Sie sind auf Seite 1von 12

Quantum Phase Transitions

The Classical and Quantum Phase Transition for The Ising Model

The University of Alabama

Alexander Daniels
May 7, 2018

Introduction

Preface

A quantum phase transition differs from a classical phase transition in that it occurs at or
near T = 0, and is therefore a phase transition with respect to the unexcited (ground) states of
a system. This is in contrast to classical phase transitions which occur at finite temperatures
much larger than zero for excited states of a system . In this T ≈ 0 limit, thermal fluctuations
are non-existent and quantum fluctuations i.e., fluctuations away from a particular ground
state, are dominant. In this paper, a brief review of classical phase transitions, and a discus-
sion of a simple example - the transverse field (Quantum) Ising model is presented, along with
a brief remark on the superconductor-insulator transition as a notable example of quantum
phase transitions in modern experimental work.

A Brief Review of Classical Phase Transitions


1

Phases & Transitions A phase is a set of states of a physical system. These states share a
consistent macroscopic behavior or order, despite small perturbations of an external param-
eter, like temperature or an externally applied magnetic field. Such states are characterized
by a well-defined thermodynamic potential. There are usually several distinct phases for a
particular system, each with a particular (macroscopic) order. If we perform a measurement
of an observable as a function of an externally controlled parameter of our choosing - say,
temperature or pressure, we would produce a phase diagram much like that shown in Fig.1.
For this figure, note that the observable we measure may simply be the appearance of the
system. The plot is obtained after varying temperature for several fixed values of pressure
and vice versa. In this figure, one may observe that in the lower region, the distinct phases

1 For a detailed discussion, see [6]. This summary is inspired by the brief summary in [2].

1
Pressure p

Solid

Liquid

Gas

Tt Tc Temperature T

FIG. 1: A generic classical phase diagram illustrating the existence of several phases,
first-order transition lines separating them, and the critical and triple points located at Tc
and T t respectively.

are divided. Such divisions in the diagram consist of states of the system which are unstable
with respect to infinitesimal variations in the external parameters (temperature or pressure)
i.e., for minute changes of a parameter, one will observe large deviations in the behavior of
the system. Such states accompany a phase transition - a phenomenon at the boundary or
point between distinct phases of the physical system.

Types of Transitions and Critical Points Recall that according to statistical mechanics,
the macrostate of a system is the set of all microstates which satisfy a collection of constraints.
The microstate of the system is a particular configuration of constituent atoms or domains
(with an appropriately chosen scale). The observable quantity is defined to be an expectation
of the action of the corresponding operator on the microstate. Armed with a description of
the microstates, we may construct a partition function which displays non-analytic behavior
(divergent derivatives of first or higher order) at a phase transition.2 First-order divergences
correspond to first-order (discontinuous) transitions, while higher-order divergences corre-
spond to second-order (continuous) transitions.3 Continuous phase transitions occur near
the vicinity of a critical point.

2 For a proper discussion, see [3].


3 In particular, the entropy changes discontinuously or continuously for first-order and second-order respec-

tively.

2
Landau Theory, Symmetry, and the Order Parameter

The Order Parameter The order parameter itself is an average, whose finite value indi-
cates an ordered state of the system. Hence, a plot of the order parameter against an external
or control parameter illustrates the phase diagram for a system, as shown in FIG. 2. It is also
approximately zero near a critical point i.e., it vanishes when a distinction between phases
cannot be made4 . This property allows for a Taylor expansion of a relevant functional in the
neighborhood of a critical point. Classically, whenever the order parameter φ is specified for
a system, there is a corresponding ordering field h such that h ⇒ 0 =⇒ φ ⇒ φ0 = 0 for T ≥ Tc
and φ0 6= 0 for T < Tc .

