Understanding capacity fading from structural degradation in Prussian blue analogues for wide-temperature sodium-ion cylindrical battery

Materials characterizations

The type and amount of host elements in the PBA crystal structure can be changed by using different precursors during synthesis. So, three PBAs with high sodium content, medium sodium content, and sodium free, were synthesized through the co-precipitation method and are denoted as HSPB, MSPB, and LSPB. Powder X-ray diffraction (XRD) of the as-synthesized PBAs was analysed by Rietveld refinement using GSAS software³⁵. Detailed descriptions of the crystal structure are provided in Supplementary Table 1–3 in the Supporting Information. As shown in Fig. 1a, HSPB exhibits rhombohedral phase (lattice parameters: a = 7.4832 Å, c = 17.5592 Å, V = 851.54 ų) with distinct sharp peaks, indicating good crystallinity and a lower amount of [Fe(CN)₆]^4- defects in the framework. The diffraction peaks of the (220) planes around 24° are split because of the decreased symmetry caused by the increased number of Na⁺ per formula unit²⁵. When there is less or no sodium in the framework, MSPB (a = 10.2676 Å, V = 1082.44 ų) and LSPB (a = 10.22648 Å, V = 1069.49 ų) have the cubic phase (Fm-3m) with no impurities (Fig. 1b, c). The broader peaks of the MSPB compared to the HSPB or LSPB mean smaller crystallite sizes. Scanning electron microscope (SEM) images and elemental mapping images for the three PBA samples are shown in Supplementary Figs. 1–5. As a chelating agent and sodium supplement, sodium citrate played a crucial part in the crystallization process, which was beneficial for slowing the rate of precipitation and raising the amount of sodium in the PBAs^14,17. After adding the chelating agent, single microcubes (3–5 μm) were obtained in the HSPB. It was easier for MSPB to aggregate into larger secondary particles because of its nanosized primary particles (Supplementary Fig. 4). The elemental mapping images and X-ray photoelectron spectroscopy (XPS) spectra of LSPB provided in the Supplementary Fig. 5 prove that there was no potassium and sodium element in the LSPB. As illustrated in Fig. 1d–f, in-situ heating XRD characterization (flowing high purity Ar atmosphere) was performed to explore the thermal durability and how the structure changed during the dehydration of the three PBAs (25–400 °C). The different phase transition processes observed in the initial products are primarily influenced by their sodium content and crystal defects. The XRD data for the three PBAs exhibited no change below 150 °C. When the temperature was raised above 150 °C, the XRD pattern of HSPB (highest sodium content with a rhombohedral phase) changed significantly. From this temperature on, the peak positions of HSPB shifted to higher degrees, and new peaks were generated, which could be indexed as a new trigonal structure without water. A similar change was also found in MSPB (enhanced structural stability by more sodium and fewer defects), but MSPB was stable up to 200 °C, and then the phase transition occurred. If the temperature exceeded 250 °C, the crystal structure of LSPB started to collapse due to thermal decomposition because of the reduced structural stability. This means that PBAs with more sodium content have a lower phase transition temperature when heated but can maintain the crystal structure up to 400 °C. The PBAs without sodium ions in the framework will decompose rapidly after a certain temperature (such as 250 °C for LSPB).

a–c Rietveld refinements of HSPB, MSPB, and LSPB. The golden spheres represent sodium ions, and the purple and green octahedrons represent low-spin and high-spin iron. d–f In-situ heating XRD of HSPB, MSPB, and LSPB. g Fe K-edge XANES spectra. h FTIR spectra. i TGA curves. The displayed percentages within panel are mass loss.

