Arratia's Rescaling

Limiting Point Processes for Rescalings of Coalescing and Annihilating Random Walks
on Zd
Author(s): Richard Arratia
Source: The Annals of Probability , Dec., 1981, Vol. 9, No. 6 (Dec., 1981), pp. 909-936
Published by: Institute of Mathematical Statistics
Stable URL: https://www.jstor.org/stable/2243755
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Institute of Mathematical Statistics is collaborating with JSTOR to digitize, preserve and

extend access to The Annals of Probability
This content downloaded from

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
All use subject to https://about.jstor.org/terms
The Annals of Probability
1981, Vol. 9, No. 6, 909-936
LIMITING POINT PROCESSES FOR RESCALINGS OF COALESCING

AND ANNIHILATING RANDOM WALKS ON Zd
BY RICHARD ARRATIA
University of Southern California
Let p(x, y) be an arbitrary random walk on Zd. Let At be the system of

coalescing random walks based on p, starting with all sites occupied, and let
aqt be the corresponding system of annihilating random walks. The spatial
rescalings P(O E (t)1/dtI for t - 0 form a tight family of point processes on Rd.
Any limiting point process as t -m00 has Lesbesgue measure as its intensity,
and has no multiple points. When p is simple random walk on Zd these
rescalings converge in distribution, to the simple Poisson point process for d
-2, and to a non-Poisson limit for d = 1. For a large class of p, we prove that
P(O E n11)/P(O E ) - 1? as t -m 00. A generalization of this result, proved for
nearest neighbor random walks on Z', and for all multidimensional p, implies
that the limiting point process for rescalings P(O E (t)l/dIqt of the system of
annihilating random walks is the one half thinning of the limiting point
process for the corresponding coalescing system.
1. Introduction. We consider two interacting particle systems on the d-dimensional

integer lattice Zd: coalescing random walks at, and annihilating random walks ijt. Each
process consists of identical particles, one starting from each site x E Zd. Each particle
undergoes a continuous time random walk on Zd, with mean one exponential holding times
between jumps, based on some fixed transition kernelp. These random walks are indepen-
dent, except that whenever a particle jumps to a site which is already occupied by another,
there is interference. In the coalescing system t, the two particles coalesce into one (one
particle vanishes); in the annihilating system qt both particles in a collision vanish
state space for each system is Y = {all subsets of Zd}, where x E 8 or x E r1t if there is a
particle present at site x at time t. The basic ergodic theory of these particle systems is
easy; the configuration ) is a trap, and starting from Zd or any other initial configuration
&t E-)d 8,, 'qt E-d So
Here, the convergence in distribution of t to 8, the probability measure on Y which is

concentrated on the single configuration 4), means that for any finite K C Zd,
P(t nK4())- 0 ast- oo.
See Griffeath (1979) for an exposition of this and other basic results about interacting
particle systems.
Since St Ed o it is natural to consider spatial rescalings
Pt aett,
choosing at so that the density of particles per unit volume in Rd is always on

of infinite particle systems are also considered in Holley and Stroock (1979) and Bramson
and Griffeath (1979). For a > 0, x = (xi, ***, Xd) E Rd, and A C Rd, write ax = (axi,
**, aXd), aA = {ax: x E A) C Rd. The following notation, which depends on the
underlying random walk p, will be used throughout this paper. Let
(1) pt = P(O E at), at =ptl/,

and for any B C Rd, t - 0 let
Bt = {x E Zd: atx E B) = (p /d B) n Zd.
Received August 7, 1980.

AMS 1970 subject classification. Primary 60K35.
Key words and phrases. Interacting Particle System.
909

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
910 RICHARD ARRATIA
Thus, for any compact convex B C Rd having Lebesgue measure m(B) > 0,
(2) Elattt n RI = Elt n BtI =PtIBtI -3 m(B) as to- oo.

Consider Vt = at6t as the random measure
(3) Vt = Exit, fa&x
on Rd having an atom of mass one at the rescaled location of each particle. Statement (2)
shows that the family { Vt, t - 0) is tight, so every sequence Vt, with ti -X o must
subsequence which converges to a limiting point process It on R dIn general, for arbitrar
p, we can show that any limit ,u is a simple point process (i.e. /A has no multiple po
having Lebesgue measure m as its intensity. What limits ,u are possible?
This question can be answered completely when p is simple random walk on Zd: for
each d = 1, 2, * , there is a point process jid on Rd such that
Cit6f a-)d I-d as t -) mo.
Theorem 1 states that for d>- 2, the limit above exists and is the basic Poisson point
process. For d = 1, the limiting point process pu on the line is not Poisson (Arratia 1979);
formula (20) given in Section 2 specifies i,1 in terms of its zero function. The proof of
Theorem 1 depends upon knowing the asymptotic behavior of Pt = P(0 E Ct), obtained for
simple random walks on Zd, d>- 2, in Bramson and Griffeath (1980b):
Pt (Vt)-112 d = 1
(4) (Irtllog t)-1 d = 2

(ydt)1 d>- 3
where Yd is the probability that a d-dimensional simple random walk never returns to its
origin. The other key ingredient for Theorem 1 is a negative correlation result, that
p(X Y Ett) cp(XE=- )p(y E=- ) = p2,
which we prove for arbitrary random walk p, using Harris's correlation inequality (Harris
1977).
What relation is there between the systems t, coalescing random walks, and qt,
annihilating random walks? There is a coupling such that, for all t>- 0, for every w,
(5) qt C Ot.
Recall that we start with all sites occupied: qjo = 40 = Zd. The coupling is easy to construct
directly: in each system, particles undergo the same random walks, and when two particles
collide, one of them disappears in it, while both of them disappear in qt. It may be bra
to suggest that since twice as many particles vanish per collision in Tt compared with it,
then the ratio of the density of particles in the two systems should go to one half as t goes
to infinity:
(6) P(0 E 7t)/P( E it) = P(0 E IOtl? E it) '/2 ast-3 mo.
Indeed, by the standard duality of coalescing and annihilating random walks with the finite
voter model t starting with a single individual at x, (6) is equivalent to
(7) P(OE -itlOE -t)=P(ltolisoddll '1>0)-1/2 ast-3 >o.

Since I jO I is a time change of simple (symmetric) random walk on the line, starting at one
with absorption at zero, (7) is highly plausible for any p. The only case for which (7) was
previously known is that of p being a nearest neighbor walk on the integers; in this case
the holding times for I tto I before absorption are all exponential-with mean 1/2. The rea
is invited to try to prove (7) on his own for a special case such as simple random walk on
Z2. Theorem 3 establishes this one half density relation for any genuinely multidimension
p, and for random walks on the integers having E p(0, x) I x I = oc. In the remaining c

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
COALESCING AND ANNIHILATING RANDOM WALKS 911
p a non-nearest neighbor walk on the integers with finite expectation, relation (7) remains
unproved.
For a large class of random walks p, Theorem 3 reduces the problem of finding the
asymptotic density of particles in the annihilating system qt to a problem abou
survival probability for the finite voter model tt. Partial results describing P(t
appear in Sudbury (1976), Kelly (1977), and Sawyer (1979). For simple random walk on
Zd, d 2 2, asymptotics forpt = P(O E At) = P(t2 # p) were finally established by Bramson
and Griffeath (1980b). Thus, for qit = ,td, the system of annihilating simple random walks
starting from all sites occupied,
P(OEEit) 1/(2v't) d=1
log tl(2irt) d = 2
1/ (2 Ydt 0 d 2 3.
The case with d = 1 above is in Griffeath (1979); the cases with d 2 2 are an immediate
consequence of Theorem 3 and the asymptotics (4) found by Bramson and Griffeath.
The one half density relation suggests that for large t, rjt may be approximately a "one
half thinning" of At, i.e. a subset of At obtained by tossing a fair coin for each particle in At
to decide whether to retain or delete that particle. Theorem 4 gives a precise version of
this: if p is a multidimensional random walk on Zd, or a nearest neighbor random walk on
Z1, then for any compact K C Rd, with Kt given by (1),
ZACK, (P(ftnKt=A)>3BCA 12-JAI-P(tnKt=BIftnKt=A)I) >O0 ast-+oo.

A more palatable restatement of this appears as Corollary 3: there exist versions of
8(attt), the one half thinnings of the rescaled point processes atft, such that for any
compact K C Rd.
P((atqit) IK # (o8at 6) IK) -O 0 as t-* 00,
where IK denotes the restriction of a measure to K. This implies a weaker result, Cor
4: for multidimensional random walk, or nearest neighbor random walk on the line,
if at6t ad l (along some sequence) then atqt - A

along the same sequence, where 01i is the one half thinning of the point process It on Rd
Combining the one half thinning result with the convergence results for rescalings of
coalescing simple random walks on Zd (a Poisson limit for d 2 2, a non-Poisson limit when
d = 1) and the asymptotic formulas (4) for pt = a d, we get:
For the system qt of annihilating simple random walks on

Zd starting with particles everywhere,
d = 1: (Ot)-1/2 atd 8d
d = 2: (Int/log t)-1/2qt ad 0l2 = Poisson, intensity 1/2
d- 3: (Ydt) /dqt ad 0/Id = Poisson, intensity 1/2.
Here 61, is the one half thinning of the point process i,' specified by formula (20). For d
>- 2, the one half thinning O/id of the intensity one Poisson process jd on Rd is the Poisson
point process with intensity one half.
The one half thinnings O/id that are the rescaled limits for rescaled annihilating simple
random walks are examples of a "compound point process." In general, if 8, ,81, ,82, * are
i.i.d., R '-valued random variables which are independent of a point process ,u = E x,,, then
the 13-compound of ji is the random measure E 813,&. In the one half thinning example
above, 13 is the fair coin variable with P(,8 = 1) = P(,8 = 0) = 1/2. For d 2 2, another
example of a 13-compound of the Poisson point process jid on Rd, with 1 exponentially
distributed, arises as the limit of rescalings of a system yt of coalesing simple random walks

