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

Present address: ]Department of Physics, 104 Davey Lab, The Pennsylvania State University, University Park, Pennsylvania 16802, USA Present address: ]Graduate School of Engineering Science, Akita University, Akita, 010-8502, Japan

Bulk-edge Correspondence in the Adiabatic Heuristic Principle

Koji Kudo1,2 [    Yoshihito Kuno1 [    Yasuhiro Hatsugai1,2 1Department of Physics, University of Tsukuba, Tsukuba, Ibaraki 305-8571, Japan
2Graduate School of Pure and Applied Sciences, University of Tsukuba, Tsukuba, Ibaraki 305-8571, Japan
Abstract

Using the Laughlin’s argument on a torus with two pin-holes, we numerically demonstrate that the discontinuities of the center-of-mass work well as an invariant of the pumping phenomena during the process of the flux-attachment, trading the magnetic flux for the statistical one. This is consistent with the bulk-edge correspondence of the fractional quantum Hall effect of anyons. We also confirm that the general feature of the edge states remains unchanged during the process while the topological degeneracy is discretely changed. This supports the stability of the quantum Hall edge states in the adiabatic heuristic principle.

Introduction

— Characterization of quantum matter with topological invariants is a modern notion in condensed matter physics Thouless et al. (1982); Kohmoto (1985); Hasan and Kane (2010); Qi and Zhang (2011). The adiabatic deformation of gapped systems is a conceptual basis in the theory of topological phases beyond the Landau’s symmetry breaking paradigm Wen (1989, 2017). Meanwhile, augmented by the symmetry, this notion leads to more unified picture exemplified by the “periodic table” for topologically nontrivial states Kitaev (2009); Schnyder et al. (2008); Qi et al. (2008); Ryu et al. (2010) and demonstrates the existence of rich topological phases. The adiabatic deformation also gives a useful way to characterize concrete models by reducing them to simple systems Greiter and Wilczek (1990, 1992, 2021); Hatsugai (2005, 2006, 2007); Kariyado et al. (2018).

The adiabatic heuristic argument of the quantum Hall (QH) effect Greiter and Wilczek (1990, 1992, 2021) is the historical example in which the adiabatic deformation has been successfully used. The fractional QH (FQH) effect Tsui et al. (1982); Laughlin (1983) is a topological ordered phase Wen (1995) with fractionalized excitations Haldane (1983); Halperin (1984); Arovas et al. (1984). Even though it is intrinsically a many-body problem of correlated electrons unlike the integer QH (IQH) effect Klitzing et al. (1980); Laughlin (1981); Thouless et al. (1982); Halperin (1982), the composite fermion theory Jain (1989, 2007) gives a unified scheme to describe their underlying physics: the FQH state at the filling factor ν=p/(2mp±1)\nu=p/(2mp\pm 1) with p,mp,m integers can be interpreted as the ν=p\nu=p IQH state of the composite fermions. By continuously trading the external flux for the statistical one Wilczek (1982a, b), both states are adiabatically connected through intermediate systems of anyons (adiabatic heuristic principle Greiter and Wilczek (1990, 1992, 2021)). Even though the ground state degeneracy Haldane (1985); Wen and Niu (1990) is wildly changed in the periodic geometry Greiter and Wilczek (1992); Kudo and Hatsugai (2020), the energy gap remains open and its many-body Chern number Niu et al. (1985) works well as an adiabatic invariant Kudo and Hatsugai (2020).

Generally bulk topological invariants such as the Chern number are intimately related to the presence of gapless edge excitations. This is the so-called bulk-edge correspondence Wen (1990a); Hatsugai (1993a, b), which is a universal feature of topological phases Kane and Mele (2005); Haldane and Raghu (2008); Hasan and Kane (2010); Qi and Zhang (2011); Kariyado and Hatsugai (2015); Delplace et al. (2017); Sone and Ashida (2019); Hatsugai and Fukui (2016); Kuno and Hatsugai (2020); Mizoguchi et al. (2021); Yoshida et al. (2020). The edges of the QH systems demonstrate the nontrivial transport properties enriched by the bulk topology, which has attracted a great interest for over decades Laughlin (1981); Halperin (1982); MacDonald (1990); Wen (1990b, a, 1991); Johnson and MacDonald (1991); Wen (1992); Chamon and Wen (1994); Wen (1995); Meir (1994); Kane et al. (1994); Kane and Fisher (1995); Wan et al. (2002); Joglekar et al. (2003); Wan et al. (2003); Chang (2003); Hu et al. (2008, 2009); Wang et al. (2013a); Repellin et al. (2018); Fern et al. (2018); Wei et al. (2020); Ito and Shibata (2021); Khanna et al. (2021). The main goal in this work is to reveal how the quantum Hall edge states are evolved during the process of the flux-attachment in the adiabatic heuristic principle.

