betaFeedback

WorksPhysics and Astronomy

Understanding insulating ferromagnetism in LaCoO3 films under tensile strain

Abstract

LaCoO3 thin films grown under epitaxial tensile strain exhibit a robust ferromagnetic insulating state that is absent in the bulk. Despite many studies, both experimental and computational, the microscopic origin of this phenomenon is not well understood. In this work, density functional theory calculations are used to systematically investigate the magnetic ground state of stoichiometric LaCoO3 under epitaxial strain equivalent to that imposed by a SrTiO3 substrate. The results identify a ferromagnetic insulating ground state characterized by a unique ordered array of high-spin (HS) and low-spin (LS) Co3+ ions. The spin state ordering is best described as 2 x 2 columns that consist of alternating HS and LS Co3+ ions, separated by planes of LS Co3+ ions. This leads to HS-LS-LS repeating sequence of Co3+ ions in both pseudocubic [100] and [010] directions. Analysis of the electronic structure confirms the presence of an insulating gap. Evaluation of the superexchange interactions reveal ferromagnetic interactions between HS Co3+ ions via 90 degree paths, and antiferromagnetic interactions via 180 degree paths, both of which are facilitated by empty sigma* (eg) orbitals on the diamagnetic LS Co3+ ions. The strength and number of 90 degree ferromagnetic interactions are sufficient to overcome the competing 180 degree antiferromagnetic interactions stabilizing a ferromagnetic insulating state.

Introduction

The interplay between magnetism and electronic transport in transition metal oxides (TMOs) is not only a rich source of complex fundamental materials science, but also underpins a wide range of applications from spintronics to thermoelectric energy harvesting Morosan et al. [1]. The combination of ferromagnetism and insulating behavior is of particular interest, but this combination is rarely seen in TMOs, where ferromagnetism is typically stabilized by the double-exchange mechanism, which requires delocalized electrons and leads to metallic conductivity Kanamori [2]. Stabilization of an insulating ferromagnetic state requires a subtle interplay of charge and/or orbital ordering, superexchange interactions, and lattice degrees of freedom Tokura and Nagaosa [3]; Tokura [4]. Despite their rarity, ferromagnetic insulators are of great interest because of their ability to facilitate the transport of pure spin currents with minimal charge flow, a prerequisite for next-generation spintronic technologies, including quantum information processing and energy-efficient devices Hellman et al. [5]; Tannous and Comstock [6]; Wang et al. (2017); Deng et al. (2018).

In particular, Lanthanum cobalt oxide (LaCoO3_3) has attracted significant attention due to its unusual magnetic properties. In its bulk form, LaCoO3_3 is a diamagnetic insulator at low temperatures, with all cobalt ions (Co3+^{3+}) in the low-spin (LS) state (t2g6eg0t_{2g}^{6}e_g^{0}, S=0S = 0) Asai et al. (1994). Transitions from the LS state to either the intermediate-spin (IS) (t2g5eg1t_{2g}^{5}e_g^{1}, S=1S = 1) or high-spin (HS) (t2g4eg2t_{2g}^{4}e_g^{2}, S=2S = 2) state can be driven by changes in temperature Asai et al. (1989), defects Hansteen et al. (1998), or external pressure Panfilov et al. (2018). These transitions are intimately linked to changes in the crystal structure and the volume of the CoO6_6 octahedra, which modify the balance between Hund’s exchange energy (ΔEX\Delta E_{\mathrm{X}}) and the crystal-field splitting (ΔCF\Delta C_{\mathrm{F}}) that governs the Co spin states.

Interestingly, when grown as thin films on substrates that induce tensile strain, such as (LaAlO3_3)0.3_{0.3}(Sr2_2TaAlO6_6)0.7_{0.7} (LSAT) or SrTiO3_3 (STO), LaCoO3_3 (Ghazisaeidi) exhibits ferromagnetic insulating behavior with Curie temperatures TCT_{\mathrm{C}} in the range of 65–80 K Li et al. (2018); Liu et al. (2021); Shin et al. (2022). In contrast, films grown on substrates that impose compressive strain, such as LaAlO3\mathrm{LaAlO}_3, remain nonmagnetic or weakly paramagnetic down to low temperatures Choi et al. (2012). It is generally assumed that tensile strain modifies the distribution of spin states at low temperature, but a satisfactory explanation of how that translates into an insulating ferromagnetic ground state remains elusive. In addition to epitaxial strain, oxygen vacancies have frequently been invoked as a contributing factor, since oxygen deficiency can stabilize Co2+\mathrm{Co}^{2+} ions and promote ferromagnetic coupling through mixed-valence interactions. This has led to interpretations in which the observed magnetism arises from oxygen-vacancy–driven phases or from an interplay between strain and non-stoichiometry. However, ferromagnetic insulating behavior has been reported in films, where electron energy loss spectroscopy detected no evidence of Co2+\mathrm{Co}^{2+} ions, thereby ruling out the presence of oxygen vacancies Choi et al. (2012); Yoon et al. (2021). Thus it would seem that epitaxial strain alone is sufficient to stabilize a ferromagnetic insulating phase.

Several fundamental issues regarding the origins of ferromagnetism in LaCoO3\mathrm{LaCoO}_3 films remain unresolved. How does the imposition of tensile strain affect the structure, in particular the Co–O bond distances that are intimately tied to the spin state? What mixture of spin-states (low, intermediate, or high-spin) is present and what is the ordered pattern of those spin states. Studies that invoke rocksalt-type ordering of HS and LS Co3+\mathrm{Co}^{3+} ions Seo et al. (2012); Chen et al. (2023); Li et al. (2023) are difficult to reconcile with conventional superexchange considerations Anderson (1959), which generally predict antiferromagnetic coupling through nearly linear Co(HS)–O–Co(LS)–O–Co(HS) bonds.

Given the difficulties associated with interrogating the structures of thin films at low temperatures, first-principles studies can offer important insights into the origins of ferromagnetism in LaCoO3\mathrm{LaCoO}_3 films. However, existing first-principles studies still pose important limitations Yoon et al. (2021); Seo et al. (2012); Rondinelli and Spaldin (2009); Meng et al. (2018); Geisler and Pentcheva (2020). Many employ simulation cells that are too small to fully capture the complexities of octahedral tilting and exclude potentially viable patterns of spin-state ordering. Moreover, the common practice of applying a uniform Hubbard UeffU_{\mathrm{eff}} to all Co sites may obscure important site-dependent variations in electronic structure and spin state. These constraints have hindered a clear identification of the magnetic ground state and the dominant exchange mechanisms.

In this work, we focus exclusively on the role of epitaxial strain and investigate the magnetic ground state of stoichiometric LaCoO3\mathrm{LaCoO}_{3} using density functional theory calculations, with sufficiently large supercells, under tensile strain equivalent to that imposed by an STO substrate. We first determine the strain-stabilized magnetic ground state and then analyze its electronic structure. Next, we extract the magnetic coupling parameters to quantitatively characterize the magnetic interactions. Finally, we examine the superexchange pathways responsible for the magnetic ordering and provide a theoretical explanation for the observed behavior.

