research36 min studyLesson 831 in Mathematics

Hausdorff measures and dimension

Fix a geometric normalization and measure sets that occupy noninteger dimensions.

Before you begin

On this page

The question for today

What you’ll understand

State the normalized Hausdorff covering construction and its Euclidean calibration.
Prove a self-similar dimension using both a covering bound and a measure bound.
Distinguish Hausdorff dimension from the value of measure at the critical exponent.

Lebesgue measure depends on the ambient dimension: a curve in a plane has zero planar area even when it has substantial length. Hausdorff measure instead assigns a separate measuring scale to each dimension, including noninteger dimensions. The normalization matters whenever an exact length or area is claimed. We will fix it explicitly, then calculate the dimension of a separated Cantor construction without assuming that a suggestive scaling equation is already a proof.

Normalize the covering cost before using geometric constants

For positive real s, let the cost of a covering set of diameter d be $\omega_s(d/2)^s$ , where $\omega_s=\pi^{s/2}/\Gamma(1+s/2)$ . Restrict coverings to diameters at most delta, minimize their total cost, and let delta decrease to zero. The result is the normalized Hausdorff outer measure. For integer m this normalization makes $\mathcal H^m$ equal ordinary m-dimensional Lebesgue measure on an m-plane. We separately use counting measure for $\mathcal H^0$ .

The diameter bound is essential: one large covering set cannot substitute for arbitrarily fine measurement. The resulting outer measure restricts to a countably additive Borel measure, and measurable sets may be integrated against it. A different positive normalization would leave dimension unchanged but alter exact measure constants. We keep the same normalization in density ratios, area and coarea formulas, and current masses throughout this module.

\mathcal H^s(E)=\lim_{\delta\downarrow0}\inf\left\{\sum_i\omega_s\left(\frac{\operatorname{diam}U_i}{2}\right)^s:E\subseteq\bigcup_iU_i,\ \operatorname{diam}U_i\le\delta\right\}.

For integer m, this convention gives the usual volume on Euclidean m-planes.↔ Scroll wide equations horizontally.

Dimension detects a transition between measuring scales

Hausdorff dimension is the threshold above which the Hausdorff measures vanish. Below that threshold the measures are infinite, apart from the usual empty-set convention. At the threshold itself, the measure can be zero, finite positive, or infinite. A claim about dimension therefore does not automatically specify the critical measure. Smooth m-dimensional pieces have dimension m, but irregular sets can have noninteger dimension or combine pieces of different dimensions.

Dilation by a positive factor multiplies s-dimensional measure by that factor to the power s. Lipschitz maps similarly give an upper bound with the Lipschitz constant raised to s; they cannot increase Hausdorff dimension. An inverse Lipschitz map gives a matching reverse estimate, so bi-Lipschitz changes preserve dimension. These statements follow by transporting fine covers, which also explains why their hypotheses concern small-scale distance distortion rather than differentiability alone.

\dim_H E=\inf\{s:\mathcal H^s(E)=0\},\qquad \mathcal H^s(\lambda E)=\lambda^s\mathcal H^s(E).

The threshold determines dimension; scaling controls exact measure only after the normalization and exponent are fixed.↔ Scroll wide equations horizontally.

A lower bound needs more than a convenient cover

A construction with many small pieces immediately supplies an upper bound by covering with those pieces. To prove a matching lower bound, distribute a probability measure over the set and estimate how much mass any short interval can capture. If every covering set of diameter r carries at most a constant times $r^s$ , summing over a cover forces its total s-cost to stay bounded below. This is the mass-distribution argument, and it prevents a hidden more efficient cover from lowering the dimension.

For the two-branch construction retaining the left and right quarters, level-k intervals have length $4^{-k}$ and probability mass $2^{-k}$ . Their separation controls arbitrary interval covers, not merely the construction's own preferred covers. The resulting estimates establish finite positive measure at exponent one half, but do not calculate its exact normalized value. Keeping this distinction avoids treating the self-similar counting relation as an unexplained formula for critical Hausdorff measure.

\mu(U)\le C(\operatorname{diam}U)^s\ \Longrightarrow\ \mathcal H^s(E)\ge\frac{\omega_s}{2^sC}\quad\text{if }\mu(E)=1.

The lower bound follows by summing the measure estimate over every sufficiently fine cover.↔ Scroll wide equations horizontally.

Work it through

The two-quarter Cantor set has dimension one half

Start with [0,1] and repeatedly retain the left and right quarters of every remaining interval. Prove the dimension of the limiting set C using a cover estimate and a lower mass estimate.

At level k there are $2^k$ intervals of length $4^{-k}$ . Their unnormalized t-cost is $(2\,4^{-t})^k$ , which tends to zero when t exceeds one half. Thus the dimension is at most one half.
Give every level-k interval probability $2^{-k}$ . For an interval I of length r with $4^{-(k+1)}<r\le4^{-k}$ , separation lets I meet at most one level-k interval. Hence its mass is at most $2^{-k}<2\sqrt r$ . Any set of diameter r lies in such an interval.
The mass-distribution bound with constant two gives positive normalized half-dimensional measure. The level covers have bounded half-dimensional cost, so that measure is also finite. Positivity forces the dimension to be at least one half, completing the two-sided argument.

2^k(4^{-k})^{1/2}=1,\qquad 0<\mathcal H^{1/2}(C)<\infty.

A bounded construction cover gives the upper measure bound; the probability estimate supplies the missing lower bound.↔ Scroll wide equations horizontally.

The result

The set has Hausdorff dimension 1/2 and finite positive normalized half-dimensional measure, without an asserted exact critical measure constant.

The living notebook

Explore the model

Interactive lab

Interactive lab · probability

Count binary addresses without losing their geometry

Pause and Reset, then set probability to 0.5 and trials to 4. There are sixteen equally likely binary addresses, although the five bars group them by the number of right branches. On paper, level four has sixteen intervals of length 1/256: total ordinary length 1/16, but unnormalized half-dimensional covering cost one. The instrument counts addresses, not geometric distances or Hausdorff measure. For a fresh comparison Reset and set both controls; control edits preserve elapsed time.

Exact probabilityHighlighted outcomeMean np

Mean · np: 5
Variance · np(1−p): 2.5
Total probability: 1

Success probability p0.5Trials n10

t = 0 s

These are exact binomial probabilities for independent trials with one fixed success probability. Playback tours the distribution; it does not resample outcomes. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.

Look forThe same collection can have shrinking ordinary length and a persistent cost at a fractional measuring exponent.

Pause & reason

Check your understanding

Interactive checks need JavaScript. You can still read the explanation.

Read the checkpoint explanation

Dimension identifies the transition exponent, but does not determine the measure at that exponent. Exact critical measure requires additional geometric estimates and a declared normalization.

Your turn

Try it yourself

Let C be the worked Cantor set and let $E=C\cup[2,5]$ . Find the dimension of E and its normalized one-dimensional measure. How does dilation by nine change $\mathcal H^{1/2}(C)$ ?

Take a moment to make a prediction or work it out on paper. Open a hint when you need one.

Give me a hint

Use the dimension of a finite union and the fact that an exponent larger than a set’s dimension gives zero measure.

Compare with the solution

The interval has dimension one, so the union has dimension one. Since C has dimension one half, its one-dimensional measure is zero; therefore $\mathcal H^1(E)=3$ . Dilation by nine multiplies half-dimensional measure by $9^{1/2}=3$ . This scales the unknown finite positive critical value without supplying its exact numerical size.

Keep the idea

Take it with you

Normalized Hausdorff measure agrees with ordinary volume on integer-dimensional planes.
A covering argument alone gives an upper dimension bound.
Critical Hausdorff measure needs separate information beyond the dimension.

Read further

Original sources and deeper explanations to follow at your own pace.

Optional progress, saved in this browser when available. Revisit any lesson at any time.

Lesson not marked complete.