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

Non-Hermitian non-Abelian topological insulators with PTPT symmetry

Motohiko Ezawa Department of Applied Physics, University of Tokyo, Hongo 7-3-1, 113-8656, Japan
Abstract

We study a non-Hermitian non-Abelian topological insulator preserving PTPT symmetry, where the non-Hermitian term represents nonreciprocal hoppings. As it increases, a spontaneous PTPT symmetry breaking transition occurs in the perfect-flat band model from a real-line-gap topological insulator into an imaginary-line-gap topological insulator. By introducing a band bending term, we realize two phase transitions, where a metallic phase emerges between the above two topological insulator phases. We discuss an electric-circuit realization of non-Hermitian non-Abelian topological insulators. We find that the spontaneous PTPT symmetry breaking as well as the edge states are well observed by the impedance resonance.

I Introduction

Topological insulators are one of the most fascinating ideas in contemporary physicsHasan ; Qi . They are characterized by topological numbers such as the winding number, the Chern number and the 2\mathbb{Z}_{2} index. However, all of these topological numbers are Abelian.

Non-Abelian topological charges are discussed in three-band models protected by PTPT symmetryWuScience ; Tiwari ; Guo ; YangPRL ; Leng or C2TC_{2}T symmetryBouhon ; DWang . They are realized in nodal line semimetalsWuScience ; Tiwari ; YangPRL ; Leng ; MWang ; BJiang ; HPark in three dimensions and Weyl pointsBouhon in two dimensions. Non-Abelian topological insulators in one dimension are studied for three-band modelsGuo and four-band modelsJiang4 . They are experimentally observed in photonic systemsYangPRL ; DWang , phononic systemsBJiang ; BPeng and transmission linesGuo ; Jiang4 . In addition, a generalization to multi-band theories is proposed in nodal line semimetalsWuScience . As far as we aware of, there is no study on non-Hermitian non-Abelian topological phases so far.

Non-Hermitian topological physics have attracted much attentionBender ; Bender2 ; Malzard ; Konotop ; Rako ; Gana ; Zhu ; Yao ; Jin ; Liang ; Nori ; Lieu ; UedaPRX ; Coba ; Jiang ; JPhys ; Ashida . In non-Hermitian systems eigenvalues and eigenfunctions are complex in general. However, they are restricted to be real if PTPT symmetry is imposedBender ; Mosta ; Ruter ; Yuce ; LFeng ; Gana ; Weimann . There is a PTPT symmetry breaking transition, where the eigenvalues and eigenfunctions become complex. Nonreciprocal hopping is such a hopping that the right-going and left-going hopping amplitude are differentHatano . It makes a system non-Hermitian.

In this paper, we study a non-Hermitian non-Abelian topological insulator in an NN band model with PTPT symmetry. We show that a spontaneous PTPT symmetry breaking is induced by increasing the nonreciprocal hoppings from a phase transition from a real-line-gap topological insulator to an imaginary-line-gap topological insulator in the case of a perfect-flat band model. Furthermore, by introducing a band bending term, we may generalize the model to have a metal with two critical points, where a metallic phase emerges between the above two topological insulator phases. Finally, we show how to implement the present model in electric circuits. The edge states and the spontaneous PTPT symmetry breaking are well signaled by the impedance resonance.

Refer to caption

Figure 1: Illustration of the tight-binding Hamiltonian. (a) Hermitian and (b) non-Hermitian models. Interactions between the α\alpha and β\beta chains yield a non-Abelian topological number. All other chains shown in green act as spectators. Red arrows represent nonreciprocal hoppings.

II Non-Hermitian Non-Abelian topological insulators

II.1 Hermitian Hamiltonian

Refer to caption

Figure 2: Energy spectrum of the non-Hermitian Hamiltonian in nanoribbon geometry. Eigenvalues of (a1) the perfectly flat α\alpha and β\beta bands with ξ=0\xi=0 and (a2) the bended bands with ξ=0.5\xi=0.5 shown in blue. The red dots represent the topological edge states. The band structure as a function of ξ\xi with (b1) γ=0\gamma=0 and (b2) γ=0.25\gamma=0.25. The red lines represent the topological edge states. (c1) and (c2) Real part of the energy. (d1) and (d2) Imaginary part of the energy. We have set ξ=0\xi=0 for (c1) and (d1), while we have set ξ=0.5\xi=0.5 for (c2) and (d2). The bulk bands are colored in blue, while the edge states are colored in red. The spectator band is colored in green. When ξ=0\xi=0, there are two phases, a real-line-gap topological insulator (real-TI) phase and an imaginary-line-gap topological insulator (im-TI) phase. When ξ0\xi\not=0, a metallic phase emerges between these two topological insulator phases. We have set εα=1\varepsilon_{\alpha}=1 and εβ=2\varepsilon_{\beta}=2.

We start with a Hermitian system capable to describe a non-Abelian topological insulator based of the one-dimensional lattice in Fig.1(a). We consider generators of 𝔰𝔬(N)\mathfrak{so}\left(N\right) rotation LαβL_{\alpha\beta} indexed by α\alpha and β\beta, whose abab components are defined by

(Lαβ)ab=δαbδβaδαaδβb.\left(L_{\alpha\beta}\right)_{ab}=\delta_{\alpha b}\delta_{\beta a}-\delta_{\alpha a}\delta_{\beta b}. (1)

We consider a PTPT-invariant Hamiltonian in the momentum space given byWuScience

Hαβ(k)=Rαβ(k2)diag.(ε1,ε2,,εN)Rαβ(k2)t,H_{\alpha\beta}\left(k\right)=R_{\alpha\beta}\left(\frac{k}{2}\right)\text{diag.}\left(\varepsilon_{1},\varepsilon_{2},\cdots,\varepsilon_{N}\right)R_{\alpha\beta}\left(\frac{k}{2}\right)^{\text{t}}, (2)

where 0k<2π0\leq k<2\pi, 1α,βN1\leq\alpha,\beta\leq N, ε1,ε2,,εN\varepsilon_{1},\varepsilon_{2},\cdots,\varepsilon_{N} are real, and

Rαβ(k2)=ek2LαβR_{\alpha\beta}\left(\frac{k}{2}\right)=e^{\frac{k}{2}L_{\alpha\beta}} (3)

is a rotation matrix given by

(Rαβ(k2))ab\displaystyle\left(R_{\alpha\beta}\left(\frac{k}{2}\right)\right)_{ab} =δab+(δaαδbα+δaβδbβ)cosk2\displaystyle=\delta_{ab}+\left(\delta_{a\alpha}\delta_{b\alpha}+\delta_{a\beta}\delta_{b\beta}\right)\cos\frac{k}{2}
+(δaβδbαδaαδbβ)sink2.\displaystyle+\left(\delta_{a\beta}\delta_{b\alpha}-\delta_{a\alpha}\delta_{b\beta}\right)\sin\frac{k}{2}. (4)

The Hamiltonian (2) is explicitly written as

Hαβ(k)\displaystyle H_{\alpha\beta}\left(k\right) =εα+εβ2\displaystyle=\frac{\varepsilon_{\alpha}+\varepsilon_{\beta}}{2}
+εαεβ2(δaαδbαδaβδbβ)cosk\displaystyle+\frac{\varepsilon_{\alpha}-\varepsilon_{\beta}}{2}\left(\delta_{a\alpha}\delta_{b\alpha}-\delta_{a\beta}\delta_{b\beta}\right)\cos k
+εαεβ2(δaβδbα+δaαδbβ)sink.\displaystyle+\frac{\varepsilon_{\alpha}-\varepsilon_{\beta}}{2}\left(\delta_{a\beta}\delta_{b\alpha}+\delta_{a\alpha}\delta_{b\beta}\right)\sin k. (5)

It is decomposed into two parts,

Hαβ(k)=jα,βHjHαβ(k),H_{\alpha\beta}\left(k\right)=\bigoplus_{j\neq\alpha,\beta}H_{j}\oplus H_{\alpha\beta}^{\prime}\left(k\right), (6)

where

Hj\displaystyle H_{j} =εj𝕀1,\displaystyle=\varepsilon_{j}\mathbb{I}_{1}, (7)
Hαβ(θ)\displaystyle H_{\alpha\beta}^{\prime}\left(\theta\right) =[εαεβ2(cosksinksinkcosk)+εα+εβ2𝕀2].\displaystyle=\left[\frac{\varepsilon_{\alpha}-\varepsilon_{\beta}}{2}\left(\begin{array}[]{cc}\cos k&\sin k\\ \sin k&-\cos k\end{array}\right)+\frac{\varepsilon_{\alpha}+\varepsilon_{\beta}}{2}\mathbb{I}_{2}\right]. (10)

The Hamiltonian is nontrivial only for the α\alpha and β\beta bands, with eigenvalues εα\varepsilon_{\alpha} and εβ\varepsilon_{\beta}. All other bands are spectators with respect to the α\alpha and β\beta bands. See Fig.1.

The energy spectrum of the bulk Hamiltonian does not change by the rotation (3) and is given by

E(k)=ε1,ε2,,εN.E\left(k\right)=\varepsilon_{1},\quad\varepsilon_{2},\quad\cdots,\quad\varepsilon_{N}. (11)

The eigenfunctions for the 2×22\times 2 matrix Hαβ(k)H_{\alpha\beta}^{\prime}\left(k\right) are

ψa+\displaystyle\psi_{a}^{+} =δaαsink2+δaβcosk2,\displaystyle=\delta_{a\alpha}\sin\frac{k}{2}+\delta_{a\beta}\cos\frac{k}{2}, (12)
ψa\displaystyle\psi_{a}^{-} =δaαcosk2+δaβsink2,\displaystyle=-\delta_{a\alpha}\cos\frac{k}{2}+\delta_{a\beta}\sin\frac{k}{2}, (13)

while those for HjH_{j} are ψa=δaj\psi_{a}=\delta_{aj}.

The α\alpha and β\beta bands are perfectly flat. They are (\ell11) fold degenerate in a finite chain, where \ell is the number of sites in the chain. See Fig.2(a1).

Refer to caption

Figure 3: (a1)\sim(e1) Real part of the energy and (a2)\sim(e2) imaginary part of the energy. (a1) and (a2) for a real-line-gap topological insulator (real-TI) with γ=0\gamma=0; (b1) and (b2) for a phase transition point with γ=γ1=0.3\gamma=\gamma_{1}=0.3; (c1) and (c2) for a metal with γ=0.4\gamma=0.4; (d1) and (d2) for a phase transition point with γ=γ0=0.5\gamma=\gamma_{0}=0.5; (e1) and (e2) for an imaginary-line-gap topological insulator (im-TI) with γ=0.6\gamma=0.6. We have set ξ=0.2\xi=0.2 for all figures. See also the caption of Fig.2.

II.2 Non-Hermitian Hamiltonian

We generalize the Hermitian non-Abelian system (2) to a non-Hermitian non-Abelian system, keeping PTPT symmetry. We consider the Hamiltonian

Hαβ(k;γ,ξ)=Hαβ(k)+iγσy+ξσxsink,H_{\alpha\beta}^{\prime}\left(k;\gamma,\xi\right)=H_{\alpha\beta}^{\prime}\left(k\right)+i\gamma\sigma_{y}+\xi\sigma_{x}\sin k, (14)

whose eigenenergies are

Eαβ(k;γ,ξ)=εα+εβ±g(k;γ,ξ)2,E_{\alpha\beta}^{\prime}\left(k;\gamma,\xi\right)=\frac{\varepsilon_{\alpha}+\varepsilon_{\beta}\pm\sqrt{g(k;\gamma,\xi)}}{2}, (15)

with

g(k;γ,ξ)=(εαεβ)2γ2+4ξ(εαεβ+ξ)sin2k.g(k;\gamma,\xi)=\left(\varepsilon_{\alpha}-\varepsilon_{\beta}\right)^{2}-\gamma^{2}+4\xi\left(\varepsilon_{\alpha}-\varepsilon_{\beta}+\xi\right)\sin^{2}k. (16)

We explain the meanings of the γ\gamma term and the ξ\xi term. The Hamiltonian (14) is Hermitian when γ=0\gamma=0. When ξ=0\xi=0 in addition, the band structure is highly degenerate as in Fig.2(a1). This degeneracy is resolved by introducing the ξ\xi term as shown in Fig.2(a2). We show the band structure with γ=0\gamma=0 as a function of ξ\xi in Fig.2(b1). The perfect flat bands at εα\varepsilon_{\alpha} and εβ\varepsilon_{\beta} become bended and have dispersions. We also show the band structure with γ=0.25\gamma=0.25 as a function of ξ\xi in Fig.2(b2).

We show Re[Eαβ(k;γ,ξ=0)E_{\alpha\beta}^{\prime}\left(k;\gamma,\xi=0\right)] in Fig.2(c1) and Im[Eαβ(k;γ,ξ=0)E_{\alpha\beta}^{\prime}\left(k;\gamma,\xi=0\right)] in Fig.2(d1) as a function of γ\gamma. They are real for |γ|γ0|\gamma|\leq\gamma_{0} with

γ0=εαεβ2,\gamma_{0}=\frac{\varepsilon_{\alpha}-\varepsilon_{\beta}}{2}, (17)

where PTPT symmetry is preserved. On the other hand, they are complex for |γ|>γ0|\gamma|>\gamma_{0}, and hence PTPT symmetry is spontaneously broken there. Namely, although the Hamiltonian is PTPT-symmetric, eigenvalues and eigenfunctions are no longer real in the spontaneous symmetry broken phase. We show the real and imaginary parts of the energy as a function of γ\gamma in Fig.2(c2) and (c2), where the bulk band has a finite width. We also show the real and imaginary parts of the energy as a function of the momentum kk in Fig.3, where the bands have dispersions.

Especially, we have

Eαβ(π2;γ,ξ)=εα+εβ±h(γ,ξ)2,E_{\alpha\beta}^{\prime}\left(\frac{\pi}{2};\gamma,\xi\right)=\frac{\varepsilon_{\alpha}+\varepsilon_{\beta}\pm\sqrt{h(\gamma,\xi)}}{2}, (18)

with

h(γ,ξ)=(εαεβ+2ξ2γ)(εαεβ+2ξ+2γ).h(\gamma,\xi)=\left(\varepsilon_{\alpha}-\varepsilon_{\beta}+2\xi-2\gamma\right)\left(\varepsilon_{\alpha}-\varepsilon_{\beta}+2\xi+2\gamma\right). (19)

By solving the condition that Eαβ(π2;γ,ξ)E_{\alpha\beta}^{\prime}\left(\frac{\pi}{2};\gamma,\xi\right) is complex, or h(γ,ξ)<0h(\gamma,\xi)<0, we find a phase transition point γ1\gamma_{1} in addition to the phase transition point γ0\gamma_{0} as

γ1=εαεβ2ξ=γ0ξ.\gamma_{1}=\frac{\varepsilon_{\alpha}-\varepsilon_{\beta}}{2}-\xi=\gamma_{0}-\xi. (20)

When ξ>0\xi>0, the bulk energy is real for |γ|γ1\left|\gamma\right|\leq\gamma_{1}, complex for |γ|>γ1\left|\gamma\right|>\gamma_{1}. On the other hand, when ξ<0\xi<0, the bulk energy is real for |γ|γ0\left|\gamma\right|\leq\gamma_{0}, complex for |γ|>γ0\left|\gamma\right|>\gamma_{0}.

II.3 Tight-binding Hamiltonian

The tight-binding Hamiltonian (10) is written in the coordinate space as

Hαβ=H0+Hγ+Hξ,H_{\alpha\beta}^{\prime}=H_{0}+H_{\gamma}+H_{\xi}, (21)

with

H0\displaystyle H_{0} =\displaystyle= εαεβ2j=11(|αjαj+1|+|βjβj+1|,\displaystyle\frac{\varepsilon_{\alpha}-\varepsilon_{\beta}}{2}\sum_{j=1}^{\ell-1}(\left|\alpha_{j}\right\rangle\left\langle\alpha_{j+1}\right|+\left|\beta_{j}\right\rangle\left\langle\beta_{j+1}\right|, (22)
+i|αjβj+1|i|βjαj+1|)+h.c.,\displaystyle+i\left|\alpha_{j}\right\rangle\left\langle\beta_{j+1}\right|-i\left|\beta_{j}\right\rangle\left\langle\alpha_{j+1}\right|)+\text{h.c.},
Hγ\displaystyle H_{\gamma} =\displaystyle= γj=1(|αjβj||βjαj|),\displaystyle\gamma\sum_{j=1}^{\ell}(\left|\alpha_{j}\right\rangle\left\langle\beta_{j}\right|-\left|\beta_{j}\right\rangle\left\langle\alpha_{j}\right|), (23)
Hξ\displaystyle H_{\xi} =\displaystyle= iξj=11(|αjβj+1||βjαj+1|)+h.c.,\displaystyle i\xi\sum_{j=1}^{\ell-1}(\left|\alpha_{j}\right\rangle\left\langle\beta_{j+1}\right|-\left|\beta_{j}\right\rangle\left\langle\alpha_{j+1}\right|)+\text{h.c.}, (24)

where the first two terms in H0H_{0} represent normal hoppings, while the last two terms represent spin-orbit-like imaginary hoppings. The ξ\xi term modifies the spin-orbit-like imaginary hoppings. The γ\gamma term represents nonreciprocal hoppings, which make the system non-Hermitian.

The tight-binding Hamiltonians for the spectator bands are simply given by

Hjα,β=j=1εj|jj|+j=11tj|jj+1|+h.c.,H_{j\neq\alpha,\beta}=\sum_{j=1}^{\ell}\varepsilon_{j}\left|j\right\rangle\left\langle j\right|+\sum_{j=1}^{\ell-1}t_{j}\left|j\right\rangle\left\langle j+1\right|+\text{h.c.}, (25)

where εj\varepsilon_{j} is the on-site energy and tjt_{j} is the hopping parameter.

In this sense, it is enough to consider only the α\alpha and β\beta bands for an arbitrary NN band system. We illustrate the tight-binding model in Fig.1.

Refer to caption

Figure 4: Non-Abelian topological charge marked in red as a function of γ\gamma for various ξ\xi. (a) ξ=0\xi=0, (b) ξ=0.2\xi=-0.2, (c) ξ=0.4\xi=-0.4, (d) ξ=0.2\xi=0.2 and (e) ξ=0.4\xi=0.4. It is quantized at 1/21/2 except for the metallic phase. In the figures, real-TI (im-TI) stands for real(imaginary)-line-gap topological insulator phase.

II.4 Edge states for non-Hermitian model

We illustrate the tight-binding model (21) in Fig.1(b). In a finite chain, two localized states emerge at the edges with the energy

E(ξ,γ)=εα+εβ2±iγE(\xi,\gamma)=\frac{\varepsilon_{\alpha}+\varepsilon_{\beta}}{2}\pm i\gamma (26)

in the presence of the γ\gamma term and the ξ\xi term. They are degenerate only in the Hermitian limit (γ=0\gamma=0). In contrasted to the bulk band, the eigenenergy (26) is complex once γ\gamma is introduced even for the PTPT symmetric phase. We show Eq.(26) as a function of γ\gamma in Fig.2. In contrast to the bulk band, the eigenenergy (26) has no ξ\xi dependence: See Figs.2(d1) and (d2).

When ξ=0\xi=0, the eigenfunctions for the edge states ψα(j)\psi_{\alpha}\left(j\right) and ψβ(j)\psi_{\beta}\left(j\right) at the jj site are perfectly localized at the edges and given by

ψα(j)=12δ1,j,ψβ(j)=i2δ1,j\psi_{\alpha}\left(j\right)=\frac{1}{\sqrt{2}}\delta_{1,j},\qquad\psi_{\beta}\left(j\right)=\frac{-i}{\sqrt{2}}\delta_{1,j} (27)

for the left edge, and

ψα(j)=12δ,j,ψβ(j)=i2δ,j\psi_{\alpha}\left(j\right)=\frac{1}{\sqrt{2}}\delta_{\ell,j},\qquad\psi_{\beta}\left(j\right)=\frac{i}{\sqrt{2}}\delta_{\ell,j} (28)

for the right edge. Here, 11 in δ1,j\delta_{1,j} represent the left edge, while \ell in δ,j\delta_{\ell,j} represents the right edge. The perfectly localized edge states for ξ=0\xi=0 are transformed to edge states with finite penetration depth for ξ0\xi\neq 0.

II.5 Non-Hermitian Non-Abelian topological charges

We define a non-Hermitian non-Abelian Berry connection or a non-Hermitian Berry-Wilczek-Zee (BWZ) connection byWZ

AαβRL(θ)=ψαR|θ|ψβL,A_{\alpha\beta}^{\text{RL}}\left(\theta\right)=\left\langle\psi_{\alpha}^{\text{R}}\right|\partial_{\theta}\left|\psi_{\beta}^{\text{L}}\right\rangle, (29)

where

H|ψαL=εα|ψαLH\left|\psi_{\alpha}^{\text{L}}\right\rangle=\varepsilon_{\alpha}\left|\psi_{\alpha}^{\text{L}}\right\rangle (30)

is the left eigenfunction, and

H|ψαR=εα|ψαRH^{\dagger}\left|\psi_{\alpha}^{\text{R}}\right\rangle=\varepsilon_{\alpha}\left|\psi_{\alpha}^{\text{R}}\right\rangle (31)

is the right eigenfunction.

Refer to caption

Figure 5: (a) Real and imaginary parts of the bulk-band energy in the (γ,ξ)(\gamma,\xi) plane. (b) Topological phase diagram in the (γ,ξ)(\gamma,\xi) plane. Metallic phase appears except for ξ=0\xi=0. In the figure, real-TI (im-TI) stands for real(imaginary)-line-gap topological insulator phase.

We define a non-Hermitian BWZ phase by

ΓαβRL=12π02πRe[AαβRL(θ)]𝑑θ,\Gamma_{\alpha\beta}^{\text{RL}}=\frac{1}{2\pi}\int_{0}^{2\pi}\text{Re}\left[A_{\alpha\beta}^{\text{RL}}\left(\theta\right)\right]d\theta, (32)

which we use as a non-Hermitian non-Abelian topological charge. The eigenfunctions are analytically solved as

|ψ±L\displaystyle\left|\psi_{\pm}^{\text{L}}\right\rangle =1(ψ1L)2+(ψ2L)2{ψ1L,ψ2L},\displaystyle=\frac{1}{\sqrt{\left(\psi_{1}^{\text{L}}\right)^{2}+\left(\psi_{2}^{\text{L}}\right)^{2}}}\left\{\psi_{1}^{\text{L}},\psi_{2}^{\text{L}}\right\}, (33)
ψ1L\displaystyle\psi_{1}^{\text{L}} =γ0cosk±γ02γ2+ξ(2γ+ξ)sin2k,\displaystyle=-\gamma_{0}\cos k\pm\sqrt{\gamma_{0}^{2}-\gamma^{2}+\xi\left(2\gamma+\xi\right)\sin^{2}k}, (34)
ψ2L\displaystyle\psi_{2}^{\text{L}} =γ(γ0+ξ)sink,\displaystyle=\gamma-\left(\gamma_{0}+\xi\right)\sin k, (35)

and

|ψ±R\displaystyle\left|\psi_{\pm}^{\text{R}}\right\rangle =1(ψ1R)2+(ψ2R)2{ψ1R,ψ2R},\displaystyle=\frac{1}{\sqrt{\left(\psi_{1}^{\text{R}}\right)^{2}+\left(\psi_{2}^{\text{R}}\right)^{2}}}\left\{\psi_{1}^{\text{R}},\psi_{2}^{\text{R}}\right\}, (36)
ψ1R\displaystyle\psi_{1}^{\text{R}} =γ0cosk±γ02γ2+ξ(2γ+ξ)sin2k,\displaystyle=\gamma_{0}\cos k\pm\sqrt{\gamma_{0}^{2}-\gamma^{2}+\xi\left(2\gamma+\xi\right)\sin^{2}k}, (37)
ψ2R\displaystyle\psi_{2}^{\text{R}} =γ+(γ0+ξ)sink.\displaystyle=\gamma+\left(\gamma_{0}+\xi\right)\sin k. (38)

When γ=0\gamma=0 and ξ=0\xi=0, using (33) and (36), we calculate the non-Hermitian BWZ connection (29) numerically and find that

AαβRL=12(0110),A_{\alpha\beta}^{\text{RL}}=\frac{1}{2}\left(\begin{array}[]{cc}0&-1\\ 1&0\end{array}\right), (39)

which leads to the non-Hermitian BWZ phase (32) as

ΓαβRL=12(0110),\Gamma_{\alpha\beta}^{\text{RL}}=\frac{1}{2}\left(\begin{array}[]{cc}0&-1\\ 1&0\end{array}\right), (40)

where we have explicitly written only the nontrivial 2×22\times 2 submatrix within the N×NN\times N matrix. Hence, the topological charges are given by

ΓαβRL=12Lαβ\Gamma_{\alpha\beta}^{\text{RL}}=-\frac{1}{2}L_{\alpha\beta} (41)

with Eq.(1). The topological charges ΓαβRL\Gamma_{\alpha\beta}^{\text{RL}} obey essentially the same non-Abelian algebra as LαβL_{\alpha\beta}.

When γ0\gamma\neq 0 and ξ=0\xi=0, we calculate the non-Hermitian BWZ connection (29) to find that it is no longer a constant. However, the non-Hermitian BWZ phase (32) is calculated as in Eq.(40), and hence the topological charge is quantized as in Eq.(41) for any γ\gamma. Nevertheless, the eigenfunctions as well as the eigenvalues are real (i.e., real-line-gap topological insulator phase) only for γ2γ02\gamma^{2}\leq\gamma_{0}^{2}, while the eigenvalues and the eigenfunctions are complex (i.e. imaginary-line-gap topological insulator phase) for γ2>γ02\gamma^{2}>\gamma_{0}^{2}. Hence, PT symmetry is preserved only for γ2γ02\gamma^{2}\leq\gamma_{0}^{2}, and it is spontaneously broken for γ2>γ02\gamma^{2}>\gamma_{0}^{2}. The system undergoes a phase transition at γ=±γ0\gamma=\pm\gamma_{0}.

When γ0\gamma\neq 0 and ξ0\xi\neq 0, we have numerically calculated the topological charge (32) with the use of (33) and (36). We have shown the (2,1)\left(2,1\right) component of the 2×22\times 2 matrix ΓαβRL\Gamma_{\alpha\beta}^{\text{RL}} for various values of ξ\xi in Fig.4. It is quantized to be 1/2 for γ2γ12\gamma^{2}\leq\gamma_{1}^{2} and γ2>γ02\gamma^{2}>\gamma_{0}^{2} when ξ>0\xi>0, while γ2γ02\gamma^{2}\leq\gamma_{0}^{2} and γ2>γ12\gamma^{2}>\gamma_{1}^{2} when ξ<0\xi<0, where γ1=γ0ξ\gamma_{1}=\gamma_{0}-\xi as in Eq.(20). On the other hand, it is not quantized for the metallic phase that emerges between γ0\gamma_{0} and γ1\gamma_{1}, as in Fig.4. It is concluded that the topological charges are quantized and given by Eq.(41) in the insulator phases.

II.6 Topological phase diagram

In non-Hermitian systems, there are line-gap insulators and point-gap insulatorsUedaPRX ; Kawabata in general. In the point-gap insulator, there is a gap in |E||E|. On the other hand, there are two types of line-gap insulators. A real-line gap topological insulator has a gap in Re[E]\left[E\right], while an imaginary-line-gap topological insulator has a gap in Im[E]\left[E\right]. We first consider the case ξ>0\xi>0. For |γ|<γ1\left|\gamma\right|<\gamma_{1}, the system is a non-Hermitian line-gap topological insulator along the Re[E]\left[E\right]. The systems is metallic for γ1|γ|γ0\gamma_{1}\leq\left|\gamma\right|\leq\gamma_{0}. For |γ|>γ0\left|\gamma\right|>\gamma_{0}, the system is a non-Hermitian line-gap topological insulator along the Im[E]\left[E\right]. If ξ<0\xi<0, the system is a real-line-gap topological insulator for |γ|<γ0\left|\gamma\right|<\gamma_{0}, it is a metal for γ0|γ|γ1\gamma_{0}\leq\left|\gamma\right|\leq\gamma_{1} and it is an imaginary-line-gap topological insulator for |γ|>γ1\left|\gamma\right|>\gamma_{1}. We show the topological phase diagram in Fig.5(b). It is consistent with the real and imaginary parts of the energy in the γ\gamma-ξ\xi plane as shown in Fig.5(a).

Refer to caption

Figure 6: (a) Illustration of the electric circuit corresponding to the lattice in Fig.1(b). The hopping along the α\alpha-chain (β\beta-chain) is represented by the inductance LL (the capacitance CC). (b) Negative impedance converter RXR_{X} represents an imaginary hoppingHofmann . (c) Operational amplifier circuit CγC_{\gamma} represents a nonreciprocal hoppingNonR .

Refer to caption

Figure 7: Real and imaginary parts of impedance ZaaZ_{aa} at the edge as a function of frequency ω\omega. (a1) \sim(a4) Hermitian model with γ=0\gamma=0. (b1)\sim(b4) Non-Hermitian model with γ=0.025\gamma=0.025, (c1)\sim(c4) with γ=0.04\gamma=0.04 (phase transition point γ1\gamma_{1}), (d1)\sim(d5) with γ=0.045\gamma=0.045, (e1)\sim(e4) with γ=0.05\gamma=0.05 (phase transition point γ0\gamma_{0}), and (f1)\sim(f4) with γ=0.075\gamma=0.075. We have used a finite chain with open boundary condition (red) and periodic boundary condition (cyan). (a1)\sim(f2) ξ=0\xi=0. (a3)\sim(f4) ξ=0.2\xi=0.2. We have set εα=1\varepsilon_{\alpha}=1 and εβ=1.1\varepsilon_{\beta}=1.1. The length of the chain is 20.

III Electric circuit simulation

An electric circuit is described by the Kirchhoff current law. By making the Fourier transformation with respect to time, the Kirchhoff current law is expressed as

Ia(ω)=bJab(ω)Vb(ω),I_{a}\left(\omega\right)=\sum_{b}J_{ab}\left(\omega\right)V_{b}\left(\omega\right), (42)

where IaI_{a} is the current between node aa and the ground, while VbV_{b} is the voltage at node bb. The matrix Jab(ω)J_{ab}\left(\omega\right) is called the circuit Laplacian. Once the circuit Laplacian is given, we can uniquely setup the corresponding electric circuit. By equating it with the Hamiltonian HH asTECNature ; ComPhys

Jab(ω)=iωHab(ω),J_{ab}\left(\omega\right)=i\omega H_{ab}\left(\omega\right), (43)

it is possible to simulate various topological phases of the Hamiltonian by electric circuitsTECNature ; ComPhys ; Hel ; Lu ; EzawaTEC ; EzawaLCR ; Research ; Zhao ; EzawaSkin ; Garcia ; Hofmann ; EzawaMajo ; Tjunc ; NonR . The relations between the parameters in the Hamiltonian and in the electric circuit are determined by this formula.

In the present problem, only the α\alpha-chain and the β\beta-chain are active in the tight-binding Hamiltonian as in Fig.1. Thus, we need only a 2×22\times 2 matrix. The circuit Laplacian follows from the Hamiltonian (14) as

Jαβ(k)=iω[(Lω2coskf+fCcosk)+εα+εβ2𝕀2],J_{\alpha\beta}^{\prime}\left(k\right)=i\omega\left[\left(\begin{array}[]{cc}-\frac{L}{\omega^{2}}\cos k&f_{+}\\ f_{-}&C\cos k\end{array}\right)+\frac{\varepsilon_{\alpha}+\varepsilon_{\beta}}{2}\mathbb{I}_{2}\right], (44)

with

f±=1ωRX(1+ξ)sink±γ.f_{\pm}=\frac{1}{\omega R_{X}}\left(1+\xi\right)\sin k\pm\gamma. (45)

We may design the electric circuit to realize this circuit Laplacian as in Fig.6. The main part consists of the α\alpha-channel and the β\beta-channel corresponding to the α\alpha-chain and the β\beta-chain in the lattice in Fig.1. Additionally, each node in the ii-channel is connected to the ground via a set of inductor LiL_{i} and capacitor CiC_{i}, where i=αi=\alpha or β\beta, in order to realize the diagonal term (εα+εβ)\varpropto(\varepsilon_{\alpha}+\varepsilon_{\beta}) in Eq.(44).

Hopping terms along the α\alpha-chain and the β\beta-chain are described by the diagonal terms in Eq.(44), where ±cosk=±(eik+eik)/2\pm\cos k=\pm(e^{ik}+e^{-ik})/2 represents the plus (minus) hopping in the tight-bind model. To simulate the positive and negative hoppings in the Hamiltonian, we replace them with the capacitance iωCi\omega C and the inductance 1/iωL1/i\omega L, respectively.

Hopping terms across the α\alpha-chain and the β\beta-chain are described by the off-diagonal terms f±f_{\pm} in Eq.(44), which consist of two terms proportional to sink\sin k and γ\gamma.

(i) The term proportional to sink\sin k produces the cross hopping, where sink=(eikeik)/2i\sin k=(e^{ik}-e^{-ik})/2i represents an imaginary hopping in the tight-bind model. The imaginary hopping is implemented by a negative impedance converter RXR_{X} with current inversionHofmann , as is constructed based on an operational amplifier with resistors: See Fig.6(b). The voltage-current relation is given by

(I1I2)=1RX(νν11)(V1V2),\left(\begin{array}[]{c}I_{1}\\ I_{2}\end{array}\right)=\frac{1}{R_{X}}\left(\begin{array}[]{cc}-\nu&\nu\\ -1&1\end{array}\right)\left(\begin{array}[]{c}V_{1}\\ V_{2}\end{array}\right), (46)

with ν=Rb/Ra\nu=R_{b}/R_{a}, where RXR_{X}, RaR_{a} and RbR_{b} are the resistances in an operational amplifier. We note that the resistors in the operational amplifier circuit are tuned to be ν=1\nu=1 in the literatureHofmann so that the system becomes Hermitian, where the corresponding Hamiltonian represents a spin-orbit interaction. It produces the Hamiltonian

H=1ωRX(iiii)H=\frac{1}{\omega R_{X}}\left(\begin{array}[]{cc}i&-i\\ i&-i\end{array}\right) (47)

for the Hermitian limit.

(ii) The term γ\varpropto\gamma produces the nonreciprocal hopping terms, which are vertical hoppings represented by red arrows in Fig.1(b). The nonreciprocal hopping is constructed by a combination of an operational amplifier and capacitorsNonR ,

(IijIji)=iωCγ(1111)(ViVj),\left(\begin{array}[]{c}I_{ij}\\ I_{ji}\end{array}\right)=i\omega C_{\gamma}\left(\begin{array}[]{cc}-1&1\\ -1&1\end{array}\right)\left(\begin{array}[]{c}V_{i}\\ V_{j}\end{array}\right), (48)

as in Fig.6(c). It corresponds to the Hamiltonian

H=Cγ(1111).H=C_{\gamma}\left(\begin{array}[]{cc}-1&1\\ -1&1\end{array}\right). (49)

In this way, the tight-binding Hamiltonian for the present non-Hermitian non-Abelian topological system is implemented in the electric circuit given in Fig.6.

III.1 Impedance resonance

The band structure as well as edge states are well observed by impedance resonance, which is definedTECNature ; ComPhys ; Hel by

Zab=Va/Ib=Gab,Z_{ab}=V_{a}/I_{b}=G_{ab}, (50)

where G=J1G=J^{-1} is the Green function. Taking the nodes a=ba=b at an edge, we show the real and imaginary parts of the impedance for a finite chain as a function of ω\omega in Fig.7, which are marked in red. For comparison, we also show the impedance for a periodic boundary condition in cyan, where the edge states are absent.

We first study the Hermitian case (γ=0\gamma=0) with ξ=0\xi=0, where the impedance is shown in Fig.7(a1) and (a2). The edge impedance resonance is clear by comparing the periodic boundary condition and the open boundary condition. There are only two bulk peaks in cyan at Re[Eαβ(k;γ,ξ)E_{\alpha\beta}^{\prime}\left(k;\gamma,\xi\right)]. On the other hand, there is an additional peak in red due to the edge states between two bulk peaks, as corresponds to Fig.2(a1).

Next, we show the impedance for various nonreciprocity γ\gamma with ξ=0\xi=0 in Fig.7(a1)\sim(f1) and (a2)\sim(f2). The edge impedance resonance rapidly decreases as the increase of γ\gamma, as shown in Fig.7(b1). This is due to the imaginary contribution in Eq.(26). Then, the distance between two bulk peaks becomes narrower, which is consistent with Re[Eαβ(k;γ,ξ=0)E_{\alpha\beta}^{\prime}\left(k;\gamma,\xi=0\right)] as shown in Fig.2(c1). The two bulk peaks merge into one peak at the spontaneous PTPT symmetry breaking point γ0\gamma_{0}, as shown in Fig.7(e1). The bulk impedance resonance is very strong due to the gap closing of the bulk band. We also observe the edge impedance resonance in the imaginary-line gap topological insulating phase, where the impedance resonance is weak comparing to Fig.7(a1) as shown in Fig.7(f1). This is also the imaginary contribution in Eq.(26).

We also show the impedance for finite ξ\xi in Fig.7(a3)\sim(f3) and (a4)\sim(f4), as corresponds to Fig.2(c2). The bulk impedance peaks become broad, which reflects the broadening of the bulk bands. As a result, the edge impedance peak becomes clearer as in Fig.7(a3) in comparison to Fig.7(a1). There are strong cyan resonances at the phase transition point γ1\gamma_{1} point as shown in Fig.7(c3) and (c4). It is due to the gap closing of the bulk band. In Fig.7(d3) and (d4), the impedance structure is complicated, which reflects the metallic band structure. The effect of the ξ\xi term is negligible for the imaginary-line-gap topological phase as shown in Fig.7(f3) and (f4) since the peak of the impedance is broad even for ξ=0\xi=0 in Fig.7(f1) and (f2). Here, note that ξ\xi appears only in the form of (1+ξ)(1+\xi) in Eq.(45).

IV Conclusion

We have proposed a non-Hermitian non-Abelian topological insulator model by imposing PTPT symmetry in one dimension. It describes a real-line-gap topological insulator with real eigenvalues in the Hermitian limit. The system undergoes a spontaneous breakdown of PTPT symmetry as the non-Hermitian term increases, and turns out to describe an imaginary-line-gap topological insulator, when the bulk bands are perfectly flat. When we introduce a bulk bending term, there are two phase transitions with the emergence of a metal with complex eigenvalues between the above two topological insulators. Finally, we have presented how to construct these models in electric circuits. We have shown that the spontaneous PTPT symmetry breaking as well as topological edge states are well signaled by measuring the frequency dependence of the impedance.

Acknowledgement

The author is very much grateful to N. Nagaosa for helpful discussions on the subject. This work is supported by the Grants-in-Aid for Scientific Research from MEXT KAKENHI (Grants No. JP17K05490 and No. JP18H03676). This work is also supported by CREST, JST (JPMJCR16F1 and JPMJCR20T2).

References

  • (1) M. Z. Hasan and C. L. Kane, Rev. Mod. Phys. 82, 3045 (2010).
  • (2) X.-L. Qi and S.-C. Zhang, Rev. Mod. Phys. 83, 1057 (2011).
  • (3) Q. Wu, A. A. Soluyanov, T. Bzdusek, Science 365, 1273 (2019)
  • (4) A. Tiwari and T. Bzdušek, Phys. Rev. B 101, 195130 (2020)
  • (5) E. Yang, B. Yang, O. You, H.-c. Chan, P. Mao, Q. Guo, S. Ma, L. Xia, D. Fan, Y. Xiang, S. Zhang, Phys. Rev. Lett. 125, 033901 (2020)
  • (6) P. M. Lenggenhager, X. Liu, S. S. Tsirkin, T. Neupert and T. Bzdusek, arXiv:2008.02807
  • (7) Q. Guo, T. Jiang, R.-Y. Zhang, L. Zhang, Z.-Q. Zhang, B. Yang, S. Zhang, C. T. Chan, Nature 594, 195 (2021)
  • (8) A. Bouhon, Q. Wu, R.-J. Slager, H. Weng, O. V. Yazyev and T. Bzdusek, Nature Physics 16, 1137 (2020)
  • (9) D. Wang, B. Yang, Q. Guo, R.-Y. Zhang, L. Xia, X. Su, W.-J. Chen, J. Han, S. Zhang, C. T. Chan, Light: Science & Applications 10, 83 (2021)
  • (10) H. Park, S. Wong, X. Zhang, S. S. Oh, arXiv:2102.12546
  • (11) Bin Jiang, Adrien Bouhon, Zhi-Kang Lin, Xiaoxi Zhou, Bo Hou, Feng Li, Robert-Jan Slager, Jian-Hua Jiang, arXiv:2104.13397
  • (12) M. Wang, S. Liu, Q. Ma, R.-Y. Zhang, D. Wang, Q. Guo, B. Yang, M. Ke, Z. Liu, and C. T. Chan, arXiv:2106.06711
  • (13) T. Jiang, Q. Guo, R.-Y. Zhang, Z.-Q. Zhang, B. Yang, C. T. Chan, arXiv:2106.16080
  • (14) B. Peng, A. Bouhon, B. Monserrat, R.-J. Slager, arXiv:2105.08733
  • (15) C. M. Bender and S. Boettcher, Phys. Rev. Lett. 80, 5243 (1998).
  • (16) R. El-Ganainy, K. G. Makris, M. Khajavikhan, Z. H. Musslimani, S. Rotter and D. N. Christodoulides, Nat. Physics 14, 11 (2018).
  • (17) C. M. Bender, D. C. Brody, and H. F. Jones, Phys. Rev. Lett. 89, 270401 (2002).
  • (18) S. Malzard, C. Poli, H. Schomerus, Phys. Rev. Lett. 115, 200402 (2015).
  • (19) V. V. Konotop, J. Yang, and D. A. Zezyulin, Rev. Mod. Phys. 88, 035002 (2016).
  • (20) T. Rakovszky, J. K. Asboth, and A. Alberti, Phys. Rev. B 95, 201407(R) (2017).
  • (21) B. Zhu, R. Lu and S. Chen, Phys. Rev. A 89, 062102 (2014).
  • (22) S. Yao and Z. Wang, Phys. Rev. Lett. 121, 086803 (2018).
  • (23) L. Jin and Z. Song, Phys. Rev. B 99, 081103 (2019).
  • (24) S.-D. Liang and G.-Y. Huang, Phys. Rev. A 87, 012118 (2013).
  • (25) D. Leykam, K. Y. Bliokh, Chunli Huang, Y. D. Chong, and Franco Nori, Phys. Rev. Lett. 118, 040401 (2017).
  • (26) S. Lieu, Phys. Rev. B 97, 045106 (2018).
  • (27) Z. Gong, Y. Ashida, K. Kawabata, K. Takasan, S. Higashikawa and M. Ueda, Phys. Rev. X 8, 031079 (2018).
  • (28) E. Cobanera, A. Alase, G. Ortiz, L. Viola, Phys. Rev. B 98, 245423 (2018).
  • (29) H. Jiang, C. Yang and S. Chen, Phys. Rev. A 98, 052116 (2018)
  • (30) A. Ghatak, T. Das, J. Phys.: Condens. Matter 31, 263001 (2019).
  • (31) Y. Ashida, Z. Gong, M. Ueda, Advances in Physics 69, 3 (2020)
  • (32) A. Mostafazadeh, J. Math. Phys 43, 205 (2002).
  • (33) C. E. Ruter, K. G. Makris, R. El-Ganainy, D. N. Christodoulides, M. Segev and D.Kip, Nat. Phys. 6, 192 (2010)
  • (34) C. Yuce, Physics Letters A 379, 12213 (2015).
  • (35) L. Feng, R. El-Ganainy and L. Ge, Nature Photonics 11, 752 (2017).
  • (36) S. Weimann, M. Kremer, Y. Plotnik, Y. Lumer, S. Nolte, K. G. Makris, M. Segev, M. C. Rechtsman and A. Szameit, Nature Materials 16, 433 (2017)
  • (37) N. Hatano and D. R. Nelson, Phys. Rev. Lett. 77, 570 (1996): Phys. Rev. B 56, 8651 (1997): Phys. Rev. B 58, 8384 (1998).
  • (38) F. Wilczek and A. Zee Phys. Rev. Lett. 52, 2111 (1984)
  • (39) K. Kawabata, K. Shiozaki, M. Ueda, and M. Sato, Phys. Rev. X 9, 041015 (2019).
  • (40) S. Imhof, C. Berger, F. Bayer, J. Brehm, L. Molenkamp, T. Kiessling, F. Schindler, C. H. Lee, M. Greiter, T. Neupert, R. Thomale, Nat. Phys. 14, 925 (2018).
  • (41) C. H. Lee , S. Imhof, C. Berger, F. Bayer, J. Brehm, L. W. Molenkamp, T. Kiessling and R. Thomale, Communications Physics, 1, 39 (2018).
  • (42) T. Helbig, T. Hofmann, C. H. Lee, R. Thomale, S. Imhof, L. W. Molenkamp and T. Kiessling, Phys. Rev. B 99, 161114 (2019).
  • (43) Y. Lu, N. Jia, L. Su, C. Owens, G. Juzeliunas, D. I. Schuster and J. Simon, Phys. Rev. B 99, 020302 (2019).
  • (44) K. Luo, R. Yu and H. Weng, Research (2018), ID 6793752.
  • (45) E. Zhao, Ann. Phys. 399, 289 (2018).
  • (46) M. Ezawa, Phys. Rev. B 98, 201402(R) (2018).
  • (47) M. Serra-Garcia, R. Susstrunk and S. D. Huber, Phys. Rev. B 99, 020304 (2019).
  • (48) T. Hofmann, T. Helbig, C. H. Lee, M. Greiter, R. Thomale, Phys. Rev. Lett. 122, 247702 (2019).
  • (49) M. Ezawa, Phys. Rev. B 100, 045407 (2019)
  • (50) M. Ezawa, Phys. Rev. B 102, 075424 (2020)
  • (51) M. Ezawa, Phys. Rev. B 99, 201411(R) (2019).
  • (52) M. Ezawa, Phys. Rev. B 99, 121411(R) (2019).
  • (53) T. Helbig, T. Hofmann, S. Imhof, M. Abdelghany, T. Kiessling, L. W. Molenkamp, C. H. Lee, A. Szameit, M. Greiter, R. Thomale, Nature Physics 16, 747 (2020)