I am deriving the error estimates of one nonconforming finite element method, and for the error of the gradient of true solution, $\mathbf{q}_h$, which is a piecewise constant vector, I have the following results:

$\Omega$ is an open bounded domain, and for any function $v\in H^1(\Omega)$ and $v=0$ on a nontrivial portion of boundary $\Gamma_1$, we have \begin{equation} \sum_{T\in\mathcal{T}}(\mathbf{q}_h,\nabla v)_{0,T}\lesssim C\Vert\nabla v\Vert_0 \end{equation} where $(\cdot,\cdot)_{0,T}$ denotes the $L^2$ inner product on each element $T$ in triangulation $\mathcal{T}$ of $\Omega$.

From this inequality, can we have any estimate about the norm of $\mathbf{q}_h$. For example, can we say that \begin{equation} \Vert \mathbf{q}_h\Vert_0\lesssim C? \end{equation}


Define $C:=\min\{\tilde C>0:(q,\nabla v)\leq \tilde C\,|u|_1\}$.

We know that, for all $|u|_1\neq 0$,

$$\dfrac{(q\,,\nabla u)}{|u|_1}\leq C$$

and, thanks Cauchy--Schwarz inequality,

$$\dfrac{(q\,,\nabla u)}{|u|_1}\leq \|q\|_0$$

for all $q$.

What happen if we suppose that exists $\tilde q$ such that $\|\tilde q\|_0>C$?

  • $\begingroup$ I don't quite understand your answer, since that $C$ and $\m{q}_h$ is settled and we can only choose different $u$ to test this inequality. I have worked out other ways to estimate my specific $\m{q}_h$ using some additional condition. Thanks for your answer. $\endgroup$ – Y. Liang Feb 6 '18 at 5:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.