Landau Theory Landau suggested the existence of a generalized thermodynamics poten-


tial f that depends on the control parameter x and order parameter φ. He also postulated
that it is non-singular in φ and the equilibrium state is given by ∂φ |φ=φeq f = 0. The former
property allows a Taylor expansion for φ ≈ 0 (near the critical point) such that (also given a
corresponding ordering field h):5

a 2 b 4
µ ¶
f ≈ f 0 φ + φ + ... − φh
2 4

Finally, he also noted that symmetries may be accounted for by appropriate choice of terms
in the expansion. For example, if the free energy is invariant under a change of sign (say, of
spin), then the free energy should only include even terms in φ which respect such a (or an
Ising) symmetry. This is the case for the above potential.

The Landau Functional If we take the order parameter as spatially varying above an ap-
propriate scale such that:
φ = φ(~
r)
r | mod a À 1 and a is the grain size.
where |~

The spatial variation of φ modifies the form of the (general) potential, yielding:

a 2 b 4
· ¸
¢2
f (T, H , φ(~ φ + φ + ζ20 5~ ~r φ(~
r ) − φ(~
¡
r )) = f n + f 0 r )h
2 4
which preserves spatial inversion symmetry (no odd powers of gradient) [2].

Finally, one may intergrate the above Landau functional to obtain the free energy:
Z
F = d~ r f (~
r)

4 See Chapter 17 of [6]


5 [2]

3
Critical Phenomena & Ginzburg Landay Theory 6

Near the critical point, a number of physical quantities of relevance display singularities.
These singularities are often expressed as power laws which feature a set of critical expo-
nents in their description. In general, Ginzburg-Landau Theory allows us to calculate these
quantities.

Since continuous transitions are characterized by ’cooperative behavior’ over some scale, we
expect that the correlation of the order parameter between two points in space might indi-
cate a phase transition. In fact, the correlation length ξ diverges at the critical point, signaling
global or long-range order in the system, corresponding to a distinct order or phase of the
system.

In particular, the correlation function

G(~ r 0 ) = 〈φ(~
r ,~ r 0 )〉 − 〈φ(~
r )φ(~ r 0 )〉 = 〈φ(~
r )〉〈φ(~ r 0 )〉 − φ20
r )φ(~
Using the Landau functional, one may calculate the correlation function to find that
G(r ) ∝ r 2−d i.e., the correlation function diverges at the critical point for d > 2. This means
that a phase transition occurs for systems with a dimensionality that is greater than two.

One typically finds that the correlation function takes the form:7

~
à !
~
r − r 0
r ,~
G(~ r 0 ) ∝ exp
ξ
demonstrating how the correlation length might enter the description.

The Classical Ising Model

The Ising model, although straightforward in its formulation, is versatile in its application;
by studying the Ising model, one gains great insight into many-body systems of greater com-
plexity. It is a fundamental in understanding classical and quantum phase transitions and
shall serve as a context in which the theory of classical and quantum phase transitions shall
unfold. We shall see that for the classical Ising model (with or without a longitudinal external
magnetic field), thermal fluctuations govern the behavior of the system, and hence the order
parameter. For the quantum case, quantum fluctuations are responsible for phase transitions
for T = 0.

First consider a one-dimensional chain of spin 1/2 particles with periodic boundary condi-
tions and a magnetic field which is parallel or anti-parallel to their spin. Neighboring spins

6 See § 2 of [2]
7 See § 2 of [2]

4
Order Parameter φ Phase 1 Phase 2

Control Parameter x
FIG. 2: A Sketch of a General Phase Diagram

interact with each other in such a way that their behavior suggests a preference for unity be-
tween them. The Ising chain hamiltonian, with a longitudinal magnetic field, is written:8

N
σiz σzj − µB σiz
X X
H N (σ1 ..σN ) = −J
〈i j 〉 i =1

where σi = σiz = |↑〉i or |↓〉i With knowledge of the Hamiltonian, one may immediately write
down the partition function, as is customary. The partition function takes the form:
à !
N z
= exp β −J σi σi +1 − µB
X X X X X
Z N (B, T ) = ...
σ1 =±1 σ2 =±1 σN =±1 i =1 i =1

Evaluating this partition function is quite cumbersome. Even so, one may gain valuable in-
formation and by utilizing the mean field approximation. In this approximation, each spin
interacts with the mean field - a representation of the average orientation of neighboring
spins. By doing so, we reduce the pairwise spin interaction term to an effective single-spin
term.