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
912 RICHARD ARRATIA
on Zd in which mass is preserved. Start with a particle of mass 1 at each site x E Zd.
Whenever two particles in yt collide, they coalesce into a single particle whose ma
sum of the colliding masses. All particles, regardless of their masses, still undergo identical
random walks, independent apart from the coalescing interference. The state space now is
(Z+)Zd, where yt(X) = n means that there is a particle of mass n at site x E Zd, n = 0
(representing no particle present), 1, 2, .... This system arises naturally in studying
coalescing and annihilating random walks; the coupling (5) may be achieved by taking
,t = {X: yt(X) >0), ?} t = {x: yt(x) is odd).
The appropriate resealing of yt is the random measure lit on Rd defined by
(8) lOt = ExErtPtYt(X)8atx,
having atoms of mass Pt, 2pt, ..., carried on the lattice atZd. Extending a moment
calculation by Sawyer (1979), Bramson and Griffeath (1980b) show for simple random
walk on Zd, d>-2, that for any a-0,
(9) limt a P(yt(x) > - yt(x) > o = e

Pt
Theorem 2 says that in a rescaled limit, these exponential masses are independent of the
particle locations and each other. More precisely, for simple random walk on Zd with d -
2, /It converges in distribution to the fl-compound of the simple, intensity one Poisson point
process, where ,/ is exponential with mean one. In terms of Laplace transforms, Theorem
1 and 2 are equivalent to, and are proved by showing: for any continuous non-negative f
having compact support in Rd,
LW1(f) Eeiff da Ld (f) = e-
L11(f) 3 Ee-fdt - I'd (-log Lo Pf) = e dm
(where Vt and lut are specified in (3) and (8).)
2. Rescaling Coalescing Random Walks
2.1 Arbitrary Random Walks. We continue with the notation Vt = atft, where Pt
- P(0 E it), at = (pt)l/d, and It = (Zd is a system of coalescing random walks on Zd, startin
with all sites occupied, based on an arbitrary random walk p. This resealing is appropriate
in the sense that the family { Vt, t 2 0) of point processes is tight, with respect to the vague
topology.
[Here are the details: Let Be denote the E-neighborhood of B in the Euclidean metric on
Rd. We have, for any t-0, and B bounded
(10) Evt(B) = I atZd n B IPt < m(Bath)pt = m(B I) < m.

Using Chebyshev's inequality, this implies, for compact B, that
lima a supt:o P(vtB > a) = 0,
which is equivalent to tightness for the family {Vt}. See Kallenberg (1975) for a general
reference on random measures.]
Consider a limiting point process ,I on Rd, that is, suppose that Vt, Id M for som
sequence t -- oo. It is easily seen that for any compact, convex B C Rd,
Ei(B) c lrn sup Evt(B) = m(B).

If we are merely given some translation invariant spatially ergodic at C Zd having Pt -O 0,
without knowing that the it are obtained from coalescing random walks, then a limiting
point process ji might have Eli(B) < m(B). This can happen iff the random variables

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
vt(B) are not uniformly integrable, which indicates clustering in It. Another way to detect
clustering in It, relative to the scale of distance at = P(O E t)1/d, would be through the
existence of multiple points-atoms of mass 2, 3, ..., in a limiting point process ,u. Lemma
1 will be the key to showing that no such clustering occurs-the full strength of Lemma 1
is not needed here; an estimate such as: for x # y E Zd
(11) P(x, yE tt) -cp't

for some finite constant c, would be enough to establish Corollaries 1 and 2. It is necessary,
in our proof of Theorem 1, to have (11) with c = 1 in order to conclude that for simple
random walk on Zd, d- 2, the limiting point process ,I is a simple Poisson process,
than some mixture of Poisson point processes (i.e. a Cox process).
LEMMA 1. For any A C Zd, and for an arbitrary random walk p on Zd, let (A be the
system of coalescing random walks starting with a particle at each site x E A. For any
X # y E Zd,
(12) P(x, yE t) C P(x E it )P(y E it)
In the special case A = Zd, this becomes
PROOF. Let (VB, B C Zd) be the family of voter models based on p, all constructed
together via a single random substructure Y. (B9 is a collection of Poisson flows T(x, y) on
[0, oo) having rate p(x, y) = p(O, y - x); the event times of T(x, y) tell the voter at x when
to discard his opinion and adopt the opinion held by the voter at y.) Thus for any B, C C
Zd, for each w, ~tBUC = gt U tc. By the usual duality between coalescing random walks and
the voter model, (12) is equivalent to
(13) pw n A & , n A & )cpw n A & )fltynA70+).

We apply Harris's (1977) elegant theorem on positive correlations: for a monotone Markov
process on a finite partially ordered state space E, a necessary and sufficient condition for
the set of measures on E having positive correlations to be preserved by the semigroup is
that the process can only jump up or down. Ignore momentarily the requirement that the
state space E be finite. Take
E= {(B, C):BC CCZd}
with the ordering (B, C) c (B', C') iff B C B' and C C C'. Define the process starting at
(B, C) E E to be
X(BC) = B CF;
this process is monotone and has only jumps up and down. For an initial distribution
having positive correlations, take the deterministic configuration ({x}, Zd _ {y}) which
has zero correlations; thus we are considering the process
= (tX, fZd_,Y,) = (tx, Zd _ by)
The conclusion, that for every t - 0 the distribution of Xt has positive correlations, says
that for any increasing functions f, g on E,
E(f(Xt)g(Xt)) > Ef(Xt)Eg(Xt).

Take
f((B, C)) = lBfnA,1, g((B, C)) = lAcC;

these are increasing functions on E. The previous inequality becomes
(14) lttxn A 0 0, rt nA = ) 'lttxn A 3& )P(rt nA = 0).

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
914 RICHARD ARRATIA
Replacing the event (tY n A = 4) by its complement yields formula (13).

To comply with the requirement in Harris's theorem that the state space E be finite, an
approximation is needed. For any r > max(I x I, l y l) let Sr = {z E Zd: I Z I< r}. For ea
w modify the substructure B? to create a new substructure 96r, by deleting all event time
of the clocks T(z,,,) having I z I > r or I z' I > r. Construct a family of voter models (rGP
C Zd) based on fir; in this family no site outside Sr can give or receive influence. Let Er
= {(B, C) B C C C Sr) C E, and for (B, C) E Er define the process rXtBC) = (rdt, r4tC);
this is a monotone Markov process on the finite state space Er. Take unit mass on ({x},
Sr - { y)) as the initial distribution. Define increasing functions fr and gr on Er by
(15) fr(B, C) = lBnlA#4b gr(B, C) = lAflSrCC.
The theorem on positive correlations yields
(16) PGrx n A 7& 0, rt n A = 0) 2 PGrx n A o O)P(,r n A = 0).

Define Tr = inf{t: At # rdt or Ayt # rdy} . Almost surely, Tr ()o as r-* 00; this is equiv
to the claim that the substructure X for the voter model has no influence from oo. Thus,
taking limits as r -X o in (16) yields formula (14). 0
A slight generalization of this is needed for Theorem 2; Lemma 1 is exactly the special
case m = n = 1 of the following lemma.
LEMMA 2. For an arbitrary random walk p on Zd, A C Zd, let yA be the system of
coalescing random walks with mass conserved, starting with a particle of mass one at
eachsitexEA. ThenVtmn>-OxyEZd,
(17) p (-y (x) 2_ m, -Y (y) 2- n) c< p(-yA(x) 2_ M)p(-y
PROOF. In terms of the usual coupling with the family of voter models, (17) is
equivalent to
Pq flAl A m, t l nAl n) - P(l | n AlI m)p(l rt n Al n).

The proof of this is exactly the proof given for Lemma 1, with the increasing functions f
and gr of (15) replaced by
fr, m(B, C) = 1IBnAI?:m, grn(B, C) = 1j(Ansr)-CI<n. E
COROLLARY 1. Let vt be the random measure on Rd defined by (3), for an arbitrary

random walk p on Zd. For any bounded B C Rd.
(18) var(vtB) c< E(,vtB).
If, along some sequence 4t - 00, vt, id t
El(B) = m(B),
i.e. the intensity of any limiting point process ,A is Lebesgue measure m on Rd.
PROOF. With Bt given by (1), vt(B) = I (t n Bt 1. Identify (t with its indicator function
i.e. write (t(x) = 1 if x E it; (t(x) = 0 otherwise. Lemma 1 says that E( t(x)(t(y)) s if x
# y. Thus
E(vtB)2 = E(QxEB, (t(x))2 = xEtB, E(Qt(x)) + ZxyEB, E( t(x) t(y))
c I Bt Ipt + I Bt 12Pt = E(vtB) + (E(vtB))2,

which shows (18). Let B C Rd be bounded and convex. A calculation like (10) shows that
EvtB -* m(B) as t-- oc. It follows that E(AB) c limi E(vPtB) = m(B); the intensity
limiting point process is absolutely continuous with respect to Lebesgue measure. Since

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
m(aB) = 0 implies EI(aB) = 0 and hence 1i(OB) = 0 a.s., B is a ,i-continuity set. Thus as
random variables, vtB ad ,IB, and formula (18) shows that the vtB are uniformly
integrable, so that EiB = lim Evt B = m(B). 0
COROLLARY 2. With the same hypotheses as Corollary 1, any limiting point process
IL is simple (or orderly), i.e.
P(IL({x}) > 1 for some x E Si) = 0.
PROOF. It follows easily from Lemma 1 that for any bounded B C Rd,
P(vtB > 1) = P(at n Bt > 1) C xZxyEB, P(x, y E Et) 0 I BtI2pt = (EvtB)2.
Write Sr = {x E Rd: I X I < r). There is a constant c depending only on d, such that fo
r E (0, 1), Si can be covered by nfr s c/m(Sr) translates U1, U2, ***, U, of Sr. Now
P(,I({x}) > 1 for some x E Si) < P(1I(Uj) > 1 for somej, 1 cj c nr)
' P(ji( Uj) > 1) = Ej 1im, P(vt( Uj) > 1)
' Ej limi (Evt, Uj)2
= E m(Uj)2 ' (c/m(Sr))(m(Sr))2 = C.m(Sr)

