Spin-Glass Theory For Pedestrians: Tommaso Castellani and Andrea Cavagna

arXiv:cond-mat/0505032v1 [cond-mat.
dis-nn] 2 May 2005
Spin-Glass Theory for Pedestrians

Tommaso Castellani† and Andrea Cavagna‡
† Dipartimento di Fisica, Univesita’ di Roma la Sapienza, Piazzale Aldo Moro 5,
00185 Roma
‡ Istituto Sistemi Complessi, INFM - CNR, Via dei Taurini 19, 00185 Roma
Abstract. In these notes the main theoretical concepts and techniques in the
field of mean-field spin-glasses are reviewed in a compact and pedagogical way,
for the benefit of the graduate and undergraduate student. One particular spin-
glass model is analyzed (the p-spin spherical model) by using three different
approaches. Thermodynamics, covering pure states, overlaps, overlap distribution,
replica symmetry breaking, and the static transition. Dynamics, covering the
generating functional method, generalized Langevin equation, equations for the
correlation and the response, the Mode Coupling approximation, and the dynamical
transition. And finally complexity, covering the mean-field (TAP) free energy,
metastable states, entropy crisis, threshold energy, and saddles. Particular attention
has been paid on the mutual consistency of the results obtained from the different
methods.
CONTENTS 2
Contents
1 Introduction 3
2 Basic concepts 5
2.1 Disorder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Self-averaging quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Annealed and quenched averages . . . . . . . . . . . . . . . . . . . . . . 7
2.4 The replica trick . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.5 Pure states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.6 Overlap, self-overlap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.7 Overlap distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 Statics 13
3.1 The p-spin spherical model . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2 First try: the replica symmetric calculation . . . . . . . . . . . . . . . . . 14
3.3 The key connection between replicas and physics . . . . . . . . . . . . . . 18
3.4 Replica symmetry breaking . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5 The 1RSB solution and the static transition . . . . . . . . . . . . . . . . 22
4 Equilibrium dynamics 24
4.1 The generating functional formalism . . . . . . . . . . . . . . . . . . . . 25
4.2 Dynamics of p-spin spherical model . . . . . . . . . . . . . . . . . . . . . 29
4.3 Equations for the correlation and the response . . . . . . . . . . . . . . . 31
4.4 Diagrammatic technique and Mode Coupling approximation . . . . . . . 32
4.5 The dynamical transition . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
5 Complexity 39
5.1 What is the TAP free energy ? . . . . . . . . . . . . . . . . . . . . . . . . 39
5.2 Definition of complexity (and a problem with the modulus). . . . . . . . 42
5.3 The calculation of the complexity. . . . . . . . . . . . . . . . . . . . . . . 43
5.4 Threshold energy and saddles. . . . . . . . . . . . . . . . . . . . . . . . . 46
5.5 The equation for the self-overlap . . . . . . . . . . . . . . . . . . . . . . . 47
5.6 Life with many metastable states . . . . . . . . . . . . . . . . . . . . . . 49
5.7 Low temperatures, entropy crisis . . . . . . . . . . . . . . . . . . . . . . 51
5.8 High temperatures, the threshold . . . . . . . . . . . . . . . . . . . . . . 52
6 Conclusions 53
1. Introduction
The aim of these notes is to provide graduate and undergraduate students in statistical
physics with a sort of handbook of the main theoretical concepts in the physics of
spin-glasses. It is important to emphasize that this is not an overview of the entire
field of disordered systems and spin-glasses: the whole experimental phenomenology
is missing; not a word is present on the large amount of numerical investigations and
results; only one model is analyzed, compared to the vast number of different models
on the spin-glass market; very little is said about the connections between spin-glasses
and structural glasses (although something is said); and finally, the focus is entirely on
mean-field spin-glasses, leaving completely untouched what may be regarded as one of
the most challenging open problems of the field, that is whether or not the mean-field
picture has some validity also in finite dimensional systems.
The student may thus rightfully ask what is contained in these notes. The basic
idea is to present the most important theoretical techniques developed in the context of
spin-glasses in a coherent, detailed, but at the same time very compact way. For this
reason we study just one specific model, which we use as an ideal arena where to discuss,
apply and compare different theoretical methods. Although the model we consider has
its own relevance in the field, the important point for us is to tell the student a consistent
and self-contained story, where each conceptual step has to be logically connected to
the previous one.
In order to do this we had to necessarily disregard many important topics in the
field, and at the same time to be very brief when introducing new ideas, hoping that
their practical implementation would help to grasp their relevance. The perfect example
is ergodicity breaking and pure states: an entire chapter, rather than few lines, should
be devoted to these tricky, but crucial concepts. In this way, however, the notes would
be unbearably long, and the main line of the story would quickly be lost. We opt for a
synthetic exposure, leaving the student the freedom to go deeper on certain subjects by
a careful use of the extensive list of references.
The concepts and techniques developed in spin-glasses have found in recent years a
wide range of applications in statistical physics and beyond, from biology to economics,
passing through computer science and optimization theory. Our hope is that these notes
may help the student to familiarize with the concepts, to practically learn how to handle
them in a non-superficial way, and to eventually apply them to their own field of interest.
The basic knowledge required to follow these notes is just a reasonable preparation in
standard statistical mechanics.
The three pillars of our discussion are Statics, Dynamics, and Complexity. The test
system where all the calculations are done and the consistency of the different results
Spin-Glass Theory for Pedestrians 4
is analyzed, is the mean-field p-spin spherical model (PSM). This model is particularly
apt to our purposes for more than one reason. First, the static (i.e. thermodynamic)
analysis of the PSM gives results which are drastically different from the dynamical
ones. For example, the two approaches give two different transition temperatures. This
naturally leads to the introduction of a third technique, dominated by the concept
of complexity, which very nicely reconcile static and dynamic results. Therefore, the
PSM is the ideal model where to develop and compare the three approaches. Secondly,
the PSM is probably a simpler model than the more famous and extensively studied
Sherrington-Kirkpatrick model, which after more than 25 year still puzzles us with its
enormous variety of weird, yet very interesting, results. Finally, the PSM has some
features which are intriguingly similar to structural glasses, most notably it is described
by a set of dynamical equations which are identical to those provided by the Mode
Coupling Theory for glasses. Therefore the PSM seems a good model to try and bridge
the gap between spin-glasses and structural glasses.
As we have said, the target of these notes are graduate and undergraduate students.
For this reason we tried to be as complete as possible when giving the details of the
calculations, typically providing more technical steps than it is usual in a technical paper.
We hope that in such a way it will always be possible for the student to work out the
final result. It is impossible to develop a genuine familiarity with spin-glasses without
a serious training in the most technical aspects of the field. We therefore encourage
the student to perform and check the calculations in these notes, in order to become as
independent as possible when studying similar subjects in her/his future. At the same
time, we tried not to lose contact with the broader picture, and to always stick a sense
to any calculation we perform. In particular, we stressed as much as possible the mutual
consistency of results obtained with different techniques. The theory of spin-glasses is
infamous for being crowded with not-too-obvious formal steps, so it is always nice to
find the same result with two (or more) different, independent methods.
These notes are the expanded version of the lectures that one of us (AC) delivered
in Bangalore, at the Conference and School on Unifying Concepts in Glassy Physics
(UCGP III), in June 2004, where also other lectures on different areas in the physics of
glassy systems were presented. Wherever we could, we tried to make contact, avoiding
overlaps, with the notes of the other participants. In particular, we mention the Mode
Coupling approximation in the section on Dynamics, in order to connect with the notes
of David Reichmann, and we restricted ourselves to equilibrium dynamics, given that
the subject of aging, and off-equilibrium dynamics in general, is extensively treated in
the notes of Giulio Biroli. We finally hope that the student will get the similarities
between the chapter on the TAP approach and the energy landscape method analyzed
by Francesco Sciortino.
We thank the organizers of UCGP III for giving us the opportunity to bring together
in a single work what we hope will be a useful collection of ideas and results in such a
fascinating field of science.
2. Basic concepts
Before we start studying a specific spin-glass model, we need to introduce a couple of

simple concepts and tricks, which we will extensively use all along these notes. Each of
them would deserve much more space than we can afford, and therefore we encourage
the student to exploit the references. A background in statistical mechanics and in
particular in the theory of critical phenomena is very helpful. A nice and compact
overview on this subject can be found in references [1] and [2]. We also add here that
a very nice and informal introduction to spin glasses can be found in [3], whereas the
classic review for spin-glass theory is still reference [4].
2.1. Disorder
There are two main classes of disordered systems. The one spin-glasses belong to is
that of quenched disorder. In these systems the disorder is explicitly present in the
Hamiltonian, typically under the form of random couplings J among the degrees of
freedom σ,
H = H(σ; J) . (1)
The disorder J is completely specified by its probability distribution p(J) dJ which is
the same for each different coupling constant in the system. A famous example is the
Edwards-Anderson model [5],
X
H=− Jij σi σj , (2)
<ij>
where the spins σi = ±1 are the degrees of freedom, and the couplings Jij are Gaussian
random variables. This is a finite dimensional model, since the sum is performed over
nearest-neighbor spins. The disorder is quenched, meaning that the J are constant on
the time scale over which the σ fluctuate. This will have a crucial consequence on the
way we will have to perform the averages over J, compared to σ. Spin-glasses are indeed
systems with quenched disorder.
Disorder creates frustration: it becomes impossible to satisfy all the couplings at
the same time, as it would be in a ferromagnetic system. Formally a system is frustrated
if there exists a loop on which the product of the couplings is negative. In a frustrated
loop, if we fix an initial spin, and starting from it we try to chain-fix the other spins
one after the other according to the sign of the couplings, we are bound to return to
the initial spin and flip it. The only way to avoid frustration is to consider a lattice
where there are no loops, for example a tree. Frustration is the main reason for the
proliferation of metastable states in disordered systems.
In some system the disorder is not present in the Hamiltonian, but is in a way
self-generated. This is the case of structural glasses, whose Hamiltonian typically takes
the form, X
H= V (ri − rj ) (3)
ij
where the degrees of freedom ri are the positions of the particles, and the function
V (r) is a deterministic potential (for example, Lennard-Jones). Even though there
is no quenched disorder in the Hamiltonian, at low temperature, in a frozen glassy
configuration of the system, each particle sees a different, disordered environment around
itself. In this sense the disorder is self-generated. The origin of this phenomenon is the
large number of non-crystalline local minima of the Hamiltonian.
It may seem odd that systems with quenched and self-generated disorder do have
any property in common, given their very different definitions. However, we shall
see that some spin-glass models do have a phenomenology quite similar to the one
of structural glasses.
2.2. Self-averaging quantities

In these notes we deal with spin-glasses, i.e. systems with quenched disorder in the
Hamiltonian. Therefore, the first key question is: How do we deal with the disorder?
The problem is that, in principle, each observable depends on J, including the free
energy of the system,
Z
1
FN (J) = − log Dσe−βH(σ;J) (4)
βN
where N is the size of the system. This is very unpleasant, since it seems to suggest that
the physical properties of spin-glasses are different for each different realization of the
disorder J, i.e. for each different sample. Were this true, it would be a disaster: we want
to build a theory for spin-glasses, and not just for a specific piece of material ! In fact,
both common sense and experience tells us that for sufficiently large systems, physical
properties do not depend on J anymore. Quantities like that are called self-averaging
[4], and the free energy is one of them,
lim FN (β, J) = F∞ (β) . (5)

N →∞
In this case it is clear that the average over the disorder of a self-averaging quantity is
equal to its J-independent value,
1
F = − lim logZ(J) = F∞ (β) (6)
N →∞ βN
where, Z
A= dJ p(J) A(J) . (7)
This is good, since it means that analytically we can average over J, and that the result
we obtain in this way is in agreement with the physical value of the observable. Self-
averageness is basically the same as asking that the distribution of physical quantities
is (for N large) sharply peaked around their average value, that is that the variance of
their distribution must go to zero for N → ∞. The free energy is self-averaging, and in
particular,
2 1
F2 − F = O (8)
N
If a quantity has, for example, a bimodal distribution, it is not self-averaging. Indeed
its average is a very poor indicator of the physical values of the quantity itself.
A simple argument to work out equation (8) can be given in finite dimension. We
divide our system in a number K of macroscopic sub-systems, with 1 ≪ K ≪ N. The
total (extensive) free energy will be the sum of the free energies of the sub-systems, plus
a contribution coming from the interactions at the interfaces between the sub-systems.
Once we compute the free energy density, this surface contribution can be neglected in
the limit N → ∞. Moreover, the sub-systems free energies are independent random
variables and therefore we can apply the central limit theorem to the sum, and get (8).
2.3. Annealed and quenched averages

In order to average the free energy we have to compute the integral,
Z Z
1
F =− dJ p(J) log Dσ e−βH(σ;J) , (9)
βN
which looks pretty bad, since we have to integrate a log over J. We could be tempted
to define the following different quantity,
Z Z
1
Fa = − log dJ p(J) Dσ e−βH(σ;J) , (10)
βN
which is certainly much simpler to compute. Unfortunately, this is not the right solution
to our problem. The difference between the two formulas above is in the role played
by the disorder J: in (9) we first integrate over the degrees of freedom, then take the
log, and finally integrate over the random couplings. In this way, the couplings J are
fixed, i.e. quenched, for each integration over the spins. In other words, couplings and
spins do not fluctuate together: for each realization of the disorder we compute the free
energy, and eventually we average it over J. This kind of average is called quenched,
and it is precisely what we need to do.
On the other hand, it is clear that in (10) the disorder J and the degrees of freedom
σ have been put on the same footing, fluctuating together. This is not what we want,
since it means that the time scale of variation of J and σ is the same, and therefore
the disorder becomes yet another degree of freedom, and it is no longer quenched. This
second kind of average is called annealed, and, even though it may be correct at high
temperatures, where the frustration induced by the disorder is irrelevant, it is normally
wrong at low temperatures, where the spins freeze in a state determined by the quenched
value of the couplings. A different way to see this point is that in the annealed case we
are in fact averaging the partition function Z, rather than the free energy F , over J.
The fact is that F is an extensive quantity, while Z is not (it is exponential in N !), and
therefore Z is not in general self-averaging.
Therefore, we have to find a way to treat integrals like the one in (9). This is where
the replica trick comes into play.
2.4. The replica trick

The replica trick [5] (which becomes in fact a method, when it is explained more deeply
than here [4]), stems from the following simple formula,
1
log Z = lim log Z n . (11)
n→0 n
If n remains a real number (as it should), there is no advantage at all in computing the
r.h.s. compared to the l.h.s., of course. However, if we now promote n to be an integer,
we can write: Z
Z = Dσ1 . . . Dσn e−βH(σ1 ,J)···−βH(σn ,J)
n (12)
which is in fact much simpler to compute. What we do is to replicate the system n

times, compute everything as a function of n, and finally cross our finger in taking the
limit n → 0. It is crucial to understand that all the Hamiltonians in (12) have the same
realization of the quenched disorder, and in this sense are replicas one of the other.
A different useful form of the replica trick is the following,
Z Z
1 n−1
hAi = Dσ A(σ) e −βH(σ,J) = lim Z Dσ A(σ) e−βH(σ,J) =
Z n→0
Z
lim Dσ1 . . . Dσn A(σ1 ) e−βH(σ1 ,J)···−βH(σn ,J) . (13)
n→0
Of course, the label 1 we used for the replica into the observable A is completely
arbitrary, and thus we have to be careful that our result must not depend on this
particular index, otherwise we have a complete nonsense.
To conclude this small section, let us have a look to a case where the replica trick
does work. Imagine that we ignore the rule (xa )b = xab with a, b real, but that we know
that xm = x · · · · · x, m times. Given y = x · x, we want to know what is y 1/2 . We can
use the replica trick:
y 1/2 = lim y n = lim x · x · · · · · x · x = lim x2n = x . (14)

n→1/2 n→1/2 n→1/2
2.5. Pure states

In the low temperature phase, and in the limit N → ∞ we can have ergodicity breaking:
the system at equilibrium explores only a sub-part of the phase space [6, 7, 4]. When
this happens the Gibbs measure can be split into sub-components, called pure states,
X
h·i = wα h·iα (15)
α
where α is an index running over all the states, and wα is the statistical (Gibbs) weight
of state α. To better understand this formula, we must assume that it is possible to
assign each configuration in the phase space with nonzero thermodynamics weight, to
one and only one state. In this case we can write for any observable A,
Z Z
1 βH(Σ) 1X
hAi = Dσ e A(σ) = Dσ eβH(Σ) A(σ) =
Z Z α σ∈α
X Zα 1 Z X
Dσ eβH(Σ) A(σ) = wα hAiα (16)
α
Z Zα σ∈α α
where we have defined, Z

