betaFeedback

WorksPhysics and Astronomy

Topological magnon modes on honeycomb lattice with coupling textures

Abstract

Topological magnon modes are expected to be useful for novel applications such as robust information propagation, since they are immune to backscattering and robust against disorder. Although there are several of theoretical proposals for topological magnon modes and growing experimental efforts for realizing them by now, it is still desirable to add complementary insights on this important phenomenon. Here, we propose a new scheme to achieve topological magnon where only nearest-neighbour exchange couplings on honeycomb lattice are necessary. In both ferromagnets and antiferromagnets, tuning exchange couplings between and inside hexagonal unit cells induces a topological state accompanied by a band inversion between p-orbital and d-orbital like magnon modes. Topological magnon modes appear at the interface between a topological domain and a trivial domain with magnon currents, which counterpropagate depending on pseudospins originated from orbital angular momenta of magnon modes. This mimics the spin-momentum locking phenomenon in the quantum spin Hall effect.

Introduction

Recently topology has became a unified key concept in material sciences[1, 2, 3] [4, 5, 6]. The most prominent feature of topological systems is the surface or edge states. Because of topological protection, these surface or edge states are immune to back-scattering and robust against disorder, which can be exploited for achieving innovative functionalities. The current intense study of topological systems was ignited by the discovery of topological insulators in quantum electronic solids, but this idea has been developed into bosonic systems and various wave phenomena7–13.

Magnons are quanta of spin-wave excitations in magnetic systems. As quasiparticles, magnons are charge neutral bosons and free of dissipation due to Ohmic heating, thus useful for various applications14–16. Novel features of topological magnon modes protected by bulk topology17–21 have also been considered. So far, several possible mechanisms for realizing topological magnon modes are proposed17, 22–26. Although these proposals are enlightening, experimental realization and firm confirmation are still not easy. Amongst the known proposals, those regarded as promising are using the Dzyaloshinskii-Moriya (DM) interaction in Kagome lattice18, 19 or honeycomb lattice20, 21. In both cases, the DM interaction is crucial for achieving the nontrivial topology, playing the role analogous to the spin-orbit coupling for electrons in the quantum spin Hall effect[2, 3]. However, having a sufficiently strong DM interaction is not necessarily easy, limiting the experimental realization of topological magnon modes. To the best of our knowledge, though there are some experimental progresses in materials with the DM interaction27, 28, firm evidences for topological magnon modes are still lacking.

In this work, we propose a new scheme to achieve topological magnon modes on honeycomb lattice, where nontrivial topology is achieved upon tuning nearest-neighbor (n.n.) exchange couplings, whereas no DM interaction is required. It is known29–31 that a ferromagnet on honeycomb lattice with n.n. exchange couplings exhibits Dirac-type linear magnon dispersions. Generally, a Dirac-type linear dispersion can be a nice starting point to have topologically nontrivial states. Here, we demonstrate that introducing a C6vC_{6v}-symmetric texture32 in the strength of exchange coupling opens a frequency band gap in the magnon frequency band structure, which yields a band inversion and the nontrivial topology. We also extend this idea to antiferromagnets. Unlike ferromagnets, the dispersion of magnon for an antiferromagnet on honeycomb lattice is doubly degenerate owing to the combination of time-reversal symmetry and inversion symmetry, where no Dirac-type dispersion exists31, 33, 34. We notice that this degeneracy is lifted in the canted antiferromagnetic state caused by an external magnetic field except at the KK and K′K' points, which yields Dirac-type dispersions. Therefore, a topological state can be induced by a coupling texture in the canted antiferromagnet with the same mechanism in ferromagnets. In both ferromagnets and antiferromagnets, topological magnon modes appear at the interface between topological and trivial domains, with directions of magnon currents governed by pseudospins32, 35, 36, which mimics the spin-momentum locking phenomenon in the quantum spin Hall effect[2, 3].

Results

Topological magnon modes in ferromagnets

We start from the Heisenberg model

H=−∑⟨i,j⟩JijSi⋅Sj,(1)H=-\sum_{\langle i,j\rangle}J_{ij}\mathbf{S}_i\cdot\mathbf{S}_j, \tag*{(1)}

where the summation runs over n.n. sites of honeycomb lattice, and Jij>0J_{ij}>0 denote ferromagnetic exchange couplings. As illustrated in Figure 1, we use a unit cell containing six sites for the convenience of modulation in strength of exchange coupling, instead of a conventional rhombic unit cell with two sites. In specific, we assign Jij=J0J_{ij}=J_0 inside unit cells and Jij=J1J_{ij}=J_1 between unit cells [see Figure 1], which respects the C6vC_{6v} symmetry. The six-site cluster in a unit cell can be regarded as an “artificial molecule”.

Image not extracted

Figure 1. Ferromagnet on honeycomb lattice with nearest-neighbor exchange couplings. Hexagonal unit cells indicated by the dashed line are chosen with J0/J1J_0/J_1 representing exchange couplings inside/between hexagonal unit cells, which preserves C6vC_{6v} symmetry. Numbers 1,...,6 index atomic sites inside the hexagonal unit cell. Unit vectors are represented by a1\mathbf{a}_1 and a2\mathbf{a}_2, and ∣a1∣=∣a2∣=a0\lvert\mathbf{a}_1\rvert=\lvert\mathbf{a}_2\rvert=a_0.