Computational Methods

Density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP) Kresse and Furthmüller (1996a,b). The projector augmented-wave (PAW) method was employed. Structural relaxations were carried out using the meta–generalized gradient approximation (meta-GGA) exchange–correlation functional in the regularized-restored strongly constrained and appropriately normed (r2r^{2}SCAN) formulation Furness et al. (2020). For density of states (DOS) calculations, on-site electron correlations were treated within the DFT+UU formalism (r2r^{2}SCAN+UU) using the Dudarev approach Dudarev et al. (1998), with an effective Hubbard parameter Ueff=U−J=1U_{\mathrm{eff}} = U - J = 1 eV applied to the Co 3d3d states. A plane-wave kinetic energy cutoff of 550 eV was used throughout all calculations. For bulk LaCoO3\mathrm{LaCoO}_{3}, the Brillouin zone was sampled using a 6×6×66 \times6 \times6 Monkhorst–Pack kk-point mesh. Supercell calculations were performed using a 1×1×41 \times1 \times4 Γ\Gamma-centered kk-point mesh. Structural relaxations were continued until the total energy converged to within 10−610^{-6} eV and the Hellmann–Feynman forces on each atom were smaller than 5 meV/Å. For the calculation of magnetic exchange coupling parameters, a stricter electronic convergence criterion of 10−810^{-8} eV was employed, and a 2×2×22 \times2 \times2 kk-point mesh was used for all magnetic configurations. The valence electron configurations used in the PAW potentials are as follows: La: 5p65d16s25p^{6}5d^{1}6s^{2}; Co: 3d74s23d^{7}4s^{2}; and O: 2s22p42s^{2}2p^{4}.

Results

Bulk LaCoO3\mathrm{LaCoO}_{3} has a rhombohedral structure with space group R3ˉcR\bar{3}c at room temperature Kobayashi et al. (2000), as shown in Figure 1(a). As a reference, the bulk properties of LaCoO3_3 are first calculated. The fully relaxed lattice parameters are summarized in Table 1. Figure 1(e) and Figure 1(f) show the calculated total DOS and the projected DOS (PDOS) for the Co ions, respectively. The results indicate an insulating ground state with a band gap of approximately 0.64 eV, in good agreement with previously reported values Chainani et al. (1992). Integration of the Co-projected DOS reveals that all Co ions adopt an LS configuration, characterized by fully occupied t2gt_{2g} orbitals and empty ege_g states. This result is consistent with previous experimental studies of bulk LaCoO3_3 at low temperatures Asai et al. (1994). Figure 1(b) illustrates the corresponding pseudocubic bulk unit cell containing eight Co ions and incorporating the characteristic a−a−a−a^{-}a^{-}a^{-} octahedral tilting pattern. We introduce this pseudocubic representation because it matches the crystallographic orientation of LaCoO3_3 films grown on the STO substrate. This unit cell will therefore serve as the structural reference for the strained-bulk calculations discussed in the following section.

Methodaa (Å)β\beta (°)rCo−Or_{\mathrm{Co-O}} (Å)θCo−O−Co\theta_{\mathrm{Co-O-Co}} (°)apca_{\mathrm{pc}} (Å)
DFT (This work)5.359460.9701.929163.233.817
Exp Bull and Knight (2016) @ 4.2 K5.3405860.9881.923163.13.8046
Exp Xu et al. (2001) @ 100 K5.378160.811.922164.13.8029

Table 1. Lattice parameters, bond lengths, bond angles, and pseudocubic lattice constant of bulk LaCoO3_3. Current DFT calculations agree well with experimental values.

To assess the effect of epitaxial strain on the spin-state transition of a single CoO6_6 octahedron, an in-plane biaxial tensile strain ranging from 0 to 4.0% was applied, and the total energies of both spin configurations were compared. Above a critical tensile strain of approximately 1.8%, the HS configuration becomes energetically favorable compared to the LS state (see Supporting Information). The results demonstrate that sufficiently large tensile strain alone can drive a spin-state transition of a single CoO6_6 octahedron. Moreover, in these calculations, the IS state was not considered, as biaxial strain within this range did not stabilize an IS configuration.

Based on this finding, a 3×3×13 \times3 \times1 supercell was constructed from the pseudocubic unit cell, as shown in Figure 2(a). This supercell size provides sufficient degrees of freedom to accommodate different spin configurations and magnetic interactions while keeping the computational cost manageable. In particular, a larger in-plane cell allows a finite concentration of high-spin Co ions to interact without becoming unrealistically dilute. Strained LaCoO3_3 was then modeled under epitaxial conditions corresponding to an STO substrate by fixing the in-plane lattice constants along the [100] and [010] directions to the STO lattice parameter (a=3.905a = 3.905 Å), which corresponds to a tensile strain of approximately 2.4% relative to bulk LaCoO3_3. The out-of-plane lattice constant was fully relaxed to accommodate the imposed biaxial strain.

Selected magnetic configurations of the 3×3×1 supercell of pseudocubic LaCoO₃

Figure 2. Selected magnetic configurations of the 3×3×13 \times3 \times1 supercell of pseudocubic LaCoO3_3. (a) An in-plane epitaxial strain of 2.4% was applied, with the out-of-plane lattice constant allowed to relax for all initial magnetic configurations. (b) The initial magnetic configurations of the previously presented models Yoon et al. (2021); Seo et al. (2012); Kwon et al. (2014), mapped onto our supercell size, along with the lowest-energy configuration from our calculations (configuration 4), are shown. The red, blue, and gray colors represent HS (↑), HS (↓), and LS Co ions, respectively. The ionic arrangement in the top layer is indicated by solid lines, and the bottom layer is shown using dashed lines. (c) The energy differences among the magnetic configurations indicate that, under epitaxial tensile strain, our magnetic configuration is lower in energy than the previously proposed models.

Crystal structure and electronic density of states (DOS) of bulk LaCoO$_3$

Figure 1. Crystal structure and electronic density of states (DOS) of bulk LaCoO3_3. (a) and (b) show the Rˉ3c\bar{R}3c crystal structure and its corresponding pseudocubic unit cell with octahedral tilting respectively. The DOS of bulk LaCoO3_3 reveals a band gap of 0.64 eV in (c). (d) shows the partial density of states (PDOS) of Co3+^{3+}. Integration of the PDOS confirms fully occupied t2gt_{2g} orbitals, consistent with a LS state.

This supercell allows for a broad range of magnetic configurations of the Co3+^{3+} ions. For each of the 36 Co ions in the first Co layer of the supercell, three possible initial spin states were considered: HS (↑), HS (↓), and LS states. To reduce the configuration space, only magnetic arrangements preserving symmetry along both the [100] and [010] in-plane directions were retained, as there is no physical justification for preferential spin-state ordering along a single in-plane axis. In addition, configurations inconsistent with established magnetic exchange interaction rules Goodenough (1958) were not included in the primary set of configurations considered here. Upon applying periodic boundary conditions in the in-plane directions, symmetry-equivalent configurations were eliminated. Considering the presence of the second Co layer along the out-of-plane direction of the supercell further reduced the number of physically distinct magnetic configurations to 95 (See supporting information).

Following full structural relaxations, a ferromagnetic ground state was identified, characterized by 2×22 \times2 rocksalt-type HS/LS regions separated by planes of LS Co ions (Figure 2(b)). The relative energies of the various configurations are presented in Figure 2(c), where the ground-state structure is energetically favored by at least 18.1 meV compared to other configurations reported in the literature Yoon et al. (2021); Seo et al. (2012); Kwon et al. (2014) which were also evaluated for comparison. Energy comparisons for additional configurations are provided in the Supporting Information. This ground state configuration exhibits an HS fraction of 22.2% (16 out of 72 Co ions), corresponding to a net magnetization of 0.88 μB\mu_{\mathrm{B}} per Co ion, and an out-of-plane lattice constant of c=3.77c = 3.77 Å, in agreement with previously reported experimental values (c≈3.76c \approx3.76–3.793.79 Å) Fan et al. (2023); Russell et al. (2026).

The ground state configuration was further analyzed to examine the local structural distortions, specifically bond distances and bond angles, and to confirm its insulating character. Figure 3(a) presents the top view of the ground state model, where HS CoO6_6 octahedra are highlighted in red and LS octahedra in blue. As shown, each 2×22 \times2 HS/LS region (indicated by red squares), referred to here as a ferromagnetic column, is separated by planes of LS Co3+^{3+} ions. We refer to this ground state configuration as the ferromagnetic columnar model. A closer inspection of the spin pattern reveals that two LS ions are always positioned between adjacent HS ions belonging to neighboring ferromagnetic columns in both the x and y directions. In this model, two distinct types of LS Co3+^{3+} ions emerge, differentiated by their local environments and octahedral volumes: LS(1)^{(1)} ions, positioned within the separating LS planes, and LS(2)^{(2)} ions, located within the rocksalt-type HS/LS columns.

