betaFeedback

WorksPhysics and Astronomy

Floquet Non-Abelian Topological Insulator and Multifold Bulk-Edge Correspondence

Abstract

Topological phases characterized by non-Abelian charges are beyond the scope of the paradigmatic tenfold way and have gained increasing attention recently. Here we investigate topological insulators with multiple tangled gaps in Floquet settings and identify uncharted Floquet non-Abelian topological insulators without any static or Abelian analog. We demonstrate that the bulk-edge correspondence is multifold and follows the multiplication rule of the quaternion group Q8Q_8. The same quaternion charge corresponds to several distinct edge-state configurations that are fully determined by phase-band singularities of the time evolution. In the anomalous non-Abelian phase, edge states appear in all bandgaps despite trivial quaternion charge. Furthermore, we uncover an exotic swap effect – the emergence of interface modes with swapped driving, which is a signature of the non-Abelian dynamics and absent in Floquet Abelian systems. Our work, for the first time, presents Floquet topological insulators characterized by non-Abelian charges and opens up exciting possibilities for exploring the rich and uncharted territory of non-equilibrium topological phases.

The past few decades have witnessed a remarkable surge of research in topological phases of matter [1, 2], culminating in the renowned Altland-Zirnbauer tenfold way [3, 4, 5, 6, 7]. Based on the underlying symmetries and spatial dimensions, gapped bulk Hamiltonians are characterized by Abelian-type topological invariants (Z\mathbb{Z} or Z2\mathbb{Z}_2) with their own manifestation of protected boundary states. Very recently, the notion of band topology has been extended to tangled multi-gap scenarios [8, 9, 10, 11, 12, 13, 14, 15]. For instance, in the presence of space-time inversion (PT) symmetry, one-dimensional (1D) insulators involving multiple bandgaps may carry non-Abelian quaternion charges [8] and host richer topological phases as experimentally observed in transmission line networks [16, 17]. Yet in its infancy, the tangled multi-gap topology plays a vital role in describing, e.g., the disclination defects of nematic liquids [18, 19, 20, 21, 22], the admissible nodal lines [23, 24, 25, 26, 27] and the reciprocal braiding of Dirac/Weyl/exceptional points [11, 28, 29, 30, 31].

Floquet engineering provides a powerful knob in manipulating band structures [32, 33, 34, 35, 36, 37, 38, 39, 40, 41], offering unprecedented control over the topological properties of materials and the exploration of non-equilibrium phenomena. With a time-periodic Hamiltonian H(t)=H(t+T)H(t)=H(t+T) (TT is the driving period), the stroboscopic dynamics is dictated by an effective Floquet Hamiltonian. Notably, Floquet systems exhibit intriguing topological features with no static analog arising from the replicas of quasienergy bands, such as the emergence of anomalous chiral edge modes [42, 43, 44, 45] despite the triviality of all bulk bands. Incorporating the multi-gap scenario, this paper aims to address three fundamental questions regarding Floquet multi-gap topology. (i) Does a Floquet topological insulating phase characterized by non-Abelian charge exist, and if so, how can it be implemented through periodic driving? (ii) What novel bulk-edge correspondence does such a non-Abelian phase possess, and how can it be described? (iii) Are there any unique topological or dynamical phenomena associated with this phase?

Here we answer these questions affirmatively. Firstly, we propose the realization of the simplest Floquet non-Abelian topological insulator (FNATI) in a 1D three-band system with PT symmetry. Secondly, the FNATI is characterized by a quaternion charge, which, on its own, cannot predict the existence or the number of edge states. Moreover, each quaternion charge corresponds to multiple edge-state configurations, demonstrating a multifold bulk-edge correspondence that obeys the multiplication rule of the quaternion group. The full topology or edge-state configuration is completely captured by the phase-band singularities of the time-evolution operator over one driving period. Intriguingly, we identify an anomalous FNATI hosting edge modes inside all bandgaps with a trivial bulk quaternion charge. Thirdly, we reveal the emergence of interface modes with swapped driving sequences as a genuine non-Abelian effect. It indicates the non-commutative nature of the FNATI. This is in sharp contrast to Floquet Abelian topological insulators, where such interface modes are absent due to the same spectral structures regardless of the choice of time frame. We emphasize that the intriguing properties of FNATI stem from the presence of multiple tangled bandgaps. Our findings expand the scope of Floquet topological insulators into the non-Abelian regime and open up new avenues for investigating the vast and unexplored territory of non-equilibrium topological phases.

Results

Multi-gap topology and driving protocol. Let us recap the static three-band topological insulator characterized by the quaternion charge Q8Q_8 [8]. In the presence of PT symmetry, the Hamiltonian becomes real-valued in momentum space H(k)=H∗(k)H(k)=H^*(k) when expressed on a suitable basis. Consequently, the eigenstates represent three real vectors that are orthonormal to each other and span a coordinate frame, as sketched in Figure 1a. When simultaneously considering both bandgaps, the configuration space of the Hamiltonian is M3=O(3)O(1)3M_3 = \frac{O(3)}{O(1)^3}, where O(N)O(N) denotes the orthogonal group of NN dimension. The O(1)=Z2O(1)=\mathbb{Z}_2 factor comes from the gauge freedom (±1\pm1) for each real eigenstate. With the variation of momentum kk from −π-\pi to π\pi, the eigenstate frame rotates on the unit sphere. The mapping from the 1D Brillouin zone k∈[−π,π]=S1k \in[-\pi,\pi]=S^1 to M3M_3 is governed by the fundamental group of M3M_3, which describes the frame rotation. This fundamental group is given by the quaternion group π1(M3)=Q8\pi_1(M_3)=Q_8 [8]. As a non-Abelian group, Q8Q_8 has eight elements and five conjugacy classes {1,±i,±j,±k,−1}\{1,\pm i,\pm j,\pm k,-1\} with multiplication rule i2=j2=k2=ijk=−1i^2=j^2=k^2=ijk=-1, ij=−jiij=-ji, ik=−kiik=-ki, jk=−kjjk=-kj. The quaternion charge captures the multi-gap band topology and governs the number of edge states in both bandgaps [16]. The trivial phase with charge q=1q=1 has no edge states under open boundaries.

Figure 1. Multi-gap topology in Floquet systems. a Sketch of the frame rotation of eigenstates for topological insulator characterized by quaternion charge. The loci of eigenstate on the unit sphere with varying momentum kk from −π-\pi to π\pi are marked in red/green/blue for the first/second/third band, respectively. b Building blocks for our Floquet driving. H1H_1 (H2H_2) contains only intracell (intercell) coupling terms. c Labeling of the quasienergy bandgaps. The third gap straddles the FBZ edge at ±πT\pm\frac{\pi}{T}. d,e Quasienergy spectra and spatial distributions (represented by color shades) of their eigenstates for quaternion charge q=jq=j with open boundaries. In d, the edge states emerge at both the first and second gaps. In e the edge states emerge at the third gap. The lattice length is L=50L=50. The parameters are listed in Methods.

We consider a 1D lattice with three sites, denoted as AA, BB, CC per unit cell. Our Floquet driving is based on two ingredients, namely H1H_1 and H2H_2 as depicted in Figure 1b. H1H_1 (H2H_2) contains only intracell (intercell) couplings and respects PT symmetry,

H1=∑n=1L∑X,YsXYcX,n†cY,n+h.c.,H2=∑n=1L−1∑X,YvXYcX,n†cY,n+1+h.c.(1)\begin{aligned} H_1 &= \sum_{n=1}^{L}\sum_{X,Y}s_{XY}c^\dagger_{X,n}c_{Y,n}+\mathrm{h.c.},\\ H_2 &= \sum_{n=1}^{L-1}\sum_{X,Y}v_{XY}c^\dagger_{X,n}c_{Y,n+1}+\mathrm{h.c.} \tag*{(1)} \end{aligned}

Here cX,n†c^\dagger_{X,n} and cX,nc_{X,n} represent the creation and annihilation operators at site XX (X=A,B,CX=A,B,C) of the nn-th unit cell, respectively. The lattice length is LL. The coupling parameters sXYs_{XY} and vXYv_{XY} used in this paper are listed in the Methods. Without loss of generality, the driving period is set as T=1T=1. We adopt a symmetric driving protocol: H(t)=H1H(t)=H_1 for t∈mT+[0,T/4]∪[3T/4,T]t\in mT+[0,T/4]\cup[3T/4,T] and H(t)=H2H(t)=H_2 for t∈mT+[T/4,3T/4]t\in mT+[T/4,3T/4] (m∈Zm\in\mathbb{Z}). The dynamics of the system is governed by the time evolution operator U(t)=Te−i∫0tH(τ)dτU(t)=\mathcal{T}e^{-i\int_0^tH(\tau)d\tau}, where T\mathcal{T} means time-ordering. The stroboscopic evolution of the system is described by the Floquet operator,

U(T)=e−iH1T/4e−iH2T/2e−iH1T/4.(2)U(T)=e^{-iH_1T/4}e^{-iH_2T/2}e^{-iH_1T/4}. \tag*{(2)}

The effective Floquet Hamiltonian HFH_F is defined through U(T)=e−iHFTU(T)=e^{-iH_FT}. Due to the symmetric driving, the Floquet Hamiltonian respects PT symmetry, i.e., HF(k)=HF∗(k)H_F(k)=H_F^*(k). After the diagonalization: HF∣un⟩=ϵn∣un⟩H_F|u_n\rangle=\epsilon_n|u_n\rangle (nn is the band index), we obtain the quasienergy ϵn\epsilon_n and eigenstates ∣un⟩|u_n\rangle. The quasienergies are well-defined modulo 2π/T2\pi/T, and form the quasienergy bands. In the following, we set ϵn\epsilon_n to be in the first Brillouin zone (FBZ) of quasienergy ϵn∈(−π/T,π/T]\epsilon_n\in(-\pi/T,\pi/T].

