Phase transitions for the long-time behavior of interacting diffusions

collision local time of random walks interacting diffusions large deviations Palm distribution phase transitions quasi-stationary distribution representation formula self-duality

Mathematics Subject Classification ID

Large deviations (60F10) Diffusion processes (60J60) Interacting random processes; statistical mechanics type models; percolation theory (60K35)

Abstract: Let

({X_{i} (t)}_{i i n m a t h b b Z^{d}})_{t g e q 0}

be the system of interacting diffusions on

[0, i n f t y)

defined by the following collection of coupled stochastic differential equations: �egin{eqnarray}dX_i(t)=sumlimits_{jin mathbb{Z}^d}a(i,j)[X_j(t)-X_i(t)] dt+sqrt{bX_i(t)^2} dW_i(t), eqntext{iin mathbb{Z}^d,tgeq 0.}end{eqnarray} Here,

a (c d o t, c d o t)

is an irreducible random walk transition kernel on

m a t h b b Z^{d} i m e s m a t h b b Z^{d}

,

b i n (0, i n f t y)

is a diffusion parameter, and

({W_{i} (t)}_{i i n m a t h b b Z^{d}})_{t g e q 0}

is a collection of independent standard Brownian motions on

m a t h b b R

. The initial condition is chosen such that

{X_{i} (0)}_{i i n m a t h b b Z^{d}}

is a shift-invariant and shift-ergodic random field on

[0, i n f t y)

with mean

T h e t a i n (0, i n f t y)

(the evolution preserves the mean). We show that the long-time behavior of this system is the result of a delicate interplay between

a (c d o t, c d o t)

and

b

, in contrast to systems where the diffusion function is subquadratic. In particular, let

h a t a (i, j) = 1 / 2 [a (i, j) + a (j, i)]

,

i, j i n m a t h b b Z^{d}

, denote the symmetrized transition kernel. We show that: (A) If

h a t a (c d o t, c d o t)

is recurrent, then for any

b > 0

the system locally dies out. (B) If

h a t a (c d o t, c d o t)

is transient, then there exist

b_{*} g e q b_{2} > 0

such that: (B1)d The system converges to an equilibrium

u_{T h e t a}

(with mean

T h e t a

) if

0 < b < b_{*}

. (B2) The system locally dies out if

b > b_{*}

. (B3)

u_{T h e t a}

has a finite 2nd moment if and only if

0 < b < b_{2}

. (B4) The 2nd moment diverges exponentially fast if and only if

b > b_{2}

. The equilibrium

u_{T h e t a}

is shown to be associated and mixing for all

0 < b < b_{*}

. We argue in favor of the conjecture that

b_{*} > b_{2}

. We further conjecture that the system locally dies out at

b = b_{*}

. For the case where

a (c d o t, c d o t)

is symmetric and transient we further show that: (C) There exists a sequence

b_{2} g e q b_{3} g e q b_{4} g e q ... > 0

such that: (C1)

u_{T h e t a}

has a finite

m

th moment if and only if

0 < b < b_{m}

. (C2) The

m

th moment diverges exponentially fast if and only if

b > b_{m}

. (C3)

b_{2} l e q (m - 1) b_{m} < 2

. uad(C4)

l i m_{m o i n f t y} (m - 1) b_{m} = c = s u p_{m g e q 2} (m - 1) b_{m}

. The proof of these results is based on self-duality and on a representation formula through which the moments of the components are related to exponential moments of the collision local time of random walks. Via large deviation theory, the latter lead to variational expressions for

b_{*}

and the

b_{m}

's, from which sharp bounds are deduced. The critical value

b_{*}

arises from a stochastic representation of the Palm distribution of the system. The special case where

a (c d o t, c d o t)

is simple random walk is commonly referred to as the parabolic Anderson model with Brownian noise. This case was studied in the memoir by Carmona and Molchanov [Parabolic Anderson Problem and Intermittency (1994) Amer. Math. Soc., Providence, RI], where part of our results were already established.

Review text

Let \((\{X_ i(t)\}_{i\in\mathbb Z^ d})_{t\geq 0}\) be the system of interacting diffusions defined by the following collection of coupled stochastic differential equations: \[ d X_ i(t)=\sum_{j\in \mathbb Z^ d}a(i,j)[X_ j(t)-X_ i(t)]\,d t+\sqrt{b X_ i(t)^ 2}\,d W_ i(t),\qquad i\in\mathbb Z^ d, t\geq 0. \] Here \(a(\cdot,\cdot)\) is an irreducible random walk transition kernel on \(\mathbb Z^ d\times\mathbb Z^ d\), \(b\in(0,\infty)\) is a diffusion parameter, and \((\{W_ i(t)\}_{i\in\mathbb Z^ d})_{t\geq 0}\) is a collection of independent standard one-dimensional Brownian motions. The initial condition is chosen such that \(\{X_ i(0)\}_{i\in\mathbb Z^ d}\) is a shift-invariant and shift-ergodic random field on \([0,\infty)\) with mean \(\Theta\in(0,\infty)\) (the evolution preserves the mean). The authors show that the long-time behavior of this system is the result of a delicate interplay between \(a(\cdot,\cdot)\) and \(b\), in contrast to systems where the diffusion function is subquadratic. Let \(\widehat a(i,j)=\frac 12[a(i,j)+a(j,i)]\) denote the symmetrized transition kernel. The authors show that: (A) If \(\widehat a\) is recurrent, then for any \(b>0\) the system locally dies out. (B) If \(\widehat a\) is transient, then there exist \(b_*\geq b_ 2>0\) such that: (B1) The system converges to an equilibrium \(\nu_\Theta\) with mean \(\Theta\) if \(0<b<b_*\). (B2) The system locally dies out if \(b>b_ *\). (B3) \(\nu_\Theta\) has a finite second moment if and only if \(0<b<b_ 2\). (B4) The second moment diverges exponentially fast if and only if \(b>b_ 2\). The equilibrium \(\nu_\Theta\) is shown to be associated and mixing for all \(b\in(0,b_ *)\). The authors argue in favor of the conjecture that \(b_ *>b_ 2\). They further conjecture that the system locally dies out at \(b=b_*\). For the case where \(a\) is symmetric and transient the authors further show that: (C) There exists a sequence \(b_ 2\geq b_ 3\geq b_ 4\geq\dots>0\) such that: (C1) \(\nu_\Theta\) has a finite \(m\)th moment if and only if \(0<b<b_ m\). (C2) The \(m\)th moment diverges exponentially fast if and only if \(b>b_ m\). (C3) \(b_ 2\leq (m-1) b_ m<2\). (C4) \(\lim_{m\to\infty} (m-1)b_ m=\sup_{m\geq2}(m-1)b_ m\). The proof of these results is based on self-duality and on a representation formula through which the moments of the components are related to exponential moments of the collision local time of random walks. Via large deviation theory, the latter leads to variational expressions for \(b_ *\) and the \(b_ m\)'s, from which sharp bounds are deduced. The critical value \(b_ *\) arises from a stochastic representation of the Palm distribution of the system. The special case where \(a\) is simple random walk is commonly referred to as the parabolic Anderson model with Brownian noise. This case was studied by Carmona and Molchanov in 1994, where part of the results of the present paper were already established.

Recommendations

Cites work

Cited in

(25)

This page was built for publication: Phase transitions for the long-time behavior of interacting diffusions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2373566)