Eigra

Measurement

Measuring an observable does not reveal a value that already existed before the measurement: quantum mechanics postulates what happens, building on the eigenbasis decomposition seen earlier.

The postulate

Let A^\hat A be a Hermitian operator (an observable), with eigenbasis ∣ψn⟩| \psi_n \rangle and eigenvalues ana_n, and let ∣ψ⟩=∑ncn∣ψn⟩| \psi \rangle = \sum_n c_n | \psi_n \rangle be the state of the particle, with cn=⟨ψn∣ψ⟩c_n = \langle \psi_n | \psi \rangle. Measuring A^\hat A:

  • can only give one of the eigenvalues ana_n;
  • gives the result ana_n with probability ∣cn∣2|c_n|^2 — the Born rule;
  • immediately projects the state onto ∣ψn⟩| \psi_n \rangle: this is the collapse of the wavefunction.
∣ψ⟩→ an ∣ψn⟩with probability ∣cn∣2.| \psi \rangle \xrightarrow{\ a_n\ } | \psi_n \rangle \qquad \text{with probability } |c_n|^2.

Repeating the measurement on identically prepared systems therefore gives, on average,

⟨A^⟩=∑nan∣cn∣2=⟨ψ∣A^∣ψ⟩.\langle \hat A \rangle = \sum_n a_n |c_n|^2 = \langle \psi | \hat A | \psi \rangle.

Why Hermitian

None of this holds if A^\hat A is not Hermitian. A Hermitian operator has two properties measurement relies on directly: its eigenvalues ana_n are real — a measurement outcome is a real number, not a complex one — and its eigenstates form a complete orthonormal basis, which guarantees ∑n∣cn∣2=1\sum_n |c_n|^2 = 1: the probabilities of the different outcomes sum to 1, as they must. A non-Hermitian operator offers neither guarantee: its eigenvalues can be complex, and its eigenstates generally form neither an orthogonal basis, nor even a basis at all.

The degenerate case

If several eigenstates ∣ψn(1)⟩,…,∣ψn(dn)⟩| \psi_n^{(1)} \rangle, \dots, | \psi_n^{(d_n)} \rangle share the same eigenvalue ana_n (an eigenspace of dimension dnd_n), nothing distinguishes which one the measurement selects. The probability of ana_n is then summed over the whole eigenspace, and the state collapses onto that entire eigenspace rather than onto any single eigenvector. Writing P^n=∑k∣ψn(k)⟩⟨ψn(k)∣\hat P_n = \sum_k | \psi_n^{(k)} \rangle \langle \psi_n^{(k)} | for the projector onto it,

P(an)=∑k∣⟨ψn(k)∣ψ⟩∣2=∥P^n∣ψ⟩∥2,∣ψ⟩→ an P^n∣ψ⟩∥P^n∣ψ⟩∥.P(a_n) = \sum_k |\langle \psi_n^{(k)} | \psi \rangle|^2 = \| \hat P_n | \psi \rangle \|^2, \qquad | \psi \rangle \xrightarrow{\ a_n\ } \frac{\hat P_n | \psi \rangle}{\| \hat P_n | \psi \rangle \|}.

When dn=1d_n = 1, this reduces exactly to the non-degenerate case above.

Operator Â

State |ψ⟩

Solving…
Fill in an operator (Re/Im per cell) and a state, in a basis of dimension 2 to 4. Each row of the result is a distinct eigenvalue: its Born-rule probability, and the state collapses to if it is obtained — the eigenvector itself when the value isn't degenerate, the projection onto the eigenspace otherwise. A non-Hermitian operator is flagged before it's even sent: those are exactly the two properties the previous section relied on.