Let $\mathcal{H}$ be a Hilbert space and $A:\mathcal{D}(A)\to\mathcal{H}$ be a densely-defined, closable linear operator. It is called positive (or better non-negative), if $\langle Ax,x\rangle\in\mathbb{R}$ for all $x\in\mathcal{D}(A)$ and
$$\langle Ax,x\rangle\geq 0\qquad\qquad\forall x\in\mathcal{D}(A)$$
Now, consider the closure $\overline{A}:\mathcal{D}(\overline{A})\to\mathcal{H}$. My question:
If $A$ is positive, does this imply that also $\overline{A}$ is positive?
By definition, $\omega\in\mathcal{D}(\overline{A})$, if there exists a sequence $(\omega_{n})_{n}$ in $\mathcal{D}(A)$ converging to $\omega$ such that $(A\omega_{n})_{n}$ is convergent. In this case, we set $\overline{A}\omega:=\lim_{n\to\infty} A\omega_{n}$. My first idea was trying to prove that
$$\langle A\omega_{n},\omega_{n}\rangle\to\langle\overline{A}\omega,\omega\rangle\qquad \mathrm{as}\quad n\to\infty$$
because this would prove the claim (a non-negative series in $\mathbb{R}$ always converges to a non-negative element). However, I am not sure whether the limit above is correct, since in general, if $a_{n}$ and $b_{n}$ are two convergent sequences with limits $a$ and $b$, respectively, then in general it is not true that $\langle a_{n},b_{n}\rangle$ converges to $\langle a,b\rangle$.
To see that the convergence
$$\langle A\omega_{n},\omega_{n}\rangle\to\langle\overline{A}\omega,\omega\rangle$$
we make the following estimates:
$$\vert \langle A\omega_{n},\omega_{n}\rangle-\langle\overline{A}\omega,\omega\rangle\vert = \vert \langle A\omega_{n},\omega_{n}\rangle-\langle\overline{A}\omega,\omega+\omega_{n}-\omega_{n}\rangle\vert =\\=\vert\langle A\omega_{n}-\overline{A}\omega,\omega_{n}\rangle-\langle\overline{A}\omega,\omega-\omega_{n}\rangle\vert\leq \underbrace{\Vert A\omega_{n}-\overline{A}\omega\Vert}_{\to 0}\cdot\underbrace{\Vert\omega_{n}\Vert}_{\to \Vert\omega\Vert}+\Vert\overline{A}\omega\Vert\cdot\underbrace{\Vert\omega-\omega_{n}\Vert}_{\to 0}\xrightarrow{n\to\infty}0$$
where we used the triangle inequality, the Cauchy-Schwarz inequality as well as the fact that convergence implies convergence of norms, i.e. $\omega_{n}\to\omega$ implies $\Vert\omega_{n}\Vert\to\Vert\omega\Vert$ (by the inverse triangle inequality).