Magnetic structure and octahedral volumes of the ferromagnetic columnar model

Figure 3. Magnetic structure and octahedral volumes of the ferromagnetic columnar model. (a) Ferromagnetic domains, composed of 2×22 \times2 rocksalt-type HS/LS Co3+\mathrm{Co}^{3+} ions, are separated by planes of LS ions. LS(1)\mathrm{LS}^{(1)} refers to the LS ions located in the separating LS planes (cyan), while LS(2)\mathrm{LS}^{(2)} denotes the LS ions within the columns (dark blue). (b) Deviations of the octahedral volumes from bulk values are shown for the top and bottom layers. The HS octahedra exhibit a larger volume increase compared to the LS octahedra.

Figure 3(b) presents the octahedral volume deviation from the bulk (Vb=9.57V_{\mathrm{b}} = 9.57 Å3^{3}) within the top and bottom layers of the film. The HS octahedra exhibit an approximate 8.5% increase in volume compared to the bulk, while the LS(1)^{(1)} and LS(2)^{(2)} octahedra show a more modest expansion of 3.0% and 0.4%, respectively. The average Co–O bond lengths and Co–O–Co bond angles for HS, LS(1)^{(1)}, and LS(2)^{(2)} Co3+^{3+} ions are summarized in Table 2. The HS CoO6_6 octahedra exhibit an in-plane Co–O bond elongation of approximately 4.1%, whereas the in-plane bond lengths of the two types of LS (LS(1)LS^{(1)} and LS(2)LS^{(2)}) octahedra increase by about 1.5% and remain nearly unchanged, respectively. Correspondingly, the octahedral volumes and in-plane Co-O bond lengths of HS Co3+^{3+} ions are noticeably larger than those of LS ions, consistent with the occupancy of half-filled ege_g orbitals inherent to the HS configuration. The in-plane Co-O-Co bond angles change only slightly relative to the bulk structure, and the HS-O-LS(1)^{(1)} and HS-O-LS(2)^{(2)} angles are essentially identical; therefore, the spin state configuration is primarily determined by bond lengths rather than bond angles.

Spin StateVˉocta\bar{V}_{\mathrm{octa}} (Å3^3)rˉCo−OX\bar{r}^{X}_{\mathrm{Co-O}} (Å)rˉCo−OY\bar{r}^{Y}_{\mathrm{Co-O}} (Å)rˉCo−OZ\bar{r}^{Z}_{\mathrm{Co-O}} (Å)
Bulk (LSLS)9.571.931.931.93
LS(1)LS^{(1)}9.861.961.961.92
LS(2)LS^{(2)}9.611.931.931.90
HSHS10.392.012.011.92
DirectionθˉHS−O−LS\bar{\theta}_{HS-O-LS} (°)θˉLS−O−LS\bar{\theta}_{LS-O-LS} (°)θLS−O−LSBulk\theta^{Bulk}_{LS-O-LS} (°)
In-plane (xyxy)165.21163.59163.23
Out-of-plane (zz)159.94155.82163.23

Table 2. Average structural parameters of CoO6_6 octahedra in the ground state configuration. HS Co ions exhibit larger octahedral volumes and longer in-plane Co–O bond lengths compared to LS(1)^{(1)} ions (in the LS planes) and LS(2)^{(2)} ions (within the columns), while their out-of-plane bond lengths remain nearly unchanged. The average in-plane Co–O–Co bond angles are comparable to those in bulk, whereas the out-of-plane angles are noticeably reduced. The HS–O–LS(1)^{(1)} and HS–O–LS(2)^{(2)} bond angles are nearly identical, so they are not differentiated in this table.

To assess the electronic structure, the density of states (DOS) was calculated for the ferromagnetic columnar model, as shown in Figure 4(a). The total DOS exhibits a band gap of approximately 0.98 eV, confirming the insulating nature of this model. The projected DOS (Figure 4(b)) further shows distinct orbital occupancies for LS(1)\mathrm{LS}^{(1)}, LS(2)\mathrm{LS}^{(2)}, and HS Co ions, illustrating the characteristic splitting of t2gt_{2g} and ege_g orbitals. Integration of the PDOS confirms that all LS Co ions have fully occupied t2gt_{2g} orbitals and empty ege_g orbitals, while the HS Co ions exhibit partial ege_g occupancy, providing quantitative verification of the spin state differentiation. Overall, the ferromagnetic columnar model is confirmed to be an insulating ferromagnet, with net magnetization and electronic properties consistent with experimental observations. These results establish the validity of this model as a representation of tensile-strained LaCoO3\mathrm{LaCoO}_{3} films and provide a foundation for analyzing the magnetic interactions that stabilize the ferromagnetic ordering.

Density of states and projected density of states for the ferromagnetic columnar model

Figure 4. Density of states (DOS) and projected density of states (PDOS) for the ferromagnetic columnar model. Integration of the PDOS quantitatively confirms the spin state assignments of the Co ions. (a) Total DOS confirms the insulating nature of the system, with a band gap of 0.98 eV. (b) PDOS for LS(1)^{(1)} ions, located in the separating LS planes, shows fully occupied t2gt_{2g} orbitals. PDOS for LS(2)^{(2)} ions, positioned within the rocksalt-type HS/LS columns, also exhibits fully occupied t2gt_{2g} orbitals. PDOS for HS Co ions reveals partially occupied ege_g orbitals, consistent with the HS state. The results further indicate a slight charge/spin disproportionation between LS(2)^{(2)} and HS Co ions within the column.

The ferromagnetic columnar model contains four distinct superexchange interactions, two within each column and two between neighboring columns, as shown in Figure 5-Figure 6. These are: (i) a 180∘180^{\circ} Co(HS)–O–Co(LS)–O–Co(HS) pathway with exchange coupling denoted by J0J_{0}, and (ii) a 90∘90^{\circ} Co(HS)–O–Co(LS)–O–Co(HS) pathway with exchange coupling denoted by J1J_{1}; both J0J_{0} and J1J_1 are within a FM column and are therefore mostly perpendicular to the substrate. In addition, (iii) a longer 180° Co(HS)–O–Co(LS)–O–Co(LS)–O–Co(HS) pathway with exchange coupling denoted by J2J_2, and (iv) a longer 90° L-shaped Co(HS)–O–Co(LS)–O–Co(LS)–O–Co(HS) pathway with exchange coupling denoted by J3J_3; both J2J_2 and J3J_3 are exchange couplings between FM columns and are therefore mostly in the plane parallel to the substrate. To quantify these magnetic couplings, the total energies of different spin configurations were mapped onto an Ising Hamiltonian of the form

Magnetic interactions within a column

Figure 5. Magnetic interactions within a column. (a) Antiferromagnetic superexchange pathway along a 180∘180^\circ bond involving a single LS Co ion situated between two HS Co ions. Only the dx2−y2d_{x^2-y^2} orbitals of the three Co atoms are shown; the oxygen pyp_y orbitals, which mediate the superexchange, are omitted for clarity. Each arrow between Co atoms denotes an effective hopping process that scales as tpd2t_{pd}^2, where tpdt_{pd} is the Co–O hybridization parameter. (b) Ferromagnetic superexchange pathway along a 90∘90^\circ bond, where two HS Co ions interact via a corner LS Co ion. The pathway involves hybridization through orthogonal ege_g orbitals (dx2−z2d_{x^2-z^2} and d3y2−r2d_{3y^2-r^2}) on the LS Co, coupled respectively to pxp_x and pyp_y orbitals on the bridging oxygen atoms (not shown). Hund’s coupling on the LS Co ion favors parallel spin alignment, stabilizing a ferromagnetic interaction between the HS Co ions.

