betaFeedback

WorksPhysics and Astronomy

First-principles study of spin orbit coupling contribution to anisotropic magnetic interaction

Abstract

Anisotropic magnetic exchange interactions lead to a surprisingly rich variety of the magnetic properties. Considering the spin orbit coupling (SOC) as perturbation, we extract the general expression of a bilinear spin Hamiltonian, including isotropic exchange interaction, antisymmetric Dzyaloshinskii-Moriya (DM) interaction and symmetric ΓΓ term. Though it is commonly believed that the magnitude of the DM and ΓΓ interaction correspond to the first and second order of SOC strength % λ respectively, we clarify that the term proportional to λ2λ^{2} also has contribution to DM interaction. Based on combining magnetic force theorem and linear-response approach, we have presented the method of calculating anisotropic magnetic interactions, which now has been implemented in the open source software WienJ. Furthermore, we introduce another method which could calculate the first and second order SOC contribution to the DM interaction separately, and overcome some shortcomings of previous methods. Our methods are successfully applied to several typical weak ferromagnets for 3d3d, 4d4d and 5d5d transition metal oxides. We also predict the conditions where the DM interactions proportional to λλ are symmetrically forbidden while the DM interactions proportional to λ2λ^{2} are nonzero, and believe that it is widespread in certain magnetic materials.

Introduction

Magnetic properties can be typically described by a quadratic spin Hamiltonian, which is the basis of most magnetic theoretical investigations [1–3]. Generally, spin-orbit coupling (SOC) always exists and leads to the anisotropic magnetic interactions with low symmetry. The general form of the bilinear expression of a spin exchange Hamiltonian could be written as

H=∑i<jJijSi⋅Sj+∑i<jDij⋅[Si×Sj]+∑i<jSi⋅Γij⋅Sj(1)H = \sum_{i<j} J_{ij}\mathbf{S}_i \cdot\mathbf{S}_j + \sum_{i<j} \mathbf{D}_{ij} \cdot[\mathbf{S}_i \times\mathbf{S}_j] + \sum_{i<j} \mathbf{S}_i \cdot\mathbf{\Gamma}_{ij} \cdot\mathbf{S}_j \tag*{(1)}

where the first term describes the isotropic Heisenberg Hamiltonian, the second one represents the Dzyaloshinskii-Moriya (DM) [4–6] interaction, and the third one is marked as Γ\Gamma term [5]. The antisymmetric DM interaction, which comes from the combination of low symmetry and SOC, is introduced by Dzyaloshinskii [4] and Moriya [5] in a phenomenological model and a microscopic model respectively. It is commonly believed that DM and Γ\Gamma term are contributed by first and second order of SOC respectively [1, 5]. Generally, DM interaction favors twisted spin structures and is constrained by the crystal symmetry. For example, when a inversion center located at the bond center of two magnetic ion sites, the DM interaction between these two magnetic ions should be zero due to its antisymmetric property [5, 6]. Now the DM interaction is invoked to explain numerous interesting magnetic systems featuring non-collinear spin textures, such as weak ferromagnets [4, 5], helimagnets [7], skyrmion formation [8–10] and chiral domain walls [11, 12]. In addition, the DM interaction also plays an important role in multiferroic materials [13–17], topological magnon materials [18–20] and spintronics [21]. It is worth mentioning that, since DM interactions are very sensitive to small atomic displacements and symmetry restrictions, it can also be used to reveal the interplay of delicate structural distortions and complex magnetic configurations [22].

Recently, the first-principles study of magnetic exchange interactions especially DM interaction, has also attracted much interest [34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50] [3, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33]. A popular numerical method is the energy-mapping analysis [3, 23, 24] to estimate magnetic interactions from the energy differences of various magnetic structures. However, this approach becomes inconvenient for the complicated systems where it is not clear how many exchange interactions needs to be considered, since in some magnetic compounds the magnetic moments may couple over a variety of distances, and even the ninth-nearest-neighbor coupling plays an important role [25, 26]. Meanwhile, in itinerant magnetic systems, the magnetism is not so localized and the calculated magnetic moments may depend on the magnetic configurations, which also significantly affects the accuracy of the calculated DM interactions. Another approach using total energy differences could extract DM strength by directly calculating the energies of spirals with the finite vector qq [27–32]. Meanwhile, an efficient approach is proposed based on the magnetic force theorem [34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46] [33]. Katsnelson et al. Katsnelson et al. [35] have derived the expression for DM interaction term based on Green's function approach. This method is applied to a large number of magnetic materials such as the antiferromagnets with weak ferromagnetism [36], thin magnetic films [38], diluted magnetic semiconductors [39] and various other magnetic materials [41, 42, 43, 44, 45]. This Green’s function approach was previously formulated in first-principles codes with direct definition of a localized orbitals basis set such as linear muffin-tin orbitals method [51]. Furthermore, they have also developed the method of calculating DM interactions using Wannier function formalism [37, 40]. In addition, DM interactions could also be estimated by computing the long-wave length limit of the spin susceptibility [47], the expectation value of the spin current density [48, 49], or utilizing Berry phase [50].

In this paper, up to the second-order perturbation of SOC, we revisit the general expression of anisotropic magnetic interactions. We clarify that the second order term of SOC has contribution to both the antisymmetric DM interaction and symmetric Γ\Gamma interaction, and reveal that their distinction arises from different hopping processes as shown in Figure 1 and following. Note that in these approaches using total energy differences, such as energy-mapping [3, 23, 24], spirals approach [28, 29] and the approaches through calculating energy variations due to spin rotations [35, 52, 53], can get the entire DM interaction but do not distinguish the contribution from first or second order of SOC. On the other hand, one can only consider DM interactions with the first order of SOC in their perturbation schemes [36, 49]. We extend the method of calculating Heisenberg interactions based on combining magnetic force theorem and linear-response approach [52, 53, 54, 55] [25, 33] to estimate DM and Γ\Gamma interactions, and the algorithm of our proposed method is now implemented in the open source called WienJ [56], as an interface to the linearized augmented plane wave (LAPW) software Wien2k [57]. Furthermore, to overcome some shortcomings of previous methods, we develop a new method that can estimate the first and second order SOC contribution to the DM exchange couplings separately. While our methods can also calculate Γ\Gamma interaction, here we only present the results of Heisenberg and DM interactions since they could be compared with many previous works. We have applied our methods to several representatives of canted antiferromagnetic materials La2_2CuO4_4, Ca2_2RuO4_4 and Ca3_3LiOsO6_6 for 3d, 4d and 5d transition metal oxides, and the calculation results are consistent with the experiment. Particularly, we find that the DM interaction proportional to λ2\lambda^2 can not be ignored in 4d transition metal oxide Ca2_2RuO4_4, and the DM interactions proportional to λ\lambda and λ2\lambda^2 have the same magnitude in 5d transition metal oxide Ca3_3LiOsO6_6. As shown in the following, the DM interactions proportional to λ\lambda and λ2\lambda^2 involve different exchange channels. Thus, based on the symmetry analysis, we explore the possibility that the DM interactions proportional to λ\lambda are symmetrically forbidden while the DM interactions proportional to λ2\lambda^2 still exist. We believe that this case is widespread in certain magnetic materials, and our method would play more important role in these magnetic systems.

Schematic pictures of exchange paths for anisotropic magnetic interactions between site A and site B

Figure 1. Schematic pictures of exchange paths for anisotropic magnetic interactions between site A and site B. The dotted line represents the SOC excitation process, while the solid line represents the hopping process. (a) represents the DM interactions for the first order of SOC. The DM interactions for the second order of SOC have two types of perturbation processes (b) and (c). Meanwhile, the perturbation processes for the symmetric Γ\Gamma term are also shown in (d) and (e) for comparison. Here GS and ES represent ground state and excited state respectively. It is worth mentioning that, the perturbation processes of DM interactions involve the hopping between GSs, which denoted by the green solid line. Meanwhile, Γ\Gamma terms would only involve the hopping processes between GS and ES.

Method

Anisotropic magnetic interactions by perturbation theory

We start from an effective model:

H=H0+Ht+HU+Hsoc=∑iασϵαciασ†ciασ+∑ijαβσtαβijciασ†cjβσ+U∑ini,↑ni,↓+∑iλli⋅si(2)H = H_0 + H_t + H_U + H_{\mathrm{soc}} = \sum_{i\alpha\sigma} \epsilon_\alpha c^\dagger_{i\alpha\sigma} c_{i\alpha\sigma} + \sum_{ij\alpha\beta\sigma} t^{ij}_{\alpha\beta} c^\dagger_{i\alpha\sigma} c_{j\beta\sigma} + U \sum_i n_{i,\uparrow} n_{i,\downarrow} + \sum_i \lambda\mathbf{l}_i \cdot\mathbf{s}_i \tag*{(2)}

where H0H_0, HtH_t, HUH_U and HsocH_{\mathrm{soc}} represent the on-site orbital energy, the hopping term, the Hubbard UU term and the SOC term, respectively. Here i,ji,j represent the site index, while α,β\alpha,\beta represent the orbital index and σ\sigma represents the spin index. We consider the spin exchange interaction between the magnetic ions located at the site AA and site BB. We label the ground state and the unoccupied states at site AA as nn and mm, respectively. Similarly, the ground state and the excited states at site BB are labeled as n′n' and m′m', respectively.

When SOC is not considered, the Heisenberg interactions Heff=JSA⋅SBH_{eff}=J\mathbf{S}_A\cdot\mathbf{S}_B can be obtained by considering the hopping term as perturbations for the case of U>>tU >> t [58]. Considering the SOC term λl⋅s\lambda\mathbf{l}\cdot\mathbf{s} as perturbation, the first-order SOC contribution to effective spin model Heff(1)H_{eff}^{(1)} has the expression of antisymmetric DM interaction as Heff(1)=D(1)(SA×SB)H_{eff}^{(1)}=\mathbf{D}^{(1)}(\mathbf{S}_A\times\mathbf{S}_B) where D(1)\mathbf{D}^{(1)} could be written as [5]

(Dα)(1)=−4iλtnn′αU(∑mlmnαϵm−ϵntmn′−∑m′lm′n′αϵm′−ϵn′tm′n)(3)(D^\alpha)^{(1)} = -\frac{4i\lambda t_{nn'}^\alpha}{U}\left(\sum_m \frac{l^\alpha_{mn}}{\epsilon_m-\epsilon_n}t_{mn'}-\sum_{m'}\frac{l^\alpha_{m'n'}}{\epsilon_{m'}-\epsilon_{n'}}t_{m'n}\right) \tag*{(3)}

Meanwhile, by considering perturbation theory up to the second-order SOC correction (see details in Appendix A), we find that the second-order SOC correction has the contribution to both DM term D(2)\mathbf{D}^{(2)} and the Γ\Gamma term, where D(2)\mathbf{D}^{(2)} could be written as

