WorksPhysics and Astronomy
Phase stability and ionic transport in post-spinel CaVO cathode
Abstract
Calcium-ion batteries (CBs) represent an alternative to lithium-ion technology but their advancement is limited by the lack of high-performance intercalation cathodes. Identified via computational screening, post-spinel CaVO has emerged as a promising candidate, though its practical application is hindered by limited electrochemical capacity. Hence, we investigate the thermodynamic and ionic transport characteristics of CaVO () in this work, by integrating the cluster expansion formalism with Monte Carlo simulations and density functional theory based calculations. We construct the temperature-composition phase diagram of CaVO revealing several stable phases ( through ) that can appear during electrochemical operations at different voltages. Importantly, we observe the formation of the phase at across a 370-590~K temperature window via invariant reactions, which agrees with observations in the experimental voltage profiles. Further, migration barrier calculations confirm that Ca mobility is severely impeded within the () and () phases. With the strong Ca-vacancy ordering contributing to the high barrier in and the persistent two-phase region stretching across the () and the phases, we expect the accessible electrochemical capacity in the CaVO system to be kinetically limited to at most half the theoretical capacity at 298~K, in agreement with experiments. Strategies including cation doping and particle size reduction can be considered to flatten the potential energy landscape of and improve Ca mobility. Our computational findings highlight the interplay between stability and transport and provide design strategies that can enable the practical use of CaVO as a CB cathode.
Introduction
Lithium-ion batteries have enabled the portable electronics revolution and are now central to electric mobility and grid-scale storage, but the technology is approaching intrinsic limitations in both energy density and resource availability. [1, 2, 3, 4, 5] To meet the growing global demand for sustainable and cost-effective energy storage, significant attention is turning toward beyond-lithium chemistries. In this context, multivalent batteries, based on Mg, Ca, or Zn, offer an appealing alternative by combining earth abundance with the prospect of high volumetric energy densities from metal anodes. [6, 7, 8, 9, 10, 11] Among the multivalent chemistries, calcium is particularly attractive as it exhibits a reduction potential close to that of lithium (-2.87 V vs SHE), high theoretical gravimetric (2205 mAh/g) and volumetric (2073 Ah/L) capacities, and is the fifth most abundant element in the Earth’s crust. [9, 12] Recent breakthroughs in electrolyte design, which now permit reversible Ca plating and stripping, have transformed the feasibility of calcium batteries (CBs) from speculation to tangible opportunity. [13, 14, 15, 16]
Although Ca-based chemistry offers compelling advantage, the practical realization of CBs remains constrained by several unresolved challenges. A foremost difficulty lies in the development of electrolytes that are simultaneously stable against the highly reducing Ca metal anode and the oxidizing potentials required for high-voltage cathodes. [11, 17] Equally critical is the identification of positive electrode (cathode) materials that are not only thermodynamically stable but also capable of reversible Ca intercalation at relevant voltages and capacities. [18, 19] Moreover, the intrinsically poor transport in most solid frameworks leads to prohibitively high migration barriers (), limiting rate performance and cycle life. Together, these challenges highlight the complexity of advancing CB technology from conceptual promise to practical deployment.
Significant effort has been devoted to identifying viable Ca-intercalation cathodes, guided by both computational and experimental studies. [20, 21, 22, 23, 24, 25, 26] For instance, a high-throughput screening of ternary transition-metal oxides and chalcogenides by Lu et al. identified post-spinel and layered as particularly promising candidates. [22] Building on this theoretical prediction, Black et al. provided experimental validation of the redox electrochemical activity in post-spinel , while complementary density functional theory (DFT [27, 28]) calculations by the authors quantified the within the lattice. [25]
Nevertheless, Black et al.’s work indicated only partial calcium extraction, with capacities restricted to mAh at 298 K and mAh at 323 K, and full Ca deintercalation not achieved even at 323 K. [25] Thus, experimental data in indicates that poor mobility, alongside possible side reactions with the electrolyte, may play a role in restricting the electrochemical capacity, as indicated by the strong temperature-dependence in the extent of Ca that is extracted.
Additionally, the evolution of the voltage profile during cycling of , [25] as characterized by the emergence of distinct plateaus and a decrease in polarization, suggests a possible dynamic restructuring of the host lattice or the onset of phase separation. Although operandо X-ray diffraction (XRD) has confirmed the extraction of mol of Ca from and the formation of oxidized phases, the precise interplay between the thermodynamic stability of the oxidized phases and any kinetic barriers associated with them is not fully understood. Thus, it is an open question whether the current electrochemical inability to reach the full theoretical capacity in is a thermodynamic impossibility or a kinetic trap. Understanding these factors is therefore critical to fully evaluate the potential of as a practical CB cathode, motivating the systematic investigation of our work.
Here, we present a comprehensive computational investigation of Ca intercalation in () that integrates DFT calculations, the cluster expansion (CE [29]) formalism, grand canonical Monte Carlo (GCMC) simulations, and nudged elastic band (NEB [30, 31]) computations to elucidate both the thermodynamic phase stability and the transport kinetics across the full Ca compositional range. We construct the 0 K convex hull and a finite temperature phase diagram to identify stable single phases and possible two phase equilibria at different temperatures, and compute the corresponding voltage profile(s). In addition, calculations are performed for all thermodynamically accessible single phase regions to pinpoint compositions that can pose kinetic limitations associated with Ca diffusion. Our combined thermodynamic and kinetic analysis reveals that Ca extraction is limited by the interplay between persistent phase separation during (de)intercalation and composition-dependent Ca mobility variations. Specifically, the () and () two-phase region extends over a broad temperature range that encompasses electrochemical operating conditions with Ca mobility severely impeded in the phase due to the strong Ca-vacancy ordering, which acts as the kinetic bottleneck for full reversible Ca (de)intercalation from CaVO. By establishing explicit links between phase behavior and ionic transport, our study provides a quantitative framework for rationalizing capacity limitations in Ca based intercalation hosts, such as CaVO, and for guiding the design of improved Ca-cathodes.
