MEYER SETS, PISOT NUMBERS, AND SELF-SIMILARITY IN

SYMBOLIC DYNAMICAL SYSTEMS

VALÉRIE BERTHÉ AND REEM YASSAWI

Abstract. Aperiodic order refers to the mathematical formalisation of quasicrystals. Sub-

stitutions and cut and project sets are among their main actors; they also play a key role in
the study of dynamical systems, whether they are symbolic, generated by tilings, or point
sets. We focus here on the relations between quasicrystals and self-similarity from an arith-
metical and dynamical viewpoint, illustrating how efficiently aperiodic order irrigates various
domains of mathematics and theoretical computer science, on a journey from Diophantine
approximation to computability theory. In particular, we see how Pisot numbers allow the
definition of simple model sets, and how they also intervene for scaling factors for invariance
by multiplication of Meyer sets. We focus in particular on the characterisation due to Yves
Meyer of Pisot and Salem numbers as the parameters of the dilations preserving some Meyer
set.

1. Introduction
In his beautiful text from the book devoted to the Abel Laureates [HP19], Y. Meyer recalls:
“I read the extraordinary book Ensembles parfaits et Séries trigonométriques by J.P. Kahane
and R. Salem. I was enthusiastic. The role played by Pisot numbers in the problem of
uniqueness for trigonometric series had been discovered by Salem and Zygmund and was
detailed in this book. (. . . ) I was fascinated by the interplay between number theory and
harmonic analysis which is so profound and beautiful in this result. I decided to extend the
work of Salem and Zygmund to the problem of spectral synthesis.” Thus Yves Meyer began
a journey that would culminate in the two books [Mey70, Mey72], and that would lead him
to quasicrystals and tilings via model sets.
We recall that a real algebraic integer β > 1 is a Pisot number (also called a Pisot-
Vijayaraghavan number) if all of its other algebraic conjugates λ satisfy |λ| < 1. Yves Meyer
thus started his journey by extending the work of Salem and Zygmund to the problem of
spectral synthesis for the Cantor set Eβ with β > 2, where
( )
X
Eβ := εi β −i with εi ∈ {0, 1} .
i
The aim of spectral synthesis is to recreate a function from its Fourier coefficients: a compact
set Γ is called a set of spectral synthesis if any continuous bounded function f in L1 with
spectrum in Γ (i.e., whose Fourier transform has support in Γ), can be approximated in
the weak star topology by the set PΓ = { γ∈Γ aγ e2iπγx } of finite trigonometric polynomials
P
with support in Γ. As was similarly done for the case of sets of uniqueness, namely those
sets outside of which a Fourier series uniquely determines a function, Yves Meyer showed
Date: April 5, 2023.
This work was supported by the Agence Nationale de la Recherche through the project “Codys” (ANR-18-
CE40-0007), and the EPSRC grant numbers EP/V007459/2 and EP/S010335/1.
1
2 V. BERTHÉ AND R. YASSAWI

that Pisot numbers also play a crucial role in the problem of spectral synthesis. Indeed, for
β > 2, Yves Meyer proved the striking fact that Eβ is a set of spectral synthesis if β is a
Pisot number. Strong spectral synthesis even holds with respect to uniform convergence on
compact sets, as stated below.
Theorem 1.1. [Mey72, Theorem V, Chapter VII] Let β > 2 be a Pisot number. For each
bounded continuous function f : R → C whose spectrum lies in Eβ , there exists a sequence
(Pk )k of finite trigonometric sums such that
• for each k, the frequencies of Pk (i.e., its spectrum) belong to the set of finite sums
{ kj=1 εj β −j with εi ∈ {0, 1}},
P

• the sequence (Pk )k converges uniformly to f on each compact subset of R, and

• supR |Pk ] converges to supR |f |.
This resolved a longstanding conjecture of Salem and Zygmund and extended the result
that Carl Herz had obtained for the triadic Cantor set E3 . In addition, the tools that Yves
Meyer developed allowed him to discern the relationship between the arithmetical nature of a
set Λ, not necessarily compact, and the properties of the finite trigonometric polynomials with
support in Λ. As an illustration, let us start with the simplest case Λ = Z. The almost periodic
functions
P with spectrum in Z, i.e., those that are uniform limits of trigonometric polynomials
in { n∈Z an e2iπnx }, behave like periodic functions. Yves Meyer introduced a class of sets
Λ, considered as pseudo-lattices, which have the property that any almost-periodic function
with spectrum (frequencies) in Λ behaves like a periodic function. Yves Meyer investigated
these sets Λ in the context of Diophantine approximation for the study of Pisot and Salem
numbers in the books [Mey70, Mey72]; see also [Mey95, Mey12]. Here, in view of Theorem
1.1, these sets are given by the discrete grids Λβ (obtained from the sets Eβ up to a suitable
rescaling) with β > 2 and
( )
X
i
Λβ = εi β with εi ∈ {0, 1}, F finite ;
i∈F
they provide explicit approximation schemes, underlying Theorem 1.1. They are examples of
sets that are now referred to as Meyer and model sets. Let us briefly recall their definitions.
Definition 1.2 (Delone and Meyer sets). A Delone set is a subset Λ of Rn which is both
uniformly discrete and relatively dense, i.e., there exist r, R > 0 so that each ball with radius
r contains at most one point of Λ and each ball of radius R contains at least one point of Λ. A
Delone set Λ is a Meyer set if there is a finite set F such that Λ − Λ ⊂ Λ + F , or equivalently,
if Λ − Λ is also a Delone set [Mey72, Lag96, Moo97].
Meyer sets play the role of lattices for crystalline structures. With atoms being located at
points of Λ, the set Λ−Λ corresponds to interatomic distances, and restrictions on interatomic
distances produces long-range order.
In [Mey72] Yves Meyer introduced cut and project schemes as explicit constructions of
particular Meyer sets, i.e., model sets. In order to introduce the idea in the simplest setting,
we consider a full rank lattice L in Rd , i.e., a lattice generated by a set of d linearly independent
vectors, together with a decomposition of Rd = Rn × Rk , with d = n + k, where Rn is called
the physical space and Rk is called the internal space. Let πphy and πint denote the natural
projections onto Rn and Rk respectively. We further assume that the restriction πphy to
L is injective and that πint (L) is dense in Rk . We project the lattice L ⊂ Rd onto the
 MEYER 3

physical space. The lattice L thus brings some order and regularity with its underlying
higher-dimensional periodicity; its projection onto the internal space provides the aperiodic
structure.
Given a cut and project scheme, model sets are then formed by projections together with
a way of selecting points. This selection is done thanks to an acceptance window W that lives
in the internal space Rk . It is usually assumed to be compact (or relatively compact) and
with non-empty interior. We now can define model sets.
Definition 1.3 (Model sets). A subset Λ of Rn is called a model set if there exist a cut and
project scheme (Rn × Rk , L) and a compact set W of Rk with non-empty interior such that
Λ = {πphy (P ); P ∈ L, πint (P ) ∈ W }.
A model set Λ is a Delone set: the uniform discreteness comes from the compactness of
the acceptance window W and the relative denseness from its non-empty interior. One of the
main features of model sets is that they provide pure point diffraction (or, in more dynamical
terms, pure discrete spectrum), such as highlighted in [Moo].
Meyer established the following powerful connections. For details, see [Mey72, Mey95] and
for more about model sets, see for instance the surveys [Moo97, Moo00, LM01] and [BG13,
Chap. 7].
Theorem 1.4. Model sets are Meyer sets. Conversely, if Λ is a Meyer set, then it is a
relatively dense subset of a model set, i.e., there exist a model set M and a finite set F such
that Λ ⊂ M + F .
The work of Meyer on model sets found a second life with the discovery of quasicrystals.
In 1974, Penrose discovered his tilings1, and in 1981 the connection between these tilings and
Meyer’s work was made explicit thanks to the work of de Bruijn [dB81], who proved that
the vertices in the Penrose tilings formed a model set. Michel Duneau, Denis Gratias, André
Katz, and Robert Moody then discovered that the quasicrystals described by Dan Shechtman
can be modeled using model sets.
A quasicrystal is a physical solid whose atoms are arranged as a translation invariant lattice
which does not satisfy the crystallographic restriction. Shechtman’s discovery of quasicrystals
in 1984 [SBGC84] mobilised the material sciences and physics to provide suitable models for
the description of quasicrystals. Mathematicians came to realise that the theory of aperiodic
tilings, with roots in both the theory of Wang tiles (Section 6), and also in recreational math-
ematics, where R. Penrose’s aperiodic tilings displayed pentagonal symmetry and thus could
not satisfy the crystallographic restriction, provided a context in which to study mathemati-
cal models of quasicrystals. In particular, one studies these physical objects by looking at an
appropriate cut and project set of their atom locations, and this leaves us with Meyer sets.
Nowadays the term aperiodic order, which first appeared in Moody’s work [Moo97], refers to
the study of mathematical models of quasicrystals, continuous or discrete. As with the study
of quasicrystals, aperiodic order is the study of mathematical objects which simultaneously
have long-range order and aperiodicity. The connections to Meyer’s concepts and to aperiodic
tilings are now well recognised and established, in particular thanks to R. V. Moody [Moo00],
and the fundamental contributions to aperiodic order brought by Yves Meyer developed in his
founding books [Mey70, Mey72] continue to nourish the field. See also [GQ, Moo, OU] in this
1A tiling is a covering of a surface (usually a Euclidean plane) using copies of geometric tiles, placed next
to each other, without holes or overlaps.
4 V. BERTHÉ AND R. YASSAWI