Zα = Dσ eβH(Σ) (17)
σ∈α
that is the partition function restricted to state α, and

Zα
wα = (18)
Z
the statistical weight of state α.
As an example we can consider the Ising model below Tc . In the thermodynamic
limit the ergodicity is broken, and we have two states, with positive and negative
spontaneous magnetization,
1 1
h·i = h·i+ + h·i− (19)
2 2
that is w+ = w− = 1/2, in absence of external magnetic field. It is crucial to split the
measure, otherwise we would not see any spontaneous magnetization,
1 1
hσi = hσi+ + hσi− = 0 (20)
2 2
A very important feature of pure states is the clustering property. In essence, this
property states the very physical concept that the statistical correlation between two
different points goes to zero when their distance goes to infinity,
hσi σj i → hσi ihσj i for |i − j| → ∞ . (21)
In other words, a very basic physical requirement is that connected correlation functions
decay to zero at large distances [7, 4]. As we have said, this property only holds in pure
states. Take, for example, the paramagnetic state in the Ising model below Tc , that is
the Gibbs ergodic measure over the full phase space:
1 1 1 1
hσi σj i = hσi σj i+ + hσi σj i− → hσi2+ + hσi2− = m2 6= 0 . (22)
2 2 2 2
Therefore, the paramagnetic state is not a pure state below the critical temperature.
The example of the Ising model is particularly simple because we know a priori
what is the structure of pure states below Tc . In particular, we know how to select a
state, i.e. how to project the system onto any one of the two states: we simply apply a
magnetic field. In disordered systems the situation is not as simple as that, since we do
not know what is the field projecting the system onto any particular state. This crucial
fact is at the heart of the difficulty in studying disordered systems: we lack the magnetic
field as a crucial tool to select states. Of course, given a state, there is a (disordered)
magnetic field selecting that state. The problem is that we do not know what this field
is !
A final important remark. In finite-dimensional systems, only equilibrium states
can break the ergodicity, i.e. states with the lowest free energy density. In other words,
the system cannot remain trapped for an infinite time in a metastable state, because in
finite dimension free energy barriers surrounding metastable states are always finite. The
extra free energy of a droplet of size r of equilibrium phase in a background metastable
phase has a positive interface contribution which grows as r d−1 , and a negative volume
contribution which grows as r d ,
∆F = σ r d−1 − δf r d , (23)
where here σ is the surface tension and δf is the bulk free energy difference between
the two phases. This function has always a maximum, whose finite height gives the free
energy barrier to nucleation of the equilibrium phase (note that at coexistence δf = 0
and the barrier is infinite).
Therefore, if initially in a metastable states the system will, sooner or later, collapse
in the stable state with lower free energy density. For this reason, in finite dimension we
cannot decompose the Gibbs measure in metastable components. When this is done, it
is always understood that the decomposition is only valid for finite times, i.e times much
smaller than the time needed for the stable equilibrium state to take over. On the other
hand, in mean-field systems (infinite dimension), barriers between metastable states
may be infinite in the thermodynamic limit, and it is therefore possible to call ’pure
states’ also metastable states, and to assign them a Gibbs weight wα . We will analyze
a mean-field spin-glass model, so that we will be allowed to perform the decomposition
above even for metastable states.
2.6. Overlap, self-overlap

In non-disordered magnetic systems, a good order parameter is normally the total
average magnetization,
N
1 X
m= hσi i (24)
N i=1
which is zero in the high temperature phase, and different from zero in the low
temperature phase, where the ± symmetry is broken. In disordered systems we may be
tempted to use a similar order parameter,
N
1 X
m= hσi i (25)
N i=1
However, due to the disorder the local magnetizations in the low temperature phase
are all frozen in different directions (if the disorder distribution is unbiased, as we shall
assume), and thus the magnetization defined above is zero at all temperatures, even
though the ± symmetry is physically broken for each spin in our sample. A better order
parameter is the Edward-Anderson parameter [5],
N
1 X
qEA = hσi i2 (26)
N i=1
Such a quantity is nonzero if the local magnetizations mi are locally nonzero, and thus
is a good order parameter. In fact qEA is a particular case of a more general quantity
called overlap.
In our study of spin glasses we will often need a tool to measure the similarity of
two configurations, or two states. To this aim we introduce the overlap. Given two
configurations σ and τ , we define their mutual overlap as,
N
1 X
qστ = σi τi (27)
N i=1
With Ising spins si = ±1 we have that,


 1 if σ e τ almost coincide

qστ = −1 if σ e τ are anti-correlated (28)

 0 if σ e τ are totally uncorrelated
The overlap is thus a measure of the similarity among different configurations. We can
also compute the overlap of a configuration with itself, the self-overlap,
N
1 X
qσσ = σi σi (29)
N i=1
With Ising spins qσσ = 1. In the following we will always deal with systems where the
self-overlap of configurations is 1.
The overlap can measure also the similarity between states: if the Gibbs measure is
split into sub-components α due to ergodicity breaking, we define the overlap between
states α and β as,
N
1 X
qαβ = hσi iα hσi iβ (30)
N i=1
which can also be written as,
N Z Z
1 X1 −βH(σ) 1
qαβ = Dσ σi e Dτ τi e−βH(τ ) =
N i=1 Z α σ∈α Z β τ ∈β
Z Z
1
Dσ Dτ e−βH(σ) e−βH(τ ) qστ (31)
Zα Zβ σ∈α τ ∈β
This expression shows that by measuring the overlap among states, we are in fact
measuring the overlaps among configurations belonging to the states, and summing
over all pairs of configurations, each one weighted with its own statistical weight.
The self-overlap of a state is simply,
N
1 X
qαα = hσi i2α (32)
N i=1
The self-overlap will be very important in what follows. It is a measure of the size of
the state in the phase space: the larger qαα , the smaller the state, i.e. the smaller the
number of configurations belonging to the state. On the other hand, a very small self-
overlap indicates a very broad state. In particular, the paramagnetic state (unbroken
ergodicity) has self-overlap equal to zero.
In the limit T → 0 each states concentrate on its lowest energy configuration. In
this case, the self-overlap of each state is qαα = 1, since it is just the self-overlap of a
configuration. When the temperature T grows, more configurations participate to the
state and the self-overlap becomes smaller than one.
2.7. Overlap distribution

As we shall see, in mean-field spin-glasses there are many inequivalent pure states at low
temperatures. In this case, it is useful to introduce the probability distribution of all the
possible values of the overlaps among states. We first compute the overlap distribution
by considering two physical systems with the same disorder (also called real replicas),
and averaging the value of the overlap qστ among the two real replicas,
Z
1
P (q) = 2 DσDτ e−βH(σ) e−βH(τ ) δ(q − qστ ) . (33)
Z
Using the definitions of the previous sections, we have,
Z Z
X 1 1
P (q) = wα wβ DσDτ e−βH(σ) e−βH(τ ) δ(q − qστ ) , (34)
αβ
Z Z
α σ∈α β τ ∈β
and using the clustering property we finally obtain,

X
P (q) = wα wβ δ(q − qαβ ) . (35)
αβ
In this formula (which can also be taken as a definition of the P (q)) the sum is extended
over all the possible pairs of states, including pairs of the same state, giving its self-
overlap. Once again, the simple Ising model can help us. At low temperature we have
two pure states, so we have in principle four possible overlaps,

1 X 1 X 2
q++ = hσi i2+ = mi = m2 (36)
N i N i
1 X 1 X 2
q−− = hσi i2− = mi = m2 (37)
N i N i
1 X 1 X
q+− = q−+ = hσi i+ hσi i− = − mi mi = −m2 . (38)
N i N i
Therefore the function P (q) has two peaks, at−m2 and +m2 , each with weight 1/2. It
is important to stress that the number of peaks of the P (q) is not equal to the number
of states, but to the number of possible values taken by the overlap. If we had a very
large number of states, all with the same self-overlap and mutual overlap, we would still
have a bimodal P (q).
To conclude, we note that the particular structure of states of a given sample
depends on the particular realization J of the quenched disorder. For this reason
both the pure states weights, and the distribution P (q) depend on the disorder J. In
particular, P (q) is not a self-averaging quantity when the structure of states is nontrivial.
For the proof and discussion of this crucial statement see [8].
3. Statics
We have now all the tools to start a thermodynamic study of a specific spin-glass. We
will use the replica method to compute the free energy of the system, and will discover
that replicas have (surprisingly enough) a rather deep physical meaning: they will act
as probes exploring the unknown phase space, and sending us important information on
the structure of states in it.
The spin-glass model we will analyze is the p-spin spherical model (PSM). Among
spin-glasses it is the one which bears more similarities with structural glasses, suggesting
that some concepts which are exactly valid for the PSM may be exported to the case of
glasses.
3.1. The p-spin spherical model

The Ising version (i.e. with ±1 spins) of the PSM was introduced in [9], while its
spherical, and simpler, counterpart appeared in [10]. The Hamiltonian of the spherical
PSM is,
XN
H=− Ji1 ...ip σi1 . . . σip p≥3 (39)
i1 >...>ip =1
where the spins are now real continuous variables. In order to keep the energy finite, we
have to put a constraint on the spins,
N
X
σi 2 = N (40)
i=1
this is the spherical constraint, from which the model takes its name. With this
constraint the self-overlap of each configuration is one. The Hamiltonian is a sum
of p-body interactions, and the sum is extended over all groups of spins, not only the
nearest-neighbor, so the model has no spatial structure, and it is in fact a mean-field
model. For such models the droplet argument given above does not work (each spin
interacts with N other spins, there are no surfaces), and thus the free energy barriers
around metastable states may be infinite. For this reason mean-field models are the
ideal play-ground to study metastability.
Each random coupling J is a Gaussian variable, with distribution,

1 2 2N p−1
dp(J) = exp − J dJ (41)
2 p!
where the factors 2 and p! are a matter of convention, whereas the factor N p−1 is essential
in order to have the Hamiltonian of order N, and thus extensive energy and free energy,
p 1
J2 ∼ p−1 ⇒ H∼N (42)
N 2
The relevance of the PSM in the context of glassy physics is due to the great
role played by metastable states in such a model. A hint of this fact comes from the
ferromagnetic version of the PSM, that is Ji1 ...ip = 1/N p−1 for each coupling: unlike its
p = 2 counterpart, this model has a first order transition between a high T paramagnetic
phase and a low T ferromagnetic one (solving the ferromagnetic mean-field PSM is a
trivial exercise). In particular, there are two relevant temperatures: a temperature Td
below which a ferromagnetic state develops, but with a free energy higher than the
paramagnetic one, and a lower temperature Ts , where the ferromagnetic state becomes
stable and the thermodynamic transition takes place. From a dynamical point of view,
however, the higher temperature Td is quite relevant, since for T < Td the system
may remain trapped by the ferromagnetic state, even though metastable, if the initial
magnetization is positive and large enough.
The first order transition at Ts in the ferromagnetic PSM is driven by entropy, since
the energy of the ferromagnetic states is always lower than the paramagnetic one. We
can roughly understand this point by noting that the p-body interaction indeed increases
very much the entropic contribution of the paramagnet, compared to the canonical p = 2
case. Metastability, entropy driven transitions, and purely dynamical transitions will
be also key ingredients of the disordered PSM we are about to study.
3.2. First try: the replica symmetric calculation

We start our static study of the PSM by performing an annealed calculation of the free
energy. We know it is wrong at low temperatures, but it will be anyway a useful warm-
up exercise. In what follows we will often write the indices for the p = 3 case, such
that Ji1 ...ip becomes Jijk . However, to give formulas that are valid even in the general
case, we will write all the factors containing a term p for the generic p case, for example
we will write N p /p! rather than N 3 /6. Another short-cut we will use is to disregard
all normalizing factors that, once taken the log and divided by N, go to zero in the
thermodynamic limit. Finally, we have to remember that all our integrals over σ are
restricted to the surface of a sphere by the spherical constraint. The average partition
function is given by,
Z Z Y p

2 N
Z = Dσ dJijk exp −Jijk + Jijk βσi σj σk =
i<j<k
p!
Z " !p #
β2 X
Dσ exp p−1
σi2 =
4N i
2

β
exp N Ω, (43)
4
where Ω is the surface of the sphere. In the equations above we have used the formula,
N
X N
X
p! = (44)
i<j<k ijk
which is valid in the thermodynamic limit. The annealed free energy is therefore given
by,
Fa = −β/4 − T S∞ , (45)
with the infinite temperature entropy, S∞ = log(Ω)/N. This is, in fact, the correct
free energy at high temperatures, i.e. in the paramagnetic phase. However, it can
be proved that at lower temperatures the annealed-paramagnetic solutions has a free
energy larger than the free energy found by the quenched computation: as anticipated
above the annealed approximation in general only holds at higher temperatures, while
at low temperature the quenched computation must be performed. Note that the fact
that the annealed entropy becomes negative at low temperatures would not be by itself
a sufficient reason to discard it, since the model is continuous, and a negative entropy
is thus perfectly legal.
In order to perform the quenched calculation we must compute the average of the
replicated partition function. Since now on the indices i, j, k, . . . will refer to sites, while
a, b, . . . will refer to replicas. We have,
Z " n
#
YZ N p X
Z n = Dσia dJijk exp −Jijk 2
+ Jijk β σia σja σka =
ijk
p! a
Z " n
#
Y β 2 p! X a b a b a b
Dσia exp σ σσ σσ σ =
ijk
4N p−1 ab i i j j k k
Z " n N
!p #
2
β X X
Dσia exp σia σib . (46)
4N p−1 ab i
We can see here the powerful replica trick at work: we started from a set of coupled
sites and uncoupled replicas, and averaging over the disorder we decoupled the sites,
but coupled the replicas (unfortunately in non-mean field models the replica trick is not
enough to decouple the sites). In particular, the overlap between two different replicas
of the system very naturally appeared in the calculation,
1 X a b
Qab = σ σ (47)
N i i i
Note that Qaa = 1 due to the spherical constraint. We introduce now a factor 1 in our
calculation, !
Z X
1 = dQab δ NQab − σia σib , (48)
i
and finally we use an exponential representation for the δ-function, to obtain,

