scieee Science in your language
[en] (orig)

The persistence of synchronization under environmental noise

Abstract

It is shown that the synchronization of dissipative systems persists when they are disturbed by additive noise no matter how large the intensity of the noise provided asymptotically stable stationary stochastic solutions are used instead of asymptotically stable equilibria.

Read accessible full text

The persistence of synchronization under environmental noise

Author: Caraballo Garrido, Tomás; Kloeden, Peter E.
Year: 2005
DOI: 10.1098/rspa.2005.1484
Source: https://idus.us.es/bitstreams/c0f2c11e-567c-40a3-9a4e-427d3f06779a/download
The pe sis ence o synch oniza ion unde
en i onmen al noise
By Tom´
as Ca aballo aand Pe e E. Kloeden b
aDepa amen o de Ecuaciones Di e enciales y An´alisis Num´e ico,
Uni e sidad de Se illa, Apdo. de Co eos 1160,
41080-Se illa, Spain
E-mail: ca abal[email p o ec ed]
bFachbe eich Ma hema ik, Johann Wol gang Goe he Uni e si ¨a
D-60054 F ank u am Main, Ge many
E-mail: kloe[email p o ec ed]ank u .de
I is shown ha he synch oniza ion o dissipa i e sys ems pe sis s when hey a e
dis u bed by addi i e noise no ma e how la ge he in ensi y o he noise p o ided
asymp o ically s able s a iona y s ochas ic solu ions a e used ins ead o asymp o -
ically s able equilib ia.
Keywo ds: Synch oniza ion, addi i e noise, andom a ac o , s a iona y
s ochas ic p ocess, one-sided Lipschi z dissipa i e condi ion
1. In oduc ion
Synch oniza ion o coupled sys ems is a e y well known phenomenon in biology
and physics, and also in he social sciences. A eadable desc ip i e accoun o i s
di e si y o occu ence can be ound in he ecen book o S oga z (2003), which
con ains an ex ensi e lis o e e ences. The synch oniza ion o coupled dissipa i e
sys ems has been in es iga ed ma hema ically in he case o au onomous sys ems
by A aimo ich and Rod igues (1998), Ca alho e al. (1998) and Rod igues (1996),
bo h o asymp o ically s able equilib ia and gene al a ac o s, such as chao ic a -
ac o s. Analogous esul s also hold o nonau onomous sys ems (Kloeden (2003)),
bu equi e a new concep o a nonau onomous a ac o .
In his no e we in es iga e he e ec o addi i e noise on he synch oniza ion
o coupled dissipa i e sys ems wi h asymp o ically s able equilib ia, which esul s
in a coupled sys em o I o s ochas ic di e en ial equa ions. Such noise is o en
conside ed as modelling backg ound en i onmen al e ec s. We show ha synch o-
niza ion pe sis s independen ly o noise in ensi y in e ms o asymp o ically s able
s a iona y s ochas ic solu ions a he han equilib ia. Such asymp o ically s able
s a iona y s ochas ic solu ions a e a special case o andom a ac o s (see C auel
& Flandoli (1994), C auel e al. (1997), Schmal uss (1992)) which a e a p obabilis ic
coun e pa o he de e minis ic nonau onomous a ac o used in Kloeden (2003).
We p esen some equi ed backg ound ma e ial in Sec ion 2 and hen o mula e
he p oblem o synch oniza ion o dissipa i e sys ems o bo h he de e minis ic
and s ochas ic cases in Sec ion 3. In Sec ion 4 we show ha he uncoupled sys ems
wi h addi i e noise ha e andom a ac o s consis ing o single s a iona y s ochas ic
A icle submi ed o Royal Socie y T
EX Pape
2T. Ca aballo & P.E. Kloeden
p ocesses and hen es ablish he co esponding esul o coupled sys ems in Sec ion
5. Finally, he asymp o ic beha iou o he coupled sys ems in he limi as he
coupling pa ame e becomes a bi a ily la ge a e p esen ed in Sec ion 6.