(Dα)(2)=2λ2tnn′αU∑m,m′lm′n′βlmnγ−lm′n′γlmnβ(ϵm′−ϵn′)(ϵm−ϵn)tmm′−2λ2tnn′αU∑m1,m2lm1nβlm2nγ−lm1nγlm2nβ(ϵm1−ϵn)(ϵm2−ϵn)tm1m2+2λ2tnn′αU∑m1′,m2′lm1′n′βlm2′n′γ−lm1′n′γlm2′n′β(ϵm1′−ϵn′)(ϵm2′−ϵn′)tm1′m2′(4)(D^\alpha)^{(2)} = 2\frac{\lambda^2 t_{nn'}^\alpha}{U}\sum_{m,m'}\frac{l^\beta_{m'n'}l^\gamma_{mn}-l^\gamma_{m'n'}l^\beta_{mn}}{(\epsilon_{m'}-\epsilon_{n'})(\epsilon_m-\epsilon_n)}t_{mm'} -2\frac{\lambda^2 t_{nn'}^\alpha}{U}\sum_{m_1,m_2}\frac{l^\beta_{m_1n}l^\gamma_{m_2n}-l^\gamma_{m_1n}l^\beta_{m_2n}}{(\epsilon_{m_1}-\epsilon_n)(\epsilon_{m_2}-\epsilon_n)}t_{m_1m_2} +2\frac{\lambda^2 t_{nn'}^\alpha}{U}\sum_{m'_1,m'_2}\frac{l^\beta_{m'_1n'}l^\gamma_{m'_2n'}-l^\gamma_{m'_1n'}l^\beta_{m'_2n'}}{(\epsilon_{m'_1}-\epsilon_{n'})(\epsilon_{m'_2}-\epsilon_{n'})}t_{m'_1m'_2} \tag*{(4)}

Meanwhile, the expression of parameter Γ(2)\Gamma^{(2)} could be seen in Eq. (12) of Appendix.

Here we present schematic pictures of the exchange processes in Figure 1. It is easy to see that, the bilinear spin exchange Hamiltonian should contain two hopping processes between two sites A and B as shown in Figure 1. Considering up to the second-order perturbation of SOC, we find that there are several different exchange processes. Among them, the first-order SOC correction has only DM contribution D(1)D^{(1)}, as shown in Figure 1(a), which represents the first term of Eq. (3). When swap the sites A and B in Figure 1(a), one can obtain the second term of Eq. (3). Meanwhile, the second-order SOC correction has not only the contribution to DM interaction, but also the contribution to Γ\Gamma term as shown in Figure 1(b)-(Figure 1e). While the type-I D(2)D^{(2)} as shown in Figure 1(b) represents the first term of Eq. (4), the type-II D(2)D^{(2)} as shown in Figure 1(c) represents the second term of Eq. (4), and the third term of Eq. (4) could be obtained by swapping the sites A and B. It is worth mentioning that,

the exchange processes for DM interactions (Figure 1 (a)-(Figure 1c)) involve the hopping between ground states, which denoted by the green solid line in Figure 1 and tnn′t_{nn'} in Eq. (3)(4) respectively. In sharp contrast, there are only hoppings between ground states and excited states in the processes of Γ\Gamma terms as shown in Figure 1 (d)-(Figure 1e). Therefore, we emphasize that the essential difference between DM and Γ\Gamma term is from their different hopping processes, rather than the commonly believed different orders of SOC [1, 5].

Magnetic interactions in the first-principles approach

Firstly we present the method to calculate magnetic interactions based on the force theorem and linear-response approach [52, 53] [33], which could be written as the following form [52]

JRl+τ,Rl′+τ′αβ=∑nkn′k′fnk−fn′k′εnk−εn′k′⟨ψnk∣[σ×Bτ]α∣ψn′k′⟩⟨ψn′k′∣[σ×Bτ′]β∣ψnk⟩ei(k′−k)(Rl−Rl′)(5)J^{\alpha\beta}_{\mathbf{R}_{l}+\tau,\mathbf{R}_{l'}+\tau'}=\sum_{nkn'k'}\frac{f_{nk}-f_{n'k'}}{\varepsilon_{nk}-\varepsilon_{n'k'}}\langle\psi_{nk}|[\boldsymbol{\sigma}\times\mathbf{B}_{\tau}]_{\alpha}|\psi_{n'k'}\rangle\langle\psi_{n'k'}|[\boldsymbol{\sigma}\times\mathbf{B}_{\tau'}]_{\beta}|\psi_{nk}\rangle e^{i(\mathbf{k}'-\mathbf{k})(\mathbf{R}_{l}-\mathbf{R}_{l'})} \tag*{(5)}

This method has been successfully applied to calculate Heisenberg interactions in various magnetic materials [52, 53, 54, 55] [25, 33]. Considering the case of α≠β\alpha\ne\beta, we extend this method to estimate DM and Γ\Gamma interactions, and the algorithm of this method is now implemented in the open source called WienJ [56], as an interface to Wien2k [57]. It is worth mentioning that, the general expression of bilinear spin exchange parameter JαβJ^{\alpha\beta}, which could be written in JJ, DD and Γ\Gamma as Eq. (1), has 9 independent components. However, one can only yield 4 out of 9 components of JαβJ^{\alpha\beta} for a given magnetic configuration. For example, for the collinear magnetic configuration with all spin moments lying along the zz-axis, only the four spin exchange parameters JxxJ^{xx}, JyyJ^{yy}, JxyJ^{xy} and JyxJ^{yx} can be estimated. Therefore, to obtain the full nine spin exchange parameters JαβJ^{\alpha\beta} (i.e. JJ, DD and Γ\Gamma terms), one need perform different first-principles self-consistent calculations for at least three independent orientations of the magnetization [25]. Based on the self-consistent results from different spin orientations, the magnetic interactions could be calculated from Eq. (5) [25]. However, these self-consistent calculations by choosing three different spin orientations would produce 12 parameters, resulting in that sets of parameters JαβJ^{\alpha\beta} are not necessarily unique, and naturally leading to the calculation deviation.

To reduce this calculation deviation, we also propose a new method when SOC is relatively small. Firstly, we perform the standard LSDA (+U+U) calculations. Based on the eigenvalues εnk\varepsilon_{nk} and eigenstates ψnk(0)\psi_{nk}^{(0)} from LSDA (+U+U) calculations, we take SOC as a perturbation, and estimate the first-order and second-order SOC corrections wavefunction ψnk(1)\psi_{nk}^{(1)} and ψnk(2)\psi_{nk}^{(2)} in Wien2k [57]. Then all JαβJ^{\alpha\beta} elements can be calculated with no need to do the separate self-consistent calculations with different spin orientations. Meanwhile, this method can produce the first and second order SOC contribution to the DM and Γ\Gamma interaction separately.

In the following, we will apply our two methods to several typical examples corresponding to 3d3d, 4d4d and 5d5d transition metal oxides respectively in the next section.

Results

First-principles examples of typical materials

A. La2_2CuO4_4

La2_2CuO4_4
JJRef. [53]Ref. [36]this work
J1J_127.229.225.76
J2J_2−3.00−4.1, −3.9−3.80, −3.38
J3J_3−0.050−0.11

Table 1. The calculated Heisenberg exchange parameters JJ (in meV) for La2_2CuO4_4. The calculated spin exchange parameters in the previous theoretical work are also shown for comparison.

As a benchmark on the accuracy of our methods in calculating Heisenberg and DM interactions, we first study the famous La2_2CuO4_4, which has been studied in a number of theoretical work [36, 37, 53, 59, 60, 61, 62, 63]. The LSDA +U+U (=7=7 eV) [64] calculation is applied. Without SOC considered, the calculated Heisenberg exchange parameters have no difference between these two approaches, which are summarized in Table I. The calculated nearest neighbor magnetic coupling J1J_1 are dominant with the value of about 25.76 meV. We can find that the spin exchange coupling parameters decrease rapidly with the increasing distance between two Cu ions. The next nearest neighbor magnetic coupling J2J_2 shows ferromagnetic behavior and is one order of magnitude smaller than J1J_1. The third nearest neighbor J3J_3 is antiferromagnetic and almost negligible. The results agree well with the previous theoretical work [36, 53].

The weak ferromagnetism of La2_2CuO4_4 is originated from the canting of the magnetic moments, which can be described by the competition of Heisenberg interaction and DM interaction. Based on the two approaches in above section, the nearest neighbor DM parameters are calculated as shown in Table II. As shown in Table II, the DM interactions proportional to λ2\lambda^2 are negligible due to the small SOC in 3d3d orbital, and the DM interactions proportional to λ\lambda are almost the same as the calculated DM parameters in WienJ. According to the calculated Heisenberg and DM parameters, the value of the canting angle is estimated to be about 1.7×10−31.7\times10^{-3}, which is in a good agreement with the experimental value of 2.22.2-2.9×10−32.9\times10^{-3} [62, 63]. For comparison with previous theoretical works, Mazurenko et al. [36] proposed the angle value of 0.7×10−30.7\times10^{-3} using Green’s function technique. With the construction of a tight-binding parametrization of the Hamiltonian with SOC, Katsnelson et al. [37] calculated canting angle to be 5.0×10−35.0\times10^{-3}. Comparatively our results agree very well with the experiment and quite promising.

RRWienJthe second method in this work
DDD(1)D^{(1)}D(2)D^{(2)}
(0.5,−0.5,0)(0.5,-0.5,0)(−0.09,−0.14,0)(-0.09,-0.14,0)(−0.09,−0.14,0)(-0.09,-0.14,0)(0,0,0)(0,0,0)
(−0.5,−0.5,0)(-0.5,-0.5,0)(−0.09,0.14,0)(-0.09,0.14,0)(−0.09,0.14,0)(-0.09,0.14,0)(0,0,0)(0,0,0)
(−0.5,0.5,0)(-0.5,0.5,0)(−0.09,−0.14,0)(-0.09,-0.14,0)(−0.09,−0.14,0)(-0.09,-0.14,0)(0,0,0)(0,0,0)
(0.5,0.5,0)(0.5,0.5,0)(−0.09,0.14,0)(-0.09,0.14,0)(−0.09,0.14,0)(-0.09,0.14,0)(0,0,0)(0,0,0)

Table II. The calculated nearest neighbor DM interaction parameters (in meV) for La2_2CuO4_4 via the two approaches in this work. RR is the radius vector from two sites of magnetic ions in units of the lattice constant. The columns D(1)D^{(1)} and D(2)D^{(2)} represent the calculated DM interaction proportional to λ\lambda and λ2\lambda^2. Due to the small SOC, the DM interactions proportional to λ2\lambda^2 are zero with an accuracy of 0.01 meV.