Z
Zn = DQab Dλab Dσia ·
" #
β 2N X p X XX
· exp Q +N λab Qab − σia λab σib =
4 ab ab ab i ab
Z
= DQab Dλab exp [−N S(Q, λ)] (49)
with,
β2 X p X 1
S(Q, λ) = − Qab − λab Qab + log det(2λab ) (50)
4 ab ab
2
In (49) the integration over Qab is performed over all the matrices with a 6= b, while the
integration over λab includes also a = b to enforce the spherical constraint. The sums in
the exponentials are over all the indices, including a = b.
The great advantage of this form of the integral is that we can use the saddle point
(or Laplace, or steepest-descent) method [11], to solve it in the limit N → ∞. This
simplification is the big effect of mean-field, and it is the result of the decoupling of the
sites operated by the use of the replica trick. The price we had to pay is that we coupled
replicas, and this looks somewhat weird at this stage of the computation.
The saddle-point method states that in the limit N → ∞ the integral (49) is
concentrated in the minimum of the integrand. However, we have to be careful here,
for a twofold reason. First, the free energy is in principle given by,
Z
1
−βF = lim lim log DQab Dλab exp [−N S(Q, λ)] (51)
N →∞ n→0 nN
and thus we should first take the limit n → 0, and then N → ∞. Unfortunately, we
are unable to do this: S is not an explicit function of n, and moreover we need to send
N → ∞ first to solve the integral. As a conclusion, we need to exchange the order of
the two limits, solve the integral, find a parametrization of the matrix Qab , and finally
take the n → 0 limit at the end. Of course, this is mathematically risky, to say the
least.
The second point we have to pay attention to, is what do we actually mean by
“minimum” of S. The problem here is that the number of independent elements of Qab
is n(n − 1)/2, which becomes negative is the limit n → 0. It is hard to say what is a
minimum of a function with a negative number of variables ! There is however a criterion
we can use to select the correct saddle point: the corrections to the saddle point result
are given by the Gaussian integration around the saddle point itself. This integration
gives as a result the square root of the determinant of the second derivative matrix of
S, and thus, in order to have a sensible result, we must have all the eigenvalues of this
matrix positive. Summarizing, we have to select saddle points with a positive-defined
second derivative of S [12].
At this point we can proceed with the saddle point calculation. We first minimize
(maximize ?) S with respect to λab . By using the general formula,
∂
log detMab = (M −1 )ab (52)
∂Mab
we get,
2λab = (Q−1 )ab (53)
and thus, " #
1 β2 X p
F = lim − Q + log detQab (54)
n→0 2βn 2 ab ab
where Qab satisfies the saddle point equation,
∂F β 2 p p−1
0= = Q + (Q−1 )ab . (55)
∂Qab 2 ab
Note that Qaa = 1 due to the spherical constraint.
What we have obtained is a free energy F , function of an order parameter, Qab ,
which is definitely weirder than the simple magnetization m = hσi we would have in the
ferromagnetic Ising model. This order parameter is the overlap between configuration
belonging to different replicas, and its physical meaning will be clearer later on. For
now, we limit ourselves to find a solution of the saddle point equation. To do this we
have first to find a parametrization of the matrix Qab , and to write (55) as a function
of the elements of Qab and of its dimension n.
Given that all replicas are equivalent (they just come from a formal trick !), it seems
wise to assume a replica symmetric form for the matrix Qab . This is what Sherrington
and Kirkpatrick did in their first mean-field spin glass model [13], that is,
Qab = q0 + (1 − q0 )δab . (56)
This means that all the elements of Qab are equal to q0 , but on the diagonal, where they
are 1. The value of q0 must be found from the saddle point equations. We have,
1 q0
(Q−1 )ab = δab − (57)
1 − q0 (1 − q0 )[1 + (n − 1)q0 ]
and thus (55) becomes, in the limit n → 0,

β 2 p p−1 q0
q0 − =0 (58)
2 (1 − q0 )2
We first see that q0 = 0 is always solution of this equation. This is the paramagnetic
solution, and from (54) we get F = −β/4, which is the same as the annealed result
(except for the phase space volume S∞ we did not include here). Thus, the annealed
calculation gives the same result as the quenched calculation when the overlap matrix
Qab is the identity. This is obvious, because when Qab = δab replicating or not the
system is exactly the same.
However, we also have a non-paramagnetic solution q0 6= 0. Recasting the equation
in the following form,
2
q0p−2 (1 − q0 )2 = T 2 (59)
p
we clearly see that at high temperatures there is no nontrivial solution, while by
decreasing T we arrive at a critical value T ⋆ below which a pair of nonzero solutions
forms. Of these two solution the only acceptable one is the larger one, which increases
with decreasing T (the self overlap must increase if the number of configurations
belonging to a state decreases, and this is exactly what we expect when we decrease
the temperature). Therefore we seem to have a transition at T ⋆ , and in particular a
discontinuous transition, since the value of q0 at the transition is different from zero,
i.e. there is a jump of the order parameter at the transition. Moreover, the free energy
associated to this new solution is lower than the paramagnetic one, therefore it would
seem we have found the new non-paramagnetic state at low temperatures.
All this seems very interesting, but there is a problem: the nontrivial solution we
have found is unstable [12, 10]. As we have said above, when we select a saddle point, we
have to be sure that all the eigenvalues of the second derivative of F around the saddle
point are positive. Unfortunately, this is not the case for this solution: both roots of
equation (59) have one negative eigenvalue below T ⋆ .
What can we do ? Remember that we did not search the whole space of Qab to
find a solution, but rather assumed a certain parametrization, which looked more or
less sensible, and plugged it into the saddle point equation. The fact that the replica
symmetric ansatz gave us a nontrivial solution at low T , but which is unstable, clearly
means that the low temperature phase of the model must be describe by a replica
symmetry breaking form of the order parameter Qab . Before looking for this new solution,
it is finally the moment to try and understand what is the physical meaning of the weird
order parameter Qab .
3.3. The key connection between replicas and physics

Let us consider the following quantity,
1 X
q (1) = hσi i2 (60)
N i
which, as we have seen in the previous chapter, is a quite natural definition of an order
parameter, since it is just a generalization of the average magnetization m. By using
the technology developed in the first chapter, we can rewrite q (1) in the following way,
1 XX X
q (1) = wα wβ hσi iα hσi iβ = wα wβ qαβ =
N i αβ αβ
Z X Z
dq wα wβ δ(q − qαβ ) q = dq P (q) q (61)
αβ
(1)
Therefore q is the first moment of the overlap distribution, averaged over the disorder.
By using the clustering property, we can easily find a generalization of this formula [4],
Z
(k) 1 X
q = k hσi1 . . . σik i = dq P (q) q k
2 (62)
N i ...i
1 k
The important fact is that we can compute these quantities also using the replica trick.
In particular,
Z
(1) 1 X 1 X 1 2 −β P H(σa )
q = hσi i = lim Dσia
2 σ ·σ e a (63)
N i n→0 N i i i
If we now go on with the calculation along the lines of the previous paragraphs,
introducing the overlap matrix Qab , we get,
Z
q = DQab e−N S(Qab) Q12 = QSP
(1)
12 (64)
where QSPab is the saddle point value of the the overlap matrix (since now on we will
drop the suffix SP), and where we have exploited the fact that S is of order n, and
therefore does not contribute when n → 0. Of course, there is something wrong about
this formula: replicas 1 and 2 cannot be different from the others! If we decided to call
them 4 and 7, we would get a different result when Qab is not replica symmetric: this
is nonsense ! What is going on here ? To understand this point we note that if the
saddle point overlap matrix is not symmetric, then there must be other saddle point
solutions with the same free energy, but corresponding to matrices obtained from Qab
by a permutation of lines and columns [4]. This is a general result: when a saddle point
breaks a symmetry corresponding to a given transformation, all the points obtained
by applying the transformation to that particular saddle point, are equally valid. This
means that we must average over all these saddle points, and this is equivalent to
symmetrize equation (64) [14, 15], obtaining,
2 X
q (1) = lim Qab (65)
n→0 n(n − 1)
a>b
This result is already telling us that there is a connection between the physical order
parameter q (1) , and the matrix of the overlap among replicas Qab . To go further, we can
generalize (65), to get,
2 X
q (k) = lim Qkab (66)
n→0 n(n − 1)
a>b
A comparison with equation (62), gives for a generic function f (q) the relation,
Z
2 X
dq f (q) P (q) = lim f (Qab ) (67)
n→0 n(n − 1)
a>b
and in particular choosing f (q) = δ(q−q ′ ), we finally find the crucial equation connecting
physics to replicas,
2 X
P (q) = lim δ(q − Qab ) . (68)
n→0 n(n − 1)
a>b
This equation is telling us that the average probability that two pure states of the system
have overlap q is equal to the fraction of elements of the overlap matrix Qab equal to q.
In other words, the elements of the overlap matrix (in the saddle point) are the physical
values of the overlap among pure states, and the number of elements of Qab equal to q
is related to the probability of q.
This is a key connection, and we understand now that Qab has an enormous physical
meaning. As a first application, let us analyze the meaning of the replica symmetric
ansatz, Qab = q0 for each a 6= b. From (68) we see that this structure of the overlap
matrix implies that the average overlap distribution is given by,
P (q) = δ(q − q0 ) , (69)
that is there is one single possible value of the overlap among states. As we have seen,
the overlap distribution should also include the self-overlap of the states, and therefore
this value q0 must be the self-overlap of the unique state in the system. The conclusion
is that a replica symmetric form of the overlap matrix in the free energy calculation,
can only be valid if there is one single equilibrium state. This state will typically be
the paramagnetic state, and its self-overlap will be q0 . On the other hand, if at low
temperatures there is ergodicity breaking, with the emergence of many inequivalent
pure states, then the correct form of Qab cannot be replica symmetric.
Now that we know what is the meaning of the overlap matrix, there is a slight
chance to understand how to find a replica symmetry breaking form of it.
3.4. Replica symmetry breaking

Thanks to equation (68) the relations of overlap among states translate into relations
of overlaps among replicas. Therefore, in order to give an ansatz on the form of Qab we
have to guess what may be the structure of states in the low temperature phase of a
spin-glass model. Vast programme ! - as someone once said [16].
Our starting point is a fact we already know: if there is ergodicity breaking, that is
if there are many states, configurations in the phase space are organized into states. In
other words, we can think of states as blobs of configurations in the phase space, with
each configuration belonging to just one blob. The self overlap of a state is just the
average overlap of the configurations belonging to it, i.e. it is a measure of the largeness
of the blob. On the other hand, the overlap between different states is basically the
overlap between configurations belonging to them.
Given this, the simplest possible spectrum of overlaps we can have when there
are many states is the following: q = 1, if we consider twice the same configuration,
q = q1 < 1 if we consider different configurations belonging to the same state, q = q0 < q1
if we consider configurations belonging to different states. In this way we are assuming
that all states have the same self-overlap q1 , and mutual overlap q0 . Moreover, a physical
requirement is that q1 > q0 , since configurations belonging to the same state must be
closer than those in different states.
What is the corresponding structure of Qab ? What is clear by now is that replicas
act as probing configurations of the structure of the states, so we must reproduce for
replicas the same clustering procedure we have seen for the configurations: replicas may
belong to the same group, having overlap Qab = q1 , or to different groups, with overlap
Qab = q0 . Finally, when we select twice the same replica we obtain Qaa = 1. To this
structure corresponds the matrix [17],
 
1 q1 q1
 q 1 q
 1 1 q0 ···  
 
 q1 q1 1 
 
Qab =  1 q1 q1  (70)
 

 q0 q1 1 q1 


 q 1 q1 1 

.. ..
. .
where we have assumed, to make an example, that the number m of replicas in each
group is m = 3. As we have seen, the parameter m is connected to the probability of
having a given value of the overlap, therefore it will become a variational parameter in
the saddle point equations, as q1 and q0 . This structure of Qab reflects what we have
said above. Note that any permutation of lines or column (replica permutation) would
also correspond to the same structure, but it would simply be much harder to visualize.
P
This matrix has the important property that a Qab does not depend on b, which is an
essential requirement, since replicas must be all equivalent [19].
It is clear that the clustering process we have described can be iterated [18]: states
can be grouped into clusters, which can be grouped into super-clusters, and so on. The
structure of states one obtains in this way is called ultrametric, and unfortunately we
do not have time to describe it here [8]. The important point is that for the PSM
the simple structure described above is sufficient [10]. This kind of replica symmetry
breaking (RSB) is called one step RSB, or 1RSB.
Let us compute the overlap distribution associated to the 1RSB structure of Qab .
From (68) we get,
m−1 n−m
P (q) = δ(q − q1 ) + δ(q − q0 ) (71)
n−1 n−1
with,
1≤m≤n (72)
We clearly see that while the elements q1 , q0 fix the position of the peaks, the parameter
m fixes their height, and therefore the probability of the overlaps. We have now to take
the limit n → 0. Here lies possibly the weirdest twist of the replica method. Relation
(72) seems to resist strenuously to our will to send n to zero. However, in this limit it
is clear that also m must be promoted to be a real number, rather than an integer. To
see how to do this we can appeal to physics (for once !), and accept the fact that the
probability (71) must be positive even in the limit n → 0,
P (q) = (1 − m) δ(q − q1 ) + m δ(q − q0 ) (73)
For this to be positive we must have m < 1 and m > 0. Therefore, the correct limit of
(72) for n → 0, is,
0≤m≤1. (74)
Summarizing, with the 1-step replica symmetry breaking ansatz we have parametrized
the overlap matrix Qab by means of two values of the overlap,
0 ≤ q0 ≤ q1 ≤ 1 (75)
and one value of the probability parameter m. We have now to fix them via the saddle
point equation.
3.5. The 1RSB solution and the static transition

The first thing to do is to compute the free energy as a function of q1 , q0 , m. We had,
" #
1 β2 X p
F = lim − Q + log det Q (76)
n→0 2βn 2 a,b ab
The first piece is easy to compute and gives in the limit n → 0,

1X p X p
Qab = Qab = 1 + (m − 1)q1p − mq0p (77)
n a
ab
The second piece is a bit harder: the 1RSB matrix Qab has three different eigenvalues
and degeneracies (the student should be able to work them out),
λ1 = 1 − q d1 = n − n/m (78)
λ2 = m(q1 − q0 ) + (1 − q1 ) d2 = n/m − 1 (79)
λ3 = nq0 + m(q1 − q0 ) + (1 − q) d3 = 1 (80)
From this, taking carefully the limit n → 0, we finally obtain,
β2 m−1
− 2βF1RSB = [1 + (m − 1)q1p − mq0p ] + log(1 − q1 ) +
2 m
1 q0
+ log[m(q1 − q0 ) + (1 − q1 )] + (81)
m m(q1 − q0 ) + (1 − q1 )
to be compared with the replica symmetric (RS) free energy,
β2 q0
−2βFRS = [1 − q0p ] + log(1 − q0 ) + (82)
2 1 − q0
It is interesting to note that the RS form is obtained either from q1 → q0 , or m → 1. In
the first limit, many states merge to form a single paramagnetic state. This is exactly
what happens in the Ising model when T → Tc− , and it is a consequence of the second
order nature of the transition in that model. The m → 1 limit has a different, less
trivial, interpretation, as we shall see in a minute.
We have now to study the saddle point equations with respect to q1 , q0 , m. First,
the equation ∂q0 F = 0 gives as a solution q0 = 0. This solution is correct in absence
of external magnetic field: q0 is the mutual overlap among different states, and it is
natural to think that without external field the distribution of states in the phase space
is symmetric, and thus all states must be orthogonal to each other.
The two remaining equations ∂q1 F = 0 and ∂m F = 0 are,
2
β p−1 q1
(1 − m) pq − =0
2 1 (1 − q1 )[(m − 1)q1 + 1]

