# Ground-state properties of fermionic mixtures with mass imbalance in optical lattices

## Abstract

Ground-state properties of fermionic mixtures confined in a one-dimensional optical lattice are studied numerically within the spinless Falicov-Kimball model with a harmonic trap. A number of remarkable results are found. (i) At low particle filling the system exhibits the phase separation with heavy atoms in the center of the trap and light atoms in the surrounding regions. (ii) Mott-insulating phases always coexist with metallic phases. (iii) Atomic-density waves are observed in the insulating regions for all particle fillings near half-filled lattice case. (iv) The variance of the local density exhibits the universal behavior (independent of the particle filling, the Coulomb interaction and the strength of a confining potential) over the whole region of the local density values.

PACS numbers: 05.30.Fk, 71.10.-w, 71.30.+h

## 1 Introduction

The ability to confine ultracold Bose and Fermi gases inside artificial crystals generated by standing-wave laser light fields, i.e., optical lattices, offers the possibility to create ideally clean and highly tunable strongly interacting quantum many-body systems [1]. The low-energy properties of these systems can be described using models borrowed from condensed matter systems [2] whose parameters and dimensionality can be controlled with high precision. This fact opens new routes for understanding the physics of long-standing problems of strongly correlated systems, like the phase separation, metal-insulator transitions, superconductivity, etc. On the other hand, the quadratic confining potential, present in addition to the regular ”lattice” potential leads to a number of fundamentally new phenomena. For example, it was shown [3, 4] that in the presence of a confining potential the Mott-insulating phase is restricted to the domain that coexists with a metallic phase, in contrast to the global character typical of solid-state systems. Moreover, mixtures of two-component atoms with different masses (e.g., and ) introduce an additional parameter, namely, the difference between the hopping amplitudes associated with each species of atoms in the optical lattice. This may affect the stability of the possible quantum phases or even induce new ones [5, 6, 7]. For these reasons the mixtures of ultracold atoms in optical lattices belong to the most intensively studied subjects of contemporary experimental and theoretical physics.

In this paper we investigate the ground-state properties of mixtures of two species of fermionic atoms with strongly different masses in a harmonic potential. Since the tunneling rate decreases exponentially with the square root of mass of the atom, and particles of different species on the optical lattice interact only through the on-site interaction, this leads naturally to the Falicov-Kimball Hamiltonian [8] with a confining potential [3]

(1) |

where () and () are the creation (annihilation) operators of heavy () and light () particles at lattice site . The number of lattice sites is and is selected so that all the fermions are confined in the trap. We denote the total number of fermions in the trap as and consider equal number of heavy () and light () atoms.

The first term of Eq. 1 is the kinetic energy corresponding to quantum-mechanical hopping of the light atoms between the nearest-neighbor sites and . These intersite hopping transitions are described by the matrix elements , which are if and are the nearest neighbors and zero otherwise (in the following all parameters are measured in units of ). The second term represents the on-site Coulomb interaction () between the light and heavy atoms. The last term is the energy of light and heavy atoms in the harmonic trapping potential. In accordance with similar studies for the asymmetric Hubbard model [9], we consider here the same trapping potential for both species of atoms.

Since in this spinless version of the Falicov-Kimball model with a confining potential the -heavy atom occupation number of each site commutes with the Hamiltonian (1), the -heavy atom occupation number is a good quantum number, taking only two values: or 0, according to whether or not the site is occupied by the heavy atom. Therefore, the Hamiltonian (1) can be written as

(2) |

where .

Thus for a given configuration of heavy atoms defined on the one-dimensional lattice, the Hamiltonian (2) is the second-quantized version of the single-particle Hamiltonian , so the investigation of the model (2) is reduced to the investigation of the spectrum of for different configurations of heavy atoms.

It is well known that in the absence of harmonic confinement () the one-dimensional Falicov-Kimball model exhibits a rich spectrum of solutions including phase separated ( small) as well as most homogeneous distributions of heavy atoms [10]. However, due to the confining potential the lattice sites become, inequivalent, and thus, it is of fundamental importance to analyse the interplay between the on-site Coulomb interaction and the confining potential.

To describe the system at nonzero we have calculated various local quantities, like the local density of heavy atoms (), the local density of light atoms (), the total site occupation (), the variance of the local density () and the local double occupation (), as functions of total number of the confined atoms, the Coulomb interaction and the confining potential . The ground states are calculated by a well-controlled numerical method that we have elaborated [11] for a description of the conventional Falicov-Kimall model (). Later, the method was successfully used for various generalizations of the Falicov-Kimball model and different physical problems [12]. Its generalization on systems with a harmonic potential is straightforward too.

## 2 Results and discussion

In Fig. 1 we present results of our numerical calculations for and obtained on the one-dimensional cluster of sites at , and different fillings. We have added also the profiles of the local compresibility that has been proposed by Rigol et al. [3] as a local order parameter to characterize the Mott-insulator regions. This quantity is defined as [3]

(3) |

where

(4) |

is the density-density correlation function and , with the correlation length of in the unconfined system at half-filling for a given . The factor is chosen within a range where becomes qualitatively insensitive to its precise value [4]. The insulating regions are then characterized by . For the values of used here we usually have - with .

