Table Of ContentUniversity of Groningen
Abelian Sandpiles and the Harmonic Model
Schmidt, Klaus; Verbitskiy, Evgeny
Published in:
Communications in Mathematical Physics
DOI:
10.1007/s00220-009-0884-3
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2009
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Schmidt, K., & Verbitskiy, E. (2009). Abelian Sandpiles and the Harmonic Model. Communications in
Mathematical Physics, 292(3), 721-759. https://doi.org/10.1007/s00220-009-0884-3
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the
author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license.
More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne-
amendment.
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the
number of authors shown on this cover page is limited to 10 maximum.
Download date: 17-03-2023
Commun.Math.Phys.292,721–759(2009) Communicationsin
DigitalObjectIdentifier(DOI)10.1007/s00220-009-0884-3 Mathematical
Physics
Abelian Sandpiles and the Harmonic Model
KlausSchmidt1,2,EvgenyVerbitskiy3,4
1 MathematicsInstitute,UniversityofVienna,Nordbergstrasse15,A-1090Vienna,Austria.
E-mail:[email protected]
2 ErwinSchrödingerInstituteforMathematicalPhysics,Boltzmanngasse9,A-1090Vienna,Austria
3 PhilipsResearch,HighTechCampus36(M/S2),5656AE,Eindhoven,TheNetherlands.
E-mail:[email protected]
4 DepartmentofMathematics,UniversityofGroningen,POBox407,9700AK,Groningen,TheNetherlands
Received:15January2009/Accepted:14April2009
Publishedonline:15August2009–©TheAuthor(s)2009.Thisarticleispublishedwithopenaccessat
Springerlink.com
Abstract: We present a construction of an entropy-preserving equivariant surjective
mapfromthed-dimensionalcriticalsandpilemodeltoacertainclosed,shift-invariant
subgroupofTZd (the‘harmonicmodel’).Asimilarmapisconstructedforthedissipative
abelian sandpile model andisusedtoprove uniqueness and theBernoulliproperty of
themeasureofmaximalentropyforthatmodel.
Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 722
1.1 Fourmodels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 722
1.2 Outlineofthepaper . . . . . . . . . . . . . . . . . . . . . . . . . . . . 723
2. APotentialFunctionandits(cid:1)1-Multipliers . . . . . . . . . . . . . . . . . . 723
3. TheHarmonicModel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 733
3.1 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734
3.2 Homoclinicpoints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734
3.3 Symboliccoversoftheharmonicmodel . . . . . . . . . . . . . . . . . 735
3.4 Kernelsofcoveringmaps . . . . . . . . . . . . . . . . . . . . . . . . . 739
4. TheAbelianSandpileModel . . . . . . . . . . . . . . . . . . . . . . . . . 744
5. TheCriticalSandpileModel . . . . . . . . . . . . . . . . . . . . . . . . . . 747
5.1 Surjectivityofthemapsξg: R∞ −→ Xf(d) . . . . . . . . . . . . . . . 747
5.2 Propertiesofthemapsξ , g ∈ I˜ . . . . . . . . . . . . . . . . . . . . . 754
g d
6. TheDissipativeSandpileModel . . . . . . . . . . . . . . . . . . . . . . . . 755
6.1 Thedissipativeharmonicmodel . . . . . . . . . . . . . . . . . . . . . 755
6.2 Thecoveringmapξ(γ): R(∞γ) −→ Xf(d,γ) . . . . . . . . . . . . . . . . 756
7. ConclusionsandFinalRemarks . . . . . . . . . . . . . . . . . . . . . . . . 758
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 759
722 K.Schmidt,E.Verbitskiy
1. Introduction
Foranyintegerd ≥2let
(cid:2) (cid:4)
(cid:1) (cid:1)
1 1 (cid:3)d
h = ··· log 2d−2 cos(2πx ) dx ···dx , (1.1)
d i 1 d
0 0 i=1
h =1.166,h =1.673,etc.Itturnsoutthatford ≥2,h isthetopologicalentropyof
2 3 d
threedifferentd-dimensionalmodelsinmathematicalphysics,probabilitytheory,and
dynamicalsystems.Ford =2,thereisevenafourthmodelwiththesameentropyh .
