Preventing Structural Rearrangements on Battery Cycling: A First

Aug 5, 2016 - (2, 20, 21) The complete structural transformation that occurs in layered LiMnO2 makes it a good model system to understand the properti...
1 downloads 12 Views 3MB Size
Download PDF
Article pubs.acs.org/JPCC

Preventing Structural Rearrangements on Battery Cycling: A FirstPrinciples Investigation of the Effect of Dopants on the Migration Barriers in Layered Li0.5MnO2 Ieuan D. Seymour, David J. Wales, and Clare P. Grey* Department of Chemistry, University of Cambridge, Cambridge CB2 1EW, United Kingdom S Supporting Information *

ABSTRACT: Layered LiMnO2 is a potential Li ion cathode material that is known to undergo a layered to spinel transformation upon delithiation, as a result of Mn migration. A common strategy to improve the structural stability of LiMnO2 has been to replace Mn with a range of metal dopants, although the mechanism by which each dopant stabilizes the structure is not well understood. In this work we characterize ion-migration barriers using hybrid eigenvectorfollowing (EF) and density functional theory to study how trivalent dopants (Al3+, Cr3+, Fe3+, Ga3+, Sc3+, and In3+) affect Mn migration during the initial stage of the layered to spinel transformation in Li0.5MnO2. We demonstrate that dopants with small ionic radii, such as Al3+ and Cr3+, can increase the barrier for migration, but only when they are located in the first cation coordination sphere of Mn. We also demonstrate how the hybrid EF approach can be used to study the migration barriers of dopant species within the structure of Li0.5MnO2 efficiently. The transition state searching methodology described in this work will be useful for studying the effects of dopants on structural transformation mechanisms in a wide range of technologically interesting energy materials.

■

(Pmmn);19 however, the layered structure LiMnO2 with the same O3 oxygen stacking as LiCoO2 can be produced as a metastable phase from ion exchange with NaMnO2.2,3 The structure of layered LiMnO2 (C2/m) is similar to other rhombohedral (R3̅m) LiMO2 compounds, with a reduction in the symmetry of the cell as a result of a cooperative Jahn− Teller distortion of the Mn3+ octahedron. Layered LiMnO2 initially generated considerable interest due to the low cost of Mn and the high theoretical capacity of around 285 mAh/g. However, it was found that after a large first charge capacity the reversible capacity in subsequent cycles dropped rapidly, with a large hysteresis in the voltage between charge and discharge. After extended cycling, the entire material was found to transform to a spinel-like, LiMn2O4, structure, with the appearance of two voltage plateaus around 4 V associated with the removal of Li from tetrahedral sites.2,20,21 The complete structural transformation that occurs in layered LiMnO2 makes it a good model system to understand the properties of other layered LiMO2 systems. A common strategy to stabilize the layered structure of LiMnO2 against the transformation to spinel has been to introduce dopants, M, for Mn to produce LiMn1−yMyO2. Two main categories of dopants have been proposed to stabilize the

INTRODUCTION Layered Li transition metal oxides have been the most widely adopted materials for Li ion battery cathodes since LiCoO2 was first proposed by Mizushima et al.1 Significant research efforts have been dedicated to finding cheaper, less toxic, and higher capacity cathode materials than LiCoO2, with a common strategy being to partially or fully replace Co in the layered LiMO2 structure with other metals (M = Ni, Mn, Al, Li, etc.) to produce families of materials such as LiMnO2,2,3 LiNiO2,4−6 Li[Ni x Li 1 / 3 − 2 x / 3 Mn 2 / 3 − x / 3 ]O 2 , 7 LiNi y Co 1 − y Al 1 − y − z O 2 (NCA),8−10 and LiNiyMn1−yCo1−y−zO2 (NMC).11,12 To provide long-term cycling performance in a Li ion battery, the stability of the LiMO2 structure has to be maintained over numerous charge and discharge cycles. However, for many LiMO2 materials, delithiation results in the LixMO2 structure becoming metastable.13−15 Upon delithiation, transformations have been found to occur from the layered (R3̅m) structure to a spinel (Fd3m ̅ ) structure involving the migration of transition metal ions within the close-packed oxygen lattice of the layered structure. The layered to spinel transformation has been observed on the surface of NMC, 16 NCA,17 and Li[NixLi1/3−2x/3Mn2/3−x/3]O218 materials. In all of these systems, it is suggested that the structural transformation leads to a reduction in the capacity and operating voltage. The layered to spinel transformation is particularly pronounced in layered LiMnO2. The thermodynamically stable phase of LiMnO2 under ambient conditions is orthorhombic © XXXX American Chemical Society

Received: May 26, 2016 Revised: August 2, 2016

A

DOI: 10.1021/acs.jpcc.6b05307 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Figure 1. Schematic representation of local minima associated with Li and Mn migration in layered Li0.5MnO2. Local minima consist of (a) ordered chains of Li and Li vacancies (VLi), (b) 1 tetrahedral Li, (c) 1 octahedral Li in the vacant chain, (d) 1 octahedral Li plus 1 tetrahedral Li, (e) a squarepyramidal MnO5, and (f) tetrahedral Mn forming a Litet−VMn−Mntet dumbbell defect. The diffusing species are enlarged for clarity. Reproduced from ref 38.

