2021 roadmap for sodium-ion batteries

Increasing concerns regarding the sustainability of lithium sources, due to their limited availability and consequent expected price increase, have raised awareness of the importance of developing alternative energy-storage candidates that can sustain the ever-growing energy demand. Furthermore, limitations on the availability of the transition metals used in the manufacturing of cathode materials, together with questionable mining practices, are driving development towards more sustainable elements. Given the uniformly high abundance and cost-effectiveness of sodium, as well as its very suitable redox potential (close to that of lithium), sodium-ion battery technology offers tremendous potential to be a counterpart to lithium-ion batteries (LIBs) in different application scenarios, such as stationary energy storage and low-cost vehicles. This potential is reflected by the major investments that are being made by industry in a wide variety of markets and in diverse material combinations. Despite the associated advantages of being a drop-in replacement for LIBs, there are remarkable differences in the physicochemical properties between sodium and lithium that give rise to different behaviours, for example, different coordination preferences in compounds, desolvation energies, or solubility of the solid–electrolyte interphase inorganic salt components. This demands a more detailed study of the underlying physical and chemical processes occurring in sodium-ion batteries and allows great scope for groundbreaking advances in the field, from lab-scale to scale-up. This roadmap provides an extensive review by experts in academia and industry of the current state of the art in 2021 and the different research directions and strategies currently underway to improve the performance of sodium-ion batteries. The aim is to provide an opinion with respect to the current challenges and opportunities, from the fundamental properties to the practical applications of this technology.

as well as its very suitable redox potential (close to that of lithium), sodium-ion battery technology offers tremendous potential to be a counterpart to lithium-ion batteries (LIBs) in different application scenarios, such as stationary energy storage and low-cost vehicles. This potential is reflected by the major investments that are being made by industry in a wide variety of markets and in diverse material combinations. Despite the associated advantages of being a drop-in replacement for LIBs, there are remarkable differences in the physicochemical properties between sodium and lithium that give rise to different behaviours, for example, different coordination preferences in compounds, desolvation energies, or solubility of the solid-electrolyte interphase inorganic salt components. This demands a more detailed study of the underlying physical and chemical processes occurring in sodium-ion batteries and allows great scope for groundbreaking advances in the field, from lab-scale to scale-up. This roadmap provides an extensive review by experts in academia and industry of the current state of the art in 2021 and the different research directions and strategies currently underway to improve the performance of sodium-ion batteries. The aim is to provide an opinion with respect to the current challenges and opportunities, from the fundamental properties to the practical applications of this technology.  44 4.4. The SEI in Na-ion batteries 47 4.5. SSEs for Na-ion batteries 50 5. Testing protocols for Na-ion batteries 53 6. Advanced characterisation 56 6.1. Neutron characterisation of battery materials 56 6.2. Solid-state NMR for the characterisation of Na-ion batteries 59 6.3. Nanoscale characterisation of the local physical properties of Na-ion batteries, electrodes, and interfaces 62 6.4. Dissolution benchmarking of sodium-ion electrodes Introduction of the current UK and international industry targets for NIBs, with special emphasis on the future performance metrics targeted by Faradion in the next five years with their technology, followed by a technoeconomic analysis. To permit a comparison, the performance and economic targets of NIBs are presented alongside commercial Li-ion battery data.

Contents
We expect that this 2021 roadmap for NIBs will be reference reading material for researchers working in the areas of electrochemical science, battery chemistry, energy materials, advanced characterisation, manufacturing, and computational methods. We also recommend other recent reviews on the topic of NIBs to the readers of this roadmap (e.g. [2][3][4]). Furthermore, we hope to inform key industries about current trends to encourage further investment in this technology, and influence decision-making from funding agencies and private investors.  power and energy optimisation can be addressed by the development of hybrid P2 | O3 materials [15]. Figure 1 indicates that there is a large combinatorial space of cation permutations that can be explored to improve power and energy densities, cycle life, and recyclability to meet the optimisation requirements of different industrial applications.
Perhaps most significantly, layered P2 | O3 cathodes offer the opportunity to develop sustainable 'green' batteries that are only manufactured from earth-abundant materials. LIBs face ethical challenges associated with mining the metal-containing ores as well as economic issues related to material scarcity and the environmental impact of recycling toxic chemical constituents. Fabricating high-performance NIBs from safe, earth-abundant materials containing, for example, a combination of sodium, iron, manganese,  spectrometry, Raman spectroscopy, electron paramagnetic resonance spectroscopy (EPR) and sequential cyclic voltammetry (CV) are often in demand to decouple oxygen redox from electrolyte decomposition. However, ex situ techniques may be unreliable due to the reactivity and instability of the phases formed at high states of charge. This has led to the recent suggestion that the behaviour of Li-rich layered materials may be derived from the Mn 4+ → Mn 7+ + 3e − redox reaction, the resulting species being too unstable to be observed [22]. Nonetheless, there is a strong indication of oxygen redox activity, based on a combination of ex-situ, in-situ and operando techniques.

Advances in science and technology to meet challenges
As mentioned above, the biggest problems encountered in cathode materials exhibiting oxygen redox are capacity fade combined with large voltage hysteresis. Among the exceptional materials which exhibit suppressed voltage hysteresis are Na 2 Mn 3 O 7 [24], with ordering between Mn and intrinsic transition metal vacancies and Na x Mn 1−y Li y O 2 [19], with a ribbon superstructure. These examples confirm that well-defined ordering in the transition-metal layers mitigates detrimental cation migration and phase changes. In the case of Na x Mn 1−y Li y O 2 , repopulation of mobile Li at the original site is critical to the retention of the superstructure. To exploit these advantages, the design of materials containing transition-metal vacancies and/or dopants that migrate reversibly represents a promising route towards reversible oxygen redox. This strategy should be coupled with an investigation of the mobility of transition-metal vacancies and the associated electronic structure changes. The development of ultra-high-resolution XAS and RIXS will be necessary to describe the complete mechanism of oxygen redox by distinguishing different oxidised oxygen species and to provide a detailed account of transition-metal charge transfer. As an alternative strategy, the activation of materials via partially reversible oxygen redox may be used to develop materials with enhanced cycling performance not accessible by direct synthesis. This approach may be considered analogous to the activation of the Li 2 MnO 3 component of Li-rich layered materials for LIBs [25]. To suppress oxygen release, surface engineering methods, such as the use of coatings, namely Al-based oxides and phosphates, may be implemented. Typically, these coatings react with the cathode material to stabilise the surface by creating a composite of the active material and the coating compounds [26]. Furthermore, electrolytes that are stable at very high potentials must be sought to enable an accurate study of the cathode material without interference from parasitic electrolyte reactions. To address this, several approaches can be adopted based on high-voltage LIBs, namely the use of electrolyte additives or super-concentrated commercial salts in organic solvents. On the other hand, organic solvents can be replaced by more stable ionic liquids or solid electrolytes, although these approaches are far from large-scale commercialisation [27].

Concluding remarks
The anion redox reaction represents a major challenge, in terms of both understanding and successful exploitation. However, the potential rewards in terms of increased energy density, together with the realisation that its occurrence is more widespread than hitherto believed, make it an important area for future development. This may take the form of materials that undergo irreversible changes as a result of oxygen redox and thereby generate phases with enhanced electrochemical performance. Alternatively, the developments outlined in the previous section could lead to the preparation of phases exhibiting truly reversible oxygen redox reactions stabilised by surface coatings and/or the availability of novel electrolytes. Progress will be assisted by the further development and increasing availability of advanced characterisation techniques, such as RIXS, which may be combined with HAXPES and XAS to obtain information at a range of depths. precursor and a glucose additive to preserve a reducing environment [35]. In situ XRD and thermogravimetric analysis monitoring provide effective optimisation guides during synthesis.
Investigations of mixed polyanionic materials continue to offer opportunities to tune the redox potential, as described by a recent report by Tang et al on the combined electronic inductive effect of oxalate and sulphate in the Na 2 Fe(C 2 O 4 )SO 4 .H 2 O phase [36]. Further investigations of dopant effects in this system could provide a pathway to improvements in electrochemical performance.

Advances in science and technology to meet challenges
Reliably determining the crystal structures of unknown phases remains a challenge. A recent development embracing a materials genomic approach was reported by Khalifah et al, who have focused on the unsolved structure of the low-temperature phase of Na 3 V 2 (PO 4 ) 2 F 3 [37]. A plethora of enumerated plausible trial structures (of the order of 3000), informed by reasoned Na-ion ordering, were fitted to experimental data by Rietveld refinement, which enabled researchers to identify a set of closely related structural models that could reliably describe the experimental powder pattern. Subsequent density functional theory (DFT) and structural analyses then pointed to the A2 1 am space group as the proposed structure. This analytical approach represents an exciting development in the determination of structures, particularly for emerging cathode chemistries, where complex cation ordering exists.
Insights from operando investigations continue to shed light on the complex structural and redox processes ongoing in operating batteries. One recent report which exemplifies this is a study ofvanadium oxyfluorides, Na 3 V 2 (PO 4 ) 2 F 3−y O y (0 ⩽ y ⩽ 2), by Croguennec et al [38]. By following the desodiation process via operando x-ray absorption spectroscopy (XAS) and combining these observations with insights gleaned from DFT calculations, it is possible to examine the complex changes in the average local coordination environment of vanadium across a series of compositions where vanadium redox centres coexist. Ongoing upgrades to synchrotron beamline facilities will expand our ability to probe the highly complex and rapid processes that occur during charge and discharge, which will enable us to better understand materials' structure-performance relationship.

Concluding remarks
Polyatomic anion cathodes continue to provide a route to increase operating potential by capitalising on the inductive effect, as well as by conferring structural stability to the material. There is a trade-off in terms of the capacities that can be achieved with the heavier polyatomic anion component (in comparison to layered oxides), but recent work to accommodate additional Na-ions through, for example, the application of suitable dopants, is intriguing. In addition to synthetic developments that allow access to high-quality samples and new polymorphs, particle size and shape tailoring will enable further improvements, such as higher power densities. In addition to a further examination of the effect of dopants, challenges remain in developing suitable coatings and additives that will bestow stability under ambient and high-voltage conditions.

Prussian blue and analogues William R Brant and Ronnie Mogensen
Department of Chemistry-Ångström Laboratory, Uppsala University, SE-75121 Uppsala, Sweden

Status
Prussian blue analogues (PBAs), with the general formula Na x M[M ′ (CN) 6 ] 1−y .zG (where M and M ′ are transition metals, most commonly Mn, Fe, or Ni, y is the number of [M ′ (CN) 6 ] n− vacancies, and G is a neutral guest, such as H 2 O), are highly porous positive-electrode materials for NIBs (top of figure 6). Since PBAs are comprised of abundant elements and can be produced at low cost, they provide the best price-to-performance ratio [39] for a positive electrode. Thus, PBAs capitalise on the key advantages of NIBs as cheap, sustainable, and safer alternatives to lithium-ion technology. They have become viable cathode materials for sodium batteries within the last decade. During this time, the greatest challenge facing the implementation of PBAs as positive electrodes has been the ability to produce highly-sodiated compounds that are structurally stable during cycling. The structural stability and maximum capacity of the material are severely reduced by the presence of the vacancies (y) introduced during synthesis. Due to the loss of a reducible cation (M ′ ) and the necessary charge compensation required to offset the negative charge of [M ′ (CN) 6 ] n− , 1 mol% of vacancies can lead to a 3-4 mol% reduction in the maximum sodium content. Resolving this issue requires alternate synthetic approaches to the more traditionally applied co-precipitation reactions. For some M cations, in particular, Fe, co-precipitation results in a high vacancy concentration. In 2014, You et al reported a way of producing a low-vacancy PBA from a single precursor via an acid decomposition synthesis route for the first time [40]. This approach was subsequently used to produce the best-performing PBA to date, which delivered 120 mAh g −1 for over 800 cycles of charging at 0.5 C (75 mA g −1 ) and discharging at 2 C (300 mA g −1 ) [41]. Since then, extensive effort has been invested in the exploration of different M combinations, particle sizes, morphologies, and coatings, and improvements have been seen for either initial capacity or total cycling life, but not both [42]. Ultimately, the goal for PBAs is to achieve capacities of more than 160 mAh g −1 which are stable for up to 10 000 cycles. Eight thousand cycles have been demonstrated using aqueous electrolytes (figure 6 top) [43]; however, the resulting capacity is frequently limited to 60-80 mAh g −1 . Conversely, higher capacities (>80 mAh g −1 ) can be achieved using non-aqueous electrolytes due to their wider potential stability window although, the number of cycles obtained is often an order of magnitude lower.

Current and future challenges
Two connected issues are plaguing the development of PBAs: moisture sensitivity and limited reversibility when cycling more than one atom of sodium per formula unit. Capacity fading in PBAs has generally been attributed to limited electronic conductivity, the presence of water, and phase transitions of the highly sodiated compounds. Several methods have been employed to overcome their limited electronic conductivities, such as growing PBAs directly on carbon nanotubes, enabling stable cycling down to −25 • C [45]. Structural stability has seemingly been a greater challenge to overcome, since phase transitions in sodium-rich PBAs are heavily influenced by their water content. Wang et al [41] first demonstrated that, when completely dehydrated, Fe-based PBAs adopt a rhombohedral structure with a unit cell volume that decreases by ≈18%. This work was followed up by Rudola et al [44], who demonstrated that obtaining a moisture-free PBA depends heavily on the temperature and pressure used for drying and that, if exposed to air, a hydrated structure forms within minutes (bottom of figure 6). Further, a 16% volume change between cubic Fe[Fe(CN) 6 ] and rhombohedral Na 2 Fe[Fe(CN) 6 ] leads to a lower cycling stability compared to the hydrated compound, if the capacity and voltage window are limited in the hydrated case. Water also increases the average voltage output [46], which may be perceived as an advantage. However, the high voltage plateau may lie beyond the upper stability limit of water (3.9 V vs. Na + /Na). Finally, it has been shown that desodiated PBAs have a lower affinity for water [47]. Consequently, charging the cell runs an additional risk of releasing water into the electrolyte. Thus, the influence of water on electrochemical performance is complicated. The presence of water is beneficial for ensuring structural stability by minimising volume contraction during cycling. However, it limits the maximum capacity that can be utilised, as water is either released from the structure and reacts with the electrolyte, or is oxidised inside the material at high potential. Solving this challenge will require further consideration of the role of the electrolyte. Electrolyte-PBA interactions will be the greatest hurdle to overcome in the future, as all synthetic design choices (figure 7) will have to determine the optimum electrolyte combination. The synthetic possibility space for PBAs is large, and so one must take a holistic view when designing a cell based on this material. Reprinted from [43], Copyright (2020), with permission from Elsevier. Bottom: structural transitions introduced due to the removal and uptake of water in Na2Fe[Fe(CN)6], as determined from x-ray diffraction data [44].

Figure 7.
Schematic showing some of the possible combinations used to produce PBAs for sodium-ion batteries and hypothetical electrolyte combinations. The steps are: synthesis, where composition, core-shell morphology, particle size, and shape can be varied, or the PBA can be grown directly onto carbon supports; handling, where one can choose to dehydrate the PBA; finally, post-synthesis modifications, where new guest species, surface coatings, or aging are used. Two pathways described in the text are highlighted in blue and orange.

Advances in science and technology to meet challenges
Optimal material design choices for PBAs are ones that address multiple issues while maintaining stable or synergetic interactions with the electrolyte. More advanced material modifications can be achieved, given a rigorous understanding of how a modification influences other synthetic steps and compatibility with other material features or the electrolyte. For example, if one elects to utilise the hydrated PBA material, a surface coating, such as reduced graphene oxide [48] might be designed. This could keep the water from leaving the structure and prevent soluble electrolyte decomposition products from entering the structure, while permitting the passage of sodium. Thus, structural integrity would be maintained while preventing side reactions with the electrolyte (figure 7, blue path). However, this solution will only function with M and M ′ cations for which the voltage remains below the stability limit of water [46]. Alternatively, the electrolyte may intentionally include small molecules, such as acetonitrile, which co-insert with sodium, fulfilling a similar role to water in aqueous-based electrolytes (figure 7, orange path). Any potential solution must be balanced against all other synthetic choices, such that the potential benefit that the modification brings is not offset by a lack of compatibility with either the electrode or the electrolyte. Thus, a paradigm shift in research mindset must be made, away from simply aiming to boost performance with a single modification, and towards deconstructing the fundamental interactions within the material and between the material and its environment.

Concluding remarks
PBAs are a challenging class of compounds to investigate for NIBs due to the highly tuneable nature of their synthesis and the tangled web of chemical interactions within the material and with its environment. However, their potential as positive electrodes in beyond lithium-ion systems is clear, in terms of both cost and performance. Thus, the grand challenge will be to build a robust fundamental understanding of these complex interactions, enabling material developments which foster synergies that can be established using new advanced electrolytes.

Current and future challenges
In principle, n-type cathode materials are required to assemble NIBs, because the electrode reaction enables reversible cation uptake/removal. To date, the best reported electrode materials that can achieve excellent stability over thousands of cycles are based on polymerised perylene diimides (table 1, entry 1) [57,58], because aromatic diimides are robust redox-active organic moieties. However, they suffer from limited working potential values which usually do not exceed 2.5 V vs. Na + /Na. Moreover, they are usually prepared in their oxidised (desodiated) state (left-hand side of the half-reaction at the positive electrode in table 1, entries 1-3), which restricts their practical use in sodium half-cells or requires a pre-reduction (sodiation) step before full-cell assembly (table 1, entries 2 and 3) [59,60]. One of the few examples of a sodiated cathode material is the tetrasodium salt of 2,5-dihydroxyterephthalic acid (Na 4 DHTA), which has been successfully introduced in a symmetric Na-ion cell configuration [61]. Again, the output voltage of the cell was restricted to about 2 V due to the low working potential of the cathode side (2.3 V vs. Na + /Na). This potential limitation for non-polymer n-type OEMs (such as Na 4 DHTA salt) is notably related to the high electron density in the redox-active organic centre. The DHTA organic ligand exhibits at least two permanent negative charges through the two inactive carboxylate functional groups. The latter serve as a simple and effective way to prevent dissolution in aprotic electrolytes [51] but their electron-donating behaviour drags the average redox potential of the whole compound down. Although polymerisation is another excellent strategy for preventing dissolution in the electrolyte, the high-molecular-weight polymers are unlikely to be chemically prepared in their reduced (sodiated) state. It is worth noting that the low operating potential of n-type organic cathodes seems related to sodium chemistry itself. As first underlined by Song et al [62], the electrochemical uptake/removal of Na + in OEMs is negatively shifted (by several hundreds of mV) in comparison with its lithium counterpart beyond the expected standard potential gap between Na + /Na and Li + /Li redox couples. Moreover, the larger radius of Na + sometimes induces sluggish electrochemical kinetics as well as lowered utilisation of the available redox-active moieties, giving rise to diminished reversible capacity values. In short, there is currently a lack of efficient and robust sodiated organic cathodes with high working potential. representative examples exhibiting redox rocking from the aromatic to the quinoid form. X/Y can be N, O, S, P, π-systems, carboxylate, anhydride, or amide functional groups; and R, R ′ are potentially integrated within the same cyclic structure. Note that p-and n-type structures correspond to systems A and B, respectively, according to Hünig's classification [56]. (b) Corresponding cell configurations obtained by experimenting with both n-and p-type systems, shown during the discharge process. Reprinted with permission from [52]. Copyright (2020) American Chemical Society.