B. Ca2_2RuO4_4

As the example of 4d4d transition metal oxides, Ca2_2RuO4_4 crystallizes in the space group Pbca and has the layered perovskite structure [65, 66, 67, 68] [69, 70]. The ground state of Ca2_2RuO4_4 is an antiferromagnetic spin ordering with an insulating electrical behavior [66]. A weak ferromagnetic component is induced by spin canting below the magnetic transition temperature 113K [65].

To study its magnetic properties, we performed the LSDA +U+ U (=3=3 eV) [67] calculations for Ca2_2RuO4_4. The calculated nearest neighbor Heisenberg interaction is about 20.9 meV. Experimentally, the Heisenberg parameters were estimated via inelastic neutron scattering as 16 meV in Ref. [69] and 5.8 meV in Ref. [70], and our result 20.9 meV is closer to the first value. Meanwhile, the calculated DM interactions by the two approaches mentioned above are both presented in Table III. As shown in Table III, the calculated DM interactions in WienJ are also almost the same as the sum of DM interactions proportional to λ\lambda and λ2\lambda^2, i.e., D≈D(λ)+D(λ2)D \approx D(\lambda)+D(\lambda^2) as shown in Table III. Note that the strength of first-order SOC corrected DM interactions ∣D(λ)∣|D(\lambda)| is around 1.31–1.41 meV, while the ∣D(λ2)∣|D(\lambda^2)| is about 0.44 meV, therefore the DM interactions proportional to λ2\lambda^2 are non-negligible in such 4d4d magnetic system. The ratio of DM interactions and Heisenberg interaction is estimated to be ∣D∣/J≈0.05|D|/J \approx0.05, which is in good agreement with the rough estimate 0.06 from experiment [68].

RRWienJthe second method in this work
DDD(1)D^{(1)}D(2)D^{(2)}
(0.5,−0.5,0)(0.5,-0.5,0)(0.50,−1.03,0.17)(0.50,-1.03,0.17)(0.47,−1.19,0.60)(0.47,-1.19,0.60)(0.04,0.15,−0.41)(0.04,0.15,-0.41)
(−0.5,0.5,0)(-0.5,0.5,0)(0.50,−1.03,0.17)(0.50,-1.03,0.17)(0.47,−1.19,0.60)(0.47,-1.19,0.60)(0.04,0.15,−0.41)(0.04,0.15,-0.41)
(−0.5,−0.5,0)(-0.5,-0.5,0)(−0.43,−0.95,0.17)(-0.43,-0.95,0.17)(−0.40,−1.10,0.59)(-0.40,-1.10,0.59)(−0.04,0.15,−0.41)(-0.04,0.15,-0.41)
(0.5,0.5,0)(0.5,0.5,0)(−0.43,−0.95,0.17)(-0.43,-0.95,0.17)(−0.40,−1.10,0.59)(-0.40,-1.10,0.59)(−0.04,0.15,−0.41)(-0.04,0.15,-0.41)

Table III. The calculated DM interaction parameters (in meV) for Ca2_2RuO4_4 via the two approaches in this work. Here RR is the radius vector from two sites of magnetic ions in units of the lattice constant. The columns D(1)D^{(1)} and D(2)D^{(2)} represent the calculated DM interaction proportional to λ\lambda and λ2\lambda^2. It can be seen that the calculated interactions via these two approaches are close, i.e. D≈D(1)+D(2)D \approx D^{(1)}+D^{(2)}.

Ca3_3LiOsO6_6

In 5d transition-metal oxides systems, the strength of SOC is expected to be stronger than 3d or 4d materials due to the large atomic number. However, in orbital singlet states with relatively large electronic gap such as 5d35d^3 with half-filling t2gt_{2g} orbitals [71, 72], the electronic structures from fully self-consistent LSDA (+U+U) + SOC calculation and the ones from further one iteration of SOC calculation after LSDA (+U+U) calculation have small difference (see Fig. 2 in Appendix B), indicating that the effect of SOC is still small [71, 72].

As one concrete 5d35d^3 example, we focus on Ca3_3LiOsO6_6 [72–75] with the crystal structure of K4_4CdCl6_6-type. The ground state of Ca3_3LiOsO6_6 is AFM with the magnetic transition temperature 117K. Both the first-principles study and the experiment suggested that Ca3_3LiOsO6_6 has a fully opened electronic gap [74]. Though the AFM ordered state has been confirmed experimentally, the magnetization curve suggests a soft magnetism with a small spontaneous magnetization. The net magnetization is about 0.02 μB\mu_B per Os5+^{5+} ion and is suggested due to a DM interaction generated by the broken inversion symmetry [72].

