Skip to content

Lesson 6: Quantum Measurement and the Postulates of Quantum Mechanics

6.1 Learning Outcomes

By the end of this hour a student will be able to:

  1. State all four postulates of quantum mechanics in the form used in quantum computing.
  2. Apply the Born rule to compute measurement probabilities in any basis.
  3. Compute the post-measurement state using projectors.
  4. Compute expectation values \(\langle A\rangle\) and uncertainties \(\Delta A\).
  5. Explain measurement in a rotated basis as rotate → measure in \(Z\).
  6. State the uncertainty principle for two observables and apply it to the Paulis.

6.2 The Four Postulates

Postulate 1 — State space

The state of an isolated quantum system is a unit vector \(|\psi\rangle\) in a complex Hilbert space \(\mathcal{H}\) associated with the system. For \(n\) qubits, \(\mathcal{H} = (\mathbb{C}^2)^{\otimes n} = \mathbb{C}^{2^n}\).

Postulate 2 — Evolution

The evolution of a closed quantum system is described by a unitary transformation:

\[ |\psi(t_2)\rangle = U(t_1,t_2)\,|\psi(t_1)\rangle . \]

In continuous time this is the Schrödinger equation

\[ i\hbar\frac{d}{dt}|\psi(t)\rangle = H|\psi(t)\rangle \qquad\Longrightarrow\qquad U(t) = e^{-iHt/\hbar}, \]

for a time-independent Hamiltonian \(H\) (a Hermitian operator, the energy observable). In quantum computing we usually set \(\hbar = 1\) and specify gates directly as unitaries.

Postulate 3 — Measurement

A quantum measurement is described by a collection \(\{M_m\}\) of measurement operators satisfying the completeness relation

\[ \sum_m M_m^\dagger M_m = I . \]

If the state before measurement is \(|\psi\rangle\), then

\[ \boxed{\ \Pr[m] = \langle\psi| M_m^\dagger M_m |\psi\rangle\ } \]

and the state immediately after obtaining outcome \(m\) is

\[ \boxed{\ |\psi'\rangle = \frac{M_m|\psi\rangle}{\sqrt{\Pr[m]}}\ } \]

The denominator re-normalises the collapsed state. Completeness guarantees \(\sum_m \Pr[m] = 1\).

Postulate 4 — Composite systems

The state space of a composite system is the tensor product of the component state spaces. If system \(j\) is prepared in \(|\psi_j\rangle\), the joint state is \(|\psi_1\rangle\otimes|\psi_2\rangle\otimes\cdots\otimes|\psi_n\rangle\).

6.3 Projective Measurement

The special case used in nearly all of quantum computing.

Definition (Projective / von Neumann measurement). An observable is a Hermitian operator \(A\) with spectral decomposition

\[ A = \sum_m \lambda_m P_m, \qquad P_m = \sum_{k} |v_{m,k}\rangle\langle v_{m,k}| \]

where \(P_m\) projects onto the eigenspace of eigenvalue \(\lambda_m\). The projectors satisfy

\[ P_m^\dagger = P_m, \qquad P_m P_{m'} = \delta_{mm'}P_m, \qquad \sum_m P_m = I . \]

Born rule. Measuring \(A\) on \(|\psi\rangle\) yields outcome \(\lambda_m\) with probability

\[ \Pr[\lambda_m] = \langle\psi|P_m|\psi\rangle = \| P_m|\psi\rangle \|^2 , \]

and the post-measurement state is \(\dfrac{P_m|\psi\rangle}{\sqrt{\Pr[\lambda_m]}}\).

The computational-basis measurement

For a single qubit, measuring \(Z\) uses \(P_0 = |0\rangle\langle 0|\) and \(P_1 = |1\rangle\langle 1|\). For \(|\psi\rangle = \alpha|0\rangle+\beta|1\rangle\):

\[ \Pr[0] = \langle\psi|0\rangle\langle 0|\psi\rangle = |\langle 0|\psi\rangle|^2 = |\alpha|^2, \qquad \Pr[1] = |\beta|^2 , \]

and the state collapses to \(|0\rangle\) or \(|1\rangle\). This is the elementary rule stated in L01, now derived from the postulate.

For an \(n\)-qubit register \(|\psi\rangle = \sum_x \alpha_x|x\rangle\):

\[ \Pr[x] = |\alpha_x|^2 = |\langle x|\psi\rangle|^2 . \]

Partial measurement

Measuring only qubit 1 of a two-qubit state uses \(P_0 = |0\rangle\langle 0|\otimes I\) and \(P_1 = |1\rangle\langle 1|\otimes I\). The unmeasured qubit is left in the correspondingly conditioned state — this is exactly the mechanism behind teleportation (L10).

