betaFeedback

WorksPhysics and Astronomy

Conformal defects in a quasi-one-dimensional Kitaev chain: an exactly parametrized Ising defect line and a relevant vacancy

Abstract

The quasi-one-dimensional Kitaev chain of coupled hexagonal plaquettes recently introduced by Majumder is exactly solvable in terms of free Majorana fermions and undergoes a topological transition at the exchange anisotropy Jx=2 JJ_x=\sqrt{2}\,J. By eliminating the two-coordinated sites of each plaquette we reduce the zero-energy problem to a uniform Majorana chain, which yields the critical point, a single Majorana cone at k=0k=0 (hence an Ising critical point with central charge c=12c=\tfrac12), and closed forms for every defect quantity that follows. A weakened inter-plaquette bond λJxλJ_x is an exactly marginal defect with transmission T=4λ2/(1+λ2)2T=4λ^2/(1+λ^2)^2 at low energy, that is, T=sin⁡22ϕT=\sin^2 2ϕ with tan⁡ϕ=λ\tanϕ=λ: the bond traverses the ferromagnetic defect line of the Ising conformal field theory from the transparent point to the cut chain, and we give its exact energy dependence on the lattice. The intra-plaquette bonds land on the same line with different parametrizations: a yy or zz bond has tan⁡ϕ=(1+λ)/2\tanϕ=(1+λ)/2 and covers only T∈[0.64,1]T\in[0.64,1], because the plaquette provides a parallel path, whereas an xx bond has tan⁡ϕ=(1+λ)/2λ\tanϕ=(1+λ)/2λ and covers the strong-link half of the line, ϕ∈(π/4,π/2)ϕ\in(π/4,π/2). The identification is closed by an independent observable: the coefficient of ln⁡L\ln L in the entanglement entropy across the defect follows the conformal-field-theory function σ(T)/2σ(\sqrt T)/2 of Brehm and Brunner to within 0.4%0.4\% along the whole line, with no adjustable parameter. A vacancy behaves in the opposite way. It never binds a plaquette flux, because plaquettes share no sites, but it leaves its singly bonded neighbor as a conserved classical spin whose dangling Majorana is a zero-energy resonance.

Introduction

Defects in critical one-dimensional systems are classified by their fate under the renormalization group. An irrelevant defect heals, a relevant one cuts the system in two, and an exactly marginal one sits on a line of fixed points along which a universal quantity, the transmission coefficient, varies continuously [3, 4, 5] [1, 2]. For the Ising universality class the marginal line is the ferromagnetic defect line, known on the lattice since the work of Bariev and of McCoy and Perk on the two-dimensional Ising model with a modified row of bonds [6, 7, 8, 9], and whose boundary state in the folded theory was constructed by Oshikawa and Affleck [10]. Along this line the entanglement entropy across the defect carries a transmission-dependent prefactor, computed for free-fermion chains by Eisler and Peschel [11, 12] and from the conformal field theory by Brehm and Brunner [13] (see also Ref. [14]). Lattice realizations of the line are usually built on the transverse-field Ising chain itself, where a modified bond is the natural defect [15, 16]. When one asks instead what a vacancy, or a modified bond in a different microscopic realization of the same conformal field theory, does, the answer depends on the lattice, and the question of which lattice objects land on which conformal defect is a legitimate one.

Kitaev-type spin models provide realizations of the Ising conformal field theory that are microscopically far from the Ising chain: their critical points are Majorana cones of an itinerant fermion moving in a static Z2\mathbb{Z}_2 gauge field [17]. One-dimensional and ladder versions of the Kitaev model have been studied for some time [18, 19, 20, 21], and a new quasi-one-dimensional member of this family was introduced recently by Majumder [22]: a chain of hexagonal plaquettes cut out of the honeycomb lattice and linked by one xx bond per plaquette. The model keeps the bond-dependent exchange, the conserved plaquette operators, and the exact Majorana representation of the honeycomb model, and it has a chiral (class BDI [23]) topological transition at the exchange anisotropy Jx=JyJzJ_x = \sqrt{J_yJ_z} separating a trivial phase from one with Majorana end modes [22]. Because the plaquettes share no sites, the chain is structurally a cactus graph, and the objects that are natural defects in the honeycomb model—a vacancy, which there binds a flux [24, 25, 26], or a modified bond—take on a different character here.

In this paper we use this chain as a lattice laboratory for conformal defects, with everything computed exactly from free Majorana fermions, checked against exact diagonalization of the spin model, and, for the defect quantities, derived in closed form. The key to the closed forms is that the two-coordinated sites of a plaquette can be eliminated exactly, which reduces the zero-energy problem to a uniform Majorana chain (Appendix A). Our results are the following. (i) The transition at Jx=2JyJzJ_x=\sqrt{2J_yJ_z} is a single Majorana cone at k=0k=0 with linear dispersion, so it is an Ising critical point with c=12c=\frac{1}{2}; we confirm cc independently from the entanglement entropy. (ii) A weakened inter-plaquette bond is an exactly marginal defect with T=4λ2/(1+λ2)2T=4\lambda^2/(1+\lambda^2)^2 at low energy, which is the Ising ferromagnetic defect line with tan⁡ϕ=λ\tan\phi=\lambda, traversed from end to end; we give the exact energy dependence of TT on the lattice. The intra-plaquette bonds land on the same line with different parametrizations: a yy or zz bond has tan⁡ϕ=(1+λ)/2\tan\phi=(1+\lambda)/2 and covers only T≥0.64T\ge0.64, whereas an xx bond has tan⁡ϕ=(1+λ)/2λ\tan\phi=(1+\lambda)/2\lambda and covers the strong-link half of the line. The entanglement entropy across the defect follows the conformal-field-theory prefactor σ(T)/2\sigma(\sqrt{T})/2 along the whole line to within 0.4%, with no fitted parameter. (iii) A vacancy is a relevant defect: its transmission vanishes as T=E2/2T=E^2/2 for every site, through a zero-energy Majorana resonance on the neighbor that is left with a single bond, and the chain is cut in the infrared. There is no flux binding, by structure. In the topological phase the vacancy binds a Majorana zero mode. (iv) The clean ground state is exactly 4Np4^{N_p}-fold degenerate in the physical Hilbert space. The degeneracy comes from the 4Np+24N_p+2 dangling Majoranas of the two-coordinated sites, constrained to one fermion-parity sector, and we give the local operators that act within it. This is the lattice-level statement behind the 22N2^{2N} degeneracy noted in the extended Majorana space in Ref. [22], and it has a practical consequence for any numerical treatment of the chain, which we spell out.