Advances in science and technology to meet challenges
Efforts must be focused on molecular design to obtain high-potential sodiated cathode materials. This would probably entail the identification of innovative n-type organic functional groups possibly guided by simulation and modelling. Furthermore, potential gain could still be possible by considering electronic effects acting on a redox centre of interest. Thus, the redox potential is classically increased at the molecular level (discrete entities) by introducing electron-withdrawing groups. More recently, it was demonstrated for a lithiated organic cathode that the electronic effects acting on a redox-active organic centre can be very efficiently mitigated in the solid state thanks to structural effects [66]. Alternatively, other OEMs can operate as cathodes thanks to p-type functionality with anionic compensation (figure 8). The latter, which is a redox system rarely encountered in inorganic materials, typically works at a higher potential than n-type redox centres, making these organic systems very appealing for Na-based batteries. For instance, several organic p-type polymer families are known to be stable in organic liquid electrolytes. Poly(2,2,6,6tetramethylpiperidinyloxy methacrylate) or PTMA represents a relevant example in this field which exhibits bipolar (n/p) electroactivity. From its initial state (centre of the half-reaction at the positive electrode in table 1, entry 5), PTMA can electrochemically accommodate a cation at 2.3 V vs. Na + /Na (in reduction) or an anion at 3.6 V vs. Na + /Na (in oxidation) [63]. However, in the case of the closely related poly(2,2,6,6-tetramethylpiperidine-4-yl-1-oxyl vinyl ether) or PTVE, the cation insertion mechanism was shown to be unstable, and steady capacity was obtained with anion insertion alone (table 1, entry 6) [64]. Such p-type functionality gives access to high-potential materials which can be paired with n-type negative OEMs in a dual-ion cell configuration (table 1, entry 7) [65]. P-type OEMs are usually synthesised in their reduced state, making them suitable for a positive electrode without requiring an electrochemical pre-treatment. As monomers, they do not possess permanent negative charges to prevent dissolution in aprotic electrolytes, but polymeric p-type OEMs seem to be the most appropriate choice of positive electrode for Na-based batteries to achieve a combination of high potential and cycling stability. However, as underlined in figure 8(b), the electrolyte is a reservoir of ions for charge compensation within electrode materials which causes a depletion of ionic carriers during operation, requiring a larger volume of electrolyte. Strictly speaking, such a cell configuration cannot be considered to be a NIB, although promising electrochemical data upon cycling have been reported (table 1, entry 7). Table 1.
Selected organic positive electrode materials tested in non-aqueous electrolytes, either in Na half-cell or full-cell configurations. Note that specific capacity values are reported per mass of positive active material, except for entry #4, which is expressed per mass of negative active material. PTCDI: 3,4,9,10-perylenetetracarboxylic diimide; TPA: terephthalic acid; DHTPA: 2,5-dihydroxyterephthalic acid; PTMA: poly(2,2,6,6-tetramethylpiperidine-4-yl-1-oxyl methacrylate); PTVE: poly(2,2,6,6-tetramethylpiperidine-4-yl- ≈3. 45 127, 94%, 100 [64] 7 Saturated NaPF6 in DOL/DME (v/v 1:1) ≈1. 8 180, 85%, 500 [65] Concluding remarks Further work is primarily required to develop high-potential and efficient sodiated organic cathode materials beyond p-type materials for NIB technology. A better understanding of the charge/discharge storage mechanisms is also crucial in order to design functional batteries, because each OEM seems to behave differently upon cycling. The use of computational techniques could be very useful in this quest. Organic compounds will not be able to supplement inorganic materials in terms of volumetric energy density, since their densities do not exceed 1.5 g cm −3 , i.e. four to five times less than inorganic electrode materials. Furthermore, another drawback lies in their tendency to dissolve in liquid organic electrolytes, although the use of solid electrolytes could solve this limitation. Conversely, the implementation of soft organic polymer electrodes enables the fabrication of flexible and thin electrochemical devices which are potentially environmentally friendly. Finally, aqueous organic NIBs can be perceived as the ultimate solution for promoting non-flammable, cost-effective, and green batteries even if only hybrid systems are reported at the moment.

Status
Hard carbon materials are the most popular choice for NIB anodes. They are produced from oxygen-rich precursors which cannot be converted into graphite, no matter how high the carbonisation temperature. Hard carbons consist of randomly oriented graphitic domains possessing a higher interlayer spacing than graphite (i.e. >0.34 nm) connected by disordered carbon regions with different curvatures. In between these disordered and more ordered domains, hard carbons have 'closed' pores, meaning they cannot be probed by gas sorption. The larger the size of the graphitic crystallites, the larger the pore size in between. In addition to curvatures and edge sites, hard carbons also contain the remaining heteroatoms (mainly oxygenated groups). Such a complex mixture of crystalline and disordered domains with defects allows sodium diffusion pathways and sodium storage sites. The first structure of 'graphitisable' vs. 'non-graphitisable' carbons was described by Rosalind Franklin in a seminal paper entitled 'The interpretation of diffuse x-ray diagrams of carbon' published in Acta Crystallographica in 1950 (figure 9(a)) [67]. Interestingly, until today, the exact structure of hard carbon has remained unknown, as this depends on the precursor and carbonisation conditions, which result in materials with varying interlayer spacing, crystallite size, pore domains and edge termination. Such different structures store Na ions in very different ways. Hence, it is no surprise that there is much debate in the literature about the mechanism of Na insertion into hard carbons. Figure 9(b) provides a typical Na insertion load curve (sodiation curve in a half-cell built with hard carbon vs. metallic Na) showing a typical sloping region between 1 V and 0.1 V and a plateau region below 0.1 V. The first theory by Stevens and Dahn using the 'house of cards' model for hard carbon attributed the slope to Na insertion into expanded graphitic domains, and the plateau to pore filling by Na (figure 10) [68]. Yet the situation is more complex than this, as hard carbons can exhibit different features depending on the choice of precursor, pre-treatment, and carbonisation conditions and temperature.

Current and future challenges
The biggest challenge for hard carbons is to improve their storage capacity to match or even overtake that of graphite in LIBs (>372 mAh g −1 ). The practical measured capacities for hard carbons today (mostly in coin cells vs. metallic Na) vary from 200 to 450 mAh g −1 ; most of the values reported in the literature average around 300 mAh g −1 . While the theoretical energy density of the commercial graphite used in LIBs can be calculated using the LiC 6 formula, for Na in hard carbons, the situation is far more complicated, due to the heterogenous nature of hard carbons and the unknown storage mechanism, which depend on many structural parameters. Therefore, all the structural and morphological features of hard carbons must be clearly determined using multiple characterisation techniques applied in concert to accurately determine the exact features, such as the interlayer spacing (XRD), defects (Raman, positron annihilation spectroscopy), functional groups (XPS), the nature of ordered vs. disordered domains and the interconnectivity between the two (x-ray pair distribution function), opened vs. closed pores and pore sizes (SAXS, SANS, gas adsorption). All these features must be precisely correlated with the electrochemical performance, which should be accurately determined in half cells, but also using three-electrode configurations [71] to better understand the fundamental electrochemistry happening at the working anode. Finally, operando characterisation techniques should be employed to understand the structural and morphological changes occurring during the intercalation of Na ions into hard carbons. Some of these techniques involve the use of XRD/SAXS to determine the change in the interlayer spacing as proof of Na intercalation [68,72,73], operando XPS [74] to determine the interaction of Na with different functional groups, in situ electrochemical TEM to observe the interaction of Na with defects [75], operando Raman spectroscopy to understand the changes in graphitic/amorphous structure upon (de)sodiation [76], operando NMR [77][78][79], MRI [80] and EPR [81] to understand the Na chemical state/deposition inside pores at less than 0.1 V. Another major challenge in the development of high-performance hard carbon anodes is to understand the solid-electrolyte interface [82,83] and how the anode's structural features, in combination with the electrolyte of choice, affect the thickness and, more importantly, the ionic conductivity of the SEI. To probe this, we recommend operando AFM [84] as well as electrochemical impedance spectroscopy coupled with quartz crystal microbalance measurements [85], corroborated by electrochemical mass spectroscopy [86] and operando FTIR [87]. To understand macroscale phenomena related to the anode volume change, operando x-ray computed tomography [88] and electrochemical dilatometry are good options [89,90]. Once the fundamentals are understood, both in an individual electrode as well as during its operation, these could be used to build  [67]. Reproduced with permission from [69]. (b) A typical discharge curve for hard carbons shows two distinct regions: the sloping region and the plateau region. Figure 10. Schematic illustration of the proposed sodium storage mechanisms; stack thicknesses and plane sizes are not to scale, and defects and curvatures are not shown [70]. Reproduced from [70] with permission of The Royal Society of Chemistry. models for further optimisation based on DFT [91,92], molecular dynamics (MD) [93], and machine learning [94] to predict the optimal structural and morphological parameters.

Advances in science and technology to meet challenges
Several research groups have made great advances in developing hard carbon anodes with outstanding performance from low-cost precursors, based on a careful synthetic design combined with advanced ex-situ and in situ characterisation [95]. Most debates to date have arisen from contradictory Na storage mechanisms in such carbons. Currently, there are several published theories: (a) the 'intercalation-filling' model: Na + ions intercalate into graphitic layers in the sloping region, and insert themselves into nanopores in the plateau region [96]; (b) the adsorption-intercalation model: Na + ions adsorb at the surface or defect sites of the carbon electrodes within the sloping region and intercalate into the graphitic layers within the plateau region [97]; (c) the adsorption-filling model: in the sloping region, Na + ions adsorb at the defect sites while filling the nanopores in the plateau region [72], and (d) the 'three-stage' model: defect adsorption of Na + ions takes place in the sloping region, but in the plateau region, the Na + ions first intercalate into the graphitic layers and fill the nanopores at the end [98]. More recent investigations by my research team as well as other authors have gathered stronger evidence to propose an alternative model, with adsorption and intercalation occurring simultaneously during the sloping region, and pore filling during the plateau region, all the while keeping in mind the dependence on the structure and pore system of the respective carbons ( figure 10). Great progress has been achieved in using operando characterisation techniques to probe the mechanisms of different carbon materials; in particular, operando NMR has been used to probe the metallic nature of Na in the pores [79,80], SAXS/WAXS [100,101] has been used to probe the Na interaction with defects and the importance of interlayer spacing, and x-ray PDF [102] has been used to understand the connectivity between ordered and disordered domains and the way Na travels throughout the hard carbon from defects and graphitic domains into the pores. Finally, significant machine learning efforts have recently been dedicated to the 'in silico' design of optimised carbons based on experimental characterisation techniques [94]. At a commercial level, the Japanese company Kuraray sells KURANODE™ as a hard carbon for LIBs and NIBs, with good performance, but few structural details. This product is often used as a benchmark by many researchers developing optimisation strategies for NIBs at the cell level, as well as by some of the small and medium enterprises developing NIBs.

Concluding remarks
Much progress has been made in developing hard carbon anodes with outstanding performance for NIBs. Future optimisation could be possible via automation of the synthetic processes using robots coupled with machine learning. Benchmarking of the best-performing hard carbons would be ideal to unify the efforts of NIB developers and standardise protocols at an international level.

Current and future challenges
The main factors limiting the commercialisation of Na 2 Ti 3 O 7 (NTO) in NIBs are the low electronic conductivity and sluggish Na + ion (de)intercalation kinetics, which result in poor rate performance and cycling stability. The presence of Ti 4+ ions (d 0 ) makes NTO an electrical insulator (bandgap energy ≈ 3.7 eV) [112], which hampers Na + ion diffusion within the crystal structure (D Na + in NTO is ≈ 2-3 orders of magnitude lower than that in hard carbon). Thus, a major area of research encompasses the development of strategies to enhance long-term cycling stability, particularly at high current densities. These mainly involve: (a) bulk and surface structural control through doping and the introduction of defects; (b) nanostructuring; and (c) composite fabrication with carbon allotropes. Furthermore, the formation of an SEI layer on the surface of the electrode and the consumption of Na + ions by carbon additives below 1 V result in low initial coulombic efficiency (ICE) (40%-60%). These phenomena are particularly critical in full cells, where there is limited sodium available. Methodologies are therefore required to mitigate the initial capacity irreversibility, although, to date, such procedures are underdeveloped. Additionally, the high reactivity of Na metal when used in half cells contributes to the underperformance of NTO due to the formation of an unstable SEI layer, requiring the use of three-electrode cells and full cells to gain a better insight into the Na contribution to the overall performance.
Few studies have demonstrated the complex nature of the SEI composition. For example, Casas-Cabanas et al reported the formation of various SEI inorganic and organic species, such as Na 2 CO 3 , alkyl carbonates and PEO during discharge when using 1 M NaClO 4 in an EC:PC electrolyte [113]. Moreover, NaCl and NaF products have been observed before cycling, as a result of the decomposition of the NaClO 4 inorganic electrolyte salt and the Polyvinylidene fluoride (PVDF) binder, respectively. Alternative inorganic salts, such as NaBF 4 or NaFSI and binders, such as sodium alginate or sodium carboxymethylcellulose have shown superior cycling stability, although their performance still lags behind those of their Li analogues [114].
Therefore, the development of next-generation electrolytes and binders is crucial to further improve battery efficiency. Another issue is the structural relaxation of the fully sodiated phase (Na 4 Ti 3 O 7 ), attributed to the large electrostatic repulsion in the crystal structure [115]. This phenomenon may lead to self-discharge in full cells containing NTO, compromising cycling stability.

Advances in science and technology to meet challenges
To address the poor electronic conductivity and high Na + ion migration barriers in NTO, different approaches, including bulk and surface (micro)structure design and carbon composite fabrication have been considered.
Elemental doping using small concentrations (<5 at.%) of Nb [116] and elements from the lanthanide series (e.g. Yb) [117] has been shown to decrease the bandgap energy, enabling superior high-rate performance in NTO. For example, Nb-doped NTO demonstrated a 0.3 eV reduction in bandgap energy with respect to pristine NTO [116]. The enhanced electrochemical performance was attributed to doping-induced crystal structure distortions, together with the formation of Ti 3+ ions. Furthermore, oxygen vacancies, which act as n-type defects, were claimed to contribute to the higher electrical conductivity observed in Yb-doped NTO [117]. Similarly, surface treatments involving the direct or indirect use of hydrogen have been used to create oxygen vacancies, conferring high capacities at fast rates [118].
Besides changes in the crystal structure, nanostructuring strategies have been extensively developed, showing an improved electrochemical performance through more efficient Na + ion (de)intercalation, exploiting larger surface areas which can be further enhanced by creating continuous networks of material to improve electronic conductivity. For instance, Anwer et al used solvothermal synthesis to create 3D networks of 2D NTO nanosheets [119], generating microflowers (figure 12(a)) with a stable capacity of 108 mAh g −1 (200 mA g −1 ) over 1000 cycles, while NTO nanotubes 10-20 nm in diameter (figure 12(b)) have shown a high stable capacity of 100 mAh g −1 [120] (100 mA g −1 ) over 2000 cycles. Moreover, the use of carbon coatings [115] and carbon composite mixtures [121] is a common strategy to improve the electronic conductivity and pseudocapacitive contribution during Na + ion storage. For instance, nanosheet-coated carbon shell particles delivered a capacity of 110 mAh g −1 (885 mA g −1 ) over 1000 cycles [122].
Finally, understanding the impact of electrolyte composition on interfacial electrochemistry is crucial for developing approaches to further prolong the cycle life of NIBs. Although this area is still in its infancy, preliminary work has been developed to mitigate SEI film instability by improving electrolyte formulations (e.g., by the incorporation of fluoroethylene carbonate (FEC) additives) and through surface engineering, among others.

Concluding remarks
Titanium-based oxides represent a very exciting class of anode materials due to their safety, stability, low-cost, non-toxicity, and ease of processing. While progress has been made in terms of understanding the factors limiting the rate performance and cycling stability, much work is still required to realise these concepts in terms of delivering high energy and power densities in NIBs. We anticipate that the greatest advances will arise from further enhancements in electronic conductivity (by combining several of the strategies described here) and further exploration within this family of compounds, including composites. Designing more stable anode/electrolyte interfaces will also be crucial to advance this area of research. This will require a better fundamental understanding of the physicochemical properties at the interface, the behaviour and stability of the fully sodiated phase, and the driving factors behind the poor initial coulombic efficiency.

Status
The development of alloy and conversion negative electrode materials for NIBs (which characteristics are resumed in figure 13) represent a substantial fraction of the work carried out to develop a suitable anode for application in commercial systems. While, on the one hand, the high reactivity of sodium and its low melting point (98 • C), as well as the formation of dendrites during its electrodeposition, make it significantly less safe than Li, on the other hand, in contrast to the case of LIBs, graphite does not intercalate Na + in conventional electrolytes, and thus alternative anodes have to be found.
Several families of materials capable of reacting electrochemically with sodium have been studied and can be classified, as for LIBs, according to their reaction mechanism, as either insertion, alloy, or conversion materials [123]. However, the mechanism and performance of such materials in NIBs cannot be directly extrapolated from those observed in LIBs.
The alloying reaction is usually observed with p-block elements (Ge, Sn, Pb, P, Sb, Bi, etc) and some transition metals (e.g., Ag, Zn and Au) which form stable alloys with Na [124]. Alloy materials (AM) usually involve a multiple-electron exchange, thus leading to very high capacities. However, the formation of the sodiated alloy causes a strong expansion of volume, even worse than that of LIBs, which represents a real limitation on their application. The most interesting AMs are surely phosphorus and antimony, which lead to the formation of Na 3 P (2596 mAh g −1 ) and Na 3 Sb (660 mAh g −1 ), respectively. Interestingly, silicon, while very promising in LIBs, is practically inactive in NIBs. While phosphorus delivers a very high theoretical capacity, antimony has shown the greatest affinity for sodium [125].
The family of conversion materials (CM) mainly includes metal oxides and sulphides, even though other families, such as metal selenides, carbodiimides, halides, and phosphides have been explored [126], all of which underwent the so-called conversion reaction [127] with Na. CMs can deliver high specific and gravimetric capacities with excellent cycling stability, but they also face at least one critical issue related to their low initial coulombic efficiency or high volume expansion.
The study of the electrochemical and aging mechanisms in both AMs and CMs has been rather challenging, given that multi-phase systems are often obtained during cycling, and many of the species formed are amorphous, nanosized, and especially metastable. The development of specific operando tools is therefore mandatory for the study of these materials [128,129].

Current and future challenges
The main challenges for AMs and CMs in NIBs are related to volume expansion and both the ionic and electronic conductivities of these materials. The large reversible volume exchange during charge/discharge cycles produces mechanical stress, which, on the one hand, rapidly leads to the electrochemical pulverisation of the electrode material, and, on the other hand, exposes new portions of the metal surface to the electrolyte at each cycle, leading to continuous electrolyte degradation and thus difficult stabilisation of the SEI. Moreover, in some cases, such as that of phosphorus, the active material is intrinsically insulating, hindering electronic percolation in the electrode. Control of these phenomena is necessary for practical applications, and several strategies have been proposed to reduce the consequences of drastic volume changes and to boost electronic conduction.
The simplest approach is to optimise the electrode formulation using conducting additives, electrolytes, and binders. In the case of phosphorus, for instance, the addition of 1D (carbon nanotubes) or 2D (graphene) nanomaterials to the electrode formulation has been shown to improve the conductivity of C/P composites, leading to an outstanding capacity retention of 80% over 2000 cycles [130], despite noticeably impacting capacity and energy performance. Nanostructuring is also a common approach to accommodate the stress and strain in AMs and CMs without pulverising the electrode, and to improve the ionic and electronic transport pathways. For instance, tin particles smaller than 10 nm allow a decrease in the strain, effectively mitigate pulverisation, and also limit their aggregation during cycling. Such nanoparticles, homogeneously embedded in a carbon matrix, deliver a stable capacity of 415 mAh g −1 after 500 cycles at 1 A g −1 [131]. Even tuning the morphology and porosity of the materials may allow an improvement in the performance. An interesting way to do so is to transform bulk materials into 2D materials which can buffer volume expansion during alkali insertion [132]. For instance, phosphorene synthesised by exfoliating phosphorus shows a rapid charge transfer between less-conductive phosphorene interlayers. In the same spirit, layered graphene/phosphorene electrodes achieved a reversible capacity of 2440 mAh g −1 (per gram of P) over 100 cycles [133]. Advances in science and technology to meet challenges As described above, the main challenge for AMs and CMs is related to volume expansion, which results in the continuous pulverisation of electrode materials and leads to a steep deterioration in electrochemical performance. The consequences of the large volume expansion are worsened by continuous electrolyte reduction, which affects both the coulombic efficiency and the cycle life.
The different strategies presented in the previous section have led to great progress in electrochemical performance in terms of cycle life, even though a substantial decrease in volumetric energy density cannot be avoided, especially when composites with inactive materials are prepared. Moreover, if in the early papers, the effective behaviour of the presented materials was mainly tested in half-cells vs. Na metal, the more recent study of these materials in realistic full cells against a real cathode material has increased, and is now rather common (for CM, see [132]). These studies are necessary to correctly evaluate the viability of these materials, which still cause non-negligible electrolyte decomposition and irreversible sodium trapping, and thus, low coulombic efficiencies in both the first and the following cycles.
When compared with other types of anode material, up until now, the performance of AMs and CMs has not beaten that of hard carbon. This is partially due to the large amount of cathode excess required to compensate for the large initial irreversible capacity and the low coulombic efficiency in cycling. Therefore, one can conclude that the primary challenge for AMs and CMs has still to be solved [134].
Up until now, hard carbon has remained the most relevant candidate, even though its cost is significantly higher than that of graphite; in fact, the cost of graphite ranges between 500 and 2000 $/ton, which is similar to the cost of some of the common precursors of hard carbon (e.g. cellulose or sucrose). Among AMs, antimony is probably the most promising material, since it possesses a large gravimetric charge storage capacity (660 mAh g −1 ) associated with an excellent rate capability and cycle life. Such excellent performance, exceeding that of Sb in LIBs, should give it good prospects when used in composites with hard carbon. Indeed, a good compromise between energy and lifetime cost can, in principle, be attained for such composites.