Repeatability

Because \(P_m^2 = P_m\), measuring the same observable twice in a row gives the same answer with certainty:

\[ \Pr[\text{same}] = \frac{\|P_m P_m|\psi\rangle\|^2}{\|P_m|\psi\rangle\|^2} = \frac{\|P_m|\psi\rangle\|^2}{\|P_m|\psi\rangle\|^2} = 1 . \]

6.4 Measuring in a Rotated Basis

Real hardware measures only in the computational basis. To measure an arbitrary observable you must rotate first.

Recipe. To measure in the orthonormal basis \(\{|b_0\rangle,|b_1\rangle\}\), find the unitary \(V\) with \(V|b_0\rangle = |0\rangle\) and \(V|b_1\rangle = |1\rangle\), apply \(V\), then measure \(Z\).

Basis Rotation to apply before \(Z\)-measurement
\(Z\): \(\lvert 0\rangle,\lvert 1\rangle\) none
\(X\): \(\lvert +\rangle,\lvert -\rangle\) \(H\)
\(Y\): \(\lvert {+}i\rangle,\lvert {-}i\rangle\) \(S^\dagger\) then \(H\)

Check for \(X\): \(H|+\rangle = |0\rangle\) and \(H|-\rangle = |1\rangle\) ✓.

This is why the Bloch coordinate \(x\) is measured by "apply \(H\), then count zeros and ones": every tomography routine in Qiskit or Cirq does exactly this.

6.5 Expectation Values and Uncertainty

Definition (Expectation value).

\[ \langle A\rangle_\psi = \langle\psi|A|\psi\rangle = \sum_m \lambda_m \Pr[\lambda_m] . \]

This is the average over many repetitions on identically prepared states — not the result of a single measurement. For \(A\) with eigenvalues \(\pm1\) (any Pauli),

\[ \langle A\rangle = (+1)\Pr[+1] + (-1)\Pr[-1] = \Pr[+1] - \Pr[-1] . \]

Definition (Standard deviation / uncertainty).

\[ \Delta A = \sqrt{\langle A^2\rangle - \langle A\rangle^2} . \]

Theorem (Robertson uncertainty relation). For observables \(A\) and \(B\),

\[ \Delta A \cdot \Delta B \ \ge\ \frac{1}{2}\big|\langle [A,B]\rangle\big|, \qquad [A,B] = AB - BA . \]

Consequence for the Paulis. \([X,Y] = 2iZ\), so

\[ \Delta X \cdot \Delta Y \ \ge\ \big|\langle Z\rangle\big| . \]

For \(|\psi\rangle = |0\rangle\): \(\langle Z\rangle = 1\), so \(\Delta X\,\Delta Y \ge 1\). Indeed \(\langle X\rangle = \langle Y\rangle = 0\) and \(X^2=Y^2=I\) give \(\Delta X = \Delta Y = 1\) — the bound is saturated. You cannot know \(X\) and \(Y\) simultaneously.

Commuting observables are compatible. If \([A,B]=0\) they share an eigenbasis and can be measured simultaneously with no uncertainty trade-off. E.g. \(Z\otimes I\) and \(I\otimes Z\) commute — you can measure both qubits at once.

6.6 What Measurement Costs You

Three hard limits that shape every algorithm in this course:

  1. Collapse. After measuring, the superposition is gone. You get one sample, not the amplitudes.
  2. Holevo's bound. \(n\) qubits can transmit at most \(n\) classical bits of accessible information.
  3. No-cloning (L09). You cannot copy the state to measure it in several bases.

Therefore any quantum algorithm must engineer interference so that the single measurement at the end is very likely to reveal the answer. Reading out "all \(2^n\) amplitudes" is impossible.

6.7 Solved Problems

Problem 1 — Born rule in two bases

For \(|\psi\rangle = \frac{1}{\sqrt5}\big(|0\rangle + 2|1\rangle\big)\), compute (a) \(\Pr[0],\Pr[1]\) in the \(Z\) basis, (b) \(\Pr[+],\Pr[-]\) in the \(X\) basis, (c) the post-measurement state if the \(X\)-measurement gives \(-\).

Solution.

(a) \(\alpha = \frac{1}{\sqrt5}\), \(\beta = \frac{2}{\sqrt5}\):

\[ \Pr[0] = \tfrac15 = 0.2, \qquad \Pr[1] = \tfrac45 = 0.8 . \]

(b) \(\langle +|\psi\rangle = \frac{1}{\sqrt2}\cdot\frac{1}{\sqrt5}(1+2) = \frac{3}{\sqrt{10}}\), \(\langle -|\psi\rangle = \frac{1}{\sqrt2}\cdot\frac{1}{\sqrt5}(1-2) = \frac{-1}{\sqrt{10}}\).

\[ \Pr[+] = \tfrac{9}{10} = 0.9, \qquad \Pr[-] = \tfrac{1}{10} = 0.1 \qquad (\text{sum } = 1\ \checkmark) \]

(c) Post-measurement state after outcome \(-\):

\[ |\psi'\rangle = \frac{P_-|\psi\rangle}{\sqrt{\Pr[-]}} = \frac{|-\rangle\langle-|\psi\rangle}{\sqrt{1/10}} = \frac{(-1/\sqrt{10})\,|-\rangle}{1/\sqrt{10}} = -|-\rangle \simeq |-\rangle . \]

The \(-1\) is a global phase and is discarded. The state has collapsed to \(|-\rangle\); the original amplitudes are irretrievably lost.

