NASICON-type air-stable and all-climate cathode for sodium-ion batteries with low cost and high-power density — TEI-tagged view

Mingzhe ChenpersName, Weibo HuapersName, Jin XiaopersName, David CortiepersName, Weihua ChenpersName, Enhui WangpersName, Zhe HupersName, Qinfen GupersName, Xiaolin WangpersName, Sylvio IndrispersName, Shu-Lei ChoupersName, Shi-Xue DoupersName (2019date). “NASICON-type air-stable and all-climate cathode for sodium-ion batteries with low cost and high-power density.” Nature Communicationstitle·j.

Object of study — a NASICON-type Na4Fe3(PO4)2(P2O7)/C nanocomposite cathodeterm·key for sodium-ion batteriesterm

Research question — how to make low-cost, long-lasting all-climateterm·key cathodesterm·key for sodium-ion batteriesterm

Contribution — a low-cost, environmentally friendly iron-basedterm·key cathodeterm·key competitive for sodium-ion batteriesterm

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

NASICON-type air-stable and all-climateterm·key cathodeterm·key for sodium-ion batteriesterm with low cost and high-power density Abstract The development of low-cost and long-lasting all-climateterm·key cathodeterm·key materials for the sodium ion battery has been one of the key issues for the success of large-scale energy storage. One option is the utilization of earth-abundant elements such as iron. Here, we synthesize a NASICON-type tuneable Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 )/C nanocomposite which shows both excellent rate performance and outstanding cycling stabilityterm·key over more than 4400 cycles. Its air stabilityterm·key and all-climateterm·key properties are investigated, and its potential as the sodium host in full cellsterm·key has been studied. A remarkably low volume changeterm·dom of 4.0% is observed. Its high sodium diffusion coefficient has been measured and analysed via first-principles calculations, and its three-dimensional sodium ion diffusion pathwaysterm·key are identified. Our results indicate that this low-cost and environmentally friendly Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 )/C nanocomposite could be a competitive candidate material for sodium ion batteriesterm.

Introduction Cost-efficient large-scale energy storage systemsterm·dom (EESsterm·dom) have been in high demand in recent years due to the rapid development of renewable energy resources (i.e., solar, wind, geothermal, and tidal energy) 1 , 2 . Although advanced lithium-ion batteryterm (LIBsterm) technology has led to commercially viable electric vehiclesterm (EVsterm) and provided a satisfactory energy density level for large-scale EESsterm·dom, the issues of lithium costs and the uneven global distribution of lithium resources have hindered their further application in the large-scale energy storage field 3 , 4 . Sodium-ion batteriesterm (SIBsterm) have been considered as promising candidates for EESsterm·dom because of the high abundance, wide availability, and low cost of sodium resources 5 , 6 . To achieve high-performance SIBsterm, however, is still a great challenge, because it is necessary to discover or develop more suitable electrodes to meet the requirements of high energy density and high cycling stabilityterm·key in the most cost-efficient way 7 . The use of the redox chemistry of earth-abundant transition metalsterm, however, will greatly reduce manufacturing coststerm·dom, which will boost the real applications of SIBsterm in the commercial EESterm·dom market 8 – 10 .

The polyanionic or mixed-polyanion system is one of the most important candidates among the various types of electrode materials suitable for SIBsterm, and it has been the subject of extensive investigations for applications in recent years 11 . The three-dimensional (3D) framework of such electrodes can provide strong and lasting structural support for repeated Na + ion de-/insertions with relatively high operating potentials. Several vanadium-based compounds have been widely studied as an important group of promising electrodes, such as sodium superionic conductorterm·dom (NASICONterm·dom)-type Na 3 V 2 (PO 4 ) 3 12 , 13 , Na 3 (VO x ) 2 (PO 4 )F 3−2 x ( x = 0 or 1) 14 , 15 , Na 7 V 4 (P 2 O 7 ) 4 PO 4 16 , 17 , etc. These compounds exhibit satisfactory high energy densities that are comparable to those of LIBsterm, based on multi-electron redox reactions (V 3+ /V 4+ and V 4+ /V 5+ ) and high operating voltage, although the use of toxic and expensive V element remains a critical issue in real applications. Other electrodes based on costly 3 d transition metal elements such as Ni-based or Co-based electrodes also face this problem. Therefore, the most abundant and non-toxic 3 d element, iron, is the first choice as the redox centre in the polyanionic or mixed-polyanion system 18 . The Fe-based polyanionic compounds have shown high structural stability with small volume changeterm·dom during cycling and sufficiently long cycling stabilityterm·key with high energy density, which will definitely reduce the overall electrode cost and promote the real application of SIBsterm in large-scale EESsterm·dom in the near future 8 , 19 – 21 .

Another critical issue for both LIBsterm and SIBsterm is their performance and reliability at the temperatures that might be encountered under all-climateterm·key conditions. The commercial LiFePO 4 material only can deliver ~70 % of its theoretical capacity when the temperature is reduced to −20 °C, which significantly lowers its applicable energy density 22 , 23 . SIBsterm face an even more severe low-temperature problem due to their intrinsically more sluggish solid-diffusion process compared with LIBsterm 24 . Also, at high temperaturesterm·dom (over 40 °C), some cathodesterm·key will encounter inevitable capacity loss and degradation of their cycling stabilityterm·key, such as layered oxide materials for both LIBsterm and SIBsterm 25 , 26 . Cathodeterm·key Na + -host materials operated at room temperature have been extensively investigated, but their all-climateterm·key performance has received little attention in recent years 7 , 27 , 28 . For real applications of SIBsterm in the commercial market, all-climateterm·key performance is strongly required, because the energy storage systemsterm·dom must work over a wide range of atmospheric temperatures. Therefore, finding an environmentally friendly, low-cost, and all-climateterm·key cathodeterm·key material is very important for the real application of SIBsterm.