Methods
First principles calculations
All spin-polarized geometry relaxation calculations were performed using DFT as implemented in the Vienna ab initio simulation package [32, 33] with the Hubbard -corrected strongly constrained and appropriately normed (i.e., SCAN+ [34, 35, 36, 37, 38, 39, 40]) exchange-correlation functional. A correction of 1.0 eV, as obtained from our previous work, was applied to the 3 electrons of V to reduce the spurious self interactions. [37, 38] We used the frozen-core projector augmented wave (PAW [41, 42]) potentials (see Table S1 of the supporting information –SI, for the specific potentials used) and expanded the plane wave basis up to a kinetic energy cutoff of 520 eV. We sampled the irreducible Brillouin zone with -centered Monkhorst-Pack [43] -points grids of a minimum density of 48 -points per Å. Geometric relaxations of cell volume, shape, and ionic positions were carried out without symmetry constraints, with convergence criteria of eV and eV/Å for total energy and atomic forces, respectively. We obtained the starting structure of CaVO from the inorganic crystal structure database. [44]
Cluster expansion formalism
To obtain the temperature–composition phase diagram, we performed GCMC simulations using a CE Hamiltonian. In the CE formalism, the total energy of a structure is expressed as a function of the site occupation variables, , written as a summation over effective cluster interactions (ECIs) of point, pair, triplet, and higher-order clusters, as given in Eq. (1):
where is the configuration vector with if site is occupied by Ca and if it is vacant. The sum runs over symmetrically distinct clusters (points, pairs, triplets, ), is the multiplicity of cluster per primitive cell, and is the corresponding ECI. is the cluster function averaged over all symmetrically equivalent clusters of type in a given configuration. represents a constant ECI term, also referred to as the ECI of the empty cluster. The symmetrically unique ECIs were systematically fitted to a set of training structures and their corresponding DFT-calculated formation energies. Note that the DFT formation energy of a given configuration was computed from the DFT total energy of that configuration referenced to the DFT calculated energies of end-member compositions, namely, VO () and CaVO ().
To construct the training set for the CE, we enumerated unique Ca-vacancy configurations at different in CaVO using , , and supercells of the conventional cell that contains four CaVO formula units. We used the pymatgen package [45] to enumerate the symmetrically distinct Ca-vacancy configurations at a given and ranked them based on their electrostatic energies via the Ewald energy summation technique. [46] We considered all enumerated structures across 15 distinct Ca-compositions, namely, , , , , , , , , , , , , , , , besides the end-member compositions ( and ), resulting in a total dataset of 262 DFT-calculated formation energies across all considered. For constructing the CE, we utilized the clusters approach to statistical mechanics (CASM [47, 48, 49]) package, with ECIs determined using least absolute shrinkage and selection operator regression to obtain a sparse and accurate model. We evaluated the accuracy of the CE fit using the root-mean-square error (RMSE) score across all configurations, while the predictive ability of the fit was assessed using the weighted cross-validation (WCV) score.
Monte Carlo simulations
All GCMC simulations were performed using the CASM package and our CE Hamiltonian, to obtain the temperature-composition phase diagram of . The GCMC simulations employed a supercell of the conventional cell, containing 6,912 Ca/vacancy sites. At each Ca chemical potential () and temperature (), GCMC simulations were performed using the Metropolis algorithm [50] with a minimum of 10,000 and a maximum of 100,000 Monte Carlo passes. The equilibration/sampling split among the Monte Carlo passes was determined automatically using the statistical equilibration criterion of van de Walle and Asta, [51] which identifies and discards the initial equilibration portion of the Monte Carlo trajectory before ensemble averaging. Subsequently, the sampling of Monte Carlo passes continued until the requested statistical precision of 0.25 meV/f.u. on the calculated energy at a 95% confidence level was achieved, or until the maximum number of passes was reached. We scanned from K to 1005 K in steps of K. At every , we scanned the forward (i.e., increasing ) and backward (decreasing ) across four concentration regions, spanning a range of eV/atom, with a step size of eV/atom. The four concentration ranges covered during every scan are , , , and , where values of -7.0, -3.7, 0.0, 3.0, and 7.0 eV/atom correspond to the five ground state structures at , , , , and , respectively, on the 0 K convex hull.
Additionally, we used canonical Monte Carlo (CMC) based simulated annealing to verify the authenticity of the ground state configurations predicted by the CE and to ensure that all ground states at 0 K have been captured accurately. Specifically, we performed CMC simulations consisting of a minimum of 100 passes per from 2005 K to 5 K in steps of 100 K at different and under varying supercell sizes for each (for example, from to at ). During the CMC simulations, if a new configuration was predicted that was not present already in the training data, we calculated its energy using DFT and subsequently added the configuration to the training set. We repeated this process iteratively until no further ground states emerged. All CMC simulations were also performed using the CASM package.
We determined the Ca intercalation voltage (, versus ) at all directly from the obtained from GCMC simulations, using Equation (2):
where is the valence of . is a constant shift applied to align the average calculated voltage across the entire range at any with the reference average intercalation voltage of 2.73 V obtained from 0 K DFT calculations. To construct the equilibrium voltage-composition (-) profiles as well as the temperature-composition phase diagram, the was determined at each by resolving the numerical hysteresis within GCMC simulations through the selection of the phase that minimizes the grand canonical potential (i.e., via the free energy integration technique, see Section S3 of SI for details. [52, 53, 54, 55])
Migration barriers
We performed DFT-based NEB calculations to evaluate the Ca in CaVO at select stable single phase compositions identified from the calculated phase diagram. Due to the trade-off between computational cost and accuracy, we used the generalized gradient approximation (GGA [56]) exchange–correlation functional instead of SCAN/SCAN+ for all NEB calculations, as GGA is known to obtain qualitative trends in precisely. [57] For the endpoint configurations in an NEB calculation, we converged the structures till the atomic forces fell below eV/Å. Subsequently, we initialized the minimum energy path by linear interpolation of lattice vectors and atomic positions between the endpoint configurations, generating seven intermediate images with a spring constant of 5 eV/Å applied between the images. We used Γ-centered Monkhorst-Pack -point meshes containing at least 32 subdivisions along each unit reciprocal lattice vector to sample the irreducible Brillouin zone of both the endpoint and the image configurations. We considered our NEB calculations converged when the perpendicular component of the band force fell below eV/Å. For compositions where DFT-based NEB calculations were not feasible, we used machine learning (ML) models to get qualitative estimates (see Section S7 for methodological details).
Results
Structure of CaVO
Post-spinel CaVO adopts the orthorhombic CaFeO-type structure (space group ) comprising two crystallographically distinct V octahedral sites, namely, V1 and V2 (teal and purple octahedra in Figure 1a). [58] The distorted VO octahedra share both edges and corners, with Ca occupying eightfold-coordinated sites in the voids created. Ca transport within the lattice is expected to occur predominantly along the -axis tunnels. [22, 25] To construct the CE model, we use the conventional cell (Figure 1a) as the primitive lattice, as well as the larger (Figure 1b) and (Figure 1c) supercells to enumerate Ca–vacancy orderings. We use these supercell dimensions to capture longer-range interactions along the Ca-diffusion direction as well as to capture different possible values of that cannot be described within the conventional cell.

Figure 1. Crystal structure of CaVO. Teal and purple (in panel a) polyhedra denote the VO framework around V1 and V2 sites, respectively, while orange spheres denote the Ca sites. Dashed lines outline the simulation cell. (a) Conventional cell, (b) 1 × 1 × 2, and (c) 1 × 1 × 3 supercells.
Convex hull and cluster expansion
Figure 2a presents the 0 K convex hull of , computed using DFT (orange line) and the fitted CE Hamiltonian (teal line). The formation energies are referenced to the fully intercalated () and deintercalated () end members. Individual datapoints are shown as orange (DFT) and teal (CE) triangles, with larger symbols denoting structures that lie on the convex hull and are therefore thermodynamically stable ground states at 0 K. Notably, three intermediate compositions at , 0.5, and 0.67 appear as stable ground states on the 0 K convex hull, indicating the formation of distinct Ca-vacancy orderings across the Ca concentration range. Additionally, our CMC simulations indicated that no other ground state configurations appear at other near 0 K, verifying that we have a total of five ground state configurations in the system including the end members. The ground state structures at intermediate (0 K) are shown in Figure S1 of the SI.
The accuracy of the CE as a function of thermodynamic (in)stability is illustrated in Figure 2b, which correlates the prediction error (represented by blue circles) with the energy above the 0 K convex hull () for each configuration. A robust CE model is expected to exhibit enhanced precision for structures in close proximity to the convex hull (i.e., low values) to ensure reliable phase stability predictions. Consistent with this requirement, the current model demonstrates a marked concentration of low-error configurations as approaches zero. Specifically, the majority of structures with meV/f.u. of the hull fall within an error margin of meV/f.u., as indicated by the inner dashed box, while nearly all configurations relevant to phase stability under electrochemical operating conditions (i.e., meV/f.u.) are captured within a meV/f.u. threshold (outer dashed box).
Figure 2. (a) Convex hull of showing formation energies (meV/f.u.) vs. Ca content, . Orange triangles denote DFT-computed energies and teal triangles the CE-computed values. Solid lines mark the K ground states, as estimated by DFT (orange) and CE (teal). (b) Quantifying the extent of CE error made across different configurations based on their DFT-computed energy above the convex hull ().
Overall, our CE model achieves an RMSE of 13.4 meV/f.u., accurately reproduces all DFT-predicted ground states, and exhibits good transferability with a leave-one-out WCV of 15.7 meV/f.u., confirming the model’s reliability in predicting energies of configurations not included in its fit. While the overall CE fit is excellent, a small subset of data points in the –200 meV/f.u. range do exhibit relatively large deviations, as indicated by the blue circles lying outside the dashed boxes in Figure 2b. These outliers likely arise from complex local environments or significant structural relaxations during DFT that are inherently challenging to capture within a truncated CE framework.
Figure 3 illustrates the 29 ECIs that constitute our CE model, as obtained by our fit against the training set of 262 DFT configurations. The details of all clusters that exhibit a non-zero ECI in our fit, including the empty and point terms, are summarized in Table S2 of the SI. We normalize the ECIs by multiplicity in Figure 3 and plot the values against the corresponding size of the cluster to highlight the physical hierarchy of the CE. Note that clusters are categorized by their maximum distance, i.e., the size of a triplet is the distance between its farthest sites. The ECIs for pairs, triplets, and quadruplets are indicated by orange circles, purple triangles, and teal squares, respectively, with the empty and point ECI values listed in Table S2. For pair interactions, positive ECIs indicate repulsive interactions (i.e., Ca-vacancy pairs are favored) while negative ECIs signify attractive interactions (i.e., Ca-Ca and vacancy-vacancy pairs are favored).
Figure 3. Plot of ECIs, normalized by multiplicity, as a function of cluster size for CaVO. Orange circles, purple triangles, and teal squares denote pair, triplet, and quadruplet interactions, respectively. Note the break in the -axis done to accommodate the dominant pair term within the plot.
Our CE exhibits a well-converged and clear physical hierarchy, characterized by the rapid decay of ECIs with both increasing cluster size and increasing cluster complexity. For example, the most dominant Ca-Ca interaction in the system is the nearest repulsive pair interaction, at Å with a value of meV, indicating the strong and short-ranged electrostatic interaction that governs Ca ordering in the structure. In other words, Ca strongly prefers vacancies as nearest neighbors within the lattice, which is reflected in the diamond-cubic-like arrangement of Ca atoms and vacancies within the Ca sub-lattice at the ground state configuration (see Figures S1b and Figure 2a). As the size of the pair clusters increases, the ECI magnitudes generally decrease and fluctuate around zero, with values in the range of meV over the cluster-size range of –10 Å. In terms of cluster complexity, the dominant triplet (at Å) and quadruplet (also at Å) interactions remain significantly smaller than the dominant pair interaction, with ECIs of approximately 28 and meV, respectively. Overall, the ECIs indicate the dominant role played by pair interactions (with lower complexity) compared to triplet and quadruplet (higher complexity) interactions.
Temperature-composition phase diagram
Elucidating the temperature-dependent phase behavior of the system, Figure 4 illustrates the calculated temperature-composition phase diagram, as derived from GCMC simulations utilizing large supercells. The associated occupancy of Ca atoms across the four Ca-sites in the lattice, as a function of and is compiled in Figure S2. The computed phase diagram in Figure 4 reveals a complex thermodynamic landscape comprising six distinct single-phase regions, denoted as through . At temperatures below 298 K, five distinct single-phase regions exist, arising from the corresponding 0 K ground state configurations (Figure 2a), namely, , , , , and , corresponding to , 0.25, 0.5, 0.67, and 1, respectively. Notably, the -phase (situated at ) is unique in being stabilized at higher temperatures (via configurational entropic contributions) and does not have a corresponding ground state configuration at 0 K. The phases emerges only at an elevated K and persists until 590 K, with the associated invariant reactions at 370 K and 590 K being of the eutectic- and peritectic-types, respectively. A representative snapshot of the structure of the phase at 370 K is shown in Figure S3 with the corresponding crystallographic information file provided in our GitHub repository.
Figure 4. Computed temperature-composition phase diagram for the system. The colored regions represent the co-existence of two-phases that are separated by the single-phase regions ( through ).
