Professional Documents
Culture Documents
Edwards-Anderson Model
Edwards-Anderson Model
Edwards-Anderson model
C. Itoi1 , K. Horie1 and Y. Sakamoto2 ,
1
Department of Physics, GS & CST, Nihon University,
arXiv:2103.00774v1 [math-ph] 1 Mar 2021
March 2, 2021
Abstract
It is proven that the ground state in the transverse field Edwards-Anderson model is
unique for a sufficiently weak transverse field in a d-dimensional finite cubic lattice. On the
basis of a recently obtained result for the low energy states in the Edwards-Anderson model,
the Kirkwood-Thomas expansion method developed by Datta-Kennedy enables us to show
that the uniqueness of ground state is preserved against a sufficiently small transverse field
in the transverse field Edwards-Anderson model.
1 Introduction
The transverse Ising model with random exchange interactions is a simple interesting quantum
spin model. Many researchers of information, mathematics and physics have been studied this
model extensively, since D-Wave Systems actually devised and produced a quantum annealer
based on this model [7], which is theoretically proposed by Finnila-Gomez-Sebenik-Stenson-Dol
[5] and Kadowaki-Nishimori [8]. On the other hand, the Edwards-Anderson (EA) model is a
well-known Ising model with short range random exchange interactions [3]. Quite recently, it is
proven that the ground state in the Edwards-Anderson (EA) model is unique in any dimension
for almost all continuous random exchange interactions under a boundary condition which breaks
the global Z2 symmetry [6]. Uniqueness of the ground state in the EA model at zero temperature
agrees with the claims of Fisher- Huse and Newman-Stein [4, 10].
In the present paper, it is proven that the uniqueness of ground state is preserved also in the
Edwards-Anderson model with a weak transverse field. The obtained uniqueness theorem for the
ground state without transverse field gives a convergent perturbative expansion for sufficiently
weak transverse field. A convergent perturbative expansion proposed by Kirkwood and Thomas
[9] and developed by Datta and Kennedy [1, 2] enables us to obtain a rigorous result for the
unique ground state under a weak transverse field.
1
Note |BΛ | = |ΛL |d. A bond spin σb denotes
for a bond b = {i, j} ∈ BΛ and a = x, y, z. Let J = (Jb )b∈BΛ be a sequence of independent and
identically distributed (i.i.d) random variables whose expectation value and variance are given
by
with a random
Psequence J. This Hamiltonian is invariant under a discrete SU(2) transformation
U := exp[iπ i∈ΛL σix /2] acting on each spin σiz 7→ U † σiz U = −σiz . This corresponds to Z2
symmetry in the Edwards-Anderson model. Define a state |σi with a spin configuration σ :
ΛL → {1, −1} by
σiz |σi = σi |σi.
For any β > 0, the partition function as a function of (β, h, J) is defined by
X
ZΛ (β, h, J) = Tre−βHΛ (σ,h,J ) = hσ|e−βHΛ (σ,h,J ) |σi, (3)
σ∈{1,−1}ΛL
The average of an arbitrary function f : ΣL → R of the spin configuration in the Gibbs state is
given by
1
hf (σ)i = Trf (σ)e−βHΛ (σ,h,J ) .
ZΛ (β, h, J)
Note that the expectation hσi i of spin at each site i vanishes in the Z2 symmetric Gibbs state.
To remove the trivial two-fold degeneracy due to the global Z2 symmetry, assume + boundary
condition, such that
σi = 1, (4)
for i ∈ ΛL \ΛL−2 . Σ+
Λ ⊂ {1, −1}
ΛL denotes the set of spin configurations satisfying the boundary
condition (4).
Theorem 2.1 The ground state of the transverse field EA model is unique for sufficiently small
|h|.
Proof. The derivative of one point function is represented in terms of the following Duhamel
function
1 ∂
hf (σ)i = (σbz , f (σ)) − hσbz ihf (σ)i, (6)
β ∂Jb
2
where the Duhamel function (A, B) between linear operators A and B is defined by
Z 1
(A, B) := dtheβtH Ae−βtH Bi.
0
Uniform bounds kf (σ)k ≤ C in the left hand side, −2C ≤ |(σbz , f (σ)) − hσbz ihf (σ)i| ≤ 2C on the
integrand in the right hand side, and the dominated convergence theorem imply the following
commutativity between the limiting procedure and the integration
Z J2
0 = lim dJb [(σbz , f (σ)) − hσbz ihf (σ)i] (7)
β→∞ J1
Z J2
= dJb lim [(σbz , f (σ)) − hσbz ihf (σ)i]. (8)
J1 β→∞
Since the integration interval (J1 , J2 ) is arbitrary, the following limit vanishes
Lemma 3.2 Consider the transverse field Edwards-Anderson (EA) model at h = 0 in d-dimensional
hyper cubic lattice ΛL under the boundary condition (4). Let f (σ z ) be a real valued function of
a spin operators . For almost all J, there exists a unique spin configuration s+ ∈ Σ+ Λ , such that
the following limit is given by
Proof. Consider the model at h = 0. In this case, the model becomes the classical Edwards-
Anderson model, then the Duhamel function becomes normal correlation function
Lemma 3.1 indicates the following conistent property of the bond spin configuration at zero
temperature. Eq.(5) in Lemma 3.1 for an arbitrary bond b ∈ BΛ and f (σ) = σb implies
3
Since either pb = 1 or pb = 0 is valid, either a ferromagnetic σb = 1 or an antiferromagnetic
σb = −1 bond spin configuration appears almost surely on any bond b ∈ BΛ for almost all J at
zero temperature. Consider a plaquette (i, j, k, l) with an arbitrary i ∈ ΛL and j = i + e, k =
i + e′ , l = i + e + e′ for unit vectors with |e| = |e′ | = 1. Lemma for b = {i, j}, {i, k} and
f (σ) = σjz σlz , σkz σlz implies
lim [hσiz σjz σjz σlz i − hσiz σjz ihσjz σlz i] = 0, (13)
β→∞
lim [hσiz σkz σkz σlz i − hσiz σkz ihσkz σlz i] = 0. (14)
β→∞
These and (σjz )2 = (σkz )2 = 1 give the consistent property of the bond spin configuration
For any site i ∈ ΛL and for b = {i, j} ∈ BΛ , Lemma 3.1 and f (σ) = σiz imply
for almost all J. For any sites i, j ∈ ΛL and i, j are connected by bonds in BΛ . Then, the +
boundary condition σi = 1 given by (4) and a bond spin configuration (σi σj ){i,j}∈BΛ fix a spin
configuration s+ ∈ Σ+ +
Λ uniquely at zero temperature for any L. This spin configuration s gives
lim hσiz i = s+
i ,
β→∞
where Λ′L := ΛL \ (ΛL \ ΛL−2 ), and an indicator I(e) for an arbitrary event e is defined by
I(true) = 1 and I(false) = 0. Let ψ(σ) be a function of spin configuration, and express ground
state of the Hamiltonian X
|GSi = σD ψ(σ)|σi,
σ∈Σ+
Λ
4
where D := {i ∈ ΛL |s+
i = −1}. Note that ψ(σ) = 1 for h = 0 is given by the ground state spin
configuration s+ ∈ ΣΛ defined by Lemma 3.2. The eigenvalue equation defined by
is written in X X
−( Jb σbx + h σiz )|GSi = E0 |GSi.
b∈BL i∈ΛL
where σ (i,j) denotes a spin configuration replaced by (σi , σj ) → (−σi , −σj ). This eigenvalue
equation can be represented in terms of ψ(σ).
(b)
X X
Jb σD ψ(σ (b) ) + hσi σD ψ(σ) = −E0 σD ψ(σ). (17)
b∈BL i∈ΛL
Therefore
(b)
X σD ψ(σ (b) ) X
Jb + hσi = −E0 . (18)
σD ψ(σ)
b∈BL i∈ΛL
To obtain the Kirkwood-Thomas equation for the ground state, represent the function ψ(σ) in
terms of a complex valued function g(X) of an arbitrary subset X ⊂ Λ′L ,
h 1 X i
ψ(σ) = exp − g(X)σX . (19)
2 ′ X⊂ΛL
ψ(σ (b) ) h X i
= exp I(b ∈ ∂X)g(X)σX
ψ(σ) ′X⊂ΛL
∂X := {{i, j} ∈ BL |i ∈ X, j ∈
/ X or j ∈ X, i ∈
/ X}.
Note also
(b)
σD
= s+
b .
σD
These and the eigenvalue equation (18) give
X h X i X
Jb s+
b exp I(b ∈ ∂X)g(X)σ X + hσi = −E0 . (21)
b∈BL X⊂Λ′L i∈ΛL
5
then we have
X X X
Jb s+
b + E0 + Jb s+
b g(X)σX
b∈BL X⊂Λ′L b∈∂X
X h X i X
+ Jb s+
b exp
(2)
I(b ∈ ∂X)g(X)σX + hσi = 0, (23)
b∈BL X⊂Λ′L i∈ΛL
where
∞
X xk
exp(2) x := ex − 1 − x = .
k!
k=2
The orthonormalization property (16) gives
∞ k
X X X 1 X Y
E0 = − Jb s+
b − Jc s+
c δX1 △···△Xk ,φ g(Xl )I(c ∈ ∂Xl ), (24)
k!
b∈BL c∈BL k=2 X1 ,··· ,Xk ⊂Λ′L l=1
and
∞ k
−1 X
+
X 1 X Y
g(X) = +
J s
c c δX1 △···△Xk ,X g(Xl )I(c ∈ ∂Xl ) + hδ|X|,1 ,
k!