With these considerations in mind, we screened a variety of iron-basedterm·key polyanionic materials, and the recently reported mixed crystalline framework structure represented by the ortho-pyrophosphates, such as Na 4 Fe 3 (PO 4 ) 2 P 2 O 7 , came to our attention 29 – 32 . Na 4 Fe 3 (PO 4 ) 2 P 2 O 7 has not been well studied, but it possesses unexpected 3D sodium diffusion pathwaysterm·key in the sturdy crystal framework of a typical NASICON-type structure, which can be considered as the key preliminary factor for long-term cyclability and low-temperature performance 5 , 19 .

Here, in this paper, we report a low-cost NASICON-type cathodeterm·key material, Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ), with favourable Na storage properties, that demonstrates fast and stable electrochemical properties under all-climateterm·key temperatures, combined with tuneable carbon-coated nanoparticles to form a robust composite material without losing its crystallinity. The uniformly carbon-coated nanosized particles provide facile and rapid electron transport along with high ionic diffusion capability. Excellent rate performances were achieved, and impressive cycling stabilityterm·key was obtained at both room temperature and low/high temperatureterm·dom (−20 °C/50 °C). In addition, we found that this material is stable in air, even after exposure for three months, and Fe-based full SIBterm configurations are demonstrated using Fe 3 O 4 nanospheres/hard carbon as anode. In-situ synchrotron X-ray diffractionterm·mth (XRDterm·mth) and in-situ X-ray absorption near-edge structureterm·mth (XANESterm·mth) analyses further revealed the outstanding reversibility of Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) in detail. We also employed density functional theoryterm·mth (DFTterm·mth) studies as well as bond valence sumterm·mth (BVSterm·mth) calculations to examine the details of each possible sodium diffusion pathwayterm·key. Our results indicate that this new NASICON-type Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material is a promising cathodeterm·key candidate for the next generation of SIBsterm in the near future.

Results Materials characterizations Both nanosized Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) plates (denoted as NFPP-Eterm) and microporous Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) particles (denoted as NFPP-Cterm) materials were successfully synthesized via a facile one-step sol–gel method. The detailed synthesis procedures for both samples can be found in the experimental section. In order to obtain accurate phase information, high-resolution synchrotron powder X-ray diffractionterm·mth (s-XRD) was employed to determine the phase constitution and atomic site occupations. Figure 1a shows the satisfactory Rietveld refinement of NFPP-Eterm with a good weighted profile R-factor ( R wp = 6.97%). The sample was indexed to the orthorhombic Pn2 1 a space group, as previously reported by Kang’s group 29 , with lattice parameters α = 17.6433(6) Å, b = 6.3616(4) Å, c = 10.2043(4) Å, and V = 1145.342(3) Å 3 for NFPP-Eterm compound. The typical crystalline structure is presented in the inset of Fig. 1a . In the open 3D robust framework, and the [Fe 3 P 2 O 13 ] units along the a -axis each consist of three FeO 6 octahedra and three PO 4 groups, while the [Fe 3 P 2 O 13 ] ∞ infinite layers are connected by [P 2 O 7 ] groups along the a direction, thus leading to the formation of large primary tunnels along the b direction. The NFPP-Cterm sample also shows good crystallinity with the space group Pn2 1 a (as shown in Supplementary Fig. 1 ). We summarize the detailed atomic site information for both the NFPP-Eterm and the NFPP-Cterm samples in Supplementary Table 1 and Supplementary Table 2 . We also observed trace amounts of maricite (NaFePO 4 ) impurities (around 4% in both NFPP-Eterm and NFPP-Cterm samples). Supplementary Fig. 2 shows the paramagnetic properties (magnetization curve ( M-H ) and temperature dependence of the magnetic susceptibility ( M-T )) for both samples from probing the spin state of Fe. It can be seen that the hysteresis loop of NFPP-Cterm is slightly larger than that of NFPP-Eterm, which can be ascribed to its slightly higher proportion of impurities, as indicated by the refinement. The calculated effective magnetic moments ( μ eff ) of NFPP-Eterm and NFPP-Cterm are 5.04 μ B /Fe and 5.19 μ B /Fe per formula unit, respectively, which are close to the theoretical value (4.89 μ B /Fe per formula unit) for high-spin Fe 2+ ( d 6 , t 2g 4 e g 2 , S = 2) 33 . Fig. 1 Characterizations of as-prepared sample. Rietveld refinements of a NFPP-Eterm. Schematic representation of the refinement results is presented in the insets. b XPS spectra of C 1s and the corresponding deconvolution curves of both NFPP-Eterm and NFPP-Cterm samples. c Raman spectra of both samples in the Raman shift range from 100 cm −1 to 1800 cm −1 . d SEM images of NFPP-Eterm and e transmission electron microscope (TEM) image of NFPP-Eterm particles. f Bright field (BF) image of NFPP-Eterm with carbon layers. g HAADF image from aberration-corrected STEM and the crystal structureterm·dom of NFPP-Eterm viewed from the [010] direction. The insets are the corresponding signal responses along the selected lines. h The corresponding SAED pattern of NFPP-Eterm and i atomic force microscopy (AFM) image of NFPP-Eterm nanoplate particles and their corresponding heights. j The STEM-EDS mapping results for selected elements of NFPP-Eterm. Scale bars: 500 nm ( d ); 200 nm ( e ); 5 nm ( f ); 1 nm ( g ); 2 1/nm ( h ); 200 nm ( i ); 100 nm ( j ) The carbon contents of NFPP-Eterm and NFPP-Cterm were detected to be 3.6 wt. % and 4.1 wt. %, respectively, according to the thermogravimetric (TG) analysis in Supplementary Fig. 3 . We further characterized the carbon via employing X-ray photoelectron spectroscopy (XPS) and Raman spectroscopy. Peak deconvolutions were carried out to fit the C 1s spectra to three main bonding configurations (Fig. 1b ). The typical π -bonding graphite-like carbon peaks for delocalized electrons are located at 284.6 eV in both samples. Peaks with higher energies ( sp 3 carbon), namely, C-O (285.8 eV) and O=C-O (288.7 eV) were observed and fitted 13 . It was found that 65.9% of the carbon was graphitized carbon in the NFPP-Eterm sample, while this value in NFPP-Cterm was only 52.3%, which indicates that the carbon conductivity of NFPP-Eterm may higher than that of NFPP-Cterm. In addition, the Fe 2p XPS spectra in Supplementary Fig. 4 reveal the chemical compositions and surface electronic states of Fe element. No obvious difference can be detected except for a tiny binding energy shift, and the Fe deconvolution curves indicate that the valence of Fe in both samples is similar. The Raman spectra of Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) range from 150 to 800 cm −1 , as shown in Fig. 1c . The detected peaks at 218.7 cm −1 and 288.0 cm −1 can be assigned to the stretching vibrations and bending motions of the PO 4 tetrahedra, respectively, and the peak located at 402.1 cm −1 is mainly due to the stretching vibrations of FeO 6 octahedra 34 . It is worth noting that the intensity ratio of the D band to the G band ( I D / I G ) for the NFPP-Eterm sample (1.04) is larger than that for the NFPP-Cterm sample (0.94). The D-band is mainly due to defects, edges, or structural disorder while the G-band normally represents the E 2g mode of sp 2 carbon layers. This phenomenon indicates that the carbon conductivity of the NFPP-Eterm particle surface would be higher than that of NFPP-Cterm, which is in good accordance the C 1 s deconvolution curves in Fig. 1b . The typical vibration modes of both the PO 4 units and the FeO 6 octahedra were also detected in the Fourier transform infrared (FT-IR) spectrum (Supplementary Fig. 5 ). The predominant overlapped peak of the v 1 and v 3 symmetric and asymmetric stretching modes of the PO 4 units is around 1109 cm −1 , while the v 2 and v 4 symmetric and asymmetric bending modes of the PO 4 tetrahedra are responsible for the splitting of the O-P-O peaks in the range from 590 to 740 cm −1 . The peak around 543 cm −1 represents the vibration of the bonds between Fe 2+ and O 2− in the isolated FeO 6 octahedra 19 . The presence of O-C=O (2361 cm −1 ) asymmetric stretching and C=O vibrations indicates the nature of the carbon bonded to the surface of the as-obtained materials.

