Skip to content
Tutorials Center
My learning

Topics

Lesson 33 of 38 · Angular Momentum and Spin

The Stern–Gerlach Experiment and Spin Measurement

How the Stern–Gerlach experiment reveals quantized spin, why sequential measurements along different axes surprise us, and a Python simulation of spin measurements.

In 1922, Otto Stern and Walther Gerlach sent a beam of silver atoms through a magnetic field that was stronger on one side than the other, and recorded where the atoms landed on a detector plate. Classical physics predicted a smooth smear, because the atoms' magnetic moments should point in random directions. What they saw was two separate spots. The experiment showed directly that a component of an atom's magnetic moment can take only discrete values.

This lesson explains the experiment and then uses it as a thinking tool. A single Stern–Gerlach apparatus is the simplest physical realization of a quantum measurement, and chains of them lead to the most important surprise of quantum mechanics: measuring one property can erase knowledge of another.

The apparatus

A magnet shaped so that the field B_z changes with height exerts a force on an object with magnetic moment μ:

Text
F_z = μ_z ∂B_z/∂z

Atoms with a larger μ_z are pushed further. If μ_z could take any value, the beam would spread into a continuous band. If μ_z is quantized to two values, the beam splits into two.

A note on history: Stern and Gerlach's result was first interpreted in the framework of the older Bohr–Sommerfeld theory, and the electron spin hypothesis (Uhlenbeck and Goudsmit) only arrived in 1925. We now understand the splitting this way: the neutral silver atom has a single outer electron in an s state, so the orbital angular momentum is zero and the atom's magnetic moment comes from the electron spin. Its z-component takes just two values, matching S_z = ±ħ/2. The beam splits into two spots, up and down.

The real experiment needs care: the atoms must be neutral, so the force comes from the magnetic moment and not from the Lorentz force on a charge, and the field gradient must be large. For the thought experiments below we idealize it: an apparatus oriented along direction n̂ is a perfect device that sorts spin-1/2 particles into two beams, one with spin up along n̂ and one with spin down.

The measurement rules for spin

For a spin-1/2 particle with state |χ⟩, an apparatus along n̂ gives:

  • The result +ħ/2 with probability |⟨+n̂|χ⟩|² and −ħ/2 with probability |⟨−n̂|χ⟩|².
  • After the measurement the particle is in |+n̂⟩ or |−n̂⟩ (if you keep that beam).

If the particle is in the state |+m̂⟩ (up along a direction m̂) and the apparatus is along n̂, with angle γ between the two directions, then

Text
P(up along n̂) = cos²(γ/2),      P(down along n̂) = sin²(γ/2)

For γ = 0 (same direction) the result is certain: up. For γ = 180° it is certain: down. For γ = 90° it is 50/50.

Sequential measurements

Imagine three devices in a row, each blocking one of its output beams.

  1. z, then z. Send atoms through a z-device and keep only the "up" beam. A second z-device finds that 100% are up. The first measurement prepared the state, and the repeat confirms it.
  2. z, then x, then z. Keep "up along z" after the first device. The x-device splits them 50/50 into "+x" and "−x". Keep only "+x" atoms and send them into a second z-device. The result is again 50/50 up and down, even though all of these atoms were selected to be "up along z" a moment before.

The second case is the key observation. Naively, "spin up along z" should be a permanent fact, but measuring S_x destroys it. This follows from the fact that Ŝ_x and Ŝ_z do not commute ([Ŝₓ, Ŝ_z] = −iħŜᵧ): they cannot both have sharp values. The +x state is an equal-weight superposition (|↑⟩ + |↓⟩)/√2 of the z-states, so a z-measurement of it gives either result with probability 1/2.

A frequent mistake is to think that the atom "really had" a definite S_z which got "disturbed" by the x-device in some uncontrolled way. The mathematics of quantum mechanics does not give the atom a definite S_z and S_x at the same time. (Hidden-variable models that do assign them must have strange nonlocal features, as Bell's theorem shows; that is a later topic.)

Interference: two beams recombined

If, instead of blocking the "−x" beam, you recombine both x-beams coherently, without measuring which path the atom took, the state is restored to "up along z" (a pure superposition of +x and −x, summed, returns |↑⟩) and the next z-device finds 100% up. Distinguishing the paths destroys the interference; leaving them indistinguishable keeps it. This is a spin version of the double-slit logic.

A simulation

The first editor models an idealized Stern–Gerlach chain. A function measure(state, axis, rng) computes the probabilities from the projection rule, randomly picks an outcome, and returns the post-measurement state. We then run the z → x → z chain on 100,000 atoms and count how many survive each stage. This is a simulation of the quantum rule, not an experiment; it uses a random number generator to sample outcomes with the quantum probabilities.

Python Simulating a z, x, z Stern–Gerlach chain

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 last line is the surprise: although every atom entering the last device was selected as "up along z" earlier, only half come out up.

The general angle rule

How does the survival probability depend on the angle between two measurement axes? Start with spin up along z and measure along an axis tilted by γ. The theory predicts P = cos²(γ/2). The second editor checks it by Monte Carlo sampling, and then shows how inserting an intermediate measurement can restore signal: sending atoms from "up along z" to "down along z" directly gives zero, but adding a middle device at 90° lets one quarter of them through (probability 1/2 at each of the two steps).

Python Angle rule and an intermediate filter

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.

Adding a filter increases what gets through at the end. In optics, the analogous experiment is the classic polarizer trick with three crossed polarizing filters.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. What did the Stern–Gerlach experiment show about silver atoms in a non-uniform magnetic field?

    Two spots appeared: the magnetic moment component along the field gradient is quantized to two values, as expected for the spin-1/2 electron.
  2. 2. Atoms are selected 'up along z', then sent through an x-device keeping '+x', then a second z-device. What fraction exits 'up along z'?

    The +x state is an equal superposition of z-up and z-down, so a z-measurement gives each with probability 1/2. Measuring S_x erased the sharp value of S_z.
  3. 3. A spin-1/2 particle is up along z. An apparatus is rotated by 60° from z. What is the probability of 'up' along the new axis?

    P = cos²(γ/2) with γ = 60° gives cos²(30°) = 3/4.
  4. 4. Why is spin the explanation for two spots in the silver experiment, and not orbital angular momentum?

    For l = 0 there is a single m state, so an orbital moment would give one undeflected beam. The two beams come from spin 1/2.

Summary

  • A Stern–Gerlach apparatus sorts spin-1/2 particles into two beams according to the spin component along its field direction.
  • Measurement rule: the probability of up along n̂ for a particle prepared up along m̂ is cos²(γ/2), where γ is the angle between the axes.
  • Measuring S_x after S_z randomizes S_z; non-commuting observables cannot both be sharp.
  • Recombining the beams without distinguishing paths preserves coherence; recording the path destroys it.
  • Simulations sample outcomes with these probabilities; they reproduce the statistics but do not explain "why this outcome".

Next: Adding Angular Momenta