Measure Theory: $\sigma$ -Algebras, Lebesgue Integration, and Probability Foundations

Measure theory is the branch of mathematical analysis that provides the rigorous theoretical foundation for modern integration, real analysis, probability theory, and stochastic calculus. Developed by Henri Lebesgue and Émile Borel, measure theory resolves the fundamental limitations of the Riemann integral, establishing formal mathematical machinery for assigning sizes (lengths, areas, volumes, and probabilities) to complex geometric subsets of Euclidean space.

This guide details $\sigma$ -algebras, Borel sets, the construction of the Lebesgue measure, the Lebesgue integral versus the Riemann integral, convergence theorems (Monotone and Dominated Convergence), and the Radon-Nikodym derivative.

1. Quick-Reference: Riemann vs. Lebesgue Integration

+-----------------------------------------------------------------------------------------+
|                               RIEMANN vs. LEBESGUE INTEGRATION                          |
+-----------------------------------------------------------------------------------------+
| Dimension           | Riemann Integral                     | Lebesgue Integral                  |
+---------------------+--------------------------------------+------------------------------------+
| Domain Partitioning | Partitions the x-axis (Domain)       | Partitions the y-axis (Range)      |
| Measurable Sets     | Continuous / piecewise continuous    | Measurable functions over σ-algebra|
| Characteristic Func | Fails on Dirichlet function 1_ℚ(x)   | ∫ 1_ℚ dμ = 0 (ℚ has measure zero)  |
| Limit Invariance    | Weak limit interchange conditions    | Monotone & Dominated Convergence   |
| Completeness        | L¹ norm space is incomplete          | L¹(μ), L²(μ) spaces are complete   |
|                     | (Needs Cauchy completion)            | Banach and Hilbert spaces          |
+-----------------------------------------------------------------------------------------+

2. $\sigma$ -Algebras and Measurable Spaces

Let $X$ be a non-empty set. A collection $\Sigma$ of subsets of $X$ is called a $\sigma$ -algebra if it satisfies three structural axioms:

Contains the Universe: $X \in \Sigma$ .
Closed under Complement: If $E \in \Sigma$ , then $E^c = X \setminus E \in \Sigma$ .
Closed under Countable Unions: If $E_1, E_2, E_3, \dots \in \Sigma$ , then $\bigcup_{n=1}^\infty E_n \in \Sigma$ .

The pair $(X, \Sigma)$ is called a Measurable Space, and elements $E \in \Sigma$ are called Measurable Sets.

The Borel $\sigma$ -Algebra ( $\mathcal{B}(\mathbb{R})$ )

The Borel $\sigma$ -algebra on $\mathbb{R}^n$ is the smallest $\sigma$ -algebra generated by all open sets in $\mathbb{R}^n$ . It contains all open intervals, closed intervals, half-open intervals, countable unions of closed sets ( $F_\sigma$ sets), and countable intersections of open sets ( $G_\delta$ sets).

3. Construction of the Lebesgue Measure

A Measure $\mu: \Sigma \to [0, \infty]$ is a non-negative function satisfying countable additivity:

\mu\left( \bigcup_{n=1}^\infty E_n \right) = \sum_{n=1}^\infty \mu(E_n) \quad \text{for pairwise disjoint } E_n

Carathéodory Measure Extension Construction:
[ Elementary Intervals (a, b) with Length l(I) = b - a ]
                          |
                          v
[ Lebesgue Outer Measure: m*(E) = inf { ∑ l(I_k) : E ⊆ ⋃ I_k } ]
                          |
                          v (Carathéodory Criterion: m*(A) = m*(A ∩ E) + m*(A ∩ Eᶜ))
[ Lebesgue Measurable Sets ℒ(ℝ) & Complete Lebesgue Measure m ]

The Vitali Non-Measurable Set

Assuming the Axiom of Choice, there exist subsets of $\mathbb{R}$ that cannot be assigned any consistent measure (Vitali Sets). Measure theory restricts integration to the $\sigma$ -algebra of measurable sets $\mathcal{L}(\mathbb{R})$ , preventing mathematical paradoxes like the Banach-Tarski paradox.

4. The Lebesgue Integral & Powerful Convergence Theorems

While the Riemann integral divides the domain $[a, b]$ into vertical strips ( $\Delta x$ ), the Lebesgue integral partitions the range of the function into horizontal slices ( $\Delta y$ ) using Simple Functions $\phi(x) = \sum_{i=1}^k a_i \mathbf{1}_{E_i}(x)$ :

\int_X \phi \, d\mu = \sum_{i=1}^k a_i \mu(E_i)

For any non-negative measurable function $f: X \to [0, \infty]$ :

\int_X f \, d\mu = \sup \left\{ \int_X \phi \, d\mu : 0 \le \phi \le f, \, \phi \text{ is simple} \right\}

+-----------------------------------------------------------------------------------------+
|                               LEBESGUE CONVERGENCE THEOREMS                             |
+-----------------------------------------------------------------------------------------+
| Theorem                        | Condition                             | Conclusion     |
+--------------------------------+---------------------------------------+----------------+
| Monotone Convergence (MCT)     | 0 ≤ f₁ ≤ f₂ ≤ f₃ ... ↑ f pointwise   | lim ∫ f_n = ∫ f|
| Fatou's Lemma                  | Non-negative f_n ≥ 0                  | ∫ (lim inf f_n)|
|                                |                                       | ≤ lim inf ∫ f_n|
| Dominated Convergence (DCT)    | |f_n(x)| ≤ g(x) with ∫ g dμ < ∞       | lim ∫ f_n = ∫ f|
+-----------------------------------------------------------------------------------------+

5. Measure-Theoretic Probability & The Radon-Nikodym Theorem

In modern probability theory (Kolmogorov's axioms), a Probability Space is defined as a normalized measure space $(\Omega, \mathcal{F}, P)$ where $P(\Omega) = 1$ :

$\Omega$ : Sample space of elementary outcomes.
$\mathcal{F}$ : $\sigma$ -algebra of observable events.
$P$ : Probability measure assigning probabilities $P(A) \in [0, 1]$ .
A Random Variable $X: \Omega \to \mathbb{R}$ is simply a measurable function.

The Radon-Nikodym Theorem

If measure $\nu$ is absolutely continuous with respect to $\mu$ ( $\nu \ll \mu$ , meaning $\mu(A) = 0 \implies \nu(A) = 0$ ), there exists an almost-everywhere unique non-negative measurable function $\frac{d\nu}{d\mu} \in L^1(\mu)$ , called the Radon-Nikodym derivative, such that:

\nu(E) = \int_E \frac{d\nu}{d\mu} \, d\mu \quad \forall E \in \Sigma

In quantitative finance and statistics, the Radon-Nikodym derivative represents the likelihood ratio and the change-of-measure density (e.g., transitioning from physical real-world measure $\mathbb{P}$ to risk-neutral pricing measure $\mathbb{Q}$ via Girsanov's theorem).

References

Rudin, W. (1987). Real and Complex Analysis (3rd ed.). McGraw-Hill.
Folland, G. B. (1999). Real Analysis: Modern Techniques and Their Applications (2nd ed.). John Wiley & Sons.
Billingsley, P. (2012). Probability and Measure (Anniversary ed.). John Wiley & Sons.
Halmos, P. R. (1950). Measure Theory. D. Van Nostrand Company.

Measure Theory: \sigma-Algebras, Lebesgue Integration, and Probability Foundations