How do we quantify this kind of information in Quantum Mechanics? Let a quantum system be given with Hilbert space $\mathscr{H}$. Suppose it is described by the quantum state $\rho$, i.e., a density matrix. We can define the Von-Neumman entropy as $$S(\rho)=-\operatorname{Tr}\rho\log\rho.$$
It can be shown that $S(\rho)=0$ if and only if $\rho$ is pure, in other words, if exists $|\psi\rangle\in \mathscr{H}$ such that $$\rho=|\psi\rangle\langle \psi|.$$
Next, suppose that $|\psi\rangle$ is a pure state. It obviously encodes some information about the system in question.
For example, suppose the system is a non-relativistic spin 1/2 particle of mass $m$. In that case, a complete set of states is $|\mathbf{p},\sigma\rangle$ the eigenstates of momentum and spin. Then


*

*If $|\psi\rangle = |\mathbf{p},\sigma\rangle$ we have the information that momentum is $\mathbf{p}$, spin is $\sigma$ and mass is $m$.

*If $|\psi\rangle = c_+|\mathbf{p},1/2\rangle+c_-|\mathbf{p},-1/2\rangle$ we have information that momentum is $\mathbf{p}$, that mass is $m$ and we have some information about spin, though it has uncertainty associated with it.

*If $|\psi\rangle = \int f(\mathbf{p})|\mathbf{p},\sigma\rangle d^3\mathbf{p}$ we have the same reasoning above. We have information that spin is $\sigma$, mass is $m$ and we have some information about momentum.
Now, if $S(\rho)$ quantifies the information, we would necessarily have $S(|\psi\rangle\langle \psi|)=0$. Still it seems wrong to me to say that "information on this state is zero". As I said, in all these above examples we have at least some information about the system.
For starters in all states we know the mass for sure. This is information already.
Secondly, if $\mathbf{p},\sigma$ are described by the spinor wavefunction $$\Psi(\mathbf{p})=\begin{pmatrix}\psi_+(\mathbf{p})\\ \psi_-(\mathbf{p})\end{pmatrix},$$
we have some information aboute $\mathbf{p}$ and $\sigma$ contained in $\Psi(\mathbf{p})$ that should be quantified wth some measure of uncertainty.
In particular $\Psi(\mathbf{p})_{i}=\delta(\mathbf{p}-\mathbf{q})\delta_{ij}$ should mean "maximum information" - we know momentum and spin. And we should have "minimum information" - a state with maximum uncertainty of these variables.
So by all above it seems obvious that pure states have information associated. The information seems to be associated to observables. And finaly, it seems that this information intuitively should be quantified by a measure ranging from "minimum information/maximum uncertainty" and "maximum information/minimum uncertainty".
Moreover, the Von-Neumman entropy seems not to capture this since it says such a state has zero information associated which seems to not be the case.
So how do we quantify this information contained in a pure state in the end?
 A: Pure states carry indeed maximal information. In fact, they are maximal with respect to a partial ordering $\prec$ on states defined as follows.
Let $\rho,\omega$ be two states. We say that $\rho \prec \omega$ iff there exists $0\leq\lambda\leq 1$ such that $\rho\geq \lambda \omega$, where $\geq$ is the bigger or equal relation on operators (i.e. $A\geq 0$ iff $\langle\psi, A\psi\rangle \geq 0$ for all $\psi$).
Now, $\omega$ is an extremal, or pure state iff it is maximal w.r.t. $\prec$, in the sense that $\omega\prec\varpi$ iff $\varpi=\lambda\omega$, for some $0\leq \lambda\leq 1$.
It is not difficult to understand by this definition why pure states carry maximal information on the system, among the states. In fact, let $\rho$ be a mixed state. Then there exist two states $\varpi_1$ and $\varpi_2$, and $0<\lambda<1$ such that
$$\rho=\lambda\varpi_1 +(1-\lambda)\varpi_2\; .$$
In fact, there exist a (possibly infinite) set of pure states $\Omega=\{\omega_i,i\in I\}$, and coefficients $\{0<\lambda_i<1,i\in I\}$ such that
$$\rho=\sum_{i\in I} \lambda_i\omega_i\; .$$
Therefore, the information carried by the mixed state $\rho$ (its description of the system) is completely encoded in the information carried by the set $\Omega$ of pure states, and in general it is encoded by other states. On the other hand, a pure state carries maximal information since the information they carry cannot be encoded in the one carried by any other state. The set $P$ of pure states is therefore exhausting (by convex combinations) all the possible configurations of the system, and thus it contains in some sense all the possible knowledge about the system.
