Show that $L^{\infty}$ is a Banach Space with respect to the norm $||.||_{\infty}$ where $||f||_{\infty}=\inf\{ a\ge 0: \mu\left(\{x: |f(x)| \gt a\}\right)=0\}$

I am going to prove the following lemmas and then use them in the proof.

Lemma-1: $||.||_{\infty}$ is a norm on $L^{\infty}$

Proof: Suppose that $||f||_{\infty}=0$. Let $A=\{x \in X : f(x) \ne 0\} $ and $A_n=\{x \in X: |f(x)| \gt \frac{1}{n}\}$ . Then $A=\cup_{n}A_n$. Then there exists $'a'$ such that $\mu\left(\{x: |f(x)| \gt a\}\right)=0$ and $a \lt \frac{1}{n}$ (by the definition of $||f||_{\infty}$). Thus $$\mu\left(\{x: |f(x)| \gt \frac{1}{n}\}\right) \le \mu\left(\{x: |f(x)| \gt a\}\right)=0$$

$\implies \mu(A_n)=0$ and hence $\mu(A)=0$. Thus $f =0 ,\mu$-a.e.

Now $$|f(x)+g(x)| \le |f(x)| +|g(x)| \le ||f||_{\infty}|+||g||_{\infty} ,\mu-\text{a.e}$$

Thus $||f+g||_{\infty} \le ||f||_{\infty}+||g||_{\infty}$

Now $$||\lambda f||_{\infty}=\inf\{a \ge 0: \mu\left(\{x: |\lambda f(x)| \gt a\}\right)=0\}$$ $$=\inf\{|\lambda|a \ge 0:\mu\left(\{x: |\lambda||f(x)| \gt |\lambda|a\}\right)=0\}$$ $$=|\lambda|\inf\{a \ge 0:\mu\left(\{x: |f(x)| \gt a\}\right)=0\}=|\lambda|||f||_{\infty}$$

Lemma-2: $||f_n-f||_{\infty} \to 0$ iff there exists $E \in \mathcal{M}$ such that $\mu(E^c)=0$ and $f_n \to f$ uniformly on $E$.

Proof: $(\implies)$ For every $k \in \mathbb{N}$, there exists $n_0(k) \in \mathbb{N}$ such that for all $n \ge n_0(k)$, $||f_n-f||_{\infty} \lt \dfrac{1}{k}$. Then for each $n \ge n_0(k)$ there exists $E_n^{k}$ such that $\mu(E_n^k)=0$ and $|f_n(x)-f(x)| \lt \dfrac{1}{k}$ for all $x \in (E_n^k)^{c}$. Let $$E^k=\cup_{n \ge n_0(k)} E_n^k$$. Then $\mu(E^k)=0$ and for all $x \not \in E^k, |f_n(x)-f(x)| \lt \frac{1}{k}, \forall n \ge n_0(k)$. Now Let $E=\cup_{k=1}^{\infty} E^k$. Then $\mu(E)=0$. Let $\epsilon \gt 0$. Then there exists a $k_0 \in \mathbb{N}$ such that $\frac{1}{k} \lt \epsilon$ for all $k \ge k_0$. Let $x \in E^c$. In Particular $x \not \in E^{k_0}$ . Then for all $n \ge n_0(k_0), |f_n(x)-f(x)| \lt \frac{1}{k_0} \lt \epsilon$

($\impliedby$) Let $\epsilon \gt 0$. Then there exists $n_0 \in \mathbb{N}$ such that for all $n \ge n_0, |f_n(x)-f(x)| \lt \epsilon, x\in E$. Then for all $n \ge n_0$,

$$\left(\{x: |f_n(x)-f(x)| \gt \epsilon\}\right) \subset E^c$$ $$\implies \mu\left(\{x: |f_n(x)-f(x)| \gt \epsilon\}\right)=0 $$ which in turn gives us that for all $n \ge n_0$, we have $$||f_n-f||_{\infty} \lt \epsilon$$.

(Proof that $L^{\infty}$ is a Banach Space) :

Let $\{f_n\}_{n \in \mathbb{N}} \in L^{\infty}$ be a cauchy sequence. Then for $\epsilon \gt 0$, there exists a $n_0(\epsilon) \in \mathbb{N}$ such that for all $n,m \ge n_0(\epsilon)$, we have $||f_n-f_m||_{\infty} \lt \dfrac{\epsilon}{2}$. Then there exists $E^{\epsilon} \in \mathcal{M}$ such that $\mu(E^{\epsilon})=0$ and for $x \in (E^{\epsilon})^{c} ,|f_n(x)-f_m(x)| \lt \dfrac{\epsilon}{2}$.

Let $$E= \cup_n E^{\frac{1}{n}}.$$ Then $\mu(E)=0$ and for all $x \in E^c, \{f_n(x)\}$ is a cauchy sequence. Let $f(x)=\lim_n f_n(x)$ for $x \in E^c$. Let $\epsilon \gt 0$. Then there exists $k_0 \in \mathbb{N}$ such that for all $k \ge k_0,\frac{1}{k} \lt \frac{\epsilon}{2}$. Then for all $n, m \ge n_0(k_0)$, we have $$|f_n(x)-f_m(x)| \lt \frac{1}{k} \lt \frac{\epsilon}{2}$$ Letting $m \to \infty$ we have $$|f_n(x)-f(x)| \lt \epsilon, x\in E^c.$$ By Lemma-2, we have $||f_n-f||_{\infty} \to 0$. From here by use of triangle inequality, we see that $ f\in L^{\infty}$.

Is this proof alright?

Thanks for the help!!


2 Answers 2


I think your proof if correct. But here is a theorem in Folland's real analysis book I proved, (chapter 6.1, exercise 2)

Theorem 6.8

a.) If $f$ and $g$ are measurable functions on $X$, then $\|fg\|_{1}\leq\|f\|_{1}\|g\|_{\infty}$. If $f\in L^1$ and $g\in L^{\infty}$, $\|fg\|_{1} = \|f\|_{1}\|g\|_{\infty}$ if and only if $|g(x)| = \|g\|_{\infty}$ a.e. on the set where $f(x)\neq 0$.

b.) $\|\cdot\|_{\infty}$ is a norm on $L^{\infty}$.

c.) $\|f_n - f\|_{\infty}\to 0$ if and only if there exists $E\in M$ such that $\mu(E^c) = 0$ and $f_n\rightarrow f$ uniformly on $E$.

d.) $L^{\infty}$ is a Banach space.

e.) The simple functions are dense in $L^{\infty}$.

Proof a.) Let $f,g$ be measurable functions on $X$ such that $f\in L^1$ and $g\in L^{\infty}$. Then, since $g\in L^{\infty}$ then by definition $|g(x)|\leq \|g\|_{\infty}$ a.e. on $X$. So, $$|f(x)||g(x)|\leq |f(x)|\|g\|_{\infty}$$ Recall that $\int |f(x)g(x)|d\mu(x) = \|fg\|_{1}$. Now integrating over all $X$ we have \begin{align*} \int_X |f(x)||g(x)|d\mu(x) &= \int |f(x)g(x)|d\mu(x)\\ &\leq \|g\|_{\infty}\int_{X} |f(x)|d\mu(x)\\ &= \|g\|_{\infty}\|f\|_{1} \end{align*} Therefore, $$\|fg\|_{1}\leq\|f\|_{1}\|g\|_{\infty}$$ Second part:\ $\Leftarrow$ Let $A = \{x\in X: f(x)\neq 0 \}$. Suppose $|g(x)| = \|g\|_{\infty}$ for every $x\in A$. Then, $$\|fg\|_{1} = \int_{X}|f(x)||g(x)|d\mu = \|g\|_{L^\infty}\int_{A}|f(x)|d\mu = \|g\|_{L^\infty}\|f\|_{L^1}$$ since the integral is zero everywhere outside $A$, since $f(x)$ is zero on $X\setminus A$.\ $\Rightarrow$ Suppose $\|fg\|_{1} = \|f\|_{1}\|g\|_{\infty}$. So we have $$ \int |fg| d\mu = \int |f|\,\|g\|_{\infty} d\mu$$ Since $|g| \leq \|g\|_{\infty}$, we have $|fg| = |f|\,|g|\leq |f|\,\|g\|_{\infty}$, and so we must have $ |f|\,|g| = |f|\,\|g\|_{\infty}$ a.e.. So $|g(x)| = \|g\|_{\infty}$ a.e. on the set where $f(x)\neq 0$.

Proof b.) Clearly $\|f\|_{\infty} = 0$ if and only if $|f| = 0$. Now, suppose $\lambda\in K$ then $$\|\lambda f\|_{\infty} = ess \sup_{x\in X}|\lambda f(x)| = ess \sup_{x\in X} |\lambda||f(x)| = |\lambda| ess \sup_{x\in X} |f(x)| = |\lambda|\|f(x)\|$$ Finally we know that $|f(x)|\leq \|f\|_{\infty}$ and $|g(x)|\leq \|g\|_{\infty}$ a.e. on $X$, so that for a.e. on $X$ we have $$|f(x) + g(x)|\leq |f(x)| + |g(x)| \leq \|f\|_{\infty} + \|g\|_{\infty}$$ Then by definition of the essential supremum of $|f + g|$ we have $$\|f + g\|_{\infty}\leq \|f\|_{\infty} + \|g\|_{\infty}$$

