Time Reversal in Quantum Dynamics

Every symmetry met so far — translations, rotations, parity — rode into quantum mechanics on a unitary operator. This article is about the one that cannot. Run the microscopic movie backwards: Newton's mechanics accepts the reversed film as a legal solution (velocities flip, trajectories retrace), and the question is whether the Schrödinger equation extends the courtesy. The naive substitution $t \to -t$ fails immediately — the equation is first order in time, and the lone time derivative flips sign while the Hamiltonian does not. Something else must flip with it. That something is the imaginary unit. Complex conjugation repairs the sign, and the operator implementing time reversal, $\Theta$ , comes out antiunitary: it conjugates every complex number it walks past. This is not a technical blemish but the article's central theorem — Wigner proved in 1932 that no unitary operator can do the job without wrecking the energy spectrum, and his 1931 symmetry theorem had already reserved exactly one alternative slot. Time reversal is the tenant of that slot. The payoff for admitting one strange operator is startlingly concrete. Antiunitarity gives $\Theta^2 = \pm 1$ , a sign that distinguishes integer from half-integer spin — and when the sign is $-1$ , every energy level of a time-reversal-invariant system is forced to be doubly degenerate. That is Kramers' theorem: a two-line argument protecting degeneracies in every odd-electron molecule, spin-½ qubit, and topological-insulator edge — degeneracy no electric field can lift, and any magnetic field immediately does.

Key takeaways

Time reversal is motion reversal: positions stay, momenta and angular momenta (including spin) flip — the reversed movie.
The Schrödinger equation is first order in $t$ , so $t \to -t$ alone gives the wrong equation; adding complex conjugation repairs it: $\Theta$ is antiunitary, $\Theta = \hat U K$ .
A unitary time reversal would force $\Theta\hat H\Theta^{-1} = -\hat H$ — an energy spectrum unbounded below. Antilinearity is the only escape (Wigner, 1932).
Antiunitary bookkeeping: $\langle\Theta\phi|\Theta\psi\rangle = \langle\psi|\phi\rangle$ — probabilities survive, amplitudes conjugate.
$\Theta^2 = +1$ for integer spin, $-1$ for half-integer. When $\Theta^2 = -1$ and $[\Theta,\hat H] = 0$ : Kramers' theorem — every level at least doubly degenerate.
Magnetic fields break the symmetry: $\Theta\hat H(\mathbf B)\Theta^{-1} = \hat H(-\mathbf B)$ . Electric fields do not.
Key idea: reversing quantum motion requires conjugating the wavefunction — and that single antiunitary fact forces spin-dependent degeneracy structure no unitary symmetry could.

The core insight

The antiunitary operator $\Theta$ reverses the direction of motion — $\Theta\hat x\Theta^{-1} = \hat x$ , $\Theta\hat p\Theta^{-1} = -\hat p$ — and a Hamiltonian with $[\Theta,\hat H] = 0$ runs the same physics forwards and backwards; because $\Theta^2 = -1$ for half-integer spin, such Hamiltonians cannot have nondegenerate levels.

Where this lives today

The sharpest tests of time reversal are electric-dipole-moment searches. A permanent EDM must point along the spin (T-odd), so a nonzero value in a T-symmetric world is a contradiction — EDMs measure T violation directly. The 2020 neutron bound $d_n < 1.8\times10^{-26}\ e\cdot$ cm corresponds to ~ $10^{-22}$ eV in a megavolt-per-meter field — among the most sensitive null measurements in physics, already pruning theories of the matter–antimatter asymmetry. Meanwhile Kramers pairs became infrastructure: the edge states of topological insulators are partner states $|\psi\rangle$ , $\Theta|\psi\rangle$ whose crossing is protected exactly by the $\Theta^2 = -1$ argument of this article (the solid-state track owns that story), and every odd-electron qubit inherits a T-protected working doublet. When a condensed-matter paper says a band crossing is "symmetry-protected," the protector is, more often than not, the operator Wigner reluctantly admitted in 1932.