β2 p 1 1 − q1 q1
q1 + 2 log + =0 (83)
2 m 1 − (1 − m)q1 m[1 − (1 − m)q1 ]
These equations can be easily studied on a computer, but most of the physics can be
worked out also graphically. At high T the only solution is q1 = 0 and m undetermined:
this is the paramagnetic solution, which is equal to the RS one. We want to know
whether there is a nontrivial spin-glass solution with q1 6= 0. The first equation is solved
by m = 1. So let us plug m = 1 into the second equation,
β2 p
q + log (1 − q1 ) + q1 ≡ g(q1 ) = 0 (84)
2 1
The graphical study of this equation for 0 ≤ q1 ≤ 1 is trivial. The limits are g(0) = 0 and
g(1) = −∞. At high T the function is monotonous and only the q1 = 0 solution exists.
However, by lowering the temperature, g(q1 ) develops a maximum, whose height diverges
for decreasing T . Therefore, it must exist a temperature Ts , where this maximum
touches the axis at q1 ≡ qs 6= 0. Therefore, at T = Ts a new spin-glass solution appears,
with q1 = qs and m = 1. When T < Ts we have to move m from 1, and one can see that
the solution simply shifts, q1 > qs , m < 1 [10]. The important point is that, unlike the
RS nontrivial solution, this solution is stable. Moreover, its free energy its lower than
the paramagnetic one. The temperature Ts where this nontrivial 1RSB solution appears
is called static transition temperature. It has been proved that the 1RSB solution is
exact in the PSM [10]. This means that if we take higher order RSB ansatz for Qab ,
from the saddle point equations we find that all the extra parameters we introduce have
in fact a trivial value, and that the 1RSB solution is recovered.
As we have seen, at Ts the value of the self-overlap is nonzero, q1 = qs , while m = 1.
This fact has an interesting physical interpretation. For T > Ts the overlap distribution
is trivial, P (q) = δ(q). By decreasing the temperature, finally a second peak appears
at Ts , for q = qs . This value is nonzero, meaning that at the transition the states are
already well formed, that is tight. However, at Ts we have m = 1, meaning that the
height of this nontrivial peak is in fact zero: the probability of these new states is zero at
the transition, and it grows below Ts as m becomes smaller than one. In other words,
it seems that when the states appear, they are already well formed, but have a zero
thermodynamic weight.
This fact has a possible interpretation in terms of metastable states: the calculation
we have just performed is a thermodynamic one, and therefore by its very nature it is
unable to capture the contribution of metastable states. If in this systems there were
some metastable states even above Ts , and some of them became stable only below Ts ,
the behaviour of P (q) would be exactly the one described above. The states are already
present in the phase space, even above Ts , with a well defined nonzero self-overlap, but
their thermodynamic weight is zero, as long as the temperature does not drop below the
static transition. At that point the free energy of these states becomes smaller than the
paramagnetic one, therefore their weight is nonzero, and the P (q) develops a secondary
peak. For now, this is just a well motivated hypothesis. We will see in the following
chapters that it is in fact verified.
Summarizing, in the PSM we find a static transition between a high temperature
paramagnetic phase, and a low temperature spin-glass phase below Ts . In this phase
many pure states dominate the partition function. The order parameter of this unusual
transition is the overlap matrix Qab , and more precisely, within the 1RSB scheme, the
self-overlap q1 and the probability parameter m. In the paramagnetic phase the overlap
matrix has a replica symmetric form, with q1 = 0 and m undetermined, while in the
spin-glass phase there is replica symmetry breaking, with nontrivial values of q1 and
m. The nature of this transition is discontinuous if we consider the parameter q1 , but
continuous if we consider the whole probability distribution P (q).
4. Equilibrium dynamics
As we have already pointed out, the dynamical behaviour of a system will be very
different from its thermodynamic behaviour, if metastable states are present. This
is particularly true in mean-field, where metastable states may have infinite lifetime.
The results from the previous chapter seem to suggest that something nontrivial is
going on in the PSM even for T > Ts . It is therefore important that we perform
an independent dynamical study of the model, and see whether our guess about the
presence of metastable states was right. A nice introduction to the main concepts of
dynamics can be found in [20].
4.1. The generating functional formalism

Before focusing on the PSM, we give a brief summary on how to study the dynamics
of a degree of freedom x(t), described by an Hamiltonian H(x), which contains some
quenched disorder. The starting point of our dynamical study is the Langevin equation
[21],
dx ∂H
=− + η(t) (85)
dt ∂x
where η(t) is a Gaussian noise, playing the role of the thermal agitation, with
hη(t)i = 0 (86)
hη(t)η(t′)i = 2T δ(t − t′ ) (87)

The factor 2T is crucial, since it relates the strength of the noise to the friction coefficient
in the Langevin equation [20]. In the rest of this chapter we will have to integrate
repeatedly over the degrees of freedom x(t), and over the disorder η(t), which are both
functions of time. Thus, most of the integrals will be functional integrals. With the
notation Dx we actually mean a measure over all the paths, i.e. D[x(t)] [23]. The
probability P (η) of the noise can be written as,
Z
1 ′ −1 ′ ′
P (η) ∼ exp − dtdt η(t)D (t − t )η(t ) (88)
2
with
D(t − t′ ) = 2T δ(t − t′ ) (89)
Every solution x(t) of the Langevin equation depends on the particular realization of
the thermal noise η(t), and we indicate it as xη (t). From the probability distribution
on η we can therefore obtain a distribution on x. To work this out, let us compute the
average over the noise of a generic observable A, function of the degree of freedom x(t),
Z Z Z
hA(x)i = Dη P (η) A(xη ) = Dη P (η) dx δ(x − xη ) A(x) =
Z Z
= dx Dη P (η) δ(∂t x + ∂x H − η) A(x) =
Z
= dx P (x)A(x) (90)
where we have defined the probability of x as,

Z
P (x) = Dη P (η) δ(∂t x + ∂x H − η) (91)
Note that in principle we should introduce the Jacobian of the equation in the formula
above. However, it can be proved that if we discretize the Langevin equation according
to the Ito prescription, this Jacobian is in fact equal to 1, and can therefore be neglected
[21, 22].
When the Hamiltonian contains quenched disorder J (as in the case of the PSM) we
must find a way to average over J. Clearly, if we simply average the Langevin equation
we get a disappointing 0 = 0. In fact, the correct quantity which has to be averaged
over the disorder is the distribution P (x) above, such that to compute the average over
J of an observable A we can simply use the formula,
Z
hA(x)i = Dx P (x) A(x) (92)
Rather than precisely compute P (x), we average the integral of P (x), in order to deal
with a scalar quantity. This is the generating functional method [24, 25], which has
been first applied and studied in the field of spin-glasses in [26, 27, 28]. The starting
point of the method is an apparently redundant way to represent the number 1,
Z Z
1 ≡ Z = Dx P (x) = DxDη P (η) δ(∂tx + ∂x H − η) =
Z Z
1
= DxDx̂Dη exp − dtdt′ η(t)D −1 (t, t′ )η(t′ )+
2
Z Z
+ i dt x̂(t)(∂t x + ∂x H) − i dt x̂(t)η(t) ]
Z
1
= DxDx̂ exp − x̂Dx̂ + ix̂(∂t x + ∂x H)
2
Z
≡ DxDx̂ exp [S(x, x̂)] (93)
with S = − 21 x̂Dx̂ + ix̂(∂t x + ∂x H). We used the integral representation of the delta
function and the fact that the functional integral is Gaussian. In our notation we do
R
not indicate explicitly the time contractions: x̂Dx̂ = dt dt′ x(t)D(t, t′ )x(t′ ), and the
same holds for ix̂(∂t x + ∂x H).
The quantity Z is the generating functional. The fact that it is just equal to 1,
must not deceive the student. In fact, we can calculate all the interesting dynamical
quantities with this functional. Let us see how. When we introduce a time dependent
magnetic field in the system, we have an extra term,
Z
dt x(t)h(t) (94)
in the Hamiltonian. Thus we have an extra term h(t) in the original Langevin equation,
which translate into a term, Z
dt x̂(t)h(t) (95)
in the action S above. Therefore, when we derive the average of any quantity with
respect to h(t) we pull down a factor x̂(t) from the exponential, and in particular,
∂
hx(t′ )i = hx̂(t)x(t′ )i ≡ R(t, t′ ) (96)
∂h(t)
which is the dynamical response function of the system, i.e. the dynamical equivalent
of the susceptibility in thermodynamics. On the other hand, we can couple a field ĥ(t)
to x(t) in the generating functional, and get,
∂
hx(t′ )i = hx(t)x(t′ )i ≡ C(t, t′ ) (97)
∂ ĥ(t)
that is the time dependent correlation function. Summarizing, once the conjugate fields
h(t) and ĥ(t) are introduced, we have,
∂Z
R(t, t′ ) =
∂ ĥ(t′ )∂h(t)
∂Z
C(t, t′ ) = (98)
∂ ĥ(t′ )∂ ĥ(t)
Let us now consider a system with quenched disorder in the Hamiltonian. We first
define,
∂H
L(x) ≡ ∂t x + (99)
∂x
and then split the Hamiltonian into a part without disorder H0 , and a part with disorder
HJ , such that the Langevin equation becomes,
L(x) = L0 (x) + LJ (x) = η(t) (100)
with
∂H0 ∂HJ
L0 = ∂t x + LJ = (101)
∂x ∂x
and
hηηi = 2T δ(t − t′ ) ≡ D0 (t − t′ ) . (102)
The generating functional becomes,
Z
1
Z = DxDx̂ exp − x̂D0 x̂ + ix̂ [L0 (x) + LJ (x)] (103)
2
It should be clear by now that Z has, in the dynamical approach, the same role as
the partition function in thermodynamic. This may suggest that, when averaging over
J, we should consider log Z rather than Z, in order to reproduce the quenched case.
However, this is not the case. In fact, the crucial point is that Z = 1, and thus it can
be safely averaged over J ! Therefore in the dynamic approach we do not need replicas
[27]. This does not mean that the calculation will be simpler. Actually, we will see that
time plays the same role as replicas: by averaging over J we will decouple the sites, but
couple different times.
We fact that replicas are not needed in the dynamical case can be understood also
in a more direct way. As we have already said, the correct quantity to average over the
disorder J is the probability distribution PJ (x) of the degree of freedom x. In the static
case we have,
(s) e−βHJ (x)
PJ (x) = R (104)
Dx̂ e−βHJ (x̂)
In order to be averaged over J this expression must be rewritten in terms of replicas,

Z Pn
(s)
PJ (x) = lim Dx̂a6=1 e−β a HJ (x̂a ) , (105)
n→0
where x̂1 = x. On the other hand, from (93) we see that the distribution in the dynamic
case is given by, Z R
(d) dt SJ [x(t),x̂(t)]
PJ (x) = Dx̂(t) e (106)
where we have reinstated the dependence on the time t to emphasize the formal similarity
between the replicated static case and the unreplicated dynamic one. It is clear that in
the latter case we do not need replicas to average over J. However, in the dynamic case
the variable t plays a role analogous to the replica index a.
Expression (103) is interesting for two reasons: first, the coefficients of x̂2 and x̂ are
respectively the correlator of the noise, and the noise-independent part of the original
Langevin equation; second, the disorder J is only contained in LJ at the exponent, and
it can be easily integrated out. This average will renormalize the coefficients of x̂2 and
x̂, giving rise to a new effective Langevin equation [29]. More specifically, the average
over J gives,
Z
1
Z = DxDx̂ exp − x̂D0 x̂ + ix̂L0 (x) exp[ix̂LJ (x)] (107)
2
We define the quantity ∆(x, x̂) as
exp[∆(x, x̂)] ≡ exp[ix̂LJ (x)] (108)
Once the average is done, it is possible in general to isolate various pieces in ∆, and in
particular,
1
∆(x, x̂) = − x̂D1 (x, x̂)x̂ + ix̂L1 (x, x̂) + . . . (109)
2
where L1 renormalizes the disorder-independent part of the Langevin equation L0 , and
D1 renormalizes the noise correlator. In the end we have the effective Langevin equation,
L0 (x) + L1 (x, x̂) = ξ with hξξi = D0 + D1 (x, x̂) (110)

In this equation the disorder is no longer present, but we had to pay a price: the
original equation gets some nontrivial corrections. The most evident difference is that
the variable ξ, the new effective noise, is no longer delta-correlated in time. In other
words the integration over J has introduced a sort of memory in the dynamics of the
system. This phenomenon is common in statistical physics: whenever starting from a
Markovian stochastic process we integrate over some degrees of freedom (the disorder,
the fast variables, the momenta, etc.), we end up with a new effective equation which
is no longer Markovian, and where modes which were previously uncoupled, are now
coupled (a simple example of this phenomenon can be found in [20]).
4.2. Dynamics of p-spin spherical model

In the following section we will apply the technique described above to the PSM [29, 30].
The formalism is trivially generalized to the case of a vectorial degree of freedom σk . In
the Langevin equation we must add a Lagrange multiplier µ(t) in order to impose the
spherical constraint:
∂H
∂t σi (t) = − − µ(t)σi (t) + ηi (t) with hη(t)η(t′)i = 2T δ(t − t′ ) (111)
∂σi
The derivative of the Hamiltonian with respect to σi gives,
∂H p X
=− Jikl σk σl (112)
∂σi p! kl
The generating functional is given by equation (103), with x → σk , x̂ → σ̂k , and,

XZ
iσ̂ · L0 = dt iσˆk (t) [∂t σk (t) + µ(t)σk (t)]
k
Z
ip X
iσ̂ · LJ = − dt Jikl σ̂i (t)σk (t)σl (t) (113)
p! ikl
If we compare this last expression with the static formulas (see equation 3.2), we can
see that in this case the time has the same function as the replica index,
Z X XX
dt Jikl σ̂i (t)σk (t)σl (t) ←→ Jikl σia σka σla (114)
ikl a ikl
The following step is to average over the disorder, that is to compute exp(iσ̂ · LJ ). In
the statics this operation gives a coupling among replicas, in this case we will have a
coupling among times. A technical remark: before averaging, we need to symmetrize
the term σ̂σσ in (113), since the couplings are completely symmetric. We find,
exp(iσ̂ · LJ ) =
Z Y Z
1 2 p−1
= dJikl exp − Jikl 2N − Jikl dt[iσ̂i σk σl + σi iσ̂k σl + σi σk iσ̂l ] =
i>k>l
2p!
Z
dtdt′ p−1 p−2
= exp [p(iσ̂ · iσ̂)(σ · σ) + p(p − 1)(iσ̂ · σ)(σ · iσ̂)(σ · σ) ] (115)
4N p−1
where we used the notation,
N
X
σ·σ ≡ σi (t)σi (t′ ) (116)
i=1
As we anticipated, in the calculation appeared a coupling among different times, through

the overlap of the configuration at time t and t′ . In complete analogy with the static
case, we therefore introduce as an order parameter the dynamical overlap [31, 30], and
get,
exp(iσ̂ · LJ ) =
Z ! !
X X
= DQ δ NQ1 − iσ̂k (t)iσ̂k (t′ ) δ NQ2 − σk (t)σk (t′ ) ·
k k
! !
X X
′ ′
· δ NQ3 − iσ̂k (t)σk (t ) δ NQ4 − σk (t)iσ̂k (t ) ·
k k
Z
pN ′ ′ ′ p−1 ′ ′ ′ p−2
· exp dtdt [Q1 (t, t )Q2 (t, t ) + (p − 1)Q3 (t, t )Q4 (t, t )Q2 (t, t ) ]
4
P
It is clear the similarity between the overlap matrix Qab = k σka σkb /N in the static
P
approach, and Q2 (t, t′ ) = k σk (t)σk (t′ )/N: in the first case we have a static overlap
between configurations belonging to different replicas, in the second case we have a
dynamic overlap between configurations at different times. It is not only Q2 that has
a physical meaning. From their definitions and from the discussion above, we see that
Q3 and Q4 are both a response functions, hσσ̂i, with their time arguments exchanged.
Finally, it is possible to argue that the (hard to interpret) order parameter Q1 = hσ̂σ̂i
must be zero [31]. Summarizing,


 Q1 (t, t′ ) = 0

Q2 (t, t′ ) = C(t, t′ )

(117)

 Q3 (t, t′ ) = R(t′ , t)

 Q (t, t′ ) = R(t, t′ )
4
We now give an exponential representations of the δ-functions. For example,