The paper is organized as follows. Section II presents the model, its Majorana solution, and the identification of the critical point. Section III treats the weakened bond and the entanglement test, Sec. IV the vacancy, and Sec. V the degenerate ground-state manifold. Section VI summarizes. The exact reduction of the plaquette and the derivations of the closed forms are collected in Appendix A.

Model and Majorana solution

The model [22] is a chain of NpN_p hexagonal plaquettes with sites AA, BB, CC, DD, EE, FF Figure 1(a),

Figure 1. (a) The chain: hexagonal plaquettes with sites A,…,FA,\ldots,F and bond-dependent couplings JxJ_x (red), JyJ_y (blue), and JzJ_z (green); consecutive plaquettes are linked by a single xx bond D(p)D(p)–A(p+1)A(p+1), also of strength JxJ_x. Filled and open circles mark the two sublattices. Sites BB, CC, EE, and FF have only two bonds. (b) Majorana bands ε(k)\varepsilon(k) for Jr≡Jx/J=1J_r \equiv J_x/J=1, 2\sqrt{2}, and 22 (Jy=Jz=J=1J_y=J_z=J=1): two flat bands at ±2Jx\pm2J_x and, at Jr=2J_r=\sqrt{2}, a single cone at k=0k=0. (c) Bulk gap and winding number ∣ν∣|\nu| versus JrJ_r. (d) Weight per plaquette of the two zero modes of an open chain of Np=60N_p=60 plaquettes at Jr=2J_r=2.

H=∑p[JyσAyσBy+JxσBxσCx+JzσCzσDz+JyσDyσEy+JxσExσFx+JzσFzσAz]p+Jx∑pσD(p)xσA(p+1)x,(1)H=\sum_p\left[J_y\sigma_A^y\sigma_B^y+J_x\sigma_B^x\sigma_C^x+J_z\sigma_C^z\sigma_D^z+J_y\sigma_D^y\sigma_E^y+J_x\sigma_E^x\sigma_F^x+J_z\sigma_F^z\sigma_A^z\right]_p+J_x\sum_p\sigma_{D(p)}^x\sigma_{A(p+1)}^x, \tag*{(1)}

where σa\sigma^a are Pauli matrices. The inter-plaquette bond is an xx bond of the honeycomb lattice and carries the same coupling JxJ_x as the two intra-plaquette xx bonds. Following Ref. [22] we set Jy=Jz=J=1J_y=J_z=J=1 and use the anisotropy Jr≡Jx/JJ_r\equiv J_x/J as the control parameter; a weakened bond will be written as λ\lambda times its clean value.

Each site has at most one bond of each type, so the Kitaev construction applies [17]: σja=ibjacj\sigma_j^a=ib_j^ac_j with four Majoranas per site and the constraint Dj=bjxbjybjzbjxcj=1D_j=b_j^xb_j^yb_j^zb_j^xc_j=1; the bond operators ujk=ibkabkau_{jk}=ib_k^ab_k^a are conserved and

H=i4∑jkAjkcjck,Ajk=2Kjkujk,(2)H=\frac{i}{4}\sum_{jk}A_{jk}c_jc_k,\qquad A_{jk}=2K_{jk}u_{jk}, \tag*{(2)}

with KjkK_{jk} the coupling of the bond jkjk. The single-particle energies are the positive eigenvalues of the Hermitian matrix iAiA, whose hopping amplitudes are ta=2Jat_a=2J_a on a bond of type aa. The lattice is bipartite, with sublattices {A,C,E}\{A,C,E\} and {B,D,F}\{B,D,F\}, and the inter-plaquette bond joins the two sublattices, so the gauge ujk=+1u_{jk}=+1 from sublattice 1 to 2 is consistent and defines the flux-free sector. The plaquette operators Wp=σAxσBzσCyσDxσEzσFyW_p=\sigma_A^x\sigma_B^z\sigma_C^y\sigma_D^x\sigma_E^z\sigma_F^y commute with HH; the inter-plaquette bonds close no loop, so there is no other conserved flux, and the ground state has Wp=+1W_p=+1 on every plaquette [22]. We verified all of this against exact diagonalization of the spin Hamiltonian: the free-Majorana ground-state energy agrees with that of the spin model to nine digits for Np=1N_p=1 and 2 and for three sets of couplings, and ⟨Wp⟩=+1\langle W_p\rangle=+1 in the spin ground state. These checks fix the numerical factor in Eq. (2) and the gauge, and everything below rests on them.

The single-particle problem is the 6×66\times6 Bloch matrix h(k)=i∑dAdeikdh(k)=i\sum_d A_de^{ikd}. Its bands are shown in Figure 1(b). Two bands are flat at ϵ=±tx=±2Jx\epsilon=\pm t_x=\pm2J_x for every JrJ_r; in the units of Ref. [22], where the exchange is written with spin-12\frac{1}{2} operators, these are the flat bands at ±Jx\pm J_x. The remaining four bands disperse, and the gap closes only at Jr=2J_r=\sqrt{2}.

The structure of the spectrum has a simple origin, which we exploit throughout the paper (Appendix A). The sites B,C,E,FB,C,E,F have two bonds each and can be eliminated exactly at any energy EE. Each arm of the hexagon, AA–BB–CC–DD or AA–FF–EE–DD, then becomes an effective coupling between AA and DD of magnitude tytztx/∣E2−tx2∣t_yt_zt_x/\lvert E^2-t_x^2\rvert, together with energy-dependent on-site terms; the two arms carry the same amplitude in the gauge u=+1u=+1 and add. At E=0E=0 the on-site terms vanish and the plaquette collapses to a single bond of strength 2tytz/tx2t_yt_z/t_x between AA and DD, so the zero-energy problem is a two-site chain with alternating hoppings 2tytz/tx2t_yt_z/t_x (through the plaquette) and txt_x (between plaquettes). This chain is gapless if and only if the two hoppings are equal,

tx2=2tytz⟺Jx=2JyJz,(3)t_x^2=2t_yt_z\quad\Longleftrightarrow\quad J_x=\sqrt{2J_yJ_z}, \tag*{(3)}