The physical picture

Film a billiard collision and play the reel backwards. Balls retrace their paths with reversed velocities, and — Newton's law being second order in time — the reversed film is also a legal solution. Classical mechanics cannot tell the films apart. "Time reversal" is a misnomer: nothing travels to the past. It is motion reversal — at one instant, keep every position, flip every velocity, and let dynamics run forward again. If the dynamics is symmetric, the system retraces its history.

Figure

Motion reversal — positions kept, momenta flipped, history retraced

Quantum mechanically the surgery must act on a wavefunction: keep the position distribution, reverse the traffic. For a plane wave $e^{ikx}$ the traffic direction lives in the phase, and the state moving the other way is the conjugate $e^{-ikx}$ . That is the whole trick: the time-reversed wavefunction is the conjugated wavefunction, $\Theta\psi = \psi^*$ for a spinless particle in the position representation. Squared magnitudes never notice; what flips is everything pointing along the motion — momentum, current, orbital and spin angular momentum. The deep question is why conjugation is forced — why no unitary rotation of Hilbert space can reverse motion. That is the first derivation below. The license to consider an antilinear operator at all comes from Unitary Operators and Transformations: Wigner's theorem allows probability-preserving symmetries to be unitary or antiunitary, and every symmetry continuously connected to the identity must take the unitary branch. Time reversal — discrete, disconnected, order two — finally occupies the other one.

Why conjugation is compulsory

Demand the minimum: a map $\Theta$ such that evolving the reversed state forward equals reversing the evolved state — formally $\Theta\,\hat U(t)\,\Theta^{-1} = \hat U(-t)$ , with $\hat U(t) = e^{-i\hat Ht/\hbar}$ owned by The Time-Evolution Operator. Two roads lead there, and only one is open.

Derivation AAntiunitarity is forced by the Schrödinger equation

We show that motion reversal cannot be implemented unitarily, by running both candidate roads to their ends. The strategy: first try the naive substitution $t \to -t$ and watch the first-order time derivative wreck it; then repair with complex conjugation and verify the repaired state solves the original equation; finally, close the escape hatch by showing a unitary reversal operator would flip the sign of the Hamiltonian itself — a fatal disease for any theory with a ground state. The failure is instructive: it locates the arrow-of-motion in the single $i$ of the Schrödinger equation.

Step 1. The naive road. Let $\chi(x,t) \equiv \psi(x,-t)$ . Substituting into $i\hbar\,\partial_t\psi = \hat H\psi$ and using $\partial_t\chi = -(\partial_t\psi)(-t)$ :

i\hbar\,\partial_t\chi \;=\; -\,\hat H\chi

Why it fails: the equation is first order in time, so reflecting $t$ flips exactly one sign. Newton's second-order law absorbs $t\to-t$ silently — $(-1)^2 = 1$ — which is why classical reversibility never needed this discussion.

Step 2. The repair. Conjugate the original equation, then evaluate at $-t$ . Define $\tilde\psi(x,t) \equiv \psi^*(x,-t)$ . Conjugation flips $i \to -i$ ; evaluating at $-t$ flips the derivative; the two signs cancel:

i\hbar\,\partial_t\tilde\psi \;=\; \hat H^*\,\tilde\psi

Step 3. When is $\hat H^* = \hat H$ ? In the position representation $\hat p^2 = -\hbar^2\nabla^2$ is real, and a real potential $V(\hat{\mathbf x})$ is real. For such Hamiltonians $\tilde\psi$ solves the original equation: the conjugated, time-reflected movie is legal physics. The operator performing the surgery on states is $\Theta = K$ — complex conjugation in the position basis — and for spinless particles that is the whole story.