d
1.1. Fourmodels. Thed-dimensionalabeliansandpilemodelwasintroducedbyBak,
TangandWiesenfeldin[3,4]andattractedalotofattentionafterthediscoveryofthe
AbelianpropertybyDharin[8].Thesetofinfiniteallowedconfigurationsofthesandpile
model isthe shift-invariantsubset R∞ ⊂ {0,...,2d −1}Zd defined in (4.4) and dis-
cussedinSect.4.1In[10],Dharshowedthatthetopologicalentropyoftheshift-action
σR onR∞ isalsogivenby(3.4),whichimpliesthateveryshift-invariantmeasureµ
∞
of maximal entropy on R∞ has entropy (1.1). Shift-invariant measures on R∞ were
studiedinsomedetailbyAthreyaandJaraiin[1,2],JaraiandRedigin[13];however,
thequestionofuniquenessofthemeasureofmaximalentropyisstillunresolved.
Spanningtreesoffinitegraphsareclassicalobjectsincombinatoricsandgraphthe-
ory.In1991,Pemantleinhisseminalpaper[17]addressedthequestionofconstructing
uniformprobabilitymeasuresonthesetT ofinfinitespanningtreesonZd —i.e.,on
d
the set of spanning subgraphs of Zd without loops. This work was continued in 1993
byBurtonandPemantle[5],wheretheauthorsobservedthatthetopologicalentropyof
thesetofallspanningtreesinZd isalsogivenbytheformula (1.1).Anotherproblem
discussedin[5]istheuniquenessoftheshift-invariantmeasureofmaximalentropyon
T (theproofin[5]isnotcomplete,butSheffieldhasrecentlycompletedtheproofin
d
[22].
This coincidence of entropies raised the question about the relation between these
models.Apartialanswertothisquestionwasgivenin1998byR.Solomyakin[24]:she
constructed injective mappings from the set of rooted spanning trees on finite regions
ofZd into Xf(d) suchthattheimagesaresufficientlyseparated.Inparticular,thispro-
videdadirectproofofcoincidenceofthetopologicalentropiesofαf(d) andσTd without
makinguseofformula(1.1).
Indimension2,spanningtreesarerelatednotonlytothesandpilemodels(cf.e.g.,
[19] for a detailed account) and, by [24], to the harmonic model, but also to a dimer
model(moreprecisely,totheevenshift-actiononthetwo-dimensionaldimermodel)by
[5].
However,theconnectionsbetweentheabeliansandpilesandspanningtrees(aswell
asdimersindimension2),arenon-local:theyareobtainedbyrestrictingthemodelsto
finiteregionsinZd (orZ2)andconstructingmapsbetweentheserestrictions,butthese
mapsarenotconsistentasthefiniteregionsincreasetoZd.
Inthispaperwestudytherelationbetweentheinfiniteabeliansandpilemodelsand
thealgebraicdynamicalsystemscalledtheharmonicmodels.Thepurposeofthispaper
istodefineashift-equivariant,surjectivelocalmappingbetweenthesemodels:fromthe
1 Inthephysicsliteratureitismorecustomarytoviewthesandpilemodelasasubsetof{1,...,d}Zd by
adding1toeachcoordinate.
AbelianSandpilesandtheHarmonicModel 723
infinite critical sandpile model R∞ to the harmonic model. Although we are not able
to prove that this mapping is almost one-to-one it has the property that it sends every
shift-invariantmeasureofmaximalentropyonR∞toHaarmeasureonXf(d).Moreover,
itshedssomelightonthesomewhatelusivegroupstructureofR∞.
Firstly,thedualgroupof Xf(d) isthegroup
G = R /(f(d)),
d d
where R = Z[u±,...,u±] is the ring of Laurent polynomials with integer coeffi-
d 1 d
cientsinthevariablesu ,...,u ,and(f(d))istheprincipalidealin R generatedby
(cid:5) 1 d d
f(d) =2d− d (u +u−1).ThegroupG isthecorrectinfiniteanalogueofthegroups
i=1 i i d
of additionoperatorsdefinedonfinitevolumes,see[9,19](cf.Sect.7).
