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

Discriminant indicator with generalized rotational symmetry

Hiromasa Wakao1    Tsuneya Yoshida2    Yasuhiro Hatsugai2 1Graduate School of Pure and Applied Sciences, University of Tsukuba, Tsukuba, Ibaraki 305-8571, Japan
2Department of Physics, University of Tsukuba, Tsukuba, Ibaraki 305-8571, Japan
Abstract

Discriminant indicators with generalized inversion symmetry are computed only from data at the high-symmetry points. They allow a systematic search for exceptional points. In this paper, we propose discriminant indicators for two- and three-dimensional systems with generalized nn-fold rotational symmetry (n=4n=4, 66). As is the case for generalized inversion symmetry, the indicator taking a nontrivial value predicts the emergence of exceptional points and loops without ambiguity of the reference energy. A distinct difference from the case of generalized inversion symmetry is that the indicator with generalized nn-fold rotational symmetry (n=4n=4, 66) can be computed only from data at two of four high-symmetry points in the two-dimensional Brillouin zone. Such a difference is also observed in three-dimensional systems. Furthermore, we also propose how to fabricate a two-dimensional system with generalized four-fold rotational symmetry for an electrical circuit.

I Introduction

In recent decades, many efforts have been devoted to understanding the topological properties of wavefunctions [1, 2, 3, 4, 5, 6, 7, 8]. In particular, it turns out that symmetry enriches topological structure [9], which is exemplified by 2\mathbb{Z}_{2} topological insulators with time-reversal symmetry [10, 11, 12, 13, 14, 15] and topological crystalline materials [16, 17, 18, 19, 20, 21, 22, 23, 24]. Intriguingly, crystalline symmetry also plays a key role in determining topological properties. For instance, in the presence of the inversion symmetry and time-reversal symmetry, the topological invariant can be computed only from the parity eigenvalues at high-symmetry points in the Brillouin zone (BZ) [17]. This notion is generalized in Refs. [25, 26, 27, 28, 29] which introduced symmetry indicators as powerful tools for the systematic search for topological materials.

In parallel with this progress, recent extensive studies have opened up a new arena of topological physics: non-Hermitian systems [30, 31, 32, 33, 34, 35, 36, 37]. In these systems, the eigenvalues of the Hamiltonian may become complex, which induces exotic phenomena [38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51] such as the emergence of exceptional points [52, 53, 54, 55, 56, 57, 58, 59, 60, 61] and skin effects [62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72]. At the exceptional points, non-Hermitian topological properties protect band touching for both of the real and imaginary parts of the eigenvalues which are further enriched by symmetry [73, 74, 75, 76, 77, 78]. The skin effects result in anomalous sensitivity of energy spectrum on the presence or absence of boundaries [62, 63, 64, 66, 65, 67, 68, 69]. The exceptional points and skin effects have been reported for a wide range of systems such as open quantum systems [79, 80, 53, 81, 82], electrical circuits [83, 84, 85, 86, 87, 88], phononic systems [89, 90, 91, 92], photonic crystals [93, 94, 95, 96], and so on.

For the systematic search for non-Hermitian topological systems, symmetry indicators are powerful tools. Indeed, recent works have introduced indicators for non-Hermitian systems [70, 82, 97, 98, 99, 100]. Among them, the discriminant indicator [101] captures exceptional points without ambiguity of the reference energy.

Despite its usefulness, the discriminant indicator is introduced only for systems with generalized inversion symmetry [101]. For the systematic search for exceptional points, the extension to other symmetries is crucial.

In this paper, we extend the indicator to systems with generalized nn-fold rotational (CnC_{n}) symmetry for n=4n=4, 66. The indicators successfully capture the exceptional points in two dimensions without ambiguity of the reference energy, which is demonstrated by systematic analysis of toy models. We also introduce indicators which capture exceptional loops in the three-dimensional BZ. In contrast to the case with the generalized inversion symmetry [101], the indicator with generalized nn-fold rotational symmetry (n=4n=4, 66) can be computed only from data at two of the four high-symmetry points in the two-dimensional BZ. Such a difference is also observed in three-dimensional systems. We also show that the nontrivial value of the discriminant indicator with generalized C4C_{4} symmetry in three dimensions predicts the merging of the exceptional loops at the line specified by (kx,ky)=(0,0)(k_{x},k_{y})=(0,0) or (π,π)(\pi,\pi). Moreover, we propose an experimental realization of the topoelectrical system with generalized C4C_{4} symmetry.

The rest of this paper is organized as follows. In Sec. II, we derive the indicators for systems with generalized CnC_{n} symmetry (n=4n=4, 66). In Sec. III, we demonstrate that these indicators capture exceptional points and loops by numerically analyzing toy models with generalized CnC_{n} symmetry (n=4n=4, 66). In Sec. IV, we experimentally realize the topoelectrical circuit with generalized C4C_{4} symmetry. In Sec. V, we present a summary of this paper.

Refer to caption
Fig. 1: (Color online) (a) and (b) Sketches of the BZ and the path of integrations in Eq. (10). (a) The BZ for systems with generalized C4C_{4} symmetry. The high-symmetry points Γ\Gamma, MM, YY, and MM^{\prime} are denoted by blue dots. The paths Γ1\Gamma_{1}, Γ2\Gamma_{2}, Γ3\Gamma_{3}, and Γ4\Gamma_{4} connect high-symmetry points. The path Γ3\Gamma_{3}^{\prime} is the mapped path of Γ3\Gamma_{3} after applying the generalized C4C_{4} operator twice. (b) The BZ for systems with generalized C6C_{6} symmetry. The high-symmetry points Γ\Gamma, KK, MM, and KK^{\prime} are denoted by blue dots. The paths Γ1\Gamma_{1}, Γ2\Gamma_{2}, Γ3\Gamma_{3}, and Γ4\Gamma_{4} connect high-symmetry points. The path Γ2\Gamma_{2}^{\prime} is the mapped path of Γ2\Gamma_{2} after applying the generalized C6C_{6} operator three times. (c) and (d) Schematic figures of the three-dimensional BZs. Blue planes denote the area for kz=0k_{z}=0 and π\pi. (c) The BZ of the tetragonal lattice. The high-symmetry points, Γ\Gamma, MM, ZZ, and AA are denoted by red dots. (d) The BZ of the hexagonal lattice. The high-symmetry points, Γ\Gamma, MM, AA, and LL are denoted by red dots.

II Discriminant indicator with generalized rotational symmetry

In this section, we derive indicators for systems with generalized rotational symmetry. In Sec. II.1, along with a definition of generalized inversion symmetry, we provide an overview of our main results. The detailed derivation is provided in Secs. II.2-II.4. In Sec. II.5, we also discuss indicators for exceptional loops in three-dimensional systems.

II.1 Overview

Consider an N×NN\times N-Hamiltonian H(𝒌)H(\bm{k}) for two-dimensional systems with generalized CnC_{n} symmetry under the periodic boundary condition [70, 98, 72, 99, 97, 101, 51, 100, 61]. In this case, the Hamiltonian satisfies

UCnH(𝒌)UCn1=H(Rn𝒌),U_{C_{n}}H(\bm{k})U_{C_{n}}^{-1}=H^{\dagger}(R_{n}\bm{k}), (1)

with

Rn=(cos2π/nsin2π/nsin2π/ncos2π/n),R_{n}=\begin{pmatrix}\cos 2\pi/n&-\sin 2\pi/n\\ \sin 2\pi/n&\cos 2\pi/n\end{pmatrix}, (2)

and a unitary operator UCnU_{C_{n}} satisfying UCnn=1lN×NU_{C_{n}}^{n}=\mbox{1}\mbox{l}_{N\times N}.