P
b∈∂X Jb sb c∈B k=2L X1 ,··· ,Xk ⊂Λ′L l=1
=: F (g)(X) (25)
where X△Y := (X ∪ Y ) \ (X ∩ Y ) for arbitrary sets X, Y , and we have used σX σY = σX△Y . The
first term in the ground state energy is identical to that of the ground state spin configuration
+
s+ for h = 0, and the excited energy of a spin configuration σ for
P
h = 0 is 2 b∈∂X Jb sb ,
where X := {i ∈ ΛL |σi 6= s+ +
P
i }. Lemma 3.2 guarantees the positivity b∈∂X Jb sb > 0. To prove
uniqueness of the ground state in the transverse field EA model with a given J for sufficiently
small h, define a norm for the function g(X) by
X X
kgk := sup I(c ∈ ∂X) Jb s+
b |g(X)|(|h|M )
−w(X)
, (26)
c∈BL X⊂Λ b∈∂X
L
where w(X) is a number of elements of the smallest connected set which contains X. We say
that a set X is connected, if for any i, j ∈ X there exists a sequence i1 , i2 , · · · , in ∈ X, such
that i1 = i, in = j and {ik , ik+1 } ∈ BL for k = 1, · · · , n − 1. Then, the following lemma can be
proven.
4
Lemma 3.3 There exist a constant M > 0 and define δ = M , such that if |h|M ≤ 1
kF (g) − F (g ′ )k ≤ kg − g′ k/2, for kgk, kg ′ k ≤ δ, (27)
and kF (g)k ≤ δ for kgk ≤ δ.
Proof. For lighter notations, define △k := X1 △ · · · △Xk , and b : X ⇔ b ∈ ∂X. The norm
kF (g) − F (g ′ )k is represented in
kF (g) − F (g ′ )k (28)
∞ k k
X X X 1 X Y Y
= sup Jb s+ δ△k ,X [ g(Xl ) − g′ (Xl )](|h|M )−w(X)
b k!
c∈BL b∈BL ′ c:X⊂ΛL k=2 b:X1 ,··· ,Xk ⊂Λ′L l=1 l=1
∞ k k
X X 1 X Y Y
≤ sup Jb s+ g(X ) − g ′
(Xl (|h|M )
) −w(△k )
.
b l
k!
c∈BL b∈B k=2 l=1 l=1
L b:X1 ,··· ,Xk ⊂Λ′L ,c:△k
∞ k k
X X 1 X Y Y
≤ sup Jb s+ g(Xl ) − g ′ (Xl )(|h|M )−w(△k ) .
b (k − 1)!
c∈BL b∈B k=2
L b:X1 ,··· ,Xk ⊂Λ′L ,c:X1 l=1 l=1
6
∂△k ⊂ ∪kl=1 ∂Xl and permutation symmetry in the summation over X1 , · · · , Xk have been used.
Inequalities w(△k ) ≤ kl=1 w(Xl ), and
P
Yk k
Y X k Y
l−1 k
Y
′ ′
g(X ) − g (X ) ≤ |g(X )||g(X ) − g (X )| |g′ (Xj )|
l l j l l
l=1 l=1 l=1 j=1 j=l+1
kF (g) − F (g ′ )k (29)
∞ k Y
l−1 k
X X 1 X X Y
≤ sup Jb s+
b |g(Xj )||g(Xl ) − g ′ (Xl )| |g′ (Xj )|(|h|M )−w(△k )
c∈BL b∈B (k − 1)!
L k=2 b:X1 ,··· ,Xk ,c:X1 l=1 j=1 j=l+1
∞ l−1
k Y k Pk
1
|g′ (Xj )|(|h|M )− w(Xj )
X X X X Y
≤ sup Jb s+
b |g(Xj )||g(Xl ) − g ′ (Xl )| j=1
c∈BL b∈B (k − 1)!
L k=2 b:X1 ,··· ,Xk ,c:X1 l=1 j=1 j=l+1
∞ k Y
l−1
X X 1 X X
= sup Jb s+
b |g(Xj )|(|h|M )−w(Xj )
c∈BL (k − 1)!
b∈BL k=2 b:X1 ,··· ,Xk ,c:X1 l=1 j=1
k
Y
×|g(Xl ) − g′ (Xl )|(|h|M )−w(Xl ) |g ′ (Xj )|(|h|M )−w(Xj )
j=l+1
∞ k l−1 X
X 1 X X X Y
= sup Jb s+
b |g(X1 )|(|h|M )−w(X1 ) |g(Xj )|(|h|M )−w(Xj )
c∈BL k=2 (k − 1)!
l=1 b∈BL X1 :b,c j=2 Xj :b
X k
Y X
× |g(Xl ) − g ′ (Xl )|(|h|M )−w(Xl ) |g′ (Xj )|(|h|M )−w(Xj )
Xl :b j=l+1 Xj :b
∞ k l−1 k ∞ k
X 1 XY Y X 1 X
≤ kg − g ′ k kgk/∆ kg′ k/∆ = kg − g ′ k (kgk/∆)l−1 (kg ′ k/∆)k−l
(k − 1)! (k − 1)!
k=2 l=1 j=1 j=l+1 k=2 l=1
∞
X k(δ/∆)k−1
≤ kg − g ′ k = Kkg − g ′ k,
(k − 1)!
k=2
where an averaged energy gap ∆ > 0 over excited states above the unique ground state in the
unperturbed model is defined by
h X i−1 X X
∆ := sup |g(X)|(|h|M ) −w(X)
sup Jb s+
b |g(X)|(|h|M )
−w(X)
c∈BL c:X⊂Λ c∈BL c:X⊂Λ b∈∂X
L L
h X i−1
= sup |g(X)|(|h|M )−w(X) kgk, (30)
c∈BL c:X⊂Λ
L
1
and K := eδ/∆ (1 + δ/∆) − 1. The condition K = 2 fixes δ/∆, and then the following inequality
X δ
sup |g(X)|(|h|M )−w(X) ≤ ,
c∈BL c:X⊂Λ ∆
L
requires sufficiently large M > 0. To obtain the bound on kF (g)k, let us evaluate kF (0)k first.
Since
−hδ|X|,1
F (0)(X) = P +,
b∈∂X Jb sb
7
the norm is given by
X δ
kF (0)k = sup I(c ∈ ∂X)|h|δ|X|,1 (|h|M )−w(X) = 2M −1 = ,
c∈BL 2
X⊂Λ′L
Hence,
kgk δ
kF (g)k = kF (g) − F (0) + F (0)k ≤ kF (g) − F (0)k + kF (0)k ≤ + ≤ δ.
2 2
This completes the proof of Lemma 3.3.
Proof of Theorem 2.1. Lemma 3.3 and the contraction mapping theorem enable us to prove
that the fixed point equation
F (g) = g (31)
has unique solution g satisfying kgk ≤ δ. This completes the proof of Theorem 2.1.
References
[1] Datta, N., Kennedy, T., Expansions for one quasiparticle states in spin 1/2 systems J. Stat.
Phys. 108,373-399(2002)
[2] Datta, N., Kennedy, T., Instability of interfaces in the antiferromagnetic XXZ chain at zero
temperature Commun. Math. Phys. 236, 477-511 (2003)
[3] Edwards,S. F., Anderson, P. W. : Theory of spin glasses J. Phys. F: Metal Phys. 5, 965-
974(1975)
[4] Fisher, D. S., Huse, D. A., :Ordered phase of short-range Ising spin glasses Phys. Rev.
Lett. 56 1601 (1986).
[5] Finnila, A., Gomez, M., Sebenik, C., Stenson, C., Doll, J., : Quantum annealing: a
new method for minimizing multidimensional functions. Chemical physics letters 219 343-
348(1994)
[6] Itoi, C. :Uniqueness of ground state in the Edwards-Anderson spin glass model J. Phys.
Soc. Jpn. 90, 033002, 1-4 (2021)
[7] Johnson, M. W., Amin,M. H. S., Gildert, S., Lanting, T., Hamze, F., Dickson, N., Harris,
R., Berkley, A. J., Johans- son, J. , Bunyk, P., Chapple, E. M. , Enderud, C., Hilton, J. P.,
Karimi, K., Ladizinsky, E., Ladizinsky, N., Oh, T., Permi- nov, I., Rich, C. , Thom, M. C.,
Tolkacheva, E., Truncik, C. J. S. , Uchaikin, S., Wang, J., Wilson, B., Rose, G. :Quantum
annealing with manufactured spins, Nature 473 194-198. (2011)
[8] Kadowaki, T., Nishimori, H., :Quantum annealing in the transverse Ising model, Phys.Rev.
E 58 5355-5363(1998).
[9] Kirkwood, J. R., Thomas, L. E., :Expansions and Phase Transitions for the Ground State
of Quantum Ising Lattice Systems, Commun. Math. Phys.88, 569-580 (t983)
[10] Newman, C. M., Stein, D. L.: Non-mean-field behavior of realistic spin glasses.Phys. Rev.
Lett. 76515-518 (1996).