time, the strong elasticity of the electrode slurry partially relaxes, and the primary Maxwell mode elastic response dominates. It is only above the primary relaxation time that the electrode slurry will act fully viscous and flow.
To elucidate the chemical changes which occur during the slurry mixing and ageing, ex-situ FTIR analysis was performed. The main groups relating to water, hydroxide, NMP, PVDF and carbonates are discussed, as shown in Fig. 7. Separate spectra of the individual compounds for comparison can be found in the supplementary information - Fig. S1.
Like that observed on the aged powder, a broad peak at 3442 cm À1 , indicative of an OH stretching vibration, is observed. This can be contributed to surface absorbed water in the electrode slurry and from hydrogen bonding throughout the slurry. The increase in the intensity of this peak over time demonstrates that more water is being absorbed as the aging progresses and the amount of hydrogen bonding is similarly increasing. During the course of the ex-situ measurements, the OH stretching peak also experiences a broadening, which is consistent with the increased effects of hydrogen bonding. The very weak shoulder peak at $3300 cm À1 , suggests OH stretching of metal hydroxides, as discussed for the powder FTIR, in Fig. 1, similarly the two peaks at 2361 and 2341 cm À1 can be assigned to absorbed CO 2 [43].
The bands located at 3014 cm À1 and 2953 cm À1 can be assigned to the CH 2 asymmetric and symmetric vibration in NMP. [57] The peak at 2875 cm À1 can be attributed the CH stretching in sp 3 hybridised carbons, likely belonging to those in the NMP. 2792 cm À1 shows a weak peak that could be assigned to the CH stretching in sp 3 hybridised carbon in NMP. The CH 2 bonds in PVDF would similarly show at these points, however, due to the quantities of NMP and PVDF in the electrode slurries, the contributions due to PVDF would be far smaller. The highest intensity peak belongs to C@O stretching peak. c-lactams, such as NMP, typically exhibit a stretching at $1750-1700 cm À1 [46]. However, the effect of the methyl group on the nitrogen, over a single hydrogen, causes a greater resonance effect that causes the C@O stretching to shift to lower wavenumbers -$1670 cm À1 , resulting in a wavenumber that is as expected for NMP.
The remaining peaks that have been highlighted, at 1427 cm À1 , 881 cm À1 , 663 cm À1 , are the positions denoting NaOH from the powder FTIR (Fig. 1). Across all these spectra, from 0 to 360 min (Fig. 7), it is very difficult to confidently assign these to NaOH. There is a lot of overlap in these positions with those peaks of NMP (as seen in Fig. S1).
The peaks between 1600 and 1725 have been fitted (Fig. 8). The data shown in Fig. 8 is for scans between 0 and 30 min, later scans are included in the supplementary data. The fitted peaks under this region include: a growing peak at 1617 cm À1 , a peak that shifts from 1647 cm À1 to 1637 cm À1 , a large peak at $1665 cm À1 , a second growing peak at 1681 cm À1 , and a final peak at 1691 cm À1 . The main peak can be confidently assigned to the C@O bond in the NMP, initially at 1673 cm À1 and shifts to lower wavenumbers over time, 1663 cm À1 at 30 min. This shift is likely due to hydrogen bonding occurring as water is absorbed into the electrode slurry. Additionally, there is a slight broadening of this peak due to the absorbance of water creating a larger number of environments for this functional group. Secondly, the broad, low intensity peak, initially at 1617 cm À1 , can be assigned to conjugated C@C stretching, as has been previously assigned to alkaline treated PVDF in literature [58][59][60]. As the aging progresses, there appears to be a very slight shift in this peak, along with a slight increase in the intensity, suggesting that there is an increase in the number of these C@C structures within the PVDF as time progresses. Finally, the peak initially lying at 1648 cm À1 can also be assigned to nonconjugated C@C structures [60,61]. As the electrode slurry ages, there appears to be a slight shift in this peak, likely due to the effects of hydrogen bonding. There is very little overall change in the intensity of this peak, suggesting that while there may be incremental changes to the amount of non-conjugated C@C structures in the electrode slurry, the overall quantity remains constant, i.e., the rates of production and loss of C@C are equal. Curiously, by comparison of the changes to these two peaks, the quantity of conjugated C@C structures increases while the quantity of nonconjugated C@C structures remains near constant. It can be inferred that the degree of conjugation in the PVDF structure is slowly increasing as the dehydrofluorination reaction progresses. The peak at 1682 cm À1 , later at 1681 cm À1 , indicates a second C@O stretching occurs, additional to the C@O stretching of NMP. NMP has been previously reported to degrade under basic conditions, causing a ring opening reaction to occur, forming 4-(methylamino) butanoic acid [62]. As the electrode aging progresses, this initially sharp low intensity peak grows in intensity, consistent with a formation of 4-(methylamino) butanoic acid.