X-ray absorption spectroscopy (XAS) was performed on the Fe K-edge to investigate changes in the local electrical and structural properties in the three PBAs. As shown in the X-ray absorption near-edge structure (XANES) spectrum in Fig. 1g, the Fe K-edge absorption rising edges move slightly towards lower energy when there are more Na ions in the framework, suggesting that the Fe valence state of HSPB is mainly 2+, while that of LSPB is mainly 3+³⁶. The Fourier-transformed intensities of the Fe K-edge extended X-ray absorption fine structure (EXAFS) spectra of the three PBAs are provided in Supplementary Fig. 6. The two peaks at 2094 and 2127 cm⁻¹ of HSPB are the vibrations of Fe²⁺-C≡N-Fe²⁺ and Fe²⁺-C≡N-Fe^(3-x)+, which means that some ferrous ions were oxidized during synthesis (Supplementary Fig. 7). Only a peak at 2151 cm^-1 (Fe³⁺-C≡N-Fe³⁺) was found in LSPB, indicating that only Fe³⁺ ions were linked with the cyano groups³⁷. Fourier transform infrared (FTIR) spectroscopy was also performed, and the corresponding spectra of the PBAs are displayed in Fig. 1h. The strong peaks at 2160 and 2170 cm⁻¹ correspond to the stretching vibrations of the C≡N bonds³⁴. Moreover, the peaks at 1610 and around 3610 cm⁻¹ represent the bending vibration of O-H and the corresponding O-H bond stretching, indicating the presence of H₂O¹⁸. Therefore, thermogravimetric analysis (TGA) was used to determine the water content of the PBAs (Fig. 1i). The water content of the PBAs was categorized into three categories: adsorbed water, interstitial water, and coordinated water. Water adsorbed on the surface was removed first as the samples were heated, then interstitial water located within the crystal framework, and finally, water that was chemically coordinated with the transition metal ions. As a result of the addition of a chelating agent, the adsorbed water in PBAs decreased significantly from 12.5% (LSPB) to 1.5% (HSPB). More water loss of HSPB occurred around 220 °C, implying more interstitial water in the framework. The reason for the increased amount of interstitial water in sodium-rich PBAs may be the slow nucleation rate, which means a long growth period and more water movement into the framework. A higher water content (adsorbed and interstitial) for sodium-free rather than sodium-rich PBA was found, which is consistent with the previous reports^16,18,38. Consistent with the in-situ heating XRD data, LSPB underwent a rapid mass degradation after 275 °C, indicating the decomposition of the framework. HSPB maintained a relatively stable mass retention rate before 400 °C, however, which indicates a more stable skeleton and fewer defects.

Electrochemical investigations

According to the cyclic voltammetry (CV) plots (Fig. 2a, b), two strong and symmetrical redox pairs were observed around 3 V (high-spin Fe) and 3.3 V (low-spin Fe). Low-spin Fe is less redox active in LSPB, which may result from its many crystal defects and the high amounts of adsorbed water in its framework (Fig. 2c). During the discharge process, the low-spin Fe undergoes a reduction reaction from Fe³⁺ to Fe²⁺. Once the low-spin Fe in HSPB has fully contributed its capacity, i.e., at around 60 mAh g⁻¹, the voltage drops almost vertically because of the completion of the reduction reaction as shown in Fig. 2d. High charging capacity (139.6 mAh g⁻¹) was achieved for the HSPB at 50 mA g⁻¹ within the range of 2-4.2 V, benefitting from the high initial Na content (Fig. 2d). MSPB provided a capacity of 86.4 mAh g⁻¹ in the 1^st charging process while only 19.9 mAh g⁻¹ was achieved for LSPB. Since LSPB does not contain potassium ions, the initial charging capacity may be mainly derived from the capacitance contribution of conductive carbon (Supplementary Fig. 8). After discharging, the reversible capacities of HSPB, MSPB, and LSPB were 136.9, 118.0, and 110.8 mAh g⁻¹, respectively. The low reversible capacity of LSPB is mainly due to structural deficiencies, including a high density of Fe(CN)₆ vacancies, which reduce active sites for sodium-ion insertion and extraction, limiting its capacity. A study of the cycling stability of the samples was performed at 50 mA g⁻¹, as shown in Fig. 2e and Supplementary Fig. 9. For capacity retention after 200 cycles, LSPB shows the highest retention level at 90.8%, while HSPB was only able to retain 65.7%. Although high initial reversible capacity was achieved by HSPB, the capacity fade after cycling makes it less promising for practical long-cycling-life SIBs. The reason for capacity decline needs to be further explored after analysis of HSPB and comparison with the results of MSPB and LSPB, which will be discussed in the next section. When the specific current was increased to 2000 mA g⁻¹, the LSPB provided a reversible capacity of  90.9 mAh g⁻¹ (equal to 81.2% of its reversible capacity at 15 mA g⁻¹ for LSPB), significantly better than for the HSPB (Fig. 2f). In addition, the galvanostatic intermittent titration technique (GITT) confirmed the fast reaction kinetics of LSPB during the charging/discharging process. The calculated sodium diffusion coefficient (D_Na⁺) of LSPB was higher than that of HSPB, indicating a fast diffusion rate (Fig. 2g, h and Supplementary Fig. 10). Considering practical applications, high specific energy, prolonged cycle life, and good rate capability are typically required for SIBs. An additional requirement for the cathode is to provide enough sodium ions to shuttle to an anode, such as hard carbon during charging. Therefore, radar plots were used to summarize the quality of the three PBAs exhibiting different electrochemical properties (Fig. 2i). Benefitting from its Na-rich structure, HSPB provided the most impressive initial charge capacity and reversible capacity, but it had poor rate performance and cycling. Despite having more water content and crystal defects, LSPB exhibited better cycling stability than Na-rich HSPB. This result shows that the capacity fading of common PBAs cannot be simply attributed to these two factors, and more attention should be paid to the morphology, active material loss, and lattice distortion after cycling.