The most interesting result obtained at low particle fillings (we note that ) is the observation of the complete phase separation with heavy atoms in the center of the trap () and light atoms in the surrounding regions (). Such a behaviour is found for all particle fillings from to some critical value that rapidly decreases with increasing and is almost independent of the local Coulomb interaction . In the regions where () the variance of the local density and the local compresibility are finite (the metallic phase), while in the middle of the trap where ( exponentially decreases in this region) both and are equal to zero (the insulating phase). Thus in accordance with results obtained for the Hubbard model [3] (the hopping probabilities are same for both types of atoms) we have found that also in the Falicov-Kimball model insulating domains coexist with metallic regions, such that global quantities are not appropriate to describe the system.

At higher particle fillings the situation is more complex. Above the critical filling (, for ) the connected cluster by heavy atoms occupied sites (in the center of the trap) splits on smaller clusters, usually of the same size, separated by the empty site. As increases the size of clusters decreases from to 1. Of course, the redistribution of heavy atoms has dramatic consequences on the distribution of light atoms. Now, the light atoms occupy preferably the empty sites in the middle of the trap () what leads to the atomic-density waves in and profiles. In the region where the variance of the local density is finite but smaller than in surrounding metallic regions indicating [3, 4] the presence of Mott-insulating phase in the center of the trap. This conjecture supports the behaviour of the local compresibility that is equal to zero in the corresponding region.

Increasing the number of particles up to , the connected cluster of heavy atoms () starts to form in the middle of the trap. In this region both the variance of the local density and the local compresibility are finite what indicates the presence of metallic phase in the center of the trap. Upon adding more fermions, this new metallic phase widens spatially, while the Mott-insulating regions of the atomic density waves are pushed to the borders and completely disappear at . At this filling practically the whole region of the trap, except the edges (where ), is metallic. With still a higher filling this metallic phase is further stabilized, but at some critical filling () a new insulating phase (”a band insulator”) starts to develop in the center of the trap (). This trend holds also for the highest particle fillings, the width of the band-insulating phase increases and the surrounding metallic regions are gradually suppressed.

To reveal the role of the Coulomb interaction and the confining potential on a formation of metallic and insulating domains, similar calculations have been also performed for various combinations of and . The results of numerical calculations for the site occupation as a function of are displayed in Fig. 2. One can see that the Coulomb interaction and the confining potential exhibit precisely opposite effects on the stability of metallic and insulating domains. Indeed, with increasing (at fixed ) the insulating domain is suppressed and the metallic domains are stabilized, while with increasing (at fixed ) the metallic domains are suppressed and the insulating domain is stabilized.

Since the Falicov-Kimball model can be considered as a simplification of the Hubbard model (only one kind of particles, say with spin up can hop) it is interesting to compare results obtained in these two different limits. Such a comparison (see Ref. 3 and Ref. 4) reveals obvious differences in behavior of these models in the confining potential. For example, the ground-state of the Hubbard model (for ) is always metallic at low particle fillings, while the metallic regions coexist with the insulating region in the ground state of the Falicov-Kimball model. Moreover, the local density profiles exhibit obvious oscillations for the Falicov-Kimball model, while no sign of such oscillations has been observed for the repulsive Hubbard model (with the exception of Hartree-Fock [4] and variational studies [13], that were not confirmed by a projector Monte Carlo simulations, however [4]).

To exclude the possibility that oscillations are a consequence of a finite size of clusters used in our numerical calculations, we have performed an exhaustive finite-size scaling analysis on finite clusters up to sites for all particle fillings from Fig. 1. This analysis showed that the ground states found for hold also on clusters of and sites and thus they can be satisfactorily extrapolated on much larger clusters. In Fig. 3 we present numerical results for extrapolated ground states obtained for a cluster of sites and the same values of and as used in Fig. 1. These results clearly demonstrate that increasing suppresses the atomic-density oscillations in the metallic phase, but stabilizes the atomic-density waves in the insulating phase.

Since in the real experiments with ultracold atoms the hopping matrix elements (of heavy atoms) between the nearest-neighbor sites () are not strictly equal to zero, it is necessary to examine the stability of our solutions obtained for against the finite values of . For this reason we have performed exhaustive exact-diagonalization studies of the asymmetric Hubbard model [14] () in a confining potential for a wide range of model parameters ( and ) on finite clusters up to . The representative examples of local density profiles are displayed in Fig. 4 for several different values of . They clearly show that results obtained for remain stable also at finite (small) . Thus the Falicov-Kimball model with a confining potential can be used satisfactorily to model the ground-state behavior of mixtures of two-component fermionic atoms (with strongly different masses) in a harmonic potential.