The general d -dimensional Ising Hamiltonian may be written as a sum of pairwise interac-
tions for neighboring spins. This Hamiltonian, without the contribution from the external
field is the ferromagnetic Ising Hamiltonian:

σi σ j
X
HF = −J
〈i , j 〉

8 This form is valid for unspecified dimensionality and is therefore general. For one dimension, the sum over

nearest neighbors 〈i j 〉 σiz σzj reduces to iN=1 σi σi +1


P P

5
QN QN
which has a ground state (state of minimum energy) |GS〉 ≡ |0〉 equal to i =1 |↑〉i or i =1 |↓〉i .
In the mean-field approximation, we have

σi σ j = 〈σi 〉σ j + σi 〈σ j 〉 − 〈σi 〉〈σ j 〉 + (σi − 〈σi 〉)(σ j − 〈σ j 〉)

.
Noting that the average spin is translationally invariant (〈σi 〉 = 〈σ〉), the Hamiltonian simpli-
fies:9

N Jq
σi + 〈σ〉2 − J
X X
HF (σ1 ..σN ) = −J q〈σ〉 (σi − 〈σi 〉)(σ j − 〈σ j 〉)
i 2 〈i , j 〉

Neglecting the last term, which accounts for spin fluctuations, the mean field result is ob-
tained.
N Jq
H IM F = −J q〈σ〉 σi + 〈σ〉2
X
i 2

Or, if an external magnetic field is included, we have H IM F = HFM F + HLM F or:

N Jq N
H IM F (σ1 ...σN ) = −J q〈σ〉 σi + 〈σ〉2 − µ(B M F + B ) σi
X X
i 2 i =1

N Jq N Jq
By taking U M F = 〈H IM F 〉 = −J q〈σ〉(N 〈σ〉) + 2 〈σ〉2 = − 2
2 〈σ〉 , one may finally calculate
the partition function for the mean field approximation:

à !
β N
µ ¶
exp − N J q〈σ〉2 exp βµ(B M F + B ) σi
X X X
Z (N , T, B, 〈σ〉) = ...
σi =±1 σN =±1 2 i =1
¶" #N
β
µ
= exp − N J q〈σ〉2 exp βµ(B M F + B )σ
X ¡ ¢
2 σ=±1
β
µ ¶
¢¤N
= exp − N J q〈σ〉2 2 cosh βµ(B M F + B )
£ ¡
2

such that the free energy is given by:

1
F = −kT ln Z = N q J 〈σ〉2 − N kT ln 2 cosh βµ(B M F + B )
£ ¡ ¢¤
2

and therefore, the quantity of interest - average spin - may be calculated10 to obtain:

qJ
· µ ¶¸
−∂B F = 〈σ〉 = tanh βµ 〈σ〉 + B
µ
with µB M F = q J 〈σ〉.

9 Since P 2 Nq 2 q PN
〈i , j 〉 〈σ〉 = 2 〈σ〉 and 〈i , j 〉 σi = 2 i =1 σi , where q is the number of neighbors per site.
P
10 or the dipole moment D = µ〈PN σ 〉 = N µ〈σ〉
i =1 i

6
Setting x ≡ βq J 〈σ〉 + βµB , in the absence of an external field( B = 0 ) we have:
x
= tanh x
βq J

T 1
Define Tc = βq J to obtain:

T
µ ¶
x = tanh x
Tc

This expression suggests that for T ≥ Tc , x = 0 which implies that 〈σ〉 = 0. However, for T < Tc
〈σ〉 6= 0 11
3
Near the critical temperature Tc , we may expand the right hand side ( tanh ≈ x − x3 + ...). Ne-
glecting higher order terms, this may be simplified to obtain the expression which describes
the behavior of the system about the critical point:

T
µ ¶
x2 = 3 1 −
Tc
¶¸ 1
T
· µ
2
x0 = 3 1 −
Tc
The above mean field expression features a critical index of 12 for 〈σ〉 since x|B =0 = x 0 = 〈σ〉 TTc .
A plot of x 0 or 〈σ〉 versus the reduced temperature TTc yields a phase diagram, displaying an
ordered 〈σ〉 6= 0 and disordered 〈σ〉 = 0 phase.

Quantum Phase Transitions

The Quantum/Transverse-Field Ising Model

Slightly modifying the classical model12 and its notation, the quantum Ising Hamiltonian:

N N
σiz σzj − Γ σix
X X
HT = −J
〈i , j 〉 i =1

= HF + HP

Or, in simplified notation:

11 Note, however, that in order for a phase transition to exist, the number of neighbors should be larger than two

(q > 2).
12 The modification amounts to adding the paramagnetic (transverse field) Hamiltonian to the ferromag-

netic (pairwise interaction) Hamiltonian which is common to both the longitudinal (classical) and trans-
verse(quantum) models.

7
FIG. 3: A Phase Diagram depicting the Quantum Critical Point and Quantum Critical Region
for an arbitrary system. Reproduced from [2].

N
h~j • σ
~i
X
HT = −
〈i , j 〉

where h~j = J σzj ẑ − Γx̂ and σ


~i = σix x̂ + σiz ẑ.

The paramagnetic part HP (Let J ⇒ 0) has a ground state |GS〉 ≡ |0〉 equal to

N
Y 1
|→〉i = p (|↑〉i + |↓〉i )
N /2
i =1 2
.
The ground state as a superposition of spin-up and spin-down hints that HP introduces
quantum uncertainty (flucuations) into the model.

Once again, a mean-field solution is sought. Following the development of the classical
model, the mean field Hamiltonian is now:

N Jq z 2 N
HTM F = −J q〈σz 〉 σiz + 〈σ 〉 − Γ σix
X X
i 2 i =1

Moreover,

N Jq z 2
〈HTM F 〉 = − 〈σ 〉 − ΓN 〈σx 〉
2
Jq
µ ¶
= N − 〈σz 〉ẑ − Γx̂ • σx x̂ + σz ẑ
¡ ¢
2
~ σ〉
= N h M F • 〈~
=U

8
Therefore, in parallel development with the classical case, we expect that:

σ〉 = tanh β|~
³ ´
〈~ h| ĥ
q q
where |~
Jq
h| = Γ2 + ( 2 〈σz 〉)2 = Γ2 + J 02 〈σz 〉2 , so that:

´ µ J 〈σz 〉 ¶ ´µ Γ ¶
〈σ 〉 = tanh β|~ ~
³ ³
z 0 x
h| ; 〈σ 〉 = tanh β|h|
|~
h| |~
h|

To explore the regime dominated by quantum fluctuations and determine the


´ nature of the
quantum phase transition, Let T = 0 =⇒ β ⇒ ∞. In this limit, tanh β|~
³
h| ⇒ 1 yielding:

J 0 〈σz 〉
〈σz 〉 = q
Γ2 + J 02 〈σz 〉2

Or,
J0
1= q
Γ2 + J 02 〈σz 〉2

At the phase boundary, 〈σz 〉 = 0, which yields the condition that at the boundary or critical
point (for T = 0 ):

J 0 = Γ|PhaseB nd y. = Γc
That is, by adjusting Γ, a phase transition occurs at Γc . In the spirit of the reduced tempera-
ture, we may consider the ratio ΓΓc = JΓ0 which is one at the critical point.

Finally, for T 6= 0 on the phase boundary (again, when the order parameter 〈σz 〉 ⇒ 0 ):

Γc J 0 〈σz 〉
µ ¶µ ¶
〈σz 〉 = tanh
kT Γc
or,

Γc Γc
µ ¶
= tanh
J0 kT

With this information, the phase diagram may be constructed, as shown above in FIG. 4.

9
Γ
Γc

〈σz 〉 6= 0 〈σz 〉 = 0

T
1 Tc

FIG. 4: The Transverse Ising Model Phase Diagram

An Example of a Quantum Phase Transition in Modern Experiments

The Superconductor-Insulator Transition The superconducting phase, existing for select


materials, features the existence of a dissipation-less electrical current. Like any (quantum)
phase transition, its emergence or breakdown may be initiated by the variation of a control
parameter. Possible control parameters which are relevant for this discussion include the
magnetic field strength or the electron concentration. 13

FIG. 5: Several Generic Phase Diagrams for Superconducting Systems, with M denoting the
normal metal phase. Reproduced from [1].

Whenever a metal is able to conduct an electrical current, the electrons are allowed to move
throughout the substance so that they are delocalized - not bound to any one atom (or lat-
tice site). However, for an insulator, this is not the case; the electrons are localized. Hence,
a phase transition between superconducting and insulating phases would involve a process
which modifies constraints on the motility of the electrons. The order parameter which de-

13 Note that we are primarily interested in the quantum phase transition - when thermal fluctuations are small

(T ≈ 0) [1]

10
scribes this involves a phase. A spatial variance in phase corresponds with a net electron flow
(current) and is nonzero in the superconducting phase. In the disordered phase, such a gra-
dient or spatial variance of phase does not exist (suppressed long-range correlations) and the
order parameter is zero as required.

A Superconductor-Insulator (SI) transition may be seen for samples of Bi2 Sr2 CaCu2 O8+y , where
Ca atoms are substituted for Pr atoms between cuprate planes.14 This results in a new sub-
stance Bi2 Sr2 (Caz Pr1-z ) Cu2 O8+y , where z parametrizes the degree of substitution of calcium
atoms and is therefore the control parameter. By varying z, one finds evidence of a SI tran-
sition near z = .5, since the resistance R rapidly drops (to zero) for T → 0, but remains finite
above the critical concentration line for T → 0.

FIG. 6: A Plot of Temperature-dependent Resistance for various concentrations of Pr in


Bi2 Sr2 (Caz Pr1-z ) Cu2 O8+y , as indicated by several curves of differing values of z. The red line
signals a critical value of z. Reproduced from [1].

Final Remarks

Summary In this paper, a simple model - the (Mean-Field) Ising model - is solved to pro-
vide context to the basic notion of quantum phase transitions. This presentation provides a
bridge between basic classical concepts of phase transitions and the quantum case. Finally,
an example of the Superconductor-Insulator transition is presented as an important experi-
mental demonstration of this phenomenon. While there is much left out of this discussion, I
hope the reader finds this summary to be useful.15

14 See [1] P.37


15 A more complete review is provided in [2].

11
References

[1] Gantmakher, Vsevolod Feliksovich, and Valery Timofeevich Dolgopolov. Superconduc-


tor–insulator quantum phase transition. Physics-Uspekhi 53.1 (2010): 1-49.

[2] Vojta, Matthias. Quantum phase transitions. Reports on Progress in Physics 66.12 (2003):
2069.

[3] Pathria, R. K., and Paul D. Beale. Statistical Mechanics. 1996. Butter worth.

[4] Gulminelli, Francesca Phase Transition and Critical Phenomena Centre International
de Rencontres Mathématiques. Marseille, France. August 2011. CEMRACS ’11, Multiscale
Coupling of Complex Models.

[5] Chakrabarti, Bikas K. Quantum Phase Transition in Transverse Ising Models Saha Insti-
tute of Nuclear Physics. February 15, 2011.

[6] Greiner, Walter, Ludwig Neise, and Horst Stöcker. Thermodynamics and statistical me-
chanics. Springer Science & Business Media, 2012.

12

Das könnte Ihnen auch gefallen