L-4: Difference between revisions

From Disordered Systems Wiki
Jump to navigation Jump to search
 
(57 intermediate revisions by the same user not shown)
Line 1: Line 1:
<Strong> Goal </Strong>: final lecture on KPZ and directed polymers at finite dimension. We will show that for <math>d>2</math> a "glass transition" takes place.




= KPZ : from 1d to the Cayley tree=


We know a lot about KPZ, but there is still much to understand: 
= Directed Polymer in finite dimension =


* In <math>d=1</math>, we have found <math>\theta=1/3</math> and a glassy regime present at all temperatures. The stationary solution of the KPZ equation describes, at long times, the fluctuations of quantities such as <math>E_{\min}[x] - E_{\min}[x']</math>. However, it does not determine the actual distribution of <math>E_{\min}</math> for a given <math>x</math>. In particular, we have no clear understanding of the origin of the Tracy-Widom distribution. 
== State of the Art ==


* In <math>d=\infty</math>, an exact solution exists for the Cayley tree, predicting a freezing transition to a 1RSB phase (<math>\theta=0</math>).
The directed polymer in random media belongs to the KPZ universality class. The behavior of this system is well understood in one dimension and in the mean-field case, more precisely for the directed polymer on the Cayley tree. In particular:


* In finite dimensions greater than one, no exact solutions are available. Numerical simulations indicate <math>\theta > 0</math> in <math>d=2</math>. The case <math>d > 2</math> remains particularly intriguing.
* In <math>d=1</math>, we have <math>\theta=1/3</math> and a glassy regime present at all temperatures. The model is integrable through a non-standard Bethe Ansatz, and the distribution of <math>E_{\min}</math> for a given boundary condition is of the Tracy–Widom type.


==Let's do replica!==
* In <math>d=\infty</math>, for the Cayley tree, an exact solution exists, predicting a freezing transition to a 1RSB phase (<math>\theta=0</math>).
To make progress in disordered systems, we need to analyze the moments of the partition function.   From Valentina's lecture, recall that if 


<center>
In finite dimensions greater than one, no exact solutions are available. Numerical simulations indicate <math>\theta > 0</math> in <math>d=2</math> and a glassy regime present at all temperatures. The case <math>d > 2</math> remains particularly intriguing.
<math>
\frac{\overline{Z(x,t)^2}}{ (\overline{Z(x,t)})^2}=1 
</math>
</center>


then the partition function is self-averaging, and 


==Let's do replica!==
To make progress  in disordered systems, we need to analyze the moments of the partition function.  The first moment provide the annealed average and the second moment tell us about the fluctuantions. In particular, the partition function is self-averaging  if
<center>   
<center>   
<math>   
<math>   
\overline{\ln Z(x,t)} =\ln\overline{Z(x,t)}   
\frac{\overline{Z(x,t)^2}}{ (\overline{Z(x,t)})^2}=1 \, .
</math>.  
</math>   
</center>   
</center>   
 
In this case annealed and the quenched average coincides in the thermodynamic limit. This strict  condition is sufficient, but not necessary. What is necessary is to show that  for large ''t''
The condition above is sufficient but not necessary. It is enough that   
 
