Mixed alkali-ion transport and storage in atomic-disordered honeycomb layered NaKNi2TeO6

Honeycomb layered oxides constitute an emerging class of materials that show interesting physicochemical and electrochemical properties. However, the development of these materials is still limited. Here, we report the combined use of alkali atoms (Na and K) to produce a mixed-alkali honeycomb layered oxide material, namely, NaKNi2TeO6. Via transmission electron microscopy measurements, we reveal the local atomic structural disorders characterised by aperiodic stacking and incoherency in the alternating arrangement of Na and K atoms. We also investigate the possibility of mixed electrochemical transport and storage of Na+ and K+ ions in NaKNi2TeO6. In particular, we report an average discharge cell voltage of about 4 V and a specific capacity of around 80 mAh g–1 at low specific currents (i.e., < 10 mA g–1) when a NaKNi2TeO6-based positive electrode is combined with a room-temperature NaK liquid alloy negative electrode using an ionic liquid-based electrolyte solution. These results represent a step towards the use of tailored cathode active materials for “dendrite-free” electrochemical energy storage systems exploiting room-temperature liquid alkali metal alloy materials.


INTRODUCTION
Honeycomb layered oxides are a family of lamellar-structured nanomaterials characterised mainly by alkali or coinage metal atoms interleaved between sheets of transition metal atoms aligned in a honeycomb formation. This emerging class has been gaining momentous interest as a result of a variety of appealing properties innate to their structural framework. [1][2][3][4][5][6][7][8][9][10] The alkali-or coinage-metal atoms manifest weak interlayer bonds that engender an abundance of unoccupied sites that induce excellent ionic conductivities. This allows for facile reinsertion and extraction of alkali ions between the transition metal sheets, thus making them ideal cathode active material candidates for rechargeable alkali-based batteries. 1,2,[11][12][13][14][15][16][17] Furthermore, the sandwiching of nonmagnetic atoms between a hexagonal sublattice comprising magnetic atoms results in pseudo-two-dimensional magnetic structures that have the potential to achieve exotic magnetic states with varied applications in fields such as quantum computing and solidstate physics. 1,10 Most honeycomb layered oxides encompass compositions; A2M2DO6, A3M2DO6 or A4MDO6 where A is an alkali-or coinage metal atom (A = Li, Na, K, Cu, Ag, …), M is a transition metal atom (M = Ni, Co, Mg, Zn, Mn, Fe, Cr, …) and D is a highly-valent ion like Te, Sb, Bi or W. 11 In these compositions, A atoms are sandwiched between slabs comprising M atoms surrounded by D atoms in a hexagonal formation. Depending on the atomic size of the A atom, the resulting lamellar structures manifest different sequential arrangements (hereupon referred to as stacking orders). Until now, only a handful of honeycomb layered oxides adopting T2-, O1-, O3-and P2-type (in Hagenmuller-Delmas' notation) stacking orders have been identified, whereby 'T', 'O' and 'P' denote the tetrahedral, octahedral or prismatic coordination of oxygen with the A atoms, whilst the ensuing digit corresponds to the number of repeating transition metal layers in the unit cell. 18 In order to further expand the scope of known honeycomb layered oxides and capitalise on their full potential, it is imperative to not only explore hitherto uncharted territories of their compositional space but also scrutinise their emergent stacking orders.
Amongst other classes of layered transition metal oxides, fascinating structures have been developed through the mixing of two different alkali species to formulate AxA'yMO2 compositions. For instance, in the commonly studied LixNayCoO2 layered cobaltate, the intermixing of similar amounts of Na and Li results in unique configurations of Na and Li within the different layers giving rise to versatile stacking structures ranging from OP4 to OPP9 stacking sequences. 19,20 This structural versatility facilitates the development of various crystal structures with the potential to host different functionalities. Materials such as Li0.48Na0.35CoO2 have been found to exhibit a large thermoelectric power (thermopower) at room temperature, surpassing that of either of its parent materials, NayCoO2 and LixCoO2. 21 Furthermore, the unique OP4 stacking sequence has found great utility in battery application as it allows LixNayCoO2 to be utilised as a precursor in ionexchange synthesis to create a new polymorph electrode material LiCoO2 with O4stacking. 22,23 Although the intermixing of alkali ions appears to be a judicious approach to the next level development of honeycomb layered oxides, as far as we can tell, only one report on a set of metastable antimonates Li3-xNaxNi2SbO6 has been published on the topic. 24 Herein, we investigate a novel composition of Na2-xKxNi2TeO6 honeycomb layered oxides.
