Continuous Symmetries and Conservation Laws

A continuous quantum symmetry can be varied smoothly from the identity. That seemingly modest property changes the problem: the transformation can be differentiated, its local action is encoded by a self-adjoint generator, and Hamiltonian invariance becomes a commutator condition. Conservation laws, good quantum numbers, and many selection rules then follow from one structural chain.

Continuous unitary family
→ self-adjoint generator
→ infinitesimal commutator action
→ Hamiltonian invariance test
→ conserved observable
→ symmetry sectors, stable labels, and selection rules

This chapter develops that chain and records where its shortcuts can fail. It also separates the ordinary quantum-mechanical statement from the stronger classical and field-theoretic forms of Noether’s theorem.

Canonical Boundaries

This page owns the chapter map, the shared workflow, and the relation among the chapter’s results. The detailed arguments remain at their canonical homes.

Question	Canonical home	Role here
what counts as a continuous unitary family?	One-Parameter Unitary Groups	states the group and continuity assumptions
how is the generator extracted?	Generators	defines the derivative at the identity
why do commutators appear?	Infinitesimal Transformations	derives the first-order action
when does symmetry imply conservation?	Commutators and Conservation Laws	gives the operator derivation
what is the quantum Noether statement?	Quantum Noether Principle	distinguishes the QM statement from the full theorem
how is a proposed constant tested?	Constants of Motion	treats explicit time dependence and strength of conservation
when is an eigenvalue a useful label?	Good Quantum Numbers	connects commuting observables to state labels
why must some matrix elements vanish?	Selection Rules Preview	gives the symmetry-sector argument
what changes when symmetry is absent or reduced?	Broken Symmetry Preview	separates explicit breaking from nonsymmetric states

The rigorous spectral and domain statement for a single continuous unitary parameter belongs to Stone Theorem. Detailed tensor-operator selection rules belong to the later Selection Rules chapter. Spontaneous breaking in extended systems is only previewed here.

The Structural Chain

Let $U(\alpha)$ be a one-parameter unitary group. Its defining algebraic conditions are

$$\begin{aligned} U(0)&=I, \\ U(\alpha+\beta)&=U(\alpha)U(\beta), \\ U(\alpha)^{-1}&=U(-\alpha). \end{aligned}$$

Continuity near the identity allows the family to be represented by a self-adjoint generator $G$:

$$U(\alpha) = \exp\!\left(-\frac{i\alpha G}{\hbar}\right).$$

Near $\alpha=0$,

$$U(\delta\alpha) = I-\frac{i}{\hbar}\delta\alpha\,G +O(\delta\alpha^2).$$

If the Hamiltonian is invariant,

$$U(\alpha)HU^\dagger(\alpha)=H,$$

then differentiating at the identity gives

$$[G,H]=0.$$

When $G$ has no explicit time dependence, the Heisenberg equation consequently gives

$$\frac{dG_H}{dt}=0.$$

Every arrow in this chain has hypotheses. A family can be unitary without being a symmetry of the chosen Hamiltonian. A conserved operator can exist without having been identified as the generator of a stated symmetry. A vanishing expectation-value derivative in one state is weaker than an operator conservation law.

Continuity and One-Parameter Groups

The parameter $\alpha$ may be time, displacement, rotation angle, or an internal phase. Its units depend on the transformation, but the exponent must be dimensionless:

$$\frac{\alpha G}{\hbar} \quad\text{is dimensionless}.$$

The group law says that applying two transformations is equivalent to adding their parameters. It is not merely an exponential ansatz. For translations,

$$T(a+b)=T(a)T(b),$$

while time-independent evolution satisfies

$$U(t+s)=U(t)U(s).$$

In finite dimensions, matrix continuity is sufficient for the usual manipulations. On an infinite-dimensional Hilbert space, strong continuity is the standard assumption:

$$\lim_{\alpha\to\alpha_0} \left\| U(\alpha)|\psi\rangle -U(\alpha_0)|\psi\rangle \right\| =0$$

for every $|\psi\rangle$ in the Hilbert space. The generator may be unbounded, so its domain matters even though every $U(\alpha)$ is bounded and unitary.

There is also a global caution. A local generator does not by itself settle periodicity, topology, or whether a representation is ordinary or projective. Rotation angles, spinor signs under $2\pi$ rotations, and compact internal symmetries require the global group structure as well as the infinitesimal algebra.

Generators as Tangent Observables

The generator is the derivative of the unitary family at the identity:

$$G = i\hbar \left. \frac{dU}{d\alpha} \right|_{\alpha=0}.$$

With the convention used throughout this volume,

Transformation	Parameter	Unitary family	Generator
time translation	$t$	$e^{-itH/\hbar}$	Hamiltonian $H$
spatial translation	$a$	$e^{-iaP/\hbar}$	momentum $P$
rotation about $\mathbf n$	$\theta$	$e^{-i\theta\mathbf n\cdot\mathbf J/\hbar}$	$\mathbf n\cdot\mathbf J$
internal phase rotation	$\alpha$	$e^{-i\alpha Q/\hbar}$	charge-like observable $Q$

Self-adjointness is essential. It makes the exponential unitary and gives the generator a real spectral measure, allowing it to serve as an observable. In finite dimensions, first-order unitarity already yields the familiar check:

$$U^\dagger U = I +\frac{i\alpha}{\hbar}(G^\dagger-G) +O(\alpha^2),$$

so $G^\dagger=G$ at first order. For unbounded generators, this calculation is a mnemonic rather than a replacement for the self-adjoint domain analysis.

Infinitesimal Actions and Sign Conventions

For an active transformation of a state,

$$|\psi\rangle \longmapsto U(\alpha)|\psi\rangle,$$

the first-order change is

$$\delta|\psi\rangle = -\frac{i}{\hbar}\delta\alpha\,G|\psi\rangle.$$

With the active operator convention $A\mapsto UAU^\dagger$,

$$\delta A = -\frac{i}{\hbar}\delta\alpha\,[G,A].$$

With the conjugation $A\mapsto U^\dagger A U$, used when the transformed state is pulled back to the original state, the sign is reversed:

$$\delta A = \frac{i}{\hbar}\delta\alpha\,[G,A].$$

These formulas do not conflict. They answer different transformation questions. Before trusting a sign, state what changes physically, what is held fixed, and whether the operator is conjugated by $U$ or $U^\dagger$.

The commutator is the infinitesimal action of the generator on operators. In Lie-algebra language, it is the local algebraic structure behind the smooth group action. For a transformation far from the identity, use the finite exponential or an appropriate Baker–Campbell–Hausdorff expansion; the first-order formula alone is not exact.

Hamiltonian Invariance

A unitary family becomes a dynamical symmetry only after it passes the Hamiltonian test. For a time-independent family,

$$U(\alpha)HU^\dagger(\alpha)=H$$

is equivalent, under the usual regularity assumptions, to

$$[H,G]=0.$$

The two directions should be kept distinct in a derivation:

1. Finite invariance differentiated at $\alpha=0$ gives the commutator condition.
2. If a self-adjoint $G$ commutes appropriately with $H$, its exponential commutes with $H$ and generates a finite invariance.

For unbounded operators, statements such as $[H,G]=0$ may need domain qualification or the stronger language of commuting spectral measures. The formal commutator is reliable in the standard finite-dimensional and common textbook domains, but it should not be promoted into a domain-free theorem.

Conservation and the Quantum Noether Principle

For a possibly time-dependent Schrödinger-picture observable $A(t)$,

$$\frac{d}{dt}\langle A\rangle = \frac{i}{\hbar}\langle[H,A]\rangle + \left\langle \frac{\partial A}{\partial t} \right\rangle.$$

The corresponding operator criterion for a constant of motion is

$$\frac{\partial A}{\partial t} + \frac{i}{\hbar}[H,A] =0.$$

Thus $[H,A]=0$ is sufficient only when $A$ has no explicit time dependence. Conversely, an explicitly time-dependent $A(t)$ can still be conserved when its explicit derivative cancels its commutator term.

For a continuous unitary symmetry generated by $G$, the ordinary quantum Noether pattern is

$$\begin{aligned} U(\alpha)HU^\dagger(\alpha)&=H \\ &\Downarrow \\ [G,H]&=0 \\ &\Downarrow \\ \frac{dG_H}{dt}&=0, \end{aligned}$$

where the last implication assumes $\partial G/\partial t=0$.

This is not the full field-theory theorem. Classical Noether theory begins with an action and treats boundary terms. Field theory produces a local current $j^\mu$ and a continuity equation $\partial_\mu j^\mu=0$ before a conserved charge is obtained. The ordinary quantum-mechanical chapter stays with Hilbert-space transformations, Hamiltonians, generators, and commutators. The field-theory bridge begins at From Quantum Generators to Noether Currents.

What Conservation Means

Several statements are often compressed into the word “conserved.”