a–c CV plots of HSPB, MSPB, and LSPB. d The 1st cycle galvanostatic charge/discharge plots of the three samples at 50 mA g⁻¹. Cycling performance (e) and (f) rate capacity of the three samples. The corresponding calculated D_Na⁺ of the three samples when charging (g) and discharging (h). i Radar chart for the trade-offs between log D_Na⁺, rate performance, cycle stability, initial charging capacity, and reversible capacity among the three materials.

Investigation of the capacity-fading mechanism

In order to gain a deeper understanding of the origins of the capacity fading of the HSPB, in-situ synchrotron XRD, SEM, and XPS were performed on the cycled electrodes, as well as density functional theory calculations (DFT). During the charging/discharging process, all the characteristic peaks are reversible, indicating that the crystal structure of the three PBAs can be maintained for the first two cycles (Fig. 3). As shown in Fig. 3a, d, rhombohedral iron-based PBAs (HSPB) experienced a phase transition process with large lattice distortion. In the first charge process, the (200), (220), and (400) planes are gradually shifted towards higher degrees, indicating some shrinkage of the unit cell. As a result of Na⁺ extraction from the crystal framework, phase changes from rhombohedral to cubic were also observed in the HSPB electrode. Some published literatures regarded the small lattice change at high voltage as indicating a phase transition from the cubic phase to the tetragonal phase because of the little changes of c value^20,25. This lattice change is relatively small (such as a = b = 10.1926 Å, c = 10.2085 Å), and it can also be considered as a Na⁺-free cubic phase^24,25. After the 1st discharging and 2nd discharging, the splitting phenomenon of the (220) peak was still obvious, as shown in Supplementary Fig. 11. Similar lattice changes (moving to higher/lower degree when charging/discharging) were also found in MSPB and LSPB, as shown in Fig. 3e, f. A change between sodium-rich and sodium-free states in the cubic phase was observed during cycling, which provides the entire reversible capacity through the redox reaction between Fe²⁺ and Fe³⁺. Notably, LSPB exhibited an initially sodium-free cubic phase structure due to the absence of active sodium ions in the framework, as discussed above. The discharge operation was performed firstly during the in-situ XRD test, so the peak mainly shifted to the left first, followed by a shift to the right during charging, and it returned to its initial position after charging to 4.2 V. This changing trend and the main peak positions of LSPB are consistent with those of HSPB and MSPB, implying a changing process between sodium-rich and sodium-free states in the cubic phase. In addition, the cell volume change of HSPB at high voltage (where the peaks shifted a little to the left immediately after the peak positions were rightmost) is much larger than that of LSPB, which may also be an important reason for capacity fading. According to the in-situ XRD, the unit volume variations of the three samples were calculated in Fig. 3d–f. Notably, the rhombohedral structure of HSPB was subject to both shrinkage and expansion during the charging and discharging processes, with a large-volume distortion of 5.61%, which was greater than the distortion of MSPB and LSPB. During total charging and discharging processes, the unit volume can return to almost its initial value. The structure will become unstable, however, due to the gradually accumulating lattice strain from the lattice change. This phenomenon indicates that the structural variation of HSPB is greater than that of MSPB and LSPB, and it clarifies the lattice strain within two-phase transitions for rhombohedral HSPB (Fig. 3g).