In this Letter, we analyze the fractional pumping phenomena associated with the Laughlin’s argument of the anyonic FQH effect. We show that the general feature of the energy spectrum with edges shows little change during the process of the flux-attachment while the topological degeneracy is wildly changed. Furthermore, the total jump of the center-of-mass works well as an invariant of this process, which is consistent with the bulk-edge correspondence of the FQH effect of anyons. This implies that the total jump of the center-of-mass characterizes the fractional charge pumping of the adiabatic heuristic principle. Also, this supports the stability of the QH edge states in the adiabatic heuristic principle.

Charge pumping

— Let us consider the QH system on a square lattice with Nx×NyN_{x}\times N_{y} sites, where Nx/Ny=2N_{x}/N_{y}=2 and the periodic boundary condition is imposed. As shown in Fig. 1(a), local fluxes ±ξ\pm\xi are set at two plaquettes A±A_{\pm} with the same yy coordinate. Their distance is Nx/2N_{x}/2. Particles are pumped from AA_{-} to A+A_{+} as ξ\xi varies from 0 to 11 [see Fig. 1(b)], which we call the (fractional) charge pump Thouless (1983); Wang et al. (2013b); Nakajima et al. (2016); Lohse et al. (2016); Guo et al. (2012); Xu et al. (2013); Zeng et al. (2015); Hu et al. (2016); Zeng et al. (2016); Nakagawa and Furukawa (2017); Taddia et al. (2017); Nakagawa et al. (2018) throughout this Letter.

Refer to caption
Figure 1: (a) Sketch of 6×36\times 3 square lattice. The gauge ξij\xi_{ij} (red arrows) describes the two local fluxes ±ξ\pm\xi at A±A_{\pm}. (b) Charge pump from AA_{-} to A+A_{+}. (c) One-dimensional projection into xx-axis. The projected sites for (a) are shown. The angle θi\theta_{\vec{i}} is measured from axa_{x} that is the xx coordinate of A+A_{+}. (d) One-dimensional charge pump on 1/2xi1/2-1/2\leq x_{\vec{i}}\leq 1/2.

This charge pump can be mapped into the one-dimensional pump with edges [Fig. 1(d)]. As shown in Fig. 1(c), we first project the system into the xx-axis. Then, projecting it into the green line shown in Figs 1(c), we finally define a new coordinate for site i=(ix,iy)\vec{i}=(i_{x},i_{y}) as xi=(1/2)cosθix_{\vec{i}}=(1/2)\cos\theta_{\vec{i}} with θi=2π(ixax)/Nx\theta_{\vec{i}}=2\pi(i_{x}-a_{x})/N_{x}, where axa_{x} is the xx coordinate of A+A_{+}, see Fig. 1(d). In this projection, the two pin-holes A±A_{\pm} are mapped into the edges xi=±1/2x_{\vec{i}}=\pm 1/2.

The charge can be transformed from xi=1/2x_{\vec{i}}=-1/2 to xi=1/2x_{\vec{i}}=1/2 as ξ\xi increases. The pumped charge is given by the integration of ξP(ξ)\partial_{\xi}P(\xi), where PP is the center-of-mass,

P(ξ)=Tr[ρ(ξ)ixini],\displaystyle P(\xi)=\text{Tr}\,[\rho(\xi)\sum_{\vec{i}}x_{\vec{i}}n_{\vec{i}}], (1)