Furthermore, we unravel the structure of the new mixed alkali ion layered oxide NaKNi2TeO6 using atomic-resolution scanning transmission electron microscopy (STEM). Visualised for the first time, the local atomic structure reveals a unique and aperiodic stacking sequence in the layered mixed alkali ion compound. We investigate the ability of NaKNi2TeO6 to store Na or K ions when coupled with metallic sodium or potassium electrodes. NaKNi2TeO6 is also tested in combination with a liquid NaK alloy anode using an ionic liquid-based electrolyte solution for room-temperature dendrite-free rechargeable dual-cation cell.

RESULTS
In this study, we utilised Na2Ni2TeO6 and K2Ni2TeO6 as the parent materials for the creation of a novel stable mixed alkali ion phase. The P2-type stacking (crystallising in the centrosymmetric P63/mcm hexagonal space group) exhibited by both parent materials (Na2Ni2TeO6 and K2Ni2TeO6) is explicitly illustrated in Figure 1a, together with possible structural models for the resulting mixed compounds. Mixed-alkali ion honeycomb layered oxides adopting the composition of Na2-xKxNi2TeO6 (0 ≤ x ≤ 2) were synthesised via a high-temperature solid-state synthesis route described in the METHODS section. A preliminary characterisation of the average crystal structures of the as-synthesised (pristine) materials was carried out using powder X-ray diffraction (XRD) (as shown in Figure 1b). When the smaller Na atoms were replaced with larger K atoms (viz., increasing x from 0 to 2), a stepwise shift comprised of several diffraction peaks was observed. Particularly, two discrete shifts are seen as 00l peaks of the Na2-xKxNi2TeO6 as shown in Figure 1c. Since the diffraction angles of the 00l peaks are inversely proportional to the distance between adjacent transition metal slabs (hereafter referred to as the interlayer distance), each position corresponds to a phase with a different interlayer distance. Two of the phases have interlayer distances closely resembling the parent materials Na2Ni2TeO6 and K2Ni2TeO6, whilst the last phase has an intermediate interlayer distance, indicating the existence of both Na and K atoms in this phase.
In order to quantify and better illustrate how the lattice parameters of the phases present in Na2-xKxNi2TeO6 change with varying amounts of Na and K, profile fitting (Le Bail fit) of the XRD patterns was subsequently carried out. The average lattice parameters as deduced from the fit, are provided in the Supplementary Information ( Supplementary   Figure 1). Na2-xKxNi2TeO6 compositions where 0.2 ≤ x ≤ 1.8 are treated as two-phase mixtures, since the 00l Bragg peaks split into two separate peaks in these samples (as shown in Figure 1c). When K content (x, in Na2-xKxNi2TeO6) is increased with Na2Ni2TeO6 as the starting material, the relative intensity of the peaks corresponding to the Na2Ni2TeO6 phase decrease in favour of the new intermediate phase. It should be noted that when an equimolar ratio of Na and K is reached (i.e., NaKNi2TeO6), the Na2Ni2TeO6 peaks disappear and new peaks emerge (as shown in Figure 1d). With further increase in the K content, the relative intensities of the peaks corresponding to the intermediate phase decreases and some other peaks appear accordingly until pure K2Ni2TeO6 is attained.
To reiterate, amongst the Na2-xKxNi2TeO6 diffraction patterns, peaks previously not found in the parent materials were observed in the mixed alkali compound. As illustrated by  adopting the composition Na2-xKxNi2TeO6 (0 ≤ x ≤ 2). In the isostructural A2Ni2TeO6 (A = Na, K) compounds, Na atoms (in yellow) or K atoms (in orange) are sandwiched between layers or slabs consisting exclusively of TeO6 (blue) and NiO6 (purple) octahedra.
Black dashed lines denote the unit cell. Owing to the larger Shannon-Prewitt ionic radius of K + (1.38 Å) compared to Na + (1.02 Å), the interlayer distance of K2Ni2TeO6 is significantly larger than that of Na2Ni2TeO6. Various reasonable structural models can be hypothesised for the new series of compounds adopting the composition Na2-xKxNi2TeO6 Here, models where Na and K atoms are either mixed within the same layers or separated into different layers are shown. (b) XRD patterns of as-synthesised Na2-  used to index both Na2Ni2TeO6 and K2Ni2TeO6. Similar observations have been noted on a previous report on Li3-xNaxNi2SbO6 whereby disparate peaks emerged when Li and Na atoms were separated in different layers, 24 suggesting new cationic ordering in these materials. As such, the disappearance of the 102 Bragg peak and the emergence of a new set of peaks in close proximity underline the structural changes occurring in this intermediate phase.