Secondly,themapξ constructedinthispapergivesrisetoanequivalencerelation∼
Id
onR∞with
x ∼ y ⇐⇒ x −y ∈ker(ξ ),
Id
such that R∞/∼ is a compact abelian group. Moreover, R∞/∼, viewed as a dynami-
calsystemunderthenaturalshift-actionofZd,hasthetopologicalentropy(1.1).This
extends the result of [16], obtained in the case of dissipative sandpile model, to the
criticalsandpilemodel.
Finally,wealsoidentifyanalgebraicdynamicalsystemisomorphictothedissipative
sandpilemodel.Thisallowsaneasyextensionoftheresultsin[16]:namely,theunique-
nessofthemeasureofmaximalentropyonthesetofinfiniterecurrentconfigurations
in the dissipative case. Unfortunately, we are not yet able to establish the analogous
uniquenessresultinthecriticalcase.
1.2. Outlineofthepaper. Sect.2investigatescertainmultipliersofthepotentialfunc-
tion(orGreen’sfunction)ofthesimplerandomwalkonZd.InSect.3theseresultsare
usedtodescribethehomoclinicpointsoftheharmonicmodel.Thesepointsarethenused
todefineshift-equivariantmapsfromthespace(cid:1)∞(Zd,Z)ofallboundedd-parameter
sequencesofintegersto Xf(d).InSect.4weintroducethecriticalanddissipativesand-
pilemodels.InSect.5weshowthatthemapsfoundinSect.3sendthecriticalsandpile
modelR∞ onto Xf(d),preservetopologicalentropy,andmapeverymeasureofmaxi-
malentropyonR∞toHaarmeasureontheharmonicmodel.Afterabriefdiscussionof
furtherpropertiesofthesemapsinSubsect.5.2,weturntodissipativesandpilemodels
in Sect. 6 and define an analogous map to another closed, shift-invariant subgroup of
TZd.Themainresultin[16]showsthatthismapisalmostone-to-one,whichimplies
that the measure of maximal entropy on the dissipative sandpile model is unique and
Bernoulli.
2. APotentialFunctionandits(cid:1)1-Multipliers
Letd ≥1.Foreveryi =1,...,dwewritee(i) =(0,...,0,1,0,...,0)fortheithunit
vectorinZd,andweset0=(0,...,0)∈Zd.
WeidentifythecartesianproductW =RZd withthesetofformalrealpowerseries
d
inthevariablesu±1,...,u±1 byviewingeachw =(w )∈ W asthepowerseries
1 d (cid:3) n d
w un (2.1)
n
n∈Zd
724 K.Schmidt,E.Verbitskiy
with w ∈ R and un = un2···und for every n = (n ,...,n ) ∈ Zd. The involution
w (cid:9)→wn∗onW isdefined1by d 1 d
d
wn∗ =w−n, n∈Zd. (2.2)
For E ⊂ Zd we denote by π : W −→ RE the projection onto the coordinates
E d
in E.
Forevery p ≥1weregard(cid:1)p(Zd)asthesetofallw ∈ W with
d
⎛ ⎞
1/p
(cid:3)
(cid:11)w(cid:11) =⎝ |w |p⎠ <∞.
p n
n∈Zd
Similarlyweview(cid:1)∞(Zd)asthesetofallboundedelementsinW ,equippedwiththe
d
supremumnorm(cid:11)·(cid:11)∞.Finallywedenoteby Rd =Z[u±11,...,u±d1]⊂(cid:1)1(Zd)⊂ Wd
theringofLaurentpolynomials(cid:5)withintegercoefficients.Everyhinanyofthesespaces
willbewrittenash =(hn)= n∈Zd hnun withhn ∈R(resp.hn ∈Zforh ∈ Rd).
