How do you describe half of an entangled pair, honestly?
Topic 01 described an isolated system with a single vector |ψ⟩. That stops working the moment two systems interact and you only have access to one of them, which is the normal situation for a qubit sitting inside a noisy chip, or one half of an entangled pair.
Take a Bell pair. The two qubits together are in a perfectly well-defined pure state. But ask "what is the state of qubit A, ignoring B entirely?" and there is no single vector that answers that: qubit A's measurement statistics, considered alone, look exactly like a classical coin flip. Not because anything is uncertain about the pair. Because the question itself doesn't have a vector-shaped answer.
The density matrix is the object that does answer it. It generalises "state" from a vector to an operator, and it can represent both an honest pure state and this new situation: a system whose own state is entangled with something you can't see.
A pure state |ψ⟩ has density matrix ρ = |ψ⟩⟨ψ|. A mixed state (genuine classical uncertainty over which pure state, with probabilities pi summing to 1) is:
ρ = Σᵢ pᵢ |ψᵢ⟩⟨ψᵢ|
Purity is testable in one number: Tr(ρ²) = 1 for a pure state, and Tr(ρ²) < 1 for a mixed one (down to 1/d for the maximally mixed state on a d-dimensional space: total ignorance).
The partial trace is how you get from "the pair's state" to "qubit A's state, ignoring B": ρA = TrB(ρAB). Apply it to a Bell pair's density matrix and the result is the maximally mixed state on qubit A alone, confirming exactly the claim above: a perfectly pure joint state can have an honestly, provably mixed reduced state.
Purification runs the logic in reverse: any mixed state ρ on system A can be written as the reduced state of some pure state |Ψ⟩ on a larger system AB. This isn't just a trick. It says every source of classical uncertainty can be modelled as entanglement with something outside your view, which is exactly how a noisy qubit is treated when the "something else" is the uncontrolled environment.
Decoherence and the entropy topics ahead Every real qubit is mixed, not pure: the density matrix is the only honest way to describe a qubit that has been sitting in a noisy environment for any length of time, which is the entire subject of quantum error correction. It's also the object von Neumann entropy is computed from (topic 15), and the reason entanglement can be verified statistically even though neither qubit alone shows anything unusual on its own.
Check yourself
A Bell pair is in a perfectly pure joint state. What is the state of one of its qubits taken on its own?
The partial trace of a Bell pair over one qubit gives the maximally mixed state: a pure joint state can have an honestly, provably mixed half.