I have read about dual spaces and the relation $1/p+1/q=1$ as mentioned in the Wikipedia page. Are there any more theorems or relations that connect elements between the $L^p$ spaces for different $p$'s?
-
1You mean for different $p_1$ and $p_2$, how is $L^{p_1}$ related to $L^{p_2}$? – chhro Jul 23 '15 at 06:54
-
yes, exactly between different p's. – Creator Jul 23 '15 at 06:55
4 Answers
Corollary 3 of Chapter 7 of Royden:
If $E$ is measurable with finite measure and $1\leq p_1< p_2\leq \infty$, then $L^{p_2}(E)\subseteq L^{p_1}(E)$. Furthermore, $||f||_{p_1}\leq c||f||_{p_2}$ for all $f\in L^{p_2}(E)$ where $c=[m(E)]^{\frac{p_2-p_1}{p_1p_2}}$ if $p_2<\infty$ and $c=[m(E)] ^{\frac{1}{p_1}}$ if $p_2=\infty$.
Royden remarks in an example however that for measurable set $E$ of infinite measure, there are no inclusion relationships among the $L^p$ spaces.
- 2,348
- 10
- 15
Here are some interesting facts about the relations between different $L^p$-spaces over the same measure space $(X,\Sigma,\mu)$ (based on Section 6.1 of Folland, 1999):
If $0<p<q<r\leq\infty$ and if $f\in L^q$, then there exist $g\in L^p$ and $h\in L^r$ such that $f=g+h$.
If $0<p<q<r\leq\infty$, then $L^p\cap L^r\subseteq L^q$.
If $0<p<q\leq\infty$ and $\mu(X)<\infty$, then $L^q\subseteq L^p$.
Suppose that $0<p<q<\infty$. Then, $$\exists f\in L^p\setminus L^q\Longleftrightarrow\forall\varepsilon>0,\exists S\in\Sigma:0<\mu(S)<\varepsilon,$$ and $$\exists f\in L^q\setminus L^p\Longleftrightarrow\forall K>0,\exists S\in\Sigma:K<\mu(S)<\infty.$$
Perhaps the most surprising one: If $X=(0,\infty)$, $\Sigma$ is the Borel $\sigma$-algebra on $(0,\infty)$, and $\mu$ is the Lebesgue measure, then there exists for any given $p\in(0,\infty)$ some $f\in L^p$ such that $f$ is not a member of any other $L^q$-space for $q\in(0,\infty)$ and $q\neq p$!
- 24,065
Here are some examples of relations between the spaces:
If $p,q\in[1,\infty)$, there is a bijection $L^p\to L^q$, namely $L^p\ni f\mapsto |f|^{p/q-1}f\in L^q$.
If $A$ has finite measure, then $p\geq q$ implies $L^p(A)\subset L^q$.
If $f\in L^p$ and $g\in L^q$ so that $1/p+1/q=1/s$ (assuming $p,q,s\in[1,\infty]$), then the pointwise product $fg$ is in $L^s$. This follows from Hölder's inequality.
- 26,444
-
-
-
@JoonasIlmavirta is the Hölder's inequality valid for p=2 and q=1, which means s<1? – Creator Jul 24 '15 at 00:07
-
@Creator, I meant the result in the usual framework of $L^p$ spaces and thus implicitly assumed $p,q,s\geq1$. It might be true even for all or some $p,q,s>0$ but there will be an additional constant in Hölder's inequality. I'll edit for clarity. – Joonas Ilmavirta Jul 24 '15 at 10:21
This is not quite what was asked for, but I thought it worth mentioning:
Below is (the non-trivial) Proposition 11.1.9 in Kalton and Albiac's Topics in Banach Space Theory:
$\ \ \ $(i) For $1\le p\le2$, $L_q[0,1]$ embeds in $L_p[0,1]$ if and only if $p\le q\le2$.
$\ \ \ $(ii) For $2< p<\infty$, $L_q[0,1]$ embeds in $L_p[0,1]$ if and only if $q=2$ or $q=p$.
Moreover, if $L_q$ embeds in $L_p$, then it embeds isometrically.
Here, "$X$ embeds into $Y$" means there is a linear operator $T:X\rightarrow Y$ that is an isomorphism onto $T(X)$. Equivalently, $X$ embeds into $Y$ iff $Y$ contains a subspace isomorphic to $X$.
- 76,633
-
@DavidMitraThis may be interesting as I am looking for operator to take an element of $L^2$ to $L^1$. Is it possible? Could you please help? – Creator Jul 24 '15 at 00:13
-
@Creator Certainly. On $[0,1]$, just take the formal identity from $L_2$ to $L_1$. This is well-defined and continuous from the remarks in the first answer. (This won't give an embedding, though.) – David Mitra Jul 24 '15 at 07:18