Suppression of finite-size effects in one-dimensional correlated systems

Suppression of finite-size effects in one-dimensional correlated systems

A. Gendiar, M. Daniška, Y. Lee, and T. Nishino Institute of Physics, Slovak Academy of Sciences, SK-845 11, Bratislava, Slovakia
Institute of Electrical Engineering, Slovak Academy of Sciences, SK-841 04, Bratislava, Slovakia
Department of Physics, Graduate School of Science, Kobe University, Kobe 657-8501, Japan
Department of Nuclear Physics and Biophysics, Faculty of Mathematics, Physics and Informatics, Comenius University, SK-842 48 Bratislava, Slovakia
July 11, 2019

We investigate the effect of a non-uniform deformation applied to one-dimensional (1D) quantum systems, where the local energy scale is proportional to determined by a positive integer , site index , and the system size . This deformation introduces a smooth boundary to systems with open boundary conditions. When , the leading correction to the ground state energy per bond vanishes and one is left with a correction, the same as with periodic boundary conditions. In particular, when , the value of obtained from the deformed open-boundary system coincides with the uniform system with periodic boundary conditions. We confirm the fact numerically for correlated systems, such as the extended Hubbard model, in addition to 1D free-Fermion models.

03.65.Aa, 05.30.Fk, 71.10.Fd

I Introduction

The periodic boundary conditions (PBC) are often more convenient than the open boundary conditions (OBC), when asymptotic properties of one-dimensional (1D) quantum systems are studied in the thermodynamic limit. This is partially because boundary energy corrections exist under OBC, where eigenstates are not translational invariant. Systems with PBC normally contain smaller finite size effects, and this property of PBC is appropriate for accurate determination of bulk properties by means of the finite size scaling. Fisher (); Barber (); fss ()

In numerical studies of lattice models, OBC are often chosen for technical reasons. In particular, majority of the practical numerical analyses by the density matrix renormalization group (DMRG) method White (); Peschel (); Schol () are performed under OBC. Concerning to finite size systems with PBC, the crucial point in DMRG is the ring-shaped geometry, which reduces the decay rate of the singular values. Ver () Although recent progress in DMRG and the tensor product formalisms made it possible to include the PBC in a natural manner, Ver (); Gerd (); Rossini1 (); Rossini2 () numerical implementation requires additional computational resources compared with conventional DMRG analyses.

A way of suppressing the boundary energy corrections induced by OBC is to introduce smooth boundary conditions (SBC). Vekic (); Vekic2 () Recently we proposed a variant of the smooth boundary conditions, where the local energy scale of -site systems is proportional to the deformation function specified by the site index ptp1 (); ptp2 () This sine-squared deformation (SSD) Hikihara () completely suppresses the boundary effects when the ground-state energy of a free-Fermions model on the 1D lattice is considered. In this article we generalize the deformation function, which is given by , where is a positive integer. Hyperbolic () In the next section, we examine the effect of this sinusoidal deformation by up to when it is applied to the 1D free-Fermion model. It is shown that the case , the SSD, is the most efficient for the suppression of the boundary effect.

Another trial in this article is the application of SSD to correlated systems. As typical examples of correlated systems, we choose the spinless-Fermion model with nearest-neighbor interaction and the extended Hubbard model; we report the numerical result obtained by DMRG in Sec. III. When the interaction is present, the determination of the chemical potential is non trivial. We present a systematic way of solving this problem in Sec. IV. We summarize results in the last section.

Ii Sinusoidal Deformation

We start from the sinusoidal deformation applied to the free-Fermion model on the 1D lattice. Consider a tight-binding model represented by the Hamiltonian


where is the system size, and the hopping energy. Operators and represent creation and annihilation of Fermions, respectively. The parameter specifies the boundary condition, where OBC and PBC correspond to and , respectively. (The choice is known as the anti-periodic boundary conditions, which we do not treat in this article.) For each boundary condition, the one-particle energy is expressed as


where the energy index runs from to . The ground-state energy at half filling is obtained as the sum of below the Fermi energy .

Throughout this article we focus on the system size dependence on the energy per site or the energy per bond, which is under PBC and is under OBC. After a short algebra, one obtains


with OBC, where the leading order of the finite size correction is proportional to . The correction of the same order also exists for the energy per bond. With PBC, one finds


where the leading correction is of the order of . The difference between Eq. (3) and Eq. (4) chiefly comes from the presence of the boundary energy which exists only when OBC are imposed.