Problem 2 — Expectation value and uncertainty

For \(|\psi\rangle = \cos\frac{\pi}{6}|0\rangle + \sin\frac{\pi}{6}|1\rangle\), compute \(\langle Z\rangle\), \(\langle X\rangle\), \(\Delta Z\) and \(\Delta X\).

Solution. \(\cos\frac{\pi}{6} = \frac{\sqrt3}{2} \approx 0.8660\), \(\sin\frac{\pi}{6} = \frac12\). So \(\alpha = \frac{\sqrt3}{2}\), \(\beta = \frac12\), both real.

\[ \langle Z\rangle = |\alpha|^2 - |\beta|^2 = \tfrac34 - \tfrac14 = \tfrac12 \]
\[ \langle X\rangle = 2\,\mathrm{Re}(\alpha^*\beta) = 2\cdot\frac{\sqrt3}{2}\cdot\frac12 = \frac{\sqrt3}{2}\approx 0.8660 \]

Since \(Z^2 = X^2 = I\), we have \(\langle Z^2\rangle = \langle X^2\rangle = 1\). Hence

\[ \Delta Z = \sqrt{1 - \tfrac14} = \frac{\sqrt3}{2} \approx 0.8660 \]
\[ \Delta X = \sqrt{1 - \tfrac34} = \frac12 = 0.5 \]

Check the uncertainty relation. \(\langle Y\rangle = 2\mathrm{Im}(\alpha^*\beta) = 0\), and \(\Delta Z\,\Delta X = \frac{\sqrt3}{2}\cdot\frac12 = 0.433 \ge \frac12|\langle[Z,X]\rangle| = \frac12|2i\langle Y\rangle| = 0\) ✓.

Note \(\langle X\rangle^2 + \langle Y\rangle^2 + \langle Z\rangle^2 = \frac34 + 0 + \frac14 = 1\) ✓ (pure state on the Bloch sphere).

Problem 3 — Partial measurement of a two-qubit state

A register is in \(|\psi\rangle = \frac{1}{\sqrt6}\big(|00\rangle + |01\rangle + 2|11\rangle\big)\). Measure only the first qubit in the \(Z\) basis. Find the outcome probabilities and both possible post-measurement states.

Solution. Check normalisation: \(\frac{1+1+4}{6} = 1\) ✓.

Projector for outcome 0 on qubit 1: \(P_0 = |0\rangle\langle0|\otimes I\). Applying it keeps only the terms whose first bit is 0:

\[ P_0|\psi\rangle = \frac{1}{\sqrt6}\big(|00\rangle + |01\rangle\big) \]
\[ \Pr[q_1 = 0] = \|P_0|\psi\rangle\|^2 = \frac{1+1}{6} = \frac{2}{6} = \frac{1}{3} \]

Post-measurement state:

\[ |\psi_0'\rangle = \frac{P_0|\psi\rangle}{\sqrt{1/3}} = \sqrt3\cdot\frac{1}{\sqrt6}\big(|00\rangle+|01\rangle\big) = \frac{1}{\sqrt2}\big(|00\rangle + |01\rangle\big) = |0\rangle\otimes|+\rangle . \]

Similarly \(P_1|\psi\rangle = \frac{2}{\sqrt6}|11\rangle\):

\[ \Pr[q_1 = 1] = \frac46 = \frac23, \qquad |\psi_1'\rangle = \frac{(2/\sqrt6)|11\rangle}{\sqrt{2/3}} = |11\rangle . \]

Total: \(\frac13 + \frac23 = 1\) ✓.

Observation. Measuring qubit 1 changed the state of qubit 2 even though qubit 2 was not touched: it became \(|+\rangle\) or \(|1\rangle\) depending on the outcome. This is the signature of entanglement, and it is the working principle of teleportation.

Problem 4 — Designing a measurement in a rotated basis

Show how to measure the observable \(A = \frac{X+Z}{\sqrt2}\) on hardware that only measures \(Z\). What are the possible outcomes?

Solution. First, note \(A^2 = \frac{(X+Z)^2}{2} = \frac{X^2 + XZ + ZX + Z^2}{2}\). Since \(X\) and \(Z\) anticommute, \(XZ + ZX = 0\), so \(A^2 = \frac{I+I}{2} = I\). Hence the eigenvalues of \(A\) are \(\pm 1\).

Observe that \(A = \frac{1}{\sqrt2}\begin{pmatrix}1 & 1\\ 1 & -1\end{pmatrix} = H\). The Hadamard gate is this observable. Its eigenvectors satisfy \(H|h_\pm\rangle = \pm|h_\pm\rangle\); the \(+1\) eigenvector lies along the Bloch axis \(\hat n = \frac{\hat x + \hat z}{\sqrt2}\).

Measurement recipe. We need a unitary \(V\) with \(VHV^\dagger = Z\). Since \(H\) is a \(\pi\) rotation about \(\frac{\hat x+\hat z}{\sqrt2}\), and we want to bring that axis onto \(\hat z\), use a rotation of \(-\pi/4\) about the \(y\)-axis:

\[ V = R_y(-\pi/4) = \begin{pmatrix}\cos\frac{\pi}{8} & \sin\frac{\pi}{8}\\ -\sin\frac{\pi}{8} & \cos\frac{\pi}{8}\end{pmatrix} \]

Procedure: apply \(R_y(-\pi/4)\) to the qubit, then measure in the computational basis. Reading \(0\) means outcome \(+1\); reading \(1\) means outcome \(-1\).

Generalisation. To measure \(\hat n\cdot\vec\sigma\), rotate \(\hat n\) onto \(\hat z\) and measure \(Z\). Every "measure this observable" instruction reduces to rotate, then measure \(Z\).

Problem 5 — Sampling error: how many shots?

A quantum circuit outputs \(|1\rangle\) with true probability \(p = 0.75\). You run it \(N\) times and estimate \(\hat p\). How many shots are needed so that the standard error is below 0.01?

Solution. Each shot is a Bernoulli trial. The estimator \(\hat p = \frac{\text{count of }1}{N}\) has

\[ \mathrm{Var}(\hat p) = \frac{p(1-p)}{N}, \qquad \mathrm{SE} = \sqrt{\frac{p(1-p)}{N}} . \]

Require \(\mathrm{SE} \le 0.01\):

\[ \sqrt{\frac{0.75\times0.25}{N}} \le 0.01 \ \Longrightarrow\ \frac{0.1875}{N} \le 10^{-4} \ \Longrightarrow\ N \ge 1875 . \]

So about 1900 shots.

The general lesson. Statistical error falls as \(1/\sqrt N\), so improving precision by a factor of 10 costs 100× more shots. This \(1/\sqrt N\) wall is why variational algorithms (Unit III) are expensive, and why quantum advantage claims must always report shot counts.

6.8 Practice Problems

  1. For \(|\psi\rangle = \frac{1}{2}|0\rangle + \frac{\sqrt3}{2}|1\rangle\), find \(\Pr[+]\) and \(\Pr[-]\). [Ans: \(\approx 0.933, 0.067\)]
  2. Compute \(\langle Y\rangle\) for \(|{+}i\rangle\). [Ans: 1]
  3. A two-qubit state is \(\frac{1}{\sqrt2}(|00\rangle + |11\rangle)\). Measure qubit 1; give both outcomes' probabilities and the resulting states.
  4. Show \(\Delta A = 0\) if and only if \(|\psi\rangle\) is an eigenvector of \(A\).
  5. Verify \([Z,X] = 2iY\) by direct matrix multiplication.
  6. How many shots are needed to distinguish \(p=0.50\) from \(p=0.52\) at two standard errors? [Ans: ≈ 10 000]
  7. Explain why measuring \(Z\) then \(X\) then \(Z\) again on \(|0\rangle\) gives \(|1\rangle\) half the time on the final \(Z\) measurement.

6.9 Summary

  • Four postulates: state = unit vector; evolution = unitary; measurement = operators \(\{M_m\}\) with \(\sum M_m^\dagger M_m = I\); composition = tensor product.
  • Born rule: \(\Pr[m] = \langle\psi|P_m|\psi\rangle\); state collapses to \(P_m|\psi\rangle/\sqrt{\Pr[m]}\).
  • Measurement in any basis = rotate to the computational basis, then measure \(Z\).
  • \(\langle A\rangle = \langle\psi|A|\psi\rangle\) is an average over shots; \(\Delta A\,\Delta B \ge \frac12|\langle[A,B]\rangle|\).
  • Measurement is destructive and yields only \(n\) bits from \(n\) qubits — algorithms must engineer interference before reading out.
  • Statistical error scales as \(1/\sqrt N\) in the number of shots.

Next lesson: L07 — Single-qubit gates and the quantum circuit model.

Practise now: unit1-qubit-lab.html → Measurement Simulator: choose a state, a basis and a shot count, and compare empirical histograms with the Born-rule prediction.