a–c Synchrotron XRD spectra and galvanostatic charge/discharge plots of HSPB, MSPB, and LSPB. d Enlarged two-dimensional (2D) contour XRD images and volume variations for HSPB. e Enlarged 2D contour XRD images and volume variations for MSPB. f Enlarged 2D contour XRD images and volume variations for LSPB. g Schematic illustration of the phase transition mechanisms. The golden spheres represent sodium ions, and the purple and green octahedrons represent low-spin and high-spin iron.

In-situ XRD for the first two cycles indicated that HSPB showed a large lattice change due to more sodium ion shuttling, which is unfavorable for long-term cycling. In addition, the characterizations of the electrode sheet after long-term cycling can more specifically analyze the changes in the material during the cycling process. To further understand the rapid capacity fade of HSPB after long cycling, SEM observations were conducted on the electrodes of the three PBA cathodes after different cycles. As a model system, the pristine electrode sample and the electrodes after the 100^th and 200^th cycles were used to study the capacity fading mechanism because of the significant capacity decrease (~35%) in HSPB after 200 cycles. This capacity decay may be accompanied by significant structural, chemical, and morphological changes. As shown in the SEM images of the pristine electrodes (Fig. 4a, d), the HSPB preferred a quasi-cubic structure with a size of ~4 μm, while the LSPB featured irregular nanoparticles intertwined with conductive additives. The morphology of HSPB was broken after 100 cycles and almost collapsed after 200 cycles (Fig. 4b, c). This shows that fresh surfaces of HSPB will be exposed during the cycling process, which will sustainably consume the electrolyte and active Na⁺ to form the cathode-electrolyte interphase. This may be a macroscopic manifestation brought about by lattice distortion during long-term cycling. In contrast, LSPB and MSPB maintained their morphology after cycling (Fig. 4e, f and Supplementary Fig. 12). It is also worth noting that many nanoparticles were formed on the surface of the HSPB after 100 cycles. These particles were clearly different from those of the conductive additives and more like particles bubbling from the inside (Supplementary Fig. 13). It can be seen from Supplementary Fig. 14 (secondary electron images on the left and corresponding backscattered electron images on the right) that the blue circled areas are the conductive additive, which are shown as dark areas in Supplementary Fig. 14b, d. There is no difference in brightness between the red-circled areas in the surface of HSPB, indicating that they do not show the conductive additive but surface nanoparticles produced by the material itself after cycling. Therefore, combining the image contrast of the backscattering mode and the SEM image analysis of the conductive additive in Supplementary Fig. 15, it can be speculated that the attached particles on the surface are the material itself. The reaction kinetics of HSPB electrodes after cycling were further studied by electrochemical impedance spectroscopy (EIS). The results of the EIS on HSPB (the pristine cell and the cell after 100/200 cycles) are shown in Fig. 4g, along with the equivalent circuit and impedance values that have been fitted. The medium-to-high-frequency range shows two partially overlapping semicircles due to the electrolyte-electrode interface film resistance (R2) and the charge-transfer resistance (R3). In the low-frequency range, the sloping line represents sodium diffusion (Warburg impedance, W1) and the bulk resistance of the cell (R1). The fitting value of R2 in the HSPB electrode changed from 172 Ω to 918 Ω after 100 cycles, which shows the increase in the resistance brought about by the electrolyte-electrode interphase film. On comparing the impedance after 200 cycles and 100 cycles, R3 changed greatly (from 1131 Ω to 3119 Ω) during this period, indicating that the charge-transfer resistance increased (Supplementary Table 4). Furthermore, the phase transition of HSPB was investigated again after multiple cycles. It can be seen from Fig. 4h that, after 100 cycles (discharged to 2 V), the splitting phenomenon of the (220) plane disappeared on the XRD pattern of HSPB, although it returns at the end of discharge state. The peak for the (200) planes of HSPB also shifted significantly after 100 cycles, indicating the reduction of Na ions in the structure. The in-situ synchrotron XRD of the HSPB after multiple cycles also confirmed that the double peak belonging to the (220) planes changed into a broad peak (Fig. 4i).

