Let $(E,N)$ be a normed space, and $M$ be another norm. Can you find a necessary and sufficient condition on $M$ so that $\sqrt{NM}$ is also a norm on $E$?
I hink the condition is that $M$ is proportional to $N$.
Of course, the only thing to investigate is the triangle inequality: we want to prove that if for all $x, y \in E$ there holds $$ \sqrt{N(x+y)M(x+y)} \leq \sqrt{N(x)M(x)} + \sqrt{N(y)M(y)} $$ then there is $\lambda >0$ such that $M=\lambda N$.
Does anoyone know?
I tried to prove a "reverse" triangle inequality without success: $$ N(x+y)M(x+y) \geq (N(x)-N(y))(M(x)-M(y)) = N(x)M(x)+N(y)M(y) - N(x)M(y)-M(x)N(y) $$ wheras $(\sqrt{N(x)M(x)} + \sqrt{N(y)M(y)})^2 = N(x)M(x) + N(y)M(y) +2\sqrt{N(x)M(x)N(y)M(y)}$ resembles it, but I'm unable to finish.
EDIT: I tried (assuming that $\sqrt{NM}$ is a norm) that for all $x, y \in E$ there holds $(\sqrt{N(x)M(y)} - \sqrt{N(y)M(x)})^2 \leq 0$, which then shows that $N$ and $M$ are proportional.
\begin{equation} (i),,,, A(x)=0 \text{ iff }x=0 \end{equation}
\begin{equation} (ii),,,, A(\lambda x)=|\lambda|A(x) \text{ for any }\lambda \in \mathbb{R} \end{equation}
it suffices to show that the set $S:={A(x)\leq 1}$ is convex (it is also necessary).
– Ragno May 06 '24 at 14:36