Taking a limit as r -* 0,
P(,I({x}) > 1 for some x E Si) = 0. ?
2.2 Simple Random Walks. For the remainder of this section we will consider only
the case where p is simple random walk on Zd, i.e. p(x, y) = 1/(2d) if I x - y I = 1, p(x, y)
= 0 otherwise. Thus, when a particle moves, it chooses any one of the 2d neighboring sites
in the lattice Zd with equal probability. For the system of coalescing simple random walks,
the asymptotic behavior of pt = P(0 E at) is known in all cases (d = 1 in Bramson and
Griffeath 1980, d>- 2 in Bramson and Griffeath 1980b). The dth root of these asymptotics
is:
at (rt)-1/2 d = 1
(19) at (V"t/log t)-1/2 d = 2
at (ydt)-ld d - 3.
Since t-1/2 is the appropriate normalizing factor for a sim

we see that at/t-1/2 __ oo as t -a 00, when d - 2. Thus
point process should be Poisson when d>- 2, and not Poisson when d = 1. The case d = 1,
coalescing simple random walk on the line, is analyzed in Arratia (1979). The result is that
Vt d /I1, where the limiting point process ,ii on R can be realized as the state at time 1/so
of a system of coalescing standard Brownian motions on the line, starting with a particle
at each x E R. A self-duality relation for the system of coalescing Brownian motions leads
to a formula expressing IL, in terms of its zero function:
(20) P(1,i(B) = 0) = P(?j1B = 4 )
Here B C R is a finite disjoint union of intervals, and tB is a finite system of annihilating

standard Brownian motions on R, starting with a particle at each site in aB. For contrast,
the zero function for the limiting point process /Id for d - 2 is
P(/Id(B) = 0) = e-m(B)
for any Borel set B C Rd.

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
916 RICHARD ARRATIA
THEOREM 1. For d -2, as t-- 00,
vt attt ad Id,
where jLd = Pm is the simple Poisson point process on Rd whose intensi

measure m. Here, it is the system of coalescing simple random walks on Zd, starting with
all lattice sites occupied; at is given asymptotically by (19).
PROOF. To show that vt = at6 is approximately Poisson, we will run the system
for a while with the collision mechanism suspended-particles will follow independent
random walks with no interference, and lattice sites may be multiply occupied. For any
t - e (= 2.7 ... ), set
At= At(t) = t/V t d = 2
= tl/2+1/d d - 3
so that atV7At -A 00, and At = o(t), hence pt-At/pt -- 1, as t -m 00. Set
(21) s= s(t) = t- At,
and let at be the system of coalescing random walks with the collision mechanism suspended
from time s to time t = s + At. Write tt(x) = 0, 1, 2, ... for the number of particles at
E Zd; for each t there is a coupling such that V w, V X E Zd,
Let
vt at 6 E>xezd Xt(X)8ax.
We claim that for any compact B C Rd,
(22) P(Vt IB 7i tI B) -O as t -oo,

where IB denotes the restriction of a measure to B. [Proof: W
density of continuous time simple random walk on Zd, run for time t. The density of
particles in at is
E~t(y) = E (E. ,8(x)p&1(x, y)) = p.,,pp(x, y) = ps.

Thus
P( t(x) # (t(x)) C E(~t(x) - (t(X)) = p-pt,
_ | Btl|( ps,-pt.)= (EvtB)( ps-pt)/pt- O as t- m.

so
P(Vt lB # Vt B) _> EXEB, P(Qt(X) # tt(x))
This establishes the claim (22).]

For each t define a transition kernel qt(x, dy) on Rd which describes the motion of each
particle of v, = asks to its location in vt = at6:
qt(x, B) = p&1(a 1x, at-'B).
(These do not form a semigroup.) Since atv it - 00,
(23) suPxeRd qt(x, B) -O 0 as t -X 00

for any compact B. By the independence of the particle motions go
(24) E(exp[-f f(y)-(dy)] I vs) = expf log( e-(Y)qt(x, dY))vs(dx)]

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
for any Borel function f on Rd. For a similar analysis of a co

system, see Liggett (1978).
For each t define a random measure Mt on Rd by
(25) Mt(B) = f qt(x, B)v8(dx) = , P&,(x, Bt)
so that
E(-it(B) l s) = Mt(B).
(To check that this is really a random measure, having Mt(B) < 00 a.s. for compact
convex B, compute
EMt(B) = EE(-vt(B) l = E-vt(B) = ps I Bt | m(B) as t -m oo.
This also shows that the family {Mt} is tight.)

Start with any convergent sequence of {vt}, say t d - v as t -m oo. Since P(vt I B I P
-? 0, - Ed V also, as t -X oc along this sequence. Take a subsequence along which Mt E
The Laplace transform of v can be computed as the limit of the Laplace transform of it,
along the subsequence for which -i E v and Mt Ed M. Let f E C+(Rd), the set
of continuous non-negative functions on Rd having compact support. Write gt(x) =
-log f exp(-f(y))qt(x, dy). By (23), sUpX If exp(-f(y))qt(x, dy) - l -* 0 as t 00; t
(26) gt(x) 1 uniformly in x.

f (1 - exp(-f(y)))qt(x, dy)
Now along our subsequence,
Lf =E exp(-f f(x)v(dx))
= lim Lj1(f) = lim EE(exp[-f f(x)Pt(dx)] |v
(by (24)) = lim E(e-fgt(x)>,,(dx))
(by (26)) = lim E(e-f[f (l-e-AY))qt(xddY)]^s(dx))
(by (25)) = lim E(e-f(l-e-AY))Mt(dy))
= lim LM/(1 -e-f) = LM(1 - e-f).
Thus v = PM, the mixture of Poisson processes directed by the random measure M.
We want to show that v is the simple Poisson process with intensity m, i.e. that M = m
(a.s.). Take B to be compact and convex, and consider the random variables v(B) and A
- M(B). The distribution of v(B) is a mixture of ordinary Poisson distributions with
parameter A directed by A, so that Ev(B)2 = E(A2) + EA. We know that EA = Ev(B) =
m(B); the bound (18) of Corollary 1, together with vt(B) Ed v(B) yields
Ev(B)2 ' lim sup E(vtB)2 < lim sup[(EvtB)2 + E vtB] = m(B)2 + m(B).
Thus E(A2) ' (EA)2, so A = m(B), (a.s.). We have M(B) = m(B) a.s., for each compact
convex B, so M = m (a.s.). Thus v = Pm, the simple Poisson point process on Rd with
intensity one. Since this limit is obtained along a subsequence of an arbitrary convergent
sequence (with t4 - oo) from the tight family {Pt, t ' 0), the theorem is proved. E
Let yt be the system of coalescing random walks on Zd with mass conserved, startin
with a particle of mass one at each lattice site. Write yt(x) = n to indicate that there is a
particle of mass n = 0, 1, 2, * * * at x E Zd; interpret yt(X) = 0 as "no particle present." In
terms of the usual coupling of the family (Q, A C Zd) of coalescing random walks, which

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
918 RICHARD ARRATIA
has the additive property: for all A, B C Zd,
(t = t t,
the system yt may be defined b
Yt(X) ={y E: ZdY = {X}} .

For each site x, the mass yt(x) at x has mean
Eyt(x) = >y(Zd pt(y, X) = 1.
The particle locations for yt are the same as those for t: for each w,
xEtt iff yt(x) >0.
The expected mass of a given particle in yt is
E(-yt(x) I yt(x) > 0) = Eyt(x)/P(yt(x) > 0) = 1/pt.
Thus we rescale both the spatial locations and the masses of the particles; let
ALt x E.EZd Ptyt(X)8a.x = ExErt Ptyt(X)8ax.X
THEOREM 2. Let p be simple random walk on zd, d - 2. As t -X oo, It, converges in

distribution to the /3-compound of Pm. Here, /3 is the exponential distribution with mea
one on R + and Pm is the simple Poisson point process on Rd whose intensity is Lebesgue
measure m. In terms of Laplace transforms,
Lill (f ) Lp4(-log L3 f) - e-ff/(l+f)dm

for every f E Cc+(R ).
PROOF. Define a particle system (* on Zd X Z+ by: for x E Zd, n = 1, 2, * ,
(x, n) EE ,t* iff yt(x) = n.

Consider a spatial resealing v* of t* as a point process on Rd X R+:
P*
Vt - (x,n)e~t 8(atx.ptn).
We will repeat the proof of Theorem 1 to show that
(27) vt *d PmXfl,
where Pmxf is the Poisson point p

To see that this establishes Theor
define g on Rd X R+ by
g(x, a) = af(x).
Now
LVlt f = E exp(-E>e=Zd ptyt(z)f(atz)) = E exp(-f g dvt)
= L (g) -* LPmxfi(g) = exp(-f (1 - e-) d(m x /3))
(28) = exp(-f (1 -e-af(x))/3(da)m(dx))
= exp(-f (1- L/3(f(x))m(dx))
= Lp (-log L3 of

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
This shows that as random measures on R d
ht id the f-compound of Pm.
The expression (28) for the Laplace transform of the limiting random measure can be
simplified; since L,8(t) = (1 + t)-1,
Lp (-log L,8 o f) = exp(-f (1-LO(f(x))m(dx)) - eff(x)/(l+f(x))m(dx).
To prove the Poisson convergence (27), take s and t as given by (21). Let ,t* be the
system ,t* run with the collision mechanism suspended from time s to t, writing (Zx, n)
= 0, 1, 2, ... for the number of particles at (x, n), x E Zd, n = 1, 2, * .. The corresponding
point process on Rd x R+ is
t= ,t(x, n)8(atx,ptn).
We need to show that there are couplings of Ct* and ,t* such that
(29) P(t 4IBXR+ # t 4IBXR+) -O as t-*oo

for every compact B C Rd. In Theorem 1, there was a coupling such that vt V va, but such
a relation for the * system is not possible. In C*, a particular at (x, n) may jump to (y, n)
if x and y are nearest neighbors in the lattice Zd; if there already is a particle at (y, m) (for
m = 1, 2, ... , possibly even m = n) then both particles vanish in collision and are replaced
by one particle at (y, m + n). Consider a joint construction of the systems Ct, ,t*, $t
Ct* in which for each r E [s, t],
XEE (r iff (x, n) E for some n = 1, 2,* ;
r(X) = En., Cr (x n).