Concluding remarks
Many studies of many AMs and CMs in NIBs have shown that most of them form Na-rich phases, and thus produce high specific capacities and energy densities. While volume expansion is often invoked as a possible limitation of the life span of such electrodes, several studies have shown impressively long cyclabilities, especially for antimony-based systems. Their durability lies in the elastic softening of the sodiated antimony phases, which enhances the ability of the electrode to absorb and mitigate the strong volume changes upon (de)sodiation [135]. The development of composites of antimony with hard carbon seems, therefore, a viable alternative. Even though such composites might still not allow NIBs to reach the performance of current LIBs, their application is still suited to applications where the energy performance is not critical, such as in stationary energy storage.

Status
The relatively large radius and heavy mass of sodium ions require NIB anode materials with large interlayer spacings. The large volume expansion and pulverisation of some anode materials during cycling also pose significant challenges for the development of high-performance NIB anodes. The rational stacking of various 2D nanosheets could allow electrode designs that would be mechanically stable over a large number of cycles.
2D materials possess attributes that may help to address these challenges. In particular, metallic 2D materials with high conductivity may improve the reaction efficiency, and weak van der Waals bonding between planar layers may allow the fast diffusion of Na + . Furthermore, the interlayer spacing between nanosheets can increase the contact area with the electrolyte and shorten the Na + diffusion length. Electrodes fabricated by the restacking of exfoliated 2D nanosheets are mechanically flexible and can accommodate large expansions in volume [136]. Among such electrodes, the metallic phases of 2D transition-metal dichalcogenides (TMDs) MX 2 (M = Mo, W, Ti et al; X = S, Se, Te) such as molybdenum disulphide (MoS 2 ) [137,138] provide large channels (6.15 Å) for Na + intercalation and possess a high theoretical capacity of 670 mAh g −1 , which makes them promising for use in NIB anodes [139]. The unit cells of the semiconducting 2H phase and the metallic 1T phase of MoS 2 , along with an image of a uniform solution of chemically exfoliated nanosheets are shown in figures 14(a) and (b). A cross-sectional scanning electron microscope (SEM) image of the restacked 1T phase MoS 2 nanosheets is shown in figure 14(c). Individual nanosheets can be seen, with intercalation sites between them.
Current and future challenges 2D TMDs are promising anode materials for Na-ion batteries, but several challenges hinder their performance. One challenge is the lack of understanding of the charge-storage mechanism. Broadly speaking, TMD-based NIB anodes work via the intercalation/extraction of sodium ions in the layered structure. Unlike storage mechanisms in conventional 2D materials (graphene, MXenes), a phase transition occurs in TMD-based electrodes, accompanied by electrochemical intercalation of sodium ions into the structure [140]. For example, in MoS 2 anodes, the intercalation of ions leads to a transformation from the semiconducting 2H phase to the metallic 1T phase [141]. This transition from 2H-MoS 2 to 1T-MoS 2 involves a slippage of the intralayer atomic planes, and thus largely influences the intrinsic properties of the material, as well as its electrochemical performance. A series of other structural features in MoS 2 -such as lattice distortion, structural modulation, and irreversible decomposition-are also observed during sodium ion intercalation. Furthermore, information regarding the formation of the SEI layer on the MoS 2 anode is not well elucidated. Therefore, a better understanding of the working mechanism and intercalation chemistry of MX 2 compounds is required for better NIB design.

Advances in science and technology to meet NIB anode challenges
In-situ/operando characterisation techniques are being developed to understand the mechanisms at work in these materials and improve their electrochemical performance. In-situ atomic force microscopy (AFM) using a microcell has been used to study the evolution of structural changes and formation of an SEI on MoS 2 . It was observed that mechanically exfoliated MoS 2 wrinkles upon sodiation at 0.4 V, and that the SEI layer forms quickly, before the intercalation of sodium ions. This measurement also showed that the SEI layer thickness on the MoS 2 electrode is ≈20.4 nm ± 10.9 nm [142]. In addition, aberration-corrected scanning transmission electron microscopy (STEM) was used together with in-situ x-ray diffraction (XRD) to track the phase transition in MoS 2 [141]. It has been demonstrated that in Na x MoS 2 , x = 1.5 is a critical point for reversible structural evolution. This critical point indicates that the structure of MoS 2 can be partially recovered if the concentration of intercalated sodium ions is less than 1.5 per formula unit of MoS 2 . However, its structure cannot be restored once it is decomposed into Na x S and metallic Mo. In-situ spectroscopy has been used to study the growth of Na dendrites and SEI formation on TMD-based anodes. Synchrotron-based soft XAS has been used to investigate discharge products by probing states near the Fermi level [143]. XAS has also been used to reveal the the local electronic and chemical structure of surfaces and interfaces. These types of analysis could be extended to 2D TMD electrodes to study local changes in their atomic and electronic structures during charging and discharging. Similarly, the Rutherford backscattering The inset shows the spacing between the restacked nanosheets, which could facilitate the intercalation and de-intercalation of Na ions. Reprinted with permission from [137]. Copyright (2011) American Chemical Society. Reprinted by permission from Springer Nature Customer Service Centre GmbH: Springer Nature, Nature Nanotechnology [138]. © 2015. spectrometry (RBS) technique can be employed to determine the composition and thickness of the SEI that forms on anodes during electrochemical cycling [143].

Concluding remarks
In summary, the field of 2D TMDs is in a nascent phase. While 2D TMDs are interesting, several key challenges remain. Furthermore, much of the work on Na-ion batteries has been done on the semiconducting 2H phase of TMDs. This phase is hydrophobic and lacks conductivity, resulting in sluggish ionic diffusion. In contrast, the metastable metallic phase of TMDs has been less studied. The exfoliation of metallic TMDs and the restacking of 2D nanosheets into electrodes are likely to provide electrodes with better intercalation efficiency and faster kinetics.

Status
OEMs currently lie at the periphery of rechargeable battery technology. However, with the increased emphasis on sustainability, electrochemically active organic molecules, which, in principle, can be derived from biomass or recycled from commercial waste, are the subject of increased interest [144]. Although the suitability of organic polymers as electrodes for alkaline metal-ion batteries was reported in the early years of secondary battery development, it is only during the last decade that research in this area has witnessed a resurgence, with the successful application of small redox-active organic molecules for reversible ion insertion in LIBs and subsequently NIBs, for example, disodium terephthalate (Na 2 C 8 O 4 H 4 ) and its analogues [145]. This recent progress has coincided with the growing applications of small organic molecules in the preparation of several classes of solid-state materials. The features of structural diversity and the ability to modulate their electronic properties on demand, make organic systems attractive candidates to explore for battery applications.
Redox reactions in these materials typically involve molecules that have an initial neutral electronic state, but then acquire a net charge [146]. This is either accompanied by the rearrangement of π-bonds in the backbone, or other mechanisms that stabilise the redox state. Moieties such as carboxylates are well suited for anodes, on account of their low-voltage redox reaction (figure 15(a)) [147]. Since bare organic molecules have high solubility in polar solvents, usually, such active molecules are considered for use in coordination compounds, which are relatively stable solid-state materials. Organic chemistry permits the tailoring of the the local molecular environment to alter electrochemical features, such as the working voltage of the material and the density of active sites. Also, over the last few years, new redox chemistries and active moieties, such as azo (figure 15(b)) and imine have been identified and tested for charge storage at lower voltages. Although their working principle is similar to that of carboxylates, these groups are usually the central part of the molecule, unlike carboxylates, which tend to be terminal groups. Apart from these well-studied mechanisms, a few organic systems have shown reversible ion insertion using different pathways [147]. In addition to the goal of practical implementation, there is an interest in developing a fundamental understanding of the unexplored active groups.

Current and future challenges
Organic anode materials have considerable room for improvement, in terms of bothelectrochemical properties and material development (figure 16). Organic molecules show redox activity at specific active sites, which leaves a significant portion of the molecule unused. This directly contributes to lower theoretical capacities. An important aspect of the study of these materials is the improvement of cycling stability. The use of organic electrolytes in non-aqueous batteries leads to the formation of inactive side products or the solubility of small, active organic molecules for long-term cycling. The formation of a solid-electrolyte interphase (SEI) gives notably low coulombic efficiencies for the initial cycles, which is particularly relevant for practical applications [148]. Another feature of OEMs is their low density, resulting in significantly lower volumetric energy density. However, making their structures compact may result in an increase of volume expansion during (de)insertion of bulkier species, such as Na + ions. Most organic materials have poor electronic conductivity, requiring significant amounts of conductive carbon either as physical mixtures or as composites with the active solid, which directly adds to the non-contributing weight of the electrode. Although there is immense scope to modulate structural features, the carbonyl and imine groups only accommodate one Na + ion per site, and the stabilisation of the redox state requires a conjugated backbone ( figure 15(a)). On the other hand, azo bonds can insert two Na + ions (figure 15(b)), but their working potential is seen to be much higher, which reduces the overall energy density of the battery.
Apart from electrochemical performance, an in-depth understanding of the activity and the factors that influence that activity are yet to be fully realised. These are crucial to meet the upcoming targets for the applications of rechargeable batteries. In this context, the production costs and scalability of this class of materials also have room for more research. In addition, the thermal and chemical stability of organic molecules is limited and requires more development to support practical implementation.

Advances in science and technology to meet challenges
A major theme of research in this area has been to leverage the versatility of organic chemistry for molecular engineering and examine new molecules by varying their atomic or structural properties (figure 16). Among various molecules, carboxylates bonded to aromatic systems are the most popular moieties. Strategies to extend π-conjugation or vary the secondary functional groups have shown the potential to both improve cycling performance and stabilise redox states. In certain cases, upon careful design, the extended π-conjugation has participated in the electrochemical reaction, leading to the insertion of excess charges [149]. The role of secondary functional groups is to modulate the electronic properties of the molecule and provide additional interaction sites for Na + ions. It is observed that in structurally similar molecules, electron donor groups lower the working voltages, while electron-withdrawing groups tend to increase the redox potentials [150]. Heteroatom (N-atom) substitution in the backbone has resulted in higher capacities and smoother kinetics by virtue of favourable interactions with Na + ions. In systems with sulphur (S) atoms, the capacity is seen to increase on account of the additional interaction sites, and the electronic conduction is also improved [151]. Apart from molecular features, bulk properties are known to affect the cycling performance of electrode materials. The morphology and particle size of the active organic material are seen to play significant roles in the practical capacities and kinetics of ion insertion [152].
Although atomic features lead to the modulation of electrochemical properties, even the conventionally studied molecules still show issues of cycling stability and volume expansion. In recent years, various porous materials displaying long-range order based on redox-active small organic molecules have been tested with the aim of circumventing the issues of solubility and stability. These include metal-organic frameworks (MOFs) and covalent organic frameworks (COFs), which are seen to be stable to cycling and can accommodate bulky Na + ions [153]. However, research into porous materials has largely been confined to the preparation of electrochemically active hierarchical carbonaceous matter; and very few examples have used the pristine materials.

Concluding remarks
Organic battery materials are likely to contribute significantly to the pursuit of sustainable technologies. As anodes in NIBs, small organic molecules with redox-active groups such as carboxylate, azo, and imine have shown much promise. Several aspects of these compounds, such as their cycling stability, volumetric densities, and poor electronic conductivity, require more research to compete with benchmark materials. Much progress is anticipated through the incorporation of the knowledge gained and the extensive testing of novel solid-state materials based on the use of organic molecules as battery electrodes. The rich diversity of organic chemistry presents an opportunity to explore untapped redox-active groups and expand our understanding of molecular engineering to support the delivery of high-performance materials. In addition to fundamental objectives, practical parameters such as scalability and production costs also need to be addressed.

Status
In the last two decades, advances in supercomputing architectures and the development of more accurate and cost-effective quantum chemistry approaches have enabled the computational exploration of new materials to take place in a more rapid and rigorous manner. DFT calculations have emerged as a key component in the computational discovery of materials. To accelerate the discovery of materials in such a vast chemical compositional space, numerous DFT-based approaches have been reported and can be classified into three types [154]: (a) chemical substitution, (b) crystal-structure prediction, and (c) data mining.
Most new materials have been discovered by chemical substitution, i.e., a simple element replacement in known structural prototypes. This is particularly effective for Na-ion battery (NIB) materials, as these materials share similar atomic structures with their LIB counterparts. Representative examples include layered oxide cathodes of Na x TMO 2 (TM = transition metals) and solid electrolytes of Na 10 MP 2 S 12 (M = Si, Ge, Sn). In contrast, crystal-structure prediction, e.g., random structure prediction and evolutionary algorithms, begin with random atomic configurations within sensible constraints on bond lengths, followed by local optimisations that find the lowest-energy structures. In general, evolutionary algorithms should offer better efficiency, as they improve candidate guesses over time by comparing the lowest-energy solutions found at every iteration. These two methods have proven to be effective in predicting the intermediate phases of alloying electrodes of NIBs, such as Sn and Sb [155].
Data mining has now emerged as a promising method for facilitating materials discovery, as it can establish criteria for new material formations and construct quantitative structure-property relationships for predicting their properties. In conjunction with high-throughput algorithms, this approach has recently been applied to predict new potential cathodes and solid electrolytes for NIBs [156]. However, due to the high dimensionality of the problem and the relatively small number of known battery materials in databases, the predictive ability of the results obtained remains controversial. Continuous research effort has been devoted to improving the accuracy of models [156]. All the structure-prediction methods mentioned above can be supported by various open-source structure-prediction tools (figure 17).

Current and future challenges
While various structure-prediction methods exist, the exploration of NIB materials has mostly been limited to structures with existing prototypes reported in LIBs. This is because of the difficulties involved in predicting previously unknown structures, which is especially tough for NIB materials that have characteristic features, including: Large elemental phase space The exploration of potential NIB materials often faces increased complexity necessitated by the large number of elemental combinations, which are sometimes even more numerous than their LIB counterparts. For instance, Na x TMO 2 can feature a plethora of redox-active transition metals (TMs) (Ti, V, Cr, Mn, Fe, Co, Ni, and Cu) compared to Li x TMO 2 where only Co-, Ni-, and Mn-redox are available [158]. Furthermore, continuous efforts to optimise their stacking sequence (e.g., P2-and O3-type) have introduced various dopants (e.g., Ni, Cu, Mg, Zn) [158], making the compositional space much more complicated. Such broad elemental combinations lead to a significant number of candidate chemical formulae that need to be considered for the prediction of high-performance cathodes.

Materials stable only at elevated temperatures
Among the various synthesis routes reported to date, high-temperature sintering has become one of the most common ways to fabricate electrodes and solid electrolytes with high Na + conductivities. The resultant structures are often only stable at elevated temperatures. The discovery of high-temperature metastable materials is challenging with standard DFT calculations alone, as they do not include the vibrational entropic contributions or the associated temperature effects. To predict materials at a finite temperature, it is therefore Figure 17. Schematic of the procedures used for computational materials discovery in previous studies. Publicly accessible tools for structure prediction are also listed for reference. Reprinted figure with permission from [157], Copyright (2008) by the American Physical Society. necessary to perform additional phonon calculations that can describe atomic vibrations, which carries a significant computational cost.

Strong self-interaction errors in correlated TM oxides
Anionic redox in TM oxides has emerged as a new paradigm to increase the energy density of NIB cathodes. However, accurate prediction of the redox activity of TM oxides has been challenging for standard DFT calculations. This is because conventional generalised gradient approximations of DFT cannot correct the self-interaction error in O and, in particular, in TMs with highly correlated d-orbitals [159]. To counteract the self-interaction error inherent in DFT, the use of computationally intensive methods, such as the DFT + U approach or hybrid density functionals, is essential.

Advances in science and technology to meet these challenges
Advances in computational resources and theoretical methods have progressed greatly recently, such that the difficult challenges of discovering novel NIB materials are now becoming tractable. There have been important developments in the consideration of temperature effects, such as the quasi-harmonic approximation and the highly efficient ab initio molecular dynamics [160]. Another recent direction in the study of temperature effects has targeted many-body approaches, such as the path-integral Monte Carlo and self-consistent harmonic approximations. The application of these methods provides a route for evaluating the thermal properties of materials, allowing the prediction of potential NIB materials that are stable at high temperatures. In order to accurately describe strongly localised d-orbitals, various novel functionals (e.g., the DFT + U approach and hybrid density functionals) have been suggested. To fully remove the self-interaction errors, many-body approaches, such as quantum Monte Carlo and random-phase approximation, are becoming achievable using today's supercomputing architectures.
Challenges remain, however, in exploring the potential phase space of multicomponent Na-ion systems, which is still prohibitively large to search fully. Therefore, to provide accurate descriptions of materials and at the same time maintain a competitive computational cost, researchers should take inspiration from the generations of chemical knowledge in the battery arena. The use of chemical heuristics has been shown to have great potential for narrowing down phase space. For example, the constraints of charge neutrality and electronegativity balance can significantly reduce the total number of candidate materials by two orders of magnitude [161]. Recent efforts to reduce the number of candidates have developed several less expensive and less accurate ways to estimate the properties of NIB materials: Mulliken electronegativity and the solid-state energy scale were used to estimate the electrochemical stability window (ESW) [161], Voronoi-Dirichlet partitioning and the bond valence sum were used to estimate the ionic conductivity [162], and 'cationic potential' was used to estimate the stacking sequence of cathodes [163].
Another persistent issue in battery materials is the environmental sustainability of the constituent elements, of which the drive to reduce expensive and toxic Co in LIBs is the prime example. Therefore, further improvements in the NIB arena should also be driven by using only earth-abundant and non-toxic elements, with an eye to our responsibility to be environmentally friendly.

Concluding remarks
Recent advances in computational methods and increasing computational power have provided researchers with a new toolkit with which to explore a vast chemical space for targeted materials discovery. The systematic integration of these tools can pave the way for the discovery of future NIB materials ( figure 18). Major challenges remain, however, in the assessment of the properties of NIB materials. Despite the various computational approaches developed to date, calculations for structures that can only form at high temperatures or strongly correlated NIB materials are still cost intensive. This severely slows down the exploration of chemical space, which hampers the prediction of new game-changing NIB materials. The development of new strategies to circumvent this bottleneck will thus be critical.