The sinusoidal deformation introduces a position dependent energy scale to each bond of the system with OBC, where is the positive integer. ptp1 (); ptp2 () Deforming in Eq. (1), we obtain the corresponding free-Fermionic Hamiltonian


We have not obtained analytical solution for the one-particle spectrum of so far, except for the zero-energy state. Thus we perform numerical analyses in the following investigations on the ground state.

Since we are interested in the ground-state energy per site (or per bond), we introduce the normalization factor


which is the sum of the deformation factors over the entire system. When is an odd positive integer, we have


and when is an even positive integer, we have


We represent the ground-state energy of at half filling by the notation . It is expected that the normalized energy


converges to in the large limit in analogy to Eqs. (3) and (4). We refer in Eq. (9) to as the energy per bond in the following. As a convention, we set for the system with OBC, and with PBC, where these values just represent the number of bonds. Using this extended definition of , we can represent energy per bond by Eq. (9) regardless of the boundary condition or the presence of deformation.

Figure 1: (Color online) Convergence of in Eq. (9) with respect to at half filling. Data with OBC, PBC, and the deformed cases with to are shown.

We regard as the unit of the energy in the numerical analyses. Figure 1 shows the -dependence of in Eq. (9) for the undeformed systems with OBC, PBC, and the deformed systems from to . When the PBC are imposed, the convergence of with respect to is linear, and there is even-odd oscillation with respect to the particle number . Similarly the linear -dependence is observed when under the sinusoidal deformation. In the case , there is additional logarithmic correction as shown later. It should be noted that when the particle number is odd, obtained with the sinusoidal deformation for coincides with obtained with PBC. AP () This complete agreement is checked down to the smallest digit in numerical precision. Throughout this paper we use the exact diagonalization in order to reduce any numerical errors to minimum.

Figure 2: (Color online) Asymptotic behavior of for the deformed chains when with respect to of the system with PBC. A logarithmic correction is present when .

In order to confirm the -dependence of with the sinusoidal deformation under , we plot the difference between obtained with PBC (when is even) and with the sinusoidal deformation. To avoid any confusion, let and denote the ground-state energy obtained with PBC and with the sinusoidal deformation, respectively. We also use a similar notation for the normalization factors and for the normalization factor defined in Eq. (6). Figure 2 depicts the magnified difference


when is even. It is shown that the logarithmic correction is present when , and is absent when .

Figure 3: (Color online) Expectation value of the bond correlation function with respect to under sinusoidal deformation.

Figure 3 shows the spatial distribution of the bond correlation function at half filling when . The Friedel oscillations induced by the boundary are clearly observed when OBC are imposed (the asterisks), and weaker oscillations are observed with the sinusoidal deformation when . Only when , there are no oscillations at all; we checked the uniformity (the translation invariance) of the bond correlation function down to the 16-digits in numerical precision. When , the boundary effects appear again. In this case the bond correlation function toward the system boundary does not oscillate, and decreases in monotonic manner. Such behaviors for each might be related to the suppression of the boundary corrections in .

We compare the efficiency of SSD () with the SBC proposed in Ref. Vekic (); Vekic2 (). Figure 4 shows the bond correlation function for both cases at half-filling, where the length of boundary area in SBC is chosen as and when the system size is . Although bulk property is well captured by SBC already for , boundary fluctuations are still present. On the other hand, the bond correlation function is almost uniform away of the boundary.

Figure 4: (Color online) Comparison of the expectation value of the bond correlation function at half filling under SSD () and SBC in Ref. Vekic (); Vekic2 (). The number in the case of SBC specifies the length of area where the interactions are modified near the system boundary. The bottom graph shows the numerical details when (SBC) with respect to SSD.

Now we discuss the way of treating the deformed system away from half filling. For the undeformed systems with OBC or PBC, it is sufficient to include the chemical potential term into Eq. (1), where is the number operator. The value of adjusts the Fermi energy to zero, and is given by


where is the filling factor


A natural way of introducing under the sinusoidal deformation is to write down the Hamiltonian as a sum of the local terms


where is included to the bond operator


In order to confirm the validity of these constructions in Eqs. (11)-(14), we carried out numerical calculations for the selected fillings and . Figure 5 shows the -dependence of , where is the ground-state energy for each filling. We plot the data only when the particle number is even. Analogous to half filling, the bond energy with PBC coincides with that obtained with SSD () when is odd. AP () The logarithmic corrections are again present when .