Variations of morphology on the electrode: SEM images of pristine HSPB electrode (a), HSPB electrode after 100 cycles (b), and HSPB electrode after 200 cycles (c); SEM images of pristine LSPB electrode (d), LSPB electrode after 100 cycles (e), and LSPB electrode after 200 cycles (f). g Nyquist plots and the equivalent circuit modeling (inset, R1: the bulk resistance of the cell; R2: the electrolyte-electrode interface film resistance; R3: the charge-transfer resistance) of HSPB electrodes. h Laboratory-based XRD patterns of pristine electrodes and electrodes after 100 cycles. i Line spectra of in-situ synchrotron XRD and 2D contour image of HSPB after 50 cycles.

To understand the detailed reasons for the capacity decrease of HSPB, XPS was used to determine the chemical species formed on three PBA cathodes after different numbers of cycles (Fig. 5 and Supplementary Figs. 16–20). All the cycled electrodes were prepared and sealed in a glove box, followed by evacuating their containers and transferring them to the XPS instrument to avoid the influence on the electrodes of water and oxygen in the air. As shown in Fig. 5a, the Fe 2p spectra of pristine HSPB electrode exhibit two pairs of peaks located at 721.6/708.7 eV (Fe²⁺) and 723.7/710.2 eV (Fe³⁺)^39,40,41. The Fe³⁺/Fe²⁺ ratio was increased after 100 cycles (reduced state), however, as shown in Fig. 5b. This means that some redox-active ferrous ions in the HSPB framework became inert Fe³⁺ sites and could not participate in subsequent cycles, causing a gradual decline in capacity. In addition, the peak at 530.6 eV in O 1s was assigned to Fe-O, which was generated by the dissolution of transition metal ions in the cell (Fig. 5c, d)^42,43. The formation of Fe-O would result in a deterioration of the electrode and the irreversible loss of redox-active parts, which would further impair reversible capacity. Less Fe-O was generated in the MSPB/LSPB compared to the HSPB (Supplementary Fig. 17-18), indicating that inhibiting the dissolution of transition metal ions can improve cycling stability. Additionally, XPS depth analysis was performed to study the inactivation of redox centers. The intensity of the peak ascribed to Fe²⁺ was almost unchanged in pristine HSPB when the sputtering depth was increased. This peak was obviously and rapidly increased in cycled HSPB with increasing sputtering depth, however, indicating that the redox-active sites on the surface were deactivated while the interior still maintained a high redox activity after multiple cycles (Fig. 5e, f). In addition, the cycled sodium metal and the separator (disassembled from the cell with HSPB as the cathode) were also tested, and the Fe 2p signals were found, further confirming the dissolution of transition metal in HSPB (Fig. 5g, h). The dissolution of iron ions is not caused by the formation of sodium dendrites or their contact with the cathode. Additionally, no voltage fluctuations indicative of short circuits were observed in the charging/discharging curves at different cycles. Instead, the dissolution is primarily attributed to structural and morphological changes in the cathode material during cycling.