2. Backg ound ma e ial
We need some backg ound no a ion and esul s. Le (Ω,F,P) be a p obabili y space.
Following A nold (1998) a andom dynamical sys em (RDS) (θ, φ) on Ω ×Rd
consis s o a me ic dynamical sys em θon Ω and a cocycle mapping φ:R+×Ω×Rd
→Rd. Essen ially (and su icien o ou pu poses he e), θ ep esen s he d i ing
noise p ocess and φ he s a e space e olu ion o he sys em. Fo example, o a
s ochas ic di e en ial equa ion on Rdwi h a wo-sided scala Wiene p ocess W ,
i.e. de ined o all ∈R,θis de ined by θ ω(·) = ω( +·)−ω(·) on a canonical
sample space Ω = C0(R,R) whe e, by de ini ion, W (ω) := ω( ), ∈R, and φis
de ined by φ( , ω, x0) = Xx0
(ω), he solu ion o he SDE s a ing a Xx0
0(ω) = x0.
See he exposi ions in A nold (1998) and Kloeden e al. (1999) o mo e de ails.
A amily b
A={A(ω), ω ∈Ω}o nonemp y measu able compac subse s A(ω)
o Rdis called φ-in a ian i φ( , ω, A(ω)) = A(θ ω) o all ≥0 and is called a
andom a ac o i in addi ion i is pa hwise pullback a ac ing in he sense ha
H∗
d(φ( , θ− ω, D(θ− ω)), A(ω)) →0 as → −∞
o all sui able (i.e. in a gi en a ac ing uni e se as, o ins ance, in Kloeden e al.
(1999)) amilies o b
D={D(ω), ω ∈Ω}o nonemp y measu able bounded subse s
D(ω) o Rd. He e H∗
dis he Hausdo semi-dis ance on Rd. The ollowing esul
(c . A nold (1998), Kloeden e al. (1999), Schmal uss (1992)) ensu es he exis ence
o a andom a ac o .
Theo em 2.1. Le (θ, φ)be an RDS on Ω×Rd. I he e exis s a amily b
B=
{B(ω), ω ∈Ω}o nonemp y measu able compac subse s B(ω)o Rdand a Tb
D,ω ≥
0such ha
φ( , θ− ω, D(θ− ω)) ⊂B(ω),∀ ≥Tb
D,ω
o all amilies b
D={D(ω), ω ∈Ω}in he gi en a ac ing uni e se, hen he RDS
(θ, φ)has a andom a ac o b
A={A(ω), ω ∈Ω}wi h he componen subse s de ined
o each ω∈Ωby
A(ω) =
s>0[
≥s
φ( , θ− ω, B(θ− ω)).
No e ha i he andom a ac o consis s o single on se s, i.e A(ω) = {X∗(ω)}
o some andom a iable X∗, hen X∗
(ω) := X∗(θ ω) is a s a iona y s ochas ic
p ocess.
We also need he ollowing lemma a.
Lemma 2.2. Le {xn}be a sequence in a comple e me ic space (X, d)such ha
e e y subsequence {xni}has a subsequence {xnij}con e ging o a common limi
x∗. Then he sequence {xn}con e ges o x∗.
A icle submi ed o Royal Socie y
Pe sis ence o synch oniza ion unde en i onmen al noise 3
P oo . I no , hen he e would exis an η > 0 and a subsequence {xni}such ha
d(xni, x∗)≥η o all i, which is no possible because {xni}has a subsequence
{xnij}con e ging o he x∗.
Lemma 2.3. Le W be a wo-sided Wiene p ocess, i.e. de ined o all ∈R.
Then he in eg als νR
−∞ e−ν( −s)dW (ω)a e pa hwise uni o mly bounded in ν > 0
on ini e ime in e als [T1, T2]o Rand he in eg als
Z
T1
e−ν( −s)dW (ω)→0as ν→ ∞
pa hwise on ini e ime in e als [T1, T2]o R.
