Béatrie de Tilière 1
September11,2014
1
LaboratoiredeProbabilitésetModèlesAléatoires,UniversitéPierreetMarieCurie,4plaeJussieu,
F-75005Paris. beatrie.de_tiliereupm.fr. Thesenotes were writtenin partwhileatInstitut de
Mathématiques,UniversitédeNeuhâtel,RueEmile-Argand11,CH-2007Neuhâtel. Supportedinpart
bytheSwissNationalFoundationgrant200020-120218.
1 Introdution 5
1.1 Statistial mehanis and 2-dimensional models . . . 5
1.2 Thedimermodel . . . 7
2 Denitions and founding results 11 2.1 Dimermodel andtiling model . . . 11
2.2 Energy ofongurations and Boltzmann measure . . . 12
2.3 Expliitomputations . . . 13
2.3.1 Partition funtion formula . . . 13
2.3.2 Boltzmann measureformula . . . 16
2.3.3 Expliitexample . . . 17
2.4 Geometri interpretationof lozenge tilings . . . 18
3 Dimer model on innite periodi bipartite graphs 21 3.1 Height funtion . . . 22
3.2 Partition funtion,harateristi polynomial, freeenergy . . . 25
3.2.1 Kasteleyn matrix . . . 25
3.2.2 Charateristi polynomial . . . 26
3.2.3 Freeenergy . . . 30
3.3 Gibbsmeasures . . . 31
3.3.1 Limitof Boltzmann measures . . . 31
3.3.2 ErgodiGibbsmeasures . . . 32
3.3.3 Newtonpolygonand available slopes . . . 32
3.3.4 Surfaetension . . . 34
3.3.5 Construting Gibbsmeasures . . . 34
3.4 Phases ofthe model . . . 36
3.4.1 Amoebas, Harnak urvesand Ronkin funtion . . . 36
3.4.2 Surfaetension revisited . . . 38
3.4.3 Phases ofthedimermodel . . . 39
3.5 Flutuations ofthe height funtion . . . 41
3.5.2 Gaussian freeeldof theplane . . . 42
3.5.3 Convergene of the height funtion to aGaussian freeeld . . . 43
Introdution
1.1 Statistial mehanis and 2-dimensional models
Statistial mehanis isthe appliation of probability theory,whihinludes mathematial tools
fordealingwith largepopulations,totheeldofmehanis,whihisonernedwiththemotion
ofpartilesor objetswhensubjetedtoafore. Statistial mehanisprovidesaframework for
relating the mirosopi properties of individual atoms and moleules to the marosopi bulk
propertiesof materials thatan be observed ineverydaylife (soure: `Wikipedia').
In otherwords, statistial mehanis aimsat studyinglarge sale propertiesof physis system,
based on probabilisti models desribing mirosopi interations between omponents of the
system. Statistial mehanis isalso knownasstatistial physis.
Itisapriorinaturaltointrodue3-dimensionalgraphsinordertoauratelymodelthemoleular
struture of a material as for example a piee of iron, a porous material or water. Sine the
3
-dimensional version of many models turns out to be hardly tratable, muh eort has been put into thestudy of their2
-dimensional ounterpart. The latter have been shown to exhibit rih, omplexand fasinating behaviors. Here areafew examples.•
Perolation. This model desribes the ow of a liquid through a porous material. The system onsidered is a square grid representing the moleular struture of the material.Eahbondofthegridiseitheropenwithprobability
p
,orlosed withprobability1 − p
,and bondsare assumedto behave independently fromeah other. The set ofopen bonds
ina given ongurationrepresentsthe partof thematerial wetted by theliquid,and the
mainissueaddressedistheexisteneandpropertiesofinnitelusters ofopenedges. The
behavior of the system depends on the parameter
p
: whenp = 0
, all edges are losed,thereisnoinnitelusterandtheliquidannotowthroughthematerial;when
p = 1
,alledges areopen andthere is a unique inniteluster llingthewhole grid. One an show
thatthere isa spei valueof theparameter
p
,known asritialp
,equal to1/2
for thesquaregrid,belowwhihtheprobabilityofhavinganinnitelusterofopenedgesis0,and
abovewhihthe probabilityofitexistingandbeingunique is
1
. Onesaysthatthesystemundergoesa phase transition at
p = 1/2
. Referenes[Kes82 ,Gri99 , BR06, W+
07,Wer09℄arebooks or leturenotesgiving an overviewof perolation theory.
•
TheIsingmodel. Thesystemonsideredisamagnetmadeofpartilesrestritedtostayonagrid. Eahpartilehasaspinwhihpointseitherupor down (spin
± 1
). Eahong-uration
σ
ofspinsonthewholegridhasanenergyE (σ)
,whihisthesumofaninterationFigure1.1: An inniteluster ofopen edges, when
p = 1 2. Courtesyof V.Beara.
energy between pairs of neighboring spins, and of an interation energy of spins with an
external magneti eld. The probabilityof a onguration
σ
is proportional toe − kT 1 E (σ),
where
k
is the Boltzmannonstant, andT
is theexternal temperature. When there isno magneti eldand thetemperature is lose to0
,spins tendto align withtheir neighborsand a typial onguration onsists of all
+1
or all− 1
. When the temperature is very high,all ongurationshave thesame probability ofourringanda typialongurationonsistsofamixture of
+1
and− 1
. Again,there isaritial temperatureT c at whihthe
Isingmodel undergoesaphase transitionbetween theordered and disorderedphase. The
literature on the Ising model is huge, as an introdutory reading we would suggest the
bookbyBaxter[Bax89 ℄,theonebyMCoyandWu[MW73℄,theleturenotesbyVelenik
[Vel ℄ and referenestherein.
Figure1.2: An Isingonguration, when
1
T = 0.9. Courtesyof V.Beara.
These two examples illustrate some of the prinipal hallenges of
2
-dimensional statistial me- hanis,whih are:•
Find theritialparameters of themodels.•
Understandthe behavior ofthe model inthesub-ritial andsuper-ritial regimes.•
Understandthebehaviorofthesystematritiality. Critialsystemsexhibitsurprisingfea-tures,andarebelievedto be universalinthesalinglimit, i.e. independent ofthespei
featuresofthe lattie on whih themodelis dened. Verypreise preditionswereestab-
lished by physiistsin the last 30-50 years, in partiular by Nienhuis, Cardy, Duplantier
and many others. On the mathematis side, a huge step forward was the introdution
of the Shramm-Loewner evolution by Shramm in [Sh00℄, a proess onjetured to de-
sribe the limiting behavior of well hosen observables of ritial models. Manyof these
onjetures weresolved inthe following years,inpartiular byLawler, Shramm,Werner
[LSW04 ℄ andSmirnov [Smi10 ℄,Chelkak-Smirnov [CS12 ℄. The importane of these results
was aknowledged with the two Fields medals awarded to Werner (2006) and Smirnov
(2010). Interesting ollaborations between thephysis andmathematis ommunities are
emerging, withfor example thework ofDuplantier andSheeld [DS11 ℄.