Consider (yIs, 0 s u -l At), a coalescing system in which mass is conserved, starting with
a particle of mass one at each x E C,. Thus with the natural coupling, -yu(x) = n indicate
that n particles from Cs have coalesced together to form a single particle in Cs+u at site x, s
for each u E [0, At],
xE- C+u iff y-(x) ' 1.
With this coupling, for any x E Zd,
(30) {w: xt*(x n) #4 (t*(x n) for some n} = {(t(x) # (t(x)) U {y}&t(x) - 2).
Compute
Eyu(x) = EyeZdPu(y, x)P(y E s) = yPu(Y, X)Ps = Ps.
Now
P('YAt(x) > 1) = P(x E Ct) = pt.
so
P('Yst(x) > 2) <Ps -Pt.
Therefore the event in (30) has probability _ 2(ps - Pt), SO
P(V* IBXR+ #4 V* |BXR+) - ExeBi P(tt*(x, n) # * (x, n) for some n)
hBtis 2(posth () h0 as to l.
This shows that (29) holds.

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
920 RICHARD ARRATIA
The transition kernel of Rd X Rt which describes the motion of each particle in v* to

its new location in t is
qt*((x, a), B x I) = p&t(a-sx, aItB)1i(pta/ps);
for any compact B C Rd
SUP(Xa) qt*((x, a), B x R+) -O 0 as t-- oc.
Define random measures on Rd X R' by
Mt*(B x I) = X(znfs P/t(z, Bt)11(ptn) = f qt((x, a), B x I) dv* (x, a).
so that
E (v-t* (B X I) l s)=Mt* (B x I).
For IC R+ of the form [a, oo), x E Rd
E(z(zn)e,; pAt(Z, x)l,(ptn)) = XzEZdn.a/pt P((z, n) E s )pAt(z, x)
= Xn;a/p, P((O, n) E is*) = P(y8(O) > a/pt)
e-ap. (by (9) and ps/pt 1.)

Thus, for I = [a, om) and B compact convex C Rd
EjTt* (B x I) = EMt* (B x I) = E >XEB, Z(zn)Ets pAt(Z, x) 1(ptn)
| Bt I e-aps 8 m(B)ea
=(m xf3)(B x I).
Note that E( t* (B x Re)) = E(i-t(B)) - c < oo for compact B, so that the familes {t*, t
0) and {Mt*, t - 0) of random measures on Rd X R' are tight. Starting with any convergen
sequence of v*, say t* v -d Vie take a subsequence along which Mt* converges, say Mt* d
M*. A Laplace transform calculation similar to (26ff.) shows that v* is the mixture of
Poisson processes on Rd X R' directed by the random measure M*. Lemma 2 is now used
to compare first and second moments of the random variable v* (B x [a, 00)) and show that
M* = m x ,8 almost surely. This establishes (27) and concludes the proof of Theorem 2. 0
3. Annihilating Random Walks. The one half thinning relationship between the
system qt of annihilating random walks and the system it of coalescing random walks
natural when viewed in terms of the dual system, the family of voter models (t, A c Zd).
(See Griffeath (1979) for an exposition of all the material in this paragraph.) The voter
model it is the spin flip system on Zd in which the voter at any site x changes opinion at
a rate equal to the proportion of his neighbors (weighted by p) who hold the opposite
opinion; equivalently, the flip rates are
C(X, D) = EypdZd: (.)#q(y) p (X, Y).
Identify the state space tO, 1}Zd for this spin system with A= fall subsets of Zd}, and write
Ot for the voter model starting with opinion 1 held by the voters at sites x E A, opinion 0
held everywhere else. When A = {x}, write gt' for the voter model tX} starting with a lone
dissenting opinion at x. Thus {w: rt #4 4 is the event that this dissenting opinion survives
until time t, and I gtx I is the size at time t of the dynasty of converts to that dissenting
opinion. For each t - 0, there is a coupling, based on a random substructure B of event
times, of (t, the system of coalescing random walks starting with all sites occupied, jqt, the
corresponding system of annihilating random walks, and (t, A C Zd), the family of voter
models started at each possible initial configuration, such that for every W,
(31) (tx= {X: }t= {X | Ix: is odd}.

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
Thus the one half density result of Theorem 3 is equivalent to a statement about the
parity of the finite voter model:
(32) P (O E- 7qt)IP (O E- It) = P( |to I is odd I ~to 34 0) --*1/2 as t m- 0.

When the voter model is in state A, for some finite A C Zd, it grows in size by one (i.e. A
-4 A U {x) for some x Z A) at rate >.XAyeA p(x, y), and decreases in size by one at rate
ZxeA,ygA p(x, y). By the translation invariance of p, these two rates are equal. We write
(33) r(A) = Zx9A ,yEA p(X, y) + XEA,y*A p(X, y) = 2 xEA,yA p(X, y)

for the total jump rate out of state A. The configuration 4 is a trap for the voter model, i.e.
r(o) = 0; but for any finite A # 4, r(A) - 2(1 - p(O, 0)) > 0. Thus I tt I is a time change
St, a simple random walk on Z+, started at one, with absorption at zero, and with mean one
exponential holding times between jumps. In the special case where p is a nearest neighbor
random walk on Z, the state of the voter model tto before absorption is a block of
consecutive integers, with r(~t0) = 2. Using the reflection principle and the local central
limit theorem, taking Xt to be simple (unstopped) random walk on Z started at zero,
P(t I is odd I0t# ) = P(S2t is oddi S2t> 0)

= P(X2t = O)/P(X2t = 0 or 1)
-- 1/2 as t a) o.
This is the only case of p for which the one half density relation (32) had been established.
Theorem 3 extends this to arbitrary multidimensional random walks p, and to p on the
integers Z having E I x I p(O, x) = mo.
For any random walk p, the size of the finite voter model, conditional on survival, tends
in probability to infinity: for any m,
(34) P(Jltol>-mlgt?+o)) + aster*o.

This is stated and proved as Lemma 3. If p is multidimensional (Lemma 4) or p has
>xEzp(O, x) I x I = X0 (Lemma 5), then
limm-- infACZdmsA<, r(A) = 00.
Combining this with (34) yields, for these random walks, that the border size of the voter
model tO, conditional on survival, tends in probability to infinity: for any ro,
(35) P(r(~tt) > ro I tto I 4) - 1 as t-* oc.

Contrast this with the nearest neighbor case on the line, where for all t
(36) P(r(;t0) = 2I1 ~to # ) = 1.

Here is a quick proof of Theorem 3, that the one half demsity relation (32) holds for any
random walk which satisfies (35). This argument cannot be extended to derive the one half
thinning relation, Theorem 4. A second proof of (32), involving a randomization of the
substructure Y of event times for the voter model, will be given as our formal proof of
Theorem 3. This second proof of Theorem 3 is really a special case of the proof of Theorem
4, but this latter proof is sufficiently complicated that it is useful to write out the full
argument in the special case.
FIRST PROOF OF THEOREM 3. We want to show that (32) holds. GivenE eE (0, 1), set
(37) ro = 8E-Iog(1/E)
and using (35), take to so large that t > to implies
(38) P(r(?t,) > ro I lto ?# 4) > 1-E.

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
922 RICHARD ARRATIA
At any jump of t?, r (pt) changes by at most 2. Let
(39) At = E3/16.
For any c - 1,
P(I r(2) - C I > EC for some s E [t, t + At] I r(~t) = c)
(40) c P(t? has at least Ec/2 jumps during [t, t + At] I r(gt) = c)
c P(X ? Ec/2), where X is a mean At(l + E)C Poisson random variable
E(X2) [At(l + E)C][At(l + E)C + 1] 16At(At + 1/(2c) 16At
-< ~~~ ~ ~ ~ = E.
(Ec/2)2 (Ec/2)2 E 2 E2
For a continuous time Markov chain Y whose states can be partitioned into "odd" and
"even" states which alternate with each other, and with all jump rates lying in the interval
[(1 - E)C, (1 + E)C] it is easily proved that
(41) P (Y&t is odd) 2E -!e -2cAt
[Here is a proof of (41). Let f(t) = P(Yt is odd), and write [a, b] for the interval of rates.
Then
f'(t) - a(1 - f(t)) - bf(t) = -(a + b)(f(t) - a/(a + b)).

Gronwall's inequality yields
(f(t) - a/(a + b)) e-(a+b)t(f(O) - a/(a + b)) - -a/(a + b)e-(a+b)t
f(t) a (1 -e-(a+b)t
from which (41) follows.] Using ro and At as specified by (37) and (39), we get, for any c
> ro,
(42) P(Y&t is odd)> e e-2ro5t - -E.

2 2 2
Given that gto = A with

Markov chain Y as described above, so that
P ( ,At 7& Y,&t) < E,
and hence, by (42)
P( |&At I is odd) > 1/2 - 2E.
Thus
P( It?+At I is odd I r(~t?) ? ro)> 1/2 -2E,
and this together with (38) yields, for any t > to,
P( I t+tI is odd) -(1/2 - 2E)(1 - E)P(1to # )> (1/2 - 3E)P (;t1w ).

The same argument shows that for t > to,
P Ot+AtI is even, gt?+At # 4)> (1/2 - 3E)P(~tt+A # 4).
Thus, for t > to,
1 1/2 - P( Itt+AtI is odd I -t#+)I < 3E.
This establishes (32) and concludes the first proof of Theorem 3. The second proof also
starts from statement (35); the heart of this second proof is presented as Lemma 6.

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
A natural generalization of the one half density relation (32) between annihilating and
coalescing random walks is that such a relation should hold independently at each of n
sites. Let A = {X1, X2, * * , x4 C Zd, IA I = n; it should be the case that for each B C A,
P(qt nA=B IA Ct) 2-' ast-*oo.

Expressed in terms of duality (31) with the family of voter models, this is: for any qi, q2,
* qn EE {? 1 }
P(llt--Iqimod2, i= ltonl zt+, oi= lton)-32-' aster-o.