The spin-wave excitations in Hamiltonian (1) can be represented as magnons by the Holstein-Primakoff transformation37,

Siz=Si−bi†bi,(2)S_i^z=S_i-b_i^\dagger b_i, \tag*{(2)}
Si+=(2Si−bi†bi)bi,(3)S_i^+=\left(\sqrt{2S_i-b_i^\dagger b_i}\right)b_i, \tag*{(3)}
Si−=bi†2Si−bi†bi,(4)S_i^-=b_i^\dagger\sqrt{2S_i-b_i^\dagger b_i}, \tag*{(4)}

with Si±=Six±iSiyS_i^\pm=S_i^x\pm iS_i^y, where SixS_i^x, SiyS_i^y and SizS_i^z are the three orthogonal components of Si\mathbf{S}_i, bib_i and bi†b_i^\dagger are annihilation operators and creation operators of magnons. The Hamiltonian in the magnon representation is complicated since it contains many-body magnon-magnon interaction terms such as bi†bibj†bjb_i^\dagger b_i b_j^\dagger b_j. At low temperatures, the average number of magnon excitations ⟨bi†bi⟩\langle b_i^\dagger b_i\rangle is small comparing to 2Si2S_i, making it a reasonable approximation to neglect the magnon-magnon interactions. Then the Holstein-Primakoff transformation gives Si+≈2SibiS_i^+\approx\sqrt{2S_i}b_i and Si−≈2Sibi†S_i^-\approx\sqrt{2S_i}b_i^\dagger, which results in the effective Hamiltonian with terms quadratic in bib_i and bi†b_i^\dagger.

We then turn to momentum space by applying the Fourier transformation

bi,r=1N∑kbi,keik⋅r,(5)b_{i,r}=\frac{1}{\sqrt{N}}\sum_k b_{i,k}e^{i\mathbf{k}\cdot\mathbf{r}}, \tag*{(5)}

where ii is the site index inside a unit cell as shown in Figure 1, NN is the number of unit cells and r\mathbf{r} is the position of the unit cell. The Hamiltonian for magnons is then described as

H^F=∑kΨk†HF,kΨk,(6)\hat{H}_{F}=\sum_k\Psi_k^\dagger H_{F,k}\Psi_k, \tag*{(6)}

where Ψk=[b1,k,b2,k,b3,k,b4,k,b5,k,b6,k]T\Psi_k = [b_{1,k},b_{2,k},b_{3,k},b_{4,k},b_{5,k},b_{6,k}]^{T} [see Figure 1]. With the uniform spin length Si=SS_i = S, HF,kH_{\mathrm{F},k} is given by

HF,k=[E0I3−Qk−Qk†E0I3],(7)H_{\mathrm{F},k} = \begin{bmatrix} E_0 I_3 & -Q_k \\ -Q_k^\dagger& E_0 I_3 \end{bmatrix}, \tag*{(7)}

with

Qk=S[J1e−ik⋅(a1−a2)J0J0J0J1eik⋅a1J0J0J0J1e−ik⋅a2],(8)Q_k = S \begin{bmatrix} J_1 e^{-i\mathbf{k}\cdot(\mathbf{a}_1-\mathbf{a}_2)} & J_0 & J_0 \\ J_0 & J_1 e^{i\mathbf{k}\cdot\mathbf{a}_1} & J_0 \\ J_0 & J_0 & J_1 e^{-i\mathbf{k}\cdot\mathbf{a}_2} \end{bmatrix}, \tag*{(8)}

where E0=S(2J0+J1)E_0 = S(2J_0 + J_1) and I3I_3 is a 3×33 \times3 identity matrix, and a1\mathbf{a}_1 and a2\mathbf{a}_2 are unit vectors presented in Figure 1.

The eigenvalue equation with Hamiltonian (7),

HF,kψn=εF,n,kψn,(9)H_{\mathrm{F},k}\psi_n = \varepsilon_{\mathrm{F},n,k}\psi_n, \tag*{(9)}

gives the eigenstates and eigenvalues with n=1,2,…,6n = 1, 2, \ldots, 6. Then, Hamiltonian (6) can be cast into

H^F=∑k∑nεF,n,kbn,k′†bn,k′,(10)\hat{H}_{\mathrm{F}} = \sum_k \sum_n \varepsilon_{\mathrm{F},n,k} b_{n,k}^{\prime\dagger} b_{n,k}^{\prime}, \tag*{(10)}

where bn,k′†=∑iψnibi,k†b_{n,k}^{\prime\dagger} = \sum_i \psi_{ni} b_{i,k}^{\dagger} is the creation operator of nn-th eigen magnon mode and ψni\psi_{ni} is the ii-th element of ψn\psi_n.

We notice that HF,kH_{\mathrm{F},k} in Eq. (7) only differs from the electronic model on honeycomb lattice with hopping textures35,36,

Hk=[0QkQk†0],(11)H_k = \begin{bmatrix} 0 & Q_k \\ Q_k^\dagger& 0 \end{bmatrix}, \tag*{(11)}

by the diagonal element E0E_0, where t0t_0 and t1t_1 in electronic model are replaced by SJ0SJ_0 and SJ1SJ_1 respectively. Therefore, they share the same eigenstates with the shifted eigenvalues:

εF,n,k=E0−ε0,n,k,(12)\varepsilon_{\mathrm{F},n,k} = E_0 - \varepsilon_{0,n,k}, \tag*{(12)}

where ε0,n,k\varepsilon_{0,n,k} is the eigenvalue of HkH_k in Eq. (11). One has εF,n,k∈[0,2E0]\varepsilon_{\mathrm{F},n,k} \in[0,2E_0], since ε0,n,k∈[−E0,E0]\varepsilon_{0,n,k} \in[-E_0,E_0].

