# Full text of "Matrix-Product Ansatz for Excited States of Fractional Quantum Hall Systems"

## See other formats

Matrix- Pro duct Ansatz for Excited States of Fractional Quantum Hall Systems Zheng- Yuan Wang and Masaaki Nakamura Department of Physics, Tokyo Institute of Technology, 0-Okayama, Meguro-ku, Tokyo 152-8551, Japan E-mail: wang@stat.phys. titech. ac.jp, masaaki@stat.phys. titech. ac.jp (Received January 24, 2013) Recently, it was discussed that the v = 1/3 fractional quantum Hall state can be expressed by a one-dimensional lattice model with an exact matrix-prduct ground state, when toroidal boundary conditions are assumed for a narrow strip [Phys. Rev. Lett. 109 (2012) 016401]. In this article, we discuss how the excitation spectra of this system are calculated analytically based on the matrix-product formalism. We introduce variational wave functions for various momenta and optimize them. The obtained spectra of the charge neutral excitations show the magneto-roton behavior. KEYWORDS: fractional quantum Hall effect, thin-torus limit, matrix product state, exactly solvable model, excited states, magneto-roton 1. Introduction In recent years, there have been theoretical efforts to study the fractional quantum Hall (FQH) states around the thin-torus (or Tao-Thouless, TT) limit, where the original 2D system in a magnetic field with toroidal boundary conditions is reduced to ID lattice models with short-range interactions [1-4]. This approach is justified because the FQH states in this anisotropic system are known to be adiabatically connected to the isotropic cases [2]. From this point of view, the authors with Bergholtz have shown that the v = 1/3 Laughlin state [5] can be expressed by a ID model with an exact matrix-product (MP) ground state [6]. In this article, we give the details to obtain the excitation spectra of this model based on the MP wave function with variational parameters which partially mentioned in Ref. [6] , and generalize this calculation to arbitrally momenta and obtain roton behavior of the excitation spectra [7). 2. Fractional quantum Hall states on torus geometries We consider a model of N e interacting electrons in a magnetic field Bona torus boundary conditions with circumferences L\ and L 2 for x and y-directions, respectively. For simplicity, we normalize x,y,Li,L2 by magnetic length Is = y 'ch/eB. When the torus is pierced by N s magnetic flux quanta, boundary conditions require the relation L\L 2 = 2irN s . In Landau gauge (A = Byx), a complete basis of N s degenerate single-particle states in the lowest Landau level can be chosen as ipi(x,y) ^exp i (gi + L 2 j) x exp (1) 1 where the guiding center of the Gaussian function in y-direction is given by the quantized momentum in x-direction with i = 0, 1, • • • , N s — 1. In this basis, any translationally-invariant 2D Hamiltonian with two-body interactions assumes the following ID lattice model, ^ = X/ Vkrn, Vkm = ^ ^km C i+m C l+k C i+k+m C i ' ( 2 ) |m|<fe<Af s /2 i where the matrix-element V& m specifies the amplitude for a pair hopping process where two particles separated k + m sites hop m steps to opposite directions (note that m can be negative). The m = terms are the electrostatic repulsions, the matrix elements are real and antisymmetric functions for interchange of k and m. This model conserves the center of mass coordinate defined by K = Y^h=i ( m od N s ) with n = c\c i} which is identified as the total momentum in x-direction [see Eq. (1)]. For small L\ the matrix elements V^ m are simplified considerably: Since one finds that for generic interaction, the m/0 terms are exponentially suppressed as V km ~V M ^v{-^ 2 m 2 /L\). (3) Therefore, in the TT limit L\ — > 0, m 7^ terms in the original Hamiltonian (2) can be neglected, so that the Hamiltonian can be rewritten as Utt = Y^i Vko n i+j n j- Then the ground state is obtained at filling factor v = 1/3 is gapped, and three-fold degenerate charge density wave state with three sites periodicity, |^o) = 1 010| 010| 010 • • •). A system of thin but finite torus can be considered by taking the leading hopping terms Vkm ( m 7^ 0) into account for the TT state \^ ). 