Status
NIBs offer a cheaper and more sustainable large-scale energy storage solution, compared to well-established LIBs. However, contrary to the latter, which primarily use 1 M LiPF 6 in ethylene carbonate/dimethyl carbonate (EC/DMC) as a standard electrolyte solution, a leading electrolyte is yet to be established for NIBs [165][166][167][168]. This can be explained in part by the infancy of the field as well as the fact that electrolyte development is overlooked in favour of electrode research. Nevertheless, there is an urgent need to find high-performance electrolytes for NIBs, since they play a leading role in the overall lifetime, rate capability, and safety of the battery. Principally, this is due to the formation of a solid-electrolyte interphase (SEI) at the anode, which results from the degradation of the electrolyte and/or solvent during the charging/discharging process. Although well characterised for LIBs [168], the SEI composition for NIBs is poorly understood and remains one of the most critical areas for future research efforts.
To be an effective organic electrolyte, the sodium salt must adhere to a number of criteria: (a) chemical, electrochemical, and thermal stability; (b) high ionic conductivity; (c) low cost; (d) low toxicity/environmental benignity, and (e) simple, scalable synthesis. High ionic conductivity promotes facile, rapid exchange of sodium ions between the two electrodes, whereas a large highest occupied molecular orbital (HOMO) energy level ensures the stability of the salt. Such strict requirements naturally limit the possible choices of salts available; the most widely used anions are shown below, in figure 19. Traditionally, these anions have strong electron-withdrawing groups surrounding the central atom, which create a delocalised negative charge, and they are weakly coordinated to the Na + cation. Generally, sodium salts have higher melting points than their lithium counterparts, which is advantageous not only for easier drying but also safety, since the former have greater thermal stability.

Current and future challenges
Among the electrolytes shown in figure 19, the most-studied electrolyte in NIBs is NaClO 4 [167] although its strong oxidising nature prohibits its commercial use due to safety concerns. In addition, NaClO 4 is known to be difficult to dry. Similarly to LIBs, the most promising salt to date for NIBs is NaPF 6 ; however, its high moisture sensitivity can promote the formation of HF via hydrolysis, leading to further electrolyte degradation and major safety concerns. As a result of this moisture sensitivity, the quality of commercially purchased NaPF 6 is problematic, and NaF formation occurs. NaBF 4 and NaOTf (OTf = trifluoromethanesulfonate) are poor choices, as they have strong cation-anion interactions, and it has been shown that NaOTf has reduced conductivity and electrochemical stability compared to NaClO 4 and NaPF 6 [164].
An interesting emerging class of electrolytes which has previously been used in ionic liquids contains NaTFSI (TFSI = bis(trifluoromethane)sulfonimide) and NaFSI (FSI = bis(fluorosulfonyl)imide), which are attractive due to their reduced toxicity and high thermal stabilities. However, these cannot be used as single salts, as they induce corrosion on the Al current collector, as well as incurring high cost. It is rare for new electrolytes to enter the field, but the DFOB (DFOB = difluoro(oxalato)borate) and BOB (bis(oxalato)borate) anions have recently both shown promising results in terms of reduced toxicity and low cost of synthesis [169]. Additionally, NaBOB has proved to be a non-flammable electrolyte when used in trimethyl phosphate (TMP) solvent; but in both cases, more work is required to optimise their performance [170]. Clearly, developing a salt that meets all the requirements outlined above is challenging, and future research efforts should focus on anions that have easily tuneable groups, which would allow for the fundamental properties of the salt to be simply modified.
Another challenge is to characterise the SEI. Few studies exist to address this, but x-ray photoelectron spectroscopy (XPS) and time-of-flight secondary-ion mass spectrometry (ToF-SIMS) analysis have confirmed the presence of Na 2 CO 3 and ROCO 2 Na species [171]. These degradation salts are often more soluble than their lithium counterparts, and the Na + inductive effect on the solvents/anions should be weaker than for Li + . This could, in turn, affect the reduction products formed at the anode.

Advances in science and technology to meet challenges
In recent times, the use of computational modelling and simulation has become a much more powerful method for the development of high-performance electrolytes. The early modelling of organic liquid electrolytes often focussed on the conduction mechanism, with little investigation into how the bulk ions themselves interacted. However, with advances in modern computer power, more recent studies have now started to quantitatively address the role of ion pairs, although currently, the high-level theory is too demanding for this to be a feasible method for the fast screening of electrolytes. In addition, computational methods can assist in better understanding the mechanism of SEI formation, where calculations have indicated that sodium electrolytes have faster desolvation of the cation at the electrolyte/electrode interface than that of Li + in analogous systems. DFT may also be used to simulate spectra (such as IR and Raman) of the degradation products in the SEI to aid the identification of the species present [172].
In addition to computational developments, there have been spectroscopic advances which may be utilised to address the challenges described above. Modern advances in dynamic nuclear polarisation-enhanced NMR spectroscopy (DNP-NMR) have recently been used as a method to aid characterisation of the SEI in LIBs. By using lithium metal as the polarisation source, selective observation of the low-concentration SEI components was possible. An intimate knowledge of the SEI composition and distribution is crucial in understanding how dendrites form, and is critical for preventing short circuits and improving the overall safety of the battery. A major advantage of this technique is that it allows for the direct probing of interfaces to reveal structural information, and it can be envisioned that this DNP-NMR spectroscopy technique can readily be applied to NIBs to provide further insights into SEI formation and composition [173]. While there have been several investigations into the structure of the SEI for NIBs, more work is necessary to understand the interplay between electrolyte chemistry, SEI composition, and stability.

Concluding remarks
In contrast to LIBs, which are a mature technology, much work is still required to unleash the full potential of NIBs. Electrolytes play a key role, as unwanted side reactions and degradation products significantly affect the overall performance and lifetime of the battery. The current range of sodium electrolytes has significant drawbacks, such as oxidising potential, moisture sensitivity, cost, and toxicity, meaning there is an urgent need to develop novel electrolytes to overcome these shortcomings. With increased research into, and development of, novel electrolytes, these challenges can be addressed, and this roadmap highlights key areas where focus should be applied. Principally, this is in characterising the degraded salts in the SEI as well as producing electrolytes which can easily be chemically modified to tune their fundamental properties. With this increased focus, NIBs undoubtedly have a prosperous future, rivaling that of their lithium counterparts.

Status
In many ways, both R&D and the practical deployment of NIB organic electrolytes follow the development of LIB electrolytes [175][176][177]. This is logical, as the operating conditions and application areas are more or less the same, and thus, similar properties are needed; i.e., readily dissolved salts to create a large concentration of charge carriers offering fast ion transport with high fluidity and low viscosity. The latter also assists in the proper wetting of the separator and the porous electrodes. Additionally, high thermal and chemical stabilities are a must, to allow for usage in various climates and vs. different electrodes, as are wide ESWs matching the electrodes. Finally, all must be available at a low cost. As of today, this means linear and cyclic carbonates, such as DMC, EC and PC [174], or ethers, such as glymes, in particular, diglyme [174,175].
While much of NIB R&D follows from earlier LIB developments, there are some distinct practical differences to be found. One is that linear carbonates, such as DMC, used to lower the melting point and viscosity of LIB electrolytes, create soluble products upon reduction in a NIB, and hence, unstable SEIs at the anode, a problem that can be overcome by smart additive usage [176]. Another is that graphite is not capable of intercalating naked Na + ions and thus, it is not commonly used for NIBs. Therefore, PC, one of the original LIB electrolyte solvents, is frequently used for NIBs (as opposed to LIBs, as it exfoliates graphite) [174,176,177]. There is also a fundamental difference between the use of Na salts and Li salt: for the former, both the ion-ion and ion-solvent interactions are reduced to ca. 80% of the latter [178]-which paves the way for faster desolvation dynamics [179] which, at the cell level, manifests as improved power rates [174,176,177]. As with any other battery technology based on organic electrolytes, improved safety without making any departure from performance is key-and here, new NIB electrolytes, in particularm solvents, should evolve out of the LIB shadow. Another driving force should be to simplify recycling-and this, for example, means the use of F-free solvents/electrolytes [170].

Current and future challenges
Even if challenged by both solid-state electrolytes (SSEs) offering performance and safety advantages and aqueous electrolytes offering lower cost and better overall sustainability, organic solvent-based electrolytes are still likely to be the main route for years to come. They can be used in their own right and/or in hybrid electrolytes, with polymers or ionic liquids, to improve performance and safety [180] or to facilitate better SSE/electrode contact. A major challenge is to create a holistic and Na + -centred (and not Li + -derived) approach to the salt(s)-solvent(s)-additive(s) recipe for a standard functional NIB electrolyte [181]. This could also reduce the complexity and offer better control and understanding of e.g., solubility issues and ion-transport advantages. Today, the latter is, at the phenomenon level, a power performance advantage compared to LIBs, which is coupled to lower solvent desolvation energies [179] and the presence of fewer contact ion-pairs [178]. However, the reduction in strong interactions renders the organic products from solvent reduction more soluble and thus less stable, and creates/results in additional inorganic SEIs. Another large challenge comes directly from the +0.3 V shift in E red of Na + /Na vs. Li + /Li. For the same final cell voltage, there is a corresponding more stringent demand on the electrolyte/solvent E ox limit-which is not a minor issue, given the interest in high-voltage cathodes. The fact that the electrolyte must be developed not only in terms of ESW, but also to match both positive and negative electrodes, is a large challenge, as there is not yet a dominant electrochemical couple for NIBs. Improved fluidity, important for low-temperature performance as well as for the proper wetting of electrodes, must be achieved without compromising safety by using volatile solvents. Another path of research is highly concentrated electrolytes (HCEs), which improve safety, widen the ESWs, and utilise non-vehicular ion transport, which leads to improved cation transference numbers. For HCEs, Na-based systems have an earlier onset of some of these properties, compared to Li-based systems [179], which is cost-saving, as less salt is needed. However, the high viscosities of the highest concentrations of salt-based HCEs, i.e., those of very low solvent content, are a challenge for manufacturing and the cell formation stages.

Advances in science and technology to meet challenges
When the key performance indicators (KPIs) and the targets to be reached for NIBs by e.g., 2030 are listed, few are directly dependent on the solvent used, but very many are indirectly limited by the electrolyte. The first, most obvious KPI is the ionic conductivity, but the target, 1 mS cm −1 , has already been reached. The next is the cell voltage-intimately coupled to the electrolyte/solvent E ox limit, as outlined above, while the less obviously connected KPIs are cycle-life and cost. However, for the NIB cycle life, the stability of the electrolyte/solvent, both in bulk and at the interfaces/interphases, is key. This is why different operando techniques enabling the study of NIB cells represent the most important technological advances needed. One such technology is based on embedding optical fibre sensors into the jelly-roll and, while it does not affect cell performance, it enables the monitoring of not only pressure and temperature changes as a function of the state-of-charge but also the origin of heat contributions and how different additives affect these properties, as elegantly demonstrated for a HC//NVPF NIB i.e. based on hard carbon as anode and Na 3 V 2 (PO 4 ) 2 F 3 as cathode [182]. As a baseline electrolyte, this study used commercially available NP30 (1 M NaPF 6 in EC:DMC 50:50), a direct analogue to the LP30 electrolyte, and standard 18 650 cells and protocols. This points to several things that are needed to advance the science and technology of NIBs; commercially available electrolytes of high purity at a cost meeting KPI targets; standardised cells and protocols, both for electrolyte and electrode studies (at present, a chicken-and-egg problem, as no fixed-star materials exist, and different aspects are in focus using different setups), and understanding the roles/actions of additives in NIBs, as these may not be the same as those in LIBs-the differences in SEI formation being the most obvious. If a proper understanding of the limitations imposed by electrolytes, test protocols, and the actions of the additives is pursued, the growing community of NIB researchers should soon have a wealth of unambiguous data. While there is room for rethinking and emphasising the differences rather than the similarities between Na + /NIBs and Li + / LIBs, NIBs should be using the field of LIB development to the maximum extent possible for real NIB leapfrogging.

Concluding remarks
While the R&D directions of NIBs have largely followed those of LIBs, they are slowly finding their own path, including applications where the emphasis is on complementarity rather than replacement. The somewhat overemphasised similarity of Na + and Li + dissolved in solvent(s) has, until today, promoted evolution, leaving NIBs always trailing behind LIBs. However, as an example, it should be possible to use the weaker solvent interactions and generally larger first solvation shells of Na + as design criteria for novel NIB electrolytes. Likewise, the problematic large solubility in the electrolyte of solvent reduction compounds should be a proper R&D target, similarly to e.g., the polysulphide issue in Li-S batteries. As we are seeing various operando, high-throughput, and robotic screening technologies starting to be applied for battery research, these could all be used to evaluate NIB electrolyte solvents and perhaps also enable the tailoring of the electrolyte to different operating conditions and applications.

Acknowledgments
The funding received from the European Union's Horizon 2020 research and innovation programme under Grant Agreement No. 646433 (NAIADES), the Swedish Research Council, the Swedish Energy Agency (#37671-1), and the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS), are all gratefully acknowledged. The many fruitful discussions within ALISTORE-ERI, and especially with M Rosa Palacín, have been most valuable. P J is also grateful for the continuous support from several of the Chalmers Areas of Advance: Materials Science and Energy.

Status
NIBs have attracted attention in parallel to Li-ion technology due to their cost-effectiveness and the high earth abundance of sodium [71,183,184]. Despite the commercialisation of NIBs by several companies (Faradion [185], Tiamat [185,186], SHARP labs [185,186]), thermal runaway with volatile electrolytes and insufficient cyclability at room temperature demand more serious electrolyte research.
Ionic liquids (ILs) are promising alternative electrolytes due to their non-flammability and wide operating electrochemical window. They are glass-forming molten salts, consisting of versatile combinations of aromatic cations and weakly coordinating anions. The structures of the ions and their interactions (hydrogen bond, electrostatic, and Van der Waals) [187] govern their physicochemical properties, (phase transitions, viscosities, thermal stability, and HOMO-LUMO (lowest unoccupied molecular orbital) gap) [187] and electrochemistry (electrochemical window, ionicity, ionic conductivity, ionic association/dissociation, and the double-layer structure at the electrode-electrolyte interface) [187,188]. In the realm of NIBs, ILs offer specific additional advantages, such as reduced corrosion of the Al current collector, a wide range of operating temperatures, and suitable electrode-electrolyte interface processes (homogeneous SEI formation, a conductive interface, and a dendrite-free anode-electrolyte interface) [185] (figure 21). The wider electrochemical window of ILs permits a wide variety of cathode materials for NIBs. Only anion decomposition defines oxidative stability and SEI composition. ILs offer a less resistive, more uniform, and consistent SEI throughout the battery cycle with hard carbon and alloy-based anodes [186]. The composition of an IL-derived SEI (with an HC anode) (SEIs formed with polyolefin, SCO) is distinctly different from those formed in organic electrolytes (SEIs formed with organic and inorganic sodium carbonates) [186]. Reduced aluminium corrosion in ILs is notable, even with salts that corrode the Al current collector in organic solvents [186]. Several imidazolium-, pyrrolidinium-, phosphonium-, and ammonium-cation-based ionic liquids have been explored in NIBs (e.g., [ CF 3 ]. C n C 1 im + : 1-alkyl-3-methylimidazolium, C n C 1 pyrr + : N-alkyl-N-methylpyrrolidinium, C n C 1 pip + : N-alkyl-N-methylpiperidinium, and N nnnn + : tetraalkylammonium) [186].

Current and future challenges
Despite several advantages, NIB research is falling behind the current state of the art of LIB technology. Safety and the battery cycle life are significantly influenced by the electrolyte formulation and various interfacial issues, including uncontrolled dendrite formation due to the use of sodium metal, unstable and increasingly resistive electrolyte interface SEI formation due to electrolyte decomposition, cracked and thicker SEIs, pronounced dissolution of SEIs, and the significant reactivity of sodium metal in the majority of volatile organic electrolytes. Ionic liquids serve as green solvents. They mitigate safety issues and offer wider oxidative stability and relatively lower solvent degradation at the electrodes. However, the major challenges of ionic liquids are their relatively high cost, tedious purification methods, and the hygroscopic nature of widely used imidazoliumand pyrrolidinium-based ionic liquids. This is despite their excellent stability and wider temperature operability. Their relatively high viscosity and inferior ionic conductivity limit their battery cycling performance at room temperature, as compared to conventional organic electrolytes. The ionic liquids offer larger interfacial resistance and slower Na + transfer at the sodium metal anode, compared to organic electrolytes. In NIBs, the advantages of ionic liquids are only fully utilised at elevated temperatures (∼90 • C) in half cells. Another important challenge with ionic liquids is to achieve a unity sodium transference number and reduce concentration polarisation. The majority of ionic liquid electrolytes are made at a maximum of 1 M concentration, a concentration at which unity transference numbers are difficult to achieve. Recently, super-concentrated ionic liquid-salt systems have been shown to offer relatively stable SEI properties and near-unity transference numbers. However, the availability of a wider range of IL structures with high salt dissolution capability is still limited. Designing electrolytes to achieve high transference numbers and reasonably good ionic conductivity is an unavoidable challenge for ILs in NIBs. Although minimal, the current-collector corrosion is not fully mitigated by the widely used fluorinated anion-based ionic liquids.   Figures 22(b) and (c) present a survey of the battery performance of half cell (with sodium metal as the anode) and full cell (with a sodium intercalation compound as the anode) NIBs, respectively, employing pyrrolidinium-based IL electrolytes with a wide range of cathodes and anodes. Impressively, ionic liquids offer superior cyclability, rate capability, and powder density, compared to organic electrolytes. Despite the higher viscosity and inferior transference number, these results strongly encourage the investigation of IL chemistry in NIBs. The reasons for such exciting performance improvements are: the enhanced Na + diffusion at the interface, a compact SEI, dendrite-free sodium deposition/dissolution, the lower reactivity of ionic liquids with sodium metal, and reduced Al corrosion at elevated temperatures ( figure 21) [190]. Full cells employing an IL electrolyte display similar benefits ( figure 22(c)). A very energy-dense (368 Wh kg −1 ) prismatic full cell has been developed using a Na[FSA]-[C 3 C 1 pyrr][FSA] electrolyte, a Na 3 V 2 (PO 4 ) 3 cathode, and an HC anode to deliver better capacity retention (75%) than that of a Na[ClO 4 ]-EC/PC electrolyte ( figure 22(c)) [191]. A highly stable passivation layer with negligible dendrite formation was observed when a vinyl substituted imidazolium IL was used as electrolyte additive [192]. Super-concentrated 50 mol% NA[FSA]-[C 3 C 1 pyrr][FSA] exhibited dendrite-free sodium deposition/dissolution onto the SEI with a nanoengineered Na + conducting interface made from (Na) x (FSA) y components [187].

Concluding remarks
In conclusion, ionic liquid electrolytes significantly improveme the safety of sodium ion batteries, widen the operating temperature and electrochemical window, and offer impressive cyclability and interface chemistry. However, insufficient literature is available to report the applicability of a wider range of electrodes in a wider range of ILs. Fine correlations between structure and critical interface electrochemistry are still under-represented in NIB research. Some additional concerns include cost, recyclability, tedious purification methods, and the high viscosities of ILs. Binary IL-organic/aqueous solvents (IL-water) and non-fluorinated cheap anion-based ILs could improve cost-effectiveness and decrease viscosity if they could be developed without sacrificing safety aspects. Interestingly, NIBs with ILs still exhibit good specific energy (near to that of LIBs) and better power than supercapacitors at elevated temperatures, which are difficult to achieve with organic solvents. The possibility of tuning the molecular structure of ionic liquids and a detailed understanding of the mechanical properties and solubility of SEIs with ILs are to be explored to enhance the performance of NIBs with ILs.