The typical scanning electron microscope (SEM) and scanning transmission electron microscope (STEM) images in Fig. 1 and Supplementary Fig. 6 visually confirm the nanosized and microporous morphologies of the as-prepared NFPP-Eterm and NFPP-Cterm samples. The NFPP-Eterm particles feature a nanoplate-like morphology with an average particle size around 150 nm, as shown in Fig. 1d and e, while the NFPP-Cterm particles crystallized in an average particle size of 2 μm with abundant microporous architecture on the surface, as displayed in Supplementary Fig. 6a–c . We employed Brunauer–Emmett–Teller (BET) testing to detect the specific surface areas, and values of 3.52 and 9.74 m 2 g −1 were obtained for the NFPP-Eterm and NFPP-Cterm samples (Supplementary Fig. 7a ), respectively. The pore size distributions are displayed in Supplementary Fig. 7b , and it was found that the average pore size of the NFPP-Cterm particles is slightly larger than that of the NFPP-Eterm, which in good accordance with their porous nature, as observed in the SEM images. Both the samples displayed good single crystallinity with the indexed space group Pn2 1 a , and the coated carbon layers on the NFPP-Eterm and NFPP-Cterm particles (Fig. 1f and Supplementary Fig. 6d ) were measured to be 4 nm and 3 nm in thickness, respectively. Typical high-angle angular dark-field (HAADF) images exhibit an atomic-level crystal structureterm·dom viewed along specific crystallographic directions, where the positions of Fe and P heavy atomic columns can be clearly identified (Fig. 1g and Supplementary Fig. 6e ). Since the signal intensity in a HAADF image is proportional to the atomic number ( Z ), the O atoms are invisible. The relative positions and contrast of Fe, P, and Na can be better acquired with the help of linear profiles. Framework structures with void space along the [010] and [111] directions were identified for the NFPP-Eterm and NFPP-Cterm samples, respectively, and the corresponding schematic illustrations are also presented as insets. Selected area electron diffraction (SAED) patterns of NFPP-Eterm and NFPP-Cterm samples also confirmed the well-crystallized structure of both samples (Fig. 1h and Supplementary Fig. 6f ). Atomic force microscopy (AFM) was used to measure the heights of the nanoplates, and an average height of 18 nm for the NFPP-Eterm nanoparticles was observed (Fig. 1i ). Energy dispersive spectroscopy (EDS) mapping results were also acquired via STEM-EDS, and the results are displayed in Fig. 1j and Supplementary Fig. 6g . The mapping results indicate that the elements Na, Fe, P, O, and C coexist and are distributed uniformly in both the NFPP-Eterm and the NFPP-Cterm particles, which is in good agreement with the refined powder diffraction results.

