1 Definition and basic concepts

A self-affine set is a set that can be reconstructed from finitely many affine copies of itself. Each copy may be scaled differently in different directions, and may also be rotated, translated, or sheared. This anisotropic behavior distinguishes self-affine sets from more rigid geometric constructions and makes them central objects in fractal geometry and dynamical systems.

Self-affine sets are commonly defined as attractors of iterated function systems made up of affine contractions. Their study combines linear algebra, topology, measure theory, and symbolic dynamics. Many familiar fractal sets, tilings, and digit-based constructions fit naturally into this framework.

1.1 Affine transformations

An affine transformation is a map of the form x ↦ Ax + b, where A is a linear transformation and b is a translation vector. In the self-affine setting, the linear part often compresses space unevenly in different directions. This allows one axis to shrink much faster than another, producing shapes with elongated or layered structure.

Affine maps preserve straight lines and parallelism, but not necessarily angles or lengths. Because of this, self-affine sets often exhibit directional features that are absent in self-similar sets. The linear component of the map largely governs the scaling geometry, while the translation component determines how the pieces are arranged.

1.2 Invariant sets

An invariant set for a family of maps is a set that is mapped into itself by the system. For a self-affine construction, the target set is usually recovered as the union of all image pieces under the given affine maps. This invariance is the defining property that makes the set self-reproducing at smaller scales.

Such sets are usually not invariant under a single map, but under a collection of maps acting together. The resulting object can be highly irregular, yet still possess an exact recursive description. This recursive nature is what makes self-affine sets accessible to both combinatorial and analytic methods.

1.3 Iterated function systems

An iterated function system is a finite family of maps whose repeated application generates a limiting set. In the affine case, the maps are affine contractions, and the limit set is the unique compact attractor satisfying a fixed-point equation. This viewpoint provides a standard framework for defining self-affine sets.

The system encodes the geometry of the set through repeated subdivision. Each stage replaces the whole by a union of transformed copies, and the process continues indefinitely. The attractor captures the asymptotic structure of the construction.

1.3.1 Contractive maps

Contractive maps bring points closer together under repeated iteration. In a self-affine iterated function system, each affine map must shrink distances in some norm, even if the shrinking is not uniform in all directions. This ensures that the recursive process does not expand without bound.

Contractivity is essential for the existence of a well-defined attractor. It also implies that the pieces become increasingly small at deeper levels of iteration. The resulting limit set is compact and often has empty interior, though exceptions occur in tile-like cases.

1.3.2 Attractor existence and uniqueness

For a contractive iterated function system, there exists a unique nonempty compact attractor that is invariant under the system. This attractor can be obtained as the limit of successive approximations starting from any compact set. The uniqueness follows from the contraction principle in the hyperspace of compact sets.

This attractor is the self-affine set associated with the system. Its construction is stable under iteration, which makes it a natural object for analysis. The same attractor may admit multiple descriptions, including symbolic codings and digit expansions.

1.4 Self-affine versus self-similar sets

Self-similar sets are built from maps that shrink all directions by the same factor, possibly with rotations and translations. Self-affine sets generalize this by allowing different contraction rates in different directions. As a result, self-affine sets can have much more varied geometry.

The distinction affects dimension theory, measure behavior, and connectedness. Self-similar sets often enjoy cleaner separation and symmetry properties, while self-affine sets may show strong directional bias. Many techniques for self-similar sets extend only partially to the affine setting.

2 Examples

Self-affine sets appear in several standard constructions, ranging from rectangular carpet-like fractals to tile attractors and digit systems. These examples illustrate how the same general definition can produce very different geometric outcomes. Some are purely fractal, while others connect directly to tilings and arithmetic.

2.1 Bedford–McMullen carpets

Bedford–McMullen carpets are planar self-affine sets constructed by subdividing a rectangle into a grid and selecting some of the smaller rectangles. The horizontal and vertical scaling factors are typically different, which produces a pronounced anisotropic appearance. These sets have played a major role in the development of dimension theory for self-affine fractals.

They are often used as a model family because many of their geometric quantities can be computed explicitly. Their structure is governed by combinatorial choices in each row and column of the grid. Despite their simple definition, they exhibit rich multifractal and dimensional behavior.

2.2 Sierpiński-type self-affine constructions

Sierpiński-type self-affine constructions resemble the classical Sierpiński gasket or carpet, but use affine rather than uniform contractions. The pieces may be stretched differently along coordinate directions, leading to a deformed but recursively similar pattern. Such examples help illustrate how classical fractal ideas extend beyond similarity maps.

These constructions may retain some qualitative features of the classical Sierpiński sets, such as holes or branching patterns, while losing rotational symmetry. Their geometry can depend sensitively on the chosen matrices and translations. As a result, they provide useful test cases for studying connectivity and dimension.

2.3 Tile-based examples

Some self-affine sets are tiles, meaning they cover space without overlaps except along boundaries. These tile-based examples often arise from expanding linear maps and carefully chosen digit sets. Their attractors may have nonempty interior and can serve as fundamental domains for certain lattice actions.

Tile examples connect self-affine geometry with discrete geometry and tiling theory. They often display a mix of fractal boundary behavior and regular interior structure. Because they are both self-affine and space-filling in a controlled sense, they form an important bridge between fractals and tessellations.

2.4 Digit set attractors

Digit set attractors arise from expansions in which points are represented using repeated digits and a linear base transformation. The attractor is the set of all points obtained from infinite digit expansions. Such systems are closely related to number systems and canonical representations.

These attractors may coincide with self-affine tiles or with more delicate fractal sets depending on the digit choice. Their combinatorial definition makes them suitable for arithmetic analysis. In many cases, the digit structure directly determines the shape and boundary complexity of the set.

3 Geometric properties

The geometry of self-affine sets is strongly influenced by directional scaling and by the arrangement of the image pieces. Properties such as connectedness, boundary regularity, and separation depend on both the linear maps and the translation vectors. The resulting sets often combine rigid algebraic structure with irregular fractal detail.

3.1 Anisotropic scaling

Anisotropic scaling means that different directions are contracted by different amounts. This is a defining feature of self-affine geometry and often produces sets that are stretched, slanted, or layered. The same set may look nearly one-dimensional in one direction and much thicker in another.

This directional behavior affects many geometric quantities, including dimension and measure. It also makes the local appearance of the set vary with orientation. In many examples, the preferred directions are determined by the eigenspaces of the linear maps.

3.2 Connectedness and topology

The topological structure of a self-affine set may range from totally disconnected dusts to connected carpets or tiles. Connectedness depends on how the affine pieces overlap and touch. Even small changes in translations can alter whether the attractor breaks into separate components.

Topological questions often focus on components, local cut points, and the presence of holes. In some families, connectedness can be characterized combinatorially. The topology of the set is frequently more subtle than its recursive definition suggests.

3.3 Rectifiability

Rectifiability concerns whether a set can be approximated, in a measure-theoretic sense, by smooth curves or surfaces. Most self-affine sets of noninteger dimension are not rectifiable in the classical sense. Their irregular scaling often prevents them from lying in finitely many smooth pieces.

However, certain self-affine sets with positive Lebesgue measure or special boundary structure can show partial regularity. Rectifiability questions are closely tied to dimension, tangent structures, and measure distribution. They remain an active part of geometric measure theory.

3.4 Separation conditions

Separation conditions control the amount of overlap between the pieces of a self-affine system. Strong separation means the pieces are disjoint, while weaker variants allow limited intersection. Such conditions simplify dimension calculations and help prevent pathological behavior.

In self-affine settings, overlap can be especially difficult to analyze because different directions may interact unevenly. Many results depend on assumptions that the pieces are sufficiently separated. When separation fails, the geometry can become more complicated and dimension formulas may no longer be exact.

4 Dimension theory

Dimension theory is one of the central topics in the study of self-affine sets. Because these sets are irregular and anisotropic, several notions of dimension may differ from one another. Understanding these differences is essential for describing the size and complexity of the attractor.

4.1 Hausdorff dimension

Hausdorff dimension is a fine-scale notion of size based on coverings by sets of small diameter. For self-affine sets, this dimension can be difficult to compute because standard similarity-based formulas usually do not apply directly. The anisotropic nature of the construction requires more refined methods.

In many cases, the Hausdorff dimension lies strictly between topological dimension and ambient dimension. It may reflect both the arrangement of the pieces and the strength of contraction in each direction. For certain structured families, explicit formulas are available.

4.2 Box-counting dimension

Box-counting dimension estimates how many small boxes are needed to cover the set at a given scale. It is often easier to compute than Hausdorff dimension, though it may be less sensitive to fine geometry. For self-affine sets, it often provides a useful first approximation to complexity.

Because self-affine sets may have different scaling in different directions, box-counting arguments typically require careful anisotropic estimates. In some examples, the box-counting dimension agrees with the Hausdorff dimension; in others, it does not. The comparison between these dimensions is a key theme in the subject.

4.3 Affinity dimension

Affinity dimension is a dimension-like quantity derived from the linear parts of the affine maps. It is designed to capture the expected size of a self-affine set before overlaps are taken into account. The definition uses singular values of the linear transformations and a subadditive pressure-type construction.

This quantity often serves as an upper bound for Hausdorff dimension. In favorable situations, it equals the actual dimension. It has become one of the standard tools for predicting the size of self-affine attractors.

4.4 Falconer’s dimension theory

Falconer’s dimension theory concerns generic dimension results for self-affine sets, especially under typical choices of linear maps. It provides conditions under which the Hausdorff dimension matches the affinity dimension for almost all parameter values. This theory has strongly influenced modern research in the area.

The approach combines probabilistic ideas with geometric estimates. It is particularly useful when exact formulas are not available for specific systems. Falconer’s work helped establish self-affine dimension theory as a major field in fractal analysis.

4.4.1 Generic dimension results

Generic dimension results describe what happens for most parameter choices in a suitable sense. For many families of self-affine systems, the expected dimension formula holds for almost all translations or linear parameters. These results suggest that dimension drop is exceptional rather than typical.

Such theorems often depend on transversality or randomization arguments. They provide strong evidence that the affinity dimension is the correct benchmark in general. However, the proofs usually leave open the structure of the exceptional cases.

4.4.2 Exceptional parameter sets

Exceptional parameter sets are those for which the generic dimension formula fails. They may include overlaps, resonance relations, or special algebraic configurations. These sets are often small in measure but can be difficult to characterize precisely.

Understanding exceptional parameters remains one of the main challenges in the field. Even when they form a negligible subset of the parameter space, they may display rich and unexpected structure. Studying them helps clarify the limits of generic dimension theory.

5 Measures on self-affine sets

Measures on self-affine sets describe how mass is distributed across the attractor. They are important for linking geometric structure with ergodic and probabilistic methods. Many natural measures arise from symbolic codings or invariant distributions on the underlying system.

5.1 Self-affine measures

A self-affine measure is a probability measure that is invariant under the weighted action of the affine maps. Such measures typically satisfy a fixed-point equation analogous to that of the attractor itself. They provide a natural way to assign weights to the different pieces of the set.

These measures can reflect nonuniform distribution across scales and directions. They are often used to study pointwise dimension and multifractal behavior. In some cases, they are supported on the entire attractor; in others, they concentrate on smaller subsets.

5.2 Invariant probability measures

Invariant probability measures arise when a measure is preserved by the dynamics of the iterated function system or a related shift space. They give a statistical description of how orbits visit different parts of the attractor. Such measures are central in ergodic theory.

For self-affine systems, invariant measures can be constructed from Bernoulli weights or Markov processes on the symbolic side. These measures often project to geometric measures on the attractor. Their properties are closely tied to entropy, Lyapunov exponents, and dimension.

5.3 Ergodic properties

Ergodic properties concern whether long-term averages along orbits reflect space averages. In self-affine settings, ergodicity helps connect symbolic randomness with geometric regularity. It is especially useful when studying typical points and typical directions.

Ergodic theorems often yield formulas for dimensions and entropy-related quantities. They also help distinguish between typical behavior and exceptional geometric configurations. The interplay between ergodicity and anisotropic scaling is a major theme in the measure-theoretic study of self-affine sets.

5.4 Dimension of measures

The dimension of a measure describes how mass scales around typical points. For self-affine measures, this may differ from both the dimension of the support and the ambient dimension. It is often computed using combinations of entropy and Lyapunov exponents.

Measure dimension is important for multifractal analysis and for understanding singularity properties. In favorable cases, exact dimension results are available, meaning the local scaling exponent is constant almost everywhere. These results connect the dynamics of the system with the geometry of the measure.

6 Symbolic dynamics and coding

Symbolic dynamics provides a combinatorial representation of self-affine sets through sequences of symbols. Each symbol corresponds to one of the affine maps, and infinite sequences encode points in the attractor. This approach translates geometric questions into problems about sequences and shift maps.

6.1 Shift spaces

A shift space is a collection of symbol sequences closed under the shift operation. For self-affine systems, the shift space records the allowed sequences of maps used in the construction. It serves as the symbolic model for the attractor.

Shift spaces can be full or constrained, depending on whether every sequence is permitted. They are often used to organize the recursive structure of the set. The symbolic viewpoint is particularly powerful for studying periodicity, entropy, and invariant measures.

6.2 Address maps

An address map assigns to each symbolic sequence the corresponding point in the attractor. Different sequences may sometimes represent the same point, especially when overlaps occur. The address map therefore encodes both the geometry and the combinatorics of the construction.

When the coding is unique, the address map gives a clean parametrization of the set. In more complicated cases, the multiplicity of addresses reveals structural overlap. Address maps are a standard tool for passing between symbolic and geometric descriptions.

6.3 Coding of points in the attractor

Coding a point means expressing it as the limit of a sequence of chosen affine maps. This produces a symbolic label for each point in the attractor. The coding often resembles a digit expansion, with each step selecting one piece of the recursive decomposition.

Coding is useful for identifying special points, such as boundary points or periodic points. It also helps study local geometry by examining which sequences lead to a given location. The quality of the coding depends strongly on the separation properties of the system.

6.4 Subshifts of finite type

A subshift of finite type is a shift space defined by finitely many forbidden transitions. Such systems arise when only certain map sequences are allowed, often because of adjacency rules or overlap restrictions. They provide a manageable symbolic model for many self-affine constructions.

In this setting, the geometry of the attractor can be studied through a finite directed graph. This makes entropy and transition behavior easier to analyze. Subshifts of finite type frequently appear in tile boundaries, constrained digit systems, and structured fractal carpets.

7 Tilings and number systems

Self-affine sets are closely related to tilings of Euclidean space and to expansions in abstract number systems. In these contexts, the attractor may function as a fundamental region or as the set of all possible digit expansions. This links fractal geometry with algebra and discrete geometry.

7.1 Self-affine tiles

A self-affine tile is a compact set that tiles space under translations and is also self-affine under an expanding linear map. These objects often have nonempty interior, yet their boundaries may be highly irregular. They are among the best-studied examples of self-affine sets.

Self-affine tiles typically arise from algebraic constructions involving matrices and digit sets. Their dual role as fractals and tiles makes them particularly significant. They provide concrete examples where geometric, combinatorial, and arithmetic ideas meet.

7.2 Expanding matrices and digit sets

An expanding matrix is a linear map whose eigenvalues all have modulus greater than one. Its inverse therefore acts as a contraction, which is suitable for defining self-affine attractors. Digit sets specify the translation components used in the recursive decomposition.

The pair of an expanding matrix and a digit set determines the resulting attractor. The choice of digits controls whether the set is a tile, a dust, or a more intricate fractal. This framework is fundamental in the study of linear numeration systems.

7.3 Canonical number systems

A canonical number system is a representation scheme in which every element of a suitable algebraic set has a unique digit expansion relative to a base transformation. Self-affine sets often arise as the geometric realization of such systems. The digits describe the recursive decomposition of the attractor.

These systems generalize familiar positional notation to higher dimensions or algebraic settings. They connect self-affine geometry with arithmetic representation theory. Questions about uniqueness, completeness, and carry structure are central to this topic.

7.4 Boundary structure of tiles

The boundary of a self-affine tile often has far more complicated structure than its interior. It may be fractal, connected in intricate ways, or described by a lower-dimensional symbolic system. Understanding the boundary is important for both topology and dimension theory.

Boundary analysis often uses graph-directed constructions and substitution rules. In many cases, the boundary inherits a self-affine or self-similar nature of its own. Its geometry can determine how the tile meets its translated copies in a tiling.

Self-affine sets appear across several areas of mathematics and applied modeling. Their recursive structure makes them useful for describing complex patterns generated by simple rules. They also serve as test cases for broader theories of fractal dimension, dynamics, and geometric structure.

8.1 Fractal image models

In fractal image modeling, self-affine constructions can approximate natural textures such as terrain, clouds, or rough surfaces. The anisotropic scaling property is especially useful for modeling features that vary differently in horizontal and vertical directions. This makes self-affine ideas relevant in graphics and signal analysis.

The models are not exact descriptions of natural objects, but rather structured approximations. Their parameters can be adjusted to control roughness and directional behavior. As a result, self-affine geometry has influenced both theoretical and practical approaches to image generation.

8.2 Dynamical systems

Self-affine sets are closely linked to dynamical systems through iterated maps, invariant measures, and symbolic coding. They provide geometric realizations of orbit structures and attractors. This connection makes them valuable in studying long-term behavior under repeated transformation.

The dynamics can be viewed either on the geometric set or on the associated symbolic space. Concepts such as entropy, pressure, and ergodicity often enter the analysis. This interaction has helped unify fractal geometry with smooth and discrete dynamical systems.

8.3 Quasicrystals and aperiodic order

Self-affine structures are related to aperiodic tilings and quasicrystal models, where local repetition coexists with global nonperiodicity. Expanding maps and substitution rules can generate patterns with rich order but no translational periodicity. These ideas overlap with the study of mathematical diffraction and symbolic substitution systems.

Although self-affine sets are not identical to quasicrystals, both involve recursive organization and nontrivial scaling. Tile-like self-affine constructions often appear in this broader context. The shared focus on hierarchical structure makes the connection mathematically natural.

8.4 Open problems and research directions

Many open problems concern the precise relationship between affine geometry, overlaps, and dimension. Researchers continue to seek general criteria for exact dimension formulas, boundary regularity, and measure behavior. The exceptional cases, where standard predictions fail, are especially challenging.

Other directions include higher-dimensional tilings, random self-affine systems, and finer multifractal analysis. There is also active work on understanding coding multiplicity, overlap structure, and the geometry of parameter spaces. The field remains dynamic because even simple affine rules can produce unexpectedly subtle sets.