Symmetries in quantum mechanics

Ask a physicist why momentum is conserved and the deepest honest answer is not a force law but a fact about empty space: nothing in the physics changes if you slide the whole experiment sideways. That single sentence — the dynamics does not care where you put the origin — turns out to be equivalent to $\frac{d}{dt}\langle\hat{p}\rangle=0$ , and the machine that converts one statement into the other is the theory of symmetries in quantum mechanics. In classical mechanics the bridge is Noether's theorem, phrased in the language of Lagrangians. In quantum mechanics the bridge is sharper and, in a sense, more elementary, because it is built directly from the two things a quantum theory cannot do without: the inner product that defines probabilities, and the Hamiltonian that defines time evolution. The argument is short enough to state in one breath. A symmetry is any transformation $\psi\to\psi'=\hat{U}\psi$ that leaves every measurable prediction unchanged. Preserving predictions means preserving the norm, which forces $\hat{U}$ to be unitary. If the Hamiltonian also respects the symmetry, then $\hat{U}$ commutes with it, $[\hat{H},\hat{U}]=0$ . Any continuous family of such unitaries is generated by a Hermitian operator $\hat{G}$ through $\hat{U}=e^{i\alpha\hat{G}}$ , and because $\hat{U}$ commutes with $\hat{H}$ so does $\hat{G}$ — whereupon the Heisenberg equation delivers $\frac{d}{dt}\langle\hat{G}\rangle=0$ . The generator is a conserved observable. Symmetry forces unitarity, unitarity supplies a Hermitian generator, the generator is a conservation law. The same reasoning, applied to internal transformations rather than motions in space, is the seed of isospin, the Eightfold Way, and the entire gauge structure of the Standard Model.

Key takeaways

A symmetry is a transformation $\psi\to\hat{U}\psi$ that leaves all predictions invariant; preserving the norm forces $\hat{U}$ to be unitary, $\hat{U}^\dagger\hat{U}=I$ .
A symmetry of the dynamics means $\hat{U}$ commutes with the Hamiltonian, $[\hat{H},\hat{U}]=0$ , so it maps energy eigenstates to degenerate energy eigenstates.
Continuous symmetries exponentiate a Hermitian generator: $\hat{U}(\alpha)=e^{i\alpha\hat{G}}$ with $\hat{G}=\hat{G}^\dagger$ — an observable.
That generator is conserved: $\frac{d}{dt}\langle\hat{G}\rangle=i\langle[\hat{H},\hat{G}]\rangle=0$ — the quantum form of Noether's theorem. Translations are generated by momentum, rotations by angular momentum, internal phases by charge.
Symmetry operations close into a group: the central ones are $U(1)$ , $SU(2)$ , $SU(3)$ , whose $n^2-1$ generators become the force-carrying gauge bosons.
Key idea: every symmetry of the Hamiltonian is a unitary operator that commutes with it, and its Hermitian generator is automatically a conserved observable — one chain of logic that links symmetry, group theory, and conservation laws.

The core insight

Every symmetry of the Hamiltonian is a unitary operator that commutes with it, and its Hermitian generator is automatically a conserved observable — this single chain of logic links symmetry, group theory, and conservation laws.

Where this lives today

Symmetry in quantum mechanics is not a foundation that later theories outgrew — it is the load-bearing wall of the Standard Model. The gauge group $SU(3)_c\times SU(2)_L\times U(1)_Y$ is a direct application of equation (6): its $8+3+1=12$ generators are the eight gluons of QCD and the four electroweak bosons ( $W^\pm$ , $Z^0$ , and the photon), measured to sub-percent precision at LEP and the LHC. The internal $SU(2)$ isospin that Heisenberg wrote down in 1932 to explain the near-equal proton and neutron masses is the template every student meets first; its $SU(3)$ flavour extension predicted the $\Omega^-$ baryon before it was found, filling the last seat in a decuplet. Lattice QCD exploits these exact symmetries to compute hadron masses from first principles, and the tiny observed violations — parity violation in the weak interaction (1957), the CP violation tied to time reversal through the CPT theorem — are themselves precision probes of new physics. Wherever an experiment reports a "conserved quantum number" or a "forbidden transition", it is reading out equations (5) and (2).

