I am looking at the proof showing that $L^2(0,1)$ is meager in $L^1(0,1)$. Define $B_n = \{f\in L^2 : \|f\|_2 \leq n\}$. With the continuous identity map $T:L^2 \rightarrow L^1$, if one of $T(B_n)$ has non empty interior, then $T$ is an open map, thus $T$ is onto which is a contradiction. Therefore, all of the image $T(B_n)$ has empty interior (and closed) thus $T(L^2) = T(\bigcup_n B_n)$ is meager in $L^1$.
My question is, how do we get from one $T(B_n)$ has non empty interior in $L^1$ to $T$ is an open map? The solution makes this looks very trivial, but I just couldn't see it.
Okay, I think I have the answer.
Method One: Take $f\in B_n$ and $g\in L^1\setminus L^2$, then $f+\frac{g}{n}$ would be a sequence in $L^1\setminus L^2$ which converges to $f$ in $L^1$. This would show that each $T(B_n)$ has empty interior, because the tail of this sequence is contained in any $L^1$ ball of $T(f)$.
Okay, here is the second method, Let $T:L^2\rightarrow L^1$ be the inclusion map. And denote $B_1(f,r)$ and $B_2(f,r)$ be open ball in $L^1$ and $L^2$.
Now if $T(\overline{B_2(0,n)})$ has non-empty interior in $L^1$, then there exists a $L^1$-open ball that is contained in $T(\overline{B_2(0,n)})$, let us write this $L^1$-open ball as $f_0 + rB_1(0,1)$, note that $f_0$ is in $\overline{B_2(0,n)}$, thus $$rB_1(0,1) \subset T(\overline{B_2(0,n)})- f_0\subset T(\overline{B_2(0,2n)})$$ and this means for some constant $M$, we have $$MB_1(0,1) \subset T(B_2(0,1)).$$ This says the image of the unit ball in $L^2$ is a open neightborhood of $0$ in $L^1$.