The tendency of these mixed-alkali ion compositions to separate into two-phase mixtures can be rationalised by the large difference in the Shannon-Prewitt ionic radii of Na + and K + , 25 making it difficult to form a solid-solution compound from a mixture of Na2Ni2TeO6 and K2Ni2TeO6. Similar behaviour has also been observed amongst the layered nickelates NaxLi1-xNiO2, where three different polymorphs with intermediate twophase regions form depending on the stochiometric Li/Na ratio, presumably as a result of the large difference between the ionic radii of Li and Na. 26 To further elucidate the manner of alkali atom arrangement upon successful intermixing of Na and K and the corresponding structural changes previously indicated by the XRD patterns (Figure 1d), crystal structural analyses were employed on NaKNi2TeO6 whose composition is closest to a phase-pure sample. It is prudent to mention here that the elemental composition and thermal stability of NaKNi2TeO6 was ascertained by inductively-coupled plasma measurements and thermal gravimetric analyses, respectively (Supplementary Table 2 Synchrotron XRD measurements of NaKNi2TeO6 were performed for in-depth structural analyses, and its refinement result is shown in Supplementary Figure 11. We performed a Le Bail profile fitting using a hexagonal unit cell, yielding the lattice parameters of a = 5.2258(1) Å and c = 11.7875(6) Å and the reliability factors of Rwp = 8.96%, Rp = 6.27% and goodness-of-fit (GOF) = 4.12. We initially considered the reported structure of Na1.97Ni2TeO6 in the P63/mcm space group, 11 but disregarded this hexagonal space group because of the presence of the 001 and 003 reflections at ~3.01º and 9.05º. We also noticed that several peaks could not be fitted well with the initial structure considered. For example, the 100 and 003 reflections (Supplementary Figure 11b) were tailed and shifted towards higher angles, as previously observed in several layered oxides containing stacking faults (e.g., Li2NiO3 59 ). Preliminarily, we adopted several hexagonal models (some of which are shown in Supplementary Figure 12), in which Na and K are alternately arranged in a honeycomb layered framework. To ascertain information about the structure of NaKNi2TeO6, especially the stacking arrangement, atomic-resolution imaging of pristine NaKNi2TeO6 was conducted along several zone axes. Information on sample preparation, measurement protocols, and caveats undertaken are explicated in the METHODS section.  should be possible to acquire an analogous image for NaKNi2TeO6 if its honeycomb slab structure is similar to that of K2Ni2TeO6. However, Figure 2a shows that all atomic sites in the image share the same intensity, indicating "overlapping" of Ni and Te atoms in adjacent layers. It was also elusive to discern lighter (lower atomic mass) elements such as K and Na in the corresponding annular bright-field (ABF)-STEM images (shown in Figure 2b).
To clarify the structural changes responsible for the overlap of Ni and Te atoms observed in NaKNi2TeO6, the crystallites were examined from different directions (zone axes). A view from the [100] direction reveals the lamellar nature of the structure -in the HAADF-STEM image (Figure 2c), Te and Ni atoms correspond to a set of bright planes. This assignment was further confirmed by augmenting the STEM images with energy dispersive X-ray spectroscopy (EDX), as shown in Supplementary Figure 14. Elemental mapping using EDX also reveals that Na and K atoms are sandwiched between these Ni/Te-slabs. Furthermore, the STEM images reveal that Na and K are separated into different layers, instead of being randomly mixed within the same layers (Figures 2d-f).
Further analyses of atomic-resolution STEM images from the [100] direction allow the stacking order in NaKNi2TeO6 to be characterised. Notably, the shift between adjacent Ni/Te slabs is contingent on whether the interlayer space is occupied by Na or K atoms  (Figures 2a and   2b), presumably owing to their overlapping. The direction of the slab shifts seems to be random lacking any periodicity, which explains the difficulties encountered in the structural and profile analyses of the powder diffraction patterns.
Attempts to determine whether the structure is truly random or aperiodic, albeit ordered to some degree, proved elusive. From an enlarged view of the atomic-scale HAADF- STEM mapping (Figure 2h), it is also evident that the interlayer distance depends on the alkali atom species sandwiched between adjacent Ni/Te layers. The Ni/Te layers with Na atoms are separated by 0.55 nm (5.5 Å) whilst the interlayer distance for the layers with K is 0.62 nm (6.2 Å). It is worth highlighting that these interlayer distances attained closely resemble those of the parent compounds Na2Ni2TeO6 and K2Ni2TeO6 (Figure 1a). Figure 2i) quantitatively illustrate the alternating interlayer distances of Na and K atoms. Exclusively relying on the XRD patterns only yields the average of these interlayer spacings/distances (Supplementary Figure 1), accentuating the efficacy of using TEM to obtain important structural information.

