1. Purpose and scope

This dependent note isolates one specific “action–angle indeterminacy” statement that is both explicit and referee-safe in a central potential: \(\phi\) is delocalized in an \(L_z\) eigenstate, and conversely localizing \(\phi\) requires a superposition over many \(m\) modes.

The core witness occupies Sections 2–4 and can be read independently of the rest. We do not enter the self-adjoint “angle operator” debate; instead we use the standard circle/Fourier structure and the unitary phase variable \(e^{i\phi}\). We also do not make any claims about the range of forces or potentials; the point here is about which variables can be simultaneously sharp in stationary states.

Sections 5–7 provide an extended series of remarks connecting the witness to related structures — coherent states, decoherence, OAM of light, superfluid vortices, metrological limits, semiclassical dynamics, geometric quantization, and spectral statistics. These remarks are supplementary context within the satellite network: they flesh out the conceptual neighborhood of the witness but are not required for its proof. Readers seeking only the core result may proceed directly from Section 4 to the Outlook (Section 7, first paragraph).

2. The safe conjugate pair on the circle: \(\phi\) and \(L_z\)

In spherical coordinates the azimuthal angle is periodic, \(\phi\sim\phi+2\pi\). The generator of rotations about the \(z\)-axis is \[ L_z=-i\hbar\,\frac{\partial}{\partial\phi}. \] The periodicity makes the naive commutator \([\phi,L_z]=i\hbar\) subtle if one insists on an everywhere-defined self-adjoint \(\phi\) operator. A standard way to stay on safe ground is to use the unitary “phase” variable \[ E := e^{i\phi}. \] Acting on \(2\pi\)-periodic wavefunctions, \(E\) is well-defined and satisfies the canonical shift relation \[ [L_z,E]=\hbar\,E, \] which already captures the operational content: sharp \(L_z\) implies maximal delocalization in the conjugate angle.

Remark 2.1 (Number–phase pair: the oscillator counterpart). The same structure appears for the harmonic-oscillator number–phase pair \((N,\theta)\). The number operator \(N=\hat a^\dagger\hat a\) has non-negative integer spectrum, and the oscillation phase \(\theta\) is periodic — the same mathematical setting as \((L_z,\phi)\). The Susskind–Glogower operator \(\hat E_-=\sum_{n=0}^\infty|n\rangle\langle n+1|=\hat a\,(N+1)^{-1/2}\) (Susskind and Glogower, 1964) plays the role of \(E=e^{i\phi}\): it satisfies \([N,\hat E_-]=-\hat E_-\) (lowering \(n\) by one) and its adjoint \(\hat E_+\) satisfies \([N,\hat E_+]=+\hat E_+\). However, while \(\hat E_-\hat E_+ = I\) (no correction — \(\hat E_+\) is an isometry), the reverse product is \(\hat E_+\hat E_-=I-|0\rangle\langle 0|\): the vacuum projection spoils exact unitarity because the spectrum of \(N\) is bounded below, and there is no state below \(|0\rangle\) to shift into. The asymmetry \(\hat E_-\hat E_+ = I \neq \hat E_+\hat E_-\) is the precise source of the non-unitarity. This is the Fock-space manifestation of the same obstruction that prevents a self-adjoint angle operator on the circle. The consequence: a Fock state \(|n\rangle\) has a completely uniform phase distribution (the ring of Remark 6.4), just as an \(L_z\) eigenstate has uniform \(\phi\) (Section 3), and localizing the oscillator phase requires a broad superposition over number states — exactly the coherent-state construction of Example 6.1.