Although this is plausible, it is intractable; there is no random walk p for which we can
prove the required analog of (34), namely that for each finite A C Zd, for every m
P(ll:-mVxEAIlt#t)VxEA)--*as t-*oo.
Intuitively, the worst case is when the set A is a cluster; a proof along the same lines as the
proof of Lemma 3 requires a lower bound on the probability of the conditioning event of
the form: for every a > 0,
eatP(ft-#4)VxEA) = eatP(A C t) *oo aste-* o.
A different way of generalizing the one half density result (32) involves the notion of a
one half thinning Ot of It: independently toss a fair coin for each particle in It to decide
whether or not to retain that point. Equivalently, if v112 is a random subset of Zd whose
distribution is product measure with density one half, and V1/2 independent of It, then
t = It n v112,
or any random set with this distribution, will be called a one half thinning of It. With the
usual notion of convergence in distribution for random elements of {tO 1) Zd, the statement
that Tt gets close to a one half thinning of It as t gets large is: for any finite K C Zd,
0 = liMt4At JACKP(tt n K = A)(EBCA lP(?t n K = B I t n K = A)-2-AI).
This result, for each fixed K, holds almost trivially since
(43) P(ftnK=A)-*1 if A=+
-*0 if A=+,
i.e. since It converges in distribution to 4.

We cure this, and arrive at the statement of Theorem 4, by replacing K above with sets
Kt C Zd such that E I (t n K I - c > 0. The uniform integrability estimate (Corollary 1) on
the random variables I (t n Kt I now ensures that lim sup P((t n Kt = p) < 1, so that
objection (43) has been overcome.
Corollary 3 gives an almost literal translation of Theorem 4 in terms of the rescaled
point processes discussed in Section 2: for each t there exists a version Ot of the one half
thinning of (t such that, for every compact K C Rd,
(44) P((atqt)IK # (at&(t)[K) -+ 0 as t -* oo.

Using Laplace transforms, Corollary 4 follows immediately:
(45) if att -*d IL, then att -*d OIL,

the limit in distribution for rescaled annihilating random walks (along some sequence t
-* oo) exists and is the one half thinning of the limiting point process for the coalescing
random walks.
Since the notion of convergence of point processes requires only that corresponding
atoms get close in position, statement (45) is not as strong as statement (44). As an
example, we will produce {Og 1}Zd valued processes At C (t for which (45) holds, with
limiting Poisson point process, while (44) fails. Let the distribution of (t be vpt, produ

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
924 RICHARD ARRATIA
measure with density Pt -O 0. Let
tf {x E zd:xi + x2+ **+ xd is even).
Then at f and atq' converge to the simple Poisson point processes on Rd having intensiti
respectively 1 and 1/2, but for any version of 0ft, for any compact convex K C Rd having
Lebesgue measure m (K) > 0,
P(atr't 1K # atOIt JK) - P( ( fl {x E Kt:xi + * + Xd is odd) -1 - e(K)/.
LEMMA 3. For the voter model based on an arbitrary nontrivial random walk p, for
every m
P( I Pomm I#4 ` )- 1 aste-* ).
PROOF. The jump rate for At, conditional on Ut # p, is at least c = 2(1 - p (0, 0)) > 0.
By comparison to a process which always jumps at rate c, then exists At such that
p(tA has at least m - 1 jumps during [0, At], or t' = 4) > 1/2,
for all finite A C Zd. With probability 2-m1-) the first m - 1 jumps are down, so
(46) P(gAAt= 0) > 2-m

for all A C Zd having I A I < m. An upper b
model #t is available from duality with coalescing random walks at, and Lemma 1 (recall
that p = P(O E at) = P(t #o )):
d
- =- dt P(0 E 6) = ZOxEZd P(0, XE ~t)p(0, X) -<X >.oPtP(0, X) Pt'
By (46),
Pt+&t --Pt - 2--P (O <IC~oI < m)
so that
P(0 <J| |< m)-c 2m(pt -pt+,At) C 2 pt/

Thus
P(JltJ< mJ1 0))<2`PtAt 0 ast-*oo. 5
For any A C Zd, define
(47) OA = {(x, y) :x y YE Zd, p (X, y) >0, and 1= I{x,y}fnA I}.

For any C C (Zd)2, define
(48) rp(C) = >(xy)ec P (x, Y)
With this notation, formula (33) for the jump rates of the voter model is
r (A) =rp (A) = 2 >xeA,gAp (X, y).
LEMMA 4. For any genuinely multidimensional random walk p on Zd,
(49) h1fn oinfACZd:n-fAi<oo rp(OA) = 00.
PROOF. Fix two linearly independent Xi, x2 E Zd with
A min(p(0,x1),p(0, X2)) > 0.
Consider a subset of the support of p:

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
S = {(x, y) E (Zd)2:yx = xI or Y-X = X2).

For i = 1, 2 define maps
Bi :A S n aA
x -* (x + kxi, x + (k + 1)xi),
where k is the maximal interger such that x + kx, E A. The map x -3 (Bi(x), B2(x)) from
A into (S n aA)2 is 1 -1 since each of the pairs Bi determines a line in Rd, and these two
lines intersect exactly at x. Thus
A IIs n A 12, so r,(OA) 2 rp(S n fA) 2x 7 A I.
LEMMA 5. For any random walk on Z such that I I x I p (0, x) = 00,

of the previous lemma holds.
PROOF. We may assume without loss of generality thatp is symmetric and irreducible.
Thus there is a set K = {x1, X2, * **, Xk) whose elements are positive integers with greate
common divisor 1, such that
A-min1Si-kp(0, Xi) > 0.
Let 1= max(x1, ** , Xk). Since 1 = a1x1 + * * * + akxk for some integer coeffi
it follows for any integers a < b that there is a path
(50) a = So, SI, S2, *., sj = b

from a to b, with displacements I si - si-I I E K, which stays inside the interval (a - 1,
b + 1). Let r > 0 be given; take m so large that
>X=exp(0, x) 2 r.
For integers 0 c i < x write

Aix = (i + xZ) n A.
If Ai x 4) define
aix = min(A,,x)
so that ai x E A, and ai - x C A. If for each i, x with [- x c m and 0 < i < x we have Aj x

0 4, then those pairs (a, ai,x -x) are sufficient:
r,(OA) > ,Xm=e ji=-J p (ai, x, x- x) = ,J= xp (0, x)-r.
If, on the other hand, Ai x = 4 for some i, x with 1{ x ' m, 0 ' i < x, then choose any
integer j E [0, 3x) such that I Aj, 3x I I A I /(3x). The assumption Ai x = 4 guarantees, for
every n such that a = 3nx + j E A, that b = 3nx + i X A. A path of the form (50) leads to
an edge en = (cA, dn) CE A having p(cn, dn) 2 A, and
(3n - 1)x - cn < dn < (3n + 2)x.
This last relation is needed in order to conclude that for different n, the edges en are really
distinct. Now if I A I > 3xr/A, then
rp(aA) 2 >n:3nx+jEAp(cn, dn) 2 (IA 1/3x)X > r.
LEMMA 6. For anyE EC (0, 1), with
At = E/8, and r = (4/E)log(1/E),
for any random walk p on Zd, for any finite A C Zd, rp (A) 2 r implies
He, is te vr m l o Zis oddp ) I < A.
Here , tA is the voter model on Z d based on p, starting from A.

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
926 RICHARD ARRATIA
PROOF. We use a randomization-take a flow of jump times having double the usual
rate, and then select with a fair coin which times will be used to run the process. Such a
device appears in Griffeath (1979) to prove that "cancellative systems having pure births
are exponentially ergodic." This randomization can be viewed as a modification of the
usual coupling proof (run two copies of the process independently, starting one of them in
equilibrium, until they reach the diagonal, then run both together along the diagonal) that
a finite state irreducible Markov process converges to its equilibrium distribution. In the
context here, our proof is a modification of the coupling proof that a two state (even, odd)
Markov chain which flips back and forth between even and odd at rate r each way, starting
from a pure state Seven or 8odd can be coupled to the chain in equilibrium (1/2) (8even + 8odd)
by time A t with probability exactly 1 - (1/2)exp(-2rAt).
Construct the voter models DB for every B c Z d simultaneously from a substructure BA
consisting of independent exponential alarm clocks T(Xy) which ring at rates p (x, y). When
the clock for the pair (x, y) rings, the voter at site x is influenced by the voter at site y: the
configuration either flips at x or else remains unchanged, so that its components at x and
y agree.
Write a = (x, y) for a pair of sites in Z d, and let Ta = T(x y) be the random set of times
at which the clock for (x, y) rings, with
Ta = {ta(1) < ta(2) < ...
To say that Ta is a Poisson flow or Poisson point process on [0, mo) with rate or intensity
p (x, y) is to say that ta(1), ta(2) - ta(1), ta(3) - ta(2), . . . are independent and exponentially
distributed with mean 1/p (x, y). Thus, the substructure ?v is a family of independent
random flows Ta of event times, indexed by pairs a = (x, y) E (Zd)2.
Construct 9 as the one half thinning of another substructure 9' having double the
usual rates. That is, let s"' = (T'xy), x, y E Zd), where T'(x, y) = {tt (1) < t' (2) < ...) is
a Poisson flow with intensity 2p(x, y), let W = (c(x,y)(i); x, y E Zd i = 1, 2, ...) be a
collection of fair coin tossing variable (P(ca(i) = 0) = P(ca(i) = 1) = 1/2) independent of
BA, and set
Ta(w) = {t'(i, w): a(i, w) = 1, i = 1, 2, * .
When ca(i) = 0 so that t' (i) is not an event time in the substructure bA, we say "there is a
phantom effect at time t' (i)," and when ca(i) = 1 we call the effect at time t' (i) "real." For
any C C (Zd)2 having rp(C) -(xy)ecp(x, y) < co, write
Tc = UaEc Ta
for the set of all event times of .4' for edges in C; T'c is a Poisson flow with intensity
2rp(C). Let
T(W) = inf TdA, and
aO(,) = (XTl YT)= the pair a such that T = t' (1)
be the time and place of the first effect in bA' on some edge in OA. Let
G = {w: ca(1) = 0)
be the event, having probability 1/2, that the effect at time T in B4' is a phantom effect.
Thus, for all w,
T fl {X7,y7)I 1G (mod 2).

