Symmetry in Applications

Symmetry helps turn a large calculation into a smaller one with a physical interpretation. It can identify which states mix in a field, which spectral lines vanish, which scattering channels communicate, or which noise processes treat directions equally. The useful question is what symmetry constrains in the particular experiment: a Hamiltonian, a transition operator, or a dynamical map. This guide develops a few examples and points to the detailed applications.

Helpful background. Symmetry principles introduces transformations and representations. Continuous symmetries connects generators to conserved quantities, and tensor operators explains selection rules.

Choose a physical problem

Problem	Read next
Atomic multiplets, field shifts, and spectral lines	Atomic physics
Molecular rotations, vibrations, and electronic states	Molecular physics
Qubit rotations, Pauli operators, and invariant subspaces	Quantum information
Crystal symmetry, effective Hamiltonians, and band geometry	Condensed matter
Angular-momentum channels in collisions	Scattering
Covariant noise and reduced evolution	Open systems
Controlled reversals and small symmetry-breaking signals	Precision measurement

A fixed field changes which rotations are symmetries

Consider an atom with an ideal rotationally invariant Hamiltonian $H_0$ and add a model field term $-\omega J_z$. Here $\omega$ has units of angular frequency. The full Hamiltonian commutes with $J_z$ and $J^2$ but, for nonzero $\omega$, not with $J_x$ or $J_y$. In a simultaneous eigenbasis,

$$E_{\alpha jm}=E^{(0)}_{\alpha j}-\hbar\omega m.$$

The labels $j,m$ remain exact for this model while the degeneracy across $m$ is lifted. A general atomic Zeeman interaction couples separately to orbital and spin moments, so the simple $J_z$ form needs an appropriate regime or projection onto a multiplet. Coupling schemes explains the choice of basis as magnetic and spin–orbit scales change.

More generally, a transformation may relate different controls $\lambda$ through

$$U(g)H(\lambda)U(g)^{-1}=H(g\cdot\lambda).$$

This does not say that $U(g)$ commutes with the Hamiltonian at fixed $\lambda$. Only transformations leaving the actual Hamiltonian and its domain unchanged are symmetries of that fixed problem. Rotating a field along with an apparatus is different from rotating a state while the laboratory field stays fixed.

Symmetry restricts amplitudes but does not supply their size

For an electric-dipole transition, the probe operator is a rotational vector and is odd under parity. For states of definite parity $\eta_i,\eta_f\in\{+1,-1\}$,

$$\langle f|\mathbf d|i\rangle =-\eta_f\eta_i\langle f|\mathbf d|i\rangle.$$

Equal parities force the matrix element to vanish. Opposite parities permit it, subject to angular-momentum and other selection rules, but do not guarantee a nonzero value. The radial integral or another independent symmetry may still make the transition vanish.

The Wigner–Eckart theorem relates the angular components through coupling coefficients and leaves a reduced matrix element to calculate. That separation is especially useful for atomic spectra and molecular rotation spectra. Polarization specifies the component of the probe; preparation and populations determine which of the permitted transitions can be observed.

An analogous simplification occurs in spinless central-potential scattering. Rotational symmetry makes the scattering operator diagonal in $\ell,m$, with equal behavior for all $m$ at fixed $\ell$. In an elastic channel it has eigenvalues $S_\ell=e^{2i\delta_\ell}$, but the phase shifts $\delta_\ell$ still require dynamics. With several channels, symmetry leaves a matrix to solve within each permitted block. See partial waves and multichannel scattering.

A symmetry of noise need not conserve an observable

For an open system, test the physical channel $\mathcal E$:

$$\mathcal E(U\rho U^\dagger)=U\mathcal E(\rho)U^\dagger.$$

This covariance says that rotating before the noise has the same effect as rotating after it. It does not imply that angular momentum, spin polarization, or another expectation value is conserved.

For example, on qubit density matrices, the depolarizing channel

$$\mathcal E_p(\rho)=(1-p)\rho+pI/2,\qquad 0\le p\le1,$$

is covariant under every qubit unitary. Its action on a Bloch vector is $\mathbf r\mapsto(1-p)\mathbf r$: the direction is treated without preference, yet the polarization shrinks. For $p>0$, full rotational covariance plainly does not make every spin component conserved.

A symmetry of the total system–environment Hamiltonian is also not sufficient on its own. For a factorized initial state $\rho\otimes\eta$, covariance follows when the joint unitary commutes with $U_g\otimes V_g$ and the environment state satisfies $V_g\eta V_g^\dagger=\eta$. A polarized environment can select a direction even if the interaction is rotationally invariant. Open-system applications develops this argument. Individual Kraus operators can change under a symmetry while their complete channel remains covariant.