! Z Z
X
′ ′ ′
δ NQ2 (t, t ) − σk (t)σk (t ) = Dl2 exp iN dtdt (l2 Q2 − l2 σ · σ) (118)
k
and we use the saddle-point method to compute the integral. By setting to zero the
derivatives with respect to all the Q’s, we get the equations,


 il1 = p4 Qp−1
2

il2 = 4p (p − 1)Q1 Qp−2 + p4 (p − 1)(p − 2)Q3 Q4 Qp−3

2 2 ≡0
p p−2

 il3 = 4 (p − 1)Q4 Q2

 il = p (p − 1)Q Qp−2
4 4 3 2
The product Q3 Q4 is zero because of causality: if t > t′ , then R(t′ , t) = 0 and vice versa.
In order to write the effective Langevin equation we have to recognize what are the new
coefficients of σ̂σ̂ and of σ̂σ. From the definition of the l’s we have the following new
term in the generating functional,
XZ
′ p ′ p−1 ′ 1 ′ ′ p−2 ′
∆= dt dt − C(t, t ) σ̂k (t)σ̂k (t ) − p(p − 1)R(t, t )C(t, t ) iσ̂k (t)σk (t ) ,
k
4 2
(119)
Note that at this point the sites in the action of the generating functional are all
decoupled. Therefore we can write an effective Langevin equation for a scalar degree of
freedom σ [29],
Z
1
∂t σ(t) = −µ(t)σ(t) + p(p − 1) dt′′ R(t, t′′ )C(t, t′′ )p−2 σ(t′′ ) + ξ(t) (120)
2
with
p
hξ(t)ξ(t′)i = 2T δ(t − t′ ) + C(t, t′ )p−1 (121)
2
Note that the average over the disorder did not generate terms σσ, which we would not
know how to interpret. The effective Langevin equation does not contain the disorder,
and it is uncoupled in the sites. However, it is more complicated than the original one,
since the noise is no longer δ-correlated in time, and we have an explicit memory term
at the r.h.s., that is a non-local kernel which couples the external time t with all the
earlier times t′′ < t.
4.3. Equations for the correlation and the response

We now want to use the effective Langevin equation to write some self-consistent
equations for the correlation and the response function. In order to do this we have
to introduce some useful formal relations. The first one is already known,
∂hx(t)i
R(t, t′ ) = = hx(t)x̂(t′ )i (122)
∂h(t′ )
The second relation is the following,

∂x(t)
h i=
∂η(t′ )
Z Z
1 −1 ∂
Dη exp − ηD η DxDx̂ x(t) exp[x̂(∂t x + ∂x H) + x̂η]
2 ∂η(t′ )
= hx(t)x̂(t′ )i = R(t, t′ ) (123)
The last relation is a little harder to prove,
hx(t)η(t′ )i =
Z
1 −1
= DηDxDx̂ exp − ηD η x(t)η(t′ ) exp[x̂(∂t x + ∂x H) + x̂η + jη]|j=0 =
2
Z
1 −1 ∂
= DηDxDx̂ exp ηD η x(t) exp[x̂(∂t x + ∂x H) + x̂η + jη]|j=0 =
2 ∂j(t′ )
Z
∂ 1
= ′
DxDx̂ x(t) exp[− x̂Dx̂ + x̂(∂t x + ∂x H) + jDj + jDx̂ + x̂Dj]|j=0
∂j(t ) 2
Z Z
= DxDx̂x(t) dt′′ D(t′ , t′′ )x̂(t′′ ) exp[S(x, x̂′ )]
Z
= dt′′ D(t′ , t′′ )R(t, t′′ ) (124)
Note that this last equation is only valid for Gaussian noise. We are now ready to write
the two equations for the response and the correlation function. To get the equation
for the response we differentiate the effective Langevin equation with respect to the
effective noise and average,
∂R(t1 , t2 ) ∂ δσ(t1 ) δ σ̇(t1 )
= h i=h i=
∂t1 ∂t1 δξ(t2) δξ(t2)
Z t1
1
= − µ(t1 )R(t1 , t2 ) + p(p − 1) dt′′ R(t1 , t′′ )C p−2 (t1 , t′′ )R(t′′ , t2 ) +
2 t2
+ δ(t1 , t2 ) (125)
The equation for the correlation is obtained by multiplying the effective Langevin
equation by σ and averaging,
∂C(t1 , t2 ) ∂
= hσ(t1 )σ(t2 )i = hσ̇(t1 )σ(t2 )i =
∂t1 ∂t1
Z t1
1
= − µ(t1 )C(t1 , t2 ) + p(p − 1) dt′′ R(t1 , t′′ )C p−2 (t1 , t′′ )C(t′′ , t2 ) +
2 −∞
+ hξ(t1 )σ(t2 )i (126)
where we can use the third one of the relations above and get,
Z Z
p t2 ′′
′′ ′′ ′′
hξ(t1 )σ(t2 )i = dt D(t1 , t )R(t2 , t ) = 2T R(t2 , t1 ) + dt R(t2 , t′′ )C p−1 (t1 , t′′ )
2 −∞
(127)
Because of causality, the term 2T R(t2 , t1 ) is zero if t2 < t1 , as we shall assume. Finally we
have to get rid of the Lagrange multiplier µ(t). Differentiating the constant C(t, t) ≡ 1,
we obtain [∂t C(t, t′ ) + ∂t′ C(t, t′ )]t,t′ =s = 0, giving the equation [30],
Z t1
1
µ(t1 ) = p2 dt′′ R(t1 , t′′ )C p−1 (t1 , t′′ ) + T (128)
2 −∞
These are the exact dynamical equations for the PSM. When they were first derived
in [29] it was immediately noted that they were formally identical to the approximated
equations formulated by Mode Coupling Theory (MCT) for structural glasses [32, 33, 34,
35]. This observation is at the heart of the theory for the glass-transition in structural
glasses inspired by p-spin spin-glass models [36]. The physics of the PSM has probably
something to do with structural glasses, at least at the dynamical level, and provided
that MCT works well. Moreover, it looks like MC theory must also work in the PSM,
since it gives the same equations ! Let us analyze this last point more in detail.
4.4. Diagrammatic technique and Mode Coupling approximation

The dynamical equations (125) and (126) can indeed be obtained within the Mode
Coupling approximation. Within this approach we consider the perturbative expansion
of the Langevin equation and write all the physical quantities using a diagrammatic
Φ= + + +...
Figure 1. Diagrammatic representation of the perturbative solution to equation 130.

The various terms of the equation for φ(t) are represented: a line stands for the bare
propagatore R0 , a cross indicates the noise. As usual in Feynman diagrams, a vertex
stands for a time convolution.
representations. For the rest of this section see [38, 39]. For the sake of simplicity, we
consider the case of a single scalar degree of freedom φ, with an energy
µ(t) 2 g p
H= φ + φ (129)
2 p!
and we assume that the dynamics of φ is described by the Langevin equation
∂φ g
= −µ(t)φ − φp−1 + η (130)
∂t (p − 1)!
with the initial condition φ(0) = 0. Note that this Hamiltonian is a scalar version,
without disorder, of the p-spin one. The thermal noise η is defined as in the previous
case.
∂ −1
We consider the inverse operator R0 = [µ(t) + ∂t ] , which we use to write the
perturbative expansion of φ(t). In figure 1 we can see the diagrammatic representation
of this expansion in the case p = 3. In this case we can write the following equation:
g
φ(t) = R0 ⊗ η − R0 ⊗ {R0 ⊗ η · R0 ⊗ η} + . . . (131)
2!
Rt
where ⊗ stands for time convolution: (R0 ⊗ f )(t) = 0 dt′ R0 (t, t′ )f (t′ ). The explicit
expression of R0 is Z t
′
R0 (t, t ) = exp − du µ(u) (132)
t′
as we can easily see by differentiating φ(t). The correlation and response function can
be written as,
C(t, t′ ) = hφ(t)φ(t′)i (133)
∂φ(t) 1
R(t, t′ ) = h ′
i= hφ(t)η(t′ )i (134)
∂η(t ) 2T
These functions can be diagrammatically represented, in figure 2 we show the case p = 3.
In what follows we shall assume that all tadpoles (like the second diagram in figure 2) are
C(t,t’) = + + + ...
R(t,t’) = + + + ...
Figure 2. Diagrammatic representations of the perturbative expansions of the

correlation and response function. These diagrams are obtained combining all the
different terms of the diagrams for the perturbative solution.
already resummed. The contribution of these diagrams to the self-energy Σ that we are
going to write is in fact simply a constant: disregarding tadpoles is equivalent to operate
mass renormalization in a usual field theory [2, 39]. The diagrammatic expansion of C
and R can be self-consistently resummed, given the following Dyson equations (fig. 3,
upper panel),
Z t Z t1
′ ′
R(t, t ) = R0 (t, t ) + dt1 dt2 R0 (t, t1 )Σ(t1 , t2 )R(t2 , t′ ) (135)
t′ t′
Z t Z t′
′
C(t, t ) = dt1 dt2 R(t, t1 )D(t1 , t2 )R(t′ , t2 ) (136)
0 0
where the self-energies (or kernels) Σ(t, t′ ) and D(t, t′ ) are, as usual, the sum of all the
amputated connected diagrams. If we multiply by R0−1 we can write the equations in
the following way:
R0−1 ⊗ R = I + Σ ⊗ R (137)
R0−1 ⊗ C = D ⊗ R + Σ ⊗ C (138)
where I is the identity operator. Explicitly we have
Z t
∂R(t, t′ ) ′ ′
= −µ(t)R(t, t ) + δ(t − t ) + duΣ(t, u)R(u, t′) (139)
∂t t′
Z t′ Z t
∂C(t, t′ ) ′ ′
= −µ(t)C(t, t ) + duD(t, u)R(t , u) + duΣ(t, u)C(u, t′)
∂t 0 0
Up to know this was very general, and most importantly exact. The mode coupling
approximation (MCA) consists in approximating the kernels Σ(t, t′ ) and D(t, t′ ): we
neglect all the vertex corrections and keep only line corrections, that is we take the
values of Σ and G at order g 2 and substitute in them the bare response and correlation
R R0 R0 R
= + Σ
C R R
= D
Σ ~ D ~
Figure 3. The first two lines are the exact Dyson equations for the correlation and
the response. The third line gives the value of the kernels Σ and D within the Mode
Coupling Approximation.
by their renormalized values, R0 → R, C0 → C. In this way we get the following

equations:
g 2 p−1
Σ(t, t′ ) = C (t, t′ )R(t, t′ )
2
g2
D(t, t′ ) = 2T δ(t − t′ ) + [C(t, t′ )]p (140)
6
If we now plug this MCA form of Σ and D into equations (139), it is easy to see that they
become identical to the equations we wrote for the p-spin model in the previous section
within the generating functional method (we have considered a scalar field, but the same
equations can be obtained for a vectorial field). This raises an interesting question: how
is that the equations obtained with the Mode Coupling approximation are identical to
those obtained with the exact generating functional method ? The answer is that in a
mean-field disordered system, thanks to the scaling with N of the couplings J, vertex
corrections are sub-leading, and vanish when N → ∞, while line corrections remain
finite in the thermodynamic limit. In other words MCA is in fact exact for mean field
systems !
To understand this fact, we can write the Hamiltonian with a vector σi (where
P
i = 1, . . . , N). The interaction term for p = 3 is g i<j<k Jijk σj σj σk . In this case the
average value of J 2 is 1/N p−1. When averaging over the disorder, the behavior in the
limit N → ∞ is different for vertex and line corrections. In the first case we have (for
p = 3) that in the term Jijk Jjlm Jmni Jkln the average over the noise causes the indices to
couple two by two, e.g. i = l and k = m. We obtain a factor (1/N 2 )2 , which we must
multiply by N 4 having in the best case four free indices over which we have to sum. This
factor is then of order 1 and vanishes once we normalize the correlation by N. Instead
in the line correction the average over the noise causes only two indices to coincide in
the best case, giving a N factor that remains finite also after the normalization [39].
Although a bit boring, this last section proved an important point: the exact
dynamical equations of the PSM are identical to the MC equations, and in particular
to the (approximated) equations that MC theory writes for deeply supercooled liquids
close to the glass transition. This is one of the main evidences supporting the idea that
the PSM is a sort of mean-field model for structural glasses, and that some of the main
physical concepts valid in the PSM, should be valid in real glasses as well. As long as
one believes that MC theory describes reasonably well fragile glasses [40, 41], one has
to accept that the physics of these systems caught by MC must have something to do
with the physics of the PSM.
4.5. The dynamical transition

It is now time to solve the dynamical equations of the PSM. In order to do this we
must make some simplifications on the correlation and response functions. First, we
shall assume that Time Translation Invariance (TTI) holds: correlation and response
no longer depend independently on the two times, but only on their difference. This
is true only at equilibrium, and therefore we are restricting ourselves to equilibrium
dynamics. The second simplification is in fact a consequence of TTI, and is the validity
of the Fluctuation-Dissipation Theorem (FDT). These two properties can be written as,
(
C(t1 , t2 ) = C(t1 − t2 ) ≡ C(τ )
TTI: (τ ≡ t1 − t2 ) (141)
R(t1 , t2 ) = R(t1 − t2 ) ≡ R(τ )
1 dC(τ )
FDT: R(τ ) = − (142)
T dτ
Using these formulas, and a bit of algebra, we can reduce the two coupled equations
(125) and (126) to a single equations for C(τ ), namely [30],
Z τ
p
Ċ(τ ) = −T C(τ ) − du C p−1 (τ − u)Ċ(u) (143)
2T 0
A crucial point: in order to perform the integrals we have supposed C(∞) = 0, that
is we assumed that there is no ergodicity breaking: after a sufficiently long time the
dynamic configuration must be allowed to go as far as possible from the initial time
configuration. In terms of overlap, this means that the overlap between σ(τ = 0) and
σ(τ = ∞) must be zero. Recall that we have,
1 X
C(τ ) = hσk (τ )σk (0)i (144)
N
k
so that the dynamical correlation function is exactly the average overlap between the two
configurations at times 0 and τ . Therefore assuming unbroken ergodicity is equivalent
to assume that C(∞) = 0. Ergodicity will be verified self-consistently at the end of the
calculation.
The first term on the r.h.s. of equation (143) comes from the J-independent part
of the original Langevin equation, i.e. from the spherical constraint. The second term,
that comes from the integration over J, is clearly a memory term, for it causes the
properties of the system at time t to depend on all times between 0 and t.
By imposing the physical condition Ċ(τ ) ≤ 0 (the average correlation cannot
increase with time), we obtain from (143) the following relation [37],
2T 2
C p−2 (τ )[1 − C(τ )] ≤ . (145)
| {z } p
≡g(C)
This inequality can be easily studied graphically: for τ = 0 we have C = 1 and g(1) = 0,
and for τ = ∞ we have C = 0 and g(0) = 0. The function g(C) has thus a maximum
between zero and one: let us call qd the position of this maximum, given by,
p−2
qd = (146)
p−1
The r.h.s. of the inequality is a constant larger the larger the temperature. At very
high temperatures the inequality is always satisfied, since g(qd ) ≪ 2T 2 /p. This is the
paramagnetic phase, which is indeed ergodic. When we lower the temperature, the
difference between 2T 2 /p and g(C) gets smaller. From equation (143) we see that this
difference is proportional the time-derivative of C: it is large when C ∼ 1, that is for
short times, it becomes smaller when C ∼ qd , and again large for C ∼ 0, i.e. for very
long times. In other words, when we lower the temperature, we observe the formation
of a plateau of the correlation function, with C(τ ) ∼ qd .
If we lower further the temperature, we arrive at a point where the r.h.s. of the
inequality touches the curve, i.e. there is a temperature Td such that 2Td2 /p = g(qd ).
Using (146) and the definition of g(C), we have,
s
p(p − 2)p−2
Td = (147)
2(p − 1)p−1
At this temperature the correlation function remains stuck at a plateau C = qd , since
Ċ = 0. Ergodicity is therefore broken. We cannot go below Td , since all our assumptions
are violated in this phase, and in particular C(∞) 6= 0. What we have just proved is that
there is a dynamical transition at Td : the system passes from the paramagnetic state,
to a phase where ergodicity is broken. Let us give a physical interpretation of what we
have found, in terms of overlap of the configuration at time t with the configuration at
time 0.
T ≫ Td : The dynamical overlap (i.e. the correlation function) rapidly decays to zero,
such that the configuration goes as far as it wants in the phase space from its initial
position. This is the fast equilibrium dynamics in the paramagnetic state. In this
phase relaxation is exponential, and nothing particularly exciting happens.
>
T ∼ Td : In this phase the dynamics is still ergodic, but something strange happens:
for a long time (longer the closer to Td we are) the configuration stays close to its
initial value, since C(τ ) has a plateau. More precisely, the dynamics explores a
“spot” of phase space around the initial configuration, of largeness roughly equal
to qd , which is the value of C at the plateau. This “spot” cannot be a true state: if
it were, the system would stay trapped there forever, while in this case, after a long
while, the system drifts away. Eventually the overlap goes to zero, consistently with
the paramagnetic state. So what is going on ? Why the system is almost trapped
close to the dynamical transition ? We shall answer these key questions in the next
chapter.
T → T+d : The plateau becomes infinite, the correlation function does not decay
anymore, so that the system takes an infinite time to equilibrate. The configuration
remains close to its initial position for an infinite time, and it is clearly trapped by
a state of self-overlap qd . Ergodicity is broken and our dynamical equations break
down.
The behaviour of the correlation function we have just described is not a peculiarity of
the PSM. In fact, it is the typical phenomenology of glassy systems, and in particular
of structural glasses. The core aim of Mode Coupling Theory is indeed to explain this
phenomenology in fragile glasses. The interpretation of the plateau in finite dimensional
glasses is usually given in terms of cage effect: at low enough temperature, each particle
is surrounded by a cage of nearest-neighbor particles, and it takes a long time (longer
the lower the temperature) to the particle to break this cage and achieve asymptotic
relaxation. This interpretation is very nice for structural glasses, but of course it cannot
be applied to the mean-field PSM, where there is no space structure, nor cage. However
the behaviour of the correlation function is indeed the same. This suggests that the cage
effect must have a deeper interpretation, which must be valid both in finite-dimensional
and mean-field systems. We will propose such an interpolation in the next chapter.
Let us make a brief summary of the dynamical results. Using the generating
functional method we obtained two exact equations for the correlation and the response,
which are formally identical to those obtained with the MC approximation. In fact MC
is exact for the mean-field PSM. We studied the equations assuming that the system is at
equilibrium and that ergodicity is not broken, in other words we studied the properties
of the (ergodic) paramagnetic state. The correlation function decays to zero, but it
develops a plateau as the temperature lowers. In particular, as T → Td+ the plateau
diverges and ergodicity is broken in this limit. Thus, this must be the limit of existence
of the paramagnetic phase, and therefore Td marks a dynamical transition in the system.
The relaxation time (roughly, the time the correlation takes to decay to zero) diverges
at Td .
A natural question is whether this dynamical transition at Td coincides with
the static one at Ts . The answer is no. One can easily check that Td > Ts :
dynamically, the ergodicity is broken at a temperature higher than the thermodynamic
singularity. Clearly, metastability, combined with the peculiar features of mean-field,

