# Particle-Hole Pair Coherence in Mott Insulator Quench Dynamics

###### Abstract

We predict the existence of novel collapse and revival oscillations that are a distinctive signature of the short-range off-diagonal coherence associated with particle-hole pairs in Mott insulator states. Starting with an atomic Mott state in a one-dimensional optical lattice, suddenly raising the lattice depth freezes the particle-hole pairs in place and induces phase oscillations. The peak of the quasi-momentum distribution, revealed through time of flight interference, oscillates between a maximum occupation at zero quasi-momentum (the point) and the edge of the Brillouin zone. We show that the population enhancements at the edge of the Brillouin zone is due to coherent particle-hole pairs, and we find similar effects for fermions and Bose-Fermi mixtures in a lattice. Our results open a new avenue for probing strongly correlated many-body states with short-range phase coherence that goes beyond the familiar collapse and revivals previously observed in the long-range coherent superfluid regime.

###### pacs:

03.75.-b, 67.85.-d, 72.20.-iUltracold atoms in optical lattices are a versatile tool for creating strongly correlated quantum many-body states duchon13 ; bloch08 . Prominent examples include atomic Mott insulator (MI) states of bosons and fermions greiner02a ; jordens08 ; schneider08 . Starting from a superfluid (SF) with long-range phase coherence, adiabatically increasing the ratio of atom-atom interaction to tunneling strength by increasing the optical lattice depth gives rise to a MI with short-range coherence greiner02a . An infinitely deep lattice with gives a perfect Mott insulator state as depicted in Fig. 1(a), denoted in Fock space notation as for unit occupancy. For finite tunneling but still in the MI regime, the many-body wavefunction includes correlated (paired) double and zero occupied sites on top of the perfect MI [see Fig. 1(b), (c)]. These particle-hole pairs, also known as doublons and holons and denoted as , etc., behave as quasi-particle excitations. Particle-hole pairs play an important role in Mott insulator physics gerbier05 ; bakr09 ; bakr10 , and have recently been observed by in-situ imaging in a one-dimensional (1D) chain of bosonic atoms endres11 .

Ultracold atoms in optical lattices are also a versatile tool for studying many-body quantum dynamics kennett13 ; polkovnikov11 . A common technique involves quenching a system by suddenly changing its parameters kennett13 ; kollath07 . One approach, starting from larger and quenching to a smaller , has been used to study thermalization altman02 ; rigol08 , the Kibble-Zurek mechanism zurek05 ; demarco11 , and light-cone spreading of correlations lightcone12 ; barmettler13 ; natu12 . The opposite approach, starting in a SF regime with small and quenching to large gives the collapse-and-revival (CR) oscillations of matter-wave phase coherence observed in Refs. greiner02b ; porto07 ; will10 . Matter-wave CR has found applications in the study of higher-body interactions will10 ; mahmud13a , fermion impurities will11 , and coherent-state squeezing will10 ; tiesinga11 . To date, CR has been viewed as a characteristic behavior of quenched systems with long-range phase coherence such as superfluids.

In this Letter, we show that collapse-and-revival also occurs for quenched MI states, for both lattice bosons, fermions, and mixtures. Surprisingly, detectable CR oscillations should occur despite the short-range coherence of the Mott state. For a 1D optical lattice, the quasi-momentum distribution oscillates between maximum occupations (revivals) at quasi-momentum and at the edge of the Brillouin zone. (Here and is Planck’s constant.) In fact, the normalized difference in the two populations, which is conventionally defined as the “visibility” of the condensate, becomes negative, in sharp contrast to the behavior of quenched superfluids. We also find that the visibility of the quenched MI state is a sinusoidal function of time, in contrast to the exponential functional form seen for quenched superfluids. This difference provides another approach for distinguishing between systems with short- and long-range coherence.

For a quenched MI, the condensate fraction oscillations also show novel kinks associated with oscillations between symmetric and antisymmetric natural orbitals of the single-particle density matrix (SPDM). The distinctive behavior of MI CR oscillations are due to the presence of correlated particle-hole pairs, frozen in place when the lattice depth is suddenly increased, leading to phase dynamics as schematically illustrated in Fig. 1(d).

Unexpectedly, we find that fermions in a lattice at half-filling, and Bose-Fermi mixtures when the total occupation of bosons plus fermions is an integer, exhibit similar dynamics due to correlated particle-hole pairs. Finally, we show that this physics is robust under realistic dephasing mechanisms, and should be within experimental reach. Our results reveal universal coherence dynamics for quenched, strongly-correlated lattice systems with short-range phase coherence.

