From Localization To Delocalization in The Quantum Coulomb Glass

From localization to delocalization in the quantum Coulomb glass
Thomas Vojta∗ , Frank Epperlein, Svetlana Kilina, and Michael Schreiber

arXiv:cond-mat/0003395v1 [cond-mat.str-el] 24 Mar 2000
Institut für Physik, Technische Universität, D-09107 Chemnitz, F. R. Germany

∗ Contact information: phone +49 371 531 3147, fax +49 371 531 3151, email
[email protected]
(October 29, 2018)
Abstract
We numerically investigate how electron-electron interactions influence the

transport properties of disordered electrons in two dimensions. Our study
is based on the quantum Coulomb glass model appropriately generalized to
include the spin degrees of freedom. In order to obtain the low-energy prop-
erties of this model we employ the Hartree-Fock based diagonalization, an
efficient numerical method similar to the configuration interaction approach
in quantum chemistry. We calculate the d.c. conductance by means of the
Kubo-Greenwood formula and pay particular attention to the spin degrees
of freedom. In agreement with earlier results we find that electron-electron
interactions can cause delocalization. For spinful electrons this delocalization
is significantly larger than for spinless electrons.
71.55.Jv, 72.15.Rn, 71.30.+h
Typeset using REVTEX
1
I. INTRODUCTION
The discovery of a metal-insulator transition (MIT) in the two-dimensional electron gas

in Si-MOSFETs [1] has induced renewed attention to the transport properties of disordered
electrons. This MIT is in conflict with the theory of localization for non-interacting electrons
which predicts that all states are localized in 2D. The electron density in the Si-MOSFETs
is very low which makes the electron-electron interaction particularly important. Thus it is
generally assumed that some type of interaction effect is responsible for this MIT. One of
the most remarkable findings about the MIT in Si-MOSFETs is that an in-plane magnetic
field (which does not couple to the orbital motion of the electrons) strongly suppresses the
conducting phase [2]. This suggests that the spin degrees of freedom play an important role
for the transition. A complete understanding has, however, not yet been obtained. There
have been attempts to explain the experiments based on the perturbative renormalization
group [3], non-perturbative effects [4], or the transition being a superconductor-insulator
transition rather than a MIT [5].
In order to attack the problem of disordered interacting electrons numerically we have
developed [6,7] an efficient method, the Hartree-Fock based diagonalization (HFD) which
is related to the quantum-chemical configuration interaction approach. We have used this
method to study the influence of interactions on the conductance in one [8], two [6], and
three [9] dimensions. We found a delocalizing tendency of the interactions for strong disorder
but a localizing one for weak disorder. Similar results have been obtained by means of the
density-matrix renormalization group [10] in one dimension and exact diagonalization in two
dimensions [11]. Since in most of the numerical work in the literature spinless electrons were
considered, there are not many results about the importance of the spin degrees of freedom.
In this work we address this question by generalizing the HFD method to spinful elec-
trons. We then use it study the influence of the spin degrees of freedom on the Kubo-
Greenwood conductance.
II. MODEL AND METHOD
The generic model for spinless interacting disordered electrons is the quantum Coulomb
glass [12]. In this paper we use a straight-forward generalization of the quantum Coulomb
glass to spinful electrons. It is defined on a regular hypercubic lattice with g = Ld (d is
the spatial dimensionality) sites occupied by N = N↑ + N↓ = 2Kg electrons (0 < K < 1).
To ensure charge neutrality each lattice site carries a compensating positive charge of 2Ke.
The Hamiltonian is given by
X X 1 X X
H = −t (c†iσ cjσ + h.c.) + ϕi niσ + (niσ − K)(njσ′ − K)Uij + UH ni↑ ni↓ (1)
hiji,σ i,σ 2 i6=j,σ,σ′ i
where c†iσ and ciσ are the creation and annihilation operators at site i and spin σ, and hiji
denotes all pairs of nearest-neighbor sites. t is the strength of the hopping term, i.e., the
kinetic energy, and niσ is the occupation number of spin state σ at site i. We parametrize
the interaction Uij = e2 /rij by its value U between nearest-neighbor sites. The Coulomb
repulsion between two electrons at the same site is described by the Hubbard interaction UH
2
The random potential values ϕi are chosen independently from a box distribution of width
2W0 and zero mean. The boundary conditions are periodic and the Coulomb interaction is
treated in the minimum image convention (which implies a cut-off at a distance of L/2).
A numerically exact solution of a quantum many-particle system requires the diagonal-
ization of a matrix whose dimension increases exponentially with system size. This severely
limits the possible sample sizes. In order to overcome this problem we have developed the
HFD method. The basic idea is to work in a truncated Hilbert space consisting of the
corresponding Hartree-Fock (Slater) ground state and the low-lying excited Slater states.
For each disorder configuration three steps have to be performed: (i) find the Hartree-Fock
solution of the problem, (ii) determine the B Slater states with the lowest energies, and
(iii) calculate and diagonalize the Hamilton matrix in the subspace spanned by these states.
The number B of new basis states determines the quality of the approximation, reasonable
values have to be found empirically.
III. RESULTS
The conductance of a quantum many-particle system can be obtained from linear-