The general framework for statistial mehanis is the following. Consider an objet
G
(most often a graph) representing the physial system, and dene all possible ongurations of thesystem. To every onguration
σ
, assignan energyE (σ)
, thentheprobability of ourrene of theongurationσ
isgiven bytheBoltzmann measureµ
:µ(σ) = e −E (σ) Z ( G ) .
Note thatthe energy is oftenmultiplied by aparameter representing theinverse external tem-
perature. Thedenominator
Z ( G )
isthenormalizing onstant, knownasthepartition funtion:Z ( G ) = X
σ
e −E (σ) .
When the systemis innite, the above denition does not hold, but we do not want to enter
into these onsiderations here.
Thepartitionfuntionisoneofthekeyobjetsofstatistialmehanis. Indeeditenodesmuh
of the marosopi behavior of the system. Hene, its omputation is the rst question one
addresses when studying suh amodel. It turns out thatthere are very fewmodels where this
omputation an be done exatly. Having a losed form for the partition funtion opens the
waytondingmanyexat results,andto having avery deepunderstanding ofthemarosopi
behaviorof the system.
Twofamousexamplesarethe
2
-dimensionalIsingmodel,wheretheomputationofthepartition funtion is due to Onsager [Ons44 ℄, and the dimer model where it is due to Kasteleyn [Kas61 ,Kas67 ℄,andindependently toTemperleyandFisher[TF61 ℄. Thedimermodelisthemaintopi
of theseletures and isdened inthe next setion.
1.2 The dimer model
The dimer model was introdued in the physis and hemists ommunities to represent the
adsorption of di-atomi moleules on the surfae of a rystal. It is part of a larger family of
modelsdesribingtheadsorptionofmoleulesofdierentsizesonalattie. Itwasrstmentioned
inapaperbyFowlerandRushbrooke[FR37℄in1937. Asmentionedintheprevioussetion,the
rst major breakthrough inthe study of the dimer model is the omputation of the partition
funtion byKasteleyn [Kas61 ,Kas67 ℄and independently byTemperley andFisher [TF61 ℄.
Itisinteresting to observe thatfora longtime, thephysis andmathematis ommunities were
unawareof their respetive advanes. Mathematiians studied related questionsasfor example
geometri andombinatorial propertiesoftilings ofregionsoftheplanebydominoesorrhombi.
To the best of our knowledge, the latter problem was rst introdued in a paper by David
and Tomei [DT89 ℄. A major breakthrough was ahieved inthe paper[Thu90℄Thurston,where
theauthor interprets rhombus tilings as
2
-dimensional interfaes ina3
-dimensional spae. An example ofrhombus tilingis given inFigure1.3.Figure1.3: Rhombus tiling. Courtesyof R.Kenyon.
Inthelate
90
'sandearly00
's,alotofprogressesweremadeinunderstandingthemodel,seethe papersofKenyon[Ken97 , Ken00 ℄, Cohn-Kenyon-Propp [CKP01℄, Kuperberg[Kup98 ℄, Kenyon-Propp-Wilson [KPW00 ℄. In
2006
Kenyon-Okounkov-Sheeld [KOS06 ℄, followed by Kenyon- Okounkov [KO06 , KO07 ℄ wrote breakthrough papers, whih give a full understanding of themodel on innite, periodi, bipartite graphs. Suh deep understanding of phenomena isa real
treasurein statistialmehanis.
My goal for these letures is to present the results of Kasteleyn, Temperley and Fisher, of
Thurston, and of the paper of Kenyon, Okounkov and Sheeld. As you will see, the dimer
model has ramiations to many elds of mathematis: probability, geometry, ombinatoris,
analysis, algebraigeometry. Iwilltryto beasthoroughaspossible,but ofoursesome results
addressing the eld of algebrai geometry reah the limit of my knowledge, so thatI will only
Inspiration for these notes omes in large parts from the letures given by R. Kenyon on the
subjet [Ken04,Ken℄. The main otherreferenes are[Kas67 ,Thu90,KOS06 , KO06 ℄.
Definitions and founding results
2.1 Dimer model and tiling model
Inthissetion, wedenethedimermodel andtheequivalenttiling model,usingtheterminology
of statistial mehanis. The system onsidered is a graph
G = ( V , E )
satisfying the following:itis planar,simple(no loopsandno multiple edges),nite orinnite.
Congurationsofthesystemareperfetmathingsofthegraph
G
. Aperfetmathing isasubset of edgeswhih overseah vertexexatlyone. Inthephysisliterature, perfet mathings arealso referredto asdimer ongurations, adimer being a di-atomi moleule represented by an
edge of theperfet mathing. Letus denoteby
M ( G )
thesetof all dimerongurations ofthe graphG
.Figure 2.1gives an example of a dimer onguration when thegraph
G
is a nite subgraph of thehoneyomb lattieH
.Figure2.1: Dimeronguration ofa subgraph ofthehoneyomb lattie.
In order to dene theequivalent tiling model, we onsider a planar embedding of the graph
G
, and suppose that it is simply onneted, i.e. that it is the one-skeleton of a simply onnetedunion of faes. From now on, when we speak of a planar graph
G
, we atually mean a graph witha partiular planarembedding.The tiling model is dened on the dual graph
G ∗
of
G
. An embedding of the dualgraphG ∗
is
obtained by assigning a vertex to every faeof
G
and joining two verties ofG ∗ by an edge if
andonly ifthe orrespondingfaesofG
areadjaent. Thedualgraph willalso be thoughtofas
anembeddedgraph. Whenthegraphisnite,wetakeaslightlydierentdenitionofthedual:
we take
G
to be the dual ofG ∗
and remove thevertex orresponding to the outer fae, aswell
asedges onnetedto it, seeFigure 2.2.
A tile of
G ∗ is a polygon onsisting of two adjaent inner faes of G ∗ glued together. A tiling
of G ∗
G ∗
is a overing of the graph
G ∗
with tiles, suh that there are no holes and no overlaps.
Figure2.2 gives an exampleof a tilingof a nite subgraphof thetriangular lattie
T
,thedual graph of the honeyomb lattie. Tiles of the triangular lattie are60 ◦-rhombi, and are also
knownaslozenges or alissons.
Figure 2.2: Dual graph of a nite subgraph of the honeyomb lattie (left). Tiling of this
subgraph (right).
Another lassial example is thetiling modelon the graph
Z 2,the dualof thegraph Z 2. Tiles
aremade oftwo adjaent squares,and areknownasdominoes.
Dimer ongurations of the graph
G
are in bijetion with tilings of the graphG ∗
through the
followingorrespondene, seealsoFigure2.3: dimeredgesofperfetmathingsonnetpairs of
adjaent faesforming tilesofthe tiling. It isan easy exeriseto provethat thisindeed denes
abijetion.
Figure2.3: Bijetion between dimerongurations of thegraph
G
and tilings ofthe graphG ∗.
2.2 Energy of onfigurations and Boltzmann measure
Welet
G
beaplanar,simplegraph. Inthissetion, andfortheremainder ofChapter2,wetakeG
to be nite. Supposethatedges areassigned apositive weight funtionν
,that iseveryedgee
ofG
hasweightν( e )
.The energy of a dimeronguration
M
ofG
,isE (M) = − P
e ∈ M log ν( e ).The weight ν(M )
of
adimer onguration
M
ofG
,isexponential of minus its energy:ν(M) = e −E (M) = Y
e ∈ M
ν( e ).
Note that by the orrespondene between dimer ongurations and tilings, the funtion
ν
anbeseenasweighting tilesof
G ∗
,
ν(M )
isthenthe weight of thetiling orrespondingtoM
.µ
is aprobabilitymeasure onthe set ofdimerongurationsM ( G )
,dened by:∀ M ∈ M ( G ), µ(M) = e −E (M )
Z( G ) = ν(M ) Z( G ) .
The term
Z( G )
is thenormalizing onstant known asthepartition funtion. It isthe weighted sumof dimerongurations, thatis,Z ( G ) = X
M ∈M ( G )
ν (M ).
When
ν ≡ 1
,the partition funtion ounts the number ofdimer ongurations of thegraphG
, or equivalently thenumberoftilings ofthe graphG ∗,andtheBoltzmann measureissimplythe uniform measureon the setof dimerongurations.
Whenanalyzingamodelofstatistialmehanis,therstquestionaddressedisthatofomput-
ingthefree energy, whih isminus theexponential growth rateofthepartitionfuntion,asthe
size of the graphinreases. The most natural way ofattaining this goal, ifthe modelpermits,
istoobtain anexpliitexpressionforthepartitionfuntion. Reallthatthedimermodelisone
of the rare
2
-dimensional models where a losed formula an be obtained. This is the topi of thenext setion.2.3 Expliit omputations
The expliit omputation of the partition funtion is due to Kasteleyn [Kas61 , Kas67 ℄ and
independentlytoTemperleyandFisher[TF61 ℄. AproofofthisresultisprovidedinSetion2.3.1,
inthe asewhere the underlying graph
G
isbipartite. Thisisone ofthefoundingresults ofthe dimer model, paving the way to obtaining other expliit expressions as for example Kenyon'slosedformulafortheBoltzmannmeasure[Ken97 ℄,seeSetion2.3.2. InSetion2.3.3,weprovide
an exampleof omputation ofthepartition funtion and oftheBoltzmann measure.
2.3.1 Partition funtion formula
We restrit ourselves to the asewhere thegraph
G
is bipartite, theproof inthe non-bipartite ase issimilar inspirit although a little moreinvolved. The simpliationin thebipartite aseisdue to Perus [Per69℄.
Agraph
G = ( V , E )
isbipartite ifthesetofvertiesV
anbesplitintotwosubsetsW ∪ B
,whereW
denotes white verties,B
blak ones, and verties inW
are only adjaent to verties inB
. We suppose that| W | = | B | = n
, for otherwise there are no perfet mathings of the graphG
; indeeda dimeredgealwaysovers ablakanda whitevertex.Labelthewhiteverties
w 1 , . . . , w nandtheblakonesb 1 , . . . , b n,andsupposethatedgesofG
are
oriented. Thehoie oforientation will be speiedlater inthe proof. Then theorresponding
G
are oriented. Thehoie oforientation will be speiedlater inthe proof. Then theorrespondingoriented, weighted, adjaeny matrix is the
n × n
matrixK
whose lines are indexed by whiteverties,whose olumns areindexedby blakones, andwhose entry
K( w i , b j )
is:K( w i , b j ) =
ν( w i b j )
ifw i ∼ b j,andw i → b j
− ν( w i b j )
ifw i ∼ b j,andw i ← b j
0
ifthevertiesw i and b j arenot adjaent.
By denition,the determinant of the matrix
K
is:det(K) = X
σ ∈S n
sgn
(σ)K( w 1 , b σ(1) ) . . . K( w n , b σ(n) ),
where
S n is the set of permutations of n
elements. Let us rst observe that eah non-zero
term intheexpansionof
det(K)
orrespondsto theweightof adimeronguration, uptosign.Thus, the determinant of
K
seems to be the appropriate objet for omputing the partition funtion, theonly problem beingthat not all terms may be ounted withthe same sign. Notethat reversing the orientation of an edge
w i b j hanges thesign of K( w i , b j )
. The remainder of
theproofonsistsinhoosingan orientation ofthe edgesof
G
allowingto ompensate signature of permutations, so that all terms in the expansion of the determinant ofK
indeed have thesame sign.
Let
M 1 and M 2 be two perfet mathings of G
drawn one on top of the other. Dene an
alternatingyle to be ayle ofG
whose edgesalternate between edges ofM 1 andM 2. Then,
G
drawn one on top of the other. Dene an alternatingyle to be ayle ofG
whose edgesalternate between edges ofM 1 andM 2. Then,
analternating ylehaseven length,and ifthelength isequal to 2,theyleisa doubled edge,
thatis an edgeovered byboth
M 1 and M 2. Thesuperimpositionof M 1 andM 2 is aunion of
M 1 andM 2 is aunion of
disjoint alternating yles,see Figure2.4. Thisis beause,bydenition ofa perfet mathing,
eah vertex is adjaent to exatlyone edge of themathing, sothat in the superimposition of
two mathings
M 1 and M 2,eah vertex is adjaent to exatly one edgeof M 1 and one edge of
M 1 and one edge of
M 2.
M M 1
2
Figure 2.4: Superimposition of two dimer ongurations
M 1 and M 2 of a subgraph G
of the
honeyomb lattie H
.
G
of the honeyomb lattieH
.One an transformthe mathing
M 1 into the mathing M 2,byreplaing edges ofM 1 by those
M 1 by those
of
M 2inallalternating ylesoflength≥ 4
ofthesuperimposition. Thus,arguingbyindution,
itsuesto showthat thesign ofthe ontributions of M 1 andM 2 todet(K)
isthesame when
M 1 andM 2 dieralong asingle alternating yleof length ≥ 4
. Letus assume thatthis isthe
M 2 todet(K)
isthesame when
M 1 andM 2 dieralong asingle alternating yleof length ≥ 4
. Letus assume thatthis isthe
M 2 dieralong asingle alternating yleof length ≥ 4
. Letus assume thatthis isthe
ase, denote the unique yle by
C
and byw i
1 , b j
1 , . . . , w i
k , b j
k
its verties inlokwise order,seeFigure 2.5.
