A continuous-looking object can still hide mass, oscillation, or variation in places a weaker notion cannot detect. Absolute continuity asks when smallness in one sense forces smallness in another, especially when ordinary [continuity](/page/Continuity) is too local or too pointwise to control accumulation.

For functions on an interval, continuity says that moving the input a small distance moves the output a small distance. That condition is local at a point. It does not prevent the function from making many tiny changes on many tiny intervals whose total effect is large. Absolute continuity asks for a stronger bargain: if a finite collection of intervals has small total length, then the total oscillation of the function across those intervals is small.

For measures, the same idea appears in a sharper form. If a measure $\mu$ declares a set invisible, does another measure $\nu$ also ignore it? Absolute continuity of $\nu$ with respect to $\mu$ says yes. The slogan is that $\nu$ has no mass outside the measure-theoretic world seen by $\mu$.

The simplest failure is a function that is continuous everywhere, increases from $0$ to $1$, has derivative $0$ almost everywhere, and yet cannot be recovered by integrating its derivative. Such a function violates the usual calculus story.

[example: The Cantor Function Breaks Naive Calculus] Let $C\subsetneq[0,1]$ be the middle-thirds [Cantor set](/page/Cantor%20Set), and let $F:[0,1]\to[0,1]$ be the Cantor function, normalized so that $F(0)=0$ and $F(1)=1$. The function $F$ is continuous, nondecreasing, and constant on each [connected component](/page/Connected%20Component) of $[0,1]\setminus C$. If $x\in[0,1]\setminus C$, then $x$ lies in an open interval component $(\alpha,\beta)$ of $[0,1]\setminus C$, so $F$ is constant on $(\alpha,\beta)$ and therefore \begin{align*} F'(x)=0. \end{align*} Thus $F'=0$ on $[0,1]\setminus C$. Since $\mathcal L^1(C)=0$, this says $F'=0$ for $\mathcal L^1$-a.e. $x\in[0,1]$, and hence \begin{align*} \int_0^1 F'(x)\,d\mathcal L^1(x)=\int_0^1 0\,d\mathcal L^1(x)=0. \end{align*} On the other hand, \begin{align*} F(1)-F(0)=1-0=1. \end{align*} Therefore this continuous, nondecreasing, almost-everywhere differentiable function satisfies \begin{align*} F(1)-F(0)\ne \int_0^1 F'(x)\,d\mathcal L^1(x). \end{align*} The example shows that continuity and almost-everywhere differentiability alone do not justify the formula \begin{align*} F(b)-F(a)=\int_a^b F'(x)\,d\mathcal L^1(x). \end{align*} [/example]

The Cantor function isolates the missing condition. We need a form of continuity that prevents a function from hiding total variation on a set of length zero. That condition is absolute continuity.

Before the formal definition, we fix the basic measure notation used throughout the page. On the real line, $\mathcal B(I)$ denotes the Borel $\sigma$-algebra of an interval $I\subset\mathbb R$, and $\mathcal L^1$ denotes one-dimensional [Lebesgue measure](/page/Lebesgue%20Measure). Thus $L^1(I,\mathcal B(I),\mathcal L^1)$ is the space of integrable functions on $I$ with respect to Lebesgue measure; when no ambiguity is possible, this is abbreviated to $L^1(I)$.

## Definition

The pointwise definition of continuity controls one input at a time. To recover integration from differentiation, we need to control many disjoint input intervals at once, because a derivative is integrated over sets rather than inspected at single points. In the measure-theoretic language around this page, the core domination idea is even sharper: null sets for the reference measure must also be null for the controlled measure.

[definition: Absolute Continuity] For a complex measure $\nu:\mathcal E\to\mathbb C$ and a positive measure $\mu:\mathcal E\to[0,\infty]$ on a measurable space $(E,\mathcal E)$, absolute continuity means that $\mu(A)=0$ implies $\nu(A)=0$ for every $A\in\mathcal E$. [/definition]

This primary definition foregrounds the measure-theoretic meaning used throughout the page: $\nu$ is controlled by $\mu$ precisely when $\mu$-null sets cannot carry $\nu$-mass. The interval version below is the one-dimensional calculus analogue of the same domination idea.

## Absolute Continuity of Functions

The interval definition looks technical until it is compared with ordinary continuity. Ordinary continuity can be checked near each point; absolute continuity monitors the total effect of many separated changes. The distinction matters because integration is additive across disjoint sets.

The interval form is the one that repairs the [fundamental theorem of calculus](/theorems/632). It asks for uniform control over every finite family of disjoint intervals, not just one neighbourhood of one point.

[definition: Absolutely Continuous Function on an Interval] Let $I\subset\mathbb R$ be an interval, and let $f:I\to\mathbb C$ be a function. The function $f$ is absolutely continuous on $I$ if for every $\varepsilon>0$ there exists $\delta>0$ such that for every finite family of pairwise disjoint intervals with endpoints $a_k,b_k\in I$ and $a_k<b_k$, $1\le k\le n$, \begin{align*} \sum_{k=1}^n (b_k-a_k)<\delta \quad\implies\quad \sum_{k=1}^n |f(b_k)-f(a_k)|<\varepsilon. \end{align*} [/definition]

This is a uniform condition over all finite interval families. The number of intervals is not fixed, so the definition rules out infinitely distributed oscillation after passing through finite approximations. It also explains why absolute continuity is stronger than [uniform continuity](/page/Uniform%20Continuity): a single interval is only the special case $n=1$. On an unbounded interval, this is the global version of absolute continuity on the whole interval; local absolute continuity would only require the same condition on each compact subinterval.

### Integral Representation

A first sanity check is that functions recovered by integrating an $L^1$ function should be absolutely continuous. Otherwise the definition would be too strong for the main objects produced by Lebesgue integration.

[quotetheorem:10120]

The statement says that indefinite Lebesgue integrals have exactly the kind of distributed continuity required by the definition. The next question is whether every absolutely [continuous function](/page/Continuous%20Function) is produced this way, because that is what would restore the fundamental theorem of calculus.