Step 4. Close the unitary escape hatch. Suppose instead some unitary $\hat T$ satisfied $\hat T\hat U(t)\hat T^{-1} = \hat U(-t)$ . Unitaries commute with forming exponentials of $-i\hat Ht/\hbar$ , so differentiating at $t=0$ gives $\hat T(-i\hat H)\hat T^{-1} = +i\hat H$ , i.e. $\hat T\hat H\hat T^{-1} = -\hat H$ : the spectrum of $\hat H$ would be symmetric about zero. A free particle's spectrum $[0,\infty)$ — or any Hamiltonian with a ground state and unbounded excitations — flatly refuses. An antilinear $\Theta$ pulls the $-i$ through as $+i$ , giving $\Theta\hat H\Theta^{-1} = +\hat H$ with no spectral damage. Antiunitarity is not a preference; it is the only survivor. Verified: symbolically — a generic complex test function under $KpK$ returns $-p$ while $KxK$ returns $+x$ ; the spectral-flip exhibit $\mathrm{eig}(-H) = -\mathrm{eig}(H)$ numerically; and $\Theta\hat U(t)\Theta^{-1} = \hat U(-t)$ holds to $10^{-12}$ on a random time-reversal-invariant $4\times4$ Hamiltonian (verify script, 25 checks green).

The antiunitary toolkit

An antiunitary operator is antilinear — $\Theta(a|\psi\rangle + b|\phi\rangle) = a^*\Theta|\psi\rangle + b^*\Theta|\phi\rangle$ — and norm-preserving. The inner-product rule replaces the unitary one:

\langle\Theta\phi\,|\,\Theta\psi\rangle \;=\; \langle\psi\,|\,\phi\rangle \;=\; \langle\phi\,|\,\psi\rangle^{*} \qquad (1)

where: amplitudes conjugate, so probabilities $|\langle\phi|\psi\rangle|^2$ are exactly preserved — which is why Wigner's theorem admits this branch at all. Two bookkeeping consequences matter daily. First, every antiunitary factorizes as $\Theta = \hat U K$ : one fixed conjugation $K$ (basis-dependent) times a unitary $\hat U$ that cleans up after it. Second, the "adjoint" of an antilinear operator needs the modified definition flagged in Hermitian Conjugate and Adjoint Operators — the naive adjoint rule would equate a quantity antilinear in $|\psi\rangle$ with one linear in it; rule (1) is the consistent replacement, and it is why $\Theta$ has no ordinary eigenvalue problem. How observables transform is fixed by one requirement — reversed states must assign reversed statistics. With $\Theta = K$ in the position basis, a direct test-function computation gives the catalogue:

\Theta\,\hat{\mathbf x}\,\Theta^{-1} = \hat{\mathbf x}, \qquad \Theta\,\hat{\mathbf p}\,\Theta^{-1} = -\hat{\mathbf p}, \qquad \Theta\,\hat{\mathbf L}\,\Theta^{-1} = -\hat{\mathbf L} \qquad (2)

where: the momentum flip comes from $K(-i\hbar\nabla)K = +i\hbar\nabla$ — the $i$ conjugates, the real derivative does not — and $\hat{\mathbf L} = \hat{\mathbf x}\times\hat{\mathbf p}$ (the operator construction opens the next article, Orbital Angular Momentum Operators) inherits the minus from its single factor of $\hat{\mathbf p}$ . Spin, having no coordinate expression, is assigned the same law by fiat of consistency: $\Theta\,\hat{\mathbf S}\,\Theta^{-1} = -\hat{\mathbf S}$ — all angular momenta reverse together. The working realization for spin-½ is

\Theta \;=\; -i\sigma_y\,K \qquad (3)

where: the $-i\sigma_y$ is the unitary clean-up crew: $K$ alone flips $\sigma_y$ (imaginary entries) but not $\sigma_x,\sigma_z$ (real entries), and conjugating by $\sigma_y$ — which anticommutes with both — repairs exactly those two. The detailed spinor machinery belongs to the spin articles ahead (Spin-1/2 and the Pauli Matrices); here (3) is needed for exactly one purpose: computing $\Theta^2$ .

quantity	under $\Theta$	reading
$\hat{\mathbf x}$ , $V(\hat{\mathbf x})$ , energy	even	positions and static landscapes survive the reversed movie
$\hat{\mathbf p}$ , current $\mathbf j$ , $\hat{\mathbf L}$ , $\hat{\mathbf S}$	odd	everything pointing along the motion flips
electric field $\mathbf E$	even	sourced by static charges
magnetic field $\mathbf B$	odd	sourced by currents — which flip
$\|\psi\|^2$ , all probabilities	even	rule (1): amplitudes conjugate, magnitudes survive

The sign that sorts the world: $\Theta^2 = \pm 1$

Apply the reversal twice and the movie runs forward again — physically, nothing happened. But states are rays: "nothing" means $\Theta^2 = e^{i\alpha}\hat 1$ . Antilinearity crushes the freedom: with $\Theta = \hat U K$ , $\Theta^2 = \hat U\hat U^*$ , and consistency forces $e^{i\alpha} = \pm 1$ — an invariant no re-phasing can move. For spinless particles $\Theta = K$ gives $\Theta^2 = +1$ . For spin-½, equation (3) gives $\Theta^2 = (-i\sigma_y)(-i\sigma_y)^* = -1$ — verified entry by entry in the script. The general law: integer total spin $\Rightarrow +1$ ; half-integer $\Rightarrow -1$ (each spin-½ constituent contributes one factor $-1$ ; the 2π-rotation kinship of this sign is the business of the SU(2) article ahead).

Figure

Kramers pairs — the −1 class forces twofold levels that only a magnetic field can split

Go deeper · Derivation B · 1 minKramers' theorem — two lines of antilinear algebra, one unbreakable degeneracy

Derivation BKramers' theorem — two lines of antilinear algebra, one unbreakable degeneracy

We prove that when $\Theta^2 = -1$ and $[\Theta,\hat H] = 0$ , no energy eigenvalue can be simple: every eigenstate has an orthogonal partner at the same energy. The strategy is disarmingly short — show the reversed state is an eigenstate (trivial), then show it can never be the same ray, by computing the overlap $\langle\psi|\Theta\psi\rangle$ twice with rule (1) and letting $\Theta^2 = -1$ produce a contradiction unless the overlap vanishes. Watch how the proof never mentions what $\hat H$ is — only that it commutes with $\Theta$ ; that indifference is the theorem's power.

Step 1. The partner shares the energy. From $[\Theta,\hat H] = 0$ and antilinearity (real $E$ passes through unharmed):

\hat H\,(\Theta|\psi\rangle) \;=\; \Theta\hat H|\psi\rangle \;=\; \Theta(E|\psi\rangle) \;=\; E\,(\Theta|\psi\rangle)

Step 2. The partner is orthogonal. Apply rule (1) with $\phi = \Theta\psi$ : $\langle\Theta\psi|\Theta(\Theta\psi)\rangle = \langle\Theta\psi \to \psi\rangle$ ordering gives

\langle\psi|\Theta\psi\rangle \;=\; \langle\Theta(\Theta\psi)\,|\,\Theta\psi\rangle \;=\; \langle\Theta^2\psi\,|\,\Theta\psi\rangle \;=\; -\,\langle\psi|\Theta\psi\rangle

Why the first equality: rule (1) read right-to-left with $\phi = \Theta\psi$ , $\chi = \psi$ — conjugating twice returns the original ordering. The only number equal to its own negative is zero: $\langle\psi|\Theta\psi\rangle = 0$ .

Step 3. Conclude. $\Theta|\psi\rangle$ is an eigenstate at $E$ , orthogonal to $|\psi\rangle$ — a genuine second dimension. Every level of the spectrum is at least doubly degenerate: Kramers pairs. For $\Theta^2 = +1$ the same computation gives $\langle\psi|\Theta\psi\rangle = +\langle\psi|\Theta\psi\rangle$ — no information, no protection; a spinless particle's levels may be (and in one dimension, are) simple.