Let
σ
(resp.τ
)bethepermutationorrespondingtoM 1(resp. M 2). Thenbytheorrespondene
between enumeration of mathingsand termsintheexpansion ofthedeterminant, we have:
j 1 = σ(i 1 ) = τ (i 2 ), j 2 = σ(i 2 ) = τ (i 3 ), . . . , j k = σ(i k ) = τ (i 1 ).
Ifwe let
c
bethe permutation ylec = (i k . . . i 1 )
,thenwe dedue:τ (i ℓ ) = σ(i ℓ − 1 ) = σ ◦ c(i ℓ ).
(2.1)...
w i
1 b j
1
w i
2 b j w i 2
k
b j
k
M 1
M 2 C
Figure 2.5: Labeling ofthe vertiesof anexample of superimpositionyle of
M 1 and M 2.
Inorderto hekthatthe ontributions of
M 1 and M 2 todet(K)
have thesamesign,itsues
det(K)
have thesamesign,itsuestohekthatthe signoftheratiooftheontributionsispositive. Thesignofthisratio,denoted
by
Sign(M 1 /M 2 )
,is:Sign(M 1 /M 2 ) = Sign
sgn
(σ)
sgn
(τ ) K( w i
1 , b σ(i
1 ) ) . . . K( w i
k , b σ(i
k ) ) K( w i
1 , b τ(i
1 ) ) . . . K( w i
k , b τ(i
k ) )
,
whih isthe same asthe signof the produtofthenumerator andthedenominator. Now,
( ⋄ 1 ) : =
sgn(σ)
sgn(τ ) =
sgn(σ ◦ τ ) =
sgn(σ ◦ σ ◦ c)
byEquation(2.1)= ( − 1) k+1 . ( ⋄ 2 ) : = Sign ( [K( w i
1 , b σ(i
1 ) ) . . . K( w i
k , b σ(i
k ) )][K ( w i
1 , b τ(i
1 ) ) . . . K( w i
k , b τ (i
k ) )]
= Sign [K( w i
1 , b j
1 ) . . . K( w i
k , b j
k )][K( w i
1 , b j
k ) . . . K( w i
k , b j
k−1 )]
= Sign K( w i
1 , b j
1 )K( w i
1 , b j
k ) . . . K( w i
k , b j
k )K( w i
k , b j
k−1 )
Let
p
be the parity of the number of edges of the yleC
oriented lokwise. The yleC
issaidto belokwise odd if
p = 1
,lokwise even ifp = 0
. Letus showthatSign(M 1 /M 2 ) = +1
ifand onlyif
p = 1
. To this purpose, werst relate( ⋄ 2 )
and( − 1) p.
Partition
W C := { w i
1 , . . . , w i
k } as W e
C ∪ W o
C
, whereW e
C
onsistsof white verties with 0 or 2inomingedges(equivalently2or 0outgoingedges),and
W o C onsistsofwhitevertieswithone
inoming and one outgoing edge, then | W e
C | + | W o
C | = k. If a white vertex belongs to W e
C
, itontributes
1
to( ⋄ 2 )
and1
top
;ifa whitevertexbelongs toW o
C
, itontributes− 1
to( ⋄ 2 )
and0
top
. We thus have:( ( ⋄ 2 ) = ( − 1) | W o C |
( − 1) p = ( − 1) | W e C | ⇒ ( ⋄ 2 ) = ( − 1) k ( − 1) p .
Asa onsequene
Sign(M 1 /M 2 ) = ( ⋄ 1 )( ⋄ 2 ) = ( − 1) 2k+1 ( − 1) p,so thatSign(M 1 /M 2 )
is positive
ifand onlyif
p = 1
,i.e. ifand onlyiftheyleC
islokwiseodd.Following Kasteleyn [Kas67℄, an orientation of the edges of
G
suh that all yles obtained as superimpositionofdimerongurationsarelokwiseodd,isalledadmissible. Deneaontouryle tobeayleboundinganinnerfaeofthegraph
G
. Kasteleynprovesthatiftheorientation issuhthatallontourylesarelokwiseodd,thentheorientationisadmissible. Theproofisbyindutiononthe numberoffaesinludedintheyle, refertothepaper[Kas67℄ fordetails.
Anorientation ofthe edgesof
G
suhthatallontouryles arelokwiseoddisonstrutedin thefollowingway,seeforexample[CR07 ℄. ConsideraspanningtreeofthedualgraphG ∗,witha vertexorrespondingto theouterfae,takento be theroot ofthetree. Chooseanyorientation
for edges of
G
not rossed by the spanning tree. Then, start from a leafof thetree, and orientorresponding faeis lokwiseodd. Remove the leafand theedge from thetree. Iterate until
onlytherootremains. Sinethetreeisspanning,allfaesarereahed bythealgorithm, andby
onstrution all orrespondingontour ylesarelokwiseodd.
AKasteleyn-Perus matrix,or simplyKasteleynmatrix,denotedby
K
,assoiated tothegraphG
is the oriented, weighted adjaeny matrix orresponding to an admissible orientation. We havethus proved thefollowing:Theorem 1. [Kas67℄ Let
G
be a nite, planarbipartite graph withan admissible orientation of its edges,letν
be a positive weight funtionon the edges, andK
be the orresponding Kasteleyn matrix. Then,the partition funtion of the graphG
is:Z ( G ) = | det(K) | .
When the graph
G
is not bipartite, lines and olumns of the adjaeny matrix are indexed by allvertiesofG
(vertiesannotbenaturallysplitinto twosubsets). Byhoosingan admissible orientation of the edges, the partition funtion an be expressed as the square root of thedeterminant of the orresponding Kasteleyn matrix or,sine this matrix isskew-symmetri, as
thePfaanof this samematrix. For more details,referto [Kas67 ℄.
2.3.2 Boltzmann measure formula
Whenthegraph
G
isbipartite,Kenyon[Ken97 ℄givesanexpliitexpressionfortheloalstatistis of the Boltzmann measure. LetK
be a Kasteleyn matrix assoiated toG
, and let{ e 1 = w 1 b 1 , . . . , e k = w k b k }
be asubset of edgesofG
.Theorem 2. [Ken97℄ The probability
µ( e 1 , . . . , e k )
of edges{ e 1 , . . . , e k }
ourring in a dimeronguration of
G
hosen withrespet to the Boltzmannmeasureµ
is:µ( e 1 , . . . , e k ) =
Y k i=1
K ( w i , b i )
!
1 ≤ det i,j ≤ k K − 1 ( b i , w j )
.