Small symmetry-breaking signals can appear without a linear energy shift

Let a nondegenerate eigenstate of $H_0$ have definite parity, and perturb it by $\lambda V$ with $PVP^{-1}=-V$. The diagonal matrix element $\langle n|V|n\rangle$ vanishes, so there is no first-order energy shift. Nevertheless the state can acquire an opposite-parity admixture at first order. An odd observable can detect interference between that admixture and the original state, producing a response linear in $\lambda$.

Thus a null energy shift need not mean a null physical effect. Conversely, reversing a field or spin does not automatically isolate one desired interaction: any systematic effect with the same reversal behavior can imitate it. A precision experiment combines the symmetry prediction with a dynamical model and experimental controls. Precision measurement gives the relevant examples.

For a first application, choose a small Hamiltonian, identify one exact symmetry, and calculate the restriction it imposes. Then add a perturbation or probe and ask what numerical information still needs to be computed. This keeps group labels connected to measurable quantities.

Exercises

1. Covariance of a qubit noise channel

Consider the depolarizing channel

$$\mathcal E_p(\rho) = (1-p)\rho+p\frac{I}{2}.$$

Show that it is covariant under every one-qubit unitary $U$. Why is the same statement false for amplitude damping with a preferred ground state?

Solution

Using $UIU^\dagger=I$,

$$\begin{aligned} \mathcal E_p(U\rho U^\dagger) &=(1-p)U\rho U^\dagger+p\frac{I}{2} \\ &=U\mathcal E_p(\rho)U^\dagger. \end{aligned}$$

The depolarizing channel therefore has full unitary covariance. Amplitude damping distinguishes the state called the ground state and the axis defined by its energy basis. A generic rotation changes that preferred state, so the channel is not covariant under all of $SU(2)$. It can retain covariance under the subgroup of phase rotations about the preferred axis.

2. What a vector selection rule does not determine

An operator $T^{(1)}_q$ transforms as a rank-one spherical tensor. State the angular-momentum conditions for a possibly nonzero matrix element between $|jm\rangle$ and $|j'm'\rangle$. Which part of the amplitude remains dynamical?

Solution

Angular-momentum addition requires

$$|j-1|\le j'\le j+1,$$

with $j+j'+1$ an integer: a rank-one tensor connects integer $j$ only to integer $j'$, and half-integer $j$ only to half-integer $j'$. The triangle condition excludes $j=0$ to $j'=0$. The spherical component fixes

$$m'=m+q, \qquad q\in\{-1,0,1\}.$$

These are necessary symmetry conditions. The Wigner–Eckart theorem factors the magnetic-quantum-number dependence from a reduced matrix element. That reduced matrix element contains the radial, internal, and dynamical information and may vanish for reasons not implied by angular momentum alone. Parity or other symmetries can impose additional conditions.

3. A parity-odd precision perturbation

Let $H_0$ commute with parity $P$, and let $|n\rangle$ be a nondegenerate parity eigenstate. A weak perturbation obeys

$$PVP^{-1}=-V.$$

Show that the first-order energy shift vanishes. Does the perturbation have no first-order physical effect?

Solution

If $P|n\rangle=\eta_n|n\rangle$ with $|\eta_n|=1$, then

$$\begin{aligned} \langle n|V|n\rangle &= \langle n|P^{-1}(PVP^{-1})P|n\rangle \\ &=-\langle n|V|n\rangle. \end{aligned}$$

Therefore

$$E_n^{(1)}=\langle n|V|n\rangle=0.$$

The perturbation can nevertheless mix $|n\rangle$ at first order with states of opposite parity. Interference involving that admixture can generate a linear symmetry-odd observable even though the nondegenerate energy shift begins at second order. This distinction underlies many null-test and induced-transition strategies.

References

• H.-P. Breuer and F. Petruccione, The Theory of Open Quantum Systems, Oxford University Press, 2002.
• P. R. Bunker and P. Jensen, Molecular Symmetry and Spectroscopy, 2nd ed., NRC Research Press, 1998.
• M. A. Nielsen and I. L. Chuang, Quantum Computation and Quantum Information, 10th anniversary ed., Cambridge University Press, 2010.
• J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics, 3rd ed., Cambridge University Press, 2020.
• J. R. Taylor, Scattering Theory: The Quantum Theory of Nonrelativistic Collisions, Wiley, 1972.
• M. Tinkham, Group Theory and Quantum Mechanics, McGraw–Hill, 1964.