You are on page 1of 17

The 93rdStatistical

Mechanics Conference

RUTGERS UNIVERSITY

Order by Disorder
A new method for establishing phase transiti
ertain systems with continuous, n–component

Applications to transition metal oxides


Cast of
characters:
M. Biskup (UCLA Math)

And, the physicists:

Z. Nussinov Relevant papers


M. Biskup, L. Chayes, Z. Nussinov and J.
(Th. Div. Los van den Brink, Orbital order in classical
models of transition-metal compounds,
Alamos) Europhys. Lett. 67 (2004), no. 6, 990–996
M. Biskup, L. Chayes and S.A. Kivelson,
Order by disorder, without order, in a
J. van den Brink two-dimensional spin system with O(2)
symmetry, Ann. Henri Poincaré 5 (2004), no. 6, 1181–
(Lorentz ITP, Leiden) 1205.
M. Biskup, L. Chayes and Z. Nussinov, Orbital ordering in
transition-metal compounds: I. The 120-degree model,
Commun. Math. Phys. 255 (2005), no. 2, 253–292

S.A. Kivelson
UCLA Physics … more to come …
(Plenty
.)
ems of interest: Continuous spins O(2), O(3
– huge degeneracy in the ground state.
Example: 2D NNN Antiferromagnet.
(Say XY spins.)

What can this system do


O(2) × O (2)
Ground states: MW: Certainly no magnetizati
ow add “weak” NN ferromagnetic coupling.
General considerations:
O(2) × O (2).
Ground states still
xactly zero message from nearest neighbors.) ¿ Disorder?
Transition Metal Compounds
vels in 3d shell split by crystal field.
eg
d–orbitals

t2g
Single itinerant electron @ each site
h multiple orbital degrees of freedom.
ange approximation (and neglect of strain–field induced interactions amo
r, r′
H orb = J[4π̂ αr π̂ αr′ −2π̂ αr −2π̂ αr′ +1]
H = ∑ r, r ′
H orb (s r ⋅s r ′ + 1)
4 r −r′
 = direction of bond
<r, r ′ >
[Kugal–Khomskii Hamiltonian]
π̂ rx = 14 (−σ rz + 3σ rx) π̂ ry = 14 (σ rz − 3σ rx)
120º–model (eg–compounds)
V2 O3 , LiVO2, LaVO3, …
π̂ rz = 12 σ rz

x 1 x
rbital compass–model (t2g–compounds) π̂ r = 2 σ r π̂ ry = 12 σ ry
LaTiO3, …
π̂ rz = 12 σ rz
tal only approximation: Neglect spin degrees of freedom.

e S limit (for pseudo–spin operators): Go classical.

Classical 120º Hamiltonian: r


Sr an XY –spin
r
−H = ∑ (Sr[a]Sr[a] Sr = Sr ⋅â
[b] [b] [c] [c] [a ]
+ex + Sr Sr+ey + Sr Sr+ez )
r similarly for Sr[b ] & Sr[c] ,
â, b̂ and ĉ
unit vectors spaced @ 120

lassical orbital compass Hamiltonian: r


Sr = (Sr[x],Sr[y],Sr[z])
−H = ∑ (Sr[ x ]Sr+e
[x]
x
+ Sr
[ y] [ y]
Sr+ey + S [ y] [ z]
r Sr+ez )
r – usual Heisenberg spins.

licity, today focus on 2D version of orbital compas


The 2D Orbital Compass Model
r r
Sr ∈S1,writeSr = (Sr[x],Sr[y])May
. r ∈Λ L .
as well take

−H L = ∑ r r+ex r Sr+ey )
(S [x] [x]
S + S [ y] [ y]
LL torus
r∈Λ L

1
= − ∑ ( (Sr[ x ] − Sr+e
[x] 2
) + (Sr
[ y]
− S [ y] 2
r+ey ) )
+ constant.
2 r∈ΛL x

ractive couplings (ferromagnetic). other ground–states


ouples in x–direction with x–component. but these
play no rôle and will
ouples in y–direction with y–component. not be discussed.
Clear: Any constant spin–field is a ground state.
* O(2) symmetry restored *
(a) Not clear what are the “states”.
t even begin to talk about contours:
(b) No apparent “stiffness”.
¿Hints from SW–theory?
1
^ 1 1 1
Disaster: S(k) ∝ Δ (k )Δ (k )≈ k2 k2 .
x x y y x y