Themap(m,w)(cid:9)→um·wwith(um·w)n =wn−misaZd-actionbyautomorphisms
oftheadditivegroupW whichextendslinearlytoan R -actiononW givenby
d d d
(cid:3)
h·w = h un·w (2.3)
n
n∈Zd
foreveryh ∈ R andw ∈ W .Ifwalsoliesin R thisdefinitionisconsistentwiththe
d d d
usualproductin R .
d
For the following discussion we assume that d ≥ 2 and consider the irreducible
Laurentpolynomial
(cid:3)d
f(d) =2d− (u +u−1)∈ R . (2.4)
i i d
i=1
Theequation
f(d)·w =1 (2.5)
with w ∈ W admits a multitude of solutions.2 However, there is a distinguished (or
d
fundamental)solutionw(d)of (2.5)whichhasadeepprobabilisticmeaning:itisacer-
tainmultipleofthelatticeGreen’sfunctionofthesymmetricnearest-neighbourrandom
walkonZd (cf.[6,12,25,27]).
Definition2.1. For every n = (n ,...,n ) ∈ Zd and t = (t ,...,t ) ∈ Td we set
(cid:5) 1 d 1 d
(cid:12)n,t(cid:13)= d n t ∈T.Wedenoteby
j=1 j j
(cid:3) (cid:3)d
F(d)(t)= f(d)e2πi(cid:12)n,t(cid:13) =2d−2· cos(2πt ), t =(t ,...,t )∈Td, (2.6)
n j 1 d
n∈Zd j=1
theFouriertransformof f(d).
2 UndertheobviousembeddingofRd (cid:7)→(cid:1)∞(Zd,Z),theconstantpolynomial1∈Rdcorrespondstothe
elementδ(0)∈(cid:1)∞(Zd,Z)givenby (cid:10)
δ(0)= 1 if n=0,
n 0 otherwise.
AbelianSandpilesandtheHarmonicModel 725
(1)Ford =2,
(cid:1)
e−2πi(cid:12)n,t(cid:13)−1
w(2) := dt forevery n∈Z2.
n T2 F(2)(t)
(2)Ford ≥3,
(cid:1)
e−2πi(cid:12)n,t(cid:13)
w(d) := dt forevery n∈Zd.
n Td F(d)(t)
The difference in these definitions for d = 2 and d > 2 is a consequence of the fact
thatthesimplerandomwalkonZ2 recurrent,whileonhigherdimensionallatticesitis
transient.
Theorem2.2.([6,12,25,27])Wewrite(cid:11)·(cid:11)fortheEuclideannormonZd.
(i)Foreveryd ≥2,w(d)satisfies(2.5).
(ii)Ford =2,
⎧
⎨ 0 if n=0,
wn(2) =⎩− 1 log(cid:11)n(cid:11)−κ −c (cid:11)n1(cid:11)4(n41+n42)−34 +O((cid:11)n(cid:11)−4) if n(cid:14)=0, (2.7)
8π 2 2 (cid:11)n(cid:11)2
whereκ > 0andc > 0.Inparticular,w(2) = 0andw(2) < 0foralln (cid:14)= 0.
2 2 0 n
Moreover,
(cid:3)∞
4·w(2) = (P(X =n|X =0)−P(X =0|X =0)),
n k 0 k 0
k=1
where(X )isthesymmetricnearest-neighbourrandomwalkonZ2.
k
(iii)Ford ≥3,
(cid:5)
1 d n4− 3
(cid:11)n(cid:11)d−2w(d) =κ +c (cid:11)n(cid:11)4 i=1 i d+2 +O((cid:11)n(cid:11)−4) (2.8)
n d d (cid:11)n(cid:11)2
as(cid:11)n(cid:11)→∞,whereκ >0,c >0.Moreover,
d d
(cid:3)∞
2d·w(d) = P(X =n|X =0)>0 forevery n∈Zd,
n k 0
k=0
where(X )isagainthesymmetricnearest-neighbourrandomwalkonZd.