Remark 2.2 (Phase POVM: rigorous angle probabilities without a self-adjoint operator). The unitary variable \(E=e^{i\phi}\) of Section 2 captures the first circular moment of the angle distribution, but the full probability distribution for angle outcomes is given by a positive operator-valued measure (POVM). For the \(L_z\) case (doubly-unbounded integer spectrum \(m\in\mathbb{Z}\)), the phase POVM on \([0,2\pi)\) is \(d\Pi(\phi)=(2\pi)^{-1}\sum_{m,n\in\mathbb{Z}}|m\rangle\langle n|\,e^{i(m-n)\phi}\,d\phi\). (For the oscillator number operator with semi-bounded spectrum \(n\ge 0\), the analogous POVM requires more care: the “phase states” \(|\phi\rangle = (2\pi)^{-1/2}\sum_{n=0}^\infty e^{in\phi}|n\rangle\) are not orthogonal, and the resolution of identity involves a subtlety related to the vacuum-projection obstruction of Remark 2.1; see [Holevo1982] for the rigorous treatment.) One verifies \(\int_0^{2\pi}d\Pi(\phi)=I\) by orthogonality, and positivity follows because \(\langle\psi|d\Pi(\phi)|\psi\rangle=(2\pi)^{-1}|\sum_m c_m e^{-im\phi}|^2\,d\phi\ge 0\). For an \(L_z\) eigenstate \(|m_0\rangle\), the distribution is uniform (as in Section 3); for a coherent state, it is peaked at the classical phase (as in Example 6.1). The POVM determines all moments of the phase distribution — the unitary \(E\) captures only the first — and thus provides the rigorous framework underlying the Fourier tradeoff of Section 4. The obstruction to a self-adjoint angle operator differs between the two settings: for the oscillator (semi-bounded spectrum \(n\ge 0\)), it is Pauli’s argument that the generated shift would produce negative eigenvalues; for \(L_z\) (doubly-unbounded integer spectrum), it is the compactness of the circle \(S^1\) — a continuous shift cannot map the discrete spectrum \(\mathbb{Z}\) to itself (Holevo, 1982). The POVM framework handles both uniformly, replacing the nonexistent self-adjoint operator with a well-defined probability measure.

Remark 2.3 (Pegg–Barnett hermitian phase operator: self-adjointness via truncation and limit). The POVM of Remark 2.2 provides angle probabilities without a self-adjoint operator. Pegg and Barnett (1989) showed that a complementary route exists: in a truncated \((s{+}1)\)-dimensional Hilbert space \(\mathrm{span}{|0\rangle,\ldots,|s\rangle}\), define \((s{+}1)\) phase states \(|\theta_m\rangle=(s{+}1)^{-1/2}\sum_{n=0}^{s}e^{in\theta_m}|n\rangle\) with \(\theta_m=\theta_0+2\pi m/(s{+}1)\), and a Hermitian phase operator \(\hat\Phi_{\mathrm{PB}}=\sum_{m=0}^{s}\theta_m|\theta_m\rangle\langle\theta_m|\) [PeggBarnett1989]. In the truncated space, the number states \({|n\rangle}\) and phase states \({|\theta_m\rangle}\) form complementary bases with uniform overlap \(|\langle n|\theta_m\rangle|^2=1/(s{+}1)\) for all \(n,m\) — making number and phase maximally complementary observables. The matrix elements of the commutator \([\hat N,\hat\Phi_{\mathrm{PB}}]\) between physical states (those in the original \(\ell^2(\mathbb{N}0)\) space) approach the canonical value \(i\) as \(s\to\infty\), with corrections of order \(1/(s{+}1)\); all expectation values and moments of \(\hat\Phi{\mathrm{PB}}\) likewise converge to the POVM results of Remark 2.2 in this limit. The strategy — self-adjoint operator in finite dimension, physical predictions extracted via limiting — sidesteps the Pauli obstruction (Remark 2.1) and the POVM detour (Remark 2.2) by a third route: the phase operator is not defined at any single finite truncation, but as a controlled limit of well-defined finite-dimensional operators.

3. Central potentials: \(L_z\) eigenstates have uniform \(\phi\) distribution

For a central potential (or any Hamiltonian commuting with \(L_z\)), one may choose simultaneous eigenstates of \(L_z\). In the standard separation of variables, the azimuthal dependence of an angular-momentum eigenstate is the Fourier mode \(e^{im\phi}\) with integer \(m\) (for example in the spherical-harmonic factor \(Y_{\ell m}(\theta,\phi)\propto P_{\ell m}(\cos\theta)e^{im\phi}\)) [TongQMLectures].

Thus an \(L_z\) eigenstate may be written as \[ \psi(r,\theta,\phi)=F(r,\theta)\,e^{im\phi}, \qquad m\in\mathbb Z, \] and therefore \[ |\psi(r,\theta,\phi)|^2 = |F(r,\theta)|^2, \] independent of \(\phi\). In particular, the marginal distribution of \(\phi\) is uniform on \([0,2\pi)\). This is the minimal “angle indeterminacy” witness for central potentials.

