For concreteness let's assume that $u\in W^{1,2}(\Bbb R^2).$ It is well known that $$ \|u\|_4\le C \|u\|_2^{\frac 12} \|\nabla u\|_2^{\frac 12}. $$ This is also true if $u\in W^{1,2}_0(\Omega)$ for a bounded domain $\Omega$ in $\Bbb R^2$.
Is it still true if we modify this in the same way as Poincare ineqaulity, i.e. $$ \|u-\overline u\|_4\le C \|u-\overline u\|_2^{\frac 12} \|\nabla u\|_2^{\frac 12}, $$ where $\overline u=\frac 1{|\Omega|}\int u dx$, for any $u\in W^{1,2}(\Omega)$?
I have seen the version for $u\in W^{1,2}(\Omega)$ where we need to add another term $C'\|u\|_s$ but that is not what I want.
Edit: The usual proof of this interpolation inequality use the fact that $$ \|u\|_{L^{n/(n-1)}(\Bbb R^n)} \le \|\nabla u\|_{L^1(\Bbb R^n)} $$ for compactly supported $u$. In bounded domain $\Omega$, the $u$ on the LHS cannot be easily replaced with $u-\overline u$ since Rellich-Kondrachov theorem doesn't give compactness for the embedding $W^{1,1}$ into $L^{1^*}$. Thus it seems to me that we'd need to modify the proof somewhere else if the statement is true.