k
Definition2.3.Letw(d) ∈ W bethepointappearinginDefinition2.1.Weset
d
(cid:14) (cid:15)
I = g ∈ R :g·w(d) ∈(cid:1)1(Zd) ⊃(f(d)), (2.9)
d d
where(f(d)) = f(d)· Rd istheprincipalidealgeneratedby f(d).Sincewn(d) = w−(dn)
foreveryn∈Zd itisclearthat I = I∗ ={g∗ :g ∈ I }.
d d d
726 K.Schmidt,E.Verbitskiy
Theorem2.4.Theideal I isoftheform
d
I =(f(d))+I3, (2.10)
d d
where
I ={h ∈ R :h(1)=0}=(1−u )· R +···+(1−u )· R (2.11)
d d 1 d d d
with1=(1,...,1).
FortheproofofTheorem2.4weneedseverallemmas.Weset
J =(f(d))+I3 ⊂ R . (2.12)
d d d
(cid:5)
Lemma2.5.Let g = k∈Zd gkuk ∈ Rd. Then g ∈ Jd if and only if it satisfies the
followingconditions(2.13)–(2.16).
(cid:3)
g =0, (2.13)
k
(cid:3) k∈Zd
g k =0 for i =1,...,d, (2.14)
k i
k=((cid:3)k1,...,kd)∈Zd
g k k =0 for 1≤i (cid:14)= j ≤d, (2.15)
k i j
(cid:3)k=(k1,...,kd)∈Zd
g (k2−k2)=0 for 1≤i (cid:14)= j ≤d. (2.16)
k i j
k=(k1,...,kd)∈Zd
Proof. Condition(2.13)isequivalenttosayingthatg ∈I .Inconjunctionwith(2.13),
d
(2.14)isequivalenttosayingthatg ∈I2:indeed,ifg ∈I ,thenitisoftheform
d d
(cid:3)d
g = (1−u )·a (2.17)
i i
i=1
witha ∈ R fori =1,...,d.Then
i d
∂g = (cid:3) g k ·uk1···ukj−1···ukd =−a +(cid:3)d (1−u )· ∂ai ,
∂u k j 1 j d j i ∂u
j k=(k1,...,kd)∈Zd i=1 j
and ∂g (1)=0ifandonlyifa ∈I .
∂uj j d
Ifg ∈I isoftheform(2.17)andsatisfies(2.14)weset
d
(cid:3)d
aj = (1−ui)·bi,j (2.18)
i=1
withbi,j ∈ Rd.Condition(2.15)issatisfiedifandonlyif
∂2g ∂a ∂a
∂u ∂u (1)=−∂ui − ∂uj =bi,j(1)+bj,i(1)=0
i j j i
for1≤i (cid:14)= j ≤d.
AbelianSandpilesandtheHarmonicModel 727
Finally,if g satisfies(2.13)–(2.14)andisoftheform (2.17)–(2.18)withbi,j ∈ Rd
foralli, j,then(2.16)isequivalenttotheexistenceofaconstantc∈Rwith
(cid:3) ∂a
gkki2 =−2∂ui(1)=2bi,i(1)=c
k=(k1,...,kd)∈Zd i
fori =1,...,d.
The last equation shows that bi,i −b1,1 ∈ Id for i = 2,...,d. By combining all
theseobservationswehaveprovedthatgsatisfies(2.13)–(2.16)ifandonlyifitisofthe
form
(cid:3)d
g =h · (1−u )2+h (2.19)
1 i 2
i=1
with c ∈ Z, h ∈ R and h ∈ I3. The set of all such g ∈ R is an ideal which we
1 d 2 (cid:5)d d
denoteby J˜.Clearly,I3 ⊂ J˜and d (1−u )2 ∈ J˜.Since(1−u )2·(1−u−1)∈I3
d i=1 i i i d
fori =1,...,d aswell,weconcludethat
(cid:3)d (cid:3)d
f(d) = (1−u )2− (1−u−1)·(1−u )2 ∈ J˜. (2.20)
i i i
i=1 i=1
Thisshowsthat J˜⊂ J ,andthereverseinclusionalsofollowsfrom(2.20)and(2.19).