The SEI in Na-ion batteries Andrew J Naylor and Reza Younesi
Department of Chemistry-Ångström Laboratory, Uppsala University, Box 538, SE-75121 Uppsala, Sweden

Status
An SEI may be formed through the decomposition reactions of an electrolyte at the interface with the anode of a battery [193]. The low electrode potential experienced by the anode causes the polar aprotic non-aqueous electrolytes employed to be thermodynamically unstable. The formation of a durable SEI is, however, essential for a kinetically stable battery and to obtain a long cycle life. This nm-scale interphase layer often has the characteristics of a solid electrolyte and can comprise both organic and inorganic, amorphous, and crystalline products derived from the electrolyte components. Ideally, it passivates the surface of the anode, being highly electronically insulating, impermeable to solvent molecules, insoluble, and inert, but offers ionic conductivity to allow the diffusion of cations (Na + ions in an NIB) between the electrode material and the electrolyte. The SEI should preferably be formed only within the initial cycles to prevent capacity losses and electrolyte depletion.
It has been shown through photoelectron spectroscopy (PES) measurements that the SEI in NIBs can vary in thickness and composition, depending on the state of charge [113]. The composition of the electrolyte is the main factor that influences the SEI composition and properties (figure 23), though other parameters including electrode-surface composition, current density, temperature, and potential limits can have an impact. The typical decomposition species formed from carbonate-based electrolytes in NIBs can be divided into inorganic species (mainly NaF and Na 2 CO 3 but also Na 2 O and NaOH in the presence of oxygen and water contamination) and organic species (sodium alkyl carbonates (ROCO 2 Na) and polyethylene oxide oligomers) [194,195].
Despite the fact that the SEI has been extensively studied in LIBs, resulting in the tuning of electrolyte formulations to provide maximum performance, it is still thought to be one of the major contributors to battery ageing [168]. However, while the concept of the SEI and its characteristics are largely similar for Na-ion batteries, many notable differences exist between the two technologies, even though they often use highly analogous materials [166]. A fundamental understanding of SEI formation mechanisms and properties is required to design an electrolyte that will lead to the commercialisation of NIBs. Here, we address the challenges of the SEI in NIBs, which can have such an important impact on their performance and ageing.

Current and future challenges
In NIBs, the main challenge that needs to be to overcome to stabilise the SEI is its solubility [197,198]. While SEI dissolution has not been studied in great detail for LIBs, it is now a major focus of NIB research. The continuous dissolution of the SEI leads to its reformation in subsequent cycles with excessive irreversible capacity losses, poor coulombic efficiencies, and eventual failure. Because the charge density and hence the dissociation energy of sodium salts is lower than that of lithium salts, the solubility of SEI components (e.g., NaF vs. LiF) formed in NIBs is generally greater compared to that in LIBs, and this holds especially true for aqueous solutions [196]. Extensive SEI dissolution has been demonstrated for carbonate-based electrolytes and carbonaceous electrodes [197]. It has been determined that capacity loss through such dissolution is dependent on electrolyte composition and electrode surface area. It has even been debated whether the most soluble organic species can be considered to be SEI components at all [194,195].
Additional challenges in characterising the SEI arise due to its high solubility. Most commonly, spectroscopic and imaging analyses are performed ex situ on electrodes after the drying/washing of samples. Such practices may inadvertently alter the composition of the SEI. The reports of an overall more inorganic SEI for NIBs compared to LIBs [199] could also be caused either by the leaching of organic species in the cells or their dissolution during the washing step.
The formation of high-resistance carbonates on the surface of sodium metal in half cells with continuous stripping/deposition can lead to a complex evolution of resistance. The lack of a reliable reference electrode and the consequential use of sodium metal in half cells results in poor SEI formation at the working electrode, and hence, inferior performance. The diffusion of decomposition products from the sodium electrode via the electrolyte to the working electrode ('crosstalk') is probably responsible. These issues prevent the extrapolation of performance from half cell to full cell; however, no such problems have been reported for full-cell configurations [196]. Nevertheless, information from studies published using half cells can still provide valuable insights and lead to a greater understanding of the challenges at hand.

Advances in science and technology to meet challenges
The low thickness and chemical complexity of the SEI in any battery system make it particularly difficult to characterise. Numerous techniques have been used, including PES, electron microscopy, electrochemical  [196]. Reproduced from [196]. CC BY-NC-SA 3.0.  [198]. Copyright (2016) American Chemical Society. quartz-crystal microbalances (EQCMs), electrochemical impedance spectroscopy (EIS), time-of-flight secondary-ion mass spectrometry (ToF-SIMS), and vibrational spectroscopies. However, innovative methods and advanced characterisation techniques are required to gain a deep understanding of the highly soluble SEI in NIBs.
Electrochemical 'pausing testing' can be used to investigate dissolution [197,198], whereby a cell is switched to its open-circuit voltage (OCV) for a fixed duration after a period of cycling to determine how much capacity is lost through self-discharge (figure 24). Such experiments have shown that the capacity loss is higher in NIBs than in LIBs.
PES is very sensitive to surfaces and chemicals and is one of the techniques most commonly used to study SEI layers in all kinds of battery systems. Synchrotron hard x-ray photoelectron spectroscopy (HAXPES) measurements can be used to probe to a depth of 50 nm and have shown that the SEI thickness decreases with increasing pause durations (figure 24) [197,198]. HAXPES measurements with large probing depths may allow for the elimination of electrode washing, an issue already discussed, although other parameters, including the pre-disassembly relaxation time, should also be considered. The future development of in situ and operando PES techniques may enable the direct measurement of SEI layer formation at an electrode during cycling.
While infrared spectroscopic (FT-IR) analysis of the electrolyte composition from cycled cells has already been demonstrated [195], it is underutilised for detecting soluble SEI species. Complementary techniques, such as gas chromatography mass spectrometry (GC-MS) and inductively coupled plasma (ICP) methods may also provide valuable insights. However, the extraction of electrolytes from cycled cells and data interpretation can present challenges.
The development of novel electrolyte components (salts and solvents) will almost certainly result in the formation of more stable SEIs and improved battery performance. However, given such development, one must bear in mind that cost-effective, non-toxic, and safe electrolytes will be required for the commercialisation of NIBs.

Concluding remarks
While relatively little is known about SEIs in NIBs and their formation mechanisms, particularly compared to their lithium counterparts, there is a growing body of knowledge on the topic as a result of increasing interest over recent years. Progress is inhibited, though, by the challenges posed by their complex nature. Their high solubility compared to the SEIs in LIBs causes difficulties in determining their composition and in preparing samples for ex situ characterisation. Furthermore, the lack of a suitably stable reference electrode and the issues of high resistance and crosstalk when employing sodium metal in half cells lead to complications in assessing the performance of materials. Novel electrolyte formulations will advance NIB technology towards commercialisation. However, a robust comprehension of the SEI achieved through the use of advanced characterisation techniques and innovative electrochemical testing methods will allow for the targeted design of electrolytes and a stable SEI.

Status
An ideal solid-state sodium electrolyte should deliver minimal ohmic losses and (electro)chemical stability with respect to both anode and cathode materials. Both of these parameters must be retained throughout the cycling processes and lifetime of the battery. Delivering such performance requires a combination of transport properties, chemical passivity, and processability, which represent a major challenge for materials science.
Sodium electrolytes have been a long-standing challenge, and Na + mobility is generally less facile than for Li + analogues. Conduction through solid polymers, such as Na salts in polyethylene oxide [200], can realise around 10 −4 S cm −1 at room temperature and the relatively low melting point of Na makes operation above ambient temperature more fraught than in analogous Li batteries. The increased size and mass of Na + compared to Li + are both implicated in the more sluggish kinetics of the heavier group 1 metal, but it should be recalled that the heavier Ag + ion shows superionic conduction in a range of materials. Here, the polarisability of Ag + offers an advantage over the harder Na + cation. This points towards the potential of using polarisable anions, such as sulphide, in building a lattice for fast Na + conduction.
SSEs based on the material Na 3 PS 4 have shown that the replacement of P 5+ by Ge 4+ , Ti 4+ , or Sn 4+ ions enables an increase in carrier concentration in the phases Na 3+x M x P 1−x S 4 . The introduction of dopants causes a tetragonal distortion away from the body-centred cubic parent phase, Na 3 PS 4 , leading to ordering of the Na ions. Unlike other ionic conducting systems, such as Li-stuffed garnets, this is not associated with a large decrease in Na + conduction, and computational simulations indicate that vacancy generation can be realised in the tetragonal phase, and that the conductivity of Na 3.1 Sn 0.1 P 0.9 S 4 is similar to that realised in solid salt:polymer composites [201]. The efficacy of the approach is further demonstrated by the room-temperature Na + conductivity of 4 cm −1 afforded in the related phase, Na 11 Sn 2 PS 12 , where a higher concentration of SnS 4 tetrahedral units delivers a complex three-dimensional pathway for Na + conduction involving all five crystallographic Na positions, and additionally, two low-energy interstices [202].
Similar conductivities have been realised in the NASICON structural family of oxide-based materials shown in figure 25, exemplified by series such as Na 1+x Zr 2 Si x P 3−x O 12 , where the carrier concentration is adjusted by the classic approach of using an aliovalent chemical substitution, in this case, replacing P 5+ with M 4+ ions. In Na 3 Zr 2 Si 2 PO 12 [203], the conductivity can be increased towards 4 mS cm −1 through control of the material's microstructure and densification processes.

Current and future challenges
The deployment of these materials into battery technologies faces the great challenge of interface management. Addressing this requires a better understanding of the role of electrolytes in interfacial processes. The minimisation of interfacial resistance is vital to avoid ohmic losses, which both reduce the energy storage capacity and lead to potentially hazardous heating. The challenge is to develop an interface which shows low resistance whilst retaining (electro)chemical stability during the multiple processes of battery cycling involving a change in potential and the (de-)swelling of electrodes. Meeting these challenges requires moving beyond the scale of material manipulation via crystal chemistry.
The NASICON phase has been explored via different synthetic routes, and these lead to variations in conductivity spanning an order of magnitude. An interesting development is the use of a 'fluoride-assisted' route to deliver a multi-phase sample with the composition Na 3 Zr 2 Si 2 PO 12+x NaF [204]. The presence of NaF leads to an increased grain size, which reaches a maximum for x = 0.7. As the grain size increases, the conductivity is enhanced from 0.23 × 10 −3 to 1.4 × 10 −3 S cm −1 as shown in figure 26. This is accompanied by a negligible change in the activation energy, indicating that the increase in grain size and the associated reduction in grain boundary volume is the likely cause for this enhancement in conductivity. Similar enhancements of conductivity up to 10 −3 S cm −1 have been achieved using liquid-phase sintering [205] to aid the densification of Na 3 Zr 2 Si 2 PO 12 at 900 • C rather than the 1200 • C necessary for conventional ceramic synthesis.
Many solid electrolytes have limited thermodynamic stability and react with metallic sodium. An examination of the interface between Na 3 Zr 2 Si 2 PO 12 and a metallic Na anode has shown that the reduction of the electrolyte is kinetically limited; the initial reduction of Zr and Si is self-limiting and the electrolyte can be plated and stripped with Na in a symmetric cell as long as the current density is kept below the 0.2 mA cm −2 limit for Na dendrite growth [203]. A relatively high charge rate of 2 C has been demonstrated using sulphide-based electrolytes Na 3+x M x P 1−x S 4 (M = Ge, Ti, Sn), although this was achieved using a Na 2 Ti 3 O 7 anode to avoid the problem of Na dendrite growth.

Advances in science and technology to meet challenges
The enhancement of electrolyte performance requires an intimate understanding of the mechanisms of Na + migration. A recent analysis of the effect of varying the Na content in the NASICON phases, Na 1+x Zr 2 Si x P 3−x PO 12 , has indicated a more complex distribution of Na than expected, which implies that ionic conduction must proceed via a correlated mechanism with an organised movement of multiple ions, rather than the more widely anticipated hopping of individual ions onto vacant sites [206]. These insights arise from combining computational simulations of molecular dynamics and advanced analysis of neutron scattering data to deliver insights that could not come from either computation or experiment alone. Similar complementarity is seen in the characterisation of the lattice dynamics of the novel perovskite-phase Na 1.5 La 1.5 TeO 6 via both impedance and muon spin relaxation [207]. This combination of bulk and local probes shows how the local barrier for Na + migration is significantly lower than that indicated by the total impedance, which suggests that chemical modification may deliver useful increases in conductivity in this structural family.
These kinds of detailed insight need to be extended from isolated electrolyte materials to look at the interplay of chemistry and microstructure in delivering stable interfaces. This can be illustrated by two differing approaches to using composites to deliver solid-state batteries capable of extended cycling lifetimes.
A solid-state battery with excellent stability at up to 10 000 cycles, even at the high charging rate of 10 C has been realised by wetting the electrolyte interface with ionic liquids [208]. The solid electrolyte is an intimate composite of Na 3 Zr 2 Si 2 PO 12 combined with adventitious Na 3 La(PO 4 ) 3 introduced by an unsuccessful attempt to substitute La 3+ into the Zr 4+ site. The introduction of this impurity phase increases the density of the ceramic and so the presence of the passive impurity counterintuitively enhances the conductivity of the electrolyte composite beyond that of the pure Na 3 Zr 2 Si 2 PO 12 phase. The capacity of 90 mAh g −1 from the Na 3 V 2 (PO 4 ) 3 cathode combined with a metallic sodium anode is modest, but the stability and lifetime of the device show the efficacy of this approach for interface management. A similar capacity can be realised by casting ceramic particles in the NASICON phase in a polyvinylidene fluoride-hexafluoropropylene polymer to combine the conductivity of the ceramic with the mechanical properties of the polymer [209]. This solution-cast composite delivers a flexible electrolyte with a room-temperature conductivity of 2.25 × 10 −3 S cm −1 , which matches that of the pure ceramic, whilst maintaining the ease of processing associated with the polymer.

Concluding remarks
Both oxide-and sulphide-based Na + electrolytes are now reaching the desired conductivity to enable solid-state batteries. The challenges in developing viable battery technologies lie in managing the interfaces between the solid electrolyte and the electrodes, and recent proof-of-concept studies have demonstrated the efficacy of an electrode/electrolyte composite approach. These examples serve as useful demonstrations of how chemical synthesis and materials processing must proceed in tandem to deliver cycling longevity from the management of electrolyte interfaces. Understanding the roles of the components and how they stabilise the interfaces will require a coordinated synthetic, experimental, and computational approach. There will also be a key role for operando measurements of crystal structure, lattice dynamics, and ion mobility in developing long-term electrochemical cycling in an all-solid-state sodium battery.

Status
For all battery chemistries, a reliable electrochemical setup is essential. Half-cell configurations with typically two or three electrodes, using lithium metal as the counter electrode (CE) and reference electrode (RE), are standard setups for Li cells. However, several requirements need to be fulfilled to achieve reliable results (figure 27) [210,211]. The CE must have high capacity, fast reaction kinetics, and should not contaminate the electrolyte solution with reaction byproducts. Meanwhile, the RE should ideally be nonpolarisable (a current flow should not affect its potential) and should have a reliable and stable potential. Although these conditions are satisfactorily met in Li half-cells, several studies point to various issues associated with the use of Na-metal CEs and REs [196,212,213].
Most of these issues are rooted in the low stability (and high solubility) of the SEI formed on the Na metal electrode (see the 'electrolyte/interface, SEI layer' section, by Reza Younesi). Considering the difficulty of developing new battery chemistries, in which both electrodes and electrolytes must be studied in parallel, the use of reliable electrochemical protocols is crucial in order not to discard potentially interesting material candidates or to make erroneous interpretations of the results. In this section, the most common issues arising from the use of Na metal CEs and REs are described, and recent strategies to overcome them are introduced.

Current and future challenges
The instability of the Na metal-electrolyte interface has several consequences for any electrochemical test involving Na metal as both the CE and the RE. First, the constant SEI dissolution/reformation results in the contamination of the electrolyte with soluble/gaseous species. When alkyl carbonate-based solutions are in contact with Na metal, a growing amount of CO, CO 2 , methane (CH 4 ) and ethylene (C 2 H 4 ) is generated ( figure 28(a)). The presence of soluble compounds, such as dimethyl 2,5-dioxahexane dicarboxylate (DMDOHC), has been observed in the solution, even after 5 d [213]. While the nature/amount of such contaminants is dependent on the electrolyte formulation, their impact on the working electrode interface and electrochemistry is clear. For instance, the impedances of hard carbon (HC)||HC (2-electrode) and HC||Na (3-electrode) cells left at OCV for 10 d indicate that the HC interface becomes more and more resistive when Na metal is present [196]. When comparing cyclic voltammograms of 1 M NaClO 4 in PC, Lee et al concluded that electrolyte decomposition products (formed on Na surface) diffuse to the working electrode (WE), where they are oxidised. Such parasitic reactions can obscure true measurements of the electrochemical performances of a given material [214]. Pfeifer et al observed that carbonate electrolytes (e.g., DMC, EC, PC) containing either NaClO 4 or NaPF 6 react readily in contact with Na, producing coloration or cloudiness of the electrolyte due to side reactions [302].
Sodium-ion intercalation into Li 4 Ti 5 O 12 was investigated using Na metal and activated carbon (AC) as CEs, with a significant cyclability improvement in the latter case. This was ascribed to a different passivation layer being formed, depending on the CE used, and possibly to more severe PVDF binder decomposition when Na is present. The highly inhomogeneous plating of Na, together with an increase in the Na metal impedance as a result of time and cycling, could also result in significant limitations on the cyclability of half cells (figures 28(b)) and (c)). Thus, when tests are performed on two-electrode half cells, the real performance of the active materials being tested as WEs can be significantly underestimated.
Finally, as the coulombic efficiency of Na plating/stripping is usually very low, the stripping of Na from the CE commonly involves two types of metallic Na: the freshly deposited one (Na plated ) and the Na bulk originally present as a disk or a foil. This can lead to an artificial voltage step that arises during the reduction of the active material used as the WE ( figure 28(d)). Such a step can only be seen in two-electrode cells and was ascribed to the higher overpotential for Na bulk when compared with Na plated stripping [212,213]. Once again, such an electrochemical response is only associated with the CE, and could be misleading.

Advances in science and technology to meet challenges
Several strategies have been developed to circumvent the problems associated with Na metal. One of the main approaches has been to control the nature and properties of passivation layers. Artificial passivation layers using freestanding composite layers consisting of Al 2 O 3 particles and liquid electrolyte-swollen poly(vinylidene fluoride-co-hexafluoropropylene) polymers were developed, leading to an improved cycle life of Na||Na symmetric cells. However, evidence of dendrite formation and/or a cell impedance increase  Reproduced from [210]. CC BY 3.0. (c) Nyquist diagrams of impedance measurements over time of symmetric Na/Na cell at 25 • C. Reproduced from [196]. CC BY-NC-SA 3.0. (d) Illustration of an artificial voltage step phenomenon observed in a two-electrode cell due to the Na CE (and RE). Reprinted from [212], Copyright (2014), with permission from Elsevier. could still be seen at 0.5 mA cm −2 [215]. A similar layer using 2.8 nm-thick Al 2 O 3 produced via low-temperature plasma-enhanced atomic-layer deposition was found to significantly increase the cyclability of Na symmetric cells up to 0.5 mA cm −2 [216].
Another avenue used to tune the SEI stability is to adjust the electrolyte formulation. Since this approach has already been discussed in section 4.4, here, we will focus on the impact of a few selected electrolytes on the reliability of Na-metal CEs and REs. Fluoroethylene carbonate (FEC) is a popular electrolyte additive in Na-ion batteries which limits the Na reactivity towards the electrolyte and helps in minimising the irreversibility of cathode materials. However, this does not come for free, and significant Na metal impedance increases can be measured together with a continuous release of small quantities of gases, even after 5 d [213]. Glyme-based electrolytes are among the most successful for Na metal. Schafzahl et al demonstrated that improved cyclability and coulombic efficiency (reaching 97.7%) could be achieved in NaFSI/dimethoxyethane (DME) when compared with a carbonate-based electrolyte [217]. Superconcentrated glyme-based electrolytes were also investigated, reaching a coulombic efficiency for Na plating/stripping of 99.3% at 5 M NaFSI in DME [218].
Beyond the electrolyte formulation, several parameters can affect Na plating and stripping. Rupp and Vlad systematically investigated the roles of the electrolyte, separator, electrode preparation, cell architecture, cycling procedure, and the nature of the current collector [219]. The use of NaCF 3 SO 3 in diglyme with a current density of 2 mA cm −2 , a Celgard 2325 separator and a high-density polyethylene gasket (to give better control of the current lines) led to a Na plating/stripping coulombic efficiency of 99.74%. Careful Na metal preparation could also result in limited surface contamination and an improved morphology [220].
While some of the strategies presented above could improve the Na plating/stripping coulombic efficiency and cyclability, the stability of the SEI (and its impedance) and the absence of soluble/gaseous contaminants remain to be ascertained. A radically different approach, relying on alternative CEs and REs, is also considered. A promising example consists of using the NASICON-type Na 3 V 2 (PO 4 ) 3 (NVP) as an intercalation-type CE [221]. Half cells with NVP as the CE exhibited stable cycling over a long period with low polarisation compared to Na as the CE. Measurements could also be performed at temperatures above the Na melting point, and due to the very flat voltage plateau of NVP, a two-electrode cell configuration could be assembled with a reliable potential measurement. Since NVP operates within the ESW of most electrolytes, it is also likely that such a CE does not produce undesirable contaminants. However, similarly to the use of capacitive electrodes (e.g., activated carbon), the sizing of the CE as a function of the expected capacity of the WE must be carefully considered [211].
The potential of the RE in a sodium cell is determined by the activity of the Na + ions in its vicinity, unlike the case for aqueous systems, where REs are commonly reversible to anions. As mentioned previously, SEI instability on Na metal has been shown to cause unreliable voltage measurements [210]. Lee and Tang evaluated alternative REs for Na-ion batteries: tin alloy, nickel hexacyanoferrate, activated carbon (AC) and silver ion [222]. Unfortunately, continuous potential drifts were recorded for tin alloy, nickel hexacyanoferrate and AC REs. Although silver ion (Ag + /Ag) REs were found to be stable, they required the presence of a separate compartment and a porous glass frit to avoid the migration of silver cations; thus, they were not suitable for compact cell designs, such as Swagelok cells.