must be responsible for this: metastable states trap the dynamics at Td , while the
thermodynamics cannot detect these states. This is what we suspected when we studied
the statics: at Ts the equilibrium spin-glass states appeared as a secondary peak in the
P (q). This peak had zero weight, but nonzero overlap qs , suggesting that even above
Ts some metastable states existed. From the dynamics we have a clear evidence that
metastable states exist from the simple fact that Td > Ts . The mean-field nature of the
PSM makes the barrier around these states infinite, so that the equilibrium states are
never reached.
A last important comment: we assumed ergodicity, so we can see ergodicity
breaking at Td as the limit of validity of our calculation. In fact, what we should do for
T ≤ Td is to give up the assumption of TTI and FDT, and solve the full equations. This
is hard, but can be done (with some suitable approximation) [42, 43]. What is found in
this way is that if the system starts from a high temperature (random) configuration,
it never reaches the static equilibrium energy, and in this sense dynamics and statics
strongly differ. However, the correlation function behaves differently from what one may
think. The plateau is not infinite below Td , but rather has a length that increases as the
earlier of the two times in C(t, t′ ) increases. This phenomenon is known as aging, and
its description is beyond the scope of these notes [44]. What happens is the following:
if we constrain the system to be at equilibrium, then for T → Td we have divergence
of the relaxation time, and thus ergodicity breaking. Therefore we cannot study the
equilibrium properties of the system for T ≤ Td , and we have to give up equilibrium.
When this is done, what we find is a weak ergodicity breaking below Td , which is an
intrinsically off-equilibrium phenomenon [42].
Even though we have an explanation in terms of metastable states of why the two
sets of results from statics and dynamics differ, it would be nice to have a way to unify
the two pictures, and obtain both results. This will be achieved in the next chapter.
5. Complexity
We have seen that in the PSM different results are obtained from the static and dynamic
approaches. Are we able to find a unifying approach, within which it is possible to give
an interpretation of all the results collected until now ? The answer is yes. By now
we have understood that the discrepancy between statics and dynamics is due to the
presence of many metastable states. It is time to directly study these states.
5.1. What is the TAP free energy ?

Up to know we have seen the effect of the existence of many pure states only in a indirect
way. In the statics, we had to break the replica symmetry because of ergodicity breaking,
but the free energy we computed was not the individual free energy of the states, but
rather the average free energy over all the thermodynamically relevant states. On the
other hand, we have seen that dynamically the ergodicity is broken at a temperature
Td > Ts because of the presence of metastable states. However, even in that case we
could not put our finger on the individual states trapping the dynamics. To do this we
need to introduce the TAP free energy.
Pure states are objects living in an N dimensional phase space: in each state α
the local magnetizations have a well defined value depending on the site, mαi = hσi iα ,
and a state is identified by the vector of its magnetizations. Therefore, what we need is
a function defined on this space, i.e. a function of the local magnetizations mi , whose
local minima coincide with the pure states of the system. The minimization of such
a function must provide a set of equations for the vector mi , equivalent to the mean
field equation for m in the Ising model, m = tanh(βm). This function is the mean-field
free energy, which is known in the context of spin-glasses as Thouless-Anderson-Palmer
(TAP) free energy, fTAP (m1 . . . mN ) [45].
It is important to stress that the mean-field, or TAP, free energy is a function of
the magnetizations mi and not of the microscopic degrees of freedom σi . In particular
its minima do not necessarily coincide with the energy minima, that is the minima of
the Hamiltonian H(σi ). In fact, pure states cannot in general be simply identified with
minima of the energy. The problem is that different energy minima may be separated
by energy barriers which at high temperature are small compared with kB T , and thus
belong to the same pure state. Even though for T → 0 a state essentially collapses onto
its lowest energy configuration, it is important to keep the two concepts distinct. A
pure state α, identified by the vector mα1 . . . mαN , is fundamentally a subcomponent of
the Gibbs measure, h·iα . As we have stressed in the first chapter, a pure state enjoys
the crucial clustering property, property that is meaningless when referred to a simple
configuration σ1 . . . σN .
The TAP free energy density for the PSM is the following [46, 47, 48],
1 X 1 β
fT AP = − Jikl mi mk ml − log(1 − q) − [(p − 1)q p − pq p−1 + 1] (148)
Np! ikl 2β 4
with,
1 X 2
q= mi ; mi = hσi i (149)
N i
The first term is the energy, the second term is minus the entropy multiplied by the
temperature, and the third one is the so-called reaction term [49]. By setting mi = 0 for
each i, we get fT AP = −β/4, the correct result for the paramagnetic state we already
met in the statics. The mean-field equations are obtained by finding the minima of
the TAP free energy, ∂mi fT AP = 0 for i = 1 . . . N. However, in order to study these
equations, it is convenient to change variables [47]. Let us introduce the new set of
variables {σ1 . . . σN ; q}, defined in the following way,
√ X
mi = q σi σi2 = N (150)
i
These variables σi (sometimes called angular variable) are formally different from the
original spin degrees of freedom, even though they play a very similar role, and are
subject to the same spherical constraint. In terms of the new variables the TAP free
energy becomes,
1 p/2
fT AP (σi , q) = q H(σ) + R(q, β) (151)
N
where H is formally the original Hamiltonian (this clarifies why we called σi the new
variables), and R is the q dependent part in (148). Now we must minimize this free
energy with respect to the angular variables σi and the self-overlap q, taking into account
the spherical constraint. We have,
 (
∂H(σ)

 ∂fT AP =0 i = 1...N

 =0 → P∂σi 2
∂σi i σi = N (152)

 ∂f T AP 1 p p
−1 ∂R

 = 0 → N 2 q 2 H(σ) + ∂q = 0
∂q
The first N equations, at fixed values of the random couplings J, contain all the
complexity of the problem: if there are many states, i.e. many solutions of the mean-
field equations, it is because of these N equations. What is surprising is that in the PSM
these equations do not depend on the temperature ! Moreover, they formally coincide
with the minimization equations of the Hamiltonian of the model. Once we have a
solution of these first N equations, call it σiα , we can compute its zero temperature
energy Eα = H(σ α), and plug it into the equation for q. This equation does depend on
β, so that the self-overlap of a state depends on its zero temperature energy and on the
temperature.
This result is surprising. We said above that in general the minima of the mean-
field free energy do not coincide with the minima of the Hamiltonian, but we seem to
have right here an exception to this rule: in the PSM minima of the TAP free energy
are basically minima of the Hamiltonian. Their positions in the phase space does not
depend on the temperature, while their self-overlap does. In other words, in the PSM
there is a one-to-one mapping between minima of the Hamiltonian (the energy) and
states, i.e. minima of the free energy. At T = 0 a state (stable or metastable) is just a
minimum (absolute or local) of the energy. When T grows energy minima get dressed
up by thermal fluctuations, and become states. So the structure of states of the PSM
is just the structure of minima of the Hamiltonian.
It is very important to understand that this is a peculiar feature of the PSM, due
to its homogeneous nature, and that in general it is not like that. However, in the PSM
such a simplification holds. If one wants to extend such a simplification to more realistic,
finite-dimensional systems (as structural glasses), it is crucial that the temperature is
small enough and times are short. Nevertheless, in realistic systems barriers are finite,
so that identifying minima of H with states is in fact conceptually very risky.
In the PSM, thus, the zero temperature energy density, or bare energy density
E = H(σ) of the minima of the Hamiltonian, is the only relevant quantity to label
states. The self-overlap q, the free energy density f , and the finite temperature energy
E thus depend on the bare energy and on the temperature,
q = q(E, β)
f = f (E, β) (153)
∂(βf )
E = E(E, β) = (154)
∂β
Where, of course, E(E, ∞) = E: the zero-temperature average energy of a state is equal
to its bare energy.
5.2. Definition of complexity (and a problem with the modulus).

The states of the systems, i.e. the minima of the TAP free energy, have the same
structure as the minima of the Hamiltonian. Therefore, we want to study the structure
of minima of H, and in particular their number. The number of minima N grows
exponentially with the size of the system,
N ∼ eN Σ (155)
The quantity Σ is called complexity in the spin-glass community, and configurational

entropy in the glass community, where minima of the potential energy are considered.
In order to compute N (and thus Σ), we have to compute the number of solutions of
the equations,
∂H
=0 i = 1...N (156)
∂σi
By calling σα a solution of these equations, we have,
Z N
X
N = Dσ δ(σ − σα ) (157)
α=1
By using the standard formula,

X δ(σ − σα )
δ(∂H) = (158)
α
|∂∂H|
we have [50], Z
N = Dσ δ(∂H) |∂∂H| (159)
where ∂∂H is a short-cut for the determinant of the second derivative matrix of H (the
Hessian). Here we have two problems: first, having the modulus in such an equation is
algebraically very unpleasant; second, in this way we are counting all stationary points
of H, not simply minima, but also unstable saddles, which can hardly be associated to
pure states of the system. To solve the first problem we are tempted to disregard the
modulus, and define, Z
N̂ = Dσ δ(∂H) ∂∂H (160)
However, now we have a very severe problem: this quantity is a topological invariant
(the Morse constant), which has no connections whatsoever with the number of minima
[51]. By disregarding the modulus we are weighting each stationary point with the sign
of its Hessian, such that,
Z X
N̂ = Dσ δ(σ − σα ) sign(∂∂H) = +1 − 1 + 1 − 1 + 1 . . . (161)
α
The situation seems to be going from bad to worse. However, if we restrict our counting
to a fixed energy density level E, things improve a lot. What we want to do is to count
minima of H which have energy E. To do this we can use the formula,
Z
N (E) = Dσ δ(∂H) ∂∂H δ(H − E) (162)
By restricting ourselves to the level E and by keeping E low enough we can hope that
we are in a region of the phase space where minima dominate, and thus where the
Hessian is positive and the modulus can in fact be disregarded [52]. This is certainly
true close to the ground state E0 . Moreover, the quantity defined in (162) has a further
advantage in its very limitation: if we push E high enough to arrive in a region which
is no longer dominated by minima, but by saddles, we expect to have some instability
in the calculation due to the change in the sign of the Hessian. Thus, we expect that
an instability in our calculation will be telling us something relevant about the nature
of the stationary points we are counting.
5.3. The calculation of the complexity.

To find the stationary points of H with the spherical constraint we can use the Lagrange
method. In this way we obtain [53],
p X 1
− Jikl σk σl − p H(σ)σi = 0 (163)
p! kl N
Given that we want to fix the energy density H(σ)/N = E, the equations become,
p X
− Jikl σk σl − pEσi = 0 (164)
p! kl
and thus we have,

Z ! !
Y p X p(p − 1) X
N (E) = Dσ δ − Jikl σk σl − pEσi det − Jikl σl − pEδik
i
p! kl p! kl
(165)
with the complexity given by,
1
Σ(E) = lim log N (E) (166)
N →∞ N
In order to average over J we have to understand whether the self-averaging quantity

is N or Σ. In general it is the complexity, since extensive quantities are self-averaging,
while exponentials are not (in the statics we had to average F rather than Z). However,
in the PSM we have a further simplification: in absence of external magnetic field we
have that
logN = logN (167)
and thus we can simply average the number, which is much simpler than averaging the
logarithm of it. Equation (167) holds because the PSM is a 1RSB system at the static
level. It can be proved [54] that if the static overlap matrix of configurations is kRSB,
the corresponding overlap matrix of magnetizations is (k − 1)RSB. For the PSM this
implies that the complexity calculation is 0RSB and can therefore be performed at an
annealed level. We give the usual exponential representation of the δ-function,
Z N
!
Y Dµ X
δ(Xi ) = N
exp i µ i Xi (168)
i
(2π) i=1
On the other hand, for the determinant we can use an integral representation in terms
of Grassmann variables (fermions) [23],
Z N
!
X
detAik = D ψ̄ Dψ exp ψ̄i Aik ψk (169)
ik=1
where ψ̄ and ψ are anti-commuting N-dimensional Grassmann vectors,
{ψ̄i , ψi } = 0 (170)
Note that we could have used commuting variables to write the determinant, but at the
price of introducing replicas [50]. So, putting all together, we have,
Z
1 1 Dµ
Σ(E) = logN (E) = log Dσ D ψ̄Dψ exp[S(σ, µ, ψ̄, ψ)] (171)
N N (2π)N
where the action S is given by,

X p X X p(p − 1) X
S(σ, µ, ψ̄, ψ) = −ipE µi σi − i Jikl µi σk σl − pE ψ̄i ψi − Jikl ψ̄i ψk σl
i
p! i
p!
ikl ikl
(172)
Part of this action depend on the disorder and it therefore must be averaged over the
couplings Jikl . This is not difficult to do, since these are Gaussian integral of the form,
YZ
1 2 2N p−1
exp[SJ ] = dJikl exp − Jikl − Jikl (. . . ) (173)
ikl
2 p!
we only have to be careful about a few technical details: first, the terms µσσ and
ψ̄ψσ must be symmetrized before averaging; second, it can be proved that the mixed
commuting-anticommuting terms obtained from the integrals are zero [52], so that we
can effectively treat separately the commuting and anticommuting parts; third,
√ we have
to remember that we are integrating on the surface of a sphere of radius N , due to
the spherical constraint. The integral of the commuting part can be performed exactly,
while for the anticommuting part we will have to work a bit more. Once the J integral
is performed, we obtain,

1 1 2 1
Σ(E) = − log(p/2) + − E + log I (174)
2 2 N
The term in square bracket comes from the J integral of the commuting part, while I
comes from the fermionic part,
 !2 
Z
1 X X
I = D ψ̄Dψ exp − p(p − 1) ψ̄i ψi − pE ψ̄i ψi  (175)
4N i i
To treat this integral we use an inverse Gaussian integration (Hubbard-Stratonovich

transformation), and write,
Z Z " #
Nω 2 X
I = D ψ̄Dψ dω exp − + (iω − pE) ψ̄i ψi =
p(p − 1) i
Z
Nω 2
dω exp N − + log(iω − pE)
p(p − 1)
Z
dω exp [NG(ω)] (176)
where we have performed the (diagonal) fermionic integral. The crucial feature of this
formula is the factor N in the exponential: for N → ∞ we can use the saddle-point
method and write,
I = exp[NG(ω̂)] (177)
where ω̂ satisfies the saddle-point equation,

∂G(ω)
=0 (178)
∂ω ω̂
It is easy to check that the saddle-point solution ω̂ lies on the imaginary axis, and thus
it is convenient to define,
ω = iz (179)
such that,
Nz 2
G(z) = + log(z − pE) (180)
p(p − 1)
Finally we can write the complexity of the PSM as,
1 1 2 N ẑ 2
Σ(E) = − log(p/2) + − E + + log(ẑ − pE) (181)
2 2 p(p − 1)
where the solution ẑ of the saddle point equation ∂z G(ẑ) = 0 is,

s !
p 2(p − 1)
ẑ = E + E2 − (182)
2 p
The second root gives a sub-leading contribution in the thermodynamic limit [53].
5.4. Threshold energy and saddles.