The time-dependent form of bn,k′b_{n,k}^{\prime} is bn,k′e−iωF,n,ktb_{n,k}^{\prime} e^{-i\omega_{\mathrm{F},n,k}t}, where ωF,n,k\omega_{\mathrm{F},n,k} is the magnon frequency. With the Heisenberg equation of motion

dbn,k′dt=iℏ[H^F,bn,k′],(13)\frac{d b_{n,k}^{\prime}}{dt} = \frac{i}{\hbar}[\hat{H}_{\mathrm{F}},b_{n,k}^{\prime}], \tag*{(13)}

and the commutation relation of bosons

[bn,k,bn′,k′†]=δnn′δkk′,(14)[b_{n,k},b_{n',k'}^\dagger] = \delta_{nn'}\delta_{kk'}, \tag*{(14)}

we can derive the dynamic equation of magnon bn,k′b_{n,k}^{\prime}

d2bn,k′dt2=−εF,n,k2ℏ2bn,k′.(15)\frac{d^2 b_{n,k}^{\prime}}{dt^2} = -\frac{\varepsilon_{\mathrm{F},n,k}^2}{\hbar^2} b_{n,k}^{\prime}. \tag*{(15)}

The magnon frequency can be obtained by

ωF,n,k=εF,n,k/ℏ,(16)\omega_{\mathrm{F},n,k} = \varepsilon_{\mathrm{F},n,k}/\hbar, \tag*{(16)}

since εF,n,k>0\varepsilon_{\mathrm{F},n,k} > 0.

The frequency band structures of magnon modes are shown in Figure 2. For J0=J1J_0 = J_1, the frequency band structure exhibits Dirac cones same as that of the electronic structure of graphene30. These Dirac cones are folded to the Γ\Gamma point as shown in Figure 2(b), since we are using the six-site hexagonal unit cell instead of the conventional two-site rhombic unit cell. For J0≠J1J_0 \ne J_1, a frequency gap opens at the Γ\Gamma point as displayed in Figure 2(a) and Figure 2(c). As clarified in Ref. 32, the band structures for J0>J1J_0 > J_1 [Figure 2(a)] and for J0<J1J_0 < J_1 [Figure 2(c)] are distinct in topology.

Image not extracted

Figure 2. (a) Frequency band structure of magnon modes for a ferromagnet with J0>J1J_0 > J_1, where a gap opens between dd and pp modes. The parity of inversion symmetry at the M point is the same as that at the Γ\Gamma point for each band, where the even/odd parity is marked by +/−+/-. (b) Same as (a) except for J0=J1J_0 = J_1, where the gap closes and Dirac cones appear. (c) Same as (a) except for J0<J1J_0 < J_1, where the gap reopens and a band inversion of pp and dd modes at the Γ\Gamma point takes place, and below the gap the number of eigenstates with even parity at the Γ\Gamma point does not equal to that at the M point. For simplicity, couplings are chosen to satisfy 2J0+J1=3J2J_0 + J_1 = 3J and ∣J0−J1∣=0.3J\lvert J_0 - J_1\rvert= 0.3J, to make gaps in (a) and (c) overlapped. (d) Magnon modes at the Γ\Gamma point, with the spin precession denoted by gray cones, where the projections of spins on xyxy plane satisfy the parity of pp and dd basis functions respectively. (e) and (f) Schematic magnon currents of magnon modes with up- and down-pseudospin, circulating counterclockwise and clockwise respectively, where ∣p±⟩=12(∣px⟩±i∣py⟩)\lvert p_{\pm}\rangle= \frac{1}{\sqrt{2}}(\lvert p_x\rangle\pm\mathrm{i}\lvert p_y\rangle) and ∣d±⟩=12(∣dx2−y2⟩±i∣d2xy⟩)\lvert d_{\pm}\rangle= \frac{1}{\sqrt{2}}(\lvert d_{x^2-y^2}\rangle\pm\mathrm{i}\lvert d_{2xy}\rangle).

Because Hamiltonian (1) preserves the inversion symmetry, eigen wavefunctions of Hamiltonian (5) can be indexed by parity eigenvalues, even and odd, at the Γ\Gamma point and M point. In the case J0>J1J_0 > J_1 [see Figure 2(a)], the two degenerate wavefunctions below the gap are ∣px⟩\lvert p_x\rangle and ∣py⟩\lvert p_y\rangle with odd parity at the Γ\Gamma point, whereas the two degenerate wavefunctions above the gap are ∣dx2−y2⟩\lvert d_{x^2-y^2}\rangle and ∣d2xy⟩\lvert d_{2xy}\rangle with even parity. The pp and dd magnon modes are shown in Figure 2(d). The parity of each band at the M point is the same as that at the Γ\Gamma point, denoting a topologically trivial state38,39. In the case J0<J1J_0 < J_1, the frequencies of pp and dd modes are inverted at the Γ\Gamma point as shown in Figure 2(c). Now, at the Γ\Gamma point, all three eigenstates below the gap are of even parity, while at the M point, there are two eigenstates with odd parity and one eigenstate with even parity. The unequal numbers of eigenstates with even parity indicate a topological state38,39. The band structure of topological state in Figure 2(c) cannot be continually transformed from that of the topologically trivial state in Figure 2(a) without closing the gap [see Figure 2(b)].

The double degeneracy of pp (dd) modes at the Γ\Gamma point are protected by the C6vC_{6v} symmetry, where the wavefunctions of pp (dd) modes correspond to the basis functions of two-dimensional irreducible representation E1E_1 (E2E_2). In order to better understand the symmetry at the Γ\Gamma point, we introduce a pseudo time-reversal operator T=iσyK\mathcal{T} = \mathrm{i}\sigma_y\mathcal{K}, where σy\sigma_y is Pauli matrix referring to the basis functions of E1E_1 and E2E_2, K\mathcal{K} denotes the complex conjugate operator32,35. This pseudo time-reversal operator originates from the C6vC_{6v} symmetry and the time-reversal symmetry of the system, which produces the Kramers doubling in the present bosonic system. Kramers pairs associated with pp or dd orbitals are constructed as ∣p±⟩=12(∣px⟩±i∣py⟩)\lvert p_{\pm}\rangle= \frac{1}{\sqrt{2}}(\lvert p_x\rangle\pm\mathrm{i}\lvert p_y\rangle) and ∣d±⟩=12(∣dx2−y2⟩±i∣d2xy⟩)\lvert d_{\pm}\rangle= \frac{1}{\sqrt{2}}(\lvert d_{x^2-y^2}\rangle\pm\mathrm{i}\lvert d_{2xy}\rangle) respectively, which carry specific pseudospins defined on the “artificial molecule”. In the present system, the magnon current Iij=⟨I^ij⟩I_{ij} = \langle\hat{I}_{ij}\rangle is given by the current operator

I^ij=−iSJijℏ(bi†bj−bj†bi).(17)\hat{I}_{ij} = -\frac{\mathrm{i} S J_{ij}}{\hbar}(b_i^\dagger b_j - b_j^\dagger b_i). \tag*{(17)}

For magnon modes with pseudospin up/down, magnon currents circulate counterclockwise/clockwise, as shown in Figure 2(e) and 2(f).

