Unique Impacts of Scandium Doping on Electrode Performance of P’2‐ and P2‐type Na_2/3MnO₂

Abstract

Layered sodium manganese oxides are one of the promising candidates as cathode materials for high‐capacity Na‐ion batteries free from rare elements. Among the polytypes of layered sodium manganese oxides, the P’2‐type Na_2/3MnO₂ electrode exhibits the highest reversible capacity ≈220 mAh g⁻¹ based on the Mn^3+/4+ redox couple; however, the cycle life remained a significant challenge. Because of Jahn‐Teller (JT) active Mn³⁺ ions ([Ar]4s²3d⁴ ), the layered structure is cooperatively distorted and experiences complex structural changes during Mn^3+/4+ redox accompanied by Na extraction/insertion. The impact of scandium(III) dopant in Na_2/3[Mn_{1‐ x} Sc _x ]O₂ on the cooperative lattice‐distortion and electrode performance is systematically investigated, and prepartion for distortion‐free hexagonal P2‐type Na_2/3[(Mn_{1‐ x} Sc _x )_0.93□_0.07]O₂ (□ = vacancy) allows the contribution of the vacancies on the Mn site to be checked. Furthermore, a comparative study of the case of other trivalent metals of yttrium(III) and aluminum(III) as dopants is demonstrated. Thus, it is concluded that a unique synergetic effect of the Sc doping and JT distortion provides significant improvement of redox activity of P’2‐Na_2/3[Mn_{1‐ x} Sc _x ]O₂, demonstrating stable cycling for more than 300 cycles in a Na‐ion cell. The findings highlight the critical role of the coexistence of the Sc dopant and honeycomb ordering of Mn^IV(Mn^III _{1‐ x} Sc^III _x ) in [Mn_{1‐ x} Sc _x ]O₂ slabs of P’2‐Na_2/3[Mn_{1‐ x} Sc _x ]O₂ on the stable redox for long‐life Na‐ion batteries.

The structural stability during electrochemical cycling for P’2‐Na_2/3MnO₂ as a positive electrode for of Na‐ion battery is investigated by modulating the materials via metal substitution and varying the degree of Jahn‐Teller distortion. The cycle stability is maximized when combining Sc doping and the lattice distortion, demonstrating > 300 cycles in the coin‐type full cell.

1. Introduction

Layered sodium manganese oxides, Na _x MnO₂ in which typically x = 0–1, attract remarkable attention as the possible candidate of positive electrode materials to realize high capacity sodium‐ion batteries made from no rare metals at all.^{[ 1} , ^{3 ]} Layered Na _x MnO₂ materials exhibit high initial capacity; however, their cycle stability remained as a significant drawback. Once resolved, Na _x MnO₂ would be considered as one of the practical solutions based on life cycle assessments of practical batteries.^{[ 4} , ^{5 ]} During electrochemical cycling, complicated structural evolution proceeds upon reversible extraction/insertion of Na⁺ ions associated with Mn^3+/4+ redox reaction. Considering Jahn‐Teller active high spin six‐coordinate Mn³⁺ (3d⁴ _; t_2g ³–e_g ¹ ) and non‐active Mn⁴⁺ (3d³ _; t_2g ³–e_g ⁰ ) ions, sodium extraction and insertion brings continuous change in atomic ratio of (distorted Mn³⁺): (undistorted Mn⁴⁺) = x: (1–x) in Na _x MnO₂, leading to successive change in a strain at both atomic scale and particle level of Na _x MnO₂. These are regarded as one of the main causes of the capacity decay over cycling, though lattice parameter change of the Na _x MnO₂ is another main cause.

The crystal structure of layered Na _x MnO₂ has several polytypes, for instance, P2‐ and O3‐type, according to the notation proposed by Delmas and coworkers,^{[ 6 ]} where a letter denotes the coordination environment of Na⁺ and a number describes the number of MeO₂ slabs in a unit cell. Jahn‐Teller activity of Mn³⁺ distorts the local MnO₆ octahedra and/or the unit cell. In the latter case, the overall unit cell is distorted, the so‐called co‐operative Jahn‐Teller distortion (JTD),^{[ 7 ]} and the structure is denoted with “prime”, such as P’2‐ and O’3‐type. P2‐Na_2/3MnO₂ and P’2‐Na_2/3MnO₂ are thermodynamically stable in the overlapped temperature range, thus a mixed phase is often obtained, i.e., it is difficult to prepare a single phase of P’2 type.^{[ 8 ]} Our group succeeded in the optimized synthesis conditions to obtain each single‐phase product and investigated their electrochemical properties.^{[ 9 ]} A schematic illustration of P’2‐structure is shown in Figure 1a,b. MnO₂ slabs are stacked along the c‐axis, and Na ions with prismatic oxygen coordination reside between the slabs. The co‐operative JTD is tied along the b‐axis, distorting Mn's equilateral triangle to an isosceles triangle. To improve the cycle stability of P’2‐Na_2/3MnO₂, a variety of metal substitutions on the Mn site have been reported.^{[ 10} , ¹¹ , ¹² , ^{13 ]} Our group recently described that the Sc³⁺ doped material, P’2‐Na_2/3[Mn_0.94Sc_0.06]O₂, exhibits a prominent improvement of the battery performance and long‐term cycle stability by mitigating the structural changes during Na extraction and insertion.^{[ 14 ]} Nonetheless, the underlying reason for the substantial improvement brought by doped scandium remains unresolved. It is yet to be determined whether this effect is structure‐independent and generally applicable, or if it is confined to specific stoichiometry of the compositions or crystal structures.

Figure 1.

Schematic illustration of a) P’2‐layered structure and b) in‐plane Mn─O layer. c) SXRD pattern for o‐NMSO8, d) Raman spectra for o‐NMSO _x samples. SAED patterns of e) o‐NMO and f) o‐NMSO8.