Mn could migrate into the tetrahedral site in the Li layer that was not face-sharing with any LiO6 octahedra. After the migration of the Mn ion, a Li ion from an octahedral site in the layer migrates into a tetrahedral site above the newly formed Mn vacancy (VMn) in the Mn layer to form a Litet−VMn−Mntet dumbbell defect. During the migration process, a charge disproportionation was proposed to occur, resulting in a tetrahedral Mn2+ site. After the initial formation of the Litet− VMn−Mntet dumbbell defects, it was then suggested that a higher energy cooperative rearrangement of the Li and Mn species was required to form the LiMn2O4 spinel structure. We recently investigated the complex migration process associated with the first stage of the transformation using a hybrid eigenvector-following (EF) and DFT-based approach.38 By exploring the potential energy landscape with single-ended transition state searches, we found that the migration mechanism involved six local minima, as shown in Figure 1. In the first step of this process, Li was found to migrate from an octahedral site in a vacancy ordered configuration of Li0.5MnO2 (configuration a in Figure 1) to a tetrahedral site in the ordered chain of VLi (configuration b). Li was then found to migrate from the tetrahedral site to an octahedral site in the Li chain (configuration c). Instead of an Mn3+ ion dropping into the newly formed VLi, the next stage of the migration process involved the diffusion of a different Li atom from an octahedral site in the Li layer above the Mn site to a tetrahedral site (configuration d). The Mn3+ ion was then found to migrate from the octahedral site to a new square-pyramidal, MnO5, site in the Li layer (configuration e). The oxidation state of Mn remained as Mn3+ throughout the octahedral to squarepyramidal migration. Finally, the Mn3+ ion migrated from the square-pyramidal site to the tetrahedral site in the Li layer (configuration f). During the migration, a charge disproportionation reaction was observed in which the diffusing Mn3+ gained an electron, becoming Mn2+, and a neighboring octahedral Mn site in the Mn layer was oxidized from Mn3+ to Mn4+. The top of the diffusion barrier was associated with the octahedral to square-pyramidal migration of Mn, in which the transition state corresponds to the point at which Mn was located in the triangular face of the MnO6 octahedron, as shown in Figure 2. In this study we build upon this previous work to show how the migration barrier for Mn is affected by the inclusion of fixed valent M3+ dopants in Li0.5MnO2. We demonstrate how the effect on the migration barriers of a range of dopants can be

layered structure of LiMnO2: electronegative multivalent cations, such as Ni2+, and low fixed-valent cations (Al3+, Cr3+, etc.).22 LiMn1−yNiyO223 has been shown to improve capacity retention over undoped LiMnO2 and reduce the cooperative Jahn−Teller distortion. Ni is substituted into LiMnO2 as Ni2+, with the concomitant oxidation of Mn3+ to Mn4+. The higher migration barrier for Mn4+ vs Mn3+ to diffuse from the octahedral sites to intermediate tetrahedral sites in the structure, arising from the higher crystal field stabilization of the d3 (t2g3eg*0) Mn4+ ion at an octahedral site, increases the stability of the layered structure; however, for low levels of doping, the spinel transformation still occurs. The addition of Ni2+ has also been found to result in the layered polymorph of LiMnO2 becoming more thermodynamically stable than the orthorhombic polymorph. The effect of fixed valence dopants, Al3+,24−28 Cr3+,24,26,29,30 Fe3+,31 In3+,32 Zn2+,31 Mg2+,33 and Y3+,34 on the cycling properties of both polymorphs of LiMnO2 has also been studied. Increased cycling performance was reported for all of the doped systems, although the stabilization mechanism remains unclear. In all of the doped systems, apart from LiMn1−yCryO2,29 the layered to spinel transformation was found to eventually occur after extended cycling. The level of doping has an effect on the cycling stability in all systems. However, the concentration of the dopant species that can be incorporated into the LiMnO2 structure is often limited. For example, with Al doping a maximum solubility of around 7% Al can be achieved via solid state synthesis.28 In the present work, the effect that fixed valent M3+ dopants have on the Mn migration barriers associated with the layered to spinel transformation in LiMnO2 will be investigated to rationalize the structural stabilization that is observed experimentally. While previous calculations have helped to understand how the inclusion of dopants affect the phase stability,35 voltage,36,37 and Li migration barriers36,37 in LiMnO2, an understanding of how they affect the migration barrier of Mn in LiMnO2 has not yet been shown. First-principles calculations have granted insight into the migration mechanisms associated with Li and Mn in the first stages of the layered to spinel transformation for undoped LiMnO2 upon delithiation. In the original study by Reed et al.,14 minimum-energy structures along the migration pathway of an Mn atom diffusing from an octahedral, MnO6, site in the Mn layer of LixMnO2 to a tetrahedral, MnO4, site, in the Li layer were considered. It was found that the first stage of the migration pathway involved the rearrangement of Li ions to form a trivacancy in the Li layer. Once this space was created, B

DOI: 10.1021/acs.jpcc.6b05307 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

eigenvector-following (EF) approach45−48 in the OPTIM code, with energies and gradients taken from an interface with VASP. The convergence parameters used in the minimization of the Rayleigh−Ritz ratio and tangent space minimization were the same as those used in ref 38. As the main focus of this work was to study how the Mn migration barrier that was found in the previous study of undoped Li0.5MnO2 was affected by the inclusion of dopants, the migration barrier from Mn to diffuse from an octahedral site, MnO6, to the square-pyramidal site, MnO5, which was found to be the highest energy process along the reaction pathway, was investigated. To initiate single-ended transition state searches, the transition state for Mn migration in undoped Li0.5MnO2, as shown in Figure 2, was used as an initial guess in which one of the Mn3+ ions in a neighboring environment of the diffusing Mn species was replaced with a fixed valent, M3+ dopant (Al3+, Cr3+, Fe3+, Ga3+, Sc3+, and In3+). Dopants were introduced into Mn3+ sites in the first- and second-nearest-neighbor environments relative to the diffusing Mn3+ species, as shown in Figure 3.