Concluding remarks
The use of Na metal for the CE and RE in a half-cell configuration results in several issues: electrolyte contamination, an increase in cell impedance, unstable potential, an artificial voltage step, and premature cell failure, among others. Most of these issues are rooted in the poor stability of the Na metal interface in most organic electrolytes. While several parameters can be tuned to mitigate such interfacial instability (e.g., electrolyte formulation, Na pre-treatment, cell geometry, etc.), there is currently no protocol which guarantees that reliable results can be obtained using Na metal. Since highly unsafe Na metal is mostly used in half-cells for material testing, one may wonder if research efforts need to be dedicated to making this challenging electrode work properly. Alternative CEs and REs such as NVP, AC, and silver-based electrodes are very attractive candidates for reliable electrochemical setups. While care must be taken when using such electrodes, their use can easily and routinely be integrated into most battery laboratories to obtain more reliable, and perhaps unexpectedly better results, than with Na half cells.

Status
Neutrons scatter from elemental nuclei and consequently interact weakly with condensed matter. This property offers considerable advantages in addressing the challenges associated with characterising the complex structure and dynamics of Na battery materials. Neutrons have a highly penetrating nature and can probe magnetic structure, resolve light elements and distinguish elements close to each other in atomic number. As a result, neutron diffraction (ND) can provide an insight into transition-metal (TM) distributions, magnetic structure, and Na diffusion pathways. Total scattering can distinguish between locally ordered TM distributions, observe local structural changes due to oxygen redox, characterise the structure of nanoparticles, and quantify disorder in hard carbon anodes. Inelastic techniques such as inelastic neutron scattering (INS) and quasi-elastic neutron scattering (QENS) can probe dynamics including Na diffusion, and techniques such as reflectometry and small-angle neutron scattering (SANS) allow us to observe the formation of the SEI layer and characterise the change in thickness/porosity of the SEI and electrode materials. Imaging techniques including radiography, tomography, and Bragg edge tomography can reveal the evolution of the components during cycling, and muon spectroscopy can probe the dynamics and energetics of Na diffusion. In addition to these specific techniques, neutrons remain a superior characterisation technique for operando studies of commercial and custom batteries due to their highly penetrating nature. Insight into the structural changes on cycling has revealed the mechanisms responsible for capacity fade [223], cycling-induced cation mixing [224], intermediate phases [224], dependence on cycle rate [225], and has led to the optimisation of commercial materials [226].
While the capabilities of neutron characterisation are vast, there is still scope for improvement through advances in facilities, instruments, and devices. For example, higher flux will increase temporal and spatial resolutions, and permit operando studies at higher cycling rates with an improved signal-to-noise (S/N) ratio.
Operando neutron experiments are extremely challenging but are essential for a full understanding of battery operation and failure. Current abilities may be improved by developing a simultaneous characterisation facility that combines neutron scattering with electron paramagnetic resonance (EPR), NMR and X-ray spectroscopy techniques, optimising and improving existing sample environment cells, and developing new cells for scattering techniques currently without operando capabilities.

Current and future challenges
Disorder is a key feature in the most of the promising electrode materials, whether it be in structure, cation and anion distribution, stacking faults, or defects. Distinguishing between disordered states and correlating these to electrochemical properties is very challenging; however, cation disorder can improve Na transport, prevent unwanted phase changes, and reduce volume changes during cycling [227][228][229]. Furthermore, the structure of amorphous hard carbon (one of the most promising anode materials), and its Na storage mechanisms are difficult to characterise due to the lack of long-range order. While neutron characterisation, specifically total scattering, is well-equipped to address this, the routine characterisation of disordered materials remains a challenge. The analysis of such data often requires additional techniques to generate disordered models, either driven by physical/thermodynamic constraints, such as the Monte Carlo method, or driven by data, such as the reverse Monte Carlo. In addition, operando total scattering remains challenging, as any additional components from the battery setup interfere severely with data analysis.
Neutron characterisation excels in operando studies, despite the low neutron flux, which limits temporal and spatial resolutions. For diffraction, this limits the cycling rate, spatial resolution across an electrode, and the data quality achievable. For other techniques, such as quasi-elastic scattering (QENS), it has inhibited in-situ studies altogether. Cell design is also complex, as the requirements for good quality data and electrochemical cycling comparable to commercial battery operation are different and often incompatible.
In general, understanding and characterising the formation of the SEI on electrode materials, its chemical composition, evolution on cycling, and impact on diffusion are huge challenges for the battery community and are still in their infancy. Neutron reflectometry is an excellent technique for studying buried interfacial structures, and the highly penetrating nature of neutrons means that SEI formation can be observed in-situ, which is essential, as the interfacial structures are usually very sensitive to air. A major challenge for this technique is the development of an operando cell that presents a thin electrode with an extremely smooth surface.

Advances in science and technology to meet challenges
Advances in instrumentation, such as guides and detectors used to obtain a higher flux in the sample will improve the temporal resolution and data quality of operando studies. This can also allow operando experiments for techniques where the current time resolution is too poor, such as QENS, as well as improving spatial resolution for imaging studies. Advances in cell development will also benefit operando studies, with the potential to reduce data collection times, allow the performance of simultaneous multicharacterisation studies, and facilitate total-scattering operando experiments. The development of operando total scattering is a significant challenge, as additional components from operando and/or multicharacterisation cells interfere heavily with pair distribution function (PDF) data collection and analysis; thus, new cells need to be developed with components that are either easily subtracted or have a small contribution to the total scattering.
Alternatively, the temporal resolution may be improved through the development of stroboscopic measurement techniques, whereby multiple identical experiments are carried out and the data are carefully time marked to allow the data from the individual experiments to be combined, enhancing the S/N ratio and resolution through overlapping data sets.
Advances in sample environments and equipment are essential, in particular, for in-situ and operando experimentation, which enable simultaneous data collection from complementary techniques-so-called 'multimodal characterisation' . Examples include simultaneously collecting NMR or EPR data whilst performing neutron diffraction, or the combination of two different neutron techniques, such as diffraction and imaging, a facility that is currently under development. The development of new multimodal characterisation sample environments that can be utilised with different instrumentations and across techniques with the same sample is vital. A full understanding of the physical phenomena that underpin functional properties can only be obtained through this type of experimentation.
While advances in flux and sample environments will benefit disorder studies, data-analysis tool development will allow heavily disordered systems to be understood. It would be helpful to improve analysis tools such as 3D-PDF [230,231] that can identify Na sites when disordered, component analysis that can separate multiple phases in the PDF, and machine-learning algorithms that can be used to identify potentially interesting structures, including disordered ones.

Concluding remarks
Neutron characterisation techniques deliver significant capabilities to the energy materials community, giving the possibility of full, multimodal characterisation of all functional properties and related structural and spectroscopic phenomena. The development of new sample environments to permit these multimodal studies across different instrumentations and techniques is essential to support progress in this area. Equally, investment in neutron facilities to improve instrumentation, detectors, and neutron flux will result in advances in the temporal and spatial resolution of neutron-scattering techniques. This combination of developments will facilitate a comprehensive understanding of battery systems and thus promote the creation of new and improved battery systems.

Solid-state NMR for the characterisation of Na-ion batteries
Valerie R Seymour 1,2 and John M Griffin 1,2 1 Department of Chemistry, Lancaster University, Lancaster LA1 4YB, United Kingdom 2 The Faraday Institution, Quad One, Harwell Science and Innovation Campus, OX11 0RA, United Kingdom

Status
In order to develop and optimise new Na-ion battery materials, it is crucial to understand the underlying structural chemistry in detail. In general, diffraction techniques are the go-to methods for studying crystalline battery materials, and these continue to be the 'gold standard' in delivering atomic-level pictures of the long-range structure. However, many structures and mechanisms involve short-range phenomena such as disorder, dynamics, and defects that can be challenging to characterise by diffraction. In this regard, solid-state NMR stands out as a powerful complementary approach due to its sensitivity to the local chemical environment. Indeed, solid-state NMR has no requirement for long-range order, making it applicable to the study of crystalline, disordered, and amorphous materials. It is also sensitive to dynamics over a wide range of timescales, making it useful for studying ion migration and diffusion.
Solid-state NMR experiments on batteries can be carried out ex situ on materials extracted from cells, or in situ using bespoke cells that are designed to fit inside the NMR detection coil (figure 30) [232]. Ex situ experiments have the benefit that they are compatible with magic-angle spinning (MAS), meaning that high-resolution results can be obtained. In situ experiments have the advantage that they give a better chance of observing any short-lived or metastable states, since the cell is not disassembled for analysis.
Solid-state NMR studies of Na-ion battery materials are not yet as prevalent as for Li-ion battery materials, but both ex situ and in situ approaches have been applied in a number of studies. For NaMnO 2 , ex situ measurements were used to follow changes in the domain structure during cycling. The spectra revealed signals associated with the α-and β-phases, as well as the stacking-fault interfaces between these domains ( figure 31(a)) [233]. During desodiation, the peak corresponding to the β-phase reduced faster than the interfacial resonance, indicating preferential extraction from this phase and/or loss of long-range order. For the candidate cathode material Na 2 FePO 4 F, 23 Na 2D exchange NMR experiments showed that the Na ions move between crystallographic sites in the structure at a frequency of approximately 200 Hz [234]. Ex situ measurements showed the formation of the fully oxidised NaFePO 4 F phase during cycling, suggesting a two-phase desodiation mechanism. NMR performed on in situ 23 Na has been applied to hard carbon-anode materials, where it clearly revealed structural transitions during sodiation and desodiation [77]. At low sodiation levels, the isolated Na ions showed a characteristically diamagnetic chemical shift; however, at high sodiation levels, a marked change to paramagnetic chemical shifts signified the growth of quasi-metallic clusters ( figure 31(b)). Recently, in situ 23 Na NMR and MRI have also been applied to observe the growth of dendritic and mossy microstructures on Na metal anodes [235]. The results revealed a two-step mechanism in which nucleating microstructures led to an increasing overpotential which triggered a rapid breakdown of the electrolyte, causing cell failure.

Current and future challenges
One of the key challenges in the solid-state NMR of battery materials is the presence of strong paramagnetic interactions arising from the unpaired electron spin density of transition-metal ions that are present in many systems. These interactions can result in large resonance shifts and extensive line broadening which complicate the acquisition and interpretation of NMR spectra. In most cases, the resolution is maximised by performing measurements at fast MAS rates (⩾60 kHz) and using moderate or low magnetic fields. However, even under these conditions, the resolution can still be compromised, making it difficult to identify all the species that are present. In addition, the assignment of spectral resonances is often not straightforward. For extended diamagnetic systems, plane-wave DFT codes can usually predict NMR parameters with sufficient accuracy to allow assignment to a model structure. However, due to the greater complexity of paramagnetic structures, progress in the development of accurate DFT methods for the calculation of NMR parameters has been slower, and spectral assignment still largely relies on empirical trends.
Another challenge is the intrinsically low sensitivity of some NMR-active nuclei, which hinders or precludes their observation. For some nuclei (e.g., 17 O or 2 H) the natural abundance is so low that their observation is rarely possible under normal conditions. Isotopic enrichment of these nuclei is possible, but is very costly and can require modification of the synthesis procedure, depending on the isotopically enriched precursor available. Sensitivity is also a significant challenge in the detection of surface and defect species, which play an important electrochemical role in many systems. Since NMR is quantitative, the detection of such low-concentration species within a bulk sample can be very challenging, even when they contain high-sensitivity nuclei. This challenge is exacerbated by in situ measurements, in which the sensitivity is typically further reduced due to the requirement for a large-diameter detection coil and a low filling factor due to the cell components, which reduces the sensitivity further.

Advances in science and technology to meet challenges
An important area of NMR hardware development is MAS probe design. As discussed above, fast MAS can help to remove the effects of paramagnetic interactions, which broaden and complicate solid-state NMR spectra in battery materials. Fast MAS frequencies remove these interactions more efficiently, thereby improving both the resolution and sensitivity of NMR spectra, so that more information can be obtained. Very recently, a probe capable of MAS frequencies up to 170 kHz has been demonstrated [236]. Although this ultra-fast MAS technology has not yet been widely applied, it represents a significant step forward, which can be expected to support new insights into battery materials in the future.
Advances in DFT methods for the calculation of paramagnetic NMR shifts will play a key role in addressing the issue of spectral interpretation and assignment. Although this remains a significant challenge, progress has been made in recent years, and a number of methods have been successfully demonstrated for model systems [237,238]. Further development of these methods will enable them to be applied more widely beyond proof-of-concept studies, although due to the complex and highly system-dependent nature of the calculations, it is not clear whether paramagnetic shift calculations will become routine in the same way as for non-paramagnetic materials.
Concerning the challenge of detecting insensitive nuclei and surface or defect species, significant progress has been made in recent years through the development of dynamic nuclear polarisation (DNP). DNP transfers strong electron spin polarisation from either native or exogenous radical species to nuclei within the material of interest. Since the transfer is distance-dependent, exogenous radical species have a tendency to selectively enhance surface groups [239]. Radicals native to the material structure have also been used to enhance insensitive nuclei in the bulk structure [240,241]. This approach holds considerable promise for the study of unstable or reactive battery materials which may not be amenable to the addition of exogenous polarising agents. Despite the promise that DNP holds, further developments are required to properly understand polarisation transfer mechanisms, and also to widen the temperature range over which it is effective, so that materials can be studied under ambient or variable-temperature conditions.  23 Na ex situ NMR spectra of NaMnO2 at different states of charge. Resonances corresponding to α-NaMnO2, β-NaMnO2, and interfacial stacking faults are highlighted in green, blue, and red, respectively. Reprinted with permission from [233]. Copyright (2014) American Chemical Society. (b) In situ e 23 Na NMR spectra of a hard carbon anode material as a function of the state of charge. The movement of the signal from the diamagnetic (low chemical shift) regime to the paramagnetic (high chemical shift) regime shows the transition from isolated Na ions to quasi-metallic clusters. Reproduced from [77] with permission of The Royal Society of Chemistry.

Concluding remarks
Although widely applied as an analytical technique, solid-state NMR continues to be a highly active area of research in itself, which engenders continual developments in experimental hardware and methodology. These ongoing advances are helping to deepen and transform our understanding of battery material structures and mechanisms, so that their material properties can be optimised. For the study of Na-ion battery materials in particular, NMR has already provided considerable insight and will continue to do so as faster MAS rates become more widely available and advances in DNP methodology provide ever higher sensitivity and selectivity.

Nanoscale characterisation of the local physical properties of Na-ion batteries, electrodes, and interfaces
Oleg V Kolosov 1,2 and Yue Chen 1,2 1 Department of Physics, Lancaster University, Bailrigg, Lancaster LA1 4YB, United Kingdom 2 The Faraday Institution, Quad One, Harwell Science and Innovation Campus, OX11 0RA, United Kingdom

Status
The majority of characterisation techniques enabling the current fast development of rechargeable batteries fall into three major categories-the traditional electrochemical performance measurements, such as galvanostatic charge/discharge and cyclic voltammetry; crystal and nanostructural characterisation, such as x-ray diffraction and electron microscopy (EM); and NMR, XPS, and soft x-ray spectroscopy, which also provide chemical component analysis. The long history of battery technology development has amassed an in-depth understanding of electrochemical processes based on the characterisation of these electrochemical, structural, and chemical properties, derived from the above electroanalytical methods and their techniques. While these properties are essential, as shown in figure 32, the physical properties, including the mechanical, thermal, and electrical properties, of the materials deployed in Na-ion batteries are no less important.
Growing attention has recently been paid to revealing the nanomechanical properties of sodium nanodendrites in high-capacity Na-metal anodes [242], and the generation of interfacial stress in layered TMD [243] anodes due to the intercalation of the larger Na ions compared to Li ions. It is also well established that the Young's moduli of the SEIs on the negative electrodes define the battery's resistance to dendrite growth whereas the stability of polycrystalline cathodes crucially depends on the mechanical integrity of their components during the charge-discharge cycle expansion and their resistance to crack propagation [244]. The electrical conductivity of the electrodes and the ionic conductivity of the interfacial layers are the key limiting parameters for battery rate performance, and the thermal conductivity is another factor that defines battery operation in demanding applications. Therefore, measuring mechanical moduli, local stress, electrical conductance, and thermal transport in real-life batteries and model systems, and correlating these data with electrochemical and structural analyses will provide an essential platform for overcoming existing challenges and for the development of novel paradigms in Na-ion batteries.

Current and future challenges
One of the major challenges in the nanoscale measurement of the physical properties of electrodes and interfaces in NIBs is linked to the wide range of length scales involved. The electrode thicknesses are typically a few tens of micrometres (µm), the individual cathode electrode particles are a few µm across, and hard carbon particles in the battery anodes are on the nm length scale. The SEI layers are several tens of nm thick, while the cathode electrolyte interphase (CEI) can be as little as a few nm thick [245]. Na-ion incorporation occurs at interatomic distances that are on the sub-nm length scale. The successful investigation of this phenomenon requires a means of characterisation that can provide local measurements of physical properties with the commensurate nanoscale lateral resolution. Another major challenge is the extreme heterogeneity of the battery components. For example, the mechanical moduli range from sub-MPa values for the outer layers of an SEI-facing electrolyte to hundreds of GPa for cathode polycrystalline particles, and the ionic conductivity of the sodium ranges from ∼10 −2 µS cm −1 when crossing the SEI layers at the electrode-electrolyte interface to ∼10 µS cm −1 within the interlayers of the TMD anode. Lastly, but most importantly, the dynamic evolution of nanoscale physical properties, which are closely related to the sodium ion transport and electrolyte electrochemical/thermal decomposition, is exceptionally difficult to 'look' at, since sodium battery active materials and interfaces are buried deep within the volatile liquid electrolyte under battery operating conditions.
To satisfy these complex characterisation capability requirements at multiple length scales and address the electrode heterogeneity and the fluidic electrochemical environment, scanning probe microscopy (SPM) has been applied for the micro/nanoscale characterisation of the dynamic and local physical properties of battery electrodes and interfaces. In these studies, model systems are often used, e.g., highly oriented pyrolytic graphite (HOPG) [246,247] or the physical vapor deposition/chemical vapor deposition of binder-free electrodes [245] as a proxy for the battery electrode, enabling the direct observation of SEI formation and ion intercalation through nanomorphology measurements of the electrode surface. However, it is not sufficient to only investigate the surface physical properties of these structures; tools are required that can access the inner structures of electrodes and interfacial layers. The direct nanoscale characterisation of multiple complementary physical properties of real-life batteries, for ex situ and in-situ/operando cases, remains the holy grail of Na-ion characterisation requiring an innovative solution for three-dimensional (3D) sample preparation.