(2.2)Proof. The weighted sum of dimer ongurations ontaining the edges
{ e 1 , . . . , e k }
is (up tosign) the sum of all terms ontaining
K( w 1 , b 1 ) . . . K( w k , b k )
in the expansion ofdet(K)
. Byexpanding this determinant along lines(or olumns), itis easy to seebyindution that this is
equal to:
Y k i=1
K( w i , b i )
!
det(K E ) ,
where
K E isthematrix obtainedfromK
byremovingthelinesorrespondingto w 1 , . . . , w k and
theolumnsorrespondingto b 1 , . . . , b k. Now byJaobi's identity,see for example[HJ90℄:
b 1 , . . . , b k. Now byJaobi's identity,see for example[HJ90℄:
det(K E ) = det(K) det (K − 1 ) E ∗ ,
where
E ∗ isthe setof edgesnot inE
. Otherwise stated,(K − 1 ) E ∗ isthek × k
matrixobtained
k × k
matrixobtainedfrom
K − 1 by keeping the lines orresponding to b 1 , . . . , b k and the olumns orresponding to
w 1 , . . . , w k. Thus,
µ( e 1 , . . . , e k ) = Q k
i=1 K( w i , b i )
det(K E )
| det(K ) | =
Y k i=1
K( w i , b i )
!
det (K − 1 ) E ∗ .
thatis apoint proesssuh thatthejointprobabilities are oftheform
µ( e 1 , . . . , e k ) = det(M( e i , e j ) 1 ≤ i,j ≤ k ),
forsomekernel
M
. Intheaseofbipartitedimers,M ( e i , e j ) = K( w i , b j )K − 1 ( b i , w j )
,see[Sos07 ℄for an overview.
2.3.3 Expliit example
Figure2.6givesanexampleofaplanar,bipartitegraphwhoseedgesareassignedpositiveweights
and anadmissible orientation.
a a
a a
a
b b b
b b
a, b > 0
PSfrag replaements
w 1 b 1 w 2
b 2 w 3 b 3
w 4 b 4
Figure 2.6: A planar,bipartite graphwith a positive weight funtion and an admissible orien-
tation.
TheorrespondingKasteleynmatrix is:
K =
a 0 − b 0 a − b 0 0
b a a b
0 0 b − a
,
and thedeterminant is equal to
det(K) = 2a 3 b + 2b 3 a.
Setting
a = b = 1
yields that the number of perfet mathings of this graph is4
. In this ase,theBoltzmann measureis theuniform measureon tilings ofthis graph.
TheinverseKasteleyn matrix
K − 1 is:
K − 1 = 1 4
2 1 1 1
2 − 3 1 1
− 2 1 1 1
− 2 1 1 − 3
.
Using the labeling of the verties of Figure 2.6 and Theorem 2, we ompute theprobability of
ourreneofsome subset of edges:
µ( w 1 b 1 ) = | K − 1 ( b 1 , w 1 ) | = 1 2 µ( w 3 b 1 ) = | K − 1 ( b 1 , w 3 ) | = 1 4 µ( w 1 b 1 , w 3 b 4 ) =
det
K − 1 ( b 1 , w 1 ) K − 1 ( b 1 , w 3 ) K − 1 ( b 4 , w 1 ) K − 1 ( b 4 , w 3 )
= 1 16
det
2 1
− 2 1
= 1
4 .
ndallpossibleongurations(
4
ofthem),andompute theprobabilities. However,thegoalof this exampleis to showhowto usethegeneral formulae.2.4 Geometri interpretation of lozenge tilings
By means of the height funtion, Thurston interprets lozenge tilings of the triangular lattie
as disrete surfaes in a rotated version of
Z 3 projeted onto the plane. He gives a similar interpretationofdominotilingsofthesquarelattie. Thisapproahanbegeneralizedtodimer
ongurations ofbipartite graphs usingows. Thisyields an interpretationof thedimermodel
on a bipartite graphas a random interfae modelin dimensions
2 + 1
,and oersmore insightinto the model. In this setion we exhibit Thurston's onstrution of the height funtion on
lozenge tilings. We postpone the denition of the height funtion on general bipartite graphs
until Setion 3.1.
Faes of the triangular lattie
T
an be olored in blak and white, so that blak faes (resp.whiteones)areonlyadjaenttowhiteones(resp. blakones). Thisisaonsequeneofthefat
thatitsdualgraph,thehoneyomblattie,isbipartite. Orienttheblakfaesounterlokwise,
and the white ones lokwise, see Figure 2.7 (left). Consider a nite subgraph
X
ofT
whih is tileable by lozenges, and a lozenge tilingT
ofX
. Then theheight funtionh T is an integer
valuedfuntion onverties of
X
,dened indutively asfollows:•
Fixa vertexv 0 of X
,and seth T ( v 0 ) = 0
.
•
For every boundary edgeuv
of a lozenge,h T ( v ) − h T ( u ) = +1
ifthe edgeuv
is oriented fromu
tov
,implyingthath T ( v ) − h T ( u ) = − 1
whenthe edgeuv
is orientedfromv
tou
. The height funtion is well dened, in the sense that the height hange around any orientedyleis 0. Anexample ofomputation of theheight funtion isgiven inFigure2.7(right).
0
3
1
1 2
3 2 4
1 2 3 4 3 2 1 0
1 2
0 1
2 1
0 2
3 3
4 4
5 3 4 3 2 1 0
4
2 1 2
2 3
2 v 0
Figure2.7: Orientation offaesofthe triangularlattie(left). Heightfuntion orrespondingto
a lozenge tiling(right).
As a onsequene, lozenge tilings are interpreted as stepped surfaes in
Z f 3 projeted onto the
plane, where Z f 3 is Z 3 rotated so that diagonals ofthe ubesare orthogonal to the plane. The
height funtionisthen simplythe height ofthesurfae(i.e. thirdoordinate). Thisonstru-
Z 3 rotated so that diagonals ofthe ubesare orthogonal to the plane. The height funtionisthen simplythe height ofthesurfae(i.e. thirdoordinate). Thisonstru-
tiongivesamathematial senseto theintuitive feelingofubesstiking inor out,whihstrikes
us whenwathinga piture oflozenge tilings.
Height funtions haraterize lozengetilings asstated bythefollowing lemma.
Lemma 4. Let
X
be a nite simply onneted subgraph of the triangular lattieT
, whih is tileable by lozenges. Leth
be an integervalued funtion on the verties ofX
,satisfying:• h( v 0 ) = 0
,wherev 0 isa xed vertex of X
.
• h( v ) − h( u ) = 1
for any boundary edgeuv
ofX
oriented fromu
tov
.• h( v ) − h( u ) = 1
or− 2
for any interior edgeuv
ofX
oriented fromu
tov
.Then,there isa bijetion between funtions
h
satisfyingthese two onditions, andtilings ofX
.Proof. Let
T
be a lozenge tiling ofX
and letuv
be an edgeofX
, oriented fromu
tov
. Then, the edgeuv
is either a boundary edge or a diagonal of a lozenge. By denition of the height funtion,theheight hange is1
intherst ase, and− 2
intheseond.Conversely,let
h
beanintegerfuntionasinthelemma. LetusonstrutatilingT
whoseheightfuntion is
h
. Consider a blakfaeofX
,thenthere isexatly one edgeuv
ontheboundary of this faewhose height hange is− 2
. Tothis fae, we assoiate the lozenge whih isrossed bytheedge
uv
. Repeating thisproedure for allblakfaesyields a tilingofX
.Thurston [Thu90℄uses height funtions inorderto determine whethera subgraphof thetrian-
gular lattiean betiledbylozenges. Referto thepaperfor details.
Dimer model on infinite periodi bipartite graphs
ThishapterisdevotedtothepaperDimersandamoebas [KOS06 ℄byKenyon,Okounkovand
Sheeld.
Reallthatedgesofdimerongurationsrepresentdi-atomimoleules. Sineweareinterested
inthemarosopibehaviorofthesystem,ourgoalistostudythemodelonverylargegraphs. It
turnsout thatitiseasierto extratinformationfor themodeldened oninnitegraphs, rather
than very large ones. Indeed, on very large but nite graphs, Kasteleyn's omputation an be
done,but involvesomputingthedeterminant ofhuge matries, whih isofourse veryhard in
general, and won't tell us muh about the system. Computing expliitly theloal statistis of
theBoltzmann measurebeomeshardlytratablesine itrequiresinvertingverylarge matries.
Thismotivates the following roadmap.
•
Assume that the graphG = ( V , E )
is simple, planar, innite, bipartite, andZ 2-periodi.
This means that
G
is embedded inthe plane sothat translations atbyolor-preserving isomorphism ofG
, i.e. isomorphisms whih map blak verties to blak ones and white verties to white ones. For later purposes, we onsider the underlying lattieZ 2 to be
a subgraph of the dual graph G ∗, and x a basis{ e x , e y }
, allowing to reord opies of a
{ e x , e y }
, allowing to reord opies of avertex
v
ofG
as{ v + (k, l) : (k, l) ∈ Z 2 }
. RefertoFigure3.1foranexample whenG
isthe square-otagongraph.e x e y
Figure3.1: A piee ofthe square-otagon graph. The underlying lattie
Z 2 is inlight grey,the
two blakvetors represent ahoieof basis{ e x , e y }
.
•
LetG n = ( V n , E n )
bethequotientofG
bythe ationofn Z 2. Thenthesequeneofgraphs
{ G n } n ≥ 1 isanexhaustionoftheinnitegraphG
bytoroidalgraphs. ThegraphG 1 = G / Z 2
isalledthe fundamental domain,see Figure3.2. Assumethatedgesof
G 1 areassigned a
positive weight funtionν
thus deninga periodiweight funtionon theedgesof G
.
1 1
1 1
1 1
1 1
1 1 1 1
PSfragreplaements
G 1
Figure 3.2: Fundamental domain
G 1 of the square-otagon graph, opposite sides in light grey areidentied. Edges areassignedweight 1.
•
The goal of this hapter is to understand the dimer model onG
, using the exhaustion{ G n } n ≥ 1,bytakinglimitsasn → ∞
of appropriate quantities.
Note that it is ruial to take an exhaustion of
G
by toroidal graphs. Indeed, the latter are invariant bytranslations intwo diretions, akeyfatwhihallows omputationsto go throughusing Fourier tehniques. Note also that takingan exhaustion by planar graphs a priori leads
to dierent results beause of theinueneof the boundary whih annotbenegleted.
Althoughitmightnotseemsolearatthisstage,thefatthat
G
and{ G n } n ≥ 1areassumedtobe
bipartiteisalsoruialfortheresultsofthishapter, beauseitallowstorelatethedimermodel
to well behaved algebrai urves. Having a general theoryof thedimermodelon non-bipartite
graphs isone ofthe important openquestions ofthe eld.
3.1 Height funtion
Inthissetion,wedesribetheonstrutionoftheheightfuntionondimerongurationsofthe
innitegraph
G
andofthetoroidalgraphsG n,n ≥ 1
. Sineweareworkingonthedimermodel
and noton thetiling model,asin Thurston'sonstrution, theheight funtion isa funtion on
faesof
G
or equivalently, afuntion on vertiesof thedualgraphG ∗.
Thedenitionofthe heightfuntionreliesonows. Denoteby
~ E
thesetofdiretededgesofthe graphG
,i.e. every edgeofE
yieldstwo orientededges of~ E
. A owω
is a realvalued funtiondened on
~ E
,that isevery diretededge( u , v )
of~ E
isassigned aowω( u , v )
.The divergene of a ow
ω
,denoted bydiv ω
,isa real valued funtion dened onV
givingthe dierene between totaloutow and totalinow at verties:∀ u ∈ V , div ω( u ) = X
v ∼ u
ω( u , v ) − X
v ∼ u
ω( v , u ).
Sine
G
is bipartite, we split vertiesV
into white and blak ones:V = W ∪ B
. Then, every dimerongurationM
ofG
denesawhite-to-blakunitowω M asfollows. Theowω M takes
value0 onall diretededges arisingfrom edgesof
E
whihdo not belong toM
,and∀ wb ∈ M, ω M ( w , b ) = 1, ω M ( b , w ) = 0.
Sineeveryvertexofthegraph
G
isinident toexatlyoneedgeoftheperfetmathingM
,theow
ω M hasdivergene 1at everywhite vertex,and -1at every blakone, thatis:
∀ w ∈ W , div ω M ( w ) = X
b ∼ w
ω M ( w , b ) − X
b ∼ w
ω M ( b , w ) = 1,
∀ b ∈ B , div ω M ( b ) = X
w ∼ b
ω M ( b , w ) − X
w ∼ b
ω M ( w , b ) = − 1.
Let
M 0 be a xed periodi referene perfet mathing of G
, and ω M 0 be the orresponding
ow, alled the referene ow. Then, for any other mathing M
with ow ω M, thedierene
ω M − ω M 0 is adivergene-free ow,that is:
M
with owω M, thedierene
ω M − ω M 0 is adivergene-free ow,that is:
∀ w ∈ W , div(ω M − ω M 0 )( w ) = X
b ∼ w
(ω M ( w , b ) − ω M 0 ( w , b )) − X
b ∼ w
(ω M ( b , w ) − ω M 0 ( b , w ))
= div ω M ( w ) − div ω M 0 ( w ) = 1 − 1 = 0,
and similarlyfor blak verties.
We arenow readyto dene theheight funtion. Let
M
be adimerongurationofG
,thenthe height funtionh M is an integer valuedfuntion on faesof G
dened asfollows.