Fe 2p spectra of pristine HSPB electrode (a) and HSPB electrode after 100 cycles (b). O 1s spectra of pristine HSPB electrode (c) and HSPB electrode after 100 cycles (d). Fe 2p depth profiles of pristine HSPB electrode (e) and HSPB electrode after 100 cycles (f) from different sputtering depths. XPS survey spectra of pristine sodium metal, sodium metal after 50 and 100 cycles (g) (with photographs shown in the insets) that were removed from the cycled cell using HSPB as cathode. XPS survey spectra of pristine separator, separator after 50 and 100 cycles (h) (with photographs shown in the insets) that were removed from the cycled cell using HSPB as cathode.

To elucidate the time scale and mechanisms of Fe²⁺ deactivation, galvanostatic intermittent titration technique (GITT) experiments were conducted at various cycle numbers (Supplementary Fig. 21). At the high discharge plateau (C-coordinated low-spin Fe), D_Na⁺ remained stable across all cycles, indicating high low-spin Fe stability. Conversely, at the low discharge plateau (N-coordinated high-spin Fe), D_Na⁺ gradually declined with increasing cycles, suggesting progressive high-spin Fe²⁺ into Fe³⁺. GITT data confirm this change is gradual rather than sudden. To further analyze Fe²⁺ deactivation, dQ/dV analysis (Supplementary Fig. 22a, b) was performed. The redox peaks of low-spin Fe remained prominent after 200 cycles, with minimal peak intensity or position changes, corroborating its stability. In contrast, the high-spin Fe redox peaks showed significant attenuation, accompanied by a discharge voltage decrease at the low plateau, indicating increased resistance and slower ion kinetics. By separating the capacity contributions of low-spin and high-spin Fe, low-spin Fe retained 88% of its initial capacity, while high-spin Fe capacity declined to 49% after 200 cycles, with deactivation accelerating notably after 40 cycles (Supplementary Fig. 22c–f). To quantify the correlation between lattice distortion, volume changes, and capacity retention, XRD measurements were conducted at different cycling stages (Supplementary Fig. 23). The results reveal that lattice parameters undergo progressive changes, with notable volume expansion and crystal distortion. In the early cycles (20 and 40 cycles), significant structural alterations were observed, including increases in the a-axis and unit cell volume by 0.3% and a decrease in the c-axis by 0.3%. These changes correlate with rapid capacity loss, indicating that increased lattice distortion impairs the material’s ability to reversibly accommodate sodium ions. As cycling continued, the lattice distortion worsened, particularly after 80 cycles, with the a-axis and unit cell volume increasing by 0.5% and 0.6%, and the c-axis decreasing by 0.4%. After 100 cycles, the material transitioned from a rhombohedral to a cubic phase. SEM imaging and analysis are also conducted to investigate the propagation of cracks and the uniformity of degradation across the electrode (Supplementary Fig. 24). The degradation is observed to occur uniformly throughout the electrode rather than being localized to specific regions. In the early cycles, the material largely retains its original structure with only a few minor surface cracks. As the cycle number increases, the cracks become more numerous and larger, indicating the accumulation of structural stress. After 80 cycles, the cubic structure begins to break down, forming smaller particles, while cracks and fragmentation are widespread across the electrode after 200 cycles.

