Abstract
Topological materials exhibit edgelocalized scatteringfree modes protected by their nontrivial bulk topology through the bulkedge correspondence in Hermitian systems. While topological phenomena have recently been much investigated in nonHermitian systems with dissipations and injections, the fundamental principle of their edge modes has not fully been established. Here, we reveal that, in nonHermitian systems, robust gapless edge modes can ubiquitously appear owing to a mechanism that is distinct from bulk topology, thus indicating the breakdown of the bulkedge correspondence. The robustness of these edge modes originates from yet another topological structure accompanying the branchpoint singularity around an exceptional point, at which eigenvectors coalesce and the Hamiltonian becomes nondiagonalizable. Their characteristic complex eigenenergy spectra are applicable to realize lasing wave packets that propagate along the edge of the sample. We numerically confirm the emergence and the robustness of the proposed edge modes in the prototypicalÂ lattice models. Furthermore, we show that these edge modes appear in a model of chiral active matter based on the hydrodynamic description, demonstrating that active matter can exhibit an inherently nonHermitian topological feature. The proposed general mechanism would serve as an alternative designing principle to realize scatteringfree edge current in nonHermitian devices, going beyond the existing frameworks of nonHermitian topological phases.
Introduction
Since topological materials exhibit robust scatteringfree currents along the edges of samples, the notion of topology has attracted much interest both in fundamental physics and in device engineering. The first discovery of such a topological material dates back to the integer quantum Hall effect^{1}, where it has been established that the gapless edge modes precisely correspond to the bulk topological number^{2}. This phenomenon has revealed a fundamental principle known as the bulkedge correspondence. Nowadays, the topological materials have been found in much broader situations especially in the presence of symmetry, such as the timereversal symmetry in topological insulators^{3,4}. In such situations, the bulkedge correspondence is still valid and predicts the presence or absence of robust edge modes^{4}, from which a periodic table has been obtained^{5}. These discoveries have opened up a stream of material designs on the basis of the bulk band topology^{4,6}.
Although the conventional notion of topological materials is based on Hermitian Hamiltonians, effective Hamiltonians can become nonHermitian in nonconservative systems including both quantum and classical ones, such as photonics^{7,8,9,10}, ultracold atoms^{11,12}, optomechanics^{13,14}, electronic circuits^{15,16}, mechanical lattices^{17,18}, and biophysical systems^{19}. For example, in photonic systems, nonHermiticity can be introduced by engineering optical gain and loss through semiconductor amplifiers or acoustic modulators. Classification of nonHermitian topological materials has been explored in terms of bulk band topology^{10,20,21,22,23,24,25,26,27,28,29,30,31,32}, and a periodic table has been proposed in the same spirit as in the Hermitian case^{23,30,31}. However, the bulkedge correspondence is more subtle in nonHermitian systems than in Hermitian systems, as a hitherto unknown nonHermitian effect may protect unpredicted edge modes or destabilize edge modes in topologically nontrivial systems.
In this article, we reveal a ubiquitous mechanism for realizing robust gapless edge modes, which emerge independently of the bulk topology and instead are protected in an unconventional manner unique to nonHermitian systems. This indicates that the bulkbased classification cannot conclusively predict the existence or absence of edge modes in the nonHermitian case, thus implying the breakdown of the bulkedge correspondence. In the conventional topologically nontrivial systems, the proposed mechanism can further stabilize the gapless edge modes, even against symmetrybreaking disorder. These edge modes inherently exhibit large positive imaginary parts of the eigenenergies and thus are naturally applicable to topological insulator laser^{33,34,35}, where the amplified unidirectional wave packet propagates along the edge of the sample. We demonstrate the emergence of the proposed gapless edge modes and the lasing wave packets by analyzing the prototypical tightbinding models.
Our edge modes owe their robustness to the distinct topological structure of exceptional points (EPs), and thus here we term these modes as exceptional edge modes. The EP^{36} is a singular point in the parameter space at which two or more eigenvectors and eigenvalues coalesce and a parameterized Hamiltonian becomes nondiagonalizable. The EP is unique to nonHermitian systems and induces intriguing phenomena, such as interchanging eigenvectors after encircling an EP^{37}, coherent perfect absorption^{38}, and unidirectional invisibility^{39}. The existence of EPs is supported by the nontrivial topology of the branchpoint singularity in intersecting Riemann surfaces around them.^{37,40} In onedimensional systems, such as the edge modes of twodimensional bulk systems, the emergence of EPs can be guaranteed by satisfying certain symmetries, including the PT symmetry, the CP symmetry, the pseudoHermiticity, and the chiral symmetry^{41,42,43}. EPs can disappear if either the symmetry is broken or a pair of EPs coalesce; the latter is reminiscent of the pairannihilation of Weyl points^{44,45} in Hermitian systems. We discover a general mechanism that EPs join two edge dispersions like glue and make them robust against disorder, which cannot be predicted by the existing periodic tables of topological phases^{23,30,31}.
Furthermore, we explicitly show the existence of exceptional edge modes in a more realistic system based on active matter^{46}, which is a collection of selfpropelled particles and has recently attracted much interest as a useful platform to study biological and outofequilibrium physics. Recent studies^{47,48,49,50,51,52} have explored the existence of the edge modes in active matter protected by the bulk topology. Some of them^{49,50,52} have utilized chiral active matter, which moves in a circular path or selfrotates. Chiral active matter has been experimentally realized, for example, in bacteria^{53} and artificial Lshaped particles^{54}. The hydrodynamics^{55,56} and the phase separation^{57} of chiral active matter have also been analyzed in recent studies. The effective Hamiltonian of the linearized hydrodynamic equations in active matter is, in general, nonHermitian because of inherent dissipations and energy injections therein. We demonstrate that this type of nonHermitian chiral active matter provides an ideal platform to experimentally realize the proposed exceptional edge modes.