Around the Γ\Gamma point, magnon modes are predominantly occupied by the pp orbitals and dd orbitals. We can rewrite Hamiltonian

(5) in the basis of ∣p+⟩,∣d+⟩,∣p−⟩,∣d−⟩\lvert p_+\rangle,\lvert d_+\rangle,\lvert p_-\rangle,\lvert d_-\rangle up to the lowest order of momentum k\mathbf{k} as

Heff,k=[Hpd,k00Hpd,−k∗](18)H_{\mathrm{eff},\mathbf{k}} = \begin{bmatrix} H_{\mathrm{pd},\mathbf{k}} & 0 \\ 0 & H_{\mathrm{pd},-\mathbf{k}}^{*} \end{bmatrix} \tag*{(18)}

with

Hpd,k=E0[1001]+S[(J1−J0)−92J1a02k2−32iJ1a0(kx+iky)32iJ1a0(kx−iky)(J0−J1)+92J1a02k2],(19)H_{\mathrm{pd},\mathbf{k}} = E_0 \begin{bmatrix} 1 & 0 \\ 0 & 1 \end{bmatrix} + S \begin{bmatrix} (J_1-J_0)-\frac{9}{2}J_1a_0^2k^2 & -\frac{3}{2}iJ_1a_0(k_x+ik_y) \\ \frac{3}{2}iJ_1a_0(k_x-ik_y) & (J_0-J_1)+\frac{9}{2}J_1a_0^2k^2 \end{bmatrix}, \tag*{(19)}

where a0a_0 is the length of unit vector [see Figure 1]. This effective Hamiltonian is similar to the Bernevig-Hughes-Zhang model for quantum spin Hall effect and that found for topological photonic crystals[3] [32]. In Hamiltonian (19), E0E_0 simply makes a constant shift in the eigenvalues, and

M=J0−J1(20)M = J_0-J_1 \tag*{(20)}

is the effective Dirac mass, which is positive in a topologically trivial state. In the case J0<J1J_0 < J_1, MM becomes negative and a band inversion takes place, which turns the system into a topological state with a topological gap 2(J1−J0)2(J_1-J_0), as discussed above and shown in Figure 2(c). It is noticed that the topological magnon modes in the present system relies on crystalline symmetry and has a weaker topology comparing to those systems with the DM interaction. Nevertheless, we expect that stable unidirectional interface states between topological and trivial domains can be observed experimentally, similarly to photonic systems40–43.

In order to investigate topological interface magnon modes, we consider a heterostructure as illustrated in Figure 3(a), which is uniform and infinitely long in the xx direction. In the yy direction, the heterostructure contains 60 unit cells in both trivial and topological domains periodically, for the simplicity of calculation. For each interface, two interface dispersions appear inside the bulk gap in the frequency band structure shown in Figure 3(b), which is obtained numerically using the large unit cell for the heterostructure [Figure 3(a)]. The interface magnon modes at the two momenta marked by dots in Figure 3(b) are shown in Figure 4(a) and Figure 4(b) respectively. The topological magnon modes correspond to precessions of spins around the zz axis, with amplitudes decaying into the bulks exponentially. In Figure 4(c) and Figure 4(d), magnon currents of topological interface magnon modes with counterclockwise circulation and clockwise circulation in unit cells correspond to the up- and down-pseudospin respectively. The net current for up-/down-pseudospin flows to the positive/negative xx direction, manifesting the pseudospin-momentum locking in the present topological magnon modes.

Image not extracted

Figure 3. (a) Ferromagnetic heterostructure including a trivial domain and a topological domain, which is uniform and infinitely long in the xx direction. In the yy direction, 60 unit cells are contained in both trivial and topological domains periodically, where the exchange couplings are the same as in Figure 2(a) and Figure 2(c) respectively. The exchange couplings inside unit cells are denoted by blue and red lines, whereas those between unit cells are denoted by light and dark brown lines in the trivial domain and the topological domain respectively. The exchange couplings between the trivial and topological domains (denoted by gray lines) are given by the geometric mean of value of J1J_1 in Figure 2(a) and Figure 2(c). (b) Frequency band structure of magnon modes calculated based on the supercell denoted by the rectangular frame with asa_s the unit vector, where topological interface dispersions (red lines) appear in the bulk band gap.

Image not extracted

Figure 4. (a) and (b) Distribution of magnon modes at the two opposite momenta denoted by ↑ and ↓ in Figure 3(b) respectively. (c) and (d) Same as (a) and (b) except for distributions of magnon currents between sites presented by black arrows. The magnon mode denoted by ↑/↓ carries up-/down-pseudospin, showing counterclockwise/clockwise magnon currents in unit cells.