Afterward, density function theory (DFT) calculations were applied to examine the origins of capacity fading for the HSPB electrode. The coexistence of high-spin (HS) and low-spin (LS) states is well-established in PBAs. The intermediate spin (IS) states are generally unstable in PBAs. When the spin states of the divalent and trivalent iron ions are IS states, the electronic configurations are (t_2g)⁵(e_g)¹ and (t_2g)⁴(e_g)¹ respectively⁴⁴. When e_g orbitals are unevenly occupied (neither fully filled nor half-filled), the system tends to reduce its total energy by distorting the coordination environment. This distortion breaks the orbital degeneracy, further splitting energy levels and lowering the energy. In addition, computational studies comparing the (LS, HS) and (LS, LS) configurations show that (LS, HS) is energetically more stable for Fe-based PBAs⁴⁵. Figure 6a, b illustrates the projected density of states (pDOS) for high spin Fe-3d (FeHS-3d) and low spin Fe-3d (FeLS-3d). Compared to HSPB, the orbitals of FeHS-3d and FeLS-3d for LSPB and MSPB are close to the Fermi level, which implies their improved frame conductivity^46,47. The above calculated result is consistent with the rate performance collected at three electrodes^45,48. The integrated crystal orbital Hamiltonian population (ICOHP) values allow direct measurements of bond strengths⁴⁹. During the whole charging process, the values of the Fe-N/Fe-C bonds of HSPB are lower than those of MSPB and LSPB in the intermediate state, implying a relatively weak bond strength of HSPB intermediates. Additionally, the -ICOHP value differences of Fe-N/Fe-C bonds for HSPB with various different states of charge are larger than that of LSPB/MSPB, which also has a negative influence on the structural stability. To confirm the lattice changes during cycling, DFT calculations were performed to determine the structural changes in the HSPB, MSPB, and LSPB during the Na-ion extraction process⁵⁰. Figure 6d depicts the crystal structures of the three samples, where Fe atoms are coordinated with either six C or six N atoms (Fe-C and Fe-N octahedra, respectively). A further illustration of the corresponding electronic configurations and spin-charge densities of FeHS-3d and FeLS-3d can be found in Fig. 6d and Supplementary Fig. 25. The intercalation energy of HSPB is consistently lower than that of MSPB and LSPB, indicating stronger sodium-ion interactions in HSPB (Supplementary Fig. 26). While this enhances structural stability during sodium insertion, it leads to higher energy barriers during extraction, causing significant stress, microcracks, and lattice deformation over charging process. In contrast, the weaker sodium-ion interactions in MSPB and LSPB reduce structural stress during extraction, enabling smoother recovery and better structural stability over long-term cycling. As the lattice distortions are caused by the entire process of Na-ion movement in three samples, Fe-C and Fe-N octahedral sites were selected for the calculations. In general, huge distortion and moderate distortion were classified according to the change in the bond length. DFT studies show that Fe-C and Fe-N octahedra of HSPB have major distortions after the removal of Na⁺, with almost all directions diminished (Fig. 6e). On the other hand, this also means that when sodium ions intercalate, Fe-C and Fe-N bonds in HSPB are violently stretched in all six directions. In contrast to HSPB, only two directions of Fe-N bonds are hugely shortened in LSPB, and four directions of Fe-C bonds remain almost uninfluenced during Na⁺ extraction. Based on these results, HSPB undergoes a significant distortion during cycling, resulting in structural degradation and decreased capacity.

a The pDOS of FeLS-3d states for the three samples. b The pDOS of FeHS-3d states for the three samples. c The -ICOHP values of FeHS-N (left) and FeLS-C (right). d Crystal structures showing the splitting of FeLS-3d and FeHS-3d states in the three PBAs. e DFT predictions of bond changes and the stretch directions in the charging process of HSPB, MSPB, and LSPB.