ρ\rho is the zero temperature density matrix in the grand canonical ensemble and nin_{\vec{i}} is the number operator at the site i\vec{i}. (In Sec. S1 of Supplemental Material sup , we derive the pumped charge by using the current operator.) As ξ\xi varies, P(ξ)P(\xi) jumps several times due to the sudden change of the particle number Hatsugai and Fukui (2016); Kuno and Hatsugai (2020, 2021). Accordingly, the pumped charge between the period ξ[0,1]\xi\in[0,1] is given by Q=(0ξ1+ξ1+ξ2++ξN+1)dξξP(ξ)Q=\left(\int_{0}^{\xi_{1}^{-}}+\int_{\xi_{1}^{+}}^{\xi_{2}^{-}}+\cdots+\int_{\xi_{N}^{+}}^{1}\right)d\xi\,\partial_{\xi}P(\xi), where ξ1,,ξN\xi_{1},\cdots,\xi_{N} are the jumping points in the period and ξα±=ξα±0\xi_{\alpha}^{\pm}=\xi_{\alpha}\pm 0. Using the periodicity P(1)=P(0)P(1)=P(0) and ΔP(ξα)P(ξα+)P(ξα)\Delta P(\xi_{\alpha})\equiv P(\xi_{\alpha}^{+})-P(\xi_{\alpha}^{-}), we get Hatsugai and Fukui (2016)

Q=α=1NΔP(ξα)ΔPtot.\displaystyle Q=-\sum_{\alpha=1}^{N}\Delta P(\xi_{\alpha})\equiv-\Delta P_{\text{tot}}. (2)

As shown below, the total jump ΔPtot\Delta P_{\text{tot}}, i.e., the sudden changes of the particle number, comes from the (dis)appearance of edge states. Equation (2) implies that the pumped charge is given only by the information of edges.

Bulk-edge correspondence

— In this Letter, we numerically show the following bulk-edge correspondence for the FQH states of anyons:

C=ND×ΔPtot,\displaystyle C=-N_{D}\times\Delta P_{\text{tot}}, (3)

where CC is the many-body Chern number Niu et al. (1985) of the NDN_{D}-fold degenerate ground state multiplet at ξ=0\xi=0. This is consistent with the Laughlin’s argument applied to the FQH systems Laughlin (1981); Halperin (1982); Thouless et al. (1982); Thouless (1983); Niu et al. (1985); Hatsugai (1993a); Thouless (1989); Hatsugai and Fukui (2016); Zeng et al. (2016); Grushin et al. (2015); Andrews et al. (2021) that implies Q=C/NDQ=C/N_{D}. In the following, we clarify how the fractional charge pumping is deformed to the standard pumping phenomena by the flux-attachment transformation. As mentioned below, the relation in Eq. (3) results in the stability of the QH edge states in the adiabatic heuristic principle.

Fermion pumping

— As a first step, we confirm Eq. (3) for the IQH system of non-interacting fermions. The Hamiltonian is H=tijeiϕijeiξijcicjH=-t\sum_{\langle ij\rangle}e^{i\phi_{ij}}e^{i\xi_{ij}}c_{i}^{\dagger}c_{j}, where cic_{i}^{\dagger} is the creation operator for a fermion on site ii and t=1t=1. The phase factors eiϕije^{i\phi_{ij}} and eiξije^{i\xi_{ij}} describe the uniform magnetic field Hofstadter (1976); Hatsugai et al. (1999) and the local fluxes at A±A_{\pm} [see Fig. 1(a)], respectively. We plot in Fig. 2(a) the single-particle energy ϵ\epsilon with Nx×Ny=40×20N_{x}\times N_{y}=40\times 20 and ϕNϕ/(NxNy)=1/10\phi\equiv N_{\phi}/(N_{x}N_{y})=1/10, where NϕN_{\phi} is the total uniform fluxes. Each set of Nϕ(=80)N_{\phi}(=80) states forms the Landau level (LL) at ξ=0\xi=0. As ξ\xi increases, some edge states go over to the mid-gap region. In Fig. 2(b), we compute P(ξ)P(\xi) with ρ=|GG|\rho=|{G}\rangle\langle{G}| in Eq. (1), where |G|{G}\rangle is the ground state completely occupying the 1st LL under the chemical potential (Fermi energy) μ=3\mu=-3. The sudden change of the particle number NpN_{p} causes the jumps of P(ξ)P(\xi) at ξ1\xi_{1} and ξ2\xi_{2}. Both values of ΔP(ξα)\Delta P(\xi_{\alpha})’s are approximately 1/2-1/2, which is consistent with Fig. 2(a) where one edge state at xi=1/2x_{\vec{i}}=1/2 goes over across μ\mu and then another at xi=1/2x_{\vec{i}}=-1/2 goes back. Although a finite size effect gives ΔPtot=ΔP(ξ1)+ΔP(ξ2)0.96\Delta P_{\text{tot}}=\Delta P(\xi_{1})+\Delta P(\xi_{2})\approx-0.96, we confirm ΔPtot=1\Delta P_{\text{tot}}=-1 in the thermodynamic limit (see Sec. S2 of Supplemental Material sup ). This is consistent with Eq. (3) with ND=1N_{D}=1 and C=1C=1. The cases for C=2C=2 and 33 have been also confirmed.

