This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Asymmetric modulation of Majorana excitation spectra and nonreciprocal thermal transport in the Kitaev spin liquid under a staggered magnetic field

Kazuki Nakazawa, Yasuyuki Kato, and Yukitoshi Motome [email protected] Department of Applied Physics, University of Tokyo, Tokyo 113-8656, Japan
Abstract

We present the theoretical results for the Majorana band structure and thermal transport properties in the Kitaev honeycomb model under a staggered magnetic field in addition to a uniform one. The Majorana band structure is asymmetrically modulated in momentum space, and the valley degree of freedom is activated and controlled by the magnitudes and directions of the uniform and staggered magnetic fields; as a result, the Majorana excitation is either gapped or gapless, the latter of which in general has Majorana Fermi surfaces, in contrast to the point nodes in the absence of the fields. We show that the asymmetric deformation of the Majorana band leads to a nonreciprocal thermal transport. The field dependence of the linear thermal conductivity is correlated with the magnitude of the Majorana gap, while that of the nonlinear thermal conductivity is associated with the asymmetry of the Majorana excitation spectra. We show the estimates of the thermal currents and discuss that the linear response is experimentally measurable, whereas future improvement is required for the observation of the nonlinear one. We also discuss how the thermal currents are modulated by other additional effects of the magnetic field and contributions from conventional magnons, latter of which are expected in heterostructures to realize an internal staggered magnetic field.

I Introduction

The quantum spin liquid (QSL) is a spin state in insulating magnets which does not break any symmetry of the systems down to absolute zero temperature Balents ; SB ; KM . This exotic state is expected to be realized when magnetic frustration and quantum fluctuation are in action. Among many proposals, the Kitaev honeycomb model has drawn attention since its ground state provides a rare example of the QSL in more than one dimension Kitaev2006 . In the Kitaev spin liquid, the elementary excitations are described by itinerant Majorana fermions and localized Z2Z_{2} fluxes as a consequence of the fractionalization of the spin degree of freedom Kitaev2006 . The quantum entanglement between the fractional excitations would be utilized for decoherence-free topological quantum computing Kitaev2003 . Since it was pointed out that a class of materials with strong spin-orbit coupling and electron correlation is a good platform for the realization of the Kitaev model JK , extensive investigation has been performed toward the identification of the Kitaev spin liquid and the fractional excitations for the candidate materials, such as A2IrO4A_{2}{\rm IrO_{4}} (A=LiandNa)(A={\rm Li\ and\ Na}) and α{\alpha}-RuCl3 TTJKN ; MN ; Winter ; Trebst .

To capture the evidence of the fractionalization, the effect of a magnetic field has been studied both theoretically and experimentally. In theory, the magnetic field opens a gap at the Dirac-like nodal points at the Brillouin zone edges in the excitation spectrum of the itinerant Majorana fermions, and the system becomes a Majorana Chern insulator with a chiral Majorana edge mode in the weak field limit, which leads to the half-quantized thermal Hall effect Kitaev2006 ; NYM . Experimentally, the external magnetic field suppresses the parasitic long-range order stabilized by residual non-Kitaev interactions and may realize the Kitaev spin liquid in the field Wolter ; Jansa ; Banerjee2018 ; Zheng ; Nagai . Indeed, the half-quantized thermal Hall effect was observed in the field-induced quantum paramagnetic region in one of the candidates, α{\alpha}-RuCl3 Kasahara ; Yokoi ; Yamashita ; Bruin , which is regarded as strong evidence of the chiral Majorana edge modes in the gapped topological Majorana state.

Recently, modulations of the Majorana excitation spectrum by the magnetic field have been examined in a wider scope. For instance, an asymmetric modulation of the Majorana band structure was discussed as a combined effect of a site-dependent internal magnetic field and a non-Kitaev interaction TF2 . A similar but different type of asymmetric modulation was discussed by introducing an electric field in addition to the magnetic field CMR . In both cases, the sublattice symmetry is broken, which activates the valley degree of freedom and may bring about the Majorana Fermi surfaces depending on the parameters. Such “Majorana band engineering” would be useful for identifying the fractional excitations in the Kitaev spin liquid, but the systematic study remains yet to be investigated, especially the consequence on the thermal transport which is sensitive to the modulations of the Majorana excitation spectrum.

In this paper, we perform a systematic study of the Majorana excitation spectrum and the thermal transport in the Kitaev model, by introducing both uniform and staggered magnetic fields. Within the perturbation theory in terms of the magnetic fields, we show that the Majorana band structure is modulated in an asymmetric way depending on the magnitudes and directions of the magnetic fields, which allows us to control the valley degree of freedom. Furthermore, it becomes gapless with the formation of the Majorana Fermi surfaces in some parameter regions of the magnetic fields. For these situations, we calculate the longitudinal thermal transport coefficient by using the Boltzmann transport theory. We find that the thermal transport becomes nonreciprocal due to the asymmetric modulation of the Majorana band structure. By the comprehensive study while changing the uniform and staggered fields, we show that the linear component of the thermal current probes the Majorana excitation gap, while the nonlinear one is sensitive to the asymmetry of the Majorana band structure. We discuss our results with the estimates of the thermal currents expected in real materials. We also discuss other additional effects of the magnetic field, including the interaction between the Majorana fermions. Finally, we mention about contributions from the conventional magnons, which would be important in heterostructures with an antiferromagnet to realize an internal staggered magnetic field.

This paper is organized as follows. In Sec. II, we introduce the model that we use in this paper. In Sec. III, we discuss the Majorana excitation spectra while changing the uniform and staggered magnetic fields, with emphasis on the Majorana excitation gap, the effective Majorana density, and the asymmetry of the Majorana band structure. In Sec. IV, we present the results of the linear and nonlinear thermal transport coefficients for the modulated Majorana bands. In Sec. V, we discuss quantitative estimates of the thermal currents, the effect of the second-order perturbation and the Majorana interaction, and the contributions from the conventional magnons. Section VI is devoted to the summary.

II Model

Refer to caption
Figure 1: (a) Schematic of the Kitaev honeycomb model. The white and black circles represent the lattice sites on the A and B sublattices, respectively. The blue, green, and red lines represent the xx, yy, and zz bonds in Eq. (1), respectively. 𝒂1{\bm{a}}_{1} and 𝒂2{\bm{a}}_{2} denote the primitive translation vectors, i,j,ki,j,k and ,m,n\ell,m,n represent examples of the site indices in the summations in Eqs. (5) and (24), respectively, and 𝒓{{\bm{r}}} labels the zz bond in Eqs. (6) and (7). The real-space and spin-space coordinates are shown in the inset. A thermal gradient is applied along the aa or bb direction. (b) Directions of the uniform component of the magnetic field, 𝒉{\bm{h}} (red arrows), and the staggered one 𝒉st{\bm{h}}_{\rm st} (blue arrows); see Eqs. (3) and (4). The left panel shows the case with the in-plane 𝒉stc\bm{h}_{\rm st}\perp c, while the right one is for the out-of-plane 𝒉stc\bm{h}_{\rm st}\parallel c; θ\theta and φ\varphi denote the polar and azimuth angles for 𝒉\bm{h}, and ϕst\phi_{\rm st} is the azimuth angle for the in-plane 𝒉st{\bm{h}}_{\rm st}.

We start from the Kitaev model on the honeycomb lattice under the magnetic field which includes both uniform and staggered components. The Hamiltonian is given by =K+Z{\cal H}={\cal H}_{\rm K}+{\cal H}_{\rm Z}, where

K\displaystyle{\cal H}_{\rm K} =α=x,y,zi,jαJαSiαSjα,\displaystyle=-\sum_{\alpha=x,y,z}\sum_{\langle i,j\rangle_{\alpha}}J^{\alpha}S_{i}^{\alpha}S_{j}^{\alpha}, (1)
Z\displaystyle{\cal H}_{\rm Z} =i𝑯i𝑺i.\displaystyle=-\sum_{i}{\bm{H}}_{i}\cdot{\bm{S}}_{i}. (2)

Here, Siα=σiα/2S_{i}^{\alpha}={\sigma}_{i}^{\alpha}/2 is the α{\alpha} component of the spin-1/2 operator at site ii defined by the Pauli matrix σiα{\sigma}_{i}^{\alpha} (α=x,y,z\alpha=x,y,z; we set =1\hbar=1 here and for a while), JαJ^{\alpha} is the coupling constant for the Kitaev exchange interaction on the α{\alpha} bond, and 𝑯i{\bm{H}}_{i} denotes the magnetic field at site ii given by

𝑯iA\displaystyle{\bm{H}}_{i\in{\rm A}} =𝑯𝑯st=H(𝒉rst𝒉st)H𝒉A,\displaystyle={\bm{H}}-{\bm{H}}_{\rm st}=H\left({\bm{h}}-r_{\rm st}{\bm{h}}_{\rm st}\right)\equiv H{\bm{h}}_{\rm A}, (3)
𝑯iB\displaystyle{\bm{H}}_{i\in{\rm B}} =𝑯+𝑯st=H(𝒉+rst𝒉st)H𝒉B,\displaystyle={\bm{H}}+{\bm{H}}_{\rm st}=H\left({\bm{h}}+r_{\rm st}{\bm{h}}_{\rm st}\right)\equiv H{\bm{h}}_{\rm B}, (4)

for site ii belonging to the sublattice A and B, respectively (see Fig. 1). In Eqs. (3) and (4), 𝑯{\bm{H}} and 𝑯st{\bm{H}}_{\rm st} denote the uniform and staggered components, respectively; H=|𝑯|H=|{\bm{H}}|, |𝒉|=|𝒉st|=1|{\bm{h}}|=|{\bm{h}}_{\rm st}|=1, and rst=|𝑯st|/|𝑯|r_{\rm st}=|{\bm{H}}_{\rm st}|/|{\bm{H}}|.

Following the perturbation theory with respect to the uniform magnetic field up to third order Kitaev2006 , the Zeeman coupling term Z{\cal H}_{\rm Z} in Eq. (2) can be effectively described by three-spin interactions as

\displaystyle{\cal H}^{\prime} =δ{i,j,k}hixhjyhkzSixSjySkz,\displaystyle=-\delta\sum_{\{i,j,k\}}h_{i}^{x}h_{j}^{y}h_{k}^{z}S_{i}^{x}S_{j}^{y}S_{k}^{z}, (5)

where 𝒉i𝑯i/H=𝒉A(B){\bm{h}}_{i}\equiv{\bm{H}}_{i}/H={\bm{h}}_{\rm A(B)} for iA(B)i\in{\rm A(B)}, and δH3/Δf2\delta\sim H^{3}/\Delta_{\rm f}^{2} with Δf\Delta_{\rm f} being a (mean) flux excitation energy. In Eq. (5), the summation is taken for neighboring three sites {i,j,k}\{i,j,k\} as exemplified in Fig. 1(a). By using the Jordan-Wigner transformation CH ; FZX ; CN , both K{\cal H}_{\rm K} and {\cal H}^{\prime} can be written in terms of the Majorana fermions as

K\displaystyle{\cal H}_{\rm K} =i4𝒓[Jxc𝒓,Ac𝒓+𝒂1,B+Jyc𝒓,Ac𝒓+𝒂2,B+Jzc𝒓,Ac𝒓,B],\displaystyle=\frac{i}{4}\sum_{{{\bm{r}}}}\Bigl{[}J^{x}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}}+{\bm{a}}_{1},{{\rm B}}}+J^{y}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}}+{\bm{a}}_{2},{{\rm B}}}+J^{z}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}},{{\rm B}}}\Bigr{]}, (6)
\displaystyle{\cal H}^{\prime} =iδ8𝒓[h~Ayzxc𝒓𝒂3,Ac𝒓,A+h~Azxyc𝒓,Ac𝒓𝒂2,A+h~Axyzc𝒓𝒂1,Ac𝒓,A\displaystyle=-\frac{i\delta}{8}\sum_{{\bm{r}}}\Bigl{[}\tilde{h}_{\rm A}^{yzx}c_{{{\bm{r}}}-{\bm{a}}_{3},{{\rm A}}}c_{{{\bm{r}}},{{\rm A}}}+\tilde{h}_{\rm A}^{zxy}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}}-{\bm{a}}_{2},{{\rm A}}}+\tilde{h}_{\rm A}^{xyz}c_{{{\bm{r}}}-{\bm{a}}_{1},{{\rm A}}}c_{{{\bm{r}}},{{\rm A}}}
+h~Byzxc𝒓,Bc𝒓𝒂3,B+h~Bzxyc𝒓𝒂2,Bc𝒓,B+h~Bxyzc𝒓,Bc𝒓𝒂1,B],\displaystyle\quad\quad\quad\ +\tilde{h}_{\rm B}^{yzx}c_{{{\bm{r}}},{{\rm B}}}c_{{{\bm{r}}}-{\bm{a}}_{3},{{\rm B}}}+\tilde{h}_{\rm B}^{zxy}c_{{{\bm{r}}}-{\bm{a}}_{2},{{\rm B}}}c_{{{\bm{r}}},{{\rm B}}}+\tilde{h}_{\rm B}^{xyz}c_{{{\bm{r}}},{{\rm B}}}c_{{{\bm{r}}}-{\bm{a}}_{1},{{\rm B}}}\Bigr{]}, (7)

where c𝒓,A(B)c_{{{\bm{r}}},{{\rm A}}({{\rm B}})} is the Majorana fermion operator at the A(B) sublattice site on the zz bond at the position 𝒓\bm{r}, and h~A(B)αβγ=hA(B)αhB(A)βhA(B)γ\tilde{h}_{\rm A(B)}^{{\alpha}{\beta}{\gamma}}=h_{\rm A(B)}^{\alpha}h_{\rm B(A)}^{\beta}h_{\rm A(B)}^{\gamma}; 𝒂1=(1/2,3/2){\bm{a}}_{1}=(1/2,\sqrt{3}/2) and 𝒂2=(1/2,3/2){\bm{a}}_{2}=(-1/2,\sqrt{3}/2) are the primitive translational vectors and 𝒂3𝒂2𝒂1{\bm{a}}_{3}\equiv{\bm{a}}_{2}-{\bm{a}}_{1} (we set the lattice constant as unity) [see Fig. 1(a)]. Hereafter, we consider the case with the isotropic ferromagnetic Kitaev couplings, Jx=Jy=Jz=J>0J^{x}=J^{y}=J^{z}=J>0, while the results are the same for the antiferromagnetic case within the above perturbation theory. We note that the second-order perturbation with respect to the magnetic field modulates JαJ^{\alpha}, and the third-order perturbation gives rise to additional three-spin terms to Eq. (5) which correspond to interactions between the Majorana fermions. We will discuss these additional effects in Sec. V.2.

III Majorana excitation spectra

Refer to caption
Figure 2: Majorana band dispersions E𝒌±/JE_{{{\bm{k}}}\pm}/J in Eq. (16): (a) in the absence of the magnetic field, (b) under a uniform magnetic field only (δ=0.36J,θ=φ=0)(\delta=0.36J,\ \theta=\varphi=0), (c) under a staggered magnetic field only (δ=0.36J,ϕst=0)(\delta=0.36J,\ \phi_{\rm st}=0), and (d) in the presence of both uniform and staggered fields (δ=0.9J,ϕst=π/6,θ=π/3,φ=0)(\delta=0.9J,\ \phi_{\rm st}=\pi/6,\ \theta=\pi/3,\ \varphi=0). The left panels are the bird’s-eye views and the right panels are the two-dimensional plots along the symmetric lines in the first Brillouin zone depicted in the inset of (a). The green lines in the left panels of (c) and (d) denote the Majorana Fermi surfaces. The right panels of (c) and (d) include the data while changing δ\delta and rstr_{\rm st}, respectively.
Refer to caption
Figure 3: Contour plots of the Majorana gap ΔM/J\Delta_{\rm M}/J while varying the direction of the uniform component of the magnetic field, 𝒉{\bm{h}}, with the fixed staggered one 𝒉st{\bm{h}}_{\rm st} along the aa direction (ϕst=0)(\phi_{\rm st}=0) for (a) rst=0r_{\rm st}=0, (b) rst=0.2r_{\rm st}=0.2, (c) rst=0.6r_{\rm st}=0.6, and (d) rst=1.0r_{\rm st}=1.0. The cyan lines represent the boundary between gapped and gapless region, while the magenta lines in (a) and points in (d) represent the parameters where the Majorana dispersion has the nodal points at zero energy as in the pure Kitaev model in Fig. 1(a). We set δ=0.096J\delta=0.096J.
Refer to caption
Figure 4: Constant energy cuts at E𝒌±=0.35JE_{{{\bm{k}}}\pm}=0.35J around the K and K points. In (a) and (b), the direction of 𝒉{\bm{h}} is varied within the (a) acac and (b) abab planes, while 𝒉st{\bm{h}}_{\rm st} is fixed along the aa direction. In (c), the direction of 𝒉st{\bm{h}}_{\rm st} is varied within the abab plane, while 𝒉{\bm{h}} is fixed along the cc direction. In all the cases, we set rst=1r_{\rm st}=1 and δ=0.096J\delta=0.096J.

In this section, we discuss the Majorana excitation spectra in the effective Hamiltonian eff=K+{\cal H}_{\rm eff}={\cal H}_{\rm K}+{\cal H}^{\prime}. The Fourier transform of Eqs. (6) and (7) is summarized into

eff=𝒌BZ(ka>0)𝒄𝒌(Δ𝒌,Aif𝒌if𝒌Δ𝒌,B)𝒄𝒌,\displaystyle{\cal H}_{\rm eff}=\sum_{{{\bm{k}}}\in{\rm BZ}(k_{a}>0)}{\bm{c}}_{{\bm{k}}}^{\dagger}\left(\begin{array}[]{cc}\Delta_{{{\bm{k}}},{{\rm A}}}&if_{{\bm{k}}}\\ -if_{{\bm{k}}}^{*}&\Delta_{{{\bm{k}}},{{\rm B}}}\end{array}\right){\bm{c}}_{{\bm{k}}}, (10)

where the summation of the momentum 𝒌=(ka,kb){\bm{k}}=(k_{a},k_{b}) is taken over a half of the first Brillouin zone, and 𝒄𝒌=(c𝒌,A,c𝒌,B){\bm{c}}_{{\bm{k}}}^{\dagger}=\left(c_{{{\bm{k}}},{{\rm A}}}^{\dagger},c_{{{\bm{k}}},{{\rm B}}}^{\dagger}\right) and 𝒄𝒌=(c𝒌,A,c𝒌,B)t{\bm{c}}_{{\bm{k}}}={}^{t}\left(c_{{{\bm{k}}},{{\rm A}}},c_{{{\bm{k}}},{{\rm B}}}\right) are the Fourier components of the Majorana fermion operators defined as

c𝒌,A(B)=12Nu𝒓c𝒓,A(B)ei𝒌𝒓,\displaystyle c_{{{\bm{k}}},{{\rm A}}({{\rm B}})}^{\dagger}=\frac{1}{\sqrt{2N_{\rm u}}}\sum_{{\bm{r}}}c_{{{\bm{r}}},{{\rm A}}({{\rm B}})}{{\rm e}}^{i{{\bm{k}}}\cdot{{\bm{r}}}}, (11)
c𝒌,A(B)=12Nu𝒓c𝒓,A(B)ei𝒌𝒓,\displaystyle\ \ c_{{{\bm{k}}},{{\rm A}}({{\rm B}})}=\frac{1}{\sqrt{2N_{\rm u}}}\sum_{{\bm{r}}}c_{{{\bm{r}}},{{\rm A}}({{\rm B}})}{{\rm e}}^{-i{{\bm{k}}}\cdot{{\bm{r}}}}, (12)

where NuN_{\rm u} is the number of the unit cell, and

f𝒌\displaystyle f_{{\bm{k}}} =J2(ei𝒌𝒂1+ei𝒌𝒂2+1),\displaystyle=\frac{J}{2}\left({{\rm e}}^{i{{\bm{k}}}\cdot{\bm{a}}_{1}}+{{\rm e}}^{i{{\bm{k}}}\cdot{\bm{a}}_{2}}+1\right), (13)
Δ𝒌,A\displaystyle\Delta_{{{\bm{k}}},{{\rm A}}} =δ2(h~Axyzsin𝒌𝒂1h~Azxysin𝒌𝒂2+h~Ayzxsin𝒌𝒂3),\displaystyle=\frac{\delta}{2}\left(\tilde{h}_{\rm A}^{xyz}\sin{{\bm{k}}}\cdot{\bm{a}}_{1}-\tilde{h}_{\rm A}^{zxy}\sin{{\bm{k}}}\cdot{\bm{a}}_{2}+\tilde{h}_{\rm A}^{yzx}\sin{{\bm{k}}}\cdot{\bm{a}}_{3}\right), (14)
Δ𝒌,B\displaystyle\Delta_{{{\bm{k}}},{\rm B}} =δ2(h~Bxyzsin𝒌𝒂1h~Bzxysin𝒌𝒂2+h~Byzxsin𝒌𝒂3).\displaystyle=-\frac{\delta}{2}\left(\tilde{h}_{\rm B}^{xyz}\sin{{\bm{k}}}\cdot{\bm{a}}_{1}-\tilde{h}_{\rm B}^{zxy}\sin{{\bm{k}}}\cdot{\bm{a}}_{2}+\tilde{h}_{\rm B}^{yzx}\sin{{\bm{k}}}\cdot{\bm{a}}_{3}\right). (15)

Note that the Majorana operators Eqs. (11) and (12) satisfy c𝒌,A(B)=c𝒌,A(B)c_{{{\bm{k}}},{{\rm A}}({{\rm B}})}^{\dagger}=c_{-{{\bm{k}}},{{\rm A}}({{\rm B}})}. Then, the eigenenergy is given by the diagonalization of Eq. (10) as

E𝒌±=12(Δ𝒌+±4|f𝒌|2+Δ𝒌2),\displaystyle E_{{{\bm{k}}}\pm}=\frac{1}{2}\left(\Delta_{{{\bm{k}}}+}\pm\sqrt{4|f_{{\bm{k}}}|^{2}+\Delta_{{{\bm{k}}}-}^{2}}\ \right), (16)

with Δ𝒌±Δ𝒌,A±Δ𝒌,B\Delta_{{{\bm{k}}}\pm}\equiv\Delta_{{{\bm{k}}},{{\rm A}}}\pm\Delta_{{{\bm{k}}},{{\rm B}}}. Note that E𝒌±E_{{{\bm{k}}}\pm} is always particle-hole symmetric with respect to zero energy, namely E𝒌±=E𝒌E_{{{\bm{k}}}\pm}=-E_{-{{\bm{k}}}\mp}, because of the nature of the Majorana fermions, and the only positive energy parts contribute to the physical properties.

Let us first discuss the overall behavior of the dispersion relation in Eq. (16) while changing the uniform and staggered components of the magnetic fields. In the absence of the magnetic field H=0H=0, E𝒌±E_{{{\bm{k}}}\pm} has the Dirac-like linear dispersions with the gapless nodal points at the K and K’ points, as shown in Fig. 2(a) Kitaev2006 . When introducing a uniform magnetic field, the nodal points are gapped out, as shown in Fig. 2(b). In this situation, the insulating state becomes topologically nontrivial having a nonzero Chern number Kitaev2006 . Meanwhile, when we apply a staggered magnetic field only, the nodal points remain gapless, but their energies at the K and K’ points are shifted in an opposite energy direction, as shown in Fig. 2(c). This makes the two valleys at the K and K points inequivalent and E𝒌±E_{{{\bm{k}}}\pm} asymmetric in momentum space, because of the breaking of sublattice symmetry. As a consequence, the staggered magnetic field yields the Majorana Fermi surfaces around the K and K’ points, whose sizes are increased with increasing the strength of the staggered magnetic field, as shown in the right panel of Fig. 2(c). Finally, when we apply both uniform and staggered fields, the gap opening and the asymmetric energy shift occur simultaneously, as shown in Fig. 2(d). In this case, the system can be insulating or gapless with the Majorana Fermi surfaces depending on the magnitudes of the uniform and staggered components, as demonstrated in the right panel of Fig. 2(d).

Figure 3 displays the Majorana gap ΔM\Delta_{\rm M} while varying the direction of the uniform component of the magnetic field, 𝒉{\bm{h}}, with a staggered component 𝒉st{\bm{h}}_{\rm st} along the aa direction. When the staggered field is zero or small, the Majorana excitation spectrum is gapped in most cases, as shown in Figs. 3(a) and 3(b). In the gapped region, the system is the Majorana Chern insulator as in the case with the uniform magnetic field only. With an increase of the staggered field, the asymmetric energy shift at the K and K’ points becomes larger, leading to an increase of the gapless region, as shown in Figs. 3(c) and 3(d).

Figure 4 shows the asymmetric modulation of E𝒌±E_{{{\bm{k}}}\pm} by plotting the constant energy cuts at E𝒌±=0.35JE_{{{\bm{k}}}\pm}=0.35J around the K and K’ points. In Fig. 4(a), we present the results while rotating 𝒉{\bm{h}} within the acac plane with 𝒉st{\bm{h}}_{\rm st} along the aa direction. When 𝒉{\bm{h}} is parallel to 𝒉st{\bm{h}}_{\rm st} (θ=π/2\theta=\pi/2), the effective magnetic field h~A(B)αβγ\tilde{h}_{A(B)}^{\alpha\beta\gamma} in Eq. (7) becomes zero, and hence, the dispersion is equivalent to that in the absence of the magnetic field and the constant energy cuts are symmetric between the K and K’ points. The rotation of the direction of 𝒉{\bm{h}} modulates the dispersion in an asymmetric way; when the constant energy cut around the K’ point is enlarged, that around the K point is shrunk. We note that the asymmetry is always present in the kak_{a} direction except for θ=π/2\theta=\pi/2, while the dispersion becomes symmetric in the kbk_{b} direction for θ=0\theta=0, π/4\pi/4, and 3π/43\pi/4. On the other hand, when 𝒉{\bm{h}} is rotated within the abab plane, the dispersion becomes asymmetric in both kak_{a} and kbk_{b} directions, except for the symmetric case with φ=0\varphi=0, as shown in Fig. 4(b). Finally, Fig. 4(c) shows the results while rotating 𝒉{\bm{h}} within the abab plane with 𝒉st{\bm{h}}_{\rm st} along the cc direction (θ=0)(\theta=0), which demonstrates that the asymmetry is changed by the rotation of 𝒉st{\bm{h}}_{\rm st}.

IV Thermal transport

In this section, we calculate the longitudinal thermal conductivity in the presence of both uniform and staggered magnetic fields, by using the Boltzmann transport theory. We apply a weak thermal gradient to the aa or bb direction. Then, the thermal current can be expanded in terms of the thermal gradient as

jQ,ξ=κξ(1)(Tξ)+κξ(2)(Tξ)2+\displaystyle j_{Q,\xi}=\kappa_{\xi}^{(1)}\left(-\frac{\partial T}{\partial\xi}\right)+\kappa_{\xi}^{(2)}\left(\frac{\partial T}{\partial\xi}\right)^{2}+\cdots (17)

where ξ=a\xi=a or bb; κξ(1)\kappa_{\xi}^{(1)} and κξ(2)\kappa_{\xi}^{(2)} are the linear and nonlinear thermal conductivities given by

κξ(1)\displaystyle\kappa_{\xi}^{(1)} =2τΩn𝒌BZ(E𝒌n>0)f(E𝒌n)TE𝒌nv𝒌nξ2,\displaystyle=\frac{2\tau}{\Omega}\sum_{n}\sum_{{{\bm{k}}}\in{\rm BZ}(E_{{{\bm{k}}}n}>0)}\frac{\partial f(E_{{{\bm{k}}}n})}{\partial T}E_{{{\bm{k}}}n}v_{{{\bm{k}}}n\xi}^{2}, (18)
κξ(2)\displaystyle\kappa_{\xi}^{(2)} =2τ2Ωn𝒌BZ(E𝒌n>0)2f(E𝒌n)T2E𝒌nv𝒌nξ3,\displaystyle=\frac{2\tau^{2}}{\Omega}\sum_{n}\sum_{{{\bm{k}}}\in{\rm BZ}(E_{{{\bm{k}}}n}>0)}\frac{\partial^{2}f(E_{{{\bm{k}}}n})}{\partial T^{2}}E_{{{\bm{k}}}n}v_{{{\bm{k}}}n\xi}^{3}, (19)

respectively, where the summation is taken over the momentum 𝒌{\bm{k}} in the first Brillouin zone for which E𝒌n>0E_{{\bm{k}}n}>0, Ω\Omega is the system size, n=±n=\pm is the band index, f(E𝒌n)=(eE𝒌n/T+1)1f(E_{{{\bm{k}}}n})=({{\rm e}}^{E_{{{\bm{k}}}n}/T}+1)^{-1} is the Fermi distribution function, v𝒌nξE𝒌n/kξv_{{{\bm{k}}}n\xi}\equiv\partial E_{{{\bm{k}}}n}/\partial k_{\xi} is the group velocity, and τ\tau is the relaxation time of Majorana fermion introduced phenomenologically. We set the Boltzmann constant kBk_{\rm B} unity and take Jτ=1J\tau=1 in this section.

The thermal current is generated by thermally excited Majorana fermions. To discuss the field dependence of the linear response κξ(1)\kappa_{\xi}^{(1)}, we introduce the “effective Majorana density” as

DM=0n𝒌BZ(E𝒌n>0)δ(εE𝒌n)(df(ε)dε)dε,\displaystyle D_{\rm M}=\int_{0}^{\infty}\sum_{n}\sum_{{{\bm{k}}}\in{\rm BZ}(E_{{{\bm{k}}}n}>0)}\delta(\varepsilon-E_{{{\bm{k}}}n})\left(-\frac{df(\varepsilon)}{d\varepsilon}\right)d\varepsilon, (20)

in addition to the Majorana gap ΔM\Delta_{\rm M}; see Sec. IV.1. Meanwhile, κξ(2)\kappa_{\xi}^{(2)} describes the second-order nonlinear response arising from the asymmetry of the Majorana band dispersions. To measure the asymmetry, we define the “band asymmetry parameter” as

Λξ=1Ω𝒌BZ(kξ>0)nf𝒌nE𝒌n(v𝒌nξ+v𝒌nξ),\displaystyle{\Lambda}_{\xi}=-\frac{1}{\Omega}\sum_{{{\bm{k}}}\in{\rm BZ}(k_{\xi}>0)}\sum_{n}\frac{\partial f_{{{\bm{k}}}n}}{\partial E_{{{\bm{k}}}n}}(v_{{{\bm{k}}}n\xi}+v_{-{{\bm{k}}}n\xi}), (21)

where the sum of momentum is taken in half of the first Brillouin zone with kξ>0k_{\xi}>0 to extract the asymmetry. We use this quantity to discuss the field dependences of κξ(2)\kappa_{\xi}^{(2)} in Sec. IV.2. We perform the following calculations for the system with periodic boundary conditions by taking 864×864864\times 864 kk points in the first Brillouin zone.

IV.1 Linear thermal conductivity

Refer to caption
Figure 5: The linear thermal conductivities (a) κa(1)\kappa_{a}^{(1)} and (b) κb(1)\kappa_{b}^{(1)}, (c) the Majorana gap ΔM\Delta_{\rm M}, and (d) the integrated density of states DMD_{\rm M} while varying the direction of the uniform magnetic field in the acac plane (left panels) and the abab plane (right panels) under the staggered magnetic field along the aa direction (ϕst=0)(\phi_{\rm st}=0). The dot-dashed black lines represent the data for the uniform magnetic field only (rst=0r_{\rm st}=0), while the green, blue, and magenta lines are for the cases with the staggered magnetic field at rst=0.2r_{\rm st}=0.2, 0.60.6, and 1.01.0, respectively. We set δ=0.096J\delta=0.096J and temperature T=0.01JT=0.01J.
Refer to caption
Figure 6: Contour plots of the linear thermal conductivities (a) κa(1)\kappa_{a}^{(1)} and (b) κb(1)\kappa_{b}^{(1)}, (c) the Majorana gap ΔM\Delta_{\rm M}, and (d) the effective Majorana density DMD_{\rm M} while varying the direction of the uniform magnetic field under the staggered magnetic field with ϕst=0\phi_{\rm st}=0, π/6\pi/6, π/3\pi/3, π/2\pi/2, and 𝒉stc{\bm{h}}_{\rm st}\parallel c from left to right. In (c) and (d), the cyan lines and the magenta lines and points are drawn with the same notations as Fig. 3. We set δ=0.096J\delta=0.096J, rst=1r_{\rm st}=1, and T=0.01JT=0.01J.

First, we show the results for the linear thermal conductivities κa(1)\kappa_{a}^{(1)} and κb(1)\kappa_{b}^{(1)} in Figs. 5(a) and 5(b), respectively, while rotating the uniform magnetic field in the acac and abab planes. We find that they change drastically with the field angle and the angle dependence becomes more conspicuous for a larger staggered magnetic field (namely, larger rstr_{\rm st}), while κa(1)\kappa_{a}^{(1)} and κb(1)\kappa_{b}^{(1)} are almost identical. To understand these behaviors, we plot the Majorana gap ΔM\Delta_{\rm M} in Fig. 5(c). When the staggered magnetic field is absent (rst=0.0r_{\rm st}=0.0) or relatively weak (rst=0.2r_{\rm st}=0.2), the Majorana spectrum changes with the field angle, while it is gapped in most regions of θ\theta and φ\varphi. Meanwhile, when the staggered magnetic field becomes larger and comparable to the uniform one (rst=0.6r_{\rm st}=0.6 and 1.01.0), the spectrum becomes gapless and the Majorana Fermi surfaces appear in wider regions, as shown in Fig. 5(c). In the gapped region, the angle dependence of κξ(1)\kappa_{\xi}^{(1)} appears to inversely correlate with ΔM\Delta_{\rm M}, as expected for the thermal transport in the gapped system. On the other hand, as shown in Fig. 5(d), in the gapless region, we find that κξ(1)\kappa_{\xi}^{(1)} roughly correlates with the effective Majorana density DMD_{\rm M}. For example, for rst=1r_{\rm st}=1, we observe that DMD_{\rm M} is almost angle independent in the region where κξ(1)\kappa_{\xi}^{(1)} does not change so much in the case of 𝒉ac{\bm{h}}\parallel ac plane, while DMD_{\rm M} increases around φ/π=0.5\varphi/\pi=0.5 and 1.51.5 where κξ(1)\kappa_{\xi}^{(1)} is enhanced in the case of 𝒉ab{\bm{h}}\parallel ab plane.

In Fig. 6, we display κξ(1)\kappa_{\xi}^{(1)} in comparison with ΔM\Delta_{\rm M} in a more comprehensive manner while changing the direction of the staggered magnetic field ϕst\phi_{\rm st}. We take the amplitudes of the uniform and staggered magnetic fields to be the same, that is, rst=1r_{\rm st}=1. Again, we observe inverse correlation between κξ(1)\kappa_{\xi}^{(1)} and ΔM\Delta_{\rm M}. In addition, we find that κξ(1)\kappa_{\xi}^{(1)} at ϕst=0\phi_{\rm st}=0 and ϕst=π/6\phi_{\rm st}=\pi/6 behave similarly to those at ϕst=π/3\phi_{\rm st}=\pi/3 and ϕst=π/2\phi_{\rm st}=\pi/2, respectively, with a shift of φ\varphi. This is because the Majorana gap satisfies the relation

ΔM(ϕst,φ)=ΔM(ϕst+mπ3,φ2mπ3),\displaystyle\Delta_{\rm M}(\phi_{\rm st},\varphi)=\Delta_{\rm M}\left(\phi_{\rm st}+\frac{m\pi}{3},\varphi-\frac{2m\pi}{3}\right), (22)

where mm is an integer. Note, however, that the values of κξ(1)(ϕst,φ)\kappa_{\xi}^{(1)}(\phi_{\rm st},\varphi) and κξ(1)(ϕst+mπ/3,φ2mπ/3)\kappa_{\xi}^{(1)}(\phi_{\rm st}+m\pi/3,\varphi-2m\pi/3) are slightly different because the Majorana spectra are different. As plotted in the rightmost panels in Fig. 6, the results for the out-of-plane staggered magnetic field show threefold rotational symmetry for φ\varphi, in contrast to those for the in-plane ones.

IV.2 Nonlinear thermal conductivity

Refer to caption
Figure 7: The nonlinear thermal conductivities (a) κa(2)\kappa_{a}^{(2)} and (c) κb(2)\kappa_{b}^{(2)}, and the band asymmetry parameters (b) Λa{\Lambda}_{a} and (d) Λb{\Lambda}_{b} for the same parameter settings as those in Fig. 5.
Refer to caption
Figure 8: Contour plots of the nonlinear thermal conductivities (a) κa(2)\kappa_{a}^{(2)} and (c) κb(2)\kappa_{b}^{(2)}, and the band asymmetry parameters (b) Λa{\Lambda}_{a} and (d) Λb{\Lambda}_{b} for the same parameter settings as those in Fig. 6. The green lines and points represent the parameters where the Majorana excitation spectrum becomes symmetric for the given magnetic fields; see the text for details.

Next, we show the results for the nonlinear thermal conductivities κa(2)\kappa_{a}^{(2)} and κb(2)\kappa_{b}^{(2)} in Figs. 7(a) and 7(c), respectively, while rotating the uniform magnetic field in the acac and abab planes. In contrast to κξ(1)\kappa_{\xi}^{(1)}, κξ(2)\kappa_{\xi}^{(2)} vanishes when the staggered magnetic field is absent (rst=0.0r_{\rm st}=0.0), as the Majorana band dispersions are symmetric in momentum space. Upon introducing the staggered magnetic field, the Majorana dispersions become asymmetric, and κξ(2)\kappa_{\xi}^{(2)} becomes nonzero and show the field angle dependence. Interestingly, we find that κξ(2)\kappa_{\xi}^{(2)} exhibits distinct angle dependence between κa(2)\kappa_{a}^{(2)} and κb(2)\kappa_{b}^{(2)}; in particular, κb(2)\kappa_{b}^{(2)} is always zero when the uniform field is rotated in the acac plane, as plotted in the left panel of Fig. 7(c). This is in sharp contrast to the behaviors of κa(1)\kappa_{a}^{(1)} in Fig. 5. We compare these behaviors with the band asymmetric parameter Λξ{\Lambda}_{\xi} in Eq. (21) plotted in Figs. 7(b) and 7(d). As expected, we find clear correlation between κξ(2)\kappa_{\xi}^{(2)} and Λξ{\Lambda}_{\xi} in their dependences on the field angle and rstr_{\rm st}. In particular, Λb{\Lambda}_{b} vanishes for the uniform field in the acac plane because of the mirror symmetry, which results in zero κb(2)\kappa_{b}^{(2)}. Our result suggests that the Majorana band asymmetry can be detected by the measurement of nonreciprocal thermal transport, while higher experimental accuracy is needed as discussed later.

Figure 8 displays κξ(2)\kappa_{\xi}^{(2)} and Λξ{\Lambda}_{\xi} in a similar manner to Fig. 6. We again observe the overall correlation between κξ(2)\kappa_{\xi}^{(2)} and Λξ{\Lambda}_{\xi}. Moreover, from the comparison with Fig. 6(c), we find that κξ(2)\kappa_{\xi}^{(2)} is further enhanced in the gapless states with the Majorana Fermi surfaces. We note that κa(2)\kappa_{a}^{(2)} and Λa{\Lambda}_{a} look differently at ϕst=π/6\phi_{\rm st}=\pi/6: Λa{\Lambda}_{a} has the dips around ϕ=0.7π\phi=0.7\pi and 1.7π1.7\pi, but κa(2)\kappa_{a}^{(2)} does not. Similar discrepancies can be seen also for κb(2)\kappa_{b}^{(2)} and Λb\Lambda_{b} at ϕst=π/2\phi_{\rm st}=\pi/2. We speculate that these behaviors reflect different temperature effects on Λa{\Lambda}_{a} and κa(2)\kappa_{a}^{(2)}.

We also note that κξ(2)\kappa_{\xi}^{(2)} as well as Λξ{\Lambda}_{\xi} vanishes on several lines and points denoted by green in Fig. 8. On these lines and points, the Majorana excitation spectrum in Eq. (16) becomes symmetric in momentum space, and hence, the band asymmetry vanishes. Specifically, Eq. (16) becomes symmetric with respect to the aa direction when h~Axyzh~Bxyz=(h~Azxyh~Bzxy)\tilde{h}_{\rm A}^{xyz}-\tilde{h}_{\rm B}^{xyz}=-(\tilde{h}_{\rm A}^{zxy}-\tilde{h}_{\rm B}^{zxy}) and h~Ayzxh~Byzx=0\tilde{h}_{\rm A}^{yzx}-\tilde{h}_{\rm B}^{yzx}=0, while it is symmetric with respect to the bb direction when h~Axyzh~Bxyz=h~Azxyh~Bzxy\tilde{h}_{\rm A}^{xyz}-\tilde{h}_{\rm B}^{xyz}=\tilde{h}_{\rm A}^{zxy}-\tilde{h}_{\rm B}^{zxy}.

V Discussion

V.1 Quantitative estimate

Let us estimate quantitatively the linear and nonlinear thermal currents, jQ,ξ(1)j_{Q,\xi}^{(1)} and jQ,ξ(2)j_{Q,\xi}^{(2)}, which are given by the first and second terms in Eq. (17), respectively. Bearing the candidate material α{\alpha}-RuCl3\rm RuCl_{3} in mind, we set the in-plane and out-of-plane lattice constants as a0=5.98a_{0}=5.98 Å{\mathrm{\mathring{A}}} and c0=5.72c_{0}=5.72 Å{\mathrm{\mathring{A}}}, respectively Johnson2015 . In the following calculations, we use c0c_{0} to take the kck_{c} integral from 0 to 2π/c02\pi/c_{0} assuming no dispersion in the kck_{c} direction. We also assume the Kitaev interaction J/kB=80J/k_{\rm B}=80 K Sandilands2015 ; NKKMM . We here set the temperature gradient T/a=160K/m\partial T/\partial a=160~{}{\rm K/m}, referring the experimental setup Kasahara ; Hirobe . We also take the amplitude of the uniform magnetic field at H/gμB7TH/g\mu_{\rm B}\sim 7~{}{\rm T}, where gg is the electron gg-factor and μB\mu_{\rm B} is the Bohr magneton, which roughly corresponds to δ0.1J\delta\sim 0.1J when assuming Δf/kB10K\Delta_{\rm f}/k_{\rm B}\sim 10~{}{\rm K} Kasahara .

Refer to caption
Figure 9: Quantitative estimates of the (a) linear and (b) nonlinear components of thermal currents, under the magnetic field configuration (θ,φ,ϕst)=(π/2,π/2,0)(\theta,\varphi,{\phi}_{\rm st})=(\pi/2,\pi/2,0) which means 𝒉b{\bm{h}}\parallel b and 𝒉sta{\bm{h}}_{\rm st}\parallel a, with the uniform magnetic field H/gμB=6.8H/g\mu_{\rm B}=6.8 T. For comparison, the effective Majorana density and the band asymmetry parameter are shown in (c) and (d), respectively. We assume the parameters in α{\alpha}-RuCl3\rm RuCl_{3}; see the text for details. The lifetime is assumed as Jτ/=105J\tau/\hbar=10^{5}.

Figure 9(a) shows the rstr_{\rm st} dependence of the magnitude of the linear component of the thermal current, |jQ,a(1)||j_{Q,a}^{(1)}|, at three different temperatures T=0.8T=0.8, 1.61.6, and 3.23.2 K in the uniform and staggered magnetic fields applied along the bb and aa directions, respectively, which leads to the gapless spectrum with the Fermi surfaces around the K and K’ points and the nodal points at nonzero energy at the K points, similar to Fig. 2(c). Note that the sign of jQ,a(1)j_{Q,a}^{(1)} depends on the direction of the temperature gradient; see Eq. (17). We here assume an extremely clean case by taking Jτ/=105J\tau/\hbar=10^{5}. We find that |jQ,a(1)||j_{Q,a}^{(1)}| does not depend strongly on rstr_{\rm st} for 0<rst10<r_{\rm st}\lesssim 1, whereas rapidly increases for rst>1r_{\rm st}>1. This behavior is well correlated with the change of the effective Majorana density DMD_{\rm M} shown in Fig. 9(c). We also find that |jQ,a(1)||j_{Q,a}^{(1)}| at rst2r_{\rm st}\sim 2 increases almost linearly with TT, while it increases faster at lower rstr_{\rm st}. In contrast, DMD_{\rm M} depends on temperature for smaller rstr_{\rm st}, while it is almost temperature independent at rst2r_{\rm st}\sim 2. These behaviors come from the energy dependence of the Majorana density of states (DOS) as follows. The Majorana DOS changes its functional form below and above the energy of the nodal points at the K points; the DOS is almost constant below the nodal points, while it increases almost linearly above the nodal points. When rstr_{\rm st} is small, the nodal points locate at a low energy and the linearly increasing DOS above the nodal points dominates the thermal transport; hence, |jQ,a(1)||j_{Q,a}^{(1)}| increases faster than TT-linear. On the other hand, when rstr_{\rm st} becomes large, the nodal points shift to a higher energy and the constant DOS dominates the thermal transport, leading to the TT-linear dependence of |jQ,a(1)||j_{Q,a}^{(1)}|. When we take rst1.5r_{\rm st}\simeq 1.5, the estimate of |jQ,i(1)||j_{Q,i}^{(1)}| is in the order of 10410^{4} W/m2, namely, κa(1)102\kappa_{a}^{(1)}\sim 10^{2} W/(K\cdotm). Even if we compromise the relaxation time down to Jτ/=103J\tau/\hbar=10^{3}, κa(1)\kappa_{a}^{(1)} is estimated as 1\sim 1 W/m2, which is comparable to the values in the previous experiments Kasahara ; Hirobe . Hence, we expect that the characteristic field angle dependence of the linear thermal conductivity in Fig. 6 is observable. In reality, it may compete with the contribution from phonons Hirobe ; Leahy ; Yu ; Hentrich ; Kasahara ; Kasahara2 ; Hentrich2 , but it would be able to identify the Majorana contribution by the field angle dependence.

Next, let us discuss the nonlinear component of the thermal current jQ,a(2)j_{Q,a}^{(2)} for the same field configuration. In this situation, the asymmetry of the Majorana dispersion is present only in the aa direction. As shown in Fig. 9(b), we find that jQ,a(2)j_{Q,a}^{(2)} increases with the increase of the asymmetry parameter Λa\Lambda_{a} shown in Fig. 9(d). We also note that jQ,a(2)j_{Q,a}^{(2)} depends on TT in the small rstr_{\rm st} region, whereas the TT dependence becomes weaker for larger rstr_{\rm st}, because of the energy dependence of the Majorana DOS discussed above. The estimate of jQ,a(2)j_{Q,a}^{(2)} at rst1.5r_{\rm st}\simeq 1.5 and T=3.2KT=3.2~{}{\rm K} is about 22W/m222~{}{\rm W/m^{2}}. This value is about 0.4 % of |jQ,a(1)||j_{Q,a}^{(1)}|, which would be hard to observe in experiments at present, and future improvement is desired.

V.2 Other additional effects from magnetic fields

In the derivation of the effective Hamiltonian in Sec. II, we omit the contribution from the second-order perturbation in terms of the magnetic fields,

2\displaystyle{\cal H}_{2} =δ2α=x,y,zi,jαhiαhjαSiαSjα\displaystyle=-\delta_{2}\sum_{{\alpha}=x,y,z}\sum_{\langle i,j\rangle_{\alpha}}h_{i}^{\alpha}h_{j}^{\alpha}S_{i}^{\alpha}S_{j}^{\alpha}
=iδ24𝒓[h~2xc𝒓,Ac𝒓+𝒂1,B+h~2yc𝒓,Ac𝒓+𝒂2,B+h~2zc𝒓,Ac𝒓,B],\displaystyle=-\frac{i\delta_{2}}{4}\sum_{{{\bm{r}}}}\bigl{[}\tilde{h}_{2}^{x}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}}+{\bm{a}}_{1},{{\rm B}}}+\tilde{h}_{2}^{y}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}}+{\bm{a}}_{2},{{\rm B}}}+\tilde{h}_{2}^{z}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}},{{\rm B}}}\bigr{]}, (23)

where h~2α=hAαhBα\tilde{h}_{2}^{\alpha}=h_{\rm A}^{\alpha}h_{\rm B}^{\alpha} and δ2\delta_{2} is assumed to be a constant. Equation (23) indicates that the second-order contribution modifies the Kitaev couplings JαJ^{\alpha} in an anisotropic way, which would modulate both linear and nonlinear thermal responses as discussed below.

We also omit the other type of three-spin terms appearing in the third-order perturbation, which is given by Kitaev2006

4\displaystyle{\cal H}_{4} =δ4[,m,n]hxhmyhnzSxSmySnz\displaystyle=-\delta_{4}\sum_{[\ell,m,n]}h_{\ell}^{x}h_{m}^{y}h_{n}^{z}S_{\ell}^{x}S_{m}^{y}S_{n}^{z}
=δ48𝒓[h~Ac𝒓,Bc𝒓,Ac𝒓𝒂1,Ac𝒓𝒂2,A+h~Bc𝒓,Ac𝒓,Bc𝒓+𝒂2,Bc𝒓+𝒂1,B],\displaystyle=-\frac{\delta_{4}}{8}\sum_{{\bm{r}}}\bigl{[}\tilde{h}_{\rm A}c_{{{\bm{r}}},{{\rm B}}}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}}-{\bm{a}}_{1},{{\rm A}}}c_{{{\bm{r}}}-{\bm{a}}_{2},{{\rm A}}}+\tilde{h}_{\rm B}c_{{{\bm{r}}},{{\rm A}}}c_{{{\bm{r}}},{{\rm B}}}c_{{{\bm{r}}}+{\bm{a}}_{2},{{\rm B}}}c_{{{\bm{r}}}+{\bm{a}}_{1},{{\rm B}}}\bigr{]}, (24)

where the summation in the first line is taken for three sites [,m,n][\ell,m,n] on the same sublattice as shown in Fig. 1(a), and h~A(B)=hA(B)xhA(B)yhA(B)z\tilde{h}_{\rm A(B)}=h_{\rm A(B)}^{x}h_{\rm A(B)}^{y}h_{\rm A(B)}^{z}; δ4\delta_{4} is assumed to be a constant. Equation (24) describes interactions between the Majorana fermions, which hamper the exact solvability of the problem. At the level of a mean-field approximation, Eq. (24) leads to modifications of not only JαJ^{\alpha} but also h~A(B)αβγ\tilde{h}_{{{\rm A}}({{\rm B}})}^{{\alpha}{\beta}{\gamma}}. Another possible effect of the Majorana interactions is an instability of the Fermi surfaces, but this issue is not yet fully understood CMR .

Summarizing Eq. (23) and the mean-field contributions from Eq. (24), we end up with the one-body Hamiltonian in the form of Eqs. (6) and (7) with normalized JαJ^{\alpha} and h~A(B)αβγ\tilde{h}_{\rm A(B)}^{\alpha\beta\gamma} as

J~α\displaystyle\tilde{J}^{\alpha} =Jα+δ2h~2αδ4(h~AΔAA+h~BΔBB),\displaystyle=J^{\alpha}+\delta_{2}\tilde{h}_{2}^{\alpha}-\delta_{4}\left(\tilde{h}_{\rm A}\Delta_{\rm AA}+\tilde{h}_{\rm B}\Delta_{\rm BB}\right), (25)
h~~A(B)αβγ\displaystyle\tilde{\tilde{h}}_{\rm A(B)}^{{\alpha}{\beta}{\gamma}} =h~A(B)αβγδ4δh~A(B)ΔAB,\displaystyle=\tilde{h}_{\rm A(B)}^{{\alpha}{\beta}{\gamma}}-\frac{\delta_{4}}{\delta}\tilde{h}_{\rm A(B)}\Delta_{\rm AB}, (26)

where α,β,γ=x,y,z{\alpha},{\beta},{\gamma}=x,y,z; here, we assume spatially uniform mean fields ΔAAia𝒓,Aa𝒓,A\Delta_{\rm AA}\equiv i\langle a_{{{\bm{r}}},{{\rm A}}}a_{{{\bm{r}}}^{\prime},{{\rm A}}}\rangle, ΔBBib𝒓,Bb𝒓,B\Delta_{\rm BB}\equiv i\langle b_{{{\bm{r}}},{{\rm B}}}b_{{{\bm{r}}}^{\prime},{{\rm B}}}\rangle, and ΔABia𝒓,Ab𝒓,B\Delta_{\rm AB}\equiv i\langle a_{{{\bm{r}}},{{\rm A}}}b_{{{\bm{r}}}^{\prime},{{\rm B}}}\rangle to decouple the two-body terms in Eq. (24). While JαJ^{\alpha} are perturbed weakly as δ2,δ4Jα\delta_{2},\delta_{4}\ll J^{\alpha}, h~A(B)αβγ\tilde{h}_{\rm A(B)}^{\alpha\beta\gamma} are modified considerably since the first and second terms in Eq. (26) can be the same order. We note that ΔAB\Delta_{\rm AB} is estimated as 0.52\sim-0.52 in the absence of magnetic field BMS ; NUM ; LF . Thus, this argument suggests that we are able to understand the whole magnetic field dependence if we could fit the experimental data by the appropriate tuning of δ\delta, δ2\delta_{2}, and δ4\delta_{4}, though the detailed analysis is left for future study.

V.3 Contributions from magnons

One way to apply a staggered magnetic field to the Kitaev spin liquid is to make a heterostructure between a Kitaev candidate materials and a honeycomb antiferromagnet, where an internal field from the antiferromagnet acts as a staggered field. Such a situation would be realized in a heterostructure composed of α\alpha-RuCl3 and a honeycomb antiferromagnet. Note that van der Waals heterostructures of atomically thin α\alpha-RuCl3{\rm RuCl_{3}} and graphene have been fabricated Mashhadi ; Zhou ; Rizzo ; Leeb , which would be extended to the current situation. Besides, a superstructure of an ilmenite MnTiO3 including IrO6 honeycomb layers MiuraCM ; MiuraAPL could be such a platform, where the antiferromagnetic moment in MnTiO3\rm MnTiO_{3} can be regarded as a staggered internal magnetic field acting on the possible Kitaev spin liquid in the IrO6 honeycomb layers Haraguchi2018 ; Haraguchi2020 ; JM . In such situations, however, the magnon excitations in the antiferromagnet can contribute to the thermal transport. In particular, it was shown that the Dzyaloshinskii-Moriya interaction on the second-neighbor bonds leads to nonlinear thermal transport HKM ; COX ; TSM . However, as the Dzyaloshinskii-Moriya interaction arises from the spin-orbit coupling, one may suppress the magnon contribution by choosing the antiferromagnets with less spin-orbit coupling. Moreover, we expect that the contribution from Majorana fermions can be distinguished from that from magnons by their temperature dependences, since they obey different statistics. Note that this is also the case for phonons, for which we discussed that the field angle dependence is useful in Sec. V.1.

VI Summary

To summarize, we have studied the thermal transport in the Kitaev model under uniform and staggered magnetic fields by using the Boltzmann transport theory. The staggered field breaks the sublattice symmetry and activates the valley degree of freedom. Depending on the amplitudes and directions of the uniform and staggered magnetic fields, the band structure of the Majorana fermions, which are yielded by the fractionalization of spins in the quantum spin liquid state in the Kitaev model, is asymmetrically modulated in momentum space, hosting the gapless or gapped nodal points at finite energy and the Majorana Fermi surfaces. We found that the linear thermal current shows a characteristic field dependence, which correlates with the magnitude of the Majorana gap or the effective Majorana density in the presence of the Majorana Fermi surfaces. On the other hand, we showed that the asymmetric band modulation gives rise to the nonlinear thermal transport, which also depends on the field amplitude and directions. We presented the quantitative estimates by referring to the material parameters of α{\alpha}-RuCl3, and found that the linear thermal response is observable while the improvement of the experimental accuracy is required for the nonlinear component. We also discussed how other additional effects of the magnetic fields arising from the second- and third-order perturbation modify the behavior of the thermal transport. Furthermore, we discussed contributions from magnons in a heterostructure to realize the internal staggered magnetic field, which would be distinguished from the Majorana contributions.

Our findings illuminate the Majorana band modulations by the uniform and staggered magnetic fields and their observation through the thermal transport measurements, but there remain several open questions. The present analysis has been done by the perturbation theory in terms of the magnetic fields, in which the quantum spin liquid state without the localized Z2Z_{2} fluxes (flux-free state) is retained. The flux-free state, however, does not hold beyond the perturbative regime. The Z2Z_{2} fluxes are also excited by raising temperature, which was shown to affect the thermal transport NYM . Another interesting unperturbed effect is a field-induced gapless spin liquid state in the antiferromagnetic Kitaev model ZKSF ; LJCLW ; GMP ; NKKM ; HT ; RVT ; JDJ , which would also affect the thermal transport significantly TZSSS . In addition, in the candidate materials like α\alpha-RuCl3, the effect of non-Kitaev interactions, such as the Heisenberg and off-diagonal interactions, cannot be neglected. Despite some recent attempts YNKM ; MKM , it is still a challenge to access the interesting parameter region at low temperature in applied magnetic fields. Further development is required to clarify these important issues.

acknowledgement

We acknowledge S. Okumura, R. Sano, and A. Tsukazaki for fruitful discussions. This work is supported by JST CREST (JP-MJCR18T2).

References