Anda di halaman 1dari 29

Supersolid phases of (lattice) bosons

Matthias Troyer (ETH Zrich)


Hebert, Batrouni, Scalettar, Schmid, Troyer, Dorneich, PRB 65, 014513 (2002) Schmid, Todo, Troyer, Dorneich, PRL 88, 167208 (2002) Schmid & Troyer, PRL 93, 067003 (2004) Sengupta, Pryadko, Alet, Schmid, Troyer, PRL 94, 207202 (2005) Wessel & Troyer, PRL 95, (2005, in press)

alps.comp-phys.org

Can a supersolid exist?


A supersolid show simultaneously density wave order (broken translational symmetry) and superuidity Experiment on Helium
Kim and Chan (2004): evidence for superuidity in solid Helium?

Theory: can Helium be supersolid?


Penrose and Onsager (1956): no Andreev and Lifshitz (1969), Chester (1970), Legget (1970): yes Legget (2004): maybe Anderson, Brinkman, Huse (2005): yes

Numerical simulations of supersolids


continuum models for Helium
Ceperley (2004): no superuidity in perfect Helium crystals Prokof ev and Svistunov (2005): supersolidity requires defects but no nal answer yet

supersolids on lattices
easier to investigate since the lattice is rigid ecient algorithms exist

relevance of lattice supersolids


physical insight gained can be applied to continuum supersolids can be realized in physical systems

Lattice bosons
4He lms on substrates

Cooper pairs in crystal

Josephson junction arrays

Atomic BEC in 3D optical lattice [Greiner et al., Nature (2002)]

BEC in cold bosonic atoms


Ultra-cold trapped 87Rb atoms form BEC rst observed 1995

Standing waves from laser superimpose an optical lattice (2002)

Boson-Hubbard model
describes bosonic atoms in optical lattice well understood without the trap: Fisher et al, PRB 1989
H = t
i,j

b bj + b bi + U i j
i

ni (ni 1)/2
i

ni

/U

Phase diagram (V=0) n=2

Incompressible Mott-insulator Integer lling

n =1

n=0

t /U

superuid

Quantum phase transition in trapped atoms


Experiment: Greiner et al, Nature (2002) Coherence vanishes as atoms enter Mott-insulating phase Measurements of momentum distribution
by taking real space image after expansion of the gas cloud

increasing U/t

M. Greiner et al, Nature (2002)

Super solids with longer range repulsion


Extended Bose Hubbard model, e.g. strong dipolar interactions in Chromium condensates
H = t
i,j

(a aj + a ai ) i j
i

ni + U
i

ni (ni 1)/2+V
i,j

ni nj

shows simultaneously solid order and superuidity

solid (crystal) caused by large V

doped solid: interstitials

supersolid: superuid interstitials

2 k

50 Width 25

Detecting the supersolid


clear and unambiguous signal!

Supersolid has doubled unit cell: additional coherence peaks 0


c

re 1 Schematic three-dimensional interference pattern with measured absorption Figure 1 Schematic three-dimensional interference p es taken along two orthogonal directions. The absorption images were obtained after images taken along two orthogonal directions. The abso stic expansion from a lattice with a potential depth of V 0ballistic expansion fromight of with a potential depth o 10E r and a time of a lattice ms. 15 ms.

ad

b 1

Figure 3 Restoring c of coherence after b c lowering the potenti atoms are rst held potential is decrease pattern of the atoms b, Width of the cent

Can a supersolid exist in lattice models?


Nearest neighbor repulsion model
Batrouni et al (1995): yes for hard-core bosons van Otterlo et al (1995): yes for hard and soft-core bosons Srensen et al (1996): yes for soft-core bosons Batrouni et al (2000): no for hard-core bosons

Supersolid in cold atoms


Goral, Santos and Lewenstein (2002): yes with dipolar interactions Bchler and Blatter (2003): yes in boson-fermion mixtures

What is the truth?

First simulation results: yes!


Batrouni, Scalettar, Zimany, Kampf, PRL (1995)
evidence for supersolid at 3% doping nite superuid density and solid structure factor

Later simulation results: maybe?


Batrouni and Scalettar, PRL (2000)
evidence for rst order transition in canonical energies but careful nite size scaling is still needed

Ecient Cluster algorithms for quantum systems


Which system sizes can be studied? temperature 1 0.1 0.1 0.005 Metropolis 16000 spins 200 spins 32 bosons modern algorithms 16000000 spins 1000000 spins 10000 bosons 50000 spins

Open source codes available at http://alps.comp-phys.org/

New simulations: no!


Simulations with 100 x more particles clearly show phase separation at rst order phase transition instead of supersolid
Hebert, Batrouni, Scalettar, Schmid, MT, Dorneich, PRB (2002) Schmid, Todo, MT, Dorneich, PRL (2002)

superfluid solid

0.5

0.55 0.6 Density

0.65

Supersolids versus phase separation


solid doped solid
2t2 = V

supersolid

G. Schmid PhD thesis, ETH (2004) Sengupta, Pryadko, Alet, MT , Schmid, PRL (2005)

2t2 = t < V

doped particles gain energy by forming a domain wall

Stabilizing the supersolid


It matters how we dope the solid
U>>4V: particles go onto empty sublattice and phase separate 4V>> U: particles go onto occupied sublattice and form supersolid
H = t
i,j

(a aj + a ai ) i j
i

ni + U
i

ni (ni 1)/2+V
i,j

ni nj

solid
V

dopants on same sublattice! supersolid


U/4 > t

qualitatively dierent supersolid!!

Phase diagram
Soft-core bosons with U/t = 20
P. Sengupta, L. Pryadko, F. Alet, MT, G. Schmid, PRL (2005) supersolid possible at >1/2 and small on-site repulsion U < 4V no supersolid at density <1/2 (superuid domain walls)
10

8 SF PS

SS

PS

4 PS 2 0.00 0.25 0.50 ! 0.75

SF

MI
1.00

CDW II

CDW I

U=20 t=1

Two alternative routes to supersolids


add next nearest neighbor hopping:
H =H t
i,j

(a aj + a ai ) i j

E = 4t

add next nearest neighbor repulsion:

H =H +V
i,j

ni nj

striped solid

doped solid

striped supersolid

Stability of triangular lattice supersolids

Triangular lattcie hard-core bosons


Classical limit: V >> t

=0

=1

=1/3

=2/3

all other densities: innitely degenerate ground states how is the degeneracy lifted by quantum uctuations?

Mean-eld calculations
3105

SUPERFLUIDS AND SUPERSOLIDS ON FRUSTRATED . . .

nd properly proportional to ero value of the superuid neighbor V can stabilize a isotropic s . One can also ith fractional lling in the r interactions. igate the properties of the wo-dimensional lattices. We interplay between frustrar, and disorder which has . ts of quantum uctuations. as and Anderson23,24 raised ems, quantum uctuations erromagnetic order even at frustration leads to an inor mean eld level not asmmetry of the Hamiltonian quantum or thermal ucnd select a unique ground nged order. Our models ex-

Murthy, Arovas, Auerbach, PRB (1997) =1 chemical potential superuid supersolid =1/3

=2/3

=0 t/V

Recent investigations
May 11, 2005: three preprints, to be published in PRL
Heidarian & Damle, cond-mat/0505257 Melko, Paramekanti, Burkov, Vishwanath, Sheng, Balents, cond-mat/0505258 Wessel & MT, cond-mat/0505298

July 26, 2005: one follow-up paper, submitted to PRL


Boninsegni & Prokof ev, cond-mat/0507620

Quantum Monte Carlo and analytical arguments


all nd supersolid not complete agreement on nature of supersolid

Quantum Monte Carlo


Phase diagram similar to mean-eld calculations canonical
1
8 6 full

grand-canonical
0.8
PS

/V

4 2 0 -2 0

solid =2/3 solid =1/3

0.6
superfluid

PS

solid =2/3

supersolid

supersolid 0.4
PS

superfluid
PS

solid =1/3

empty 0.1 0.2 t/V 0.3 0.4 0.5

0.2 0 0 0.1 0.2 0.3 0.4

t/V

0.5

Domain wall instability


Doping > 2/3 or < 1/3
superuid domain walls have lower energy than uniform supersolid phase separation instead of supersolid

Energy t2 /V

Energy t

Domain wall instability


Doping > 2/3 or < 1/3
superuid domain walls have lower energy than uniform supersolid phase separation instead of supersolid
1.2 1.1 1 0.9

0.8 0.7 0.6 0.5 3

0.66 /V=4 0.64 0.62 0.6 0.1 0.15 0.2 0.25 t/V t/V=0.1 t/V=0.2 t/V=0.3 4 5 6 7 8

jump in density at 1st order phase transition

/V

Supersolid for 1/3 < < 2/3


particle doping =1/3 solid

hole doping =2/3 solid

Numerical results: supersolid!


Simulations show both density wave order and superuidity
0.2

0.15

/V=3 t/V=0.1

0.1 /2 S(Q) / N 0 0 0.05 0.1 1/L 0.15 0.2

0.05

Summary & Conclusions


Simulations provide high-accuracy results for large bosonic systems Supersolids in square and cubic lattices
unstable towards formation of superuid domain walls need weak on-site interaction to stabilize a supersolid longer range interactions give striped supersolid (relevant for high-Tc cuprates?)

Supersolids on triangular lattice


two dierent supersolids stable in wide lling regime can be realized in Chromium BEC on optical lattices

Lessons for Helium supersolids


Superuid domain walls are main instability for supersolids do the Helium experiments just see superuid grain boundaries?

Thanks to my collaborators
ETH Zrich Guido Schmid Universit de Nice George Batrouni Frederic Hebert CEA Saclay Fabien Alet University of Tokyo Synge Todo UMass Amherst Nikolay Prokof ev Boris Svistunov UC Riverside Leonid Pryadko Pinaki Sengupta Universitt Stuttgart Stefan Wessel

Anda mungkin juga menyukai