P oo . In eg a ing by pa s (see Mao (1997)), we ha e
Z
−∞
e−ν( −s)dWs(ω) = W (ω)−νZ
−∞
e−ν( −s)Ws(ω)ds.
Then we use he sub-exponen ial g ow h o he pa hs o Wiene p ocesses. Fo he
second pa we i s no ice ha
νZ
T1
e−ν( −s)W (ω)ds =W (ω)−e−ν( −T1)W (ω)
o any ω∈Ω. In eg a ing again by pa s i ollows
Z
T1
e−ν( −s)dWs(ω) = W (ω)−e−ν( −T1)WT1(ω)−νZ
T1
e−ν( −s)Ws(ω)ds
=e−ν( −T1)(W (ω)−WT1(ω))
+νZ
T1
e−ν( −s)(W (ω)−Ws(ω)) ds
om which he esul ollows.
3. Fo mula ion o he p oblem
Suppose we ha e wo au onomous o dina y di e en ial equa ions in Rd,
dx
d = (x),dy
d =g(y),(3.1)
which a e su icien ly egula o ensu e he o wa ds exis ence and uniqueness o
solu ions and sa is y one-sided dissipa i e Lipschi z condi ions
hx1−x2, (x1)− (x2)i ≤ −L|x1−x2|2,
hy1−y2, g(y1)−g(y2)i ≤ −L|y1−y2|2,(3.2)
on Rd o some L > 0, and hus ha e unique equilib ia ¯xand ¯y, espec i ely, which
a e globally asymp o ically s able (c . Kloeden (2004)). No ice ha he con inui y o
and g, and he one-sided dissipa i e Lipschi z condi ions (3.2) ensu e he o wa ds
exis ence and uniqueness o solu ions o (3.1).
A icle submi ed o Royal Socie y
4T. Ca aballo & P.E. Kloeden
Conside now he dissipa i ely coupled sys em
dx
d = (x) + ν(y−x),dy
d =g(y) + ν(x−y) (3.3)
wi h ν > 0. I can be shown (see A aimo ich and Rod igues (1998), and Ca alho
e al. (1998)) ha his also has a unique equilib ium (¯xν,¯yν), which is globally
asymp o ically s able. Mo eo e , (¯xν,¯yν)→(¯z, ¯z) as ν→ ∞, whe e ¯zis he unique
globally asymp o ically s able equilib ium o he “a e aged” sys em
dz
d =1
2( (z) + g(z)) .(3.4)
This phenomena is known as synch oniza ion. Analogous esul s hold o mo e gen-
e al au onomous a ac o s (c . A aimo ich and Rod igues (1998), Ca alho e al.
(1998)) as well as o nonau onomous sys ems (Kloeden (2003)) wi h app op ia ely
de ined nonau onomous a ac o s.
The aim o his no e is o show ha his synch oniza ion e ec is p ese ed un-
de addi i e noise p o ided equilib ia a e eplaced by s a iona y andom solu ions.
Speci ically, we conside wo I o s ochas ic di e en ial equa ions in Rd,
dX = (X )d +α dW1
, dY =g(Y )d +β dW 2
,(3.5)
whe e α,β∈Rd
+a e cons an ec o s wi h no componen s equal o ze o, W1
,
W2
a e independen wo-sided scala Wiene p ocesses†, and ,ga e as abo e,
in pa icula , sa is ying he one-sided dissipa i e Lipschi z condi ions (3.2). We
will show in he nex sec ion ha each o hese s ochas ic sys ems has a pa hwise
asymp o ically s able andom a ac o consis ing o a s a iona y andom a iable.
Fo example, o linea d i e ms, i.e he SDEs
dX =−X d +α dW1
, dY =−Y d +β dW 2
,(3.6)
hese andom a iables a e gi en explici ly by
¯
X =αe− Z
−∞
esdW1
s,¯
Y =βe− Z
−∞
esdW2
s.(3.7)
The synch onized sys em co esponding o SDEs (3.5) eads
dX = ( (X ) + ν(Y −X )) d +α dW1
,
dY = (g(Y ) + ν(X −Y )) d +β dW 2
.(3.8)
I will be shown ha his sys em is dissipa i e and has a unique s a iona y solu ion
(¯
Xν
,¯
Yν
), which is pa hwise globally asymp o ically s able wi h
(¯
Xν
,¯
Yν
)→(¯
Z∞
,¯
Z∞
) as ν→ ∞,
pa hwise on ini e ime in e als [T1, T2] o R, whe e ¯
Z∞
is he unique pa hwise
globally asymp o ically s able s a iona y solu ion o he “a e aged” SDE
dZ =1
2( (Z ) + g(Z )) d +1
2α dW 1
+1
2β dW2
.(3.9)
†al e na i ely, one could ake α,βscala alued and W1
,W2
ec o alued
A icle submi ed o Royal Socie y
Pe sis ence o synch oniza ion unde en i onmen al noise 5
Fo example, o he linea SDEs (3.6) we ha e
dZ =−Z d +1
2α dW1
+1
2β dW2
wi h
¯
Z∞
=1
2e− µαZ
−∞
esdW1
s+βZ
−∞
esdW2
s¶=1
2¡¯
X +¯
Y ¢,
his a e aged exp ession being due o he special linea s uc u e.
In addi ion o he one-sided Lipschi z dissipa i e condi ion on he unc ions
and gwe assume he ollowing in eg abili y condi ion:
The e exis s m0>0 such ha o any m∈(0, m0], and any con inuous unc ion
u:R→Rdwi h sub-exponen ial g ow h i ollows
Z
−∞
ems| (u(s))|2ds < +∞,Z
−∞
ems|g(u(s))|2ds < +∞.(3.10)
Wi hou loss o gene ali y, we can assume ha L≤m0.
4. The uncoupled sys ems wi h addi i e noise
We conside he i s o he uncoupled equa ions in (3.5),
dX = (X )d +α dW 1
.(4.1)
I s solu ion pa hs a e gene ally no di e en iable, so in o de o use he one-sided
dissipa i e Lipschi z condi ion (3.2) we conside he di e ence X −¯
X whe e ¯
X
is he O ns ein-Uhlenbeck s a iona y p ocess (3.7) sa is ying he i s o he linea
equa ions (3.6). This di e ence is pa hwise di e en iable since he pa hs X and ¯
X
a e con inuous and sa is y he in eg al equa ion
X −¯
X =X0−¯
X0+Z
0¡ (Xs) + ¯
Xs¢ds,
which, by he undamen al heo em o calculus, is hus equi alen o he di e en ial
exp ession
d
d ¡X −¯
X ¢= (X ) + ¯
X ,
om which i ollows ha
d
d ¯¯X −¯
X ¯¯2= 2 X −¯
X , (X )− (¯
X )®+ 2 X −¯
X , (¯
X ) + ¯
X )®
≤ −2L¯¯X −¯
X ¯¯2+L¯¯X −¯
X ¯¯2+1
L¯¯ (¯
X ) + ¯
X )¯¯2
and hus
¯¯X −¯
X ¯¯2≤¯¯X 0−¯
X 0¯¯2e−L( − 0)+e−L
LZ
0
eLs ¯¯ (¯
Xs) + ¯
Xs¯¯2ds.
A icle submi ed o Royal Socie y

6T. Ca aballo & P.E. Kloeden
Pa hwise pullback con e gence (i.e. as 0→ −∞) gi es pullback abso p ion (c .