Overall, due to the presence of the C@C structures, both conjugated and non-conjugated, evident in the fitted FTIR data, it is likely that a dehydrofluorination reaction of PVDF is occurring. To confirm this the coated and dried electrodes were analysed using XPS. These electrodes (for XPS analysis) were coated immediately after mixing of the electrode slurry and then allowed to dry for $40 min while exposed to air. This allowed the reaction to occur as the electrodes were drying and, hence, the chemical changes to be observed without the gel network percolating the entire electrode slurry, rendering it completely uncoatable. Fig. 9 shows the F1s XPS spectra, with three distinct components, two of which relate to the PVDF, while the feature at 684.3 eV can only be assigned to NaF. The presence of NaF can be explained through expulsion of HF from the PVDF structure dur- ing defluorination, which subsequently reacts with NaOH forming NaF and H 2 O, as shown by Eq. (4) (see Fig. 10).
By comparison of the XPS peak areas, an elemental quantification can be obtained, and an idea of the degree and progression of the reaction can be noted À81.3% for CF 2 CH 2 , 3.0% for CF 2 CF 2 , and 15.7 % for NaF Assuming that all of the expelled HF reacts to form NaF, after the aging reaction has been allowed to progress, $15% of the fluorine in the PVDF was expelled from the polymer     and so 15% of the gelled PVDF binder contains conjugated or crosslinked structures.
In summary, NaOH formation on the cathode upon exposure to air has been confirmed from FTIR and SEM analysis, and FTIR and XPS results indicate that a dehydrofluorination reaction of PVDF is occurring during the aging of the slurry. The base catalysed reaction of PVDF in which a carbon-carbon double bond is formed as a result of the elimination of hydrogen fluoride (HF) units from the polymer, as shown in Fig. 14, has been widely reported in PVDF literature previously, although not in battery manufacturing literature [58,[64][65][66]. The formation of these C@C structures allows a crosslinking between PVDF chains to occurs, creating a gel network throughout the electrode ink.
In similar lithium-ion battery systems, it has been shown that LiOH is evolved first, followed by the formation of Li 2 CO 3 [36]. It is possible that the formation of LiOH is the cause of gelation in analogous lithium cathode electrode slurries. In this study were unable to detect the presence of Na 2 CO 3 and it is possible that the carbonate forms over larger timescales than those used in this research.
The gelation mechanism of dehydrofluorination and subsequent crosslinking is observed also in the time constant for the rheology as indicated through the phase angle, Fig. 4b. The later process, thought to be cross-linking exhibits a $1-2% phase angle, this observation coupled with the convergence of moduli at frequencies with time indicates the irreversible formation of a chemical gel, as opposed to a physically stabilised gel which can be broken up through increased applied shear. The storage modulus, G 0 , illustrates changes in the gel structure (Fig. 4a), and the change in G 0 with time indictes three likely stages of gel development, as observed in Fig. 15. The first stage is an initially slow rate, whereby the quantities of base are still increasing to the point of saturation and the dehydrofluorination is occurring, but the crosslinking is still slow. The second stage occurs at a faster rate, whereby the highest concentration of C@C structures exists, and crosslinking can readily occur. As this reaction progresses, the quantity of C@C structures decreases until, the third stage where the remaining C@C structures crosslink at a slower rate.
An additional possibility is the hydrolysis and ring opening reaction of NMP which occurs with bases, (Fig. 16) which has been detailed in literature [62,67].   This reaction occurs with a base present and, by examining the FTIR data in Figs. 7 and 8, there is a suggestion that this reaction is occurring. However, due to the very low intensities of the peaks, the presence of 4-(methylamino) butanoic acid either cannot be confirmed, or it exists in very small quantities. Thus, it likely does not contribute significantly to the gelation of these electrode slurries.
To investigate the effect of the electrode slurry aging upon electrochemical performance, electrodes were coated at different intervals after the mixing was completed, with electrode coating at 40 min being just possible. Beyond this time the electrode slurry was too solid-like to coat. The resulting cross-sectional SEM images of the electrodes are shown in Fig. 11. Fig. 11a shows the image of an electrode coated immediately after mixing, and (b) after 40 min. These samples contained no additives. It can be observed that the longer time resulted in aluminium rich amorphous regions, (Figs. S3 and S4) which are likely caused by aluminium corrosion of the current collector in these basic conditions. The aluminium corrosion will also result in hydrogen production, causing additional micro porosity in the electrodes [63]. Fig. 12a shows the capacity curves for the material in a half cell setup for the first and fifth cycles. It is evident that there is a charge transfer plateau at $2.8 V, which can be clearly seen for the charge and discharge curves in the first cycle.