*Lattice bosons.* Ultracold bosons in an optical lattice
can be described by the Bose-Hubbard Hamiltonian

(1) |

where are indices to lattice sites, only nearest-neighbor tunneling is assumed, and . The subscript denotes parameter values before the quench. The total number of particles is , and is the mean occupation per site where is the number of sites. For non-integer occupation per site, the ground state is superfluid. At integer occupation, the system exhibits a quantum phase transition, going from a SF to MI state above a critical value of . At unit occupation, in 1D scalettar91 ; zakrzewski08 .

The system is quenched by suddenly increasing the depth of the optical lattice such that tunneling is suppressed () greiner02b . The Hamiltonian governing the post-quench dynamics is , where is the final interaction strength. After a hold-time , the lattice is turned off and the atoms are allowed to freely expand. Imaging the atoms after the time-of-flight expansion, the atomic spatial density corresponds to the quasi-momentum distribution at the moment of release, where is the SPDM and is the lattice wavevector. In a 1D lattice with a period of unit length, the edge of the Brillouin zone is at wavevector . In addition, we examine the visibility gerbier05 , and the condensate fraction , where is the largest eigenvalue of the SPDM penrose56 ; hofstetter11 .

In previous CR studies the initial state is a superfluid. Here, our initial state is a one-dimensional MI with integer and . Due to the presence of particle-hole pairs, this state, after the quench, is not an eigenstate of , and undergoes nonequilibrium evolution until the moment of release from the lattice. Much of the MI quench dynamics can be understood from an analysis of correlations in the initial state. It is convenient to define the off-diagonal coherence , which is independent of as the SPDM only depends on for a homogeneous system with periodic boundary conditions. It follows that and, similarly, .

We are able to obtain explicit analytic approximations for the evolution of the off-diagonal coherence and quasimomentum populations for bosons in a 1D lattice using the strong-coupling expansion freericks96 up to second-order in the tunneling Hamiltonian. The full derivation is given in the Supplementary material. Briefly, to zeroth-order the ground-state wavefunction is the Mott state with atoms in each of the sites. Corrections are due to states containing a “hole”, atoms in one of the sites, and a “particle”, atoms in another. We can express these particle-hole pair states as , where is the lattice distance between particle and hole. The improved ground state becomes with , from which it follows that initially . This corresponds to an exponential decay of the off-diagonal coherence with distance (see also Ref. cazalilla11 ).

Deep in the MI regime the time evolution of the and populations are therefore, to good approximation, and , respectively, for a MI with . Consequently, when is positive (negative) the population is larger (smaller) than . In fact, it is shown in the Supplementary material that with period . Hence, is larger than for . This implies that for these hold times.

In contrast, for a SF with for and with any for non-integer , the mean-field ground-state wavefunction is site separable, i.e., of the form , with . Consequently, the SPMD is also separable with . For a homogeneous lattice is independent of and there is no decay of correlation with particle-hole distance . As all are positive we find for all . This implies that . This conclusion also holds for a fully correlated model when superfluid SPDM is not constant but decays with algebraically in 1D and goes to a constant value in higher dimensions. Due to the slower than exponential decay, all the terms in the expression of contribute, and .

In the Supplementary material we also derive the quasi-momentum and visibility in the MI regime and the thermodynamic limit. They are given by

and

(3) |

In contrast, in the superfluid regime with a homogeneous pre-quench state that is separable, and , where . For a coherent state at each lattice site

(4) |

where and . For both the MI and SF quench the time evolution is periodic with period . The oscillations, however, are sinusoidal for the MI quench and “exponential” for the SF quench.

We next perform numerical i-TEBD simulations vidal07 for a MI with to extend our analysis to all values of and validate the SCE analysis in the MI regime. We calculate the SPDM for and use periodic boundary conditions. Figure 2(a) shows the quasi-momentum for a Mott state with at hold times , , and . At the momentum distribution is peaked around , while at it is peaked around , corresponding to a “revival” at the edge of the Brillouin zone and to a new kind of nonequilibrium state. This revival is a distinctive signature of the presence of nearest-neighbor particle-hole pairs in the MI state.

To highlight the difference between MI and superfluid quenches, Fig. 2(b) shows the dynamics for a superfluid with . Here, at the quasimomentum distribution is again peaked around , but is much narrower than the MI case due to the long-range coherence of the SF. During the collapse at , the atoms are equally spread over all quasi-momenta with no enhancement or revival of the population. Note that revivals have also been predicted for supersolid quenches in a system with (long-range) nearest-neighbor interactions fischer12 . Those revivals, however, are associated with long-range coherences and not particle-hole pairs.

Figure 2(c) and (d) show the time evolution of and for different values of and , one in the SF regime and three in the MI regime. For , the time trace exhibits superfluid CR oscillations consistent with Eq. 4 and the visibility is positive. Based on our numerical results, we conjecture that for there will always exist time intervals where the visibility is negative. We show an example for in Fig. 2(d). For larger the oscillations in and become more sinusoidal, consistent with Eqs. Particle-Hole Pair Coherence in Mott Insulator Quench Dynamics and 3. Examples are given for and .

In Fig. 2(e) we compare analytic predictions for in Eq. 3 with our numerical simulations. The analytic approximation gives excellent agreement for . The regime of validity for the analytic theory can be extended to smaller by performing higher-order calculations and keeping high occupancy terms, e.g., terms such as and for unit occupancy ().

Figure 2(f) shows the condensate fraction as a function of time. For in the MI regime has kinks at times and , which occur when the eigenfunction or natural orbital of the SPDM associated with the largest eigenvalue changes between a symmetric state and an anti-symmetric state . This behavior is consistent with our SCE calculations, which give

and where the second term leads to the kinks. Within the SCE approximation the kinks occur when the nearest neighbor coherence goes to zero. For in the SF regime the fraction performs smooth oscillations reaching . (The full range is not shown in the graph.) The numerical result is consistent with , based on the mean-field approximation hofstetter11 .

*Lattice fermions.* Fermions in an optical lattice are
described by the Fermi-Hubbard Hamiltonian

(6) |

where denotes two spin states and . Its phase diagram contains metallic, Mott insulator and anti-ferromagnatic phases lee06 . After the quench the Hamiltonian becomes . Figure 3 illustrates the quench dynamics, computed using a four site lattice and exact diagonalization. We first consider the half-filled case with . Figure 3(a) shows the momentum distribution at times and , where . At the momentum distribution is peaked around , while at the peak occurs at the edge of the Brillouin zone, an effect similar to the bosonic MI case. The revivals are again due to correlated doublons and holons in the initial state. Visibility oscillations of the spin-up state, shown in Fig. 3(b), reach negative values for any and become more sinusoidal for larger values of . In contrast, Figs. 3(c) and (d) show that a metallic state with and exhibits the familiar superfluid-type CR, where all quasi-momenta are uniformly occupied during collapse at .

*Lattice Bose-Fermi mixtures.* For Bose-Fermi mixtures in
a lattice, the post-quench Hamiltonian is

(7) |

where and are Bose-Bose and Bose-Fermi interaction strengths, respectively, and assuming fermions in a single spin state. For the pre-quench Hamiltonian, the usual hopping terms are also present albus03 . Bose-Fermi mixtures give rise to a plethora of phases and phenomena albus03 ; anders12 , depending on the relative interaction strengths and fermion filling. CR of a Bose-Fermi mixture has been observed in Ref. will11 . When the total occupation per site of fermions plus bosons is one, the ground state phase diagram has a MI state altman12 . The quench dynamics, computed with a four site lattice, is shown in Figs. 4(a) and (b). Here again we see revivals of the bosonic momentum distribution and the sinusoidal behavior of the bosonic visibility, with modifications due to the Bose-Fermi interaction. In contrast, for non-unit occupation, the dynamics gives the SF-like CR shown in Figs. 4(c) and (d).

*Experimental prospects and conclusions.* Our analysis in
Fig. 2 shows that the predicted visibility of the revival
should be of the same order of magnitude as the peak
observed for an equilibrium MI gerbier05 , and hence a
detection of the dynamics should be possible. These effects are
also robust under several dephasing mechanisms. First,
Fig. 5(a) shows that for small post-quench tunneling
the characteristic negative survives, but is destroyed
for larger .
Second, we find that effective multi-body
interactions will10 ; johnson09 ; johnson12 do not significantly
influence the dynamics deep in the MI regime (e.g.,
). Third, by adding
in Eq. 1, harmonically trapped systems give a MI
core with SF regions at the two edges. Assuming after
the quench, Fig. 5(b) shows that the SF regions
modify , but do not completely wash out the
distinctive MI signature for a sufficiently large Mott plateau.
Dephasing due to post-quench harmonic trapping
for SF revivals has been quantified in
Refs. hofstetter11 ; mahmud13b .

In conclusion, we have shown that bosons and fermions in 1D periodic potentials with short-range coherence will manifest a new type of collapse-and-revival oscillation due to particle-hole pair correlations. We expect qualitatively similar effects in 2D and 3D. Our results open up the possibility that the underlying correlations of a wide class of strongly correlated matter can now be revealed through coherence dynamics.

*Acknowledgments.* We acknowledge support from the US Army
Research Office under Contract No. 60661PH.

## References

- (1) E. Duchon, Y. L. Loh, and N. Trivedi, e-print arXiv:1311.0543 (2013).
- (2) I. Bloch, J. Dalibard, and W. Zwerger, Rev. Mod. Phys. 80, 885-964 (2008).
- (3) M. Greiner, O. Mandel, T. W. Hänsch and I. Bloch, Nature 415, 39 (2002).
- (4) R. Jordens, N. Strohmaier, K. G nter, H. Moritz, and T. Esslinger, Nature 455, 7210 (2008).
- (5) U. Schneider, L. Hackermuller, S. Will, Th. Best, I. Bloch, T. A. Costi, R. W. Helmes, D. Rasch, and A. Rosch, Science 322, 5907 (2008).
- (6) F. Gerbier, A. Widera, S. Foelling, O. Mandel, T. Gericke, and I. Bloch, Phys. Rev. Lett. 95, 050404 (2005); Phys. Rev. A 72, 053606 (2005).
- (7) W. S. Bakr, J. I. Gillen, A. Peng, S. Foelling, and M. Greiner, Nature 462, 74 (2009).
- (8) W. S. Bakr, A. Peng, M. E. Tai, R. Ma, J. Simon, J. I. Gillen, S. Foelling, L. Pollet, and M. Greiner, Science 329, 547 (2010).
- (9) M. Endres, M. Cheneau, T. Fukuhara, C. Weitenberg, P. Schauss, C. Gross, L. Mazza, M.C. Banuls, L. Pollet, I. Bloch, and S. Kuhr, Science 334, 200 (2011).
- (10) M. P. Kennett, ISRN Condensed Matter Physics, 2013, 393616 (2013).
- (11) A. Polkovnikov, K. Sengupta, A. Silva, and M. Vengalattore, Rev. Mod. Phys. 83, 863 (2011); J. Dziarmaga, Adv. Phys. 59, 1063 (2010).
- (12) C. Kollath, A. M. Laeuchli, and E. Altman, Phys. Rev. Lett 98, 180601 (2007).
- (13) E. Altman and A. Auerbach, Phys. Rev. Lett. 89, 250404 (2002).
- (14) M. Rigol, V. Dunjko, and M. Olshanii, Nature 452, 854 (2008).
- (15) W. H. Zurek, U. Dorner and P. Zoller, Phys. Rev. Lett. 95, 105701 (2005).
- (16) D. Chen, M. White, C. Borries and B. DeMarco, Phys. Rev. Lett. 106, 235304 (2011).
- (17) M. Cheneau, P. Barmettler, D. Poletti, M. Endres, P. Schauss, T. Fukuhara, C. Gross, I. Bloch, C. Kollath, and S. Kuhr et al., Nature 481, 484 (2012).
- (18) P. Barmettler, D. Poletti, M. Cheneau, and C. Kollath, Phys. Rev. A 85, 053625 (2012).
- (19) S. S. Natu and E. J. Mueller, Phys. Rev. A 87, 063616 (2013).
- (20) M. Greiner, O. Mandel, T. W. Hänsch and I. Bloch, Nature 415, 51 (2002).
- (21) J. Sebby-Strabley, B. L. Brown, M. Anderlini, P. J. Lee, W. D. Phillips, J. V. Porto, and P. R. Johnson, Phys. Rev. Lett. 98, 200405 (2007).
- (22) S. Will, T. Best, U. Schneider, L. Hackermüller, D.-S. Lühmann, and I. Bloch, Nature 465, 197 (2010).
- (23) K. W. Mahmud and E. Tiesinga, Phys. Rev. A 88, 023602 (2013).
- (24) S. Will, T. Best, S. Braun, U. Schneider, and I. Bloch, Phys. Rev. Lett. 106, 115305 (2011).
- (25) E. Tiesinga and P. R. Johnson, Phys. Rev. A 83, 063609 (2011).
- (26) R. T. Scalettar, G. G. Batrouni, and G. T. Zimanyi, Phys. Rev. Lett. 66, 3144 (1991).
- (27) J. Zakrzewski and D. Delande, AIP Conf. Proc. 1076, 292 (2008).
- (28) O. Penrose and L. Onsager, Phys. Rev. 104, 576 (1956).
- (29) M. Buchhold, U. Bissbort, S. Will, and W. Hofstetter, Phys. Rev. A 84, 023631 (2011).
- (30) J. K. Freericks and H. Monien, Phys. Rev. B 53, 2691 (1996).
- (31) M. A. Cazalilla, R. Citro, T. Giamarchi, E. Orignac, and M. Rigol, Rev. Mod. Phys. 83, 1405-1466 (2011).
- (32) G. Vidal, Phys. Rev. Lett. 98, 070201 (2007); 93, 040502 (2004).
- (33) U. R. Fischer and B. Xiong, Europhys. Lett. 99, 66003 (2012).
- (34) P. A. Lee and X.-G. Wen, Rev. Mod. Phys. 78, 17 (2006).
- (35) A. Albus, F. Illuminati, and J. Eisert, Phys. Rev. A 68, 023606 (2003).
- (36) P. Anders, P. Werner, M. Troyer, M. Sigrist, and L. Pollet, Phys. Rev. Lett. 109, 206401 (2012).
- (37) E. Altman, E. Demler, and A. Rosch, Phys. Rev. Lett. 109, 235304 (2012).
- (38) P. R. Johnson, E. Tiesinga, J. V. Porto, and C. J. Williams, New J. Phys. 11, 093022 (2009).
- (39) P. R. Johnson, D. Blume, X. Y. Yin, W. F. Flynn, and E. Tiesinga, New J. Phys. 14, 053037 (2012).
- (40) K. W. Mahmud, L. Jiang, E. Tiesinga, and P. R. Johnson, e-print arXiv:1310.4967 (2013).

## I Supplementary Material

We derive analytic results for quench dynamics of a one-dimensional Mott insulator using the strong-coupling expansion freericks96 . Bosons in an optical lattice are described by the Bose-Hubbard Hamiltonian of Eq. 1. We assume a homogeneous system with sites, periodic boundary conditions, and atom number , where and are positive integers.

We prepare the Mott-insulator ground state assuming a non-zero tunneling energy that is small compared to the atom-atom interaction strength . Following freericks96 we perform perturbation theory in the tunneling operator or kinetic energy. To zeroth-order the ground state wavefunction is the Mott state with atoms in each of the sites. The correction to this wavefunction is due to states containing a single hole with atoms in one of the sites and a single “particle” with atoms in another. We classify these (normalized) particle-hole pair states as

where is the number of sites that separates the hole and particle. The dots in a ket indicate sites with atoms and the sum is over all states with the same particle and hole separation. The hole is either to the left or right of the particle. (Periodic boundary conditions imply that the largest separation is .) The Bose-Hubbard Hamiltonian within the space spanned by is given by

where is the diagonal matrix element for state , with elements describes the coupling between and , and

is a tridiagonal (diagonal-constant) matrix that describes the couplings among the .

All diagonal elements of are the same, and we must first diagonalize in order to perform perturbation theory. Its eigenenergies and functions are

and

Performing perturbation theory with respect to the non-degenerate states , the correction to the ground state wavefunction is

in terms of the original particle-hole basis.

We now quench the system by suddenly setting the tunneling to zero and change . The subsequent time evolution is then given by

as the states and are eigenstates of the quenched Hamiltonian. In fact, the are degenerate with an energy relative to the state .

Our experimental observables are the quasi-momentum distribution at the point and the edge of the Brillouin zone, defined as , and , respectively. Here is the single-particle density matrix. Finally, the visibility is defined by

The single-particle density matrix can be evaluated using Eq. I and leads to

to second order in . Further analysis shows that is a positive-definite pentadiagonal matrix with diagonal-constant coefficients given by , , , and zero otherwise. In the thermodynamic limit, the quasi-momentum dynamics is given by

and the visibility is

We find that it is also useful to define , the largest eigenvalue of the single-particle density matrix at time . As is a real symmetric matrix, we can use perturbation theory to find its largest eigenvalue. We define with tri-diagonal matrix given by the diagonal and sub- and super-diagonal of . The matrix can be diagonalized analytically and its largest eigenvalue is

with corresponding eigenvector and elements

when , and

when . Here is any integer and index .

Finally, the matrix corrects the largest eigenvalue and we have

to first order in .