strongly interacting fermions in optical lattices
DESCRIPTION
PhD thesis talkTRANSCRIPT
Strongly interacting fermionsin optical lattices
Arnaud Koetsier
PhD thesis talk
2
Outline
• Topics covered in this talk:Chapter 2: introduction to Hubbard and Heisenberg modelsChapter 3: cooling into the Néel stateChapter 4: imbalanced antiferromagnets in optical lattices.
• Topics in the thesis but not covered in this talk:Chapter 1: Introduction (mainly historical)Chapter 6: BEC-BCS crossover for fermions in an optical latticeChaoter 7: Analogy of the BEC-BCS crossover for bosonsJuicy details
3
Introduction
• Fermions in an optical lattice• Described by the Hubbard model• Realised experimentally [Esslinger ’05], fermionic Mott insulator recently seen [Esslinger ’08, Bloch ’08]• There is currently a race to create the Néel state
— How to achieve the Néel state in an optical lattice?
• Imbalanced Fermi gases• Experimentally realised [Ketterle ’06, Hulet ’06]• High relevance to other areas of physics (particle physics, neutron stars, etc.)
— How does imbalance affect the Néel state?
4
Fermi-Hubbard Model
Sums depend on:Filling NDimensionality (d=3)
On-site interaction: U Tunneling: t
Consider nearest-neighbor tunneling only.
The positive-U (repulsive) Fermi-Hubbard Model, relevant to High-Tc SC
H = −tPσ
Phjj0i
c†j,σcj0,σ + UPjc†j,↑c
†j,↓cj,↓cj,↑
5
Quantum Phases of the Fermi-Hubbard ModelFi
lling
Frac
tion
0
0.5
1
Mott Insulator (need large U)
Band Insulator
Conductor
Conductor
Conductor
• Positive U (repulsive on-site interaction):
• Negative U: Pairing occurs — BEC/BCS superfluid at all fillings.
6
• At half filling, when and we are deep in the Mott phase.hopping is energetically supressedonly spin degrees of freedom remain (no transport)
• Integrate out the hopping fluctuations, then the Hubbard model reduces to the simpler Heisenberg model:
U À t
Mott insulator: Heisenberg Model
kBT ¿ U
J =4t2
U
Szi =1
2
³c†i,↑ci,↑ − c†i,↓ci,↓
´S+i =c
†i,↑ci,↓
S−i =c†i,↓ci,↑
H =J
2
Xhjki
Sj · Sk
Spin ½ operators: S = 12σ
Superexchange constant (virtual hops):
7
Néel State
• The Néel state is the antiferromagnetic ground state for
• Néel order parameter measures amount of “anti-alignment”:
• Below some critical temperature Tc, we enter the Néel state and becomes non-zero.
0 Tc0
0.5
T
⟨n⟩
0 ≤ h|n|i ≤ 0.5
h|n|i
h|n|i
nj = (−1)jhSji
J > 0
8
Start with trapped 2-component fermi gas of cold atoms. The entropy is:
How to reach the Néel state: Step 1
Total number of particles: NFermi temperature in the harmonic trap: kBTF = (3N )1/3~ω
SFG = NkBπ2 T
TF
V =1
2mω2r2
Trapping potential:
9
Adiabatically turn on the optical lattice. We enter the Mott phase: 1 particle per site.
Entropy remains constant: temperature changes!
How to reach the Néel state: Step 2
10
How to reach the Néel state: Step 3
Prepare the system so that the initial entropy in the trap equals the final entropy below Tc in the lattice:
What is the entropy of the Néel state in the lattice?
To reach the Néel state:
SFG(Tini) = SLat(T ≤ Tc)
Final temperature is below Tc : we are in the Néel state. The entropy remained constant throughout.
11
• Mott insulator:
• Perform a mean-field analysis about the equilibrium value of the staggered magnetisation :
• Landau free energy:
hni
Mean-field theory in the lattice
H ' J
2
Xhiji
½(−1)inSj + (−1)jnSi − Jn2
¾
H =J
2
Xhjki
Sj · Sk
fL(n) =Jz
2n2 − 1
βln
∙cosh(
β|n|Jz2
)
¸− 1βln(2).
Entropy:
S = −N ∂fL(hni)∂T
Self-Consistency:
∂fL(n)
∂n
¯̄̄̄n=hni
= 0hni
12
Mean-field theory results
• Entropy in the lattice:
kBTc = 3J/2
0.02 0.04 0.06 0.080.0
0.2
0.4
0.6
ln(2)
T/TF
S/Nk B
Lattice EntropyTrap Entropy
Mott
Née
l
Heating
Cooling
Lattice depth,6ER
Tc = 0.036 TF
13
• Improve on “single-site” mean-field theory by including the interaction between two sites exactly:
• Incorporate correct low-temperature and critical behaviour
Improved 2-site mean-field theory
0.02 0.04 0.06 0.080.0
0.2
0.4
0.6
ln(2)
T/TF
S/Nk B
Lattice, MFTLattice, fluc.Trap
Mott
Née
lExample:
atoms with a lattice depth of :
Tc = 0.012 TF
40K
8ER
Tini = 0.059 TF
14
• Until now, • What happens if — spin population imbalance? • This gives rise to an overal magnetization
N↑ 6= N↓
• Add a constraint to the Heisenberg model that enforces
Heisenberg Model with imbalance
Effective magnetic field (Lagrange multiplier):
H =J
2
Xhjki
Sj · Sk −Xi
B · (Si −m)
m = (0, 0,mz)
mz = SN↑ −N↓N↑ +N↓
(fermions: )S = 12
N↑ = N↓ = N/2
hSi =m
B
15
• ground state is antiferromagnetic (Néel state)Two sublattices:
• Linearize the Hamiltonian
• Magnetization:
• Néel order parameter:
• Obtain the on-site free energy subject to the constraint (eliminates )
J > 0⇒
Mean field analysis
A, B
m =hSAi+ hSBi
2
n =hSAi− hSBi
2
SA(B)i = hSA(B)i i+ δS
A(B)i
f(n,m;B)
∇Bf = 0 B
B
A B
A
A
B
AB
B
B
A
A
16
Phase Diagram in three dimensions
0 Tc0
0.5
T
⟨n⟩
Add imbalance0.0
0.2
0.4mz0 0.3 0.6
0.91.2kBT�J
0.0
0.2
0.4n
mz
k B T
/J
0 0.1 0.2 0.3 0.4 0.50
0.5
1
1.5
m
nIsing:
m
n
Canted:
n 6= 0
n = 0
17
Spin waves (magnons)
• Spin dynamics can be found from:
• Imbalance splits the degeneracy:
−π2
π
2kd
h̄ω/Jz
00
0.1
0.2
0.3
0.4
0.5No imbalance: Doubly degenerate antiferromagnetic dispersion
Antiferromagneticmagnons: ω ∝ |k|
ω ∝ k2Ferromagneticmagnons:
dS
dt=i
~[H,S]
Gap:(Larmorprecessionof n)
18
Long-wavelength dynamics: NLσM
• Dynamics are summarised a non-linear sigma model with an action
• The equilibrium value of is found from the Landau free energy:
• NLσM admits spin waves but also topologically stable excitations in the local staggered magnetisation .
S[n(x, t)] =
Zdt
Zdx
dD
½1
4Jzn2
µ~∂n(x, t)
∂t− 2Jzm× n(x, t)
¶2− Jd
2
2[∇n(x, t)]2
¾
F [n(x),m] =
Zdx
dD
½Jd2
2[∇n(x)]2 + f [n(x),m]
¾
• lattice spacing:• number of nearest neighbours:• local staggered magnetization:
d = λ/2z = 2Dn(x, t)
n(x, t)
n(x, t)
19
• The topological excitaitons are vortices; Néel vector has an out-of-easy-plane component in the core
• In two dimensions, these are merons:• Spin texture of a meron:
• Ansatz:
• Merons characterised by:Pontryagin index ±½VorticityCore size λ
Topological excitations
n =
⎛⎝ pn2 − [nz(r)]2 cosφ
nvpn2 − [nz(r)]2 sinφ
nz(r)
⎞⎠nv = 1
nv = −1nv = ±1
nz(r) =n
[(r/λ)2 + 1]2
20
Meron size
• Core size λ of meron found by plugging the spin texture into and minimizing (below Tc):
• At low temperatures, the energy of a single meron diverges logarithmically with the system area A as
merons must be created in pairs.
0 0.1 0.20.3
0.4mz0.3
0.60.9
1.21.5
kBT�J0
2
4
6Λ
d
F [n(x),m]
Jn2π
2ln
A
πλ2
mz
k B T
/J
0 0.1 0.2 0.3 0.4 0.50
0.5
1
1.5
Meronspresent
Meron core size
21
Meron pairs
Low temperatures:A pair of merons with opposite vorticity, has a finite energy since the deformation of the spin texture cancels at infinity:
Higher temperatures:Entropy contributions overcome the divergent energy of a single meronThe system can lower its free energy through the proliferation of single merons
22
Kosterlitz-Thouless transition
• The unbinding of meron pairs in 2D signals a KT transition. Thisdrives down Tc compared with MFT:
• New Tc obtained by analogy to an anisotropic O(3) model (Monte Carlo results: [Klomfass et al, Nucl. Phys. B360, 264 (1991)] )
mzk B
T/J
0 0.05 0.10
0.02
0.04
0.06
n 6= 0
mz
k B T
/J
0 0.2 0.40
0.2
0.4
0.6
0.8
1
KT transition
MFT in 2D
23
Experimental feasibility
• Experimental realisation:Néel state in optical lattice: adiabatic coolingImbalance: drive spin transitions with RF field
• Observation of Néel stateCorrelations in atom shot noiseBragg reflection (also probes spin waves)
• Observation of KT transitionInterference experiment [Hadzibabic et al. Nature 441, 1118 (2006)]In situ imaging [Gericke et al. Nat. Phys. 4, 949 (2008); Würtz et al.arXiv:0903.4837]
24
Conclusion
• Néel state appears to be is experimentally achievable (just) by adiabatically ramping up the optical lattice – accurate determination of initial T is crucial.
• The imbalanced antiferrromagnet is a rich systemspin-canting below Tcferro- and antiferromagnetic properties, topological excitationsin 2D, KT transition significantly lowers Tc compared to MFT resultsmodels quantum magnetism, bilayers, etc.possible application to topological quantum computation and information: merons possess an internal Ising degree of freedom associated toPontryagin index
• Future work:incorporate equilibrium in the NLσM actiongradient of n gives rise to a magnetization (can possibly be used to manipulate topological excitations)effect of imbalance on initial temperature needed to achieving the canted Néel state by adiabatic cooling
25
• On-site free energy:
where• Constraint equation:
• Critical temperature:
• Effective magnetic field below the critical temperature:
Free energy and Tc for imbalanced AFM
f(n,m;B) =Jz
2(n2 −m2) +m ·B
− 12kBT ln
∙4 cosh
µ |BA|2kBT
¶cosh
µ |BB |2kBT
¶¸BA (B) = B− Jzm± Jzn
B = 2Jzm
m =1
4
∙BA|BA|
tanh
µ |BA|2kBT
¶+BB|BB |
tanh
µ |BB |2kBT
¶¸
Tc =Jzmz
2kBarctanh(2mz)
26
Finding KT transition: Anisotropic O(3) model
• Dimensionless free energy of the anisotropic O(3) model [Klomfass et al, Nucl. Phys. B360, 264 (1991)] :
• KT transition:
• Numerical fit:
βf3 = −β3Xhi,ji
Si · Sj + γ3Xi
(Szi )2
γ3(β3) =β3
β3 − 1.06exp[−5.6(β3 − 1.085)].
1.0 1.2 1.4 1.6 1.8 2.00.0
0.2
0.4
0.6
0.8
1.0
b3
g 3êH1+g
3L
27
Finding KT transition: Analogy with the anisotropic O(3) model
• Landau free energy:
0.0 0.1 0.2 0.3 0.4 0.50.0
0.5
1.0
1.5
2.0
2.5
3.0
m
gHm,bLê
J
βF =− βJ
2
XhI,ji
ni · nj + βXi
f(m,ni,β)
'− βJn2
2
XhI,ji
Si · Sj + βn2γ(m,β)Xi
(Szi )2
β3 =Jβn2
2
γ3 =2β3Jγ(m,β)
Mapping of our model toAnisotropic O(3) model:
Numerical fit parameter
0.02
1/Jβ =
0.2
0.4
0.60.8
Tc
28
2-Site Mean-Field Theory
1 2
• First Term: Treats interactions between two neighboring sites exactly,
• Second Term: Treats interactions between other neighbors within mean-field theory
H = JS1 · S2 + J(z − 1)|n|(Sz1 − Sz2) + J(z − 1)n2
• Improve on standard mean-field approach by including 2 sites exactly:
29
2-site mean-field theory: Néel order parameter
0 0.5 1 1.50
0.1
0.2
0.3
0.4
0.5
k BT/J⟨n⟩
2-site1-site
Comparison with 1-site mean-field theory:
• Depletion at zero temperature due to quantum fluctuations
Tc ' 1.44kBJ
• Lowering of Tc:
30
Entropy in the lattice: Three temperature regimes
• Low T: entropy of magnon gas
• High T: 2-site mean field theory result
• Intermediate T: non-analytic critical behaviour
Where, from renormalization group theory [Zinn-Justin]
S(T À Tc) = NkB
∙ln(2)− 3J2
64k2BT2
¸
S(T ¿ Tc) = NkB4π2
45
µkBT
2√3Jhni
¶3
d = 3, ν = 0.63, A+/A− ' 0.54
t =T − TcTc
→ 0±S(T = Tc) = S(Tc)±A±|t|dν−1
Tc = 0.957J/kB•Tc: from quantum Monte-Carlo [Staudt et al. ’00]:
31
Entropy in the lattice, T>Tc
First term: Critical behavior
Other terms: To retrieve correct high-T limit of 2-site theory. → Found by expanding critical term and subtracting all terms of lower order than in T than high-T expression, which is .
•Result:
•Function with the correct properties above Tc:
S(T ≥ Tc)NkB
' α1
∙µT − TcT
¶κ− 1 + κTc
T
¸+ ln(2)
α1 =3J2
(32κ(κ − 1)k2BT 2c )κ = 3ν − 1 ' 0.89
∼ 1/T 2
32
Entropy in the lattice, T<Tc
First term and last term: Critical behavior and continuousinterpolation with T>Tc result.
Other terms: Retrieve low-T behavior of magons, again found by expanding critical term and subtracting all terms of lower order than .
• Function for with the correct properties below Tc:
S(T ≤ Tc)NkB
= −α2∙µ
Tc − TTc
¶κ− 1 + κ
T
Tc− κ(κ− 1)
2
T 2
T 2c
¸+ β0
T 3
T 3c+ β1
T 4
T 4c
T 3
33
Entropy in the lattice: Coefficients for T<Tc
• Result:
α1 =3J2
(32κ(κ− 1)k2BT 2c )κ = 3ν − 1 ' 0.89
(same as high-T expression):
α2 =6
(κ− 1)(κ− 2)(κ− 3)
µ4π2k3BT
3c
135√3J3− α1(κ− 1) + β1 − ln(2)
¶β0 =
κ
(κ− 3)
µ4π2k3BT
3c
45√3κJ3
+ α1(κ− 1)− β1 + ln(2)
¶β1 = ln 2− J2
6(A+/A− + 1) + κ(κ− 5)64κk2BT
2c A
+/A−− 4π
2k3BT3c
135√3J3
34
Maximum number of particles in the trap
For smooth traps, tunneling is not site-dependent, overfilling leads to double occupancy:
The trap limits the number of particles to avoid double occupancy:
Destroys Mott-insulator state in the centre!
N ≤ Nmax =4π
3
µ8U
mω2λ2
¶3/2Example:
atoms with a lattice depth ofand
40K
λ = 755 nm
8ER
⇒ Nmax ' 3 × 106