Topological magnon modes in antiferromagnets

Next, we explore topological magnon modes in antiferromagnets on honeycomb lattice. Antiferromagnets exhibit degenerate dispersions guaranteed by the combination of time-reversal and inversion symmetries, where no Dirac cone exists33. Therefore, it is not straightforward to realize topological magnon modes using the procedure formulated for ferromagnets directly. In order to overcome this difficulty, we introduce an external magnetic field to lift the double degeneracy. The Heisenberg model for antiferromagnets is

H=∑⟨i,j⟩JijSi⋅Sj−Bext∑iSiz(21)H=\sum_{\langle i,j\rangle}J_{ij}\mathbf{S}_i\cdot\mathbf{S}_j-B_{\mathrm{ext}}\sum_i S_i^z \tag*{(21)}

where exchange couplings Jij>0J_{ij}>0, the external magnetic field Bext>0B_{\mathrm{ext}}>0, and spin lengths Si=SS_i=S, and for simplicity, the external magnetic field is perpendicular to the two-dimensional spin plane. Without single ion anisotropy considered here, spins are automatically aligned in the plane perpendicular to the direction of external magnetic field and canted as shown schematically in Figure 5(a). The canting angle θ\theta of an antiferromagnet under external magnetic field is determined by the competition between exchange couplings and the Zeeman energy. We consider uniform exchange couplings Jij=JJ_{ij}=J as shown in Figure 5(a). The energy of a classical canted antiferromagnetic state with NN rhombic unit cells is

Image not extracted

Figure 5. (a) Antiferromagnet on honeycomb lattice under an external magnetic field BextB_{\mathrm{ext}} perpendicular to the lattice. The rhombic unit cell denoted by the dashed line contains two sites A and B. (b) Transformation between the laboratory frame and the rotating frame. xx, yy and zz (denoted by black arrows) are coordinate axes of the laboratory frame, where the external magnetic field BextB_{\mathrm{ext}} induces a canting angle θ\theta of spin relative to the xyxy plane, while x′x', y′y' and z′z' (denoted by blue arrows) are coordinate axes of the rotating frame. The transformation is achieved by rotating coordinate axes around the zz axis with an angle ϕ\phi then around the y′y' axis with an angle π/2−θ\pi/2-\theta. (c) Frequency band structure of magnon modes with double degenerate dispersions for an antiferromagnet in absence of external magnetic field. (d) Frequency band structure of magnon modes for a canted antiferromagnet induced by an external magnetic field corresponding to sin⁡2θ=0.4\sin^2\theta=0.4 (see text for details), where the degeneracy in (c) is lifted except for the KK and K′K' points.

Eclassical=−3JNS2cos⁡2θ−2BextNSsin⁡θ,(22)E_{\mathrm{classical}}=-3JNS^2\cos2\theta-2B_{\mathrm{ext}}NS\sin\theta, \tag*{(22)}

which is minimized when spins are canted at an angle sin⁡θ=Bext/(6JS)\sin\theta=B_{\mathrm{ext}}/(6JS).

In order to apply the Holstein–Primakoff transformation for antiferromagnets, we need a rotating frame where the zz coordinate axis is rotated to the local spin direction at each site [see Figure 5(b)]. The transformation between the rotating frame and the laboratory frame is

\begin{bmatrix} S^x\\ S^y\\ S^z \end{bmatrix} = \begin{bmatrix} \sin\theta\cos\phi& -\sin\phi& \cos\theta\cos\phi\\ \sin\theta\sin\phi& \cos\phi& \cos\theta\sin\phi\\ -\cos\theta& 0 & \sin\theta \end{bmatrix} \begin{bmatrix} S^{x'}\\ S^{y'}\\ S^{z'} \end{bmatrix}, \tag*{(23)}\end{bmatrix}

where SxS^x, SyS^y, SzS^z are spin components of the laboratory frame, Sx′S^{x'}, Sy′S^{y'}, Sz′S^{z'} are those of the rotating frame, ϕ\phi is the angle between the projection of a spin in the xyxy plane and the coordinate axis xx and θ\theta is the canting angle of spin. Note that projections of spins on different sublattices point to opposite directions on the xyxy plane, namely ϕ=ϕ0\phi=\phi_0 (ϕ=π+ϕ0\phi=\pi+\phi_0) for spins on A (B) sites, a classical Néel order to host magnon excitations. Now Hamiltonian (21) becomes

H=J∑⟨i,j⟩[cos⁡2θ(SixSjx−SizSjz)−SiySjy−sin⁡2θ(SixSjz+SizSjx)]−B∑i[sin⁡θSiz−cos⁡θSix],(24)\begin{aligned} H={}&J\sum_{\langle i,j\rangle}\left[\cos2\theta\left(S_i^xS_j^x-S_i^zS_j^z\right)-S_i^yS_j^y-\sin2\theta\left(S_i^xS_j^z+S_i^zS_j^x\right)\right]\\ &-B\sum_i\left[\sin\theta S_i^z-\cos\theta S_i^x\right], \tag*{(24)} \end{aligned}

which does not depend explicitly on ϕ0\phi_0. Adapting the Holstein-Primakoff transformation (2) for Six,y,zS_i^{x,y,z} and applying the Fourier transformation (5), a Bogoliubov Hamiltonian for magnon excitations is derived from Hamiltonian (24) as

H=12∑kΨk†HCAF,kΨk,(25)H=\frac{1}{2}\sum_k\Psi_k^\dagger H_{\mathrm{CAF},k}\Psi_k, \tag*{(25)}

where

HCAF,k=JS[3−sin⁡2θfk0(1−sin⁡2θ)fk−sin⁡2θfk∗3(1−sin⁡2θ)fk∗00(1−sin⁡2θ)fk3−sin⁡2θfk(1−sin⁡2θ)fk∗0−sin⁡2θfk∗3],(26)H_{\mathrm{CAF},k}=JS\begin{bmatrix} 3 & -\sin^2\theta f_k & 0 & (1-\sin^2\theta)f_k\\ -\sin^2\theta f_k^* & 3 & (1-\sin^2\theta)f_k^* & 0\\ 0 & (1-\sin^2\theta)f_k & 3 & -\sin^2\theta f_k\\ (1-\sin^2\theta)f_k^* & 0 & -\sin^2\theta f_k^* & 3 \end{bmatrix}, \tag*{(26)}

