Lebesgue Integral Definition. Integrating a characteristic of a set of infinite measure

140 Views Asked by At

I'm reading the book Real Analysis from Elias.M.Stein and Rami Shakarchi.

In the chapter about abstract integration, when the book starts to construct the integral, it limits itself to $\sigma$-finite spaces. I've been told that I can define the integral in general space without any problem but I am struggling to get a result that is obvious if one has a $\sigma$-finite space, maybe here is not true.

This book starts defining the integral for simple functions, meaning lineal combinations of characteristic of measurable sets of finite measure, putting $$\int \sum_{1}^nc_i\chi_{E_i} \ d\mu = \sum_1^nc_i\mu(E_i)$$

After that it basically define the integral of a positive function as $$\int f \ d\mu = sup\{\int g, \ 0\leq g\leq f\}$$ with $g$ simple. At this point one can observe that the first step in the definition, i.e, the definition for simple functions does not tell someone the value of $$\int \chi_E \ d\mu$$ when $\mu(E) = \infty$. However, once one has the second defintion, if the space is $\sigma$-finite, the next reasoning proves that the integral above has the value $\infty$ as one may expect; As the space is $\sigma$-finite, $X = \cup_{i=1}^{\infty}E_i$ with $E_i$ of finite measure, then $E = \cup_{i=1}^{\infty}(E\cap E_i)$. Let me call $B_n = \cup_{i=1}^{n}(E\cap E_i)$. Now by the second definition, or by monotony, $$\int \chi_E \ d\mu \geq \int \chi_{B_n} \ d\mu = \mu(B_n)$$ where the last equal follows from definition for simple functions, as $B_n$ has finite measure. To finish if one takes $n \to \infty$, by the continuity of the measure one has $\mu(B_n) \to \mu(E) = \infty$ and then $$\int \chi_E \ d\mu = \infty$$

This seems really clear but I am getting confused to get the same result if one doesn't have a $\sigma$-finite spaces. It is clear that I can't use the same reasoning.

One important observation is that I have seen books where admit sets of infinite measure in the first definition of integral, but this is not the case in this book. (Maybe because they restrict to $\sigma$-finite spaces).