In this study, to understand more details in the stoichiometric and crystallographic points of view, we systematically study the impact of scandium doping on P2 and P’2 polytypes of Na_2/3MnO_2, which differ in several aspects, including the oxidation state of Mn, the ordering of Na ions, the presence of lattice distortion, and Mn deficiency. The role of not only scandium doping but also stoichiometry of Na and Mn into the structural properties is revealed via synchrotron X‐ray diffraction (SXRD), X‐ray absorption spectroscopy (XAS), transmission electron microscopy (TEM), and electrochemical analysis as well as Na‐ion full cell demonstration. From these structural investigations, the key factors to improve the battery performance of the layered oxides in NIBs have been discussed through comparing the two polymorphous P’2/P2 with and without Sc doping.

2. Results and Discussion

2.1. Material Characterization

Single‐phase P’2 products with Sc doping are obtained via a conventional solid‐state reaction as confirmed by XRD data. We hereafter use the following abbreviations: as P’2‐type has an orthorhombic structure, we refer to P’2‐type Na_2/3[Mn_{1‐ x} Sc _x ]O₂ (x = 0, 0.06, 0.08, 0.11) as o‐NMO, o‐NMSO6, o‐NMSO8, o‐NMSO11, respectively. Since the doping with 6% Sc doped material showed improved cycle performance,^{[ 14 ]} we further examined the dependence of Sc amount on the electrode reversibility and structural change in the Na cell. Consequently, we found that o‐NMSO8 demonstrates superior electrode performance in terms of the capacity retention (see Figure S1, Supporting Information), and details will be mentioned later.

The chemical compositions of the obtained materials are carefully examined through ICP‐AES measurement, and the average oxidation state of manganese is calculated with the assumption of a stoichiometric amount of oxygen. (Table S1, Supporting Information) The oxidation state of Mn is reasonably increased with the increase of Sc³⁺ amount and decrease Mn³⁺ in Na_2/3[Mn^IV _1/3Mn^III _{2/3‐ x} Sc^III _x ]O₂ when the charge neutrality is considered there. In the XRD patterns of o‐NMSO8, all Bragg diffraction peaks can be assigned to a P’2 type layered phase with a space group of Cmcm without any impurity peaks. Rietveld analysis of the SXRD pattern was conducted for o‐NMSO8 as shown in Figure 1c. The diffraction patterns and refined parameters for all o‐NMSO samples are summarized in Tables S2–S4 and Figure S2–S3 (Supporting Information), respectively. All the refinement results give reasonably low R values, confirming that the well‐crystalline single phases of P’2‐Na_2/3[Mn_{1‐ x} Sc _x ]O₂ (x = 0 – 0.11) are successfully synthesized. The linear increase in lattice parameters confirms partial substitution of Sc for Mn in P’2 phase^{[ 14 ]} because of the larger ionic radius of Sc³⁺ (0.745 Å in 6 coordination) than that of Mn³⁺ ion (0.645 Å in 6 coordination) according to Vegard's law (Figure S4, Supporting Information). It is notable that uniaxial elongation, defined as a’ (Figure 1b) remained significant even after Sc doping. For example, the a’‐axis for o‐NMSO8 is 5.5% longer than the a‐axis, which is equivalent to 6.2% in the case of o‐NMO.

To further investigate the lattice distortion, Raman spectroscopy measurement was performed for Sc‐doped samples (Figure 1d). There are two distinct peaks at 580 and 643 cm⁻¹, A _1g‐like signals as an analogue to the non‐distorted P2‐materials.^{[ 15 ]} These two Raman bands should originate from Mn─O stretching motion and thus be strongly affected by the two different distances of Mn─O bonds in the distorted MnO₆ octahedra, leading to the peak split. This orthorhombic material is expected to show a total of nine Raman bands, but probably the other signals are too broadened due to the high Na mobility, which is consistent with the previous study for non‐doped material.^{[ 8 ]} Then, while the Sc³⁺ amount increased, the peak near 580 cm⁻¹ shifts to lower wavenumber and the other peak near 650 cm⁻¹ shifts to higher wavenumber, implying that the energy state of the Raman activity changed by the ratio of distortion. The Raman‐active vibration modes of this material are under investigation, and the results will be reported in the future.

TEM and SAED measurements were carried out to further investigate the crystal structure of non‐doped and Sc‐doped samples. The SAED patterns along [001] axis of o‐NMO and o‐NMSO8 are shown in Figure 1e,f, respectively. In addition to the agreement of the main spots assigned to the Cmcm structure, some clear diffraction spots are observed as highlighted with red triangles. The superlattice spots of o‐NMO are assigned to (6^* a, 4^* b), which is consistent with our previous study.^{[ 9 ]} The superstructure is caused by the Na/vacancy ordering as well as the Mn^3+/4+ charge ordering. Interestingly, Sc doping modulates the superstructure to (6^* a, 2^* b) as shown in Figure 1f. In the SXRD pattern, the superstructure peak is found at 2θ = 12.9° and not assigned to the known superstructure in Figure S5 (Supporting Information). A detailed understanding of the long‐range ordering is still required and will be reported elsewhere. Here, we conclude that Sc doping doesn't disturb the in‐plane ordering, combined with Na⁺‐vacancy and honeycomb ordering of 2Mn³⁺: Mn^4+, differing from other dopants, diminishing the orderings^{[ 11 ]} but also modulating the way of ordering.

Based on the systematic change of lattice constant as a function of Sc amount through Rietveld refinement of high‐quality synchrotron XRD data, we conclude that Sc³⁺ was substituted for the Mn³⁺ site selectively in the honeycomb ordered distribution of Mn³⁺ and Mn⁴⁺ ions for o‐NMO.^{[ 9} , ^{16 ]} In addition, the crystal structure is rationally modulated by larger distortion‐free Sc³⁺ doping both in the long‐range scale, including change of cooperative JTD as well as lattice constants, and in the local MeO₆ scale via Raman vibration. The impact of those changes on the electrochemical reaction is investigated and discussed in the following section.