We perform the LSDA +U+U calculations of Ca3_3LiOsO6_6 with U=2U=2 eV [71, 76–78] and calculate the magnetic interactions by applying our methods. The Heisenberg interactions J1J_1, J2J_2 and J3J_3 are estimated to be all AFM with the values of 13.1 meV, 5.5 meV and 1.1 meV respectively. The J1J_1 is the strongest spin exchange, while J2J_2 is slightly less than one half of J1J_1, and J3J_3 is an order of magnitude smaller than J1J_1. These properties are in consistent with the energy-mapping results, though our calculated spin parameters are slightly larger than theirs (9.9 meV, 4.1 meV and 0.63 meV for J1J_1-J3J_3 respectively) [75]. Meanwhile, our numerical DM interactions by the two approaches mentioned in Method section are both summarized in Table IV. Since the DM interactions between the 3rd nearest neighbor and longer-range distances for Os5+^{5+} ions are negligible, thus we only show the DM interactions for nearest neighbor and the next nearest neighbor in Table IV. It can be seen that the DM interactions proportional to λ2\lambda^2 have the same order of magnitude as the one proportional to λ\lambda in Ca3_3LiOsO6_6. According to the crystal symmetry, the nearest neighbor D1D_1 has the form of (0,0,Dz)(0,0,D_z), and the calculated D1D_1 via the two approaches are summarized in Table IV. Meanwhile, there are three different directions of D2D_2 connected by the symmetry of threefold rotation along z-axis, as shown in last three rows of Table IV. Summarizing the DM parameters of all nearest neighbors and the isotropic spin exchange parameters JJ up to the 3rd nearest neighbor, the expected magnetic moment is estimated to be 0.03 μB0.03\ \mu_B, which is in good agreement with the experimental value of 0.02 μB0.02\ \mu_B [72]. and m′m' all belong to representation A′′A''. Thus, the hopping between the orbitals with different representations are symmetry forbidden, i.e. tmn′=tm′n=0t_{mn'} = t_{m'n} = 0, while the hopping tnn′t_{nn'} and tmm′t_{mm'} between the orbitals with same representation can exist. Moreover, since these two magnetic ions are not related by the crystal symmetry, lm′n′αlmnβ−lm′n′βlmnαl_{m'n'}^\alpha l_{mn}^\beta- l_{m'n'}^\beta l_{mn}^\alpha could also be non-zero. Therefore, this case satisfies all the three conditions we listed above to be a candidate with only the DM interactions of second order SOC. The above three conditions can be satisfied in many Wyckoff positions of various space groups, and we believe that it may be widespread in a variety of magnetic materials.

RRDDD(1)D^{(1)}D(2)D^{(2)}
(0.5,0.5,0.5)(0.5, 0.5, 0.5)(0,0,−0.263)(0, 0, -0.263)(0,0,−0.223)(0, 0, -0.223)(0,0,−0.099)(0, 0, -0.099)
(−0.5,0.5,0.5)(-0.5, 0.5, 0.5)(−0.125,0.205,−0.021)(-0.125, 0.205, -0.021)(−0.083,0.143,−0.052)(-0.083, 0.143, -0.052)(−0.055,0.095,0.005)(-0.055, 0.095, 0.005)
(0.5,−0.5,0.5)(0.5, -0.5, 0.5)(−0.125,−0.205,−0.021)(-0.125, -0.205, -0.021)(−0.083,−0.143,−0.052)(-0.083, -0.143, -0.052)(−0.055,−0.095,0.005)(-0.055, -0.095, 0.005)
(0.5,0.5,−0.5)(0.5, 0.5, -0.5)(0.226,0,−0.021)(0.226, 0, -0.021)(0.165,0,−0.052)(0.165, 0, -0.052)(0.110,0,0.005)(0.110, 0, 0.005)

Table IV. The calculated DM interaction parameters (in meV) for Ca3_3LiOsO6_6 via the two approaches in this work. RR is the radius vector from two sites of magnetic ions in units of the lattice constant. The columns D(1)D^{(1)} and D(2)D^{(2)} represent the calculated DM interaction proportional to λ\lambda and λ2\lambda^2.

Materials with the second-order SOC correction to DM interactions

Here we discuss the conditions where the first order of SOC in DM interactions (i.e. D(1)D^{(1)}) are absent, while DM interactions proportional to the second order of SOC (i.e. D(2)D^{(2)}) are non-zero. According to the picture of Figure 1 and Eqs. (3),(4), we summarize these three conditions that need to be met:

  1. the hopping processes between ground state and excited state (tmn′t_{mn'} and tm′nt_{m'n}) are symmetry forbidden.

  2. the hopping between ground states (tnn′t_{nn'}) and the hopping between excited states at two sites (tmm′t_{mm'}) are non-zero.

  3. the relation of orbital angular momentum of the two magnetic ions lm′n′αlmnβ−lm′n′βlmnαl_{m'n'}^\alpha l_{mn}^\beta- l_{m'n'}^\beta l_{mn}^\alpha should be also non-zero.

According to the first condition tmn′=tm′n=0t_{mn'} = t_{m'n} = 0, the exchange processes in Figure 1(a) and (c) are forbidden, therefore D(1)D^{(1)} is constrained to zero. Meanwhile, the second and third conditions make the exchange processes in Figure 1(b) exist, thus D(2)D^{(2)} could be present. Based on these restrictions, one can easily predict promising candidates with only the second-order SOC correction to DM interactions according to their different combinations of crystal symmetry, Wyckoff sites and orbital occupation pattern. As a simple example, we consider the space group PmPm (No. 6) and put two magnetic ions AA and BB located at two different 1a Wyckoff Positions (x,0,z)(x, 0, z) and (x′,0,z′)(x', 0, z'). These Wyckoff Positions have a mirror symmetry. Therefore, according to the different eigenvalues of mirror operation, we can mark their states with the irreducible representation A′A' and A′′A'' respectively [79]. Here we assume that their ground states nn and n′n' both belong to representation A′A', while their excited states mm and m′m' all belong to representation A′′A''. Thus, the hopping between the orbitals with different representations are symmetry forbidden, i.e. tmn′=tm′n=0t_{mn'} = t_{m'n} = 0, while the hopping tnn′t_{nn'} and tmm′t_{mm'} between the orbitals with same representation can exist. Moreover, since these two magnetic ions are not related by the crystal symmetry, lm′n′αlmnβ−lm′n′βlmnαl_{m'n'}^\alpha l_{mn}^\beta- l_{m'n'}^\beta l_{mn}^\alpha could also be non-zero. Therefore, this case satisfies all the three conditions we listed above to be a candidate with only the DM interactions of second order SOC. The above three conditions can be satisfied in many Wyckoff positions of various space groups, and we believe that it may be widespread in a variety of magnetic materials.