Results
Exceptional edge modes in twolayered nonHermitian BernevigHughesZhang model
We first construct and analyze a minimal tightbinding model. For a Hermitian Hamiltonian H, timereversal symmetry means that there exists a unitary operator T satisfying TH(k)T^{âˆ’1}â€‰=â€‰H^{*}(âˆ’k), where H(k) is the Bloch Hamiltonian constructed from H^{4}. The definition of timereversal symmetry can be extended to nonHermitian systems and it has been pointed out^{21,23,30,31} that there are two types of timereversal symmetry, i.e., TH(k)T^{âˆ’1}â€‰=â€‰H^{*}(âˆ’k) and TH(k)T^{âˆ’1}â€‰=â€‰H^{T}(âˆ’k), which are equivalent in Hermitian systems while not in nonHermitian cases. One can construct the conventional timereversalsymmetric topological insulator by coupling a Chern insulator with its timereversal counterpart^{3,58}. The bulk bands of a timereversalsymmetric insulator are topologically characterized by the \({{\mathbb{Z}}}_{2}\) index^{3,4}, which corresponds to the parity of the number of the edge modes across the Fermi energy. If we construct a timereversalsymmetric system from Chern insulators with even numbers of edge modes, we obtain a topologically trivial bulk. However, we reveal that such a trivial bulk can still accompany robust gapless edge modes by introducing nonHermitian coupling between the two Chern insulators (see Fig. 1a).
To construct the minimal model for demonstrating the emergence of such edge modes, we consider the twolayered QiWuZhang (QWZ) model, H_{0}â€‰=â€‰I_{2} âŠ— H_{QWZ}â€‰+â€‰cÏƒ_{x} âŠ— I_{2}, which exhibits two chiral modes per edge in the bulk energy gap. Here, H_{QWZ} is the Hamiltonian of the QWZ model^{59}, which can be described as \({H}_{{\rm{QWZ}}}({\bf{k}})=\sin {k}_{x}{\sigma }_{x}+\sin {k}_{y}{\sigma }_{y}+(u+\cos {k}_{x}+\cos {k}_{y}){\sigma }_{z}\) in the wavenumber space (see Supplementary Methods for the realspace description). Here, I_{2} is the 2â€‰Ã—â€‰2 identity matrix and Ïƒ_{i} is the ith component of the Pauli matrices. Also, we assume that c is real, and thus the Hamiltonian is still Hermitian. By coupling H_{0} and its timereversal counterpart \({H}_{0}^{* }\) with a nonHermitian term, \(i\Sigma =i(\beta +\beta ^{\prime} ){I}_{2}\otimes {\sigma }_{x}/2+i(\beta \beta ^{\prime} ){\sigma }_{z}\otimes {\sigma }_{x}/2\) with Î², \(\beta ^{\prime}\) being real parameters, we obtain the following nonHermitian Hamiltonian
We note that this model resembles the BernevigHughesZhang model^{58} but differs from it since our model has two layers of the QWZ model and two other layers of the timereversal QWZ model, which are coupled by the nonHermitian term. Furthermore, as this Hamiltonian has the pseudoHermiticity defined as Î·H(k)Î·^{âˆ’1}â€‰=â€‰H^{â€ }(k) which can lead to another topological classification characterized by the \({\mathbb{Z}}\) invariant^{30,31}, we add a Hermitian coupling and consider the Hamiltonian \(H^{\prime} =H+\gamma {\sigma }_{x}\otimes {\sigma }_{y}\otimes {\sigma }_{x}\), to break the pseudoHermiticity. The additional Hermitian coupling corresponds to the spin coupling in condensed matter and thus can open an energy gap in the conventional trivial insulator. We can confirm that this Hamiltonian has timereversal symmetry \(TH^{\prime} ({\bf{k}}){T}^{1}={H}^{^{\prime} * }({\bf{k}})\), and thus have to consider \({{\mathbb{Z}}}_{2}\) indices as in Hermitian systems (see Supplementary Note 10). Below we focus on the parameter regimes in which the bulk bands are trivial in the conventional sense, i.e., the number of the edge modes in H_{0} is even.
To reveal the existence of robust edge modes, we calculate the band structure of our model with open (periodic) boundaries in the x (y) direction. Figure 1b shows the band structure for the wavenumber in the y direction. There, gapless edge bands exist in the bulk energy gap and they accompany EPs, where both the eigenenergies and the eigenstates coalesce. The EPs act as a glue that holds the edge band structures together and thus stabilize the existence of exceptional edge modes. This gluing is reminiscent of the branchpoint structure in nonHermitian bulk bands^{42,60}, which remains until the EPs coalesce. In general models including the present one, the edge modes between two EPs exhibit the large imaginary parts of the eigenenergies, whereas all the bulk modes can have zero imaginary parts of the eigenenergies. As discussed below, this property finds a possible application to realize a topological insulator laser^{33,34}.
To explicitly demonstrate that the appearance of EPs is independent of the bulk topology and thus violates the bulkedge correspondence, we numerically calculate the edge band structures for different strengths of the nonHermitian coupling. By modifying the strengths of the nonHermitian coupling Î², \(\beta ^{\prime}\), we can control the existence of edge modes and EPs in the bulk gap as shown in Fig. 2. On the other hand, during this modification, the bulk energy gap remains open. Therefore, the bulk topology should remain trivial at arbitrary strength of the nonHermitian coupling and thus have no relation to the exceptional edge modes. This result indicates that while the bulk band topology still can predict the existence of the ordinary edge modes without EPs, it fails to predict the existence of robust exceptional edge modes.
Although we have concentrated on timereversalsymmetric systems so far, timereversal symmetry is not the prerequisite for realizing exceptional edge modes. If the sum of the Chern numbers of the bulk bands below the energy gap is zero, a system without relevant symmetries cannot exhibit gapless edge modes protected by bulk topology. However, combining the topological systems with the opposite Chern numbers by the nonHermitian coupling, we can obtain not only a trivial bulk but also robust exceptional edge modes.
We can also realize exceptional edge modes in topologically nontrivial systems. We construct the nonHermitian BernevigHughesZhang model,
where \(i\Sigma ^{\prime}\) is the nonHermitian coupling \(i\Sigma ^{\prime} =i\beta {\sigma }_{x}\), and Î² is real. This model satisfies the timereversal symmetry and associates with a nontrivial \({{\mathbb{Z}}}_{2}\) invariant. We calculate the edge band structure and confirm the existence of the exceptional edge modes (see Supplementary Note 2). We note that the exceptional edge modes can also exist robustly against timereversalsymmetrybreaking disorder and thus can be more advantageous than conventional edge modes.
