To eliminate our dependence on carbon-based energy production, specifically in the field of mobility, it is essential to develop energy storage devices able to store the intermittent energy harvested by renewable energy sources (Armand and Tarascon 2008). A key to the deployment of electric vehicles is the ability of batteries to charge quickly while ensuring long cycle life, and optimized safety (Burnham et al., 2017; Li et al., 2020). The current use of graphite as a negative electrode only partially satisfies these needs (Choi et al., 2020). In fast-charging batteries, graphite enables the insertion of lithium at an average potential close to that of Li⁺/Li (0.1 V vs. Li⁺/Li), which can lead to the formation of lithium plating followed by dendrite growth, resulting in possible short circuits and thermal runaway. Moreover, the specific capacity of graphite is drastically reduced at high current densities due to kinetic limitations on ion mobility between graphene layers (Sivakkumar et al., 2010).

For these reasons, alternative negative electrode materials have attracted increased attention over the past decade One of the best known negative electrode materials for high-rate applications is Li₄Ti₅O₁₂ (LTO) (Prakash et al., 2010; Song et al., 2014; Thackeray and Amine 2021). Crystallizing in a spinel structure, this zero strain material offers an alternative route towards designing fast charging batteries with a relatively high specific capacity. Moreover, new approaches have been proposed to offer more options for the replacement of graphite electrode. Among others, shear-structure materials including TiNb₂O₇ (Griffith et al., 2019) or Nb₁₆W₁₈O₉₃ (Griffith et al., 2018) offer relatively large capacity values at high rate between 1V and 3V vs. Li⁺/Li. Low volume changes upon lithium insertion and an operating potential that avoids the formation of lithium-consuming solid electrolyte interphase (SEI) are the main advantages offered by these materials. In addition, a new approach has recently been proposed to transform a material traditionally used as an ionic conductor into a material for high-rate Na-ion batteries. Partial substitution of Al by Fe in ß”-Al₂O₃ offers both ionic and electronic conduction, allowing fast and reversible Na⁺ intercalation (Butts et al., 2021). This material displays cyclic voltammogram with broad peaks attributed to a surface-controlled charge storage mechanism. Similar approaches are needed to unravel different fast charging materials and their associated mechanisms.

In this work, we propose an alternative way to produce materials that can insert alkali cations with minor diffusion limitation, thus allowing high current density operation. More precisely, the A-site of AgNbO₃ perovskite synthesized by sol-gel process has been partially substituted by lanthanum (La³⁺), allowing the creation of vacancies in this crystallographic site. Insertion of Li ions is observed in every substituted Ag_1-3xLa_x□_2xNbO₃ compound (with x > 0), thus leading to experimental specific capacities close to the theoretical values. Calculation of b-value shows a non-diffusion-controlled mechanism and in situ XRD shows negligible volume expansion (zero strain electrode) during the insertion/deinsertion mechanism for this new family of materials, therefore explaining the cycling stability of Ag_1-3xLa_x□_2xNbO₃.

At room temperature, AgNbO₃ crystallizes in an orthorhombic perovskite structure (space group Pbcm) with distorted [BO₆] octahedra, similar to most perovskites (Sciau et al., 2004). Historically, AgNbO₃ has been studied in material science as antiferroelectric material for energy storage applications. Substitutions in this lead-free material have been widely examined to stabilize the antiferroelectric phase and increase the power density. Various studies have explored the substitution of Ag⁺ in AgNbO₃ by Li⁺ (Farid et al., 2019), Na⁺ (Takeda et al., 2003), and multivalent cations (Li et al., 2009; Juang et al., 2016; Gao et al., 2020) (Sr²⁺, La³⁺, Sm³⁺, etc.) to create A-site vacancies. All the substitutions by multivalent cations show an increase in the symmetry of the material until the presence of impurities at a high degree of substitution (Juang et al., 2016; Gao et al., 2019; Gao et al., 2020). In order to investigate the possible Li-ion insertion into A site vacancies of this perovskite, we synthesized La-doped AgNbO₃ with different substitution ratios from x = 0 to x = 0.20. In a perovskite lattice, the coordination number of the A-site is 12, and 6 for the B-site. The ionic radius of silver is 1.54 Å, whereas that of niobium is 0.64 Å (Shannon and Prewitt 1969; Shannon 1976). Since the ionic radius of La for a coordination number of 12 is 1.36 Å, the silver-to-lanthanum substitution should be effective in the A site of the perovskite Ag_1-3xLa_x□_2xNbO₃, where □ depicts a cation, vacancy in the A site of the perovskite, following reaction Eq. 1: L a 2 O 3 →   2 L a A g ⋅ ⋅ + 3 O o x + 4 V A g ′ (1)

