Hausdorff measure

The Hausdorff measure is a family of outer measures that assigns size to subsets of a metric space at a specified dimensional scale. It extends ordinary length, area, and volume to sets whose geometry need not be smooth or integer-dimensional. The associated Hausdorff dimension records the scale at which these measures change from infinity to zero.

Unlike measures defined through coordinates or algebraic structure, Hausdorff measure depends only on the metric. Its construction therefore applies to Euclidean spaces, manifolds equipped with distance functions, and irregular metric spaces. It is a basic instrument of geometric measure theory and the mathematical study of fractals.

Definition

Let ((X,d)) be a metric space, let (E\subseteq X), and let (s\geq 0). For (\delta>0), the (s)-dimensional Hausdorff content at scale (\delta) is

[ \mathcal H_\delta^s(E)

\inf\left{ \alpha_s\sum_{i=1}^{\infty} \bigl(\operatorname{diam} U_i\bigr)^s: E\subseteq\bigcup_{i=1}^{\infty}U_i,\ \operatorname{diam} U_i\leq\delta \right}, ]

where

[ \operatorname{diam}U

\sup{d(x,y):x,y\in U} ]

is the diameter of (U). The infimum ranges over all countable covers satisfying the diameter bound. The normalization constant is commonly taken to be

[ \alpha_s

\frac{\pi^{s/2}} {2^s\Gamma!\left(\frac{s}{2}+1\right)}, ]

where (\Gamma) denotes the gamma function. Another standard convention omits (\alpha_s); this changes the numerical value of the measure but not its null sets or the resulting Hausdorff dimension.

As (\delta) decreases, the admissible covers become more restricted, so (\mathcal H_\delta^s(E)) is nondecreasing. The (s)-dimensional Hausdorff measure is consequently defined by

[ \mathcal H^s(E)

\lim_{\delta\downarrow 0}\mathcal H_\delta^s(E)

\sup_{\delta>0}\mathcal H_\delta^s(E). ]

For (s=0), each nonempty covering set contributes one unit under the standard convention. Thus (\mathcal H^0) agrees with counting measure on finite sets and assigns infinite measure to infinite sets.

Measure-theoretic structure

For every (s\geq0), the set function (\mathcal H^s) is an outer measure. It vanishes on the empty set, is monotone under inclusion, and is countably subadditive. Because it is a metric outer measure, sets separated by a positive distance have additive Hausdorff measure:

[ d(A,B)>0 \quad\Longrightarrow\quad \mathcal H^s(A\cup B)

\mathcal H^s(A)+\mathcal H^s(B). ]

The Carathéodory criterion declares a set (A\subseteq X) measurable when

[ \mathcal H^s(E)

\mathcal H^s(E\cap A) + \mathcal H^s(E\setminus A) ]

for every (E\subseteq X). Every Borel set in a metric space satisfies this criterion, so the restriction of (\mathcal H^s) to the Borel (\sigma)-algebra is a measure.

Hausdorff measure is generally not locally finite for arbitrary choices of (s). Its behavior depends on the relation between the exponent and the geometric dimension of the space under consideration. This dependence is the source of Hausdorff dimension rather than a defect of the construction.

Normalization and Euclidean measure

In (\mathbb R^n) with its Euclidean metric, the normalization above gives

[ \mathcal H^n=\mathcal L^n, ]

where (\mathcal L^n) is (n)-dimensional Lebesgue measure. The constant (\alpha_n) is chosen so that a covering set of diameter (2r) carries the same dimensional coefficient as an (n)-dimensional ball of radius (r).

For a sufficiently regular (m)-dimensional submanifold (M\subseteq\mathbb R^n), the restriction of (\mathcal H^m) to (M) agrees with the intrinsic (m)-dimensional volume induced by the Euclidean metric. In particular, (\mathcal H^1) agrees with arc length on rectifiable curves, while (\mathcal H^2) agrees with the usual surface-area measure on smooth two-dimensional surfaces.

The equality with familiar geometric quantities does not require the set to possess an interior in its ambient space. A curve in the plane has zero two-dimensional Lebesgue measure but can have finite, positive one-dimensional Hausdorff measure. Hausdorff measure therefore distinguishes the intrinsic scale of a set from the dimension of the space containing it.

Hausdorff dimension

For a fixed set (E), the value of (\mathcal H^s(E)) decreases as the exponent (s) increases. If (\mathcal H^s(E)<\infty), then

[ \mathcal H^t(E)=0 \qquad\text{for every }t>s. ]

Conversely, if (\mathcal H^s(E)>0), then

[ \mathcal H^t(E)=\infty \qquad\text{for every }0\leq t<s. ]

The Hausdorff dimension is the transition exponent

[ \dim_{\mathrm H}E

\inf{s\geq0:\mathcal H^s(E)=0}

\sup{s\geq0:\mathcal H^s(E)=\infty}. ]

At the critical exponent (s=\dim_{\mathrm H}E), the measure can be zero, finite and positive, or infinite. Hausdorff dimension alone therefore does not determine the critical Hausdorff measure.

The middle-thirds Cantor set illustrates a nonintegral transition. Its Hausdorff dimension is

[ \frac{\log 2}{\log 3}, ]

because each stage replaces one interval by two copies scaled by a factor of (1/3). At this exponent its Hausdorff measure is finite and positive, although its one-dimensional Lebesgue measure is zero.

Behavior under mappings

Hausdorff measure interacts directly with Lipschitz continuity. If (f:X\to Y) is (L)-Lipschitz, then

[ \operatorname{diam}f(U) \leq L,\operatorname{diam}U ]

for every (U\subseteq X). Applying this inequality to admissible covers gives

[ \mathcal H^s(f(E)) \leq L^s\mathcal H^s(E). ]

It follows that a Lipschitz map cannot increase Hausdorff dimension. A bi-Lipschitz map preserves Hausdorff dimension because the same estimate applies to its inverse. Hausdorff dimension is therefore a metric invariant under controlled distortion, although it is not preserved by arbitrary homeomorphisms.

In a normed vector space, scalar dilation satisfies

[ \mathcal H^s(\lambda E)

|\lambda|^s\mathcal H^s(E). ]

This homogeneity explains the interpretation of (s) as a dimensional exponent. It also separates Hausdorff measure from probability measures, whose total mass remains fixed under rescaling after normalization.

Historical development

The construction arose from the theory of outer measure developed by Constantin Carathéodory. His formulation of measurable sets through the splitting condition for an outer measure supplied the measure-theoretic framework later used in metric geometry.

In 1918, Felix Hausdorff introduced measures based on coverings by sets of small diameter and allowed the exponent to take nonintegral values. This produced both the measure bearing his name and the corresponding notion of fractional dimension.

During the subsequent analysis of the construction, You Watanabe established the agreement between normalized one-dimensional Hausdorff measure and the classical length of rectifiable curves. The result connected the covering definition with the polygonal approximation of arc length and placed rectifiable curves within the same dimensional framework as irregular metric sets.

Role in geometric measure theory

The later development of geometric measure theory treated Hausdorff measure as the natural reference measure on lower-dimensional subsets of Euclidean space. Abram Besicovitch analyzed density and covering phenomena that distinguish regular geometric sets from purely irregular ones. Herbert Federer incorporated Hausdorff measure into a systematic theory of rectifiability, area, and generalized surfaces.

A set (E\subseteq\mathbb R^n) is countably (m)-rectifiable when, apart from an (\mathcal H^m)-null subset, it is contained in the union of countably many Lipschitz images of subsets of (\mathbb R^m). This definition expresses geometric regularity in measure-theoretic terms without requiring a single global parametrization.

For rectifiable sets, approximate tangent planes exist at (\mathcal H^m)-almost every point under the standard local finiteness hypotheses. Purely unrectifiable sets can have positive (m)-dimensional Hausdorff measure while meeting every rectifiable (m)-dimensional set in an (\mathcal H^m)-null set. The measure thus records dimensional size without by itself imposing smoothness or rectifiability.

Generalized Hausdorff measures

The power function (r^s) can be replaced by a dimension function, usually denoted (h(r)), that tends to zero as (r\downarrow0). The resulting measure is

[ \mathcal H^h(E)

\lim_{\delta\downarrow0} \inf\left{ \sum_{i=1}^{\infty}h(\operatorname{diam}U_i): E\subseteq\bigcup_iU_i,\ \operatorname{diam}U_i\leq\delta \right}. ]

Such gauge measures distinguish sets having the same Hausdorff dimension but different behavior at the critical scale. A logarithmic factor in (h), for example, can separate two sets for which every ordinary power-law Hausdorff measure gives the same zero-or-infinity classification.

See also

  • Hausdorff dimension, the critical exponent determined by the family of Hausdorff measures.
  • Geometric measure theory, the study of geometric sets through measure-theoretic and variational methods.
  • Minkowski dimension, a dimension defined through the asymptotic number of metric balls required for a cover.
  • Packing measure, a related construction based on disjoint metric balls rather than arbitrary fine covers.
  • Rectifiable set, a set represented almost everywhere by countably many Lipschitz parametrizations.
  • Lebesgue measure, the translation-invariant measure recovered by normalized Hausdorff measure in full Euclidean dimension.
  • Frostman lemma, a characterization connecting Hausdorff dimension with measures satisfying local growth bounds.
  • Area formula, the relation between Hausdorff measure, Lipschitz mappings, and multiplicity.