The intensity line profiles (shown in
Although the atomic structure of NaKNi2TeO6 exhibits significant aperiodicity, a partial structural model containing the slab shift can still be constructed based on the TEM analysis.   Correspondingly, in the ABF-STEM superimposition, the Na and K atoms are seen as bright grey spots between the darker Te/Ni-atom planes (Figure 3h, showing a clear difference between Na and K atom sites. The prismatic coordination of oxygen is clearly seen in the ABF-STEM images in both the Na and K layers. However, Na-atom sites are equidistantly spaced, whilst two adjacent K atom sites are grouped together. Kinematically simulated SAED patterns in the [11 ̅ 0] direction generated based on the model agree well with the experimentally determined diffraction pattern (Figures 3i-j).  Supplementary Tables 3 and 4. However, several issues with peak intensities and asymmetric peak profiles, which arise from the stacking disorder of the transition metal slab layers are apparent in both the XRD and ND data that warranted further detailed structural analyses using TEM.
Structural intricacies of the mixed-alkali honeycomb layered oxide framework of NaKNi2TeO6 were further divulged by atomic-resolution STEM images ( Supplementary   Figure 19), which reveal the existence of stacking disorders/faults wherein 1/3 shifts of the Te / Ni slabs are observed. To further quantitatively scrutinise the nature of the stacking faults innate in NaKNi2TeO6 as revealed by STEM (Supplementary Figure 19), the FAULTS program was employed. 54 The atomic structural model indexed in the 62 hexagonal space group was used as the initial model to perform the analyses of the

DISCUSSION
A detailed characterisation of the crystal structure of the mixed-alkali honeycomb layered oxide (NaKNi2TeO6) was achieved using aberration-corrected STEM. To our knowledge, previous reports on the local atomic structure of a mixed-alkali ion layered oxide with similar structure have not been previously reported, making information on this class of materials obscure and underutilised. In the course of this study, a prominent aspect that emerges is the difference between Na and K sites evident when NaKNi2TeO6 crystals are viewed along the [11 ̅ 0] and [100] zone axes. We find that, Na atoms are distributed in sites that assume triangular patterns (Figure 4a) whilst the K atoms reside in sites arranged in honeycomb formations (Figure 4b). Although beyond the limits of present experiments, these atomic configurations and crystallographic occupations can be used to predict emergent properties of such materials as well as the electrodynamics of the alkali or coinage atoms within the realm of their electromagnetic behaviour, electrochemistry, quantum phenomena etc. 30 For instance, the atomic arrangements of the atoms given in In particular, the free energy of the alkali atom layer is expected to be minimised (or equivalently, the entropy maximised) to achieve stability of the crystal especially when the Na or K atoms are arranged in a honeycomb fashion. This follows from the honeycomb conjecture, which states that the honeycomb lattice is the most efficient way of packing any two-dimensional (2D) surface with equal size unit cells of maximum area and minimum perimeter. 44 Thus, taking the free energy to scale with the perimeter of the unit cells (honeycomb, triangular etc.), and the entropy to scale with the area, A, the lattice exhibited by K atoms in Figure 4b satisfies this conjecture by maintaining its honeycomb pattern whilst Na atoms in Figure 4a do not.
For the sake of rigour, we make a straightforward approximation for the free energy, using the thermodynamics formula, = − B ln , where is a constant related to the geometry of the surface (Gaussian curvature), 30 = B ln is the entropy contribution to the free energy and ≃ − is the internal potential energy corresponding to various binding energies of the alkali atoms (Na, K) to each other, whose leading term is taken to be their activation energy, . Following this formula, the entire material comprising a series of such layers has to maximise entropy, even for the mixed alkali atom honeycomb layered oxides. As affirmed earlier, the highest entropy the Na + ion probability density profile in the Na-layer (mapped onto 2 × 2 unit cells) using common colour bars (shown on the right). The population contours reflects that the preferred migration pathway amongst the interstitial cationic sites. (d) MD simulation for K + ion probability density profile in the K-layer. (e) Voltage profiles of the NaKNi2TeO6based electrode tested using Na metal as counter electrode and Na0.20Pyr0.80FSI ionic liquid as electrolyte solution at 6.65 mA g -1 and (f) Voltage profiles of the NaKNi2TeO6based electrode tested using K metal as counter electrode and K0.20Pyr0.80FSI ionic liquid as electrolyte solution at 6.65 mA g -1 . Voltage-capacity profiles set at a lower cut-off voltage of 2.5 V have been furnished in the Supplementary Information   (Supplementary Figure 23).