Here, we will name the samples Ag_0.85La_0.05□_0.10NbO₃, Ag_0.70La_0.10□_0.20NbO₃, Ag_0.55La_0.15□_0.30NbO₃ and Ag_0.40La_0.20□_0.40NbO_3, respectively for x = 0.05, x = 0.10, x = 0.15 and x = 0.20. Figure 1 represents the crystallographic arrangement of Ag_0.70La_0.10□_0.20NbO₃ which shows both distortion of [NbO₆] octahedra and the partial substitution of Ag by La in A-site with the corresponding vacancies.

The synthesis of Ag_1-3xLa_x□_2xNbO₃ by a ceramic route was difficult to achieve. One reason is the reduction of Ag₂O to metallic Ag for a temperature higher than 150°C (Chun et al., 2009; Shin et al., 2013). A second reason is the thermodynamic competition between AgNbO₃ and Ag₂Nb₄O₁₁ does not allow one to easily obtain a phase pure material, even by adding an excess of Ag₂O (Valant et al., 2007). For this reason, the different stoichiometries of AgNbO₃ were synthesized by a sol-gel (SG) process which has been demonstrated to be effective for the synthesis of perovskite materials (Huang et al., 1996; Y.; Li et al., 2010; Huízar-Félix et al., 2012). Moreover, the heat treatment at lower temperature (650°C) of the amorphous AgNbO₃ particles will avoid the reduction of silver. Figure 2A,B shows both AgNbO₃ and Ag_0.70La_0.10□_0.20NbO₃ agglomerated nanoparticles which are typical morphologies of sol-gel synthesis. AgNbO₃ particles are larger than those of Ag_0.70La_0.10□_0.20NbO₃, a feature confirmed by nitrogen adsorption measurements (BET method; Supplementary Table S1). These measurements revealed a specific surface area of 3 m² g⁻¹ for AgNbO₃ and 13 m² g⁻¹ for Ag_0.70La_0.10□_0.20NbO₃. In addition, as shown in Supplementary Table S2, EDX cationic ratio between Ag, La and Nb for every Ag_1-3xLa_x□_2xNbO₃ confirms the presence of vacancies in all the powders of La-substituted AgNbO₃.

XRD was used to better understand the structural differences among all Ag_1-3xLa_x□_2xNbO₃ powders. As shown in Figure 3A., AgNbO₃ has a perovskite structure with no secondary phase, although splitting of several peaks is observed. Ag_0.85La_0.05□_0.10NbO₃ (x = 0.05) and Ag_0.70La_0.10□_0.20NbO₃ (x = 0.10) also show the perovskite structure with no secondary phase visible in the diffractograms. However, starting from x = 0.15, impurity peaks appear, perhaps due to the lack of cations on the A-site, resulting in collapse of the structure. This is in agreement with previous reports on the A-site substitution in AgNbO₃ (Li et al., 2009; Juang et al., 2016). Depending on the size of the dopant cation, the perovskite structure can accept up to a certain amount of substitution, sometimes accompanied by the appearance of extra peaks which correspond to secondary phases. The decrease in peak intensity for Ag_0.55La_0.15□_0.30NbO₃ and Ag_0.40La_0.20□_0.40NbO₃, compared to other samples, is an indicator of the stress caused by the high proportion of La in the perovskite. For the substituted oxides, a decrease in peak splitting was observed with increasing lanthanum substitution (Figures 3B–C). Further, the disappearance of peak splitting by x = 0.15 suggests a higher degree of symmetry in the substituted materials due to lower [NbO₆] octahedra distortion.