<center>   
<center>   
<math>   
<math>   
Line 36: Line 28:
</math>,   
</math>,   
</center>   
</center>   
In the following, we compute these moments via a replica calculation, considering polymers starting at <math>0</math> and ending at <math>x</math>.
To proceed, we only need two ingredients:
* The random potential <math>V(x,\tau)</math> is a Gaussian field characterized by
<center> <math> \overline{V(x,\tau)} = 0, \qquad \overline{V(x,\tau) V(x',\tau')} = D \, \delta^d(x-x') \, \delta(\tau - \tau'). </math> </center>
* Since the disorder is Gaussian, averages of exponentials can be computed using Wick’s theorem:
<center> <math> \overline{\exp(W)} = \exp\!\Big[\overline{W} + \frac{1}{2}\big(\overline{W^2} - \overline{W}^2\big)\Big], </math> </center>
for any Gaussian random variable <math>W</math>.
These two properties are all we need to carry out the replica calculation below.
==First Moment==
<center> <math> \overline{Z(x,t)} = \int_{x(0)=0}^{x(t)=x} \mathcal{D}x(\tau) \exp\Big[-\frac{1}{T}\int_0^t d\tau \frac{1}{2}(\partial_\tau x)^2\Big] \overline{\exp\Big[-\frac{1}{T} \int_0^t d\tau V(x(\tau),\tau)\Big]} </math> </center>
Due to the short-distance divergence of <math>\delta^d(0)</math>,
<center> <math> T^2 \overline{W^2} = \int d\tau_1 d\tau_2 \overline{V(x,\tau_1)V(x,\tau_2)} = D t \delta_0. </math> </center>
Hence,
<center> <math> \overline{Z(x,t)} = \frac{1}{(2\pi t T)^{d/2}} \exp\Big[-\frac{x^2}{2 t T}\Big] \exp\Big[\frac{D t \delta_0}{2 T^2}\Big] = Z_{free}(x,t,T)  \exp\Big[\frac{D t \delta_0}{2 T^2}\Big].  </math> </center>
==Second Moment==
For the second moment we need two replicas:
* Step 1
<center> <math> \overline{Z(x,t)^2} = \int \mathcal{D}x_1 \int \mathcal{D}x_2 \exp\!\Bigg[-\frac{1}{2T}\int_0^t d\tau \Big((\partial_\tau x_1)^2 + (\partial_\tau x_2)^2\Big)\Bigg] \; \overline{\exp\!\Bigg[-\frac{1}{T} \int_0^t d\tau V(x_1(\tau),\tau) - \frac{1}{T} \int_0^t d\tau V(x_2(\tau),\tau)\Bigg]}. </math> </center>
* Step 2: Wick’s Theorem
<center> <math> \overline{Z(x,t)^2} = \exp\!\Bigg[\frac{D t \delta_0}{T^2}\Bigg] \int \mathcal{D}x_1 \int \mathcal{D}x_2 \exp\!\Bigg[-\frac{1}{2T}\int_0^t d\tau \Big((\partial_\tau x_1)^2 + (\partial_\tau x_2)^2 - \frac{D}{T^2}\delta^d[x_1(\tau)-x_2(\tau)]\Big)\Bigg]. </math> </center>
* Step 3: Change of Coordinates
Let <math>X = (x_1+x_2)/2</math> and <math>u = x_1 - x_2</math>. Then:
<center> <math> \overline{Z(x,t)^2} = (\overline{Z(x,t)})^2 \frac{\displaystyle \int_{u(0)=0}^{u(t)=0} \mathcal{D}u \exp\!\Bigg[-\int_0^t d\tau \frac{1}{4T} (\partial_\tau u)^2 - \frac{D}{T^2} \delta^d[u(\tau)]\Bigg]} {Z_{free}(u=0,t,2T)}. </math> </center>
Here,
<center> <math> Z_{free}^2(x,t,T) = Z_{free}(X=x,t,T/2) \, Z_{free}(u=0,t,2T), \qquad Z_{free}(u=0,t,2T) = (4 \pi T t)^{d/2}. </math> </center>
=== Two-Replica Propagator ===
Define the propagator:
<center> <math> W(0,t) = \int_{u(0)=0}^{u(t)=0} \mathcal{D}u \exp\Big[-\int_0^t d\tau \frac{1}{4T} (\partial_\tau u)^2 - \frac{D}{T^2} \delta^d[u(\tau)]\Big]. </math> </center>
By the Feynman-Kac formula:
<center> <math> \partial_t W(x,t) = -\hat H W(x,t), \quad \hat H = -T \nabla^2 - \frac{D}{T^2} \delta^d[u]. </math> </center>
The single-particle potential is time-independent and attractive. Long-time behavior is governed by the low-energy eigenstates.


when <math>t \to \infty</math>, to ensure the equivalence between annealed and quenched averages. 


In the following, we compute this quantity, which corresponds to a two-replica calculation.  For simplicity, we consider polymers starting at <math>0</math> and ending at <math>x</math>. We recall that: 
For <math>d \le 2</math>, the attractive potential always produces a bound state with energy <math>E_0<0</math>. Hence, at long times:
<center> <math> W(x,t) \sim e^{|E_0| t}  </math> </center>
This explosion means that the quenched free energy is smaller than the annealed one at all temperatures.