Refer to caption
Figure 2: (a) Single-particle energy ϵ\epsilon on 40×2040\times 20 lattices with ϕ=1/10\phi=1/10. The green (orange) plots represent the bulk (edge) states. The blue line is μ=3\mu=-3. (b) Center-of-mass PP. The two jumps are ΔP(ξ1)ΔP(ξ2)0.48\Delta P(\xi_{1})\approx\Delta P(\xi_{2})\approx-0.48.

Normalized jumps

— As mentioned above, the jump ΔP(ξα)\Delta P(\xi_{\alpha}) are not quantized to ±1/2\pm 1/2 due to the finite size effect. Let us then properly normalize each jumps: when PP jumps positively or negatively at ξα\xi_{\alpha}, we assign it as ΔP(ξα)1/2\Delta P(\xi_{\alpha})\mapsto 1/2 or 1/2-1/2. Hereafter “\mapsto” denotes this normalization; e.g., we have ΔPtot=P(ξ1)+P(ξ2)1/21/2=1\Delta P_{\text{tot}}=P(\xi_{1})+P(\xi_{2})\mapsto-1/2-1/2=-1 in Fig. 2(b). This gives the bulk-edge correspondence in Eq. (3) even for finite systems.

Fractional anyon pumping

Refer to caption
Figure 3: (a) Energy gaps as functions of 1/ν1/\nu at ξ=0\xi=0 with Np=5N_{p}=5. The statistical angle θ\theta is determined by ν=p/[p(1θ/π)+1]\nu=p/[p(1-\theta/\pi)+1] with p=1p=1. The color expresses the denominator of θ/π\theta/\pi. (b) Ground state degeneracy NDN_{D}. (c1)-(c8) Energy spectra as functions of local fluxes ξ\xi. Each value of (ν,θ)\nu,\theta) is represented above the panel (b). We set μ=3\mu=-3 in (c1) and choose μ\mu in (c2)-(c8) so that the lowest energy at ξ=0\xi=0 is same as that in (c1). In (c3) and (c8), the NDN_{D} lowest energy states are marked by dots. In (a)-(c), we plot the lowest NcutN_{\text{cut}} energies with Ncut=2N_{\text{cut}}=2 for Np=6N_{p}=6 and Ncut=20N_{\text{cut}}=20 for Np=5,4N_{p}=5,4. (d)-(h) Center of mass PP as for the panels (c3) and (c8). ξI\xi_{\text{I}},ξII\xi_{\text{II}} and ξi\xi_{\text{i}}-ξiv\xi_{\text{iv}} represent the gap closing points.

Let us consider the fractional pumping of anyons. To this end, we take the Hamiltonian as H=tijeiϕijeiξijeiθijcicjH=-t\sum_{\langle ij\rangle}e^{i\phi_{ij}}e^{i\xi_{ij}}e^{i\theta_{ij}}c_{i}^{\dagger}c_{j}, where the phase factor θij\theta_{ij} Wen et al. (1990); Hatsugai et al. (1991); fer depends on the configuration of all particles {𝒓k}1kNp\{\bm{r}_{k}\}_{1\leq k\leq N_{p}}, which describes the fractional statistics eiθe^{i\theta}. Note that although cic_{i}^{\dagger} is the creation operator for a fermion, HH is the Hamiltonian of anyons and includes intrinsically the many-body interactions. Due to constraints of the braid group, dimH\dim H depends on θ\theta even for the same NpN_{p} Wen et al. (1990); Einarsson (1990): the Hilbert space for θ/π=n/m\theta/\pi=n/m (n,mn,m: coprime) is spanned by the basis |{𝒓k};w|{\{\bm{r}_{k}\};w}\rangle, where w=1,,mw=1,\cdots,m is an additional internal degree of freedom. When a particle hops across the boundary in the xx direction, the label is shifted from ww to w1w-1. As for the boundary in the yy direction, the phase factor eiwθe^{iw\theta} is given. Thanks to this, global requirements of anyons hold GR ; Wen et al. (1990); Einarsson (1990); Hatsugai et al. (1991). Also, we introduce ξij\xi_{ij} only for the basis with w=1w=1 def .