Effective edge Hamiltonian and robustness of exceptional edge modes
Robustness against the perturbation and the disorder is an important feature of topological edge modes. We note that conventional topological edge modes are fragile under the perturbations breaking the symmetry. To see what types of perturbations can sustain stable exceptional edge modes, we introduce a general onedimensional effective Hamiltonian parametrized by wavenumber k_{y},
which describes the generic behavior of the lowenergy dispersion of edge modes. The diagonal elements represent the linear dispersion of two edge modes without couplings, and the offdiagonal parts represent the nonHermitian coupling. The case of \(\alpha =\gamma ={\rm{Im}}\ \beta =0\) represents exceptional edge modes in the disorderfree system (\({\rm{Re}}\ \beta\, \ne\, 0\) is necessary to generate EPs in exceptional edge modes). The effective Hamiltonian has the eigenenergy \({E}^{\pm }({k}_{y})={E}_{0}\pm \sqrt{{({k}_{y}+\alpha )}^{2}{\beta }^{2}{\gamma }^{2}}\) and EPs at \({k}_{y}=\alpha \pm \sqrt{{\beta }^{2}+{\gamma }^{2}}\). The topological index associated with EPs can guarantee their presence in the complex wavenumber space in this case (see Supplementary Note 3).
As we consider the bulk gaps for the real parts of eigenenergies, the gapless edge modes remain when there exists a real wavenumber k_{y} that satisfies \({\rm{Re}}\sqrt{{({k}_{y}+\alpha )}^{2}{\beta }^{2}{\gamma }^{2}}=0\). Thus, we can conclude that \( {\rm{Im}}\sqrt{{\beta }^{2}+{\gamma }^{2}} \le  {\rm{Im}}\ \alpha \) is a necessary and sufficient condition to realize robust edge modes (see Methods). Meanwhile, the exceptional edge modes remain when \(\alpha \pm \sqrt{{\beta }^{2}+{\gamma }^{2}}\) is real, i.e., (i) \({\rm{Im}}\ \alpha =0\), (ii) \({\rm{Im}}\ ({\beta }^{2}+{\gamma }^{2})=0\), and (iii) \({\rm{Re}}\ ({\beta }^{2}+{\gamma }^{2})\, > \, 0\). From conditions (ii), (iii), we can derive \({\rm{Im}}\sqrt{{\beta }^{2}+{\gamma }^{2}}=0\) and thus can confirm that the condition for gapless edge modes must be satisfied under the condition for exceptional edge modes. We note that conditions (i), (ii) are equivalent to the condition for pseudoHermiticity^{61}, which ensures that the eigenenergies are either real or pairs of complex conjugate values (i.e., \({({k}_{y}+\alpha )}^{2}{\beta }^{2}{\gamma }^{2}\) is real in the present case). In general, disorder in the existing terms satisfies this condition. Meanwhile, nonzero \({\rm{Re}}\ \alpha\) breaks the timereversal symmetry defined as Tâ€‰=â€‰Ïƒ_{y}, TH(k_{y})T^{âˆ’1}â€‰=â€‰H^{*}(âˆ’k_{y}). However, perturbation to \({\rm{Re}}\ \alpha\) does not affect the stability of exceptional edge modes even if they accompany nontrivial bulk topology of a timereversalsymmetric system.
We can also relate the robustness of the exceptional edge modes to the symmetry and the topology of EPs. In onedimensional systems, EPs can robustly exist under the PT symmetry, the CP symmetry, the pseudoHermiticity, or the chiral symmetry^{41,42,43}. In the case of \({E}_{0}=\alpha =\gamma ={\rm{Im}}\ \beta =0\), the effective edge Hamiltonian exhibits the PT symmetry PTâ€‰=â€‰Ïƒ_{z}, PTH(k_{y})(PT)^{âˆ’1}â€‰=â€‰H^{*}(k_{y}), the CP symmetry CPâ€‰=â€‰Ïƒ_{x}, CPH(k_{y})(CP)^{âˆ’1}â€‰=â€‰âˆ’H^{*}(k_{y}), the pseudoHermiticity Î·â€‰=â€‰Ïƒ_{z}, Î·H(k_{y})Î·^{âˆ’1}â€‰=â€‰H^{â€ }(k_{y}), and the chiral symmetry Î“â€‰=â€‰Ïƒ_{x}, Î“H(k_{y})Î“^{âˆ’1}â€‰=â€‰âˆ’H^{â€ }(k_{y}). We note that the pseudoHermiticity here is in a narrower class than that considered in the previous paragraph, that is, the operator Î· is restricted to a local operator that only acts on the inner degrees of freedom. To preserve the PT symmetry and/or the chiral symmetry, we need \({\rm{Im}}\ \alpha ={\rm{Im}}\ \beta ={\rm{Re}}\ \gamma =0\). We can also confirm that the preservation of the CP symmetry and/or the pseudoHermiticity requires \({\rm{Im}}\ \alpha ={\rm{Im}}\ \beta ={\rm{Im}}\ \gamma =0\). From these equations, we can derive conditions (i), (ii) for realizing the exceptional edge modes discussed above and thus confirm that the exceptional edge modes robustly exist under sufficiently small \( {\rm{Im}}\ \gamma \) and one of the following symmetry: the PT symmetry, the CP symmetry, the chiral symmetry, or the pseudoHermiticity. If we increase \( {\rm{Im}}\ \gamma \), two EPs coalesce at critical strength of \( {\rm{Im}}\ \gamma \), and the exceptional edge modes disappear under larger \( {\rm{Im}}\ \gamma \). We can expect that the symmetry in the effective edge Hamiltonian is the same as that in the bulk and thus can utilize the symmetry as the guiding principle to predict what types of disorders remain exceptional edge modes.
To confirm the robustness of the exceptional edge modes in our model, we calculate the band structure with adding disorder (see Fig. 3). We show that the exceptional edge modes still exist robustly under certain types of disorders, i.e., the random real onsite potential and the imaginary noise in the coupling terms (see Methods for details). These disorders preserve the modified PT symmetry \(P^{\prime} TH({k}_{x},{k}_{y}){(P^{\prime} T)}^{1}\equiv{H}^{* }({k}_{x},{k}_{y})\) that has the same role as the PT symmetry in the edge band structure. Thus, the result is consistent with the discussion in the previous paragraph. Also, the onsite nonHermitian term, igÏƒ_{z} âŠ— I âŠ— I, recovers the robustness of the edge modes against the real noise in the coupling terms, which lifts the degeneracy in the edge bands without onsite terms. With the onsite nonHermitian term, as the two edge modes avoid each other in the imaginary part of the energy, they are not degenerate and thus are prohibited to open the real gaps (similar feature has been observed in the previous study^{10} at the interface between gain and loss regions). This avoidance protects the edge modes from opening gaps as understood from the perturbation theory (see Supplementary Note 4 for details). These results are consistent with the analysis of the effective edge Hamiltonian. In Supplementary Notes 11 and 12, we further discuss the symmetry of the disordered Hamiltonian and clarify its relation to the robustness of the exceptional edge modes in both topologically trivial and nontrivial systems. Especially, we demonstrate the existence of the chiralsymmetryprotected exceptional edge modes and the importance of the modification of the PT symmetry and the CP symmetry for the protection of the exceptional edge modes.
In general, in Hermitian systems, the physical significance of the periodic table obtained from the bulk band topology is guaranteed by the bulkedge correspondence that consistently predicts the presence or absence of robust gapless edge modes at open boundaries^{4}. In contrast, in nonHermitian cases, our findings force us to fundamentally alter this point of view. In particular, when gâ€‰â‰ â€‰0 and Î³â€‰=â€‰0 in our model, there exist the robust gapless edge modes as in Fig. 3d (gâ€‰â‰ â€‰0 and Î³â€‰â‰ â€‰0), whereas the bulk topological invariant is trivial as inferred from the topological classification^{23,30,31} (see Supplementary Note 10 for details). In other words, the robust gapless edge modes found here violate the bulkedge correspondence and cannot be captured by the existing periodic tables^{23,30,31} of nonHermitian topological phases, thus challenging the conventional classification based on Bloch Hamiltonians.
Application to amplifying edge modes
Next, we show that amplified exceptional edge modes with nonzero group velocity can be realized. Specifically, we find that the general form of effective edge Hamiltonians is given by
We derive the dispersion relation of this effective Hamiltonian, \(E({k}_{y})={E}_{0}+[(a1){k}_{y}\pm \sqrt{{(a+1)}^{2}{k}_{y}^{2}4\beta \beta ^{\prime} }]/2\), which exhibits EPs at \({k}_{y}=\pm 2\sqrt{\beta \beta ^{\prime} }/(a+1)\) and nonzero group velocity. Although this Hamiltonian describes the generic behavior of lasing edge modes utilizing exceptional edge modes, we construct a concrete tightbinding model represented by the following Hamiltonian:
where H_{QWZ} is the Hamiltonian of the QWZ model. Figure 4a shows the edge band structure of this system. Nonzero imaginary parts of the eigenenergies appear only in the edge modes as in our first model. Also, the edge modes exhibit nonzero slopes of the real energy dispersion \(\partial {\rm{Re}}E/\partial k\), which correspond to nonzero group velocities. Thus, we can observe the amplified wave packet propagating along the edge of the sample, which allows us to stably transfer the energy and thus may find potential applications. We note that this Hamiltonian is neither timereversal symmetric nor pseudoHermitian. The sum of the Chern numbers for the bands under the energy gap is zero in our model, which indicates the bulk triviality in the conventional sense. Therefore, the edge modes are protected not by the bulk band topology but by the EPs.
In general, exceptional edge modes are essential for this construction of a topological insulator laser. To obtain lasing edge modes, we must utilize a pair of edge modes localized at the same side whose dispersion relations cross each other without coupling terms. Also, to accomplish nonzero group velocity, the absolute values of the slopes of the edge energy bands must be different. Therefore, the degeneracy is not protected by the bulk band topology or the symmetry and thus can be resolved by Hermitian couplings as shown in Fig. 4b. On the contrary, nonHermitian couplings lead to both the enhancement and the robustness of the edge mode. Thus, the lasing edge mode must be an exceptional edge mode.
We demonstrate the realspace dynamics of our topological insulator laser (Eq. (5)) by numerical calculations (see Supplementary movies 13). Figure 5 shows the snapshots for the realspace distributions of the probability densities of the wave functions. Without nonHermitian coupling, \(\beta =\beta ^{\prime} =0\), the bulk oscillation survives. With nonHermitian coupling, \(\beta ,\beta ^{\prime}\, \ne\, 0\), the bulk oscillation becomes much smaller than the edge oscillation in a short time, and only the edge mode remains even if we start with the random initial state. Also, we can confirm that the edge mode has nonzero group velocity. Furthermore, we introduce disorder on the edge and excite only one edge site. Then, we obtain the propagating edge mode without backscattering. This implies the robustness of the exceptional edge mode against the disorder at the edge. In contrast to the previous research^{33,34}, we do not need to introduce judicious gain along the edge. This difference can potentially facilitate the realization of topological insulator laser in various physical setups.
Active matter realization of exceptional edge modes
Analogous to the conventional topological edge modes^{62}, the exceptional edge modes can also exist in continuum systems. We construct a continuum toy model and confirm the existence of exceptional edge modes by calculating the band structure (see Methods and Supplementary Note 7 for the detail of the model). Figure 6 represents the edge band structure of the continuum model. Although the bulk bands are topologically trivial as in the tightbinding model (Eq. (1)), it exhibits the robust exceptional edge modes.
To show that exceptional edge modes are indeed realizable in realistic systems, we focus on a continuum active matter model. We consider chiral active matter without the leftright symmetry in which each particle moves on a clockwise (or counterclockwise) circular trajectory (Fig. 7a). We mix clockwise and counterclockwise moving particles. We also assume that the chirality of active particles flips occasionally and the flipping rate Î³ is symmetric between clockwise and counterclockwise moving particles. The active particles have long and narrow shapes. We assume that polar interaction acts on them, which aligns the neighboring particles and effectively appears in some selfpropelled rods^{63}. Antipolar interaction is also allowed to exist between particles with opposite chirality. This setup can possibly be experimentally realized by utilizing bacteria^{53}, artificial Lshaped particles^{54}, or robotic rotors^{52}. To be concrete, we expect that exceptional edge modes can appear in bacteria swimming between the two plates at the distance shorter than the bacteria length and in Lshaped active particles that are occasionally turned over (see Supplementary Note 8 for further details). Here, the crucial requirements for the experimental realization of exceptional edge modes are the flippable chirality and the momentum coupling.
In Fig. 7b, we show the existence of exceptional edge modes by numerically diagonalizing the effective Hamiltonian of our active matter model, which is derived by linearizing the hydrodynamic equations^{46,64} (see Methods and Supplementary Methods). We confirm that a pair of EPs appear at the frequency Ï‰â€‰=â€‰0 and support the robustness of the edge modes. Meanwhile, at a glance, there are degeneracies in the bulk gap. However, the edge modes avoid each other in the imaginary part of the frequency as in Fig. 3d, and thus these apparent degeneracies are robust against the disorder. We obtain two other crossings on the \({\rm{Re}}\ E=0\) axis, which correspond to the points where unprotected edge bands appear from the bulk bands around the axis (see Supplementary Note 9 for the detail on the function of these crossing). In realistic experimental situations, we expect that the oscillation of the fluctuation of the density or the velocity field propagates at the edge of the sample in the direction depending on the chirality of particles (i.e., clockwise or counterclockwise) when we apply the perturbation with a small frequency compared to the bulk bandgap, which is almost equal to the frequency of rotation Ï‰. The imaginary parts of the eigenvalues are all nonpositive, and thus we need further modification of the active system to apply the proposed setup to lasing devices.
Discussion
We revealed the existence of robust gapless edge modes unique to nonHermitian systems by utilizing EPs. These edge modes, which we called exceptional edge modes, can exist even when the bulk topology is trivial. We also analyzed and confirmed the robustness of the edge modes by constructing the effective edge Hamiltonian. By utilizing these edge modes, we proposed a topological insulator laser whose edge modes were amplified and propagate along the edge. We also showed that the chiral active particles with chirality flipping can exhibit the exceptional edge modes and thus they can be realized in the upcoming experimental techniques of active matter, whereas the model analyzed here has only nonpositive imaginary parts of the eigenfrequencies and thus exhibits no lasing behavior.
The edge modes found here provide an alternative design principle to realize scatteringfree edge current intrinsic to nonHermitian systems, which is not based on the bulk topology and thus indicates that the conventional arguments on bulk topology, including the periodic tables^{23,30,31}, are insufficient to predict the presence or absence of robust edge modes in nonHermitian systems. Exceptional edge modes in higherdimensional systems are important to further elucidate a nontrivial role of open boundaries in nonHermitian systems. Furthermore, our active matter model demonstrates that hydrodynamics of active matter can be applied to nonHermitian topological phenomena, indicating that active matter provides a useful platform for exploring nonHermitian topology.
Methods
QWZ model and numerical calculations of the tightbinding model
The Hamiltonian of the QWZ model in the wavenumber space can be described as \(H({\bf{k}})=\sin {k}_{x}{\sigma }_{x}+\sin {k}_{y}{\sigma }_{y}+(u+\cos {k}_{x}+\cos {k}_{y}){\sigma }_{z}\)^{59}. This model exhibits the two bulk bands with the Chern number Câ€‰=â€‰Â±1 and a bandgap around Eâ€‰=â€‰0. By transforming it to the realspace basis via the inverse Fourier transformation, we obtain the Hamiltonian in the real space (see Supplementary Methods for the detailed description). To calculate the edge band structures, we transform only \(\sin {k}_{x}\) and \(\cos {k}_{x}\), and obtain the Hamiltonian of our model with the hybrid boundary conditions, i.e., the open boundary conditions in the x direction and the periodic boundary conditions in the y direction. We arrange the 50 unit cells in the x direction and consider the 1â€‰Ã—â€‰50 superribbon structures. Diagonalizing the obtained Hamiltonians, we calculate the edge band structures of our model.
We also introduce the disorder terms in Fig. 3. We use a(x){I_{2}, Ïƒ_{z}} âŠ— {I_{2}, Ïƒ_{z}} âŠ— {I_{2}, Ïƒ_{z}} and b(x){Ïƒ_{x}, iÏƒ_{y}} âŠ— {I_{2}, Ïƒ_{z}} âŠ— {Ïƒ_{x}, iÏƒ_{y}} as the random real onsite potential and the imaginary and real noise in the nonHermitian coupling for each, where brackets mean that we introduce all the combinations made by choosing either one in each bracket. a(x) and b(x) are random values for each x from uniform distributions ranging \([W,W]\in {\mathbb{R}}\) or \(i[W,W]\in i{\mathbb{R}}\). We set a(x) to be real and Wâ€‰=â€‰0.5 in Fig. 3b. Also, we set b(x) to be imaginary (real) for the imaginary (real) noise in the nonHermitian coupling and Wâ€‰=â€‰0.02 (Wâ€‰=â€‰0.1) in Fig. 3b (Fig. 3c, d). We also consider onsite imaginary potential igÏƒ_{z} âŠ— I_{2} âŠ— I_{2} and set gâ€‰=â€‰0.2 in Fig. 3d.
To calculate the realspace dynamics of lasing modes in the finite system, we perform the inverse Fourier transformation for all the terms in H(k) and impose the open boundary conditions both in the x and y directions. We arrange the 20â€‰Ã—â€‰20 sites, which have four sublattices for each of the sites. We calculate the time evolution under the Hamiltonian by using the fourthorder RungeKutta method. We set the time step dtâ€‰=â€‰0.001 and use the parameters uâ€‰=â€‰âˆ’1. The parameters of the nonHermitian coupling are set to be \(\beta =\beta ^{\prime} =0\) in Fig. 5a, Î²â€‰=â€‰0.2, \(\beta ^{\prime} =0.1\) in Fig. 5b, and Î²â€‰=â€‰0.9, \(\beta ^{\prime} =0.8\) in Fig. 5c.
Derivation of the condition for remaining gapless edge modes
From the analysis of the effective edge Hamiltonian (Eq. (3)), we conclude that \( {\rm{Im}}\sqrt{{\beta }^{2}+{\gamma }^{2}} \le  {\rm{Im}}\ \alpha \) is the necessary and sufficient condition for the existence of gapless edge modes. To show this, we start from the equation \({\rm{Re}}\sqrt{{({k}_{y}+\alpha )}^{2}{\beta }^{2}{\gamma }^{2}}=0\), which leads to the eigenenergy E^{Â±}â€‰=â€‰E_{0}â€‰Â±â€‰iÎ´ with Î´ being a real number. For a wavenumber satisfying this equation, the real parts of the two eigenenergies become the same, and thus we obtain the gapless edge modes. Then, we prove that the existence of such a wavenumber is equivalent to the condition \( {\rm{Im}}\sqrt{{\beta }^{2}+{\gamma }^{2}} \le  {\rm{Im}}\ \alpha \). We consider \(\alpha ^{\prime} ={\rm{Im}}\ \alpha\) and \(k^{\prime} ={k}_{y}+{\rm{Re}}\ \alpha\), and obtain \({({k}_{y}+\alpha )}^{2}={k}_{y}^{^{\prime} 2}{\alpha }^{^{\prime} 2}+2ik^{\prime} \alpha ^{\prime}\). Similarly, if we describe \(\sqrt{{\beta }^{2}+{\gamma }^{2}}=l+i\beta ^{\prime}\), we obtain \({\beta }^{2}+{\gamma }^{2}={l}^{2}{\beta }^{^{\prime} 2}+2il\beta ^{\prime}\). To make \(\sqrt{{({k}_{y}+\alpha )}^{2}{\beta }^{2}{\gamma }^{2}}\) a real or a pure imaginary number, we have to set \({\rm{Im}}[{({k}_{y}+\alpha )}^{2}{\beta }^{2}{\gamma }^{2}]=0\) and thus consider \(k^{\prime} =l\beta ^{\prime} /\alpha ^{\prime}\). Then, \({\rm{Re}}[{({k}_{y}+\alpha )}^{2}{\beta }^{2}{\gamma }^{2}]=({l}^{2}+{\alpha }^{^{\prime} 2})({\beta }^{^{\prime} 2}{\alpha }^{^{\prime} 2})/{\alpha }^{^{\prime} 2}\) and this is zero or negative if and only if \({\beta }^{^{\prime} 2}\le {\alpha }^{^{\prime} 2}\). Thus \( {\rm{Im}}\sqrt{{\beta }^{2}+{\gamma }^{2}} \le  {\rm{Im}}\ \alpha \) is the necessary and sufficient condition for the existence of a wavenumber satisfying \({\rm{Re}}\sqrt{{({k}_{y}+\alpha )}^{2}{\beta }^{2}{\gamma }^{2}}=0\).
Continuum model of the exceptional edge modes
To construct the continuum model of the exceptional edge modes, we consider the 2â€‰Ã—â€‰2 continuum Hamiltonian H_{cont} analyzed in the previous paper^{62} (see Supplementary Note 7 for details), which describes the lowenergy dispersion of a higherChernnumber insulator (the Chern number Câ€‰=â€‰2). We combine it and its timereversal counterpart \({H}_{{\rm{cont}}}^{* }\) by a nonHermitian coupling. The obtained Hamiltonian is described as below:
where M, Î², a, b, and \(b^{\prime}\) are real parameters. To numerically calculate the band structure, we must discretize the space and modify the derivative in the Hamiltonian into the difference between the neighboring sites. Here, we use central differences to discretize the first and secondorder x derivatives, âˆ‚_{x} and \({\partial }_{x}^{2}\), in the Hamiltonian. We impose the open boundary condition in the x direction. On the contrary, we consider the periodic boundary condition in the y direction and convert the derivative âˆ‚_{y} into ik_{y} to obtain the effective Bloch Hamiltonian. We discretize the space into 50 sites in the x direction and set the discretization step to be unity. As in the tightbinding model, by numerically diagonalizing the obtained Hamiltonian, we calculate the edge band structure for the continuum model. We use the parameters Mâ€‰=â€‰0.5, aâ€‰=â€‰1, bâ€‰=â€‰0.5, bâ€‰=â€‰0.3, and \(b^{\prime} =0.2\) in the numerical calculation.
Derivation of the effective Hamiltonian for twocomponent chiral active matter
We start with the Vicsektype model^{65} with a constant rotational force on each chiral particle. We can derive hydrodynamic equations from such a particle model via the BoltzmannGinzburgLandau approach^{66,67}. First, we derive the Boltzmann equation of our particle model. Then, we consider the Fourier modes of the density function and only keep the leading order terms. The obtained equations can be interpreted as the hydrodynamic equations of active matter considered here. As a further step, we consider fluctuations of the density and the velocity fields from the unordered steady state. By linearizing the hydrodynamic equation with respect to those fluctuations, we finally obtain the linearized eigenequation. The coefficient matrix corresponds to the effective Hamiltonian of our model and is described as,
with H_{0}, A, and C being
Further details are provided in Supplementary Methods.
To numerically calculate the band structure, we must modify the derivative in our effective Hamiltonian into the difference between the neighboring sites in the discretized space as in the continuum model. Here, we use central differences to discretize the first and secondorder x derivatives, âˆ‚_{x} and \({\partial }_{x}^{2}\), in our effective Hamiltonian. We impose the open boundary condition in the x direction. On the contrary, we consider the periodic boundary condition in the y direction and convert the derivative âˆ‚_{y} into ik_{y} to obtain the effective Bloch Hamiltonian. We discretize the space into 50 sites in the x direction and set the discretization step to be unity. As in the tightbinding model, by numerically diagonalizing the obtained Hamiltonian, we calculate the edge band structure of our active matter model. The parameters used in the numerical calculation are Ï‰_{0}â€‰=â€‰1, Î½^{o}â€‰=â€‰0.5, Î³â€‰=â€‰0.3, and Î²â€‰=â€‰0.5.
Data availability
The data that support the plots within this paper and other findings of this study are available from the corresponding author on request.
References
Klitzing, K. V., Dorda, G. & Pepper, M. New method for highaccuracy determination of the finestructure constant based on quantized Hall resistance. Phys. Rev. Lett. 45, 494â€“497 (1980).
Thouless, D. J., Kohmoto, M., Nightingale, M. P. & Nijs, M. D. Quantized Hall conductance in a twodimensional periodic potential. Phys. Rev. Lett. 49, 405â€“408 (1982).
