勒贝格单调收敛定理：函数有界性与≤∞条件的矛盾疑问

Great question—this is a super common source of confusion when first diving into measure theory, so let's break down why these two ideas don't conflict at all.

First, clarify the theorem's context and definitions #

The Lebesgue Monotone Convergence Theorem (LMCT) applies to nonnegative measurable functions, and in measure theory, measurable functions are permitted to take values in the extended real numbers ($\mathbb{R} \cup {+\infty}$). The condition 0 ≤ f_k(x) ≤ f_{k+1}(x) ≤ ∞ is a pointwise inequality using these extended reals, which is completely valid under the formal definition of measurable functions.

What's the "boundedness requirement" people are referencing? #

When discussions mention "boundedness," it's almost always one of two scenarios, neither of which clashes with LMCT's conditions:

• Mix-up with the Dominated Convergence Theorem (DCT): The DCT does require an integrable dominating function (which is often bounded almost everywhere). But LMCT has no such requirement—its core rules are monotonicity and nonnegativity, not boundedness.
• Context-specific constraint for finite integrals: Sometimes people talk about "boundedness" when focusing on cases where the limit function $f$ has a finite integral. But this is an outcome we might want, not a prerequisite for the theorem. LMCT works perfectly even if the limit function is infinite on some set (or even everywhere)—the integral just evaluates to infinity, which is a valid result in measure theory.

A concrete example to make this tangible #

Take the Lebesgue measure on $\mathbb{R}$, and define the sequence:

f_k(x) = \mathbf{1}_{[0,k]}(x) \cdot x

Here, $\mathbf{1}{[0,k]}(x)$ is the indicator function that equals 1 on $[0,k]$ and 0 elsewhere. This sequence is pointwise nondecreasing: $f_k(x) \leq f{k+1}(x)$ for all $x$, and as $k \to \infty$, $f_k(x)$ approaches $f(x) = x$ for all $x \geq 0$ (which tends to $\infty$ as $x$ grows).

LMCT guarantees that:

\lim_{k \to \infty} \int_{\mathbb{R}} f_k(x) dx = \int_{\mathbb{R}} f(x) dx

Calculating the left-hand side: $\int_0^k x dx = \frac{k^2}{2}$, which tends to $\infty$. The right-hand side is $\int_0^\infty x dx = \infty$, so the theorem holds perfectly. Even though $f(x)$ isn't pointwise bounded, there's no conflict—LMCT doesn't demand boundedness.

Key takeaway #

The LMCT's allowance for functions to take $\infty$ is a standard part of measure theory's extended real number framework, and it doesn't contradict any "boundedness requirements" you might encounter. Those requirements either belong to a different theorem (like DCT) or are context-specific constraints for obtaining a finite integral, not a mandatory condition for LMCT itself.

内容的提问来源于stack exchange，提问作者Susan_Math123