Step 4. Read the fine print. The theorem needs $[\Theta,\hat H] = 0$ . Electric fields, crystal fields, strain, spin–orbit coupling — all T-even, all powerless against the doublet. A magnetic field is T-odd: $\Theta\hat H(\mathbf B)\Theta^{-1} = \hat H(-\mathbf B) \neq \hat H(\mathbf B)$ , the hypothesis fails, and the pair splits by the Zeeman energy. Kramers degeneracy is thus a diagnostic: an odd-electron system whose levels split without an applied field is reporting an internal effective magnetic field. Verified: numerically — a random $4\times4$ Hamiltonian symmetrized to commute with $\Theta = (-i\sigma_y\otimes\hat 1)K$ shows eigenvalue pairs degenerate to $10^{-13}$ while its unsymmetrized parent has no degeneracy; $\langle\psi|\Theta\psi\rangle = 0$ exactly for an arbitrary symbolic spinor; and $\Theta(\mathbf B\cdot\boldsymbol\sigma)\Theta^{-1} = -\mathbf B\cdot\boldsymbol\sigma$ (verify script).

One classical corollary deserves its pointer rather than a re-derivation. For a spinless particle ( $\Theta^2 = +1$ ) with a nondegenerate spectrum, $\psi^*$ must be the same ray as $\psi$ — so the eigenfunction can be chosen real, and real eigenfunctions carry zero probability current. That chain is owned by Non-Degeneracy of One-Dimensional Bound States (its reality corollary); re-read it now as time-reversal reasoning: a nondegenerate stationary state has nowhere to hide a direction of motion, so $\Theta$ must fix it, so it is real.

Dynamics, propagators, and the reversed Green function

The symmetry statement at the level of evolution is the one Derivation A demanded: for $[\Theta,\hat H] = 0$ ,

\Theta\,\hat U(t)\,\Theta^{-1} \;=\; \hat U(-t) \qquad (4)

where: reversing, evolving, and reversing back is the same as un-evolving. Applied to matrix elements this becomes microreversibility — the amplitude for $|i\rangle \to |f\rangle$ forward equals the amplitude for reversed- $f$ into reversed- $i$ — the microscopic seed of detailed balance in statistical mechanics (owned by that track; one line here). On kernels, conjugation acts concretely: the free propagator satisfies $K_0(x,t)^* = K_0(x,-t)$ — checked numerically in the script against the closed form from Propagator for the Free Particle and Gaussian Spreading. And the Green-function pairing met one article ago is this symmetry wearing operator clothes: $\hat G^{-}(t) = \hat G^{+}(-t)^{\dagger}$ , the advanced function as the retarded function of the reversed world — Retarded and Advanced Green Functions proved the identity; this article supplies its name.

Limiting cases

case	result	meaning
spinless, real $V(\mathbf x)$	$\Theta = K$ , $\Theta^2 = +1$	conjugation is the whole operator; no protected degeneracy
spin-½	$\Theta = -i\sigma_yK$ , $\Theta^2 = -1$	Kramers pairs; every level doubly degenerate if T holds
$N$ spin-½ particles	$\Theta^2 = (-1)^N$	odd-electron systems protected, even-electron not
$\mathbf B \neq 0$	$\Theta\hat H\Theta^{-1} = \hat H(-\mathbf B)$	T broken; doublets split by $g\mu_BB$
$\mathbf E \neq 0$ , any strength	$[\Theta,\hat H] = 0$ still	Stark effects cannot lift Kramers degeneracy
classical limit	$(x,p) \to (x,-p)$	motion reversal of trajectories; the second-order law never noticed

Common trapTime reversal sends the system into the past

It does not — nothing here travels anywhere. $\Theta$ acts now: it takes the state at one instant and reverses every motional quantity, producing a perfectly ordinary state that then evolves forward. The content of "T-symmetry" is that this reversed state retraces the original history — equation (4), not time travel. Two sibling confusions: (i) "the universe is irreversible, so T-symmetry must be broken" — macroscopic irreversibility is statistics (entropy, initial conditions), not a property of $\hat H$ ; the microscopic laws tested to date are T-symmetric except for a sliver in the weak interaction. (ii) " $\Theta$ is just parity in time, so it should be unitary like parity" — parity is linear because space reflection leaves the Schrödinger equation's $i\partial_t$ untouched; time reflection hits it head-on, and Derivation A shows the unitary road ends in a spectrum unbounded below. The antilinearity is load-bearing.

Worked examples

Example 1 — reversing a moving wave packet (confidence builder). A Gaussian packet $\psi \propto e^{ik_0x}e^{-x^2/2\sigma^2}$ has $\langle\hat x\rangle = 0$ and $\langle\hat p\rangle = \hbar k_0$ . Compute both expectations in the reversed state $\Theta\psi = \psi^*$ .

Step 1. Position. $|\psi^*|^2 = |\psi|^2$ , so $\langle\hat x\rangle_\Theta = 0$ — unchanged, as the catalogue (2) demands for a T-even operator.

Step 2. Momentum. Conjugation sends $e^{ik_0x} \to e^{-ik_0x}$ : the plane-wave factor now travels leftward. Direct integration (script, exact): $\langle\hat p\rangle_\Theta = \mathbf{-\hbar k_0}$ .

Step 3. Read it dynamically. Let the original packet fly for time $\tau$ , reverse it, and evolve for $\tau$ again: by (4) it returns to the conjugate of its starting state — same place, mirrored motion. The free Hamiltonian is T-invariant, so the movie really does retrace; only the dispersive spreading is common to both legs (spreading is T-even — it happens in both time directions, as Wave Packet Spreading and Dispersion can confirm).

Example 2 — Kramers doublet meets a magnetic field: EPR (exam level). A molecular radical has one unpaired electron (half-integer total spin, $\Theta^2 = -1$ ) in a messy, anisotropic crystal field. (a) What does time reversal say about its levels at $\mathbf B = 0$ ? (b) A field $B = 0.35$ T is applied; find the doublet splitting and the resonance frequency for $g \approx 2.0023$ .

Step 1. At $\mathbf B = 0$ every interaction present — Coulomb, crystal field, spin–orbit — is T-even, so Kramers' theorem holds: each level is a rigorous doublet, regardless of how asymmetric the environment is. No electric perturbation can split it.

Step 2. The field breaks the protection: $\Delta E = g\mu_BB = 2.0023 \times (5.788\times10^{-5}\ \mathrm{eV/T}) \times 0.35\ \mathrm{T} = \mathbf{4.06\times10^{-5}\ eV}$ .

Step 3. The resonance: $\nu = \Delta E/h = \mathbf{9.81\ GHz}$ — squarely in the X-band where EPR spectrometers live. The entire technique presupposes this article: without the field there is a degenerate pair and no transition; the microwave probes a splitting that exists only because T was deliberately broken.

Step 4. Reader's bonus: run it backwards. An odd-electron system showing zero-field splitting is announcing an internal T-odd agent — magnetic ordering, for instance. Kramers degeneracy as instrumentation.

The man who found the antiunitary branch

Origins · 1 minEugene Wigner and how this was discovered

Eugene Wigner

1902–1995 · Berlin, then Princeton

Wigner's 1931 theorem had catalogued how a symmetry can act on quantum states: unitary, or the strange antiunitary alternative nobody had a use for. One year later he found the tenant. The 1932 paper "Über die Operation der Zeitumkehr in der Quantenmechanik" showed that reversing motion requires the antiunitary branch — then cashed the sign $\Theta^2 = -1$ into physics: Kramers had found in 1930, by brute calculation, that odd-electron ions kept a twofold degeneracy no crystal field could remove, and Wigner's two-line argument revealed a theorem about time itself, not about crystals. It is a very Wigner story: locate the exact mathematical structure — antilinearity — and physical facts fall out that computation had only stumbled on. The group-theoretic machinery he built around such arguments earned the 1963 Nobel Prize. The same antiunitary operator now guards qubit doublets and topological edge states; the branch nobody wanted holds up half of condensed-matter physics.

History

1930

— Kramers finds, computationally, an unremovable double degeneracy in odd-electron ions

1931

— Wigner's theorem: symmetries act unitarily or antiunitarily

1932

— Wigner's time-reversal paper: Θ is antiunitary; Kramers degeneracy explained

1964

— Cronin and Fitch observe CP violation in neutral kaons — with CPT, indirect T violation

1998

— CPLEAR at CERN measures direct T violation in kaon oscillation rates

2020

— the neutron electric-dipole-moment bound reaches 1.8 × 10⁻²⁶ e·cm, still consistent with zero

Going further

References. Sakurai & Napolitano, Modern Quantum Mechanics, 2nd ed., §4.4 (Discrete Symmetries, Time-Reversal) · Wigner, E.P., Group Theory and its Application to the Quantum Mechanics of Atomic Spectra, Ch. 26 · Cohen-Tannoudji, Diu & Laloë, Quantum Mechanics, complements on time reversal · Shankar, Principles of Quantum Mechanics, 2nd ed., Ch. 11 · Gottfried & Yan, Quantum Mechanics: Fundamentals, 2nd ed., §7.2.

Provenance. Transformation rules ( $\Theta x\Theta^{-1}=x$ , $\Theta p\Theta^{-1}=-p$ , $\Theta\mathbf{J}\Theta^{-1}=-\mathbf{J}$ ), antiunitarity, $\Theta=-i\sigma_y K$ , $\Theta^2=(-1)^{2j}$ , the propagator relation $\Theta U(t)\Theta^{-1}=U(-t)$ , and Kramers' theorem follow Sakurai & Napolitano, Modern QM, Sec. 4.4; cross-checked with Shankar Ch. 11, Messiah Vol. II Ch. XV, and Wigner (1959) Ch. 26. A verifier should recompute the antiunitarity inner-product reversal, the $\Theta i\Theta^{-1}=-i$ consistency with $[x,p]=i\hbar$ , the $\Theta^2=-1$ matrix computation for spin- $1/2$ , and the Kramers orthogonality $\langle\psi|\Theta\psi\rangle=0$ . No empirical numbers are quoted (EDM bounds intentionally omitted as they are experiment- and date-dependent). For this edition, verification was machine-executed (sympy/numpy, 25 checks green), including the test-function catalogue, the antiunitary inner-product rule, the exact vanishing of $\langle\psi|\Theta\psi\rangle$ for arbitrary spinors, Kramers pairing on a randomized T-invariant matrix, and the worked numbers ( $4.06\times10^{-5}$ eV, 9.81 GHz, $1.8\times10^{-22}$ eV — empirical anchors added beyond the base, each recomputed). Ownership bridges: the Wigner-theorem statement stays with Unitary Operators and Transformations, the reality corollary with Non-Degeneracy of One-Dimensional Bound States, $\hat G^{-}=\hat G^{+}(-t)^{\dagger}$ with Retarded and Advanced Green Functions, spinor detail with the spin articles, CPT with the relativistic chapter.

Builds on: The Time-Evolution Operator · Time-Dependent Schrodinger Equation · Stationary States and Energy Eigenstates · Constants of the Motion and Conservation Laws

Leads to: Spectral Decomposition of Time Evolution · Ehrenfest's Theorem · The Heisenberg Picture

If Θ is antilinear, what happens to the superposition principle in the reversed world?

Nothing observable. $\Theta(a|\psi\rangle + b|\phi\rangle) = a^*\Theta|\psi\rangle + b^*\Theta|\phi\rangle$ — superpositions map to superpositions; only the coefficients conjugate. Since physical predictions involve squared magnitudes of amplitudes, and $|a^*| = |a|$ , every interference pattern survives intact. The conjugation is detectable only in quantities that compare phases across operators — exactly the motional quantities (momenta, currents) that time reversal is supposed to flip.

Does Θ have eigenvalues and eigenstates like other symmetry operators?

Not in the useful sense, and this trips everyone. For an antilinear operator, an "eigenvalue equation" $\Theta|\psi\rangle = \lambda|\psi\rangle$ is unstable under re-phasing: replacing $|\psi\rangle \to e^{i\beta}|\psi\rangle$ changes $\lambda \to e^{-2i\beta}\lambda$ . So there is no basis-independent spectrum, no "conservation of Θ-number," and no quantum number analogous to parity's $\pm1$ . The invariant content of $\Theta$ lives elsewhere: in the sign of $\Theta^2$ , in Kramers pairing, and in selection-rule constraints on matrix elements.