Unlike the static three-band model with two bandgaps, the Floquet system exhibits an additional bandgap that spans across the FBZ boundary at ±π/T\pm\pi/T [See Fig. 1c], which arises from the replica of Floquet spectra. As a consequence, the multi-gap topology is greatly enriched in Floquet settings. The closing and reopening of this bandgap lead to the emergence of extra edge modes under open boundary conditions. For ease of reference, we denote the band from bottom to top as the first, second, and third bands, respectively, while bearing in mind their replicated nature. The gap between them is then labeled as the first, second, and third bandgap. In Figs. 1d, e, we present the quasienergy spectra and their spatial profiles (of their associated eigenstates) with open boundaries. In both cases, the bulk Floquet Hamiltonian carries a quaternion charge of q=jq=j. In Figure 1d, we observe one edge mode (for each end of the lattice) at the first and second bandgaps, respectively, which is similar to the static case [16]. While in Figure 1e, there is only a single edge mode within the third bandgap. Similar scenarios apply to other cases as well. For instance, for charge ii, edge states can exist in both the first and third bandgaps or solely in the second gap. Likewise, for charge kk, the edge state may exist in both the second and third bandgaps or only in the first gap.

Anomalous FNATI. Owing to the additional bandgap at the FBZ edge, the periodically driven system can exhibit an intriguing phase with protected edge modes even when the bulk quaternion invariant is trivial. This is in stark contrast with the static systems [16], wherein the charge q=1q=1 system does not possess any nontrivial edge states. Note that a similar anomaly also happens on 2D driven lattice with protected chiral edge states despite the triviality of bulk Chern bands [42]. We thus dub this phase anomalous FNATI. In Figure 2a, we plot the quasienergies and the spatial profiles of their associated eigenstates for this anomalous phase. We can clearly observe the existence of edge states (one for each end of the lattice) in all three bandgaps.

FNATI with anomalous edge states

Figure 2. FNATI with anomalous edge states. a Quasienergy spectra and the spatial distributions of their eigenstates with open boundaries. The bulk quaternion charge is q=1q=1. The lattice length is L=50L=50. b Phase bands of U~(k,t)\tilde{U}(k,t) in the 2D momentum-time space. The Dirac-point singularities are marked with red dots. The inset: locations of the Dirac points with quaternion charges j~\tilde{j}, −i~-\tilde{i}, and k~\tilde{k} in the (k,t)(k,t)-space. PP is the base point for the encircling paths. The quaternion charge of the bulk Floquet Hamiltonian and Dirac points are related via 1=k~⋅(−i~)⋅j~1=\tilde{k}\cdot(-\tilde{i})\cdot\tilde{j}. The parameters are listed in Methods.

In the anomalous phase, the bulk invariant of the Floquet Hamiltonian, regardless of whether it is the quaternion charge or Berry phase of Abelian type introduced for static systems [16], is unable to view the edge-state landscape. In fact, the emergence of edge state has dynamical origin and the complete information of the dynamical topology is encoded in the full-time evolution operator. A unified analysis of the bulk-edge correspondence can be achieved through the introduction of phase band and its associated momentum-time singularities [41, 46, 47, 48]. It should be noted that the time-evolution operator U(t)U(t) is not always PT symmetric. To this end, we adopt a PT symmetric operator U~(k,t)\tilde{U}(k,t) via a smooth deformation of U(t)U(t) [49]. U~(k,t)\tilde{U}(k,t) preserves the phase-band structure (including its singularities) and satisfies U~(k,t)=U(k,t)\tilde{U}(k,t)=U(k,t) at t=0,Tt=0,T. Formally, the operator U~(k,t)\tilde{U}(k,t) can be expressed in terms of the spectral decomposition,

U~(k,t)=∑n=13e−iϕn(k,t)∣ψn(k,t)⟩⟨ψn(k,t)∣(3)\tilde{U}(k,t)=\sum_{n=1}^{3}e^{-i\phi_n(k,t)}|\psi_n(k,t)\rangle\langle\psi_n(k,t)| \tag*{(3)}

where e−iϕn(k,t)e^{-i\phi_n(k,t)} denotes the eigenvalues of U~(k,t)\tilde{U}(k,t) and ϕn(k,t)∈(−π/T,π/T]\phi_n(k,t)\in(-\pi/T,\pi/T] forms the phase bands in the 2D momentum-time space.

The phase band may touch and reopen with the variation of time, leaving degenerate Dirac points in the bandgaps in the 2D momentum-time space. At t=Tt=T, the phase bands become the quasienergy bands. As shown in Figure 2b, the phase bands touch once in each gap for the anomalous phase. The presence of Dirac singularities in the phase band leads to the emergence of edge states in the corresponding gap. As the three gaps are tangled, each Dirac point can be assigned a quaternion charge through the Wilson loop along an enclosing path nearby [49]. The quaternion charge qq of the Floquet Hamiltonian HFH_F is related to the quaternion charge of the phase-band singularities via

q=∏mq~m.(4)q=\prod_m \tilde{q}_m. \tag*{(4)}

Here mm labels the Dirac point, and the multiplication is ordered such that the concatenation of the small enclosing paths coincide with the orientation of momentum integral −π→π-\pi\to\pi. As a distinction, the charge of the Dirac singularity is marked with a tilde. In the anomalous FNATI depicted in Figure 2b, the three singularities from right to left possess quaternion charge q~1=k~\tilde{q}_1=\tilde{k}, q~2=−i~\tilde{q}_2=-\tilde{i}, and q~3=j~\tilde{q}_3=\tilde{j}, respectively [See the inset]. It is evident that the relation 1=k~⋅(−i~)⋅j~1=\tilde{k}\cdot(-\tilde{i})\cdot\tilde{j} holds.

Multifold bulk-edge correspondence. The phase-band singularities and their non-Abelian nature offer a straightforward interpretation of the bulk-edge correspondence in FNATI. As indicated by Eq. (4), various patterns of edge states may corresponds to the same quaternion charge of the Floquet Hamiltonian. Such multifold bulk-edge correspondence follows the multiplication rule of the quaternion group Q8Q_8. For example, in Figure 1d,Figure 1e, the two phases with the same quaternion charge q=jq=j have distinct edge-state patterns. In their phase bands, there are two Dirac singularities with charge k~\tilde{k} and i~\tilde{i} for the former, while there is a single singularity with charge j~\tilde{j} for the latter. They satisfy the multiplication rule k~⋅i~=j~\tilde{k}\cdot\tilde{i}=\tilde{j}. Similar discussions can be applied to other cases. The multifold bulk-edge correspondence can be deduced from two key observations. Firstly, the edge state within each bandgap is determined by the phase-band closings that occur within that bandgap during time evolution. Each gap closing results in a change of the mass term and the emergence of an in-gap mode by Jackiw-Rebbi’s argument [50]. With multiple bandgaps present, the edge-state patterns are determined by considering the phase-band singularities within each bandgap. Secondly, the phase-band singularities relate to the bulk quaternion charge via Eq. (4). It then becomes evident that the edge-state pattern is linked to the quaternion charge, and the bulk-edge correspondence showcases a multifold nature that adheres to the multiplication rule.