There are, however, notable changes due to the electrode slurry aging, as seen in Fig. 12b. Over 50 cycles, the capacity of the material reduces from $138 ± 1.5 mAh/g to $81 ± 1.8 mAh/g (averages and standard deviations calculated from no less than 3 cells). The formation of NaOH due to the surface reactions of the active material ( Figs. 1 and 2), causes a loss of sodium, the charge carrier, from the active material which leads to a reduction in the capacity of the cells. Coating after 20 min mixing, results in very little change in cycle-life, indicating that the chemical changes due in the electrode slurry at this point are not enough to affect the life of the electrode. Coating after 40 min of mixing, however, results in a steady reduction in capacity from $138 ± 1.7 mAh/g to $75 ± 1.8 mAh/g over the 50 cycles.
Electrochemical impedance spectroscopy has been performed as the electrodes have cycled (Fig. 13). It is observed that the charge transfer resistance increases after formation up to 20 min, and then increases significantly after 40 min. This is consistent with the 20-minute stability window of the electrode slurry as observed in the oscillatory rheology (Fig. 13). After 50 cycles the aged slurry charge transfer resistance increases significantly, indi-cating that the ageing of the slurry influences the surface properties of the material, likely from sodium leaching and resistive sodium residuals at the surface, and hence increases the resistance for sodium transfer across the surface.

Conclusions
The powder stability of O3-type high nickel-based sodium ion cathode materials (NaNi 1/2 Mn 1/4 Ti 1/8 Sn 1/8 O 2 ) and the chemical gelation with PVDF and NMP binder in electrode slurries has been investigated through in-depth chemical analysis, electrochemical cycling, impedance spectroscopy and in-depth time dependent rheological studies of the electrode slurries. NaNi 1/2 Mn 1/4 Ti 1/8 -Sn 1/8 O 2 , a O3-type layered oxide material, prepared by a solidstate. The effects of air were characterized using FTIR and SEM, and sodium hydroxide was observed over prolonged exposure. The pristine powder was then processed into a slurry with PVDF binder and NMP solvent where over a prolonged time, an irreversible or chemically bonded gel was formed. FTIR studies on the electrode slurries show a formation of C@C bonds during the gelation process. Furthermore, sodium fluoride, observed through XPS in the electrode slurries after mixing and gelation, demonstrates a reaction between sodium hydroxide from the layered oxide and fluorine from the PVDF. These indicate a dehydrofluorination reaction of PVDF. The gelation of the electrode slurry, and the surface reactions affect the electrochemical performance. Al based amorphous regions, formed due to the basicity of the electrode slurry, were detected using cross-sectional SEM and EDS in electrodes coated 40 min after mixing. Electrochemical testing in sodium-metal anode half-cells with coatings produced 0, 20, and 40 min after mixing showed limited initial capacity differences, however, after 50 cycles, an increased capacity decline was observed for electrodes coated after 40 min of mixing. Electrodes coated 40 min after mixing similarly demonstrate greater charge transfer resistances over those produced 0, and 20 min after mixing, which increase further with cycling.
These key findings demonstrate defluorination and crosslinking reactions of the PVDF binder material, caused by the alkalinity of the sodium layered oxide electrode slurries, which in turn cause the gelation of the inks. From the time dependent rheological studies, the new concept of a three-stage gelation process of the electrode slurry was observed, likely linked to the concentration of unsaturated structures in the PVDF. First the PVDF is defluorinated, which causes cross-linking of the chains, which in turn causes a non-reversible chemical gelation of the slurry. Al-rich amorphous regions, observed using cross-section SEM, further emphasises this 3-stage gelation process, and highlight the impact of these reactions on the electrode structure. To predict the coatability from the rheological data would save significant time and cost for electrode development. These observations, indicate a 20-min processing window, which is impractical for manufacturing where 2-4 h or longer electrode slurry stability is required. Accordingly, methods to prevent the surface residue formation, to improve the stability window and prevent basic electrode slurries are required.
To improve the stability and cycling performance of these highly alkaline oxides, changes in the methodology for mixing and coating are required. The major challenges with electrode materials remains the gelation, and the affect this has on the coatability. Suppression of this gelation would mean this material or analogous materials (for both sodium and lithium-ion batteries) may be more freely tested, examined, and optimised. One potential solution, and a possible avenue for future work, is to coat the surface of these electrode materials with a sacrificial or blocking material to prevent exposure to air, and therefore preventing sodium surface species from forming. An alternative solution is use of an additive in these electrode slurries that is able to absorb water or pacify the surface.
In conclusion this work shows the mechanism for instability of high nickel containing powders and electrode slurries, and presents a new time dependent oscillatory rheology test that can be used to determine the process window for these unstable slurry systems.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.