52
Part One · Quantum mechanics

Decoherence and the measurement problem

You never see your coffee cup in two places at once. Quantum mechanics has no rule against it, and nothing in the Schrödinger equation stops working at the size of a cup. What happens instead is that the air keeps measuring the cup. Every molecule that bounces off it carries away a little information about where it is, and within a tiny fraction of a second the superposition has spread into the room, beyond any hope of seeing it interfere. This is decoherence. It explains a great deal, and this chapter is careful to say exactly what it does not.

On the plate: a superposed object, and the environment’s molecules bouncing off it.

A model you can solve exactly

Take one qubit in the superposition (∣0⟩+∣1⟩)/2(|0\rangle + |1\rangle)/\sqrt2, and NN environment qubits, each starting in ∣0⟩|0\rangle. If the system is in ∣1⟩|1\rangle, each environment qubit is turned by a small angle ε\varepsilon; if it is in ∣0⟩|0\rangle, nothing happens. Afterwards the system and environment are entangled, and the system’s own coherence, the off-diagonal entry of its density matrix (Chapter 49), is multiplied by the overlap of the two environment states:

∣ρ01∣=12cos⁡N(ε/2)|\rho_{01}| = \tfrac12\cos^N(\varepsilon/2)

Set the environment to 20 qubits, each turned by 30°. The plate shows the interference fringes the system could still make. Each environment qubit learns only a little; together they learn almost everything, and the fringes are gone. No energy was exchanged, and nothing was disturbed in the usual sense. Established

How fast

Let the environment keep interacting, so the angles grow with time. For many weak interactions, cos⁡N\cos^N becomes a falling Gaussian or exponential: the coherence dies at a rate set by how much information each collision carries and how often collisions happen. For a distance Δx\Delta x between two positions of an object, smaller than the wavelength of what hits it, the rate is Λ Δx2\Lambda\,\Delta x^2, where Λ\Lambda counts scattering strength.

Choose an environment for a dust grain 10 µm across, superposed over its own size: , , or . The estimates, from Erich Joos and Dieter Zeh’s 1985 analysis, range from about a second in deep space to 10−2610^{-26} seconds in air. They are orders of magnitude, but the conclusion is robust: macroscopic superpositions are destroyed far faster than anything could notice them.

Why the world looks classical in one particular way

Decoherence does not erase every superposition equally. Air molecules interact with an object according to its position, so states of definite position are left alone and superpositions of different positions are destroyed. The environment selects a preferred set of states, the pointer states, which survive being watched. Wojciech Zurek called this einselection.

The plate shows it on the Bloch ball: the environment couples to zz. A state along zz stays put; a state along xx shrinks onto the zz-axis. That is why the classical world is made of objects with positions, and not of superpositions of positions: position is what the environment keeps asking about.

Watching it happen

In 1996 Serge Haroche’s group in Paris put a microwave field of a few photons into a superposition of two phases, a small “Schrödinger cat”, and watched its coherence disappear as photons leaked from the cavity, faster for larger separations between the two phases, as the theory predicts. Haroche shared the 2012 Nobel Prize. In 2003 Klaus Hornberger, Markus Arndt and colleagues in Vienna sent C₇₀ molecules through an interferometer and let in a trace of gas. The interference fringes faded exponentially with the gas pressure, at the rate decoherence theory predicts for each collision. Established

What decoherence does not do

After decoherence, the system’s density matrix is diagonal: it looks exactly like a classical mixture, “the cup is here or there, with these probabilities”. But it is not one. The whole of system plus environment is still in a single pure, entangled superposition; the mixture appears only because you ignore the environment (Chapter 49). Decoherence explains why you can never see the two branches interfere, and why the branches are defined by position. It does not explain why you find the cup in one place rather than the other, or why that one.

That remaining question is the measurement problem, and every interpretation in Chapter 54 answers it differently. Debated

What you now have

The exact dephasing model, decoherence rates and time scales, pointer states and einselection, the cavity and fullerene experiments, and a precise statement of what decoherence leaves unexplained.

· · ·

Three questions to carry with you.

Does decoherence require energy exchange?tap
No. It requires only that the environment become correlated with the system, carrying away which-state information.
Why are pointer states usually positions?tap
Most environmental interactions depend on position, so position states are the ones left unchanged.
Does decoherence solve the measurement problem?tap
Not by itself. It makes interference unobservable but does not explain why a single outcome occurs.

And one to work out. Twenty environment qubits are each turned by 30°. What fraction of the coherence remains, cos⁡20(15∘)\cos^{20}(15^\circ)?

fraction =

Sources: E. Joos and H. D. Zeh, Z. Phys. B 59, 223 (1985). W. H. Zurek, Rev. Mod. Phys. 75, 715 (2003). M. Schlosshauer, Decoherence and the Quantum-to-Classical Transition (2007), with its table of decoherence rates. M. Brune et al., Phys. Rev. Lett. 77, 4887 (1996). K. Hornberger et al., Phys. Rev. Lett. 90, 160401 (2003). Checked by code: exact N-qubit dephasing model gives |ρ₀₁| = ½cos^N(ε/2); cos²⁰(15°) = 0.4995.

Next · Chapter 53
Path integrals

The electron takes every path at once. Add up all the paths with the right phases and you get Newton.

Turn the page