Theo em 2.1)
¯¯X −¯
X ¯¯2≤R2
X(θ ω) := 1 + e−L
LZ
−∞
eLs ¯¯ (¯
X(θsω)) + ¯
X(θsω)¯¯2ds
o all ≥Tb
D(ω) o app op ia e amilies b
D(ω) o bounded se s {D(θ ω), ∈R}o
ini ial condi ions.
The in eg als he e exis due o assump ion (3.10) and he ac ha O ns ein-
Uhlenbeck p ocesses inhe i he sub-exponen ial g ow h o hei Wiene p ocesses.
Thus ¯¯X (ω)−¯
X (ω)¯¯≤RX(θ ω),∀ ≥Tb
D(ω),
and so
|X (ω)| ≤ ¯¯¯
X (ω)¯¯+RX(θ ω),∀ ≥Tb
D(ω),
which means ha his sys em has a andom a ac o b
A={A(ω), ω ∈Ω}. Bu he
di e ence o any wo solu ions sa is ies he di e en ial inequali y
d
d ¯¯X1
−X2
¯¯2≤ −2L¯¯X1
−X2
¯¯2,
which means all solu ions con e ge pa hwise o each o he and hus he andom a -
ac o se s a e single on se s A(ω) = {X∗(ω)}, i.e. he andom a ac o is o med
by a s a iona y andom p ocess X∗
(ω) = X∗(θ ω) which pa hwise a ac s all o he
solu ions.
An analogous si ua ion holds o he second equa ion o he uncoupled equa ions
in (3.5).
5. The synch onized sys em wi h addi i e noise
To show ha he synch onized sys em (3.8) is s ongly dissipa i e, we use he OU
p ocesses
X∗,ν
=αe−ν Z
−∞
eνs dW 1
s, Y ∗,ν
=βe−ν Z
−∞
eνs dW 2
s,(5.1)
which a e he s a iona y solu ions o he linea equa ions
dX =−νX d +α dW 1
, dY =−νY d +β dW 2
.(5.2)
The di e ences o he solu ions o (3.8) and hese s a iona y solu ions a e pa hwise
di e en iable and sa is y he sys em o andom di e en ial equa ions
d
d ¡X −X∗,ν
¢= (X ) + ν(Y −X ) + νX∗,ν
d
d ¡Y −Y∗,ν
¢=g(Y ) + ν(X −Y ) + νY ∗,ν
,
A icle submi ed o Royal Socie y
Pe sis ence o synch oniza ion unde en i onmen al noise 7
which, wi h Uν
:= X −X∗,ν
and Vν
:= Y −Y∗,ν
, is equi alen o
d
d Uν
= (Uν
+X∗,ν
)+ν(Vν−Uν)+νY ∗,ν
,d
d Vν=g(Y )+ν(Uν−Vν)+νX∗,ν
.
Thus
1
2
d
d ¡|Uν
|2+|Vν
|2¢=Uν
, (Uν
+X∗,ν
)− (X∗,ν
)®+Vν
, g(Vν
+Y∗,ν
)−g(Y∗,ν
)®
+Uν
, (X∗,ν
) + νY ∗,ν
®+Vν
, g(Y∗,ν
) + νX∗,ν
®
−νhUν
−Vν
, Uν
−Vν
i
≤ −L¡|Uν
|2+|Vν
|2¢+L
2¡|Uν
|2+|Vν
|2¢
+2
L¯¯ (X∗,ν
) + νY ∗,ν
¯¯2+2
L¯¯g(Y∗,ν
) + νX∗,ν
¯¯2,
and hence
d
d ¡|Uν
|2+|Vν
|2¢≤ −L¡|Uν
|2+|Vν
|2¢+4
L¯¯ (X∗,ν
) + νY ∗,ν
¯¯2+4
L¯¯g(Y∗,ν
) + νX∗,ν
¯¯2.
This means ha |Uν
(ω)|2+|Vν
(ω)|2is pa hwise abso bed by he amily b
Bν=
{Bν(ω), ω ∈Ω}o closed balls in R2dcen e ed on he o igin and o adius Rν(ω),
whe e R2
ν(ω) is de ined by
1+ 4
LZ0
−∞
eLs ³| (X∗,ν (θsω) + νY ∗,ν (θsω)|2ds +|g(Y∗,ν (θsω) + νX∗,ν (θsω)|2´ds.
Hence, by Theo em 2.1, he synch onized sys em has a andom a ac o b
Aν=
{Aν(ω), ω ∈Ω}wi h Aν(ω)⊂Bν(ω). Bu he di e ence (∆X ,∆Y ) = (X1
−
X2
, Y 1
−Y2
) o any pai o solu ions sa is ies he sys em o andom di e en ial
equa ions
d
d ∆X = (X1
)− (X2
)+ν(∆Y −∆X ),d
d ∆Y =g(Y1
)−g(Y2
)+ν(∆X −∆Y ),
so
1
2
d
d ¡|∆X |2+|∆Y |2¢=∆X , (X1
)− (X2
)®+∆Y , g(Y1
)−g(Y2
)®
−νh∆X −∆Y ,∆X −∆Y i
≤ −L¡|∆X |2+|∆Y |2¢,
om which we ob ain
|∆X (ω)|2+|∆Y (ω)|2≤¡|∆X0(ω)|2+|∆Y0(ω)|2¢e−2L ,
which means all solu ions con e ge pa hwise o each o he as → ∞. Thus he
andom a ac o consis s o single on se s o med by an o de ed pai o s a iona y
p ocesses ( ¯
Xν
(ω),¯
Yν
(ω)).
A icle submi ed o Royal Socie y
8T. Ca aballo & P.E. Kloeden
6. The synch onized s a iona y solu ions as ν→ ∞
Lemma 6.1. ¯
Xν
(ω)−¯
Yν
(ω)→0as ν→ ∞ pa hwise on any bounded ime in e al
[T1, T2]o R.
P oo . Sub ac ing he second om he i s equa ion o (3.8) gi es
d¡¯
Xν
−¯
Yν
¢=¡−2ν¡¯
Xν
−¯
Yν
¢+ (¯
Xν
)−g(¯
Yν
)¢d
+α dW1
−β dW 2
,
o , wi h Dν
=¯
Xν
−¯
Yν
,
d¡Dν
e2ν ¢=e2ν ¡ (¯
Xν
)−g(¯
Yν
)¢d +αe2ν dW1
−βe2ν dW 2
,
so pa hwise
|Dν
| ≤ e−2νT1|Dν
T1|+Z
T1
e−2ν( −s)¡¯¯ (¯
Xν
s)¯¯+¯¯g(¯
Yν
s)¯¯¢ds
+|α|¯¯¯¯Z
T1
e−2ν( −s)dW1
¯¯¯¯+|β|¯¯¯¯Z
T1
e−2ν( −s)dW2
¯¯¯¯.(6.1)
By Lemma 2.3 we see ha he adius Rν(θ ω) is pa hwise uni o mly bounded on
each bounded ime in e al [T1, T2], so |Xν
(ω)|,|Yν
(ω))|and |Dν
0(ω))|a e pa hwise
uni o mly bounded on each bounded ime in e al [T1, T2]. Then by Lemma 2.3 and
condi ion (3.10) we see ha all o he in eg als in (6.1) con e ge o ze o as ν→ ∞
pa hwise on he bounded ime in e al [T1, T2].
Theo em 6.2. (¯
Xνn
,¯
Yνn
)→(Z∞
, Z∞
)pa hwise uni o mly on bounded ime in-
e als [T1, T2] o any sequence νn→ ∞, whe e Z∞
is he s a iona y s ochas ic
solu ion o he a e aged SDE
dZ =1
2( (Z ) + g(Z )) d +1
2αdW 1
+1
2βdW2
.(6.2)
P oo . De ine
Zν
:= 1
2¡¯
Xν
+¯
Yν
¢∀ ∈R
and obse e ha Zν
sa is ies he equa ion
dZν
=1
2¡ (¯
Xν
) + g(¯
Yν
)¢d +1
2α dW1
+1
2β dW2
.
Also de ine
¯
Z := 1
2¡¯
X +¯
Y ¢∀ ∈R
whe e ¯
X and ¯
Y a e he OU p ocesses sa is ying he linea SDEs (3.6).
The di e ence Zν
−¯
Z is pa hwise con inuous and hus by Lemma 2.3 and
he de ini ions o he espec i e abso bing se s, equi-bounded w. . . ν > 0 on he
bounded ime in e al [T1, T2]. Mo eo e Zν
−¯
Z is pa hwise di e en iable and
sa is ies he andom di e en ial exp ession
d
d ¡Zν
−¯
Z ¢=1
2 (¯
Xν
) + 1
2g(¯
Yν
) + 1
2¯
X +1
2¯
Y ,
A icle submi ed o Royal Socie y
Pe sis ence o synch oniza ion unde en i onmen al noise 9
Since
¯¯¯¯
d
d ¡Zν
(ω)−¯
Z (ω)¢¯¯¯¯≤1
2¯¯ (¯
Xν
(ω))¯¯+1
2¯¯g(¯
Yν
(ω))¯¯+1
2¯¯¯
X (ω)¯¯+1
2¯¯¯
Y (ω)¯¯
≤MT1,T2(ω)<∞
by Lemma 2.3, we can use he Ascoli Theo em o conclude ha o any sequence νn
→ ∞, he e is a (possibly) andom subsequence νnj(ω)→ ∞ such ha Zνnj
(ω)−
¯
Z (ω)→Z∞
(ω)−¯
Z (ω) as j→ ∞, and hus ha Zνnj
(ω)→Z∞
(ω) as j→ ∞.
Now
Zνnj
(ω)−¯
Yνnj
(ω) = 1
2³¯
Xνnj
(ω)−¯
Yνnj
(ω)´→0,
Zνnj
(ω)−¯
Xνnj
(ω) = 1
2³¯
Yνnj
(ω)−¯
Xνnj
(ω)´→0
as νnj→ ∞, so
¯
Xνnj
(ω)=2Zνnj
(ω)−¯
Yνnj
(ω)→Z∞
(ω),
¯
Yνnj
(ω)=2Zνnj
(ω)−¯
Xνnj
(ω)→Z∞
(ω)
as νnj→ ∞. Mo eo e ,
Zν
−¯
Z =Zν
−¯
Z +1
2Z
T1
(¯
Xν
s)ds +1
2Z
T1
g(¯
Yν
s)ds
+1
2Z
T1
¯
Xsds +1
2Z
T1
¯
Ysds
which con e ges pa hwise o
Z∞
=Z∞
T1+1
2Z
T1
(Z∞
s)ds +1
2Z
T1
g(Z∞
s)ds
+¯
Z −¯
ZT1+1
2Z
T1
¯
Xsds +1
2Z
T1
¯
Ysds
=Z∞
T1+1
2Z
T1
(Z∞
s)ds +1
2Z
T1
g(Z∞
s)ds
+1
2αZ
T1
dW1
s+1
2βZ
T1
dW2
s,
on he in e al [T1, T2], so Z∞
is a solu ion o he SDE (6.2) o all ∈R. The d i
o his SDE sa is ies he s ongly dissipa i e one-sided Lipschi z condi ion (3.2),
so i has a andom a ac o consis ing o a single on se o med by a s a iona y
s ochas ic p ocess which hus mus be equal o Z∞
.
Finally, we no e ha pa hwise all possible subsequences he e ha e he same
limi , so by Lemma 2.2 e e y ull sequence Zνn
ac ually con e ges o Z∞
as νn→
∞.
As a s aigh o wa d consequence o he a gumen s in he p e ious p oo we
ha e
A icle submi ed o Royal Socie y