The two-phase regions in Figure 4 are shown as colored domes, with gray lines marking their boundaries. Specifically, the two-phase region exists between (teal dome), with a narrow solid-solution (white region) beginning at K near , signifying negligible solubility of Ca in the structure at 298 K. The two-phase region remains until K at , with the system transitioning into a disordered solid solution phase above 440 K. Thus, Ca’s solubility in should increase dramatically, if electrochemical discharge into the lattice is done under elevated temperatures (e.g., 323 K) compared to room temperature. Similarly, the region exists between (yellow dome), with the two-phase region transitioning into a disordered solid solution at K. Notably, the region, spanning the composition range (light purple dome), remains stable at temperatures exceeding 1000 K (not shown in Figure 4), indicating that this two-phase region will be accessed under all electrochemical conditions irrespective of Ca intercalation into or Ca removal from . Also, the phase exhibits fairly negligible non-stoichiometry even at elevated temperatures (up to K), further contributing to the persistence of the two-phase region.
The composition range between displays a more complex set of transitions compared to . At low temperatures, the two-phase region (blue) exists up to 370 K, with negligible solubility of vacancies in the lattice (i.e., phase). Above 370 K, the emergence of the phase via a eutectic-type reaction causes the formation of two distinct two-phase regions, namely, (purple) and (pink), with a corresponding increase in vacancy solubility in the phase. As increases further, the phase disappears via a peritectic-type reaction at 590 K, resulting in the formation of a single two-phase region, (green dome), which persists till K. The extent of off-stoichiometry in both the and phases reaches an inflection point that coincides with the peritectic-type reaction at 590 K and continues to increase rapidly with further increase in . Eventually, the composition range exhibits a single disordered solid solution at K. Given that both the solubility of Ca in ( phase) and of vacancy in ( phase) are negligible at 298 K and exhibit an increase at higher temperatures, the cathode may exhibit better electrochemical cycling (i.e., with lower polarization) at elevated temperatures, in qualitative agreement with experimental observations so far. [25]
The occupancy of Ca atoms among the four distinct Ca sites in the lattice (Figure 1a) shows strong non-monotonic behavior and temperature dependence as well (see Figure S2). However, the occupancies of all Ca sites in the lattice converge to a value of 0.5 (or half occupancy) as approaches 0.5 ( phase) at all temperatures, indicating the strong nature of the Ca-vacancy ordering in the phase (Figure S1b) that is driven by strong electrostatic attraction between Ca atoms and neighboring vacancies (Figure 3). In comparison to , Ca occupancies exhibit fairly linear trends across all four Ca sites in the composition range (Figure S2), across all temperatures, with the occupancy of all four sites being equal in the single phase at 475 K.
Voltage profiles
Figure 5 presents the GCMC-computed voltage profiles for Ca intercalation in , color-coded by temperature from dark purple (300 K) to bright yellow (1000 K), and are compared against the 0 K DFT-computed voltage profile (solid gray). The inset in Figure 5 focuses on the narrow composition range of $0.832 < x < 0.836$, which is adjacent to the phase. Notably, the 0 K DFT and the 300 K GCMC voltage profiles exhibit three distinct voltage steps at (), (), and (), corresponding to the corresponding stable single-phase regions (Figure 4). Consequently, four well-defined voltage plateaus are observed, representing the two-phase coexistence regions: ( V vs. Ca/Ca), ($\sim 4.59 V), ( V), and ( V). Thus, the and/or the two-phase regions need to be accessed electrochemically for to exhibit high (de)intercalation voltage (and hence high energy density) as a cathode. While a one-one comparison with the experimental voltage profile is not possible due to significant polarisation across the charge-discharge profiles and differences in the voltage plateaus (or steps) across the charging and discharging steps, [25] we do note that our overall calculated average voltage across the entire Ca composition range (2.73 V vs. Ca/Ca) is lower than the experimental average voltage of 3.55 V across a 0.6 mol Ca extraction from . However, side reactions and overpotentials during Ca extraction from may also be contributing to the observed voltages experimentally, causing the disagreement between computations and experiments.
Figure 5. Computed voltage profiles for Ca intercalation in as a function of . The finite-temperature voltage profiles (dashed lines) are derived from GCMC simulations and are color-coded by temperature. The solid gray line represents the 0 K DFT-computed voltage profile. The inset highlights the voltage step in the region, corresponding to the -phase.
Notably, at 400 K and 500 K, the computed voltage profile exhibits an additional (small) step corresponding to the emergence of the () at 370 K, as highlighted in the inset of Figure 5. Note that this voltage step occurs at a composition that is marginally offset from the composition (by 0.001-0.002), possibly due to sampling issues in our GCMC and the associated free energy integration. Nevertheless, our computational finding of the presence of the phase is in agreement with the experimental electrochemical profiles, which show subtle slope changes and capacity plateaus within the to 0.85 range in , [25] suggesting a possible (metastable) formation of the phase below 370 K. As the temperature increases to 1000 K (bright yellow), the voltage profile evolves from sharp, discrete steps into a smoother series of plateaus, signifying the existing of the disordered single phase regions at and (Figure 4). Notably, the voltage plateau corresponding to the region (0.5 0.67) remains even at 1000 K, consistent with our calculated phase diagram, and in qualitative agreement with the experimental charging curves at 323 K. [25]
Ca migration barriers
Given the dominant two phase region across the 0.5 0.67 range (Figure 4), any limitation on Ca-mobility within the or the phases can adversely affect the amount of Ca that can be reversibly (de)intercalated into the system. Hence, we calculate the for Ca-motion using DFT-based NEB in different ground state ordered structures of and present the data for , 0.5, 0.67, and 1 in Figure 6. The endpoint values ( and ) are based on the GGA functional and obtained from the previous work of Black et al., [25] while we have computed the for and 0.67 in this work and have included the associated minimum energy pathways in Figure S4. Due to the supercell size requirements for calculating at () and (), we did not perform DFT-based NEB calculations for these phases and instead used foundational machine learned interatomic potentials (MLIPs [59, 60], without any fine-tuning) and a transfer learned (TL) model [61] to obtain estimates for these phases, similar to our previous work [26] (see Figure S5). MLIP and TL estimates of for the , and 1 ground states are also provided in Figure S5 for comparison.
Figure 6. DFT-based NEB calculated Ca in CaVO at different . Results for the endpoints ( and ) are obtained Ref. 25, while the remaining are calculated in this work. Figure S5 compiles the estimates for the different ground states in CaVO obtained using machine learned models.
Voltage profiles
The exhibit a strong, non-monotonic dependence on the calcium content, with the peak (minimum) in at (). In the high-Ca limit (i.e., , phase), the is low ( 600 meV), suggesting favorable Ca motion under typical electrochemical conditions, as long as a sufficient number of diffusion carriers (i.e., vacancies) are present in the phase. Indeed, a threshold of 525-650 meV has been used before to identify cathode materials with reasonable ionic mobility for various battery frameworks. [19] At
( phase), the increases marginally to meV, indicating that the formation of the phase during electrochemical Ca (dis)charge in CaVO should not act as a mobility bottleneck.