response theory. It is essentially determined by the current-current correlation function
of the ground state. The real (dissipative) part of the conductance (in units of e2 /h) is given
by the Kubo-Greenwood formula [13],
2π 2 X
Re Gxx (ω) = |h0|j x |νi|2 δ(ω + E0 − Eν ) (2)
ω ν
where ω denotes the frequency. j x is the x component of the current operator and ν denotes
the eigenstates of the Hamiltonian. Eq. (2) describes an isolated system while in a real
d.c. transport experiment the sample is connected to contacts and leads. This results in a
finite life time τ of the eigenstates leading to an inhomogeneous broadening γ = τ −1 of the δ
functions in (2) [14]. To suppress the discreteness of the spectrum of a finite system, γ should
be at least of the order of the single-particle level spacing. For our systems this requires a
comparatively large γ ≥ 0.05. We tested different γ and found that the conductance values
depend on γ but the qualitative results do not.
The main results of this paper are summarized in Fig. 1 which shows the disorder and
interaction dependence of the typical conductance for both spinless and spinful electrons on
a two-dimensional lattice of 4 × 4 sites. The qualitative behavior in both cases is similar:
In the strongly localized regime (small t) a moderate interaction delocalizes the electrons
while a sufficiently strong interaction always strongly suppress the conductance. This is
the precursor of a Wigner crystal or Wigner glass. With decreasing disorder (increasing
t) the interaction-induced enhancement of the conductance also decreases and eventually
vanishes. The behavior of the conductance can be attributed to the competition of two
effects: First, the interactions destroy the phase of the electrons and thus the interference
necessary for localization. This is particularly effective if the localization length is small to
begin with. Second, the interactions introduce an additional source of randomness which
tends to increase the localization.
A comparison of the cases of spinless and spinful electrons shows that the interaction in-
duced delocalization is significantly larger for spinful electrons. Moreover, the enhancement
3
100 t=
0.5
0.3
0.2
G(0) [e2/h]
-2
10 0.1
0.05
0.03
0.02
10-4 0.01
(a) (b)
-6
10
0 0.5 1 1.5 2 0 0.5 1 1.5 2
U U
FIG. 1. d.c. conductance G(0) for a system of 4 × 4 lattice sites occupied by (a) 8 spinless
electrons or (b) 8 spin-up and 8 spin-down electrons for different U and t. The disorder strength
is fixed to W0 = 1, the Hubbard energy is UH = 0.5, the broadening is γ = 0.0625, and the HFD
basis size is B = 500. The data points represent logarithmical averages over 400 samples.
seems to vanish at a larger kinetic energy (which we did not reach in the simulations). A sys-
tematic investigation of the dependence of the conductance on U and UH will be published
elsewhere.
In summary, we have studied the influence of electron-electron interactions on Anderson
localization for spinless and spinful electrons in two dimensions. For strong disorder moder-
ate interactions significantly enhance the transport. This enhancement is much stronger for
spinful than for spinless electrons. Identifying a real phase transition and thus establishing a
connection between these findings and the experiments on Si-MOSFETs requires a finite-size
scaling analysis of the conductance. This remains a task for the future.
This work was supported in part by the NSF under grant no. DMR–98–70597 and by the DFG
under grant no. SFB 393/C2. T.V. thanks the Aspen Center for Physics and the University of
Oregon for hospitality during the completion of this paper.
4
REFERENCES
[1] S. V. Kravchenko, D. Simonian, and M. P. Sarachik, Phys. Rev. Lett. 77, 4938 (1996).
[2] D. Popovich, A. B. Fowler, and S. Washburn, Phys. Rev. Lett. 79, 1543 (1997); D.
Simonian, S. V. Kravchenko, and M. P. Sarachik, ibid. 79, 2304 (1997).
[3] C. Castellani, C. DiCastro, and P. A. Lee, Phys. Rev. B 57, R9381 (1998).
[4] S. Chakravarty, S. Kivelson, C. Nayak, and K. Völker, cond-mat/9805383.
[5] D. Belitz and T. R. Kirkpatrick, Phys. Rev. B 58, 8214 (1998); F. C. Zhang and T. M.
Rice, cond-mat/9708050; P. Phillips et al., Nature 395, 253 (1998).
[6] T. Vojta, F. Epperlein, and M. Schreiber, Phys. Rev. Lett. 81, 4212 (1998).
[7] T. Vojta, F. Epperlein, and M. Schreiber, accepted for publication in Comp. Phys.
Commun., cond-mat/9809171.
[8] M. Schreiber, F. Epperlein, and T. Vojta, Physica A 266, 443 (1999).
[9] T. Vojta and F. Epperlein, Ann. Phys. (Leipzig) 7, 493 (1998).
[10] P. Schmitteckert, R.A. Jalabert, D. Weinmann, and J.-L. Pichard, Phys. Rev. Lett. 81,
2308 (1998).
[11] G. Benenti, X. Waintal, and J.-L. Pichard, cond-mat/9904096.
[12] A. L. Efros and F. G. Pikus, Solid State Commun. 96, 183 (1995); J. Talamantes, M.
Pollak, and L. Elam, Europhys. Lett. 35, 511 (1996); F. Epperlein, M. Schreiber, and
T. Vojta, Phys. Rev. B 56, 5890 (1997).
[13] K. Kubo, J. Phys. Soc. Jpn. 12, 576 (1957); D. A. Greenwood, Proc. Phys. Soc. 71, 585
(1958).
[14] See, e.g., S. Datta, Electronic transport in mesoscopic systems, Cambridge University
Press, Cambridge (1997).

From Localization To Delocalization in The Quantum Coulomb Glass

Uploaded by

Copyright:

Available Formats

From Localization To Delocalization in The Quantum Coulomb Glass

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

From Localization To Delocalization in The Quantum Coulomb Glass

Uploaded by

Copyright:

Available Formats

From localization to delocalization in the quantum Coulomb glass

Thomas Vojta∗ , Frank Epperlein, Svetlana Kilina, and Michael Schreiber

Institut für Physik, Technische Universität, D-09107 Chemnitz, F. R. Germany

We numerically investigate how electron-electron interactions influence the

Typeset using REVTEX

The discovery of a metal-insulator transition (MIT) in the two-dimensional electron gas

II. MODEL AND METHOD

The conductance of a quantum many-particle system can be obtained from linear-

You might also like