Figure 5: (Color online) Convergence of with respect to at -filling and -filling.

Figure 6: (Color online) Position dependence of the occupation number at , , and when , , and (upper graph). The left and the right lower graphs display the occupancy at close to the boundaries.

The uniformity of the ground state obtained with the sinusoidal deformation is checked by calculating the occupancy at , , and . Figure 6 shows when , , and . At half filling, , is always equal to by the particle-hole symmetry. Even away from the half filling, this uniformity is kept when . There are small fluctuations near the system boundary when and . In this way, the construction of Hamiltonian in Eqs. (13) and (14) away from half-filling is justified, especially for SSD.

Iii Correlated Systems

It is expected that the SSD reduces boundary effect even when interactions are present between particles. In order to check this conjecture, we apply the sinusoidal deformation to correlated systems, by means of the Hamiltonian Eq. (13) written in the linear combination of bond operators. We study the system size dependence of the ground-state energy and the uniformity of the system.

iii.1 Spinless Fermions

As an example, let us consider spinless Fermions on the 1D lattice, whose behavior is described by the uniform Hamiltonian

with PBC, where the system contains the repulsive Coulomb interaction between neighboring sites, in addition to the hopping amplitude . In this section we restrict ourselves to the half-filled case only, therefore the chemical potential is zero. XXZ () The construction of the Hamiltonian with OBC and its sinusoidal deformation is analogous to what we have done in the previous section. The deformed Hamiltonian can be defined by putting the bond operator as

and substituting it to Eq. (14). We calculate the ground state and the corresponding energy at half-filling up to .

Figure 7: (Color online) Energy per bond of the spinless Fermions model at up to .

Figure 7 shows with respect to . When the sinusoidal deformation is applied, we observe the same dependence as was seen for the non-interacting case. In particular when is odd, obtained for shows a good agreement with the result from PBC. The coincidence, however, becomes less accurate with increasing , and there is a deviation around in when . Figure 8 shows the occupation number and the bond correlation when at half filling. A clear uniformity is observed when as shown in the non-interacting cases.

Figure 8: (Color online) Occupation number and the bond energy when at half filling.

iii.2 Extended Hubbard Model

Now we consider the extended Hubbard model, which contains the on-site Coulomb interaction and the neighboring interaction . In this case there is spin degree of freedom (), therefore creation and annihilation operators, respectively, are represented as and . The bond operator of the extended Hubbard model is represented as

To avoid the complexity of determining the chemical potential , we consider the half-filled case where is guaranteed by the particle hole symmetry. Figure 9 shows the -dependence of for various combinations of and . The coincidence between PBC and SSD at half filling occurs when the total number of both up- and downs-spin Fermions are odd.

Figure 9: (Color online) System size dependence of of the extended Hubbard model at half filling. The full and the dashed lines connect the energies obtained by PBC and SSD, respectively.

Figure 10: (Color online) Occupation and the bond energy of the extended Hubbard model at half filling.

The occupancy and the bond correlation function at half filling are shown in Fig. 10. The occupancy with OBC is clearly influenced by the system boundaries, whereas the sinusoidal deformations () lead to almost constant distribution. The case with (asterisks) realizes the minimal position dependence. The bond correlation function is influenced by the boundaries in all cases, and the position dependence is the weakest when .

Iv Chemical Potential

So far we have not discussed the proper value of the chemical potential away from the half filling when the interaction is present. The chemical potential term has to be included to the bond operator in Eq. (III.2). In the Hartree-Fock or the Fermi liquid picture, is adjusted so that the particle number coincides with the number of negative-energy quasiparticle states. The number of particles is independently represented by the derivative of with respect to . Thus, the relation must be satisfied for the targeted particle number . Within the sinusoidal deformation, this relation is slightly modified as


according to the position dependence in the energy scale.

Figure 11: (Color online) The dependence of for the two cases and .

We plot the absolute value with respect to for the extended Hubbard model in Fig. 11, under the SSD. The numerical analysis by exact diagonalization gives for the case , , , and . Figure 12 shows the corresponding occupation and the bond correlation functions. There is a slight position dependence, since we are dealing with relatively small system size with a few particles. The position dependence becomes conspicuous if we choose either an inaccurate value of or we consider the deformations with .