Conclusion

The magnetic model plays an important role in magnetic investigations. Here we revisit the general expression of magnetic interactions, including isotropic exchange interaction, antisymmetric DM interaction and symmetric Γ\Gamma term. We clarify that the term proportional to λ2\lambda^2 has both contribution to DM interaction and Γ\Gamma term. We find that the DM and Γ\Gamma interactions can be separated with three different hopping processes rather than the orders of SOC. We present two first-principles methods to calculate the anisotropic magnetic interactions. Based on the first method, one need perform self-consistent calculations for at least three different spin orientations to obtain the full nine exchange parameters JαβJ^{\alpha\beta}. On the other hand, using the second method, one can estimate these magnetic exchange parameters with no need to do the separate self-consistent calculations for different spin orientations. This method can also calculate the first-order and second-order SOC contribution to DM interactions separately. We have successfully applied our methods to several typical weak ferromagnetic materials La2_2CuO4_4, Ca2_2RuO4_4 and Ca3_3LiOsO6_6 respectively. Furthermore, according to microscopic mechanism shown in Figure 1, we list all three conditions which can lead to that the DM interactions proportional to λ\lambda are symmetric forbidden while the DM interactions proportional to λ2\lambda^2 exist.

Acknowledgements

This work was supported by the NSFC (No. 12188101, 11834006, 12004170, 51721001, 11790311), National Key R&D Program of China (No. 2022YFA1403601), Natural Science Foundation of Jiangsu Province, China (Grant No. BK20200326), and the excellent programme in Nanjing University. Xiangang Wan also acknowledges the support from the Tencent Foundation through the XPLORER PRIZE.