•
Fixa faef 0 of G
,andset h M ( f 0 ) = 0
.
•
For any other faef 1, onsider an edge-path γ
of the dual graph G ∗
, from
f 0 to f 1. Let
( u 1 , v 1 ), . . . , ( u k , v k )
denoteedges ofG
rossing the pathγ
,where for everyi
,u i ison the
leftofthe pathγ
andv i ontheright. Thenh M ( f 1 ) − h M ( f 0 )
isthetotalux ofω M − ω M 0
h M ( f 1 ) − h M ( f 0 )
isthetotalux ofω M − ω M 0
aross
γ
,that is:h M ( f 1 ) − h M ( f 0 ) = X k i=1
[(ω M ( u i , v i ) − ω M ( v i , u i )) − (ω M 0 ( u i , v i ) − ω M 0 ( v i , u i ))].
The height funtion is well dened if itis independent of thehoie of
γ
,or equivalently ifthe height hange aroundeveryfaef ∗ ofthedualgraph G ∗is0. Letu
bethevertexofthegraph G
u
bethevertexofthegraphG
orrespondingto the fae
f ∗,and letv 1 , . . . , v k be itsneighbors. Then, bydenition, theheight
hange aroundthe faef ∗
f ∗
,inounterlokwiseorder, is:
X k i=1
[(ω M ( u , v i ) − ω M ( v i , u )) − (ω M 0 ( u , v i ) − ω M 0 ( v i , u ))] = div(ω M − ω M 0 )( u ) = 0.
Theheightfuntion isthus well dened asa onsequeneof thefatthattheow
ω M − ω M 0 is
divergenefree, uptothe hoieofa basefae
f 0 and ofareferenemathingM 0. An analogof
Lemma 4 givesa bijetion between height funtionsanddimerongurations of
G
. Remark 5.•
There is atually an easy wayof omputing theheight funtion. Reall from Setion 2.3thatthesuperimpositionoftwodimerongurations
M
andM 0 onsistsofdoublededges
andalternating ylesoflength
≥ 4
(ylesmayextendto innitywhenG
isinnite). Let usdenotebyM − M 0theorientedsuperimpositionofM
andM 0,withedgesofM
oriented
M
orientedfromwhitevertiesto blakones,and thoseof
M 0 fromblakvertiesto whiteones,then
M − M 0 onsists of doubled edges oriented in both diretions and oriented alternating
ylesof length ≥ 4
,seefor example Figure3.3. Returning tothedenition oftheheight
≥ 4
,seefor example Figure3.3. Returning tothedenition oftheheightfuntion,onenotiesthattheheighthangesby
± 1
exatlywhenrossingayle,andthesign onlydependson theorientation of the yle.
•
By takinga dierent hoie of referene ow, one an reover Thurston's height funtioninthe aseof lozenge and domino tilings, up to a global multipliative fator of
1 3
intherst ase, and
1
4
in theseond - the interested reader an try and work out this relationexpliitly.
Let us now onsider the toroidal graph
G n = G /n Z 2. In this ase, the height funtion is not well dened sine there might besome period,or heighthange, alongyles inthedualgraph
winding around the torus horizontally or vertially, see Figure 3.3. More preisely, a perfet
mathing
M
ofG n an be liftedto a perfet mathingof theinnitegraph G
,also denoted M
.
Then, theperfetmathing
M
ofG n issaid tohave height hange (h M x , h M y )
if:
h M ( f + (n, 0)) = h M ( f ) + h M x h M ( f + (0, n)) = h M ( f ) + h M y .
Note that height hange is well dened, i.e. does not depend on the hoie of fae
f
, beause theowω M − ω M 0 is divergene-free.
Figure3.3givesanexampleoftherefereneperfetmathing
M 0 induedbyaperiodireferene
mathingof
G
,andoftheheighthangeomputationofadimerongurationM
ofthetoroidalgraph
G 2 of the square-otagongraph. TheperfetmathingM
hasheight hange (0, 1)
.
0 f 0
M M
0
0 0
0 0 0
−1
−1 0 −1
0 0
0 0 0
−1 0 −1 0 −1
−1
−1
−1
−1
−1
Figure3.3: Aperfet mathing
M
ofthe toroidal graphG 2,having height hange (0, 1)
.
Remark 6. Let
T 2 = { (z, w) ∈ C 2 : | z | = | w | = 1 }
denotethetwo-dimensional unit torus,and letH 1 ( T 2 , Z ) ∼ = Z 2 be the rst homology group of T 2 in Z
. The graph G n being embedded in
Z
. The graphG n being embedded in
T 2,we takeasrepresentative of abasisof H 1 ( T 2 , Z )
thevetorsne x and ne y embedded onthe
ne y embedded onthe
torus,where reall
{ e x , e y }
were our hoie ofbasisvetor forZ 2,seeFigure3.1. Inthease of
thesquare-otagon graphandG 2 ofFigure3.3, the rstbasisvetoristhelightgrey horizontal
yle oriented from left to right, and the seond is the light grey vertial yle oriented from
bottomto top. Then, the homology lassofthe orientedsuperimposition
M − M 0 inthisbasis
is
(1, 0)
, and the height hange is(0, 1)
. More generally, if the homology lass ofM − M 0 is
(a, b)
, thentheheight hange (h M x , h M y )
is( − b, a)
. Thisis beause,asmentioned inRemark5,
theheight funtion hanges by
± 1
exatlywhen itrossesorientedyles ofM − M 0,implying
that the height hange
(h M x , h M y )
an be identied through the intersetion pairing with the homology lassoftheorientedongurationM − M 0 inH 1 ( T 2 , Z )
.
Letusreallthesetting:
G
isasimple,planar,innite,bipartiteZ 2-periodigraph,and{ G n } n ≥ 1
is the orresponding toroidal exhaustion. Edges of
G
are assigned a periodi, positive weight funtionν
,and{ e x , e y }
denotes ahoie ofbasisof theunderlying lattieZ 2.
Inthis setionwepresent theomputation ofthepartition funtion for dimerongurations of
thetoroidal graph
G n,and the omputation of the free energy,whih is minus theexponential
growth rate of the partition funtion of the exhaustion G n. More preisely, Setion 3.2.1 is
devoted to Kasteleyn theory on the torus. Then, in Setion 3.2.2 we dene the harateristi
polynomial, one of the keyobjets underlying thedimer model on bipartite graphs, yielding a
ompat losed formula for the partition funtion, see Corollary 10. With this expression at
hand,one thenderivestheexpliit formula forthe free energy,seeTheorem 11.
3.2.1 Kasteleyn matrix
InSetion 2.3, we proved thatthe partition funtionof anite, simplyonneted planargraph
is given bythedeterminant of aKasteleyn matrix. Whenthegraph isembedded on thetorus,
a single determinant is not enough, what is needed is a linear ombination of determinants
determined inthe following way.