which is the critical anisotropy found numerically in Ref. [22] and equals Jr=2J_r=\sqrt{2} for Jy=Jz=JJ_y=J_z=J. For Jx<2JyJzJ_x<\sqrt{2J_yJ_z} the intra-plaquette hopping dominates and the open chain is trivial; for Jx>2JyJzJ_x>\sqrt{2J_yJ_z} the inter-plaquette hopping dominates, the end sites A(1)A(1) and D(Np)D(N_p) are weakly attached, and the chain is topological with Majorana end modes, exactly as in the dimerized Kitaev wire [27]. The flat bands at ±tx\pm t_x are the energies at which the elimination is singular: they are the antisymmetric combinations of the inner sites of the two arms, which decouple from AA and DD. At the critical point the zero-energy chain is uniform with two sites per unit cell, so it has exactly one gap closing in the reduced Brillouin zone, at k=0k=0 in our convention: a single Majorana cone. Its velocity is reduced from the bare hopping txt_x by the energy dependence of the eliminated sites, v=tx/(tx2+ty2+tz2)1/2=Jxv=t_x/(t_x^2+t_y^2+t_z^2)^{1/2}=J_x for Jy=JzJ_y=J_z [Eq. (A3)], that is, v=2 Jv=\sqrt{2}\,J. A scan of 4×1044\times10^4 points across the Brillouin zone confirms all of this: at Jr=2J_r=\sqrt{2} the gap closes at exactly one momentum, k=0k=0, and the dispersion there is linear, ε=±v∣k∣\varepsilon=\pm v|k| with v/J=1.414214v/J=1.414214 constant over three decades of kk. Since k=0k=0 is its own time-reversal partner, the critical point is one free Majorana fermion, which is the Ising conformal field theory with c=12c=\frac{1}{2}. This is the fact on which the rest of the paper depends.

The chiral symmetry is the sublattice bipartition; in the sublattice basis the diagonal blocks of h(k)h(k) vanish, and the winding number of det⁡q(k)\det q(k) of the off-diagonal block is ν=0\nu=0 for Jr<2J_r<\sqrt{2} and ∣ν∣=1|\nu|=1 for Jr>2J_r>\sqrt{2} [Fig. 1(c)], as found in Ref. [22]. An open chain in the topological phase has two Majorana modes at ∣ε∣<10−15|\varepsilon|<10^{-15} with 98% of their weight in the three outermost plaquettes [Fig. 1(d)]. Close to the transition their localization length grows, and at Jr=1.5J_r=1.5 they are still hybridized (ε=3×10−4\varepsilon=3\times10^{-4}) in a chain of 360 sites.

A weakened bond realizes the Ising defect line

We now place a single defect in the critical chain and compute the transmission of a Majorana fermion through it, using the Fisher–Lee formula [28] with the surface Green’s functions of the two semi-infinite leads obtained by the decimation scheme of López Sancho et al. [29]. Without a defect, T=1.000000T=1.000000 from E=10−2E=10^{-2} down to 10−510^{-5}: there is a single channel, the cone. The reduction of Appendix A makes every result of this section analytic. At the critical point the surface Green’s function of the semi-infinite chain at its terminal site is exactly that of a uniform chain with hopping txt_x, g(E)=[E−i4tx2−E2]/2tx2g(E) = [E - i\sqrt{4t_x^2 - E^2}]/2t_x^2 [Eq. (14)], so a modified link between two half-chains is a textbook single-link scattering problem, and a modified bond inside a plaquette is the same problem with an effective link strength.

We first multiply the inter-plaquette bond between two given plaquettes by λ\lambda. The exact lattice transmission is [Appendix A, Eq. (15)]

Tinter(λ,E)=4λ2s2(1−λ2)2+4λ2s2,s2=1−E24tx2,(4)T_{\mathrm{inter}}(\lambda,E)=\frac{4\lambda^2s^2}{(1-\lambda^2)^2+4\lambda^2s^2},\qquad s^2=1-\frac{E^2}{4t_x^2}, \tag*{(4)}

which agrees with the numerical Fisher–Lee result to 10−910^{-9}, the accuracy set by the broadening, for all λ\lambda and all ∣E∣<2tx|E| < 2t_x. Its low-energy limit,

Tinter(λ)=4λ2(1+λ2)2,(5)T_{\mathrm{inter}}(\lambda)=\frac{4\lambda^2}{(1+\lambda^2)^2}, \tag*{(5)}

is approached with corrections of relative order E2/4tx2=E2/32E^2/4t_x^2 = E^2/32, below 10−610^{-6} for E≤10−2E \le10^{-2} [Figure 2(b)]: the defect neither heals nor cuts; it is exactly marginal, and T(λ)T(\lambda) labels a line of fixed points. Equation (5) covers the whole interval from T=1T=1 at λ=1\lambda=1 to T=0T=0 at λ=0\lambda=0, the cut chain [Figure 2(a)]. It is the same function of the bond factor as the transmission through a single modified hopping in the critical transverse-field Ising or XX chain [11, 12]: at low energy, the inter-plaquette bond behaves exactly as a bond defect of the Ising chain, even though the two lattices are very different away from the cone. Here this is not a coincidence but a consequence of Eq. (14).

Figure 2. Transmission through a weakened bond at the critical point Jr=2J_r = \sqrt{2}. (a) T(E→0)T(E \to0) versus the bond factor λ\lambda for the inter-plaquette xx bond, the intra-plaquette zz bond (CC–DD), and the intra-plaquette xx bond (BB–CC); symbols are the lattice Fisher–Lee result at E=10−4E = 10^{-4}, lines are the closed forms of Eqs. (5), (6), and (7). (b) T(E)T(E) for the inter-plaquette bond: constant over four decades, the signature of an exactly marginal defect; the exact lattice correction, Eq. (4), is below 10−610^{-6} in this window.

We next weaken a bond inside a plaquette. Since at E→0E \to0 the plaquette is a single effective link whose strength is the sum of the two arms (Appendix A), a bond factor λ\lambda on one arm turns the plaquette into a link of relative strength μ\mu in an otherwise uniform chain, and T=4μ2/(1+μ2)2T=4\mu^2/(1+\mu^2)^2. For the zz bond CC–DD (or, by the mirror symmetry of the plaquette, the zz bond FF–AA, and likewise for the two yy bonds) the arm strength is linear in λ\lambda and

Tintra(y,z)(λ)=4μ2(1+μ2)2,μ=1+λ2.(6)T_{\mathrm{intra}}^{(y,z)}(\lambda)=\frac{4\mu^2}{(1+\mu^2)^2},\qquad\mu=\frac{1+\lambda}{2}. \tag*{(6)}

Removing the bond altogether (λ=0\lambda=0) leaves T=0.64T=0.64 rather than 0: the other arm of the hexagon offers a parallel path, and the effective bond factor μ\mu is the arithmetic mean of the factors of the two arms, λ\lambda and 1. A yy or zz intra-plaquette bond therefore covers only the segment T∈[0.64,1]T\in[0.64,1] of the line. The intra-plaquette xx bonds (BB and EE–FF) are different: the xx bond sits in the middle of its arm, and the arm strength tytz/(λtx)t_y t_z/(\lambda t_x) grows as the bond is weakened, so

Tintra(x)(λ)=4μ2(1+μ2)2,μ=1+λ2λ,(7)T_{\mathrm{intra}}^{(x)}(\lambda)=\frac{4\mu^2}{(1+\mu^2)^2},\qquad\mu=\frac{1+\lambda}{2\lambda}, \tag*{(7)}

valid for E≪λtxE \ll\lambda t_x. Here μ≥1\mu\ge1 for λ≤1\lambda\le1, and μ→∞\mu\to\infty as λ→0\lambda\to0: weakening an intra-plaquette xx bond also cuts the chain, but through an effective link that is stronger than the bulk. All three closed forms agree with the lattice values to the accuracy of the broadening [Figure 2(a)].

