1 Definition and basic properties
A sigma-algebra is a collection of subsets of a given set that is closed under the operations needed to describe observable events in measure theory and probability. It provides the basic framework for deciding which sets can be assigned measures, probabilities, or integrals in a consistent way. The notion is especially important because it is stable under countable operations, which makes it well suited to limits, approximations, and infinite processes.
1.1 Set-system formulation
Let X be a set, often called the underlying space or sample space. A sigma-algebra on X is a family of subsets of X. Its members are the sets regarded as measurable. The family must contain the empty set and, as a consequence of the axioms, also the whole set X. In practice, a sigma-algebra identifies which collections of outcomes or points can be handled by a measure.
1.2 Closure under complementation
If A is a member of a sigma-algebra on X, then its complement X \ A must also belong to the same sigma-algebra. This means that whenever an event is measurable, the event that it does not occur is measurable as well. Complementation is one of the key features that makes the structure useful for probability, where both an event and its negation should be describable.
1.3 Closure under countable unions
If A1, A2, A3, and so on are all in a sigma-algebra, then their union over countably many sets is also in the sigma-algebra. This closure property is essential for studying sequences of events and limiting behavior. It allows one to form events such as “at least one of these occurs” even when the list is infinite but countable.
1.4 Closure under countable intersections
Countable intersections are included as a consequence of closure under complements and countable unions. If a sigma-algebra contains each set in a countable family, then it also contains the intersection of that family. This makes it possible to express events such as “all of these occur” for countably many measurable sets. The interplay between unions, intersections, and complements gives sigma-algebras their flexibility.
2 Examples of sigma-algebras
Sigma-algebras can range from very small families to highly intricate ones. Some are determined by simple set-theoretic constraints, while others arise from topology or from the structure of a probability model. Examples are useful because they show how the abstract definition appears in concrete settings.
2.1 Trivial and discrete sigma-algebras
The smallest possible sigma-algebra on a set X is the trivial sigma-algebra, consisting only of the empty set and X itself. At the opposite extreme, the discrete sigma-algebra is the collection of all subsets of X, though this notion is more commonly identified with the power set in finite or unrestricted contexts. These two cases represent the simplest and most inclusive possibilities.
2.2 Power set sigma-algebra
The power set of X is always a sigma-algebra, because every subset is automatically closed under complements and countable unions. On finite sets, the power set often serves as the standard sigma-algebra. In countably small or discrete settings, it provides the broadest measurable structure, allowing every subset to be treated as measurable.
2.3 Borel sigma-algebra
On a topological space, especially on the real line, the Borel sigma-algebra is generated by the open sets. It contains all open and closed sets, along with many other sets obtained from them through countable operations. In analysis and probability, the Borel sigma-algebra is a central example because it connects topology with measure.
2.4 Generated sigma-algebras
Given a collection of sets, one can form the smallest sigma-algebra containing them. This generated sigma-algebra consists of everything that must be included once the starting sets are closed under the sigma-algebra axioms. It is often used to describe the information carried by a family of events or by a particular function.
3 Construction methods
Sigma-algebras are usually built from simpler data. Rather than listing every measurable set directly, one specifies a generating family and then closes it under the required operations. This approach is economical and reveals how the measurable structure depends on the original input.
3.1 Sigma-algebra generated by a collection of sets
A collection of sets can serve as a starting point for constructing a sigma-algebra. The generated object is obtained by repeatedly applying complementation, countable unions, and the consequences of those operations. The result is uniquely determined as the smallest sigma-algebra containing the original family.
3.1.1 Smallest sigma-algebra containing a set family
The smallest sigma-algebra containing a given family of sets is the intersection of all sigma-algebras that contain that family. This construction guarantees minimality: nothing is included unless it is forced by the axioms. It is widely used in measure theory because it produces the exact measurable structure required by a problem.
3.1.2 Generating sets and closure operations
A generating family may be simple, such as intervals on the real line, or more abstract, such as level sets of a function. Once the family is chosen, closure operations create new measurable sets in stages. This process can be visualized as expanding the initial collection until no further sets are required by the sigma-algebra rules.
3.2 Sigma-algebra generated by a partition
A partition of X divides the space into disjoint pieces whose union is all of X. The sigma-algebra generated by the partition consists of all unions of those pieces. This is a particularly transparent construction, since each measurable set is built by choosing which blocks of the partition to include. It appears in simplified probability models and in coarse descriptions of information.
3.3 Product sigma-algebras
When dealing with product spaces, one often forms a product sigma-algebra from sigma-algebras on the factor spaces. It is generated by measurable rectangles, that is, products of measurable sets from each component space. Product sigma-algebras are fundamental in multivariable probability, joint distributions, and the study of random vectors.
4 Sigma-algebras in probability theory
In probability, sigma-algebras determine which events can be assigned probabilities. They formalize the intuitive idea that a model only observes certain subsets of outcomes. By separating the sample space from the measurable structure, probability theory can handle both finite experiments and more complex infinite ones.
4.1 Sample spaces and event spaces
A sample space lists all possible outcomes of an experiment, while a sigma-algebra selects the events that are considered measurable. Not every subset of a sample space must be an event, especially in infinite settings. The event space is therefore the measurable layer on top of the raw outcome set.
4.2 Measurable events
An event is measurable if it belongs to the chosen sigma-algebra. Measurable events are the ones for which probability can be defined in the model. This restriction is not a loss of generality in most applications; instead, it ensures that the collection of events behaves well under logical combinations and limits.
4.3 Probability measures on sigma-algebras
A probability measure assigns numbers between 0 and 1 to the sets in a sigma-algebra, with total mass 1 on the whole space. It is countably additive, meaning that disjoint measurable sets have probabilities that add correctly even for countably many of them. The sigma-algebra is what makes such a measure possible without ambiguity or contradiction.
4.4 Completion of a probability space
A probability space may be completed by adding all subsets of null sets to the sigma-algebra. This ensures that any set contained in a set of probability zero is also measurable. Completion is useful because it removes technical exceptions and makes the space more robust for analysis and stochastic modeling.
5 Measurable functions and random variables
Sigma-algebras are not only about sets; they also define when a function is measurable. In probability, this is the bridge between abstract events and numerical quantities. Random variables, distributions, and expectation all depend on this structure.
5.1 Definition of measurability
A function between two measurable spaces is measurable if the preimage of every measurable set in the target is measurable in the source. This definition is designed so that measurable structure is preserved under the function. It ensures that events described by the function can be translated back into the original space.
5.2 Preimages of measurable sets
The preimage operation is central because it allows one to test measurability by examining sets in the target space. If intervals, open sets, or other generating sets have measurable preimages, then measurability often follows. This makes the concept practical for verifying whether a given map is compatible with the sigma-algebras involved.
5.3 Random variables as measurable maps
In probability theory, a random variable is a measurable function from a probability space to the real numbers or another measurable space. This formalizes the idea that a random variable assigns a numerical value to each outcome in a way that respects the event structure. Once measurability is established, one can define its distribution and compute its expectation.
5.4 Distribution functions and induced measures
A random variable induces a measure on the target space by transporting the probability measure through the function. For real-valued random variables, this leads to distribution functions and probability laws. The induced measure summarizes how likely different numerical outcomes are, independently of the original sample space.
6 Relations to other structures
Sigma-algebras are closely related to several other algebraic and geometric constructions. They extend set algebras, interact with topology through Borel sets, and organize information over time in probability. These relationships help explain why sigma-algebras are so widely used.
6.1 Fields and algebras of sets
A field or algebra of sets is closed under finite unions and complements, but not necessarily under countable unions. A sigma-algebra strengthens this by demanding countable closure. Thus every sigma-algebra is an algebra of sets, but not every algebra of sets is a sigma-algebra.
6.2 Topological spaces and Borel sets
On a topological space, open sets generate the Borel sigma-algebra. This construction connects the continuous structure of the space with measurable structure. Many standard spaces in analysis use Borel sigma-algebras as the default measurable families.
6.3 Measurable spaces
A measurable space is a pair consisting of a set and a sigma-algebra on it. This abstract notion isolates the measurable structure from any particular measure. It serves as the natural domain and codomain for measurable functions and provides the setting for measure theory itself.
6.4 Filtrations and increasing families of sigma-algebras
A filtration is an increasing family of sigma-algebras indexed by time or another parameter. It models the growth of information in a stochastic process. At each stage, the sigma-algebra represents what is known up to that point, making filtrations fundamental in martingale theory and dynamic probability.
7 Common examples and applications
Sigma-algebras appear throughout probability, analysis, and stochastic modeling. Their role is often invisible in elementary computations, but they are what make the underlying theory coherent. Applications typically use them to formalize what can be observed, integrated, or conditioned upon.
7.1 Event modeling in experiments
In a random experiment, a sigma-algebra specifies which outcomes or combinations of outcomes count as events. For a finite experiment, this may simply be the power set. For infinite experiments, a smaller measurable family is often necessary to avoid pathological subsets and to keep probability assignments well defined.
7.2 Integration and expectation
Measure-theoretic integration is built on sigma-algebras, because integrable functions must be measurable. Expectation is defined as an integral with respect to a probability measure on a sigma-algebra. This framework allows one to handle discrete, continuous, and mixed distributions within the same theory.
7.3 Conditional probability and conditional expectation
Conditional probability and conditional expectation are defined relative to a sigma-algebra that represents available information. Rather than conditioning on a single event, one may condition on a whole collection of events. This gives a powerful and flexible way to describe partial knowledge and updated predictions.
7.4 Stochastic processes and information flow
In stochastic processes, sigma-algebras track how information evolves over time. Each time point or interval may correspond to a different level of observability. This structure is essential for defining adapted processes, stopping times, and many results in modern probability theory.