d
(cid:17)(cid:18)
Lemma2.6. I ⊂ J .
d d
Proof. Weassumethatg ∈ Id an(cid:5)dsetv = g·w(d).Inordertoverify(2.13)weargue
bycontradictionandassumethat g (cid:14)=0.Ifd =2then
k k
(cid:5)
g
v =− k k log(cid:11)n(cid:11)+l.o.t.,
n 2π
forlarge(cid:11)n(cid:11).Ifd ≥3,then
(cid:5)
κ g
v = d k k +l.o.t.
n (cid:11)n(cid:11)d−2
forlarge(cid:11)n(cid:11).Inbothcasesitisevidentthatv (cid:14)∈(cid:1)1(Zd).
Bytaking(2.13)intoaccountonegetsthat,foreveryd ≥2,
(cid:3)
v =(g·w(d)) = g w(d)
n n k n−k
(cid:1)k (cid:5)
g e2πi(cid:12)k,t(cid:13)
= e−2πi(cid:12)n,t(cid:13) k(cid:5)k dt.
Td 2d−2 dj=1cos(2πtj)
Hencev =(v )isthesequenceofFouriercoefficientsofthefunction
n
(cid:5)
g e2πi(cid:12)k,t(cid:13)
H(t)= k(cid:5)k .
2d−2 d cos(2πt )
j=1 j
728 K.Schmidt,E.Verbitskiy
Ifv ∈(cid:1)1(Zd),then H mustbeacontinuousfun(cid:5)ctiononTd.Sincet =0istheonly
zeroof F(d) onTd (cf.(2.6)),thenumeratorG = g e2πi(cid:12)k,·(cid:13) mustcompensatefor
k k
thissingularity.ConsidertheTaylorseriesexpansionofG att =0:
(cid:3) (cid:3)d (cid:3) (cid:3)d (cid:3) (cid:3) (cid:3)
G(t)= g +2πi t g k −2π2 t2 g k2−4π2 t t g k k
k j k j j k j i j k i j
k j=1 k j=1 k i(cid:14)=j k
+h.o.t.
TheTaylorseriesexpansionof F(d)att =0isgivenby
(cid:3)d
F(d)(t)=4π2 t2+h.o.t.
j
j=1
Supposethat
(cid:5) (cid:5) (cid:5)
h(t)= a0+ dj=1bjtj + dj=1cjt2j + i(cid:14)=jdi,jtitj +h.o.t
t2+···+t2+h.o.t
1 d
iscontinuousatt =0.Then
a =0, b =0 forall j, c =c forall j, d =0 foralli (cid:14)= j,
0 j j ij
andforsomeconstantc.Ifanyoftheseconditionsisviolated,thenoneeasilyproduces
examplesofsequencest(m) → 0asm → ∞withdistinctlimitslimm→∞h(t(m)).By
applyingthisto H weobtain(2.13)–(2.16),sothatg ∈ J byLemma2.5. (cid:17)(cid:18)
d
ToestablishtheinclusionJ ⊆ I ,wehavetoshowthatforanyg ∈ J ,g·u ∈(cid:1)1(Zd),
d d d
whereu ∈ W oftheform
d
(cid:5)
ω = id=1ni4, or ω = 1 with γ ≥d−2.
n (cid:11)n(cid:11)d+4 n (cid:11)n(cid:11)γ
Ford =2,wealsohavetotreatthecaseω =log(cid:11)n(cid:11).
n
Theseresultsareobtainedinthefollowingthreelemmas.
Lemma2.7.Supposethatd ≥2andthatω∈ W isgivenby
d
(cid:10)
0 if n=0,
ω = (cid:5)
n id=1ni4 if n(cid:14)=0.
(cid:11)n(cid:11)d+4
Ifg ∈ R satisfies(2.13),theng·ω∈(cid:1)1(Zd).
d
Proof. Let M =max{(cid:11)k(cid:11):g (cid:14)=0},andsupposethat(cid:11)n(cid:11)> M.Then
k
(cid:5) (cid:5)
(g·ω) =(cid:3)g id=1(ni −ki)4 =(cid:3)g id=1ni4+O((cid:11)n(cid:11)3)
n k (cid:11)n−k(cid:11)d+4 k(cid:11)n(cid:11)d+4(1+O((cid:11)n(cid:11)−1))
(cid:5)k (cid:2) (cid:4) (cid:16) k (cid:17) (cid:16) (cid:17)
= id=1ni4 (cid:3)g +O 1 =O 1 .
(cid:11)n(cid:11)d+4 k (cid:11)n(cid:11)d+1 (cid:11)n(cid:11)d+1
k
(cid:5)
Therefore, |(g·ω) |<∞. (cid:17)(cid:18)
n n
AbelianSandpilesandtheHarmonicModel 729
For the reverse inclusion J ⊂ I we need different arguments for d = 2 and for
d d
d ≥3.Westartwiththecased =2.
(cid:5)
Lemma2.8.Suppose that g = k∈Z2gkuk ∈ R2 satisfies (2.13). We set S+ =
{k:gk >0}and S− ={k:gk <0}.Put
(cid:3) (cid:3)
M =2 g =2 |g |
g k k
k∈S+ k∈S−
anddefinetwopolynomialsinthevariables(n ,n ):
1 2
(cid:18) (cid:19) (cid:20) (cid:18)
P (n ,n )= (n −k )2+(n −k )2 gk = (cid:11)n−k(cid:11)2gk,
+ 1 2 1 1 2 2
k(cid:18)∈S+ (cid:19) (cid:20) k∈(cid:18)S+ (2.21)
P−(n1,n2)= (n1−k1)2+(n2−k2)2 |gk| = (cid:11)n−k(cid:11)2|gk|.
k∈S− k∈S−
Letmg bethedegreeof P = P+− P−.If
M −m ≥3, (2.22)
g g
theng·ω∈(cid:1)1(Z2),where
(cid:10)
0 if n=(0,0),
ω =
n log(cid:11)n(cid:11) if n(cid:14)=(0,0).
(cid:5)
Proof. Since k∈Z2gk =0by(2.13), Mg =degP+ =degP−and
mg =degP <max(degP+,degP−)= Mg.
Letv =g·ω.Hence,forallnwith(cid:11)n(cid:11)>max{(cid:11)k(cid:11):k∈ S+∪S−},onehas
(cid:21) (cid:21) (cid:21) (cid:16) (cid:17)(cid:21)
|(g·ω)n|= 21(cid:21)(cid:21)(cid:21)log PP−+((nn11,,nn22))(cid:21)(cid:21)(cid:21)= 21(cid:21)(cid:21)(cid:21)log 1+ P+(n1,Pn−2)(n−1,Pn−2()n1,n2) (cid:21)(cid:21)(cid:21).
ThereexistconstantsC,N suchthat
(cid:21) (cid:21)
(cid:21)(cid:21)(cid:21)P+(n1,n2)− P−(n1,n2)(cid:21)(cid:21)(cid:21)≤C(cid:11)n(cid:11)mg = C < 1
P−(n1,n2) (cid:11)n(cid:11)Mg (cid:11)n(cid:11)Mg−mg 2
for(cid:11)n(cid:11)≥ N.HencewecanfindanotherconstantC˜ suchthat
˜
C
|(g·ω) |≤
n (cid:11)n(cid:11)Mg−mg
forallsufficientlylarge(cid:11)n(cid:11).SinceM −m ≥3,wefinallyconcludethatg·ω∈(cid:1)1(Z2).
g g
(cid:17)(cid:18)
Lemma2.9.Supposethatg ∈ J (cf.(2.13)–(2.16)),andthatω∈ W isgivenby
d d
(cid:10)
0 if n=0,
ω =
n 1 if n(cid:14)=0,
(cid:11)n(cid:11)γ
forsomeintegerγ ≥d−2.Theng·ω∈(cid:1)1(Zd).
Description:Publisher's PDF, also known as Version of record abelian sandpile model and is used to prove uniqueness and the Bernoulli property of.