* <math>V(x,\tau)</math> is a Gaussian field with  
For <math>d > 2</math>, The low-energy behavior depends on <math>D/T^2</math>:
<center> <math>
* High temperature: the spectrum is positive and continuous. Annealed and quenched coincide, the exponent <math>\theta=0</math>.
\overline{V(x,\tau)}=0, \quad  \overline{V(x,\tau) V(x',\tau')} = D \delta^d(x-x') \delta(\tau-\tau')
</math></center> 


* From Wick's theorem, for a generic Gaussian field <math> W </math>, we have 
* Low temperature: bound states appear.  No replica-symmetry breaking (RSB), but the quenched free energy is smaller than the annealed one. Numerical simulations show <math>\theta>0</math>.
<center><math>
=Back to REM: condensation of the Gibbs measure=
\overline{\exp(W)} = \exp\left[\overline{W} +\frac{1}{2} \left(\overline{W^2}-\overline{W}^2\right)\right] 
</math></center> 


===The first moment===
The first moment of the partition function is straightforward to compute and corresponds to a single replica: 


Thanks to the computation of <math>\overline{n(x)}</math>, we can identify an important fingerprint of the glassy phase.  Let's compare the weight of the ground state against the weight of all other states:
<center>   
<center>   
<math>
<math>  
\overline{Z(x,t) } =\int_{x(0)=0}^{x(t)=x} {\cal D} x(\tau) \exp\left[- \frac{1}{T} \int_0^t d \tau \frac{1}{2}(\partial_\tau x)^2\right]  \overline{\exp\left[- \frac{1}{T} \int d \tau V(x(\tau),\tau ) \right]}
\frac{\sum_\alpha z_\alpha}{z_{\alpha_{\min}}} = 1 + \sum_{\alpha \ne \alpha_{\min}} \frac{z_\alpha}{z_{\alpha_{\min}}} = 1 + \sum_{\alpha \ne \alpha_{\min}} e^{-\beta(E_\alpha - E_\min)} \sim 1 + \int_0^\infty dx\, \frac{d\overline{n(x)}}{dx} \, e^{-\beta x}= 1+ \int_0^\infty dx\, \frac{e^{x/b_M}}{b_M} \, e^{-\beta x}
</math>   
</math>
</center>  
</center>
 
=== Behavior in Different Phases:===
* '''High-Temperature Phase (<math> \beta <  \beta_c = 1/b_M = \sqrt{2 \log2}</math>):'''  
: In this regime, the total weight of the excited states dominates over the weight of the ground state. The ground state is therefore not deep enough to overcome the finite entropy contribution. As a result, the probability of sampling the same configuration twice from the Gibbs measure is exponentially small in the system size.


Note that the term <math> T^2 \overline{W^2} = \int d \tau_1 d\tau_2 \overline{V(x,\tau_1)V(x,\tau_2)}= D t \delta_0</math> exhibits a short-distance divergence due to the delta function. Hence, we can write: 


* '''Low-Temperature Phase (<math> \beta > \beta_c =1/ b_M = \sqrt{2 \log2}</math>):''' 
: In this regime, the integral is finite: 
<center>   
<center>   
<math>
<math>
\overline{Z(x,t) } = \frac{1}{(2 \pi t T)^{d/2}}\exp\left[ -\frac{1}{2} \frac{ x^2}{t T} \right]  \exp\left[ \frac{D  t \delta_0}{2T^2} \right] 
\int_0^\infty dx \, e^{ (1/b_M-\beta) x}/b_M = \frac{1}{\beta b_M-1} = \frac{\beta_c}{\beta - \beta_c}
</math>   
</math>   
</center>   
</center>   
In this regime, the total weight of the excited states is of the same order as the weight of the ground state. Consequently, the ground state is occupied with finite probability, reminiscent of Bose–Einstein condensation. However, unlike the directed polymer in finite dimension, this condensation involves not only the ground state but also the first excited states.
====Overlap Distribution and Replica Symmetry Breaking:====
The structure of states can be further characterized through the overlap between two configurations <math>\alpha</math> and <math>\gamma</math>, defined as


=== The second moment === 
<center> <math> q_{\alpha,\gamma} = \frac{1}{N} \sum_{i=1}^N \sigma_i^\alpha \sigma_i^\gamma, </math> </center>


For the second moment, there are two replicas:
which takes values in the interval <math>(-1,1)</math>. The distribution <math>P(q)</math> of the overlap between two configurations sampled from the Gibbs measure distinguishes the two phases:


* Step 1: The second moment is 
At high temperature (<math>\beta < \beta_c</math>), the system is replica symmetric and the overlap distribution is concentrated at zero:


<center>
<center> <math>P(q) = \delta(q).</math> </center>
<math>
 
\overline{Z(x,t)^2 } =\int {\cal D} x_1\int  {\cal D} x_2 \exp\left[- \int_0^t d \tau  \frac{1}{2T}[(\partial_\tau x_1)^2+ (\partial_\tau x_2)^2] \right]  \overline{\exp\left[- \frac{1}{T} \int_0^t d \tau_1 V(x_1(\tau_1),\tau_1 ) - \frac{1}{T} \int_0^t d \tau_2 V(x_2(\tau_2),\tau_2 )\right]} 
At low temperature (<math>\beta > \beta_c</math>), the system exhibits one-step replica symmetry breaking, and the overlap distribution becomes
</math>
 
</center>
<center> <math>P(q) = \tfrac{\beta_c}{\beta}\,\delta(q) + \Bigl(1 - \tfrac{\beta_c}{\beta}\Bigr)\,\delta(1-q).</math> </center>
 
 
== Finite Dimensional Systems==
In finite dimensions, the fluctuations of the ground-state energy are characterized by an exponent <math>\theta</math>:
 
<center> <math>\overline{\big(E_{\min} - \overline{E_{\min}}\big)^2} \sim L^{2\theta},</math> </center>


* Step 2: Using Wick's theorem, we obtain
where <math>L</math> is the linear size of the system and <math>N = L^D</math> is the number of degrees of freedom.


<center> 
When <math>\theta < 0</math>, the critical temperature vanishes with increasing system size, leading to the absence of a glass transition. This scenario occurs in many low-dimensional systems, such as the Edwards–Anderson model in two dimensions.
<math>
\overline{Z(x,t)^2 } = \exp\left[ \frac{D  t \delta_0}{T^2}  \right]\int {\cal D} x_1\int  {\cal D} x_2 \exp\left[-  \int_0^t d \tau  \frac{1}{2T}[(\partial_\tau x_1)^2+ (\partial_\tau x_2)^2 - \frac{D}{T^2} \delta^d[x_1(\tau)-x_2(\tau)]\right] 
</math>
</center> 


* Step 3: Changing coordinates** <math>X=(x_1+x_2)/2; \; u=x_1-x_2</math>, we get 
When <math>\theta > 0</math>, one must extend the definition of this exponent to finite temperatures and consider the fluctuations of the free energy <math>F(L,\beta)</math>. Three representative cases are:


<center>
* '''Directed polymer in <math>N=1,2</math>:'''
<math>
The fluctuations of the ground state exhibit a positive, temperature-independent exponent <math>\theta</math>. In this situation, only the glassy phase exists, and
\overline{Z(x,t)^2} = (\overline{Z(x,t)})^2 \frac{\int_{u(0)=0}^{u(t)=0} {\cal D} u  \exp\left[-  \int_0^t d \tau  \frac{1}{4T}(\partial_\tau u)^2- \frac{D}{T^2} \delta^d[u(\tau)]\right]}{\int_{u(0)=0}^{u(t)=0} {\cal D} u  \exp\left[-  \int_0^t d \tau  \frac{1}{4T}(\partial_\tau u)^2\right]} 
</math>
</center>


==Discussion==
<center><math>P(q) = \delta(1-q),</math></center> because producing an excitation with vanishing overlap with the ground state is very costly.
Hence, the quantity <math>\overline{Z(x,t)^2}/ (\overline{Z(x,t)})^2</math> can be computed.
* The denominator  <math>\int_{u(0)=0}^{u(t)=0} {\cal D} u  \exp\left[-  \int_0^t d \tau  \frac{1}{4T}(\partial_\tau u)^2\right]  </math> is the free propagator and gives a contribution <math> \sim (4 T  t)^{d/2}</math> .
* Let us define  the numerator
<center> <math>
W(0,t)= \int_{u(0)=0}^{u(t)=0} {\cal D} u  \exp\left[-  \int_0^t d \tau  \frac{1}{4T}(\partial_\tau u)^2- \frac{D}{T^2} \delta^d[u(\tau)]\right]
</math></center>


* '''Directed polymer in <math>N=3</math>:'''
The exponent <math>\theta</math> depends on the temperature: it vanishes above the glass transition and becomes strictly positive below it. Accordingly,


<center><math>P(q) = \delta(1-q)</math> at low temperature, and <math>P(q) = \delta(q)</math> at high temperature.</center>


<Strong>Remark 2:</Strong> From Feynman-Kac we can write the following equation
* '''Directed polymer on the Cayley tree:'''
<center> <math>
The behavior is analogous to the Random Energy Model: <math>\theta = 0</math> in both phases. At high temperature,
\partial_t W(x,t) =-  \hat H W(x,t)
</math></center>
Here the Hamiltonian reads:
<center> <math>
\hat  H= -2 T \nabla^2 - \frac{D}{T^2} \delta^d[u]
</math></center>
The single particle  potential  is <Strong> time independent and actractive </Strong>.
<center> <math>
W(x,t) = \langle x|\exp\left( - \hat H t\right) |0\rangle
</math></center>
At large times the behaviour is dominatated by the low energy part of the spectrum.


* In <math> d\le 2</math> an actractive potential always gives a bound state. In particular the ground state has a negative energy <math> E_0 <0</math>. Hence at large times
<center><math>P(q) = \delta(q),</math></center> while at low temperature the system exhibits the one-step replica symmetry breaking picture:
<center> <math>
W(x,t) = e^{ |E_0| t}
</math></center>
grows exponentially. This means that at all temperature, when  <math>  t\to \infty</math>
<center><math> \overline{\ln Z(x,t)}  \ll \ln\overline{Z(x,t)}
</math></center>


* For <math> d > 2</math> the low part of the spectrum is controlled by the strength of the prefactor <math>\frac{D}{T^2} </math>. At high temperature we have a continuum positive spectrum, at low temperature bound states exist. Hence,  when  <math>  t\to \infty</math>
<center> <math>P(q) = \tfrac{\beta_c}{\beta}\,\delta(q) + \Bigl(1 - \tfrac{\beta_c}{\beta}\Bigr)\,\delta(1-q).</math> </center>
<center><math> \begin{cases}
\overline{\ln Z(x,t)} = \ln\overline{Z(x,t)} \quad \text{for} \; T>T_c \\
\\
\overline{\ln Z(x,t)} \ll \ln\overline{Z(x,t)} \quad  \text{for} \; T<T_c
\end{cases}
</math></center>
This transition, in <math> d =3 </math>, is between a high temeprature, <math> \theta=0</math> phase and a low temeprature <math> \theta>0</math> <Strong> no RSB </Strong> phase.

Latest revision as of 20:14, 16 September 2025


Directed Polymer in finite dimension

State of the Art

The directed polymer in random media belongs to the KPZ universality class. The behavior of this system is well understood in one dimension and in the mean-field case, more precisely for the directed polymer on the Cayley tree. In particular:

  • In , we have and a glassy regime present at all temperatures. The model is integrable through a non-standard Bethe Ansatz, and the distribution of for a given boundary condition is of the Tracy–Widom type.
  • In , for the Cayley tree, an exact solution exists, predicting a freezing transition to a 1RSB phase ().

In finite dimensions greater than one, no exact solutions are available. Numerical simulations indicate in and a glassy regime present at all temperatures. The case remains particularly intriguing.


Let's do replica!

To make progress in disordered systems, we need to analyze the moments of the partition function. The first moment provide the annealed average and the second moment tell us about the fluctuantions. In particular, the partition function is self-averaging if

In this case annealed and the quenched average coincides in the thermodynamic limit. This strict condition is sufficient, but not necessary. What is necessary is to show that for large t

,

In the following, we compute these moments via a replica calculation, considering polymers starting at and ending at .

To proceed, we only need two ingredients:

  • The random potential is a Gaussian field characterized by
  • Since the disorder is Gaussian, averages of exponentials can be computed using Wick’s theorem:

for any Gaussian random variable .

These two properties are all we need to carry out the replica calculation below.

First Moment

Due to the short-distance divergence of ,

Hence,

Second Moment

For the second moment we need two replicas:

  • Step 1
  • Step 2: Wick’s Theorem
  • Step 3: Change of Coordinates

Let and . Then:

Here,

Two-Replica Propagator

Define the propagator:

By the Feynman-Kac formula:

Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \partial_t W(x,t) = -\hat H W(x,t), \quad \hat H = -T \nabla^2 - \frac{D}{T^2} \delta^d[u]. }

The single-particle potential is time-independent and attractive. Long-time behavior is governed by the low-energy eigenstates.


For Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle d \le 2} , the attractive potential always produces a bound state with energy Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle E_0<0} . Hence, at long times:

Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle W(x,t) \sim e^{|E_0| t} }

This explosion means that the quenched free energy is smaller than the annealed one at all temperatures.

For Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle d > 2} , The low-energy behavior depends on Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle D/T^2} :

  • High temperature: the spectrum is positive and continuous. Annealed and quenched coincide, the exponent Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \theta=0} .
  • Low temperature: bound states appear. No replica-symmetry breaking (RSB), but the quenched free energy is smaller than the annealed one. Numerical simulations show Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \theta>0} .

Back to REM: condensation of the Gibbs measure

Thanks to the computation of Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \overline{n(x)}} , we can identify an important fingerprint of the glassy phase. Let's compare the weight of the ground state against the weight of all other states:

Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \frac{\sum_\alpha z_\alpha}{z_{\alpha_{\min}}} = 1 + \sum_{\alpha \ne \alpha_{\min}} \frac{z_\alpha}{z_{\alpha_{\min}}} = 1 + \sum_{\alpha \ne \alpha_{\min}} e^{-\beta(E_\alpha - E_\min)} \sim 1 + \int_0^\infty dx\, \frac{d\overline{n(x)}}{dx} \, e^{-\beta x}= 1+ \int_0^\infty dx\, \frac{e^{x/b_M}}{b_M} \, e^{-\beta x} }

Behavior in Different Phases:

  • High-Temperature Phase (Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \beta < \beta_c = 1/b_M = \sqrt{2 \log2}} ):
In this regime, the total weight of the excited states dominates over the weight of the ground state. The ground state is therefore not deep enough to overcome the finite entropy contribution. As a result, the probability of sampling the same configuration twice from the Gibbs measure is exponentially small in the system size.


  • Low-Temperature Phase (Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \beta > \beta_c =1/ b_M = \sqrt{2 \log2}} ):
In this regime, the integral is finite:

Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \int_0^\infty dx \, e^{ (1/b_M-\beta) x}/b_M = \frac{1}{\beta b_M-1} = \frac{\beta_c}{\beta - \beta_c} }

In this regime, the total weight of the excited states is of the same order as the weight of the ground state. Consequently, the ground state is occupied with finite probability, reminiscent of Bose–Einstein condensation. However, unlike the directed polymer in finite dimension, this condensation involves not only the ground state but also the first excited states.

Overlap Distribution and Replica Symmetry Breaking:

The structure of states can be further characterized through the overlap between two configurations Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \alpha} and Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \gamma} , defined as

Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle q_{\alpha,\gamma} = \frac{1}{N} \sum_{i=1}^N \sigma_i^\alpha \sigma_i^\gamma, }

which takes values in the interval Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle (-1,1)} . The distribution Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle P(q)} of the overlap between two configurations sampled from the Gibbs measure distinguishes the two phases:

At high temperature (Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \beta < \beta_c} ), the system is replica symmetric and the overlap distribution is concentrated at zero:

Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle P(q) = \delta(q).}

At low temperature (Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \beta > \beta_c} ), the system exhibits one-step replica symmetry breaking, and the overlap distribution becomes

Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle P(q) = \tfrac{\beta_c}{\beta}\,\delta(q) + \Bigl(1 - \tfrac{\beta_c}{\beta}\Bigr)\,\delta(1-q).}


Finite Dimensional Systems

In finite dimensions, the fluctuations of the ground-state energy are characterized by an exponent Failed to parse (SVG (MathML can be enabled via browser plugin): Invalid response ("Math extension cannot connect to Restbase.") from server "https://wikimedia.org/api/rest_v1/":): {\displaystyle \theta} :

where is the linear size of the system and is the number of degrees of freedom.

When , the critical temperature vanishes with increasing system size, leading to the absence of a glass transition. This scenario occurs in many low-dimensional systems, such as the Edwards–Anderson model in two dimensions.

When , one must extend the definition of this exponent to finite temperatures and consider the fluctuations of the free energy . Three representative cases are:

  • Directed polymer in :

The fluctuations of the ground state exhibit a positive, temperature-independent exponent . In this situation, only the glassy phase exists, and

because producing an excitation with vanishing overlap with the ground state is very costly.

  • Directed polymer in :

The exponent depends on the temperature: it vanishes above the glass transition and becomes strictly positive below it. Accordingly,

at low temperature, and at high temperature.
  • Directed polymer on the Cayley tree:

The behavior is analogous to the Random Energy Model: in both phases. At high temperature,

while at low temperature the system exhibits the one-step replica symmetry breaking picture: