www.imstat.org/aihp 2009, Vol. 45, No. 4, 1020–1047
DOI: 10.1214/08-AIHP193
© Association des Publications de l’Institut Henri Poincaré, 2009
Cavity method in the spherical SK model 1
Dmitry Panchenko
Department of Mathematics, Texas A&M University, Mailstop 3386, Room 209, College Station, TX 77843, USA.
E-mail: panchenk@math.tamu.edu
Received 13 February 2008; revised 6 September 2008; accepted 15 September 2008
Abstract. We develop a cavity method for the spherical Sherrington–Kirkpatrick model at high temperature and small external field. As one application we compute the limit of the covariance matrix for fluctuations of the overlap and magnetization.
Résumé. Nous développons la méthode de la cavité pour le modèle sphérique de Sherrington–Kirkpatrick à haute température et champs externe faible. Nous illustrons la méthode par le calcul de la matrice de covariance des fluctuations des recouvrements et de la magnétisation.
MSC:60K35; 82B44
Keywords:Sherrington–Kirkpatrick model; Cavity method
1. Introduction
The cavity method in the Sherrington–Kirkpatrick model [3], as described for example in [5], is one of the most important tools developed for the analysis of the model at high temperature. Among many applications of the cavity method, the most important include self-averaging and Gaussian fluctuations of the overlap (see [4] or [2]). The present paper was motivated by an attempt to find an analogue of the cavity method for the spherical model and, surprisingly, the task turned out to be more challenging than expected. Of course, one can anticipate technical difficulties due to the fact that uniform measure on the sphere is not a product measure and decoupling one coordinate from the others, which is the idea of the cavity method, is not straightforward. As we shall see, the suggested interpolation has some new features compared to [5].
After we develop the cavity method, as one application we will study the fluctuations of the overlap and magne- tization and compute their covariance matrix in the thermodynamic limit. We stop short of proving a central limit theorem since our goal is to give a reasonably simple illustration of the cavity method.
We consider a spherical SK model with Gaussian Hamiltonian (process)HN(σ)indexed byspin configurationsσ on the sphereSN of radius√
NinRN. We will assume that 1
NEHN σ1
HN σ2
=ξ(R1,2), (1.1)
1Supported in part by NSF grant.
whereR1,2=N−1
i≤Nσi1σi2isthe overlapof configurationsσ1,σ2∈SNand where the functionξ(x)is three times continuously differentiable. A classical example of such Gaussian process, corresponding to the choice ofξ(x)=xp for integerp≥1,is thep-spin Hamiltonian
HN(σ)= 1 N(p−1)/2
1≤i1,...,ip≤N
gi1,...,ipσi1· · ·σip,
where(gi1,...,ip)are i.i.d. standard Gaussian random variables. One could also consider a linear combination ofp-spin Hamiltonians.
This model was studied in [1] and mathematically rigorous results were proved in [6]. Under the additional as- sumptions onξ,
ξ(0)=0, ξ(x)=ξ(−x), ξ(x) >0 ifx >0, (1.2)
the limit of the free energy FN= 1
NElog
SN
exp
βHN(σ)+h
i≤N
σi
λN(σ) (1.3)
was computed in [6] for arbitraryinverse temperatureβ >0 andexternal fieldh∈R.HereλN denotes the uniform probability measure onSN.The main results of the present paper will be proved for small enough parametersβandh, i.e. for very high temperature and small external field, andwithoutthe assumptions in (1.2). For example, our results will apply top-spin Hamiltonian for both even and oddp≥1.
To provide some motivation, let us first describe some implications of the results in [6]. For smallβ andh,they imply that under (1.2) the limit of the free energy takes a particularly simple form:
Nlim→∞FN= inf
q∈[0,1]
1 2
h2(1−q)+ q
1−q +log(1−q)+β2ξ(1)−β2ξ(q)
. (1.4)
The critical point equation for the infimum on the right hand side of (1.4) is h2+β2ξ(q)= q
(1−q)2. (1.5)
For small enoughβ the infimum in (1.4) is achieved atq=0 ifh=0 and at the unique solutionq of (1.5) ifh =0.
Theorem 1.2 in [6] suggests that the distribution of the overlapR1,2with respect to the Gibbs measure is concentrated nearq and by analogy with the classical SK model (see Chapter 2 in [5] or [2]) one expects that the distribution of
√N (R1,2−q) is approximately Gaussian. In the setting of the classical SK model this is proved using the cavity method and the main goal of the present paper will be to develop the analogue of the cavity method for the spherical SK model. Many computations will be more involved and, as a result, instead of using the cavity method to prove a central limit theorem we will only compute the covariance matrix of the overlaps and other related quantities. This main application is formulated in Theorem 5 in Section 3.
We note that our results also imply (1.4) without the assumption (1.2). Namely, since we will prove that for small βandhthe overlapR1,2is concentrated near the unique solutionqof (1.5), it is a simple exercise to show that in this case
Nlim→∞FN=1 2
h2(1−q)+ q
1−q +log(1−q)+β2ξ(1)−β2ξ(q)
. (1.6)
To prove this, one only needs to compare the derivatives of both sides with respect toβ.Finally, we suggest to an interested reader not familiar with the original cavity method to learn about it in Chapter 2 in [5], because we believe that one can gain much more by first learning a technically simpler case from which all the main ideas originate.
2. Cavity method
For specificity, from now on we assume thath =0 andβis small enough so thatqis theuniquesolution of (1.5). The case ofh=0 is also significantly simpler because one can use the second moment method as in [5], Section 2.2. All the results below are proved without the assumption (1.2). Given a configurationσ∈SN,we will denoteε=σN and fori≤N−1 denote
ˆ σi=σi
N−ε2
N−1,
so that a vector σˆ =(σˆ1, . . . ,σˆN−1)∈SN−1,i.e. | ˆσ| =√
N−1. We consider a Gaussian HamiltonianHN−1(σˆ) independent ofHN(σ)such that
1
N−1EHN−1
σˆ1 HN−1
σˆ2
=ξ(Rˆ1,2), (2.1)
whereRˆ1,2=(N−1)−1
i≤N−1σˆi1σˆi2.We define theinterpolating Hamiltonianby Ht(σ)=√
t βHN(σ)+√
1−t βHN−1(σˆ)+h
i≤N−1
ˆ σi
1+t
N−ε2
N−1 −1
+hε+√
1−t εzβ
ξ(q)−1
2(1−t )ε2b, (2.2)
wherezis a standard Gaussian r.v. independent ofHN andHN−1and
b=h2(1−q)+β2(1−q)ξ(q). (2.3)
Define thepartition functionalong the interpolation by Zt=
SN
expHt(σ)dλN(σ)
and for a functionf:SNn →Rdefine theGibbs averageoff with respect to the Hamiltonian (2.2) by ft= 1
Ztn
SNn
fexp
l≤n
Ht
σl
dλnN. (2.4)
Letνt(f )=Eftbe the annealed Gibbs average.
The main purpose of this interpolation is to devise a way of computingν(f ):=ν1(f )for some particular choices of f – typically, functions of the overlaps – and finding the right expression forHtis always motivated by the properties that such interpolation should have in order to make these computations work.
First of all, any cavity method aims to decouple the last coordinate ε from the other coordinates to make the computation ofν0(f )easier, in some sense. Notice the special form of the Hamiltonian (2.5) att=0,
H0(σ)=βHN−1(σˆ)+h
i≤N−1
ˆ
σi+εa−1
2ε2b, (2.5)
where we introduced the notation a=zβ
ξ(q)+h. (2.6)
The first two terms depend only onσˆ∈SN−1and, therefore, the Gibbs average (2.4) att=0 for functions of the type f1(σˆ)f2(ε)will decouple (see (2.15) below). In other words, att=0 we can think ofεandσˆas independent variables,
even thoughσˆ formally depends onε.This explains a somewhat unusual dependence ofHN−1on(σˆ1, . . . ,σˆN−1)in the above interpolation instead of the more expected dependence on(σ1, . . . , σN−1).This is probably the main new feature compared to the classical SK model on the hypercube{−1,+1}N.
Another desired property of the interpolation is thatνt(f )does not change much fort∈ [0,1]so thatν(f )can indeed be approximated byν0(f ).The proof of this will have several ingredients and, of course, the appropriate choice of parameterqin (1.5) will play an important role at some point. The main ingredients, however, are the computation and control of the derivative ofνt(f )with respect tot. This computation will be a first step toward clarifying the role of all the terms in (2.2). Forqin (1.5) we define
r=h(1−q). (2.7)
Given(σl)l≥1we denoteεl=σNl ,Rˆl=(N−1)−1
i≤N−1σˆiland defineal andal,l by al=1−ε2l, 2al,l=ξ(q)−1
2
εl2+εl2
qξ(q)+ξ(q)
+εlεlξ(q). (2.8)
Theorem 1. We have νt(f )=h
2
l≤n
νt
f al(Rˆl−r)
−h 2nνt
f an+1(Rˆn+1−r)
+2β2
1≤l<l≤n
νt
f al,l(Rˆl,l−q)
−2nβ2
l≤n
νt
f al,n+1(Rˆl,n+1−q) +n(n+1)β2νt
f an+1,n+2(Rˆn+1,n+2−q)
+νt(fR), (2.9)
where the remainderRis bounded by
|R| ≤ L N
β2+h
1+
l≤n+2
ε4l
+Lβ2
1≤l =l≤n+2
1+εl2
(Rˆl,l−q)2.
This computation will be carried out in Section 4. In order for the interpolation to be useful, the above derivative should be small. Notice that all the terms in (2.9) have factors of four typesal, al,l, (Rˆl−r)or(Rˆl,l−q)which is why our goal will be to show that the last coordinateεis of order one and the overlapRˆl,l and magnetizationRˆlare concentrated near constantsq andr. The corresponding results will be proved in Sections 5 and 6.
Theorem 2. Ifβandhare small enough,we can find a constantL >0such that νt
exp 1
Lε2
≤L (2.10)
for allt∈ [0,1].
Theorem 3. Ifβandhare small enough then for anyK >0we can findL >0such that νt
I
| ˆR1,2−q| ≥L logN
N
1/4
≤ L
NK, (2.11)
νt
I
| ˆR1−r| ≥L logN
N
1/4
≤ L
NK (2.12)
for allt∈ [0,1].
A step in the proof of Theorem 3, where the conditions (1.5) and (2.7) that define parametersqandrfinally appear, will further clarify the role of all the terms in (2.2). In some sense, this is where all the pieces will fall together.
Finally, let us explain what happens at the end of the interpolation att=0.We start by writing the integration over SN as a double integral over εand(σ1, . . . , σN−1).LetλρN denote the area measure on the sphereSNρ of radiusρ inRN,and let|SNρ|denote its area, i.e.|SNρ| =λρN(SNρ).Then,
SN
f (σ)dλN(σ)= 1
|S
√N N |
S
√N N
f (σ1, . . . , σN)dλ
√N
N (σ1, . . . , σN)
= 1
|S
√N N |
√N
−√ N
dε
1−ε2/N
S
√N−ε2 N−1
f (σ1, . . . , σN−1, ε)dλ
√N−ε2
N−1 (σ1, . . . , σN−1)
= √N
−√ N
|S
√N−ε2 N−1 |
|S
√N N |
dε
1−ε2/N
SN−1
f
ˆ σ1
N−ε2
N−1, . . . ,σˆN−1
N−ε2
N−1, ε
dλN−1(σˆ)
=aN √N
−√ N
dε
1−ε2 N
(N−3)/2
SN−1
f
ˆ σ
N−ε2
N−1, ε
dλN−1(σˆ), (2.13) whereaN= |SN1−1|/(|SN1|√
N )→(2π)−1/2as can be seen by takingf =1.In particular, if f (σ)=f1(ε)f2(σˆ),
then
SN
f (σ)dλN(σ)=aN √N
−√ N
f1(ε)
1−ε2 N
(N−3)/2
dε
SN−1
f2(σˆ)dλN−1(σˆ). (2.14) Since the Hamiltonian (2.5) decomposed into the sum of terms that depend only onεor only onσˆ,(2.14) implies that
f0= f10f20, (2.15)
where f1(ε)
0= 1 Z1
√N
−√ N
f1(ε)
1−ε2 N
(N−3)/2
exp
aε−1 2bε2
dε, (2.16)
Z1= √N
−√ N
1−ε2
N
(N−3)/2
exp
aε−1 2bε2
dε and
f2(σ)ˆ
0= 1 Z2
SN−1
f2(σˆ)exp
HN−1(σ)ˆ +h
i≤N−1
ˆ σi
dλN−1(σˆ), (2.17)
Z2=
SN−1
exp
HN−1(σˆ)+h
i≤N−1
ˆ σi
dλN−1(σˆ).
As we mentioned above, this decoupling is a crucial feature of the cavity method. Using (2.15), (2.16), we will be able to compute the momentsν0(ε1k1· · ·εnkn)for integer ki ≥0, which is an important part of the second moment computations and of the cavity method in general. The following holds. Let us recall (2.3), (2.6) and defineγ0= 1, γ1=a/(b+1)and, recursively, fork≥2
γk= a
b+1γk−1+k−1
b+1γk−2. (2.18)
Theorem 4. For small enoughβ >0, ν0
ε1k1· · ·εnkn
−Eγk1· · ·γkn≤ L
N, (2.19)
where the constantLis independent ofN.
This theorem will be proved in Section 5 below.
3. Second moment computations
We are now ready to demonstrate the promised application of the cavity interpolation, namely, the computation of the covariance matrix for the fluctuations of a certain collection of overlaps and magnetizations. Let us introduce the following seven functions
f1=(R1,2−q)2, f2=(R1,2−q)(R1,3−q), f3=(R1,2−q)(R3,4−q),
f4=(R1,2−q)(R1−r), f5=(R1,2−q)(R3−r), f6=(R1−r)2, (3.1) f7=(R1−r)(R2−r)
and letvN=(ν(f1), . . . , ν(f7)).In this section we will compute the vectorNvNup to the terms of order o(1). As we mentioned above, it is likely that with more effort one can prove the central limit theorem for the joint distribution of
√N (R1,2−q), √
N (R1,3−q), √
N (R3,4−q), √
N (R1−r), √
N (R2−r),
so the computation of this section identifies the covariance matrix of the limiting Gaussian distribution. To describe our main result let us first summarize several computations based on Theorem 4. The definition (2.18) implies that
γ1= a
b+1, γ2= a
b+1 2
+ 1
b+1, γ3= a
b+1 3
+ 3a
(b+1)2. (3.2)
The definition (2.3) and (1.5) imply that 1
b+1 = 1
1+(1−q)(β2ξ(q)+h2)=1−q.
Therefore, E a
b+1 =(1−q)Ea=(1−q)h=r, (3.3)
E a
b+1 2
=(1−q)2Ea2=(1−q)2
β2ξ(q)+h2
=q, (3.4)
where we used (1.5) again, and W:=E
a b+1
3
=(1−q)3Ea3=(1−q)3
3β2ξ(q)h+h3
, (3.5)
U:=E a
b+1 4
=(1−q)4Ea4=(1−q)4
h4+6β2h2ξ(q)+3β4ξ(q)2
. (3.6)
For simplicity of notations let us write x∼y ifx=y+O
N−1 .
Then it is trivial to check that Theorem 4 and (3.2)–(3.6) imply the following relations:
ν0(ε1)∼r, ν0(ε1ε2)∼q, ν0(ε21)∼1, ν0(ε1ε2ε3)∼W, ν0(ε1ε22)∼W+h(1−q)2, ν0(ε13)∼W+3h(1−q)2,
(3.7) ν0(ε12ε22)∼U+1−q2, ν0(ε1ε2ε32)∼U+q−q2,
ν0(ε1ε32)∼U+3q−3q2, ν0(ε1ε2ε3ε4)∼U.
Let us recall the definitionsalandal,lin (2.8). Using relations (3.7) it is now straightforward to compute the following nine quantities
ν0
a1,2(ε1ε2−q)
∼Y1, ν0
a1,3(ε1ε2−q)
∼Y2, ν0
a3,4(ε1ε2−q)
∼Y3, ν0
a1(ε1ε2−q)
∼Y4, ν0
a3(ε1ε2−q)
∼Y5, ν0
a1,2(ε1−r)
∼Y6, (3.8)
ν0
a2,3(ε1−r)
∼Y7, ν0
a1(ε1−r)
∼Y8, ν0
a2(ε1−r)
∼Y9,
whereY1, . . . , Y9are functions ofq, r, h, U, W.We omit the explicit formulas forYj’s since they do not serve any particular purpose in the sequel. Let us define a(7×7)-matrixMthat consists of four blocks
M=
M1 O1 O2 M2
, (3.9)
whereO1is a(3×2)-matrix andO2is a(4×3)-matrix both entirely consisting of zeros, M1=
⎛
⎝2β2Y1 −8β2Y2 6β2Y3 hY4 −hY5
2β2Y2 2β2(Y1−2Y2−3Y3) 6β2(−Y2+2Y3) h2(Y4+Y5) h2(Y4−3Y5) 2β2Y3 8β2(Y2−2Y3) 2β2(Y1−8Y2+10Y3) hY5 h(Y4−2Y5)
⎞
⎠,
M2=
⎛
⎜⎝
2β2(Y1−2Y2) 2β2(−2Y2+3Y3) (h/2)Y4 (h/2)(Y4−2Y5) 2β2(2Y2−3Y4) 2β2(Y1−6Y2+6Y3) (h/2)Y5 (h/2)(2Y4−3Y5)
−2β2Y6 2β2Y7 (h/2)Y8 −(h/2)Y9
2β2(Y6−2Y7) 2β2(−2Y6+3Y7) (h/2)Y9 (h/2)(Y8−2Y9)
⎞
⎟⎠.
Finally, we define a vectorv=(v1, . . . , v7)by
v1=(1−q)U+1−4q2+3q3, v2=(1−q)U+q(1−q)(1−2q), v3=(1−q)U−q2(1−q), v4=W−1
2rU+1 2r
2−6q+3q2 ,
(3.10) v5=W−1
2rU+1 2r
−2q+q2
, v6= −1
2rW+1+1
2r2(−4+3q), v7= −1
2rW+q+1
2r2(−2+q).
We are now ready to formulate the main result of this section.
Theorem 5. For small enoughβandhwe have (I−M)vTN= 1
NvT+o N−1
. (3.11)
HerevTdenotes the transpose of vectorv. In the statement of the Theorem we implicitly assumed thatβ andhare small enough so that(I−M)is invertible. Notice that each entry in the matrixMhas either a factor ofβ2orhand, therefore, for small enoughβ andhthe matrix(I−M)will indeed be invertible. Theorem 5 implies
vTN= 1
N(I−M)−1vT+o N−1
.
In the remainder of this section we will prove Theorem 5.
For each functionfl in (3.1), we will definefˆl by replacing each occurrence ofR byR,ˆ i.e.fˆ1=(Rˆ1,2−q)2, fˆ2=(Rˆ1,2−q)(Rˆ1,3−q), etc. Next, we introduce functions
f1=(ε1ε2−q)(R1,2−q), f2=(ε1ε2−q)(R1,3−q), f3=(ε1ε2−q)(R3,4−q),
f4=(ε1ε2−q)(R1−r), f5=(ε1ε2−q)(R3−r), f6=(ε1−r)(R1−r), (3.12) f7=(ε1−r)(R2−r).
As in the classical cavity method in [5], we introduce these functions because, first of all, by symmetry, ν(fl)=ν
fl
(3.13) and, second of all, emphasizing the last coordinate inflis perfectly suited for the application of the cavity method. As above, for each functionflwe will definefˆlby replacing each occurrence ofRbyR,ˆ i.e.fˆ1=(ε1ε2−q)(Rˆ1,2−q), etc.
To simplify the notations we will writex≈y whenever
|x−y| =o 1
N +ν0
(Rˆ1,2−q)2 +ν0
(Rˆ1−r)2
. (3.14)
The proof of Theorem 5 will be based on the following.
Theorem 6. For small enoughβandh,for alll≤7, ν0(fˆl)≈ν0
fl
+ν0fˆl
. (3.15)
We will start with a couple of lemmas.
Lemma 1. Iff≥0andf∞is bounded independently ofN then for anyK >0we can findL >0such that νt(f )≤L
N−K+ν0(f )
. (3.16)
Proof. The derivativeνt(f )in (2.9) consists of a finite sum of terms of the typeνt(fpεg)wherepεis some polyno- mial in the last coordinates(εl)andgis one of the following:
Rˆl,l−q, Rˆl−r, (Rˆl,l−q)2, N−1. (3.17)
Theorem 2 and Chebyshev’s inequality imply νt
I
|εl| ≥logN
≤LN−K
and combining this with Theorem 3 yields that for anygin (3.17), νt
I
|pεg| ≥N−1/8
≤LN−K. Therefore, one can control the derivative
νt(f )≤LN−K+LN−1/8νt(f )≤L
N−K+νt(f )
(3.18)
and (3.16) follows by integration.
Lemma 2. For small enoughβ andhand alll≤7we have ν(fl)≈ν0(fˆl) and ν0
fl
≈ν0fˆl
. (3.19)
Proof. We will only consider the casel=1, f1=(R1,2−q)2, since other cases are similar. We have ν
(R1,2−q)2
−ν0
(R1,2−q)2≤sup
t
νt
(R1,2−q)2
≤sup
t
LN−K+LN−1/8νt
(R1,2−q)2
≤
LN−K+LN−1/8ν0
(R1,2−q)2
, (3.20)
where in the second line we used (3.18) and then (3.16). We will use that
R1,2= ˆR1,2+s(ε1, ε2)Rˆ1,2+N−1ε1ε2, (3.21)
where
s(ε1, ε2)=
1−ε21 N
1−ε22
N
−1.
Since √
1+x−1−x 2
≤Lx2 forx∈ [−1,1] (3.22)
we have
s(ε1, ε2)+ 1 2N
ε21+ε22≤ L N2
ε41+ε42
. (3.23)
Since by (3.21)
R1,2−q=(Rˆ1,2−q)+
1−ε12 N
1−ε22
N
−1
Rˆ1,2+ 1
Nε1ε2, (3.24)
squaring both sides and using (3.23) yields (R1,2−q)2−(Rˆ1,2−q)2≤ 1
Npε| ˆR1,2−q| + 1 N2pε, where from now onpεdenotes a quantity such that
|pε| ≤L
1+
l
εl4
. Therefore,
ν0
(R1,2−q)2
−ν0
(Rˆ1,2−q)2≤ 1 Nν0
pε| ˆR1,2−q| + 1
N2ν0(pε)=o N−1
by Theorems 2 and 3. Thus, (3.20) implies the first part of (3.19). To prove the second part of (3.19) we notice that f1− ˆf1=(ε1ε2−q)(R1,2− ˆR1,2)≤ 1
Npε
by (3.21) and (3.23). Since each term in the derivativesν0(f1)andν0(fˆ1)will contain another factor from the list
(3.17), Theorems 2 and 3 imply the result.
Proof of Theorem 6. We start by writing ν
fl
−ν0
fl
−ν0
fl≤sup
t
νt fl. If we can show that
sup
t
νt
fl≈0 (3.25)
and, thus,ν(fl)≈ν0(fl)+ν0(fl),then Lemma 2 and (3.13) will imply ν0(fˆl)≈ν(fl)=ν
fl
≈ν0
fl +ν0
fl
≈ν0
fl
+ν0fˆl ,
which is precisely the statement of Theorem 6. To prove (3.25) we note that by (2.9) the second derivativeνt(fl)will consist of the finite sum of terms of the typeflpεg1g2whereg1, g2are from the list (3.17). Clearly,
|g1g2| ≤L 1
N2 +(Rˆl,l−q)2+(Rˆl−r)2
and since eachflcontain another small factor(Rl,l−q)or(Rl−r),Theorems 2 and 3 imply (3.25).
We are now ready to prove Theorem 5.
Proof of Theorem 5. Let us first note thatν0(fˆl)is defined exactly the same way asν(fl)forN−1 instead ofN.In other words,
v0N:=
ν0(fˆ1), . . . , ν0(fˆ7)
=vN−1
and, therefore, it is enough to prove that (I−M)v0TN = 1
NvT+o N−1
. (3.26)
Replacing 1/N by 1/(N−1) on the right hand side is not necessary since the difference is of orderN−2. Each equation in the system of Eqs (3.26) will follow from the corresponding equation (3.15). Namely, we will show that
ν0
f1 , . . . , ν0
f7
≈ 1
Nv and ν0fˆ1
, . . . , ν0fˆ7T
≈Mv0TN. (3.27)
Then (3.15) will imply thatv0TN ≈N−1vT+Mv0TN. However, since the definition (3.14) means that the error in each equation is of order o(N−1+ν0(fˆ1)+ν0(fˆ6)), this system of equation can be rewritten as
(I−M−EN)v0TN = 1 Nv+o
N−1 ,
where the matrixENis such thatEN =o(1). Therefore, whenever the matrixI−Mis invertible (for example, for smallβandh) we have forN large enough
v0TN =(I−M−EN)−1 1
NvT+o N−1
= 1
N(I−M)−1vT+o N−1
.
Hence, to finish the proof we need only to show (3.27). We will only carry out the computations forl=1 since all other cases are similar. Let us start by proving thatν0((ε1ε2−q)(R1,2−q))≈v1.Using (3.24) and (2.15), we write
ν0
(ε1ε2−q)(R1,2−q)
= 1 Nν0
ε1ε2(ε1ε2−q)
+ν0(ε1ε2−q)ν0(Rˆ1,2−q)
+qν0
(ε1ε2−q)
1−ε21 N
1−ε22
N
−1
+ν0
(ε1ε2−q)
1−ε21 N
1−ε22
N
−1
ν0(Rˆ1,2−q).
Using (3.23), one can bound the last term by 1
Nν0(pε)ν0(Rˆ1,2−q)=o N−1 by Theorems 2 and 3. The term
ν0
(ε1ε2−q) ν0
(Rˆ1,2−q)
=o N−1
by Theorem 3 and the second relation in (3.7), i.e.ν0(ε1ε2−q)∼0.Finally, we use
1−ε12 N
1−ε22
N
−1+ ε21 2N + ε22
2N ≤ 1
N2pε to observe that
qν0
(ε1ε2−q)
1−ε21 N
1−ε22
N
−1
≈ − 1 2Nqν0
(ε1ε2−q)
ε12+ε22
= −1 Nqν0
(ε1ε2−q)ε12 by symmetry and, therefore,
ν0
(ε1ε2−q)(R1,2−q)
≈ 1 N
ν0
ε1ε2(ε1ε2−q)
−qν0
(ε1ε2−q)ε12
= 1 N
ν0
ε12ε22
−qν0(ε1ε2)−qν0
ε13ε2
+q2ν0
ε21
≈v1
by using (3.7) and comparing with the definition ofv1in (3.10).
Next, we need to show the second part of (3.27) forl=1,i.e.
ν0
(ε1ε2−q)(Rˆ1,2−q)
≈ Mv0TN
1,
where(·)1denotes the first coordinate of a vector. We use (2.9) forn=2 to writeν0((ε1ε2−q)(Rˆ1,2−q))as hν0
a1(ε1ε2−q)(Rˆ1,2−q)(Rˆ1−r)
−hν0
a3(ε1ε2−q)(Rˆ1,2−q)(Rˆ3−r) +2β2hν0
a1,2(ε1ε2−q)(Rˆ1,2−q)2
−8β2hν0
a1,3(ε1ε2−q)(Rˆ1,2−q)(Rˆ1,3−q) +6β2hν0
a3,4(ε1ε2−q)(Rˆ1,2−q)(Rˆ3,4−q) +ν0
(ε1ε2−q)(Rˆ1,2−q)R
≈2β2Y1ν0(fˆ1)−8β2Y2ν0(fˆ2)+6β2Y3ν0(fˆ3)+hY4ν0(fˆ4)−hY5ν0(fˆ5)+ν0
fˆ1R
= Mv0TN
1+ν0
fˆ1R ,
where in second to last line we used (3.8) and the last line follows by comparison with the definition ofMin (3.9).
Finally, since clearlyν0(fˆ1R)≈0 by Theorems 2 and 3, this finishes the proof of Theorem 5.
4. Derivative along the interpolation
In this section we prove Theorem 1. We start by writing νt(f )=E
f
l≤n
∂
∂tHt
σl
t
−nE
f ∂
∂tHt(σn+1)
t
(4.1)
and
∂
∂tHt(σ)= β 2√
tHN(σ)− β 2√
1−tHN−1(σˆ)+h
i≤N−1
ˆ σi
N−ε2
N−1 −1
− 1 2√
1−tεzβ
ξ(q)+1
2ε2b. (4.2)
In order to use a Gaussian integration by parts (see, for example, (A.41) in [5]) we first compute the covariance Cov
Ht
σ1 , ∂
∂tHt σ2
=β2 2
N ξ(R1,2)−(N−1)ξ(Rˆ1,2)−ε1ε2ξ(q)
by (1.1) and (2.1). We will rewrite this using Taylor’s expansion ofξ(R1,2)nearRˆ1,2.By assumption,ξis three times continuously differentiable and (3.21), (3.23) imply
ξ(R1,2)−ξ(Rˆ1,2)−ξ(Rˆ1,2)(R1,2− ˆR1,2)≤ L N2
ε41+ε42
and
ξ(R1,2)−ξ(Rˆ1,2)+ 1 2N
ε21+ε22Rˆ1,2ξ(Rˆ1,2)− 1
Nε1ε2ξ(Rˆ1,2) ≤ L
N2
ε41+ε42 . Therefore,
N ξ(R1,2)−(N−1)ξ(Rˆ1,2)=ξ(Rˆ1,2)−1 2
ε21+ε22Rˆ1,2ξ(Rˆ1,2)+ε1ε2ξ(Rˆ1,2)+R1, (4.3)
where from now onR1will denote a quantity such that
|R1| ≤L N
1+
l≤n+2
εl4
.
Sinceξ is three times continuously differentiable,
ξ(Rˆ1,2)−ξ(q)=ξ(q)(Rˆ1,2−q)+R2, ξ(Rˆ1,2)−ξ(q)=ξ(q)(Rˆ1,2−q)+R2, Rˆ1,2ξ(Rˆ1,2)−qξ(q)=
ξ(q)+qξ(q)
(Rˆ1,2−q)+R2, whereR2denotes a quantity such that
|R2| ≤L(Rˆ1,2−q)2.
Using this in (4.3) and recalling the definition ofal,l in (2.8) we get Cov
Ht
σl , ∂
∂tHt σl
=β2 2
2al,l(Rˆl,l−q)−1 2
ε2l +ε2l
qξ(q)+ξ(q)
+β2R3, (4.4)