Upload
duongtruc
View
222
Download
0
Embed Size (px)
Citation preview
Nucleation for the zero range process on a finite set ofsites
Johel Beltran, PUCP - IMCA
Joint work withM. Jara and C. Landim (IMPA)
J. Beltran () Nucleation for the zero range process 1 / 9
A configuration: η ∈ {0, 1, 2, . . . }S .
Given η and x , y ∈ S , the transition η → ηx ,y occurs at rate
g(η(x))r(x , y)
J. Beltran () Nucleation for the zero range process 2 / 9
A configuration: η ∈ {0, 1, 2, . . . }S .
Given η and x , y ∈ S , the transition η → ηx ,y occurs at rate
g(η(x))r(x , y)
J. Beltran () Nucleation for the zero range process 2 / 9
A configuration: η ∈ {0, 1, 2, . . . }S .
Given η and x , y ∈ S , the transition η → ηx ,y occurs at rate
g(η(x))r(x , y)
J. Beltran () Nucleation for the zero range process 2 / 9
A configuration: η ∈ {0, 1, 2, . . . }S .
Given η and x , y ∈ S , the transition η → ηx ,y occurs at rate
g(η(x))r(x , y)
J. Beltran () Nucleation for the zero range process 2 / 9
A configuration: η ∈ {0, 1, 2, . . . }S .
Given η and x , y ∈ S , the transition η → ηx ,y occurs at rate
g(η(x))r(x , y)
J. Beltran () Nucleation for the zero range process 2 / 9
{r(x , y) ≥ 0 : x , y ∈ S} is such that
r is irreducible.
The uniform distribution on S is invariant for r .
g : {0, 1, 2, . . . } → R+ is such that g(0) = 0;
g(n)→ 1;
for some n0 ∈ N, g(n) > 1 ∀n ≥ n0; and
limn→∞ n(g(n)− 1) = b, with b > 1
g(n) ∼ 1 +b
n
J. Beltran () Nucleation for the zero range process 3 / 9
{r(x , y) ≥ 0 : x , y ∈ S} is such that
r is irreducible.
The uniform distribution on S is invariant for r .
g : {0, 1, 2, . . . } → R+ is such that g(0) = 0;
g(n)→ 1;
for some n0 ∈ N, g(n) > 1 ∀n ≥ n0; and
limn→∞ n(g(n)− 1) = b, with b > 1
g(n) ∼ 1 +b
n
J. Beltran () Nucleation for the zero range process 3 / 9
{r(x , y) ≥ 0 : x , y ∈ S} is such that
r is irreducible.
The uniform distribution on S is invariant for r .
g : {0, 1, 2, . . . } → R+ is such that g(0) = 0;
g(n)→ 1;
for some n0 ∈ N, g(n) > 1 ∀n ≥ n0; and
limn→∞ n(g(n)− 1) = b, with b > 1
g(n) ∼ 1 +b
n
J. Beltran () Nucleation for the zero range process 3 / 9
{r(x , y) ≥ 0 : x , y ∈ S} is such that
r is irreducible.
The uniform distribution on S is invariant for r .
g : {0, 1, 2, . . . } → R+ is such that g(0) = 0;
g(n)→ 1;
for some n0 ∈ N, g(n) > 1 ∀n ≥ n0; and
limn→∞ n(g(n)− 1) = b, with b > 1
g(n) ∼ 1 +b
n
J. Beltran () Nucleation for the zero range process 3 / 9
{r(x , y) ≥ 0 : x , y ∈ S} is such that
r is irreducible.
The uniform distribution on S is invariant for r .
g : {0, 1, 2, . . . } → R+ is such that g(0) = 0;
g(n)→ 1;
for some n0 ∈ N, g(n) > 1 ∀n ≥ n0; and
limn→∞ n(g(n)− 1) = b, with b > 1
g(n) ∼ 1 +b
n
J. Beltran () Nucleation for the zero range process 3 / 9
{r(x , y) ≥ 0 : x , y ∈ S} is such that
r is irreducible.
The uniform distribution on S is invariant for r .
g : {0, 1, 2, . . . } → R+ is such that g(0) = 0;
g(n)→ 1;
for some n0 ∈ N, g(n) > 1 ∀n ≥ n0;
and
limn→∞ n(g(n)− 1) = b, with b > 1
g(n) ∼ 1 +b
n
J. Beltran () Nucleation for the zero range process 3 / 9
{r(x , y) ≥ 0 : x , y ∈ S} is such that
r is irreducible.
The uniform distribution on S is invariant for r .
g : {0, 1, 2, . . . } → R+ is such that g(0) = 0;
g(n)→ 1;
for some n0 ∈ N, g(n) > 1 ∀n ≥ n0; and
limn→∞ n(g(n)− 1) = b, with b > 1
g(n) ∼ 1 +b
n
J. Beltran () Nucleation for the zero range process 3 / 9
{r(x , y) ≥ 0 : x , y ∈ S} is such that
r is irreducible.
The uniform distribution on S is invariant for r .
g : {0, 1, 2, . . . } → R+ is such that g(0) = 0;
g(n)→ 1;
for some n0 ∈ N, g(n) > 1 ∀n ≥ n0; and
limn→∞ n(g(n)− 1) = b, with b > 1
g(n) ∼ 1 +b
n
J. Beltran () Nucleation for the zero range process 3 / 9
We keep S fixed and N ↑ ∞.
Conditions on g produce a condensation effect!
J. Beltran () Nucleation for the zero range process 4 / 9
We keep S fixed and N ↑ ∞.
Conditions on g produce a condensation effect!
J. Beltran () Nucleation for the zero range process 4 / 9
We keep S fixed and N ↑ ∞.
Conditions on g produce a condensation effect!
J. Beltran () Nucleation for the zero range process 4 / 9
We are interested in {UNt : t ≥ 0},
UNt := ηtN2/N
∈ ES
whereES := {u ∈ RS
+ :∑x∈S
u(x) = 1} .
Theorem[B., Jara, Landim]For each N ≥ 1, let ηN be a configurations of N particles and let v ∈ ES
such that ηN/N → v . Let {UNt : t ≥ 0} start at ηN/N. Then
{UNt : t ≥ 0} (Law)−−−→ {Ut : t ≥ 0}
where {Ut}t≥0 is a Markov process starting at v .
J. Beltran () Nucleation for the zero range process 5 / 9
We are interested in {UNt : t ≥ 0},
UNt := ηtN2/N ∈ ES
whereES := {u ∈ RS
+ :∑x∈S
u(x) = 1} .
Theorem[B., Jara, Landim]For each N ≥ 1, let ηN be a configurations of N particles and let v ∈ ES
such that ηN/N → v . Let {UNt : t ≥ 0} start at ηN/N. Then
{UNt : t ≥ 0} (Law)−−−→ {Ut : t ≥ 0}
where {Ut}t≥0 is a Markov process starting at v .
J. Beltran () Nucleation for the zero range process 5 / 9
We are interested in {UNt : t ≥ 0},
UNt := ηtN2/N ∈ ES
whereES := {u ∈ RS
+ :∑x∈S
u(x) = 1} .
Theorem[B., Jara, Landim]For each N ≥ 1, let ηN be a configurations of N particles and let v ∈ ES
such that ηN/N → v .
Let {UNt : t ≥ 0} start at ηN/N. Then
{UNt : t ≥ 0} (Law)−−−→ {Ut : t ≥ 0}
where {Ut}t≥0 is a Markov process starting at v .
J. Beltran () Nucleation for the zero range process 5 / 9
We are interested in {UNt : t ≥ 0},
UNt := ηtN2/N ∈ ES
whereES := {u ∈ RS
+ :∑x∈S
u(x) = 1} .
Theorem[B., Jara, Landim]For each N ≥ 1, let ηN be a configurations of N particles and let v ∈ ES
such that ηN/N → v . Let {UNt : t ≥ 0} start at ηN/N.
Then
{UNt : t ≥ 0} (Law)−−−→ {Ut : t ≥ 0}
where {Ut}t≥0 is a Markov process starting at v .
J. Beltran () Nucleation for the zero range process 5 / 9
We are interested in {UNt : t ≥ 0},
UNt := ηtN2/N ∈ ES
whereES := {u ∈ RS
+ :∑x∈S
u(x) = 1} .
