1 Measurable Spaces
1.1 Definitions of measurable space
A measurable space is a pair \((X,\Sigma)\) where \(X\) is a set and \(\Sigma\) is a collection of subsets of \(X\) called a sigma-algebra. The sigma-algebra specifies which subsets of \(X\) are regarded as measurable, and therefore which set operations are permitted without leaving the measurable universe.
This framework replaces geometric notions like length or area with an abstract “measurability” structure that can be attached to many kinds of sets.
1.2 Sigma-algebras and their role
A sigma-algebra \(\Sigma\subseteq \mathcal P(X)\) is required to satisfy:
- \(X\in \Sigma\).
- If \(A\in\Sigma\), then \(X\setminus A\in\Sigma\).
- If \(A_1,A_2,\dots \in\Sigma\), then \(\bigcup_{n=1}^\infty A_n\in\Sigma\).
These axioms encode stability under complements and countable unions, which is essential because many measure-theoretic constructions rely on countable processes.
1.3 Examples of measurable spaces
Common examples include:
- Trivial sigma-algebra \(\Sigma=\{\varnothing,X\}\).
- Power set sigma-algebra \(\Sigma=\mathcal P(X)\), where every subset is measurable.
- Borel sigma-algebra on \(\mathbb R\), generated by open sets (often denoted \(\mathcal B(\mathbb R)\)).
- Product sigma-algebras formed from measurable structures on factors, used to define measurability on higher-dimensional spaces.
These examples illustrate how the same definition can yield very different measurable structures depending on which sets are declared measurable.
2 Measurable Functions
2.1 Definition via preimages
Let \((X,\Sigma)\) and \((Y,\mathcal T)\) be measurable spaces. A function \(f:X\to Y\) is measurable (from \(\Sigma\) to \(\mathcal T\)) if for every set \(B\in\mathcal T\), the preimage \[ f^{-1}(B)=\{x\in X:\ f(x)\in B\} \] belongs to \(\Sigma\).
Intuitively, measurability means that the function does not create “non-measurable surprises”: whenever you test measurability in the codomain, the corresponding region in the domain remains measurable.
2.2 Measurability on real-valued functions
If \(Y=\mathbb R\) with \(\mathcal T=\mathcal B(\mathbb R)\), then a real-valued function \(f:X\to\mathbb R\) is measurable precisely when preimages of Borel sets are measurable in \(X\). In practice, one often checks preimages of simpler generating families such as intervals; measurability is built to be compatible with these.
2.3 Measurable mappings between general spaces
The definition extends directly to any codomain measurable structure \(\mathcal T\). If \(\mathcal T\) is complicated, measurability can still be established by verifying that preimages of a generating class of \(\mathcal T\) are measurable.
This flexibility is one reason the concept is central: it adapts to many target spaces without changing the underlying logic.
2.4 Equivalent characterizations
2.4.1 Checking measurability with generating sets
A standard strategy is: if \(\mathcal G\subseteq\mathcal T\) is a collection of sets whose sigma-algebra closure equals \(\mathcal T\) (i.e., \(\sigma(\mathcal G)=\mathcal T\)), then it is enough to verify that \(f^{-1}(G)\in\Sigma\) for each \(G\in\mathcal G\).
This reduces potentially infinite verification over \(\mathcal T\) to a more manageable test set, provided the chosen family generates \(\mathcal T\).
3 Relationship to Continuity
3.1 Continuous functions as measurable functions
If \(f:X\to Y\) is continuous between topological spaces and \(\Sigma,\mathcal T\) are the corresponding Borel sigma-algebras, then \(f\) is measurable. The reason is that preimages of open sets (which generate the Borel sigma-algebra) are open, hence Borel measurable.
Thus continuity is a sufficient condition for measurability in the usual real or metric settings.
3.2 Borel measurability
A function \(f:X\to\mathbb R\) is Borel measurable when preimages of Borel sets in \(\mathbb R\) are measurable in \(X\) with respect to a given sigma-algebra on \(X\). In standard analysis, one frequently takes \(X\) itself to be a topological space with its Borel sigma-algebra.
Borel measurability is commonly used in integration theory and in describing random variables, where distributions are defined through measurable sets.
3.3 Discontinuity structure and measurability
Measurable functions may have complicated discontinuity sets. For example, a function can be discontinuous at many points while still remaining measurable, because measurability depends on preimages of sets, not on pointwise behavior alone.
In real analysis, discontinuities of measurable functions can be organized using set-theoretic tools (such as expressing certain inequalities via measurable sets), but they are not necessarily constrained to be “small” in the topological sense.
3.4 Separating measurability from stronger notions
Measurability sits between continuity and arbitrary set-theoretic functions. It is weaker than continuity (since discontinuous measurable functions exist) but structured enough to support integration and probabilistic definitions.
Stronger properties—such as being continuous, uniformly continuous, or having certain regularity—impose additional restrictions that measurability alone does not guarantee.
4 Measurability of Constructions
4.1 Sums, differences, and scalar multiples
If \(f,g:X\to\mathbb R\) are measurable and \(a\in\mathbb R\), then:
- \(f+g\) is measurable,
- \(f-g\) is measurable,
- \(af\) is measurable.
These results follow from the fact that operations can be expressed through preimages of Borel sets using the sigma-algebra properties. The class of measurable functions is thus an algebra over scalars.
4.2 Products and quotients (where defined)
Products of measurable functions are measurable. Quotients require care: if \(g\) is measurable, then \(f/g\) is measurable on the set where \(g\neq 0\). One typically defines the quotient everywhere by setting it to a fixed value on \(\{g=0\}\), ensuring measurability on all of \(X\).
This “define on the problematic set in a controlled way” approach is common in measure theory.
4.3 Composition of measurable functions
If \(f:X\to Y\) is measurable and \(h:Y\to Z\) is measurable, then the composition \(h\circ f:X\to Z\) is measurable. This reflects the stability of measurability under function composition: preimages under the composition can be rewritten as nested preimages, both measurable by assumption.
4.4 Limits of measurable functions
4.4.1 Pointwise limits and almost-sure variants
If \(f_n:X\to\mathbb R\) are measurable and \(f_n(x)\to f(x)\) pointwise, then \(f\) is measurable. The proof uses that sets of the form \(\{x:\ f(x)<a\}\) can be approximated using countable unions/intersections involving \(\{x:\ f_n(x)<a\}\).
In probability theory, “almost-sure” variants appear: if two functions agree except on a null set, they are treated as equivalent for expectations and many convergence theorems.
4.5 Indicator and simple functions
Given a measurable set \(A\in\Sigma\), the indicator function \(\mathbf 1_A\) defined by \(\mathbf 1_A(x)=1\) if \(x\in A\) and \(0\) otherwise is measurable. More generally, simple functions are finite linear combinations of indicators of measurable sets; they are measurable by construction.
These building blocks are central for constructing integrals first on simple functions and then extending by approximation.
5 Special Classes of Measurable Functions
5.1 Simple functions
A function \(s:X\to\mathbb R\) (or \(\overline{\mathbb R}\)) is simple if it takes only finitely many values and each level set is measurable. Equivalently, \(s=\sum_{k=1}^m a_k \mathbf 1_{A_k}\) for measurable \(A_k\).
Simple functions provide the starting point for defining the integral and for proving approximation results.
5.2 Step functions
On ordered domains such as \(\mathbb R\), step functions are simple functions whose measurable sets align with intervals or half-lines, such as \((a,b]\). They model functions that are constant on regions and change value only at prescribed cut points.
Step functions are especially common in elementary integration and in proofs of measurability for piecewise-defined maps.
5.3 Monotone and continuous-from-above/below sequences
If \(f_n\uparrow f\) pointwise (i.e., \(f_n(x)\) increases to \(f(x)\)) and each \(f_n\) is measurable, then \(f\) is measurable; similarly for \(f_n\downarrow f\). These monotone limits are fundamental because many integral convergence theorems use them directly.
5.4 Lower semicontinuous and upper semicontinuous functions
A real-valued function \(u:X\to\mathbb R\) is lower semicontinuous if for every \(a\), the set \(\{x:\ u(x)>a\}\) is open (in topological contexts). Lower semicontinuity implies Borel measurability: the relevant preimages can be written in terms of open sets.
Upper semicontinuity yields an analogous measurability statement using sets \(\{x:\ u(x)<a\}\).
5.5 Bounded vs unbounded measurability
Measurability is independent of boundedness: measurable functions may be bounded or unbounded. When using the extended real line \(\overline{\mathbb R}\), one treats values \(+\infty\) and \(-\infty\) carefully, but the measurability definition still relies on preimages of Borel sets (now in \(\overline{\mathbb R}\)).
Thus the measurable structure supports both finite and infinite-valued functions.
6 Measurability in Probability Theory
6.1 Random variables as measurable functions
In probability theory, a random variable is typically modeled as a measurable function \(X\to\mathbb R\) (or another measurable space) defined on a probability space \((\Omega,\mathcal F,\mathbb P)\). Measurability ensures that events determined by the random variable correspond to measurable sets in \(\mathcal F\).
This is what makes probabilities of statements like “\(X\le a\)” well-defined.
6.2 Distribution functions and measurability
For a real-valued random variable \(X\), the distribution function \(F(t)=\mathbb P(X\le t)\) is defined using measurability of the event \(\{X\le t\}\). Measurability guarantees that these sets lie in \(\mathcal F\), so \(\mathbb P(X\le t)\) is meaningful.
6.3 Events defined through random variables
Many probability statements are expressed through measurable transformations of random variables. Examples include:
- \(\{X\in B\}\) for Borel sets \(B\),
- \(\{X>Y\}\) for random variables \(X,Y\),
- events of the form \(\{\phi(X)\in C\}\).
In each case, measurability ensures that the event can be assigned a probability.
6.4 Measurable transforms of random variables
If \(X\) is a random variable and \(\phi\) is measurable, then \(\phi(X)\) is again a random variable. This is a direct consequence of measurability being preserved under composition.
It formalizes the idea that one can safely process data with measurable operations—computing functions of outcomes—without leaving the probabilistic framework.
7 Measurable Sets and Preimages
7.1 Preimage of sigma-algebras
Given \(f:X\to Y\) and a sigma-algebra \(\mathcal T\) on \(Y\), the collection \[ f^{-1}(\mathcal T)=\{f^{-1}(B):\ B\in\mathcal T\} \] is a sigma-algebra on \(X\). This sigma-algebra represents exactly which subsets of \(X\) become measurable when one only observes the mapping \(f\).
This viewpoint clarifies why measurability is defined via preimages: it is about transporting measurable structure backward along a function.
7.2 Pullback sigma-algebras
The sigma-algebra \(f^{-1}(\mathcal T)\) is often called the pullback sigma-algebra. It is the largest sigma-algebra on \(X\) that makes \(f\) measurable as a map into \((Y,\mathcal T)\).
In other words, measurability can be reframed as: the domain sigma-algebra must contain the pullback of the codomain sigma-algebra.
7.3 Image measure intuition (without advanced prerequisites)
When a probability measure is placed on \(X\), one can sometimes “push it forward” to define a measure on \(Y\) via a measurable map \(f\). Informally, the probability assigned to a set \(B\subseteq Y\) is taken to be the probability of all points in \(X\) that map into \(B\), i.e., \(\mathbb P(f\in B)=\mathbb P(f^{-1}(B))\).
Although full construction details belong to measure theory’s deeper machinery, the driving principle is again preimages.
7.4 Graph measurability
A function \(f:X\to Y\) can be studied through its graph \(\{(x,f(x)):\ x\in X\}\). Under appropriate measurable structures on \(X\times Y\), measurability of \(f\) is closely tied to measurability properties of its graph.
This provides an alternate geometric interpretation: measurability governs how the graph sits inside the product measurable space.
8 Tools and Theorems
8.1 Monotone class ideas (high-level)
Monotone class reasoning is a technique for proving that certain properties hold for all sets in a sigma-algebra by showing the property is preserved under increasing limits and complements in a controlled way. In practice, it allows one to extend measurability checks from a generating collection to the whole sigma-algebra.
The approach is especially useful when direct verification for every measurable set is difficult.
8.2 Dynkin systems and pi-lambda style reasoning
Dynkin systems (also called \(\lambda\)-systems) and related \(\pi\)-\(\lambda\) arguments provide another method to upgrade statements from a generating class to the generated sigma-algebra. These tools are central in proofs that certain families of sets coincide or that measurability can be verified on smaller collections.
They formalize the idea that sigma-algebras are determined by their behavior on well-chosen test sets.
8.3 Closure properties summary
Measurable functions form a robust class:
- they are closed under algebraic operations that can be expressed via Borel-measurable transformations,
- they are closed under pointwise limits,
- they are closed under composition.
These closure properties underpin many theorems in integration and probability, where one repeatedly builds new functions from old ones.
8.4 Common measurability pitfalls
Typical mistakes include:
- assuming that pointwise convergence alone gives measurability without having measurability of the approximating sequence,
- checking continuity at points rather than verifying preimages of measurable sets,
- ignoring the need to define quotients carefully on the zero set of the denominator,
- confusing “measurable with respect to \(\Sigma\)” with “measurable with respect to a finer or coarser sigma-algebra” without checking inclusion.
Because measurability depends on the sigma-algebra structure, these details matter.
9 Extensions and Advanced Views
9.1 Measurable functions on product spaces
To define measurability on \(X\times Y\), one equips the product with a suitable sigma-algebra, such as a product sigma-algebra generated from the factors. Measurability of maps from the product is then interpreted relative to that structure.
This setting is essential for multivariate integration, where measurability ensures that regions in the plane (or higher dimensions) behave properly under maps.
9.2 Carathéodory measurability (outline-level)
In more advanced measure constructions, Carathéodory measurability describes a compatibility between a function and an outer measure. Conceptually, it identifies which sets behave well with respect to cutting and measuring along the function’s level structure.
While the detailed development uses outer measures and coverings, the theme remains the same: measurability is a condition ensuring “well-defined measure behavior.”
9.3 Measurability with respect to sub-sigma-algebras
If \(\mathcal G\subseteq \Sigma\) is a sub-sigma-algebra, one can ask whether a function \(f:(X,\Sigma)\to Y\) is measurable with respect to \(\mathcal G\), meaning preimages of sets in \(\mathcal T\) lie in \(\mathcal G\). This captures situations where only partial information about the domain is available.
In probability terms, it corresponds to measurability with respect to a filtration or information structure.
9.4 Outer-measure viewpoint (conceptual)
An outer-measure perspective treats “measure” initially as defined on all subsets rather than only measurable ones. Measurable sets are then characterized as those that satisfy a consistency property with respect to that outer measure, ensuring additive behavior on the measurable collection.
Conceptually, measurable functions align with this viewpoint because measurability governs which preimages should qualify as measurable sets under the outer-measure-to-measure refinement process.