Magnetic interactions between columns

Figure 6. Magnetic interactions between columns. (a) Antiferromagnetic superexchange pathway along a longer 180° bond, involving two intermediate LS Co ions between a pair of HS Co ions. The interaction is mediated by three consecutive Co–O–Co links and arises from a virtual hopping process of ege_g electrons, resulting in an antiferromagnetic interaction that scales as tpd2t_{pd}^{2}. (b) Ferromagnetic superexchange pathway along a 90° “L-shaped” bond connecting HS Co ions in neighboring columns via two LS Co ions. Three distinct paths contribute to this interaction, increasing its overall strength.

Structural model of CoO$_6$ octahedra

Figure 1 (PDF p. 10).

E=−(J0∑⟨i,j⟩0sisj+J1∑⟨i,j⟩1sisj+J2∑⟨i,j⟩2sisj+J3∑⟨i,j⟩3sisj),(1)E=-\left(J_0\sum_{\langle i,j\rangle_0}s_i s_j+J_1\sum_{\langle i,j\rangle_1}s_i s_j+J_2\sum_{\langle i,j\rangle_2}s_i s_j+J_3\sum_{\langle i,j\rangle_3}s_i s_j\right), \tag*{(1)}

where sis_i represents the Ising spin variable associated with each Co site. Assuming that all HS Co3+^{3+} ions are equivalent, Eq. (1) can be simplified to

ΔE=−S2∑i=03niJi,(2)\Delta E = -S^2 \sum_{i=0}^{3} n_i J_i, \tag*{(2)}

where SS is the spin of the HS Co3+^{3+} ion (S=2S = 2), nin_i denotes the number of corresponding superexchange interactions of type ii, and ΔE\Delta E is the total energy difference between the ferromagnetic ground state configuration and a magnetic configuration in which selected spins are flipped (see Supporting Information). By solving the resulting set of linear equations obtained from gbc multiple spin configurations, the values of the exchange parameters J0J_0, J1J_1, J2J_2, and J3J_3 were extracted. The resulting coupling constants are summarized in Table 3.

ParameterValue (meVmeV)Angle
J0J_0−5.754±0.327-5.754 \pm0.327180∘180^\circ
J1J_15.514±0.1065.514 \pm0.10690∘90^\circ
J2J_2−3.093±0.637-3.093 \pm0.637180∘180^\circ
J3J_32.736±0.1592.736 \pm0.15990∘90^\circ

Table 3. Calculated superexchange coupling parameters for the ferromagnetic columnar model. J0J_0 denotes the antiferromagnetic superexchange interaction along the out-of-plane (zz) direction, while J1J_1 represents the ferromagnetic 90∘90^\circ superexchange interaction; both couplings act within a column. J2J_2 corresponds to the longer-range antiferromagnetic 180∘180^\circ superexchange interaction, and J3J_3 to a ferromagnetic L-shaped superexchange interaction, both operating between neighboring columns. Consistent with the discussion above, the intra-column interactions (J0J_0 and J1J_1) are stronger than the inter-column interactions (J2J_2 and J3J_3).

To gain insight into the mechanism underlying ferromagnetic ordering in the ground state, we analyze the superexchange interactions in spatial arrangement of HS and LS Co ions revealed by DFT; see Figure 5-Figure 6. Our goal is to understand the signs (ferromagnetic or antiferromagnetic) and the relative strengths of the magnetic interactions arising from different superexchange pathways between HS ions.

We begin with the simplest case of a 180∘180^\circ Co(HS)–O–Co(LS)–O–Co(HS) superexchange pathway shown in Figure 5(a). The HS Co ion is in a t2g4eg2t_{2g}^{4}e_g^{2} configuration with S=2S = 2, while the LS Co ion is in a t2g6eg0t_{2g}^{6}e_g^{0} configuration with S=0S = 0. Superexchange arises from an ege_g electron hopping through oxygen pσp_\sigma orbitals, specifically the dx2−y2d_{x^2-y^2} orbitals along the xx-axis, which strongly overlap with the pyp_y orbitals of the bridging O atoms. (The relevant oxygen orbitals are omitted from the figures for clarity). Hopping of t2gt_{2g} electrons is generally much weaker due to the smaller hybridization with O pπp_\pi orbitals. In this case, it is completely suppressed by Pauli exclusion as the t2gt_{2g} orbitals of LS Co are fully occupied. The virtual hopping process involves a transient double occupancy of the dx2−y2d_{x^2-y^2} orbital on the LS Co ion, which is possible only when the spins on the two HS Co ions are antiparallel. This results in an antiferromagnetic interaction of the form J0S1⋅S2J_0\mathbf{S}_1 \cdot\mathbf{S}_2, where S1\mathbf{S}_1 and S2\mathbf{S}_2 denote the spin-2 moments of the HS Co ions.

This is essentially the same as conventional antiferromagnetic superexchange in a 180° bond between two transition metal (TM) ions separated by O, which scales as tpd4t_{pd}^{4}, where tpdt_{pd} denotes the TM–O hybridization matrix element. In the Co(HS)–O–Co(LS)–O–Co(HS) geometry considered here, however, the exchange interaction J0J_0 is smaller, scaling like tpd8t_{pd}^{8}. The associated energy denominators in perturbation theory depend on several material-specific parameters, including the charge transfer energy, the intra-orbital and inter-orbital Coulomb repulsions and the Hund’s coupling on the LS Co, and the crystal field splitting differences between HS and LS Co ions. A detailed discussion of how these parameters enter in the calculation can be found in Appendix A.

We note that extended 180° superexchange pathways involving Co(HS)-O-Co(LS)-O-Co(HS) are not present in the xy-plane of the “columnar structure" found in the DFT, but they are present along the z-axis. The same AFM interaction J0J_0 can be derived along the z-axis, following the discussion above, by considering the virtual hopping of electrons in d3z2−r2d_{3z^2-r^2} via the intervening O pzp_z orbitals.

We next turn to the 90° Co(HS)–O–Co(LS)–O–Co(HS) superexchange pathway shown in Figure 5(b). Here, we focus on the dx2−y2d_{x^2-y^2} orbitals of the two HS Co ions that couple to the ege_g orbitals of the intervening LS Co ion through an O pxp_x orbital along one leg and an O pyp_y orbital along the other. This geometry gives rise to two distinct virtual hopping processes, each critically dependent on the specific LS Co orbitals involved. One process favors AFM coupling, while the other promotes FM interactions. As we show next, the net effect of these competing contributions ultimately favors a FM superexchange in this 90° configuration.

(a) The first process involves the delocalization of antiparallel spins on the HS ions with an intermediate virtual state with a doubly occupied dx2−y2d_{x^2-y^2} orbital on the LS Co. The details of the energetics, specifically, the energy denominators involved in the perturbation theory, are relegated to Appendix A. This analysis closely parallels the 180° superexchange discussed above and results in an AFM interaction.

(b) The second process corresponds to the delocalization of parallel-spin electrons on the two HS Co ions. By Pauli exclusion, the intermediate state in the process necessarily involves two orthogonal ege_g orbitals on the corner LS Co ion. We choose these two orbitals to be specific linear combinations, dx2−z2d_{x^2-z^2} and d3y2−r2d_{3y^2-r^2}, that can hybridize with the O pxp_x and pyp_y orbitals on the two legs of the 90° bond geometry; see Figure 5(b). The key point here is that the energy of the intermediate state on the LS Co is lower by Hund’s coupling JHJ_{\mathrm{H}} relative to case (a). The reduced energy denominator in case (b) relative to (a) leads to a net FM superexchange.