Electrochemical properties investigations of Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) materials The electrochemical properties of both the NFPP-Eterm and the NFPP-Cterm samples were tested in coin cells with sodium metal anodes. The loading density of active material for each prepared electrode was measured to be about 2.0 mg cm −2 to avoid weighting deviations. The electrolyte consisted of ethylene carbonate/propylene carbonate (EC/PC, 1:1 by volume) with 1 M NaClO 4 as the sodium salt and 5 vol. % fluoroethylene carbonate (FEC) as additive 35 . The NFPP-Eterm electrode showed excellent electrochemical performance at various current densities (Fig. 2a ). It delivered discharge capacities of 113.0 mAh g −1 and 108.3 mAh g −1 at 0.05 C and 0.1 C (1 C = 120 mA g −1 ), and even up to 20 C, there was still 80.3 mAh g −1 remaining, which is completely comparable with the known sodium superionic conductorterm·dom (NASICONterm·dom)-structured polyanionic materials, such as the well-studied Na 3 V 2 (PO 4 ) 3 material 13 . We also compared the C-rate performances of recently published iron-basedterm·key polyanionic materials, and it can be found that our as-obtained material shows the best C-rate performance among them up to 20 C (Supplementary Fig. 8 ) 19 , 29 , 34 , 36 – 46 . At small current densities, there was no obvious discrepancy between the NFPP-Eterm and NFPP-Cterm electrodes, indicating the high ionic conductivity of the Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) particles. The inset of Fig. 2a presents the cycling stabilityterm·key of both electrodes at 0.05 C for 50 cycles, and there was almost 100 % retention for both samples, reflecting the robust framework of this mixed-polyanion system. Also, the initial Coulombic efficiencyterm·dom (ICEterm·dom) was nearly 100% for both samples, which will greatly facilitate their real application in full SIBsterm. The high rate performance of NFPP-Eterm was better than that of NFPP-Cterm, which can be explained by the nanosized particles, which shortened the sodium diffusion distances. Excellent reversibility was obtained with no obvious discrepancy between the first and second cycles except for small changes in peak position due to solid electrolyte interphase (SEIterm·dom) layer formation (Fig. 2b ), which indicates that topotactic single phase variation takes place during cycling without iron atom migration to their favoured face-sharing tetrahedral sites, which may result in a non-equilibrium phase transition (usually seen in alluaudite or pyrophosphateterm frameworks). The same situation between the first two cycles of NFPP-Cterm electrode was observed and is shown in Supplementary Fig. 9a . Also, the reduced charge transfer resistance after the first cycle in the electrochemical impedance spectroscopy (EIS) spectra indicates the formation of a SEIterm·dom layer on the particle surface (Supplementary Fig. 9b, c ). Much alleviated electrochemical polarization compared with that of NFPP-Cterm can be observed the NFPP-Eterm electrode in Fig. 2c . In addition, the as-prepared NFPP-Eterm electrode showed outstanding cycling stabilityterm·key at various C-rates. Figure 2d displays the cycling performance at 0.2 C and 0.5 C for 250 cycles and 430 cycles, respectively, and capacity retention of 85.0% and 84.0% could be obtained. It is worth noting that there was almost no mid-voltage decay (retention of 98.9%) within the first 430 cycles (Supplementary Fig. 10a ), and this value is very important in terms of the total energy density for practical use. We also tested both NFPP-Eterm and NFPP-Cterm electrodes at high rate (20 C, Fig. 2e ), and similar capacity degradation was obtained for both electrodes, indicating that the morphology and particle size are the only reasons for the electrochemical discrepancy. Capacity retention of 69.1% and 57.2% for NFPP-Eterm and NFPP-Cterm, respectively, was achieved after 4400 cycles. Then, we extracted the electrodes from the coin cells to acquire more details. No obvious cracks can be seen on the surfaces of the NFPP-Eterm and NFPP-Cterm electrodes (Supplementary Fig. 10b–e ). From Supplementary Fig. 11 , it can be seen that the primary morphologies were not destroyed even after 4400 high-rate cycles. Good crystallinity still could be founded in the high-resolution TEM (HRTEM) images as well as the SAED patterns of both samples. Supplementary Fig. 12 shows the STEM-EDS mapping results after 4400 cycles. All the existing elements (Na, P, Fe, O, C, and F (from the binder)) were uniformly distributed, and the porous structure of NFPP-Cterm was well maintained. Fig. 2 Electrochemical profiles of both samples. a Rate capability of both samples from 0.1 C to 20 C. The inset is the cycling stabilityterm·key at low current density (0.05 C). b Cyclic voltammetry (CV) curves for the first 5 cycles of NFPP-Eterm electrode (scan rate 0.05 mV s −1 ). c Charge-discharge curves at different rates for both samples. d Cycling stabilityterm·key of NFPP-Eterm electrodes over 250 cycles at 0.2 C and 430 cycles at 0.5 C. e Long-term cycling stabilityterm·key (4400 cycles) at high rate (20 C) for both NFPP-Eterm and NFPP-Cterm electrodes. f Galvanostatic intermittent titration technique (GITTterm·mth) curves of NFPP-Eterm material for both charge and discharge processes. The inset is the chemical diffusion coefficient of Na + ions as a function of voltage calculated from the GITTterm·mth profile (after 30 cycles, current density: 0.05 C). g The calculated capacitance contribution (shadowed area) to the CV curve of NFPP-Eterm at the scan rate of 0.3 mV s −1 Galvanostatic intermittent titration technique GITTterm·mth testing was carried out in a coin cell after it was given 30 cycles to reach its thermal equilibrium state (Fig. 2f ). It can be seen that a single phase (solid-solution) reaction mechanism appeared in both the charge and the discharge processes. The theoretical capacity (128.9 mAh g −1 ) is based on a triple electron transfer process, so the sodium diffusion coefficients fluctuated during the continuous sodium de-/insertion from/into the framework (inset image in Fig. 2f ). The sodium diffusion coefficients were in the range from 10 −13 to 10 −10 cm 2 s −1 within the voltage window of 2.7–4.1 V, a performance which is highly comparable to those recognized NASICON-type cathodeterm·key materials with similar orders of magnitude. We summarize the details of the calculation in Supplementary Fig. 13 and Supplementary Note 1 . In addition, the capacitance was determined from various scan rates (Supplementary Fig. 14 and Supplementary Note 2 , from 0.05 to 0.4 mV s −1 ). It is believed that non-faradic and faradic processes always coexist in the charge storage mechanism. The faradic process can provide a fixed working potential, while the non-faradic processterm·dom (commonly regarded as pseudocapacitanceterm·dom) can help by providing fast charge transitions with extended cycling life. The b value was determined to be around 0.72 with four redox peaks identified, and the non-faradic processterm·dom was calculated to be responsible for 23.4% of the total current at 0.3 mV s −1 (Fig. 2g ). This can be regarded as one of the key factors behind the outstanding electrochemical properties of NFPP-Eterm electrode.

Air stabilityterm·key, all-climateterm·key, and full cellterm·key performances We further tested the air stabilityterm·key and all-climateterm·key performance of NFPP-Eterm material to provide more practical details relevant to its real applications. Figure 3a shows an XRDterm·mth comparison between the fresh powder and the powder exposed to air for three months. No obvious peak shifts or variations can be seen. In addition, from the XPS results in Fig. 3b , there are no detectable discrepancies or peak shifts that can be observed. We collected EIS spectra and found that only a negligible difference appeared in the charge transfer resistance, which might be ascribed to the individual cell assembly processes (Fig. 3c ). The morphologies remained unchanged (Supplementary Fig. 15 ), and the particles surfaces were still well crystallized even after being in contact with air for three months (Fig. 3d ). From Fig. 3e and f, it is clear that almost the same electrochemical performance can be achieved for the NFPP-Eterm electrodes under all the various conditions, and the Coulombic efficiency was maintained at around 100% for each cycle. From the characterizations conducted above, it can be deduced that this Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) mixed polyanionic materialterm·dom has an air stable nature, which is quite different from the case of the pyrophosphatesterm, which are usually unstable in air and face electrochemical degeneration 39 , 47 . Thus, this material is also very promising for aqueous SIBterm systems 48 . We assume that both the carbon coating layer and the robust phosphate anionic groups contribute to the air stabilityterm·key of this Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material. In addition, testing of its all-climateterm·key performance was carried out at both −20 °C and 50 °C. It can be seen that at low C-rates at −20 °C, the specific capacity of NFPP-Eterm could reach 95.0 mAh g −1 and 84.7 mAh g −1 at 0.1 C and 0.2 C, respectively, but a fast capacity drop was encountered with increased current density. There was almost no difference between room temperature and 50 °C for the NFPP-Eterm electrodes, however, indicating that their superior electrochemical performance can be well maintained in hot climate regions. The charge/discharge curves for temperature comparison are displayed in Supplementary Fig. 16 . The cycling stabilityterm·key of NFPP-Eterm electrode at −20 °C and 50 °C remained outstanding, with 92.1% and 91.4% capacity retention at 0.5 C, respectively, indicating that neither low nor high temperaturesterm·dom had any further influence on the crystal structureterm·dom apart from the kinetics factors. In addition, we also fabricated polypyrrole (PPy)-coated Fe 3 O 4 nanospheres to make all-iron-based low-cost SIBterm full cellsterm·key. The preparation details for the anodes can be found in the experimental section. Supplementary Fig. 17 shows the XRDterm·mth pattern and morphology of the as-obtained PPy-coated Fe 3 O 4 nanospheres. Figure 3g displays the electrochemical behaviour of the as-obtained PPy-coated Fe 3 O 4 nanospheres, and capacity of around 250 mAh g −1 and 210 mAh g −1 was achieved and stabilized at 100 mA g −1 and 150 mA g −1 , respectively. As shown in Fig. 3h , an all-iron-based low-cost SIBterm full cellterm·key was activated and operated at 100 mA g −1 in the voltage window of 0.1–4.0 V. It should be pointed out that, due to the low ICEterm·dom of Fe 3 O 4 , the anodes in the full cellsterm·key were pre-cycled. The first cycle reversible efficiency reached as high as 93.1 %, and no obvious curve discrepancies can be observed, indicating the excellent reversibility of the as-fabricated full cellterm·key. All the specific capacities of the full cellterm·key are based on the anodes. In Fig. 3i , the full cellterm·key achieves capacity retention of 76.9% at 100 mA g −1 over 500 cycles and high Coulombic efficiency near 100%. In addition, with more practical concerns, we also fabricated the full cellsterm·key with commercial hard carbon as anode (purchased from KURARAY Co., Ltd., Japan (Type 2)). SEM images and the electrochemical performance at 100 mA g −1 are displayed in Supplementary Fig. 18a–c . We then fabricated a Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 )//Hard carbon full cellterm·key with the loading mass ratio of 1.8:1 for capacity balance according to the individual specific capacities of the electrodes. The anode electrodes were presodiated to reduce the dramatic initial irreversible capacity loss. The results are shown in Supplementary Fig. 18d, e . It can be seen that the charge/discharge curves of the Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 )//Hard carbon full cellterm·key are not as sloping as for those using PPy-coated Fe 3 O 4 , so the energy density can be improved with an elevated mid-working voltage platform. Nevertheless, the capacity is continuously dropping within the initial 10 cycles. We cycled the fabricated full cellterm·key for up to 120 cycles at 100 mA g −1 and found that the capacity retention is less than 50 %. The charge capacity was always higher than discharge capacity for every cycle, which might be the main reason for the continuous capacity drop. The Coulombic efficiency of commercial hard carbon (97% each cycle) is likely to be another reason for this. Nevertheless, we are still putting much effort into working towards better full cellterm·key performance using commercial hard carbon as anode with an optimised electrolyte system, more suitable loading ratio, etc. to be used in our future work. Fig. 3 Air stabilityterm·key, low/high-temperature performance, and full cellterm·key performance. a XRDterm·mth comparison of NFPP-Eterm powder in the fresh state and after exposure to air for three months (Cu Kα radiation, λ = 1.5406 Å). b Fe 2p, C 1s, and P 2p XPS fitted spectra of NFPP-Eterm powder in the fresh state and after exposure to air for three months. c EIS spectra of NFPP-Eterm powder in the fresh state and after exposure to air for three months. d HRTEM image of the powder after exposure to air for three months. The inset is the fast Fourier transform (FFT) pattern of the marked area. e C-rate capability of NFPP-Eterm electrodes: fresh, exposed to air for three months, and fresh at both −20 °C and 50 °C. f Cycling performances of the four electrodes in e . g Charge/discharge curves of polypyrrole (PPy)-coated Fe 3 O 4 at various current densities. The insets are (top) SEM image of as-prepared Fe 3 O 4 nanospheres and (bottom) the cycling performance. h Voltage profiles of the all-iron-based Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 )//Fe 3 O 4 full cellterm·key for the first five cycles. i Cycling stabilityterm·key of the Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 )//Fe 3 O 4 full cellterm·key. The specific capacities of the full cellterm·key were all based on the anodes. Scale bars: 10 nm ( d ); 200 nm ( g , inset) Sodium-storage mechanism In order to obtain a more comprehensive understanding of the structural superiority of Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material, both in-situ synchrotron-based X-ray diffractionterm·mth patterns and X-ray absorption spectra (XAS) were collected on the DESY beamline, Germany. The wavelength was changed to 1.5406 Å (Cu Kα) for better comparison. A two-dimensional (2D) XRDterm·mth pattern and the original in-situ pattern without wavelength change are displayed in Supplementary Fig. 19 . Several peaks in the initial states were indexed and identified, as shown in Fig. 4a, b . It can be seen that all the indexed peaks changed reversibly during the charge/discharge process with the general patterns remaining unchanged, indicating that the robust crystal framework can be well maintained after electrochemical activation. Major peaks such as (200), (011), (210), and (222) gradually moved to higher 2 θ values during sodium insertion and returned to their original values during sodium de-insertion, which can be attributed to the continuous lattice volume variations during cycling. No asymmetric variations were observed, indicating that there were no crystal distortions or cation migration during sodium de-/insertion. Therefore, it can be deduced that a topotactic one-phase transition occurred in the Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) electrode during cycling. We also obtained HRTEM images of the fully charged and fully discharged NFPP-Eterm electrode (Fig. 4c, d ). Lattice fringes for the (200) and (202) planes are clearly observed and identified, indicating the high structural crystallinity of the NFPP-Eterm material during all of the electrochemical reactions. The remaining Na + ions in the crystal structureterm·dom (around 25 %) can be regarded as belonging to the binding pillars. In Fig. 4e, f , the cell parameters were calculated from selected peaks during the first cycle with reversible lattice breathing. The corresponding volume changeterm·dom was calculated to be only 4.0% during the first charge process. Such a small volume changeterm·dom in the crystal structureterm·dom guarantees long-term cycling stabilityterm·key, as demonstrated. Furthermore, the valence state of Fe in Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) during electrochemical reactions was evaluated using in-situ X-ray absorption near-edge structureterm·mth (XANESterm·mth) analysis, as shown in Fig. 4g–i . We employed FePO 4 and LiFePO 4 as the references for the valences of Fe 2+ and Fe 3+ , respectively. The 2D contour plot (Fig. 4g ) of the normalized XANESterm·mth spectra clearly shows the reversible variation in the valence of Fe in the X-ray energy range around 7120 eV. Figure 4h also shows that the XANESterm·mth spectrum is shifted to the right (higher energy area) during charging, indicating oxidation of iron from Fe 2+ to Fe 3+ . The spectrum of the fully charged state was very close to that of the FePO 4 reference sample, indicating that most of the iron in the crystal structureterm·dom had become Fe 3+ . The pre-edge of Fe K-edge during charging also showed discernible variation (inset in Fig. 4g ), with the peak shifted towards higher energy. Then, the XANESterm·mth spectrum moved back to the lower energy area during the discharge process (Fig. 4i ), and the corresponding pre-edge spectra shifted accordingly (inset in Fig. 4i ). It should be pointed out that the spectra of both the initial state and the discharge state were almost the same, but with obvious discrepancy compared to the LiFePO 4 reference sample. This can be ascribed to the individual fingerprint information on the P 2 O 7 and PO 4 groups. Also, the possibility of oxidation of the remaining Fe 2+ in the crystal structureterm·dom during cycling still cannot be neglected. On the basis of the in-situ XRDterm·mth and in-situ XANESterm·mth analyses, we can confirm that both the robust reversible crystal framework and the continuous valence change of Fe contributed to the excellent electrochemical performance shown above. Fig. 4 Sodium-storage mechanism investigations. a In-situ synchrotron-based XRDterm·mth patterns and b corresponding charge-discharge curves of NFPP-Eterm electrode during the first cycle. c , d HRTEM images of fully charged and fully discharged NFPP-Eterm electrodes. e Variations of a , b , and c lattice parameters during the charge/discharge process. f Volume changeterm·dom details during the charge/discharge process. g In-situ XANESterm·mth spectra at the Fe K-edge of NFPP-Eterm electrode (2D contour plot) as a function of charge-discharge curve. h Charge process and i discharge process of typical Fe K-edge XANESterm·mth spectra. The insets in h and i are the corresponding pre-edge spectra. Scale bars: 5 nm ( c ); 5 nm ( d ) Na-ion dynamics In order to deeply understand the intrinsic properties of the crystal structureterm·dom that are responsible for the outstanding C-rate performance, both BVSterm·mth and DFTterm·mth calculations were employed to investigate the localized migration energy barriersterm·key. The BVSterm·mth calculation is a well-established empirical toolterm·mth for preliminary examination of ionic states and diffusion pathwaysterm·key. Both the bond-valence map and the bond-valence electron voltage map are shown in Supplementary Fig. 20 . The isosurface value was set as 1 where the Na + ions can possibly be found (Supplementary Fig. 20a, b ), and the isosurface near −3.5 eV shows the lowest initial energy regions, indicating possible Na + ion diffusion pathwaysterm·key (Supplementary Fig. 20c, d ). The corresponding 2D slice images reveal details of the lowest energy regions in various directions (Supplementary Fig. 21 ). We found that along the a , b , and c axes, the lowest energy regions are well connected, which means that this material possibly possesses more than one-dimensional Na + ion diffusion pathwaysterm·key. Therefore, we conducted a DFTterm·mth study to further determine the details. We discovered that all the Na + ions in a single unit can be assigned to three different types, based on their individual binding energies. The detailed information, as well as calculation method, is summarised in Supplementary Table 3 . In Fig. 5 , various images of the crystal structureterm·dom of Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material with three different types of Na + ions are presented. It was calculated that the diffusion energy barriersterm·key within the same Na + ion type were 0.553 eV, 0.02 eV, and 0.365 eV for A to A type, B to B type, and C to C type, respectively, which are all very low energy barriersterm·key for the transfer of Na + ions (Fig. 5b ). In particular, it can be concluded that the most favourable diffusion tunnel in the first stage is along the a direction, since almost no energy barrierterm·key (0.02 eV) was detected. Then, we performed the calculations between different Na + ion types, which are equivalent to three-dimensional diffusion pathwaysterm·key, since a , b , and c directions are all involved. From Fig. 5e , it can be seen that all the energy barriersterm·key for Na + ion diffusion are lower than 0.9 eV, which are all possible diffusion pathwaysterm·key in the presented crystal structureterm·dom, providing solid evidence for this newly recognized NASICON-type structure with 3D diffusion pathwaysterm·key or tunnels in this Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material. The NASICONterm·dom framework is constructed from corner-sharing MO 6 F 6− x (M=Ti, V, etc.) octahedra and XO 4 (X = P, S, Si, etc.) tetrahedra. The XO 4 tetrahedra maintain the charge balance and interact with the bipolar MO 6 F 6− x octahedra, while the MO 6 F 6− x octahedra can offer the main electrostatic repulsion that accounts for the various voltage platforms for Na + intercalation/extraction. The critical factor for a NASICON-type structure is the arrangement of sodium sites and whether the 3D diffusion pathwaysterm·key of Na + can be accessed with relatively low energy barriersterm·key. We also performed ionic conductivity testing, and it was found that the total conductivity in various temperatures of the Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material was of the same order of magnitude compared with the well-recognised NASICON-type Na 3 V 2 (PO 4 ) 3 electrode (Supplementary Fig. 22 ) 49 . Both the BVSterm·mth and the DFTterm·mth calculations revealed the intrinsic reasons for the outstanding high rate performance of this material, which are completely comparable to those of the known NASICON-type materials mentioned above. Considering the long-term cycling stabilityterm·key, high-rate capability, air stabilityterm·key, and all-climateterm·key performance of this low-cost mixed polyanionic materialterm·dom, we believe that this new NACISON-type Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material is a strong competitor among the various sodium hosts competing for real applications in large-scale ESSs 50 – 54 . Fig. 5 Crystal structureterm·dom, Na + ion diffusion paths, different types of Na + ions in a single unit, and corresponding migration energy barriersterm·key. a , c , d Various images of the crystal structureterm·dom of Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material with three different types of Na + ions. b The migration energy barriersterm·key within the same Na + ion groups. e The migration energy barriersterm·key between different Na + ion groups (equivalent to three-dimensional diffusion pathwaysterm·key) Discussion We have successfully synthesized tuneable nanosized Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) plates and microporous Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) particles via a facile one-step sol-gel method with high phase purity and uniform carbon coating. As cathodeterm·key, excellent rate performances of 113.0 and 80.3 mAh g −1 at 0.05 C and 20 C, respectively, were achieved, and impressive cycling stabilityterm·key was obtained without noticeable voltage decay (69.1% capacity retention after 4400 cycles at 20 C) with almost 100% ICEterm·dom achieved. The air stabilityterm·key and all-climateterm·key (−20 °C and 50 °C) performance were investigated, and this material showed outstanding air stabilityterm·key and all-climateterm·key electrochemical properties (84.7 mAh g −1 at 0.2 C) at −20 °C. A low-cost all-iron-based full cellterm·key was fabricated and characterized with a view towards real applications. Both the in-situ XRDterm·mth and the in-situ XANESterm·mth results revealed its excellent crystal reversibility and structural stability during cycling. A GITTterm·mth study showed high sodium diffusion coefficients, and its NASICON-type structure and 3D diffusion pathwaysterm·key were further explored and verified by both BVSterm·mth and DFTterm·mth calculations in detail, revealing impressive low activation energy barriersterm·key. Our studies suggest that this low-cost mixed polyanionic Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) material is a strong competitor among the various sodium hosts and should receive more comprehensive investigation towards its real application in large-scale energy storage systemsterm·dom in the near future.

Methods Synthesis of Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) particles Both the nanosized Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) plates (NFPP-Eterm) and the microporous Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) particles (NFPP-Cterm) were prepared via typical sol-gel methods. All the chemicals used in this paper were analytically pure, and purchased from SIGMA-ALDRICH without further purifications. In a typical process, 4 mmol sodium acetate, 4 mmol ammonium phosphate, 0.1 g glucose, and 0.1 g stearic acid were added to 20 ml deionized water with magnetic stirring until a transparent solution was obtained (denoted as solution A). 3 mmol iron (II) acetate, 0.8768 g ethylenediaminetetraacetic acid, and 0.05 g cetyltrimethylammonium bromide were added to 20 ml deionized water with magnetic stirring until a transparent solution was obtained (denoted as solution B). Then solution B was added dropwise to solution A with vigorous stirring. The mixed solution was then heated in a water bath at 90 °C until all the excess water was removed and the sol-gel precursor was obtained. The precursor was ground into a fine powder and then annealed at 500 °C for 24 h under high purity Ar gas atmosphere with an intermediate grinding. The final nanosized Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) plates were denoted as NFPP-Eterm in this paper. For the microporous Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) particles, all the preparation procedures were the same, except that 0.6305 g citric acid monohydrate replaced the 0.8768 g ethylenediaminetetraacetic acid in solution B. The final product was denoted as NFPP-Cterm in this paper. Both the final NFPP-Eterm and NFPP-Cterm samples were transferred into an Ar-filled glove box after annealing until further use.