The physical picture

The subject is a chain of implications, each link forced by the one before. Start from the only thing a symmetry is required to do — leave measurable predictions alone — and unitarity, commutation with $\hat{H}$ , a Hermitian generator, and a conservation law fall out in order. The first figure lays that chain out at a glance.

The second figure carries a real number, and it is the cleanest quantitative surprise in the subject: the group of quantum symmetries is not always the group of geometric motions you started with. Rotations in space form $SO(3)$ , but the operators that rotate a spin-½ state form $SU(2)$ , which wraps around $SO(3)$ twice. Turn a spinor through a full $2\pi$ and it does not come back — its state vector is multiplied by $-1$ ; only a $4\pi$ turn restores it. For a spin lying perpendicular to the rotation axis — so that $\langle\hat{\mathbf n}\cdot\vec\sigma\rangle=0$ , the standard neutron-interferometry geometry — the overlap between original and rotated state is exactly $\cos(\theta/2)$ , plotted below.

Figure

A spin-½ state, its spin perpendicular to the rotation axis, rotated by angle θ: the amplitude to remain unrotated is cos(θ/2), which passes through 0 at θ = π, reaches −1 after a full 2π turn, and returns to +1 only after 4π — the two-to-one double cover of SU(2) over the rotation group SO(3)

The core equations

The theory is a handful of operator identities. Requiring a transformation to leave probabilities alone is requiring it to leave the inner product alone, $\langle\hat{U}\psi|\hat{U}\phi\rangle=\langle\psi|\phi\rangle$ for all states. On operators, that is unitarity.

\hat{U}^\dagger\hat{U}=\hat{U}\hat{U}^\dagger=I \qquad (1)

where: $\hat{U}$ — the symmetry operator; $\hat{U}^\dagger$ — its adjoint; $I$ — the identity. Reading: $\hat{U}$ preserves lengths and angles in Hilbert space, so no probability is created or destroyed — the bar every candidate symmetry must clear first. A transformation can be unitary and still not be a symmetry of the dynamics — that requires the Hamiltonian be unchanged, $\hat{U}\hat{H}\hat{U}^\dagger=\hat{H}$ , which rearranges into a vanishing commutator.

[\hat{H},\hat{U}]=\hat{H}\hat{U}-\hat{U}\hat{H}=0 \qquad (2)

where: $\hat{H}$ — the Hamiltonian; $[\,\cdot\,,\cdot\,]$ — the commutator. Reading: $\hat{U}$ and $\hat{H}$ share eigenstates, so $\hat{U}$ maps each energy level onto itself — shuffling degenerate states without moving energies, organizing the spectrum into multiplets. A continuous symmetry is a family $\hat{U}(\alpha)$ smoothly connected to the identity. Expanding near $\alpha=0$ , the first-order term defines the generator.

\hat{U}(\epsilon)=I+i\epsilon\hat{G}+O(\epsilon^2),\qquad \hat{G}=\hat{G}^\dagger \qquad (3)

where: $\epsilon$ — an infinitesimal parameter; $\hat{G}$ — the generator. Reading: imposing $\hat{U}^\dagger\hat{U}=I$ to order $\epsilon$ gives $\hat{G}=\hat{G}^\dagger$ , so the generator is Hermitian — an observable you can measure. Finite transformations are recovered by compounding infinitely many infinitesimal ones — exponentiating the generator.

\hat{U}(\alpha)=\lim_{N\to\infty}\left(I+\tfrac{i\alpha}{N}\hat{G}\right)^{N}=e^{i\alpha\hat{G}} \qquad (4)

where: $\alpha$ — the finite parameter (an angle, a displacement). Reading: the Hermitian operator $\hat{G}$ contains the whole one-parameter family, and $[\hat{H},\hat{U}(\alpha)]=0$ for every $\alpha$ gives $[\hat{H},\hat{G}]=0$ . The conservation law is now automatic. The Heisenberg equation for any operator without explicit time dependence, in natural units ( $\hbar=1$ ), reads $\frac{d\hat{G}}{dt}=i[\hat{H},\hat{G}]$ ; its expectation value gives Noether's theorem.

