Skip to content
Tutorials Center
My learning

Topics

Lesson 25 of 38 · Operators, Observables and Uncertainty

Expectation Values and Measurement

Compute expectation values, expand a state in eigenfunctions, find measurement probabilities |c_n|², and simulate repeated energy measurements in Python.

Quantum mechanics does not predict what a single measurement will give. It predicts the probability of each possible result, and hence the average over many identical experiments. This lesson shows exactly how: you write the state as a sum of eigenfunctions of the observable, and the squared size of each coefficient is the probability of the corresponding eigenvalue.

The same recipe works for any observable and any state, so it is the most heavily used piece of the theory.

The expectation value

The expectation value of an observable A in a state ψ is the average result you would get over many measurements on identically prepared systems:

Text
⟨A⟩ = ∫ ψ*(x) Â ψ(x) dx          (ψ normalized: ∫ |ψ|² dx = 1)

Two familiar cases:

Text
⟨x⟩ = ∫ x |ψ|² dx                    (the mean position)
⟨p⟩ = ∫ ψ* (-iħ dψ/dx) dx           (the mean momentum)

The operator acts on ψ first and the result is multiplied by ψ*; the order matters when  is a derivative.

The expectation value need not be a value you can ever measure. For a fair die the mean is 3.5, but you never roll 3.5. Likewise a particle in a box in the state (ψ₁ + ψ₂)/√2 has an average energy 2.5 E₁ that is not one of the allowed energies.

Expanding a state in eigenfunctions

Let φ_n be the orthonormal eigenfunctions of an observable Â, with eigenvalues a_n. Because the eigenfunctions of a Hermitian operator form a complete set, any state can be written

Text
ψ = Σ c_n φ_n,        c_n = ∫ φ_n* ψ dx

The coefficient c_n is the overlap of the state with φ_n. Normalization of ψ implies

Text
Σ |c_n|² = 1

The measurement rules

  1. Possible results. A measurement of A returns one of the eigenvalues a_n.
  2. Probabilities. The probability of getting a_n is P_n = |c_n|². (If the eigenvalue is degenerate, add the |c|² of all its eigenfunctions.)
  3. Expectation value. ⟨A⟩ = Σ |c_n|² a_n. This equals ∫ ψ* Â ψ dx, which you can check by inserting the expansion and using orthonormality.
  4. State after measurement. Right after a measurement that gave a_n, the state is φ_n (the collapse or projection postulate). Measuring A again immediately gives a_n with certainty.

The spread of results is described by the variance (ΔA)² = ⟨A²⟩ - ⟨A⟩²; ΔA is the standard deviation. It is zero exactly when the state is an eigenfunction of Â.

Rule 4 is a postulate, and physicists disagree about its interpretation: the Copenhagen view takes it at face value, other views (many-worlds, de Broglie–Bohm, objective-collapse theories) explain the same observed statistics differently. The statistical predictions in rules 1–3 are the same for all of them and are confirmed by experiment. We will use the standard rules without taking a side.

Worked example

A particle in an infinite well is prepared in ψ = (3ψ₁ + 4ψ₂)/5. The coefficients are c₁ = 3/5 and c₂ = 4/5, and |c₁|² + |c₂|² = 9/25 + 16/25 = 1, so ψ is normalized.

  • Energy measurement gives E₁ with probability 9/25 = 0.36, and E₂ = 4E₁ with probability 16/25 = 0.64. Nothing else can occur.
  • ⟨E⟩ = 0.36 E₁ + 0.64 · 4E₁ = 2.92 E₁.
  • ⟨E²⟩ = 0.36 E₁² + 0.64 · 16 E₁² = 10.6 E₁², so (ΔE)² = 10.6 - 2.92² = 2.0736 in units of E₁², and ΔE = 1.44 E₁.

Time dependence. Each energy eigenstate evolves with the phase e^{-iE_n t/ħ}, so ψ(t) = Σ c_n e^{-iE_n t/ħ} φ_n. The coefficients' magnitudes |c_n| do not change, so the probabilities of energy outcomes and ⟨E⟩ are constant in time. Other quantities, such as ⟨x⟩, oscillate because the relative phases between the terms change.

Computing overlaps numerically

The first editor takes a smooth, not-an-eigenstate wavefunction of the box, the parabola ψ ∝ x(L − x), and finds the coefficients by numerical integration. The state looks a lot like the ground state, so |c₁|² should be close to 1. It checks Σ|c_n|² and compares ⟨E⟩ computed from the coefficients with ⟨E⟩ computed straight from ψ, using the identity ⟨E⟩ = (ħ²/2m) ∫ |ψ'|² dx (integration by parts, since ψ = 0 at the walls). We use units ħ = m = L = 1.

Python Expanding a state in box eigenfunctions

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The coefficients with even n vanish because the parabola is symmetric about the center of the box, while the even-n box states are antisymmetric about it, so their overlap integrals cancel. The ground state carries about 99.86% of the probability, but ⟨E⟩ is 5, while E₁ = π²/2 ≈ 4.93: the small admixture of higher states raises the average.

Simulating a measurement series

A single measurement is random, but the averages are not. The second editor simulates measuring the energy of many identically prepared particles in ψ = (3ψ₁ + 4ψ₂)/5. It is a statistical simulation: it draws random outcomes with the probabilities |c_n|², which are the quantum prediction, and shows that the sample average approaches 2.92 E₁ as the number of runs grows.

Python Simulated energy measurements

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

For small N the mean wanders (it can even be far from 2.92). For a million runs it matches to about three decimals. Each individual run gives exactly E₁ or 4E₁, never 2.92.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. A state is ψ = 0.6 φ₁ + 0.8 φ₂, where φ₁ and φ₂ are orthonormal eigenfunctions of  with eigenvalues 2 and 5. What is ⟨A⟩?

    P₁ = 0.6² = 0.36 and P₂ = 0.8² = 0.64, so ⟨A⟩ = 0.36·2 + 0.64·5 = 0.72 + 3.20 = 3.92.
  2. 2. Right after a measurement of A returns the eigenvalue a₃, you measure A again at once. What do you get?

    The state has collapsed to the eigenfunction φ₃, so a repeated measurement gives a₃ with probability 1.
  3. 3. For an isolated particle in the state c₁φ₁e^(−iE₁t/ħ) + c₂φ₂e^(−iE₂t/ħ), how do the energy-measurement probabilities change with time?

    The time factors are pure phases, so |c_n|² is constant. Only quantities that depend on the relative phase, such as ⟨x⟩, can oscillate.
  4. 4. What is the variance (ΔA)² in terms of expectation values?

    The variance is the mean of the square minus the square of the mean. It is zero only in an eigenstate of Â.

Summary

  • ⟨A⟩ = ∫ψ* Âψ dx is the average over many identical measurements.
  • Expand ψ = Σ c_n φ_n in eigenfunctions of Â; then P_n = |c_n|² is the probability of the eigenvalue a_n, and ⟨A⟩ = Σ|c_n|² a_n.
  • After a measurement giving a_n, the state is φ_n, so an immediate repeat gives a_n again.
  • Energy probabilities do not change in time for an isolated system; relative phases do, so other expectation values can.

Next: Commutators and the Uncertainty Principle