Ψk=[bA,k,bB,k,bA,−k†,bB,−k†]T\Psi_k=[b_{A,k},b_{B,k},b_{A,-k}^\dagger,b_{B,-k}^\dagger]^T, fk=∑ieik⋅δif_k=\sum_i e^{i\mathbf{k}\cdot\boldsymbol{\delta}_i} with δi\boldsymbol{\delta}_i being the n.n. vectors shown in Figure 5(a). Note that the terms linear in bib_i or bi†b_i^\dagger cancel out for sin⁡θ=Bext/(6JS)\sin\theta=B_{\mathrm{ext}}/(6JS), which minimizes Eq. (22). Because the classical Néel order is not the ground state for quantum antiferromagnets, bbbb and b†b†b^\dagger b^\dagger terms that do not conserve the total number of magnon appear, as such the basis vector Ψk\Psi_k including both creation and annihilation operators is chosen.

Dispersions of magnon modes can be derived by applying the bosonic Bogoliubov transformation44, 45 Ψk=TkΓk\Psi_k=T_k\Gamma_k to diagonalize HCAF,kH_{\mathrm{CAF},k},

Ψk†HCAF,kΨk=Γk†EkΓk,(27)\Psi_k^\dagger H_{\mathrm{CAF},k}\Psi_k=\Gamma_k^\dagger E_k\Gamma_k, \tag*{(27)}

where Γk=[γ1,k,γ2,k,γ1,−k†,γ2,−k†]T\Gamma_k=[\gamma_{1,k},\gamma_{2,k},\gamma_{1,-k}^\dagger,\gamma_{2,-k}^\dagger]^T is a new set of creation and annihilation operators for magnons and EkE_k is a diagonalized matrix. Magnons both before and after the Bogoliubov transformation should obey the same commutation relation for bosons in Eq. (14), demanding that the matrix TkT_k satisfies TkI^Tk†=I^T_k\hat{I}T_k^\dagger=\hat{I} with I^=diag⁡(1,1,−1,−1)\hat{I}=\operatorname{diag}(1,1,-1,-1). Here, TkT_k is not a unitary matrix since (Tk†)−1=I^TkI^(T_k^\dagger)^{-1}=\hat{I}T_k\hat{I}, which leaves HCAF,kTk=(Tk†)−1Ek≠TkEkH_{\mathrm{CAF},k}T_k=(T_k^\dagger)^{-1}E_k\ne T_kE_k in general. Therefore, one cannot calculate EkE_k from the characteristic equation of matrix HCAF,kH_{\mathrm{CAF},k} directly, since columns of TkT_k are not eigenvectors of HCAF,kH_{\mathrm{CAF},k}.

Multiplying I^(Tk†)−1\hat{I}(T_k^\dagger)^{-1} to both sides of Tk†HCAF,kTk=EkT_k^\dagger H_{\mathrm{CAF},k}T_k=E_k from left, we obtain the relation

I^HCAF,kTk=TkI^Ek,(28)\hat{I}H_{\mathrm{CAF},k}T_k=T_k\hat{I}E_k, \tag*{(28)}

namely columns of TkT_k are eigenvectors of I^HCAF,k\hat{I}H_{\mathrm{CAF},k}. The diagonalized matrix I^Ek\hat{I}E_k can be derived from the characteristic equation of matrix I^HCAF,k\hat{I}H_{\mathrm{CAF},k}, det⁡∣I^HCAF,k−I^Ek∣=0\det|\hat{I}H_{\mathrm{CAF},k}-\hat{I}E_k|=0, which is the conventional way to obtain I^Ek\hat{I}E_k and TkT_k.

Because Ψk\Psi_k and Γk\Gamma_k include both creation and annihilation operators of magnons, dispersions in Eq. (27) are redundant, which should be eliminated by symmetries. First, we define a particle-hole operator

C=[0I2I20]KP,(29)\mathcal{C}=\begin{bmatrix}0&I_2\\I_2&0\end{bmatrix}\mathcal{K}\mathcal{P}, \tag*{(29)}

where I2I_2 is a 2×22\times2 identity matrix, K\mathcal{K} is the complex conjugate operator and P\mathcal{P} is the space-inversion operator. For I^HCAF,k\hat{I}H_{\mathrm{CAF},k}, we can directly check from Eq. (26) and Eq. (29) that I^HCAF,kC=−CI^HCAF,k\hat{I}H_{\mathrm{CAF},k}\mathcal{C}=-\mathcal{C}\hat{I}H_{\mathrm{CAF},k}, leading to I^Ek=diag⁡(ε1,k,ε2,k,−ε1,k,−ε2,k)\hat{I}E_k=\operatorname{diag}(\varepsilon_{1,k},\varepsilon_{2,k},-\varepsilon_{1,k},-\varepsilon_{2,k}). With the same method, we can also check that PI^HCAF,k=I^HCAF,−kP\mathcal{P}\hat{I}H_{\mathrm{CAF},k}=\hat{I}H_{\mathrm{CAF},-k}\mathcal{P}, which leads to εn,k=εn,−k\varepsilon_{n,k}=\varepsilon_{n,-k} (n=1,2n=1,2). In this way, Hamiltonian (25) is diagonalized into44–46

H^CAF=∑n=1,2;kεn,kγn,k†γn,k,(30)\hat{H}_{\mathrm{CAF}}=\sum_{n=1,2;k}\varepsilon_{n,k}\gamma_{n,k}^\dagger\gamma_{n,k}, \tag*{(30)}

with

γn,k†=∑i(uA,n,kbi,k†−vA,n,−k∗bi,−k−uB,n,kbi,k†−vB,n,−k∗bi,−k),(31)\gamma_{n,k}^\dagger=\sum_i\left(u_{A,n,k}b_{i,k}^\dagger-v_{A,n,-k}^*b_{i,-k}-u_{B,n,k}b_{i,k}^\dagger-v_{B,n,-k}^*b_{i,-k}\right), \tag*{(31)}

where (uA,n,k,uB,n,k,vA,n,−k∗,vB,n,−k∗)(u_{A,n,k},u_{B,n,k},v_{A,n,-k}^*,v_{B,n,-k}^*) is the nn-th column of TkT_k.