In contrast to the and phases, in the high-vacancy limit (, phase), the is meV, suggesting inherently poor Ca motion, despite the presence of a large number of vacancies. Unfortunately, the reaches a maximum value of meV at ( phase). Such high values, exceeding 1 eV in the and phases, indicate that Ca mobility is severely hindered in the low-to-intermediate concentration regimes (), highlighting the kinetic bottleneck(s) associated with the reversible electrochemical cycling of Ca in the CaVO system. Particularly, the mobility limitation associated with the phase, which will co-exist alongside the phase as drops below 0.67 during Ca extraction, may be the predominant reason to the limited capacity observed during electrochemical extraction of Ca from the CaVO structure. [25]
Consequently, the accessible electrochemical capacity of CaVO is likely limited to half of its theoretical value, unless strategies are devised to overcome the substantial kinetic bottlenecks associated with deep Ca de-intercalation from CaVO.
In terms of MLIP and TL estimates, we find that the ML predictions do not systematically reproduce the trends observed in the DFT-calculated (Figure S5). For example, at both MLIPs substantially overestimate the DFT-calculated (892-1093 meV vs. 600 meV), while the TL model is in better agreement (467 meV). At , all ML models do overestimate but are in better alignment with DFT- (690-878 meV vs. 661 meV). On the other hand, at and 0.5, the ML models underestimate the DFT-. For the ε phase () where we do not have a DFT-, the ML-predicted resemble the distribution of values obtained for the ζ () phase (MLIPs: 921-1130 meV; TL: 636 meV). Therefore, we suggest the ML models to have also overestimated the at and we expect the true DFT- for the ε phase to exhibit a value intermediate to that of the ζ (600 meV) and δ (661 meV) phases. Hence, we do not expect the possible formation of the ε phase during electrochemical cycling to be a potential bottleneck for Ca motion.
Discussion
We have performed a comprehensive first-principles investigation of the thermodynamics and Ca transport in CaVO cathode to elucidate the origin of experimentally observed electrochemical capacity limitations. Ground state configurations were identified through exhaustive enumeration of configurational degrees of freedom (Figure 1 and S1) and the resultant construction of the 0 K convex hull. A CE Hamiltonian trained on our DFT calculated Ca-vacancy configurations (Figure 2 and Figure 3) was subsequently employed in GCMC simulations to map the temperature–composition phase diagram (Figure 4), thereby resolving the phase stability and the temperature-dependent voltage profiles (Figure 5). We computed for all thermodynamically accessible ordered single-phase configurations (Figure 6), indicating kinetic bottlenecks associated with Ca motion alongside the formation of the γ () phase, which is likely the predominant reason for the observed reductions in electrochemical capacity upon Ca deintercalation from CaVO.
Our computed phase diagram is generally aligned with the experimental observations by Black et al. [25]. For example, our calculations predict the existence of distinct stable phases (α through ζ, Figure 4) separated by miscibility gaps, such as the γ + δ gap that persists to high temperatures, and the δ + ζ gap that can influence electrochemical operation. Operando XRD analysis by Black et al. [25] revealed that during oxidation the Ca-stoichiometry ( in CaVO) of the oxidized phase(s) that formed remains constant even as the total Ca content in the electrode decreased, which is a likely signature of a two-phase behavior, in agreement with our predictions of miscibility gaps in the range. Additionally, XRD refinement of Ca occupancies in the oxidised phase at 298 K indicated a Ca content of ∼ 0.69 mol, [25] which is in alignment with our predicted stability of the δ phase at (Figure 2a and Figure 4).
Although quantitative comparison of calculated voltage profiles (Figure 5) with experimental electrochemical profiles has proven difficult, we do observe a few qualitative similarities. For example, our prediction of the formation of the ε phase (Figure 4 and S3) is in agreement with the characteristic voltage step observed near in experiments, [25] suggesting the (metastable) formation of the ε phase under electrochemical charging. Additionally, we do not expect the ε phase to exhibit any kinetic limitations associated with Ca removal from CaVO, based on the similarities in the ML-predicted values at and (Figure S5). However, we note here that all ML models considered in ths work do not exhibit any qualitative agreement with the overall DFT-calculated trends, indicating further refinement of the models are necessary for more accurate predictions.
Our calculations indicate that the primary bottleneck that limits Ca extraction from CaVO is the high for Ca motion in the γ () phase (Figure 6). As highlighted by its large (negative) formation energy at 0 K ( meV, Figure 2a), the phase exhibits a strong diamond-cubic-like ordering of Ca atoms and vacancies (Figure S1b) with all four Ca sites being exactly half-occupied (Figure S2), as driven by the strong electrostatic attraction between Ca atoms and neighboring vacancies (Figure 3), which contributes to the observed high . Ideally, Ca extraction from can proceed until since the two-phase co-existence (Figure 4) also has the phase that does not exhibit any Ca-mobility limitations (Figure 6). However, Ca extraction within the two-phase region may be limited because of the high in the phase, which can explain the limited extraction of mol of Ca from at 298 K. [25] Further, electrochemical measurements conducted at 323 K do enable higher extraction of Ca (up to 0.6 mol from [25], with possible side reactions as well contributing to the observed capacity), pointing to Ca extraction being a thermally activated process, a signature of the presence of kinetic bottlenecks in the system.
Elucidating the thermodynamic phase stability of electrode materials has long been central to rational battery design. A classic example is (LFP), in which control of phase behavior leads to dramatic improvements in rate capability. [62] Although equilibrium thermodynamics predicts a two-phase reaction separating into Li-rich and Li-poor domains in LFP, [63] particle size reduction enables access to a non-equilibrium solid solution during cycling, effectively bypassing the nucleation barrier and sustaining rapid electrochemical kinetics. [64] Similar strategies have since been widely applied, and numerous studies have focused on understanding the thermodynamic phase stability of cathode materials across other battery chemistries, including sodium-ion systems. [65, 66, 67] Analogous considerations are likely crucial for optimizing the rate performance and cyclability of . By identifying the thermodynamically stable phases and quantifying the associated kinetic barriers, this work provides a framework for exploring size-dependent effects and metastable reaction pathways as potential routes to overcome the poor kinetics observed at intermediate Ca compositions in .
In this context, the relatively high Ca migration barrier at is the result of an energetically unfavorable and spatially constrained diffusion landscape. One possible strategy to mitigate this limitation is to disrupt such ordering of Ca and vacancies through partial substitution at the Ca sites with electrochemically inactive (or active) cations. Similar to observations in P2-type layered sodium transition-metal oxides, [68] where Li substitution suppresses Na-vacancy ordering and promotes a more disordered alkali-ion distribution, introducing suitable dopants in could break the long-range Ca–vacancy correlations, flatten the potential energy landscape, and facilitate Ca motion. Consequently, such an approach may provide a viable pathway to improve the rate capability, accessible capacity, and overall electrochemical performance of as a Ca cathode material.
Conclusion
We systematically investigated the thermodynamic and kinetic landscape of Ca (de)intercalation in , a candidate cathode material for CBs, to elucidate the origins of the limited electrochemical capacity observed experimentally. Using first principles calculations combined with cluster expansion and Monte Carlo simulations, we constructed a temperature–composition phase diagram to resolve the phase stability across in and across temperatures, with the computed voltage profile(s) showing qualitative alignment with experiments. The 0 K convex hull and finite-temperature phase diagram revealed six distinct phases around (), 0.25 (), 0.5 (), 0.67 (), 0.83 (), and 1.0 (), highlighting the complex phase behavior governing Ca (de)intercalation, with the phase appearing over a 370–590 K temperature window via invariant reactions. Further, we demonstrated that the and phases (i.e., at ), exhibit high migration barriers for Ca diffusion, which, together with the stable Ca–vacancy ordering in the phase, gives rise to a significant kinetic bottleneck for full Ca extraction from . Our results establishes a fundamental basis for design strategies to improve the electrochemical performance of , including doping, particle size reduction, and high-temperature electrochemical operations, which can eventually enable the use of as a cathode and the practical deployment of CBs for energy storage.
Acknowledgments
G.S.G. acknowledges financial support from the Science and Engineering Research Board (SERB), Department of Science and Technology, Government of India, under sanction number IPA/2021/000007. D.B.T. and J.A.D. acknowledge financial assistance from the Indian Institute of Science (IISc). The authors acknowledge the computational resources provided by the Supercomputer Education and Research Centre (SERC) at IISc, and the Jülich Supercomputing Centre at Forschungszentrum Jülich, Germany, for access to the JURECA supercomputer under projects hpc-prf-emdft and hpc-prf-desal.
Data and code availability
All computed data associated with this work are available freely to all via our GitHub repository.
Conflicts of interest
The authors have no conflicts of interest to declare.
References
- [1]M Stanley Whittingham. Ultimate limits to intercalation reactions for lithium batteries. Chemical Reviews, 114(23):11414–11443, 2014.DOI
- [2]Dominique Larcher and Jean-Marie Tarascon. Towards greener and more sustainable batteries for electrical energy storage. Nature Chemistry, 7(1):19–29, 2015.DOI
- [3]Björn Nyqvist and Måns Nilsson. Rapidly falling costs of battery packs for electric vehicles. Nature Climate Change, 5(4):329–332, 2015.DOI
- [4]Jean-Marie Tarascon. Is lithium the new gold? Nature Chemistry, 2(6):510–510, 2010.DOI
- [5]Zachary P Cano, Dustin Banham, Siyu Ye, Andreas Hintennach, Jun Lu, Michael Fowler, and Zhongwei Chen. Batteries and fuel cells for emerging electric vehicle markets. Nature Energy, 3(4):279–289, 2018.DOI
- [6]Pieremanuele Canepa, Gopalakrishnan Sai Gautam, Daniel C Hannah, Rahul Malik, Miao Liu, Kevin G Gallagher, Kristin A Persson, and Gerbrand Ceder. Odyssey of multivalent cathode materials: open questions and future challenges. Chemical Reviews, 117(5):4287–4341, 2017.DOI
- [7]Alexandre Ponrouch, Jan Bitenc, Robert Dominko, Niklas Lindahl, Patrik Johansson, and M Rosa Palacín. Multivalent rechargeable batteries. Energy Storage Materials, 20:253–262, 2019.DOI
- [8]Lauren E Blanc, Dipan Kundu, and Linda F Nazar. Scientific challenges for the implementation of zn-ion batteries. Joule, 4(4):771–799, 2020.DOI
- [9]John Muldoon, Claudiu B Bucur, and Thomas Gregory. Quest for nonaqueous multivalent secondary batteries: magnesium and beyond. Chemical Reviews, 114(23):11683–11720, 2014.DOI
- [10]Alexandre Ponrouch, Carlos Frontera, Fanny Bardé, and M Rosa Palacín. Towards a calcium-based rechargeable battery. Nature Materials, 15(2):169–172, 2016.DOI
- [11]M Elena Arroyo-de Dompablo, Alexandre Ponrouch, Patrik Johansson, and M Rosa Palacín. Achievements, challenges, and prospects of calcium batteries. Chemical Reviews, 120(14):6331–6357, 2019.
- [12]Abdel El Kharbachi, Olena Zavorotynska, M Latroche, Fermin Cuevas, Volodymyr Yartys, and M Fichtner. Exploits, advances and challenges benefiting beyond li-ion battery technologies. Journal of Alloys and Compounds, 817:153261, 2020.
- [13]Da Wang, Xiangwen Gao, Yuhui Chen, Liyu Jin, Christian Kuss, and Peter G. Bruce. Plating and stripping calcium in an organic electrolyte. Nature Materials, 17(1):16–20, 2018.DOI
- [14]Zhenyou Li, Olaf Fuhr, Maximilian Fichtner, and Zhirong Zhao-Karger. Towards stable and efficient electrolytes for room-temperature rechargeable calcium batteries. Energy & Environmental Science, 12(12):3496–3501, 2019.DOI
- [15]Shengda D Pu, Chen Gong, Xiangwen Gao, Ziyang Ning, Sixie Yang, John-Joseph Marie, Boyang Liu, Robert A House, Gareth O Hartley, Jun Luo, et al. Current-density-dependent electroplating in ca electrolytes: From globules to dendrites. ACS Energy Letters, 5(7):2283–2290, 2020.DOI
- [16]Abhinandan Shyamsunder, Lauren E Blanc, Abdeljalil Assoud, and Linda F Nazar. Reversible calcium plating and stripping at room temperature using a borate salt. ACS Energy Letters, 4(9):2271–2276, 2019.DOI
- [17]Damien Monti, Alexandre Ponrouch, Rafael B Araujo, Fanny Barde, Patrik Johansson, and M Rosa Palacín. Multivalent batteries—prospects for high energy density: Ca batteries. Frontiers in Chemistry, 7:79, 2019.DOI
- [18]Rosalind J Gummow, George Vamvounis, M Bobby Kannan, and Yinghe He. Calcium-ion batteries: current state-of-the-art and future perspectives. Advanced Materials, 30(39):1801702, 2018.DOI
- [19]Ziqin Rong, Rahul Malik, Pieremanuele Canepa, Gopalakrishnan Sai Gautam, Miao Liu, Anubhav Jain, Kristin Persson, and Gerbrand Ceder. Materials design rules for multivalent ion mobility in intercalation structures. Chemistry of Materials, 27(17):6016–6021, 2015.DOI
- [20]Dereje Bekele Tekliye, Ankit Kumar, Xie Weihang, Thelakkatū Devassy Mercy, Pieremanuele Canepa, and Gopalakrishnan Sai Gautam. Exploration of nasicon frameworks as calcium-ion battery electrodes. Chemistry of Materials, 34(22):10133–10143, 2022.