To get greater insight into the structure of the synthesized materials, Le Bail structural refinements were performed for each composition. First, as indicated in Supplementary Figure S1, the refinement of AgNbO₃ confirms that the Pbcm space group well represents the structure of AgNbO₃. Cell parameters of AgNbO₃ were extracted from the refinement as shown in Supplementary Figure S2 and Supplementary Table S3. We found that the cell parameters and the cell volume increase with increasing substitution of La for Ag_1-3xLa_x□_2xNbO₃, which is in agreement with a previous report (Li et al., 2009). Moreover, this is consistent with a decrease of NbO₆ octahedra distortions previously described above. Refinements using a higher symmetry space group (Pm3m) have been performed to check that no phase transition appears with the substitution of silver by lanthanum. However, they did not lead to conclusive results. For this reason, we assume that the same space group (Pbcm) can be used to describe the crystallographic structure of all the Ag_1-3xLa_x□_2xNbO₃ powders.

To characterize the ability of this family of materials to reversibly store lithium ions, cyclic voltammetry at 0.1 mV s⁻¹ was first performed. Note that this sweep rate should be considered as a low rate for power applications (around 5.5 h of charge). As depicted in Figure 4A, AgNbO₃ does not show any electrochemical activity between 1.2 and 3.0 V vs. Li ⁺ /Li at the selected scan rate. A negligible specific capacity of 6 mAh.g⁻¹ is reported. However, for each substituted material, broad and reversible peaks appear between 2.2 and 1.2 V vs. Li ⁺ /Li, characteristic of lithium insertion into the host structure. Considering no insertion is visible for the unsubstituted oxide, we propose the following reaction Eq. 2 to describe the present electrochemical reaction. A g 1 − 3 x L a x □ 2 x Nb O 3 + y L i x + y e − →   L i y A g 1 − 3 x L a x □ 2 x Nb O 3 (2)

At 0.1 mV s⁻¹, the specific capacity increases with the number of vacancies. During the first cycle at 0.1 mV s⁻¹, specific capacities are close to the theoretical capacity considering the insertion of one lithium ion per vacancy: 16 mAh g⁻¹, 35 mAh g⁻¹, 54 mAh g⁻¹, 76 mAh g⁻¹ for Ag_0.85La_0.05□_0.10NbO₃, Ag_0.70La_0.10□_0.20NbO₃, Ag_0.55La_0.15□_0.30NbO₃, Ag_0.40La_0.20□_0.40NbO₃, respectively. Considering only the mass of active materials the gravimetric capacity is found to be equal to 38 mAh cm⁻³, 109 mAh cm⁻³, 206 mAh cm⁻³, 219 mAh cm⁻³ and 264 mAh cm⁻³ for AgNbO₃, Ag_0.85La_0.05□_0.10NbO₃, Ag_0.70La_0.10□_0.20NbO₃, Ag_0.55La_0.15□_0.30NbO₃, Ag_0.40La_0.20□_0.40NbO₃, respectively. If the entire volume of the practical electrode is now considered (porosity volume ∼70%), the gravimetric capacity is found to be equal to 12 mAh cm⁻³, 32 mAh cm⁻³, 62 mAh cm⁻³, 66 mAh cm⁻³ and 79 mAh cm⁻³ for AgNbO₃, Ag_0.85La_0.05□_0.10NbO₃, Ag_0.70La_0.10□_0.20NbO₃, Ag_0.55La_0.15□_0.30NbO₃, Ag_0.40La_0.20□_0.40NbO₃, respectively. Interestingly, no reduction of silver is observed in these materials indicating the strong interaction of silver cations within the perovskite structure, contrary to most silver-based materials reported in the literature, i.e., silver vanadates (West and Crespi 1995; Marschilok et al., 2010; Sauvage et al., 2010).