From the form of ẑ we clearly see that something weird happens when the absolute
value of the energy E becomes too small. In fact, for the complexity to be a well
defined physical quantity we must have ẑ real. If we define the threshold energy as
[47, 42], s
2(p − 1)
Eth = − (183)
p
we can write,
q
p 2
ẑ = E + E 2 − Eth (184)
2
We see that ẑ is real, and thus the complexity physically defined, only for,
E ≤ Eth (185)
What have we obtained ? If we plot the complexity, we see that it is an increasing

function of E, with negative second derivative. The complexity is zero at an energy
E0 : below this energy the complexity is negative, and thus the number of states is
exponentially small in the thermodynamic limit. The energy E0 corresponds thus to
the ground state of the system, the lowest part in our landscape. On the other hand,
the complexity grows up to E = Eth beyond which it is no longer defined, since ẑ takes
an imaginary part. Therefore, the interval [E0 : Eth ] is the physical band of states of
the PSM, and all the states with E > E0 are metastable.
A natural question at this point is: what happens above the threshold energy ?
Why is the complexity no longer defined in that regime ? In order to answer these
questions, we have to remember that we disregarded the modulus of the determinant
of the Hessian, and that we therefore expected to have some problems if minima were
no longer dominant in the energy regime under consideration. This is exactly what
is going on here: above Eth minima are not dominant anymore, but unstable saddles
are, so the Hessian gets the contribution of the negative eigenvalues of saddles. To see
this we have to recall that the anticommuting (fermionic) part of our total integral was
basically nothing else that the average determinant of the Hessian,

∂H
∆ = det (186)
∂σi ∂σk
What we have obtained above can thus be rewritten as,

N ẑ 2
∆ = exp (ẑ − pE)N (187)
p(p − 1)
We see that as long as E ≤ Eth , and thus ẑ is real, ∆ is positive (it is easy to see
that the term ẑ − pE is positive). This means that below the threshold the Hessian
is on average positive-defined, and this is the same as saying that on average minima
dominate in this energy regime. On the other hand, for E > Eth , we have ẑ = a + i b,
and if we plug this into ∆ we obtain [55],
∆(E) = (−1)k(E)N exp[Ng(E)] (188)
where k(E) and g(E) are two not-too-complicated functions of E. In this energy regime,
thus, the sign of the determinant oscillates when N goes to infinity. This is exactly what
we would expect from the determinant of a matrix with k(E)N negative eigenvalues. In
fact, it is possible to calculate the eigenvalue spectrum of the Hessian and prove that
k(E) is exactly the fraction of negative eigenvalues of the Hessian [55].
The physical picture is therefore the following: below the threshold the energy
landscape is dominated by minima, and the Hessian is positive on average. In this
phase disregarding the modulus is harmless, and the complexity we find in this way
is well defined. On the other hand, above the threshold the landscape is dominated
by unstable saddles, and the average determinant gets an oscillating part. Having
disregarded the modulus, we detect this transition as the point where the complexity
develops an imaginary contribution. However, if we are not too picky, we can define
a new physical complexity in this phase, by isolating the factor (−1)kN and taking
the logarithm of the real part exp[Ng]. By doing this we are in fact computing the
complexity of saddles dominating at energy E > Eth [56].
5.5. The equation for the self-overlap

What we have said above about the threshold and saddles may seem a bit exotic. In
order to check all that, it is sound to consider the remaining equation for the self-overlap
q. Once we specify the bare energy E of a minimum, we can work out the self-overlap of
the associated finite T state. We expect that a bona fide pure state, i.e. a minimum of
the TAP free energy, must have a well defined self-overlap, indicating roughly the size
of the state in the phase space. On the other hand, we definitely do not expect saddles
to have a well defined self-overlap, since saddles are not trapping stationary points, and
it is hard (although perhaps not impossible [57]) to define their size.
Given a solution with bare energy E of the first N equations, the corresponding
equation for q reads,
p p 1 β
− q 2 −1 E + − p(p − 1)q p−1 − p(p − 1)q p−2 = 0 (189)
2 2β(1 − q) 4
By introducing the auxiliary variable,

1 p−2
y= q 2 (1 − q) (190)
T
we can rewrite the equation as,
p
2
−E ± E 2 − Eth
p(p − 1)y 2 + 2pEy + 2 = 0 giving y = (191)
p−1
where Eth is the same threshold energy as we have introduced in the calculation of the
P
complexity. Given that the self-overlap q = i m2i /N must definitely be real, equation
(189) does not admit solutions for E > Eth and the self-overlap is not physically defined
above the threshold. This is indeed what we expected: it is natural to associate a state
to a minimum of the energy when we turn on the temperature, but it is not natural at
all to do the same with a saddle.
What is the self-overlap at the threshold energy ? For E = Eth we have
2 2
y = Eth /(p − 1)2 and thus from (190),
p−2 2
qth (1 − qth )2 = T 2 (192)
p(p − 1)
This equation gives the self-overlap of the threshold states as a function of the
temperature. We know that the PSM has a purely dynamic transition at Td , where
the correlation function, instead of decaying to zero, remains trapped for an infinite
time at a plateau, C(τ ) → qd , for τ → ∞. Our interpretation of this phenomenon was
that the dynamic configuration remains trapped in a region of the phase space of size
(self-overlap) qd . It is tempting to compare this value qd with the self-overlap of the
threshold states at Td : indeed these are the highest metastable states, so it is reasonable
to expect they are responsible for trapping the dynamics at Td . Recalling the definition
of Td , we have,
p(p − 2)p−2 (p − 2)p−2

Td2 = =⇒ 2
qth (1 − qth )2 = (193)
2(p − 1)p−1 (p − 1)p
This equation could easily be solved on a computer. However, we have a good guess
p−2
for the solution: if we plug into it qd = p−1 , we see that it is identically satisfied. The
important conclusion is that,
qth (Td ) = qd (194)
This result confirms all our expectations: the dynamical overlap at the transition
Td has an asymptotic limit equal to the self-overlap of the threshold states. These
states therefore are the ones trapping the dynamics, and forbidding it to relax to the
equilibrium values.
Another way for the q equation to stop having solution, is by increasing the
temperature, T > T ⋆ (E), at fixed bare energy E. This means that, even though minima
of the energy do not depend on the temperature, states, i.e. minima of the free energy,
do. When the temperature becomes too large, the paramagnetic states becomes the only
pure ergodic states, even though the energy landscape is broken up in many basins of the
energy minima. This is just one particularly evident demonstration of the fundamental
different between pure states and energy minima.
5.6. Life with many metastable states

We have computed the complexity Σ(E), a function of the bare energy E of the minima,
which does not depend on the temperature. All these minima become states when T 6= 0.
As we have seen, the bare energy E and the temperature T are the only variables we
need in order to compute the other properties at finite temperature. In particular the
free energy density of the states is a function f = f (E, T ).
We now ask what is the role (if any) of the complexity when computing the
equilibrium properties of the system. A first intuitive answer is that there must be
no role at all: after all, the complexity is zero for ground (stable) states, which rule the
equilibrium properties of the system, and it is only different from zero for metastable
states, which we expect to have no influence on equilibrium. In fact, it is not like that.
We must remember that the PSM is a mean-field model, where even metastable states
have an infinite lifetime, and contribute as stable states in partitioning the phase space
in ergodic sub-components. So, metastable states do play a role also in determining
equilibrium properties. We could expect, thus, that what follows is valid only for mean-
field systems. Strictly speaking, yes. However, in real system, like supercooled liquids
at low temperatures, many of the following observations apply as well, provided that
we pay great attention to the “states vs energy minima” issue. More precisely, in the
temperature regime where T is low enough so that the dynamics is activated, but high
enough so that the system is still ergodic and at equilibrium, the dynamics consists in
vibrations inside a potential energy minimum, with some rare jumps among minima. In
this regime, which is the one close to the Mode Coupling temperature, a phase space
decomposition as the one we are going to explain below is applicable [58].
Let us compute the equilibrium partition function Z of the system,
Z XZ X
Z = Dσ exp[−βH(σ)] = Dσ exp[−βH(σ)] = Zα (195)
α σ∈α α
where Zα is the partition function restricted to state α (stable or metastable it may be).
We have,
Zα = e−βN fα (196)
and thus, X
Z= e−βN fα (197)
a
In these formulas the free energy density of state α is fα = f (Eα , T ), where Eα is the
bare energy of state α. We want to pass from a sum over all states to an integral over
all bare energies,

XZ
Z= dE δ(E − Eα ) exp[−βNf (E, T )] =
α
Z
= dE N (E) exp[−βNf (E, T )] =
Z
= dE exp{−βN[f (E, T ) − T Σ(E)]} ≡
Z
≡ dE exp[−βNΦ(E, T )] (198)
where we have defined,

Φ(E, T ) ≡ f (E, T ) − T Σ(E) (199)
and where we have used the very definition of the number of states at energy E,
X
N (E) = δ(E − Eα ) (200)
α
In the equation above we can use the saddle-point method in the limit N → ∞: the
integral is concentrated on the value Eeq (T ) which minimize the exponent. The total
equilibrium free energy density Feq (T ) is therefore given by,
1
Feq (T ) = − logZ = min[f (E, T ) − T Σ(E)] = Φ(Eeq (T ), T ) (201)
βN E
with
∂Φ
(Eeq (T )) = 0 (202)
∂E
From its definition we see that Φ is clearly a sort of generalized free energy, with f
playing the role of the energy, and the complexity playing the role of the entropy,
f = E − T S with S entropy
(203)
Φ = f − T Σ with Σ complexity
and putting together these formulas we have,
Φ = E − T (S + Σ) (204)
so the complexity is the extra contribution to the total entropy due to the presence
of an exponentially large number of metastable states. We recall that E is the
finite temperature energy density of the states, which is different from the bare (zero-
temperature) energy E. In fact, E = E(E, T ) and E(E, 0) = E.
From what said above we see that the total equilibrium free energy density is found
by minimization with respect to E of the potential Φ(E, T ), in which the complexity
plays a major role. The bare energy density Eeq (T ) obtained minimizing Φ, fixes the
equilibrium states of the system. The free energy density of these equilibrium states
will then be feq = f (Eeq (T ), T ). What is a bit surprising is that,
Feq = f (Eeq , T ) − T Σ(Eeq ) < f (Eeq , T ) (205)

since the complexity is positive. In other words, the free energy density of equilibrium
states is larger than the global equilibrium free energy density. This funny thing is due to
the presence of an exponentially large number of metastable states: equilibrium is given
by an ensemble of states, each one with rather large free energy density feq , but whose
collective contribution to equilibrium is enhanced by their complexity, which lowers the
global free energy Feq .
This situation may seem paradoxical: equilibrium is given by a mixture of
metastable states, but each of them is surrounded by infinite free energy barriers, so
dynamically the system would not be able to exit from anyone of these states ! On the
one hand, this is just a particular way of breaking the ergodicity, which is of course
strictly valid only in mean-field. On the other hand, this situation makes much more
sense in finite dimensions, where these metastable states may trap the dynamics for a
time sufficiently long to allow us to define a complexity, but sufficiently short to make
the system ergodic. This may indeed be the situation in structural glasses close to the
glass transition [58, 59, 60].
5.7. Low temperatures, entropy crisis

The interval of definition of Φ(E, T ) is the same as Σ(E), that is E ∈ [E0 : Eth ].
Assuming that at a given temperature T the energy Eeq (T ) minimizing Φ lies in this
interval, what happens if we lower the temperature ? Remember that the complexity is
an increasing function of E, as of course is f (E, T ). When T decreases we favor states
with lower free energy and lower complexity, and therefore Eeq decreases. As a result,
it must exist a temperature T0 , such that,
Eeq (T0 ) = E0 (206)
and thus,
Σ(Eeq (T )) = Σ(E0 ) = 0 (207)
Below T0 the bare energy Eeq cannot decrease any further: there are no other states
below the ground states E0 . Thus, Eeq (T ) = E0 for each temperature T ≤ T0 . As a
result, if we plot the complexity of equilibrium states Σ(Eeq (T )) as a function of the
temperature, we find a discontinuity of the first derivative at T0 , where the complexity
vanishes.
A thermodynamic transition takes place at T0 : below this temperature equilibrium
is no longer dominated by metastable states, but by the lowest lying states, which have
zero complexity and lowest free energy density. The temperature T0 can be computed
by studying numerically equation (202). The following result should not be surprising
at this point,
T0 = Ts (208)
The temperature where equilibrium is given for the first time by the lowest energy
states, is equal to the static transition temperature. Above T0 the partition function is
dominated by an exponentially large number of states, each with high free energy and
thus low statistical weight, such that they are not captured by the overlap distribution
P (q). At T0 the number of these states becomes sub-exponential and their weight
nonzero, such that the P (q) develops a secondary peak at qs 6= 0.
In supercooled liquids, we can give an interesting interpretation of what is going
on. As we have seen the total entropy is the sum of the entropy S inside each state and
the complexity Σ. But in liquids the entropy of each energy minimum is (at low enough
temperatures) very similar to the entropy of the crystal SCR , while the total entropy is
just the entropy SLQ observed in the supercooled liquid phase. Thus we can write,
Σ(T ) = SLQ (T ) − SCR (T ) (209)
i.e. the complexity is the excess entropy of the liquid compared to the crystal. Therefore,
the temperature T0 is the temperature where the entropy of the liquid seems to become
equal to the temperature of the crystal, as first observed by Kauzmann in 1948 [61].
This scenario (vanishing complexity at T0 ) is normally known as entropy crisis.
Of course in real systems we cannot observe T0 , since it is far below the dynamical
glass transition, where the system falls out of equilibrium. The possible existence of T0
in real liquids relies on low temperature extrapolations of high temperature equilibrium
data.
5.8. High temperatures, the threshold

When we raise the temperature we privilege states with higher free energy and
complexity. Also in this case, thus, we must have a temperature beyond which we
exit from our range of definition of the complexity. Indeed, there is a temperature Tth ,
such that,
Eeq (Tth ) = Eth (210)
i.e. the bare energy density of equilibrium states becomes equal to the threshold energy
at Tth . What happens above Tth ? If we close our eyes and insist minimizing the potential
Φ, we see that the system would try to thermalize in an energy regime dominated by
unstable saddles, and not by minima. This fact suggests that the dynamics above Tth
is no longer trapped by minima, and that therefore it is ergodic. More precisely, we can
argue that while below Tth equilibrium is in fact given by a superposition of metastable
states with infinite barriers surrounding them, above Tth the system enters a phase
dominated no longer by minima, but by saddles [62, 63, 56]. A crucial result, which can
easily be proved, and which confirms this scenario, is the following,
Tth = Td (211)
The temperature Tth associated to the transition from minima to saddle (going up in
temperature) is thus the same as the temperature Td marking the passage from an
ergodic to nonergodic dynamics (going down in temperature). In the light of this, it
becomes clearer the interpretation of the dynamics for T > Td , but close to Td : the
landscape visited by the system in this phase is dominated by unstable saddle points,
which have however a very small number of negative modes, since T ∼ Td = Tth implies
E ∼ Eth . These objects cannot trap the dynamics for infinite times, but they can slow it
down. In particular, the finite, but very long plateau of the dynamic correlation function
C(τ ), can be interpreted as a pseudo-relaxation of the system into a saddle with very few
(order one) unstable modes [57]. At Td the bare energy is Eth and unstable saddles turn
into stable trapping minima. The plateau becomes infinite and ergodicity is broken.
The identification of Td with Tth , and its resulting interpretation, is a crucial point
in the physics of the PSM. It connects the dynamical and topological properties of the
system in a very general way and it suggests that even in different systems where a
glassy transition occurs, the topological properties of the underlying energy landscape
may be the responsible for the slowing down of the system [68]. We have seen in the
previous chapter that the dynamical equations of the PSM are just the Mode Couping
equations, strongly suggesting that systems well described by MCT close to the glass
transition, as fragile glasses, may have a dynamical behaviour similar to the PSM. As a
consequence, one can try to extend to fragile glasses the topological approach developed
in this chapter, which, as we have seen, is so closely related to dynamics. The cage
effect, which as we have seen cannot explain the plateau in a mean-field model, can thus
be reinterpreted in general as the effect of quasi-stable saddles probed by the system
close to the glass transition [57].
A final remark. Mode coupling theory predicts a sharp transition at Td , but this
cannot be strictly true out of mean-field, where barriers are finite. In fact, even in
fragile glasses, at Td one just observes a very steep crossover, but not a transition [40].
However, it may be that the underlying description of the landscape, in terms of minima-
to-saddle transition, is still valid [62]. In this way, a unique topological phenomenon
would be responsible for the dynamical transition in the mean-field PSM, and of the
sharp crossover in finite-dimensional fragile glasses [64, 66, 65, 67].
The complexity has at last unified all our results. The two transition temperatures
are nothing else than the manifestation, at the static and dynamical level, of the lower
and upper edges of the band of metastable states.
6. Conclusions
It was long enough, so let us be brief in these conclusions. We have seen that the
PSM has two transitions. There is a thermodynamic transition at a temperature Ts ,
where the free energy switches from a paramagnetic state, to many spin-glass states.
Moreover, Ts is also the temperature where the complexity of equilibrium states vanishes.
What happens at Ts is a perfect realization of the entropy crisis scenario described by
Kauzmann for supercooled liquids. Below Ts equilibrium is given by a non-exponential
number of lowest free energy states, which are detected and described by a standard
thermodynamic approach. Above Ts an exponentially large number of metastable states
dominates the partition function, due to their nonzero complexity. In this phase, we
have the funny result that the free energy density of equilibrium states is larger than
the global equilibrium free energy density. Thermodynamics is totally unaware of these
states, and predicts a trivial paramagnetic state, with P (q) = δ(q).
At higher temperatures we have a purely dynamic transition Td > Ts . When we
arrive at this temperature coming from above, the dynamics gets trapped by metastable
states, and the correlation time diverges. The equations describing such dynamical
behavior are the same as the MCT equations for supercooled liquids. This suggests
that what happens in the PSM at Td is similar to what happens in real glasses close
to the MCT temperature. The crucial difference, of course, is that in the PSM there
can be no barrier crossing, since barriers are infinite, while in real glasses activation is
present. On the other hand, the fact that standard MCT predicts a sharp dynamical
transition at Td seems to suggest that this theory too, as the PSM, does not account for
activated events.
We have finally seen that there is a close relationship between the topological
properties of the model and its dynamical behavior. In particular, the slowing down
of the dynamics above but close to Td is connected to the presence of saddles, whose
instability decreases with decreasing energy. In fact, we have seen that the threshold
energy level Eth separating saddles from minima, can be associated to the temperature
Tth = Td , marking the passage from ergodicity to ergodicity breaking. In this context
the dynamical transition can be seen as a topological transition. The plateau of the
dynamical correlation function, which has an interpretation in terms of cage effect in
liquids, may be reinterpreted as a pseudo-thermalization inside a saddle with a very
small number of unstable modes.
A very final warning. We should never forget that the PSM is a mean-field model,
with no spatial structure at all. As a consequence, all physical modelizations and
interpretations coming from the PSM, inevitably have a mean-field flavor. In particular,
this is true for the topological interpretation of the dynamical transition: no fluctuations
are taken into account, not to mention spatial heterogeneities, which may play a very
important role. However, the arguments we gave in terms of phase space and topological
concepts have at least the virtue of being simple and effective. If not pushed too far,
they provide a nice tool to understand in a unifying way the physics of glassy systems.
[1] J. Cardy, Scaling and Renormalization in Statistical Physics, Cambridge University Press,
Cambridge (1996).
[2] J. J. Binney, N. J. Dowrick, A. J. Fisher and M. E. J. Newman, The modern theory of critical
phenomena, Clarendon Press (1992).
[3] J. Hertz, Spin Glass Physics, http://www.nordita.dk∼hertz/nssg.html (1998).
[4] M. Mézard, G. Parisi and M. A. Virasoro, Spin Glass Theory and Beyond, World Scientific,
Singapore (1987).
[5] S. F. Edwards and P. W. Anderson, J. Phys. F5 965 (1975).
[6] D. Ruelle, Statistical Mechanics, Benjamin, New York (1969).
[7] G. Parisi, Statistical Field Theory, Addison Wesley, New York, (1988).
[8] M. Mézard, G. Parisi, N. Sourlas, G. Toulouse and M.A. Virasoro, J. Phys. 45 843 (1984).
[9] B. Derrida, Phys. Rev. Lett. 45 79 (1980); B. Derrida, Phus. Rev. B 24 2613 (1981).
[10] A. Crisanti and H.-J. Sommers, Z. Phys. B 87, 341 (1992).
[11] C.M. Bender and S.A. Orszag, Advanced Mathematical Methods for Scientist and Engineers,
McGraw-Hill International Editions, Singapore (1978).
[12] J. R. L. De Almeida and D. J. Thouless, J. Phys. A11 983 (1978).
[13] D. Sherrington and S. Kirkpatrick, Phys. Rev. Lett. 35 1792 (1975).
[14] G. Parisi, Phys. Rev. Lett. 50, 1946 (1983).
[15] C. De Dominicis and P. Young, J. Phys. A 16 2063 (1983).
[16] C. De Gaulle, see, for example, http://oferriere.free.fr/Ilsontdit/ilsontditg.htm.
[17] G. Parisi, Phys. Lett. 73A 203 (1979).
[18] G. Parisi, J. Phys. A 13 L115 (1980).
[19] G. Parisi, J. Phys. A 13 1101 (1980).
[20] R. Zwanzig, Nonequilibrium Statistical Mechanics, Oxford University Press (2001).
[21] C.W. Gardiner, Handbook of Stochastic Methods, Springer (1985).
[22] N.G. van Kampen, J. Stat. Phys. 24, 175 (1981).
[23] J. Zinn-Justin, Quantum Field Theory and Critical Phenomena, Clarendon Press, Oxford Third
Edition (1996).
[24] C. P. Martin, E. Siggia and H. A. Rose, Phys. Rev. A 8, 423 (1973).
[25] H.K Janssen, Z. Phys. B 23, 377 (1976).
[26] C. De Dominicis, J. Phys. (Paris) C 1, 247 (1976).
[27] C. De Dominicis, Phys. Rev. B 18, 4913 (1978).
[28] C. De Dominicis and L. Peliti, Phys. Rev. B 18, 353 (1978).
[29] T.R. Kirkpatrick and D. Thirumalai, Phys. Rev. B 36, 5388 (1987).
[30] A. Crisanti, H. Horner and H.-J. Sommers, Z. Phys B 92, 257 (1993).
[31] H. Sompolinsky and A. Zippelius, Phys. Rev. B25 6860 (1982).
[32] U. Bengtzelius, W. Götze and A. Sjølander, J. Phys. Chem. 17, 5915 (1984).
[33] E. Leutheusser, Phys. Rev. A 29 2765 (1984).
[34] For a review of MCT see: W. Götze, and L. Sjøgren, Rep. Prog. Phys. 55, 241 (1992).
[35] D. Reichman, UCGPIII notes on “Mode Coupling Theory in glassy systems” (2005).
[36] T. R. Kirkpatrick, D. Thirumalai and P. G. Wolynes, Phys. Rev. A 40, 1045 (1989); D. Thirumalai
and T. R. Kirkpatrick, Phys. Rev. B 38, 4881 (1988); D. Thirumalai and T. R. Kirkpatrick, J.
Phys. A 22, L149 (1989).
[37] We haveR essentially used the fact that in the asymptotic limit τ ≫ 1, where C(τ ) ≪ 1, we can
τ
write 0 du C p−1 (τ − u)Ċ(u) = C p−1 (τ )[C(τ ) − 1] + αĊ(τ ), where α is a positive term.
[38] J.-P. Bouchaud, L. Cugliandolo, J, Kurchan and M. Mézard, Physica A 226, 243 (1996).
[39] L. Cugliandolo, preprint cond-mat/0210312 (2002).
[40] C. A. Angell, J. Phys. Chem. Sol. 49, 863 (1988).
[41] For a series of thorough numerical tests of MCT, see: W. Kob and H. C. Andersen, Phys. Rev.
Lett. 73, 1376 (1994); Phys. Rev. E 51, 4626 (1995); Phys. Rev. E 52, 4134 (1995).
[42] L. F. Cugliandolo and J. Kurchan, Phys. Rev. Lett. 71 173 (1993).
[43] S. Franz and M. Mezard, Physica A 210, 48 (1994).

[44] G. Biroli, UCGPIII notes on “Aging in glassy sytems” (2005).
[45] D. J. Thouless, P. W. Anderson and R. G. Palmer, Phil. Mag. 35 593 (1977).
[46] H. Rieger, Phys. Rev. B 46 14655 (1992).
[47] J. Kurchan, G. Parisi and M.A. Virasoro, J. Phys. I France 3 1819 (1993).
[48] A. Barrat, cond-mat/9701031 (1997).
[49] For a discussion of the meaning of the reaction term see [4] and [46].
[50] A. J. Bray and M. A. Moore J. Phys C13 469 (1980).
[51] J. Kurchan, J. Phys. A. 24 4969 (1991).
[52] A. Cavagna, I. Giardina and G. Parisi, Phys. Rev. B, 57 11251 (1998).
[53] A. Crisanti and H.-J. Sommers, J. Phys. I (France) 5, 805 (1995).
[54] A. Annibale, A. Cavagna, I. Giardina, and G. Parisi, Phys. Rev. E 68, 061103 (2003).
[55] A. Cavagna, J.P. Garrahan, and I. Giardina, Phys. Rev. B 61, 3960 (2000).
[56] A. Cavagna, I. Giardina and G. Parisi, J. Phys. A: Math. Gen. 34 5317 (2001).
[57] A. Cavagna, I. Giardina, and T. Grigera. J. Phys. A 36, 10721 (2003).
[58] M. Goldstein, J. Chem. Phys 51, 3728 (1969).
[59] F.H. Stillinger and T.A. Weber, Phys. Rev. A 25, 978 (1982).
[60] F. Sciortino, UCGPIII notes on “Energy landscape approach in glassy systems” (2005).
[61] A.W. Kauzmann, Chem. Rev. 43, 219 (1948).
[62] A. Cavagna, preprint cond-mat/9910244-v1 (1999); Europhys. Lett. 53, 490 (2001).
[63] S. Franz, and M.A. Virasoro, J. Phys. A 33, 891 (2000).
[64] K. Broderix, K.K. Bhattacharya, A. Cavagna, A. Zippelius, and I. Giardina, Phys. Rev. Lett. 85,
5360 (2000).
[65] T.S. Grigera, A. Cavagna, I. Giardina, and G. Parisi, Phys. Rev. Lett. 88, 055502 (2002).
[66] L. Angelani, R. Di Leonardo, G. Ruocco, A. Scala, and F. Sciortino, Phys. Rev. Lett. 85, 5356
(2000).
[67] An acute analysis of the numerical drawbacks of the saddle approach is given in J.P.K. Doye, and
D.J. Wales, J. Chem. Phys. 116, 3777 (2002).
[68] A very nice account of the general connections between topology and dynamics in high dimensional
systems can be found in J. Kurchan and L. Laloux, J. Phys. A 29 1929 (1996).

Spin-Glass Theory For Pedestrians: Tommaso Castellani and Andrea Cavagna

Uploaded by

Copyright:

Available Formats

Spin-Glass Theory For Pedestrians: Tommaso Castellani and Andrea Cavagna

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Spin-Glass Theory For Pedestrians: Tommaso Castellani and Andrea Cavagna

Uploaded by

Copyright:

Available Formats

arXiv:cond-mat/0505032v1 [cond-mat.

dis-nn] 2 May 2005

Spin-Glass Theory for Pedestrians

Before we start studying a specific spin-glass model, we need to introduce a couple of

2.2. Self-averaging quantities

lim FN (β, J) = F∞ (β) . (5)

2.3. Annealed and quenched averages

2.4. The replica trick

which is in fact much simpler to compute. What we do is to replicate the system n

y 1/2 = lim y n = lim x · x · · · · · x · x = lim x2n = x . (14)

2.5. Pure states

where we have defined, Z

that is the partition function restricted to state α, and

hσi σj i → hσi ihσj i for |i − j| → ∞ . (21)

2.6. Overlap, self-overlap

With Ising spins si = ±1 we have that,

2.7. Overlap distribution

and using the clustering property we finally obtain,

two pure states, so we have in principle four possible overlaps,

3.1. The p-spin spherical model

3.2. First try: the replica symmetric calculation

and finally we use an exponential representation for the δ-function, to obtain,

Qab = q0 + (1 − q0 )δab . (56)

and thus (55) becomes, in the limit n → 0,

3.3. The key connection between replicas and physics

P (q) = δ(q − q0 ) , (69)

3.4. Replica symmetry breaking

P (q) = (1 − m) δ(q − q1 ) + m δ(q − q0 ) (73)

3.5. The 1RSB solution and the static transition

The first piece is easy to compute and gives in the limit n → 0,

to be compared with the replica symmetric (RS) free energy,

4.1. The generating functional formalism

hη(t)η(t′)i = 2T δ(t − t′ ) (87)

where we have defined the probability of x as,

L(x) = L0 (x) + LJ (x) = η(t) (100)

In order to be averaged over J this expression must be rewritten in terms of replicas,

We define the quantity ∆(x, x̂) as

exp[∆(x, x̂)] ≡ exp[ix̂LJ (x)] (108)

L0 (x) + L1 (x, x̂) = ξ with hξξi = D0 + D1 (x, x̂) (110)

4.2. Dynamics of p-spin spherical model

The generating functional is given by equation (103), with x → σk , x̂ → σ̂k , and,

As we anticipated, in the calculation appeared a coupling among different times, through

We now give an exponential representations of the δ-functions. For example,

4.3. Equations for the correlation and the response

The second relation is the following,

4.4. Diagrammatic technique and Mode Coupling approximation

Figure 1. Diagrammatic representation of the perturbative solution to equation 130.

Figure 2. Diagrammatic representations of the perturbative expansions of the

by their renormalized values, R0 → R, C0 → C. In this way we get the following

4.5. The dynamical transition

singularity. Clearly, metastability, combined with the peculiar features of mean-field,

5.1. What is the TAP free energy ?

5.2. Definition of complexity (and a problem with the modulus).

The quantity Σ is called complexity in the spin-glass community, and configurational

By using the standard formula,

5.3. The calculation of the complexity.

and thus we have,

In order to average over J we have to understand whether the self-averaging quantity