Consider the fundamental domain
G 1 and let K 1 be a Kasteleyn matrix of this graph, that is
K 1 is the oriented, weighted, adjaeny matrix of G 1 for a hoie of admissible orientation of
theedges. Notethatadmissible orientationsalsoexistforgraphsembeddedinthetorus. Letus
K 1 is the oriented, weighted, adjaeny matrix of G 1 for a hoie of admissible orientation of
theedges. Notethatadmissible orientationsalsoexistforgraphsembeddedinthetorus. Letus
nowlookat the signsof the weighted mathings in the expansion of
det(K 1 )
. Consider a xedreferene mathing
M 0 of G 1,and let M
be any other mathing of G 1. Then, bythe results of
[Kas67 ,DZM
M
be any other mathing ofG 1. Then, bythe results of [Kas67 ,DZM
+
96,GL99 ,Tes00 ,CR07℄,the signofthemathing
M
onlydependson theparityofthe vertial andhorizontalheighthange. Moreover,of thefour possible paritylasses,three
havethesamesignin
det(K 1 )
andonehasoppositesign. Byanappropriatehoieofadmissible orientation,one an makethe(0, 0)
lasshave positive sign.Let
γ x,γ y bethe basisvetor e x,e y of the underlying lattieZ 2,embedded onthetorus. Sine
we have hosenZ 2 to be asubgraphof thedualgraph G ∗,γ x,γ y areorientedylesofthedual
e x,e y of the underlying lattieZ 2,embedded onthetorus. Sine
we have hosenZ 2 to be asubgraphof thedualgraph G ∗,γ x,γ y areorientedylesofthedual
Z 2,embedded onthetorus. Sine
we have hosenZ 2 to be asubgraphof thedualgraph G ∗,γ x,γ y areorientedylesofthedual
G ∗,γ x,γ y areorientedylesofthedual
γ y areorientedylesofthedual
graph
G ∗
1
,seeFigure 3.4. We refer toγ x asa horizontalyle, and to γ y asa vertialone.
G 1
γ γ
x y
Figure3.4: Fundamental domain
G 1 of the squareotagongraphwiththeorientedpathsγ x , γ y
inthedualgraph
G ∗
1
. Thetwo opies ofγ x,respetively γ y,aregluedtogether.
For
σ, τ ∈ { 0, 1 }
, letK 1 θτ be the Kasteleyn matrix in whih the weights of theedges rossing
the horizontal yle
γ x are multiplied by ( − 1) θ, and those rossing the vertial yle γ y are
γ y are
multiplied by
( − 1) τ. Observing that hanging the signsalong a horizontal dual yle has the
eet of negating theweight of mathings withodd horizontal height hange, and similarly for
vertial;thefollowing tableindiates thesignofthemathings intheexpansionof
det(K 1 θτ )
,asa funtionof the horizontaland vertial height hange mod
2
.(0, 0) (1, 0) (0, 1) (1, 1)
det(K 1 00 ) + − − −
det(K 1 10 ) + + − +
det(K 1 01 ) + − + +
det(K 1 11 ) + + + −
(3.1)
From Table3.1, we dedue
Z( G 1 ) = 1
2 − det(K 1 00 ) + det(K 1 10 ) + det(K 1 01 ) + det(K 1 11 )
.
(3.2)A similar argument holds for the toroidal graph
G n, n ≥ 1
. The orientation of edges of G 1
denes a periodi orientation of edges of
G
, and thus an orientation of edges ofG n. Let γ x,n,
γ y,n betheorientedyles inthedualgraph G ∗ n,obtainedbytakingn
timesthebasisvetore x,
γ y,n betheorientedyles inthedualgraph G ∗ n,obtainedbytakingn
timesthebasisvetore x,
n
timesthebasisvetore x,
n
times the basisvetore y respetively,embedded on the torus. For σ, τ ∈ { 0, 1 }
, let K n στ be
thematrix
K ninwhihtheweightsof edgesrossingthehorizontal yleγ x,n aremultipliedby
( − 1) θ,andthose rossingthe vertialyle γ y,n aremultiplied by( − 1) τ. Then,
( − 1) θ,andthose rossingthe vertialyle γ y,n aremultiplied by( − 1) τ. Then,
( − 1) τ. Then,
Theorem 7. [Kas67, DZM
+
96,GL99 , Tes00, CR07, CR08℄
Z ( G n ) = 1
2 − det(K n 00 ) + det(K n 10 ) + det(K n 01 ) + det(K n 11 ) .
3.2.2 Charateristi polynomial
Let
K 1 be a Kasteleyn matrix of the fundamental domain G 1. Given omplex numbers z
and
w
, an altered Kasteleyn matrix K 1 (z, w)
is onstruted as follows. Let γ x, γ y be the oriented
z
andw
, an altered Kasteleyn matrixK 1 (z, w)
is onstruted as follows. Letγ x, γ y be the oriented
horizontal and vertial yles of
G ∗
1
. Then, multiply edge-weights of edges rossingγ x by z
whenever the white vertex is on the left, and by
z − 1 whenever the blak vertex is on theleft.
Similarly,multiply edge-weightsof edgesrossing thevertial path
γ y byw ± 1,seeFigure 3.5.
The harateristi polynomial
P (z, w)
of the graphG 1 is dened as the determinant of the alteredKasteleyn matrix:
P (z, w) = det(K 1 (z, w)).
Aswewillsee,theharateristipolynomialontainsmostoftheinformationonthemarosopi
behaviorof the dimermodel onthegraph
G
.The next very useful lemma expresses the harateristi polynomial using height hanges of
dimerongurations of
G 1. Referto Setion 3.1for denitions andnotations onerningheight
hanges. LetM 0beareferenedimerongurationofG 1,andsupposetheadmissibleorientation
oftheedgesishosensuhthatperfetmathingshaving (0, 0)
mod(2, 2)
heighthangehave+
G 1,andsupposetheadmissibleorientation
oftheedgesishosensuhthatperfetmathingshaving (0, 0)
mod(2, 2)
heighthangehave+
signintheexpansionof
det(K 1 )
. WeonsiderthealteredKasteleynmatrixK 1 (z, w)
onstrutedfrom
K 1. Letω M 0 be therefereneoworrespondingto thereferenedimerongurationM 0,
M 0,
andlet
x 0 denote thetotal uxofω M 0 arossγ x,similarlyy 0 isthetotaluxofω M 0 arossγ y.
γ x,similarlyy 0 isthetotaluxofω M 0 arossγ y.
ω M 0 arossγ y.
Then, we have:
Lemma 8. [KOS06℄
P (z, w) = z x 0 w y 0 X
M ∈M ( G 1 )
ν(M )z h M x w h M y ( − 1) h M x h M y +h M x +h M y ,
(3.3)where, for every dimeronguration