29
Part One · Quantum mechanics

Operators, observables, commutators

Measure a particle’s position and then its momentum, or its momentum and then its position: in everyday life the order would not matter. In quantum mechanics it does, and the difference is not vague. It is exactly iℏi\hbar, a fixed number, the same for every particle in the universe.

This chapter is about operators, the objects that represent measurable quantities, and about their commutators, which measure how much order matters. One commutator, [x^,p^]=iℏ[\hat x, \hat p] = i\hbar, is the seed of most of what is strange about quantum physics.

Three kinds of operator

Hermitian operators, A^†=A^\hat A^\dagger = \hat A, represent quantities you can measure: real eigenvalues, perpendicular eigenvectors. Unitary operators, U^†U^=1\hat U^\dagger\hat U = \mathbb 1, represent changes that preserve probability: rotations, time evolution, quantum gates. Projectors, P^2=P^\hat P^2 = \hat P, represent yes-or-no questions: is the particle in this region, is the spin up?

Unitary operators are exponentials of Hermitian ones, U^=e−iA^s\hat U = e^{-i\hat A s}, and the Hermitian operator is called the generator of the change (Chapter 20). Time evolution is generated by energy, rotations by angular momentum and translations by momentum.

Momentum from translations

Shift a wavefunction to the right by a distance aa: ψ(x)→ψ(x−a)\psi(x) \to \psi(x - a). For a small shift, Taylor’s theorem (Chapter 2) gives ψ(x)−a ψ′(x)\psi(x) - a\,\psi'(x), so the generator of translations is a derivative. Calling it p^/ℏ\hat p/\hbar defines momentum:

p^=−iℏ∂∂x,e−iap^/ℏ ψ(x)=ψ(x−a)\hat p = -i\hbar\frac{\partial}{\partial x}, \qquad e^{-ia\hat p/\hbar}\,\psi(x) = \psi(x - a)

Shift the packet by 1.0. The momentum operator is the infinitesimal version of that move.

The commutator

Apply x^\hat x then p^\hat p to any function, and then the other way round:

[x^,p^] ψ=x(−iℏψ′)−(−iℏ (xψ)′)=iℏ ψ[\hat x, \hat p]\,\psi = x\left(-i\hbar\psi'\right) - \left(-i\hbar\,(x\psi)'\right) = i\hbar\,\psi

The difference is iℏi\hbar times the function itself, whatever the function. So [x^,p^]=iℏ[\hat x, \hat p] = i\hbar, Dirac’s replacement for the classical {x,p}=1\{x, p\} = 1 (Chapter 13). Established

The plate shows it geometrically. Shift a packet in position, then boost it in momentum, and compare with boosting first and then shifting. You land on the same state up to a phase, eiap0/ℏe^{iap_0/\hbar}, that comes directly from the commutator.

No finite matrices can do this

Could position and momentum be finite matrices, like spin? Take the trace, the sum of the diagonal entries, of both sides of [X,P]=iℏ1[X, P] = i\hbar\mathbb 1. The trace of any commutator is zero, because tr(XP)=tr(PX)\mathrm{tr}(XP) = \mathrm{tr}(PX). The trace of iℏ1i\hbar\mathbb 1 is iℏNi\hbar N. They cannot be equal, so position and momentum need infinitely many dimensions.

Try to cheat with 6×6 matrices cut from the harmonic oscillator. The plate shows the diagonal of [X,P]/iℏ[X, P]/i\hbar: ones all the way down, as they should be, until the last entry, which is −(N−1)-(N-1) to make the trace vanish. The error never goes away; it just moves to the edge.

Multiplying exponentials

For numbers, eAeB=eA+Be^Ae^B = e^{A+B}. For operators that do not commute, it fails. When the commutator [A,B][A, B] commutes with both, which is true for x^\hat x and p^\hat p, there is a clean replacement, the Baker–Campbell–Hausdorff formula:

eAeB=eA+B+12[A,B]e^{A}e^{B} = e^{A + B + \frac12[A,B]}

It is the extra phase you saw on the plate two passages ago. In general there are more terms, nested commutators, and BCH is how quantum computers estimate the error of chopping a simulation into small time steps (Part Three).

Angular momentum

The components of angular momentum do not commute either:

[L^x,L^y]=iℏ L^zand cyclic[\hat L_x, \hat L_y] = i\hbar\,\hat L_z \quad\text{and cyclic}

So no state can have definite values of two components at once (except zero). That is why a spin measured along zz has an undetermined xx-component, as you saw in Chapter 27. But L^2\hat L^2 commutes with every component, so a state can have definite total angular momentum and one definite component. Atomic orbitals are labelled by exactly those two numbers, ℓ\ell and mm.

What you now have

Hermitian, unitary and projection operators, momentum as the generator of translations, [x^,p^]=iℏ[\hat x,\hat p] = i\hbar and the trace argument, BCH, and the angular-momentum algebra.

· · ·

Three questions to carry with you.

Why is p̂ = −iħ ∂/∂x?tap
Momentum generates translations, and the generator of an infinitesimal shift ψ(x) → ψ(x − a) is the derivative.
Why can’t [X, P] = iħ hold for finite matrices?tap
The trace of a commutator is zero but the trace of iħ times the identity is iħN.
Which angular-momentum quantities can be definite together?tap
L² and one component, usually L_z, since they commute.

And one to work out. Using [x^,p^]=iℏ[\hat x,\hat p] = i\hbar, [x^,p^2]=c iℏ p^[\hat x, \hat p^2] = c\,i\hbar\,\hat p. What is cc?

c =

Sources: J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics. B. C. Hall, Quantum Theory for Mathematicians (Stone–von Neumann theorem). Checked by code: truncated oscillator [X, P] diagonal (1, …, 1, −(N−1)), trace 0; BCH for the Heisenberg algebra; [L_x, L_y] = iL_z.

Next · Chapter 30
The uncertainty principle

It isn't that we can't measure carefully enough. A particle with a sharp position and a sharp momentum doesn't exist.

Turn the page