Corrections

We list here typos and errors that have been found in the published version of the
book. We welcome any additional corrections you may find! We indicate page
and line numbers for the published version of the book; negative line numbers are
counted from the bottom of the page. Most typos have already been corrected in
the version available on our web page (the address is given in the footer). When
this has not been done, the relevant pages are indicated in brackets.
¯ ¯
• p. 6, l. 5: ¯ should be ¯ .
¯ ¯
U
e =U 1 U
e1 =U 1

• p. 10, l. −9: − Rv dv should be + Rv dv.

• p. 10, l. −7: s(u, v) should be s(e, v) and −R log(v/v 0 ) should be +R log(v/v 0 ).

• p. 20, l. −1: litterature should be literature.

• p. 30, l. −1: ΩΛ,Nk should be ΩΛk ,Nk .

• p. 31, l. 11: see the exercise below.

• p. 32, l. 11: the equation should read

1 def
lim Boltz (ρ) = −ρ log ρ − (1 − ρ) log(1 − ρ) .
SBoltz (Λ; N ) = shard
2V

• p. 65, l. 4, 8 and 13: min should be inf.

• p. 66, (2.11): The factor N 1/2 can be removed from the upper bound. The
same is true when applying this bound just after Exercise 2.1.

• p. 66, l. −2: min should be inf.

We thank all readers who have sent us their corrections, among which (our sincere apologies
to those we forgot): Costanza Benassi, Quentin Berthet, Adolfo Camacho Melo, Tecla Cardilli, Loren
Coquille, Yair Daon, Margherita Disertori, Hugo Duminil-Copin, Arved Niklas Fanta, Maëllie Godard,
Christian Hansch, Qidong He, Roman Kotecký, Benjamin Lees, Antoine Maier, Mirmukhsin Makhmu-
dov, Mauro Mariani, Philippe Moreillon, Sébastien Ott, Ron Peled, Elias Reutelsterz, Costanza Rojas-
Molina, Sylvie Roelly, Vedran Sohinger, Daniel Ueltschi, Zichun Ye.

© S. Friedli and Y. Velenik

www.unige.ch/math/folks/velenik/smbook
2

• p. 70, Fig. 2.4: −h/β should be −h/(2d β).

• p. 71, l. −3: ψβ should be ψCW

β
.

• p. 72, l. 5: min should be inf.

• p. 75, l. −8: 2/N can be replaced by 1/N . In the following line, 4d β/N can be
replaced by 2d β/N .

• p. 83, l. −2: "such that dr=1 |(i r − j r ) mod n|" should be "such that i − j has
P

only one nonzero component and the latter is equal to ±1 modulo n".

• p. 90, l. 3: k 0 is actually independent of h.

η
• p. 91, l. 8: the notation HΛ has never been introduced (although its meaning
should be pretty clear from the context).

• p. 92, l. 1: Λ ⊂ Zd should be Λ ⋐ Zd .

• p. 93, l. −10: magnetization should be magnetization density.

• p. 103, l. −1: “prove that 〈σ A 〉+

Λ;J,h
is increasing” should be “prove that 〈σ A 〉+
Λ;J,h
∅
and 〈σ A 〉Λ;J,h are nondecreasing”.

• p. 103, l. −1: increasing should be nondecreasing.

• p. 108,© l. 10: ∆2k−1 ⊂ Λ1n : n ≥ 1 , ∆2k ⊂ Λ2n : n ≥ 1 should be ∆2k−1 ∈ Λ1n : n ≥ 1 ,

© ª © ª © ª

∆2k ∈ Λ2n : n ≥ 1 .
ª

• p. 112, l. −5: |〈σ0 〉+

Λ
| ≤ 1 should be |〈σi 〉+
Λ
| ≤ 1.
n ;β,h n ;β,h

• p. 116: The code for Figure 3.10 has some bug that make it looks rather poor.

• p. 121, l. 1: “There exists” should be “There exist”.

def
• p. 122, l. −6: it should read “log µ = limn→∞ n1 logC n ”.

• p. 129, l. −3: “with the free boundary condition” should be “with free bound-
ary condition”.

• p. 130, l. 8: “Observing that” should be “Since”.

• p. 136, l. −7: “For any A, B ⊂ Λ” should be “For any A ⊂ Λ”. Moreover, the
claim in this exercise (Exercise 3.31) can be strengthened, see the additional
exercises for Chapter 3 below.