Synthesis of polypyrrole (PPy)-coated Fe 3 O 4 nanospheres The Fe 3 O 4 nanospheres were prepared via a solvothermal method using trisodium citrate as stabilizer and FeCl 3 as raw material in ethylene glycol solution. Specifically, 4 mmol FeCl 3 ·6H 2 O and 1.3 g trisodium citrate were dissolved in 30 ml ethylene glycol by stirring for 2 h; then, 0.28 sodium acetate was added into the solution. After 1 h of ultrasonication and 1 h of magnetic stirring, the yellow solution was transferred to and sealed in a Teflon-lined autoclave. The autoclave was maintained at 180 °C for 12 h and then cooled down to room temperature naturally. The Fe 3 O 4 nanoparticles were obtained by centrifugation for several times with water and ethanol. Then, 30 mg of the as-obtained Fe 3 O 4 nanospheres was added to 30 mL deionized water with paddle stirring in a three-neck round-bottom flask. 600 μL pyrrole was added to the solution, followed by 2 h of ultrasonication and 12 h of stirring. The final PPy-coated Fe 3 O 4 nanospheres were obtained via freeze-drying under vacuum for 24 h.

Materials characterization X-ray diffractionterm·mth (XRDterm·mth) was employed to characterize the crystalline structure of the obtained samples in the 2 θ range of 15°–45° (GBC MMA diffractometer, Cu Kα radiation, λ = 1.5406 Å, 1 ° min −1 , step size of 0.02 ° s −1 ). Synchrotron powder diffraction data were collected at the Australian Synchrotron beamline at the wavelength ( λ ) of 0.7748 Å, calibrated with the standard reference material (National Institute of Standards and Technology (NIST) LaB 6 660b), and analysed via GSAS-II software 55 . Schematic representations of the synchrotron XRDterm·mth data were obtained by VESTA software 56 . The morphologies of both samples were examined through a field emission scanning electron microscope (FESEM, JEOL JSM-7500FA). The detailed structural information on the crystals was determined by both high-resolution transmission electron microscopy (HRTEM, JEOL 2010, 200 kV) and scanning transmission electron microscopy (STEM, JEOL JEM-ARM200F, 200 kV) with EDS mapping and a selected area electron diffraction (SAED) module. The angular range of collected electrons for the HAADF images was around 70–250 mrad. The EDS mapping results were obtained via STEM using NSS software. Thermogravimetric (TG) analysis was performed on a Mettler Toledo TGA/DSC1 with a heating rate of 10 °C min −1 in air. The surface information was collected via X-ray photoelectron spectroscopy (XPS) with Al Kα radiation. Raman spectra (JY HR800 spectrometer, 10 mW helium/neon laser at 632.8 nm) were collected to characterize the carbon of both samples. The vibration states of existing functional groups were examined via Fourier transform infrared spectroscopy (FT-IR, spectral resolution of 4 cm −1 on a Thermo Nicolet Nexus 670 FT-IR spectrometer). The magnetic measurements were performed via a Quantum Design physical properties measurement system (PPMS) combined with a vibrating sample magnetometer (VSM) option at various temperature (5–300 K) from –20000 Oe to 20000 Oe. The Brunauer–Emmett–Teller (BET) testing was carried out using a Micromeritics Tristar II 3020 surface area analyzer. The valence states of Fe in the Na 4 Fe 3 (PO 4 ) 2 (P 2 O 7 ) were characterized via XANESterm·mth analysis. The in-situ XRDterm·mth was conducted at the P02.1 Powder Diffraction beamline and XANESterm·mth spectra were recorded at the P64 XAS beamline of the DESY synchrotron. Atomic force microscopy (AFM) on the nanoplate particles was performed on an MPF-3D, Asylum Research, Santa Barbara, USA.

Electrochemical testing The NFPP-Eterm and NFPP-Cterm electrodes were prepared by mixing 80 wt. % active material, 10 wt. % conductive carbon black, and 10 wt. % polyvinylidene fluoride (PVDF, binder) to form the electrode slurry with a proper amount of N-methyl-2-pyrrolidone (NMP) as solvent. The slurry was uniformly coated on aluminium foil, followed by vacuum drying at 120 °C overnight. Both the Fe 3 O 4 electrode and purchased hard carbon (KURARAY Co., Ltd., Japan (Type 2)) was prepared by mixing 80 wt. % active material, 10 wt. % conductive carbon black, and 10 wt. % carboxymethyl cellulose (CMC, binder) to form the electrode slurry with a proper amount of deionized water as solvent. The slurry was uniformly coated on copper foil, followed by vacuum drying at 80 °C overnight. The dried slurry on Al/Cu foil was pressed under 10 MPa and punched into small disc electrodes. The loading density was about 2.0 mg cm −2 for the positive electrodes. The loading density of the negative electrodes was about 1.0–1.2 mg cm −2 . The weight ratio of the two electrodes was balanced with reference to the corresponding reversible capacities, and the current density is based on the anode mass. Na metal and Fe 3 O 4 /hard carbon electrodes were used as counter electrodes in the half cell and full-cell measurements, respectively. The electrolyte was 1 M NaClO 4 in ethylene carbonate (EC)–propylene carbonate (PC) solution (1:1 by volume) with 5 vol. % addition of fluoroethylene carbonate (FEC). Coin cells (CR2032) were assembled in a glove box filled with ultra-pure Ar, and the concentrations of O 2 and H 2 O were kept under 0.1 ppm. The electrochemical properties were examined by a NEWARE test system. The voltage window was set between 1.9 V–4.1 V, and 1 C = 120 mA g −1 . The high/low-temperature performance was investigated via employing a high/low-temperature test box (Si Yang Precision Equipment (Shanghai) Co. Ltd.). The low-temperature was set at −20 °C, and the high temperatureterm·dom was set at 50 °C. Electrochemical impedance spectroscopy (EIS) was employed in the range of 100 kHz to 10 mHz, with the amplitude of the AC voltage set at 10 mV (Bio-Logic VMP-3 electrochemical workstation). The ionic conductivity ( σ ) was obtained by Eq. ( 1 ) 1 σ = L ∕ R * S Where L is the thickness of the electrodes, R is the resistance tested by the same EIS mentioned above, and S is the area of the electrode film. Cyclic voltammetry tests were carried out at various scanning speeds from 0.05 mV s −1 to 0.4 mV s −1 in the voltage window of 1.9 V–4.1 V. Galvanostatic intermittent titration technique (GITTterm·mth) measurements were conducted on a Land battery testing system after 10 cycles to let the electrolyte/electrode reach its equilibrium state (current density: 6 mA g −1 , 0.05 C).

Bond-valence calculations and DFTterm·mth calculations The charge distribution and possible Na + ion motions were calculated using soft-BV parameters. First-principles calculations were performed using the Vienna ab -initio simulation package (VASP) 57 , which is based on density functional theoryterm·mth and the plane-wave pseudopotential method 58 , with the following settings. The generalized gradient approximation (GGA) 59 with the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional was used with the plane-wave cut-off energy set at 400 eV for all calculations. The criterion of convergence is that the residual forces should be less than 0.02 eV Å −1 and the change in the total energy less than 10 −5 eV. All calculations were performed using periodic boundary conditions, and the Brillouin-zone integrations were performed using a 3 × 3 × 3 Monkhorst-Pack grid.

Supplementary information Supplementary Information Peer Review File