Question on $L^p$ spaces defining metric

451 Views Asked by At

First question here so really excited and hope you can help me, thanks!

In my intro to functional analysis class we just now covered $L^p$ spaces and I was presented with this homework question:

Given $ 0 < p < 1 $, show the formula $$ \rho(f,g)= \int |f-g|^p\mathsf dx $$ defines a metric and the resulting topological vector space is complete.

I don't think I quite got $L^p$ spaces yet and so I really hope there is someone who can take it easy on me and help me with this. Thanks all

EDIT: I should have mentioned triangle inequality and completeness of the space are where I am stuck

1

There are 1 best solutions below

4
On BEST ANSWER

Given a measure space $(X,\mathcal M,\mu)$, we define $$L^1(X) = \left\{f : X\to\mathbb C \;\mid\; \int_X |f|\mathsf d\mu < \infty\right\}$$ and $$\mathcal L^1(X) = L^1(X) / \sim $$ where $\sim$ is the equivalence relation on the bracketed set defined by $f\sim g$ iff $f=g$ a.e. That is, $$\mu(\{x\in X : f(x)\ne g(x)\}) = 0. $$ In other words, we identify two integrable functions $f$ and $g$ if they are equal almost everywhere. Generally this distinction isn't made explicit, and $L^1$ is used to denote both of these (since it is cumbersome to refer to equivalence classes instead of functions). Given $0<p<\infty$, define $$L^p(X) = \left\{f:X\to\mathbb C \mid f^p \in L^1(X)\right\}. $$ That is, $f\in L^p(X)$ iff $$\int_X |f|^p \mathsf d\mu < \infty,$$ and define $$\mathcal L^p(X) = L^p(X)/\sim. $$

Suppose now $0<p<1$. Since a nonnegative measurable function $f$ has zero integral iff $f=0$ a.e. it is clear that for $f,g\in\mathcal L^p(X)$ $$\int_X |f-g|^p\mathsf d\mu = 0 \iff f=g, $$ where we interpret $f=g$ as $f=g$ a.e. (note that without the technical point about equivalence classes, we would only have a pseudometric space). The second condition of a metric space is clear from $|f-g|^p = |g-f|^p$. If $f,g,h\in\mathcal L^p(X)$ then \begin{align} \rho(f,h) &= \int_X |f-h|^p \mathsf d\mu\\ &= \int_X |f-g+g-h|^p \mathsf d\mu\\ &\leqslant \int_X (|f-g|^p + |g-h|^p)\mathsf d\mu \tag1\\ &= \int_X |f-g|^p\mathsf d\mu + \int_X |g-h|^p\mathsf d\mu, \end{align} where $(1)$ is justified by the inequality $(a+b)^p \leqslant a^p + b^p$ for $a,b\geqslant0$.

Now, $\rho$ induces a topology on $X$ with the basis consisting of all open balls, i.e. sets of the form $$B(f;\varepsilon) := \{g\in X : \rho(f,g) <\varepsilon\}. $$

Edit: I read the question as "show that $X$ is a topological vector space," as opposed to "show that the resulting topological vector space is complete." So to address that notion:

We need to show that if $\{f_n\}$ is a Cauchy sequence in $\mathcal L^p(X)$ then there exists $f\in\mathcal L^p(X)$ with $\lim_{n\to\infty} f_n = f$. Since $\{f_n\}$ is a Cauchy sequence, we can construct a subsequence $\{f_{n_k}\}$ such that $\rho(f_{n_k}, f_{n_{k+1}}) \leqslant 2^{-k}$ for $k\in\mathbb N$. Define $$ f = f_{n_1} + \sum_{k=1}^\infty (f_{n_{k+1}}-f_{n_k}). $$ Then for any $N\in\mathbb N$, $$\rho\left(\sum_{k=1}^N f_{n_{k+1}},\; \sum_{k=1}^N f_{n_k}\right)\leqslant \sum_{k=1}^N \rho(f_{n_{k+1}},f_{n_k})\leqslant \sum_{k=1}^N 2^{-k}\leqslant 1, $$ by the triangle inequality. Put $$g_N= |f_{n_1}|+\sum_{k=1}^N |f_{n_{k+1}}-f_{n_k}|.$$ Then $g_N\leqslant g_{N+1}$ and $$g:=\lim_{N\to\infty} g_N = |f_{n_1}|+\sum_{k=1}^\infty |f_{n_{k+1}}-f_{n_k}|, $$ so by monotone convergence \begin{align} \int_X \left(|f_{n_1}|+\sum_{k=1}^\infty |f_{n_{k+1}}-f_{n_k}|\right)^p \mathsf d\mu &= \lim_{N\to\infty}\int_X \left(|f_{n_1}|+\sum_{k=1}^N |f_{n_{k+1}}-f_{n_k}|\right)^p\mathsf d\mu\\ &\leqslant \int_X |f_{n_1}|^p\mathsf d\mu + \lim_{N\to\infty}\int_X \left(\sum_{k=1}^N |f_{n_{k+1}}-f_{n_k}|\right)^p\mathsf d\mu\\ &\leqslant \int_X|f_{n_1}|^p\mathsf d\mu + \lim_{N\to\infty}\sum_{k=1}^N \int_X |f_{n_{k+1}}-f_{n_k}|^p\mathsf d\mu\\ &= \int_X|f_{n_1}|^p\mathsf d\mu + \sum_{k=1}^\infty \rho(f_{n_{k+1}},f_{n_k})\\ &\leqslant \int_X|f_{n_1}|^p\mathsf d\mu + 1 < \infty, \end{align} so that $g = \lim_{N\to\infty}g_N \in\mathcal L^p(X)$. Now $$|f| = \left| f_{n_1} + \sum_{k=1}^\infty (f_{n_{k+1}}-f_{n_k})\right| \leqslant |f_{n_1}| + \sum_{k=1}^\infty |f_{n_{k+1}}-f_{n_k}| = g, $$ so that $f\in\mathcal L^p(X)$. Finally, as $\{f_n\}$ is a Cauchy sequence with a subsequence that converges to $f$, it follows that $\lim_{n\to\infty} f_n=f$.