Define H to be the event
H n ind ependeXT, YT here 1 (mod 2) e .
Although G and H are not independent, there is an event F having P(F) > 1 - E,

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
conditional on which G and H are independent. F is a g7'-measurable event which

guarantees thatT " '-At and each of the sites xT, y, can neither give nor receive influence at
any time before At other than . To make this precise, for any x EE Z d let
Nx 3 a({x)) = {(x, y) :p(x, y) >0, y # x) U {(y, x) :p(x, y) > 0, y # x),
and for any pair a = (x, y) E (Zd)2, let
Na = Nx U Ny.
Note that rp(Nx) c 2 and rp(NO) c 4. For each a E dA, define the event
Fa = {(w a, = a, Tr C <At, [O. At] n (TN',- {(T))=44.

Define the event F by
F = UaEOA Foa
It can be seen that
for w E F, I I t- + 1H (mod 2).
Here is the argument that P( F) > 1 - 2E. For any a E dA and t E [0, A t],
P (FaI T = t, at = a)
= P(TNmnOA fn (t, At] = 4, TN-dA fn [0, At] = 4 IT = t, aT = a)
= P(TNndA fn (t, At] = 4, TN-dA fn [0, At] =
2 e-2rp(N,)&t 2 e- e8E> 1 -=E.
Integrate over t E [0, A t] to get
P(Fa,) 2 (1-E) P(T ' At, aT = a)
and sum over a E dA to get
P(F) 2 (1 -E) P(T _ At).
Since bA' has double the usual rates, and rp(dA) 2 r-(4/E)log(1/E), and At
P(T > At) = e-2rp(aA)t < e-2r E
Thus P(F) 2 (1 -E)2 > 1 -2E.

Next we show that G and Hare independent, conditional on F, and that G is independent
of F. Notice that for any a = (x, y) E dA, the event Fa f { I A- {x, y} I 1 mod 2} is
measurable with respect to the a-field generated by NA', (ca(i) i 2 2), and (cfl(i), 8 E (Z d)2
- {a}, i 2 1), i.e. with respect to the information in i?' and W excluding the variable ca(1).
Thus we can compute
P(F n G n H) =a=(x, y)EOA P( FaYCa()= ,t{XY)} 1)
= 1/2 P( Fa I t^At _ {X, Y} I -1
= /2P(FfnH).
The same argument using the complements of G or H establishes the claims of indepen-
dence.
Finally,
P(ItI isodd) -P(Fn GnHC) +P(Fn GcnH)
= l/2P(FfnHC) + 1/2P(FfnH)
=-1/2 P(F) > '/2(1 - 2E) = 1/2 - E

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
928 RICHARD ARRATIA
and similarly
P(I t|is even)> 1/2-E. E
THEOREM 3. Let p be any random walk on Z d for which statement (35) about the
voter model holds. In particular, p may be any genuinely multidimensional random walk,
or a random walk on Z having l I x I p (0, x) = co. Then
(51) P (O E q~t) / P ( E (t) ---1/2 as t - o m.
Here, qt is the system of annihilating random walks, starting from al

(t is the corresponding system of coalescing random walks.
PROOF. Given E E (0, 1), set r = (4/E)log(1/E) and At = E/8. Lemma 6 guarantees that
for finite A C Zd, rp (A) 2 r implies
II1/2 I_/ p(I

APIAtl IsodlI<,E.
is odd)
Since the flip rates for the voter model at any site are at most 1,
P(t At# ) -Ae't> 1 - E, for any A #.
Thus rp (aA) 2 r also implies
11/2- P(j ~tA is even, tt # k)I < 2E.
By relation (35), (which follows easily for those particular cases of p mentioned by using
Lemmas 3, 4, and 5,) there exists to such that t >to implies
P(rp(a(gt?)) 2r g 4to 0) > 1-E

Thus for t > to,
P I gt?+AtI is odd) 2 (1 - E)(1/2- 2,E)P(~t # ) > (1/2- 3E)P(gtAw # )
The same inequality holds for P( I 't+&t I is even, gt+At # 4), so that
? 1/2-P I 't%+AtI is odd I 4t4+1t + I)| < 3E.
Thus P( | ~t is odd I #to 4 >) 1/2 as t a0, which is equivalent to (51) by duality. 1
Now we embark on a proof of Theorem 4. The main part of the argument lies in Lemma
8, which generalizes the technique of Lemma 6.
LEMMA 7. Letp be a genuinely multidimensional random walk on Zd; let r > 0 and
n > 0 be given. There exists m such that for any disjoint finite sets A1, A2, * , An C
I Ai1 I m for i = 1 to n implies the existence of Ci, C2, * * Cn C (Zd)2 satisfying, for
i, j c n,
(52a) Ci C dAi, and rp(Ci) 2 r;
(52b) there is a basis {di, d2, * * *, dJ for (Z/2)' su

I{x, y) n Aj I = 1 iff the jth component of di is
(52c) if i # j, (x, y) E Ci, and (u, v) E Cj, then {
Informally, condition (52b) says that for the voter models gAt if the first jump is caused
by an alarm clock for a pair (x, y) E Ci, then di will be the parity change in (I A , I mod 2,
i = 1 to n) E (Z/2)-. Condition (52c) may be restated: if i 1 j, a E8 Cl, then C1 f Na" # 4.
PROOF. Continue with the notations xl, x2, A, S from the proof of Lemma 4. Let
ml= r ro/ -,

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
(m1 is the number of pairs from S needed in each Ci to yield a rate ?ro), let
m2 = 6min,
(m2 is the number of pairs from S needed for sets Cl so that subsets Cl of size exactly ml
may be chosen to satisfy the "no interference" condition (52c)), and take
m = (m2n2)2
for the value m whose existence this lemma asserts. Assume now that we are given disjoint
sets Ai C Zd, satisfying m < I Ai I < o for i = 1 to n. Let
Ao = Zd _ Ui=i Ai,
N= {O,1, ,n}.
Define a subset of the border between Ai and A,
Bi0= (oAi) n (aAj) n s.

Define an undirected graph G on N by
G = {(i,j) E N2: I Bij I 2 m2)

We prove that G is a connected graph by showing that any nonempty connected componen
I c N must contain 0. Suppose to the contrary that 0 M I. Let A = Ui=1Ai, so that m c I
< o and by Lemma 4,
(53) | Ans I >- = m2n2.

In the partition
aA nfs= UiEI UjEN-I Bij,
each Bij has I Bij < M2, since the choice of I as a connected component means (i, j) 0 G.
This partition involves at most n2 sets Bij, giving I dA n s J < n2m2 in contradiction to (5
Thus, 0 is an element of every connected component of G.
Since G is connected, there is a subgraph T C G which is a connected tree on N. Fix
such a T and define a map
s:N- {O)} N
s (i) = the unique vertex: (i, s (i)) is the first edge of a path in T from i to 0.
Let el, e2, * , e, be the standard basis for (Z/2)' and let eo = 0. For i = 1 to n, define
di = ei + e.(i).
To see that {d1, d2, * *, ch) is a basis for (Z/2)', check that each ei is in its linear span.
Just follow the path in Tfrom i to 0, i.e., with h(i) = min{k:sk(i) = 0),
el= =k L0 d-k()
For i = 1 to n, let
Ci = Bi.,(i),
so that the C1 are disjoint subsets of S, and the Ci in place of the Ci satisfy (52b), and i Ci |
2 M2 = 6min. Observe that for any pair a EC S,
I(N {}a)) nfSl=6.

Choose any subset C1 C C1 having I C1 i = ml. Proceed recursively, at the ith st
n, choosing
Ci C Ci -(Ul.-1<i UEC, (Na n Ci))

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
930 RICHARD ARRATIA
to have I l = mi. This is possible si

(i - 1)ml values of a, and each Na n cl has at most 6 elements. This produces sets C1, C2,
. .. , C,, such that i < j, a E Ci, and,8 EE Cj imply,8 M Nat. This establishes (52c). Notice th
rp(C,) >- A I CI = A (r/?C1) r. E
LEMMA 8. For any genuinely multidimensional random walk p, E E (0, 1), and
positive interger no there exists m with the following property:
Given disjoint finite subsets Al, A2, * * , An C Zd, n S no, with each I Ai I 2 m, define the
random joint parity vector in (Z/2)':
Qt- (I tA, I mod 2, i= 1 to n).

For At = E/8no,
(54) .q e (Z/2) n 12 - P(Qut = q)| < 4E.

PROOF. Set r = (4no/E)log(no/E) and take m as determined by Lemma 7 (depending on
p, r and no; this m will also work for n = 1, 2, * * *, no in place of no.) Let Al, A2, , * * An be
any disjoint finite subsets of Zd having I Ai I 2 m, with n S no. The unqualified indices i, j
will always be taken to range over 1, 2, * , n. By Lemma 7, there are sets of pairs Ci C
MAi having rp(C,) 2 r, satisfying also (52b) and (52c). Condition (52b) says that if the fi
effect in B7 to change any of the gtA, is at a E Ci, then the change in Qt produced by tha
jump will be di.
We continue with notation from the proof of Lemma 6: there is a double rate substruc-
ture 7', an independent collection of fair coins l used to thin 7', and the resulting ordinary
substructure 9 is used to construct the family of voter model (PA, A C Zd). Define
ri = T'c,
to be the time of the first effect in the double rate substructure 7' to involve a pair in Ci.
Define the random pair
A = (xi, yi) E Ci
to be the pair a such that ri = iX ' (1), i.e., the place at which the effect at ri occurs. Define
the 0, 1-valued variable
ci= cf3,(1)
to be the value of the coin in W which determines whether the effect at time -r is real (ci
= 1) or phantom (cl = 0). The vector of these ci is a random element of (Z/2)n
C () = (C1, C2, * * , CO)
for which each possible value has probability 2 .

Consider the set of sites involved at the random times ii:
B (w) = U=.1 {xi, yi).
Note that I B I = 2n since i #j implies ,/i M NqB. Define a modified parity vector Qt*, similar
to Qt, but which "resets" the opinions on B to what they were initially, just before counting
up the ttA, (mod 2):
Qt*(w) = (| - B I + I Ai n B I mod 2, i = 1 to n).
For any a = (x, y) E Ci define the event
FaW = {/31= at, i-i ' At, (T' -{rj) n [O, At] = o ,
which says that ri -- At is caused by an effect in 7' at a, and that sites x and y are not
involved in any other effects in 7' during [0, At]. Thus, on Fa there cannot be any
interaction before At between what happens at time Ti (a real or phantom effect, indicated
by cM(O) = 1 or 0) and the rest of the evolution of the family of voter models. Define F,

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
representing successful coupling, with no interference possible, by
F = UaIECI ... UaEC,, (nflg Fa,).
A little reflection shows that for w E F,
QAt(w) = Q~t(w) + Ei[., ci(w) di.
Taking D to be the n x n matrix over Z/2 whose rows are d1, d2, dn, this may be
rewritten as
Qwt(w) = Q *t(w) + c (w)D, for w E F.
By arguing that conditional on F, Q t and c are independent, and that c is independent of

F, it can be seen that for all qj, q2 E (Z/2)'
(55) P(Q~t = qj, C = q2, F) = 2-'P(Q~t = qj, F).

Thus, for any q E (Z/2)',
P(Q&t(w) = q) ' P(Q&t(w) = q, F)
= EqE(Z/2)- P(Q t = qj, c = (q - ql)D-1, F)
= 2-nP(Q t = qj, F)
= 2-nP(F).
Once it is shown that
P(F) > 1-2E,
we are done since
EqE=Z/2)n 12-n P(Q&t(W) = q)I c 2P(Fc) < 4E.

Thus, all that rermains is to demonstrate the independent claim (55), and to show that
P(F) > 1 - 2E.
Here is the proof that P(F) > 1 - 2E. For any choices of ai E Ci and ti E [0, At], using
i 0 j impliesCj nNa, = 4,
P(fl=1 F., I ri = ti, aT, = ai for i = 1 to n)
= P(T'N nc n (ti, At] = >, TN -c, n [O, At] = 4),
for i = 1 to n 1 ri = ti, a,, = ai for i = 1 to n)
= P(T'Nncfn (ti, At] = 4, T -cf n [O, At] =4)for i= 1 to n)
- H exp(-2r,(N,4)At)
2 exp(-8nAt) > 1 - E.
Integrating over ti E [0, At], i = 1 to n yields
P(nf!=1 Fa,) 2 (1 - E)P(Ti ' At, a,, = ai for i = 1 to n).
Summing this over Ci x C2 x ... x Cn yields
P(F) 2 (1 -E)P(ri - At for i = 1 to n).
For each i,
P(i- > At) = exp(-2rp(Ci)At) - exp(-2rAt) = exp(-2 (-log n o = E/nO.

k E E/8fo/

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
932 RICHARD ARRATIA
Since n < no,
P(i ' At for i = 1 to n) > 1 -,
so that
P(F) > (1- E)2 > 1 - 2E.
To show the independence result (55), start with any choices of a, E Ci, and q1, q2 E
(Z/2) . On the event nfl1 Faon c is determined by coin tossing variables ca,(1), i = 1 to n.
The event n Fa, is measurable with respect to 7' (without using any of the information in
W, saying which of the event times in B?' are real and which are phantom). On the event
nfFa,, QIt is measurable with respect to B' and that part of ' which excludes the first
ca, (1) for each pair of ai. Thus, conditional on nFl ga,, Qut and c are independent:
P(c = q2, nfFa, QAt = ql)
= P((ca,(l), i = 1 ton) = q2, QAt = qi InlFa,)P(nfFa,)
= P((cal(1), i = 1 to n) = q2 I nFla,)P(QAt = qI I nFF,)P(nFaF,)
= 2-nP(Qt = qi, nFla,).
Summing the outer equality over all choices of the ai E C, yields (55) and completes the
proof. ?
The counterpart to Lemma 8, for a nearest neighbor random walk p on the line, is easy
because the jump rate (36) for this voter model is constant. In contrast to Lemma 8, where
At = E/8no, in Lemma 8' below At must be taken very large to accommodate either E small
or no large.
LEMMA 8'. Let p be a nearest neighbor random walk on Z1, and let E> 0 and no be
given. There exist m and At > 0 such that, for any n - no and ao < a, < ... < an having
ai- ai- 2 m for i = 1 to n, the joint voter model parity
Q -(| t4a,.,a,) I mod 2, i = 1 to n)

satisfies
E=-(Z/2)" | 2 - P(Qzt = q) I < E.

PROOF. Let Rt be the random walk on (Z/2) n, starting at Qo, and having jumps,
expressed in terms of the standard basis {el, e2, .. .,en),
qua-3- q + ei at rate 1, fori = 1 or n
q -- q + ei + ei+l at rate 2, for i = 1 to n - 1.
Up until time
X n= min-_n inf{t:t '"' = ,
Q and R can be coupled so that Qt = Rt. Take m so large that P(Dk m) = 4 ) < E/(4no), and
thus P(T c At) < E/4. The distribution PRt-l of Rt converges to rn, having mass 2-n at
each point of (Z/2) n. Take At so large that the total variation distancel PR-1 -O ,l1, which
is the same regardless of which pure state R is started in, is less than E/2. Now
IL7Tn-PQ- c II IT -PR I 11 + 1I PR` -k 11 <PE/2 + E/2 =E.

THEOREM 4. Let p be a genuinely multidimensional random walk on Zd or a nearest
neighbor walk on Z1. Let K C Rd be compact and convex, and define sets Kt C Zd by (1),

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
so that
(56) SUpt:-o E((t n Kt) C c < co.
The system rqt of annihilating random walks,

totically the one half thinning of the corresponding system ,t of coalescing random walks,
i.e.
(57) limt,. SACK! P(tt n Kt = A)(>BCA 1 2 1A-J -P(t lKt= B IBtfnKt=A)1) = 0.
PROOF. We show that (57) holds for each choice of c < Xo, uniformly in arbitrary choices
Kt C Zd satisfying (56); the geometric structure of the Kt does not enter into the argument,
except in the case d = 1 where it is necessary to assume that each Kt is an interval.
For the family of voter models (tA , A C Zd) with the standard additive coupling induced
by a substructure 9:
At = nxe-A aXt
introduce abbreviation for the set of individuals whose initial opinions survive until time
t:
St= {x E Zd: Dx )
and for the set of those whose dynasty of followers at time t has odd cardinality:
(58) Sodd = {x E Zd: I a I is odd).

The usual coupling on [0, t] is coalescing random walks at, annihi
(both starting from Zd, all sites occupied) and the family of voter models yields, for all w,
(59) S= t, 7t = St
Fix c < Xo and sets Kt c Zd satisfying (56); in d = 1 also require that the Kt be intervals.
Introduce the abbreviation, for finite A C Zd, t > 0
gt(A) = XBCA 121A1 -2P(St f K = B|St n Kt = A) |.
Since each gt (A) is the total variation distance between two probability measures on the
collection of all subsets of A,
O-gt(A) - 2.
Using the coupling (59), our goal (57) becomes
(60) limtos y,,, P(st n Kt = A)gt(A) = O.

Let E E (0, 1) be given; we will find to, depending on E, p, and c, such that the sum in (60)
is less than 10E when t > to.
Set no = c/E, so that by Chebyshev's inequality and (56),
(61) P(IKtfnStI>no)<E.
In the multidimensional case, let m, depending on no, p, and E, be determined by Lemma

8, and set At = E/(8no). In the nearest neighbor, d = 1 case, let m and At, depending on no
and E, be determined by Lemma 8'. Fix any t > At and set
s = t - At.
For n _ no and any disjoint finite Al, A2, * * , An C Zd satisfying I Ai1 2 m [requir
the d = 1 case, that the Ai as well as UAi be intervals] Lemma 8 or Lemma 8' says
SqE(Z/2)tp t s & up to i = 1 to n) = q mod 2)t1 < 4E.
Since the part of the substructure 9 up to time s is independent of the part of 9 Rafter s, for

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
934 RICHARD ARRATIA
anyAC Kt with IAI =n,A= {xi, ...,x,,),
Zq12`-P((ldxI i= iton) -qItxs=A fori= lton,KtnfS,=A)I<4E.

Average this over all allowable choices for the AI, * , A,, to get:
Eq 2 - P((I xt , i = ito n) q II x I >- m for i = 1 ton, Kt nS = A) I <4E.

[In the d = 1 case each tx, is an interval, so Kt n Ss = A, together with Kt being an interval,
implies that UI=, dSx is an interval.] Using the notation (58), this can be rewritten
(62) BCA | 2 -P(St 2n Kt = B I s n K, = A, I S2 m Vx E A) < 4E.

Consider the events
Et = {I txj - 2m Vx E Kt nsS, Kt nS,= K, n st).

Our final goal will be to show that
(63) P(Et) -1 ast -oo.
To show how (60) follows from (62) and (63), define
Wt = {A C Kt: P (Et Kt n St = A )> 1 -} E.
Choose to so that t 2 to implies P(Et') <2. Since
E2 > P(EtC) = ,AcK, P(St n Kt = A)P(EtcI st n Kt = A)
?2 EACK,:Ag P(St n Kt = A).E = EP(St n Kt o o),

t > to implies
(64) P(St n Kt E Wt) > 1-s
NowforA E Wtwith IAI = n no,
P(KtfnSs=A, ISxI>2mVx-EAIKt nSt=A)>1-E
and (62) together imply
(65) EBcA 2 P(St nKt=BKtfnSt = A) I < 2E + 4E.
[This argument has the form: if a measure t = XAL + (1 -

probability measures, then for the total variation distance from a probability measure v,
lIP- iLII - ILI 11 + (1- X) IIP - sL2 A - M(2) + Ill - /L2 1,

with X <E andi IJ-L2 II < 4E.] Formula (65) may be rewritten: for A E Wt with I A l = n
no,
gt(A) <-6E,
so the sum in (60) may be estimated, for t > to:
,c K, P(St n Kt = A)gt(A) S >A,- eAsI P(St n Kt = A) 6E
+ ?ACK,:A %orAj>P(St n Kt = AP) *A 2
(using (64) and (61)) s 16E + 2E.2 = 10E.

Thus all that remains is to verify (63). Recall that s = t - At; At is fixed as t varies; we
writept = P(O E (t) = P(tx # p) so that condition (56) is: I Kt I pt S c. Define events, for k
- 1, 2, ...
E,,k=f{ljxj-kVxEKtnSs, KtnSs=KtnSt},

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
so that Et Etm D Et,k if k 2 m. To get P(Et') -O 0, write
Etak = UxeK, ({O < I sx < k) U {Ix Ik, tt=4}),

so that by translation invariance and (56),
(66) P(Etc)-C infk:k::m ,P(E t'k) C (C/Pt) infk k::,[P(0 <1| s l< k) +P(| Isl| 2k, t = ?)].
Since the flip rate at any site in the voter model is at most 1,
Pt/Ps =PP(#t 0 4I sx# () ? e1t.

By conditioning on to,
P(I DSI | 2k, Dt = 4) C P(| I | k) SUPA:JAJ:k P(D 5 = 4)) <Ps SUPA:JAJ~k P(IA = ) ).
Thus, continuing with (66),
P(Etc) < cent infk klm[P( IS I < k I Is 3 4)) + supAjAIsk P(kAt = 4))].
Lemma 3 says that the first term above goes to zero as t -X 00, for any fixed k. To see that
the second term can be made arbitrarily small by choosing k large enough, note that the
embedded Markov chain for MA must pass through a sequence Ak, Aki1, *.-, Al of
configurations having I i = i before reaching 4. The jump rate out of Ai is rp (dAt) ' I) A I
- i. For a pure death process which starts at k and jumps from i to i - 1 at rate i, the time
until extinction has mean Z i=i (1/i) and variance Z 1/i2; the probability that this process,
starting at k, is extinct by time At tends to 0 as k increases. A comparison of MA with this
process shows that
SUpA:IAISk P(DA = 4) 0 as k -*oo.
Thus P (EY) -* 0 as t -a oo, concluding this proof. U
COROLLARY 3. Let p be any genuinely multidimensional random walk on Zd, d> 2,

or a nearest neighbor random walk on Zd, d = 1. For each t - 0 there exists versions of
Tqt, and of ft, a one half thinning of it, such that for any compact K C Rd.
P (qt n Kt 7& E) ,t n Kt) O- .
PROOF. When the sum in (57) is c, there exist versions of qt and Edt such that
P(qt n Kt 3# (Jt n Kt) = E.
Let S8 be the sphere of radius n centered at the origin in Rd, and fix an increasing sequence
tn -X00 such that for t 2 tn, with K = Sn,, the sum in (57) is less than 1/n. For t E [t,, tn+1)
select versions of mt and 0-~t such that
P(71t n St- 3. Fit n St <-

n
These are the required versions. E
COROLLARY 4. Let p be a genuinely multidimensional random walk, or a nearest

neighbor random walk on the line. If for some sequence ti - o,
at, t, Ed AL,
then
at,7qt, Ed OI,
where the point process 01i is the one half thinning of the point process it on Rd.

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC
936 RICHARD ARRATIA
PROOF. Write C+(Rd) for the class of all nonnegative continuous functions on Rd
having compact support. The Laplace transform L4 a random measure ju on Rd is defined
by LLf = E (exp(-f f d,)) for f E C+. The one half thinning OI of a random measure is
characterized by
L(,L)f = L,(-log(l - 1/2 (1 -e-f))).
Suppose f E C+ with support contained in a compact set K. Let g = -log(l - 1/2(1 -e-f)),
so that Le, f = LLg. Take versions of -qt and O(t as in Corollary 3. We have
I Lat60lf LI ' I La1z11 f -Leoat f I + Lg - L,4g1.

The first term above is dominated by P(qt n Kt # o f Kt) which goes to zero as t
approaches infinity; since g E C+ the second term goes to zero along the subsequence t4 by
the hypothesis attt, -*d I. Thus La,f f-* Le, f for every f E Cc, so atqt, id et-. E
REFERENCES
[1] ARRATIA, R. (1979). Coalescing Brownian Motions on the Line. Ph.D. thesis, University of
Wisconsin at Madison.
[2] BRAMSON, M. and GRIFFEATH, D. (1979). Renormalizing the 3-dimensional voter model. Ann.
Probability 7 418-432.
[3] BRAMSON, M. and GRIFFEATH, D. (1980). Clustering and dispersion rates for some interacting
particle systems on ZV. Ann. Probability 8 183-213.
[4] BRAMSON, M. and GRIFFEATH, D. (1980b). Asymptotics for interacting particle systems on Zd.
Z. Warsch. verw. Gebiete 53 183-196.
[5] GRIFFEATH, D. (1979). Additive and Cancellative Interacting Particle Systems. Lecture Notes
in Math. 724. Springer, New York.
[6] HARRIS, T. (1977). A correlation inequality for Markov processes in partially ordered state
spaces. Ann Probability 5 451-454.
[7] HOLLEY, R. and STROOCK, D. (1979). Central limit phenomena of various interacting systems.
Ann. of Math. 110 333-393.
[8] KALLENBERG, 0. (1975). Random Measures. Akademie-Verlag, Berlin.
[9] KELLY, F. P. (1977). The asymptotic behavior of an invasion process. J. Appl. Probability 14
584-590.
[10] LIGGETT, T. (1978). Random invariant measures for Markov chains, and independent particle
systems. Z. Wahrsch. verw. Gebiete 45 297-313.
[11] SAWYER, S. (1979). A limit theorem for patch sizes in a selectively-neutral migration model. J.
Appl. Probability 16 482-495.
[12] SUDBURY, A. (1976). The size of the region occupied by one type in an invasion process. J. Appl.
Probability 13 355-356.
DEPARTMENT OF MATHEMATICS
UNIVERSITY OF SOUTHERN CALIFORNIA
LOS ANGELES, CALIFORNIA 90007

129.67.185.115 on Wed, 31 Aug 2022 09:05:18 UTC

Arratia's Rescaling

Uploaded by

Document Information

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

Arratia's Rescaling

Uploaded by

Copyright:

Available Formats

Limiting Point Processes for Rescalings of Coalescing and Annihilating Random Walks

Stable URL: https://www.jstor.org/stable/2243755

Institute of Mathematical Statistics is collaborating with JSTOR to digitize, preserve and

This content downloaded from

LIMITING POINT PROCESSES FOR RESCALINGS OF COALESCING

University of Southern California

Let p(x, y) be an arbitrary random walk on Zd. Let At be the system of

1. Introduction. We consider two interacting particle systems on the d-dimensional

&t E-)d 8,, 'qt E-d So

Here, the convergence in distribution of t to 8, the probability measure on Y which is

P(t nK4())- 0 ast- oo.

choosing at so that the density of particles per unit volume in Rd is always on

(1) pt = P(O E at), at =ptl/,

Bt = {x E Zd: atx E B) = (p /d B) n Zd.

Received August 7, 1980.

This content downloaded from

(2) Elattt n RI = Elt n BtI =PtIBtI -3 m(B) as to- oo.

(3) Vt = Exit, fa&x

Cit6f a-)d I-d as t -) mo.

(4) (Irtllog t)-1 d = 2

p(X Y Ett) cp(XE=- )p(y E=- ) = p2,

(7) P(OE -itlOE -t)=P(ltolisoddll '1>0)-1/2 ast-3 >o.

This content downloaded from

P(OEEit) 1/(2v't) d=1

ZACK, (P(ftnKt=A)>3BCA 12-JAI-P(tnKt=BIftnKt=A)I) >O0 ast-+oo.

P((atqit) IK # (o8at 6) IK) -O 0 as t-* 00,

if at6t ad l (along some sequence) then atqt - A

For the system qt of annihilating simple random walks on

d = 2: (Int/log t)-1/2qt ad 0l2 = Poisson, intensity 1/2

d- 3: (Ydt) /dqt ad 0/Id = Poisson, intensity 1/2.

This content downloaded from

,t = {X: yt(X) >0), ?} t = {x: yt(x) is odd).

The appropriate resealing of yt is the random measure lit on Rd defined by

(8) lOt = ExErtPtYt(X)8atx,

(9) limt a P(yt(x) > - yt(x) > o = e

LW1(f) Eeiff da Ld (f) = e-

L11(f) 3 Ee-fdt - I'd (-log Lo Pf) = e dm

(where Vt and lut are specified in (3) and (8).)

2. Rescaling Coalescing Random Walks

(10) Evt(B) = I atZd n B IPt < m(Bath)pt = m(B I) < m.

lima a supt:o P(vtB > a) = 0,

Ei(B) c lrn sup Evt(B) = m(B).

This content downloaded from

(11) P(x, yE tt) -cp't

(12) P(x, yE t) C P(x E it )P(y E it)

In the special case A = Zd, this becomes

(13) pw n A & , n A & )cpw n A & )fltynA70+).

E= {(B, C):BC CCZd}

= (tX, fZd_,Y,) = (tx, Zd _ by)

E(f(Xt)g(Xt)) > Ef(Xt)Eg(Xt).

f((B, C)) = lBfnA,1, g((B, C)) = lAcC;

(14) lttxn A 0 0, rt nA = ) 'lttxn A 3& )P(rt nA = 0).

This content downloaded from

Replacing the event (tY n A = 4) by its complement yields formula (13).

(15) fr(B, C) = lBnlA#4b gr(B, C) = lAflSrCC.

The theorem on positive correlations yields

(16) PGrx n A 7& 0, rt n A = 0) 2 PGrx n A o O)P(,r n A = 0).

(17) p (-y (x) 2_ m, -Y (y) 2- n) c< p(-yA(x) 2_ M)p(-y

Pq flAl A m, t l nAl n) - P(l | n AlI m)p(l rt n Al n).

fr, m(B, C) = 1IBnAI?:m, grn(B, C) = 1j(Ansr)-CI<n. E

COROLLARY 1. Let vt be the random measure on Rd defined by (3), for an arbitrary

(18) var(vtB) c< E(,vtB).

If, along some sequence 4t - 00, vt, id t