configuration leading to a minimised free energy and a stable crystalline structure is the one where both types of alkali atoms are arranged in a honeycomb fashion. The next favoured configuration is the one that allows for only one type of alkali atom to be in a honeycomb fashion, case in point being the configuration observed in NaKNi2TeO6 as displayed in Figures 4a and 4b. Moreover, since the activation energy of K is lower than that of Na, 1,45 it is apparent, by setting = 0, that the Na layer can still minimise its free energy by disrupting its honeycomb configuration into e.g. triangular patterns shown in  Figures 4c and 4d also show the similar behavior within the structure of NaKNi2TeO6. A few high-density areas are identified in the population profile, indicating favourable sites of Na + or K + ions. Particularly, the highdensity areas are well-connected for the Na-layer, whereas a modest connectivity amongst the high-density sites in K-layer is observed, resulting in higher diffusion of Na + ion compared to K + ion. It is worth to mention that this behaviour is different than the parent Na-or K-systems i.e. A2Ni2TeO6, where A = Na, K. This can be traced to the vastly different NiO6 and TeO6 octahedral stacking sequences from the parent Na or K-systems, which leads to a differing local environment. 1,44 Recall, we observed in STEM (Figure   2g) that, the layers where K atoms occupy the interlayer space, Te / Ni slabs are not shifted with respect to each other whereas, for layers where Na atoms reside, shifts of the Te / Ni slabs are observed. This behaviour is maintained in the simulation results despite high temperatures where the alkali ions are dynamic. Supplementary Video 1 indeed shows the dynamic behaviour of Na + and K + atoms, when simulated at 600 K. Further investigation of the nature of Na and K ion transport and its mechanism is beyond the scope of this work.
Presumably, NaKNi2TeO6 like other honeycomb layered oxides, holds potential in many fields. Nonetheless, the primary focus of this study is to ascertain its feasibility as positive electrode active material for alkali-metal batteries. Thus, electrochemical energy storage tests were carried out to verify its ability to transport and store mixed alkali ions. For such reasons, Na and K half-cells were assembled, as further explicated in the METHODS section and Supplementary Information (Supplementary Figure 23). Even though tellurium is not a constituent element of choice for energy storage systems entailing Earthabundant elements, the insights obtained herein are not exclusive to the mixed-alkali tellurate systems. Indeed, the present study should be considered as a fundamental scientific research work rather than an applied one. Moreover, we do not rule out the use of tellurates in niche applications where functionality may be prioritised over cost. Figure   4e shows the voltage-capacity plots of NaKNi2TeO6 in Na half-cells. The theoretical capacity for a full Na + extraction from NaKNi2TeO6 is approximately 67 mAh g -1 .
However, a reversible capacity of ca. 80 mAh g -1 was attained upon subsequent cycling, suggesting the occurrence of K + extraction. In the case of the K half-cells (Figure 4f), an initial capacity of 45 mAh g -1 was realised and maintained upon successive cycling. This capacity presumably arises from predominant K + extraction given that K metal was used.
Post-mortem imaging of NaKNi2TeO6 electrodes subsequently cycled in Na-and K-half cells were performed using high-resolution STEM, in order to ascertain the nature of the intercalation and de-intercalation process of the alkali-ions. Figure 5a shows the ex situ HAADF-STEM images of a NaKNi2TeO6 electrode upon subsequent cycling (i.e., fully discharged sample at the third cycle) in Na-half cells taken at the [100] zone axis and the corresponding ABF-STEM images are shown in Figure 5b. SAED patterns taken along the [100] axis are shown in Figure 5c. Subsequent cycling of NaKNi2TeO6 in Na halfcells leads to the replacement of K + with Na + to yield a Na-rich phase composition (Na2Ni2TeO6), as affirmed by the equidistant interlayer spacings (0.55 nm) of Na atoms along the [001] axis as quantitatively illustrated by the intensity line profiles (Figure 5d) for the highlighted area in the HAADF-STEM images (shown in Figure 5a). Streaks are evinced in the SAED patterns taken at the [100] axis (Figure 5c), indicating modulation in the arrangement of Na atoms along the ab plane as is exemplified in Na2Ni2TeO6. 57 Figures 5e and 5f show, respectively, the ex situ HAADF-and ABF-STEM micrographs of a NaKNi2TeO6 electrode upon cycling in K-half cells taken at [100] zone axis. Figure   5g shows the corresponding SAED patterns. Alternating interlayer spacings (0.55 nm and 0.62 nm) of Na and K atoms along the [001] axis is observed, indicating that the mixed alkali layered framework is retained owing to reversible extraction and insertion of K alone. Voltage-capacity plots of NaKNi2TeO6 upon subsequent cycling in K half-cells reveal a reversible initial capacity of approximately 50 mAh g -1 (Figure 4f). The theoretical capacity for a full K + extraction from NaKNi2TeO6 is approximately 67 mAh g -1 , which indicates that the capacity arises predominantly from K + extraction. Given that  (Supplementary Figures 24 and 25).
K metal was used as anode (counter electrode), reversible extraction and reinsertion of K can be envisaged, as is validated experimentally judging from the intensity line profile (Figure 5h) that quantitatively show alternating interlayer distances occupied by Na and K atoms. Further, SAED patterns (Figure 5g) show streaks indicative of the aperiodic stacking nature along the c-axis.
These electrochemical measurements indicate that NaKNi2TeO6 mixed-alkali honeycomb layered oxide is amenable to electrochemical binary alkali-ion transport and storage, pointing towards the possibility of developing a viable mixed Na-and K-ion electrochemical cell that relies on electrolytes and electrode materials that can accommodate both Na and K binary-cation transport. Given that cells utilising both cation and anion as charge carriers (dual-ion batteries (DIBs)) have already shown remarkable metrics in terms of energy density, power density and cycling life, 42,46-48 the present battery chemistry exploiting binary alkali metal cations could be a promising successor to DIB technology. 45,46,60,61 Indeed, taking into account the abundance of Na and K, a cell with suitable specific energy and cyclability can be designed. Moreover, it offers the possibility of utilising a NaK liquid metal alloy (albeit at its nascent stage of development) as anode material which can be effective in accommodating the cations, thwarting the formation of dendrites that have long plagued the direct utilisation of alkali metal anodes in secondary batteries (as illustrated in Figure 6a). 60,61 Voltage-capacity plots of NaKNi2TeO6 when initially cycled in a NaK half-cell, using a mixed electrolyte based on Voltage-capacity profiles of NaKNi2TeO6 dual-cation cathode material in a cell using NaK as anode during initial cycling at a specific current of 6.7 mA g -1 . The voltage range was set at 4.6 -1.3 V. The electrolyte used was a pyrrolidinium-based dual-metal-cation ionic liquid, comprising equimolar amounts of Na and K (i.e., Na0.10K0.10Pyr0.80FSI).  (Figure 6b). The theoretical capacity of NaKNi2TeO6 is 134 mAh g -1 , assuming a full extraction of Na + and K + in a NaK cell.
Although about 60% of theoretical capacity is attained, the performance is promising considering no electrode optimisation (carbon-coating, nanosizing, etc.) has been undertaken. The corresponding voltage-capacity plots at various specific currents are shown in Figure 6c, indicating NaKNi2TeO6 sustains decent rate capabilities. Moreover, reversible electrochemical behaviour is observed for 20 cycles at a specific current of 13.4 mA g -1 , as shown in Figure 6d, and with good Coulombic efficiency (Supplementary Figure 27). To fully tap the potential of NaKNi2TeO6, further electrode optimisation strategies are warranted, which is a subject of future work. Ionic liquids were utilised owing to their higher stability at high-voltage regimes, in comparison to etherand ester-based solvent electrolytes. [48][49][50] Figure 6e shows voltage-capacity comparison plots of NaKNi2TeO6 along with cathode materials reported for NaK battery system. NaKNi2TeO6 is the first material to have both Na and K initially stabilised in its layered framework and offers a high discharge voltage along with relatively high capacity. The high capacity of NaKNi2TeO6 could be associated with the redox process of Ni. Indeed, X-ray photoelectron spectroscopy measurements (shown in Supplementary Figure 28) indicate the participation of Ni to the charge compensation process, whereas Te is dormant, as has been noted in related honeycomb layered oxides such as K2Ni2TeO6 and Na4NiTeO6. 2,4 With a judicious choice of constituent elements, we speculate that mixed alkali compositions could exhibit even higher voltage and capacity.
Since both Na and K reinsertion can be envisaged during cycling of NaKNi2TeO6 in NaK cells, further structural insights were attained from high-resolution STEM ex situ measurements. Figures 7a and 7b show the ex situ HAADF-STEM images along the axis revealing NaKNi2TeO6 to preferentially insert Na atoms into the lattice. SAED patterns taken along (d) [100] axis and (e) [11 ̅ 0] axis showing streaks (underpinned in green arrows) indicative of the modulation in the Na arrangement as is exemplified in Na2Ni2TeO6. 57 [100] and [1 1 ̅ 0] zone axes, respectively, for a NaKNi2TeO6 crystallite taken after subsequent cycling (i.e., fully discharged sample at the third cycle) in NaK cell.
Equidistant interlayer spacings are apparent, suggesting that one type of alkali-ion is reversibly reinserted into NaKNi2TeO6 upon successive cycling. Intensity line profiles reveal equidistant interlayer spacings (0.55 nm) corresponding to Na atoms in the lattice (Figure 7c and Supplementary Figure 29). Further, SAED patterns taken along the [100] and [11 ̅ 0] zone axes (Figures 7d and 7e) evidence streaks reminiscent of the modulation in the Na arrangement in Na-rich phase (Na2Ni2TeO6). 57 Whilst only Na-rich phases are observed in TEM measurements, ex situ XRD measurements of discharged electrodes (Supplementary Figure 30) reveal K-rich and mixed alkali phases with Narich phase being predominant. These results reveal that there is a propensity of NaKNi2TeO6 to predominantly reinsert Na ions when cycled in NaK cells. Presumably, this could be rooted in the stability of Na2Ni2TeO6 over other phases, as affirmed by theoretical computations (Supplementary Figures 31, 32 and 33 It is imperative to mention that the concept of mixed-alkali battery materials can be beneficially applied to tailor the ability to transport and store alkali metal ions. For instance, the partial substitution of Li atoms in layered transition metal cathode oxides with Na, K, Rb or Cs is a well-investigated route to enhance their structural stability and increase Li-ion diffusion. [31][32][33][34][35][36][37][38][39] However, the theoretical capacity of materials with equimolar amounts (i.e., 50% atomic fraction) of different alkali metal atoms, such as NaKNi2TeO6, is drastically attenuated when large amounts of Na are replaced with K in a cathode for a Na-battery or vice versa. This is generally ascribed to the fact that the number of cations participating in extraction and reinsertion would be diminished. As a way to enhance performance, the utilisation of an alloy such as NaK in the case of NaKNi2TeO6 would facilitate the participation of both cation species thus yield high theoretical capacity. A schematic showing the conceptual design of such a battery is shown in Figure 6a. In addition, the liquid nature of NaK does not allow the formation of dendrite on the anode thus rendering the design 'a dendrite-free' metal anode cell. 45,46 Therefore, this concept showcases the potential for NaKNi2TeO6 and related mixed-alkali layered oxide materials as functional materials. In conclusion, the successful design of mixed-alkali honeycomb layered oxides, for instance NaKNi2TeO6, not only offers a conduit to engineering new functional materials but also promises to expand the compositional space of known honeycomb layered oxides.
The results of this study reaffirm the correlation between the ionic radii of the alkali atoms and the interlayer distance even for the mixed-alkali system, which can be exploited to configure the intricate interlayer structure of the mixed-alkali honeycomb layered oxides, by the same token as (Li/Ag)CoO2. 40 Detailed local atomic information provided through a series of scanning transmission electron microscopy reveal characteristics of a unique aperiodic stacking structure, suggesting structural versatility that could unlock the potential of this material for electromagnetic, quantum and electrochemical functionalities. 1,30,41 Further, we expound on the feasibility of NaKNi2TeO6 for battery applications that utilise mixed cation transport. The mixed triangular and honeycomb atomics conformations may have profound impact on the electrodynamics of the alkali ions. An attempt to rationalise this view has been made by theoretical computations. We hope that this work will serve as a cornerstone for further augmentation of mixed-alkali layered oxides in various realms of science and technology. diameter. Neutron diffraction data taken using the backscattering bank were evaluated and refined using the FULLPROF suite, JANA2006, and Z-Rietveld softwares. 43,55 VESTA was used to display the refined crystal structures. 58  STEM-EDX (energy dispersive X-ray spectroscopy) spectrum images were obtained with two JEOL JED 2300T SDD-type detectors with 100 mm 2 detecting area whose total detection solid angle was 1.6 sr. Elemental maps were extracted using Thermo Fisher Scientific Noran (NSS) X-ray analyser.

METHODS
Thermal stability measurements: Thermogravimetric and differential thermal analysis (TG-DTA) was performed using a Bruker AXS 2020SA TG-DTA instrument in the temperature ranges of 25 to 900 °C. The measurements were performed at a ramp rate of 5 °C min −1 under argon using a platinum crucible. Thermal stability measurements reveal NaKNi2TeO6 mixed alkali honeycomb layered oxide to be stable to up to 800 °C ( Supplementary Figure 9), as further confirmed by XRD measurements (see Supplementary Figure 10).
Electrochemical measurements: Coin cell assembly protocols were performed in an Arpurged glove box (MIWA, MDB-1KP-0 type) with H2O and O2 contents less than 1 ppm.