Figure 2. Schematic representation of the transition state structure for Mn diffusion between an octahedral, MnO6 site (configuration d) and a square-pyramidal, MnO5, site (configuration e) in the structure of Li0.5MnO2. The diffusing species are enlarged for clarity. Reproduced from ref 38.

characterized using geometry optimization and hence rationalize the differences in the experimentally observed structural stability of different doped systems. This approach will be useful for studying the effects of dopants on the migration barriers in a range of technologically interesting materials in the future.

■

METHODS Computational Details. The same treatment of the electronic structure that was found to give well-converged results for the migration barriers in undoped Li0.5MnO2 in our previous study38 was adopted in the current work, so only a brief description will be given here. Energies and gradients were calculated with DFT using the Perdew−Burke−Ernzerhof (PBE)39 spin-polarized generalized gradient approximation (GGA) functional in the VASP code.40 The projector augmented wave approach (PAW)41 was used with a plane wave cutoff of 500 eV. For the PAW sets, the s semicore states were treated as valence for Li and Sc, the p semicore states for Mn, Fe, and Cr, and the d semicore states for Ga and In. The standard PAW set was used for O. To treat electron correlation in the systems with 3d electrons, a Hubbard U parameter (DFT +U)42 was included on the Mn, Fe, and Cr sites within the Liechtenstein treatment.43 In this formalism the Coulomb matrix (U) and exchange matrix (J) are treated separately, with an effective U value defined as Ueff = U − J. Ueff parameters of 3.75 , 5.3, and 3.7 eV were chosen for Mn3+/4+, Fe3+, and Cr3+, respectively, based on the work of Wang et al.44 A J value of 1 eV was used in each case. All calculations were performed in the spin-polarized, ferromagnetic state. Transition state searches for M3+-doped systems were performed under fixed volume conditions in supercells of Li18M1Mn35O72. The unit cell parameters were fixed to those of the undoped Li18Mn36O72 cell that was used in the previous work,38 which was formed from a 3a × 3b × 2c expansion of the primitive Li0.5MnO2 unit cell. A 1 × 2 × 2 Monkhorst Pack k-point grid was used to sample reciprocal space. The energy from single point calculations was converged to 10−6 eV. To investigate the effect of dopants on the local bonding arrangement, geometry optimizations were performed in which the atomic positions were allowed to relax under the same fixed volume conditions as used for the transition state calculations. The root-mean-square forces were converged to below 10−5 eV Å−1. Transition State Searching. We adopt the same methodology that was used in our previous work on undoped Li0.5MnO2,38 in which the migration barriers associated with the layered to spinel transformation were investigated with a hybrid

Figure 3. Schematic representations of the local environment around an Mn ion (X) diffusing between an octahedral, MnO6 site (configuration d) and a square-pyramidal, MnO5, site (configuration e) in the Mn3+/Mn4+ charge ordered structure of Li0.5MnO2. The firstand second-nearest-neighbor Mn3+ positions are labeled.

To investigate the migration of the dopant species themselves, the transition state structure of undoped Li0.5MnO2 (Figure 3) was once again used as an initial starting guess in which the diffusing Mn species (X) was replaced by the dopant species.

■

RESULTS AND DISCUSSION Local Structure. In our previous work on undoped Li0.5MnO2,38 the lowest energy configuration of Li and Li vacancies (VLi) was found to consist of ordered chains of Li and VLi along the b direction of the cell (Figure 1a). The Li and VLi ordering in the structure was accompanied by a charge ordering of Mn3+ and Mn4+ along the b-axis of the cell. To investigate the effect that M3+ doping had on the local bonding environment, one of the Mn3+ atoms in the ordered Li0.5MnO2 chain structure was replaced with a trivalent M3+ dopant (Al3+, Cr3+, Ga3+, Fe3+, Sc3+, and In3+), as shown in Figure 4, with the changes in the local bond lengths collected in Table 1. As a result of the d4 (t2g3eg*1) electronic configuration of octahedral Mn3+, each of the Mn3+ sites in the undoped, ordered Li0.5MnO2 structure experiences a strong positive Jahn−Teller distortion in which four of the Mn−O bonds are shortened and two are lengthened (Table 1). In the ordered structure of Li0.5MnO2 considered in the present work, the Mn3+/Mn4+ charge ordering results in the lengthened Mn−O bonds being aligned along the [101] direction in the cell, producing the same cooperative distortion that is observed in C

DOI: 10.1021/acs.jpcc.6b05307 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Mn Diffusion: First-Nearest-Neighbor Dopant. The migration barrier of Mn from an octahedral site was calculated in the supercell with one dopant introduced into the first coordination sphere, as shown in Figure 3. The resulting migration barriers are shown in Figure 5.

Figure 4. Ordered Li and vacancy (VLi) chain structure of Li0.5MnO2 with a metal, M3+, dopant replacing a Mn3+ ion on the octahedral site. The arrows highlight the charge ordered chains of Mn3+ and Mn4+.

Table 1. Variation in Local Bond Length with the Inclusion of a Metal, M3+, Dopant in the Ordered Li and Vacancy Chain Structure of Li0.5MnO2a M3+−O bond length (Å)

1st nn Mn3+−O bond length (Å)

dopant

ionic radius (Å)

short

long

av

short

long

av

Mn Al Cr Ga Fe Sc In

0.645 0.535 0.615 0.62 0.645 0.745 0.8

1.969 1.937 2.021 2.007 2.031 2.087 2.141

2.290 2.038 2.088 2.118 2.124 2.168 2.239

2.076 1.971 2.043 2.044 2.062 2.114 2.174

1.969 1.971 1.966 1.968 1.967 1.962 1.962

2.290 2.249 2.266 2.266 2.270 2.287 2.305

2.076 2.064 2.066 2.068 2.067 2.071 2.076

Figure 5. Energy along the Mn migration pathway between an octahedral site (configuration d) and a square-pyramidal site (configuration e) containing one M3+ dopant in the first-nearestneighbor environment of the Li0.5MnO2 structure. The migration barrier for the undoped (Mn) Li0.5MnO2 is reproduced from ref 38.