Indeed, spherical version of th


Very IR–divergent. model has no phase transition

red bounds only give upper bounds on the scattering


And MW–theorem (strictly speaking) does not apply.
he O(2) symmetry is in the ground states, not the Hamiltonian itself

Theorem. For the d = 2 orbital compass model,


for all  sufficiently large, (at least) two
limiting translation invariant Gibbs states.
One has < (S0 ) >close
[x] 2 (x)
β to one.
< (S0[y] )2 >close
The other has
( y)
β to one.
KeyInideas:
the physics literature since the early 80’s
J. Villain, R. Bidaux, J. P. Carton and R.
Conte, Order as an Effect of Disorder, J. Phys.
(Paris) 41 (1980),
E. F. Shender, no.11, 1263–1272.
Antiferromagnetic Garnets
with Fluctuationally Interacting
Sublattices, Sov. Phys. JETP 56 (1982)
178–184 .
C. L. Henley, Ordering Due to Disorder in a Really clarified matters;
Frustrated Vector Antiferromagnet, Phys.
Rev. Lett. 62 (1989) 2056–2059. put things on a firm
foundation in a general
itely many papers (mostly quantum) in which context.
specific calculat
All of this
ntribution to physics general theory: Modest.
works
even in d = 2.
(But TMO models of some topical interest.)

∞, weighting of various ground states


e into account more than just energetics:
ations of spins will contribute to overall statistical
se (spin–fluctuation) degrees of freedom will
elves organize into spin–wave like modes.
Can be calculated (or estimated).
Key ideas:
 < ∞, weighting of various ground states
ke into account more than just energetics: Gaussian like SW–free energeti
uations of spins contribute to overall statistical weight.
will tell us which of the grou
e (spin–fluctuation) degrees of freedom will states are actually preferred
es organize into spin–wave like modes.
@ finite temperature
Can be calculated (or estimated).

1) Remarks
Not as drastic an approximation as it sounds;
• Infrared divergence virtually non–existent
the level of free–energetics.
(2) In math–phys, plenty of “selection due
to finite–temperature excitations”.
• Excitation spectrum always with (huge) gap.
But:
• Finite (or countable) number of ground states.
Sr[ x ] = cos(θ + ϕ[r)= fixed “ground state
’s do calculation. Write:
Sr[ y] = sin(θ + ϕ r)
H, neglect terms of higher order than quadratic in

Well, got:Sr = cos(θ + ϕ r)= cosθ cosϕ r −sinθ sinϕ r


[x]

so Sr[ x ] −Sr[+x]ex = cosθ (cosϕ r −cosϕ r+ex )−sinθ (sinϕ r −sinϕ r+ex ).

Neglect quadratic (and beyond).


Will square this.
2 2
[x]
Get (S
[x]
r −S
r+ ex
2
) ≈ [sin θ ](ϕ r −ϕ r+ex )

and similarly
2 2
[y]
(S
[ y]
r −S
r+ ey
2
) ≈ [cos θ ](ϕ r −ϕ r+ey ) .
roximate Hamiltonian is therefore:
1 2
−β H A (θ ) = − β ∑ cos θ (ϕ r − ϕ r+ex ) + sin θ (ϕ r − ϕ r+ey ) ⎤.
⎡ 2
2 2

2 r ⎢
⎣ ⎥

1
ϕ
Go to transform variables:
r = ∑
Vol. k
^
ϕ k
e −ikr
.

So, after some manipulations,


(θ ) = −β ∑ | ϕ k | 2 ⎡
^
−β H A ⎣cos 2
θ (1 − cos k x ) + sin 2
θ (1 − cos k y ) ⎤
⎦.
k

total weight can easily be calculated:


1
ZA (θ )≡ ∫e −β H A (θ )
dϕ ∝ ∏ .
β (cos θ (1−coskx )+ sin θ (1−cosky ))
2 2
k

Fact that we are


talking about “” means Take logs:
that we do not
integrate over the k =
lim −log Z2A (θ ) = const. + 1 ∫ d 2 k log (( cos2 θ (1 − cos kx ) + sin 2 θ (1 − cos ky )))
L→ ∞ L 2
ant as small as possible.
Scales with 
Φ(θ ).
A free energy;
Pause to refresh: No difficulty doing these
integrals; some infrared “action” but no big deal
(logarithmic). We are only interested in a free
Nowenergy.
log is a (strictly) concave function:
log ((cos2 θ (1−coskx )+ sin2 θ (1−cosky ))) ≥ [cos2 θ]log(1−coskx )+[sin2 θ]log(1−cosky ).

ky integrals on RHS, these come out the same. We l


Φ(θ ) ≥ [cos 2 θ ] Φ(0) +[ sin 2 θ] Φ(0)
Φ(θ ≠ 0) π
se strict concavity,
–or–π
Φ(θ ≠ 2 )
> Φ(θ = 0) = Φ(θ = 2 ).
lculation indicates:
π
= 0 and θ =
There are “states” θat
2
which will dominate any other “state”.

Outline of a proof.

1) Define a fluctuation scale.


 β − 2 << Δ(β ) << β − 3
1 1

situation where each deviation variable ϕrr has:


< Δ(β )
– and hence –
ϕ r − ϕ r ′ < 2Δ(β )
3
Fact: βΔ << 1 that quadratic approximation is “good”.
means
2
>> 1 that the effective Gaussian variables
Fact: βΔmeans
– namely
βϕ r –
are allowed to get large.
Not hard to see:
Proposition: With the constraints (globally)
enforced, the spin wave formula for the – β → ∞.
dependent limiting free energy is asymptotic as
(2) Define a running length
Not important; for technical
scale, B – and another,
reasons, this is not ().
interrelated spin–scale, 
A block B of scale B is defined to be good if
Definition.

ϕ r ,ϕ r ′ ; r, r ′ ∈Λ B
ach neighboring pair of fluctuation variables
satisfies ϕ r − ϕ r ′ ≤ Δ.
b) θ r , r ∈Λsatisfies
Each spin variable B

π
θ r < ε – or– θ r − 2 < ε .(You can add .)

Clear: There are two types of good blocks.


Also: Two types of bad blocks.
ergetic disaster. (Should be suppressed
exponentially with rate ~ 2 )
More important, more interesting:
rgetics good, but  not particularly near
π
2 0 or

Can use method of chessboard estimates.


[FSS][FLIS(1)]
[FLIS(2)][LF] [DLS][FLIS(3)]
A] Gives estimates on probability of bad
blocks in terms of constrained partition
function where all blocks are bad blocks. N2
⎡Z L (A ) ⎤
L2
pe (i) indeed suppressed exponentially. P (A ) ≤ ⎢ ⎥
⎣ ZL ⎦
ype (ii) has probability <
~
bounded eby
2
− N ( Φ(θ )−Φ(0))

e N is the block size. This goes to Nzero


→ ∞.as

Note: Type (ii) is independent of .


Reiterate: Two
blocks form contour element which separate distinct types of
ions of the two distinct types of goodness. goodness. A single
box cannot exhibit
both types of
[FSS][FLIS(1)]
[FLIS(2)][LF] [DLS][FLIS(3)] goodness. Thus,
regions of the
B] Gives estimates on probability distinctive types of
goodness must be
of contour bad blocks by the separated by closed
product of the previously contours.
mentioned probability estimates.
• Standard Peierls
argument, implies
existence of two distinct
Remarks:
states. Same sort of thing true
for antiferromagnet, 120º–model & 3D
orbital compass model (sort of).
Interesting feature: Limiting behavior of
model as T goes to zero is not the same as
the behavior of the model @ T = 0.
lar, non–trivial stiffness at T = 0 (and presumably  as well
The End

You might also like