3. Matrix-product ground state for v = 1/3 Laughlin state The exactly solvable model for v = 1/3 states [6] can be obtained by truncating the Hamiltonian (2) up to the third nearest interactions [k + \m\ < 3) as an approximation in the vicinity of the TT limit. When we further assume L2 — > 00 limit, the matrix elements of the pseudo potential for v = 1/3 Laughlin state [8] satisfies the conditions = V10V30 and V20 > 0. Then the Hamiltonian is rewritten as = Yli[Q\Qi + -^/-^h where Qi = aoQ_|_iQ + 2 + aiCiQ+3 and Pi = f3oCiCi + 2 with ao,ai,/3o £ R. This Hamiltonian apparently has positive expectation values (^1/3) > 0, and the following wave function as the exact ground state I*) = II^ 1 + te f+i c !+2 c i+3 c i) l*o) , t = -ai/oo, (4) i since Qi 1^1/3) = P% ^1/3) = 0, Vi is satisfied. This is three-fold degenerate but can be shown to be unique ground state for each center-of-mass sectors. The wave function (4) can be written in a MP form 1*1/3) = tr[<7i<72 ■ ■ ■ 9N e ] [9,10] where the matrix gi is identified in the following way. First, we introduce the 5 = 1 spin representation for the three-sites unit cell as 1 010) — > |o), 1 001 ) — > |+) and 1 100) — > |— ). One finds that the possible configurations of the nearest two spins are |oo) + t \-\ — ) + |o+) + 1— o) + |— +), so that if we express the relations in the following 3x3 matrix 9i /__!-), f. \o) z /_+|+> fo-\-)i /oo|o)j /o+|+), /+_!-), j+oio), / ++ |+) (5) its non-vanishing elements are identified as f OQ = f Q+ = /_ Q = / h = 1 and / + _ = t. Next, we reduce the obtained 3x3 matrix to a 2 x 2 form by changing the base from (t |— ) , |o) , |+)) 2 to (t\-) + |o) ,|+)) as t\~)i (6) The matrix (6) has a 4 x 4 transfer matrix G = gj <8> gj with only two eigenvalues X± = (1 ± yj At 2 + 1) /2. The argument for the Hamiltonian with an exact MP wave function can also be extended to general v = 1/q Laughlin states [11]. 4. Variational wave functions for excited states We derive excitation spectra at v = 1/3 by a variational method. In the TT limit, the charge neutral excited states include one 01-type and one 0-type domain walls that carry fractional charges e* = e/3 and e* = — e/3, respectively. These states \^>^ K=1 ) = |010|0|01|010|010 - - ->, |*o^ =2 ) = |010|0|010|01|010 - - •>. are classified by the center of mass coordinate relative to the ground state which is related to the momentum of the in- direction. Now we consider the variational wave functions that consists of configurations generated from \^q~ k ) by applying the Hamiltonian %i/ 3 . For AK = 1, there are following exchange processes, |010|001|010|010 • • •) -»• |010|001|001|100 • • •) -»• 1 010| 000| 1 10| 100 • • •) -»• |010|001|001| 100 • • •). Since the interactions in are limited within the third neighbors, there is no exchange process across the defects. We can rewrite these processes in the 5 = 1 spin variables by inserting vacancies appropriately (between the two l's) [12,13] as 1010101 010 I 010 I 010- O OOO [010 1 1 001 1 010 I 100- •) -»• |010 |0| 010 1 001 1 100 - - -> - o o + — •) -> |010 |0| 010 1 001 1 100 - - -> . (7) We redefine the site indices so that the defect comes to the edge, and introduce an ansatz with variational parameters u and v as \ty AK=1 ) = tr[g[g 2 g 3 ■ ■ ■ 9m -i9'n\-> where 9i = u\o) 1 t\~) 1 w|o)i // _ 9N e = t\~) (8) For the case AK > 2, the ansatz should include three variational parameters u, v and w as l^ A ^> = tr bif 2 • • • 9 K -i9k9 k +2 • • • 9 Ne - 1 9N e ] with u \°+)i,i+i + v \+°)i,i+i in 9i = 00 tu\-o) iti+1 +tv\o-) tw\-+hi When we calculate the compressibility, we need to shrink the system size as N s (9) lby