[Math] Not all norms are equivalent in an infinite-dimensional space

functional-analysisnormed-spaces

How to prove that not all norms are equivalent in an infinite-dimensional vector space?

In particular, I would like to prove that for a space $X$ of continuous real-valued functions defined on interval $[0,1]$, every two norms $\|\ .\|_p$ ($p \in [1, \infty]$) are not equivalent.

Best Answer

Consider the two spaces $L^{p_1}(-1, 1), L^{p_2}(-1, 1)$ with $1\le p_1<p_2 < \infty$. Let $f \in L^{p_1}\cap L^{p_2}(-1, 1)$. For all $\lambda\ge 1$ define $$ f_\lambda(x)=\lambda^\frac1{p_1}f(\lambda x).$$ Then you have $$\tag{1} \|f_\lambda\|_{p_1}=\|f\|_{p_1} $$ and $$\tag{2} \|f_\lambda\|_{p_2} = \lambda^{\frac{p_2-p_1}{p_1p_2}}\|f\|_{p_2}$$ If the two norms were equivalent on $L^{p_1}\cap L^{p_2}$ you would have, by definition, $$ c\|f_\lambda\|_{p_2} \le \|f_\lambda\|_{p_1}\le C\|f_\lambda\|_{p_2}$$ but (1) and (2) show that this is not possible (to see why, let $\lambda \to \infty$).

Related Question