In all of the doped structures, the Mn migration mechanism remained the same as in the undoped case, in which the Mn3+ diffused from an octahedral, MnO6, site in the Mn layer (configuration d in Figure 1) to a square-pyramidal MnO5 site (configuration e) in the Li layer. The transition state for this pathway corresponded to the point at which the Mn ion passes through the triangular face of the MnO6 octahedron (Figure 2). In Figure 5, the path length of zero corresponds to the Mn in the octahedral site (configuration d), and unity corresponds to the Mn in the square-pyramidal site (configuration e). The energy for each doped structure is quoted relative to the energy of the Mn in the octahedral site (configuration d) in each case. Comparing the migration barriers of the undoped (Mn) cell (0.196 eV) to the doped cells, we see that Al and Cr in the first coordination environment have the biggest effect on the Mn diffusion, increasing the migration barrier to 0.309 and 0.285 eV, respectively. Doping Fe (0.257 eV), Ga (0.259 eV), or Sc (0.227 eV) into the first coordination environment also increases the migration barrier, but to a lesser extent. In contrast, the introduction of In (0.168 eV) reduced the barrier. Plots of the variation in the migration barriers versus the ionic radius of the dopant as well as the changes in the average Mn−O bond length are shown in Figures 6 and 7, respectively. From Figure 6 it can be seen that a decrease in the ionic radius of the dopant results in an increase in the migration barrier. Similarly, an observed decrease in the average Mn−O bond distance for Mn ions in the first cation shell surrounding the M3+ dopant correlates with an increase in the migration barrier (Figure 7). The smallest dopants lead to the shortest average Mn−O bond lengths for the neighboring Mn3+ ion (Table 1), predominantly due to a reduction in the Jahn−Teller lengthened bond; the shorter Mn−O bonds will lead to an increase in the ionic bond strength, which is inversely proportional to the distance between the Mn and O centers. The increased ionic bond strength therefore stabilizes the Mn in the octahedral configuration, increasing the migration barrier for Mn to move to the square-pyramidal site. The barrier for the undoped cell (Mn) does not follow the trend, as a result of the cooperative Jahn−Teller distortion of the neighboring Mn3+

The Mn3+−O bond length of the 1st nearest neighbor (nn) octahedral Mn3+ site and the M3+−O bond lengths around the dopant are shown. The positive Jahn−Teller distortion of the neighboring Mn3+ species results in four shortened bonds and two lengthened bonds for all species. Average bond length = 1/3 × long + 2/3 × short. a

fully lithiated LiMnO2. In the Mn layer of ordered Li0.5MnO2, each Mn3+ octahedron is edge-sharing with two Mn3+ octahedra along Jahn−Teller shortened bonds and four Mn4+ octahedra along Jahn−Teller lengthened bonds. Table 1 shows that the presence of the fixed valence M3+ dopants in the octahedral Mn3+ site changes the M−O bond lengths of the site itself and of the neighboring Mn3+ sites. The bonding of the neighboring Mn4+ sites is less affected, as shown in Table S1 of the Supporting Information. For all of the dopant systems considered, the electronic configuration is such that a Jahn−Teller distortion would not be expected at the dopant center itself. However, around the dopant site there is a small splitting of the M3+−O bond lengths into longer and shorter bonds as a result of the cooperative distortion of the neighboring Mn3+ sites. The changes in the bond lengths of the neighboring Mn3+ site depend on the ionic radius of the dopant, but because of the Jahn−Teller distortion on the Mn3+ ion, the trend is not necessarily intuitive: for the Jahn−Teller lengthened Mn−O bonds, for all dopants apart from In, the bond lengths are reduced with respect to undoped Li0.5MnO2, with smaller ionic radii dopants leading to the biggest reduction in the bond length. The Jahn−Teller shortened Mn−O bonds are also affected by the presence of dopants, but to a lesser extent, with larger dopants reducing the Mn−O bond length. The Mn−O bonds of the neighboring Mn3+ site are reduced with respect to undoped Li0.5MnO2 for all dopants except for Al, with larger ionic radii dopants leading to the biggest reduction. D

DOI: 10.1021/acs.jpcc.6b05307 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

site. In the ordered structure (configuration a), the cooperative distortion between the neighboring Mn3+ site and the diffusing Mn3+ site means that there is a preferential elongation of the Jahn−Teller lengthened bonds on both sites. Even though the ionic radius of Mn3+ neighbor is similar to that of Fe, the preferential elongation of the Jahn−Teller lengthened bond has the same effect on the Mn−O length as a larger dopant, such as In. Potential Energy Surface: First-Nearest-Neighbor Dopant. The migration of Mn relies on the migration of Li to form a Li trivacancy in the Li layer, as outlined in the Introduction.14 In the previous study,38 starting from the ordered chain structure of Li0.5MnO2 (configuration a), it was found that six minima were involved in the migration pathway, four of which (configurations a, b, c, and d) were connected by Li migration processes. The migration barrier for undoped Li0.5MnO2 is reproduced in Figure 8. Although the inclusion of dopants may result in the presence of slightly different minima along the reaction pathway, at low levels of doping, the six minima that were found for undoped Li0.5MnO2 are a good approximation for the minima structures of doped systems. To investigate how the inclusion of dopants affected the relative energies of all of the minima along the complete pathway, the geometries of all of the minima containing a M3+ dopant in the same first-nearest-neighbor environment were allowed to relax. The resulting energies are also shown in Figure 8. The energies of minima associated with Li migration are slightly increased by the addition of Al and Cr, whereas the addition of In and Sc increases the energy more significantly. For configuration b, containing one tetrahedral Li in the vacant chain, the Al- and Cr-doped cells are 0.066 and 0.053 eV higher in energy, respectively, while the Sc and In cells are 0.104 and 0.110 eV higher, respectively. A possible explanation for the increase is that the size of the tetrahedral site in configuration b is inversely proportional to the size of the dopant as shown in the Supporting Information, Figure S1. From the previous theoretical work of Kang et al.,49 the Li migration barrier in layered compounds was found to be sensitive to the tetrahedral volume, with smaller Li tetrahedron leading to larger migration barriers. Explicit calculation of the migration barrier in each