All of these bonds realize the ferromagnetic defect line of the Ising conformal field theory. For that line the transmission is T=sin⁡22ϕT=\sin^2 2\phi, where ϕ∈[0,π/2]\phi\in[0,\pi/2] parametrizes the boundary state of the folded theory [10, 13], with ϕ=π/4\phi=\pi/4 the transparent point and ϕ=0,π/2\phi=0,\pi/2 the two cut chains. Equation (5) is sin⁡22ϕ\sin^2 2\phi with tan⁡ϕ=λ\tan\phi=\lambda: the inter-plaquette bond runs over ϕ∈[0,π/4]\phi\in[0,\pi/4]. The intra-plaquette yy and zz bonds have tan⁡ϕ=(1+λ)/2\tan\phi=(1+\lambda)/2 and run over ϕ∈[arctan⁡12,π/4]\phi\in[\arctan\frac{1}{2},\pi/4], and the intra-plaquette xx bonds have tan⁡ϕ=(1+λ)/2λ\tan\phi=(1+\lambda)/2\lambda and run over the complementary half ϕ∈(π/4,π/2)\phi\in(\pi/4,\pi/2). The three bond types of the lattice thus cover the whole line: the inter-plaquette bond alone covers the weak-link half ϕ∈[0,π/4]\phi\in[0,\pi/4], the intra-plaquette yy and zz bonds retrace the upper part of that half, and the intra-plaquette xx bonds are the only ones that reach the strong-link half ϕ∈(π/4,π/2)\phi\in(\pi/4,\pi/2). On this line the boundary entropy gg [30] is constant: the identity component of the boundary state has coefficient one for every ϕ\phi [Eq. (3.36) of Ref. [10]], consistent with an exactly marginal deformation, which cannot change gg. The universal quantity that does vary with the transmission is the prefactor of the entanglement entropy across the defect.

That prefactor provides an independent test of the identification. For the Ising defect line, the entanglement entropy of the two halves of an infinite chain joined by

λ\lambdaTTaa (lattice)σ(T)/2\sigma(\sqrt{T})/2
1.01.00000.083660.08333
0.90.98900.082930.08264
0.70.88280.076050.07588
0.50.64000.059480.05942
0.30.30300.033190.03321
0.10.03920.006220.00623

Table 1. Coefficient aa of ln⁡L\ln L in the entanglement entropy across the inter-plaquette defect (lattice, Np=128N_p=128–512512) compared with σ(T)/2\sigma(\sqrt{T})/2 from Eq. (8), with TT from Eq. (5).

the defect is S=σ(s)ln⁡L+constS=\sigma(s)\ln L+\mathrm{const} with s=Ts=\sqrt{T} and [13]

σ(s)=2π2∫0∞du(1+s2sinh⁡2u−1),(8)\sigma(s)=\frac{2}{\pi^2}\int_0^\infty du\left(\sqrt{1+\frac{s^2}{\sinh^2u}}-1\right), \tag*{(8)}

which reduces to σ(1)=16=c/3\sigma(1)=\frac{1}{6}=c/3 without a defect [31, 32, 33] and vanishes as −Tln⁡T-\sqrt{T}\ln T when the chain is cut; it coincides with one half of the effective central charge obtained by Eisler and Peschel for the XX chain [11, 12]. For an open chain cut at the defect the coefficient is halved. We computed the entanglement entropy of the free-Majorana ground state of open chains of Np=32N_p=32 to 512512 plaquettes, bipartitioned at the defect, from the correlation matrix [34], and fitted the coefficient of ln⁡L\ln L in Np=128N_p=128–512512. This is the entropy of the itinerant Majorana sector in the fixed gauge; the dangling modes discussed in § 1.4 contribute only size-independent constants and do not affect the coefficient of ln⁡L\ln L. Figure 3 compares the fitted coefficient with σ(T)/2\sigma(\sqrt{T})/2, with TT taken from Eq. (5). The agreement is between 0.1% and 0.4% along the entire line (Table I), the residual being of the size expected from 1/L1/L corrections in a three-size fit. At λ=1\lambda=1 the coefficient gives 12a=1.00412a=1.004, that is, c=12c=\frac{1}{2} again, by a route independent of the band structure. Three separate calculations—the scattering matrix, the entropy of the correlation matrix, and the conformal-field-theory integral—are tied together by the single lattice parameter λ\lambda with nothing adjusted. This closes the chain of identifications: single cone →\to Ising →\to ferromagnetic defect line →tan⁡ϕ=λ\to\tan\phi=\lambda.

Figure 3. Entanglement entropy across the inter-plaquette defect in an open critical chain of L=6NpL=6N_p sites, bipartitioned exactly at the defect. (a) ScutS_{\mathrm{cut}} versus LL for several λ\lambda. (b) Coefficient of ln⁡L\ln L, fitted in L=768L=768–30723072, versus the transmission T(λ)T(\lambda) of Eq. (5), compared with the conformal-field-theory prediction σ(T)/2\sigma(\sqrt{T})/2 of Ref. [13] Eq. (8). No parameter is adjusted.

