The difference between a density matrix and a state vector is that the latter corresponds to having more complete information about the system. Let’s see how this plays out in a simple example.

A density matrix that is diagonal with entries of {1}/{2} means that we are talking of just a two dimensional state space (such as that of a single electron whose position we are ignoring), and that we have chosen a particular observable, such as spin component in the z direction, to determine the basis with respect to which you are representing the state.

In this context there are actually many different pure states for which the probability of having spin up or down in the z direction are equal. It could for example be certain to have a particular value in (say) the x direction, or the y, or any other direction perpendicular to the z axis. Each of these corresponds to a “ket” whose components (relative to the z eigenstates) are of the form {1}/{\sqrt{2}} multiplied by complex phase factors (and the different relative phases correspond to different directions of spin in the xy plane). On the other hand, representing the state by that density matrix means that we have prepared the electron in such a way that its spin has equal probability of being measured up and down in the z direction, but *also that all ways of getting that result are equally likely* so there is also equal probability of getting up or down in any other direction.