Figure 6. Variation in the Mn migration barrier with the ionic radius of the M3+ dopants included into the first cation coordination sphere of Mn in the Li0.5MnO2 structure (triangles). The energy barrier for undoped Li0.5MnO2 (Mn) reproduced from ref 38 is shown with a square.

Figure 7. Variation in the Mn migration barrier with the average Mn− O bond length around Mn in the first-nearest-neighbor coordination to a M3+ dopant in the vacancy ordered structure of Li0.5MnO2 (triangles). The energy barrier for undoped Li0.5MnO2 (Mn) reproduced from ref 38 is shown with a square.

Figure 8. Energy of local minima along the initial layered to spinel transformation for doped M3+ structure of Li0.5MnO2. The energies of local minima and transition states are represented with filled and open symbols, respectively. Minima are (a) ordered Li18Mn35M1O72, (b) 1 tetrahedral Li, (c) 1 octahedral Li, (d) 1 tetrahedral + 1 octahedral Li, (e) a square-pyramidal MnO5, and (f) tetrahedral MnO4. The full migration barrier of undoped Li0.5MnO2 (Mn) reproduced from ref 38 is shown with the blue dashed line. E

DOI: 10.1021/acs.jpcc.6b05307 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C case would be required to fully understand the effect that the dopant has on the Li migration; however, the smaller tetrahedral volume and higher energy of the Li minima structures suggest that In and Sc will increase the barrier for Li diffusion. It is interesting to note that if the overall migration barrier is considered, the maximum in energy along the pathway associated with the Mn diffusion barriers found in Figure 8 is now similar for all dopants. Based on the above arguments, this result suggests that the smaller dopants, such as Al and Cr, increase the overall migration barrier by increasing the barrier to Mn migration, whereas the larger dopants, such as In and Sc, increase the overall migration barrier by increasing the barrier to Li migration. Mn Diffusion: Second-Nearest-Neighbor Dopant. To investigate how the location of the dopant affected the migration barrier of Mn from the octahedral to the squarepyramidal site, the M3+ dopants were introduced into the second cation coordination sphere, as described in Figure 3, with the resulting migration barriers shown in Figure 9.

Figure 10. Energy along the migration pathway of metal, M3+, dopants from octahedral to tetrahedral sites within the structure of Li0.5MnO2. The migration barrier for Mn in the undoped Li0.5MnO2 is reproduced from ref 38.

undoped case (Mn) is reproduced in Figure 10, in which two transition states occur as a result of the two migration processes (Mn octahedral → Mn square pyramidal and Mn square pyramidal → Mn tetrahedral). For Al, Fe, and Ga, the mechanism involves a direct migration from the octahedral to the tetrahedral site in which the transition state occurs when the M3+ dopant passes through the triangular O face of the MO6 octahedron. The migration barrier for Al (0.287 eV) is slightly higher than for Mn in undoped Li0.5MnO2 (0.196 eV), whereas the migration barriers for Fe (0.130 eV) and Ga (0.017 eV) are smaller than for Mn in undoped Li0.5MnO2. As for Mn2+, the tetrahedral site is more thermodynamically stable than the octahedral configuration for Al, Fe, and Ga. For Cr3+, the first migration mechanism located was the diffusion of Cr3+ from the CrO6 octahedral site (configuration d) to the CrO5 square-pyramidal site (configuration e), analogous to the first stage of Mn diffusion in undoped Li0.5MnO2. The migration of Cr3+ from the square-pyramidal site to the tetrahedral was then examined using the same procedure of doping Cr3+ into the transition state structure for the corresponding pathway in undoped Li0.5MnO2 (see Figure S2). Throughout the migration process, Cr maintained the Cr3+ oxidation state, as shown by the integrated unpaired spin density (see Figure S3). The migration barrier for Cr (0.922 eV) is significantly larger than for Mn in undoped Li0.5MnO2, and unlike the other dopants, the tetrahedral configuration of Cr3+ is thermodynamically less stable than the octahedral configuration. For Sc and In, transition states could not be located from the initial starting guess structure, which suggests that the transition state structures for Sc3+ and In3+ migration are different from those of the other dopants investigated. During the initial unsuccessful transition state searches for these systems, the Li in the tetrahedral site (configuration d) appeared to be unstable and migrated back toward the octahedral site (configuration c). A transition state search was therefore initiated in which the Li and vacancy arrangement in configuration c was used as the starting structure and the In and Sc dopants were displaced by 0.2 Å toward the vacancy in the Li layer. For both the Sc and In systems, a cooperative migration mechanism was found in which the octahedral Li and dopant ions in configuration c migrated to the tetrahedral sites in the Li layer, (configuration f), as shown in Figure S4. The corresponding migration barriers that were located are shown in Figure 11, in addition to the corresponding migration barrier for undoped Li0.5MnO2 that involves three successive migration processes (Li octahedral →

Figure 9. Energy along the Mn migration pathway between an octahedral site (configuration d) and a square-pyramidal site (configuration e) containing one M3+ dopant in the second-nearestneighbor environment of the Li0.5MnO2 structure. The migration barrier for the undoped (Mn) Li0.5MnO2 is reproduced from ref 38.