- [21]Dereje Bekele Tekliye and Gopalakrishnan Sai Gautam. Fluoride frameworks as potential calcium battery cathodes. Journal of Materials Chemistry A, 12(30):18993–19007, 2024.
- [22]Wang Lu, Juefan Wang, Gopalakrishnan Sai Gautam, and Pieremanuele Canepa. Searching ternary oxides and chalcogenides as positive electrodes for calcium batteries. Chemistry of Materials, 33(14):5809–5821, 2021.
- [23]Sanghyeon Kim, Liang Yin, Myeong Hwan Lee, Prakash Parajuli, Lauren Blanc, Timothy T. Fister, Haesun Park, Bob Jin Kwon, Brian J. Ingram, Peter Zapol, Robert F. Klie, Kisuk Kang, Linda F. Nazar, Saul H. Lapidus, and John T. Vaughey. High-voltage phosphate cathodes for rechargeable ca-ion batteries. ACS Energy Letters, 5(10):3203–3211, 10 2020.DOI
- [24]Boosik Jeon, Jongwook W. Heo, Jooeun Hyoung, Hunho H. Kwak, Dongmin M. Lee, and Seung-Tae Hong. Reversible calcium-ion insertion in nasicon-type nav2(po4)3. Chemistry of Materials, 32(20):8772–8780, 10 2020.DOI
- [25]Ashley P Black, Carlos Frontera, Arturo Torres, Miguel Recio-Poo, Patrick Rozier, Juan D Forero-Saboya, François Fauth, Esteban Urones-Garrote, M Elena Arroyo-de Dompablo, and M Rosa Palacín. Elucidation of the redox activity of ca2mno3.5 and cav2o4 in calcium batteries using operando xrd: charge compensation mechanism and reversibility. Energy Storage Materials, 47:354–364, 2022.DOI
- [26]Dereje Bekele Tekliye, Achinthya Krishna Bheemaguli, and Gopalakrishnan Sai Gautam. Geometry-based discovery of calcium battery cathodes accelerated by foundational machine-learned models. arXiv preprint arXiv:2605.29029, 2026.
- [27]Pierre Hohenberg and Walter Kohn. Inhomogeneous electron gas. Physical Review, 136(3B):B864, 1964.
- [28]Walter Kohn and Lu Jeu Sham. Self-consistent equations including exchange and correlation effects. Physical Review, 140(4A):A1133, 1965.DOI
- [29]Juan M Sanchez, Francois Ducastelle, and Denis Gratias. Generalized cluster description of multicomponent systems. Physica A: Statistical Mechanics and its Applications, 128(1-2):334–350, 1984.DOI
- [30]Graeme Henkelman and Hannes Jónsson. Improved tangent estimate in the nudged elastic band method for finding minimum energy paths and saddle points. The Journal of chemical physics, 113(22):9978–9985, 2000.DOI
- [31]Daniel Sheppard, Rye Terrell, and Graeme Henkelman. Optimization methods for finding minimum energy paths. The Journal of chemical physics, 128(13), 2008.DOI
- [32]Georg Kresse and Jürgen Hafner. Ab initio molecular dynamics for liquid metals. Physical review B, 47(1):558, 1993.
- [33]Georg Kresse and Jürgen Furthmüller. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Physical Review B, 54(16):11169, 1996.pubmed.ncbi.nlm.nih.gov/9984901
- [34]Jianwei Sun, Adrienn Ruzsinszky, and John P Perdew. Strongly constrained and appropriately normed semilocal density functional. Physical Review Letters, 115(3):036402, 2015.arxiv.org/abs/1504.03028
- [35]Sergei L Dudarev, Gianluigi A Botton, Sergey Y Savrasov, CJ Humphreys, and Adrian P Sutton. Electron-energy-loss spectra and the structural stability of nickel oxide: An LSDA+U study. Physical Review B, 57(3):1505, 1998.DOI
- [36]Vladimir I Anisimov, Jan Zaanen, and Ole K Andersen. Band theory and mott insulators: Hubbard U instead of stoner I. Physical Review B, 44(3):943, 1991.
- [37]Gopalakrishnan Sai Gautam and Emily A Carter. Evaluating transition metal oxides within DFT-SCAN and SCAN+U frameworks for solar thermochemical applications. Physical Review Materials, 2(9):095401, 2018.DOI
- [38]Olivia Y Long, Gopalakrishnan Sai Gautam, and Emily A Carter. Evaluating optimal U for 3d transition-metal oxides within the SCAN+U framework. Physical Review Materials, 4(4):045401, 2020.DOI
- [39]Dereje Bekele Tekliye and Gopalakrishnan Sai Gautam. Accuracy of metagga functionals in describing transition metal fluorides. Physical Review Materials, 8(9):093801, 2024.
- [40]S Swathilakshmi, Reshma Devi, and Gopalakrishnan Sai Gautam. Performance of the r2scan functional in transition metal oxides. Journal of chemical theory and computation, 19(13):4202–4215, 2023.DOI
- [41]Georg Kresse and Daniel Joubert. From ultrasoft pseudopotentials to the projector augmented-wave method. Physical Review B, 59(3):1758, 1999.DOI
- [42]Peter E Blöchl, Ove Jepsen, and Ole Krogh Andersen. Improved tetrahedron method for brillouin-zone integrations. Physical Review B, 49(23):16223, 1994.pubmed.ncbi.nlm.nih.gov/10010769
- [43]Hendrik J Monkhorst and James D Pack. Special points for brillouin-zone integrations. Physical Review B, 13(12):5188, 1976.DOI
- [44]Mariette Hellenbrandt. The inorganic crystal structure database (ICSD)—present and future. Crystallography Reviews, 10(1):17–22, 2004.DOI
- [45]Shyue Ping Ong, William Davidson Richards, Anubhav Jain, Geoffrey Hautier, Michael Kocher, Shreyas Cholia, Dan Gunter, Vincent L Chevrier, Kristin A Persson, and Gerbrand Ceder. Python materials genomics (pymatgen): A robust, open-source python library for materials analysis. Computational Materials Science, 68:314–319, 2013.DOI
- [46]Paul P Ewald. Die berechnung optischer und elektrostatischer gitterpotentiale. Annalen der physik, 369(3):253–287, 1921.DOI
- [47]Brian Puchala, John C Thomas, Anirudh Raju Natarajan, Jon Gabriel Goiri, Sesha Sai Behara, Jonas L Kaufman, and Anton Van der Ven. Casm—a software package for first-principles based study of multi-component crystalline solids. Computational Materials Science, 217:111897, 2023.
- [48]Anton Van der Ven, John C Thomas, Brian Puchala, and Anirudh Raju Natarajan. First-principles statistical mechanics of multicomponent crystals. Annual Review of Materials Research, 48(1):27–55, 2018.DOI
- [49]B Puchala and A Van der Ven. Thermodynamics of the zr-o system from first-principles calculations. Physical Review B—Condensed Matter and Materials Physics, 88(9):094108, 2013.DOI
- [50]Nicholas Metropolis, Arianna W Rosenbluth, Marshall N Rosenbluth, Augusta H Teller, and Edward Teller. Equation of state calculations by fast computing machines. The journal of chemical physics, 21(6):1087–1092, 1953.
- [51]Axel Van De Walle and Mark Asta. Self-driven lattice-model monte carlo simulations of alloy thermodynamic properties and phase diagrams. Modelling and Simulation in Materials Science and Engineering, 10(5):521–538, 2002.DOI
- [52]Yoyo Hinuma, Ying S Meng, Kisuk Kang, and Gerbrand Ceder. Phase transitions in the lini0. 5mn0. 5o2 system with temperature. Chemistry of Materials, 19(7):1790–1800, 2007.DOI
- [53]Yoyo Hinuma, Ying S Meng, and Gerbrand Ceder. Temperature-concentration phase diagram of p 2-na x coo 2 from first-principles calculations. Physical Review B—Condensed Matter and Materials Physics, 77(22):224111, 2008.
- [54]A Van der Ven, C Marianetti, D Morgan, and G Ceder. Phase transformations and volume changes in spinel lixmn2o4. Solid State Ionics, 135(1-4):21–32, 2000.DOI
- [55]Anton Van der Ven, John C Thomas, Qingchuan Xu, and Jishnu Bhattacharya. Linking the electronic structure of solids to their thermodynamic and kinetic properties. Mathematics and computers in simulation, 80(7):1393–1410, 2010.DOI
- [56]John P Perdew, Kieron Burke, and Matthias Ernzerhof. Generalized gradient approximation made simple. Physical review letters, 77(18):3865, 1996.pubmed.ncbi.nlm.nih.gov/10062328
- [57]Reshma Devi, Baltej Singh, Pieremanuelle Canepa, and Gopalakrishnan Sai Gautam. Effect of exchange-correlation functionals on the estimation of migration barriers in battery materials. npj Computational Materials, 8(1):160, 2022.DOI
- [58]A Niazi, Sergey L Bud’ko, Deborah L Schlagel, JQ Yan, Thomas A Lograsso, A Kreyssig, S Das, S Nandi, Al Goldman, Andreas Honecker, et al. Single-crystal growth, crystallography, magnetic susceptibility, heat capacity, and thermal expansion of the antiferromagnetic s= 1 chain compound cav 2 o 4. Physical Review B—Condensed Matter and Materials Physics, 79(10):104432, 2009.
- [59]Ilyes Batatia, Simon Batzner, Dávid Péter Kovács, Albert Musaelian, Gregor NC Simm, Ralf Drautz, Christoph Ortner, Boris Kozinsky, and Gábor Csányi. The design space of e (3)-equivariant atom-centred interatomic potentials. Nature Machine Intelligence, 7(1):56–67, 2025.DOI
- [60]Benjamin Rhodes, Sander Vandenhaute, Vaidotas Šimkus, James Gin, Jonathan Godwin, Tim Duignan, and Mark Neumann. Orb-v3: atomistic simulation at scale. arXiv preprint arXiv:2504.06231, 2025.arxiv.org/abs/2504.06231
- [61]Reshma Devi, Keith T Butler, and Gopalakrishnan Sai Gautam. Leveraging transfer learning for accurate estimation of ionic migration barriers in solids. npj Computational Materials, 2026.DOI
- [62]Rahul Malik, Fei Zhou, and Gerbrand Ceder. Kinetics of non-equilibrium lithium incorporation in lifepo4. Nature materials, 10(8):587–590, 2011.DOI
- [63]Fei Zhou, Thomas Maxisch, and Gerbrand Ceder. Configurational electronic entropy and the phase diagram of mixed-valence oxides:¡? format:?, the case of li x fepo 4. Physical review letters, 97(15):155704, 2006.DOI
- [64]Hao Liu, Fiona C Strobridge, Olaf J Borkiewicz, Kamila M Wiaderek, Karena W Chapman, Peter J Chupas, and Clare P Grey. Capturing metastable structures during high-rate cycling of lifepo4 nanoparticle electrodes. Science, 344(6191):1252817, 2014.DOI
- [65]Zeyu Deng, Gopalakrishnan Sai Gautam, Sanjeev Krishna Kolli, Jean-Noël Chotard, Anthony K Cheetham, Christian Masquelier, and Pieremanuele Canepa. Phase behavior in rhombohedral nasicon electrolytes and electrodes. Chemistry of Materials, 32(18):7908–7920, 2020.DOI
- [66]Ziliang Wang, Sunkyu Park, Zeyu Deng, Dany Carlier, Jean-Noël Chotard, Laurence Croguennec, Gopalakrishnan Sai Gautam, Anthony K Cheetham, Christian Masquelier, and Pieremanuele Canepa. Phase stability and sodium-vacancy orderings in a nasicon electrode. Journal of Materials Chemistry A, 10(1):209–217, 2022.DOI
- [67]Damien KJ Lee, Zeyu Deng, Gopalakrishnan Sai Gautam, and Pieremanuele Canepa. Thermodynamics of sodium–lead alloys for negative electrodes from first-principles. Chemistry of Materials, 36(14):6831–6837, 2024.DOI
- [68]Mingfeng Xu, Giovanni Gammationi, Michael Häfner, Eduardo Villalobos-Portillo, Carlo Marini, and Matteo Bianchini. Understanding and optimizing li substitution in p2-type sodium layered oxides for sodium-ion batteries. Advanced Functional Materials, page 2425499, 2025.