Finally, we have also calculated the variance of the local density as a function of the local density . It is known from the study of the one-dimensional Hubbard model in a harmonic potential [3] that this quantity shows universality with respect to the confining potential for . It should be noted that the universal behavior was observed only for strong Coulomb interactions , where systems have a Mott-insulating phase at . In Fig. 5 we present numerical results for the variance of the local density obtained within the Falicov-Kimball model with a harmonic potential as a function of the local density calculated for various particle fillings and various values of and . One can see that all numerical data for the variance of the local density collapse on the same curve (given by for and for ). Thus in contrast to the similar studies on the Hubbard model [3] we have found that the variance of the local density of systems described by Falicov-Kimball model with a harmonic potential exhibits universality not only with respect to the confining potential but also with respect to the Coulomb interaction . In addition, we have found that the universality hold not only for , but for all values of . A similar universal behaviour has been observed also for other local quantities. For example, the local double occupation for and for , independent of values of and . Also the universal behavior is found for the local compresibility when . However, the value of the critical exponent for the Falicov-Kimball model is equal to 1, unlike the nontrivial value 0.68-0.78 found for the Hubbard model [3].

In summary, we have studied the ground-state properties of fermionic mixtures with mass imbalance in a one-dimensional optical lattice within the spinless Falicov-Kimball model with a harmonic potential. We have found that the system exhibits the phase separation at low particle fillings. In this case the heavy atoms occupy the center of the trap while the light atoms are localized in the surrounding (metallic) regions. At higher fillings we have observed a formation of Mott-insulating domains with atomic-density waves. In all cases Mott-insulating phases coexist with metallic phases. One of the most interesting results is, however, the observation of the universal behavior of the variance of the local density (independent of the particle filling, the Coulomb interaction and the strength of the confining potential) over the whole region of the local density values.

This work was supported by Slovak Grant Agency VEGA under Grant No.2/7057/27 and Slovak Research and Development Agency (APVV) under Grant LPP-0047-06.

Figure Caption

Fig. 1. Profiles of various local quantities along the trap calculated for and four different values of particle fillings. Different panels (from top to bottom) correspond to: the local density of heavy atoms , the local density of light atoms , the total site occupation , the local double occupation , the variance of the local density and the local compresibility .

Fig. 2. The site occupation as a function of site position calculated for different (the first column) and different (the second column).

Fig. 3. The site occupation as a function of site position calculated for extrapolated ground states for different particle fillings. The parameters involved are and . The insets show details of profiles for the region of atomic-density waves.

Fig. 4. The exact-diagonalization results (the asymmetric Hubbard model with a harmonic trap) for the site occupation as a function of site position calculated for different values of and on a finite cluster of sites.

Fig 5. The variance of the local density as a function of the local density calculated for different values of and . The numerical data are fitted by for and for .

### References

- M. Greiner et al., Nature (London) 415, 39 (2002); I. Bloch, Phys. World 17, 25 (2004); D. Jacksch and P. Zoller, Ann.Phys. (N.Y.) 315, 52 (2005); W. Hofstetter, Philos. Mag. 86, 1891 (2006); T. Stoferle et al., Phys. Rev. Lett. 96, 030401 (2006).
- J.I. Cirac and P. Zoller, Science 301, 176 (2003);
- M. Rigol, A Muramatsu, G.G. Batrouni and R.T. Scalettar, Phys. Rev. Lett. 91, 130403 (2003).
- M. Rigol and A Muramatsu, Phys. Rev. A. 69, 053612 (2004).
- C. Ates and K. Ziegler, Phys. Rev. A. 71, 063610 (2004); K. Ziegler, Nuclear Physics A 790, 718C (2007).
- S. J. Gu, R. Fan, and H. Q. Lin, Phys. Rev. B 76, 125107 (2008).
- M.A. Cazalilla, A.F. Ho and T. Giamarchi, Phys. Rev. Lett. 95, 226402 (2003).
- L.M. Falicov and J.C. Kimball, Phys. Rev. Lett. 22, 997 (1969).
- T. Gottwald and P.G.J van Dongen, Eur. J. Phys. 61, 277 (2008).
- P. Lemberger, J. Phys. A 25, 715 (1992); C. Gruber and D. Ueltschi and J. Jedrzejewski, J. Stat. Phys. 76, 125 (1994); J.K. Freericks, Ch. Gruber and N. Macris, Phys. Rev. B 53, 16189 (1996); J.K. Freericks, E.H. Lieb and D. Ueltschi, Phys. Rev. Lett. 88, 106401 (2002).
- P. Farkašovský, Eur. J. Phys. B 20, 209 (2001); P. Farkašovský, Int. J. Mod. Phys. B 17, 4897 (2003).
- P. Farkašovský, H. Čenčarikova and N. Tomašovičova, Eur. J. Phys. B 45, 479 (2005); P. Farkašovský and H. Čenčarikova, Eur. J. Phys. B 47, 517 (2005).
- Y. Fujihara, A. Koga and N. Kawakami, J. Phys. Soc. Jpn.B 76, 034716 (2007).
- R. Lyzwa and Z. Domanski, Phys. Rev. B 50, 11381 (1994); P. Farkašovský, Phys. Rev. B 77, 085110 (2008).