A general expression for the net FM superexchange interaction J1S1⋅S2J_{1}\mathbf{S}_{1}\cdot\mathbf{S}_{2} with J1<0J_{1}<0 along a 90° pathway is derived in Appendix A. For α=JH/Ueff≲1\alpha=J_{\mathrm{H}}/U_{\mathrm{eff}}\lesssim1, this simplifies to

J1=−2α3J0.(3)J_{1}=-\frac{2\alpha}{3}J_{0}. \tag*{(3)}

Here J0J_{0} is the 180° AFM interaction derived above, and α\alpha is the ratio of the Hund’s coupling JHJ_{\mathrm{H}} on LS Co to Ueff=U′+2Δ/3U_{\mathrm{eff}}=U^{\prime}+2\Delta/3, where U′U^{\prime} is the inter-orbital Coulomb repulsions on LS Co and Δ\Delta the crystal field splitting difference between HS and LS Co ions. We have included a factor of 2 enhancement in the expression for J1J_{1} because, for the HS CO ions located at opposite corners of a square, there are two different 90° pathways through the two LS ions located on the other diagonal of the square.

It is well known from the classic work of Goodenough, Kanamori, and Anderson that 90° bonds, between two transition metal ions with oxygen at the corner, lead to FM superexchange Khomskii (2014). In that case, the Hund’s coupling between electrons in two orthogonal pσp_{\sigma} oxygen orbitals stabilizes FM exchange. Our analysis is very similar, with the following differences. (i) the Co-O-Co bonds are all 180°, but we have 90° Co(HS)–O–Co(LS)–O–Co(HS) bonds. (ii) This forces us to look at a linear combination of the standard ege_g orbitals on the LS Co atom, which can hybridize with the relevant pσp_{\sigma} O orbitals and thus with the HS Co’s. These are the dx2−z2d_{x^{2}-z^{2}} and d3y2−r2d_{3y^{2}-r^{2}} orbitals on LS Co in the analysis above. (iii) It is the Hund’s coupling between electrons in these two orthogonal ege_g orbitals on LS Co that stabilizes FM exchange.

The AFM coupling J0J_{0} along the zz-axis and the ferromagnetic exchange J1J_{1} in the xyxy, xzxz and yzyz planes describe the magnetic interactions within a column. We next need to look at the longer range couplings between the columns, before turning to the question of the magnetic ground state of the entire system.

To understand the interactions between columns, we need to look at superexchange pathways of the form Co(HS)–O–Co(LS)–O–Co(LS)–O–Co(HS). The two inter-column interactions that arise are (1) a 180° linear configuration, shown in Figure 6(a), and (2) an L-shaped 90° configuration, shown in Figure 6(b). The 180° interaction (Figure 6(a)) results in an AFM interaction J2>0J_2 > 0 between columns, which scales as tpd12t_{pd}^{12} weaker than the tpd8t_{pd}^{8} interactions within a column. The AFM nature of J2J_2 arises as before: Pauli exclusion implies electrons with parallel spins cannot occupy the same orbital in intermediate virtual states.

The L-shaped configuration (Figure 6(b)) leads to a FM interaction J3<0J_3 < 0. This is again stabilized by selecting two orthogonal ege_g orbitals on the corner LS Co ion, so that the Hund’s coupling energy gained by parallel spins reduces the energy denominator. Now, we must take into account a factor of three enhancement in the FM interaction because of three distinct exchange paths connecting the two HS Co ions in this geometry (the two L-shaped paths and a “zig-zag” path). Nevertheless, the FM ∣J3∣\lvert J_3\rvert will be smaller than AFM ∣J2∣\lvert J_2\rvert by a factor of α=JH/Ueff\alpha= J_H/U_{\mathrm{eff}}, as in the case of intra-column interactions.

We have checked using conjugate gradient minimization of a classical Heisenberg Hamiltonian with the HS Co ions arranged as given by DFT, and retaining all the interactions J0J_0 through J3J_3, that the the ground state has ferromagnetic long range order. The combination of the FM interactions J1J_1 within the columns and J3J_3 between columns, together with the multiplicity of FM neighbors, is able to stabilize FM long range order despite the presence of competing AFM interactions J0J_0 and J2J_2.

We conclude with a comment on analyzing this system at finite temperature and computing its transition temperature TcT_c. The difficulty here stems from the fact that the spin state of the Co ions is itself a function of temperature. Even in bulk LaCoO3\mathrm{LaCoO_3}, which has a diamagnetic ground state with all Co ions in the LS configuration (S=0S = 0), the spin-state population changes with temperature Goodenough (1958): as TT increases, the concentration of HS Co3+\mathrm{Co^{3+}} rises rapidly at the expense of LS Co3+\mathrm{Co^{3+}}, reaching a 50:50 mixture near 110 K. This evolution reflects the higher entropy of the S≠0S \ne0 states relative to the unique S=0S = 0 state. As a result, such spin-state changes complicate any finite-temperature analysis based solely on the T=0T = 0 spatial ordering of HS and LS Co ions.

Conclusions

In summary, a microscopic mechanism for the ferromagnetic insulating state in tensile-strained, stoichiometric LaCoO3_3 thin films has been established using first-principles calculations. By explicitly isolating the role of epitaxial strain, the analysis shows that sufficiently large tensile strain can drive a spin state transition of Co3+^{3+} ions from the low-spin to the high-spin state, even in the absence of oxygen vacancies. When strain corresponding to a SrTiO3_3 substrate is imposed, the energetically favored ground state consists of ferromagnetically aligned columnar regions with rocksalt-type high-spin/low-spin ordering, separated by planes of low-spin Co ions. Electronic structure calculations further confirm the insulating nature of this configuration. The calculated superexchange coupling parameters indicate that 180∘180^\circ Co–O–Co pathways through low-spin Co3+^{3+} ions are antiferromagnetic, whereas 90∘90^\circ pathways favor ferromagnetic interactions. Accounting for all such interactions together, ferromagnetic couplings are dominant, providing the energetic basis for the stabilization of the columnar magnetic ground state.

Overall, this work establishes epitaxial strain as a sufficient driving force for ferromagnetic insulating behavior in LaCoO3\mathrm{LaCoO}_{3} thin films and offers a consistent microscopic picture that reconciles electronic structure, lattice distortion, and magnetic interactions. These insights provide a foundation for strain-engineered control of spin states and magnetism in correlated oxide heterostructures.

Acknowledgments

This work was supported by the Center for Emergent Materials (CEM), a National Science Foundation MRSEC under NSF Award Number DMR-2011876. Computational resources were provided through the Ohio Supercomputer center.

Data Availability

The data supporting this study’s findings are available within the article.

Derivation of Superexchange Interactions

In this Appendix, we estimate the magnitudes and signs of the exchange couplings generated by the superexchange pathways discussed in the main text. We will consider the HS-LS-HS geometries shown in Figure 5(a) and Figure 5(b) where the two magnetic ions are HS Co (t2g4eg2t_{2g}^{4}e_{g}^{2}, spin S=2S = 2) and the intermediate ion is LS Co (t2g6eg0t_{2g}^{6}e_{g}^{0}, spin S=0S = 0). The ege_g electrons on HS Co hop into the empty ege_g orbitals of LS Co via oxygen pσp_{\sigma} orbitals (which are omitted in the Figures for clarity). We can ignore the hopping of the t2gt_{2g} electrons; these are generally weaker due to the smaller hopping amplitudes via oxygen π\pi orbitals, but in fact these completely suppressed due to Pauli blocking on the LS site.

We write the Hamiltonian as H=H0+VH = H_0 + V, where H0H_0 contains the on-site interaction energies, while the perturbation VV describes the Co–O hopping with amplitude tpdt_{pd}. We consider processes where the electron virtually moves from the HS Co to the LS Co ege_g manifold through oxygen, and then returns. To second order in VV, we can eliminate O to obtain an effective HS–LS hopping amplitude

t∼tpd2ΔCT,(4)t \sim\frac{t_{pd}^{2}}{\Delta_{\mathrm{CT}}}, \tag*{(4)}

where ΔCT\Delta_{\mathrm{CT}} is the charge-transfer energy. The HS–HS exchange interaction then arises in fourth order in this effective hopping tt, or equivalently eighth order in tpdt_{pd}.

Let us first look at the 180° pathway shown in Figure 5(a). To keep track of the virtual processes, it is convenient to label the relevant charge configurations by the occupancies on the three Co sites (HS–LS–HS) as shown in Table A.4. The ground state is denoted by the label S1 in this Table, while intermediate states that arise in the perturbation theory involve either one (states S2 and S3) or two electrons (in state S4) that have been transferred to the LS ege_g orbitals. In the 180° geometry, the electrons originating from the two HS ions must hop onto the same LS orbital. Thus, the doubly occupied LS state S4 can arise only from an antiparallel spin configuration on the HS sites. This, of course, is the microscopic origin of the antiferromagnetic exchange.

The energies of the intermediate states, measured relative to the ground state, for the 180° geometry are shown in Table A.4. Here UU is the intra-orbital Hubbard interaction, U′U' is the inter-orbital Coulomb energy, JHJ_H is the (ferromagnetic) Hund’s coupling and Δ\Delta is the difference between LS and HS crystal-field splittings.

StateHS Co3+^{3+}LS Co3+^{3+}HS Co3+^{3+}Energy
S1S1t2g4eg2t_{2g}^{4}e_{g}^{2}t2g6eg0t_{2g}^{6}e_{g}^{0}t2g4eg2t_{2g}^{4}e_{g}^{2}00
S2S2t2g4eg1t_{2g}^{4}e_{g}^{1}t2g6eg1t_{2g}^{6}e_{g}^{1}t2g4eg2t_{2g}^{4}e_{g}^{2}U′+JH+ΔU' + J_H + \Delta
S3S3t2g4eg2t_{2g}^{4}e_{g}^{2}t2g6eg1t_{2g}^{6}e_{g}^{1}t2g4eg1t_{2g}^{4}e_{g}^{1}U′+JH+ΔU' + J_H + \Delta
S4S4t2g4eg1t_{2g}^{4}e_{g}^{1}t2g6eg2t_{2g}^{6}e_{g}^{2}t2g4eg1t_{2g}^{4}e_{g}^{1}2U′+U+2JH+2Δ2U' + U + 2J_H + 2\Delta (↑↓\uparrow\downarrow)

Table A.4. Charge configurations distinguished by the occupancies on the HS–LS–HS ions and their energies for the 180° geometry in the AFM alignment. Here Δ\Delta is the difference between LS and HS crystal-field splittings, UU and U′U' are the intra-orbital and inter-orbital Coulomb interactions, respectively, and JHJ_H is Hund’s coupling.

To calculate the exchange coupling, we write the fourth-order correction to the energy of an (unperturbed) ground state ∣n(0)⟩|n^{(0)}\rangle in the standard Rayleigh–Schrödinger perturbation theory. This is given by

En(4)=∑k2,k3,k4Vnk4Vk4k3Vk3k2Vk2nEnk2Enk3Enk4−∑k2,k4∣Vnk4∣2∣Vnk2∣2Enk42Enk2−2∑k3,k4Vnk4Vk4k3Vk3nVnnEnk32Enk4+∑k4∣Vnk4∣2Vnn2Enk43.(5)E_n^{(4)} = \sum_{k_2,k_3,k_4}\frac{V_{nk_4}V_{k_4k_3}V_{k_3k_2}V_{k_2n}}{E_{nk_2}E_{nk_3}E_{nk_4}}-\sum_{k_2,k_4}\frac{\lvert V_{nk_4}\rvert^2\lvert V_{nk_2}\rvert^2}{E_{nk_4}^{2}E_{nk_2}}-2\sum_{k_3,k_4}\frac{V_{nk_4}V_{k_4k_3}V_{k_3n}V_{nn}}{E_{nk_3}^{2}E_{nk_4}}+\sum_{k_4}\frac{\lvert V_{nk_4}\rvert^2V_{nn}^{2}}{E_{nk_4}^{3}}. \tag*{(5)}

where we use the notation

Vnm≡⟨n(0)∣V∣m(0)⟩,Enm≡En(0)−Em(0),(6)V_{nm}\equiv\langle n^{(0)}|V|m^{(0)}\rangle,\qquad E_{nm}\equiv E_n^{(0)}-E_m^{(0)}, \tag*{(6)}

In our problem the perturbation VV describes HS–LS hopping, and Vnn=0V_{nn}=0 for the unperturbed charge configurations and thus Eq. (A.2) simplifies to

En(4)=∑k2,k3,k4Vnk4Vk4k3Vk3k2Vk2nEnk2Enk3Enk4−∑k2,k4∣Vnk4∣2∣Vnk2∣2Enk42Enk2.(7)E_n^{(4)} = \sum_{k_2,k_3,k_4}\frac{V_{nk_4}V_{k_4k_3}V_{k_3k_2}V_{k_2n}}{E_{nk_2}E_{nk_3}E_{nk_4}}-\sum_{k_2,k_4}\frac{\lvert V_{nk_4}\rvert^2\lvert V_{nk_2}\rvert^2}{E_{nk_4}^{2}E_{nk_2}}. \tag*{(7)}

For HS–LS–HS superexchange, each Vnm∼tV_{nm}\sim t whenever ∣m(0)⟩|m^{(0)}\rangle and ∣n(0)⟩|n^{(0)}\rangle differ by a single allowed hop. The important spin-dependent fourth-order processes are “closed loops” that start in the ground state configuration S1S1 of Table A.4, visit states S2S2 or S3S3, then pass through S4S4, back to S2S2 or S3S3, before returning to S1S1. There are four equivalent ways to realize such a loop – two choices for which HS electron hops first, and two choices for the return sequence – giving an overall multiplicity factor of 4.

StateHS Co3+^{3+}LS Co3+^{3+}HS Co3+^{3+}Energy
S1t2g4eg2t_{2g}^{4}e_{g}^{2}t2g6eg0t_{2g}^{6}e_{g}^{0}t2g4eg2t_{2g}^{4}e_{g}^{2}0
S2t2g4eg1t_{2g}^{4}e_{g}^{1}t2g6eg1t_{2g}^{6}e_{g}^{1}t2g4eg2t_{2g}^{4}e_{g}^{2}U′+JH+ΔU' + J_H + \Delta
S3t2g4eg2t_{2g}^{4}e_{g}^{2}t2g6eg1t_{2g}^{6}e_{g}^{1}t2g4eg1t_{2g}^{4}e_{g}^{1}U′+JH+ΔU' + J_H + \Delta
S4t2g4eg1t_{2g}^{4}e_{g}^{1}t2g6eg2t_{2g}^{6}e_{g}^{2}t2g4eg1t_{2g}^{4}e_{g}^{1}3U′+JH+2Δ3U' + J_H + 2\Delta (↑↑)
S4t2g4eg1t_{2g}^{4}e_{g}^{1}t2g6eg2t_{2g}^{6}e_{g}^{2}t2g4eg1t_{2g}^{4}e_{g}^{1}3U′+2JH+2Δ3U' + 2J_H + 2\Delta (↑↓)

Table A.5. Charge configurations and energies for the 90∘90^\circ geometry. In the two-electron intermediate state S4, the two transferred electrons occupy orthogonal LS ege_g orbitals and thus two spin configurations are possible, as indicated. Here Δ\Delta is the difference between LS and HS crystal-field splittings, U′U' is the inter-orbital Coulomb interaction, and JHJ_H is Hund’s coupling.

The second term in Eq. (A.4) is a product of two independent second-order processes and depends only on the S2 and S3 denominators, without ever involving S4. It is thus identical for the FM and AFM configurations of the two HS Co ions. Since the exchange coupling is obtained from the energy difference EFM−EAFME_{\mathrm{FM}} - E_{\mathrm{AFM}}, this second term does not contribute to the exchange.