Cyclic voltammetry was used to characterize the performance of the lanthanum-substituted oxides at a higher sweep rate of 2 mV s⁻¹ (Figure 4B). This sweep rate corresponds to 15 min of charge/discharge, which is consistent with the current goal for the development of electric vehicles (Colclasure et al., 2019). As was observed at 0.1 mV s⁻¹, the CVs shown at 2 mV s⁻¹ present very broad and reversible redox peaks.

As expected, the capacity of AgNbO₃ remains very low compared to the substituted oxides. Its capacity is 5 mA h.g⁻¹. Ag_0.85La_0.05□_0.10NbO₃ is the least substituted material and the one with the lower capacity at both 0.1 mV s⁻¹ and 2 mV s⁻¹ (32 mAh g⁻¹). Interestingly, at 2 mV s⁻¹, Ag_0.40La_0.20□_0.40NbO₃ has a lower capacity than Ag_0.70La_0.10□_0.20NbO₃ and Ag_0.55La_0.15□_0.30NbO₃, even though Ag_0.40La_0.20□_0.40NbO₃ shows the higher theoretical capacity most likely attributed to the higher amount of vacancies per unit cell. Such a decrease in capacity is explained by the lower degree of crystallinity which does not allow adequate diffusion of lithium ions in this substituted oxide. Moreover, the defects detected by the presence of peaks that did not belong to the orthorhombic lattice provides another possible explanation to the lower capacity of Ag_0.40La_0.20□_0.40NbO₃ at 2 mV s⁻¹. In addition, these CVs show the polarization between the oxidation and reduction peaks to be very small for each compound, even for CV at 2 mV s⁻¹. This result suggests that the intercalation process is not limited severely by low ionic diffusion, a consideration which will be discussed later.

CVs at different scan rates were performed to assess the capacity retention as a function of increasing scan rate. Scan rates from 0.1 mV s⁻¹ to 50 mV s⁻¹ were chosen, corresponding to 5.5h and 40s of charge, respectively. Figures 4C,D shows the specific capacity and relative capacity of every material at the mentioned scan rates. Ag_0.40La_0.20□_0.40NbO₃, the material with the highest number of vacancies, provides the highest capacity at a low sweep rate (Figure 4C). However, when the scan rate is increased, a rapid decrease in capacity is observed starting at 0.2 mV s⁻¹. As shown in Figure 4D, at 50 mV s⁻¹, Ag_0.40La_0.20□_0.40NbO₃ retains only 34% of the initial capacity measured at 0.1 mV s⁻¹. Similarly, Ag_0.55La_0.15□_0.30NbO₃ demonstrates a relatively high capacity at low scan rates but a rapid drop at higher rates. In contrast, while less substituted materials such as Ag_0.85La_0.05□_0.10NbO₃ and Ag_0.70La_0.10□_0.20NbO₃ exhibit lower capacities at 0.1 mV s⁻¹, they demonstrate better capacity retention. At 50 mV s⁻¹, Ag_0.85La_0.05□_0.10NbO₃, and Ag_0.70La_0.10□_0.20NbO₃ exhibit capacity retentions of 53 and 64%, respectively, for a 500 times faster rate of charge/discharge compared to their capacities measured at 0.1 mV s⁻¹. All the specific capacities are summarized in Supplementary Table S4.

In order to get a better understanding of the charge storage mechanism, b-values from the power law relationship in Eq. 3 have been calculated, i= a . v b (3) where i is the intensity of reduction/oxidation peak, v the corresponding scan rate, a and b are constant values. If b is close to 0.5, charge storage can be described as limited by ionic diffusion in the host material, as described in the Randles-Ševčík equation (Ševčík 1948; Randles 1948). When the value of b is equal or close to 1, charge storage occurs in agreement with non-diffusion limited mechanism, as in the case of pseudocapacitive materials (Choi et al., 2020). Regarding Ag_1-3xLa_x□_2xNbO₃ compounds, as shown in Figure 5A,B, a linear relationship between specific current and scan rate (0.1 mV s⁻¹ and 50 mV s⁻¹) is observed, the b-values of which are listed in Table 1. The values of b vary between 0.87 and 0.95, which seems to indicate weak diffusion limitation during charge/discharge processes. Additionally, the smaller b values calculated for Ag_0.55La_0.15□_0.30NbO₃ and Ag_0.40La_0.20□_0.40NbO₃ support the difference in capacity retention observed previously, as compared to Ag_0.85La_0.05□_0.10NbO₃, and Ag_0.70La_0.10□_0.20NbO₃ .

As displayed in Figure 6, the behavior of the different oxides during lithium insertion was investigated by in situ XRD, to better understand the filling of the vacant A-sites of the Ag_1-3xLa_x□_2xNbO₃ perovskite. Ag_0.70La_0.10□_0.20NbO₃ was used because no structural defects were observed in this phase compared to Ag_0.55La_0.15□_0.30NbO₃ and Ag_0.40La_0.20□_0.40NbO₃. First, no new peak appears (and/or disappears) on the XRD patterns during charge and discharge. This behavior indicates that a solid solution type mechanism predominates upon charge storage, as previously reported for intercalation materials dedicated to power application with a b value close to 1 (Come et al., 2014). Moreover, it is observed that the peaks initially present do not move during lithiation. The three diffraction peaks observed in the studied 2θ range present the three crystallographic directions, implying that the perovskite structure does not modify its lattice parameters during lithium cation insertion in the vacant A sites. Thus, this material can be considered as a zero strain oxide during lithium insertion, similar to the well-known Li₄Ti₅O₁₂ (Thackeray and Amine 2021).

In another series of measurements, self-discharge and NMR experiments carried out on cycled Ag_0.70La_0.10□_0.20NbO₃ electrodes provide further evidence of Li⁺ insertion in the material. For these experiments, electrodes were cycled to 1.8 and 1.2 V vs. Li ⁺ /Li at 0.1 mV s⁻¹ and subsequently polarized for 10 h to reach a thermodynamic equilibrium. As shown in Figure 7A,B, both electrodes (cycled at 1.2 and 1.8 V vs. Li ⁺ /Li) present relatively low self-discharge. The electrode polarized at 1.8V shows a potential of 1.82 V vs. Li ⁺ /Li after almost 80 h of OCV. The electrode polarized at 1.2 V shows a potential of 1.59 V vs. Li ⁺ /Li in the same condition. These low self-discharge levels at 1.2 and 1.8 V are a strong indication that lithium cations are inserted in a stable way into the vacancies of the material. Ex situ ⁷Li MAS NMR measurements have been performed after these self-discharge experiments. ⁷Li MAS NMR spectra were also recorded for the pristine and the fully discharged electrode (3.0V) in addition to the charged samples at 1.2 V vs. Li ⁺ /Li, at a similar scan rate of 0.1 mV s⁻¹. As expected, the pristine electrode has a negligible quantity of Li detected. For the partially and fully charged samples, a peak is observed at around −1 ppm. The integrated intensity of the NMR peak, which is intrinsically proportional to the detected amount of lithium in the materials, is also found to be proportional to the specific capacity of the materials at 1.8 and 1.2 V (Figure 7B,C). The spectrum of the fully discharged electrode at 3.0V displays a small amount of lithium assigned to the partial irreversibility of the reaction during the first cycle and is also found to be in agreement with the first cycle of the CV. Due to the amount of lithium inserted in the materials, the signal to noise ratio remains relatively low even for a large number of scans. For this reason, spectra with a narrower linewidth and thus better resolution, were not possible. Thus, it is difficult to conclude as to the number of lithium sites, even at the fully charged state, as only one apparent resonance is observed in ⁷Li MAS NMR. Taken together, the combination of ex situ NMR and self-discharge provide convincing evidence that lithium ions are inserted in this perovskite material, most probably in the vacant A site.

The cycling characteristics for Ag_1-3xLa_x□_2xNbO₃ were determined as cycle life is a key property for the development of batteries based on high rate electrode materials. As observed in Figure 8, all the materials exhibit exceptional stability upon cycling. After 800 cycles at 2 mV s⁻¹, Ag_0.85La_0.05□_0.10NbO₃, Ag_0.70La_0.10□_0.20NbO₃, Ag_0.55La_0.15□_0.30NbO₃ and Ag_0.40La_0.20□_0.40NbO₃ show a capacity retention of 95, 95, 97 and 88%, respectively. The in situ XRD experiments and b-value calculation, which show zero-strain materials and non-diffusion limited charge storage mechanisms, respectively, are consistent with the outstanding cyclability observed for these materials.

In contrast to the classical design of insertion materials, our approach was to synthesize an oxide with vacancies in order to create potential sites for cation insertion in a perovskite. As proven by ex situ MAS NMR, lithium ions can be reversibly stored in the A site of Ag_0.70La_0.10□_0.20NbO₃ and by extension to every Ag_1-3xLa_x□_2xNbO₃ composition. To further demonstrate this point, Supplementary Figure S3A–D presents the difference between lithium and sodium insertion by cyclic voltammetry at different scan rates. While Li⁺ insertion leads to characteristic redox peaks, in the case of Na⁺, no redox reaction can be observed. It is likely that sodium ions cannot be inserted into the structure of Ag_0.70La_0.10□_0.20NbO₃ and therefore cannot generate redox reactions due to the larger ionic radius This is also the case for the pseudocapacitive T-Nb₂O_5, which shows highly reversible Li-ion intercalation but negligible Na-ion insertion (Kim et al., 2012). Moreover, kinetic analysis provides compelling evidence that a not diffusion-limited insertion occurs during the charge and discharge processes.

This work presents the partial substitution of AgNbO₃ perovskite by lanthanum in the A-site, resulting in the creation of cationic vacancies. We report on a gradual increase of the capacity at low scan rates as a function of the increase of the substitution. In addition, the different oxides present excellent behavior at high scan rates. At 50 mV.s⁻¹Ag_0.70La_0.10□_0.20NbO₃ retains 64% of its capacity measured at 0.1 mV s⁻¹. Moreover, the combination of ex situ MAS NMR and in situ XRD experiments confirm that the insertion of lithium ions in the A site of the perovskite is responsible for the observed capacity and does not cause any structural change of the material. Together, these observations help to explain the good cycling behavior observed. In addition to structural insight, kinetic analysis indicates a non-diffusion limited ion insertion process for these oxides.

Nevertheless, some questions still remain. A better understanding of the solvation/desolvation occurring at the surface of particles during alkali ion insertion is critical towards understanding the small limitation of diffusion in charge storage process (Wang et al., 2019). Further experiments such as EQCM or operando atomic force microscopy can provide more insight in the charge storage mechanism. From another point of view, the promising observations reported here should be explored with other vacancy-designed materials and with different alkali cations. Moreover, these oxides present a charge storage mechanism which are between double layer capacitor and battery type electrodes (as observed for MnO₂ (Toupin et al., 2004; Brousse et al., 2006), T-Nb₂O₅ (Augustyn et al., 2013; Griffith et al., 2016) or MXenes (Naguib et al., 2012; Anasori et al., 2017) and represent an interesting model material for fundamental understanding in general.

In conclusion, the new approach reported in this work presents a different type of material design which widens the opportunities for creating the next generation of negative electrodes for high power Li- ion batteries. Although the oxide presented here contains silver, we believe this study can open the way toward the search of more practical materials for high-power battery electrodes.