• p. 144, l. −9: This should be replaced by

ω0 exp h j ∈Λ σ j + β { j ,k}∈E b \{0,i } ω j ωk 1 + ω0 ωi tanh β

P © P P ª¡ ¢
ω∈Ω+ Λ Λ
¢ .
ω∈Ω+ exp h j ∈Λ σ j + β { j ,k}∈E b \{0,i } ω j ωk 1 + ω0 ωi tanh β
P © P P ª¡
Λ Λ

• p. 147, l. −5: ω̃ j (n) should be ω̃ j .

• p. 148, l. 4: ω̃ j (n) should be ω̃ j and ω j (n) should be ω j .

w w
• p. 151, l. 18: 2NΛ (E ) should be 2NΛ (E )−1 .

© S. Friedli and Y. Velenik

www.unige.ch/math/folks/velenik/smbook
 3

• p. 155: In the equation following (3.82), {i , j } should be {i } (three times) and

m m
i ←→ j should be i ←→ ∂ex Λ.

• p. 156, l. 6: PΛ;β should be P+

Λ;β
(6 times).

• p. 161, l. 11: Theorem 3.59 should be Theorem 3.58. A couple of lines be-
low, [βc (d − 1), βc (d )) should be (βc (d ), βc (d − 1)] and [βc (2), βc (3)) should be
(βc (3), βc (2)].

• p. 164, l. 4: e −c1 n should be e −c1 R .

• p. 181, l. 9: In (4.5), f Λn ;β (ρ) should be f Λn ;β (Nn /|Λn |). 3 lines below, one
might specify explicitly that ρ 7→ f β (ρ) is continuous and convex on [0, 1].
q q
N N N
• p. 183, l. 5: 2π |Λ| (1 − |Λ| ) should be 2πN (1 − |Λ| ).

1−ρ 1−ρ
• p. 183, l. −3: log( ρ ) should be |log( ρ )|.

• p. 184, l. 15: Proposition B.9 should be Lemma B.11.

• p. 184, l. −4: In (4.23), βNΛ (η) should be µNΛ (η).

• p. 185, l. 3: “β” should be “β ≥ 0”.

∂p β ∂p β
• p. 185, l. −8 ∂µ should be ∂µ (µ).

• p. 186, l. 12, 14, 15: c(1 − c) should be c 2 .

• p. 189, l. 5: “interval” should be “closed interval”.

• p. 189, l. 13: Theorem 4.5 should be Theorem 4.13.

• p. 191, l. 13: The sets in Dα (Λ) should be translates of each other.

′ ′′
• p. 192, l. 9: Nmax should be Nmax .

• p. 196, l. −7: (4.1) should be “the exponent in (4.22)”.

• p. 198, l. 12: 14 βc (d ) should be 4βc (d ).

• p. 207: The paragraph following (4.62) should read as follows:

Once more, for fixed β, once v is taken large enough, pbβvW (v) is well
approximated by the solution to
³ ¡1 1
¢ 1 ´
p+ 2 − 2β (v − 1) = β−1 ,
v2
1
which is essentially van der Waals’ expression (1.23), with a = 2 −
1
2β and b = 1.

• p. 207, l. −8: s β should be s l.g. .

• p. 208, Fig. 4.16: The maximizers should be expressed in terms of the variable
x, not the density ρ (so, ρ g should be x g , etc.).

© S. Friedli and Y. Velenik

www.unige.ch/math/folks/velenik/smbook
4

• p. 208, l. 3: f β should be f βvW (twice).

• p. 208, l. −4: ρ β (µ) should be ρ vW

β
(µ).

• p. 210, l. 4: p β should be p βvW . After (4.66), s should be s l.g. (9 times!).

• p. 214, l. 13: “a cube of sidelength ℓ2n , given by” should be “a cube given by”.

• p. 221, l. 15: e −HΛ;µ∗ (η) should be e −βHΛ;µ∗ (η) (twice).

• p. 260, l. 9 and 11: X ∋ v should be X ∋ v (twice).

• p. 262, l. 5: the semicolon should be a comma.

• p. 264, l. 15: decribe should be describe.

• p. 265: In the second part of Notation 6.2, Λ is a subset of S.

• p. 266: In Exercise 6.2, countably should be at most countably.

• p. 267: In Lemma 6.3, one should add that ϕ is measurable.

• p. 276, l. 15: A ∈ FΛ should be A ∈ F .

• p. 277, l. −11: the comma should be a colon.

• p. 279, l. 14 and 16: the sign before the term H∆ (η ∆ τ′′Λ\∆ ωΛc ) is incorrect
(twice).

• p. 280: The two lines before Definition 6.18 should belong to a new para-
graph.

• p. 284 [p. 265]: In the caption of Fig. 6.1, one can assume, for simplicity, that
A has its support in Λ.

• p. 287, l. −1: it would be better to use η i rather than ωi in the right-hand side.

• p. 290, l. −2: k ⪰ i should be k ≻ i .

• p. 291, l. 2: the last identity should be an inequality (since we are taking the
supremum over i .)

• p. 291, l. 4: the factor a k should not be present.

• p. 291, l. 13: Theorem 6.35 should be Theorem 6.31.

• p. 293, l. 6: (6.46) should be (6.37).

• p. 301, l. 8: One should replace the assumption that G (π) ̸= ∅ by the as-
sumption that π is quasilocal. This guarantees that G (π) is a closed subset
of M1 (Ω) (Lemma 6.27). (The condition G (π) ̸= ∅ is not needed anymore by
Theorem 6.26.)

• p. 308, l. 12: Corollary 3.60 should be Theorem 3.60.

# |Λ| # |Λ|
• p. 389, l. 6: Z# (Λ) = e −e Ξ# (Λ) should be Z# (Λ) = e −βe Ξ# (Λ). The same
is true 9 lines below.

© S. Friedli and Y. Velenik

www.unige.ch/math/folks/velenik/smbook
 5

′
• p. 390, l. 6: e −β∥γ∥ should be e −β∥γ ∥ .

b #stable should be ψ
• p. 391, l. 1: ψ b #n,stable .

• p. 434, l. −4: Replace “Let ϵ > 0 be such that e ϵ /(1 + m 2 ) < 1 and n be large
enough to ensure that |η i | ≤ e ϵn for all i ∈ ∂ext B(n). Then,” by “Let ϵ > 0 be
such that e ϵ /(1 + m 2 ) < 1 and, for all n large enough, |η i | ≤ e ϵn for all i ∈
∂ext B(n). Then, for all such n,”.

• p. 462, l. −15: the exterior-most expectation is missing the proper indices

η
〈·〉B(N ) (twice).

• p. 462, l. −7: J˜o j , J o j should be J˜0 j , J 0 j .

• p. 470, l. 17: the comma before Namely should be a colon.

• p. 480, l. 2: J m (i , Θ(i )) should be J im,Θ(i ) .

ν
• p. 499-500: In (10.41) and in the statement of Theorem 10.25, should be
4d
ν 1 (2π)d
d
(three occurrences). Moreover, in (10.41), should be .
4d (2π) |TL | |TL |
• p.521, l. −2 [p. 484, l. −3]: The functions ( f n )n≥1 should be uniformly strongly
convex, in the sense that infn≥1 infx∈R f n′′ (x) > 0.

• p. 528, l. −3: One should add that ϕ satisfies ϕ(ω0 ) = z 0 .

• p. 574, l. −6: c(1 − c) should be c 2 .

• p. 579, l. 8: 2d k≥r k d −1 2−k should be 2d k>r (2k + 1)d −1 2−k .

P P

• p. 579, l. −17: remove “a sequence (δn )n≥1 decreasing to 0”.

• p. 580, l. −4: Exercise 6.65 should be Exercise 6.23.

• p. 582, l. 6: Theorem 6.27 should be Lemma 6.27.

• General: We incorrectly (for reasons we don’t remember) write Marc Kac’s

name as Kać in the book.

© S. Friedli and Y. Velenik

www.unige.ch/math/folks/velenik/smbook
 © S. Friedli and Y. Velenik
www.unige.ch/math/folks/velenik/smbook
Additional exercises

Here, we suggest a variety of additional relevant exercises. To facilitate using the

latter for teaching, we do not plan on providing solutions.

Chapter 3
▶ The following two exercises are complements to Section 3.3.

Exercise. Consider the one-dimensional Ising model on the torus Tn . Show that

per ∂ per
〈σ0 〉V = ψ (β, h)
n ;β,h ∂h Vn
per
and deduce an expression for 〈σ0 〉β,h = limn→∞ 〈σ0 〉V .
n ;β,h

Exercise. Let Λn = {−n, . . . , n}. Using the transfer matrix, compute ψ+ Λn (β, h),
〈σ0 〉+
Λ ;β,h
and 〈σ0 σi 〉 +
Λ ;β,h
, as well as the corresponding limits as n → ∞.
n n

▶ The following exercises and remark are complements to Lemma 3.31:

Exercise. Prove that, in any dimension d , the function h 7→ 〈σ0 〉+

β,h
is strictly in-
creasing on R and that, for all Λ ⊂ Zd containing 0, all β ≥ 0 and all h ∈ R,

tanh(h) ≤ 〈σ0 〉+
Λ;β,h ≤ tanh(h + 4d β).

∂
Hint: Use FKG for the upper bound. Then, show that ∂h 〈σ0 〉+
Λ;β,h
≥ 1 − (〈σ0 〉+
Λ;β,h
)2 .
Deduce from that both the lower bound and the strict monotonicity.

Remark. An improved upper bound is provided in Theorem 3.53. When h is large,

improved estimates follow from the expansion in Theorem 5.10. Moreover, related
results are established in Theorem 4.12 and Proposition 10.16. ⋄

Exercise. Prove that, in any dimension d , the function β 7→ 〈σ0 〉+

β,h
is strictly in-
creasing when h > 0 or when h = 0 and β > βc (d ).

© S. Friedli and Y. Velenik

www.unige.ch/math/folks/velenik/smbook
8

▶ The following exercise is a complement to Section 3.7.2. Remember that the q-

state Potts model (q ≥ 2) has spins taking value in the set {0, . . . , q −1} and a Hamilto-
nian given by −β {i , j }∈E b δσi σ j . The corresponding Gibbs measure in B(n) with
P
B(n)
q−Potts,0
boundary condition identically equal to 0 is denoted by µB(n);β .

Exercise. 1. Check that the 2-state Potts model reduces to the Ising model at
inverse temperature β/2.

2. Check that the Hamiltonian of the Potts model is invariant under permuta-
tion of the spin values.

3. Adapt Peierls’ argument to the 2-dimensional Potts model, in order to show

that, for all β large enough,

q−Potts,0 1
lim inf µB(n);β (σ0 = 0) > ,
n→∞ q

which shows that the symmetry under permutation is broken at low temper-
atures. Hint: One way to proceed is to use contours that keep some informa-
tion on the values of the spins on each side.

▶ The following exercise is a complement to Exercise 3.31.

Exercise. Deduce from Exercise 3.31, under the same assumptions, that

|〈σ A 〉Λ;K | ≤ 〈σ A 〉Λ;K′ .

© S. Friedli and Y. Velenik

www.unige.ch/math/folks/velenik/smbook
 9

Chapter 9
▶ The following exercise is a complement to Theorem 9.14.

Exercise. The goal of this exercise is to provide an alternative proof of exponen-

tial decay of the 2-point function of O(N ) models at sufficiently high temperature.
It relies on a suitable high-temperature expansion. Fix 0, k ∈ Zd and take n large
enough to ensure that 0, k ∈ B(n).

1. Writing e βSi ·S j = e −β (e β(Si ·S j +1) − 1 + 1) and expanding the product over the
edges of B(n), show that

|〈S0 · Sk 〉∅
B(n);β
| ≤ QB(n);β (0 ←→ k),

where QB(n);β is the probability measure on subsets of EB(n) defined by

Z Y ¡ β(S (ω)·S (ω)+1)
QB(n);β (E ) ∝
¢ Y
e u v
−1 dωi
ΩB(n) {u,v}∈E i ∈B(n)

and 0 ←→ k denotes the event that 0 and k are connected.

2. Let E be a random subset of EB(n) distributed according to QB(n);β . Show that,

for any e ∈ EB(n) and E ⊂ EB(n) \ {e}, QΛ;β e ∈ E ¯ E \ {e} = E ≤ e 2β − 1.
¡ ¯ ¢

3. Deduce from this that there exists β0 > 0 such that, for any β < β0 ,

QB(n);β (0 ←→ k) ≤ e −c∥k∥2

for some c = c(β) > 0 uniformly in n.

© S. Friedli and Y. Velenik

www.unige.ch/math/folks/velenik/smbook