As shown in Figure 9, when the M3+ dopant is introduced into the second cation coordination environment, there is little variation in the resulting Mn3+ migration barrier regardless of the type of dopant introduced. This observation shows that the variation in the migration barrier is a highly localized effect resulting from changes in the Mn−O bond length of the firstnearest-neighbor Mn sites. Dopant Migration. The stability of the M3+ dopants in Li0.5MnO2 was investigated by calculating the migration barriers for the M3+ ions to diffuse from the octahedral sites in the Mn layer to tetrahedral sites in the Li layer, i.e., configuration d to f in Figure 1. To provide an initial guess for the transition state of each migrating M3+ species, the transition state associated with Mn diffusion from the octahedral to square-pyramidal site (Figure 2) in undoped Li0.5MnO2 was used, with the diffusing Mn replaced with M3+. The resulting migration barriers are shown in Figure 10. The migration mechanism was found to vary depending on the dopant. For the horizontal axis in Figure 10, zero corresponds to the M3+ species in the octahedral site (configuration d) and unity corresponds to the M3+ in the tetrahedral site (configuration f), and the energies are quoted relative to configuration d. The migration barrier for the F

DOI: 10.1021/acs.jpcc.6b05307 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

LiMn1−yAlyO2. From Figure 5, the increase in the migration barrier when Al3+ is in the first cation coordination sphere of Mn suggests that LixMn1−yAlyO2 will be more stabilized against the first stage of the layered to spinel transformation than undoped LixMnO2. This result is consistent with experiment, in which the first discharge capacity of LiMn 0.95 Al 0.05 O 2 (approximately 170 mAh/g) was found to be higher than that of LiMnO2 (∼140 mAh/g).24 As shown in Figure 5, only dopants in the first cation coordination sphere increase the migration barrier for Mn. For the highest possible solubility of Al3+ doping at LiMn0.93Al0.07O2, assuming that the Al3+ ions are randomly distributed in the structure, the probability of a Mn site having at least one Al3+ in the six first-nearest-neighbor positions is 35.3%. Therefore, although the Al3+ ions may enhance the stability locally, the majority of the Mn sites in the structure do not experience the effect, and so the first stage of the layered to spinel transformation involving the migration of Mn into tetrahedral sites in the Li layer is still expected to occur. If the Al3+ ions are not introduced into the structure randomly but cluster together, for example at the surface of the particles, then a similar effect will occur in which locally the Al3+ rich areas will experience enhanced stability, whereas the Al3+ depleted areas will be susceptible to transformation. As shown in Figure 10, the migration barrier for Al3+ (0.287 eV) to diffuse from the octahedral site in the Mn layer to the tetrahedral in the Li layer is only slightly higher than that for Mn3+ (0.196 eV) in undoped Li0.5MnO2, which suggests that during the first stage of the process Litet−VMn−Altet defects are also likely to be formed. It can also be seen from Figure 10 that there is a small energy difference between the octahedral and tetrahedral sites for Al3+ (−0.515 eV), which is similar to that of Mn (−0.457 eV). Therefore, we expect that in the second stage of the phase transformation process the migration barrier of Al3+ from tetrahedral sites to octahedral sites in the spinel structure should be similar to that of Mn in undoped Li0.5MnO2, which is consistent with the fact that all LixMn1−yAlyO2 materials are eventually seen to transform to the spinel structure.26 LiMn1−yFeyO2. From experimental results, the inclusion of 5% doping of Fe3+ is found to have a similar effect to Al3+, resulting in a first cycle discharge capacity around 190 mAh/ g.31 As for Al3+, the inclusion of Fe3+ into the first cation coordination sphere of Mn was found to increase the migration barrier, but to a lesser extent. However, as discussed previously for Al3+, at small doping fractions, the probability of a Mn site having a first-nearest-neighbor Fe3+ is low, suggesting that the first stage of the transformation is still going to occur. Higher fractions (y = 0.2) of Fe3+ dopants can be introduced into the LiMn1−yFeyO2 structure than for LixMn1−yAlyO2. However, higher Fe3+ concentrations were found to result in lower discharge capacities.31 The origin of the lower discharge capacity at higher Fe3+ doping fraction may be due to the fact that the migration barrier for Fe3+ is lower than that of Mn3+, which suggests that the formation of Litet−VMn−Fetet defects in the first stage of the process should be even more facile than Litet−VMn−Mntet defects. The energy difference between the octahedral and tetrahedral sites is once again small for Fe3+ (−0.401 eV) due to the lack of octahedral stabilization energy for the d5 electronic configuration, which means that the second stage of the transformation process should also occur for LiMn1−yFeyO2, as is seen experimentally.31 LiMn1−yGayO2. Although Ga3+ is suggested as a possible dopant to increase the structural stability of LiMnO2 in ref 50, electrochemical data was not provided. However, based on the

Figure 11. Energy along the migration pathway of metal, Sc3+ and In3+, dopants from octahedral to tetrahedral sites within the structure of Li0.5MnO2. The migration barrier for Mn in the undoped Li0.5MnO2 is reproduced from ref 38. For the horizontal axis, zero corresponds to energy for the M3+ species in the octahedral site (configuration c) and unity corresponds to the M3+ in the tetrahedral site (configuration f), and the energies are quoted relative to configuration c.