For electrode fabrication, pristine NaKNi2TeO6 was mixed with carbon black and polyvinylidene fluoride (PVdF) to attain a final weight ratio of active material : carbon : binder in the cathode of 70: 15: 15. The mixture was suspended in N-methyl-2pyrrolidinone (NMP) to obtain a viscous slurry, which was then cast on tungsten foils with a typical mass loading of ~5 mg cm -2 . Note that tungsten foil was preferred to aluminium foil as current collector, owing to its high oxidative stability. Electrodes with a geometric area of 1 cm 2 were punched and dried at 120 °C in a vacuum oven. The average thickness of the electrodes was around 50 μm. Electrochemical properties of the materials were evaluated in 2032-type coin cells using NaKNi2TeO6-based positive electrode (i.e., the working electrode) separated from the K, Na or NaK metal negative electrodes (i.e., the counter/reference electrodes) by glass fibre discs soaked in electrolyte.
The total volume of electrolyte used per coin cell was 80 microlitres. Sodium-potassium (NaK) liquid alloy was prepared at room temperature in an argon-purged glove box.
Sodium and potassium were mixed physically in a weight percentage of 54.1 and 40.9, respectively, in a glass vial. The shiny sodium-potassium alloy was formed spontaneously upon mixing sodium (purity of 99 %, Kishida Chemicals) and potassium (purity of 99.95%, Kishida Chemicals) lumps. Sodium-potassium alloy was immersed in superdehydrated hexane (water content of < 10 ppm) in order to avert the formation of oxides at the surface. A glass syringe was used to take a portion of the liquid alloy for assembly of the cells. Sodium-potassium alloy (which has a high surface tension) was immobilised in the coin cells using aluminium meshes (100-mesh size and purity of 99% (Nilaco in Pyr13FSI ionic liquid was used for Na half-cells, 48,49 whereas a 0.5 mol dm -3 equimolar mixture of NaFSI and KFSI in Pyr13FSI ionic liquid was used in assembling NaK cells. 47,50 Protocols relating to the preparation of these electrolytes have been detailed elsewhere. 8,48,49 Molar ratio of the ionic liquids were calculated based on their molecular weight and density at 298 K. The densities for 0.5 mol dm -3 NaFSI + 0.5 mol dm -3 KFSI/Pyr13FSI, 1.0 mol dm -3 NaFSI/Pyr13FSI, 1.0 mol dm -3 KFSI/Pyr13FSI, and 0.5 mol dm -3 KTFSI/Pyr13TFSI were 1.4323, 1,4119, 1.4227, and 1.4283 g cm -3 , respectively. The concentration (expressed in molar ratio) of the electrolytes are respectively as follows: Na0.10K0.10Pyr0.80FSI, Na0.20Pyr0.80FSI, K0.20Pyr0.80FSI and K0.13Pyr0.87TFSI. For clarity to readers, pyrrolidinium-based ionic liquid is abbreviated in literature as Pyr(r)13TFSA (Pyr(r)13TFSI or Pyr(r)13Tf2N)).
The physicochemical properties of the mixed ionic liquid (0.5 mol dm -3 NaFSI + 0.5 mol dm -3 KFSI in Pyr13FSI) used, in the present study, as an electrolyte for the NaK half-cell are shown in Supplementary Table 8. Galvanostatic cycling tests were carried out applying specific currents ranging from 6.7 mA g -1 to 670 mA g -1 . Unless otherwise stated, the cut-off voltage was set at 1.2 V to 4.35 V for the Na half-cells or 1.3 V to 4.6 V as for the K half-cells. A cut-off voltage of 1.3 V to 4.35 V was set for the sodium-potassium full-cells. All the electrochemical measurements were performed at room-temperature (25 ±1 o C) with temperature maintained using a temperature-controlled oven (ESPEC).
Theoretical computations: Molecular dynamics (MD) simulations at constant pressure and temperature (N, P, T) were carried out using the Vashishta-Rahman type interatomic pair potential that has been used remarkably well for a variety of system including related honeycomb layered oxides such as Na2M2TeO6 (M = Mg, Zn, Co, Ni), 16,17,52 (eq. 1) where qi is the charge and σi is the ionic radius of the i th ion.
is the inter-atomic distance. The parameters, Aij, Pij, and Cij, are the short-range interaction parameters, between ion pairs i and j. Simulations were done using the software package LAMMPS. 53 The pressure and temperature in the system were controlled using Nose-Hoover type thermostatting and barostatting techniques. The interaction between pairs of atoms was chosen from the previous report, 16,17,52 as listed in Supplementary Table 5, and a few of them (indicated by foot note in Supplementary Table 5 V. The attained XPS spectra were fitted using Gaussian functions and data processing protocols were performed using COMPRO software.