Lebesgue measure is the version of length, area, and volume that survives countable limiting operations. The elementary length of an interval is easy to state, but analysis rarely manipulates only finitely many intervals. Sets appear as countable unions, countable intersections, limits of increasing families, exceptional null sets, and level sets of functions. A useful theory of size must assign values to those sets while keeping the familiar rules for boxes, translations, and decompositions.

The key change from elementary geometry is that size is constructed before integration. Instead of defining integrals by subdividing intervals and then using them to infer areas, measure theory first builds a robust size function on sets. Once measurable sets are available, functions can be measured through their level sets and built from simple functions. This is why Lebesgue measure sits at the beginning of modern real analysis, probability, PDE, and functional analysis.

[example: Why Countable Additivity Is Forced] Let $E=(0,1]\subset \mathbb{R}$. For each $k\in\mathbb{N}$, set \begin{align*} I_k=\left(2^{-k},2^{-k+1}\right]. \end{align*} The intervals are pairwise disjoint, and every $x\in(0,1]$ lies in exactly one of them. Indeed, there is a unique $k$ such that $2^{-k}<x\le 2^{-k+1}$, so \begin{align*} (0,1]=\bigcup_{k=1}^{\infty}I_k. \end{align*} The usual length of $I_k$ is $2^{-k}$. For finite partial unions one obtains \begin{align*} \left|\bigcup_{k=1}^{m}I_k\right| =\sum_{k=1}^{m}2^{-k} =1-2^{-m}. \end{align*} Letting $m\to\infty$ forces \begin{align*} |(0,1]|=\sum_{k=1}^{\infty}|I_k|=\sum_{k=1}^{\infty}2^{-k}=1. \end{align*} Finite additivity only sees the partial unions $(2^{-m},1]$. Countable additivity is the rule that lets the limiting union keep the correct length. [/example]

This example shows why the class of measurable sets cannot be a small geometric list. Once countable additivity is required, the theory must be closed under countable unions and compatible with countable limiting processes. The construction below starts with simple geometric covers and then isolates the sets for which cutting and recombining do not change the measured size.

## Construction from Outer Measure

### Boxes as Measuring Rulers

The construction starts with sets whose volumes are already known. Boxes are rigid enough that their volume is unambiguous, but flexible enough to cover arbitrary subsets of $\mathbb{R}^n$. We need a precise box convention before the covering construction can measure anything.

[definition: Rectangular Box] A rectangular box in $\mathbb{R}^n$ is a set of the form \begin{align*} Q=\prod_{i=1}^n(a_i,b_i), \end{align*} where $a_i,b_i\in\mathbb{R}$ and $a_i\le b_i$ for each $i\in\{1,\ldots,n\}$. [/definition]

A box with one collapsed side should have zero volume, and a genuine box should have the product of its side lengths. Without a stable numerical cost for each box, the covering construction would have no way to compare a coarse cover with a refined one. The volume assignment below supplies that cost and makes countable covers into measurable estimates.

[definition: Volume of a Rectangular Box] Let $\mathcal R_n$ be the collection of rectangular boxes in $\mathbb{R}^n$. The volume of a rectangular box is the function \begin{align*} \operatorname{vol}:\mathcal R_n&\to[0,\infty)\\ \prod_{i=1}^n(a_i,b_i)&\mapsto \prod_{i=1}^n(b_i-a_i). \end{align*} [/definition]

This definition agrees with ordinary length when $n=1$, area when $n=2$, and volume when $n=3$. The important point is not the formula alone, but the fact that the formula gives a numerical cost to each member of a countable cover. The next subsection turns those costs into an outer estimate for arbitrary sets.

### Infimum Over Countable Covers

A single cover of a set usually overestimates its size. If a set can be covered in many different ways, the best outside estimate is obtained by taking the infimum over all countable box covers. This leads to a quantity defined for every subset, including sets that will later turn out not to be measurable.

[definition: Lebesgue Outer Measure] The Lebesgue outer measure is the function \begin{align*} \mathcal{L}^{n,*}:\mathcal{P}(\mathbb{R}^n)&\to[0,\infty]\\ E&\mapsto\inf\left\{\sum_{k=1}^{\infty}\operatorname{vol}(Q_k):E\subset\bigcup_{k=1}^{\infty}Q_k,\ Q_k\in\mathcal R_n\right\}. \end{align*} Here $\mathcal{P}(\mathbb{R}^n)$ denotes the power set of $\mathbb{R}^n$. [/definition]

Outer measure is deliberately generous about its domain. It measures every subset from outside, but it does not promise countable additivity for every subset. The next example shows how the infimum viewpoint already recovers the expected zero size of countable sets.

[example: Countable Sets Have Outer Measure Zero] Let $E=\{x_k:k\in\mathbb{N}\}\subset\mathbb{R}^n$, where $x_k=(x_{k,1},\ldots,x_{k,n})$. Fix $\varepsilon>0$ and define \begin{align*} \ell_k=\left(\varepsilon 2^{-k-1}\right)^{1/n}. \end{align*} For each $k$, cover $x_k$ by the box \begin{align*} Q_k=\prod_{i=1}^n\left(x_{k,i}-\frac{\ell_k}{2},x_{k,i}+\frac{\ell_k}{2}\right). \end{align*} Then $E\subset\bigcup_{k=1}^{\infty}Q_k$ and \begin{align*} \operatorname{vol}(Q_k)=\ell_k^n=\varepsilon 2^{-k-1}. \end{align*} Thus this cover has total volume \begin{align*} \sum_{k=1}^{\infty}\operatorname{vol}(Q_k) =\sum_{k=1}^{\infty}\varepsilon 2^{-k-1} =\frac{\varepsilon}{2}<\varepsilon. \end{align*} Since $\varepsilon>0$ was arbitrary, $\mathcal{L}^{n,*}(E)=0$. [/example]

This computation is not merely a curiosity about countable sets. It shows that dense sets such as $\mathbb{Q}\subset\mathbb{R}$ can have zero outer measure. Measure is therefore not topological largeness; it is a volume notion that can ignore dense but countably thin sets.

### The Cut Test

Outer measure is subadditive by construction, but additivity can fail if arbitrary sets are allowed to cut other sets in pathological ways. The correct measurable sets are the ones whose cuts preserve outer measure exactly. This is the Caratheodory idea: a set is measurable when every test set splits cleanly across it.

[definition: Lebesgue Measurable Set] A set $E\subset\mathbb{R}^n$ is Lebesgue measurable if for every set $A\subset\mathbb{R}^n$, \begin{align*} \mathcal{L}^{n,*}(A)=\mathcal{L}^{n,*}(A\cap E)+\mathcal{L}^{n,*}(A\setminus E). \end{align*} [/definition]