Sc doping provides an impact on the particle morphology, and how those changes affect the powder properties are investigated. Actually, Sc doping suppresses crystal growth, and the average particle size is monotonously decreased from 5.4 µm (o‐NMO) to 2 µm (o‐NMSO11) as shown in Figure 2a. Such a sintering inhibitor effect is often observed when metal elements as additives are incorporated in the structure.^{[ 11 ]} The obtained materials have a plate‐like and highly crystalline morphology. Homogenous distribution of scandium in the primary particle is confirmed by TEM/EDS mapping, exhibiting the equivalent color contrast between Mn and Sc maps (Figure 2b–e). It is conclusive by combining XRD results that the majority of Sc³⁺ ions successfully replace Mn³⁺ in the MnO₂ slab in the entire particles. The water resistivity of the obtained samples is examined by checking the structural change before and after immersing the powders into the water for 10 min. The shift of 002 diffraction peak of o‐NMSO8 is remarkably lower than that of o‐NMO (Figure 2f). No change of the 002 peak width of o‐NMSO8 indicates that the crystallinity remained largely intact during exposure, whereas a slight asymmetric broadening for o‐NMO indicates an increase in lattice strain due to the heterogeneous Na loss, progressing from the surface to the bulk of the particles. It means the Na loss from the bulk structure to the immersed water is dramatically suppressed by Sc doping. It is contrary to the increased reaction area resulting from the smaller particle by Sc doping (Figure 2a). There are several previous literatures reporting that metal substitution is effective to improve the moisture stability of sodium layered materials, such as Ca doping into O3‐NaMeO₂,^{[ 17 ]} Cu doping into P2‐Na_2/3MeO₂,^{[ 18 ]} and F doping into P’2‐Na_2/3MnO₂.^{[ 19 ]} The improvement mechanism of the water resistance is being discussed from the view of enhancing bulk stability and forming the surface protection layer.^{[ 20 ]} The former bulk improvement could be attributed by stronger bonding between doped Sc and oxygen.^{[ 21 ]} In addition, Yang et al. recently proposed the use of ionic potential as an indicator of moisture stability.^{[ 22 ]} A lower average ionic potential correlates with stronger Na–O bonding, which in turn suppresses reactivity with moisture and reduces Na⁺ ion loss. The calculated values for the NMSOx samples are summarized in Table S1 (Supporting Information), confirming that Sc doping effectively decreases the average ionic potential, thereby contributing to the enhanced moisture stability observed in this study. With the EIS investigation for the symmetric cell of the pristine samples (Figure 2g), the o‐NMSO8 electrode has a lower charge transfer resistance than the o‐NMO electrode. More interestingly, an additional semi‐circle at the high‐frequency region appeared for o‐NMSO8. A possible formation of surface film could play as the protection layer against water exposure, as well as promoting a less resistive passivation layer like CEI (cathode‐electrolyte interphase) in Na cell. This layer is presumed to consist of excess Sc species segregated and accumulated at grain boundaries during synthesis because dopant accumulation at grain boundaries is expected, as reported for Al^{[ 23 ]} and Ti,^{[ 24 ]} when the dopant doesn't fit to the structure or the concentration exceeds its solubility limit. Such a layer may act as a protective interface, contributing to the enhanced moisture stability and prolonged cycle life observed in o‐NMSO8. These results are further supported by a quantitative comparison through equivalent circuit fitting. The circuit model and fitting curves are summarized in Figure S6 (Supporting Information). The fitting reveals a clear decrease in charge transfer resistance (R₃), consistent with a smaller particle size, and a noticeable increase in interfacial resistance (R₂). Although this study is not able to provide a clear distinction between bulk and surface contributions to the battery performance, it is noteworthy that the unique doping with Sc³⁺ ions contributes to engineering the stable interface.^{[ 24 ]}

Figure 2.

a) Particle size distribution for o‐NMSOx samples. b,c) TEM images and EDS mapping of o‐NMSO8 for d) Mn mapping, and e) Sc mapping. f) XRD patterns of o‐NMO and o‐NMSO8 before and after water exposure for 10 min. g) EIS spectra for the symmetric cells of o‐NMO and o‐NMSO8.

2.2. Electrochemical Performance

The electrode behavior of the materials is investigated in coin‐type aprotic Na cells at 25 °C. The charge/discharge curves or o‐NMO and o‐NMSO8 are shown in Figure 3a,b. The materials are electrochemically cycled between 1.5 and 4.4 V and exhibit more than 200 mAh g⁻¹ of reversible capacity, which corresponds to ≈0.8 mol of Na extraction and insertion. The amount of Na extraction during the first charge is lower than the Na insertion during the first discharge, as the Na metal counter electrode compensates for the intrinsic Na deficiency of P2‐type cathode materials. The slightly lower initial charge capacity for o‐NMSO8 compared to o‐NMO is primarily attributed to the amount of electrochemically active Mn³⁺/Mn⁴⁺ redox, as determined by ICP analysis (Table S1, Supporting Information). Sc doping leads to partial oxidation of Mn ions and a slight increase in Na loss under the synthesis conditions employed in this study. The slight loss of reversible capacity, from 220 mAh g⁻¹ (o‐NMO) and 216 mAh g⁻¹ (o‐NMSO8), and smoothening of charge/discharge curves by Sc doping are observed and consistent with o‐NMSO6.^{[ 14 ]} To further elucidate the impact of Sc doping on the electrochemical behavior, cyclic voltammetry and dQ/dV analyses were performed, as summarized in Figure S7 (Supporting Information). These results further support the observation of smoother electrochemical reactions in the Sc‐doped sample. Given that the superlattice spots are observed by SAED for the pristine o‐NMSO8 (Figure 1f), it is notable that the Sc doping effectively smoothens the voltage curves, suggesting suppression of Na/vacancy ordering during electrochemical Na extraction. Sc doping successfully improves cycle stability as seen in Figure 3c compared with non‐doped materials. The optimized amount of doped Sc is found to be 8% as further improvement on the capacity retention is achieved compared with 6% and 11% doping.^{[ 14 ]} (Figure S1, Supporting Information) We interpreted that the CEI‐like Sc layer should be effective to mitigate the side reaction with liquid electrolyte as well as enhance moisture stability. (Figure 2f) The improvement on discharge rate capability is confirmed by Figure 3d, which would be attributed to a larger specific surface area due to smaller particles and also smoother bulk diffusion of Na⁺ ions without Na/vacancy ordering for o‐NMSO8. The result is also consistent with the EIS result (Figure 2g; Figure S6, Supporting Information) that o‐NMSO8 has lower charge transfer resistance than o‐NMO. Although the galvanostatic intermittent titration technique (GITT) is a well‐established method for determining Na⁺ diffusion coefficients, it was not performed due to experimental limitations. Future work will include GITT measurements to address Na⁺ ion transport behavior.

Figure 3.

The charge/discharge curves of a) o‐NMO and b) o‐NMSO8. c) Cycle performance of o‐NMSOx half cells, and coulombic efficiencies were also plotted together. d) Rate performance of o‐NMSOx. TEM and electron diffraction images of the powders after 50 cycles for e) o‐NMSO8 and f) o‐NMO.

We discuss two possible mechanisms leading to the capacity decay on the positive electrode: (i) the bulk degradation associated with the structural changes, such as P2‐O2 phase transformation,^{[ 25 ]} and (ii) the surface degradation linked to the side reaction with a liquid electrolyte.^{[ 10 ]} With these electrochemical data, we first discuss that the structural change during Na extraction/insertion should be smoother and vaguer for the doped material, evidenced as the smoothened voltage curves. By comparing the TEM and SAED images of the samples after 50 cycles, as shown in Figure 3e,f, the crystalline orthorhombic structure and particle morphology are retained for o‐NMSO8 over 50 cycles. In the case of o‐NMO, however, the remarkable cracks even deep inside the bulk region are observed, which is due to the pronounced change in the lattice volume as discussed later. The halo‐like signal in SAED for o‐NMO shows the particles being partly amorphized. We, therefore, conclude that the scandium doping is efficient to stabilize the P’2 structure over electrochemical reactions. The improvement mechanism will be further discussed with the advanced characterizations in the next sessions.

To check the reversibility and capacity retention of Na extraction of o‐NMSO8, the upper cut‐off voltage (UCV) is varied as shown in Figure 4a. When extending the potential range above 3.47 V of UCV, corresponding to the potential of OP4 phase formation,^{[ 9 ]} the voltage gap starts increasing significantly, but the capacity retention is not simply dependent on the phase transition into OP4 phase. Surprisingly, the capacity retention under the UCV = 4.0 V condition is exceptionally worse than that on any other UCVs (Figure 4b). The cells are disassembled after 50 cycles, and ex‐situ XRD measurements are conducted for the electrodes. As clearly seen in the inset of Figure 4c, the photograph shows that the separator is discolored from white to yellow, which should be resulted from a decomposition of carbonate‐based solvent,^{[ 26 ]} in the case of UCV = 4.0 V, though no coloration for UCV = 4.4 V. The coloration is a definite indication of a serious side reaction with the electrolyte during 50 cycles in the potential range between 1.5 and 4.0 V. From the ex‐situ XRD patterns, the UCV 4.4 V electrode after 50 cycles has the left shoulder on the 002 diffraction peak located at 2θ = 16°, but in the case of UCV = 4.0 V, 002 diffraction peak is totally split into the two peaks. These results indicate the existence of inhomogeneous electrode reactions, resulting in the two phases of Na‐rich and Na‐poor (charged and discharged) domains in the composite electrodes. An effective CEI is likely formed in the voltage range of 4.0–4.4 V, and thus the cycling with UCV = 4.0 V suffers from continuous Mn dissolution in the absence of an effective CEI. The same trend was also reported in the undoped materials.^{[ 9 ]}

Figure 4.

a) Initial charge/discharge curves and b) cycle performance of o‐NMSO8 at different cutoff voltages. c) ex‐situ XRD patterns of o‐NMSO8 with the cut‐off V of 4.0 and 4.4 V. d) Cycle performance of o‐NMSO8 at 1.5–4.0 V with and without pre‐cycling.

Pre‐cycling of battery is often used to ensure the formation of a robust passivation layer, such as CEI and/or gradual phase transition, leading to longer cycle life, higher capacity, and less gas generation of Li‐rich Mn oxides for Li‐ion batteries.^{[ 27} , ^{28 ]} Considering the effectiveness of precycling of the battery, we applied the precycling protocol, the initial 5 cycles up to 4.4 V, then cycling under the UCV = 4.0 V condition from 6^th cycle. The capacity retention is dramatically improved by precycling (Figure 4d). As mentioned in Figure 4c, we confirmed no significant deterioration of o‐NMSO8 crystal structure but the inhomogeneous electrode reaction, and separator coloration only in the UCV = 4.0 V case; therefore we believe that the mechanism of the better capacity retention by the precycle treatment is due to the suitable modification of the interface at the electrolyte/electrode including cathode electrolyte interphase (CEI). In fact, when we tested the o‐NMSO8//Na cells filled with the ionic liquid of NaTFSA‐[C1C2Pyrr]FSA (1 : 4 mol/mol) as an electrolyte having better anodic stability than NaPF₆ PC solution,^{[ 29 ]} the capacity retention of 4.0 V cycling is decent with no precycling.(Figure S8, Supporting Information) The detailed study of the electrolyte dependence is ongoing in our laboratory. Thus, the precycling approach is an important option to improve practical electrode performance of sodium layered oxides; moreover, this approach enables greater flexibility in electrolyte selection and enhances the cycle life of the batteries.

Based on the electrochemical investigation in Na half‐cell, coin‐type full cells of the NMSO samples are fabricated with hard carbon as negative electrodes and tested as shown in Figure 5a–c. The performance of the hard carbon (HC) electrodes in the half cell is shown in Figure S9 (Supporting Information), and the electrode used for the full cell is pre‐sodiated to through the 3^rd sodiation in Na half cell to ensure SEI formation on the anode electrode.^{[ 30 ]} The pre‐sodiation degree of hard carbon is fixed to the excess discharge capacity (≈1/3 mol) of the first cycle in Figure 3a,b for o‐NMO and o‐NMSO8, respectively. In other word, the presodiation is able to compensate deficient 0.33 mol Na sites in P’2‐type Na_2/3[Mn_{1‐ x} Sc _x ]O₂. While this method effectively mitigates initial capacity loss by compensating for Na deficiency and balancing the N/P ratio, future work will focus on tailoring the pre‐sodiation conditions to further enhance energy density and long‐term cycling performance. Both full cells of o‐NMO//(pre‐sodiated HC) and o‐NMSO8//(pre‐sodiated HC) configuration show >170 mAh g⁻¹ (per cathode weight) with > 2.1 V of average discharge voltage. The capacity retention of the NMSO8 full cell is markedly greater than that of the NMO full cell, showing ≈60% of the initial discharge capacity retained after 300 cycles.^{[ 31 ]} The trend is in agreement with the retention for o‐NMO//(excess Na) and o‐NMSO8//(excess Na) cells. The evolution of coulombic efficiency (CE) is consistent with the observed capacity decay. The CE of o‐NMSO8 remains above 99% up to 300 cycles, indicating highly reversible electrochemical reactions and stable interfacial stability. In contrast, o‐NMO exhibits the CE of ≈98% up to 100 cycles, followed by an increased fluctuation. This instability correlates with the lower capacity retention and suggests possible Mn dissolution, which leads to micro short‐circuiting.

Figure 5.

Charge/discharge curves of o‐NMSOx // HC full cells: a) o‐NMO, b) o‐NMSO8. c) The long‐term evolution of capacity retention and coulombic efficiency of the full cells. d) Ex‐situ XRD patterns of cathode electrodes after full cell after 300 cycles. HAXPES data of e) Mn 2s and f) P 1s for HC anode after 300 cycles, in which the photoelectron intensity is normalized using the O 1s peak of lattice oxygen at 529.5 eV as a reference.

The faster capacity decay of the non‐doped sample can be attributed to the structural degradation during cycling, such as a phase transition to OP4 structure occurring ≈3.5 V in charge.^{[ 9 ]} The ex‐situ XRD measurement is conducted for the positive electrodes after 300 cycles as shown in Figure 5d. While the crystallinity of the non‐doped sample is clearly lost after 300 cycles, Sc‐doped sample keeps the crystalline layered structure, which is consistent with the TEM observation (Figure 3e,f). It can be understood that the Sc doping successfully mitigates the strain of the primary and secondary particles caused by the change in lattice volume of the layered oxides during repeated Na extraction/insertion. On the anode electrodes after 300 cycles, Mn deposition is observed for both samples evidenced by HAXPES data in Figure 5e. It was reported that dissolved Mn(II) poisons the surface of graphite anode material and causes irreversible reaction on the Mn deposits, leading to a serious capacity loss in the case of LIBs.^{[ 32} , ^{33 ]} The amount of Mn dissolution should be decreased by Sc doping because of the lower intensity of HAXPES Mn 2s peak, which is a bulk analysis of surface species (Figure 5e). This improvement can be attributed to the suppression of cracking (Figure 3), caused by the phase transitions and/or CEI acting as a protection layer. Furthermore, the passivation and surface chemistry of SEI on hard carbon is also changed; the sharper P 1s peak with clearer peak splitting for o‐NMSO8 compared with o‐NMO indicates less decomposition of NaPF₆ ^{[ 34 ]} (Figure 5f) while no significant change was found in the ex‐situ XRD patterns (Figure S10, Supporting Information).

2.3. The Improvement Mechanism of Sc Doping on the Cycle Stability

To understand the improvement mechanism of the capacity retention by Sc doping, the structural evolution during 1^st charge and discharge is carefully examined with XRD. Figure 6a displays operando XRD data during the first cycle, revealing the phase evolution of o‐NMSO8. Single‐phase reaction of P’2‐I phase proceeds from y = 0.67 to 0.2 in Na _y MO₂ (M = Mn) accompanied with lattice contraction, then OP4 phase is dominant afterward. When approaching 1^st full discharge, a bi‐phasic reaction appears with P’2‐II having a further in‐plane distortion via elongation of the b‐axis.

Figure 6.

a) Color map from In situ XRD patterns of o‐NMSO8. b) Ex‐situ XRD patterns of o‐NMSO8: c) the full charge state at 4.4 V and d) the full discharge state at 1.5 V.

On the other hand, the structural changes of o‐NMSO8 are much simpler without any stepwise phase transitions, unlike those of the non‐doped sample. Based on the operando study, the characteristic data points are selected to discuss their structures from ex‐situ XRD results as shown in Figure 6b. The target structure at each voltage is successfully extracted, and the phases are consistent with the structural evolution confirmed by the in situ XRD. We observe a clear transformation from P’2 to OP4 above 3.35 V upon charge and back to P’2 below 3.24 V upon discharge. When fully discharged, the material comes back to the highly crystalline P’2 phase. In order to discuss the full charge state, ex‐situ SXRD measurement was conducted (Figure 6c). o‐NMSO8 shows mainly OP4 phase with a trace of P’2 phase as seen in the diffractograms ≈2θ = 10 degree. o‐NMO shows one additional peak, which could be indexed to the O1‐like phase.^{[ 35 ]} causing a significant intra‐plane shrinkage to form the MnO₂ sheet via nearly full Na extraction. One of the main causes of o‐NMO having poor cycle life (Figure 5c) can be derived from the serious strain in the particles caused by O1‐like domains. Substituting redox‐inactive Sc³⁺ ions could be effective to the homogeneous distribution of Na‐pillar in the entire layered oxide,^{[ 25 ]} which mitigates such a degradation and helps the structure remain highly crystalline, as confirmed with XRD patterns for the cycled electrode (Figure 5d). In addition, Sc doping keeps a stronger distortion among other 3d metal doping at the end of discharge, which is supported by our previous data as summarized in Figure S11 (Supporting Information). The distortion is more pronounced than that of Co³⁺ and Ni^2+/3+ doping, even higher than that of JT‐active Cu²⁺ doping, which also agrees with our previous study, showing the correlation between the lattice distortion and the potential of the plateau near the full discharge.^{[ 11 ]}

Mn K‐edge XANES spectra are obtained to discuss Mn redox reaction during 1^st charge and discharge. (Figure 7a,b) Mn K‐edge absorption is shifted toward higher energy upon charge and back to lower energy upon discharge, confirming that the reversible Mn^3+/4+ redox couple is responsible for the electrochemical reaction accompanied by Na extraction/insertion. The chemical shift of Mn K edge with and without Sc doping is plotted in Figure 7c. o‐NMSO8 shows slightly higher energy of the Mn K‐edge at the pristine state, showing higher oxidation of Mn than o‐NMO. It is in line with the partial replacement of Mn³⁺ with Sc³⁺ and the constant atomic ratio of Na/(Mn+Sc) for all samples, as already mentioned based on ICP data (Table S1, Supporting Information). Because o‐NMSO8 contains a lower amount of redox‐active Mn³⁺ than that of o‐NMO, it is likely that o‐NMSO8 utilizes the wider valence change of Mn^3+/4+ couple reversibly. The local distortion of manganese can be discussed with EXAFS spectra shown in Figure 7d. In contrast to undistorted hexagonal P2‐Na_xMnO₂ material having each one peak of Mn─O and Mn‐Mn bond length,^{[ 36 ]} NMSO8 shows split peaks of Mn─O and Mn‐Mn bonds. This peak split agrees with the existence of co‐operative JTD as well as the SXRD of the pristine state (Figure 1c). With the oxidation of Mn upon charge, the gap of peak splitting decreased, whereas the peak split remains at full charge after forming to OP4 phase. Considering the anion redox activity through s‐block of Group 1 and 2 cation dopants like Li⁺ and Mg²⁺, respectively, we expected to activate the oxide anion redox by Group 3 cation, Sc³⁺ doping (Figure S12, Supporting Information). Since XAS data confirmed that the manganese is oxidized and reduced constantly in the entire range of operation potential, we reasonably think that the oxide anion redox would be negligible and insignificant contribution to the redox activity of the o‐NMSO8 electrode.

Figure 7.

a,b) Ex‐situ XANES spectra of o‐NMSO8. c) Chemical energy shifts of Mn K‐edge during 1^st charge and discharge. d) Ex‐situ EXAFS spectra o‐NMSO8. The tested electrodes were prepared by charging and discharging in Na cells, followed by taking the electrodes from the cells.

2.4. Comparison with Other Doping on o‐NMO

To further illustrate the role of Sc³⁺ ion as a dopant, the other trivalent metal cations of Al³⁺ and Y³⁺, which are classified as Group 13 and 3 next to Mg and below Sc, respectively, in the periodic table, were used as dopants for o‐NMO. Through the same synthesis condition, the phase pure P’2‐type Na_2/3[Mn_0.92Al_0.08]O₂ (o‐NMAO8) powder is obtained as seen in Figure 8a, whereas Y‐doped sample shows the multiple impurity peaks under the synthetic conditions in this study. (Figure S13 and Table S5, Supporting Information) The limited incorporation of Y³⁺ into the P2‐type layered structure is probably due to its larger ionic radius (0.90 Å) than Mn³⁺ ions, while it is much smaller than Na⁺ ions. In addition, no reports have demonstrated successful Y³⁺ doping into P2‐type structures, suggesting that Y³⁺ tends to form separate phases such as the O3‐type layered phase^{[ 37 ]} and a cubic phase.^{[ 38 ]} The initial reversible capacity of o‐NMAO8 is higher than that of the non‐doped sample, reaching 238 mAh g⁻¹ but the material suffers a rapid capacity loss over the cycling. (Figure 8b) The cycle stability of non‐doped and Al³⁺ doped samples exhibited similar capacity fading (Figure 8c). It confirms the unique benefit of Sc³⁺ ion to improve the reversibility, differing from Al³⁺ doping. One possible explanation is that the smaller ionic radius of Al³⁺ ions (0.535 Å in 6 coordination) is probably insufficient to serve as a structural pillar to suppress the phase transformation, unlike the case observed with Sc doping. (Figure 6c).

Figure 8.

Electrochemical properties of o‐NMAO8: a) SXRD pattern, b) Long term charge/discharge profile, c) Comparing the cycle performance of different TM³⁺ (= Mn, Sc, Al) replacement. Electrochemical properties of h‐NMSO8: d) SXRD pattern, e) Long term charge/discharge profile, f) Comparing the cycle performance for h‐NMSO8 with h‐NMO. The schematic illustration of metal distribution for g) o‐NMO, h) o‐NMSO8 and i) h‐NMSO8.

The influence of not only doping metal elements but also vacancy of the Mn site is examined by comparing experimental data of P’2 and P2 Sc‐doped samples. As we described previously, orthorhombic P’2‐Na_2/3MnO₂ can be thermally transformed into hexagonal P2 phase by an additional heat treatment.^{[ 9 ]} It is due to the oxidation of Mn ions and simultaneous oxygen uptake. The amount of excess oxygen is defined as δ in P2‐Na_2/3MnO_{2+ δ} . It simultaneously causes vacancies in both Na and Mn sites, namely described with layered notation, Na_{2/3 ‐δ ’}(Mn_{1 ‐δ ’}□ _{δ ’})O₂. (□ = vacancy).^{[ 9 ]} The same thermal treatment is applied to o‐NMSO samples, and we successfully prepared Sc‐doped P2‐type oxide, which contains >6% of vacancy of M site in MO₂ (M═Mn and Sc) slab based on the Rietveld refinement. Hereafter, hexagonal non‐doped and 8% Sc doped P2‐samples are abbreviated as h‐NMO and h‐NMSO8, respectively. (Figure 8d; Figures S14 and S15, Supporting Information)

To our knowledge, this is the first report of the structural and electrochemical investigation of Sc doped P2 materials as well as Sc doping into P’2 material. Notably, the capacity retention of the P2 electrode is not improved by Sc doping, unlike the case of P’2 (Figure 8e,f), implying a synergetic improvement on cycle stability by combining the lattice distortion and Sc doping. The result provides an important insight about how Sc doping is effective for P’2‐material. The schematic illustrations of in‐plane metal and charge distribution on MnO₂ layer for o‐NMO, o‐NMSO8, and h‐NMSO8 are summarized in Figure 8g,h, and 8i, respectively. Considering that the vacancy formation also smoothened the charge/discharge curves, both vacancies and redox‐inactive Sc³⁺ dopants may play similar roles in inducing randomness in the charge distribution of MnO₂ planes, thereby suppressing the ordering of Na ions and vacancies. Doped Sc³⁺ dopant acts as the 3d‐electron‐vacant site with keeping the P’2 phase, while the actual vacancy (both metal‐ and 3d‐electron‐vacancy) induces the increase in the oxidation number from JT‐active Mn³⁺ to non‐distorted Mn⁴⁺ ion with keeping the honeycomb ordering of MO₂ slab. Similar improvement verifies that smoothening the electrochemical curves is effective in enhancing the cycle stability.^{[ 25 ]} It can reasonably explain why Sc doping into vacancy‐containing sample (h‐NMSO8) shows no additional improvement. The approach of the phase control from P’2 to P2 phases of doped samples enables a fair comparison of electrochemical properties with a minimal change of the particle morphology and size distribution, which often hinders discussing the doping effect on the performance.

3. Conclusion

Scandium substitution for Mn site in P’2‐Na_2/3MnO₂ effectively modulates the structure, such as the lattice constant, Mn─O stretching mode, and the crystal growth, while keeping cooperative Jahn‐Teller distortion and the superstructure. It provides a significant improvement of the structural stability. The as‐prepared material exhibits enhanced moisture stability, which could de‐bottleneck the process of battery production. Basic electrochemical testing in the half cell reveals a significant improvement on the cycle stability by Sc doping. The structural durability against repeated Na insertion/extraction is greatly improved and thus, the crystallinity is remarkably retained after 50 cycles than the non‐doped sample.

The way to utilize Sc doped P’2‐Na_2/3MnO₂ as a positive electrode material is investigated, and we've proposed the pre‐cycling at the optimal voltage range, like 4.4 V as a crucial step to have a stable CEI layer. Both Sc doping and pre‐cycling are effective in elongating the cycle stability. Thus, the side reaction with liquid electrolyte is suppressed, which is confirmed through the suppression of Mn dissolution by HAXPES measurement on the cycled anode electrodes. The improvement of the structural stability by Sc doping is evident by ex‐situ and in situ XRD.

Doped Sc³⁺ ions played as a 3d‐electron‐vacant pillar, leading to less strain to the primary and secondary particles. The comparison with Sc doped undistorted material further unveils the synergetic effect of Sc³⁺ doping into the distorted structure. This study illustrates the general strategy to extend the structural stability of the layered metal oxides, which reversibly interact with the mobile cations, and the way to maximize the battery performance with those materials.

4. Experimental Section

Material Synthesis

P’2‐type Na_2/3[Mn_{1‐ x} Sc _x ]O₂ (x = 0, 0.06, 0.08, 0.11), P’2‐type Na_2/3[Mn_0.92TM_0.08]O₂ (TM = Al, Y), and P2‐Type Na_2/3[Mn_{1‐ x} Sc _x ]O₂ (x = 0, 0.08) were synthesized by solid state reaction. Stoichiometric powder mixture of sodium carbonate (Na₂CO₃, Nacalai Tesque, 99.8%), manganese oxide (Mn₂O₃, prepared from MnCO₃, Aldrich, by heating under air at 700 °C for 12 h), scandium oxide (Sc₂O₃, Kojundo Chemical Lab., 99.9%), aluminum hydroxide (Al(OH)₃, Kanto Chemical, ≥ 95.0%), yttrium oxide (Y₂O₃, Wako Pure Chemical Industries, 99.99%) were ball‐milled for 12 h at 600 rpm. The milling was performed using a planetary ball mill (Pulverisette 7 Classic Line, Fritsch GmbH) equipped with a 45 mL zirconia jar and zirconia balls (2 and 5 mm in diameter). The ball‐to‐powder weight ratio was maintained at 10:1. The jar was filled with acetone to ≈67% of its volume to facilitate wet milling. The mixture was dried at 80 °C in air and pressed into pellets. For P’2‐type materials, the pellets were heated under air to 1050 °C at a ramping rate of 1 °C min⁻¹ and then quenched without dwell time by immediately transferring the products into an Ar‐filled glove box. To obtain the P2‐type compounds, the as‐prepared P’2‐type were used as the starting materials and pressed into pellets. The pellets were heated under air to 700 °C (ramping rate of 1 °C min⁻¹) and without dwell time, slowly cooled to 500 °C (ramping rate of 0.5 °C min⁻¹), then quenched in the same way.^{[ 9 ]} The as‐synthesized samples were kept inside an Ar‐filled glove box to avoid contact with moisture and oxygen in air.

The following are used abbreviations; P’2‐type has an orthorhombic structure, P’2‐type Na_2/3[Mn_{1‐ x} Sc _x ]O₂ (x = 0, 0.06, 0.08, and 0.11) is referred to as as o‐NMO, o‐NMSO6, o‐NMSO8, o‐NMSO11, respectively. P’2‐type Na_2/3[Mn_0.92TM_0.08]O₂ (TM = Al, Y) are referred as o‐NMAO8 and o‐NMYO8, respectively. As P2‐type has a hexagonal structure, P2‐type Na_2/3[Mn_{1‐ x} Sc _x ]O₂ (x = 0, 0.08) are referred to as h‐NMO and h‐NMSO8.

Material Characterization

The crystal structure of the as‐prepared materials before and after cycling in Na cells was examined with lab‐scale and synchrotron XRD. Lab‐scale XRD measurement was carried out using a diffractometer equipped with an X‐ray tube of Cu target with Ni filter (SmartLab, Rigaku Corporation). Synchrotron XRD was measured at the beam line BL02B2,^{[ 39 ]} SPring‐8 in Japan, equipped with a large Debye‐Scherrer camera. The wavelength of the incident X‐ray beam was set to 0.799 Å using a silicon monochromator, which was calibrated with a CeO₂ standard. Structural parameters were refined by the Rietveld method with the program RIETAN‐FP and the program Conogragh. Schematic illustrations of the crystal structures of the sample were drawn using the program VESTA.^{[ 40 ]} Structural changes during initial charging and discharging processes for o‐NMSO8, h‐NMSO8 were examined by an operando XRD technique using a battery‐cell attachment (Rigaku Corporation), and the data were collected by an X‐ray diffractometer (MultiFlex, Rigaku Corporation) equipped with a high‐speed position‐sensitive detector (D/teX Ultra, Rigaku Corporation) and non‐monochromatized but Ni‐filtered Cu Kα radiation as an X‐ray source.

Raman analysis was performed using a 532.0 nm laser with 2 mW of power and a Raman spectrometer (Raman 11i, Nanophoton) at room temperature. The samples were placed between a glass slide and a cover glass inside an Ar‐filled glove box to avoid exposure to the air prior to analysis.

Particle morphology of the samples was observed by using a scanning electron microscope (SEM, JCM‐6000, JEOL Ltd.) equipped with an energy‐dispersive X‐ray spectrometer. The outermost surface structure and morphology of the samples were observed with the transmission electron microscope (TEM, JEM‐2100F, JEOL Ltd.) at an accelerating voltage of 200 keV. Selected area electron diffraction (SAED) patterns of the samples were also collected with the microscope. The precise chemical compositions of the prepared materials were examined by inductively coupled plasma atomic emission spectrometry (ICP‐AES, SPS3520UV, Hitachi High‐Tech Science Corporation) analysis after dissolving the samples in HCl solutions. Moisture stability is investigated by comparing the XRD patterns before and after immersing the powder in water for 10 min with following steps of filtering and overnight drying.

X‐ray photoelectron spectroscopy (XPS) measurements were performed by using an X‐ray photoelectron spectrometer (JPS 9010MC, JEOL Ltd.). XPS spectra were collected at room temperature using non‐monochromatic Mg Kα radiation (1253 eV) under operation at 12 kV and 10 mA. The photoelectron energy of the spectra was calibrated by the binding energy of lattice oxygen to 529.5 eV.

Hard X‐ray absorption spectroscopy was performed at the beamline BL14‐B2, SPring‐8 in Japan. The spectra were collected with a silicon monochromator (Si(111)) in transmission mode. The intensities of the incident and transmitted X‐ray were measured using ionization chambers at room temperature. The energy was calibrated by setting the first inflection point in the spectrum of a Cu foil to 8979 eV. Electrode samples were galvanostatically cycled with the coin cells. The positive electrodes were taken out from the coin cells, rinsed with propylene carbonate (PC) and diethyl carbonate (DEC) solvent, and then dried at room temperature in an Ar‐filled glovebox. The several electrodes on Al‐foil current collectors were stacked and sealed in an oxygen and moisture‐barrier polymer film in an Ar‐filled glovebox. Absorption spectroscopy data were normalized and treated using the ATHENA program in the Demeter software package.^{[ 41 ]} Fourier transform of χ(k)k ³–k plots, where χ and k are oscillatory components of normalized absorption and angular wavenumber, respectively, was performed using the ATHENA program. The manganese oxidation states were also examined by hard X‐ray photoelectron spectroscopy (HAXPES), which was conducted at the high excitation energy of 7938.9 eV using a Scienta Omicron photoelectron energy analyzer, R4000, at the BL46XU beamline in SPring‐8, Japan. The binding energies in the HAXPES spectra were calibrated and normalized using the O 1s peak of lattice oxygen at 529.5 eV as a reference.

Electrochemical Characterization

Working electrodes were prepared by the following procedures in an Ar‐filled glove box. A slurry with a mixture of 80 wt.% active material, 10 wt.% acetylene black (AB, Strem Chemicals, Inc.), and 10 wt.% poly(vinylidene fluoride) (PVdF, Polyscience, Inc.) dispersed in N‐methyl pyrrolidone (Kanto Chemical Co., Ltd.) as a solvent was prepared and pasted on aluminum foil and then dried at 110 °C under vacuum. Single‐side‐coated electrodes were punched into 10‐ or 15‐mm disks in diameter for R2032‐type coin cells with an Al‐clad type of stainless‐steel cap for the positive electrode side (Hohsen Corp.). The active material loading was ≈2.5 mg cm⁻². The separator used was a glass fiber filter (GB‐100R, ADVANTEC, Co.), the electrolyte solution was 1.0 mol dm⁻³ NaPF₆ dissolved in PC (Kishida Chemical Co., Ltd.), and the counter electrode was sodium metal (purity > 99%, Kanto Chemical Co., Ltd.). The electrode preparation and fabrication of the coin cells were carried out in an Ar‐filled glove box. Galvanostatic charge/discharge tests were conducted with a charge/discharge measurement system (TOSCAT‐3100, TOYO System Co., Ltd.) in the voltage range of 1.5–4.4 V versus Na at C/20 (≈13.0 mA g⁻¹) at 25 °C.

The electrochemical performance of the positive electrode materials was also evaluated in Na‐ion full cells featuring hard carbon negative electrodes filled with 1.0 mol dm⁻³ NaPF₆ dissolved in PC solution, cycling in the voltage range of 0.5 – 4.35 V versus Na at C/5 (≈75 mA g⁻¹) at 25 °C. Hard carbon (HC) negative electrodes were fabricated by a slurry with a mixture of 85% hard carbon (Carbotron P(J), Kureha), 10 wt.% AB and 5 wt.% sodium polyacrylate binder (PANa, Kishida Chemical Co. Ltd.) with deionized water as a dispersion medium to form a homogeneous slurry, followed by applied on aluminum foil and then dried at 150 °C under vacuum. To check the cycle characteristics of HC electrodes, galvanostatic charge/discharge cycling was carried out in the voltage range of 0.002–2.0 V versus Na. Full cell balance was achieved at a capacity ratio of the negative to positive electrode materials (N/P ratio) of 1.05–1.15: 1. Prior to assembling full cells, both the positive and negative electrodes were pre‐cycled for 5 cycles under the half‐cell's charge/discharge conditions.

Electrochemical impedance spectroscopy (EIS) was performed using a VMP3 potentiostat (BioLogic, Seyssinet‐Pariset, France) at a cell voltage of ≈0 V such that the electrodes in the symmetric cells were in equal states of charge. The frequency was scanned from a range of 1 MHz to 10 mHz. The Randles‐type equivalent circuit is used^{[ 42 ]} and the fitting was performed using EC‐Lab v11.61.

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supporting Information

Moriya K., Kumakura S., Kim E. J., et al. “Unique Impacts of Scandium Doping on Electrode Performance of P’2‐ and P2‐type Na_2/3MnO₂ .” Adv. Mater. 38, no. 1 (2026): e11719. 10.1002/adma.202511719

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.