The swapping of integration and taking limit of the integrand, like (int lim_{n ightarrow infty} f_n = lim_{n ightarrow infty} int f_n ), is usually taken for granted as a valid operation in engineering courses. However, if we require mathematical rigorousness, such manipulation relies on additional constraints in order to be feasible. This is governed by a set of convergence theorems about measurable functions. By referring to Halsey Royden's "Real Analysis (3th ed., 2004)", this article compiles these theorems to present an overview of their similarities and differences.
Let ({f_n}_{n geq 1}) be a sequence of measurable functions defined on a measurable set (E). The measure of (E) is ({ m m}(E)). Let (g) be integrable on (E) and ({g_n}_{n geq 1}) be a sequence of integrable functions which converges a.e. to (g). The integrals in the following are in the sense of Lebesgue integral.
Theorem |
Requirements on $${f_n}_{n geq 1}$$ | $${ m m}(E)$$ |
Convergence of |
Boundedness of | Swapping of integration and taking limit |
---|---|---|---|---|---|
Bounded convergence theorem |
Measurable | $${ m m}(E)in [0, infty)$$ | $f_n ightarrow f$ on $E$ | $$abs{f_n(x)} leq M$$ | [displaystyle{int_E f = lim_{n ightarrow infty} int_E f_n}] |
Fatou's Lemma |
1. Measurable 2. Nonnegative |
$${ m m}(E) in [0,infty]$$ | $f_n ightarrow f$ a.e. on $E$ | None | [displaystyle{int_E f leq underline limint_E f_n}] |
Monotone convergence theorem |
1. Measurable 2. Nonnegative 3. Increasing |
$${ m m}(E) in [0,infty]$$ | $f_n ightarrow f$ a.e. on $E$ | None | [displaystyle{int_E f = lim_{n ightarrow infty} int_E f_n}] |
Lebesgue convergence theorem |
Measurable | $${ m m}(E) in [0,infty]$$ | $f_n ightarrow f$ a.e. on $E$ | 1. $abs{f_n} leq g$ 2. $int_E g < infty$ |
[displaystyle{int_E f = lim_{n ightarrow infty} int_E f_n}] |
Extended Lebesgue convergence theorem |
Measurable | $${ m m}(E) in [0,infty]$$ | $f_n ightarrow f$ a.e. on $E$ | 1. $abs{f_n} leq g_n$ 2. $int_E g_n < infty$ 3. $g_n ightarrow g$ a.e. on $E$ 4. $int_E g < infty$ 5. $int_E g = lim_{n ightarrow infty} int_E g_n$ |
[displaystyle{int_E f = lim_{n ightarrow infty} int_E f_n}] |