Proof c.) Suppose $\|f_n - f\|\to 0$ for all $n$. Then given $\epsilon > 0$ there exists an $N\in\mathbb{N}$ such that for $n\geq N$ $$\|f_n - f\|_{\infty} < \epsilon$$ Thus for any $n\geq N$, we have \begin{equation} |f_n(x) - f(x)|\leq \|f_n - f\|_{\infty} < \epsilon \tag{1} \end{equation} a.e. on $X$. Let $M_n = \|f_n - f\|_{\infty}$ for all $n\geq N$ and set $$A_n = \{x\in X: |f_n(x) - f(x)| > M_n \}$$ Then we have $\mu(A_n) = 0$. Now, let $A = \bigcup_{n\geq N}A_n$ then $\mu(A) = 0$. Let $E = A^c$ Then for $x\in E$ we have from $(1)$ $$|f_n(x) - f(x)| \leq \|f_n - f\|_{\infty} < \epsilon \ \ \forall n\geq N$$ Thus, $$|f_n(x) - f(x)| < \epsilon \ \ \forall n \geq N$$ and so $f_n\to f$ uniformly on $E$ and clearly $\mu(E^c) = 0$. \ Conversely, suppose $E\in M$ and $\mu(E^c) = 0$ and $f_n\to f $ uniformly on $E$, then for each $\epsilon > 0$ there exists an $N\in\mathbb{N}$ such that $$|f_n(x) - f(x)| < \epsilon \ \ \forall n\geq N \ \ \text{and} \ \ x\in E$$ Hence we also have $$|f_n(x) - f(x)| < \epsilon \ \ \text{a.e. on} \ \ X$$ Thus by definition of the essential supremum of $|f_n - f|$ we have $$\|f_n - f\|_{\infty} < \epsilon$$

Proof d.) We want to show that $L^{\infty}$ is complete as a metric space in the metric associated with $\|\cdot\|_{\infty}$. Let $\{f_n\}_{n\in\mathbb{N}}$ be a Cauchy sequence in $L^{\infty}$. Thus given $\epsilon > 0$, there exists an $N\in\mathbb{N}$ such that \begin{equation} \|f_m - f_n\| < \epsilon \ \ \forall m,n\geq N \tag{1} \end{equation} For each $m,n\in\mathbb{N}$, set $$F_{m,n} = \{x\in X: |f_m(x) - f_n(x)| > \|f_m - f_n\|_{\infty} \}$$ Then clearly $\mu(F_{m,n}) = 0$ for all $m,n\in\mathbb{N}$. Set $F = \bigcup_{m,n\in\mathbb{N}}F_{m,n}$ and $E = F^c$. Note that $\mu(E^c) = \mu(F) = 0$. Moreover, \begin{align*} E &= \bigcap_{m,n\in\mathbb{N}}\{x\in X: |f_m(x) - f_n(x)|\leq \|f_m - f_n\|_{\infty} \}\\ &= \{x\in X: |f_m(x) - f_n(x)\leq \|f_m - f_n\|_{\infty} \ \forall \ m,n\in\mathbb{N} \} \end{align*} Let $\epsilon > 0$ and choose $N\in\mathbb{N}$ such that $(1)$ holds. Then for $x\in E$ and for all $m,n\geq N$ we have \begin{equation} |f_m(x) - f_n(x)| \leq \|f_n - f_m\| < \epsilon \tag{2} \end{equation} This shows that for every $x\in E$, $\{f_n(x)\}$ is a Cauchy sequence in $K = (\mathbb{R}$ or $\mathbb{C}$). Since $K$ is a complete metric space, there exists a limit $f(x) = \lim_{n\to \infty}f_n(x)$. Note $f(x)$ is defined in $E$ (i.e. outside of $F$). Thus for $x\in F$, $f(x) = 0$. Note that $f = \lim_{n\to \infty}f_n(x)\chi_{E}$ is measurable. Now recall earlier that for $\epsilon > 0$ there is an $N\in\mathbb{N}$ such that for $x\in E$ we have for all $m,n\geq N$ $$|f_m(x) - f_n(x)| \leq \epsilon$$ Let $n\to \infty$ then we have for $m\geq N$ $$|f_m(x) - f(x)| \leq \epsilon$$ Thus for $m\geq N$ $$\|f_m(x) - f(x)\|_{\infty} = ess\sup |f_m - f| \leq \epsilon$$ This shows that $f_m \to f$ in $L^{\infty}$ norm.\ Finally, we note that $f\in L^{\infty}$ from the triangle inequality: $$\|f\|_{\infty} \leq \|f_m\|_{\infty} + \|f_m - f\|_{\infty} \leq \|f_m\|_{\infty} + \epsilon < \infty$$ for $m\geq N$. Thus $L^{\infty}$ is a Banach space.

Proof e.) Let $f\in L^{\infty}$ where $f$ is bounded. Then by Theorem 2.10 there exists a sequence $\{\phi_n\}$ of simple functions that converge uniformly to $f$. Thus given $\epsilon > 0$ $$\|f - \phi_n\|_{\infty} = ess\sup|f(x) - \phi_n(x)| \leq \epsilon$$ Thus since $\epsilon$ is arbitrary we are done.

  • $\begingroup$ For part d), conclusion from part c) can be used by noting that $f_n \to f$ uniformly as their sup, $||.||_\infty$ on E is lower than $\epsilon$. $\endgroup$ Dec 2, 2019 at 18:07

I would probably add one more lemma, namely:

If $f \in L^{\infty}(X)$, then $|f(x)|\leq||f||_{\infty}$, a.e. on $X$.

This can be done by definition of infimum, like you did in Lemma 1.

Very nice proof that $L^{\infty}(X)$ is Banach space you can find in Real Analysis, Theory of measure and integration, by James Yeh.


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.