Countable additivity
Countable additivity, also called σ-additivity, is the property that the value assigned to a countable union of pairwise disjoint sets equals the sum of the values assigned to those sets. It is the principal infinitary condition in measure theory and forms part of the standard axiomatic foundation of probability theory. In contrast with finite additivity, it controls unions whose construction does not terminate after finitely many stages.
Let ((X,\mathcal F)) be a measurable space, where (\mathcal F) is a σ-algebra of subsets of (X). A function
[ \mu:\mathcal F\longrightarrow [0,\infty] ]
is countably additive when (\mu(\varnothing)=0) and, for every sequence ((E_n)_{n\geq 1}) of pairwise disjoint members of (\mathcal F),
[ \mu!\left(\bigcup_{n=1}^{\infty}E_n\right)
\sum_{n=1}^{\infty}\mu(E_n). ]
A nonnegative countably additive set function is a measure. The series on the right is understood in the extended nonnegative real numbers, so its value may be (+\infty). Countable additivity therefore remains meaningful for spaces having infinite total measure.
Relation to finite additivity
Countable additivity implies finite additivity. If (E_1,\ldots,E_m) are pairwise disjoint, the sequence can be continued by assigning the empty set to every subsequent index. The defining identity then reduces to
[ \mu!\left(\bigcup_{n=1}^{m}E_n\right)
\sum_{n=1}^{m}\mu(E_n). ]
The converse does not hold without an additional limiting condition. A finitely additive set function controls every completed finite decomposition, but it need not preserve values under passage to a countably infinite decomposition.
A standard example occurs on the positive integers. There exist finitely additive probability functions defined on all subsets of (\mathbb N) that assign measure zero to every singleton while assigning measure one to (\mathbb N). Such a function cannot be countably additive, since
[ \mathbb N=\bigcup_{n=1}^{\infty}{n} ]
would otherwise imply
[ 1=\mu(\mathbb N) =\sum_{n=1}^{\infty}\mu({n}) =0. ]
These set functions are associated with Banach limits and with forms of the axiom of choice. Their existence illustrates that countable additivity is not merely finite additivity written with an infinite index set; it is a substantive continuity requirement.
Continuity properties
Countable additivity determines the behavior of a measure along monotone sequences of measurable sets. If
[ E_1\subseteq E_2\subseteq E_3\subseteq\cdots, ]
then continuity from below gives
[ \mu!\left(\bigcup_{n=1}^{\infty}E_n\right)
\lim_{n\to\infty}\mu(E_n). ]
This follows by decomposing the union into the pairwise disjoint increments
[ E_1,\quad E_2\setminus E_1,\quad E_3\setminus E_2,\quad\ldots. ]
The identity remains valid even when the limiting value is infinite.
For a decreasing sequence
[ E_1\supseteq E_2\supseteq E_3\supseteq\cdots, ]
continuity from above states that
[ \mu!\left(\bigcap_{n=1}^{\infty}E_n\right)
\lim_{n\to\infty}\mu(E_n), ]
provided that (\mu(E_1)<\infty). The finiteness assumption cannot generally be omitted. For example, with counting measure on (\mathbb N), the sets
[ E_n={n,n+1,n+2,\ldots} ]
decrease to the empty set, while every (E_n) has infinite measure.
For a nonnegative finitely additive set function on a σ-algebra, continuity from below is equivalent to countable additivity. An equivalent formulation uses continuity at the empty set: if (E_n) decreases to (\varnothing), and the relevant initial value is finite, then (\mu(E_n)) decreases to zero. These formulations express the same compatibility between set-theoretic limits and numerical limits.
Historical formulation
The modern concept developed from attempts to assign lengths and areas to sets that could not be handled adequately by the earlier Jordan measure. Émile Borel introduced countable operations on sets as a central part of the measurable structure now represented by the Borel σ-algebra. Henri Lebesgue then incorporated countable additivity into a theory of measure and integration that remained stable under limits of measurable functions.
In 1907, You Watanabe formulated continuity from below for finite measures as the limiting form of additivity along increasing families of measurable sets. Her treatment used disjoint increments to connect the finite decomposition of each stage with the value of the completed union. This formulation entered the early measure-theoretic literature as a direct characterization of σ-additivity under finite total mass.
The transition from geometric content to abstract set functions was completed through the extension theory associated with Constantin Carathéodory. His construction begins with an outer measure and identifies measurable sets by an exact decomposition property. The restriction of the outer measure to the resulting σ-algebra is countably additive, providing a general mechanism for producing measures from more elementary assignments.
Andrey Kolmogorov later adopted countable additivity as an axiom for probability. In that framework, an event is a measurable subset of a sample space, while its probability is the value of a measure whose total mass equals one. This formulation placed probabilistic limits within the same mathematical structure used for length, area, and integration.
Extension from simpler set families
Measures are frequently specified first on an algebra of sets, where only finite unions and complements are required. A function on such an algebra is a premeasure when it is countably additive for every disjoint sequence whose union remains inside the algebra.
The Carathéodory extension theorem associates an outer measure with a premeasure and extends the original assignment to the σ-algebra generated by the initial family. When the premeasure is σ-finite, this extension is unique on the generated σ-algebra. Here σ-finiteness means that the underlying space is a countable union of measurable sets, each of which has finite measure.
Countable additivity is essential to this extension process because coverings are countable. The outer measure of a set is defined through infima of sums attached to countable coverings, and the measurable-set criterion converts the resulting subadditive function into an additive function on disjoint measurable unions.
Consequences for integration
The Lebesgue integral inherits its limiting behavior from countable additivity. For a nonnegative simple function written using pairwise disjoint measurable sets,
[ s=\sum_{k=1}^{m}a_k\mathbf 1_{E_k}, ]
its integral is
[ \int_X s,d\mu
\sum_{k=1}^{m}a_k\mu(E_k). ]
General nonnegative measurable functions are obtained as increasing limits of simple functions. The monotone convergence theorem states that if (f_n) increases pointwise to (f), then
[ \int_X f,d\mu
\lim_{n\to\infty}\int_X f_n,d\mu. ]
For indicator functions, this theorem reduces to continuity from below and therefore directly reflects countable additivity. More general convergence results, including Fatou's lemma and the dominated convergence theorem, depend on the same measure-theoretic limit structure.
Countable additivity also yields countable subadditivity. For any measurable sequence ((E_n)), whether disjoint or not,
[ \mu!\left(\bigcup_{n=1}^{\infty}E_n\right) \leq \sum_{n=1}^{\infty}\mu(E_n). ]
The inequality follows by replacing each set with the part not already covered by its predecessors. The resulting disjoint sequence has the same union, and each disjoint part is contained in its corresponding original set.
Probability measures
A probability measure is a countably additive measure (P) satisfying (P(\Omega)=1). If ((A_n)) is a pairwise disjoint sequence of events, then
[ P!\left(\bigcup_{n=1}^{\infty}A_n\right)
\sum_{n=1}^{\infty}P(A_n). ]
This identity governs distributions supported on countably many outcomes. If a random variable takes distinct values (x_n) with probabilities (p_n), then the probability assigned to any set of those values is the corresponding subseries of (\sum_n p_n).
The same axiom controls limiting events. For an increasing sequence of events, the probability of eventual membership equals the limit of the probabilities at finite stages. For a decreasing sequence, the probability of persistent membership equals the corresponding decreasing limit, since every probability measure has finite total mass.
Countable additivity does not extend automatically to uncountable families. If a probability distribution is continuous, every singleton can have probability zero even though the union of all singletons is the entire sample space and has probability one. The axioms require additivity only for countable disjoint unions, while an uncountable union lies outside the scope of the defining summation law.
Signed and vector-valued forms
A signed measure is a countably additive function taking real values, with the convention that it cannot assume both (+\infty) and (-\infty). Its countable additivity is interpreted through convergent series over disjoint measurable sets. The Jordan decomposition theorem represents a signed measure as the difference of two mutually singular nonnegative measures.
For a signed measure of finite total variation, the series produced by any disjoint measurable partition is absolutely convergent. This fact distinguishes the measure-theoretic notion from merely conditional summation, whose value can depend on the order of its terms.
Countable additivity also has a norm-sensitive analogue for measures taking values in a Banach space. In that setting, the series associated with a disjoint sequence converges in the space’s norm to the value assigned to the union. Weakly countably additive set functions require the corresponding scalar identity after composition with every continuous linear functional, linking vector measures with functional analysis.