removing from the root state. Then only one 01-type domain wall (|- ■ • 010|010|01|010 • • •)) should be considered, so that the ansatz becomes |^ _ ) = tr[g'{'g 3 g 4 ■ ■ ■ g Ne _ 1 }- These variational wave functions for N e — > 00 can be optimized analytically by using the undetermined coefficient method. Fig. 1 (a) shows the excitation spectra obtained by the MP ansatz for the infinite system as functions of L\. These well agree with the results obtained from the infinite systems by exact diagonalization (ED). The deviation in the large L\ region is due to the finite size effect which appears through A_/A + . We also get a finite value of E(N S — 1) for iV e — > 00, reflecting that the Laughlin state is incompressible k~ 1 = lim 7Vs ^ 00 (iV s /47rZ2)[ J E(Ar s _i) + J E(Ar s + i)_2 J E s (iV s )] -> 00 where E(N S + 1) = E S {N S ) = 0. Fig. 1 (b) shows the neutral excitation spectra obtained by the MP ansatz for AK = 1-6. The AK = 2 state is always the lowest, and the excitation energy converges to a constant value when AK is increased as shown in the inset. This captures a character of the magneto-roton behavior in the FQH state [7]. These results are also consistent with a recent analysis based on spherical coordinate and the Jack polynomials [14]. AK=1 MP ED AK=2 MP ED E(N S - 1) MP ED 5 6 7 0.05 0.04 ty 0.03 < 0.02 0.01 0.00 (b) { \ (AK=1 2 ■ 3 4 5 J 6 : A£ 9 10 11 12 5 6 10 11 12 Fig. 1. (a) Comparison of excitation spectra obtained by the MP anzats for infinite systems and those by exact diagonalization (ED) for a finite system N s — 27. (b) Neutral excitation spectra for different momenta AK — 1-6 derived from the MP anzats for a infinite system. The inset shows the excitation spectra as function of AK at L± = 8 which show a magneto-roton minimum. 5. Conclusion We have studied the excitation spectra of the u = 1/3 FQH states considering toroidal boundary conditions in the vicinity of the Tao-Thouless limit. In this region, the original 2D system can be mapped onto a ID lattice model with short-range interactions which has the exact MP ground state. We have introduced "ansatz" for the excited state as the MP state with defects and variational parameters, and derived the excitation spectra analytically. The excitation spectrum with a fractional charge reflects the incompressibility of fractional quantum Hall states. The charge neutral excitations behave like the roton mode. 6. Acknowledgment We thank Emil J. Bergholtz for the related collaboration in Ref. [6]. Z.-Y. W. and M. N. acknowledge support from the Global Center of Excellence Program "Nanoscience and Quan- tum Physics" of the Tokyo Institute of Technology. M. N. also acknowledges support from Grant-in-Aid No.23540362 by MEXT. References [1] R. Tao and D. J. Thouless: Phys. Rev. B 28 (1983) 1142. [2] E. H. Rezayi and F. D. M. Haldane: Phys. Rev. B 50 (1994) 17199. [3] E. J. Bergholtz and A. Karlhede: Phys. Rev. Lett. 94 (2005) 026802. [4] A. Seidel, H. Fu, D. -H. Lee, J. M. Leinaas and J. Moore: Phys. Rev. Lett. 95 (2005) 266405. [5] R. B. Laughlin: Phys. Rev. Lett. 50 (1983) 1395. [6] M. Nakamura, Z. -Y. Wang and E. J. Bergholtz: Phys. Rev. Lett. 109 (2012) 016401. [7] S. M. Girvin, A. H. MacDonald and P. M. Platzman: Phys. Rev. Lett. 54 (1985) 581; Phys. Rev. B 33 (1986) 2481. [8] S. A. Trugman and S. A. Kivelson: Phys. Rev. B 31 (1985) 5280. [9] M. Fannes, B. Nachtergale and R. F. Werner: Europhys. Lett. 10 (1989) 633; Commun. Math. Phys. 144 (1992) 443. [10] A. Kliimper, A. Schadschneider and J. Zittartz: Z. Phys. B 87 (1992) 281; Europhys. Lett. 24 (1993) 293. [11] Z. -Y. Wang and M. Nakamura: arXiv:1206.3071v2. [12] M. Nakamura, Z. -Y. Wang and E. J. Bergholtz: J. Phys.: Conf. Ser. 302, 012020 (2011). [13] Z. -Y. Wang, S. Takayoshi and M. Nakamura: Phys. Rev. B 86 (2012) 155104. [14] B. Yang, Z.-X. Hu, Z. Papic and F. D. M. Haldane: Phys. Rev. Lett. 108 (2012) 256807. 4