In Figure 3, we present a comprehensive list of bandgap closings and their corresponding edge-state patterns (per edge) for all quaternion charges. Notably, we observe similar bandgap closings and edge-state configurations for quaternion charges that belong to the same conjugacy class, such as ±i\pm i. This is due to the anti-commutating relations of quaternion charge. For example, the existence of a type j~\tilde{j} and type k~\tilde{k} singularity in the phase band (regardless of their ordering) results in two edge states (located at the first and third bandgap). Switching them yields conjugate quaternion charges: j~⋅k~=−k~⋅j~\tilde{j}\cdot\tilde{k}=-\tilde{k}\cdot\tilde{j}. In addition to the patterns that appeared in the static case (the fourth column) [16], Floquet systems can exhibit unique edge-state patterns (the fifth column) because of the additional phase-band touchings at the FBZ edge. It should be noted that for a particular gap, multiple touchings can exist [49], such as i=k~⋅(−k~)⋅i~i=\tilde{k}\cdot(-\tilde{k})\cdot\tilde{i}, which has two Dirac points in the first bandgap. However, such cases can be reduced to the simpler cases listed in Figure 3 by eliminating two Dirac points of the same gap pairwise, resulting in a factor ±1\pm1. Furthermore, in the case of charge q=−1q=-1, two fickle edge states [16] (with an unspecified bandgap) may appear according to −1=i~2=j~2=k~2-1=\tilde{i}^{2}=\tilde{j}^{2}=\tilde{k}^{2}, or three separate edge states may appear according to −1=i~⋅j~⋅k~-1=\tilde{i}\cdot\tilde{j}\cdot\tilde{k}. We note that the multifold bulk-edge correspondence can be

Figure 3. Multifold bulk-edge correspondence for FNATI. The first column lists the quaternion charge for the Floquet Hamiltonian, and the five conjugacy classes of Q8Q_8 correspond to different rows. The second and third columns sketch the phase band structure with singularities depicted by red dots. The cases with band crossing at the FBZ edge are listed in the third column. The fourth/fifth columns sketches the edge-state patterns corresponding to the phase bands in the second/third column. The black dots represent edge states. Empty circles mark the fickle edge states. The configurations unique to Floquet systems are listed in fifth column. The list applies equally to the domain-wall problem.

extended directly to the domain-wall problem. For two Floquet systems with charge qLq_L and qRq_R, the patterns of topological domain-wall states between them are dictated by the quotient qL/qRq_L/q_R. Different from the static case, the bandgap closings across the domain wall are fully captured by the phase-band singularities, yielding a multifold bulk-domain-wall correspondence [49]. Last, our numerical analysis indicates that the edge/domain-wall states are robust against small disorders. In the presence of domain-wall decorations, it is possible for additional trivial bound states to emerge. However, the non-trivial domain states, governed by the non-Abelian topology, remain unaffected [49].

Interface modes induced by swapped driving. Besides the multifold bulk-edge correspondence, here we uncover a counter-intuitive phenomenon of FNATI as a manifestation of the Floquet non-Abelian topology. As illustrated in Figure 4a, we consider a system with two sides featuring swapped (mismatched) driving sequences. Within one full period, the Hamiltonians on the left and right sides are given by H1→H2→H1H_1 \to H_2 \to H_1 and H2→H1→H2H_2 \to H_1 \to H_2, respectively. The bulk Floquet operators for these two sides, denoted as ULU_L and URU_R, are related via a similarity transformation UL=V−1URVU_L = V^{-1}U_RV where V=e−iH2T/4e−iH1T/4V = e^{-iH_2T/4}e^{-iH_1T/4} accounts for the time shift. This shift does not alter the quasienergy spectra irrespective of the boundary conditions. In Floquet Abelian topological phases [39, 40], the bulk topological invariant is defined for each individual quasienergy bandgap (or branch-cut). The two subsystems should have the same topological invariant for each bandgap to match with the edge states, and no stable interface modes are expected to exist [49]. However, in our non-Abelian system with multiple intertwined gaps, this is not always true. In Figure 4b, we plot the quasienergy spectra in the presence of an interface of swapped driving using the same parameters as Figure 2. We can clearly observe a pair of domain-wall states (located near ±πT\pm\frac{\pi}{T}) in the third gap. The emergence of the interface modes signifies the noncommutativity of the two driving sequences and it is a genuine non-Abelian effect of the Floquet dynamics. We have also verified the appearance of domain-wall states for other cases with different quaternion charges.

Interface modes with swapped driving

Figure 4. Interface modes with swapped driving as a manifestation of the non-Abelian topology. a The Floquet setting: two subsystems connected by an interface have swapped driving sequences. b Quasienergy spectra and spatial profiles of their eigenstates with open boundaries at the two ends. A pair of domain-wall states emerge at the third bandgap. The lattice length is L=51L=51, and the interface is located at n=26n=26. Other parameters are the same as Figure 2. c Trajectory of the transformation QQ (blue curve) in the parameter space of SO(3)SO(3) (represented by a solid ball). As kk varies from −π-\pi to π\pi, QQ follows a closed nontrivial loop that connects a pair of antipodal points in SO(3)SO(3). For comparison, the black curve represents a trivial loop in SO(3)SO(3).

To understand this fascinating effect, we define the bulk eigenstates of ULU_L and URU_R on two sides as SL=(∣u1L⟩,∣u2L⟩,∣u3L⟩)S_L = (|u_{1L}\rangle, |u_{2L}\rangle, |u_{3L}\rangle) and SR=(∣u1R⟩,∣u2R⟩,∣u3R⟩)S_R = (|u_{1R}\rangle, |u_{2R}\rangle, |u_{3R}\rangle). With PTPT symmetry, they are related by an SO(3)SO(3) transformation SL=OSRS_L = OS_R, with O=SLSRTO = S_LS_R^{T} (TT denotes transposition). It is important to note that OO may alter the quaternion charge. In particular, when OO corresponds to the non-identity element in the fundamental group π1(SO(3))=Z2\pi_1(SO(3)) = \mathbb{Z}_2, the bulk quaternion charge for the two sides satisfy qL=−qRq_L = -q_R. In fact, according to the exact sequence of homotopy groups [51]:

⋯→π1(SO(3))→j1π1(SO(3)D2)→∂1π0(D2)→i00,(5)\cdots\to\pi_1(SO(3)) \xrightarrow{j_1} \pi_1\left(\frac{SO(3)}{D_2}\right) \xrightarrow{\partial_1} \pi_0(D_2) \xrightarrow{i_0} 0, \tag*{(5)}

the kernel of ∂1\partial_1-mapping is {1,−1}\{1,-1\} in π1(SO(3)D2)=Q8\pi_1\left(\frac{SO(3)}{D_2}\right) = Q_8. Thus j1j_1 maps the non-identity element of Z2\mathbb{Z}_2 to −1-1 in Q8Q_8. For our case, OO traces a nontrivial path in SO(3)SO(3) as the momentum kk varies from −π-\pi to π\pi [See Figure 4c]. ULU_L has a quaternion charge of qL=1q_L = 1, while URU_R has qR=−1q_R = -1, indicating the emergence of domain-wall states. Alternatively, one can scrutinize this effect through the phase-band singularities [49]. With time variation, the phase band undergoes crossings with charge k~,−i~,j~\tilde{k},-\tilde{i},\tilde{j} and −k~,i~,−j~-\tilde{k},\tilde{i},-\tilde{j} for ULU_L and URU_R, respectively. They satisfy Eq. (4), qL=1=k~⋅(−i~)⋅j~q_L = 1 = \tilde{k}\cdot(-\tilde{i})\cdot\tilde{j} and qR=−1=(−k~)⋅i~⋅(−j~)q_R = -1 = (-\tilde{k})\cdot\tilde{i}\cdot(-\tilde{j}).

Discussion

In summary, our exploration of the FNATI represents a significant advancement in the field of out-of-equilibrium topological phases. We have demonstrated the implementation of FNATI through step-like driving and fully characterized its topological properties via phase-band singularities. Our analysis revealed a multifold bulk-edge correspondence, governed by the multiplication rule of the quaternion group Q8Q_8. We have identified an anomalous phase that possesses edge modes within all bandgaps despite a trivial bulk, which lacks a static analog. Furthermore, we have uncovered a novel interface effect resulting from swapped driving that serves as a key signature of the non-Abelian topology.

Our findings offer novel insights into Floquet topological phases with multiple intertwined bandgaps. Our results can be extended to higher-band cases, such as a four-band topological insulator characterized by the non-Abelian group Q16Q_{16} [17]. In higher dimensions, the phase-band singularities may extend to nodal lines, and it would be intriguing to explore their interwinding in momentum-time space and the associated non-Abelian effects. From a wider perspective, the multi-gap topology belongs to the fragile topology and relies on the partition of energy bands [10]. A different partition (2+1)(2+1) yields a different flag manifold RP2\mathbb{R}P^2 and is relevant for the Floquet Euler phase [14]. Other tantalizing extensions include the 3D nodal-line metals [8] characterized by non-Abelian frame charges and non-Hermitian phases with non-Abelian band braidings [52, 53, 54] by relaxing the PT symmetry to allow for a complex flag manifold. A natural issue is how these non-Abelian features interplay with Floquet driving. Besides the step-like driving utilized for illustration purposes, smooth Floquet driving should also be suitable for implementing FNATI. With the high feasibility of Floquet engineering [32, 33, 34, 35, 36, 37, 38, 39, 40, 41] and the ability to manipulate tight-binding models in various platforms, such as ultracold atoms [45, 55, 56, 57, 58], photonic or acoustic materials [43, 44, 59, 60], we anticipate that the uncharted FNATI and non-Abelian effect will be observed in near-future experiments.