Figure 12: (Color online) The upper and lower graphs show the occupancy and the bond correlation at .

V Conclusion and discussion

We have shown that the sinusoidal deformation applied to 1D quantum Hamiltonians improves convergence of the ground state energy per bond (or per site) toward the thermodynamic limit, compared to the uniform systems with open boundary conditions. Such suppression of the boundary effects is confirmed also for interacting systems, typically for the extended Hubbard model.

We have not determined the ‘canonical’ form of the sinusoidal deformation in case where there are long-range interactions. It is reported that a small but finite residual boundary effect appears in spin chains which includes the next-nearest-neighbor interaction. Hikihara () Application of the sinusoidal deformation to higher-dimensional quantum systems could be a future problem.

A theoretical puzzle of the sinusoidal deformation remains in coincidence of the ground-state energy calculated with PBC and SSD. The agreement is almost perfect, which strongly suggests a hidden algebraic relation between these two cases.

We thank T. Hikihara, A. Feiguin, and K. Okunishi for valuable discussions. This work was supported by ERDF OP R&D, Project ”QUTE - Centre of Excellence for Quantum Technologies” (ITMS 26240120009), CE QUTE SAV, APVV-51-003505, VVCE-0058-07, and VEGA-2/0633/09. T. N. acknowledges the support of Grant-in-Aid for Scientific Research (C) No. 22540388.


  • (1) M.E. Fisher and M.N. Barber, Phys. Rev. Lett. 28, 1516 (1972).
  • (2) M.N. Barber, in Phase Transitions and Critical Phenomena Vol. 8, ed. C. Domb and J.L. Lebowitz (Academic Press, 1983).
  • (3) T.W. Burkhardt and J.M.J. van Leewen, Real-Space Renormalization, Topics in Current Physics, Vol. 30 (Springer, Berlin, 1982), and references therein.
  • (4) S.R. White, Phys. Rev. Lett. 69, 2863 (1992); Phys. Rev. B 48, 10345 (1993).
  • (5) I. Peshel, X. Wang, M. Kaulke, and K. Hallberg, Density-Matrix Renormalization, Lecture Notes in Physics, vol. 528 (Springer, Berlin, 1999).
  • (6) U. Schollwöck. Rev. Mod. Phys. 77, 259 (2005).
  • (7) F. Verstraete, D. Porras, and J.I. Cirac, Phys. Rev. Lett. 93, 227205 (2004).
  • (8) P. Pippan, S.R. White, and H.G. Evertz, Phys. Rev. B 81, 081103(R) (2010).
  • (9) D. Rossini, V. Giovannetti, R. Fazio, arXiv:1102.3562.
  • (10) D. Rossini, V. Giovannetti, R. Fazio, arXiv:1102.3545.
  • (11) M. Vekić and S. R. White, Phys. Rev. Lett. 71, 4283 (1993).
  • (12) M. Vekić and S. R. White, Phys. Rev. B 53, 14552 (1996).
  • (13) A. Gendiar, R. Krcmar, and T. Nishino, Prog. Theor. Phys. 122, 953 (2009).
  • (14) The deformation function used in Ref. [13] is , although all the deformed Hamiltonians are denoted with the wrong factor . This primitive mistake by the authors of Ref. [13] was corrected in the Erratum: A. Gendiar, R. Krcmar, and T. Nishino, Prog. Theor. Phys. 123, 393 (2010).
  • (15) T. Hikihara and T. Nishino, to appear in phys. Rev. B; arXiv:1012.0472.
  • (16) Another generalization is to replace the sinusoidal function by the hyperbolic function: H. Ueda, A. Gendiar, V. Zauner, T. Iharagi, T. Nishino, preprint arXiv/1008.3458.
  • (17) If the particle number is even, coincides with that obtained for the anti-periodic boundary condition, which is specified by in Eq. 2. The fact was pointed out in Ref. [16].
  • (18) The Hamiltonian in Eq. (15) can be mapped to model by means of Jordan-Wigner transformation. The model under sine-squared deformation is studied in Ref. [16] under the presence of magnetic field, which plays the role of chemical potential.
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
Add comment
Loading ...
This is a comment super asjknd jkasnjk adsnkj
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test description