In contrast, the first term contains the “closed loop" with the intermediate state S4 of the two-electron LS and is therefore sensitive to spin alignment. In the 180∘180^\circ pathway, the FM configuration cannot access S4 because double occupancy of the same LS orbital by two parallel-spin electrons is forbidden by the Pauli exclusion principle. Hence, the first term vanishes for the FM configuration, and the antiferromagnetic exchange J0J_0 is determined just by the first term evaluated in the AFM configuration. Using the intermediate state energies shown in Table A.4, we obtain

H=J0 S1⋅S2J0=4t4(U′+JH+Δ)2(3U′+4JH+2Δ)(8)H = J_0\, \mathbf{S}_1 \cdot\mathbf{S}_2 \\ J_0 = \frac{4t^4}{(U' + J_H + \Delta)^2(3U' + 4J_H + 2\Delta)} \tag*{(8)}

where we used U=U′+2JHU = U' + 2J_H Khomskii (2014) for the intra-orbital Coulomb repulsion.

Next, we turn to the 90∘90^\circ pathway; see Figure 5(b). In contrast to the 180∘180^\circ case, the S1 intermediate state now involves two electrons occupying orthogonal LS ege_g orbitals. Therefore, both spin configurations (↑↑\uparrow\uparrow) and (↑↓\uparrow\downarrow) are allowed, and consequently both FM and AFM alignments contribute to the first term in Eq. (A.4). As before, the second term in Eq. (A.4) depends only on the one-electron charge-transfer denominators and is identical for FM and AFM configurations; it therefore cancels when we take the difference EFM−EAFME_{\mathrm{FM}} - E_{\mathrm{AFM}} to extract the exchange coupling.

We summarize in Table A.5 the energies of the states S1 through S4 for the 90° case. The difference between FM and AFM alignments arises entirely from the energy of the two-electron intermediate state S4. For the 90° geometry, the Hund’s coupling lowers the energy of the parallel-spin configuration. This leads to EFM(4)<EAFM(4)E_{\mathrm{FM}}^{(4)} < E_{\mathrm{AFM}}^{(4)}, and hence the net exchange is ferromagnetic.

Using the energies listed in Table A.5, we find that the 90° exchange is

H=J1S1⋅S2,H = J_1 \mathbf{S}_1 \cdot\mathbf{S}_2,
J1=−2×4t4(U′+JH+Δ)2JH(3U′+2JH+2Δ)(3U′+JH+2Δ).(9)J_1 = -2 \times\frac{4t^4}{(U' + J_H + \Delta)^2}\frac{J_H}{(3U' + 2J_H + 2\Delta)(3U' + J_H + 2\Delta)}. \tag*{(9)}

Here the factor of 2 arises from the geometry of two HS ions located at diagonally opposite corners of a square, so that there are two different 90° pathways through LS ions (on the other diagonal of the square).

Let us finally look at the relative magnitudes of the 90° FM and 180° AFM exchange couplings. Using Eqs. (A.5)–(A.6), one obtains

J1=−2J0JH(3U′+4JH+2Δ)(3U′+2JH+2Δ)(3U′+JH+2Δ).(10)J_1 = -2J_0 \frac{J_H(3U' + 4J_H + 2\Delta)}{(3U' + 2J_H + 2\Delta)(3U' + J_H + 2\Delta)}. \tag*{(10)}

Using Ueff≡U′+2Δ/3U_{\mathrm{eff}} \equiv U' + 2\Delta/3 and defining α≡JH/Ueff\alpha\equiv J_H/U_{\mathrm{eff}}, we find

J1=−2J03α(1+4α/3)(1+2α/3)(1+α/3).(11)J_1 = -\frac{2J_0}{3}\frac{\alpha(1 + 4\alpha/3)}{(1 + 2\alpha/3)(1 + \alpha/3)}. \tag*{(11)}

To get a feel for this expression, we look at the regime α≪1\alpha\ll1, where we obtain

J1≈−23J0α+⋯(12)J_1 \approx-\frac{2}{3}J_0\alpha+ \cdots \tag*{(12)}

Thus, in the α≪1\alpha\ll1 limit, the 90° ferromagnetic exchange is smaller than the 180° antiferromagnetic exchange by a factor of α=JH/Ueff\alpha= J_H/U_{\mathrm{eff}}. We further find that this correction remains tiny even up to α∼1\alpha\sim1.

Supplementary Material

Strain Effects on a Single Co Ion

To examine the effect of epitaxial strain on the spin state of an individual Co ion, biaxial in-plane tensile strain in the range of 0–4% was applied to the LaCoO3\mathrm{LaCoO_3} unit cell. For each strain value, the total energies of the low spin (LS) and high spin (HS) configurations were calculated and compared, as shown in Supplementary Figure S1. The intermediate spin (IS) state was not considered, as the equivalence of the in-plane xx and yy directions under biaxial strain disfavors stabilization of an IS configuration. The results show that above approximately 1.8% epitaxial tensile strain, the HS Co3+\mathrm{Co^{3+}} ion becomes energetically favorable relative to the LS state. This behavior is consistent with a strain-induced reduction of the crystal field splitting, which eventually becomes smaller than Hund’s exchange coupling, thereby driving a spin-state transition. These results demonstrate that sufficiently large epitaxial tensile strain alone can induce a transition from the LS to the HS state.

Energy difference between LS and HS states of a single Co ion in LaCoO3 as a function of applied tensile strain

Figure S1. Energy difference between LS and HS states of a single Co ion in LaCoO3 as a function of applied tensile strain. The HS state becomes energetically favorable above 1.8% strain, demonstrating that epitaxial strain can induce a spin state transition. The intermediate spin (IS) state is not considered due to symmetry equivalence in the xx and yy directions.

Total Number of Configurations

To determine the total number of possible configurations, we begin with a 6×66 \times6 square lattice as illustrated in Supplementary Figure S2(a). Because we wish to preserve symmetry along both the X and Y directions, it is sufficient to consider only the 12 independent squares indicated in Supplementary Figure S2(b), which must be specified to define the full configuration.

Each site can be in one of three spin states: high spin state up (HS↑\mathrm{HS}\uparrow), high spin state down (HS↓\mathrm{HS}\downarrow), or low spin state (LS), yielding a total of 315=14,348,9073^{15} = 14{,}348{,}907 possible configurations.

Next, to identify the physically meaningful configurations, we impose the Goodenough–Kanamori rules. The two main rules are as follows:

  1. Parallel alignment of two neighboring HS ions is forbidden in both rows and columns.

  2. Diagonal neighbors must have the same spin sign.

To implement these rules systematically, we represent the three states numerically as −1-1 (HS↓\mathrm{HS}\downarrow), +1+1 (HS↑\mathrm{HS}\uparrow), and 0 (LS). Under this representation:

  • Rule 1 requires that the sum of any two neighboring spins in a row or column cannot be +2+2 or −2-2.

  • Rule 2 requires that diagonal neighbors must not sum to 0 unless at least one of them is LS (i.e., diagonal HS pairs of opposite spins are forbidden).

By applying these constraints to the initial set of configurations, the number of valid configurations is reduced to 3213 (comprising 3212 configurations plus the single configuration with all LS ions). However, because all 15 squares are equivalent under the periodic boundary conditions, and because both a configuration and its spin-inverted counterpart are included in this count, we divide by 30=2×1530 = 2 \times15 to account for these symmetries. Moreover, additional random simulations indicate that placing two HS ions adjacent to each other in a row or column is energetically unfavorable. This yields a total of 95 unique configurations. These configurations were generated by representing each combination as a 1×151 \times15 array. For example, one of the valid configurations is given by [1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, −1-1, 0, 0], as illustrated in Supplementary Figure S3. We also note that the magnetic arrangement in the second layer is directly related to that of the first layer in order to preserve the overall symmetry.

Square lattices illustrating the arrangement of Co ions

Figure S2. Square lattices illustrating the arrangement of Co ions. (a) A 6×66 \times6 square grid representing the Co ions in a single layer, with the irreducible unit highlighted in blue. (b) The irreducible grid is shown, with individual squares labeled from 1 to 15 for reference.

An example of a valid configuration shown both in numerical and corresponding spin-state formats

Figure S3. An example of a valid configuration shown both in the numerical format (−1-1, 0, 1) and in the corresponding spin states (HS↑\uparrow, HS↓\downarrow, and LS).

Energy Comparison of Magnetic Configurations

The relative total energies of the eight high-symmetry magnetic configurations in the 3×3×13 \times3 \times1 supercell were calculated and are shown in Supplementary Figure S4. These configurations were chosen because they preserve the highest possible symmetry of the supercell and are physically meaningful.

In total, 95 symmetry-inequivalent magnetic configurations were examined; however, all configurations not shown in Supplementary Figure S4 lie significantly higher in energy and do not compete with the ground state configuration. Therefore, only the energetically relevant high-symmetry configurations are discussed here.

Relative total energies of eight high-symmetry magnetic configurations

Figure S4. Relative total energies of eight high-symmetry magnetic configurations considered in the 3×3×13 \times3 \times1 supercell. The red, blue, and gray colors represent HS (↑\uparrow), HS (↓\downarrow), and LS Co ions, respectively. The ionic arrangement in the top layer is indicated by solid lines, and the bottom layer is shown using dashed lines.

ConfigurationEquation
Ground State8J0+40J1+8J2+40J38J_0 + 40J_1 + 8J_2 + 40J_3
1Flip 2 from one column4J0+20J1+4J2+24J34J_0 + 20J_1 + 4J_2 + 24J_3
2Flip 3 from one column2J0+22J1+2J2+10J32J_0 + 22J_1 + 2J_2 + 10J_3
3Flip 4 from each column−8J0+8J1+8J2+8J3-8J_0 + 8J_1 + 8J_2 + 8J_3
4Flip 1 from one and 2 from the other6J0+10J1+2J2+18J36J_0 + 10J_1 + 2J_2 + 18J_3
5Flip 1 from one and 5 from the other4J0+10J1−8J34J_0 + 10J_1 - 8J_3
6Flip 2 from one and 3 from the other−2J0+6J1+2J2+6J3-2J_0 + 6J_1 + 2J_2 + 6J_3
7Flip 3 from one and 4 from the other2J0−6J1+2J2+2J32J_0 - 6J_1 + 2J_2 + 2J_3

Table S1. Magnetic configurations considered for the extraction of superexchange coupling parameters and the corresponding energy equations. The coefficients nin_i in the energy expressions are obtained by subtracting the interaction counts of the ground state configuration from those of each respective magnetic configuration.

Superexchange Coupling Calculations

To determine all superexchange coupling parameters, the original 3×3×13 \times3 \times1 supercell is insufficient to capture the out-of-plane 180∘180^\circ superexchange interaction (J0J_0) due to the imposed periodic boundary conditions along the zz direction and the resulting self-interaction of the HS Co ions. To overcome this limitation, a smaller but equivalent supercell was derived from the ground state configuration, as shown in Supplementary Figure S5. In this supercell, the in-plane dimensions were reduced by a factor of two, while the out-of-plane dimension was doubled, enabling explicit treatment of the out-of-plane 180∘180^\circ superexchange pathway. In this supercell, two ferromagnetic columns are present.

Constructed supercell and magnetic superexchange interactions

Figure S5. Constructed supercell derived from the 3×3×13 \times3 \times1 ferromagnetic columnar model and the corresponding magnetic superexchange interactions. (a) The supercell is generated by halving the in-plane dimensions and doubling the out-of-plane dimension of the ground state configuration, allowing explicit treatment of out-of-plane exchange pathways. (b) Schematic illustration of the four distinct superexchange interactions. The intra-column interactions include the short-range out-of-plane 180∘180^\circ interaction (J0J_{0}) and the short-range 90∘90^\circ interaction (J1J_{1}). The inter-column interactions consist of the longer-range 180∘180^\circ interaction (J2J_{2}) and the L-shaped 90∘90^\circ interaction (J3J_{3}).

As illustrated in Supplementary Figure S5, four distinct superexchange interactions are present in this geometry. To extract these coupling parameters, the spin orientation of selected HS Co ions was flipped to generate seven additional magnetic configurations relative to the ground state, yielding a total of seven independent energy equations. The corresponding spin configurations and equations are summarized in Supplementary Table S1. The resulting overdetermined system of equations was solved using a least-squares fitting procedure, allowing all four superexchange coupling constants to be determined. The quality of the fit is demonstrated in Supplementary Figure S6.

Predicted $\Delta e$ versus actual $\Delta e$ for seven magnetic configurations

Figure S6. Accuracy of the least-squares fit used to extract the superexchange coupling parameters. The DFT-calculated energy differences for the seven magnetic configurations are plotted against the values predicted by the fitted Ising model. The excellent linear correlation, with a slope close to unity and negligible residuals, confirms the robustness and accuracy of the extracted exchange parameters with R2=0.9994R^2 = 0.9994.

References

  1. [1]Emilia Morosan, Douglas Natelson, Andriy H. Nevidomskyy, and Qimiao Si. Strongly Correlated Materials. Advanced Materials, 24(36):4896–4923, 2012. ISSN 1521-4095. doi: 10.1002/adma.201202018. URL https://onlinelibrary.wiley.com/doi/abs/10.1002/adma.201202018. _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1002/adma.201202018.DOI
  2. [2]Junjiro Kanamori. Superexchange interaction and symmetry properties of electron orbitals. Journal of Physics and Chemistry of Solids, 10(2):87–98, July 1959. ISSN 0022-3697. doi: 10.1016/0022-3697(59)90061-7. URL https://www.sciencedirect.com/science/article/pii/0022369759900617.DOI
  3. [3]Y. Tokura and N. Nagaosa. Orbital Physics in Transition-Metal Oxides. Science, 288(5465):462–468, April 2000. doi: 10.1126/science.288.5465.462. URL https://www.science.org/doi/10.1126/science.288.5465.462. Publisher: American Association for the Advancement of Science.DOI
  4. [4]Yoshinori Tokura. Correlated-Electron Physics in Transition-Metal Oxides. Physics Today, 56(7):50–55, July 2003. ISSN 0031-9228. doi: 10.1063/1.1603080. URL https://doi.org/10.1063/1.1603080.
  5. [5]Frances Hellman, Axel Hoffmann, Yaroslav Tserkovnyak, Geoffrey S. D. Beach, Eric E. Fullerton, Chris Leighton, Allan H. MacDonald, Daniel C. Ralph, Dario A. Arena, Hermann A. Dürr, Peter Fischer, Julie Grollier, Joseph P. Heremans, Tomas Jungwirth, Alexey V. Kimmel, Bert Koopmans, Ilya N. Krivorotov, Steven J. May, Amanda K. Petford-Long, James M. Rondinelli, Nitin Samarth, Ivan K. Schuller, Andrei N. Slavín, Mark D. Stiles, Oleg Tchernyshyov, André Thiaville, and Barry L. Zink. Interface-induced phenomena in magnetism. Reviews of Modern Physics, 89(2):025006, June 2017. doi: 10.1103/RevModPhys.89.025006. URL https://link.aps.org/doi/10.1103/RevModPhys.89.025006. Publisher: American Physical Society.DOI
  6. [6]Charbel Tannous and R. Lawrence Comstock. Magnetic Information-Storage Materials. In Safa Kasap and Peter Capper, editors, Springer Handbook of Electronic and Photonic Materials, pages 1–1. Springer International Publishing, Cham, 2017. ISBN 978-3-319-48933-9.DOI

Paper details

vlnc:W_uyngerwt
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents