# Hausdorff dimension

In mathematics, the **Hausdorff dimension** (also known as the **Hausdorff–Besicovitch dimension**) is an extended non-negative real number associated with any metric space. The Hausdorff dimension generalizes the notion of the dimension of a real vector space. That is, the Hausdorff dimension of an *n*-dimensional inner product space equals *n*. This means, for example, the Hausdorff dimension of a point is zero, the Hausdorff dimension of a line is one, and the Hausdorff dimension of the plane is two. There are, however, many irregular sets that have noninteger Hausdorff dimension.

The concept was introduced in 1918 by the mathematician Felix Hausdorff. Many of the technical developments used to compute the Hausdorff dimension for highly irregular sets were obtained by Abram Samoilovitch Besicovitch. The Hausdorff dimension is a successor to the simpler, but usually equivalent, box-counting or Minkowski–Bouligand dimension.

## Intuition

The intuitive concept of dimension of a geometric object *X* is the number of independent parameters one needs to pick out a unique point inside. However, any point specified by two parameters can be instead specified by one, because the cardinality of the real plane is equal to the cardinality of the real line (this can be seen by an argument involving interweaving the digits of two numbers to yield a single number encoding the same information.) The example of a space-filling curve shows that one can even take one real number into two both surjectively (so all pairs of numbers are covered) and *continuously*, so that a one-dimensional object completely fills up a higher-dimensional object.

Every space filling curve hits some points multiple times, and does not have a continuous inverse. It is impossible to map two dimensions onto one in a way that is continuous and continuously invertible. The topological dimension, also called Lebesgue covering dimension, explains why. This dimension is *n* if, in every covering of *X* by small open balls, there is at least one point where *n*+1 balls overlap. For example, when one covers a line with short open intervals, some points must be covered twice, giving dimension *n* = 1.

But topological dimension is a very crude measure of the local size of a space (size near a point). A curve that is almost space-filling can still have topological dimension one, even if it fills up most of the area of a region. A fractal has an integer topological dimension, but in terms of the amount of space it takes up, it behaves like a higher-dimensional space.

The Hausdorff dimension measures the local size of a space taking into account the distance between points, the metric. Consider the number N(*r*) of balls of radius at most *r* required to cover *X* completely. When *r* is very small, N(*r*) grows polynomially with 1/*r*. For a sufficiently well-behaved *X*, the Hausdorff dimension is the unique number *d* such that N(*r*) grows as 1/*r ^{d}* as

*r*approaches zero. More precisely, this defines the box-counting dimension, which equals the Hausdorff dimension when the value

*d*is a critical boundary between growth rates that are insufficient to cover the space, and growth rates that are overabundant.

For shapes that are smooth, or shapes with a small number of corners, the shapes of traditional geometry and science, the Hausdorff dimension is an integer agreeing with the topological dimension. But Benoît Mandelbrot observed that fractals, sets with noninteger Hausdorff dimensions, are found everywhere in nature. He observed that the proper idealization of most rough shapes you see around you is not in terms of smooth idealized shapes, but in terms of fractal idealized shapes:

Clouds are not spheres, mountains are not cones, coastlines are not circles, and bark is not smooth, nor does lightning travel in a straight line.

^{[1]}

For fractals that occur in nature, the Hausdorff and box-counting dimension coincide. The packing dimension is yet another similar notion which gives the same value for many shapes, but there are well documented exceptions where all these dimensions differ.

## Formal definitions

### Hausdorff content

Let *X* be a metric space. If *S* ⊂ *X* and *d* ∈ [0, ∞), the *d*-dimensional **Hausdorff content** of *S* is defined by

In other words, is the infimum of the set of numbers δ ≥ 0 such that there is some (indexed) collection of balls covering *S* with *r _{i}* > 0 for each

*i*∈

*I*that satisfies . (Here, we use the standard convention that inf Ø =∞.)

### Hausdorff dimension

The **Hausdorff dimension** of *X* is defined by

Equivalently, dim_{H}(*X*) may be defined as the infimum of the set of *d* ∈ [0, ∞) such that the *d*-dimensional Hausdorff measure of *X* is zero. This is the same as the supremum of the set of *d* ∈ [0, ∞) such that the *d*-dimensional Hausdorff measure of *X* is infinite (except that when this latter set of numbers *d* is empty the Hausdorff dimension is zero).

## Examples

- The Euclidean space
**R**^{n}has Hausdorff dimension*n*. - The circle
**S**^{1}has Hausdorff dimension 1. - Countable sets have Hausdorff dimension 0.
- Fractals often are spaces whose Hausdorff dimension strictly exceeds the topological dimension. For example, the Cantor set (a zero-dimensional topological space) is a union of two copies of itself, each copy shrunk by a factor 1/3; this fact can be used to prove that its Hausdorff dimension is ln(2)/ln(3) ≈ 0.63. The Sierpinski triangle is a union of three copies of itself, each copy shrunk by a factor of 1/2; this yields a Hausdorff dimension of ln(3)/ln(2) ≈ 1.58.
- Space-filling curves like the Peano and the Sierpiński curve have the same Hausdorff dimension as the space they fill.
- The trajectory of Brownian motion in dimension 2 and above has Hausdorff dimension 2 almost surely.
- An early paper by Benoit Mandelbrot entitled
*How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional Dimension*and subsequent work by other authors have claimed that the Hausdorff dimension of many coastlines can be estimated. Their results have varied from 1.02 for the coastline of South Africa to 1.25 for the west coast of Great Britain. However, 'fractal dimensions' of coastlines and many other natural phenomena are largely heuristic and cannot be regarded rigorously as a Hausdorff dimension. It is based on scaling properties of coastlines at a large range of scales; however, it does not include all arbitrarily small scales, where measurements would depend on atomic and sub-atomic structures, and are not well defined. - The bond system of an amorphous solid changes its Hausdorff dimension from Euclidean 3 below glass transition temperature
*T*_{g}(where the amorphous material is solid), to fractal 2.55±0.05 above*T*_{g}, where the amorphous material is liquid.^{[2]}

## Properties of Hausdorff dimension

### Hausdorff dimension and inductive dimension

Let *X* be an arbitrary separable metric space. There is a topological notion of inductive dimension for *X* which is defined recursively. It is always an integer (or +∞) and is denoted dim_{ind}(*X*).

**Theorem**. Suppose *X* is non-empty. Then

Moreover,

where *Y* ranges over metric spaces homeomorphic to *X*. In other words, *X* and *Y* have the same underlying set of points and the metric *d*_{Y} of *Y* is topologically equivalent to *d*_{X}.

These results were originally established by Edward Szpilrajn (1907–1976). The treatment in Chapter VII of the Hurewicz and Wallman reference is particularly recommended.

### Hausdorff dimension and Minkowski dimension

The Minkowski dimension is similar to, and at least as large as, the Hausdorff dimension, and they are equal in many situations. However, the set of rational points in [0, 1] has Hausdorff dimension zero and Minkowski dimension one. There are also compact sets for which the Minkowski dimension is strictly larger than the Hausdorff dimension.

### Hausdorff dimensions and Frostman measures

If there is a measure μ defined on Borel subsets of a metric space *X* such that μ(*X*) > 0 and μ(*B*(*x*, *r*)) ≤ *r ^{s}* holds for some constant

*s*> 0 and for every ball

*B*(

*x*,

*r*) in

*X*, then dim

_{Haus}(

*X*) ≥

*s*. A partial converse is provided by Frostman's lemma. That article also discusses another useful characterization of the Hausdorff dimension.

### Behaviour under unions and products

If is a finite or countable union, then

This can be verified directly from the definition.

If *X* and *Y* are non-empty metric spaces, then the Hausdorff dimension of their product satisfies^{[3]}

This inequality can be strict. It is possible to find two sets of dimension 0 whose product has dimension 1.^{[4]} In the opposite direction, it is known that when *X* and *Y* are Borel subsets of **R**^{n}, the Hausdorff dimension of *X* × *Y* is bounded from above by the Hausdorff dimension of *X* plus the upper packing dimension of *Y*. These facts are discussed in Mattila (1995).

## Self-similar sets

Many sets defined by a self-similarity condition have dimensions which can be determined explicitly. Roughly, a set *E* is self-similar if it is the fixed point of a set-valued transformation ψ, that is ψ(*E*) = *E*, although the exact definition is given below.

Theorem. Supposeare contractive mappings on

R^{n}with contraction constantr< 1. Then there is a unique_{j}non-emptycompact setAsuch that

The theorem follows from Stefan Banach's contractive mapping fixed point theorem applied to the complete metric space of non-empty compact subsets of **R**^{n} with the Hausdorff distance.^{[5]}

### The open set condition

To determine the dimension of the self-similar set *A* (in certain cases), we need a technical condition called the *open set condition* (OSC) on the sequence of contractions ψ_{i}.

There is a relatively compact open set *V* such that

where the sets in union on the left are pairwise disjoint.

The open set condition is a separation condition that ensures the images ψ_{i}(*V*) do not overlap "too much".

**Theorem**. Suppose the open set condition holds and each ψ_{i} is a similitude, that is a composition of an isometry and a dilation around some point. Then the unique fixed point of ψ is a set whose Hausdorff dimension is *s* where *s* is the unique solution of^{[6]}

The contraction coefficient of a similitude is the magnitude of the dilation.

We can use this theorem to compute the Hausdorff dimension of the Sierpinski triangle (or sometimes called Sierpinski gasket). Consider three non-collinear points *a*_{1}, *a*_{2}, *a*_{3} in the plane **R**^{2} and let ψ_{i} be the dilation of ratio 1/2 around *a _{i}*. The unique non-empty fixed point of the corresponding mapping ψ is a Sierpinski gasket and the dimension

*s*is the unique solution of

Taking natural logarithms of both sides of the above equation, we can solve for *s*, that is: *s* = ln(3)/ln(2). The Sierpinski gasket is self-similar and satisfies the OSC. In general a set *E* which is a fixed point of a mapping

is self-similar if and only if the intersections

where *s* is the Hausdorff dimension of *E* and *H ^{s}* denotes Hausdorff measure. This is clear in the case of the Sierpinski gasket (the intersections are just points), but is also true more generally:

**Theorem**. Under the same conditions as the previous theorem, the unique fixed point of ψ is self-similar.

## The Hausdorff Dimension Theorem

For any given *r* ≥ 0, and integer *n * ≥ *r*, there are *at least* continuum-many fractals of Hausdorff dimension *r*.
^{[7]}

## See also

- List of fractals by Hausdorff dimension Examples of deterministic fractals, random and natural fractals.
- Intrinsic dimension
- Packing dimension

## Notes

- ↑ {{#invoke:citation/CS1|citation |CitationClass=book }}
- ↑ {{#invoke:Citation/CS1|citation |CitationClass=journal }}
- ↑ {{#invoke:Citation/CS1|citation |CitationClass=journal }}
- ↑ {{#invoke:citation/CS1|citation |CitationClass=book }}
- ↑ {{#invoke:citation/CS1|citation |CitationClass=book }}
- ↑ {{#invoke:Citation/CS1|citation |CitationClass=journal }}
- ↑ Soltanifar, M. (2006).
*On A Sequence of Cantor Fractals*, Rose Hulman Undergraduate Mathematics Journal, Vol 7, No 1, paper 9.

## References

- {{#invoke:Citation/CS1|citation

|CitationClass=journal }}

- {{#invoke:citation/CS1|citation

|CitationClass=book }}

- {{#invoke:Citation/CS1|citation

|CitationClass=journal }}

- {{#invoke:Citation/CS1|citation

|CitationClass=journal }}

- {{#invoke:citation/CS1|citation

|CitationClass=book }}

### Historical

- {{#invoke:Citation/CS1|citation

|CitationClass=journal }}

- {{#invoke:Citation/CS1|citation

|CitationClass=journal
}}

Several selections from this volume are reprinted in {{#invoke:citation/CS1|citation
|CitationClass=book
}} See chapters 9,10,11

- {{#invoke:Citation/CS1|citation

|CitationClass=journal }}