Methods

Calculation of quaternion charge

Calculation of quaternion charge. The quaternion charge q∈Q8q \in Q_8 characterizes the rotations of the eigen-state frame as the momentum kk varies from −π-\pi to π\pi. In the case of the Dirac singularity in the phase band, the quaternion charge describes the frame rotation along the enclosing path. According to Ref. [8], the generalized Wilson operator can be used to obtain the quaternion charge. Formally, the Wilson loop along a closed path Γ\Gamma is defined as follows:

WΓ=Pe∮ΓAall(k)⋅dk.(6)W_\Gamma= \mathcal{P}e^{\oint_\Gamma A_{\mathrm{all}}(k)\cdot dk}. \tag*{(6)}

Here, [Aall(k)]mn=⟨um(k)∣∂k∣un(k)⟩[A_{\mathrm{all}}(k)]_{mn} = \langle u_m(k)|\partial_k|u_n(k)\rangle represents the Berry-Wilczek-Zee connection. The band indices mm and nn take values from 1 to 3, and ∣un(k)⟩|u_n(k)\rangle is the eigenstate of the Floquet Hamiltonian HFH_F. For the Dirac-point case, we use the phase band instead. Aall(k)A_{\mathrm{all}}(k) is anti-symmetric and can be decomposed into the so(3)\mathfrak{so}(3) Lie-algebra basis:

Aall(k)=∑i=1,2,3βiLi,(7)A_{\mathrm{all}}(k) = \sum_{i=1,2,3}\beta_iL_i, \tag*{(7)}

where (Li)jk=−ϵijk(L_i)_{jk} = -\epsilon_{ijk} and ϵijk\epsilon_{ijk} is the anti-symmetric tensor. We then lift AallA_{\mathrm{all}} to the spin(3)\mathfrak{spin}(3)-valued 1-form [8] by replacing LiL_i with tit_i, where ti=−i2σit_i = -\frac{i}{2}\sigma_i and σi\sigma_i represents the Pauli matrix. This gives us:

A‾all(k)=∑i=1,2,3βiti.(8)\overline{A}_{\mathrm{all}}(k)=\sum_{i=1,2,3}\beta_i t_i. \tag*{(8)}

Finally, the non-Abelian charge is defined by:

q=Pe∮ΓA‾all(k)⋅dk.(9)q=\mathcal{P}e^{\oint_{\Gamma}\overline{A}_{\mathrm{all}}(k)\cdot dk}. \tag*{(9)}

The elements of the quaternion group are represented by 1→σ01\to\sigma_0, i→−iσxi\to-i\sigma_x, j→−iσyj\to-i\sigma_y, and k→−iσzk\to-i\sigma_z.

Smooth deformation of time evolution. To define the quaternion charge of the Dirac singularity, real wave functions of the phase bands are required. To this end, we smoothly deform U(k,t)U(k,t) into U~(k,t)\widetilde{U}(k,t) such that U~∗(k,t)U~(k,t)=1\widetilde{U}^{*}(k,t)\widetilde{U}(k,t)=1 for all t∈[0,1]t\in[0,1] (we set T=1T=1 for convenience). The time evolution operator for our driving protocol is

U={e−iH1t,t∈[0,1/4],e−iH2(t−1/4)e−iH1/4,t∈[1/4,3/4],e−iH1(t−3/4)e−iH2/2e−iH1/4,t∈[3/4,1].(10)U=\begin{cases} e^{-iH_1t}, & t\in[0,1/4],\\ e^{-iH_2(t-1/4)}e^{-iH_1/4}, & t\in[1/4,3/4],\\ e^{-iH_1(t-3/4)}e^{-iH_2/2}e^{-iH_1/4}, & t\in[3/4,1]. \end{cases} \tag*{(10)}

We can define the PT symmetric operator U~(k,t)\widetilde{U}(k,t) as

U~={e−iH1t,t∈[0,1/4],e−iH1/8e−iH2(t−1/4)e−iH1/8,t∈[1/4,3/4],e−i2(t−12)H1e−iH2/2e−i2(t−12)H1,t∈[3/4,1].(11)\widetilde{U}=\begin{cases} e^{-iH_1t}, & t\in[0,1/4],\\ e^{-iH_1/8}e^{-iH_2(t-1/4)}e^{-iH_1/8}, & t\in[1/4,3/4],\\ e^{-\frac{i}{2}(t-\frac{1}{2})H_1}e^{-iH_2/2}e^{-\frac{i}{2}(t-\frac{1}{2})H_1}, & t\in[3/4,1]. \end{cases} \tag*{(11)}

To visualize the smoothness of the deformation, let us consider the continuous interpolation between them:

U(s,k,t)={e−iH1t,t∈[0,1/4],e−iH11−s8e−iH2(t−1/4)e−iH1(1−s2)14,t∈[1/4,3/4],e−i[(1−s2)t+2s−34]H1e−iH2/2e−i(s2t+2s−34)H1,t∈[3/4,1],(12)U(s,k,t)=\begin{cases} e^{-iH_1t}, & t\in[0,1/4],\\ e^{-iH_1\frac{1-s}{8}}e^{-iH_2(t-1/4)}e^{-iH_1(1-\frac{s}{2})\frac{1}{4}}, & t\in[1/4,3/4],\\ e^{-i[(1-\frac{s}{2})t+\frac{2s-3}{4}]H_1}e^{-iH_2/2}e^{-i(\frac{s}{2}t+\frac{2s-3}{4})H_1}, & t\in[3/4,1], \end{cases} \tag*{(12)}

such that U(s=0,k,t)=U(k,t)U(s=0,k,t)=U(k,t) and U(s=1,k,t)=U~(k,t)U(s=1,k,t)=\widetilde{U}(k,t). During the deformation from s=0s=0 to s=1s=1, the phase bands and the Dirac singularities are unchanged.

Coefficients of the tight binding model. The coefficients used in the paper are listed in Table I. In all the cases, we take sAA=sBB=sCC=0s_{AA}=s_{BB}=s_{CC}=0, sAB=sBA=rs_{AB}=s_{BA}=r, sBC=sCB=ss_{BC}=s_{CB}=s, sCA=sAC=ts_{CA}=s_{AC}=t and vAB=vBA=iuv_{AB}=v_{BA}=iu, vBC=vCB=ivv_{BC}=v_{CB}=iv, vCA=vAC=iwv_{CA}=v_{AC}=iw, where rr, ss, tt, uu, vv and ww are all real numbers.

qqrrssttuuvvwwvAAv_{AA}vBBv_{BB}vCCv_{CC}
jj [Fig. 1d]1122000-22
jj [Fig. 1e]104300-101
1 [Fig. 2a]1030-3-2002
ii [Fig. S2a]101011-101
-1 [Fig. S2b]101-1010-22
jj [Fig. 4a(left)]104300-101
jj [Fig. S4a(right)]1122000-22
jj [Fig. S4b(right)]1041.401.4-202
ii [Fig. S5–S7]102011-202
kk [Fig. S5–S7]111-110-111

Table I. TABLE I. Coefficients used in the paper.

Data availability

All data is available upon reasonable request.

Code availability

The code that support the findings of this study are available at https://doi.org/10.5281/zenodo.8294074.

References

Acknowledgments

This work is supported by the National Key Research and Development Program of China (Grant No. 2022YFA1405800) and the start-up grant of IOP-CAS. T. L. also acknowledges the support from the Project Funded by China Postdoctoral Science Foundation (Grant No. 2023M733719).

Author contributions

H.H. conceived the main idea, and performed the theoretical analysis with T. L. T. L. did the numerical calculations. Both authors contribute to the writing of the paper.

Competing interests

The authors declare no competing interests.

Tables

Supplemental Material for “Floquet Non-Abelian Topological Insulator and Multifold Bulk-Edge Correspondence”

This supplemental material provides additional details on the multiplication rule of Dirac singularities in the phase band, examples with multiple Dirac touchings in the same gap, the phase-band picture of the interface effect, the quotient relation in domain-wall problem, the Zak phase of quasienergy bands, and the stability of edge/domain-wall states.

Multiplication rule of Dirac singularities

In the main text, we have established a relationship between the quaternion charge qq of the Floquet Hamiltonian and the charges of the Dirac singularities. This is expressed through the equation:

q=∏mq~m(13)q = \prod_m \tilde{q}_m \tag*{(13)}

In the equation, we stipulate that the multiplication is from left to right. Since the quaternion group Q8Q_8 is non-Abelian, the ordering in the product is important. We use Figure S1 to illustrate our choice of ordering in the 2D momentum-time space. Firstly, all the enclosing paths of the Dirac points share the same starting point PP. Secondly, the composition of these paths, i.e., the concatenation of paths in the fundamental group, should be smoothly deformed to the boundaries of the 2D (k,t)(k,t) space clockwise. For example, in Figure S1, the path concatenation associated with the multiplication q~1⋅q~2⋅…⋅q~n\tilde{q}_1 \cdot\tilde{q}_2 \cdot\ldots\cdot\tilde{q}_n is equivalent to circling around the boundaries of the 2D (k,t)(k,t) space clockwise. On the other hand, the path concatenation associated with the multiplication q~n⋅q~n−1⋅…⋅q~2⋅q~1\tilde{q}_n \cdot\tilde{q}_{n-1} \cdot\ldots\cdot\tilde{q}_2 \cdot\tilde{q}_1 is equivalent to circling around the boundaries counterclockwise. Notably, the former coincides with the integral taken from −π-\pi to π\pi in calculating the quaternion charge of the Floquet Hamiltonian.