In the following, we focus on a family of the ν=1\nu=1 IQH states connected by trading the magnetic fluxes for statistical ones Jain (1989); Greiter and Wilczek (1990, 1992); Kudo and Hatsugai (2020): ν=p/[p(1θ/π)+1]\nu=p/[p(1-\theta/\pi)+1] with p=1p=1. Fixing Nx×Ny=10×5N_{x}\times N_{y}=10\times 5, Np=5N_{p}=5 and ξ=0\xi=0, we plot the energy gaps as functions of 1/ν1/\nu in Fig. 3(a). Due to the lattice, the topological degeneracy is lifted. We here define the low-energy states with EnE1<0.2E_{n}-E_{1}<0.2 as the ground state multiplet. The ground state at ν=s/t\nu=s/t (s,ts,t: coprime) in Fig. 3(a) gives the degeneracy ND=tN_{D}=t numerically [see Fig. 3(b)] and the Chern numbers of the multiplets are always C=1C=1 Kudo and Hatsugai (2020). Namely, the many-body Chern number is used as an adiabatic invariant. The gap closing at ν1/2\nu\approx 1/2 is expected due to finite-size effects Kudo and Hatsugai (2020).

Let us investigate the pumping phenomena. As for each parameter (1/ν,θ/π)(1/\nu,\theta/\pi) shown in Fig. 3(b), we plot in Figs. 3(c1)-(c8) the eigenvalues of the Hamiltonian including the chemical potential, HμNpH-\mu N_{p}, as functions of ξ\xi with 4Np64\leq N_{p}\leq 6 Np . Figure 3(c1) is in the same setting of Fig. 2 but for smaller system sizes. In Fig 3(c1), the particle number NpN_{p} of the unique ground state is changed as ξ\xi increases due to the (dis)appearance of the edge state as mentioned previously. The gap between the two red lines at ξ0.5\xi\approx 0.5 is a finite-size effect. As shown in Figs. 3(c1)-(c8), even though the topological degeneracy is wildly changed as 1/ν1/\nu and θ/π\theta/\pi vary, the general feature of the spectra remains unchanged. The degenerate ground states at ξ=0\xi=0 are lifted as ξ\xi increases and then one or two states float up in energy to cross with another state having one particle less.

Now we focus on the anyonic system in Fig. 3(c3) and show its the bulk-edge correspondence. Here θ/π=6/7\theta/\pi=6/7, ν=7/8\nu=7/8 and ND=8N_{D}=8 at ξ=0\xi=0. To define the center-of-mass of the ground state multiplet suitably, we define the density matrix as ρ(ξ)=(1/ND)k=1ND|Gk(ξ)Gk(ξ)|\rho(\xi)=(1/N_{D})\sum_{k=1}^{N_{D}}|{G_{k}(\xi)}\rangle\langle{G_{k}(\xi)}|, where |Gk(ξ)|{G_{k}(\xi)}\rangle is the kk-th lowest energy state. Using it with Eq. (1), we plot ND×P(ξ)N_{D}\times P(\xi) in Fig. 3(d). There are two jumps at ξI\xi_{\text{I}} and ξII\xi_{\text{II}}, where the NDN_{D}th and ND+1N_{D}+1th lowest energy states cross each other in the spectrum. Because of NDρ=k=1ND|GkGk|N_{D}\rho=\sum_{k=1}^{N_{D}}|{G_{k}}\rangle\langle{G_{k}}|, the obtained jumps are solely given by PP with ρ=|GNDGND|\rho=|{G_{N_{D}}}\rangle\langle{G_{N_{D}}}| shown in Fig. 3(e). This figure gives ΔPtot=ΔP(ξI)+P(ξII)1\Delta P_{\text{tot}}=\Delta P(\xi_{I})+P(\xi_{II})\mapsto-1, which implies ND×ΔPtot1N_{D}\times\Delta P_{\text{tot}}\mapsto-1 in Fig. 3(d). This is consistent with Eq. (3) with C=1C=1. Because of Q=ΔPtotQ=-\Delta P_{\text{tot}}, we ahve the fractional pumped charge Q=1/8Q=1/8. In this argument, we assume the absence of the gap closing between states with the same NpN_{p} apart from ξ=0\xi=0 since there are no symmetry except for the charge U(1)U(1). The gap at ξ0.7\xi\approx 0.7 between |GND|{G_{N_{D}}}\rangle and |GND1|{G_{N_{D}-1}}\rangle is very small but is finite Two as shown in the inset.