Li tetrahedral, Mn octahedral → Mn square pyramidal, and Mn square pyramidal → Mn tetrahedral). The transition state for this Sc/In migration pathway occurred when the diffusing Li+ ion was located in the triangular O face of the LiO6 octahedra, as show in Figure S4. The migration mechanism for the Sc3+ and In3+ is therefore initiated by the migration of Li analogous to the other M3+ systems; however, for Sc3+ and In3+ the tetrahedral Li site is not stable, and instead a spontaneous migration of the Sc3+ and In3+ occurs once the Li ion passes through the transition state. The migration barriers for Sc (0.143 eV) and In (0.013 eV) via this cooperative mechanism are noticeably lower than that of Mn in undoped Li0.5MnO2 (0.436 eV). For the Al3+, Fe3+, Cr3+, and Fe3+, migration barrier calculations were performed starting from the transition state structures found for Sc3+ to see whether a low-energy cooperative migration process could be located from configuration c to f; however, in all cases configuration d involving a stable Li tetrahedral site was found to be a local minimum, as was the case for undoped Li0.5MnO2. General Discussion. The use of the hybrid EF approach in the current work provided an efficient method for studying the effect of dopants on the migration mechanisms associated with the layered to spinel transformation in Li0.5MnO2. In the hybrid EF approach, the most computationally demanding part of the calculation is associated with the location of the transition state. By using the previously found transition state structure for undoped Li0.5MnO2 as an initial starting guess in doped Li0.5MnO2, the computational cost of locating the transition states was decreased. For the doped structures, five to seven EF steps were required to locate the transition states for Mn migration in Figure 5, as opposed to ten EF steps for undoped Li0.5MnO2 in ref 38, in which no previous information for the transition state structure was available. Based on the migration barriers found in this work, the specific stabilizing effect of the individual trivalent dopants on the structure of delithiated LixMn1−yMyO2 observed experimentally can be rationalized. In the following discussion, we propose how the dopants will affect both the initial stage of the layered to spinel transformation, which involves the migration of Mn from octahedral to tetrahedral sites in the layered structure, and the second stage involving the cooperative rearrangement of Mn from tetrahedral sites to octahedral sites in the spinel structure.14 G

DOI: 10.1021/acs.jpcc.6b05307 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

small doping concentrations of y = 0.1, the structure becomes more disordered on delithiation, which may be due to the migration of Mn to tetrahedral sites in the Li layer.29 This result is consistent with the fact that at a doping concentration of y = 0.1 only 46.9% of Mn have a Cr3+ dopant as a first-nearest neighbor. However, unlike in the LiMn1−yAlyO2 system, for LiMn1−yCryO2 high Cr3+ doping levels are possible, as Cr and Mn form a complete solid solution.29 To increase the probability of a Mn site having first-nearest-neighbor M3+ ion to >95%, and therefore significantly increase the barrier for the first stage of the transformation process, doping levels around y = 0.4 would be required. The observation that LiMn1−yCryO2 does not undergo the second stage of the layered to spinel transformation is related to the fact that the migration barrier for Cr3+ diffusion (0.922 eV) is significantly higher than Mn in undoped Li0.5MnO2, as the tetrahedral site is energetically more unfavorable than the octahedral site by 0.548 eV. The large barrier for Cr3+ migration is consistent with that found for Cr migration in Li0.5CrO2 (1.10 eV).15 The large migration barrier and stability of the octahedral site for Cr3+ are due to a ligand field stabilization effect22 of the d3 Cr3+ ion, and so in an octahedral environment, only the lower energy, nonbonding t2g states are populated (t2g3eg*0). In other layered systems, such as Li[Li0.2Cr0.4Mn0.4]O2, where migration of Cr ions from octahedral sites to tetrahedral sites is found to occur, this effect is accompanied by an oxidation of Cr3+ to the d0 ion, Cr6+.52 However, in the current work Cr remains in the Cr3+ state throughout the migration process, and so in the tetrahedral site, the electronic structure involves the occupation of the higher energy t2* antibonding orbital (e2t2*1), which is energetically unfavorable. The square-planar, CrO5, configuration that was also observed for Mn in undoped Li0.5MnO2 differs in energy from the tetrahedral configuration by only 1 meV. The significantly higher migration barrier for Cr3+ suggests that Litet−VMn−Crtet dumbbell defects will not be formed in the first stage of the transformation process. The fixation of the Cr ions in the layered structure must therefore hinder the cooperative migration of Mn in the second stage of the transformation process to spinel.

results for the LixMn1−yAlyO2 and LixMn1−yFeyO2 systems, a model of the structural stability can be proposed. The increase in the migration barrier of Mn when Ga3+ is in the first coordination sphere is similar to Fe3+, suggesting that Ga3+ doping will result in a slightly enhanced stability against the first stage of the transformation, which will be sensitive to the Ga3+ concentration. However, unlike Li x Mn 1−y Al y O 2 and LixMn1−yFeyO2, the barrier for Ga3+ migration (0.017 eV) is much lower than for Mn (0.196 eV), suggesting that Litet− VMn−Gatet will form rapidly in the structure during delithiation under ambient conditions. The migration of Ga3+ from octahedral to tetrahedral sites in the structure of LiNi0.908Co0.085Ga0.003O2 has been previously observed experimentally through XAS.51 The energy preference for the Ga3+ (−0.684 eV) ions to occupy the tetrahedral sites is higher than for Mn, which suggests that the second stage of the transformation involving tetrahedral → octahedral migration may be more energetically unfavorable for Ga3+. As proposed for LiNi0.908Co0.085Ga0.003O2, the rapid formation of tetrahedral Ga3+ defects may hinder the migration of Mn in both the first and second stages of the migration process, hindering the layered to spinel transformation in LixMn1−yGayO2. LiMn1−yScyO2. To the best of our knowledge, no electrochemical cycling data have been published for LixMn1−yScyO2 systems. However, the similarities between the migration barriers of the LixMn1−yScyO2 and LixMn1−yFeyO2 systems allow some predictions to be made. The small increase in the Mn migration barrier with the inclusion of Sc3+ into the structure suggests that Sc3+ doping will only result in a modest increase in the stability against the first stage of the transformation, as is seen experimentally for Fe3+ doping which has a similar effect on the Mn migration barrier. As the barrier for cooperative migration of Sc and Li (0.143 eV) is considerably smaller than that of sequential Li and Mn3+ migration (0.436 eV) in undoped Li0.5MnO2, it is expected that the formation of Litet−VMn−Sctet defects during the first stage of the process should be fairly rapid. The energy difference between the octahedral and tetrahedral sites for Sc3+ (−0.209 eV) is once again small, which suggests that the second stage of the transformation to spinel should be relatively facile, as seen experimentally for LiMn1−yFeyO2.31 LiMn1−yInyO2. The small decrease in the Mn migration barrier with the inclusion of In3+ suggests that In3+ doping will actually facilitate the first stage of the transformation to a small extent. However, the potential destabilizing effect of In may be particularly limited by the low solubility in this system. Although experimental data are not available for layered LiMn1−yInyO2, the previous study on the orthorhombic polymorph of LiMnO2 showed that only around 3.2% In could be included in the orthorhombic LiMnO2 structure before the impurity phase LiInO2 was formed.32 If a similar level of solubility were present in the layered LiMnO2 structure, then only 17.7% of Mn would have an In3+ dopant in the firstnearest-neighbor configuration. Similar to the Ga3+ system, the migration barrier for In3+ is very low (0.013 eV), which suggests that Litet−VMn−Intet will form more rapidly than Litet−VMn− Mntet defects. The energy preference for In3+ (−0.689 eV) ions to occupy tetrahedral sites over octahedral sites is also similar to the Ga3+ case and may suggest that In3+ doping will hinder the second stage of the layered to spinel transformation, although the effect may be minimal at low doping concentrations. LiMn1−yCryO2. For LiMn1−yCryO2, the full transformation to the spinel structure is not observed experimentally; however, at

■

CONCLUSION In this study we demonstrate how hybrid eigenvector-following transition state searches combined with density functional theory can be used to investigate the effects that trivalent dopants (Al3+, Cr3+, Fe3+, Ga3+, Sc3+, and In3+) have on the initial stage of the layered to spinel transformation in delithiated Li0.5MnO2. The transition state structure that was found for Mn migration in undoped Li0.5MnO2 was used as an initial starting guess for in doped structures, significantly improving the computational efficiency. The introduction of dopants into the first cation coordination sphere of Mn3+ was found to change the energy of the migration barrier for Mn to diffuse from an octahedral to a square-pyramidal site, relative to undoped Li0.5MnO2. The variation in the migration barrier correlates with the local changes in the Mn−O bond length as a result of the different ionic radii of the dopants, with small ionic radius dopants such as Al3+ and Cr3+ producing the biggest increase in the barrier for Mn diffusion. Dopants in the second cation coordination sphere had little effect on the Mn migration barrier, highlighting the highly localized stabilization effect. Larger dopants, such as In3+, were found to increase the overall migration barrier for Mn migration by increasing the energy of local minima associated with Li diffusion. H

DOI: 10.1021/acs.jpcc.6b05307 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C

Energy Storage (NECCES), an Energy Frontier Research Center funded by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award #DE-SC0012583. I.D.S. acknowledges funding from NECCES and the Geoffrey Moorhouse Gibson Studentship in Chemistry from Trinity College Cambridge.

On the basis of diffusion pathways taken by Li and Mn in undoped Li0.5MnO2, the migration pathway for Al3+, Fe3+, and Ga3+ from octahedral sites in the Mn layer to tetrahedral sites in the Li layer was proposed to involve similar local minima consisting of ordered octahedral Li → one tetrahedral Li → one octahedral Li → one octahedral plus one tetrahedral Li → one tetrahedral M3+ plus one tetrahedral Li. In contrast, the migration pathway for In3+ and Sc3+ dopants was proposed to occur via the mechanism ordered octahedral Li → one tetrahedral Li → one octahedral Li → one tetrahedral M3+ plus one tetrahedral Li, as the presence of the second Li tetrahedral site above the octahedral M3+ dopant was found to be unstable. The migration barriers for Al3+ and Fe3+ were found to be similar to those for Mn diffusion in undoped Li0.5MnO2, whereas there was almost no diffusion barrier for Ga3+ to diffuse from octahedral to tetrahedral sites. The migrations barrier for In3+ and Sc3+ via the second mechanism were found to be noticeably lower than for Mn. The migration barrier for Cr3+ from the octahedral to tetrahedral sites was found to be much higher and occurred via an intermediate square-pyramidal, CrO5, configuration. Based on the variation in the migration barriers calculated, the relative efficiency of different trivalent dopant species in stabilizing the layered structure of LiMnO2 against transformation to spinel can be rationalized, with Cr3+ emerging as the only dopant that increases the migration barrier for Mn migration, but does not migrate to tetrahedral sites within the Li layer. Our analysis of the stabilizing role of dopants in Li0.5MnO2, as well as the transition state searching methodology, will be useful in studying a range of battery materials in which the inclusion of dopants is a key factor in determining the structural stability and safety of the materials.

■

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcc.6b05307. The variation in the local Mn4+−O bond lengths around the M3+ dopants; a plot of the variation of the Li tetrahedral volume with the ionic radius of the dopant; the transition state structure for square-pyramidal (CrO5) to tetrahedral (CrO4) Cr3+ diffusion; the integrated differential spin density around Cr during the octahedral (CrO6) to tetrahedral (CrO4) migration; the minima and transition state structures for octahedral to tetrahedral diffusion of In3+ and Sc3+ (PDF)

■

REFERENCES

(1) Mizushima, K.; Jones, P. C.; Wiseman, P. J.; Goodenough, J. B. LixCoO2 (0 < x