Why does a magnetic field break time reversal but an electric field doesn't — both are just fields?

Trace them to their sources. $\mathbf E$ comes from charges standing still — play the movie backwards and static charges are still static charges: T-even. $\mathbf B$ comes from currents, and reversed motion reverses every current: T-odd. Equivalently, at the Hamiltonian level $\Theta$ flips the spin and momentum operators but not the external field parameters, leaving $\Theta\hat H(\mathbf B)\Theta^{-1} = \hat H(-\mathbf B)$ . Note the fix: if you also reverse the sources of the field (the current in the coil), symmetry is restored — Kramers degeneracy in the lab splits because the coil current is not part of the reversed system.

Kramers' theorem needs half-integer spin. What happens in an even-electron molecule?

$\Theta^2 = +1$ , the overlap argument yields $0 = 0$ , and no protection exists — an even-electron system in a low-symmetry environment generically has fully nondegenerate levels (singlets). This sharp even/odd alternation is experimentally loud: odd-electron radicals always show EPR doublets; closed-shell molecules are EPR-silent at zero field. The electron count, through nothing but a sign in $\Theta^2$ , dictates the degeneracy structure before any Hamiltonian details are known.

Is the arrow of time — entropy increasing — evidence that Θ-symmetry is broken?

No. Microscopic T-symmetry and macroscopic irreversibility coexist without contradiction: the reversed movie of a shattering glass is a legal solution of T-symmetric dynamics, merely one requiring absurdly conspiratorial initial conditions. Irreversibility is about which initial conditions the universe actually supplies (statistical mechanics' jurisdiction), not about the equations. The genuine, measured breaking of T in fundamental physics is tiny and specific — kaon and B-meson systems via the weak interaction — and has nothing to do with why coffee cools.

How does Θ interact with the other discrete symmetry, parity?

They are independent surgeries: parity $\hat\Pi$ reflects space (unitary, $\hat\Pi^2 = 1$ , honest eigenvalues $\pm1$ — see its own article), $\Theta$ reverses motion (antiunitary, no eigenvalue structure). They commute on spinless systems, and their product is a distinct symmetry with its own uses. In relativistic physics the trio $C$ , $P$ , $T$ is bound by the CPT theorem — the combination is exact in any local Lorentz-invariant theory, which is precisely why observed CP violation (kaons, 1964) implies T violation without measuring time directly. That story belongs to the relativistic chapter.

Why is the Θ² sign the same as the 2π-rotation sign for spinors — coincidence?

No — same mathematics, met twice. Both are statements about the double-valuedness of the rotation group's spin-½ representation: a $2\pi$ rotation gives $-1$ on spinors, and $\Theta^2$ computes $\hat U\hat U^*$ for $\hat U = -i\sigma_y$ , which is a $\pi$ -rotation about $y$ applied together with its conjugate — effectively completing a full turn. The SU(2) article ahead makes the double cover precise; when you get there, recognize $\Theta^2 = (-1)^{2j}$ as an old friend.

Time Reversal in Quantum Dynamics

Where this lives today

The physical picture

Why conjugation is compulsory

The antiunitary toolkit

The sign that sorts the world: $\Theta^2 = \pm 1$

Dynamics, propagators, and the reversed Green function

Limiting cases

Worked examples

The man who found the antiunitary branch

Going further

Before this

Where this leads

Time Reversal in Quantum Dynamics

Where this lives today

The physical picture

Why conjugation is compulsory

The antiunitary toolkit

The sign that sorts the world: Θ2=±1\Theta^2 = \pm 1Θ2=±1

Dynamics, propagators, and the reversed Green function

Limiting cases

Worked examples

The man who found the antiunitary branch

Going further

Before this

Where this leads

The sign that sorts the world: $\Theta^2 = \pm 1$