A vacancy is a relevant defect

Removing a site removes four Majoranas and, in the free-fermion language, one row and one column of AA. The structure of the chain settles several things before any number is computed. Removing AA or DD removes the only inter-plaquette bond and cuts the chain literally, T=0T=0. Removing BB, CC, EE, or FF opens the hexagon into a path and the chain stays connected; by the symmetry A↔DA\leftrightarrow D, B↔EB\leftrightarrow E, C↔FC\leftrightarrow F there are two inequivalent connected vacancies, BB and CC. Unlike the honeycomb model, where a vacancy binds a flux by merging the three plaquettes around it [24, 25], here a vacancy cannot bind a flux: plaquettes share no sites, so the vacancy destroys its own plaquette instead of merging neighbors. What is left is a neighbor with a single bond (CC for a vacancy on BB, BB for a vacancy on CC), whose remaining Pauli component commutes exactly with HH (we checked [H,σCz]=0[H,\sigma_C^z]=0 in the spin model) and acts as a conserved classical spin, a longitudinal field of either sign on its one neighbor. In Majorana terms it is an itinerant Majorana cc hanging from the chain by a single bond. The free-Majorana energy with the vacancy, in the gauge u=+1u=+1, agrees with the spin exact diagonalization to 10−910^{-9} both for a vacancy at the end (Np=2N_p=2, 11 spins) and in the interior (Np=3N_p=3, 17 spins, Lanczos), so the description remains exact.

At the critical point the vacancy is not a conformal defect: the transmission vanishes at low energy for every site, including the ones that keep the chain connected, with the law

T(E)=E2/2(9)T(E) = E^2/2 \tag*{(9)}

(fitted exponent 2.0000, amplitude 0.4998 from E=3×10−2E=3\times10^{-2} to 3×10−43\times10^{-4}; Figure 4(a)). The mechanism is the dangling Majorana: a single Majorana coupled to the chain by one bond is a resonance at zero energy, and at E=0E=0 it reflects the incoming cone mode perfectly, as a Fano antiresonance [35]. The law follows from the reduction of Appendix A. The dangling site contributes a self-energy td2/Et_d^2/E to its neighbor, with td=tyt_d=t_y or tzt_z the bond that survives; the remaining arm of the damaged plaquette is a link of strength tytz/txt_yt_z/t_x; and at low energy the transmission through this link next to a site with a divergent on-site term is T=(2Etytz/txtd2)2T=(2Et_yt_z/t_xt_d^2)^2 [Eq. (A7)], which at the critical point with ty=tz=tdt_y=t_z=t_d reduces to T=4E2/tx2=E2/2T=4E^2/t_x^2=E^2/2. The same law holds for a vacancy on BB, CC, EE, or FF, since the four cases differ only by the mirror symmetries of the plaquette. In the language of the conformal field theory, a boundary Majorana γ\gamma coupled to the bulk fermion ψ(0)\psi(0) is an operator of dimension 12\frac{1}{2}, hence relevant, and flows to the cut fixed point. The two descriptions agree: a weakened bond is marginal and sits on a line, whereas a vacancy is relevant and cuts the chain in the infrared; they are different objects from the conformal point of view even though both are single-site defects on the lattice.

Figure 4. A single vacancy. (a) Transmission at the critical point through a vacancy on site BB or CC (the two inequivalent sites that keep the chain connected), compared with E2/2E^2/2. (b) Weight per plaquette of the zero modes of an open chain of Np=61N_p = 61 plaquettes with the vacancy on site CC of the central plaquette, at the critical point and in the topological phase.

In the topological phase the vacancy does bind a Majorana zero mode Figure 4(b). At Jr=2J_r=2 an open chain of 61 plaquettes with the vacancy on site CC of the central plaquette has three modes with ∣ϵ∣<10−6|\epsilon|<10^{-6}: the two end modes, and one with 98% of its weight in the five plaquettes around the vacancy and 67% on the single-bond site. At the critical point there is one exact zero mode, forced by the odd number 6Np−16N_p-1 of Majoranas, but it is delocalized (81% in the bulk, 0.8% on the single-bond site): there is no bound state at criticality.

The degenerate ground-state manifold

Exact diagonalization of the clean open chain gives a ground state that is exactly 4-fold degenerate for Np=1N_p=1 and 16-fold degenerate for Np=2N_p=2, that is, 4Np4^{N_p}-fold, separated from the next level by a gap of order JJ (0.54JJ and 0.49JJ, respectively); the degeneracy is preserved (up to the factor 2 of the classical spin) by a vacancy. It is not a flux degeneracy: flipping a WpW_p costs 0.3–0.7 JJ in all sizes and changes the single-particle spectrum. Its origin is the coordination. Sites BB, CC, EE, FF have only two bonds, so one Pauli component on each of them (σBz\sigma_B^z, σCy\sigma_C^y, σEz\sigma_E^z, σFy\sigma_F^y) never appears in HH: the corresponding bb Majorana is absent from Eq. (2) and is an exact zero mode of the extended problem, the same mechanism that produces the local conserved quantities of the one-dimensional Kitaev models of Refs. [18, 21]. This is the origin of the 22N2^{2N} degeneracy of the extended Majorana space noted in Ref. [22]; the statement here is that the degeneracy survives the projection Dj=1D_j=1 to the physical spin space [36, 37], where it is 4Np4^{N_p}, and that the count is exact. On an open chain the end sites A(1)A(1) and D(Np)D(N_p) also lack their inter-plaquette bond and contribute a dangling bxb^x each, for a total of 4Np+24N_p+2 dangling Majoranas, that is, 2Np+12N_p+1 fermion modes; the global constraint ∏jDj=1\prod_jD_j=1, which acts trivially on the uu’s, fixes the parity of these modes jointly with that of the itinerant sector, and the physical ground manifold is the fixed-parity sector, of dimension 22Np+1/2=4Np2^{2N_p+1}/2=4^{N_p}.

The operators that move within the manifold are the gauge-invariant bilinears of two dangling Majoranas dressed by the bond string between them. They are short products of Pauli matrices that commute exactly with HH, and we found the complete set by direct search: within each plaquette, σByσCz\sigma_B^y\sigma_C^z and σEyσFz\sigma_E^y\sigma_F^z; through the three-coordinated sites, σCxσDy\sigma_C^x\sigma_D^y and σFxσGx\sigma_F^x\sigma_G^x; across the inter-plaquette bond, \sigma_C^x_{(p)}\sigma_D^y_{(p)}\sigma_A^z_{(p+1)}\sigma_B^x_{(p+1)}; and at the two ends, σBzσFx\sigma_B^z\sigma_F^x, σDyσFx\sigma_D^y\sigma_F^x and σCxσDy\sigma_C^x\sigma_D^y, σDzσEx\sigma_D^z\sigma_E^x. Applied repeatedly to any one ground state they generate the full 4Np4^{N_p}-dimensional manifold, which we verified numerically up to Np=3N_p=3 (64 states, residual 10−1110^{-11}).

None of the defect results above depends on this manifold: the transmission and the ln⁡L\ln L coefficient of the entropy are properties of the itinerant cc sector, and the dangling modes add only size-independent constants to the entropy. The practical consequence is for numerics. A density-matrix renormalization-group calculation on this chain converges to an arbitrary vector of the manifold: we find that three random starts at Np=2N_p=2 reach the same energy to 10−1010^{-10} but three different states, with mutual overlaps of 0.3. Any dynamical or correlation quantity computed from “the” ground state of Eq. (1), such as the structure factors of Ref. [22], depends on which vector of the manifold the algorithm happened to reach, unless the degeneracy is lifted by a perturbation or the observable is manifestly independent of the dangling sector. How a uniform field lifts the manifold is a question in its own right, since the first-order matrix elements vanish and the lifting occurs at second order through the critical itinerant sector; we treat it in a separate paper.

Summary and outlook

The quasi-one-dimensional Kitaev chain realizes the Ising conformal field theory through a single Majorana cone, and its bonds realize the Ising defect line with analytic parametrizations that follow from a single exact step, the elimination of the two-coordinated sites of each plaquette: tan⁡ϕ=λ\tan\phi=\lambda for the inter-plaquette bond, which covers the line from transparency to the cut; tan⁡ϕ=(1+λ)/2\tan\phi=(1+\lambda)/2 for an intra-plaquette yy or zz bond, which covers only T≥0.64T\ge0.64 because the other arm of the plaquette provides a parallel path; and tan⁡ϕ=(1+λ)/2λ\tan\phi=(1+\lambda)/2\lambda for an intra-plaquette xx bond, which covers the strong-link half of the line. The same step gives the critical anisotropy Jx=2JyJzJ_x=\sqrt{2J_yJ_z}, the velocity of the cone, and the exact energy dependence of the inter-plaquette transmission on the lattice. The conformal prefactor of the entanglement entropy follows the lattice transmission along the whole line to within 0.4% with nothing fitted. A vacancy is the opposite kind of object: it binds no flux, it leaves a conserved classical spin whose dangling Majorana is a zero-energy resonance, and it cuts the chain in the infrared with T=4E2/tx2=E2/2T=4E^2/t_x^2=E^2/2, binding a Majorana zero mode only in the topological phase. The clean ground state is exactly 4Np4^{N_p}-fold degenerate in the physical Hilbert space, a fact that leaves the defect physics untouched but that any numerical treatment of the chain has to confront.

Two of these statements are, we believe, general beyond this chain. First, the transmission of a weakened link between two segments of a Kitaev-type chain whose critical point is a single cone is the Ising-chain law 4λ2/(1+λ2)24\lambda^2/(1+\lambda^2)^2, independently of the microscopic structure of the segments; the lattice enters only through the effective link factor, as the two intra-plaquette results illustrate. Second, in any Kitaev-type lattice in which plaquettes share no sites, a vacancy cannot bind a flux and instead leaves a dangling itinerant Majorana, so it is a relevant rather than a marginal defect. Kitaev ladders and other cactus-like geometries [18, 19, 20, 21] offer further tests of both statements.

Acknowledgments

A. R. P. thanks INCT/CNPq – Spintrônica e Nanoestruturas Magnéticas Avançadas (INCT-SpinNanoMag) and Rede Mineira de Nanomagnetismo/FAPEMIG. This work was supported by the Brazilian agencies CNPq, FAPEMIG, CAPES and FAPES.

Exact reduction of the plaquette and closed forms

In the gauge u=+1u=+1 the matrix iAiA of Eq. (2) is a Hermitian hopping matrix with amplitudes of magnitude ta=2Jat_a=2J_a on the bonds of type a=x,y,za=x,y,z and zero on-site energies. We write hh for hh and EE for the energy; the resolvent is G(E)=(E+i0+−h)−1G(E)=(E+i0^+-h)^{-1}.

Elimination of an arm. Consider a path AA–BB–CC–DD with amplitudes hABh_{AB}, hBCh_{BC}, hCDh_{CD} and no other bonds on BB and CC. Eliminating BB and CC at energy EE (Gaussian elimination on the resolvent, or equivalently the Schur complement) adds to the remaining sites the energy-dependent terms

ΣAA=∣hAB∣2EE2−∣hBC∣2,ΣDD=∣hCD∣2EE2−∣hBC∣2,\Sigma_{AA}=\frac{\lvert h_{AB}\rvert^2E}{E^2-\lvert h_{BC}\rvert^2},\qquad \Sigma_{DD}=\frac{\lvert h_{CD}\rvert^2E}{E^2-\lvert h_{BC}\rvert^2},
ΣAD=hABhBChCDE2−∣hBC∣2.(10)\Sigma_{AD}=\frac{h_{AB}h_{BC}h_{CD}}{E^2-\lvert h_{BC}\rvert^2}. \tag*{(10)}

A plaquette has two such arms between AA and DD, AA–BB–CC–DD with bond types (y,x,zy,x,z) and AA–FF–EE–DD with bond types (z,x,yz,x,y). Both run from sublattice 1 to 2 to 1 to 2, so in the gauge u=+1u=+1 their amplitude products hABhBChCDh_{AB}h_{BC}h_{CD} and hAFhFEhEDh_{AF}h_{FE}h_{ED} are equal, and the two contributions to ΣAD\Sigma_{AD} add. With Jy=JzJ_y=J_z the plaquette therefore reduces to the two sites AA and DD with

ϵ(E)=(ty2+tz2)EE2−tx2,∣teff(E)∣=2tytxtz∣E2−tx2∣(11)\epsilon(E)=\frac{(t_y^2+t_z^2)E}{E^2-t_x^2},\qquad\lvert t_{\mathrm{eff}}(E)\rvert=\frac{2t_yt_xt_z}{\lvert E^2-t_x^2\rvert} \tag*{(11)}

on both sites and between them, respectively, while the inter-plaquette bond D(p)D(p)–$A(p+1)A(p+1) keeps its amplitude txt_x. The elimination is singular at E=±txE=\pm t_x; these are the two flat bands of Figure 1(b), the combinations of the inner sites of the two arms that are antisymmetric under the mirror exchanging the arms and therefore decouple from AA and DD.

Critical point, single cone, and velocity. At E=0E=0 the on-site terms vanish and ∣teff(0)∣=2tytz/tx\lvert t_{\mathrm{eff}}(0)\rvert=2t_yt_z/t_x. The zero-energy problem is a chain with two sites per cell and alternating hoppings 2tytz/tx2t_yt_z/t_x and txt_x, which is gapless if and only if they are equal, Eq. (3). Then the chain is uniform with two sites per unit cell of length 1 and the off-diagonal Bloch element is tx(1−eik)t_x(1-e^{ik}) up to phases, whose modulus 2tx∣sin⁡(k/2)∣2t_x\lvert\sin(k/2)\rvert vanishes once, at k=0k=0: a single cone. Away from E=0E=0 the effective problem is det⁡[E−ϵ(E)−heff(k,E)]=0\det[E-\epsilon(E)-h_{\mathrm{eff}}(k,E)]=0; to first order in EE, ϵ(E)≃−(ty2+tz2)E/tx2\epsilon(E)\simeq-(t_y^2+t_z^2)E/t_x^2 and ∣teff(E)∣≃2tytz/tx\lvert t_{\mathrm{eff}}(E)\rvert\simeq2t_yt_z/t_x, so E[1+(ty2+tz2)/tx2]=±tx∣k∣E[1+(t_y^2+t_z^2)/t_x^2]=\pm t_x\lvert k\rvert and

v=tx3tx2+ty2+tz2=tx2=Jx(Jy=Jz, tx2=2tytz),(12)v=\frac{t_x^3}{t_x^2+t_y^2+t_z^2}=\frac{t_x}{2}=J_x\qquad(J_y=J_z,\ t_x^2=2t_yt_z), \tag*{(12)}

which is v=2 Jv=\sqrt{2}\,J at Jr=2J_r=\sqrt{2}, as found numerically in Sec. II.

Surface Green’s function. Let g(E)g(E) be the resolvent at the terminal site DD of a semi-infinite critical chain that ends with a complete plaquette. Using Eq. (11), gg obeys the two-step recursion of the effective alternating chain, g=[E−ϵ−teff2/(E−ϵ−tx2g)]−1g=[E-\epsilon-t_{\mathrm{eff}}^2/(E-\epsilon-t_x^2g)]^{-1}, that is,

xtx2g2−(x2−teff2+tx2)g+x=0,x=E−ϵ(E).(13)xt_x^2g^2-(x^2-t_{\mathrm{eff}}^2+t_x^2)g+x=0,\qquad x=E-\epsilon(E). \tag*{(13)}

Substituting Eq. (11) with ty=tzt_y=t_z and tx2=2tytzt_x^2=2t_yt_z, one verifies by direct polynomial reduction that the solution of Eq. (13) coincides identically in EE with the surface Green’s function of a uniform chain of hopping txt_x,

g(E)=E−i4tx2−E22tx2,∣E∣<2tx.(14)g(E)=\frac{E-i\sqrt{4t_x^2-E^2}}{2t_x^2},\qquad\lvert E\rvert<2t_x. \tag*{(14)}

The lead that ends at a site AA has the same gg, because the mirror A↔DA\leftrightarrow D, B↔EB\leftrightarrow E, C↔FC\leftrightarrow F maps hh to −h-h and the sublattice (chiral) symmetry maps hh to −h-h as well; their combination is a symmetry of hh that exchanges the two terminations. We have checked Eq. (14) against the numerical decimation to 10−910^{-9} over ∣E∣≤1\lvert E\rvert\leq1. For ty≠tzt_y\ne t_z the identity holds only at E=0E=0, where g(0)=−i/txg(0)=-i/t_x, which is all that the defect classification requires.

Inter-plaquette bond. A single link of amplitude τ\tau between two leads with terminal Green’s functions gL=gR=gg_L=g_R=g transmits T=4τ2(Im⁡g)2/∣1−τ2g2∣2T=4\tau^2(\operatorname{Im}g)^2/\lvert1-\tau^2g^2\rvert^2. With τ=λtx\tau=\lambda t_x and Eq. (14) this is

Tinter(λ,E)=4λ2s2(1−λ2)2+4λ2s2,s2=1−E24tx2.(15)T_{\mathrm{inter}}(\lambda,E)=\frac{4\lambda^2s^2}{(1-\lambda^2)^2+4\lambda^2s^2},\qquad s^2=1-\frac{E^2}{4t_x^2}. \tag*{(15)}

the transmission of a single modified hopping in a uniform chain, which reduces to Eq. (5) at E→0E\to0.

Intra-plaquette bonds. At E→0E\to0 a bond factor λ\lambda on one bond of one arm multiplies that arm’s contribution to teff(0)t_{\mathrm{eff}}(0) by λ\lambda if the bond is yy or zz [the factor enters hABh_{AB} or hCDh_{CD} in Eq. (10)] and by 1/λ1/\lambda if it is the middle xx bond [the factor enters hBCh_{BC} in both numerator and denominator, and E2−λ2tx2→−λ2tx2E^2-\lambda^2t_x^2\to-\lambda^2t_x^2 for E≪λtxE\ll\lambda t_x]. The plaquette becomes a link of strength μtx\mu t_x with μ=(1+λ)/2\mu=(1+\lambda)/2 or μ=(1+λ)/2λ\mu=(1+\lambda)/2\lambda, respectively, in an otherwise uniform chain, and Eq. (15) at E→0E\to0 with λ→μ\lambda\to\mu gives Eq. (6) and Eq. (7). For the xx bond the on-site terms ΣAA,ΣDD∝E/(E2−λ2tx2)\Sigma_{AA},\Sigma_{DD}\propto E/(E^2-\lambda^2t_x^2) of the weakened arm are negligible only for E≪λtxE\ll\lambda t_x, which sets the range of validity quoted in the text: the lattice values in Figure 2(a) were computed at E=10−4E=10^{-4} and agree with Eq. (7) down to λ=0.05\lambda=0.05.

Vacancy. Removing CC (the case of BB, EE, or FF follows by the mirrors) leaves BB attached to AA by the single bond td=tyt_d=t_y, and the arm AA–FF–EE–DD as the only path through the plaquette, of strength tytztx/txt_yt_zt_x/t_x at low energy. The dangling site contributes ΣAA=td2/E\Sigma_{AA}=t_d^2/E, which diverges as E→0E\to0. Treating the two sites AA and DD as the scattering region between leads with g=−i/txg=-i/t_x, the Fisher–Lee transmission is T=4tx2∣GAD∣2T=4t_x^2\lvert G_{AD}\rvert^2 with GAD=tarm/[(E−td2/E+itx)(E+itx)−tarm2]G_{AD}=t_{\mathrm{arm}}/[(E-t_d^2/E+it_x)(E+it_x)-t_{\mathrm{arm}}^2], and to leading order in EE

T(E)=(2Etytztxtd2)2=4E2tx2=E22,(16)T(E)=\left(\frac{2Et_yt_z}{t_xt_d^2}\right)^2=\frac{4E^2}{t_x^2}=\frac{E^2}{2}, \tag*{(16)}

where the second equality uses ty=tz=tdt_y=t_z=t_d and tx2=2tytzt_x^2=2t_yt_z, and the third tx=22t_x=2\sqrt{2}. Equation (16) is the exponent-two law of Eq. (9) with the amplitude 12\frac{1}{2} observed in Figure 4(a).

References

References

  1. [3]S. Eggert and I. Affleck, Magnetic impurities in half-integer-spin Heisenberg antiferromagnetic chains, Phys. Rev. B 46, 10866 (1992).
  2. [4]C. Bachas, J. de Boer, R. Dijkgraaf, and H. Ooguri, Permeable conformal walls and holography, J. High Energy Phys. 06, 027.
  3. [5]T. Quella, I. Runkel, and G. M. T. Watts, Reflection and transmission for conformal defects, J. High Energy Phys. 04, 095.
  4. [6]R. Z. Bariev, Effect of linear defects on the local magnetization of a plane Ising lattice, Sov. Phys. JETP 50, 613 (1979), [Zh. Eksp. Teor. Fiz. 77, 1217 (1979)].
  5. [7]B. M. McCoy and J. H. H. Perk, Two-spin correlation functions of an Ising model with continuous exponents, Phys. Rev. Lett. 44, 840 (1980).DOI
  6. [8]M. Henkel, A. Patkós, and M. Schlottmann, The Ising quantum chain with defects (i), the exact solution, Nucl. Phys. B 314, 609 (1989).DOI
  7. [9]G. Delfino, G. Mussardo, and P. Simonetti, Scattering theory and correlation functions in statistical models with a line of defect, Nucl. Phys. B 432, 518 (1994).
  8. [10]M. Oshikawa and I. Affleck, Boundary conformal field theory approach to the critical two-dimensional Ising model with a defect line, Nucl. Phys. B 495, 533 (1997).
  9. [11]V. Eisler and I. Peschel, Entanglement in fermionic chains with interface defects, Ann. Phys. (Berlin) 522, 679 (2010).DOI
  10. [12]I. Peschel and V. Eisler, Exact results for the entanglement across defects in critical chains, J. Phys. A: Math. Theor. 45, 155301 (2012).
  11. [13]E. M. Brehm and I. Brunner, Entanglement entropy through conformal interfaces in the 2d Ising model, J. High Energy Phys. 09, 080.
  12. [14]K. Sakai and Y. Satoh, Entanglement through conformal interfaces, J. High Energy Phys. 12, 001.DOI
  13. [15]I. Peschel, Entanglement entropy with interface defects, J. Phys. A: Math. Gen. 38, 427 (2005).DOI
  14. [16]F. Iglói, Z. Szatmári, and Y.-C. Lin, Entanglement entropy with localized and extended interface defects, Phys. Rev. B 80, 024405 (2009).DOI
  15. [17]A. Kitaev, Anyons in an exactly solved model and beyond, Ann. Phys. (N.Y.) 321, 2 (2006).DOI
  16. [18]X.-Y. Feng, G.-M. Zhang, and T. Xiang, Topological characterization of quantum phase transitions in a spin-1/2 model, Phys. Rev. Lett. 98, 087204 (2007).
  17. [19]A. Saket, S. R. Hassan, and R. Shankar, Manipulating unpaired Majorana fermions in a quantum spin chain, Phys. Rev. B 82, 174409 (2010).DOI
  18. [20]A. Saket, S. R. Hassan, and R. Shankar, Topological aspects of an exactly solvable spin chain, Phys. Rev. B 87, 174414 (2013).DOI
  19. [21]W. DeGottardi, D. Sen, and S. Vishveshwara, Topological phases, Majorana modes and quench dynamics in a spin ladder system, New J. Phys. 13, 065028 (2011).
  20. [22]R. Majumder, Fractional excitations in Kitaev quasi-one-dimensional chain (2026), arXiv:2606.20309 [cond-mat.str-el].arxiv.org/abs/2606.20309
  21. [23]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
  22. [24]A. J. Willans, J. T. Chalker, and R. Moessner, Disorder in a quantum spin liquid: Flux binding and local moment formation, Phys. Rev. Lett. 104, 237203 (2010).
  23. [25]A. J. Willans, J. T. Chalker, and R. Moessner, Site dilution in the Kitaev honeycomb model, Phys. Rev. B 84, 115146 (2011).DOI
  24. [26]G. Santhosh, V. Sreenath, A. Lakshminarayan, and R. Narayanan, Localized zero-energy modes in the Kitaev model with vacancy disorder, Phys. Rev. B 85, 054204 (2012).DOI
  25. [27]A. Y. Kitaev, Unpaired Majorana fermions in quantum wires, Phys. Usp. 44, 131 (2001).DOI
  26. [28]D. S. Fisher and P. A. Lee, Relation between conductivity and transmission matrix, Phys. Rev. B 23, 6851 (1981).DOI
  27. [29]M. P. López Sancho, J. M. López Sancho, and J. Rubio, Highly convergent schemes for the calculation of bulk and surface Green functions, J. Phys. F: Met. Phys. 15, 851 (1985).
  28. [30]I. Affleck and A. W. W. Ludwig, Universal noninteger “ground-state degeneracy” in critical quantum systems, Phys. Rev. Lett. 67, 161 (1991).pubmed.ncbi.nlm.nih.gov/10044510
  29. [31]C. Holzhey, F. Larsen, and F. Wilczek, Geometric and entanglement entropy in conformal field theory, Nucl. Phys. B 424, 443 (1994).DOI
  30. [32]G. Vidal, J. I. Latorre, E. Rico, and A. Kitaev, Entanglement in quantum critical phenomena, Phys. Rev. Lett. 90, 227902 (2003).pubmed.ncbi.nlm.nih.gov/12857342
  31. [33]P. Calabrese and J. Cardy, Entanglement entropy and quantum field theory, J. Stat. Mech. 2004, P06002 (2004).DOI
  32. [34]I. Peschel, Calculation of reduced density matrices from correlation functions, J. Phys. A: Math. Gen. 36, L205 (2003).DOI
  33. [35]U. Fano, Effects of configuration interaction on intensities and phase shifts, Phys. Rev. 124, 1866 (1961).DOI
  34. [36]F. L. Pedrocchi, S. Chesi, and D. Loss, Physical solutions of the Kitaev honeycomb model, Phys. Rev. B 84, 165414 (2011).
  35. [37]F. Zschocke and M. Vojta, Physical states and finite-size effects in Kitaev’s honeycomb model: Bond disorder, spin excitations, and NMR line shape, Phys. Rev. B 92, 014403 (2015).DOI

Paper details

vlnc:W_x4cb6wqz
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 8

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