Let us here mention the finite size effect in Fig. 3(d). The value of ND×ΔPtotN_{D}\times\Delta P_{\text{tot}} before normalizing is about 0.46-0.46, which is far away from 1-1. Although this value in the IQH system in Fig. 3(c1) is about the same magnitude (about 0.60-0.60, see the data point at ν=1\nu=1 of Fig. 4), it approaches 1-1 as the system size increases as confirmed in Sec. S2 of Supplemental Material sup . Since the bulk gaps of the two systems are comparable and their system sizes are same, the deviation from 1-1 in Fig. 3(d) is also expected to be the finite size effect.

Let us next focus on the system in Fig. 3(c8), where θ/π=2/3\theta/\pi=2/3, ν=3/4\nu=3/4 and ND=4N_{D}=4 at ξ=0\xi=0. Unlike the previous case, there are four gap-closing points, ξiξiv\xi_{\text{i}}\cdots\xi_{\text{iv}}, as for the NDN_{D} lowest energy states. However, P(ξ)P(\xi) with ρ=(1/ND)k=1ND|GkGk|\rho=(1/N_{D})\sum_{k=1}^{N_{D}}|{G_{k}}\rangle\langle{G_{k}}| in Fig. 3(f) jumps only at ξi\xi_{\text{i}} and ξiv\xi_{\text{iv}} because the jumps at ξii,ξiii\xi_{\text{ii}},\xi_{\text{iii}} cancel each other, see Figs. 3(g) and 3(h). Consequently, the total jump is given by ND×ΔPtot1N_{D}\times\Delta P_{\text{tot}}\mapsto-1, which is consistent with Eq. (3) with C=1C=1. This implies the fractional pumped charge Q=1/4Q=1/4. The gap at ξ0.5\xi\approx 0.5 is very small but is finite as mentioned before, see the inset in Fig. 3(c8).

The results shown in Figs. 3(c3) and 3(c8) suggest that ND×ΔPtotN_{D}\times\Delta P_{\text{tot}} is the invariant of the bulk gap in the process of the flux-attachment. To demonstrate it, we plot the total jumps as functions of 1/ν1/\nu in Fig. 4, where both data before/after normalizing each jumps are shown. The normalized data justify that ND×ΔPtotN_{D}\times\Delta P_{\text{tot}} works well as the invariant. This nature is also indicated by the unnormalized data in Fig. 4: the plots are smooth as 1/ν1/\nu and θ/π\theta/\pi vary even though (i) the degeneracy NDN_{D} is wildly changed and (ii) the dimension of the Hamiltonian is discretely changed depending on the denominator of θ/π\theta/\pi: e.g., with Np=5N_{p}=5, dimH=(NxNyNp)=(505)=2118760\dim H=\binom{N_{x}N_{y}}{N_{p}}=\binom{50}{5}=2118760 for θ/π=n\theta/\pi=n while dimH=11(505)=23306360\dim H=11\binom{50}{5}=23306360 for θ/π=n/11\theta/\pi=n/11 (this is due to the additional internal degree ww of the basis |{𝒓k};w|{\{\bm{r}_{k}\};w}\rangle as mentioned above). We stress that this nontrivial smoothness in Fig. 4 implies the stability of the QH edge states in the adiabatic heuristic principle.

Refer to caption
Figure 4: The total jump of the center-of-mass ND×ΔPtotN_{D}\times\Delta P_{\text{tot}}. The solid and dotted lines mean the normalized and the unnormalized data, respectively. The systems in Figs. 3(c1)-(c8) are marked. The sudden change in the dotted line, represented by a red arrow, is due to the gap collapse of the ND1N_{D}-1th lowest energy state, compare Figs. 3(c5) and 3(c6). The jumps are calculated by discretizing the period ξ[0,1]\xi\in[0,1] into NξN_{\xi} meshes with Nξ=48N_{\xi}=48 (Only at (c6) point, we set Nξ=240N_{\xi}=240).

Conclusion

— In this Letter, we demonstrate the bulk-edge correspondence of the FQH states of anyons. The results indicate that the total jump of the center-of-mass, which corresponds to the many-body Chern number, is an invariant with respect to the flux-attachment. This implies the stability of edge states in the adiabatic heuristic principle. Recently, direct observation of the center of mass in pumping phenomena has been conducted in cold atoms Nakajima et al. (2016); Lohse et al. (2016). The behavior of the center-of-mass that we focus on would be observed in cold atoms although the experimental realization of the two-dimensional anionic system is still a challenging problem.

Acknowledgements.
We thank the Supercomputer Center, the Institute for Solid State Physics, the University of Tokyo for the use of the facilities. The work is supported in part by JSPS KAKENHI Grant Numbers JP17H06138, JP19J12317 (K.K.), and JP21K13849 (Y.K.).

References

Supplemental Material

S1 Current operator and the pumped charge

In this appendix, we derive the pumped charge QQ by using the current operator. We consider the Hamiltonian described in the paragraph Fractional anyon pumping in the main text:

H(ξ)=ti,jeiϕi,jeiξi,jeiθi,jcicj,\displaystyle H(\xi)=-t\sum_{\langle\vec{i},\vec{j}\rangle}e^{i\phi_{\vec{i},\vec{j}}}e^{i\xi_{\vec{i},\vec{j}}}e^{i\theta_{\vec{i},\vec{j}}}c^{\dagger}_{\vec{i}}c_{\vec{j}}, (S1)

where cic_{\vec{i}}^{\dagger} is the creation operator for a fermion on site i\vec{i}, and the phase factors eiϕi,je^{i\phi_{\vec{i},\vec{j}}}, eiξi,je^{i\xi_{\vec{i},\vec{j}}} and eiθi,je^{i\theta_{\vec{i},\vec{j}}} describe the uniform fluxes, the local fluxes ±ξ\pm\xi and the statistical phase θ\theta, respectively. Using a unitary operator,

U(α)=ieiαxini,\displaystyle U(\alpha)=\prod_{\vec{i}}e^{-i\alpha x_{\vec{i}}n_{\vec{i}}},

let us modify the Hamiltonian in Eq. (S1):

H(α,ξ)\displaystyle H(\alpha,\xi) U(α)H(ξ)U(α)\displaystyle\equiv U(\alpha)H(\xi)U^{\dagger}(\alpha)
=ti,jeiα(xixj)eiϕi,jeiξi,jeiθi,jcicj,\displaystyle=-t\sum_{\langle\vec{i},\vec{j}\rangle}e^{-i\alpha(x_{\vec{i}}-x_{\vec{j}})}e^{i\phi_{\vec{i},\vec{j}}}e^{i\xi_{\vec{i},\vec{j}}}e^{i\theta_{\vec{i},\vec{j}}}c^{\dagger}_{\vec{i}}c_{\vec{j}}, (S2)

where U(α)ciU(α)=eiαxiciU(\alpha)c_{\vec{i}}U^{\dagger}(\alpha)=e^{i\alpha x_{\vec{i}}}c_{\vec{i}} is used. We then define the current operator in the xx direction as

x\displaystyle\mathcal{I}_{x} =iti,j(xixj)eiα(xixj)eiϕi,jeiξi,jeiθi,jcicj\displaystyle=i\frac{t}{\hbar}\sum_{\langle\vec{i},\vec{j}\rangle}(x_{\vec{i}}-x_{\vec{j}})e^{-i\alpha(x_{\vec{i}}-x_{\vec{j}})}e^{i\phi_{\vec{i},\vec{j}}}e^{i\xi_{\vec{i},\vec{j}}}e^{i\theta_{\vec{i},\vec{j}}}c^{\dagger}_{\vec{i}}c_{\vec{j}}
=1αH(α,ξ).\displaystyle=\frac{1}{\hbar}\partial_{\alpha}H(\alpha,\xi).

We now assume the following density matrix:

ρ(α,ξ)=1NDΦ(α,ξ)Φ(α,ξ),\displaystyle\rho(\alpha,\xi)=\frac{1}{N_{D}}\Phi(\alpha,\xi)\Phi^{\dagger}(\alpha,\xi),

where Φ(α,ξ)=(|G1(α,ξ),,|GND(α,ξ))\Phi(\alpha,\xi)=(|{G_{1}(\alpha,\xi)}\rangle,\cdots,|{G_{N_{D}}(\alpha,\xi)}\rangle) is the ground state multiplet of the Hamiltonian H(α,ξ)H(\alpha,\xi). The measured current IxI_{x} computed from x\mathcal{I}_{x} is reduced to the Berry curvature Thouless (1983); Hatsugai and Fukui (2016):

Ix(α,ξ)=iNDB(α,ξ),\displaystyle I_{x}(\alpha,\xi)=-\frac{i}{N_{D}}B(\alpha,\xi),
B(α,ξ)=αAξ(α,ξ)ξAα(α,ξ),\displaystyle B(\alpha,\xi)=\partial_{\alpha}A_{\xi}(\alpha,\xi)-\partial_{\xi}A_{\alpha}(\alpha,\xi),
As(α,ξ)=tr[Φ(α,ξ)sΦ(α,ξ)],s=α,ξ,\displaystyle A_{s}(\alpha,\xi)=\text{tr}\left[\Phi^{\dagger}(\alpha,\xi)\partial_{s}\Phi(\alpha,\xi)\right],\ s=\alpha,\xi,

where “tr” indicates the trace of a NDN_{D}-dimensional matrix. The gauge transformation in Eq. (S2) implies Φ(α,ξ)=U(α)Φ(ξ)\Phi(\alpha,\xi)=U(\alpha)\Phi(\xi), i.e.,

Aα(α,ξ)=tr[Φ(ξ)(iixini)Φ(ξ)]=iNDP(ξ),\displaystyle A_{\alpha}(\alpha,\xi)=-\text{tr}\,\left[\Phi(\xi)^{\dagger}\left(\sum_{\vec{i}}ix_{\vec{i}}n_{\vec{i}}\right)\Phi(\xi)\right]=iN_{D}P(\xi),
Aξ(α,ξ)=tr[Φ(ξ)ξΦ(ξ)]Aξ(ξ),\displaystyle A_{\xi}(\alpha,\xi)=\text{tr}\,\left[\Phi(\xi)^{\dagger}\partial_{\xi}\Phi(\xi)\right]\equiv A_{\xi}(\xi),

where P(ξ)P(\xi) is the center-of-mass defined in Eq. (1). By fixing the gauge Φ(ξ)\Phi(\xi) properly Thouless (1983); Hatsugai (2004); Hatsugai and Fukui (2016), one can prepare Aξ(ξ)A_{\xi}(\xi) as a well-defined function. Because of αAξ(ξ)=0\partial_{\alpha}A_{\xi}(\xi)=0, we have B(α,ξ)=iNDξP(ξ)B(\alpha,\xi)=iN_{D}\partial_{\xi}P(\xi), namely,

Ix(α,ξ)=ξP(ξ).\displaystyle I_{x}(\alpha,\xi)=\partial_{\xi}P(\xi). (S3)

The pumped charge QQ is given by the integration of the current Ix(α,ξ)I_{x}(\alpha,\xi) over ξ\xi. This clearly justifies the formulation using P(ξ)P(\xi) in the main text.

S2 Finite size effect

Let us discuss a finite size effect of numerical simulation. In Fig. S1, we plot the total jump ΔPtot\Delta P_{\text{tot}} at ν=1\nu=1, where we fix the aspect ratio, the chemical potential, and flux per plaquette as Nx/Ny=2N_{x}/N_{y}=2, μ=3\mu=-3, and ϕ=1/10\phi=1/10, respectively. As indicated in the figure, the systems in Figs. 2, 3(c1) and 4 are included. As the system size NxN_{x} increases, ΔPtot\Delta P_{\text{tot}} approaches -1. This implies that the edge modes are somewhat spread over in a finite size system, which results in the deviation from 1-1 at ν=1\nu=1 in Figs. 2 and 4.

Refer to caption
Figure S1: Finite-size scaling analysis of the total jump ΔPtot\Delta P_{\text{tot}} at ν=1\nu=1. The parameters are fixed as Nx/Ny=2N_{x}/N_{y}=2, μ=3\mu=-3, and ϕ=1/10\phi=1/10. Toward NxN_{x}\rightarrow\infty, the data extrapolate to ΔPtot=0.98\Delta P_{\text{tot}}=-0.98.