References

References

  1. [34]P. Bruno, “Exchange interaction parameters and adiabatic spin-wave spectra of ferromagnets: A renormalized magnetic force theorem,” Phys. Rev. Lett. 90, 087205 (2003).DOI
  2. [35]M. Katsnelson and A. Lichtenstein, “First-principles calculations of magnetic interactions in correlated systems,” Phys. Rev. B 61, 8906 (2000).DOI
  3. [36]V. Mazurenko and V. Anisimov, “Weak ferromagnetism in antiferromagnets: α-Fe2O3 and La2CuO4,” Phys. Rev. B 71, 184434 (2005).DOI
  4. [37]M. Katsnelson, Y. Kashnin, V. Mazurenko, and A. Lichtenstein, “Correlated band theory of spin and orbital contributions to Dzyaloshinskii-Moriya interactions,” Phys. Rev. B 82, 100403 (2010).DOI
  5. [38]L. Udvardi, L. Szunyogh, K. Palotás, and P. Weinberger, “First-principles relativistic study of spin waves in thin magnetic films,” Phys. Rev. B 68, 104436 (2003).DOI
  6. [39]H. Ebert and S. Mankovsky, “Anisotropic exchange coupling in diluted magnetic semiconductors: Ab initio spin-density functional theory,” Phys. Rev. B 79, 045209 (2009).DOI
  7. [40]V. Dmitrienko, E. Ovchinnikova, S. Collins, G. Nisbet, G. Beutier, Y. Kvashnin, V. Mazurenko, A. Lichtenstein, and M. Katsnelson, “Measuring the Dzyaloshinskii-Moriya interaction in a weak ferromagnet,” Nature Physics 10, 202 (2014).DOI
  8. [41]V. Mazurenko, S. Skornyakov, V. Anisimov, and F. Mila, “First-principles investigation of symmetric and antisymmetric exchange interactions of SrCu2(BO3)2,” Phys. Rev. B 78, 195110 (2008).DOI
  9. [42]S. Mankovsky and H. Ebert, “Accurate scheme to calculate the interatomic Dzyaloshinskii-Moriya interaction parameters,” Phys. Rev. B 96, 104416 (2017).DOI
  10. [43]I. Solovyev, N. Hamada, and K. Terakura, “Crucial role of the lattice distortion in the magnetism of LaMnO3,” Phys. Rev. Lett. 76, 4825 (1996).pubmed.ncbi.nlm.nih.gov/10061390
  11. [44]A. N. Rudenko, V. V. Mazurenko, V. I. Anisimov, and A. I. Lichtenstein, “Weak ferromagnetism in Mn nanochains on the CuN surface,” Phys. Rev. B 79, 144418 (2009).DOI
  12. [45]Y. O. Kvashnin, A. Bergman, A. I. Lichtenstein, and M. I. Katsnelson, “Relativistic exchange interactions in CrX3 (X= Cl, Br, I) monolayers,” Phys. Rev. B 102, 115162 (2020).DOI
  13. [46]I. Solovyev, “Self-consistent linear response for the spin-orbit interaction related properties,” Phys. Rev. B 90, 024417 (2014).DOI
  14. [47]T. Koretsune, N. Nagaosa, and R. Arita, “Control of Dzyaloshinskii-Moriya interaction in Mn1-xFexGe: a first-principles study,” Scientific reports 5, 13302 (2015).DOI
  15. [48]T. Kikuchi, T. Koretsune, R. Arita, and G. Tatara, “Dzyaloshinskii-Moriya interaction as a consequence of a Doppler shift due to spin-orbit-induced intrinsic spin current,” Phys. Rev. Lett. 116, 247201 (2016).DOI
  16. [49]T. Koretsune, T. Kikuchi, and R. Arita, “First-principles evaluation of the dzyaloshinskii-moriya interaction,” Journal of the Physical Society of Japan 87, 041011 (2018).DOI
  17. [50]F. Freimuth, S. Blügel, and Y. Mokrousov, “Berry phase theory of Dzyaloshinskii–Moriya interaction and spin–orbit torques,” Journal of physics: Condensed matter 26, 104202 (2014).DOI
  18. [51]O. K. Andersen and O. Jepsen, “Explicit, first-principles tight-binding theory,” Physical Review Letters 53, 2571 (1984).DOI
  19. [52]X. Wan, Q. Yin, and S. Y. Savrasov, “Calculation of magnetic exchange interactions in mott-hubbard systems,” Phys. Rev. Lett. 97, 266403 (2006).pubmed.ncbi.nlm.nih.gov/17280438
  20. [53]X. Wan, T. A. Maier, and S. Y. Savrasov, “Calculated magnetic exchange interactions in high-temperature superconductors,” Phys. Rev. B 79, 155114 (2009).DOI
  21. [54]X. Wan, J. Dong, and S. Y. Savrasov, “Mechanism of magnetic exchange interactions in europium monochalcogenides,” Physical Review B 83, 205201 (2011).DOI
  22. [55]X. Wan, V. Ivanov, G. Resta, I. Leonov, and S. Y. Savrasov, “Exchange interactions and sensitivity of the ni two-hole spin state to hund’s coupling in doped ndnio2,” Physical Review B 103, 075123 (2021).DOI
  23. [56]D. Wang, X. Bo, F. Tang, and X. Wan, “WienJ,” https://github.com/diwang0214/WienJ.
  24. [57]P. Blaha, K. Schwarz, G. K. H. Madsen, D. Kvasnicka, and J. Luitz, WIEN2k, An augmented plane wave+ Local Orbitals Program for calculating Crystal Properties (Karl-heinz Schwarz, TU Wien, Austria, 2001).
  25. [58]P. W. Anderson, “Antiferromagnetism. theory of superexchange interaction,” Physical Review 79, 350 (1950).DOI
  26. [59]D. Coffey, T. Rice, and F. Zhang, “Dzyaloshinskii-Moriya interaction in the cuprates,” Physical Review B 44, 10112 (1991).pubmed.ncbi.nlm.nih.gov/9999014
  27. [60]L. Shekhtman, O. Entin-Wohlman, and A. Aharony, “Moriyas anisotropic superexchange interaction, frustration, and Dzyaloshinskys weak ferromagnetism,” Phys. Rev. Lett. 69, 836 (1992).
  28. [61]T. Yildirim, A. B. Harris, A. Aharony, and O. Entin-Wohlman, “Anisotropic spin Hamiltonians due to spin-orbit and Coulomb exchange interactions,” Physical Review B 52, 10239 (1995).DOI
  29. [62]T. Thio, T. Thurston, N. Preyer, P. Picone, M. Kastner, H. Jenssen, D. Gabbe, C. Chen, R. Birgeneau, and A. Aharony, “Antisymmetric exchange and its influence on the magnetic structure and conductivity of La2CuO4,” Physical Review B 38, 905 (1988).pubmed.ncbi.nlm.nih.gov/9945290
  30. [63]M. Kastner, R. Birgeneau, T. Thurston, P. Picone, H. Jenssen, D. Gabbe, M. Sato, K. Fukuda, S. Shamoto, Y. Endoh, et al., “Neutron-scattering study of the transition from antiferromagnetic to weak ferromagnetic order in La2CuO4,” Physical Review B 38, 6636 (1988).pubmed.ncbi.nlm.nih.gov/9945340
  31. [64]P. Wei and Z. Q. Qi, “Electronic structure of La2CuO4 and YBa2Cu3O6: A local-spin-density approximation with on-site Coulomb-U correlation calculations,” Physical Review B 49, 12159 (1994).pubmed.ncbi.nlm.nih.gov/10010089
  32. [65]M. Braden, G. André, S. Nakatsuji, and Y. Maeno, “Crystal and magnetic structure of Ca2RuO4: Magnetoelastic coupling and the metal-insulator transition,” Phys. Rev. B 58, 847 (1998).DOI
  33. [66]H. Fukazawa, S. Nakatsuji, and Y. Maeno, “Intrinsic properties of the mott insulator Ca2RuO4+δ (δ=0) studied with single crystals,” Physica B: Condensed Matter 281, 613 (2000).DOI
  34. [67]J. Jung, Z. Fang, J. He, Y. Kaneko, Y. Okimoto, and Y. Tokura, “Change of Electronic Structure in Ca2RuO4 Induced by Orbital Ordering,” Phys. Rev. Lett. 91, 056403 (2003).pubmed.ncbi.nlm.nih.gov/12906613
  35. [68]S. Kunkemöller, E. Komleva, S. Streltsov, S. Hoffmann, D. Khomskii, P. Steffens, Y. Sidis, K. Schmalz, and M. Braden, “Magnon dispersion in Ca2Ru1-xTixO4: Impact of spin-orbit coupling and oxygen moments,” Phys.

Paper details

vlnc:W_sapwky29
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