\frac{d}{dt}\langle\hat{G}\rangle=i\,\langle[\hat{H},\hat{G}]\rangle=0 \qquad (5)

where: $\langle\hat{G}\rangle$ — the expectation value in any state. Reading: the generator's expectation is frozen in time. Translations freeze $\langle\hat{p}\rangle$ , rotations freeze $\langle\hat{L}\rangle$ , a phase freezes charge — one equation, many conservation laws. Finally, the symmetry operations close under composition into a group, and for continuous groups the generators close under commutation into a Lie algebra fixed by structure constants.

[\hat{T}_a,\hat{T}_b]=i\,f_{abc}\,\hat{T}_c,\qquad \dim SU(n)=n^2-1 \qquad (6)

where: $\hat{T}_a$ — the Hermitian generators; $f_{abc}$ — the real structure constants; $n^2-1$ — the number of $SU(n)$ generators. Reading: $U(1)$ has one (charge), $SU(2)$ three (isospin), $SU(3)$ eight (colour) — and in a gauge theory each generator becomes a force-carrying boson.

Derivation AWhy a symmetry operator must be unitary and commute with the Hamiltonian

The goal here is to show that the two defining properties of a quantum symmetry — unitarity and commutation with the Hamiltonian — are not separate postulates but forced consequences of two innocuous demands: that a symmetry preserve the probabilities a measurement can return, and that it be a symmetry of the dynamics rather than of a single instant. We build the argument from the inner product outward, because that is where all physical content lives, and finish by reading off the degeneracy structure that commutation implies.

Step 1. Demand that the transformation $\psi\to\hat{U}\psi$ leave every transition probability unchanged: $|\langle\hat{U}\phi|\hat{U}\psi\rangle|^2=|\langle\phi|\psi\rangle|^2$ for all pairs of states. The cleanest way to enforce this for all states at once is $\langle\hat{U}\phi|\hat{U}\psi\rangle=\langle\phi|\psi\rangle$ , i.e. $\langle\phi|\hat{U}^\dagger\hat{U}|\psi\rangle=\langle\phi|\psi\rangle$ identically, which forces $\hat{U}^\dagger\hat{U}=I$ . Why: probabilities are squared inner products, so "predictions unchanged" is literally "inner products unchanged", and that statement, holding for arbitrary states, is the operator equation of unitarity.

Step 2. Note the subtlety Wigner resolved: preserving only the moduli $|\langle\phi|\psi\rangle|$ admits a second class where $\hat{U}$ is antiunitary (antilinear, with complex conjugation), which is how time reversal is realized. For any symmetry continuously connected to the identity, antiunitary is impossible — the identity is unitary — so every continuous symmetry is unitary. Why: this is Wigner's theorem, and why parity and charge conjugation are unitary while time reversal alone is antiunitary.

Step 3. Now demand invariance of the dynamics, not just of states. Time evolution is $e^{-i\hat{H}t}$ ; requiring that transforming-then-evolving equals evolving-then-transforming, $\hat{U}e^{-i\hat{H}t}=e^{-i\hat{H}t}\hat{U}$ for all $t$ , is equivalent to $\hat{U}\hat{H}\hat{U}^\dagger=\hat{H}$ . Multiply on the right by $\hat{U}$ to get $\hat{U}\hat{H}=\hat{H}\hat{U}$ , that is $[\hat{H},\hat{U}]=0$ . Why: a symmetry of a snapshot need not survive evolution; commuting with $\hat{H}$ is exactly the condition that the symmetry is respected at every later time.

Step 4. Read off the consequence. If $\hat{H}|\psi\rangle=E|\psi\rangle$ , then $\hat{H}(\hat{U}|\psi\rangle)=\hat{U}\hat{H}|\psi\rangle=E(\hat{U}|\psi\rangle)$ , so $\hat{U}|\psi\rangle$ is a second eigenstate with the same energy. Unless $\hat{U}|\psi\rangle\propto|\psi\rangle$ , the level is degenerate, and $\hat{U}$ rotates within the degenerate subspace. Why: this is the mechanism by which symmetry produces the multiplets — the $2\ell+1$ states of a given angular momentum, the near-degenerate proton and neutron of isospin — that structure every spectrum in physics.

Go deeper · Derivation B · 2 minFrom infinitesimal transformations to Hermitian generators and conserved quantities

Derivation BFrom infinitesimal transformations to Hermitian generators and conserved quantities

The aim of this derivation is to turn the abstract statement "a continuous symmetry" into a concrete, measurable object — a Hermitian generator — and then to show without any further input that this observable is conserved. The route is the standard one for Lie groups: linearize near the identity to extract the generator, impose unitarity to fix its Hermiticity, exponentiate to recover finite transformations, and feed the result into the Heisenberg equation of motion, where the conservation law appears for free.

Step 1. Write a transformation infinitesimally close to the identity as $\hat{U}(\epsilon)=I+i\epsilon\hat{G}+O(\epsilon^2)$ , which merely defines the generator $\hat{G}$ as the first derivative $\hat{G}=-i\,d\hat{U}/d\epsilon$ at $\epsilon=0$ . The factor of $i$ is inserted by convention so that $\hat{G}$ will come out Hermitian rather than anti-Hermitian. Why: every element of a continuous group near the identity is captured by its tangent direction, and that tangent vector is the generator — one operator that encodes an entire one-parameter family.

Step 2. Impose unitarity to first order: $\hat{U}^\dagger\hat{U}=(I-i\epsilon\hat{G}^\dagger)(I+i\epsilon\hat{G})=I+i\epsilon(\hat{G}-\hat{G}^\dagger)+O(\epsilon^2)$ . For this to equal $I$ we need $\hat{G}=\hat{G}^\dagger$ : the generator is Hermitian, hence an observable with real eigenvalues. Why: this is the pivotal line of the whole subject — it is what turns a symmetry (a piece of geometry) into a physical quantity (something a detector can read), and it is the reason momentum and angular momentum are observables at all.

Step 3. Compound $N$ infinitesimal steps of size $\alpha/N$ and let $N\to\infty$ : $\hat{U}(\alpha)=\lim_{N\to\infty}(I+i\tfrac{\alpha}{N}\hat{G})^N=e^{i\alpha\hat{G}}$ . Since $\hat{G}$ is Hermitian, $e^{i\alpha\hat{G}}$ is manifestly unitary for real $\alpha$ . Why: this exponential map is how one Hermitian generator reconstructs the entire continuous symmetry — the same formula that turns the Hamiltonian into the time-evolution operator $e^{-i\hat{H}t}$ , so time translation is just another symmetry.

Step 4. Suppose the symmetry is one of the dynamics, so $[\hat{H},\hat{U}(\alpha)]=0$ for all $\alpha$ . Differentiate at $\alpha=0$ : $[\hat{H},\hat{G}]=0$ . Insert this into the Heisenberg equation $\frac{d}{dt}\langle\hat{G}\rangle=i\langle[\hat{H},\hat{G}]\rangle$ (valid for a generator with no explicit time dependence) to obtain $\frac{d}{dt}\langle\hat{G}\rangle=0$ . Why: this is Noether's theorem in its quantum dress — the generator of any continuous symmetry of the Hamiltonian is conserved — and it required nothing beyond unitarity plus the equation of motion.

Go deeper · Derivation C · 1 minMomentum as the generator of spatial translations

Derivation CMomentum as the generator of spatial translations

This derivation makes the general machinery concrete by identifying the generator of the most familiar continuous symmetry — a rigid shift of the whole system in space — and showing it is precisely the momentum operator. The payoff is twofold: it shows the abstract generator $\hat{G}$ is an object you already know, and it turns "space is homogeneous" into " $\hat{p}$ is conserved" in three lines, a template for how every spacetime symmetry produces its conservation law.

Step 1. Define the translation operator by how it acts on wavefunctions: $[\hat{U}(a)\psi](x)=\psi(x+a)$ shifts the argument by $a$ . Taylor-expand the right side, $\psi(x+a)=\sum_{n=0}^{\infty}\frac{a^n}{n!}\frac{d^n\psi}{dx^n}=e^{a\,d/dx}\psi(x)$ . Why: a finite translation is nothing but the Taylor series that reconstructs a function from its derivatives at a point, and Taylor's theorem has quietly become an operator exponential.