Statement	Scope	What it establishes
$d\langle A\rangle/dt=0$ in one state	state dependent	one mean value is stationary
$d\langle A\rangle/dt=0$ for all states	dynamical	all expectation values are stationary
$dA_H/dt=0$	operator level	the Heisenberg observable is fixed
the spectral projectors of $A$ are preserved	distribution level	every measurement probability for $A$ is fixed

A conserved observable need not have a definite value. A wave packet can have a nonzero spread in momentum while the entire momentum distribution remains unchanged. Conservation also does not mean that $A$ commutes with every observable, only that it satisfies the relevant dynamical criterion.

Good Quantum Numbers

If a time-independent self-adjoint operator $Q$ commutes with $H$, the Hamiltonian preserves each eigenspace of $Q$. Energy eigenstates can then be chosen with a $Q$ label:

$$\begin{aligned} H|E,q,\lambda\rangle &=E|E,q,\lambda\rangle, \\ Q|E,q,\lambda\rangle &=q|E,q,\lambda\rangle. \end{aligned}$$

The extra label $\lambda$ matters. A good quantum number need not form a complete label. Several independent states may share the same $E$ and $q$.

For several proposed labels $Q_1,\ldots,Q_r$, it is not enough that each commute with $H$. They must also be mutually compatible if states are to carry all labels sharply:

$$[Q_i,Q_j]=0 \qquad \text{for all }i,j.$$

This explains why $L^2$ and one component such as $L_z$ can label central-potential states, while $L_x$, $L_y$, and $L_z$ cannot all be sharp. It also explains why changing the Hamiltonian can change the preferred labels. Spin–orbit coupling, external fields, or crystal anisotropy may preserve only a smaller commuting set.

Selection Rules from Symmetry Labels

Let $Q$ generate a continuous symmetry, and suppose

$$Q|i\rangle=q_i|i\rangle, \qquad Q|f\rangle=q_f|f\rangle.$$

If a transition operator $T$ has a definite commutator with $Q$,

$$[Q,T]=q_T T,$$

then

$$\begin{aligned} (q_f-q_i)\langle f|T|i\rangle &= \langle f|[Q,T]|i\rangle \\ &= q_T\langle f|T|i\rangle. \end{aligned}$$

A nonzero matrix element therefore requires

$$q_f-q_i=q_T.$$

This is a symmetry constraint, not a dynamical prediction of the transition rate. Satisfying the condition is generally necessary, not sufficient: radial integrals, additional symmetries, kinematics, and accidental cancellations can still make the matrix element vanish.

If the Hamiltonian is perturbed so that $Q$ is no longer conserved, the old rule may become approximate. The size of a formerly forbidden amplitude then depends on how the perturbation mixes the old symmetry sectors.

Exact, Reduced, and Broken Symmetry

Suppose

$$H=H_0+\lambda V, \qquad [H_0,G]=0.$$

If $[V,G]\ne0$, then

$$[H,G]=\lambda[V,G],$$

so the full Hamiltonian does not have the original symmetry for nonzero $\lambda$. The old generator is not exactly conserved, old multiplets may split, and old selection rules may be violated.

That is explicit breaking. It must not be confused with choosing a state that is not invariant even though the Hamiltonian is. Nor should every nonsymmetric state in a finite system be called a spontaneously broken phase. Spontaneous symmetry breaking requires additional structure, typically a many-body or infinite-volume limit and a family of stable states or phases. The Broken Symmetry Preview establishes these distinctions without relocating their many-body canonical treatment.

Residual symmetry is often more useful than saying simply that a symmetry is gone. A magnetic field may reduce full rotation symmetry to rotations about one axis, leaving one angular-momentum component conserved even though the others are not.

Three Standard Examples

Translation invariance

For a particle on the line,

$$T(a)=e^{-iaP/\hbar}.$$

If $T(a)HT^\dagger(a)=H$ for every $a$, then $[H,P]=0$ and momentum is conserved. For

$$H=\frac{P^2}{2m}+V(X),$$

this requires $V$ to be translation invariant. A generic position-dependent potential breaks the symmetry. The detailed wavefunction and operator derivations belong to Translations and Momentum.

Rotational invariance

For rotations about the unit vector $\mathbf n$,

$$R_{\mathbf n}(\theta) = \exp\!\left( -\frac{i\theta}{\hbar}\mathbf n\cdot\mathbf J \right).$$

If the Hamiltonian is invariant under every spatial rotation, then

$$[H,J_i]=0, \qquad i=x,y,z.$$

The three generators do not commute with one another, so they cannot all label a state sharply. The compatible pair $J^2,J_z$ is used instead. This distinction between symmetry generators and a commuting labeling set is central to angular momentum.

Time translations

For a time-independent Hamiltonian,

$$U(t)=e^{-itH/\hbar}.$$

The Hamiltonian generates time evolution and is trivially conserved because $[H,H]=0$. If $H(t)$ depends explicitly on time, the evolution operator is generally not a one-parameter group depending only on elapsed time, and energy need not be conserved. Time ordering and two-time propagators then replace the simple group law.

A Reliable Problem-Solving Workflow

1. Identify the transformation. State what happens to states, observables, coordinates, or external fields.
2. Write the group law. Check whether the parameter is additive, periodic, multidimensional, or discrete.
3. Fix the convention. Record whether operators transform as $UAU^\dagger$ or $U^\dagger A U$.
4. Find the generator. Differentiate at the identity and check the dimensions of $\alpha G/\hbar$.
5. Test the Hamiltonian. Evaluate finite invariance or compute $[H,G]$ with the necessary domain assumptions.
6. Include explicit time dependence. Use the full constants-of-motion equation, not only the commutator shortcut.
7. Choose compatible labels. Distinguish conserved generators from a mutually commuting set of observables.
8. Extract consequences. Determine protected sectors, degeneracies, selection rules, or residual symmetries.
9. State the approximation. If a perturbation breaks the symmetry, say which conclusions are exact and which are approximate.

The order prevents a common reversal: guessing a conserved quantity from familiar notation and only afterward asking whether the stated Hamiltonian actually has the required symmetry.

Chapter Map

Page	Central question	Best use
One-Parameter Unitary Groups	what analytic structure connects finite transformations to a generator?	start here for the chapter’s mathematical object
Generators	how is the observable $G$ extracted and interpreted?	translation, rotation, and time-evolution examples
Infinitesimal Transformations	how does $G$ act on states and operators?	sign conventions and commutator derivations
Commutators and Conservation Laws	why does invariance imply conservation?	the core operator argument
Quantum Noether Principle	what is the precise QM analogue of Noether’s theorem?	conceptual framing and QFT bridge
Constants of Motion	how is conservation tested in practice?	explicit time dependence and strength of statements
Good Quantum Numbers	when do conserved observables label states?	simultaneous eigenstates and incomplete labels
Selection Rules Preview	how do symmetry labels force matrix elements to vanish?	preparation for tensor operators
Broken Symmetry Preview	what survives when the Hamiltonian or state lacks the full symmetry?	exact, approximate, residual, and spontaneous cases

Reading Paths

First pass through the logic

1. One-Parameter Unitary Groups
2. Generators
3. Infinitesimal Transformations
4. Commutators and Conservation Laws
5. Good Quantum Numbers

For practical spectral calculations

1. Constants of Motion
2. Good Quantum Numbers
3. Selection Rules Preview
4. Angular Momentum Operators

For mathematical control

1. Lie Groups
2. Lie Algebras
3. One-Parameter Unitary Groups
4. Stone Theorem

For the field-theory bridge

1. Quantum Noether Principle
2. From Quantum Generators to Noether Currents
3. Selection Rules to Ward Identities

Distinctions Worth Keeping

Do not identify	With	Reason
continuous unitary family	symmetry of $H$	invariance must still be tested
Hermitian-looking differential expression	self-adjoint generator	domains and boundary conditions matter
conserved expectation in one state	operator constant of motion	the former is state dependent
conserved observable	sharp observable	a conserved distribution can have nonzero variance
good quantum number	complete quantum-number set	degeneracy may remain
symmetry-allowed matrix element	nonzero matrix element	symmetry gives necessary constraints, not full dynamics
nonsymmetric state	spontaneously broken phase	the latter needs stability and an appropriate large-system limit
infinitesimal algebra	global group	topology and projective phases are global data

Common Mistakes

• Omitting the continuity assumption before introducing a generator.
• Calling any unitary change of basis a physical continuous symmetry.
• Forgetting the factor of $\hbar$ or giving $\alpha G/\hbar$ nonzero dimensions.
• Switching between $UAU^\dagger$ and $U^\dagger A U$ without changing the sign of the infinitesimal commutator.
• Using $[H,A]=0$ when $A$ has explicit time dependence.
• Assuming all generators of a non-Abelian symmetry can be simultaneously diagonalized.
• Treating a good label as unique or complete without checking degeneracy.
• Reading a selection rule as a prediction of transition strength.
• Calling small explicit breaking spontaneous symmetry breaking.
• Ignoring domains when an unbounded generator or Hamiltonian is involved.

Cross-Links

• Symmetry Principles
• Active and Passive Transformations
• Symmetry Constraints on Hamiltonians
• Spatial Symmetries
• Translations and Momentum
• Angular Momentum Operators
• XXZ Spin Chain
• Compatible Observables
• Conservation Laws
• Heisenberg Equations of Motion
• Noether Theorem in Quantum Mechanics

References

• M. H. Stone, “On One-Parameter Unitary Groups in Hilbert Space,” Annals of Mathematics 33, 643–648, 1932.
• P. A. M. Dirac, The Principles of Quantum Mechanics, 4th ed., Oxford University Press, 1958.
• E. P. Wigner, Group Theory and Its Application to the Quantum Mechanics of Atomic Spectra, Academic Press, 1959.
• M. Reed and B. Simon, Methods of Modern Mathematical Physics I: Functional Analysis, revised and enlarged ed., Academic Press, 1980.
• R. Shankar, Principles of Quantum Mechanics, 2nd ed., Springer, 1994.
• J. J. Sakurai and J. Napolitano, Modern Quantum Mechanics, 3rd ed., Cambridge University Press, 2020.
• B. C. Hall, Quantum Theory for Mathematicians, Springer, 2013.
• E. Noether, “Invariante Variationsprobleme,” Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen, 235–257, 1918.

Exercises

1. Let $U(\alpha)=e^{-i\alpha G/\hbar}$, and suppose $U(\alpha)HU^\dagger(\alpha)=H$ for every $\alpha$. Derive the generator commutator condition.

Solution

Differentiate the invariance condition:

$$\left. \frac{d}{d\alpha} \left(UHU^\dagger\right) \right|_{\alpha=0} =0.$$

At the identity,

$$\left.\frac{dU}{d\alpha}\right|_0 =-\frac{iG}{\hbar}, \qquad \left.\frac{dU^\dagger}{d\alpha}\right|_0 =\frac{iG}{\hbar}.$$

Therefore

$$-\frac{i}{\hbar}GH +\frac{i}{\hbar}HG =0,$$

which is $[G,H]=0$, equivalently $[H,G]=0$.

2. A particle has $H=P^2/(2m)+\tfrac12m\omega^2X^2$. Is spatial translation generated by $P$ a symmetry? Evaluate the relevant commutator.

Solution

Using $[X,P]=i\hbar$,

$$\begin{aligned} [P,X^2] &= [P,X]X \\ &\quad+X[P,X] \\ &=-2i\hbar X. \end{aligned}$$

Hence

$$\begin{aligned} [P,H] &= \frac12m\omega^2[P,X^2] \\ &= -i\hbar m\omega^2X. \end{aligned}$$

For $\omega\ne0$ this is nonzero, so translations are not a symmetry and momentum is not conserved. The oscillator center selects a preferred position.

3. Show that the explicitly time-dependent operator $A(t)=X-tP/m$ is a constant of motion for a free particle with $H=P^2/(2m)$.

Solution

The explicit derivative is

$$\frac{\partial A}{\partial t} =-\frac{P}{m}.$$

Because $[P^2,X]=-2i\hbar P$,

$$\frac{i}{\hbar}[H,A] = \frac{i}{\hbar} \left[\frac{P^2}{2m},X\right] = \frac{P}{m}.$$

The two terms cancel:

$$\frac{\partial A}{\partial t} + \frac{i}{\hbar}[H,A] =0.$$

Thus $A_H$ is constant even though the Schrödinger-picture expression has explicit time dependence. This is why the commutator-only shortcut would be insufficient.

4. Suppose $Q|i\rangle=2\hbar|i\rangle$, $Q|f\rangle=5\hbar|f\rangle$, and $[Q,T]=3\hbar T$. Does the $Q$ selection rule force $\langle f|T|i\rangle$ to vanish?

Solution

The necessary condition for a nonzero matrix element is

$$q_f-q_i=q_T.$$

Here

$$q_f-q_i = 5\hbar-2\hbar =3\hbar =q_T.$$

The matrix element is symmetry allowed. The selection rule does not prove that it is nonzero; dynamics or another symmetry may still make it vanish.

5. A Hamiltonian is rotationally invariant, so $[H,J_i]=0$ for $i=x,y,z$. Explain why $J_x,J_y,J_z$ are not three simultaneous good quantum numbers, and name a standard compatible set.

Solution

The angular-momentum components obey

$$[J_i,J_j] = i\hbar\epsilon_{ijk}J_k.$$

They are individually conserved by a rotationally invariant Hamiltonian, but they do not commute with one another. They therefore cannot all have sharp values in a common basis. Since

$$[J^2,J_z]=0,$$

a standard compatible labeling set is $H,J^2,J_z$, supplemented by any additional commuting observables needed to resolve degeneracy.