Advances in science and technology to meet challenges
Among the vast number of techniques used for sodium battery interface analysis [248], SPM occupies a unique place, as it can provide both the required nanoscale spatial resolution, as well as being sensitive to the local physical properties of the studied material. Using SPM in the electrochemical environment on the model HOPG substrate allowed the study of the initial stages of SEI growth [246], revealing the dynamics of ion intercalation. The downside of SPM is that it requires nanoscale flat samples and mainly allows the investigation of surfaces rather than 3D structures. Focused ion beam (FIB) sectioning, widely used in EM measurements, is inadequate for SPM, as the sections produced are too small to be measured efficiently and, more significantly, the Ga ions can drastically change the physical properties of the studied material during sectioning. As shown in figure 33(a), a much more suitable approach is to use advanced nanosectioning using multiple Ar-ion beams, known as beam-exit cross-section polishing (BEXP), which produces a close to atomically flat surface with minimal damage, due to the non-reactive nature of Ar atoms. In the BEXP method, the open-angle tilted section adjacent to the intact surface of the sample produces a perfect area for the material-sensitive SPM, additionally expanding the interfacial area by a factor of 5-10, allowing the study of interfaces a few nm thick (such as the superlattice in figure 33(b)) with a total section area on the sub-mm length scale. By using a dedicated material-sensitive SPM contrast, such as nanomechanical via sample excitation, ultrasonic force microscopy (UFM), or tip-excitation waveguide UFM [249], the conductance of the cross-sectioned samples can reveal the local physical properties of the internal structure at nanoscale resolution. Figures 33(c)-(e) show a diagram of a commercial Ni-rich cathode polished by BEXP and the prepared sample's cross-sectional UFM images. In figures 33(c) and (d), the cathode particles at different states of charge (SOC) in the binder and the conductive carbon black matrix can be clearly distinguished. Additionally, the crack network within the grain section is visible in the nanomechanical UFM images, which could serve as a perfect model for operando SPM studies of dynamic crack propagation in cathode polycrystalline particles, revealing the fundamentals of capacity degradation in sodium-transition-metal oxide cathodes.
Another approach for examining 3D structures is to use direct in situ sectioning of the sample through nanotomography [250], in which the tip mills out the surface along the section plane, followed by nanomechanical, nano-electric, or surface-potential measurements. Such an approach is more suitable for the sectioning of softer layers, such as the SEI, in which follow-up nanomechanical SPM probes all the layers, starting from the double layer in the electrolyte, through to the soft and then more compact SEI layers at the interface with the anode. As shown in figure 33(f), when BEXP and nanotomography are combined, the organic adsorption layer that forms on the atomically flat cut slope of the HOPG anode can be scraped by the SPM tip. This allows for the uncovering of the subsurface physical properties of electrodes and interfaces, which could be the 'buried treasure' in the study of Na-ion batteries.
Overall, SPM will allow the investigation of the highly important Na-ion intercalation phenomena in carbon-based [247] and other electrodes by observing both the nanoscale morphological (operando dimensional change) and physical parameters (operando mechanical compliance that was shown to affect the mechanical properties during atomic intercalation) [251]. Other physical parameters, such as the surface potential and conductance changes, when measured ex-situ, can provide additional information in such studies.

Concluding remarks
The real-space physical-property nanomapping of battery electrodes and interfaces via multifunctional SPM, combined with the efficient and artefact-free sample preparation approach via BEXP and nanotomography will provide a valuable characterisation toolbox for the development of the next generation of Na-ion batteries. Furthermore, it will allow the direct investigation of SEI stability and composition, the development of more robust cathode/anode-electrolyte interfaces, and the exploration of Na-ion intercalation, enabling the selection of a winning electrode-electrolyte combination.

Status
Loss of battery electrode inventory to dissolution processes is one of the major sources of capacity fading. In the lithium-ion field, much research has been conducted into the charge-discharge profiles and electrolyte formulations that contribute to dissolution, in particular, for electrodes based on manganese, nickel, and cobalt, which can be translated to sodium-ion chemistries (with some caveats) when similar electrode materials and electrolytes are used [252]. Na + ions have a 34% larger radius than Li + (1.02 vs. 0.76 Å) which causes a greater strain on the electrode materials during insertion and extraction. Sodium-ion electrodes also span a larger compositional range than current LIB materials, encompassing transition-metal oxides, alloying reactions, and conversion-reaction materials, and therefore enhance both the possible sources and consequences of electrode dissolution. While direct loss of active electrode material due to dissolution into the liquid phase is typically in the single-digit percentage of overall capacity loss, the presence of dissolved transition metals in the electrolyte is known to rapidly accelerate SEI formation, leading to near-complete passivation of the negative electrode. This so-called dissolution-migration-deposition (DMD) process is estimated to account for up to 30% of the overall capacity loss in manganese-oxide-based intercalation batteries [253].
Several strategies already exist to counteract electrode dissolution, such as heteroatom doping [254], coatings, and the use of concentrated electrolyte formulations, figure 34. These indeed improve the long-term stability, but typically at the expense of the power and/or energy densities of the electrode. Fast throughput techniques that can determine electrode dissolution rates as a function of electrode potential, time, and electrolyte formulation are therefore essential and urgent. Understanding the effect of electrolyte, solvent, and electrode composition on dissolution and battery longevity will lead to better-informed materials discovery for durable and efficient NIBs.

Current and future challenges
Due to their propensity to dissolve manganese, LiMn 2 O 4 and Na x MnO 2 electrodes are widely used for model studies. A primary challenge is to accurately identify the cause of electrode dissolution. For example, Hunter's mechanism of Mn 3+ disproportionation into soluble Mn 2+ and insoluble Mn 4+ was confirmed in an aqueous solution but may differ in non-aqueous, fluoride-containing carbonate-based electrolytes. Recent studies have suggested that in non-aqueous electrolytes, Mn(III) is the dominant dissolved oxidation state. Other elements may also dissolve into the electrolyte, for example, vanadium from NaV 3 O 8 cathodes [255].
As the rate of dissolution is proportional to the electrochemically exposed surface area, bulk characterisation techniques, such as x-ray diffraction may not be adequate. In situ surface-sensitive methods, such as electron energy-loss spectroscopy, extended x-ray absorption fine structure, and high-angle annular dark-field imaging show that de-intercalated λ-MnO 2 is chemically unstable [256]. However, only the surface layers break down to Li 2 Mn 2 O 4 and Mn 3 O 4 , where the average Mn oxidation state is much closer to three. The stability of the electrode therefore cannot be predicted, based on the bulk composition or the cell potential, which would indicate only an insoluble product. Such structural changes also induce stress in the host material due to changes in the lattice constant, leading to cracks and again increasing the exposed surface area. Although such methods have not yet been employed to elucidate sodium-ion electrode dissolution, the intercalation of Na + induces more strain than the intercalation of Li + due to its larger radius, potentially leading to more cracks, a greater exposed surface area, and hence, enhanced dissolution. Heteroatom dopants, such as iron and nickel in, for example, Na 0.67 Fe 0.5 Ni 0.15 Mn 0.65 O 2 may increase stability against manganese dissolution [257], but may themselves dissolve following chemical or electrochemical pathways and increase the potential range in which dissolution takes place. The dissolution problem is further compounded in sodium-ion anodes, in which elements such as tin, antimony, phosphorus, titanium, copper, iron, or sulphur can produce soluble byproducts during operation, whereas graphite does not.
Solvents, electrolytes, and contamination from moisture or CO 2 can further accelerate the degradation of sodium cathode host materials at a much higher rate than their lithium counterparts [258]. The dissolution of ions into the electrolyte solution depends on the particular electrolyte formulation, ion solubility, and potential programming of the battery cell, which encompasses a huge array of variables. The lack of direct, real-time dissolution measurements greatly compromises the time and potential resolution of dissolution information. To date, most dissolution diagnosis has been conducted post-mortem. For example, inductively coupled plasma (ICP)-based elemental analysis of the electrolyte has been used to identify ppm-range  concentrations of dissolved iron from sodium iron hexacyanoferrate electrodes as a function of the surface coating strategy [259]. Energy-dispersive x-ray spectroscopy (EDS) is also commonly used, for example, to analyse the separator in order to identify trapped ions, or to analyse the negative electrode to identify DMD products from Na 0.67 MnO 2 electrodes [260]. The preparation and ex situ handling of such samples increases experimental error and may compromise the reliability of the conclusions. Furthermore, full and half cells prepared in CR2032 coin cells do not isolate a single cause of dissolution, as the effects of electrode arrangement, cell pressure, the counter-electrode material and current collectors must also be considered.
Open questions are whether electrode dissolution is electrochemically or chemically induced, i.e., which are the most significant dissolution pathways, and their accurate quantification. Methods of detection must be highly sensitive, as dissolution only affects a small part of the overall electrode mass, and setups must rigorously exclude agents known to accelerate dissolution, such as water, as otherwise, artefacts will arise. An adequate dissolution monitoring system must therefore be carefully designed.
Advances in science and technology to meet challenges ICP-based instruments are routinely available, and allow highly sensitive elemental analysis, with detection limits of sub-ppb using optical emission spectroscopy (ICP-OES) or even sub-ppt using mass spectrometry (ICP-MS). Primarily used for aqueous samples, these are, to a certain degree, forgiving towards non-aqueous battery solvents and electrolyte salts (up to 0.1 M (ICP-MS) or 1 M (ICP-OES) without dilution). The observation of dissolution in non-aqueous electrolytes is challenging, due to the need for oxygen and water to be excluded from the electrolyte. Prohibitive factors include instrument corrosion from highly reactive electrolytes and the high cost of many high-quality non-aqueous electrolyte solutions.
Recent advances in electrochemical flow cell design have enabled direct observations of the dissolution of candidate electrode surfaces in non-aqueous environments under potential or galvanostatic control. Typically, the electrochemical flow cell outlet is directly connected to the ICP nebuliser by tubing, in the so-called direct-injection analysis. Successful examples include the flow-through scanning flow cell (FT-SFC) introduced by Mayrhofer et al [261], which incorporates a V-shaped channel design for efficient transport of dissolution products, figure 35. This type of cell has also been manufactured by 3D printing and successfully used for fast and high-throughput dissolution detection [262]. However, intricate new cell designs are not always necessary. Markovic's group [263] fitted an ICP-MS inlet to a rotating disc electrode tip in a stationary probe rotating disc electrode (SPRDE-ICP-MS), adding the advantage of uniform and well-characterised convection. Hence, hyphenation of existing apparatus to increase time-and potential-dependent dissolution information output in accelerated degradation tests becomes possible.
While ICP-based analytical instruments identify the atomic mass/charge ratio (ICP-MS) or element-specific emission spectra (ICP-OES), valuable information about the oxidation state of the dissolved species, i.e., speciation, is lost. Other in situ techniques may additionally allow speciation, e.g., rotating ring-disc electrode [264], electron paramagnetic resonance, or UV-vis [265] measurement of the electrolyte in operando cells. However, interference from solvents, electrolytes, or other solution species must be accounted for.

Concluding remarks
Post-mortem analyses and ex-situ observations of battery electrodes limit our understanding of the degradation processes that are the main causes of capacity fading and the premature failure of batteries. Today, a detailed understanding of the atomistic causes of the dissolution of electrode materials remains elusive, even for the most basic model electrodes. The development of online analysis techniques that quantify time-resolved dissolution as a function of potential and other experimental parameters can revolutionise our understanding of battery electrode degradation, as an essential step towards its mitigation and the rational design of resilient new battery electrode materials. Advances in spectroelectrochemical cell design and analytical instrumentation can unlock detailed dissolution probing, revealing not only elemental quantities but also chemical speciation.

Status
Research into Na-ion batteries (NIBs) began in the 1970s and 1980s and was usurped by LIBs in the 1990s. NIBs have recently re-emerged as a potential sustainable and lower-cost alternative to LIBs [266]. They offer an energy storage solution which is not as reliant upon geographically localised raw or critical materials. Several commercial entities have built up around this technology, with demonstrator cells and prototype battery packs, offering the potential for lithium-ion and lead-acid battery substitutions in current markets [71]. Sodium can be extracted from seawater, anode materials can be obtained from renewable sources, and cathodes generally contain iron and manganese, rather than high levels of cobalt and nickel. The manufacturing processes are expected to be the same as those for LIBs, and therefore, no investment is required in new manufacturing lines. The cost difference between LIBs and NIBs can therefore be considered to be due to materials and performance only [184]. NIBs are unlikely to ever reach the energy densities (gravimetric and volumetric) of LIBs, due to their lower operating voltages and lower active component densities [267]. The 'drop-in' manufacturing approach also creates limitations on NIB development, in terms of the design of components and cells. The thickness and design of the electrodes are limited by these tape-casting methods. Opportunities for differentiation consequently lie in high power, safe, and long-life NIB technologies. A sustainable circular-economy perspective on materials, manufacturing, and end-of-life could also offer significant advantages over current battery technologies. Alternative manufacturing methods may offer solutions for improved electrode kinetics, the engineering of more efficient electronic and ionic transport pathways, and simplified recycling. Sustainability needs to be considered within a greater holistic outlook, however; the use of low-cost and highly abundant materials may mean, for example, that the reclamation and reuse of these materials never becomes economically viable. Consideration of the hierarchy of recycling, re-use, or second life within the circular economy is required. Therefore, lifetime, in particular, is a key performance parameter for sodium ion batteries, as the value of the materials reclaimed from recycling will not be sustainable if low-cost and abundant materials are used [268].

Current and future challenges
There are several research challenges which need to be addressed in the field of NIB technologies, one of which is long life, which is required to improve sustainability. Advancements in LIB energy density, both volumetric and gravimetric, are required for applications reliant upon this technology, and there is an immense amount of work ongoing to develop new anode and cathode materials. However, the lifespan of the cells is governed by various chemical and cell engineering optimisations relating not only to the materials but also to the design of electrodes and cells and their operation (figure 36) [269]. For NIBs, mass balance optimisation between positive and negative electrodes is critical for high-energy and long-life batteries. Each cathode and anode material has a first-cycle coulombic inefficiency related to electrolyte decomposition and irreversible sodium movement from the positive to the negative electrode. As a result, an interface (surface-electrolyte interface (SEI)) forms on the anode, causing less sodium to be available (reducing capacity) for shuttling between the positive and the negative electrodes. The mass balance and the voltages chosen therefore have a large effect upon the effectiveness of this interface formation at stabilising the electrolyte against decomposition, and consequently, the energy density of the cell [270]. The stability of the interfaces on the positive and negative electrodes is reliant on many factors, and it is known that the stability of NIB interfaces is inferior to those of LIBs due to the highly soluble nature of sodium salts. Much work is required to understand the formation and changes that occur at these surfaces and interfaces during the lifetime of the cell, and some progress has been made in this area (figure 37) [271]. These interfaces are not only important for longevity but also for safety. Dendrites and plating can occur very readily in sodium-ion cells, and this is not only controlled by the electrolyte composition and the ability to form low-resistance, stable interfaces, but also through electrode design. For example, low and inhomogeneous electronic conductivities within the electrode can lead to changes in current density across the electrode, resulting in sodium plating and dendrite growth in areas with high resistance. Improving the electronic conductivity of the electrode with conductive additives, such as carbon black, can aid electron transport but additionally leads to increased electrolyte side reactions and hence ongoing growth in the SEI. This, in turn, blocks pore networks for ionic transport and increases local resistance [272]. Understanding the combined effects of  these changes upon the lifetime, capacity, and ultimately the degradation mechanisms will be essential to extend the lifetime of cells and make further NIB optimisations.

Advances in science and technology to meet challenges
In order to achieve longer lifetime NIBs, a greater understanding of the interaction of all the cell components and competing ageing and degradation processes is required. This will be different for each cell chemistry, voltage window, and electrode design. An understanding of the interfaces is of the greatest importance for both cathodes and anodes. These interfaces control the transport properties of sodium into the electrodes and influence the type and extent of microstructural changes during the ageing of the electrode. The compositional makeup of these interfaces requires optimisation through electrode and electrolyte design. Changes in these interfaces and their effects on the physical and chemical properties will also affect performance and safety. This offers opportunities for advanced, data-driven, and multiphysics models to be partnered with experimental data to provide greater in-depth information about the changes in the fundamental kinetic and thermodynamic properties of the cell for various compositions and designs.
Advances in parameterisation methods for LIBs could be translated to NIBs, offering novel, unique insights into cell construction and component optimisation [273].
Moving away from traditional LIB manufacturing processes is essential to achieve step changes in sustainable cell designs. Novel processes are required to provide improvements in electrode 3D electronic and ionic transport pathways. With improvements in electrode design, less current collector and packaging materials are required, which improves energy density and reduces waste at the end of battery life. Novel methods of embedding electronic conductors in active materials without using high-cost and highly resistive polymeric binders to adhere them to metallic current collectors could provide the step change required. In addition, designs that enable simple manufacturing and subsequent disassembly for material reclamation will enable greater value from material recovery and reuse, creating a sustainable circular economy for NIBs. The ultimate design intent would be to use no binder, however moving to renewable sources for water-soluble binders, such as those based on algae and cellulose, may also improve recyclability prospects. Designing for recycling and disassembly is a key area which can capitalise on the benefits of this set of materials and unlock the potential of NIBs.

Concluding remarks
An economically viable NIB should be considered with respect to the sustainability potential of a circular economic model rather than purely the first-life cost of materials and manufacturing. Consideration of lifetime performance is required for both first life and second life, with recovery and recycling of cell components at the end of life. The current focus is upon matching LIB performance criteria. However, with rapidly changing markets for energy storage solutions, NIBs offer different opportunities for new cell design and manufacturing processes, allowing a holistic approach to sustainability. A focus upon developing longer-life batteries is required to achieve this. The optimisation of electrolyte and electrode designs to reduce interface resistance and dendrite growth will improve both longevity and safety, while new manufacturing processes and cell designs are required for sustainable end-of-life solutions. Materials and components can then be easily and repeatedly manufactured, reconditioned, disassembled, and remanufactured. In this way, the challenges of both the performance and economics of sodium-ion batteries can be overcome by combining novel materials, processes, and products with advanced material recovery, repurposing, and recycling.

Status
The manufacture of batteries is often overlooked at a research level, due to the importance of finding new, higher-energy electrode materials, and solving the challenges faced in cell chemistries at smaller scales. However, when producing large-format cells, its importance in achieving the optimal cell performance is paramount. As demand increases further, together with the growing applications of LIBs, the availability of manufacturing facilities has also increased. Europe has multiple 'Gigafactories' in the pipeline, with examples such as Northvolt, who are expected to reach 32 GWh by 2023. Similar or larger-scale factories in China and beyond are also in the pipeline, backed by larger companies such as Tesla. As noted by the Faraday Institution, eight 'Gigafactories' will be needed by 2040 to satisfy UK battery demands, which requires one to be established at least every two years from 2020 [274]. Projects such as those recently proposed by AMTE Power and Britishvolt are the start of promising large-scale manufacturing of LIBs within the UK, with the hope that this can complete the chain to take materials from powder to power, thus attracting end users of battery technology by shortening supply chains [275].
NIBs are beginning to reach a technology readiness which requires large-scale production, as demonstrated by Faradion, with recent orders from Australia and India [276]. NIBs will see use in large, decentralised power applications, and small-scale deployments will be required to assist in load balancing from intermittent-power farms, such as wind and solar [277]. Factories can switch between lithium and sodium chemistries, due to their 'drop-in' nature, as long as the technologies remain similar enough in their production ( figure 38). This helps to reduce barriers to entry and initial costs, as the creation of new facilities is not required; however, the gap between commercial and laboratory technologies still needs to be bridged to facilitate material advances and their application in larger cells. Additionally, research institutions often lack access to facilities which can help to realise materials at larger scales. By improving this connection, especially through links between industry and academia, it is hoped that an acceleration of research towards manufacturing readiness can be achieved.