Kane, C. L. & Mele, E. J. Z_{2} topological order and the quantum spin Hall effect. Phys. Rev. Lett. 95, 146802 (2005).
Hazan, M. Z. & Kane, C. L. Colloquium: topological insulators. Rev. Mod. Phys. 82, 3045â€“3067 (2010).
Kitaev, A. Periodic table for topological insulators and superconductors. AIP Conf. Proc. 1134, 22â€“30 (2009).
Ando, Y. Topological insulator materials. J. Phys. Soc. Jpn. 82, 102001 (2013).
Ruschhaupt, A., Delgado, F. & Muga, J. G. Physical realization of PTsymmetric potential scattering in a planar slab waveguide. J. Phys. A 38, L171â€“L176 (2005).
ElGanainy, R., Makris, K. G., Christodoulides, D. N. & Musslimani, Z. H. Theory of coupled optical PTsymmetric structures. Opt. Lett. 32, 2632â€“2634 (2007).
RÃ¼ter, C. E. et al. Observation of paritytime symmetry in optics. Nat. Phys. 6, 192â€“195 (2010).
Zhao, H. et al. NonHermitian topological light steering. Science 365, 1163â€“1166 (2019).
Ashida, Y., Furukawa, S. & Ueda, M. Paritytimesymmetric quantum critical phenomena. Nat. Commun. 8, 15791 (2017).
Li, J. et al. Observation of paritytime symmetry breaking transitions in a dissipative Floquet system of ultracold atoms. Nat. Commun. 10, 855 (2019).
Xu, H., Mason, D., Jiang, L. & Harris, J. G. E. Topological energy transfer in an optomechanical system with exceptional points. Nature 537, 80â€“83 (2016).
Jing, H., Ã–zdemir, Åž. K., LÃ¼, H. & Nori, F. Highorder exceptional points in optomechanics. Sci. Rep. 7, 3386 (2017).
Schindler, J., Li, A., Zheng, M. C., Ellis, F. M. & Kottos, T. Experimental study of active LRC circuits with PT symmetries. Phys. Rev. A 84, 040101 (2011).
Ezawa, M. NonHermitian boundary and interface states in nonreciprocal higherorder topological metals and electrical circuits. Phys. Rev. B 99, 121411 (2019).
Kane, C. L. & Lubensky, T. C. Topological boundary modes in isostatic lattices. Nat. Phys. 10, 39â€“45 (2013).
Nash, L. M. et al. Topological mechanics of gyroscopic metamaterials. Proc. Natl Acad. Sci. USA 112, 14495â€“14500 (2015).
Murugan, A. & Vaikuntanathan, S. Topologically protected modes in nonequilibrium stochastic systems. Nat. Commun. 8, 13881 (2017).
Hu, Y. C. & Hughes, T. L. Absence of topological insulator phases in nonHermitian PTsymmetric Hamiltonians. Phys. Rev. B 84, 153101 (2011).
Esaki, K., Sato, M., Hasebe, K. & Kohmoto, M. Edge states and topological phases in nonHermitian systems. Phys. Rev. B 84, 205128 (2011).
Leykam, D., Bliokh, K. Y., Huang, C., Chong, Y. D. & Nori, F. Edge modes, degeneracies, and topological numbers in nonHermitian systems. Phys. Rev. Lett. 118, 040401 (2017).
Gong, Z. et al. Topological phases of nonHermitian systems. Phys. Rev. X 8, 031079 (2018).
Shen, H., Zhen, B. & Fu, L. Topological and theory for nonHermitian Hamiltonians. Phys. Rev. Lett. 120, 146402 (2018).
Kunst, F. K., Edvardsson, E., Budich, J. C. & Bergholtz, E. J. Biorthogonal bulkboundary correspondence in nonHermitian systems. Phys. Rev. Lett. 121, 026808 (2018).
Xiong, Y. Why does bulk boundary correspondence fail in some nonHermitian topological models. J. Phys. Commun. 2, 035043 (2018).
Yao, S. & Wang, Z. Edge states and topological invariants of nonHermitian systems. Phys. Rev. Lett. 121, 086803 (2018).
Lee, C. H. & Thomale, R. Anatomy of skin modes and topology in nonHermitian systems. Phys. Rev. B 99, 201103 (2019).
Yokomizo, K. & Murakami, S. Nonbloch band theory of nonHermitian systems. Phys. Rev. Lett. 123, 066404 (2019).
Zhou, H. & Lee, J. Y. Periodic table for topological bands with nonHermitian symmetries. Phys. Rev. B 99, 235112 (2019).
Kawabata, K., Shiozaki, K., Ueda, M. & Sato, M. Symmetry and topology in nonHermitian physics. Phys. Rev. X 9, 041015 (2019).
Borgnia, D. S., Kruchkov, A. J. & Slager, R. J. NonHermitian boundary modes and topology. Phys. Rev. Lett. 124, 056802 (2020).
Harari, G. et al. Topological insulator laser: theory. Science 359, 1230 (2018).
Bandres, M. A. et al. Topological insulator laser: experiments. Science 359, 1231 (2018).
Song, A. Y. et al. \({\cal{P}}{\cal{T}}\)symmetric topological edgegain effect. Phys. Rev. Lett. 125, 033603 (2020).
Kato T. Perturbation Theory for Linear Operators (Springer, New York, 1966).
Dembowski, C. et al. Experimental observation of the topological structure of exceptional points. Phys. Rev. Lett. 86, 787â€“790 (2001).
Longhi, S. \({\cal{PT}}\)symmetric laser absorber. Phys. Rev. A 82, 031801 (2010).
Lin, Z. et al. Unidirectional Invisibility Induced by \({\cal{P}}{\cal{T}}\)symmetric periodic structures. Phys. Rev. Lett. 106, 213901 (2011).
Persson, E., Rotter, I., StÃ¶ckmann, H.J. & Barth, M. Observation of resonance trapping in an open microwave cavity. Phys. Rev. Lett. 85, 2478â€“2481 (2000).
Okugawa, R. & Yokoyama, T. Topological exceptional surfaces in nonHermitian systems with paritytime and parityparticlehole symmetries. Phys. Rev. B 99, 041202 (2019).
Budich, J. C., CarlstrÃ¶m, J., Kunst, F. K. & Bergholtz, E. J. Symmetryprotected nodal phases in nonHermitian systems. Phys. Rev. B 99, 041406 (2019).
Kawabata, K., Bessho, T. & Sato, M. Classification of exceptional points and nonHermitian topological semimetals. Phys. Rev. Lett. 123, 066405 (2019).
Matsuura, S., Chang, P. Y., Schnyder, A. P. & Ryu, S. Protected boundary states in gapless topological phases. N. J. Phys. 15, 065001 (2013).
Armitage, N. P., Mele, E. J. & Vishwanath, A. Weyl and Dirac semimetals in threedimensional solids. Rev. Mod. Phys. 90, 015001 (2018).
Marchetti, M. C. et al. Hydrodynamics of soft active matter. Rev. Mod. Phys. 85, 1143â€“1189 (2013).
Souslov, A., van Zuiden, B. C., Bartolo, D. & Vitelli, V. Topological sound in activeliquid metamaterials. Nat. Phys. 13, 1091â€“1094 (2017).
Shankar, S., Bowick, M. J. & Marchetti, M. C. Topological sound and flocking on curved surfaces. Phys. Rev. X 7, 031039 (2017).
Dasbiswas, K., Mandadapu, K. K. & Vaikuntanathan, S. Topological localization in outofequilibrium dissipative systems. Proc. Natl. Acad. Sci. USA 115, 9031â€“9040 (2018).
Souslov, A., Dasbiswas, K., Fruchart, M., Vaikuntanathan, S. & Vitelli, V. Topological waves in fluids with odd viscosity. Phys. Rev. Lett. 122, 128001 (2019).
Sone, K. & Ashida, Y. Anomalous topological active matter. Phys. Rev. Lett. 123, 205502 (2019).
Yang, X., Ren, C., Cheng, K. & Zhang, H. P. Robust boundary flow in chiral active fluid. Phys. Rev. E 101, 022603 (2020).
DiLuzio, W. R. et al. Escherichia coli swim on the righthand side. Nature 435, 1271â€“1274 (2005).
KÃ¼mmel, F. et al. Circular motion of asymmetric selfpropelling particles. Phys. Rev. Lett. 110, 198302 (2013).
FÃ¼rthauer, S., Strempel, M., Grill, S. W. & JÃ¼licher, F. Active chiral fluids. Eur. Phys. J. E 35, 89 (2012).
Banerjee, D., Souslov, A., Abanov, A. G. & Vitelli, V. Odd viscosity in chiral active fluids. Nat. Commun. 8, 1573 (2017).
Ai, B., Shao, Z. & Zhong, W. Mixing and demixing of binary mixtures of polar chiral active particles. Soft Matter 14, 4388â€“4395 (2018).
Bernevig, B. A., Hughes, T. L. & Zhang, S. C. Quantum spin Hall effect and topological phase transition in HgTe quantum wells. Science 314, 1757â€“1761 (2006).
Qi, X. L., Wu, Y. S. & Zhang, S. C. Topological quantization of the spin Hall effect in twodimensional paramagnetic semiconductors. Phys. Rev. B 74, 085308 (2006).
Zhou, H. et al. Observation of bulk Fermi arc and polarization half charge from paired exceptional points. Science 359, 1009â€“1012 (2018).
Mostafazadeh, A. PseudoHermiticity versus PT symmetry: the necessary condition for the reality of the spectrum of a nonHermitian Hamiltonian. J. Math. Phys. 43, 205â€“214 (2002).
Xu, G., Weng, H., Wang, Z., Dai, X. & Fang, Z. Chern semimetal and the quantized anomalous Hall effect in HgCr_{2}Se_{4}. Phys. Rev. Lett. 107, 186806 (2011).
BÃ¤r, M., GroÃŸmann, R., Heidenreich, S. & Peruani, F. Selfpropelled rods: insights and perspectives for active matter. Annu. Rev. Condens. Matter Phys. 11, 441â€“466 (2020).
Toner, J. & Tu, Y. Longrange order in a twodimensional dynamics XY model: how birds fly together. Phys. Rev. Lett. 75, 4326 (1995).
Vicsek, T., CzirÃ³k, A., BenJacob, E., Cohen, I. & Shochet, O. Novel type of phase transition in a system of selfdriven particles. Phys. Rev. Lett. 75, 1226 (1995).
Bertin, E., Droz, M. & GrÃ©goire, G. Boltzmann and hydrodynamic description for selfpropelled particles. Phys. Rev. E 74, 022101 (2006).
Peshkov, A., Bertin, E., Ginelli, F. & ChatÃ©, H. BoltzmannGinzburgLandau approach for continuous descriptions of generic Vicseklike models. Eur. Phys. J. Spec. Top. 223, 1315â€“1344 (2014).
Acknowledgements
We thank Zongping Gong, Kohei Kawabata, Kyogo Kawaguchi, Daiki Nishiguchi, Shun Otsubo, Kazumasa Takeuchi, and Hiroki Yamaguchi for valuable discussions. K.S. is supported by Worldleading Innovative Graduate Study Program for Materials Research, Industry, and Technology (MERITWINGS) of the University of Tokyo. Y.A. is supported by JSPS KAKENHI Grant Numbers JP16J03613 and JP19K23424. T.S. is supported by JSPS KAKENHI grant numbers JP16H02211 and JP19H05796. T.S. is also supported by Institute of AI and Beyond of the University of Tokyo.
Author information
Authors and Affiliations
Contributions
K.S., Y.A., and T.S. planned the project. K.S. performed the analytical and numerical calculations. K.S., Y.A., and T.S. analyzed and interpreted the results and wrote the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Peer review information Nature Communications thanks the anonymous reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Publisherâ€™s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the articleâ€™s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the articleâ€™s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Sone, K., Ashida, Y. & Sagawa, T. Exceptional nonHermitian topological edge mode and its application to active matter. Nat Commun 11, 5745 (2020). https://doi.org/10.1038/s41467020194880
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467020194880
This article is cited by

A topological fluctuation theorem
Nature Communications (2022)

NonHermitian topology in rockâ€“paperâ€“scissors games
Scientific Reports (2022)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.