This is not a specific problem, but a general question. Often when we're showing convergence of functions (particularly pointwise) or even of sets in certain cases, a set of the following form appears: $$ \bigcap_{N \in \mathbb{N}} \bigcup_{n \geq N} E_{n} $$ Often we will encounter situations where we want to show that something goes "wrong" on a set of this form, or a set of this form is measurable (which is obviously true if $E_{n}$ are measurable); I have a general question about a set of this form.

In what contexts does this show up? In my understanding, something is in this set if it "happens" infinitely often. This is a slightly probabilistic interpretation, but if some property holds for infinitely many sets, then something like this shows up. Is that a correct interpretation? I'd appreciate some elaboration.

Example: If we have a sequence of functions $\{f_{n}\}$, where each $f_{n}$ is supported on a measurable set $E_{n}$, then

$\bigcap_{N} \bigcup_{n \geq N} E_{n} $

is precisely the set of points where the sequence $f_{n}$ is not eventually $0$. Thus, points $x$ at which arbitrarily many $f_{n}$ are non-zero (points in infinitely many supports $E_{n}$). Is this correct?


What you've written is called $\limsup E_n$ and $x \in \limsup E_n$ iff $x \in E_n$ for infinitely many $n$. This understanding already shows your example is correct.

There is a similar notion that $\liminf E_n = \bigcup_{j \geq 1} \bigcap_{n \geq j} E_n$. $x \in \liminf E_n$ iff there is an $N$ such that $x \in E_n$ whenever $n \geq N$ (i.e. "$x \in E_n$ eventually").

These sets occur naturally when considering convergence as you noted. In fact, $\liminf E_n$ is defined to capture definitions like that of convergence which require some behaviour to occur "eventually". Then $\limsup E_n$ is just the complement of $\liminf E_n$.

For example, in the case of convergence we require that for fixed $\varepsilon > 0$, your sequence lies within distance $\varepsilon$ of the limit "eventually". More precisely, if $E_n^\varepsilon = \{x: |f_n(x) - f(x)| < \varepsilon\}$ then $$\liminf E_n^\varepsilon = \{x: \exists N \text{ such that } \forall n \geq N, |f_n(x) - f(x)| < \varepsilon\}$$ so that we can write $$\{x: f_n(x) \to f(x)\} = \bigcap_{k \geq 1} \liminf E_n^{k^{-1}}.$$

Conversely, if we want to find $x$ such that $f_n(x) \not \to f(x)$ then there is an $\varepsilon > 0$ such that $|f_n(x) - f(x)| > \varepsilon$ infinitely often which is the same as saying $x \in \limsup (E_n^\varepsilon)^c$. Hence we can write $$\{x: f_n(x) \not \to f(x)\} = \bigcup_{k \geq 1} \limsup (E_n^{k^{-1}})^c.$$

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.