volume. This area has now exploded and covers a wide range of domains such as illustrated
by the books [Sen95, BM00, KLS15] and the book series Aperiodic order by M. Baake and
U. Grimm, in particular [BG13].
The notion of long-range order is crucial here: crystallographers now mean an ordering
of atoms which produces a diffraction pattern with sharp bright spots, such as confirmed
by the general definition of crystals adopted by the Crystallographic Union in 1992. These
diffraction patterns translate mathematically to the study of the spectrum of point sets. In
a nutshell, aperiodic order goes with diffraction and pure discrete spectrum, as highlighted
in the contribution by R. Moody in this volume [Moo]. But the beauty and the strength of
aperiodic order is that it goes well beyond spectral considerations.
There is in particular a further crucial aspect here, namely the prevalence of self-similarity
in aperiodic order. In mathematics, a self-similar object is exactly or approximately similar
to a part of itself, or it can be divided into smaller copies of itself; hence it looks roughly the
same on any scale and is invariant upon being scaled larger or smaller. It can be that parts of
a figure are small replicas of the whole, or else a self-similar group. The paradigmatic example
is a compact set K which is the solution of an iterated function system, i.e., K = ∪Si (K),
where i belongs to a finite set and where the Si ’s are contractions. A classical way to generate
self-similar objects is by iterating a rule. For example, one can iterate functions, to obtain
fractals such as the Sierpinski gasket, one can substitute a side of a triangle by a polygonal
line, such as in the Koch curve. Among the main methods for generating mathematical
models of quasicrystals are substitutions which will be the main object of Section 4. In
aperiodic order we iterate these substitutions to obtain self-similar sequences, points sets or
tilings. Indeed, in tiling theory, a tile represents an atom and a tiling represents large atomic
configurations. In point sets, points are idealisations of atomic positions, and letters are used
to code interatomic distances, and, considering specially chosen control points in the tiles
allows one to equivalently associate Delone point sets with tilings.
Algebraicity provides a particularly relevant arithmetic expression of self-similarity, and
algebraic numbers and particularly Pisot numbers are at the heart of aperiodic order. The
aim of this survey is twofold. We want to stress the impact of Pisot numbers in aperiodic order,
and also to highlight the explosion of the involved domains, by focusing on an approach based
on dynamical systems. The involved mathematical fields cover here an impressive spectrum,
with the toolbox of aperiodic order containing Diophantine approximation and arithmetic,
together with topological methods via topological dynamics or cohomology, computabillity
theory, spectral theory, algebra, discrete geometry and harmonic analysis. There are several
reasons that explain this abundance. Firstly, aperiodic order applies to various types of
quasiperiodic structures which can serve as models for quasicrystalline structures, whether
they are geometric with tilings and point sets, or symbolic with infinite words. Secondly, the
two main types of mathematical models for quasicrystals, namely model sets obtained via cut
and project schemes, and substitutions, are prominent concepts in the study of dynamical
systems, allowing dynamicists to express self-similarity in both arithmetic and geometric
terms. As an illustration, consider the role played by substitutions via renormalization in the
study of interval exchanges [Yoc06].

In this survey we made the choice to focus on several complementary topics that illustrate
the strength of the concepts of cut and project schemes and substitutions in the dynamics of
aperiodic order. In Section 2 we see how Pisot numbers allow the definition of simple model
sets, and we recall the characterisation, due to Yves Meyer, of Pisot and Salem numbers as the
 MEYER 5

parameters of the dilations preserving some Meyer set: this is Theorem 2.2 and it will serve
as a common thread. In Section 3 we discuss how Meyer and model sets are provided by Pisot
numbers and beta-numeration. Substitutions, their definitions, stability by multiplication and
self-similarity are considered in Section 4. In Section 5 we discuss the spectral properties of
substitutions and the Pisot conjecture, simple one-dimensional model sets given by Sturmian
words and more exotic acceptance windows provided by Rauzy fractals. In Section 6, we
discuss how computabillity enters the picture with the question of the existence of local rules
for aperiodic tilings and the domino problem.

Acknowledgements. We would like to thank Nathalie Aubrun, Thomas Fernique, Pierre-

Antoine Guihéneuf, Étienne Moutot, Wolfgang Steiner and Pascal Vanier for their careful
reading and their suggestions. We also thank Uwe Grimm for kindly allowing us to use Figure
2.6. Lastly, we thank the anonymous referees for their careful reading and their constructive
suggestions.

2. Pisot numbers and aperiodic order

Algebraic numbers, and particularly Pisot numbers, are at the heart of aperiodic order.
Algebraicity intervenes in two ways: firstly, Pisot numbers allow us to define model sets Λ,
and secondly, scaling invariance, i.e., the existence of a scaling factor β with βΛ ⊂ Λ, is a
step towards self-similarity. This is seen in Theorem 2.2, a striking result by Meyer which
has led to a rich theory in the study of associated dynamical systems, some of which we will
exposit in later sections. Let us first recall the following definition.
Definition 2.1 (Salem number). A real algebraic integer β > 1 is a Salem number if at least
one of its algebraic conjugates λ satisfies |λ| = 1 while the other conjugates have modulus
smaller than 1.
The following result will be a common thread throughout this survey. See also [Lag99] for
a similar statement for Delone sets.
Theorem 2.2. [Mey95, Theorem 6] If Λ is a Meyer set, β > 1 a real number and if βΛ ⊂ Λ,
then β is either a Pisot number or a Salem number. Conversely, for each dimension n and
for each Pisot or Salem number β, there exists a model set Λ ⊂ Rn such that βΛ ⊂ Λ.
The proof of the first statement goes via the concept of harmonious sets which provide a
further characterisation by duality of Meyer sets.
Definition 2.3 (Harmonious set). A harmonious set is a subset F of a locally compact abelian
group G such that every weak character2 on the subgroup generated by F may be approxi-
mated uniformly by a continuous character on G.
For more on the subject, see [Mey72], [Mey95], [Mey20]. One considers here characters of
the additive group generated by the Meyer set Λ, equipped with the discrete topology. The
following beautiful result tells us that harmonious sets are non-other than Meyer sets (see
[Mey72, Chap. II Section 5], [Mey95, Theorem 4]).
Theorem 2.4. Let Λ be a Delone set. Then Λ is a Meyer set if and only if Λ is harmonious.
2A weak character is a homomorphism for which no continuity is required.
6 V. BERTHÉ AND R. YASSAWI

Using Theorem 2.4, we can prove the first part of Theorem 2.2. We follow here [Mey95].
Take a Meyer set Λ and let β > 1 be such that βΛ ⊂ Λ. Take a non-zero element λ0 in Λ.
Then the set {β k λ0 : k ≥ 0} is included in Λ. We then note that a subset of a harmonious
set is again a harmonious set, and that the set {β k λ0 : k ≥ 0} is harmonious in the one-
dimensional space λ0 R. Since Λ is a Meyer set, then Λ is a harmonious set. We now can
conclude by using the following.
Theorem 2.5. [Mey72, Theorem XX, Chapter 1] Let β > 1. The set of powers {β k , k ≥ 0}
is harmonious if and only if β is a Pisot or a Salem number.
This sketch of proof is an illustration of the fact that “weak characters pave the road
which goes from Bochner to Shechtman”, as nicely phrased by Yves Meyer in [Mey20]. He
moreover adds: “On the way we are visiting coherent sets of frequencies, harmonious sets,(...)
and finally we arrive at model sets”, a visit to coherent sets of frequencies that we will also
make in Section 3. We will not say more on the first statement of Theorem 2.2, but rather
concentrate on the second one, that is, the construction of the model set related to β. But
before providing some elements of the proof of the second statement, we first illustrate it with
a paradigmatic and illustrative example of how algebraic numbers provide us with the first
and simplest examples of model sets. We later revisit this running example in other settings
where aperiodic order is studied. Although it has a simplicity which does not reflect the
subtleties needed in the general case, it nevertheless gives the reader a flavour of why these
results may be true. See also [GQ] in this volume.
Example 2.6. Consider the golden ratio φ, which is the largest root of x2 − x − 1 = 0, and its
algebraic conjugate φ′ , and define the lattice
L := {m(1, 1) + n(φ, φ′ ) : m, n ∈ Z} = {(m + nφ, m + nφ′ ) : m, n ∈ Z}.
In the first coordinate (the physical space), the lattice L projects injectively by πphy (L) onto
Z[φ] in R, and in the second (the internal space), its projection πint (L) is dense in R. Now
take W = [−1, φ − 1]; then
Λ := {m + nφ : m, n ∈ Z, −1 ≤ m + nφ′ ≤ φ − 1}
is a model set; see Figure 2.6.
Moreover, this model set is scaling invariant under multiplication by φ. We use the fact
that φ is a quadratic unit. Indeed, assume that m + nφ ∈ Λ. As φ2 = φ + 1, we have
φ(m + nφ) = n + (m + n)φ. Similarly, n + (m + n)φ′ = φ′ (m + nφ′ ). Multiplying the
given inequality −1 ≤ m + nφ′ ≤ φ − 1 by φ′ and using φφ′ = −1 and φ + φ′ = 1 yields
φ − 1 = −φ′ ≥ n + (m + n)φ′ ≥ −1 − φ′ ≥ −1, proving the stability of Λ under multiplication
by φ.
Finally, it can be shown that this cut and project scheme generates both a one-dimensional
tiling of R using two distinct tiles, and if we label these tiles A and B, we obtain an infinite
word, i.e., an infinite concatenation of the letters A and B. We will see in Section 4 that
this discrete object is the fixed point of a substitution, and a discrete model of self-similar
aperiodicity. See also [GQ, OU] in this volume.
More generally, given two distinct irrational parameters ν, ϵ (which are not necessarily
algebraic) and a bounded acceptance interval W , one similarly gets a one-dimensional model
set {a + bν | a, b ∈ Z, a + bϵ ∈ W } ⊂ Z[ν]. As in the example above, such cut and project
schemes, obtained by selecting points with integer coordinates located within a bounded
 MEYER 7

B A A B A B A A B A B A

Figure 1. A model set generated by the golden mean φ; it consists of the

circled points on the horizontal axis. The tile length of A is φ and that of B
is 1.

distance of a line, and thus having one-dimensional physical and internal spaces, lead to one-
dimensional tilings. Depending on whether or not the length of the window belongs to Z[ϵ],
the lengths in the tiling take two or three values, according to the size of the acceptance
window [GMP03], and if we code the intervals with distinct letters, we obtain two- or three-
letter infinite words. Such codings correspond either to Sturmian words (see Section 5.2), in
the two interval case, or to codings of so-called three-interval exchanges in the three-interval
case. Special attention has been given to the combinatorial properties of this particularly rich
class of infinite words: see for instance [GMP03, GMP06] and the survey [PM07]. Such a
model set has a priori no scale invariance. Imposing the requirement of self-similarity implies
that ϵ is a quadratic integer and ν = −ϵ′ , i.e., the algebraic conjugate of ϵ and the scaling
factor γ must be a quadratic Pisot number in the same algebraic field Q(ϵ). This brings us
back to Meyer’s connection between self-similar model sets and Pisot numbers in Theorem
2.2.
Pisot numbers allow the extension of such a simple and seminal construction to higher
dimensions, which is a key point in the proof of the second statement of Theorem 2.2, which
we now summarise from [Mey95], and which is illustrated with Example 2.6. Briefly, when
β is a Pisot or a Salem number, the construction of a model set is algebraically realised by
taking the lattice L to be the ring of algebraic integers in Q(β) and the canonical embeddings
σi associated with the (other) algebraic conjugates of β provide the projection πint onto
the internal space, with the projection πphy onto the physical space being the identity. The
acceptance window is given by the conditions |σi (x)| ≤ 1. The model set is thus the set of
algebraic integers λ in Q(β) such that |σi (λ)| ≤ 1 for all i ≥ 2. This approach by Yves Meyer
is now classical and has had a particularly fruitful influence on subsequent constructions.

3. Pisot numbers and beta-numeration

Arithmetic
Pexamples of Meyer sets which clearly have scaling invariance are provided by the
sets Λβ = ε β i with ε ∈ {0, 1}, F finite that we met in the introduction, for β > 2
i∈F i i
Pisot. These grids have remarkable properties that lead to the notion of a coherent set of
frequencies (see [Mey72, p.110]).
Definition 3.1 (Coherent set of frequencies). A set Λ of real numbers is a coherent set of
frequencies if there exist a constant C > 0 and a compact set K of real numbers such that
sup |P (x)| ≤ C sup |P (x)|
R K
8 V. BERTHÉ AND R. YASSAWI

for all finite trigonometric sums P with frequencies in Λ.

For more on the contributions of Yves Meyer in this setting, see the enlightening text by
A. Cohen in [HP19] and see also [Mey20] which revisits the relationship between coherent
sets of frequencies and Bochner’s property.
The following result has to be compared with Theorem 2.5.
Theorem 3.2. [Mey72, Proposition 7, Chapter 8] Let β > 2. The following are equivalent:
• the number β is a Pisot number;
• the set Λβ is harmonious;
• the set Λβ is a coherent set of frequencies.
Note that this statement does not mean that Λβ is a Meyer set since it is not relatively
dense (β > 2), and hence not Delone. Indeed, consider the strictly increasing sequence (xn ) of
numbers that have at least one representation as an element of Λβ , and then its first difference
sequence (xn+1 − xn ). The following holds.
Proposition 3.3. [FW02, Lemma 1.1] If β > 2, and (xn ) is an increasing sequence of
numbers that have at least one representation as an element of Λβ , then its first difference
sequence takes arbitrarily large values.
More generally, sets of finite sums of the form ΛF = { i εi β i | εi ∈ F }, for specific finite
P
sets F of non-negative integers, when β is assumed to be a Pisot number, have been widely
considered. Researchers in arithmetic and numeration came indeed to notions similar to
Meyer sets independently via base expansions. In particular, the study of the first difference
sequence was initiated in the fundamental work [EJK90] for ΛF = Λβ with 1 < β < 2 and
F = {0, 1}. The study of the limit inferior of the first difference sequence has attracted much
attention. In particular, lim inf xn+1 − xn > 0 if and only if β is a Pisot number; see [Bug96]
and also [EJK98].
Returning to the golden mean Example 2.6, if we take β = φ and F = {0, 1}, one can check
that ΛF = Z[φ] and that the entries of the first difference sequence are either 1 or φ − 1. Thus
ΛF is nothing other than our lattice L and the entries 1 and φ − 1 of the difference sequence
give us the lengths of the tiles that we saw in Figure 2.6, when scaled by a factor of φ.
The case where β ≥ 2 and F = {1, · · · , m} has also been widely studied. P See e.g.
[AK13], and the references therein, for the study of set of all differences εi β i where
εi ∈ {0, 1, ±1, · · · , ±m} in relation with singular Bernoulli convolution distributions; note
that this latter set gives information about ΛF − ΛF ; in particular the absence of accumu-
lation points in this set is proved to be equivalent to β being Pisot or β ≥ m + 1 [AK13,
Theorem 1.1]. In the language of aperiodic order, this gives the following:
Proposition 3.4. Let β > 1. For F = {0, 1, · · · , m}, the set {P (β); P ∈ F [X]} is a Meyer
set if and only if β ≤ m + 1 and β is a Pisot number.
Let us provide a sketch of the proof of Proposition 3.4. We first briefly recall how the
Pisot property intervenes here as being crucial in guaranteeing the uniform discreteness of
ΛF = {P (β); P ∈ F [X]} for F a finite set of integers. This is a classical argument known
as Garsia’s Lemma; see [Gar62, Lemma 1.51] or [Sol97, Lemma 6.6]. If β is assumed to be
Pisot, there exists C such that |P1 (β (i) ) − P2 (β (i) )| ≤ C for any algebraic conjugate β (i) ̸= β
and any P1 , P2 ∈ F [X]. Next, since i (P1 − P2 )(β (i) )(P1 − P2 )(β) ∈ Z and is non-zero for
Q
P1 , P2 ∈ F [X] with P1 (β) ̸= P2 (β), we deduce a positive uniform lower bound for P1 (β)−P2 (β)
with P1 (β) ̸= P2 (β), from which we conclude uniform discreteness.
 MEYER 9

It remains to consider the property of relative denseness where the cases β ≤ m + 1 and
β > m + 1 have to be distinguished by [FW02, Lemma 1.1]. Indeed, if β > m + 1, the
difference sequence (xn+1 − xn ) takes arbitrarily large values (this extends Proposition 3.3),
which prevents relative denseness (see the proof of [EK98, Lemma 2.1] such as stressed in
[FW02]). Next if β ≤ m + 1 and β is a Pisot number, the difference sequence can take only
finitely many distinct values; it has even a strong combinatorial structure such as described
in [FW02], where it is shown that the difference sequence can be generated by a substitution
over a finite alphabet (see Section 4.1).
Lastly, assume that ΛF = {P (β); P ∈ F [X]} is a Meyer set. As βΛF ⊂ ΛF , we deduce
from Theorem 2.2 that β is a Pisot or a Salem number. By the same proof as that of Theorem
2.2 and also by Theorem 2.4, ΛF is harmonious, and so Λβ is also harmonious, from which
we deduce that β is a Pisot number, by Theorem 3.2.

Thus researchers have studied the sets ΛF for over fifty years, due to their connections to
quasi-crystals, infinite Bernoulli convolutions and expansions in non-integer bases. Dynami-
cists have also studied these objects, via the beta-transformation. The beta-numeration is a
numeration system that allows the representation (in a greedy way) of real numbers as sums
of powers of a real number β, with β > 1, in the same way as real numbers can be expanded
in decimal numeration. It was introduced and studied by A. Renyi [Rén57] and W. Parry
[Par60] in dynamical terms. Indeed, given a positive irrational number β > 1, consider the
β-transformation Tβ acting on the unit interval defined by Tβ (x) = {βx}. Digits of x in its
β-expansion are then obtained as ai = ⌊βTβ−i x⌋, for i ≥ 1 (hence ai ≤ ⌊β⌋), which yields
x = i≥1 ai β −i . The beta-numeration has been particularly well studied when β is assumed
P
to be a Pisot number, both from a dynamical and a combinatorial viewpoint; there are indeed
specific properties that occur when β is a Pisot number. Returning to Example 2.6 where
β is the golden mean, digits are in {0, 1} and the product of two consecutive digits in the
β-expansion equals 0 (since we use the greedy representation). For Pisot numbers β, the
appropriate generalisation of this statement is that the finite sequences of digits that occur
in the possible representations have a simple expression in terms of finite automata (see e.g.
[BS02]). From a dynamical viewpoint, points in the interval are replaced by the sequence of
coefficients in their beta-expansion.
Moving from the unit interval to R+ , one can also expand non-negative real numbers. In
this context, the countable set of beta-integers Zβ plays a particular role: elements of Zβ
are the integer parts of beta-expansions, that is, they are polynomials in β with coefficients
in {0, · · · , ⌊β⌋} with combinatorial constraints on their digits determined by the algebraicity
of β. In this dynamical setting, the constraints are given on the digits εi that can appear
when expressing the beta-expansion of a number. These constraints are imposed to guarantee
uniqueness of expansions, and are particularly easy to express when β is an algebraic integer.
For example, returning to Example 2.6, the equation φ2 = φ + 1 tells us that a fragment
φn+1 + φn of φ-expansion can be replaced by φn+2 . For this example then we consider Zφ to
be polynomials in φ with coefficients in {0, 1} but where we do not see consecutive monomials.
In the case where 1 < β < 2, one sometimes has Zβ = Λβ , for example in the case of our
running example where β is equal to the golden ratio. However in general one only has
Zβ ⊂ Λβ ; see [FS92].
This set Zβ is self-similar, in that βZβ ⊂ Zβ . As stressed by Meyer in [Mey12], “Model
sets are non-uniform grids”, and beta-numeration has revealed itself as Pa very efficient tool
for the modeling of families of quasicrystals thanks to the beta-grids Zβ ei , where the ei
10 V. BERTHÉ AND R. YASSAWI

are vectors of the canonical basis (see e.g. [BFGK98, VGG04] and also [Thu89]). The next
statement, due to [BFGK98, VGG04], and which also follows from arguments in [Sch80] and
[FS92, Proposition 2], is again reminiscent of Theorem 2.2.
Proposition 3.5. If β is a Pisot number, then Zβ is uniformly discrete, and Zβ ∪ (−Zβ ) is
a Meyer set. Conversely, if Zβ ∪ (−Zβ ) is a Meyer set, then β has to be a Pisot or a Salem
number.
Furthermore, the set Zβ can be endowed with laws that are close to multiplication and
addition. This allows one to recover properties that have the flavour of the algebraic rules of
lattices. For some families of Pisot numbers β, mainly Pisot quadratic units and for some cubic
Pisot numbers, an internal law can even be produced formalising this quasi-stability under
subtraction and multiplication: the set of beta-integers is then endowed with an addition law
and a multiplication law, which are compatible with the combinatorial structure which makes
it a quasiring [BFGK98]. Let us quote also [ABF05], where a method based on automata is
given to find a set F such that Zβ − Zβ ⊂ Zβ + F .

4. Substitutive dynamics and self-similarity

We have seen in Section 2 that infinite words occur naturally as codings of one-dimensional
quasicrystals. There is a particularly simple yet rich combinatorial object that allows the
expression of scale invariance and self-similarity properties: this is the notion of a substitution.
We define a substitution rigorously below, but roughly speaking, a substitution is an inflation
rule, either combinatorial or geometric, that, following the cases, replaces a letter by a word,
or a tile by a geometric pattern made of a finite union of tiles. Iteration of substitutions
generates hierarchically ordered structures such as infinite words, shifts, point sets, or else
tilings, that all display strong self-similarity properties.
Substitutions originated with the first works of A. Thue in 1906, with the so-called Prouhet-
Thue-Morse substitution, which is an illustrious example of an automatic sequence [AS03].
Substitutions generate symbolic dynamical systems and this is the setting in which we define
them, that of symbolic dynamics.
Symbolic dynamical systems are discrete dynamical systems where the space is a Cantor
space. Typical symbolic systems are shifts where the space X consists of infinite words
with entries from a finite alphabet, and the dynamics is the shift map S moving entries of
an infinite word one unit to the left. The set X must be closed in the subspace topology
inherited from the product topology on AZ , where A is given the discrete topology. Shifts
provide representations of dynamical systems: given a dynamical system we can discretise the
state space using a finite (or countable) partition, and code trajectories of points as sequences
of symbols, as first done by J. Hadamard [Had98] and M. Morse [Mor21], and the Jewett-
Krieger theorem gives us very general conditions in which we can represent a dynamical
system as a shift.
Substitutions soon outreached in an unexpected way this combinatorial and dynamical
setting. Indeed, beginning in the sixties, substitutions were used to input computation in
tilings, enabling researchers to produce the first examples of aperiodic tilings, such as Robin-
son tilings, and thereby proving the undecidability of the domino problem (see Section 6).
Also, in the eighties, Pisot substitutions attracted much attention in the context of aperiodic
order as mathematical models of quasicrystals displaying self-similarity, yielding in particular
the Pisot substitution conjecture (see Section 5.1).
 MEYER 11

4.1. Substitutions: how dynamicists came to model sets. Let A be a finite alphabet
and let A+ be the set of non-empty finite words on A. A (word) substitution is a map
σ : A → A+ which extends to σ : A+ → A+ and σ : AN → AN via concatenation. A
substitution defines a language, which is the set of all finite words w0 . . . wk which appear as
a subword of σ n (a), for a ∈ A and n ∈ N. The shift space Xσ is the set of infinite words all
of whose finite subwords belong to the language of σ, and with the shift map S : Xσ → Xσ ,
the pair (Xσ , S) is called a (substitution) shift.
Example 4.1. As a prominent example of a word substitution in aperiodic order, consider the
Fibonacci substitution σ defined on the alphabet A = {A, B} as σ(A) = AB and σ(B) = A.
The sequence of finite words (σ n (A))n are nested, i.e., σ n (A) is a prefix of σ n+1 (A) for each
n, and it converges to the so-called Fibonacci word, whose first terms are
ABAABABAABAABABAABABAABAABABAABAABABAABABAABAABABA . . .
The Fibonacci word is a one-dimensional discrete model of a quasicrystal; indeed, one of
the first models that have been considered. For, it codes a one-dimensional tiling obtained as
a model set via a selection strip in Z2 as illustrated in Figure 2.6, noting that the Fibonacci
word will be the coding of the right part of the tiling seen there, starting at the origin;
see below for how to generate the actual tiling. The Fibonacci word also corresponds to a
discretization of a line in discrete geometry [RK01], and from a symbolic dynamics viewpoint,
it belongs to the family of Sturmian words which are discussed in Section 5.2. For more on
the Fibonacci word, see e.g. [Lot02, Pyt02].
The incidence matrix Mσ of a substitution σ is the square matrix whose entries count the
number of occurrences of letters  in the images of letters. For example, for the Fibonacci
1 1
substitution σ, we have Mσ = . This matrix, called the abelianization of σ as the
1 0
order of letters in substitution words is forgotten, allows us to apply tools from linear algebra
to analyse the self-similarity properties of the Fibonacci word, as well as its combinatorial
and dynamical properties. The Perron–Frobenius theorem tells us that if Mσ is primitive,
then it admits a strictly dominant eigenvalue that is positive, and this eigenvalue is called
the inflation factor, or expansion factor of the substitution. Also, under the assumption of
primitivity of Mσ , the shift (Xσ , S) has particularly useful ergodic properties, such as unique
ergodicity or minimality. For more details, see [Que10].
Analogously, substitutions can also be defined on tiles, and iterating the substitution pro-
duces self-affine tilings. Tilings are considered in any dimension and over general combinato-
rial structures with group actions. Two main natural models of tilings are Wang tiles coming
from computer science (see Section 6) and shifts coming from dynamical systems. In the tiling
setting, we now work in a d-dimensional geometric space, here Rd , rather than in a symbolic
setting, with letters and words.
Definition 4.2 (Self-affine tiling). Let {Ti ; 1 ≤ i ≤ m} be a set of tiles (usually assumed to be
polygons) in Rd , called prototiles. A self-affine tiling T = {xi + Ti ; xi ∈ Λi , i ≤ m} in Rd is a
tiling fixed by a linear expanding map ϕ : Rd → Rd that maps every prototile onto a union
of prototiles (expanding means that all its eigenvalues are greater than one in modulus). The
prototiles are translated by vectors in the sets Λi in order to form the tiling T . With the map
ϕ comes a substitution rule Σ that explains how to divide a tile inflated by ϕ into a union of
translated prototiles and an incidence matrix MΣ which indicates for each prototile M how
12 V. BERTHÉ AND R. YASSAWI

ϕ(M ) is a union of prototiles. More precisely, there exist finite sets Dij ⊂ Rd (1 ≤ i, j ≤ k)
such that for any j (1 ≤ j ≤ k)
Σ(Tj ) = {u + Ti , u ∈ Dij , i, j ≤ k} with ϕTj = ∪ki=1 (Ti + Dij ).
Here all sets on the right-hand side must have disjoint interiors; it is also possible for some
of the Dij to be empty. The matrix MΣ then counts the cardinalities of the sets Dij .
If MΣ admits a strictly dominant eigenvalue that is positive, then as with substitutions,
this eigenvalue is called the inflation factor of the tiling. It equals the absolute value of the
determinant of ϕ, which controls the volume expansion. The tiling is said self-similar if ϕ is
a similarity.
Example 4.3. We revisit Example 2.6 and 4.1. Here d = 1. We define a tiling with the
Fibonacci substitution σ : A 7→ AB, B 7→ A by starting with two intervals (prototiles),
labelled respectively A and B, with lengths ℓA and ℓB . To obtain a geometric realisation Σ
of the Fibonacci substitution σ using two one-dimensional interval tiles, we thus need to find
tiles A and B, with lengths ℓA and ℓB , to be determined, where
• Σ maps the tile A to the two concatenated tiles AB,
• Σ maps the tile B to the tile A, and
• Σ corresponds to an expansive map on R.
The substitution Σ maps A, whose length is ℓA , to two tiles of combined length ℓA + ℓB , and
the tile B of length ℓB to a tile A of length ℓA . Since we want the substitution rule Σ to act
as a map that expands the space with an inflation factor λ, this implies that
λℓA = ℓA + ℓB and λℓB = ℓA ,
i.e.,
 
1 1
λ(ℓA , ℓB ) = (ℓA , ℓB ) = (ℓA , ℓB )MΣ ,
1 0
so that λ must be a left eigenvector for MΣ , and since φ is the only expanding eigenvalue of
MΣ , λ = φ, and we can take ℓA = φ and ℓB = 1. We thus set A = [0, φ] and B = [0, 1]. This
defines a self-affine tiling (and even a self-similar one). The matrix MΣ equals the matrix Mσ ,
and one has D11 = {0}, D12 = {φ}, D21 = {0}, and D22 = ∅. Iterating this tiling substitution
rule gives us a geometric tiling of the line; see Figure 2.6.
Self-affine tilings occur in a wide range of problems in dynamics; for instance, they are
related to Markov partitions for hyperbolic maps and radix representations [BBLT06]. More
generally, for tilings, not necessarily generated by substitutions, the associated shift dynamics
is given by Rd -translations yielding as dynamical systems orbit closures of tilings [Rob04,
Sol97]. The closure is taken in the local topology where coincidences on a large ball around
the origin up to a small translation are considered. A remarkable topological property here is
that the local structure of the tiling space (i.e., the associated dynamical system endowed with
the Rd -action) is described as the product of a Cantor set and a Euclidean space. Locally, it
looks like a disk crossed with a totally disconnected set. Tiling spaces then yield particularly
interesting types of topological spaces obtained as inverse limits of branched manifolds with
totally disconnected compact fibers [Sad08]. This makes cohomological methods particularly
adapted in this setting, and particularly the C̆ech cohomology which gives information on
how a tiling can be deformed. This goes beyond the case of substitutive tilings. Consider
as an illustration the works of A. Forrest, J. Hunton, and J. Kellendonk in [FHK02] for
 MEYER 13

cut and project tilings which, along the way, illustrate the efficiency of topological methods
combined with tools from C ∗ -algebras (see also [Put18] for an elementary exposition). More
generally, topological dynamics and topological notions (e.g., maximal equicontinuous factor
or proximality relation) have led to beautiful developments, for instance in the computation
of the Ellis semigroup [ABKL15, KY20].

4.2. More on self-affine tilings. In this section, we revisit Meyer’s Theorem 2.2 in the
setting of self-similar tilings. According to Definition 4.2, we consider an expanding linear
map ϕ : Rd → Rd and a self-affine tiling T fixed by ϕ. Dynamicists ask which numbers can
appear as inflation factors, or which maps can appear as expanding maps, of self-affine tilings,
and which of these are inflation factors for substitutions that lead to quasicrystals. This has
led to a vast literature, and as in Theorem 2.2, it is articulated in two steps: firstly, what are
the conditions that the inflation factors should satisfy? Secondly, can it be possible to realise
a self-similar structure, and even a Meyer or a model set, with each dilation factor satisfying
these conditions? We turn to the last question in Section 5.3.
We first introduce two arithmetic definitions.

Definition 4.4 (Perron number). The algebraic integer β is said to be a Perron number if
|β| > |α| for any other of its algebraic conjugates.

For example, the Perron–Frobenius theorem tells us that the dominant eigenvalue of a
primitive non-negative matrix is a Perron number.

Definition 4.5 (Pisot family). Let Λ be a finite set of algebraic integers of modulus larger
than or equal to 1. The set Λ is said to form a Pisot family if for every λ ∈ Λ and for every
algebraic conjugate λ′ of λ with |λ′ | ≥ 1, then λ′ ∈ Λ.

In the one-dimensional case, it follows from the work of D. Lind [Lin84] that λ is an
inflation factor of a self-similar tiling if and only if λ is a Perron number. In the two-
dimensional self-similar case (the dilation is the same in every direction), the expansion map
ϕ, when viewed as acting on C, equals λz where λ is a complex Perron number, as shown by
W. Thurston [Thu89]. In the case where the expansion map ϕ is diagonalizable, R. Kenyon
and B. Solomyak [KS10] [Ken90] have shown that the eigenvalues of ϕ are algebraic integers.
The general result (without the diagonalizabiliy condition) has then been established by J.
Kwapisz [Kwa16] who proves that, with a suitable notion of Perron, the map ϕ of a self-affine
tiling is integral algebraic and Perron.
Now that it is established what Perron numbers have to do with being self-affine (or even
being self-similar), Pisot numbers enter into play when further properties are required such
as some of those shared by Meyer sets, extending Meyer’s Theorem 2.2. In this setting, the
connection with the notion of the Pisot family was successfully developed by J.-Y. Lee and B.
Solomyak [LS12] for spectral considerations in the case where the map ϕ is diagonalizable over
the complex numbers, with all eigenvalues being algebraic conjugates of the same multiplicity.
They proved that the associated tiling shift has relatively dense discrete spectrum (i.e., the
set of eigenvalues has full rank) if and only if the spectrum of the linear map ϕ forms a Pisot
family, which is also equivalent to the discrete set of control points for the tiling to be a Meyer
set.
More generally, we will see in Section 5.1 how Pisot numbers enter the picture in this substi-
tutive setting not only as inflation factors, but also for model sets via spectral considerations.
14 V. BERTHÉ AND R. YASSAWI

5. Pisot substitutions and aperiodic order

Now that inflation factors are well characterised, the question is to be able to construct self-
similar tilings or point sets having this inflation factor that are also Meyer or even model sets.
There exist several strategies for such realisations. Among them, Rauzy fractals, discussed
in Section 5.3, provide suitable windows for explicit cut and project schemes. But before
discussing them, we introduce the so-called Pisot substitution conjecture.
5.1. Spectral properties and the Pisot conjecture. The spectral study of substitu-
tive dynamical systems is a fundamental question (see for instance the monographs [Que10,
Pyt02]). This is particularly relevant in the context of aperiodic order, as stressed by the
contribution by R. Moody in this volume [Moo]. Weak mixing, namely the absence of eigen-
values, i.e., the absence of Bragg peaks, indicates a certain level of disorder and conversely,
the existence of spectral eigenvalues provides dynamical factors which consist of group trans-
lations. For instance, constant length substitutions yield p-adic factors. In particular, discrete
spectrum deals with the possibility of providing substitutive dynamical systems with a repre-
sentation as a group translation, that is, a dynamical system acting on a space of a geometric
nature (such as a torus for instance), that is (measurably) isomorphic to the substitutive
system under study.
We have seen through the work of Yves Meyer that the connection with Pisot numbers
arose from the inception of quasiorder, and this has led to the following natural definition.
Definition 5.1 (Pisot substitution). If σ is a word substitution whose incidence matrix has
a characteristic polynomial which is the irreducible minimal polynomial of a Pisot number,
then σ is called a Pisot substitution.
From the eighties onwards, Pisot substitutions have attracted much attention in the context
of mathematical quasicrystals. We recall that the Fibonacci substitution provided one of
the first examples of a one-dimensional quasicrystal and more generally substitutions were
considered as promising examples of one-dimensional quasicrystals; as we have seen, they
are very simple algorithmic rules that create configurations that display long-range order.
However not all substitutions yield quasicrystals. E. Bombieri and J. E. Taylor asked already
in [BT86, BT87] which substitutions produce quasicrystals and they highlighted the Pisot
algebraic restriction for substitutions to yield pure discrete spectrum. See also [Sol97] where
a self-similar tiling is proved to admit a discrete component in its spectrum if and only if its
inflation factor is a Pisot number.
This culminated with the Pisot substitution conjecture, stated below for substitutions de-
fined on symbols. There exist various formulations, see the survey [ABBS08].
Conjecture 5.2 (Pisot substitution conjecture). Let σ be a Pisot irreducible substitution
(i.e., the characteristic polynomial of its incidence matrix is the minimal polynomial of a
Pisot number). Then, the shift (Xσ , S) has pure discrete spectrum, i.e., it is isomorphic, in
a measure-theoretic sense, to a translation on a compact abelian group.
In a nutshell, the Pisot arithmetic condition induces order, where order is expressed here in
spectral and dynamical terms as being isomorphic to the simplest dynamical systems, namely
group translations. The conjecture is known to be true for the case of substitutions on two
letters [BD02, HS03]. The difficulty for a larger alphabet comes from the arithmetic of higher
degree algebraic numbers, which is more difficult to manage than for quadratic numbers.
Also, in two dimensions, it is easier to identify the occurrences of so-called coincidences, the
 MEYER 15

existence of which allows us to project injectively an infinite word onto a group, here the
circle; to do so, words are embedded as discrete lines in the plane and coincidences between
these lines are to be found, which is easier in a plane (see Example 5.6 for an illustration);
generic behaviour is that once there is a coincidence, there are infinitely many occurrences of
the coincidence, and this implies that this projection is an injection.
Theorem 5.3. [BD02, HS03] Two-letter Pisot substitutions have pure discrete spectrum.
The still open Pisot substitution conjecture, even if solved in the closely related context of
beta-numeration by M. Barge [Bar18], shows that important parts of the picture are still to
be developed. Once again, one particularly appealing feature concerning the works developed
around the Pisot substitution conjecture is that they involve several mathematical approaches
and reformulations such as the homological Pisot conjecture, which aims to take into account
topological invariance, or the coincidence rank conjecture cohomology, involving topological
dynamics, arithmetic, combinatorics, fractal geometry, etc. See as an illustration [ABBS08]
or [Thu19].
It follows from algebraic considerations, involving the eigenvalues of the incidence matrix,
that Pisot substitution shifts and Pisot tiling dynamical systems must have a non-trivial
rotation factor. Criteria for obtaining these dynamical eigenvalues are now well understood;
see for instance [Sol07]. The difficult part consists in producing a measurable isomorphism.
Fortunately, there is a wide range of algorithmic conditions, called coincidence conditions
(such as briefly evoked above), for checking pure discrete spectrum, which date back to the
work of M. Dekking on constant-length substitutions, as described in [Que10]. They occur
for infinite words, for higher-dimensional and for non-lattice based self-affine tilings, and even
in the non-unimodular case. See for instance [Sol97, LM01, ST09, AL11]. Also relevant
is the notion of almost-automorphy, introduced by Veech [Vee65], which implies discrete
spectrum for substitutions, and which incidentally was shown to imply spectral synthesis
[Vee69], bringing us back again to Meyer’s work.

5.2. Sturmian words and beyond substitutions. As noted, the Pisot substitution con-
jecture has been proved for two–letter irreducible Pisot substitutions (Theorem 5.3). Such
substitutions produce model sets. Amongst them are a remarkable and widely studied family,
that of the Sturmian substitutions, which generate shifts that belong to the class of Sturmian
shifts that we now describe.
Sturmian shifts are symbolic representations of irrational circle rotations. More precisely,
consider the rotation Rα acting on T = R/Z, with Rα (x) = x + α mod 1, for α irrational.
If we code the orbit of a point under the action of Rα using a two-interval partition of
semi-open intervals whose lengths are respectively α and 1 − α, then we obtain a Sturmian
word, generating a Sturmian shift Xα consisting of all Sturmian words of angle α, that is, all
Sturmian words coding Rα . This seminal class of symbolic dynamical systems, introduced by
M. Morse and G. Hedlund in [MH40], laid some of the foundations for symbolic dynamics.
For a thorough description of Sturmian words, see [BS02] and [Pyt02, Chapter 6].
The Fibonacci substitution from Section 4.1 is the most cited example of a Sturmian
substitution. When it is used to generate a tiling, then as we saw in Figure 2.6 the associated
point set is a model set. Sturmian words have provided the simplest examples of cut and
project sets: the window is here an interval of unit length and the associated tilings are one-
dimensional, with the cut and project scheme relying on two lines as described in Section
2. Phrased in a more geometric way, they code discrete lines in digital geometry [RK01].
16 V. BERTHÉ AND R. YASSAWI

There is an impressive literature devoted to the study of Sturmian words and to possible
generalisations; let us mention for instance episturmian words, also called Arnoux-Rauzy
words, which have attracted a lot of attention [CFM08, AHS16].
One key feature is the scale invariance of Sturmian shifts. Not all Sturmian words are
substitutive. Indeed the angle α of a substitutive Sturmian shift Xα is a quadratic number
[CMPS93] (see also [BEIR07] for a characterisation of Sturmian words that are fixed points of
substitutions obtained in terms of Rauzy fractals discussed in Section 5.3). However Sturmian
words can be all generated in terms of sequences of substitutions which we elaborate below.
More precisely, Sturmian words are perfectly understood and described via a representation
based on substitutions and Euclid’s algorithm: the continued fraction expansion of the angle
α provides an infinite product of square matrices of size two with non-negative integer entries,
each of these matrices can be seen as the incidence matrix of a substitution, and the action of
a substitution can be seen as a combinatorial interpretation of a step of Euclid’s algorithm,
as described below.

Theorem 5.4. [AR91] We consider the substitutions over the alphabet {0, 1} defined by
σ0 : 0 7→ 01, 1 7→ 1, and σ1 : 0 7→ 0, 1 7→ 10. If α has continued fraction expansion α =
[0; a1 + 1, a2 , · · · ], then the Sturmian shift Xα is generated by the infinite word
a
lim σ0a1 σ1a2 σ0a3 σ1a4 · · · σ0 2n−1 σ1a2n (0).
n→∞

Seen through this lens, Sturmian words are described using a renormalization scheme gov-
erned by continued fractions via the geodesic flow acting on the modular surface. This crys-
tallises with the study of interval exchanges in relation with the Teichmüller flow through
the work, among others, of W. Veech, H. Masur, J.-C. Yoccoz, and A. Avila (see e.g.
[Yoc06, Buf14]).
Theorem 5.4 states that a Sturmian word is the limit of an infinite composition of these
substitutions; such an approach has been formalised using the language of S-adic words, which
are infinite words generated when a sequence S of substitutions is applied, as opposed to when
a single substitution is iterated [BD14]. The S-adic setting pertains to non-stationary dynam-
ics (i.e., time inhomogeneous dynamics), which consists in working with iterated sequences
of transformations drawn according to a further dynamical system, similarly as for random
dynamics, random Markov chains and random products of matrices. This formalism allows
us to extend the Pisot conjecture beyond algebraicity [BST19]. The Pisot condition is then
replaced by the requirement that the second Lyapunov exponent of the infinite associated
products of matrices is negative. This extended Pisot conjecture has been proved to hold for
large relevant families of systems based on continued fractions expansions, such as the Brun
algorithm. As a striking outcome, this yields symbolic codings for almost every translation of
T2 [BST19], paving the way for the development of equidistribution results for the associated
two-dimensional Kronecker sequences.

5.3. Rauzy fractals. In this section we describe a construction for word substitutions having
a Pisot inflation factor that generates suitable and effective windows for model sets; this
generalises the Sturmian case for which the window in the internal space is an interval. In fact,
it is now well understood that a substitution tiling with pure discrete spectrum is a model set,
such as described by Lee in [Lee07]. She shows the equivalence between specific model sets, the
 MEYER 17

so-called inter-model sets3, and pure point dynamical spectrum in the context of primitive
substitution point sets, where the construction of the cut and project scheme involves an
abstract internal space. The aim of this section is to present more explicit constructions of
these internal spaces as Rauzy fractals. This approach appeals in its combination of arithmetic
and dynamics.
Recall that for (Xσ , S) to have pure discrete spectrum, it must be isomorphic to a group
translation. Thus, given a Pisot substitution σ, one wants to provide a geometric repre-
sentation for (Xσ , S) as a translation on the torus, or more generally on a locally compact
abelian group. And as a candidate for a fundamental domain for the expected translation,
one associates with the substitutive dynamical system (Xσ , S) a Rauzy fractal.
Rauzy fractals were first introduced in [Rau82] in the case of the so-called Tribonacci
substitution in order to prove the following statement.
Theorem 5.5. [Rau82] Let σ be the Tribonacci substitution defined over the alphabet A =
{A, B, C} as as σ(A) = AB, σ(B) = AC and σ(C) = A. The symbolic dynamical system
(Xσ , S) is measure-theoretically isomorphic to the translation Rβ on the two-dimensional torus
T2 defined as Rβ : T2 → T2 , x 7→ x + (1/β, 1/β 2 ), where β, the Perron eigenvalue for σ,
satisfies β 3 = β 2 + β + 1.
We now sketch how to build a fractal from the shift (Xσ , S) as a fundamental domain for
the translation of Theorem 5.5.
Example 5.6. Let α be one of the complex roots of X 3 − X 2 − X − 1; we have |α| < 1. We
will associate to a point x ∈ Xσ an α-expansion as follows. The point x ∈ Xσ can indeed
be uniquely desubstituted, either as x = σ(y), or as x = Sσ(y) for some y ∈ Xσ . We can
associate to x an initial digit ϵ0 ∈ {0, 1}, depending on whether we shift or not. We repeat
this procedure, desubstituting y to obtain a second digit ϵ1 . Iterating we obtain a sequence
(ϵn ) ∈ {0, 1}N , and it can be Pverified that this sequence satisfies ϵn ϵn+1 ϵn+2 = 0 for each
n. We then project x to z := n ϵn αn ; note the connection to beta-numeration (evoked in
Section 3) that appears; this connection is studied in [Thu89]. Geometrically, this amounts
to the projection πint along the expanding eigendirection of the matrix Mσ with respect to
the decomposition of R3 into the expanding eigendirection and the contracting eigenplane of
the matrix Mσ . The image πint (Xσ ) of Xσ is the Rauzy fractal R; see Figure 5.3. Depending
on what we see at the 0-index x0 of x, we can further specify in which of the three regions
R0 = αR, R1 = α3 + α2 R and R2 = α3 + α4 + α3 R the point πint (x) lives. The projection
is injective outside of the boundaries between these three regions, and shifting inside Xσ
corresponds to an exchange from one region to another; see [Mes98] for details. The domain
R is the internal space of a cut and project scheme which generates the Tribonacci tiling.
Rauzy fractals can more generally be associated with Pisot substitutions (see [AI01, BK06,
Sie04] and the surveys [BS05, Pyt02]), as well as with Pisot beta-transformations and beta-
shifts under the name of central tiles [Aki02]. Rauzy fractals can be defined in a unimodular
case [BK06] (when the inflation factor β of a substitution is a Pisot unit) as well as in a
p-adic setting [Sie04, Sin06] (the primes p that occur are prime divisors of the norm of β and
the Rauzy fractal lives in a finite product of Euclidean and p-adic spaces). Moreover, the
geometric properties of Rauzy fractals reflect the self-similarity of the associated substitutive
3Inter-model sets are model sets satisfying a topological condition that is less restrictive than the condition
for the boundary of having zero measure satisfied by a regular model set.
18 V. BERTHÉ AND R. YASSAWI

Figure 2. The Rauzy fractal

dynamical system: they are solutions of (graph-directed) iterated function systems. Note also
that Rauzy fractals have also been studied as quasicrystals [VM01].
Rauzy fractals do not only produce geometric representations of substitutive dynamical
systems, but they also have very interesting Diophantine applications. We refer to [IFHY03,
AFSS10] for representative examples. Rauzy fractals are particularly convenient for providing
arithmetic descriptions of periodic orbits, yielding relevant generalisations of Lagrange’s and
Galois’ theorems for continued fractions. We recall that Lagrange’s theorem states that the
continued fraction expansion of x is eventually periodic if and only if either x ∈ Q, or x is
a quadratic number. Furthermore, if x > 1 is a quadratic number, and x′ is its algebraic
conjugate, then the continued fraction expansion of x is purely periodic if and only if x
is irrational and −1 < x′ < 0: this is Galois’ theorem. Galois’ theorem can be proved
dynamically by using the Gauss map T acting on [0, 1] as x 7→ {1/x}. A key idea used here
to provide a dynamical proof of Galois’ theorem (and thus to describe purely periodic orbits)
is to transform the map T that is not one-to-one into an injective map thanks to a suitable
realisation of its natural extension.
The same type of methods can also be applied in the beta-numeration case allowing a
characterisation of periodic orbits for a Pisot number β. The condition on algebraic conjugates
is then expressed in terms of Rauzy fractals, such as stated below, noting that Galois’ theorem
expresses the pure periodicity of the continued fraction expansion of a quadratic real number
x in terms of the location of x and its algebraic conjugate on the line. Here the location is
expressed in terms of belonging to a Rauzy fractal.
Theorem 5.7. If β is a Pisot number, then x ∈ [0, 1] has an eventually periodic β-expansion
if and only if x ∈ Q(β) ∩ [0, 1]. Moreover, a real number x ∈ Q(β) ∩ [0, 1) has a purely periodic
β-expansion if and only if the vector composed of x and its conjugates belong to the Rauzy
fractal associated with β.
The analogue of Lagrange’s theorem has been proved in [Ber77, Sch80] and the analogue of
Galois’ theorem in [ABBS08] (see also the references therein). The proof is here again based
on an explicit realisation of the natural extension of the beta-transformation Tβ .

6. How computability came to model sets

We have seen with Meyer’s Theorem 2.2 that algebraicity enters naturally into play for
describing self-similarity for aperiodic order. We will see in this section that computability
 MEYER 19

is an increasingly pertinent viewpoint to consider. The tiling setting is the one which is best
suited for this discussion with the notion of a local rule, which describes the constraints on
how adjacent tiles can be assembled. More generally, local rules naturally model the ener-
getic finite-range interactions between atoms. Aperiodic order emerges when local properties
enforce global order together with aperiodicity, and for tilings, this results in the research of
local rules.
The notion of local rule is natural when considering Wang tiles. A Wang tile is a unit square
with edges marked with symbols (or colors). A tiling by Wang tiles consists in assembling
copies of the tiles (by translations only) so that symbols on shared edges match. These tiles
were introduced by Wang in 1961 for the study of fragments of first order logics in the context
of the domino problem, which asks for the existence of an algorithm deciding whether a finite
set of Wang tiles can tile the plane. The domino problem is known to be undecidable, as
proved by R. Berger in 1966. In dynamical terms, this amounts to ask whether a given
two-dimensional shift of finite type is non-empty. This result is based on two ingredients:
firstly, the simulation of Turing computations by tilings of the plane (usually via local rules
or substitutions), and secondly, the existence of aperiodic sets of tiles (in the sense that
they can only produce aperiodic tilings). Since the first examples of aperiodic shifts of finite
type were based on hierarchical structures [Ber66, Rob71], substitutive structures have been
known to be able to force aperiodicity. But more than that, in dimension d ≥ 2, under
natural assumptions, it is known that most substitution tilings admit local rules (possibly
after decoration) [Moz89, GS98], as expressed in the following statement that covers all known
examples of hierarchical aperiodic tilings.

Theorem 6.1. [GS98] Every substitution tiling of the d-dimensional Euclidean plane, d > 1,
can be enforced with finite matching rules, subject to a mild condition.

After the connection between higher-dimensional substitutions and local rules was estab-
lished, the following question became natural: can we describe a tiling, which is not neces-
sarily substitutive, in terms of local rules? This question emerged in mathematics as early as
the 1990s in the study of quasicrystals [Kat95, Lev88, Le95, Soc90] (see more references in
[BF15, FS19]). This has known particularly rich developments for tilings obtained by the cut
and project schemes. One important feature here is that it comes down both to computability
and algebraic considerations on the associated parameters.
Consider for instance a tiling obtained as an approximation of a d-dimensional space E
in Rn , via the cut and project method. When d = 2, this yields a rhombus tiling of the
Euclidean plane, approximating a real plane embedded in a higher-dimensional space. The
question is then how to force tiles to approximate the desired plane E by specifying local
rules. The case of lines in the planes (d = 1, n = 2) corresponds to Sturmian words, which
cannot be defined with local rules, in direct contrast with the higher-dimensional case. The
study of the connections between the existence of local rules for such planar tilings and the
parameters of its slope started with [dB81, Lev88, Le95, Le97, Soc90]. The first conditions
on the parameters of cut and projection schemes were of an algebraic nature. In particular, it
was proved in [Le97] that a slope enforced by undecorated local rules is necessarily algebraic
(this is however not sufficient, see e.g. [BF15, BF20]). However, computability comes into
play when the tiles can be decorated (i.e., in dynamical terms, when we go from shifts of
finite type to sofic ones). Decorations indeed allow the transfer of information through the
tiling, and this was used in [FS19] to prove the following statement.
20 V. BERTHÉ AND R. YASSAWI

Theorem 6.2. [FS19, Corollay 2] A slope can be enforced by colored weak local rules if and
only if its slope is computable.
This computational approach is reminiscent of the proof of the undecidability of the domino
problem. Colored local rules are used to encode simulations of Turing computations which
check that only planar tilings that approximate the desired planes can be formed. The
fundamental tool is the use of so-called simulation theorems, which state that any effective
one-dimensional shift can be obtained as the subaction of a two-dimensional sofic shift [Hoc09,
AS13, DRS10]). The Penrose tiling is an example where all these viewpoints gather: it can
be considered simultaneously as generated by a substitution, it is a model set, thus obtained
via a cut and project scheme, and it can be endowed with local rules.
Nourished by this computability viewpoint, aperiodic order has known particularly fruitful
developments with the study of tilings, spreading also in the direction of higher-dimensional
word combinatorics, symbolic dynamics or else group combinatorics, with symbolic dynamics
on groups and the domino problem for them. In this latter setting, the decidability of the
domino problem is reinterpreted as a group property, inspired by Higman’s embedding theo-
rem which states that any finitely generated and recursively presented group can be embedded
as a subgroup of a finitely presented group, such as developed in [Jea16, JV19].

In conclusion, Meyer’s Theorem 2.2 and its mathematical successors show that Pisot
numbers, and more generally arithmetic and Diophantine approximation, are natural ac-
tors of aperiodic order when self-similarity occurs. As an illustration, observe that the most
classical combinatorial and dynamical measures of disorder developed for cut and project
structures can be evaluated in terms of Diophantine properties such as developed e.g. in
[HKWS16, HJKW19] with in particular nice connections with the Littlewood conjecture
[HKW18]. However, the last developments of the study of quasiorder show that computabil-
ity is also a viewpoint that has to be considered, with the possibility of encoding some
computation in tilings. We have seen that this has emerged as early as in the 1960s with
the introduction of Wang tiles. This concept then crystallised in a striking way 50 years
later with the work of M. Hochman [Hoc09] and with the characterisation, obtained by M.
Hochman and T. Meyerovitch, of the entropies of multidimensional finite type shifts. If these
entropies are logarithms of Perron numbers in the one-dimensional case, M. Hochman and
T. Meyerovitch made a fundamental change of perspective by proving that in the higher-
dimensional case, these entropies are characterised as the recursive numbers that are right
recursively enumerable [HM10].

References
[ABBS08] S. Akiyama, G. Barat, V. Berthé, and A. Siegel. Boundary of central tiles associated with Pisot
beta-numeration and purely periodic expansions. Monatsh. Math., 155(3-4):377–419, 2008.
[ABF05] S. Akiyama, F. Bassino, and C. Frougny. Arithmetic Meyer sets and finite automata. Inform. and
Comput., 201(2):199–215, 2005.
[ABKL15] J.-B. Aujogue, M. Barge, J. Kellendonk, and D. Lenz. Equicontinuous factors, proximality and
Ellis semigroup for Delone sets. In Mathematics of aperiodic order, volume 309 of Progr. Math.,
pages 137–194. Birkhäuser/Springer, Basel, 2015.
[AFSS10] B. Adamczewski, Ch. Frougny, A. Siegel, and W. Steiner. Rational numbers with purely periodic
beta-expansion. Bull. London Math. Soc., 42:538–552, 2010.
[AHS16] A. Avila, P. Hubert, and A. Skripchenko. On the Hausdorff dimension of the Rauzy gasket. Bull.
Soc. Math. France, 144:539–568, 2016.
 MEYER 21

[AI01] P. Arnoux and S. Ito. Pisot substitutions and Rauzy fractals. Bull. Belg. Math. Soc. Simon Stevin,
8(2):181–207, 2001.
[AK13] S. Akiyama and V. Komornik. Discrete spectra and Pisot numbers. J. Number Theory, 133(2):375–
390, 2013.
[Aki02] S. Akiyama. On the boundary of self affine tilings generated by Pisot numbers. J. Math. Soc.
Japan, 54(2):283–308, 2002.
[AL11] S. Akiyama and J.-Y. Lee. Algorithm for determining pure pointedness of self-affine tilings. Adv.
Math., 226(4):2855–2883, 2011.
[AR91] P. Arnoux and G. Rauzy. Représentation géométrique de suites de complexité 2n + 1. Bull. Soc.
Math. France, 119(2):199–215, 1991.
[AS03] J.-P. Allouche and J. Shallit. Automatic sequences. Cambridge University Press, Cambridge, 2003.
Theory, applications, generalizations.
[AS13] N. Aubrun and M. Sablik. Simulation of effective subshifts by two-dimensional subshifts of finite
type. Acta Appl. Math., 126:35–63, 2013.
[Bar18] M. Barge. The Pisot conjecture for β-substitutions. Ergodic Theory Dynam. Systems, 38:444–472,
2018.
[BBLT06] G. Barat, V. Berthé, P. Liardet, and J. Thuswaldner. Dynamical directions in numeration. Ann.
Inst. Fourier (Grenoble), 56:1987–2092, 2006. Numération, pavages, substitutions.
[BD02] M. Barge and B. Diamond. Coincidence for substitutions of Pisot type. Bull. Soc. Math. France,
130:619–626, 2002.
[BD14] V. Berthé and V. Delecroix. Beyond substitutive dynamical systems: S-adic expansions. In Nu-
meration and substitution 2012, RIMS Kôkyûroku Bessatsu, B46, pages 81–123. Res. Inst. Math.
Sci. (RIMS), Kyoto, 2014.
[BEIR07] V. Berthé, H. Ei, S. Ito, and H. Rao. On substitution invariant Sturmian words: an application of
Rauzy fractals. Theor. Inform. Appl., 41(3):329–349, 2007.
[Ber66] R. Berger. The undecidability of the domino problem. Mem. Amer. Math. Soc., 66:72, 1966.
[Ber77] A. Bertrand. Développements en base de Pisot et répartition modulo 1. C. R. Acad. Sci. Paris
Sér. A-B, 285(6):A419–A421, 1977.
[BF15] N. Bédaride and Th. Fernique. When periodicities enforce aperiodicity. Comm. Math. Phys.,
335(3):1099–1120, 2015.
[BF20] N. Bédaride and Th. Fernique. Canonical projection tilings defined by patterns. Geom. Dedicata,
208:157–175, 2020.
[BFGK98] Č. Burdı́k, Ch. Frougny, J. P. Gazeau, and R. Krejcar. Beta-integers as natural counting systems
for quasicrystals. J. Phys. A, 31(30):6449–6472, 1998.
[BG13] M. Baake and U. Grimm. Aperiodic order. Vol. 1, volume 149 of Encyclopedia of Mathematics and
its Applications. Cambridge University Press, Cambridge, 2013.
[BK06] M. Barge and J. Kwapisz. Geometric theory of unimodular Pisot substitutions. Amer. J. Math.,
128:1219–1282, 2006.
[BM00] M. Baake and R. V. Moody, editors. Directions in mathematical quasicrystals, volume 13 of CRM
Monograph Series. American Mathematical Society, Providence, RI, 2000.
[BS02] J. Berstel and P. Séébold. Sturmian words. In M. Lothaire, editor, Algebraic Combinatorics on
Words, volume 90 of Encyclopedia of Mathematics and Its Applications, pages 45–110. Cambridge
University Press, 2002.
[BS05] V. Berthé and A. Siegel. Tilings associated with beta-numeration and substitutions. Integers,
5(3):A2, 46 pp. (electronic), 2005.
[BST19] V. Berthé, W. Steiner, and J. M. Thuswaldner. Geometry, dynamics, and arithmetic of S-adic
shifts. Annales de l’Institut Fourier, 69:1347–1409, 2019.
[BT86] E. Bombieri and J. E. Taylor. Which distributions of matter diffract? An initial investigation.
volume 47, pages C3–19–C3–28. 1986. International workshop on aperiodic crystals (Les Houches,
1986).
[BT87] E. Bombieri and J. E. Taylor. Quasicrystals, tilings, and algebraic number theory: some preliminary
connections. In The legacy of Sonya Kovalevskaya (Cambridge, Mass., and Amherst, Mass., 1985),
volume 64 of Contemp. Math., pages 241–264. Amer. Math. Soc., Providence, RI, 1987.
[Buf14] A. I. Bufetov. Limit theorems for translation flows. Ann. of Math. (2), 179:431–499, 2014.
22 V. BERTHÉ AND R. YASSAWI

[Bug96] Y. Bugeaud. On a property of Pisot numbers and related questions. Acta Math. Hungar., 73(1-
2):33–39, 1996.
[CFM08] J. Cassaigne, S. Ferenczi, and A. Messaoudi. Weak mixing and eigenvalues for Arnoux-Rauzy
sequences. Ann. Inst. Fourier (Grenoble), 58(6):1983–2005, 2008.
[CMPS93] D. Crisp, W. Moran, A. Pollington, and P. Shiue. Substitution invariant cutting sequences. J.
Théor. Nombres Bordeaux, 5(1):123–137, 1993.
[dB81] N. G. de Bruijn. Algebraic theory of Penrose’s nonperiodic tilings of the plane. I, II. Nederl. Akad.
Wetensch. Indag. Math., 43(1):39–52, 53–66, 1981.
[DRS10] B. Durand, A. Romashchenko, and A. Shen. Effective closed subshifts in 1D can be implemented
in 2D. In Fields of logic and computation, volume 6300 of Lecture Notes in Comput. Sci., pages
208–226. Springer, Berlin, 2010.
P. Erdös, I. Joó, and V. Komornik. Characterization of the unique expansions 1 = ∞ −ni
P
[EJK90] i=1 q and
related problems. Bull. Soc. Math. France, 118(3):377–390, 1990.
[EJK98] P. Erdös, I. Joó, and V. Komornik. On the sequence of numbers of the form ϵ0 +ϵ1 q+· · ·+ϵn q n , ϵi ∈
{0, 1}. Acta Arith., 83(3):201–210, 1998.
[EK98] P. Erdős and V. Komornik. Developments in non-integer bases. Acta Math. Hungar., 79(1-2):57–83,
1998.
[FHK02] A. Forrest, J. Hunton, and J. Kellendonk. Topological invariants for projection method patterns.
Mem. Amer. Math. Soc., 159(758), 2002.
[FS92] C. Frougny and B. Solomyak. Finite beta-expansions. Ergodic Theory Dynamical Systems, 12:45–
82, 1992.
[FS19] Th. Fernique and M. Sablik. Weak colored local rules for planar tilings. Ergodic Theory Dynam.
Systems, 39(12):3322–3346, 2019.
[FW02] D-J. Feng and Z.-Y. Wen. A property of Pisot numbers. J. Number Theory, 97(2):305–316, 2002.
[Gar62] A. M. Garsia. Arithmetic properties of Bernoulli convolutions. Trans. Amer. Math. Soc., 102:409–
432, 1962.
[GMP03] L.-S. Guimond, Z. Masáková, and E. Pelantová. Combinatorial properties of infinite words associ-
ated with cut-and-project sequences. J. Théor. Nombres Bordeaux, 15(3):697–725, 2003.
[GMP06] J.-P. Gazeau, Z. Masáková, and E. Pelantová. Nested quasicrystalline discretisations of the line.
In Physics and number theory, volume 10 of IRMA Lect. Math. Theor. Phys., pages 79–131. Eur.
Math. Soc., Zürich, 2006.
[GQ] D. Gratias and M. Quiquandon. Cristallographie, quasicristaux et yves meyer. This volume.
[GS98] C. Goodman-Strauss. Matching rules and substitution tilings. Ann. of Math. (2), 147(1):181–223,
1998.
[Had98] J. Hadamard. Sur la forme des lignes géodésiques à l’infini et sur les géodésiques des surfaces réglées
du second ordre. Bull. Soc. Math. France, 26:195–216, 1898.
[HJKW19] A. Haynes, A. Julien, H. Koivusalo, and J. Walton. Statistics of patterns in typical cut and project
sets. Ergodic Theory Dynam. Systems, 39(12):3365–3387, 2019.
[HKW18] A. Haynes, H. Koivusalo, and J. Walton. Perfectly ordered quasicrystals and the Littlewood con-
jecture. Trans. Amer. Math. Soc., 370:4975–4992, 2018.
[HKWS16] A. Haynes, H. Koivusalo, J. Walton, and L. Sadun. Gaps problems and frequencies of patches in
cut and project sets. Math. Proc. Cambridge Philos. Soc., 161:65–85, 2016.
[HM10] M. Hochman and T. Meyerovitch. A characterization of the entropies of multidimensional shifts of
finite type. Ann. of Math. (2), 171:2011–2038, 2010.
[Hoc09] M. Hochman. On the dynamics and recursive properties of multidimensional symbolic systems.
Invent. Math., 176:131–167, 2009.
[HP19] H. Holden and R. Piene, editors. The Abel Prize 2013–2017. Springer, Cham, 2019.
[HS03] M. Hollander and B. Solomyak. Two-symbol Pisot substitutions have pure discrete spectrum.
Ergodic Theory Dynam. Systems, 23:533–540, 2003.
[IFHY03] S. Ito, J. Fujii, H. Higashino, and S.-i. Yasutomi. On simultaneous approximation to (α, α2 ) with
α3 + kα − 1 = 0. J. Number Theory, 99(2):255–283, 2003.
[Jea16] E. Jeandel. Computability in symbolic dynamics. In Pursuit of the universal, volume 9709 of
Lecture Notes in Comput. Sci., pages 124–131. Springer, 2016.
 MEYER 23

[JV19] E. Jeandel and P. Vanier. A characterization of subshifts with computable language. In 36th Inter-
national Symposium on Theoretical Aspects of Computer Science, volume 126 of LIPIcs. Leibniz
Int. Proc. Inform., pages Art. No. 40, 16. Schloss Dagstuhl. Leibniz-Zent. Inform., Wadern, 2019.
[Kat95] A. Katz. Matching rules and quasiperiodicity: the octagonal tilings. In Beyond quasicrystals (Les
Houches, 1994), pages 141–189. Springer, Berlin, 1995.
[Ken90] R. W. Kenyon. Self-similar tilings. ProQuest LLC, Ann Arbor, MI, 1990. Thesis (Ph.D.)–Princeton
University.
[KLS15] J. Kellendonk, D. Lenz, and J. Savinien, editors. Mathematics of aperiodic order, volume 309 of
Progress in Mathematics. Birkhäuser/Springer, Basel, 2015.
[KS10] R. Kenyon and B. Solomyak. On the characterization of expansion maps for self-affine tilings.
Discrete Comput. Geom., 43(3):577–593, 2010.
[Kwa16] J. Kwapisz. Inflations of self-affine tilings are integral algebraic Perron. Invent. Math., 205(1):173–
220, 2016.
[KY20] J. Kellendonk and R. Yassawi. The ellis semigroup of bijective substitutions. Groups, Geometry
and Dynamics, 2020.
[Lag96] J. C. Lagarias. Meyer’s concept of quasicrystal and quasiregular sets. Comm. Math. Phys.,
179(2):365–376, 1996.
[Lag99] J. C. Lagarias. Geometric models for quasicrystals I. Delone sets of finite type. Discrete Comput.
Geom., 21(2):161–191, 1999.
[Le95] T. Q. T. Le. Local rules for pentagonal quasi-crystals. Discrete Comput. Geom., 14(1):31–70, 1995.
[Le97] T. Q. T. Le. Local rules for quasiperiodic tilings. In The mathematics of long-range aperiodic
order (Waterloo, ON, 1995), volume 489 of NATO Adv. Sci. Inst. Ser. C Math. Phys. Sci., pages
331–366. Kluwer Acad. Publ., Dordrecht, 1997.
[Lee07] J.-Y. Lee. Substitution Delone sets with pure point spectrum are inter-model sets. J. Geom. Phys.,
57(11):2263–2285, 2007.
[Lev88] L. S. Levitov. Local rules for quasicrystals. Comm. Math. Phys., 119(4):627–666, 1988.
[Lin84] D. A. Lind. The entropies of topological Markov shifts and a related class of algebraic integers.
Ergodic Theory Dynam. Systems, 4(2):283–300, 1984.
[LM01] J.-Y. Lee and R. V. Moody. Lattice substitution systems and model sets. Discrete Comput. Geom.,
25(2):173–201, 2001.
[Lot02] M. Lothaire. Algebraic combinatorics on words, volume 90 of Encyclopedia of Mathematics and its
Applications. Cambridge University Press, Cambridge, 2002.
[LS12] J.-Y. Lee and B. Solomyak. Pisot family self-affine tilings, discrete spectrum, and the Meyer prop-
erty. Discrete Contin. Dyn. Syst., 32(3):935–959, 2012.
[Mes98] A. Messaoudi. Propriétés arithmétiques et dynamiques du fractal de Rauzy. J. Théor. Nombres
Bordeaux, 10(1):135–162, 1998.
[Mey70] Y. Meyer. Nombres de Pisot, nombres de Salem et analyse harmonique. Lecture Notes in Mathe-
matics, Vol. 117. Springer-Verlag, Berlin-New York, 1970. Cours Peccot donné au Collège de France
en avril-mai 1969.
[Mey72] Y. Meyer. Algebraic numbers and harmonic analysis. North-Holland Publishing Co., Amsterdam-
London; American Elsevier Publishing Co., Inc., New York, 1972. North-Holland Mathematical
Library, Vol. 2.
[Mey95] Y. Meyer. Quasicrystals, Diophantine approximation and algebraic numbers. In Beyond quasicrys-
tals (Les Houches, 1994), pages 3–16. Springer, Berlin, 1995.
[Mey12] Y. Meyer. Quasicrystals, almost periodic patterns, mean-periodic functions and irregular sampling.
Afr. Diaspora J. Math., 13(1):1–45, 2012.
[Mey20] Y. Meyer. From Salomon Bochner to Dan Shechtman. Trans. R. Norw. Soc. Sci. Lett., 1:1–22,
2020.
[MH40] M. Morse and G. A. Hedlund. Symbolic dynamics II. Sturmian trajectories. Amer. J. Math., 62:1–
42, 1940.
[Moo] R. V. Moody. Meyer sets and diffraction. This volume.
[Moo97] R. V. Moody. Model sets and their duals. In The mathematics of long-range aperiodic order (Wa-
terloo, ON, 1995), volume 489 of NATO Adv. Sci. Inst. Ser. C Math. Phys. Sci., pages 403–441.
Kluwer Acad. Publ., Dordrecht, 1997.
24 V. BERTHÉ AND R. YASSAWI

[Moo00] R. V. Moody. Model sets: a survey. In From Quasicrystals to More Complex Systems. Springer
Verlag, 2000. F. Axel, F. Dénoyer, and J.-P. Gazeau, Centre de physique Les Houches.
[Mor21] H. M. Morse. Recurrent geodesics on a surface of negative curvature. Trans. Amer. Math. Soc., 22,
1921.
[Moz89] S. Mozes. Tilings, substitution systems and dynamical systems generated by them. J. Analyse
Math., 53:139–186, 1989.
[OU] A. Olevskii and A. Ulanovskii. Meyer sets and related problems. This volume.
[Par60] W. Parry. On the β-expansions of real numbers. Acta Math. Acad. Sci. Hungar., 11:401–416, 1960.
[PM07] E. Pelantová and Z. Masáková. Quasicrystals: algebraic, combinatorial and geometrical aspects.
In Physics and theoretical computer science, volume 7 of NATO Secur. Sci. Ser. D Inf. Commun.
Secur., pages 113–131. IOS, Amsterdam, 2007.
[Put18] I. F. Putnam. Cantor minimal systems, volume 70 of University Lecture Series. American Mathe-
matical Society, Providence, RI, 2018.
[Pyt02] N. Pytheas Fogg. Substitutions in Dynamics, Arithmetics and Combinatorics, volume 1794 of
Lecture Notes in Mathematics. Springer Verlag, 2002. Ed. by V. Berthé and S. Ferenczi and C.
Mauduit and A. Siegel.
[Que10] M. Queffélec. Substitution dynamical systems—spectral analysis, volume 1294 of Lecture Notes in
Mathematics. Springer-Verlag, Berlin, second edition, 2010.
[Rau82] G. Rauzy. Nombres algébriques et substitutions. Bull. Soc. Math. France, 110(2):147–178, 1982.
[Rén57] A. Rényi. Representations for real numbers and their ergodic properties. Acta Math. Acad. Sci.
Hungar., 8:477–493, 1957.
[RK01] A. Rosenfeld and R. Klette. Digital straightness. Electronic Notes in Theoretical Computer Science,
46:1 – 32, 2001. IWCIA 2001, 8th International Workshop on Combinatorial Image Analysis.
[Rob71] R. M. Robinson. Undecidability and nonperiodicity for tilings of the plane. Invent. Math., 12:177–
209, 1971.
[Rob04] E. A. Robinson, Jr. Symbolic dynamics and tilings of Rd . In Symbolic dynamics and its applications,
volume 60 of Proc. Sympos. Appl. Math., Amer. Math. Soc. Providence, RI, pages 81–119, 2004.
[Sad08] L. Sadun. Topology of tiling spaces, volume 46 of University Lecture Series. American Mathematical
Society, Providence, RI, 2008.
[SBGC84] D. Shechtman, I. Blech, D. Gratias, and J. W. Cahn. Metallic phase with long-range orientational
order and no translational symmetry. Phys. Rev. Lett., 53:1951–1953, Nov 1984.
[Sch80] K. Schmidt. On periodic expansions of Pisot numbers and Salem numbers. Bull. London Math.
Soc., 12(4):269–278, 1980.
[Sen95] M. Senechal. Quasicrystals and geometry. Cambridge University Press, Cambridge, 1995.
[Sie04] A. Siegel. Pure discrete spectrum dynamical system and periodic tiling associated with a substi-
tution. Ann. Inst. Fourier, 54(2):288–299, 2004.
[Sin06] B. Sing. Iterated function systems in mixed Euclidean and p-adic spaces. In Complexus mundi,
pages 267–276. World Sci. Publ., Hackensack, NJ, 2006.
[Soc90] J. E. S. Socolar. Weak matching rules for quasicrystals. Comm. Math. Phys., 129(3):599–619, 1990.
[Sol97] B. Solomyak. Dynamics of self-similar tilings. Ergod. Th. Dyn. Sys., 17:695–738, 1997.
[Sol07] B. Solomyak. Eigenfunctions for substitution tiling systems. In Probability and number theory—
Kanazawa 2005, volume 49 of Adv. Stud. Pure Math., pages 433–454. Math. Soc. Japan, Tokyo,
2007.
[ST09] A. Siegel and J. M. Thuswaldner. Topological properties of Rauzy fractals. Mém. Soc. Math. Fr.
(N.S.), (118):140, 2009.
[Thu89] W. P. Thurston. Groups, tilings and finite state automata. Lectures notes distributed in conjunction
with the Colloquium Series, in AMS Colloquium Lectures, 1989.
[Thu19] J. M. Thuswaldner. S-adic sequences. A bridge between dynamics, arithmetic, and geometry.
arXiv:1908.05954, preprint, 2019.
[Vee65] W. A. Veech. Almost automorphic functions on groups. Amer. J. Math., 87:719–751, 1965.
[Vee69] William A. Veech. Properties of minimal functions on abelian groups. Amer. J. Math., 91:415–440,
1969.
[VGG04] J.-L. Verger Gaugry and J.-P. Gazeau. Geometric study of the beta-integers for a Perron number
and mathematical quasicrystals. J. Théor. Nombres Bordeaux, 16:125–149, 2004.
 MEYER 25

[VM01] J. Vidal and R. Mosseri. Generalized quasiperiodic Rauzy tilings. J. Phys. A, 34(18):3927–3938,
2001.
[Yoc06] J.-C. Yoccoz. Continued fraction algorithms for interval exchange maps: an introduction. In Fron-
tiers in number theory, physics, and geometry. I, pages 401–435. Springer, Berlin, 2006.

Université de Paris, IRIF, CNRS, F-75013 Paris, France

Email address: [email protected]

School of Mathematics and Statistics, The Open University, United Kingdom

Email address: [email protected]