Then the dispersions of magnon modes for the canted antiferromagnet shown in Figure 5(a) can be derived from the above procedure explicitly as33

ε1,k=SJ9+(2sin⁡2θ−1)∣fk∣2+6sin⁡2θ∣fk∣(32)\varepsilon_{1,k}=SJ\sqrt{9+(2\sin^{2}\theta-1)|f_k|^{2}+6\sin^{2}\theta|f_k|} \tag*{(32)}
ε2,k=SJ9+(2sin⁡2θ−1)∣fk∣2−6sin⁡2θ∣fk∣(33)\varepsilon_{2,k}=SJ\sqrt{9+(2\sin^{2}\theta-1)|f_k|^{2}-6\sin^{2}\theta|f_k|} \tag*{(33)}

It is clear that without an external magnetic field (θ=0\theta= 0), the dispersion is doubly degenerate in the whole Brillouin zone [see Fig. 5(c)] where no Dirac cone exists, unlike ferromagnets [see Fig. 2(b)]. A finite external magnetic field lifts the degeneracy, except for the KK and K′K' points where fK=fK′=0f_K = f_{K'} = 0 and εi,K=εi,K′=3SJ\varepsilon_{i,K} = \varepsilon_{i,K'} = 3SJ, as shown in Figure 5(d).

By choosing a hexagonal unit cell for canted antiferromagnets with a coupling texture as shown in Figure 6(a), where the canting angle is determined by sin⁡θ=Bext/2S(2J0+J1)\sin\theta= B_{\mathrm{ext}}/2S(2J_{0}+J_{1}) now, Hamiltonian (26) becomes

Image not extracted

Figure 6. (a) Same as Figure 5(a) except for that a coupling texture is introduced, where hexagonal unit cells are chosen with exchange couplings J0/J1J_{0}/J_{1} inside/between unit cells. (b) Frequency band structure of magnon modes for a canted antiferromagnet with J0>J1J_{0} > J_{1}. (c) Same as (b) except for J0=J1J_{0} = J_{1}. The KK and K′K' points in Figure 5(d) are folded into the Γ\Gamma point. (d) Same as (b) except for J0<J1J_{0} < J_{1}. Canting angle sin⁡2θ=0.4\sin^{2}\theta= 0.4 is considered for (b), (c) and (d). (e) Magnon modes for canted antiferromagnets at the Γ\Gamma point. In the rotating frame defined in Figure 5(a), where the rotating frame at A and B sublattice are different by a π\pi rotation of ϕ\phi, magnon modes look exactly the same way as Figure 2(d). For simplicity, ϕ=±π/2\phi= \pm\pi/2 is taken.

HCAF,k=[(1−sin⁡2θ)E0I6−sin⁡2θHF,k(1−sin⁡2θ)(E0I6−HF,k)(1−sin⁡2θ)(E0I6−HF,k)(1−sin⁡2θ)E0I6−sin⁡2θHF,k](34)H_{\mathrm{CAF},k}= \begin{bmatrix} (1-\sin^{2}\theta)E_{0}I_{6}-\sin^{2}\theta H_{\mathrm{F},k} & (1-\sin^{2}\theta)(E_{0}I_{6}-H_{\mathrm{F},k}) \\ (1-\sin^{2}\theta)(E_{0}I_{6}-H_{\mathrm{F},k}) & (1-\sin^{2}\theta)E_{0}I_{6}-\sin^{2}\theta H_{\mathrm{F},k} \end{bmatrix} \tag*{(34)}

where E0=S(2J0+J1)E_{0}=S(2J_{0}+J_{1}), I6I_{6} is a 6×66\times6 identity matrix and HF,kH_{\mathrm{F},k} is Hamiltonian (7) for ferromagnets.

Eigenvalues of canted antiferromagnets can be derived in terms of eigenvalues of ferromagnets with the same coupling texture. Here we consider the particle solution of Eq. (28) which has positive energy εn,k\varepsilon_{n,k} with n=1,2,…,6n=1, 2, \ldots, 6. All four blocks of HCAF,kH_{\mathrm{CAF},k} in Eq. (34) are diagonalized in the basis of eigenstates of HF,kH_{\mathrm{F},k}, so that for an eigenstate ψ\psi of HF,kH_{\mathrm{F},k} with eigenvalue εF,n,k\varepsilon_{\mathrm{F},n,k}, the state (ψ,αψ)T(\psi,\alpha\psi)^{\mathrm{T}} is an eigenstate of I^HCAF,k\hat{I}H_{\mathrm{CAF},k} in Eq. (28), provided

(1−sin⁡2θ)(1+αE0)−(sin⁡2θ+α−αsin⁡2θ)εF,n,k=εn,k(35)(1-\sin^{2}\theta)(1+\alpha E_{0})-(\sin^{2}\theta+\alpha-\alpha\sin^{2}\theta)\varepsilon_{\mathrm{F},n,k}=\varepsilon_{n,k} \tag*{(35)}
−(1−sin⁡2θ)(1+αE0)+(1−sin⁡2θ+αsin⁡2θ)εF,n,k=αεn,k(36)-(1-\sin^{2}\theta)(1+\alpha E_{0})+(1-\sin^{2}\theta+\alpha\sin^{2}\theta)\varepsilon_{\mathrm{F},n,k}=\alpha\varepsilon_{n,k} \tag*{(36)}

leading to

εn,k=(2sin⁡2θ−1)εF,n,k2+2E0(1−sin⁡2θ)εF,n,k(37)\varepsilon_{n,k}=\sqrt{(2\sin^{2}\theta-1)\varepsilon_{\mathrm{F},n,k}^{2}+2E_{0}(1-\sin^{2}\theta)\varepsilon_{\mathrm{F},n,k}} \tag*{(37)}

For sin⁡2θ>1/3\sin^2\theta> 1/3, ϵn,k\epsilon_{n,k} is monotonic with ϵF,n,k\epsilon_{\mathrm{F},n,k}. The pp-dd band inversion is achieved in canted antiferromagnets by tuning coupling texture the same as in the ferromagnetic case. For sin⁡2θ=0.4\sin^2\theta= 0.4, the full frequency band structures of canted antiferromagnets with the coupling textures are derived as shown in Figure 6(b)–(Figure 6d), and pp and dd magnon modes in canted antiferromagnets are shown in Figure 6(e). For sin⁡2θ<1/3\sin^2\theta< 1/3, the present scheme does not apply straightforwardly.

In order to see the topological interface magnon modes, we consider a heterostructure shown in Figure 7(a) similar to Figure 3(a) for the ferromagnet. Topological interface dispersions appear inside the band gap of the frequency band structure as shown in Figure 7(b). Magnon modes at the two momenta marked by dots in Figure 7(b) are depicted in Figure 8(a) and Figure 8(b) respectively. We can calculate the magnon current for particle and hole parts of canted antiferromagnet using the same way of ferromagnetic case. The total magnon current is the magnon current of the particle part minus that of the hole part. Similar to topological interface magnon modes of the ferromagnetic case, in Figure 8(c) and Figure 8(d) magnon current circulates counterclockwise/clockwise in unit cells dominated by the up-/down-pseudospin, which also governs the direction of net magnon currents, demonstrating the pseudospin-momentum locking phenomenon in these topological interface magnon modes. Because the external magnetic field induces a ferromagnetic component Ssin⁡θS\sin\theta on each site, the magnitude of magnon current is proportional to sin⁡2θ\sin^2\theta.

Image not extracted

Figure 7. (a) Canted antiferromagnetic heterostructure containing a trivial domain and a topological domain, which is uniform and infinitely long in the xx direction. In the yy direction, 60 unit cells are contained in both trivial and topological domains periodically, where the exchange couplings are the same as in Figure 6(b) and Figure 6(d) respectively, which are denoted in the same way as in Figure 3(a), and the couplings between the trivial and topological domains (denoted by gray lines) are given by geometric mean of J1J_1. (b) Frequency band structure of magnon modes for the heterostructure, where topological interface dispersions (red lines) appear in the bulk band gap.

Image not extracted

Figure 8. (a) and (b) Distribution of magnon modes at the two opposite momenta denoted by ↑\uparrow and ↓\downarrow in Figure 7(b) respectively. (c) and (d) Same as (a) and (b) except for current distributions of magnon currents between sites presented by black arrows. The magnon mode denoted by ↑/↓\uparrow/\downarrow carry up-/down-pseudospin, showing counterclockwise/clockwise magnon currents in unit cells.

Discussion

We propose a method to achieve topological magnon modes in magnetic systems on honeycomb lattice, including both ferromagnet and antiferromagnet. The frequency band structures are gapless for uniform nearest-neighbor exchange couplings. In ferromagnets, a topological frequency gap opens when exchange couplings inside the hexagonal unit cells are smaller than exchange couplings between unit cells, associated with a pp-dd band inversion at the Γ\Gamma point. In antiferromagnets, the degeneracy in the frequency band structure due to the combination of time-reversal symmetry and inversion symmetry has to be lifted by applying an external magnetic field. The resulting canted antiferromagnets become topological upon tuning exchange couplings inside and between hexagonal unit cells same as in ferromagnets. Magnon currents of topological magnon modes propagate along the interface between a trivial domain and a topological domain in opposite directions governed by pseudospins, the circulation direction of magnon current in unit cells, manifesting the pseudospin-momentum locking phenomenon. In the future, candidate materials should be found to realize the scheme proposed in the present work. Besides specific materials, one may also observe topological magnon modes in artificial systems which can be achieved by depositing magnetic atoms on a metallic substrate using the STM technique or trapping magnetic atoms in an optical lattice using laser beams.

Methods

The effective Hamiltonians of magnon modes in ferromagnet and antiferromagnet on honeycomb lattice are obtained by the Holstein-Primakoff approach. Besides, a local rotating frame and the Bogoliubov transformation are adopted for canted antiferromagnet. Frequency band structures and magnon modes are obtained by direct diagonalizations of Hamiltonian.

References

Acknowledgements

This work is supported by CREST, JST (Core Research for Evolutionary Science and Technology, Japan Science and Technology Agency), Grant No. JPMJCR18T4.

Author contributions statement

X.H. supervised the project. H.H. conducted the numerical calculation. T.K. joined the discussions and writing of manuscript. All authors contributed to the discussions and the preparation of the manuscript.

Additional information

Competing interests: The authors declare that they have no competing interests. The corresponding author is responsible for submitting a competing interests statement on behalf of all authors of the paper. This statement must be included in the submitted article file.

References

  1. [1]Haldane, F. D. M. Model for a quantum hall effect without landau levels: Condensed-matter realization of the "parity anomaly". Phys. Rev. Lett. 61, 2015–2018, DOI: 10.1103/PhysRevLett.61.2015 (1988).DOI
  2. [2]Kane, C. L. & Mele, E. J. Quantum spin hall effect in graphene. Phys. Rev. Lett. 95, 226801, DOI: 10.1103/PhysRevLett.95.226801 (2005).DOI
  3. [3]Bernevig, B. A., Hughes, T. L. & Zhang, S.-C. Quantum spin hall effect and topological phase transition in hgte quantum wells. Science 314, 1757–1761, DOI: 10.1126/science.1133734 (2006).DOI

Paper details

vlnc:W_j75gngsu
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