Theorem[B., Jara, Landim]For each N ≥ 1, let ηN be a configurations of N particles and let v ∈ ES
such that ηN/N → v . Let {UNt : t ≥ 0} start at ηN/N. Then
{UNt : t ≥ 0} (Law)−−−→ {Ut : t ≥ 0}
where {Ut}t≥0 is a Markov process starting at v .
J. Beltran () Nucleation for the zero range process 5 / 9
Let {ex : x ∈ S} be the canonical basis for RS .
Suppose v ∈ E and v(x) > 0 if and only if x ∈ A.Then for 0 ≤ t ≤ τ , (Ut(x))x∈A follows the generator
LAφ(u) =∑x ,y∈A
{b
cA(r , x , y)
u(y)∂ex + aA(r , x , y)∂ex∂ey
}
where cA(r , x , x) < 0 and cA(r , x , y) > 0 for x 6= y .
J. Beltran () Nucleation for the zero range process 6 / 9
Let {ex : x ∈ S} be the canonical basis for RS .Suppose v ∈ E and v(x) > 0 if and only if x ∈ A.
Then for 0 ≤ t ≤ τ , (Ut(x))x∈A follows the generator
LAφ(u) =∑x ,y∈A
{b
cA(r , x , y)
u(y)∂ex + aA(r , x , y)∂ex∂ey
}
where cA(r , x , x) < 0 and cA(r , x , y) > 0 for x 6= y .
J. Beltran () Nucleation for the zero range process 6 / 9
Let {ex : x ∈ S} be the canonical basis for RS .Suppose v ∈ E and v(x) > 0 if and only if x ∈ A.Then for 0 ≤ t ≤ τ , (Ut(x))x∈A follows the generator
LAφ(u) =∑x ,y∈A
{b
cA(r , x , y)
u(y)∂ex + aA(r , x , y)∂ex∂ey
}
where cA(r , x , x) < 0 and cA(r , x , y) > 0 for x 6= y .
J. Beltran () Nucleation for the zero range process 6 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N → v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1 and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0 .
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N
→ v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1 and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0 .
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N → v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1 and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0 .
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N → v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1 and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0 .
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N → v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1 and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0 .
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N → v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1
and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0 .
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N → v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1 and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0 .
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N → v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1 and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0
.
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
Fix some v ∈ {ex : x ∈ S}. Then
UNt := ηtN2/N → v , ∀t ≥ 0 .
DefineXNt := ηtNb+1/N , t ≥ 0
and, for each ε > 0 define ∆(ε) := {u ∈ ES : ‖u − ex‖ > ε}.
Theorem[B., Landim]
If ‖ηN/N − v‖ ≤ `N/N for 1� `N � Nb+1
b(κ−1)+1 and the uniformdistribution is reversible for r , then we have
EηN[∫ T
0 1∆(`N/N)(XNt ) dt
]→ 0, for every T > 0 .
The trace of {XNt : t ≥ 0} on ES \∆(`N/N) converges to a Markov
process {Xt : t ≥ 0} on {ex : x ∈ S}.
J. Beltran () Nucleation for the zero range process 7 / 9
References
Tunneling and metastability of continuous time Markov chains; J. B.,C. Landim. Journal of Stat. Phys. Vol 140, No 6, pp 1065 – 1114,(2010).
Metastability of reversible condensed zero range processes on a finiteset; J. B., C. Landim. Probability Theory and Related Fields. Vol152, No 3-4, pp. 781 – 807, (2012).
Metastability for a non-reversible dynamics: the evolution of thecondensate in totally asymmetric zero range processes; C. Landim
J. Beltran () Nucleation for the zero range process 8 / 9
References
Tunneling and metastability of continuous time Markov chains; J. B.,C. Landim. Journal of Stat. Phys. Vol 140, No 6, pp 1065 – 1114,(2010).
Metastability of reversible condensed zero range processes on a finiteset; J. B., C. Landim. Probability Theory and Related Fields. Vol152, No 3-4, pp. 781 – 807, (2012).
Metastability for a non-reversible dynamics: the evolution of thecondensate in totally asymmetric zero range processes; C. Landim
J. Beltran () Nucleation for the zero range process 8 / 9