Figure S1. Path concatenation and ordering of Dirac singularities. The quaternion charge qq of the Floquet Hamiltonian is related to the charges of the singularities through q=q~1⋅q~2⋅…⋅q~nq = \tilde{q}_1 \cdot\tilde{q}_2 \cdot\ldots\cdot\tilde{q}_n.

Examples with multiple phase-band touchings in the same gap

In the main text, we have listed all the phase-band patterns with at most one touching in each gap in Figure 3. In principle, multiple touchings within the same gap are possible, as demonstrated by two such cases in Figure S2 and b. Specifically, in Figure S2, the quaternion charge of the bulk Floquet Hamiltonian and the Dirac points are related through i=−k~⋅k~⋅i~i = -\tilde{k} \cdot\tilde{k} \cdot\tilde{i}, while in Figure S2, they are related through −1=(−i~)⋅k~⋅(−i~)⋅k~-1 = (-\tilde{i})\cdot\tilde{k}\cdot(-\tilde{i})\cdot\tilde{k}. Furthermore, we have plotted the corresponding quasienergy spectra and spatial distributions of eigenstates under open boundary conditions in Figure S2 and Figure S2. For the former, we observe the existence of a single edge state in the second gap. For the latter, we observe a pair of edge states for the first and second gap, respectively.

Phase bands with multiple touchings in the same gap

Figure S2. Phase bands with multiple touchings in the same gap. a Phase bands with bulk quaternion charge q=iq=i for the Floquet Hamiltonian. It is related to the charges of the Dirac points through i=−k~⋅k~⋅i~i=-\tilde{k}\cdot\tilde{k}\cdot\tilde{i}. b Phase bands with bulk quaternion charge q=−1q=-1 for the Floquet Hamiltonian. It is related to the charges of the Dirac points through −1=(−i~)⋅k~⋅(−i~)⋅k~-1=(-\tilde{i})\cdot\tilde{k}\cdot(-\tilde{i})\cdot\tilde{k}. c, d Quasienergy spectra and the spatial distributions of eigenstates with open boundaries corresponding to the phase bands in a, b, respectively. The lattice length is L=50L=50. The parameters are listed in Methods.

Interface effect from the phase-band picture

An alternative explanation of the emergence of interface modes is from the phase-band picture. The bulk Floquet operators on the left and right sides are UL=e−iH1T/4e−iH2T/2e−iH1T/4U_L=e^{-iH_1T/4}e^{-iH_2T/2}e^{-iH_1T/4} and UR=e−iH2T/4e−iH1T/2e−iH2T/4U_R=e^{-iH_2T/4}e^{-iH_1T/2}e^{-iH_2T/4}, respectively. We denote their bulk quaternion charges as qLq_L and qRq_R. Let us formally consider the two PT symmetric operators as below:

U~(1)(k,t)={e−iH1t,t∈[0,1/2],e−iH1/4e−iH2(t−1/2)e−iH1/4,t∈[1/2,1],(14)\widetilde{U}^{(1)}(k,t)=\begin{cases} e^{-iH_1t}, & t\in[0,1/2], \\ e^{-iH_1/4}e^{-iH_2(t-1/2)}e^{-iH_1/4}, & t\in[1/2,1], \end{cases} \tag*{(14)}
U~(2)(k,t)={e−iH1t,t∈[0,1/2],e−iH2(t−1/2)/2e−iH1/2e−iH2(t−1/2)/2,t∈[1/2,1].(15)\widetilde{U}^{(2)}(k,t)=\begin{cases} e^{-iH_1t}, & t\in[0,1/2], \\ e^{-iH_2(t-1/2)/2}e^{-iH_1/2}e^{-iH_2(t-1/2)/2}, & t\in[1/2,1]. \end{cases} \tag*{(15)}

U~(1)(k,t)\widetilde{U}^{(1)}(k,t) and U~(2)(k,t)\widetilde{U}^{(2)}(k,t) are related by an SO(3)SO(3) transform, and thus have the same phase bands. At t=Tt=T, the phase bands of U~(1)(k,T)\widetilde{U}^{(1)}(k,T) and U~(2)(k,T)\widetilde{U}^{(2)}(k,T) coincide with the bulk quasienergy bands of ULU_L and URU_R, respectively:

U~(1)(k,T)=UL,U~(2)(k,T)=UR.(16)\widetilde{U}^{(1)}(k,T)=U_L,\qquad\widetilde{U}^{(2)}(k,T)=U_R. \tag*{(16)}

Our numerical analysis shows that there are three Dirac singularities with charges k~\tilde{k}, −i~-\tilde{i}, and j~\tilde{j} for U~(1)(k,t)\tilde{U}^{(1)}(k,t) and three Dirac singularities with charges −k~-\tilde{k}, i~\tilde{i}, and −j~-\tilde{j} for U~(2)(k,t)\tilde{U}^{(2)}(k,t). According to Eq. (13), the quaternion charge of the bulk Floquet Hamiltonian is qL=k~⋅(−i~)⋅j~=1q_L=\tilde{k}\cdot(-\tilde{i})\cdot\tilde{j}=1, and qR=−k~⋅i~⋅(−j~)=−1q_R=-\tilde{k}\cdot\tilde{i}\cdot(-\tilde{j})=-1. Therefore, domain-wall states are expected to emerge at the interface.

Quotient relation in the domain-wall problem

In this section, we address the domain-wall problem for Floquet non-Abelian topological insulators (FNATIs). As depicted in Figure S3, let us denote qLq_L and qRq_R as the bulk quaternion charge of the Floquet Hamiltonian on the left and right side, respectively. In the static case, the appearance of domain-wall states between two topologically distinct phases is governed by the quotient relation Δq=qL/qR\Delta q=q_L/q_R. We demonstrate that in the Floquet setting, the quotient relation still holds. However, due to the additional bandgap across the Floquet Brillouin zone edge, the quotient relation implies a multifold bulk-domain-wall correspondence. For specific values of qLq_L and qRq_R, their quotient Δq\Delta q is definite, but the domain-wall states may exhibit different configurations, as summarized in Figure 3 of the main text. In fact, the edge-state patterns listed in Figure 3 can be seen as a special case of the domain-wall problem, where the system interfaces with the vacuum. Considering its multifold nature, determining the domain-wall states requires additional information about the phase-band singularities. These singularities are related to the bulk quaternion charge through qL=q~L1q~L2⋯q_L=\tilde{q}_{L1}\tilde{q}_{L2}\cdots and qR=q~R1q~R2⋯q_R=\tilde{q}_{R1}\tilde{q}_{R2}\cdots, respectively. We proceed to demonstrate the quotient relation using three steps.

Figure S3. Quotient relation of Floquet non-Abelian topological insulators (FNATIs). a Sketch of the domain wall, where the left and right sides correspond to bulk Hamiltonians with quaternion charges qLq_L and qRq_R, respectively. The domain wall is described by the quotient relation Δq=qR−1qL\Delta q=q_R^{-1}q_L. b Geometric visualization of the quaternion charge as a closed path (loop) in the configuration space M3M_3. The black dot is the base point. The change in quaternion charge from qRq_R (blue) to qLq_L (orange) is given by Δq\Delta q (dotted black loop), and the concatenated path qRΔqq_R\Delta q is equivalent to qLq_L. c Relationship between the quaternion charge and the phase-band singularities.

Firstly, between two distinct topological phases, there are bandgap closings at the domain wall, corresponding to the topological phase transition. This holds true for both static and Floquet cases. Importantly, in Floquet systems, two systems being topologically distinct implies different phaseband singularities or edge-state patterns. In Figure 3 of the main text, for the same quaternion charge, there may be multiple distinct phases due to their different phase-band singularities.

Secondly, the details of the bandgap closings, such as their position and times, are encoded in the quotient relation Δq=qR−1qL\Delta q = q_{R}^{-1}q_{L}. (Here qL/qRq_{L}/q_{R} is treated as qR−1qLq_{R}^{-1}q_{L}; the other choice qLqR−1q_{L}q_{R}^{-1} is also fine as they are conjugate). To visualize this, we remind that the quaternion charge is the first homotopy invariant of the configuration space M3=O(3)O(1)3M_{3} = \frac{O(3)}{O(1)^{3}} of PT-symmetric Hamiltonians. Geometrically, the quaternion charge can be represented as a closed path in M3M_{3} space, as depicted in Figure S3. At the domain wall, the quaternion charge undergoes a change from qRq_{R} to qLq_{L} through gap closings. This change is described by an intermediate path (the dotted black loop) denoted as Δq\Delta q. In homotopy language, concatenating the two paths qRΔqq_{R}\Delta q, first following qRq_{R} and then Δq\Delta q, yields the path qLq_{L}: qL=qRΔqq_{L} = q_{R}\Delta q. Thus, we have Δq=qR−1qL\Delta q = q_{R}^{-1}q_{L}. In Floquet systems, the quaternion charge on each side is further related to the phase-band singularities through Eq. (4) in the main text, as sketched in Figure S3. Consequently, we have Δq=qR−1qL=(q~R1q~R2…)−1(q~L1q~L2…)\Delta q = q_{R}^{-1}q_{L} = (\tilde{q}_{R1}\tilde{q}_{R2}\ldots)^{-1}(\tilde{q}_{L1}\tilde{q}_{L2}\ldots).

Thirdly, the gap closings recorded in the Δq\Delta q above give rise to domain-wall states through the Jackiw-Rebbi argument [50]: the number and locations of the gap closings determine the number of edge states in that gap. In other words, whenever there is a gap closing, a domain-wall state appears. By accounting for all gap closings inside Δq\Delta q represented by the phase-band singularities, we can determine the patterns of domain-wall states.

To compare with the static case and demonstrate the multifold nature of Floquet systems, let us consider a domain wall between two Floquet Hamiltonians of the same charge, qL=qR=jq_{L} = q_{R} = j, as shown in Figure S4. The quotient is Δq=1\Delta q = 1. The domain wall between qL=jq_{L} = j (with singularity j~\tilde{j}) and qR=jq_{R} = j (with singularity k~i~\tilde{k}\tilde{i}) possesses three domain-wall states as Δq=(i~k~)j~\Delta q = (\tilde{i}\tilde{k})\tilde{j} [See Fig. S4a]. On the other hand, the domain wall between qL=jq_{L} = j (with singularity j~\tilde{j}) and qR=jq_{R} = j (with singularity j~\tilde{j}) possesses no domain-wall state [See Fig. S4b]. The two different configurations are consistent with the charge 1 class listed in Figure 3 of the main text.

Quasienergy spectra and spatial distributions of the eigenstates in the presence of a domain wall

Figure S4. Quasienergy spectra and spatial distributions of the eigenstates in the presence of a domain wall. The bulk quaternion charge for both sides is qL=qR=jq_{L} = q_{R} = j. a A domain wall between qL=jq_{L} = j (with singularity j~\tilde{j}) and qR=jq_{R} = j (with singularity k~i~\tilde{k}\tilde{i}) possesses three domain-wall states (one in each gap), as per Δq=1=(i~k~)j~\Delta q = 1 = (\tilde{i}\tilde{k})\tilde{j}. b A domain wall between qL=jq_{L} = j (with singularity j~\tilde{j}) and qR=jq_{R} = j (with singularity j~\tilde{j}) does not possess any domain-wall state.

Zak phase of the quasienergy band

Besides the quaternion charge, the Zak phase is another topological invariant that is assigned to each quasienergy band. The Zak phase can take two values, 0 or π\pi. To calculate the Zak phase, we diagonalize the Floquet Hamiltonian HF(k)=S(k)diag⁡(ϵ1,ϵ2,ϵ3)ST(k)H_{F}(k) = S(k)\operatorname{diag}(\epsilon_{1},\epsilon_{2},\epsilon_{3})S^{T}(k). Here, the eigenvectors are organized in the SO(3)SO(3) transformation S(k)=(∣u1⟩,∣u2⟩,∣u3⟩)S(k) = (\lvert u_1\rangle, \lvert u_2\rangle, \lvert u_3\rangle). The Zak phase can be extracted from the evolution of these eigenvectors with respect to the lattice momentum kk. As kk ranges from −π-\pi to π\pi, the eigenvector ∣uj⟩\lvert u_j\rangle (j=1,2,3j = 1, 2, 3) may acquire an additional sign in a continuous manner, ∣uj(k=π)⟩=−∣uj(k=−π)⟩\lvert u_j(k = \pi)\rangle= -\lvert u_j(k = -\pi)\rangle. This corresponds to the Zak phase taking π\pi for the jj-th band. Otherwise, the Zak phase takes 00. Formally, one can consider the matrix S(π)TS(−π)S(\pi)^{\mathrm{T}}S(-\pi), which takes the form of diag⁡(λ1,λ2,λ3)\operatorname{diag}(\lambda_1, \lambda_2, \lambda_3). λj\lambda_j (j=1,2,3j = 1, 2, 3) =±1= \pm1. The Zak phase of the jj-th band is 00 (π\pi) if λj=1\lambda_j = 1 (λj=−1\lambda_j = -1) [17]. It should be noted that in the static case [16], the Zak phase also takes other values, such as −π-\pi and 2π2\pi. They are introduced solely to distinguish between two conjugate elements in the same class. ±π\pm\pi are associated with different gauge choices of the eigenstates, and treated as the same here. We list the obtained Zak phases for all quaternion charges below:

qq11{±i}\{\pm i\}{±j}\{\pm j\}{±k}\{\pm k\}−1-1
Third band00π\piπ\pi0000
Second band00π\pi00π\pi00
First band0000π\piπ\pi00

Table S1. Zak phase of the quasienergy band

In the same conjugacy class (e.g., ii and −i-i), the Zak phase pattern is the same. The constraint that the summation of Zak phases for all three bands equals 00 (mod 2π2\pi) leads to four possible Zak phase patterns. Since there are five conjugacy classes in the quaternion group QQ, the class −1-1 falls outside the scope of the Zak-phase description. Furthermore, the Zak phase is defined for the Floquet Hamiltonian, while the topology of the system is fully encoded in the time-evolution operator or phase-band singularities. As discussed in the main text, the bulk-edge (or domain-wall) correspondence is multifold. Different patterns of edge states (domain-wall states) can have the same quaternion charge and Zak phase configuration. Therefore, the quaternion charge or the Zak phase is insufficient to predict the edge states, and descriptions based on phase-band singularities are necessary.

Stability of edge/domain-wall states

In this section, we conduct numerical studies to test the stability of edge states and domain-wall states. We start by varying the boundary conditions and showcase the emergence of edge states. Starting from the periodic boundary condition, we gradually decrease the coupling strengths between the first and the last (NN-th) unit cell: vXY,N=(1−α)vXYv_{XY,N} = (1-\alpha)v_{XY}. Here vXYv_{XY} represents the intercell coupling in the bulk [See the model, Eq. (1) in the main text], and vXY,Nv_{XY,N} denotes the intercell coupling between the first and the last unit cell. In Figure S5, b, we plot the quasienergy bands for quaternion charges ii (with phase-band singularity i^\hat{i}) and kk (with phase-band singularity k^\hat{k}), respectively. By varying α\alpha continuously from 00 (periodic boundary condition) to 11 (open boundary condition), we observe the appearance of edge states in the second gap for charge ii and the first gap for charge kk, detached from neighboring bands.

Stability of edge states in the presence of disorder

Figure S5. Stability of edge states in the presence of disorder. a, b Quasienergy spectra for quaternion charges ii and kk as a function of the boundary coupling strength α\alpha. The intercell coupling across the boundary is vXY,N=(1−α)vXYv_{XY,N}=(1-\alpha)v_{XY}, where α=0\alpha=0 and α=1\alpha=1 correspond to periodic and open boundary conditions, respectively. The in-gap edge modes (red lines) emerge from the neighboring bulk bands. c, d Quasienergy spectra under open boundary condition as a function of the disorder strength WW. The edge states are encircled by red ellipses. The parameters used are listed in Table 1.

We proceed to examine the robustness of the edge states by introducing onsite disorder: sXY,n=sXY+δs_{XY,n} = s_{XY} + \delta (n=1,2,…Nn = 1, 2, \ldots N, X=A,B,CX = A, B, C). Here, δ\delta is taken from the uniform distribution of [−W,W][-W,W], with WW representing the disorder strength. The numerical results are presented in Figure S5, d. It is evident that the edge states persist as long as the bulk quasienergy gaps are not closed by the disorder.

To test the stability of the domain-wall states, we consider four different types of domain walls, as depicted in Figure S6-Figure S6. In Figure S6, b, the intracell couplings sXYs_{XY} of the interface unit cell are set to be the same as those on the left and right sides, respectively. In Figure S6, d, the intracell couplings are set to zero for the interface unit cell, except for the additional next-nearest-neighbor couplings (with strength v=0.5v=0.5) in the latter case. We set the left/right side to possess quaternion charges ii (with phase-band singularity i~\tilde{i}) and kk (with phase-band singularity k~\tilde{k}) as an example. For all four types of domain wall, we observe the emergence of domain-wall states in the first and second gaps, as shown in Figure S6-Figure S6. This is consistent with the quotient relation, Δq=qR−1qL=−k~i~\Delta q=q_{\mathrm{R}}^{-1}q_{\mathrm{L}}=-\tilde{k}\tilde{i}. Furthermore, we test the robustness of these domain-wall states against onsite disorder in Figure S6-Figure S6. Similar to the edge states at the boundary, the domain-wall states survive weak disorder.

Figure S6. Stability of domain-wall states against different domain-wall settings and disorder. a-d Domain-wall configurations. The left (right) side has bulk quaternion charge ii (kk). The black arrows in d indicate the next-nearest-neighbor couplings. e-h Quasienergy spectra and spatial distributions of eigenstates for the four domain-wall settings. i-l Quasienergy spectra as a function of the disorder strength WW. The domain-wall states are encircled by red ellipses. The parameters used are listed in Table 1.

It is worth mentioning that the decorations of the domain wall may induce some additional bound states, as illustrated in Figure S7. We consider a tunable onsite potential with strength pp at the interface unit cell: sAA=sBB=sCC=ps_{AA}=s_{BB}=s_{CC}=p. In Figure S7 to Figure S7, we take four values of p=3,9,50,1000p=3,9,50,1000 and plot the quasienergy spectra and spatial distributions of the eigenstates. The domain-wall states located in the first and second gaps always exist. However, there also appear some trivial bound states. These bound states have no topological origin. Their quasienergies depend on the added potential and may merge into the bulk bands by adjusting the potential. In contrast, the domain-wall states are robust. They originate from the non-Abelian topology and are described by the quotient relation.

Quasienergy spectra and spatial distributions of eigenstates for different onsite potentials at the interface unit cell

Figure S7. Quasienergy spectra and spatial distributions of eigenstates for different onsite potentials at the interface unit cell. (a-d) The potential strengths are p=3p = 3, 9, 50 and 1000, respectively. The domain-wall states in the first and second gaps always exist (encircled by red ellipses). Trivial bound states may arise due to the local onsite potential. The parameters used are listed in Table 1.

References

  1. [1]X.-L. Qi and S.-C. Zhang, Topological insulators and superconductors. Rev. Mod. Phys. 83, 1057–1110 (2011). doi: 10.1103/RevModPhys.83.1057DOI
  2. [2]M. Z. Hasan and C. L. Kane, Colloquium: topological insulators. Rev. Mod. Phys. 82, 3045–3067 (2010). doi: 10.1103/RevModPhys.82.3045DOI
  3. [3]A. Altland and M. R. Zirnbauer, Nonstandard symmetry classes in mesoscopic normal-superconducting hybrid structures. Phys. Rev. B 55, 1142(1997). doi: 10.1103/PhysRevB.55.1142
  4. [4]S. Ryu, A. P. Schnyder, A. Furusaki and A. W. W. Ludwig, Topological insulators and superconductors: tenfold way and dimensional hierarchy. New J. Phys. 12, 065010(2010). doi: 10.1088/1367-2630/12/6/065010DOI
  5. [5]A. P. Schnyder, S. Ryu, A. Furusaki and A. W. W. Ludwig, Classification of topological insulators and superconductors in three spatial dimensions, Phys. Rev. B 78, 195125(2008). doi: 10.1103/PhysRevB.78.195125
  6. [6]A. Kitaev, Periodic table for topological insulators and superconductors, AIP Conference Proceedings 1134, 22 (2009). doi: 10.1063/1.3149495
  7. [7]C. -K. Chiu, J. C. Y. Teo, A. P. Schnyder and S. Ryu, Classification of topological quantum matter with symmetries. Rev. Mod. Phys. 88, 035005(2016). doi: 10.1103/RevModPhys.88.035005DOI
  8. [8]Q. Wu, A. A. Soluyanov and T. Bzdusek, Non-Abelian band topology in noninteracting metals. Science 365, 1273–1277 (2019). doi: 10.1126/science.aau8740DOI
  9. [9]J. Ahn, S. Park and B.-J. Yang, Failure of Nielsen-Ninomiya Theorem and Fragile Topology in Two-Dimensional Systems with Space-Time Inversion Symmetry: Application to Twisted Bilayer Graphene at Magic Angle. Phys. Rev. X 9, 021013(2019).doi: 10.1103/PhysRevX.9.021013
  10. [10]A. Bouhon, T. Bzdusek and R.-J. Slager, Geometric approach to fragile topology beyond symmetry indicators. Phys. Rev. B 102, 115135 (2020).doi: 10.1103/PhysRevB.102.115135DOI
  11. [11]C.-X. Guo, S. Chen, K. Ding and H. Hu, Exceptional Non-Abelian Topology in Multiband Non-Hermitian Systems. Phys. Rev. Lett. 130, 157201 (2023).doi: 10.1103/PhysRevLett.130.157201
  12. [12]B. Jiang, A. Bouhon, Z.-K. Lin, X. Zhou, B. Hou, F. Li, R.-J. Slager and J.-H. Jiang, Experimental observation of non-Abelian topological acoustic semimetals and their phase transitions. Nat. Phys. 17, 1239–1246 (2021).doi: 10.1038/s41567-021-01340-x
  13. [13]B. Jiang, A. Bouhon, S.-Q. Wu, Z.-L. Kong, Z.-K. Lin, R.-J. Slager and J.-H. Jiang, Experimental observation of meronic topological acoustic Euler insulators. arXiv:2205.03429 (2022).doi: 10.48550/arXiv.2205.03429
  14. [14]R. -J. Slager, A. Bouhon and F. N.Unal, Floquet multi-gap topology: Non-Abelian braiding and anomalous Dirac string phase. arXiv:2208.12824 (2022). doi: 10.48550/arXiv.2208.12824
  15. [15]F. N. Unal, A. Bouhon and R.-J. Slager, Topological Euler Class as a Dynamical Observable in Optical Lattices. Phys. Rev. Lett. 125, 053601 (2020). doi: Phys-RevLett.125.053601
  16. [16]Q. Guo, T. Jiang, R.-Y. Zhang, L. Zhang, Z.-Q. Zhang, B. Yang, S. Zhang and C. T. Chan, Experimental observation of non-Abelian topological charges and edge states. Nature 594, 7862 (2021). doi: 10.1038/s41586-021-03521-3DOI
  17. [17]T. Jiang, Q. Guo, R.-Y. Zhang, Z.-Q. Zhang, B. Yang and C. T. Chan, Four-band non-Abelian topological insulator and its experimental realization. Nat. Commun. 12, 6471 (2021). doi: 10.1038/s41467-021-26763-1DOI
  18. [18]G. P. Alexander, B. G. Chen, E. A. Matsumoto and R. D. Kamien, Disclination loops, point defects, and all that in nematic liquid crystals. Rev. Mod. Phys. 84, 497514 (2012). doi: 10.1103/RevModPhys.84.497DOI
  19. [19]G. E. Volovik and V. P. Mineev, Investigation of singularities in superfluid He3 in liquid crystals by the homotopy topology methods. Zh. Eksp. Teor. Fiz. 72, 2256–2274 (1977). doi: 10.1051/jphys:019770038088700
  20. [20]V. Poenaru and G. Toulouse, The crossing of defects in ordered media and the topology of 3-manifolds. J. Phys. Lett. 38, 887–895 (1977). doi: 10.1051/jphys:019770038088700
  21. [21]L. A. Madsen, T. J. Dingemans, M. Nakata and E. T. Samulski, Thermotropic biaxial nematic liquid crystals. Phys. Rev. Lett. 92, 145505 (2004). doi: 10.1103/PhysRevLett.92.145505DOI
  22. [22]K. Liu, J. Nissinen, R.-J. Slager, K. Wu and J. Zaanen, Generalized liquid crystals: giant fluctuations and the vestigial chiral order of I, O and T matter. Phys. Rev. X 6, 041025 (2016). doi: 10.1103/PhysRevX.6.041025DOI
  23. [23]A. T. Twar... and T. Bzdusek, Non-Abelian topology of nodal-line rings in PT-symmetric systems. Phys. Rev. B 101, 195130 (2020). doi: 10.1103/PhysRevB.101.195130
  24. [24]E. Yang, B. Yang, Q. You, H.-C. Chan, P. Mao, Q. Guo, S. Ma, L. Xia, D. Fan, Y. Xiang and S. Zhang, Observation of Non-Abelian Nodal Links in Photonics. Phys. Rev. Lett. 125, 033901 (2020). doi: 10.1103/PhysRevLett.125.033901DOI
  25. [25]D. Wang, B. Yang, Q. Guo, R.-Y. Zhang, L. Xia, X. Su, W.-J. Chen, J. Han, S. Zhang and C. T. Chan, Intrinsic in-plane nodal chain and generalized quaternion charge protected nodal link in photonics. Light.: Sci. Appl. 10, 83 (2021). doi: 10.1038/s41777-021-00523-8DOI
  26. [26]H. Park, S. Wong, X. Zhang and S. S. Oh, Non-Abelian Charged Nodal Links In Dielectric Photonic Crystal. ACS Photonics 2021, 8, 9, 2746–2754 (2021). doi: 10.1021/acsphotonics.1c00876DOI
  27. [27]M. Wang, S. Liu, Q. Ma, R.-Y. Zhang, D. Wang, Q. Guo, B. Yang, M. Ke, Z. Liu and C. T. Chan, Experimental observation of non-Abelian earring nodal links in phononic crystals. Phys. Rev. Lett. 128, 246601 (2022). doi: 10.1103/PhysRevLett.128.246601DOI
  28. [28]A. Bouhon, Q. S. Wu, R.-J. Slager, H. Weng, O. V. Yazyev and T. Bzdusek, Non-Abelian reciprocal braiding of Weyl points and its manifestation in ZrTe. Nat. Phys. 16, 1137–1143 (2020). doi: 10.1038/s41567-020-0967-9DOI
  29. [29]B. Peng, A. Bouhon, R.-J. Slager and B. Monserrat, Multi-gap topology and non-Abelian braiding of phonons from first principles. Phys. Rev. B 105, 085115 (2022). doi: 10.1103/PhysRevB.105.085115DOI
  30. [30]B. Peng, A. Bouhon, B. Monserrat and R.-. Slager, Phonons as a platform for non-Abelian braiding and its manifestation in layered silicates. Nat. Commun. 13, 423 (2022). doi: 10.1038/s41467-022-28046-9DOI
  31. [31]H. Qiu, Q. Zhang, T. Liu, X. Fan, F. Zhang and C. Qiu, Minimal non-abelian nodal braiding in ideal metamaterials. Nat. Commun. 14, 1261 (2023). doi: 10.1038/s41467-023-36952-9DOI
  32. [32]T. Kitagawa, E. Berg, M. Rudner and E. Demler, Topological characterization of periodically driven quantum systems. Phys. Rev. B 82, 235114(2010). doi: 10.1103/PhysRevB.82.235114DOI
  33. [33]L. Jiang, T. Kitagawa, J. Alicea, A. R. Akhmerov, D. Pekker, G. Refael, J. I. Cirac, E. Demler, M. D. Lukin, and P. Zoller, Majorana Fermions in Equilibrium and in Driven Cold-Atom Quantum Wires. Phys. Rev. Lett. 106, 220402 (2011). doi: 10.1103/PhysRevLett.106.220402DOI
  34. [34]A. Eckardt, Colloquium: Atomic quantum gases in periodically driven optical lattices. Rev. Mod. Phys. 2, 305 (1930). doi: 10.1103/RevModPhys.89.011004
  35. [35]J. Cayssol, B. Dra, F. Simon and R. Moessner, Floquet topological insulators. Phys. Status Solidi RRL 7, 101 (2013). doi: 10.1002/pssr.201206451DOI
  36. [36]M. S. Rudner and N. H. Lindner, Band structure engineering and non-equilibrium dynamics in Floquet topological insulators. Nature Reviews Physics 2, 229 (2020). doi: 10.1038/s42254-020-0170-zDOI
  37. [37]N. H. Lindner, G. Refael and V. Galitski, Floquet topological insulator in semiconductor quantum wells, Nat. Phys. 7, 490 (2011). doi: 10.1038/nphys1926DOI
  38. [38]M. M. Malz... and E. Zhao, Counter-propagating edge modes and topological phases of a kicked quantum Hall system. Phys. Rev. Lett. 112, 026805 (2014). doi: 10.1103/PhysRevLett.112.026805
  39. [39]R. Roy and F. Harper, Periodic table for Floquet topological insulators. Phys. Rev. B 96, 155118 (2017). doi: 10.1103/PhysRevB.96.155118DOI
  40. [40]S. Yao, Z. Yan and Z. Wang, Topological invariants of Floquet systems: General formulation, special properties, and Floquet topological defects. Phys. Rev. B 96, 195303 (2017). doi: 10.1103/PhysRevB.96.195303DOI
  41. [41]F. N. Unal, B. Seradjeh and A. Eckardt, How to Directly Measure Floquet Topological Invariants in Optical Lattices. Phys. Rev. Lett. 122, 253601 (2019). doi: 10.1103/PhysRevLett.122.253601DOI
  42. [42]M. S. Rudner, N. H. Lindner, E. Berg and M. Levin, Anomalous Edge States and the Bulk-Edge Correspondence for Periodically Driven Two-Dimensional Systems. Phys. Rev. X 3, 031005 (2013). doi: 10.1103/PhysRevX.3.031005DOI
  43. [43]L. J. Maczewsky, J. M. Zeuner, S. Nolte and A. Szameit, Observation of photonic anomalous Floquet topological insulators. Nat. Commun. 8, 13756 (2017). doi: 10.1038/ncomms13756DOI
  44. [44]S. Mukherjee, A. Spracklen, M. Valiente, E. Andersson, P. Öhberg, N. Goldman and R. R. Thomson, Experimental observation of anomalous topological edge modes in a slowly driven photonic lattice. Nat. Commun. 8, 13918 (2017). doi: 10.1038/ncomms13918DOI
  45. [45]K. Wintersperger, C. Braun, F. N. Unal, A. Eckardt, M. D. Liberto, N. Goldman, I. Bloch and M. Aidelsburger, Realization of an anomalous Floquet topological system with ultracold atoms. Nat. Phys. 16, 1058 (2020). doi: 10.1038/s41567-020-0949-y
  46. [46]F. Nathan and M. S. Rudner, Topological singularities and the general classification of Floquet Bloch systems. New J. Phys. 17, 125014 (2015). doi: 10.1088/1367-2630/17/12/125014DOI
  47. [47]H. Hu, B. Huang, E. Zhao and W. V. Liu, Dynamical Singularities of Floquet Higher-Order Topological Insulators. Phys. Rev. Lett. 124, 057001 (2020). doi: 10.1103/PhysRevLett.124.057001DOI
  48. [48]H. Hu and E. Zhao, Topological invariants for quantum quench dynamics from unitary evolution. Phys. Rev. Lett. 124, 160402 (2020). doi: 10.1103/PhysRevLett.124.160402DOI
  49. [49]See Supplementary Material for additional details on the multiplication rule in Eq. (4), examples with multiple Dirac singularities in the same gap, the phase-band explanation of the interface effect, the quotient relation in domain-wall problem, the Zak phase of quasisenergy bands, and the stability of edge/domain-wall states.
  50. [50]S. -Q. Shen, Topological insulators , pages 19-22 (Berlin: Springer, 2012).DOI
  51. [51]A. Hatcher, Algebraic Topology (Cambridge University Press, Cambridge, UK, 2002).
  52. [52]C. C. Wojcik, X. -Q. Sun, T. Bzdušek and S. Fan, Homotopy characterization of non-Hermitian Hamiltonians. Phys. Rev. B. 101, 205417 (2020). doi: 10.1103/PhysRevB.101.205417DOI
  53. [53]Z. Li and R. S. K. Mong, Homotopical characterization of non-Hermitian band structures. Phys. Rev. B 103, 155129 (2021). doi: 10.1103/PhysRevB.103.155129DOI
  54. [54]H. Hu and E. Zhao, Knots and Non-Hermitian Bloch Bands. Phys. Rev. Lett. 126, 010401 (2021). doi: 10.1103/PhysRevLett.126.010401DOI
  55. [55]N. R. Cooper, J. Dalibard and I. B. Spielman, Topological bands for ultracold atoms. Rev. Mod. Phys. 91, 015005 (2019). doi: 10.1103/RevModPhys.91.015005DOI
  56. [56]G. Jotzu, M. Messer, R. Desbuquois, M. Lebrat, T. Uehlinger, D. Greif and T. Esslinger, Experimental realization of the topological Haldane model with ultracold fermions. Nature 515, 237–240 (2014). doi: 10.1038/nature13915DOI
  57. [57]C. D. Brown, S.-W. Chang, M. N. Schwarz, T. -H. Leung, V. Kozii, A. Avdoshkin, J. E. Moore and D. Stamper-Kurn, Direct geometric probe of singularities in band structure. Science 377, 1319-1322 (2022). doi: 10.1126/science.abm6442DOI
  58. [58]T. -H. Leung, M. N. Schwarz, S.-W. Chang, C. D. Brown, G. Unnikrishnan and D. Stamper-Kurn, Interaction-enhanced group velocity of bosons in the flat band of an optical kagome lattice. Phys. Rev. Lett. 125, 133001 (2020). doi: 10.1103/PhysRevLett.125.133001DOI
  59. [59]Y. -K. Sun, X. -L. Zhang, F. Yu, Z.-N. Tian, Q. -D. Chen and H.-B. Sun, Non-Abelian Thouless pumping in topological waveguides. Nat. Phys. 18, 1080–1085 (2022). doi: 10.1038/s41567-022-01669-xDOI
  60. [60]O. You, S. Liang, B. Xie, W. Gao, W. Ye, J. Zhu and S. Zhang, Observation of Non-Abelian Thouless Pump. Phys. Rev. Lett. 128, 244302 (2022). doi: 10.1103/PhysRevLett.128.244302DOI

Paper details

vlnc:W_zscpa2tj
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents