1
$\begingroup$

Consider the measure space $(\mathbb{N}, \mathcal{P}(\mathbb{N}),\mu)$, where $\mathcal{P}(\mathbb{N})$ is the power set of $\mathbb{N}$ and $\mu$ is the counting measure. Let $f: \mathbb{N} \to [0,\infty]$. How do I prove that $$\int_\mathbb{N} f d\mu = \sum_{n = 1}^\infty f(n)$$ using only the definition of Lebesgue integral of $f$ (as $\sup_{0 \le s \le f}\int_\mathbb{N}sd\mu,$ where $s$ is a simple function)?


There are several similar questions on this website, but they use the monotone convergence theorem or there are not complete answers.

1 Answers 1

1

Here is a sketch of the proof. Also I will assume the followings:

  1. We know well about what the sum $\sum_{n=1}^{\infty} f(n)$ means.

  2. The integral $\int_{\Bbb{N}} s \, d\mu$ is well-defined for simple functions $s$ and is given by $$ \int_{\Bbb{N}} \left( \sum_{j=1}^{m} a_j \mathbf{1}_{E_j} \right) \, d\mu = \sum_{j=1}^{m} a_j \mu(E_j) $$ for constants $a_1, \cdots, a_m$ and subsets $E_1, \cdots, E_m \subseteq \Bbb{N}$.


The key observation is that, for $E \subseteq \Bbb{N}$, the following holds:

$$ \mu(E) = \sum_{n=1}^{\infty} \mathbf{1}_E(n). $$

Proving this identity may depend on how exactly the counting measure $\mu$ is defined in your context. For any sensible definition, however, we may invoke mathematical induction to show that this is true for all $E \subset \{1, \cdots, n\}$ for all $n \in \Bbb{N}$. If $E$ is infinite, it will not be hard to show that both sides are $+\infty$.

From this, if $s = \sum_{j=1}^{m} a_j \mathbf{1}_{E_j}$ is a simple function such that $0 \leq s \leq f$, then

\begin{align*} \int_{\Bbb{N}} s \, d\mu = \sum_{j=1}^{m} a_j \mu(E_j) &= \sum_{j=1}^{m} a_j \sum_{n=1}^{\infty} \mathbf{1}_{E_j}(n) \\ &= \sum_{n=1}^{\infty} \sum_{j=1}^{m} a_j \mathbf{1}_{E_j}(n) = \sum_{n=1}^{\infty} s(n) \leq \sum_{n=1}^{\infty} f(n). \end{align*}

This shows that $\int_{\Bbb{N}} f \, d\mu \leq \sum_{n=1}^{\infty} f(n)$. For the converse direction, notice that

$$ \int_{\Bbb{N}} f \, d\mu \geq \int_{\Bbb{N}} \left( \sum_{j=1}^{m} f(j) \mathbf{1}_{\{j\}} \right) \, d\mu = \sum_{j=1}^{m} f(j). $$

Taking $m \to \infty$, we have $\int_{\Bbb{N}} f \, d\mu \geq \sum_{n=1}^{\infty} f(n)$ and hence the equality is established.

  • 0
    Why is the converse direction true? I do not understand the steps.2017-10-17