The above non-Hermitian Hamiltonian may host exceptional points which are characterized by the discriminant number [59, 60, 78]

ν=d𝒌2πi𝒌lnΔ(𝒌).\nu=\oint\frac{d\bm{k}}{2\pi i}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta(\bm{k}). (3)

The integral is taken along a closed path in the BZ. The discriminant Δ(𝒌)\Delta(\bm{k}) of the characteristic polynomial P(𝒌,E)=det[H(𝒌)E1lN×N]=j=0NajEjP(\bm{k},E)=\det[H(\bm{k})-E\mbox{1}\mbox{l}_{N\times N}]=\sum_{j=0}^{N}a_{j}E^{j} is defined as

Δ(𝒌):=n>n(ϵnϵn)2.\Delta(\bm{k}):=\prod_{n>n^{\prime}}(\epsilon_{n}-\epsilon_{n^{\prime}})^{2}. (4)

Here, ϵn\epsilon_{n} denote the eigenvalues of H(𝒌)H(\bm{k}). Δ(𝒌)\Delta(\bm{k}) can be rewritten as

Δ(𝒌)=det(a0a1a2aN000a0a1aN1aN00a0a1a2aN1aNb1b2b3bN000b1b2bN1bN0000b1b2bN),\Delta(\bm{k})=\text{det}\begin{pmatrix}a_{0}&a_{1}&a_{2}&\cdots&a_{N}&0&\cdots&0\\ 0&a_{0}&a_{1}&\cdots&a_{N-1}&a_{N}&\cdots&0\\ \vdots&\ddots&\ddots&\ddots&\vdots&\ddots&\ddots&\vdots\\ 0&\cdots&a_{0}&a_{1}&a_{2}&\cdots&a_{N-1}&a_{N}\\ b_{1}&b_{2}&b_{3}&\cdots&b_{N}&0&\cdots&0\\ 0&b_{1}&b_{2}&\cdots&b_{N-1}&b_{N}&\cdots&0\\ \vdots&\ddots&\ddots&\ddots&\vdots&\ddots&\ddots&\vdots\\ 0&\cdots&0&0&b_{1}&b_{2}&\cdots&b_{N}\end{pmatrix}, (5)

with bj=jajb_{j}=ja_{j} [102, 103]. Clearly, Δ(𝒌)\Delta(\bm{k}) is determined by the coefficients of the characteristic polynomial. If ν\nu takes a non-zero value, Δ(𝒌)\Delta(\bm{k}) vanishes at a point inside of the loop. Such a point corresponds to an exceptional point because Eq. (4) indicates the band touching on this point.

In the presence of the generalized CnC_{n} symmetry, the discriminant Δ(𝒌)\Delta(\bm{k}) satisfies

Δ(𝒌)=Δ(Rn𝒌),\Delta(\bm{k})=\Delta^{\ast}(R_{n}\bm{k}), (6)

which is proven in Sec. II.2.

By making use of Eq. (6), we obtain discriminant indicators for two-dimensional systems with generalized CnC_{n} symmetry (n=4n=4, 66),

(1)νC4:=ξC4,2D=sgnΔ(Γ)sgnΔ(M),(-1)^{\nu_{C_{4}}}:=\xi_{C_{4},2D}=\text{sgn}\Delta(\Gamma)\text{sgn}\Delta(M), (7a)
(1)νC6:=ξC6,2D=sgnΔ(Γ)sgnΔ(M).(-1)^{\nu_{C_{6}}}:=\xi_{C_{6},2D}=\text{sgn}\Delta(\Gamma)\text{sgn}\Delta(M). (7b)

Here, νC4\nu_{C_{4}} and νC6\nu_{C_{6}} denote the discriminant number computed along the closed path illustrated in Fig. 1(a) [Fig. 1(b)]. Symbols Γ\Gamma and MM denote the high-symmetry points in the BZ [see Fig. 1]. Because of the constraint written in Eq. (6), the discriminant becomes real at Γ\Gamma and MM points.

We note that for n=2n=2, the problem is reduced to the case for the generalized inversion symmetry, which is discussed in Appendix A. For systems with generalized C3C_{3} symmetry, the Hamiltonian becomes Hermitian 111 This can be seen as follows, H(𝒌)=UC33H(𝒌)UC33=H(R33𝒌)=H(𝒌)H(\bm{k})=U_{C_{3}}^{3}H(\bm{k})U_{C_{3}}^{-3}=H^{\dagger}(R_{3}^{3}\bm{k})=H^{\dagger}(\bm{k}).. In Secs. II.3 and II.4, we derive these indicators. These indicators can be computed without the input of the reference energy, and predict the presence of exceptional points.

II.2 Constraints on the discriminant with generalized rotational symmetry

For the systems with generalized CnC_{n} symmetry, the polynomial P(𝒌,E)P(\bm{k},E) satisfies

P(𝒌,E)=det[UCnH(Rn𝒌)UCnE1lN×N]=detUCndet[H(Rn𝒌)E1lN×N]detUCn=det[H(Rn𝒌)E1lN×N].\begin{split}P(\bm{k},E)&=\det[U_{C_{n}}^{\dagger}H^{\dagger}(R_{n}\bm{k})U_{C_{n}}-E\mbox{1}\mbox{l}_{N\times N}]\\ &=\det U_{C_{n}}^{\dagger}\det[H^{\dagger}(R_{n}\bm{k})-E\mbox{1}\mbox{l}_{N\times N}]\det U_{C_{n}}\\ &=\det[H^{\ast}(R_{n}\bm{k})-E\mbox{1}\mbox{l}_{N\times N}].\end{split} (8)

Here, we have used detAT=detA\det A^{T}=\det A and det(AB)=detAdetB=detBdetA\det(AB)=\det A\det B=\det B\det A. We note that the discriminant can be computed only from coefficients of the polynomial. It results in Eq. (6). It is straight forward to extend the above arguments to three-dimensional cases. Specifically, for three-dimensional systems with generalized CnC_{n} symmetry about the zzaxis, the discriminant satisfies

Δ(𝒌,kz)=Δ(Rn𝒌,kz),\Delta(\bm{k}_{\|},k_{z})=\Delta^{\ast}(R_{n}\bm{k}_{\|},k_{z}), (9)

with 𝒌T=(kx,ky)\bm{k}_{\|}^{T}=(k_{x},k_{y}).

II.3 Two-dimensional system with generalized fourfold rotational symmetry

In order to derive Eq. (7a), let us consider the discriminant number evaluated along the closed path illustrated in Fig. 1(a). This integral can be decomposed into integrals along the path Γi\Gamma_{i} (i=1,,4i=1,\cdots,4),

pi=Γid𝒌2πi𝒌lnΔ(𝒌).p_{i}=\int_{\Gamma_{i}}\frac{d\bm{k}}{2\pi i}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta(\bm{k}). (10)

Here, Γi\Gamma_{i} (i=1,,4i=1,\cdots,4) are the paths between the high-symmetry points [see Fig. 1(a)]. First, we note that p2+p3p_{2}+p_{3} vanishes because applying the generalized C4C_{4} operator twice maps the path Γ3\Gamma_{3} to the path Γ3\Gamma_{3}^{\prime} which is equivalent to the path Γ2\Gamma_{2} [see Fig. 1(a)] due to the periodicity of the BZ.

Second, let us evaluate p1+p4p_{1}+p_{4}. The integration of Δ(𝒌)\Delta(\bm{k}) along the path Γ4\Gamma_{4} is mapped to that of Δ(𝒌)\Delta^{\ast}(\bm{k}) along the opposite direction of Γ1\Gamma_{1} by the generalized C4C_{4} symmetry. As a result, the integration p1+p4p_{1}+p_{4} is simplified as

2πi(p1+p4)=Γ1𝑑𝒌𝒌lnΔ(𝒌)+Γ4𝑑𝒌𝒌lnΔ(R41𝒌)=Γ1𝑑𝒌𝒌lnΔ(𝒌)Γ1𝑑𝒌𝒌lnΔ(𝒌)=2iImΓ1𝑑𝒌𝒌lnΔ(𝒌)=2iΓ1𝑑𝒌𝒌argΔ(𝒌)=2i[argΔ(M)argΔ(Γ)+2πN0],\begin{split}2\pi i(p_{1}+p_{4})&=\int_{\Gamma_{1}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta(\bm{k})\\ &+\int_{\Gamma_{4}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta^{\ast}(R_{4}^{-1}\bm{k})\\ &=\int_{\Gamma_{1}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta(\bm{k})-\int_{\Gamma_{1}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta^{\ast}(\bm{k})\\ &=2i\text{Im}\int_{\Gamma_{1}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta(\bm{k})\\ &=2i\int_{\Gamma_{1}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\arg\Delta(\bm{k})\\ &=2i\left[\arg\Delta(M)-\arg\Delta(\Gamma)+2\pi N_{0}\right],\end{split} (11)

with some integer N0N_{0}. From the second to the third lines, we have used Eq. (6). We note that this discussion can be extended to the cases of generalized C6C_{6} symmetry.

From these results, we obtain the relation

(1)νC4=eiπi=14pi=ei[argΔ(Γ)+argΔ(M)]=sgnΔ(Γ)sgnΔ(M).\begin{split}(-1)^{\nu_{C_{4}}}&=e^{i\pi\sum_{i=1}^{4}p_{i}}\\ &=e^{i\left[-\arg\Delta(\Gamma)+\arg\Delta(M)\right]}\\ &=\text{sgn}\Delta(\Gamma)\text{sgn}\Delta(M).\end{split} (12)

Here, we used the fact that Δ(𝒌)\Delta(\bm{k}) is real at Γ\Gamma and MM points because of Eq. (6). Therefore, we obtain the discriminant indicator with generalized C4C_{4} symmetry in Eq. (7a). We note that even if the exceptional points exist on the red line in Fig. 1(a), we can avoid by the continuous deformation of the integration path [24]. We also note that if Δ=0\Delta=0, exceptional points emerge at these points.

II.4 Two-dimensional system with generalized sixfold rotational symmetry

In order to derive Eq. (7b), let us consider the discriminant number composed along the closed path illustrated in Fig. 1(b). We decompose the path for the integral into four paths pip_{i} (i=1,,4i=1,\cdots,4) in a similar way to the previous section [see Eq. (10)].

Firstly, we discuss the integration of p2+p3p_{2}+p_{3} in a similar way to the previous case [see Eq. (11)]. In the presence of the generalized C6C_{6} symmetry, the integration of Δ(𝒌)\Delta(\bm{k}) along Γ2\Gamma_{2} is mapped to the one of Δ(𝒌)\Delta^{\ast}(\bm{k}) along Γ2\Gamma_{2}^{\prime}. Additionally, from the periodicity in the BZ, this integration is mapped to one along the opposite direction of Γ3\Gamma_{3}. Thus, in a similar way to the previous case, p2+p3p_{2}+p_{3} is simplified as

2πi(p2+p3)=Γ2𝑑𝒌𝒌lnΔ(R63𝒌)+Γ3𝑑𝒌𝒌lnΔ(𝒌)=Γ2𝑑𝒌𝒌lnΔ(𝒌)+Γ3𝑑𝒌𝒌lnΔ(𝒌)=Γ3𝑑𝒌𝒌lnΔ(𝒌)+Γ3𝑑𝒌𝒌lnΔ(𝒌)=2i[argΔ(K)argΔ(M)+2πN0],\begin{split}&2\pi i(p_{2}+p_{3})\\ &=\int_{\Gamma_{2}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta^{\ast}(R_{6}^{3}\bm{k})+\int_{\Gamma_{3}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta(\bm{k})\\ &=\int_{\Gamma_{2}^{\prime}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta^{\ast}(\bm{k})+\int_{\Gamma_{3}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta(\bm{k})\\ &=-\int_{\Gamma_{3}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta^{\ast}(\bm{k})+\int_{\Gamma_{3}}d\bm{k}\cdot\bm{\nabla}_{\bm{k}}\ln\Delta(\bm{k})\\ &=2i\left[\arg\Delta(K^{\prime})-\arg\Delta(M)+2\pi N_{0}\right],\end{split} (13)

with some integer N0N_{0}.

Secondly, let us simplify p1+p4p_{1}+p_{4}. The integration of Δ(𝒌)\Delta(\bm{k}) along Γ1\Gamma_{1} is mapped to the integration of Δ(𝒌)\Delta^{\ast}(\bm{k}) along the opposite direction of Γ4\Gamma_{4} by the generalized C6C_{6} operator. Hence, the integration p1+p4p_{1}+p_{4} is

2πi(p1+p4)=2i[argΔ(Γ)argΔ(K)+2πN0],2\pi i(p_{1}+p_{4})=2i\left[\arg\Delta(\Gamma)-\arg\Delta(K^{\prime})+2\pi N_{0}^{\prime}\right], (14)

with some integer N0N_{0}^{\prime}.

Putting Eqs. (13) and (14) together, we obtain Eq. (7b). Here, we used the fact that Δ(𝒌)\Delta(\bm{k}) is real at the Γ\Gamma and MM points because of Eq. (6).

II.5 Three-dimensional system

Let us see the discriminant number in the three-dimensional BZ and exceptional points from loops. In the presence of generalized C4C_{4} symmetry about the zzaxis, the following indicator captures the exceptional loops crossing one of the planes at kz=0k_{z}=0 and π\pi;

ξC4,3D=sgn[Δ(Γ)Δ(M)Δ(Z)Δ(A)].\xi_{C_{4},3D}=\text{sgn}\left[\Delta(\Gamma)\Delta(M)\Delta(Z)\Delta(A)\right]. (15)

Here, Γ\Gamma, MM, ZZ, and AA denote high-symmetry points as shown in Fig. 1(c). We note that Δ(Γ)\Delta(\Gamma), Δ(M)\Delta(M), Δ(Z)\Delta(Z), and Δ(A)\Delta(A) are real because of Eq. (9). All possible configurations of sgnΔ(𝒌)\text{sgn}\Delta(\bm{k}) giving ξCn,3D=1\xi_{C_{n},3D}=-1 are shown in Appendix B. The above indicator is obtained by applying the argument in Sec. II.3 to the two-dimensional subsystems at kz=0k_{z}=0 and π\pi where generalized C4C_{4} symmetry is preserved. In a similar way, we can introduce the indicator for systems with generalized C6C_{6} symmetry,

ξC6,3D=sgn[Δ(Γ)Δ(M)Δ(A)Δ(L)],\xi_{C_{6},3D}=\text{sgn}\left[\Delta(\Gamma)\Delta(M)\Delta(A)\Delta(L)\right], (16)

where Γ\Gamma, MM, AA, and LL are high-symmetry points in Fig. 1(d). We note that Δ(Γ)\Delta(\Gamma), Δ(M)\Delta(M), Δ(A)\Delta(A), and Δ(L)\Delta(L) are real because of Eq. (9).

III Applications to toy models

In this section, we demonstrate that the discriminant indicator with generalized CnC_{n} symmetry (n=4, 6n=4,\ 6) predicts the existence of the exceptional points and loops by numerically analyzing the two- and three-dimensional toy models with generalized CnC_{n} symmetry.

III.1 Two-dimensional toy model with generalized fourfold rotational symmetry

Refer to caption
Fig. 2: (Color online) (a) Schematic figure of the toy model with generalized C4C_{4} symmetry. The unit cell is enclosed by the black dashed box. The cross denotes the center of the generalized fourfold rotation. (b) Color plot of the map of the discriminant indicator ξC4,2D\xi_{C_{4},2D}. In the yellow regions (blue regions), the indicator takes a value of 11 (1-1). (c) Color map of argΔ(𝒌)/π\arg\Delta(\bm{k})/\pi against kxk_{x} and kyk_{y} for (δ,δt)=(0.1, 0.5)(\delta,\ \delta t)=(0.1,\ 0.5) [see the red dot in (b)]. Green and black dots represent exceptional points with ν=1\nu=1 and 1-1, respectively. (d) Sketch of the BZ. (e) and (f) The real and imaginary band structures for (δt,δ)=(0.1, 0.5)(\delta t,\ \delta)=(0.1,\ 0.5), respectively. We plot along the line kx=ky=kk_{x}=k_{y}=k in (d). Red dots denote exceptional points in the band structure.

Let us consider the two-dimensional toy model with generalized C4C_{4} symmetry in Fig. 2(a), whose Hamiltonian is defined in Appendix C. In Fig. 2(a), the red (blue) dots denote the on-site potentials iδi\delta (iδ-i\delta). The red (blue) lines denote the hoppings 1+δt1+\delta t (1δt1-\delta t).

The above model respects generalized C4C_{4} symmetry whose rotation axis is illustrated by a cross in Fig. 2(a). The explicit form of UC4U_{C_{4}} is written in Appendix C.

The numerical results for discriminant indicators are displayed in Fig. 2(b). The discriminant indicator takes vales of ξC4,2D=1\xi_{C_{4},2D}=1 and 1-1 in yellow and blue regions, respectively. We note that ξC4,2D\xi_{C_{4},2D} vanishes for δt=0\delta t=0 or δ=0\delta=0 since Δ(𝒌)\Delta(\bm{k}) vanishes at the high-symmetry point in the BZ. Additionally, at the boundary of the phases, ξC4,2D\xi_{C_{4},2D} vanishes for the same reason. For ξC4,2D=1\xi_{C_{4},2D}=-1, the exceptional points can be observed by computing argΔ(𝒌)/π\text{arg}\Delta(\bm{k})/\pi. As an example, we show the map of argΔ(𝒌)/π\text{arg}\Delta(\bm{k})/\pi in momentum space for (δ,δt)=(0.1, 0.5)(\delta,\ \delta t)=(0.1,\ 0.5) [see Fig. 2(c)]. Green and black dots in Fig. 2(c) represent exceptional points characterized by ν=1\nu=1 and 1-1, respectively. This result indicates that the discriminant number along the path in Fig. 1(a) takes a value of 1-1.

Figures 2(e) and 2(f) display the band structure along the line kx=ky=kk_{x}=k_{y}=k illustrated in Fig. 2(d). Figs. 2(e) and 2(f), we can find exceptional points.

III.2 Two-dimensional toy model with generalized sixfold rotational symmetry

Refer to caption
Fig. 3: (Color online) (a) Schematic figure of the toy model with generalized C6C_{6} symmetry. The unit cell is enclosed by the black dashed hexagon. The cross denotes the center of the generalized sixfold rotation. (b) Color plot of the map of the discriminant indicator ξC6,2D\xi_{C_{6},2D} for δ=0.1\delta=0.1. In the yellow regions (blue regions), the indicator takes a value of 11 (1)(-1). (c) Color map of argΔ(𝒌)/π\arg\Delta(\bm{k})/\pi for (δ,γ1,γ2)=(0.1, 0.3, 0.3)(\delta,\ \gamma_{1},\ \gamma_{2})=(0.1,\ 0.3,\ 0.3) [see the red dot in panel (b)] in the BZ. (d) Magnified version of panel (c) in the region enclosed by a green dashed triangle. (e) and (f) The real and imaginary band structures for 0.6kx0.70.6\leq k_{x}\leq 0.7 and 0.7ky0.80.7\leq k_{y}\leq 0.8 [see the green box in (d)]. The red point in the band structure represents an exceptional point.

We consider the two-dimensional toy model with generalized C6C_{6} symmetry in Fig. 3(a), whose Hamiltonian is defined in Appendix C. In this figure, the red (blue) dots denote the on-site potentials iδi\delta (iδ-i\delta). The red (blue) lines denote the non-reciprocal hoppings. Specifically, the hopping along the red (blue) arrows is 1+γ11+\gamma_{1} (1+γ21+\gamma_{2}). The hopping opposite to the red (blue) arrows is 1γ11-\gamma_{1} (1γ21-\gamma_{2}). For the blue lines without arrows, the hopping is 11 for both directions.

The above model respects generalized C6C_{6} symmetry whose rotation axis is illustrated by a cross in Fig. 3(a). The explicit form of UC6U_{C_{6}} is written in Appendix C.

We show the numerical result for the discriminant indicator ξC6,2D\xi_{C_{6},2D} in Eq. (7b) for δ=0.1\delta=0.1 [see Fig. 3(b)]. In Fig. 3(b), there exist two phases where the discriminant indicator takes values of ξC6,2D=1\xi_{C_{6},2D}=1 (yellow area) and 1-1 (blue area). We note that ξC6,2D\xi_{C_{6},2D} vanishes for γ1=0\gamma_{1}=0 and γ2=0\gamma_{2}=0 since Δ(𝒌)\Delta(\bm{k}) vanishes at the high-symmetry point in the BZ. Additionally, at the boundary between phases, ξC6,2D\xi_{C_{6},2D} vanishes for the same reason. In the blue-colored region, the indicator predicts the presence of exceptional points. To be concrete, we compute argΔ(𝒌)/π\text{arg}\Delta(\bm{k})/\pi in the momentum space for (δ,γ1,γ2)=(0.1, 0.3, 0.3)(\delta,\ \gamma_{1},\ \gamma_{2})=(0.1,\ 0.3,\ 0.3) [see Figs. 3(c) and 3(d)]. Green and black dots in Fig. 3(d) represent exceptional points characterized by ν=1\nu=1 and 1-1, respectively. Figure 3(d) indicates that the discriminant number computed along the path in Fig. 1(b) takes a value of 1-1.

Figures 3(e) and 3(f) plot the band structure for (δ,γ1,γ2)=(0.1, 0.3, 0.3)(\delta,\gamma_{1},\gamma_{2})=(0.1,\ 0.3,\ 0.3), which indicates the emergence of the exceptional point at the point denoted by the red dots in Figs. 3(e) and 3(f).

III.3 Three-dimensional toy model with generalized fourfold rotational symmetry

Refer to caption
Fig. 4: (Color online) (a) Color plot of the map of the discriminant indicator ξC4,3D\xi_{C_{4},3D}. In the yellow regions (blue-colored regions), the indicator takes a value of 11 (1)(-1). (b)-(c) Color maps of argΔ(𝒌)\text{arg}\Delta(\bm{k}) for (d1,d2)=(1,1)(d_{1},d_{2})=(1,1) in the BZ [see the red dot in (a)]. (b) and (c) are displayed at the plane for kz=0k_{z}=0 and kz=πk_{z}=\pi in the BZ, respectively. Green and black dots represent the exceptional loops crossing the plane for fixed kzk_{z} with ν=1\nu=1 and ν=1\nu=-1, respectively. (d) Momentum points 𝒌\bm{k} satisfying absΔ(𝒌)<0.05\text{abs}\Delta(\bm{k})<0.05.

Consider the three-dimensional toy model with generalized fourfold rotational symmetry whose Hamiltonian reads

H(𝒌)=i=13Ri(𝒌)σi,H(\bm{k})=\sum_{i=1}^{3}R_{i}(\bm{k})\sigma_{i}, (17)

with

R1(𝒌)\displaystyle R_{1}(\bm{k}) =icoskz,\displaystyle=i\cos k_{z}, (18a)
R2(𝒌)\displaystyle R_{2}(\bm{k}) =(coskxcosky)+i,\displaystyle=(\cos k_{x}-\cos k_{y})+i, (18b)
R3(𝒌)\displaystyle R_{3}(\bm{k}) =d1coskz+sinkz+d2(coskx+cosky)+1.\displaystyle=d_{1}\cos k_{z}+\sin k_{z}+d_{2}(\cos k_{x}+\cos k_{y})+1. (18c)

Here, σi\sigma_{i} (i=1i=1, 22, 33) denote the Pauli matrices. The above model respects generalized C4C_{4} symmetry about the zzaxis with UC4=σ3U_{C_{4}}=\sigma_{3}.

Figure 4(a) displays the numerical result for the discriminant indicator ξC4,3D\xi_{C_{4},3D}. In Fig. 4(a), there exist two phases where the discriminant indicator takes values of ξC4,3D=1\xi_{C_{4},3D}=1 (yellow area) and ξC4,3D=1\xi_{C_{4},3D}=-1 (blue area). For ξC4,3D=1\xi_{C_{4},3D}=-1, the exceptional loops crossing the plane for kz=0k_{z}=0 or π\pi can be obtained by computing the map of argΔ(𝒌)/π\text{arg}\Delta(\bm{k})/\pi. As an example, we show the maps of argΔ(kx,ky,0)\text{arg}\Delta(k_{x},k_{y},0) and argΔ(kx,ky,π)\text{arg}\Delta(k_{x},k_{y},\pi) in momentum space for (d1,d2)=(1,1)(d_{1},d_{2})=(1,1) [see Figs. 4(b) and 4(c)]. In Figs. 4(b) and 4(c), green and black dots represent exceptional loops characterized by ν=1\nu=1 and 1-1, respectively. By comparing these figures, we obtain the emergence of the exceptional loops crossing the plane for kz=πk_{z}=\pi.

Figure 4(d) displays the momentum points 𝒌\bm{k} satisfying absΔ(𝒌)<0.05\text{abs}\Delta(\bm{k})<0.05, indicating the emergence of exceptional loops. We note that exceptional loops predicted by the nontrivial value of ξC4,3D\xi_{C_{4},3D} merge at the line specified by (kx,ky)=(0,0)(k_{x},k_{y})=(0,0) or (π,π)(\pi,\pi), originating from the symmetry constraint on the exceptional points for fixed kzk_{z} in three dimensions (for more details see Appendix D).

Refer to caption
Fig. 5: (Color online) (a) Color plot of the map of the discriminant indicator ξC6,3D\xi_{C_{6},3D}. In the yellow regions (blue regions), the indicator takes a value of 11 (1)(-1). (b) and (c) Color maps of argΔ(𝒌)\text{arg}\Delta(\bm{k}) for (d1,d2)=(1,1)(d_{1},d_{2})=(1,1) [see the red dot in panel (a)] in the BZ. Panel (b) and Panel (c) are displayed at the plane for kz=0k_{z}=0 and kz=πk_{z}=\pi in the BZ, respectively. Green and black dots represent the exceptional loops crossing the plane for fixed kzk_{z} with ν=1\nu=1 and ν=1\nu=-1, respectively. (d) Momentum points 𝒌\bm{k} satisfying absΔ(𝒌)<0.05\text{abs}\Delta(\bm{k})<0.05.

III.4 Three-dimensional toy model with generalized sixfold rotational symmetry

Consider the three-dimensional toy model with generalized sixfold rotational symmetry whose Hamiltonian reads

H(𝒌)=i=13Ri(𝒌)σi,H(\bm{k})=\sum_{i=1}^{3}R_{i}(\bm{k})\sigma_{i}, (19)

with

R1(𝒌)\displaystyle R_{1}(\bm{k}) =id1C(kx,ky)+icoskz+i,\displaystyle=id_{1}C(k_{x},k_{y})+i\cos k_{z}+i, (20a)
R2(𝒌)\displaystyle R_{2}(\bm{k}) =S(kx,ky)+isinkz,\displaystyle=S(k_{x},k_{y})+i\sin k_{z}, (20b)
R3(𝒌)\displaystyle R_{3}(\bm{k}) =d2C(kx,ky)+2iS(kx,ky)\displaystyle=d_{2}C(k_{x},k_{y})+2iS(k_{x},k_{y})
+5coskz+sinkz+1.\displaystyle\quad+5\cos k_{z}+\sin k_{z}+1. (20c)

Here, C(kx,ky)C(k_{x},k_{y}) and S(kx,ky)S(k_{x},k_{y}) are defined as

C(kx,ky)\displaystyle C(k_{x},k_{y}) =cos(kx)+cos(kx/2+3ky/2)\displaystyle=\cos(-k_{x})+\cos\left(k_{x}/2+\sqrt{3}k_{y}/2\right)
+cos(kx/23ky/2),\displaystyle\quad+\cos\left(k_{x}/2-\sqrt{3}k_{y}/2\right), (21a)
S(kx,ky)\displaystyle S(k_{x},k_{y}) =sin(kx)+sin(kx/2+3ky/2)\displaystyle=\sin(-k_{x})+\sin\left(k_{x}/2+\sqrt{3}k_{y}/2\right)
+sin(kx/23ky/2).\displaystyle\quad+\sin\left(k_{x}/2-\sqrt{3}k_{y}/2\right). (21b)

The above model respects generalized C6C_{6} symmetry about the zzaxis with UC6=σ3U_{C_{6}}=\sigma_{3}.

Figure 5(a) displays the numerical result of the discriminant ξC6,3D\xi_{C_{6},3D}. In Fig. 5(a), there exist two phases where the discriminant indicator takes values of ξC6,3D=1\xi_{C_{6},3D}=1 (yellow area) and ξC6,3D=1\xi_{C_{6},3D}=-1 (blue area). For ξC6,3D=1\xi_{C_{6},3D}=-1, the exceptional loops crossing the plane for kz=0k_{z}=0 or π\pi can be obtained by computing the map of argΔ(𝒌)/π\text{arg}\Delta(\bm{k})/\pi. As an example, we show the maps of argΔ(kx,ky,0)\text{arg}\Delta(k_{x},k_{y},0) and argΔ(kx,ky,π)\text{arg}\Delta(k_{x},k_{y},\pi) in the momentum space for (d1,d2)=(1,1)(d_{1},d_{2})=(1,1) [see Figs. 5(b) and 5(c)]. In Fig. 5(c), green and black dots represent exceptional loops characterized by ν=1\nu=1 and 1-1, respectively. By comparing Fig. 5(b) with 5(c), we obtain the emergence of the exceptional loops crossing the plane for kz=πk_{z}=\pi.

Figure 5(d) displays the momentum points 𝒌\bm{k} satisfying absΔ(𝒌)<0.05\text{abs}\Delta(\bm{k})<0.05, indicating the emergence of exceptional loops. We note that, in contrast to the case of generalized C4C_{4} symmetry in Sec. III.3, generalized C6C_{6} symmetry does not predict the merging of the exceptional loops since the exceptional loops can cross the boundary of the BZ (for more details see Appendix D).

IV Topoelectrical systems with generalized fourfold rotational symmetry

Refer to caption
Fig. 6: (Color online) Schematic figure of the topoelectrical system with generalized C4C_{4} symmetry. The unit cell is enclosed by the dashed black box. We omit power sources connecting each nodes.

We have shown that our indicators capture the exceptional points and loops in toy models. In this section, we discuss the relevance of generalized C4C_{4} symmetry to the topoelectrical circuits [105, 106, 87].

Consider the two-dimensional topoelectrical circuit with generalized C4C_{4} symmetry in Fig. 6. To describe this system, under the periodic boundary condition, we define the voltages at each node as 𝑽(𝒌,ω)\bm{V}(\bm{k},\omega) and electric currents between the node and the power source as 𝑰(𝒌,ω)\bm{I}(\bm{k},\omega). Here, ω\omega denotes the frequency of the currents from the power source into the node. The admittance of the red (blue) resistors shown in Fig. 6 is RrR_{\mathrm{r}} (RbR_{\mathrm{b}}) with Rr=1+δrR_{\mathrm{r}}=1+\delta r and Rb=1δrR_{\mathrm{b}}=1-\delta r. The admittance of red (blue) inductors is iω1Lr1-i\omega^{-1}L_{\mathrm{r}}^{-1} (iω1Lb1-i\omega^{-1}L_{\mathrm{b}}^{-1}), with Lr=(1+δ)1L_{\mathrm{r}}=(1+\delta)^{-1} and Lb=(1δ)1L_{\mathrm{b}}=(1-\delta)^{-1}. The relation between 𝑰(𝒌,ω)\bm{I}(\bm{k},\omega) and 𝑽(𝒌,ω)\bm{V}(\bm{k},\omega) is

𝑰(𝒌,ω)=J(𝒌,ω)𝑽(𝒌,ω).\bm{I}(\bm{k},\omega)=J(\bm{k},\omega)\bm{V}(\bm{k},\omega). (22)

Here, J(𝒌,ω)J(\bm{k},\omega) denotes the admittance matrix,

J(𝒌,ω)=61l4×4+(iω)1diag(Lb1,Lb1,Lr1,Lr1)(0RrRrRrRr0RrRrRrRr0RrRrRrRr0)(0RbeikxRbeikxRbeikyRbeikx0RbeikyRbeikxRbeikxRbeiky0RbeikyRbeikyRbeikxRbeiky0).\begin{split}J(\bm{k},\omega)&=6\mbox{1}\mbox{l}_{4\times 4}+(i\omega)^{-1}\text{diag}(L_{\mathrm{b}}^{-1},L_{\mathrm{b}}^{-1},L_{\mathrm{r}}^{-1},L_{\mathrm{r}}^{-1})\\ &-\begin{pmatrix}0&R_{\mathrm{r}}&R_{\mathrm{r}}&R_{\mathrm{r}}\\ R_{\mathrm{r}}&0&R_{\mathrm{r}}&R_{\mathrm{r}}\\ R_{\mathrm{r}}&R_{\mathrm{r}}&0&R_{\mathrm{r}}\\ R_{\mathrm{r}}&R_{\mathrm{r}}&R_{\mathrm{r}}&0\end{pmatrix}\\ &-\begin{pmatrix}0&R_{\mathrm{b}}e^{ik_{x}}&R_{\mathrm{b}}e^{ik_{x}}&R_{\mathrm{b}}e^{ik_{y}}\\ R_{\mathrm{b}}e^{-ik_{x}}&0&R_{\mathrm{b}}e^{-ik_{y}}&R_{\mathrm{b}}e^{-ik_{x}}\\ R_{\mathrm{b}}e^{-ik_{x}}&R_{\mathrm{b}}e^{ik_{y}}&0&R_{\mathrm{b}}e^{ik_{y}}\\ R_{\mathrm{b}}e^{-ik_{y}}&R_{\mathrm{b}}e^{ik_{x}}&R_{\mathrm{b}}e^{-ik_{y}}&0\end{pmatrix}.\end{split} (23)

We note that the constant diagonal terms just shift the band structure and do not contribute to the emergence of exceptional points. Thus, we define J(𝒌,ω)J^{\prime}(\bm{k},\omega) by removing the constant diagonal terms of the admittance matrix,

J(𝒌,ω)=J(𝒌,ω)[TrJ(𝒌,ω)]1l4×4/4.J^{\prime}(\bm{k},\omega)=J(\bm{k},\omega)-[\text{Tr}J(\bm{k},\omega)]\mbox{1}\mbox{l}_{4\times 4}/4. (24)

Clearly, J(𝒌,ω)J^{\prime}(\bm{k},\omega) corresponds to the Hamiltonian in Sec. III.1.

V Summary

Extending the argument of the work in [101], we have introduced the discriminant indicators for two- and three-dimensional systems with generalized CnC_{n} symmetry (n=4n=4, 66). In two dimensions, the indicators can be computed from the parity of the discriminant at the Γ\Gamma and MM points [see Eqs. (7a) and (7b)], which is in contrast to the case of generalized inversion symmetry. These indicators taking a nontrivial value predict the emergence of exceptional points and exceptional loops without the input of the reference energy. We have numerically confirmed that these indicators capture the exceptional points and the loops for two- and three-dimensional toy models.

Acknowledgements.
The authors thank R. Okugawa for fruitful discussion. This work is supported by JSPS KAKENHI Grants No. JP17H06138, No. JP20H04627, and No. JP21K13850 and also by JST CREST, Grant No. JPMJCR19T1, Japan.
Refer to caption
Fig. 7: (Color online) Sketch of the BZ with generalized C2C_{2} symmetry and the path of integrations in Eq. (10). The high-symmetry points are represented by blue dots. The path Γi\Gamma_{i} (i=1,,6i=1,\cdots,6) connects the high-symmetry points.

Appendix A Discriminant indicator with generalized twofold rotational symmetry

We briefly discuss the indicators for systems with generalized C2C_{2} symmetry, although the problem is reduced to the one for generalized inversion symmetry.

With generalized C2C_{2} symmetry in two-dimensional systems, the discriminant number is simplified to the computation at the high-symmetry points in the BZ [101]. This symmetry enables us to rewrite Eq. (3) as the discriminant indicator,

(1)νC2:=ξC2,2D=sgn[Δ(Γ)Δ(X)Δ(M)Δ(Y)],(-1)^{\nu_{C_{2}}}:=\xi_{C_{2},2D}=\text{sgn}\left[\Delta(\Gamma)\Delta(X)\Delta(M)\Delta(Y)\right], (25)

where νC2\nu_{C_{2}} denotes the discriminant number defined along the path in Fig. 7. Symbols Γ\Gamma, XX, MM, and YY denote the high-symmetry points [see Fig. 7]. We note that Δ(Γ)\Delta(\Gamma), Δ(X)\Delta(X), Δ(M)\Delta(M), and Δ(Y)\Delta(Y) are real because of Eq. (6).

We see the derivation of Eq. (25). Let us consider the discriminant number [see Eq. (4)] composed along the closed path illustrated in Fig. 7. In a way similar to what we did in Sec. II.3, we decompose the closed path into six parts. The integral for the paths Γi\Gamma_{i} (i=1,,6i=1,\cdots,6) is denoted by pip_{i} (i=1,,6i=1,\cdots,6) in Eq. (10).

We note that p1=p4p_{1}=-p_{4} holds due to the periodicity of the BZ. We also note that p2+p3p_{2}+p_{3} is simplified as

2πi(p2+p3)=2i[argΔ(Γ)argΔ(Y)+2πN0],2\pi i(p_{2}+p_{3})=2i\left[\text{arg}\Delta(\Gamma)-\text{arg}\Delta(Y)+2\pi N_{0}\right], (26)

with some integer N0N_{0}. This is because Γ3\Gamma_{3} is mapped to Γ2\Gamma_{2} by applying the operator of generalized C2C_{2} symmetry. In a similar way, we have

2πi(p5+p6)=2i[argΔ(M)argΔ(X)+2πN0],2\pi i(p_{5}+p_{6})=2i\left[\text{arg}\Delta(M)-\text{arg}\Delta(X)+2\pi N_{0}^{\prime}\right], (27)

with some integer N0N_{0}^{\prime}.

Putting Eqs. (26) and (27) together, we obtain Eq. (25). Here, we have used the fact that Δ(𝒌)\Delta(\bm{k}) is real at the Γ\Gamma, XX, MM, and YY points because of Eq. (6).

Refer to caption
Fig. 8: (Color online) (a)-(h) The sketch of the configuration of the sign of the discriminant giving ξCn,3D=1\xi_{C_{n},3D}=-1. The squares whose vertices are Γ\Gamma, MM, ZZ, and AA (Γ\Gamma, MM, LL, and AA) denote the area in the BZ with generalized C4C_{4} (C6C_{6}) symmetry. The blue (red) dots are the sign of the discriminant sgnΔ(𝒌)=1\text{sgn}\Delta(\bm{k})=1 (1-1). Green lines represent the examples of exceptional loops.

Appendix B Specific configuration with the nontrivial value of the discriminant indicator

In this appendix, we illustrate the specific configuration yielding ξCn,3D=1\xi_{C_{n},3D}=-1 (see Fig. 8). With this nontrivial value of the indicators, the exceptional loops emerge crossing one of the planes for kz=0k_{z}=0 or π\pi exist. For example, in Fig. 8(a), there exist an exceptional loop crossing the plane for kz=πk_{z}=\pi, and two loops crossing both of the planes for kz=0k_{z}=0 and π\pi. In this case, at the plane for kz=0(π)k_{z}=0\ (\pi), the indicator for two-dimensional systems takes a trivial (nontrivial) value. This result indicates that the discriminant indicator takes a value of 1-1.

Appendix C Details of the Hamiltonian in toy models

Here, we show the details of the Hamiltonian in Sec. III. The bulk Hamiltonian describing the toy model in Fig. 2(a) is

HC4(𝒌)=(iδ(1+δt)(1δt)eikx(1+δt)(1δt)eikx(1+δt)(1δt)eiky(1+δt)(1δt)eikxiδ(1+δt)(1δt)eiky(1+δt)(1δt)eikx(1+δt)(1δt)eikx(1+δt)(1δt)eikyiδ(1+δt)(1δt)eiky(1+δt)(1δt)eiky(1+δt)(1δt)eikx(1+δt)(1δt)eikyiδ).H_{C_{4}}(\bm{k})=\begin{pmatrix}i\delta&-(1+\delta t)-(1-\delta t)e^{-ik_{x}}&-(1+\delta t)-(1-\delta t)e^{-ik_{x}}&-(1+\delta t)-(1-\delta t)e^{-ik_{y}}\\ -(1+\delta t)-(1-\delta t)e^{ik_{x}}&i\delta&-(1+\delta t)-(1-\delta t)e^{ik_{y}}&-(1+\delta t)-(1-\delta t)e^{ik_{x}}\\ -(1+\delta t)-(1-\delta t)e^{ik_{x}}&-(1+\delta t)-(1-\delta t)e^{-ik_{y}}&-i\delta&-(1+\delta t)-(1-\delta t)e^{-ik_{y}}\\ -(1+\delta t)-(1-\delta t)e^{ik_{y}}&-(1+\delta t)-(1-\delta t)e^{-ik_{x}}&-(1+\delta t)-(1-\delta t)e^{ik_{y}}&-i\delta\end{pmatrix}. (28)

This model preserves generalized C4C_{4} symmetry [see Eq. (1)] where UC4U_{C_{4}} is defined as

UC4=(0001001010000100).U_{C_{4}}=\begin{pmatrix}0&0&0&1\\ 0&0&1&0\\ 1&0&0&0\\ 0&1&0&0\end{pmatrix}. (29)