Step 2. Rewrite the derivative using the momentum operator $\hat{p}=-i\,d/dx$ (natural units, $\hbar=1$ ), so $d/dx=i\hat{p}$ and therefore $\hat{U}(a)=e^{a\,d/dx}=e^{ia\hat{p}}$ . Comparing with the canonical form $\hat{U}(\alpha)=e^{i\alpha\hat{G}}$ identifies the generator as $\hat{G}=\hat{p}$ . Why: this is the identification that gives momentum its deepest meaning — not "mass times velocity" but "the generator of translations", the definition that survives into quantum field theory and general relativity.

Step 3. Check the conservation law directly. For $\hat{H}=\hat{p}^2/2m+V(\hat{x})$ , the commutator $[\hat{H},\hat{p}]=[V,\hat{p}]=i\,V'(\hat{x})$ , using $[\hat{x},\hat{p}]=i$ . Hence $\frac{d}{dt}\langle\hat{p}\rangle=i\langle[\hat{H},\hat{p}]\rangle=-\langle V'(\hat{x})\rangle=\langle F\rangle$ — Ehrenfest's theorem — which vanishes exactly when $V$ is constant, i.e. when space is homogeneous. Why: it closes the circle explicitly — translational invariance ( $V'=0$ ) is identical to momentum conservation, and the same commutator reproduces Newton's second law for expectation values.

Limiting cases

Case	Result	Meaning
Continuous symmetry of $\hat{H}$	Hermitian generator $\hat{G}$ , $\frac{d}{dt}\langle\hat{G}\rangle=0$	translation → $\hat{p}$ , rotation → $\hat{L}$ , phase → charge
Discrete symmetry (parity $\hat{P}$ )	no generator; $\hat{P}^2=I$ so eigenvalues $\pm1$	a multiplicative quantum number, conserved iff $[\hat{H},\hat{P}]=0$
Symmetry not shared by $\hat{H}$	$\hat{U}$ unitary but $[\hat{H},\hat{U}]\neq0$	not a symmetry of the dynamics; $\langle\hat{G}\rangle$ drifts (Ehrenfest source term)
Antiunitary symmetry (time reversal $\hat{T}$ )	Wigner: $\hat{T}$ antiunitary, not $e^{i\alpha\hat{G}}$	involves complex conjugation; no Hermitian generator, no additive conserved charge
Abelian vs non-Abelian group	$f_{abc}=0$ vs $f_{abc}\neq0$	$U(1)$ phases commute; $SU(2)$ , $SU(3)$ generators do not, forcing self-interacting gauge fields

Common trapConfusing the symmetry operator with its generator

The single most common slip is to treat the unitary $\hat{U}$ itself as "the conserved quantity". It is not. The unitary $\hat{U}=e^{i\alpha\hat{G}}$ implements the transformation; it is generally not Hermitian, has complex eigenvalues on the unit circle, and is not an observable — you cannot build a detector that measures $\hat{U}$ . The conserved physical quantity is the Hermitian generator $\hat{G}$ , which is Hermitian precisely because $\hat{U}$ is unitary. Momentum, angular momentum, and charge are generators, not the exponentials they generate. A related trap: a symmetry of a state (say, a spherically symmetric wavefunction) is a different statement from a symmetry of the Hamiltonian; only the latter, $[\hat{H},\hat{U}]=0$ , guarantees a conservation law.

Worked examples

1 · A spin-½ state under rotation, and the 4π return (fully worked)

Rotations of a spin-½ particle are generated by $\hat{\mathbf{J}}=\tfrac12\vec{\sigma}$ , with $\vec\sigma$ the Pauli matrices, so the rotation operator about a unit axis $\hat{\mathbf n}$ through angle $\theta$ is $\hat{U}(\theta)=e^{i\theta(\hat{\mathbf n}\cdot\vec\sigma)/2}$ . Find the explicit matrix, then evaluate a full $2\pi$ turn and a $4\pi$ turn.

Step 1. Use the identity $(\hat{\mathbf n}\cdot\vec\sigma)^2=I$ , which holds because $\sigma_i\sigma_j=\delta_{ij}I+i\epsilon_{ijk}\sigma_k$ . Then the exponential's even powers collapse to $\cos$ and odd powers to $\sin$ :

\hat{U}(\theta)=\cos\!\tfrac{\theta}{2}\,I+i\sin\!\tfrac{\theta}{2}\,(\hat{\mathbf n}\cdot\vec\sigma).

This is the general $SU(2)$ element — any $2\times2$ unitary of determinant $1$ has this form.

Step 2. Evaluate a full rotation, $\theta=2\pi$ : $\cos\pi=-1$ , $\sin\pi=0$ , so $\hat{U}(2\pi)=-I$ . A full $360^\circ$ turn multiplies every spin-½ state by $\mathbf{-1}$ — the state vector is not returned to itself, even though the physical direction of space has come back.

Step 3. Evaluate a double rotation, $\theta=4\pi$ : $\cos2\pi=+1$ , $\sin2\pi=0$ , so $\hat{U}(4\pi)=+I$ . Only after $\mathbf{720^\circ}$ does the spinor return. For a spin perpendicular to the rotation axis ( $\langle\hat{\mathbf n}\cdot\vec\sigma\rangle=0$ ), the overlap after any angle is $\langle\psi|\hat{U}(\theta)|\psi\rangle=\cos(\theta/2)$ , the curve in the second figure: $+1$ at $0$ , $\mathbf{0}$ at $\pi$ , $\mathbf{-1}$ at $2\pi$ , $+1$ at $4\pi$ .

Step 4. Interpret. The overall sign is unobservable for a single spinor (states are rays), but it is measurable by interference: split a neutron beam, rotate one arm by $2\pi$ , recombine, and the relative $-1$ flips constructive interference to destructive — seen experimentally in 1975. The physics: $SU(2)$ double-covers $SO(3)$ — two elements ( $\pm I$ ) map to the one "do nothing" rotation, which is why half-integer spin exists.

2 · Translational invariance and the conservation of momentum

A particle moves in a potential $V(x)$ . Use the generator formalism to decide when momentum is conserved, and connect the result to a familiar classical law.

Step 1. The generator of translations is $\hat{p}=-i\,d/dx$ . Momentum is conserved iff it commutes with the Hamiltonian $\hat{H}=\hat{p}^2/2m+V(\hat{x})$ . Since $\hat{p}$ commutes with any function of $\hat{p}$ , only the potential term can contribute: $[\hat{H},\hat{p}]=[V(\hat{x}),\hat{p}]$ .

Step 2. Evaluate the commutator on a test function: $[V,\hat{p}]\psi=V(-i\psi')-(-i)(V\psi)'=-iV\psi'+i(V'\psi+V\psi')=iV'(x)\psi$ . Hence $[\hat{H},\hat{p}]=i\,V'(\hat{x})$ .

Step 3. Feed it into the Heisenberg equation: $\dfrac{d}{dt}\langle\hat{p}\rangle=i\langle[\hat{H},\hat{p}]\rangle=i\cdot i\langle V'\rangle=-\langle V'(\hat{x})\rangle=\langle F\rangle$ . This is Ehrenfest's theorem — the quantum analogue of $\dot p=F$ .

Step 4. Read off the symmetry statement. Momentum is conserved, $\frac{d}{dt}\langle\hat{p}\rangle=0$ , exactly when $V'(x)=0$ — a homogeneous space with no preferred origin. A free particle conserves momentum; a particle on a slope ( $V=mgx$ ) does not, and the drift $\langle F\rangle=-mg$ is precisely the classical force. Translational symmetry $\Leftrightarrow$ momentum conservation, made quantitative.

3 · Counting the generators of the Standard Model gauge group

The Standard Model is built on the internal symmetry group $SU(3)_c\times SU(2)_L\times U(1)_Y$ . How many independent generators does it have, and what are they physically?

Step 1. Use $\dim SU(n)=n^2-1$ and $\dim U(1)=1$ . Colour $SU(3)_c$ contributes $3^2-1=\mathbf{8}$ generators; weak isospin $SU(2)_L$ contributes $2^2-1=\mathbf{3}$ ; hypercharge $U(1)_Y$ contributes $\mathbf{1}$ . The total is $8+3+1=\mathbf{12}$ .

Step 2. In a gauge theory each generator becomes a spin-1 boson. The eight $SU(3)$ generators are the eight gluons; the four electroweak generators mix into the $W^+$ , $W^-$ , $Z^0$ and the photon. Twelve generators, twelve gauge bosons — the counting is exact.

Step 3. Because $f_{abc}\neq0$ for $SU(3)$ and $SU(2)$ , those gauge fields carry their own charge and self-interact (gluons couple to gluons), whereas the Abelian $U(1)$ photon does not. The arithmetic $n^2-1$ is not bookkeeping — it fixes the particle content of the strong and electroweak forces.

Physics of the portrait

If any single person turned "symmetry" into the organizing principle of quantum mechanics, it was Eugene Wigner. Where Noether tied symmetry to conservation in the classical Lagrangian, Wigner asked the sharper quantum question — how may a symmetry act on a wavefunction? — and answered it with a theorem that still fixes the rules.

Origins · 1 minEugene Wigner and how this was discovered

Eugene Wigner

1902–1995 · group theory and symmetry in quantum mechanics

Eugene Paul Wigner, born in Budapest in 1902 and trained as a chemical engineer, became the physicist who taught quantum mechanics to speak the language of group theory. Definitively in his 1931 book Gruppentheorie und ihre Anwendung auf die Quantenmechanik der Atomspektren, he showed that the degeneracies and selection rules of atomic spectra were consequences of the rotation group's representations — the $2\ell+1$ -fold structure of an energy level is the dimension of an irreducible representation of $SO(3)$ . His deepest single result, now called Wigner's theorem, settles the question this whole article rests on: any symmetry transformation of a quantum system must be represented by an operator that is either unitary or antiunitary. That dichotomy is why continuous symmetries and parity are unitary (Hermitian generators, additive conserved charges) while time reversal alone is antiunitary. Wigner also classified elementary particles as irreducible representations of the Poincaré group, labelled by mass and spin — the modern definition of a particle. He shared the 1963 Nobel Prize in Physics for exactly this: the discovery and application of fundamental symmetry principles.

History

1918

— Emmy Noether proves that continuous symmetries imply conservation laws

1926

— Heisenberg, Born, Jordan, and Schrödinger complete quantum mechanics

1927

— Wigner and Weyl bring group representation theory into atomic spectroscopy

1931

— Wigner's theorem fixes symmetries as unitary or antiunitary operators

1932

— Heisenberg proposes isospin, the first internal SU(2) symmetry

1961

— Gell-Mann and Ne'eman classify hadrons by SU(3), the Eightfold Way

Going further

Texts. Thomson, Modern Particle Physics, §9.1 (Symmetries in quantum mechanics) and §9.1.1 (Finite transformations) — the reference this article tracks most closely · Griffiths, Introduction to Elementary Particles, 2nd ed., §4.1–4.2 (Symmetries, groups, angular momentum) · Halzen & Martin, Quarks and Leptons, Ch. 2 (symmetries and quarks) · Peskin & Schroeder, An Introduction to Quantum Field Theory, §2.2 and Ch. 15 (Noether's theorem, non-Abelian gauge symmetry) · Schwartz, Quantum Field Theory and the Standard Model, Ch. 8–10 · Sakurai & Napolitano, Modern Quantum Mechanics, Ch. 3–4 (rotations, symmetry) for the single-particle foundations. The key results to keep in mind across this article are: (1) $\hat{G}=\hat{G}^\dagger$ follows from $\hat{U}^\dagger\hat{U}=I$ to first order in $\epsilon$ ; (2) $\hat{U}(a)=e^{ia\hat{p}}$ follows from Taylor's theorem with $\hat{p}=-i\partial_x$ ; (3) $[\hat{H},\hat{p}]=iV'(\hat{x})$ and hence $\frac{d}{dt}\langle\hat{p}\rangle=\langle F\rangle$ ; (4) $\hat{U}(\theta)=\cos\tfrac{\theta}{2}I+i\sin\tfrac{\theta}{2}(\hat{\mathbf n}\cdot\vec\sigma)$ with $\hat{U}(2\pi)=-I$ and $\hat{U}(4\pi)=+I$ ; (5) the generator counts give $\dim SU(3)=8$ , $\dim SU(2)=3$ , for a total of $12$ .

Prerequisites. Non-relativistic quantum mechanics review · Natural units. Leads to. Isospin · $SU(3)$ flavour symmetry · The local gauge principle.

Ask the tutor: Want to see how this same unitary-generator logic, applied to an internal phase instead of a motion in space, forces the existence of the photon — the local gauge principle? Ask me to walk you from $U(1)$ symmetry to electromagnetism.

Is quantum Noether's theorem the same as the classical one?

They express the same truth — continuous symmetry implies conservation law — but by different machinery. The classical theorem works in the Lagrangian: an invariance of the action gives a conserved current via the Euler–Lagrange equations. The quantum version is more direct: it needs only the inner product (to force unitarity and hence a Hermitian generator) and the Heisenberg equation (to make the generator's expectation constant). No action or variational calculus is required at the single-particle level. In quantum field theory the two fuse — the Noether current becomes an operator whose integrated charge is the very generator that exponentiates the symmetry.

Why does unitarity force the generator to be Hermitian?

Write the transformation near the identity as $\hat{U}=I+i\epsilon\hat{G}$ , with adjoint $\hat{U}^\dagger=I-i\epsilon\hat{G}^\dagger$ . Multiplying gives $\hat{U}^\dagger\hat{U}=I+i\epsilon(\hat{G}-\hat{G}^\dagger)+O(\epsilon^2)$ . Unitarity demands this equal the identity, so the first-order term must vanish: $\hat{G}=\hat{G}^\dagger$ , Hermitian. The factor of $i$ in the definition is exactly what makes this work. This one line is why the generators of symmetries are always observables with real eigenvalues: momentum, angular momentum, and charge are Hermitian because translations, rotations, and phase changes are unitary.

What is the difference between a symmetry of a state and a symmetry of the Hamiltonian?

A symmetry of a state means a particular wavefunction is left unchanged by $\hat{U}$ — an s-orbital is invariant under rotations. A symmetry of the Hamiltonian means $[\hat{H},\hat{U}]=0$ , so the dynamics itself is unchanged. Only the second guarantees a conservation law, because only $[\hat{H},\hat{U}]=0$ feeds the Heisenberg equation. The distinction is not academic: in spontaneous symmetry breaking the Hamiltonian keeps a symmetry that its ground state does not share, and the conservation law still holds even though no low-energy state looks symmetric — the source of much muddle about the Higgs mechanism.

Why do spins use SU(2) rather than the rotation group SO(3)?

Because half-integer angular momentum exists, and $SO(3)$ cannot accommodate it. The rotation group $SO(3)$ has only integer-spin representations; a rotation by $2\pi$ is the identity in $SO(3)$ . But experiment shows a spin-½ state picks up a sign under $2\pi$ and needs $4\pi$ to return. The group that captures this is $SU(2)$ , which "double-covers" $SO(3)$ : two elements, $+I$ and $-I$ , both project to the identity rotation. Half-integer spins are representations of $SU(2)$ that are not representations of $SO(3)$ , which is why the electron, proton, and quarks are described by two-component spinors.

Do discrete symmetries like parity have generators?

No — and that is the defining difference. A generator exists only for a continuous symmetry, where you can take an infinitesimal step and read off $\hat{G}=-i\,d\hat{U}/d\epsilon$ . Parity, charge conjugation, and reflections are discrete: there is no "infinitesimal parity", so no Hermitian generator and no additive charge. What they conserve instead is a multiplicative quantum number. Since $\hat{P}^2=I$ , parity has eigenvalues $\pm1$ , and if $[\hat{H},\hat{P}]=0$ the total parity is conserved as a product of $\pm1$ factors. Selection rules like "electric-dipole transitions must change parity" are the practical face of this.

What is Wigner's theorem and why does time reversal break the pattern?

Wigner's theorem states that any symmetry of a quantum system — anything preserving all transition probabilities $|\langle\phi|\psi\rangle|^2$ — must be represented by an operator that is either unitary or antiunitary (antilinear plus complex conjugation). Every symmetry continuously connected to the identity is unitary, since the identity is unitary and you cannot deform continuously into the antiunitary class; so translations, rotations, parity, and charge conjugation are unitary with Hermitian generators. Time reversal is the exception: flipping the direction of time reverses momenta and spins while leaving positions alone, and the only consistent operator that does this is antiunitary. That is why time reversal has no Hermitian generator and no additive conserved charge.