Current and future challenges
Although some good parallels can be drawn from LIB manufacture, there are still limitations to how this can be applied to NIBs, due to differences in the materials used. When changing any physical property of a material, such as particle size, crystal structure, morphology etc., this also alters the processing parameters, thus impacting electrochemical performance. Similarly, changing the processing procedures of an electrode material, for example, slurry preparation or electrode modifications, can also influence both its structure and its maximum performance (figure 39) [278]. Therefore, every material must be treated differently to achieve this fine balance between every parameter.
Most of the challenges within manufacture lie at a materials level. Complex multi-step procedures can be used to produce grams of material for research purposes. However, these techniques may not be feasible when producing the kilogram amounts required for larger cells. Material morphologies, such as the creation of high-surface-area particles, will assist with electrochemical performance. Additionally, some layered sodium oxides of interest are extremely sensitive to air and moisture, which can make handling of large quantities more complex, and imposes additional requirements such as additives in slurries to improve their longevity. It is possible to modify these materials to improve their stability in air or to reduce the sodium content, but this can be detrimental to electrochemical performance [279].
Electrode slurries at the coin-cell level also have standardised formulations; little attention is paid to metrology and dry powder mixing, and they include large amounts of binder and carbon to overcome material limitations, such as poor conductivity. Reducing these amounts to increase the energy density of cells is vital at large scales, as commercial cells sometimes reach up to 98 wt.% active material. However, such changes require specific testing and optimisation of every electrode slurry. After coating, the influence of drying and the application of calendering as a technique to improve electrode density and adherence are also not routinely considered within the research environment. These steps also require individual tailoring to maximise electrochemical performance without negatively impacting electrode structure.
Although materials and processing are key to cell performance, other considerations, such as electrode balancing and geometry, tab placement, and safety aspects related to long-term storage and long-term cycling which are not routinely studied at smaller scales also become relevant. Due to the infancy of NIB technology, little research exists for many of these other areas, though it is fortunate that many of these can  be directly transferred from LIB research, and challenges can be jointly tackled. Some advantages and benefits which can be discovered are, however, solely for NIBs, such as the ability to internally short a pouch cell with no performance impact, as demonstrated by Faradion [277].

Advances in science and technology to meet challenges
In general, a better understanding is required of how to produce and process materials, as there is very little information that specifically relates to NIBs at larger scales. Ideally, techniques already used to scale up electrode materials to significant quantities could be implemented. Co-precipitation is the most widely utilised approach for materials scale up, and several researchers have produced well-defined structures of layered sodium oxides for NIBs, which required additional firing steps after the initial precipitation, which can be energy intensive [280]. Prussian blue analogues (PBAs) can also be produced by co-precipitation, and do not require further processing, reducing cost and time. Co-precipitation is not necessarily suitable for all electrode materials, depending upon their stability and composition. Therefore, new routes such as microwave synthesis could be optimised to provide alternatives.
Electrode processing is also a significant focus, as it contributes a large amount to the cost of cell production. Tape casting is the technique most widely utilised to produce electrode sheets; however, the use of N-methyl pyrrolidone (NMP) increases expense, due to the need for solvent-recovery systems. The development of newer electrode slurry combinations, particularly aqueous chemistries, would reduce this cost. Additionally, several researchers have begun to develop 'solvent-free' casting techniques in a bid to remove this issue entirely and have greater control of electrode morphologies, though these advances are likely to take some time to become commercially viable [281,282].
Perhaps one of the biggest advances which will help manufacturing immensely is the use of simulation and modelling. Computational modelling of cells and packs has become much more viable in recent years and by varying parameters, can allow for rapid analysis of changes to electrode composition and cell design, obtaining incremental performance gains when applied experimentally. One example of modelling to identify improvements in pouch cells is that of tab placement, where the benefits of an opposite-ended tab design were shown to improve electrode usage and temperature management [283]. Greater practical application is required to demonstrate the benefits of these findings, and whether there are correlations with simulations.
Finally, in terms of NIBs, there needs to be a much greater understanding of the long-term performance of cells, especially as size increases, which will become better understood with time. Additionally, the use of half cells with metallic sodium does not realise true cell chemistries, which instead use hard carbon anodes, though the use of half cells is also a common practice for new LIB materials. Any performance issues or irregularities in a half-cell setup are not a true representation of what may occur in a scaled-up cell configuration; thus, a greater emphasis on testing materials in three-electrode or small full-cell setups is essential to ensure that materials can be scaled up without issue.

Concluding remarks
Unquestionably, as the field of NIBs expands over the next decade, greater cooperation between materials development and manufacturing will be required to ensure that advances can be easily transferred from small scale to large scale. Though pouch cells have been considered in this text, there are multiple formats for batteries which should all benefit from this focus. If LIB manufacturing lines can be converted to produce NIBs, this would help to drastically reduce costs further and allow for better knowledge transfer, making NIBs an attractive low-cost option. However, there are still many NIB-specific challenges which need to be tackled, mainly by improving knowledge about the processing of electrode materials, and their inherent long-term safety, which both require greater study.

Status
Faradion Limited, established in 2011, was the first non-aqueous Na-ion battery company in the world. Since then, several Na-ion companies have been founded, covering a wide range of sodium-based chemistries. Faradion's technology utilises mixed-phase O3/P2-type Na-Mn-Ni-Ti-Mg layered oxide cathodes, hard carbon anodes, and non-aqueous electrolytes [284]. Other companies reliant on carbon-based anodes and non-aqueous electrolytes include the French company, Tiamat, whose cathodes are based on the polyanionic Na 3 V 2 (PO 4 ) 2 F 3 [285], the Chinese company HiNa Battery, which utilises O3-type Na-Cu-Fe-Mn layered oxide cathodes [286], and the Swedish company Altris AB, based on Prussian blue analogue (PBA) cathodes. In contrast, US-based Natron Energy has built its Na-ion technology around aqueous solvents and PBA-based cathodes and anodes [287].
From the above, it can be seen that there is already a diverse array of Na-ion chemistries available in the industry. This is important, as different markets/applications have different requirements and might require unique solutions -the battery of choice needs just the right combination of cost ($ kWh −1 ), energy density, stability/reliability, power rating, charge acceptance, and temperature performance. As such, figure 40 presents a Ragone plot covering several main market applications with their specific energy/power envelopes (gauged by the datasheets of existing battery technologies that are currently deployed in the market for these specific applications) [288,289]. It should be stressed that these are just a few examples of applications-there are others with different performance requirements. We have also indicated the maximum performance of Na-ion cells from some industry players, as given by their datasheets or published reports [284,285,287]: for Tiamat and Natron, the values mentioned are correct as of mid-2018 and mid-2019, respectively, while for Faradion, values are accurate as of mid-2020. A Na-ion cell might be appropriate for utilisation for a particular application, as long as the application's envelope lies within its energy/power capabilities (gauged by the area below its respective dotted lines in figure 40) and it can meet other requirements such as cycle life. It is evident that current Na-ion technology could be attractive for most applications, apart from some types of consumer electronics and long-range electric vehicles.

Current and future challenges
Referring to figure 40, it can be seen that different Na-ion chemistries are better suited for different applications. For example, Natron's aqueous Na-ion chemistry has the lowest specific energy/power-this is expected, as aqueous electrolyte-based batteries cannot match the energy densities of their non-aqueous counterparts owing to the limited voltage window of water as an electrolyte solvent. Thus, Natron's Na-ion systems would be ill-suited for higher-energy applications, such as electric vehicles. However, Natron batteries' expected low cost, exceptional cycling stability, safety and efficiency would make them attractive for different ESS applications [287]. Tiamat's use of fluorinated vanadium-based polyanionic cathodes delivers Na-ion chemistry with excellent power capabilities albeit at the expense of energy density and raw material handling and availability [285]. Polyanionic cathodes represent an attractive avenue for high power and durable Na-ion systems. The challenge is to base them on more Earth-abundant elements, such as Fe and Mn. Unfortunately, pure Fe/Mn-based polyanionic cathodes generally either demonstrate relatively low operating potentials which limit their specific energies, or rely on bulky anionic groups [290], which limit their tap densities and, in turn, volumetric energy densities [284]. On the anode side, hard carbon can be prepared from a variety of precursors, including low-cost options, and has an attractive combination of desirably low operating potentials and high capacity (current hard carbon capacities almost match those of graphite Li-ion anodes). Hence, it is unlikely to be displaced as the anode of choice for high-energy applications in the foreseeable future. Furthermore, an intense research focus on hard carbon has resulted in significant improvements, not only in its cycling stability and rate performance but also in its charge acceptance capabilities-these, along with excellent compatibility with high-flash-point electrolytes, confer a high degree of safety to Na-ion systems [284].
Faradion's Na-ion cells currently deliver a good mix of energy and power densities with attractive costs, due, in large part, to cathodes devoid of costly and scarce elements, such as Co and V [284]. Table 2 illustrates how Faradion Na-ion cells compare with two types of state-of-the-art Li-ion cells: Li-Ni-Mn-Co oxide (NMC)//graphite and LiFePO 4 (LFP) cathode//graphite chemistries. Ragone plot illustrating the current performance of Na-ion cells of some industry players [284,285,287], along with the two-year target for Faradion's Na-ion cells. The requirements of some applications have been illustrated as well. For this purpose, the applications' envelopes were constructed using datasheets of commercially available batteries that are currently being used for these applications, as indicated in the legend [288]. Acronyms used: ESS = energy storage system (for different grid-storage applications); 2W = two-wheeler; 3W = three-wheeler; eVs = electric vehicles; SCiB = super charge ion battery; PbA = lead-acid; LiPo = lithium-ion polymer; NCA = Li-Ni-Co-Al oxide//graphite; DOD = depth-of-discharge.

Advances in science and technology to meet challenges
Faradion's Na-ion cells currently deliver energy and power metrics in between those of NMC//graphite and LFP//graphite cells (table 2). They can respond at 10 C (continuous discharge), which compares favourably with LFP//graphite pouch cells (1-3 C) [293]. Faradion's Na-ion chemistry could be considered as a replacement for most applications where LFP//graphite is currently deployed. Table 2 also outlines targets for an 'energy cell' and a 'power cell,' that would solidify Na-ion technology as the go-to battery choice for different applications. The 'energy cell' is based on Faradion's Na-ion chemistry and seeks to satisfy some of the more energy-demanding applications, such as mobile phones, laptops, and some EVs. For this purpose, Faradion has devised a clear roadmap to achieve these targets by enhancing reversible capacities, increasing electrode densities, and boosting coulombic efficiencies with novel electrolytes and additives. In fact, we recently showed that the rates of increase of Faradion's Na-ion specific energies (and cycle lives) are significantly faster than those of Li-ion [284].
Even though current Na-ion technology can deliver high power, for applications such as power tools or drones, both power and energy densities need to be simultaneously acceptable. For such a 'power cell,' it might be challenging to meet the power and energy requirements using polyanionic cathodes at acceptable costs. We thus anticipate layered oxide cathodes will be used for such cells, relying on a majority, or a pure, P2 phase; it is well known that many P2-based Na-ion oxide cathodes can deliver excellent rate performance. Faradion's low cell BOM (bill-of-materials) is based on the absence of expensive cobalt, the use of aluminium (as opposed to expensive copper) current collectors, and the natural abundance of sodium, resulting in active materials and electrolyte salts that are cost-effective to produce. Future cell BOMs ($ kWh −1 ) can primarily be reduced in three ways: (a) increasing energy density; (b) large-scale industrialisation of the material supply chain with (c) increasing production volumes, all reducing cost ($) with scale. The technical approach to achieving these target energy densities (stated in table 2) is listed above. The development of material supply chains and an increase in production volumes are already happening with the ever-increasing realisation of the performance benefits of Na-ion batteries (figure 40) and the ever-rising demand. Table 2. Performance metrics of current NMC//graphite and LFP//graphite Li-ion technologies and those of Faradion's Na-ion cells. Industry targets for a Na-ion energy cell and a Na-ion power cell are also shown.

Current technology
Na-ion energy cell: production c pouch cells Na-ion power cell: production d

Concluding remarks
Na-ion technology is rapidly catching up with the established Li-ion battery. It is revealed here that the energy/power densities of Faradion's Na-ion chemistry, which is devoid of scarce elements, such as Co and V, already surpass those of LFP//graphite systems. For most applications, the different types of Na-ion battery already possess the required performance attributes at the cell level, from negligible self-discharge to excellent temperature performance and the ability to be shipped/stored at 0 V [284], to name a few. Such performance must now be shown in actual demonstrations of large-scale systems, cycled according to the duty cycles required for an application in real-world settings. These demonstrations by various Na-ion companies are already underway, and with each successful demonstration, the market's confidence in this technology will progressively grow. If the near-term industrial targets for Na-ion energy and power cells can be met, mass penetration of Na-ion batteries in the market will be almost inevitable.

Status
Batteries using conventional Lithium-ion chemistries (LIBs) have long since been the power source solution of choice in a range of applications due to their high volumetric and gravimetric energy densities. Energy-dense, state-of-the-art LIBs in the pouch and cylindrical formats are currently achieving energy densities of up to 250 Wh kg −1 [294]. Beyond the current LIB designs, bulk all-solid-state batteries (ASSBs) are also being developed [295] to provide a significant increase in energy density (>400 Wh kg −1 ) and safety and to address the range anxiety with electric vehicles. However, it is not believed that the first generation will be a cost-effective solution for stationary (grid, renewables, telecoms, and domestic) applications, and timescales to market for a competitively performing product are expected to be way beyond 2025. Other 'beyond-lithium' chemistries, such as Li-air and rechargeable magnesium batteries, are still at very early stages of R&D [294]. Current analyses are focused on non-aqueous battery systems. For stationary and telecoms applications, size and weight are less limiting than, say, cost, safety, ease of maintenance, reliability, and cycle life. Therefore, this techno-economic assessment focuses on applications where it is believed that NIBs will be competitive, at least in the near term (5-10 years) based on their current state of development. Some of these application areas include stationary and telecoms. Their inherently low material cost and high technology safety make NIBs a very attractive proposition for stationary applications [252]. In addition to being made in similar formats to LIBs, NIBs can be made on existing LIB manufacturing lines-significantly reducing development costs and timescales. Current high-performance LIBs are considered less accessible (and safe) for these applications, although they have been proposed for second-life use [296]. Based on economics, Pb-acid and LiFePO 4 (LFP) are the current cathode chemistries of choice for stationary applications, compared to either LiNi 1-x-y Mn x Co y O 2 (NMC) or LiNi 1-x-y Co x Al y O 2 (NCA).
NIBs using a layered oxide, NaMO 2 (where M stands for Ni, Mn, Mg, or Ti) P2/O3-type structure and hard carbons (HCs) have been significantly advanced by Faradion and look very promising for stationary applications [297]. Low-cost NIBs based on Prussian blue analogue (PBA) anodes and cathodes and sodium vanadium fluorophosphates (NVPF) are also being developed by Natron energy and Tiamat respectively [298], but are considered less competitive for these applications, predominantly on a gravimetric and volumetric energy basis.
Using various cost models including BatPac, € kWh −1 has been estimated for NIB (NaMO 2 /HC), LFP and NMC (40%Ni:40%Mn:20%Cobalt), 18 650 cells, respectively [267,296]. The BatPac model included cell-design elements, depreciation, and warranty in addition to material and processing costs. The outputs from these models indicate comparable costs for LFP cells and NIBs (and lower costs for NMC cells). The models also indicate that the € kWh −1 for NIBs can be significantly decreased with volume-scale manufacturing and if the Wh kg −1 and Wh l −1 can be increased.

Current and Future challenges
Global energy demand is predicted to grow by up to 40% in the next ten years, including an increase of 2500 TWh in the transport sector, and the global power capacity of solar and wind are planned to be 2 TW and 1.5 TW, respectively, in the same timescale [299]. Combined with a global push to reduce fossil-fuel reliance, this will create a massive market for energy storage. Competitive battery technologies covering a wide performance, safety, and cost matrix will be required to support this growth [300].
In the short term, the first generation of NIBs has to prove that it is a contender in the energy-storage market, in terms of competitive performance, safety, cost, and maintenance, compared to incumbent technologies, such as Pb-acid, LIBs (LFP and NMC), and redox-flow batteries. The cost of development, production, and time to market will also have to be factored in. One of the major challenges in this space will be to match or, if possible, outperform state-of-the-art LIB cathodes such as NMC on a cost/performance, (€ kWh −1 ) basis and fit in with existing infrastructure. The majority of new stationary energy-storage installations use LIBs (currently, mostly LFP), which have benefitted from cost reductions and innovations in the EV market [300], and a bulk cost reduction could allow for a quick entrance into this application. Another major short-term challenge facing all new energy-storage markets is regulation and policy. Concerns around safety and the decisions to invest in different types of infrastructure will have a huge impact on the demand for energy storage and the associated costs of installation and use [300]. Future challenges will lie in securing and maintaining a competitive edge over other developing technologies. LIB roadmaps are looking at novel materials and architectures to increase energy density and reduce cost per kWh. Ease of recycling is another challenge that will need to be addressed, since the financial and environmental costs of disposal increase with the scale of adoption.

Advances in science and technology to meet challenges
In recent years, there has been a significant increase in worldwide NIB R&D activities [252,267], driven by potential cost and safety benefits compared to LIBs, as well as compatibility with LIB manufacturing processes and applications. Next-generation cells will seek to improve energy density beyond 170 Wh kg −1 [301] and improve other cell characteristics without compromising safety. This will further reduce the € kWh −1 cost and make them more competitive for automotive and other applications [252,267].
Advanced NIBs based on current-generation chemistries will improve in gravimetric and volumetric energy density, simply based on material, electrode, cell design, and engineering improvements. One area of improvement will be to increase the physical and active content and loadings and the density of the electrodes (and tap densities of active materials) to maximise the volumetric capacity. Electrolyte loadings can also be optimised to reduce cell weight.
First-cycle efficiencies (FCEs) for hard carbons are also quite low compared to that of graphite (<78% compared with >92%); investigating the use of a range of hard carbon/electrolyte combinations with higher FCEs will further increase the energy density. Hard carbons with lower average discharge voltages will increase the average cell operating voltage and the overall gravimetric and volumetric energy densities of the cells. Implementing these design and engineering optimisations could increase the energy density of first-generation NIBs to 160 Wh kg −1 [301].
Next-generation cathode chemistries are also being investigated [252]. These will seek to reduce the amount of Ni in the cathode by replacing it with Fe and/or Mn to significantly reduce the cost. Lower processing costs are also an area to explore to improve € kWh −1 for first-generation NIBs.
Alloying elements such as Sb, Si, Ge, As, Sb, Se, Pb, and Bi are also being investigated as alternatives to hard carbon anodes [252]; it is believed that the second generation of anodes will use these as blends with hard carbon. Titanium phosphates (e.g. NaTi 2 (PO 4 ) 3 ) and titanium oxide (Na 2 Ti 3 O 7 ) insertion compounds are also being investigated as alternatives to hard carbons [252].
Compared to LIBs, the NIB anode and cathode spaces are relatively unmined, so there is still significant potential to further increase the energy density of second-generation NIBs beyond 170 Wh kg −1 ; 190 Wh kg −1 has been predicted by Barker [301] Concluding remarks This short techno-economic assessment has shown that the current generation of NIBs, if advanced to mass production, can be competitive against LFP (and replace Pb-Acid) and NMC cells for stationary and telecoms applications. NIBs show massive advantages in safety, compared to either LFP or NMC cells, and have the potential for reduced cost (predominantly because Na is significantly more abundant than Li) and improved cycle life. The current generation of NIBs has the potential to exceed 150 Wh kg −1 through electrode and cell design optimisation. As a new technology, NIBs also have the added advantage of being a 'drop-in' replacement for LIBs, as the manufacturing processes are compatible.
Other potential benefits include the potential to store NIBs at 0 V without impacting safety and performance; this will allow for cheaper storage and transportation.
Due to the significant worldwide research in NIB materials, advances will probably lead to second-generation energy densities of >180 Wh kg −1 .
However, timescales for commercialisation may be long, and the timing of introduction to the market will be crucial to displace existing technologies. Satisfying existing regulatory and policy requirements is also a factor to be considered in the timescales for the commercialisation of NIBs. If Ni can be removed from the cathode formulation, then upcycling and disposal could be potentially easy and inexpensive.