In this setion we address the question of the utuations of the height funtion around its
mean. We restritourselvesto the aseof the uniform dimermodel on theinnitehoneyomb
lattie
H
,withnomagnetield. Thegoalistoprovethatthelimitingutuationsoftheheight funtion are desribed by a Gaussian free eldof the plane, a behavior whih is harateristiof theliquidphase.
Resultspresentedhere, and extensionsan be found in[Ken00 ,Ken01 ,Ken08 , dT07℄.
3.5.1 Uniform dimermodel on the honeyomb lattie
Letusspeifytheresultsobtainedintheprevioussetionstotheasewhere
H
isthehoneyomb lattie,edges areassignedweights1
,and thereisno magneti eld. Thehoie offundamental domain isgiven inFigure3.10.1/w
γ x γ y
z 1 1
1 1
1
Figure3.10: Choieof fundamentaldomain of thehoneyomb lattie
Theharateristi polynomialis:
P (z, w) = 1 + z + 1 w .
The Gibbs measure
µ
obtained asweak limit of the Boltzmann measuresµ n on M ( G n )
(with
no magnetield) hasthefollowing expressionon ylindersets:
µ( e 1 , . . . , e k ) = det K − 1 ( b i , w j ) 1 ≤ i,j ≤ k
,
(3.10)where
K − 1 ( b 0,0 , w x,y ) = 1 (2πi) 2
Z
T 2
1
1 + z + w 1 z y w − x dz z
dw w ,
and
w x,y representsthewhitevertexofthe(x, y)
opyofthefundamentaldomain,andsimilarly
for b x,y.
on the unit torus
T
:(z 0 , w 0 ) = (e i 2π 3 , e i 2π 3 ),
and(z 0 , w 0 ).
The uniform dimer model on thehoneyomb lattie isthus intheliquid phase.
InSetion 2.4, we dened Thurston'sheight funtionon lozenge tilingsof thetriangular lattie
T
,whih wasa funtion onvertiesofT
. Equivalently, itis afuntion on dimerongurations of the honeyomb lattieH
,dened on faesofH
. For onveniene, from now on, we onsider1
3
of Thurston's height funtion.The following lemma gives an expression of the height funtion using indiator funtions of
edges. We will use this inthe proof of the onvergene of the height funtion. Let
u , v
be two faes ofH
,and letγ
be an edge-path in thedual graph fromu
tov
. Denote bye 1 , . . . , e m the
dualedgesofedgesofγ
whihhaveablakvertexontheleft,andbyf 1 , . . . , f nthosewhihhave
awhite vertexon theleft. Then,
Lemma 21.
γ
. Then, returning to the orientation of the edges of the triangular lattie introdued in theonstrutionofThurston'sheightfuntion,seeSetion2.4,weknowthattheedge
u i v i isoriented
fromu i to v i. Thus bydenition, we have:
v i. Thus bydenition, we have:
h( v i ) − h( u i ) = ( 1
3
iftheedgee i isnot inthedimeronguration
− 2 3 iftheedge e i isinthedimeronguration.
.
Thisan be summarized as:
h( v i ) − h( u i ) = − I e
i + 1 3
.A similar argument holds for edges whih have a white vertex on the left. Summing over all
edges inthe path
γ
yields the result.3.5.2 Gaussian free field of the plane
We nowdene the Gaussian freeeld ofthe plane, whih is, aswe will see, thelimiting objet
of theheight funtion.
The Green's funtion of the plane, denoted by
g
satises∆ x g(x, y) = δ x (y)
, whereδ x is the
Diradistribution at
x
. Upto an additive onstant,g
is given byg(x, y) = − 1
implying that the bilinear form
G
is positive denite. ThequantityG(ϕ 1 , ϕ 1 )
is known astheDirihlet energy of
f 1.
A random generalized funtion
F
assigns to every test funtionϕ
inC c,0 ∞ ( R 2 )
, a real randomvariable
F ϕ
. Itisassumedtobelinearandontinuous,whereontinuitymeansthatonvergeneof the funtions
ϕ n j (1 ≤ j ≤ k
) to ϕ j, implies onvergene in law of the random vetor
(F ϕ n 1 , . . . , F ϕ n k )
to(F ϕ 1 , . . . , F ϕ k )
. Arandomgeneralized funtionissaidtobeGaussianiffor every linearly independent funtionsϕ 1 , . . . , ϕ k ∈ C c,0 ∞ ( R 2 )
, therandom vetor(F ϕ 1 , . . . , F ϕ k )
isGaussian.
Theorem22. [Bo55 ℄If
B : C c,0 ∞ ( R 2 ) × C c,0 ∞ ( R 2 ) → R
isabilinear,ontinuous,positivedenite form,then there existsaGaussian random generalized funtionF
,whose ovariane funtionisgiven by:
E [F ϕ 1 F ϕ 2 ] = B (ϕ 1 , ϕ 2 ).
Bydenition aGaussian free eldof the plane GFFisa Gaussianrandom generalized funtion
whose ovariane funtionis the bilinear form
G
dened above,i.e.E [
GFF(ϕ 1 )
GFF(ϕ 2 )] = − 1
There areother ways of deningtheGaussian freeeld, seefor example [GJ81 ℄ and[She07 ℄.
3.5.3 Convergene of the height funtion to a Gaussian free field
Let
H ε be the graph H
whose edge-lengths have been multiplied by ε
, and onsider the
un-normalized height funtion
h
on faesofH ε. Dene,
H ε : C c,0 ∞ ( R 2 ) → R
ϕ 7→ H ε ϕ = ε 2 X
v ∈ F( H ε )
ϕ( v )h( v ).
Theorem 23. The random generalized funtion
H ε onverges weakly in law to √ 1 π
times a
Gaussian free eld,i.e. for every
ϕ 1 , . . . , ϕ k ∈ C c,0 ∞ ( R 2 )
,(H ε ϕ 1 , . . . , H ε ϕ k )
onverges in law(aswhere
g( u , v , u ′ , v ′ ) = g( v , v ′ ) + g( u , u ′ ) − g( v , u ′ ) − g( u , v ′ )
,g
isthe Green'sfuntionof the plane,and the sumisover all
(k − 1)!!
pairings of{ 1, . . . , k }
.Proof. We prove this proposition ina partiular ase. The proof of thegeneral ase issimilar,
although notationally muh more umbersome. Let us assume that
k = 2
, and suppose thatu 1 , v 1 , u 2 , v 2 are suh that we an hoose u ε
2
respetively asinFigure 3.11.ε
ε lim → 0
E [(h( v ε
1 ) − h( u ε
1 ))(h( v ε
2 ) − h( u ε
2 ))] = − 2 Re Z v 1
u 1
Z v 2 u 2
1
4π 2 (z 2 − z 1 ) 2 dz 1 dz 2
= − 1 2π 2 log
( v 1 − v 2 )( u 1 − u 2 ) ( v 1 − u 2 )( u 1 − v 2 )
= 1
π g( u 1 , v 1 , u 2 , v 2 ).
One Proposition 24 is available, one then introdues test funtions and prove onvergene of
all moments of (3.11) . This implies handling moments of height dierenes involving verties
whiharelose. Theasymptoti formulafor theinverse Kasteleynmatrix doesnot holdinthis
ase, and onehasto do areful boundsto see thateverything still works.
********************************
We have reahed the end of these leture notes overing the foundations of the dimer model
and thepaper Dimersand amoeba ofKenyon,Okounkovand Sheeld [KOS06 ℄, givinga full
desription of the phase diagram of the dimer model on periodi, bipartite graphs. Although
these arespetaularresults, thereare others,suh asthespei behaviorof thedimermodel
dened on isoradial graphs [Ken02 , KO06 ℄,theunderstanding of limit shapes[CKP01,KO07 ℄,
the appliation of dimertehniques to study theIsing model[BdT10b,BdT10a ℄. We do hope
that these notes will enourage the reader to learn more on this very rih model of statistial
mehanis.
[Bax89℄ Rodney J. Baxter. Exatly solved models in statistial mehanis. Aademi Press
In. [Harourt Brae Jovanovih Publishers℄, London, 1989. Reprint of the 1982
original.
[BdT10a℄ Cédri Boutillier and Béatrie de Tilière. The ritial
Z
-invariant Ising model viadimers: loality property. Comm.Math. Phys.,pages 144,2010.
[BdT10b℄ Cédri Boutillier and Béatrie de Tilière. The ritial
Z
-invariant Ising model viadimers: the periodi ase. Probab. Theory Related Fields,147:379413, 2010.
[Bo55℄ S.Bohner. Harmoni analysisandthetheoryof probability. UniversityofCalifornia
press, 1955.
[BR06℄ B. Bollobás andO. Riordan. Perolation. Cambridge Univ Pr,2006.
[CKP01℄ Henry Cohn,RihardKenyon,and JamesPropp. Avariational priniplefordomino
tilings. J.Amer.Math.So.,14(2):297346 (eletroni), 2001.
[CR07℄ David Cimasoni and NiolaiReshetikhin. Dimers onsurfae graphs and spin
stru-tures. I. Comm. Math.Phys., 275(1):187208,2007.
[CR08℄ David Cimasoni and NiolaiReshetikhin. Dimers onsurfae graphs and spin
stru-tures. II. Comm.Math.Phys., 281(2):445468, 2008.
[CS12℄ D. Chelkak and S. Smirnov. Universality in the 2D Ising model and onformal
invariane of fermioni observable. Invent. Math.,189:515580, 2012.
[DS11℄ B. Duplantier andS.Sheeld. Liouville quantum gravity and KPZ. Invent. Math.,
185(2):333393, 2011.
[DT89℄ G.DavidandC.Tomei.TheproblemofCalissons.Amer.Math.Monthly,96(5):429
431, 1989.
[dT07℄ B. de Tilière. Saling limit of isoradial dimer models and the ase of triangular
quadri-tilings. Ann. Inst. HenriPoinaré (B)Probab. Stat.,43(6):729750,2007.
[DZM
+
96℄ N. P. Dolbilin, Yu. M. Zinov
′
ev, A. S. Mishhenko, M. A. Shtan
′
ko, and M. I.
Shtogrin. Homologial properties of two-dimensional overings of latties on
sur-faes. Funktsional. Anal. iPrilozhen., 30(3):1933, 95,1996.
[FPT00℄ M. Forsberg, M. Passare, and A.Tsikh. Laurent determinants and arrangements of
hyperplane amoebas. Adv. Math.,151(1):4570, 2000.
[FR37℄ RH Fowler and GS Rushbrooke. An attempt to extend the statistial theory of
perfet solutions. Transations of the Faraday Soiety,33:12721294, 1937.
Springer-Verlag, 1981.
[GKZ94℄ I.M. Gelfand,M.M. Kapranov, and A.V. Zelevinsky. Disriminants, resultants,and
multidimensional determinants. Springer, 1994.
[GL99℄ A.Galluioand M.Loebl. OnthetheoryofPfaanorientations. I. Perfet
math-ings and permanents. Eletron. J.Combin., 6(1):R6,1999.
[Gri99℄ G. Grimmett. Perolation. Springer Verlag,1999.
[HJ90℄ R.A. Hornand C.R.Johnson. Matrixanalysis. Cambridge UniversityPress,1990.
[Kas61℄ P. W. Kasteleyn. The statistis of dimers on a lattie : I. The number of dimer
arrangementson a quadratilattie. Physia,27:12091225, De1961.
[Kas67℄ P.W.Kasteleyn.Graphtheoryandrystalphysis.InGraphTheoryandTheoretial
Physis, pages 43110.AademiPress, London,1967.
[Ken℄ R. Kenyon. Letures on dimers, statistial mehanis. IAS/Park City Math. Ser,
16:191230.
[Ken97℄ RihardKenyon.Loalstatistisoflattiedimers.Ann.Inst.HenriPoinaréProbab.
Stat., 33(5):591618,1997.
[Ken00℄ R. Kenyon. Conformal invariane of domino tiling. Ann. Probab., 28(2):759795,
2000.
[Ken01℄ R. Kenyon. Dominos and the Gaussian free eld. Ann. Probab., 29(3):11281137,
2001.
[Ken02℄ Rihard Kenyon. The Laplaianand Dira operators on ritial planar graphs.
In-vent. Math.,150(2):409439, 2002.
[Ken04℄ RihardKenyon. An introdution tothedimermodel. InShool and Conferene on
ProbabilityTheory,pages267304.ICTPLet.Notes,XVII,AbdusSalamInt.Cent.
Theoret. Phys., Trieste, 2004.
[Ken08℄ R.Kenyon.Heightutuationsinthehoneyombdimermodel. Comm.Math.Phys.,
281(3):675709, 2008.
[Kes82℄ H.Kesten. Perolation theory for mathematiians. Birkhäuser,1982.
[KO06℄ Rihard Kenyon and Andrei Okounkov. Planar dimers and Harnak urves. Duke
Math. J.,131(3):499524, 2006.
[KO07℄ R.KenyonandA.Okounkov. LimitshapesandtheomplexBurgersequation. Ata
Math., 199(2):263302,2007.
[KOS06℄ RihardKenyon,AndreiOkounkov,andSottSheeld. Dimersandamoebae. Ann.
of Math. (2),163(3):10191056, 2006.
[KPW00℄ R.W. Kenyon, J.G. Propp, and D.B. Wilson. Trees and mathings. Eletron. J.
Combin, 7(1):R25, 2000.
[Kup98℄ Greg Kuperberg. An exploration of the permanent-determinant method. Eletron.
J. Combin.,5:Researh Paper46,34 pp.(eletroni),1998.
erasedrandomwalksanduniformspanningtrees.Ann.Probab.,32(1):939995,2004.
[Ma01℄ P.A. MaMahon. Combinatoryanalysis. Chelsea PubCo, 2001.
[Mik00℄ G. Mikhalkin. Realalgebraiurves,the moment mapand amoebas. Ann. ofMath.
(2), 151(1):309326, 2000.
[MR01℄ G. Mikhalkin and H. Rullgård. Amoebas of maximal area. Int. Math. Res. Not.,
2001(9):441, 2001.
[MW73℄ B.MCoyandF.Wu. The two-dimensionalIsingmodel. HarvardUniv.Press,1973.
[Ons44℄ Lars Onsager. Crystalstatistis. I.A two-dimensional modelwithanorder-disorder
transition. Phys. Rev., 65(3-4):117149, Feb 1944.
[Per69℄ J.K. Perus. One more tehnique for the dimer problem. J. Math. Phys., 10:1881,
1969.
[Sh00℄ O.Shramm. Salinglimitsofloop-erasedrandomwalksanduniformspanningtrees.
Israel J. Math., 118(1):221288,2000.
[She05℄ Sott Sheeld. Random surfaes. Astérisque, (304):vi+175,2005.
[She07℄ S.Sheeld. Gaussianfreeeldsfor mathematiians. Probab.TheoryRelated Fields,
139(3):521541, 2007.
[Smi10℄ S.Smirnov. Conformalinvarianeinrandomlustermodels.I.Holomorphifermions
inthe Ising model. Ann. of Math. (2),172(2):14351467, 2010.
[Sos07℄ A. Soshnikov. Determinantal random point elds. Russian Mathematial Surveys,
55(5):923, 2007.
[Tes00℄ G. Tesler. Mathings ingraphs onnon-orientable surfaes. J.Combin. Theory Ser.
B, 78(2):198231,2000.
[TF61℄ HNV Temperley and M.E.Fisher. Dimer problem instatistialmehanis-anexat
result. Philosophial Magazine, 6(68):10611063, 1961.
[Thu90℄ W.P.Thurston. Conway'stilinggroups. Amer.Math.Monthly,97(8):757773,1990.
[Vel℄ Y.Velenik. Le modèle d'Ising. http://www.unige.h/math/folks/velenik/ours.html.
[W
+
07℄ W.Werneretal. Leturesontwo-dimensional ritialperolation. Arxiv,0710.0856,
2007.
[Wer09℄ W. Werner. Perolation etmodèle d'Ising. Cours spéialisés, SMF,16,2009.