Dual-pronged strategy and practical cells investigation

From the results of the characterization and theoretical calculations of PBAs during cycling, it is evident that the primary degradation mechanism is due to significant lattice distortion during charge-discharge processes and subsequent irreversible phase transformation. These factors lead to the disintegration of the microscopic structure and the loss of active sites, particularly the deactivation of high-spin divalent iron ions. Therefore, a small amount of manganese ions was introduced into the high-spin sites to adjust the local coordination environment of high-spin Fe²⁺. By regulating the solution environment, the nucleation of PBAs crystals was controlled, and a modified sodium-rich Prussian blue analogs (M-HSPB) were successfully synthesized. The XRD results for M-HSPB are shown in Fig. 7a, showing good crystallinity. After the introduction of some transition metal ions (Mn²⁺) to modulate the coordination environment, the material retained rhombohedral phase without any change. By regulating the solution environment during synthesis, the M-HSPB crystals did not grow to approximately 4 μm like HSPB but exhibited a particle size distribution around 1 μm. Characterizations of LSPB indicate that smaller crystal nucleation helps maintain the integrity of the microstructure, reducing the surface crack during cycling. Elemental analysis confirms the uniform incorporation of Mn ions into the M-HSPB structure. The initial water content of the M-HSPB was determined through TGA, with a lower water content (Supplementary Fig. 27). A lower initial water content indicates fewer structural defects in M-HSPB, which helps maintain a stable framework during cycling. The in-situ synchrotron XRD was performed on the M-HSPB electrodes (Supplementary Fig. 28). Notably, during the charging and discharging processes, the volume distortion of M-HSPB is 4.45%, smaller than the distortion of HSPB. Smaller volume changes have a big benefit for long-cycle stability. These findings suggest that M-HSPB shows the ability to minimize stress during phase transitions. In addition, the introduction of Mn²⁺ into the HSPB structure significantly alters the electronic environment of Fe. Mn²⁺ doping at Fe-N sites modifies the charge distribution of the remaining Fe atoms, increasing the positive charge at Fe-N coordination sites (Supplementary Fig. 29). This change enhances the tendency of Fe to lose electrons, improving its redox activity. Additionally, the altered charge distribution boosts electronic conductivity, collectively explaining the enhanced cycling stability and redox performance of M-HSPB. With the dual regulation of coordination environment and nucleation, the prepared M-HSPB was employed in practical sodium-ion cylindrical batteries. As shown in Fig. 7d, the prepared M-HSPB was used as the cathode, while commercially available hard carbon (HC) served as the anode. After balancing the capacities, HC | | M-HSPB 18650/33140 type sodium-ion cylindrical batteries were assembled. As shown in Fig. 7e and Supplementary Fig. 30, the assembled HC | | M-HSPB 18650-type cells exhibited good rate capability, retaining 86% of its capacity at 3 C. Moreover, due to the modulation of the local coordination environment and control of crystal nucleation, the prepared HC | | M-HSPB exhibited stable cycling performance (Supplementary Fig. 31). After 1000 cycles, the capacity retention reached 80.1%, showing significantly improved stability compared to HSPB (Fig. 7f). Additionally, the voltage of cells remained stable during cycling, resulting in an energy retention rate of 77.6% after 1000 cycles. The 18650 batteries using HC | | M-HSPB display good cycle stability than other full-cells with different reported PBA-based materials (Supplementary Table R5). Due to the lower internal resistance and wider market potential of larger cylindrical batteries, the performance of HC | | M-HSPB in 33140 cylindrical batteries (Supplementary Fig. 32) was also investigated. Supplementary Fig. 33 presents the computed tomography picture of 33140 cylindrical battery, which adopts a full-tab design, clearly showing the cathode, anode, and separator. With the use of a high-temperature resistant separator, the 33140-type cells were able to operate normally at 100 °C. As shown in Fig. 7g, the discharge capacity retention at high temperatures was over 97%. In a low-temperature environment, 88% capacity retention was achieved at −20 °C compared to performance at 25 °C, and 68% capacity retention was maintained even at −40 °C (Fig. 7h). This demonstrates the potential of PBAs for applications of practical SIBs with a wide temperature range. The cylindrical cells were mainly designed to validate the practicality of the materials rather than to achieve high specific energy. The 18,650 cell (~26.5 g) achieved a specific energy of 68 Wh kg⁻¹, while the 33,140 cell (~174 g) reached 108 Wh kg⁻¹. While the theoretical specific energy of PBAs-based sodium-ion batteries can exceed 160 Wh kg⁻¹ with advanced engineering, currently available commercial batteries of this type typically deliver 100–120 Wh kg⁻¹. These are primarily used in power-focused and wide-temperature applications rather than for maximizing specific energy, demonstrating their competitiveness in niche markets.

a XRD pattern of M-HSPB. SEM image of M-HSPB (b). c Element mapping and corresponding SEM image of M-HSPB. d The potential matchup shown by the charging/discharging profiles of M-HSPB and HC. e The rate performance of HC | | M-HSPB 18650-type cell at room temperature. f Cycling performance of HC | | M-HSPB 18650-type cell at room temperature. g The electrochemical performance of HC | | M-HSPB 33140-type cell under high temperature (25 °C–100 °C). h The electrochemical performance of HC | | M-HSPB 33140-type cell under low temperature (−40 °C–25 °C).