Additionally, the bulk Hamiltonian describing the toy model in Fig. 3(a) is

HC6(𝒌)=(iδ(1γ1)(1γ2)ei𝒌𝒂2ei𝒌𝒂2(1+γ2)ei𝒌(𝒂1+𝒂2)(1γ1)(1+γ1)iδ(1+γ1)(1+γ2)ei𝒌(𝒂1+𝒂2)ei𝒌(𝒂1+𝒂2)(1γ2)ei𝒌𝒂1(1+γ2)ei𝒌𝒂2(1γ1)iδ(1γ1)(1γ2)ei𝒌𝒂1ei𝒌𝒂1ei𝒌𝒂2(1γ2)ei𝒌(𝒂1+𝒂2)(1+γ1)iδ(1+γ1)(1+γ2)ei𝒌𝒂2(1γ2)ei𝒌(𝒂1+𝒂2)ei𝒌(𝒂1+𝒂2)(1+γ2)ei𝒌𝒂1(1γ1)iδ(1γ1)(1+γ1)(1+γ2)ei𝒌𝒂1ei𝒌𝒂1(1γ2)ei𝒌𝒂2(1+γ1)iδ).\begin{split}&H_{C_{6}}(\bm{k})\\ &=\begin{pmatrix}i\delta&-(1-\gamma_{1})&-(1-\gamma_{2})e^{i\bm{k}\cdot\bm{a}_{2}}&-e^{i\bm{k}\cdot\bm{a}_{2}}&-(1+\gamma_{2})e^{i\bm{k}\cdot(-\bm{a}_{1}+\bm{a}_{2})}&-(1-\gamma_{1})\\ -(1+\gamma_{1})&-i\delta&-(1+\gamma_{1})&-(1+\gamma_{2})e^{i\bm{k}\cdot(-\bm{a}_{1}+\bm{a}_{2})}&-e^{i\bm{k}\cdot(-\bm{a}_{1}+\bm{a}_{2})}&-(1-\gamma_{2})e^{-i\bm{k}\cdot\bm{a}_{1}}\\ -(1+\gamma_{2})e^{-i\bm{k}\cdot\bm{a}_{2}}&-(1-\gamma_{1})&i\delta&-(1-\gamma_{1})&-(1-\gamma_{2})e^{-i\bm{k}\cdot\bm{a}_{1}}&-e^{-i\bm{k}\cdot\bm{a}_{1}}\\ -e^{-i\bm{k}\cdot\bm{a}_{2}}&-(1-\gamma_{2})e^{-i\bm{k}\cdot(-\bm{a}_{1}+\bm{a}_{2})}&-(1+\gamma_{1})&-i\delta&-(1+\gamma_{1})&-(1+\gamma_{2})e^{-i\bm{k}\cdot\bm{a}_{2}}\\ -(1-\gamma_{2})e^{-i\bm{k}\cdot(-\bm{a}_{1}+\bm{a}_{2})}&-e^{-i\bm{k}\cdot(-\bm{a}_{1}+\bm{a}_{2})}&-(1+\gamma_{2})e^{i\bm{k}\cdot\bm{a}_{1}}&-(1-\gamma_{1})&i\delta&-(1-\gamma_{1})\\ -(1+\gamma_{1})&-(1+\gamma_{2})e^{i\bm{k}\cdot\bm{a}_{1}}&-e^{i\bm{k}\cdot\bm{a}_{1}}&-(1-\gamma_{2})e^{i\bm{k}\cdot\bm{a}_{2}}&-(1+\gamma_{1})&-i\delta\end{pmatrix}.\end{split} (30)

Here, 𝒂1=(12,32)\bm{a}_{1}=\left(\frac{1}{2},\frac{\sqrt{3}}{2}\right) and 𝒂2=(12,32)\bm{a}_{2}=\left(-\frac{1}{2},\frac{\sqrt{3}}{2}\right) are the unit vectors.

This model preserves generalized C6C_{6} symmetry [see Eq. (1)] where UC6U_{C_{6}} is defined as

UC6=(000001100000010000001000000100000010).U_{C_{6}}=\begin{pmatrix}0&0&0&0&0&1\\ 1&0&0&0&0&0\\ 0&1&0&0&0&0\\ 0&0&1&0&0&0\\ 0&0&0&1&0&0\\ 0&0&0&0&1&0\end{pmatrix}. (31)

Appendix D The merging of exceptional loops

Refer to caption
Fig. 9: (Color online) (a) and (b) Sketch of sections of the BZ at given kzk_{z} for systems with generalized C4C_{4} and C6C_{6} symmetry, respectively. The square and the hexagon denote the section of the BZ. When the indicator is nontrivial, exceptional loops emerge in the three-dimensional BZ. Correspondingly, exceptional points emerge in the two-dimensional subspace as denoted by green and blue dots. Here, blue (green) dots represent the exceptional points with ν=1\nu=1 (1-1). In these panels, shaded area are denoted by the region enclosed by the integration path in Figs. 1(a) and 1(b).

In this appendix, we show the details of the merging of exceptional loops arising by generalized rotational symmetry in three dimensions. Consider the case in which the discriminant indicator ξCn,3D\xi_{C_{n},3D} (n=4n=4, 66) takes the nontrivial value. In this case, the exceptional loops emerge, meaning that the presence of the exceptional point in a two-dimensional subspace is specified by a given kzk_{z} (see Fig. 9). We note that an odd number of exceptional points exist in the shaded area in the BZ due to the nontrivial value of the two-dimensional indicator for this subspace. We also note that applying the generalized rotation, an exceptional point is mapped to another one with the opposite sign of the velocity ν\nu (see Fig. 9).

The merging of the exceptional loops observed in Fig. 4(d) can be understood by shifting kzk_{z}. Because of the indicator taking a value of ξC4,3D=1\xi_{C_{4},3D}=-1, the exceptional points observed in Fig. 9(a) should annihilate each other. Such annihilation is allowed only at (kx,ky)=(0,0)(k_{x},k_{y})=(0,0) and (π,π)(\pi,\pi) in the presence of generalized C4C_{4} symmetry. In the following, we see the details.

Refer to caption
Fig. 10: (Color online) The path of the exceptional points in the BZ by shifting kzk_{z} with generalized (a)-(d) C4C_{4} and (e)-(h) C6C_{6} symmetry. Blue (green) dots represent the exceptional points with ν=1\nu=1 (1-1). The red arrow is the path of exceptional points by shifting kzk_{z}.

Consider a proper kzk_{z} where the exceptional points emerge [see Fig. 9(a)]. Shifting kzk_{z}, we can see that these exceptional points in the shaded area move to the boundaries, which is categorized into four cases [see Figs. 10(a)-10(d)]. In the case illustrated in Fig. 10(a), the exceptional point does not meet another one, and thus, pair annihilation does not occur. In the case illustrated in Fig. 10(b), the exceptional point meets another one but with the same vorticity, and thus, pair annihilation does not occur. In the case illustrated in Figs. 10(c) and 10(d), the exceptional point meets another one with opposite vorticities. The above argument elucidates that two exceptional loops merge at the line specified by (kx,ky)=(0,0)(k_{x},k_{y})=(0,0) and (π,π)(\pi,\pi).

In a similar way, let us consider the case of generalized C6C_{6} symmetry. In this case, the indicator ξC6,3D=1\xi_{C_{6},3D}=-1 does not necessary mean the merging of exceptional loops. Let us illustrate the details. As is the case for the generalized C4C_{4} symmetry, the movement of the exceptional point for the pair annihilation is categorized into four cases in Figs. 10(e)-10(h). In the case illustrated in Fig. 10(e), the exceptional point does not meet another one, and thus, pair annihilation does not occur. In the case illustrated in Figs. 10(f) and 10(g), the exceptional point meets another one with the opposite vorticity. Thus, pair annihilation occurs. In the case illustrated in Fig. 10(h), the exceptional point meets another one with the same vorticity, and thus, pair annihilation does not occur. The above argument indicates that the indicator ξC6,3D=1\xi_{C_{6},3D}=-1 does not necessarily mean the merging of the exceptional points.

References