Remark 3.1 (Real spherical harmonics: directional lobes from the minimal m-superposition). The complex spherical harmonics \(Y_{\ell,m}\) have definite \(m\) and therefore uniform \(\phi\)-dependence (the main result above). The “real” spherical harmonics used in chemistry — \(p_x\propto\sin\theta\cos\phi\), \(p_y\propto\sin\theta\sin\phi\), \(d_{xy}\propto\sin^2\theta\sin 2\phi\), etc. — are the real and imaginary parts of \(Y_{\ell,m}\), hence equal-weight superpositions of \(m\) and \(-m\). This minimal two-mode superposition already breaks azimuthal uniformity: the probability density acquires \(\cos^2(m\phi)\) or \(\sin^2(m\phi)\) angular lobes, at the cost of angular-momentum indeterminacy \(\mathrm{Var}(L_z)=m^2\hbar^2\). The directional orbital shapes of atomic and molecular physics are thus the simplest instance of the Fourier tradeoff quantified in Section 4.

Remark 3.2 (Dipole selection rules as Fourier orthogonality). The azimuthal Fourier structure \(e^{im\phi}\) of Section 3 has its most experimentally consequential manifestation in electromagnetic selection rules. The electric dipole transition matrix element involves \(\langle n’l’m’|r\,Y_1^q|nlm\rangle\), where \(q=0\) for \(\pi\)-transitions (\(z\)-polarization) and \(q=\pm 1\) for \(\sigma\)-transitions (circular polarization). The azimuthal integral \(\int_0^{2\pi}e^{-im’\phi}\,e^{iq\phi}\,e^{im\phi}\,d\phi=2\pi\,\delta_{m’,m+q}\) gives the selection rule \(\Delta m=q=0,\pm 1\) directly from Fourier orthogonality on the circle. A separate selection rule, \(\Delta l=\pm 1\), arises from the polar integral (parity and Clebsch–Gordan orthogonality of the associated Legendre functions); this constrains the radial-angular coupling but is not a consequence of the azimuthal Fourier structure. For each allowed transition, the emitted photon carries angular momentum component \(q\hbar\) along the quantization axis: \(+\hbar\) for \(\sigma^+\) (\(\Delta m=+1\)), \(-\hbar\) for \(\sigma^-\) (\(\Delta m=-1\)), and \(0\) for \(\pi\) (\(\Delta m=0\)). The selection rule \(m’=m+q\) is thus angular momentum conservation projected onto the \(z\)-axis. The Zeeman effect provides the canonical experimental demonstration: in a magnetic field the \(m\)-degeneracy is lifted, and spectral lines split into components with \(\Delta m=0,\pm 1\) — directly encoding the azimuthal Fourier structure in the emitted light. This is the observational face of the Fourier tradeoff quantified in Section 4: the orthogonality of Fourier modes constrains which modes can couple, and this orthogonality is precisely the mathematical content of the uniform \(\phi\) distribution.

Remark 3.3 (Orbital angular momentum of light: the photonic analog of azimuthal structure). The azimuthal Fourier structure \(e^{im\phi}\) of Section 3 is not limited to matter wavefunctions: optical beams carry orbital angular momentum (OAM) with the same \(e^{i\ell\phi}\) dependence. Allen, Beijersbergen, Spreeuw, and Woerdman (1992) showed that Laguerre–Gauss beams \(\mathrm{LG}{p,\ell}\) carry OAM of \(\ell\hbar\) per photon — independently of the spin angular momentum (polarization), so the total angular momentum along the beam axis is \((\ell+\sigma)\hbar\) with \(\sigma=\pm 1\) for circular polarization [Allen1992]. For \(|\ell|\ge 1\), the azimuthal intensity is uniform in \(\phi\), producing the characteristic ring-shaped “doughnut” beam — the optical version of the \(L_z\) eigenstate’s uniform \(\phi\) distribution (the \(\ell=0\) modes are Gaussian-like, with a central maximum). Superposing \(\mathrm{LG}{p,\ell}\) and \(\mathrm{LG}_{p,-\ell}\) creates intensity patterns with \(2|\ell|\) azimuthal petals — the photonic analog of the real spherical harmonics of Remark 3.1. The OAM alphabet \({\ell=0,\pm 1,\pm 2,\ldots}\) is unbounded, unlike spin polarization, enabling higher-dimensional quantum information protocols; Mair, Vaziri, Weihs, and Zeilinger (2001) demonstrated entanglement of photonic OAM states [Mair2001]. The azimuthal Fourier tradeoff of Section 4 applies identically: localizing the angular position of a photon requires superposing many \(\ell\) modes, and the phase vortex at the beam axis — where the phase winds by \(2\pi\ell\) around the optical axis, topologically forcing zero intensity — is the photonic manifestation of the single-valuedness condition that quantizes \(m\) in Section 3.

Remark 3.4 (Quantized vortices in superfluids: the macroscopic face of azimuthal quantization). The azimuthal structure \(e^{im\phi}\) of Section 3 is not limited to single-particle wavefunctions: it governs the macroscopic order parameter of a superfluid or Bose–Einstein condensate, \(\Psi(\mathbf{r})=\sqrt{\rho(\mathbf{r})}\,e^{iS(\mathbf{r})/\hbar}\). Single-valuedness of \(\Psi\) around any closed loop forces the circulation to be quantized: \(\oint\mathbf{v}\cdot d\boldsymbol{\ell}=(h/m)\,n\) with integer \(n\), where \(\mathbf{v}=(\hbar/m)\nabla S\) is the superfluid velocity (Onsager, 1949; Feynman, 1955) [Feynman1955]. At a vortex with winding number \(n\), the phase winds by \(2\pi n\), topologically forcing \(|\Psi|\to 0\) at the core — the same mechanism as the OAM beam’s on-axis singularity (Remark 3.3), with the core radius set by the healing length \(\xi=\hbar/\sqrt{2m\mu_c}\) (equivalently \(\xi=1/\sqrt{8\pi\rho a}\) in units \(\hbar=m=1\), where \(\rho\) is the number density and \(a\) the s-wave scattering length) (typically tens to hundreds of nanometers in dilute BECs). In dilute-gas BECs, singly quantized vortices (\(n=1\)) have been directly imaged (Matthews et al., 1999; Madison et al., 2000); multiply charged vortices (\(n>1\)) are dynamically unstable in uniform condensates and typically decay into \(n\) single-quantum vortices on millisecond timescales under dilute-BEC conditions (the precise timescale depends on trap geometry, condensate density, and interaction strength) [Matthews1999]. In type-II superconductors, the circulation quantization becomes magnetic flux quantization: each flux tube carries \(\Phi_0=h/(2e)\) (reflecting the Cooper-pair charge \(2e\)), and the tubes self-organize into Abrikosov’s triangular lattice (1957; Nobel Prize 2003). The action–angle connection is direct: the winding number \(n\) is the quantized action variable, while the overall condensate phase is the conjugate angle — delocalized when particle number \(N\) is sharp (the macroscopic version of the number–phase tradeoff of Remark 2.1). The Anderson–Josephson relation \(d\varphi/dt=-\mu/\hbar\) (where \(\mu\) is the chemical potential) makes the phase dynamics explicit: Josephson oscillations between weakly coupled condensates are the macroscopic manifestation of the number–phase tradeoff.

4. Fourier tradeoff: localizing \(\phi\) forces a broad \(m\)-superposition

Any square-integrable \(2\pi\)-periodic function admits a Fourier series \[ \psi(\phi)=\sum_{m\in\mathbb Z} c_m e^{im\phi}, \qquad \sum_{m\in\mathbb Z}|c_m|^2<\infty. \] If only one Fourier mode is present (sharp \(m\), hence sharp \(L_z\)), then \(|\psi(\phi)|^2\) is constant; conversely, a state that is peaked in \(\phi\) necessarily uses many Fourier modes (broad \(m\)-support).

Example 4.1 (Dirichlet-kernel packet). The normalized superposition of modes \(-M\le m\le M\), \[ \psi_M(\phi)=\frac{1}{\sqrt{2\pi(2M+1)}}\sum_{m=-M}^{M} e^{im\phi}, \] is peaked near \(\phi=0\) with an angular width that scales like \(1/M\), while its \(m\)-distribution is spread across \({-M,\dots,M}\). This makes the “sharpening \(\phi\) \(\Rightarrow\) broadening \(L_z\)” tradeoff completely explicit without invoking any disputed angle-operator formalism.

The Fourier tradeoff above can be made into a sharp quantitative bound using only the self-adjoint observables \(\cos\phi\) and \(\sin\phi\):

Proposition 4.2 (Circular uncertainty relation). For any state on the circle, define the circular concentration \(R=|\langle e^{i\phi}\rangle|\in[0,1]\). Adding the Robertson inequalities for the two self-adjoint pairs \((L_z,\cos\phi)\) and \((L_z,\sin\phi)\) — using \([L_z,\cos\phi]=i\hbar\sin\phi\) and \([L_z,\sin\phi]=-i\hbar\cos\phi\) — and the identity \(\mathrm{Var}(\cos\phi)+\mathrm{Var}(\sin\phi)=1-R^2\) (which follows from \(\langle\cos^2\phi\rangle+\langle\sin^2\phi\rangle=1\) and \(\langle\cos\phi\rangle^2+\langle\sin\phi\rangle^2=R^2\)), gives \[ \mathrm{Var}(L_z)\cdot(1-R^2)\ge\frac{\hbar^2}{4}\,R^2. \] When \(R=0\) (uniform distribution, as in an \(L_z\) eigenstate) the bound is trivial. As \(R\to1\) (sharply localized angle) the bound forces \(\mathrm{Var}(L_z)\to\infty\): angular localization requires spreading across many \(m\)-modes. This quantifies the Fourier tradeoff above without invoking a self-adjoint angle operator.

This inequality was first derived (in equivalent form) by Judge (1964) [Judge1964] and developed systematically by Carruthers and Nieto (1968) [CarruthersNieto1968]; the present derivation follows their approach. The contribution of this section is the explicit verification (Example 4.3) and the near-saturation analysis (Remark 4.4) in the context of central-potential quantum mechanics.

Example 4.3 (Verifying the bound for the Dirichlet-kernel packet). For the state \(\psi_M\) of Example 4.1, the circular concentration is \(R=\langle e^{i\phi}\rangle = 2M/(2M+1)\) (by orthogonality, only the \(2M\) consecutive pairs \((m,m+1)\) with both in \({-M,\dots,M}\) contribute). The angular-momentum variance is \(\mathrm{Var}(L_z)=\hbar^2 M(M+1)/3\) (using \(\sum_{m=1}^M m^2 = M(M+1)(2M+1)/6\) and \(\langle L_z\rangle=0\) by symmetry). The ratio of the left-hand side to the right-hand side of the bound in Proposition 4.2 is \[ \frac{\mathrm{Var}(L_z)\,(1-R^2)}{(\hbar^2/4)\,R^2} =\frac{(M+1)(4M+1)}{3M}, \] which equals \(10/3\approx 3.3\) at \(M=1\) and grows as \(4M/3\) for large \(M\). The inequality is satisfied with increasing slack: the Dirichlet kernel is far from a minimum-uncertainty state for the circular relation. Physically, narrower angular packets (\(R\to 1\)) require disproportionately more angular-momentum spread than the bound demands.

Remark 4.4 (Near-optimal angular localization: the von Mises state). The rectangular Fourier profile of the Dirichlet kernel wastes angular-momentum variance on sidelobes, driving the ratio LHS/RHS to \(4M/3\). The angular analog of a Gaussian — the von Mises wavefunction \(\psi(\phi)\propto \exp(\kappa\cos\phi)\) — has Fourier coefficients \(c_m\propto I_m(\kappa)\) (modified Bessel functions of the first kind) that decay smoothly [MardiaJupp2000, Ch. 3].

Derivation of near-saturation. The probability density is \( \psi ^2 \propto \exp(2\kappa\cos\phi)\), which is a von Mises distribution with concentration parameter \(\mu = 2\kappa\). Its circular concentration is \(R = I_1(\mu)/I_0(\mu)\) (the ratio of modified Bessel functions), which for large \(\mu\) satisfies \(R = 1 - 1/(2\mu) + O(\mu^{-2})\) by the uniform asymptotic expansion \(I_\nu(\mu) \sim e^\mu/\sqrt{2\pi\mu}\,(1 - (4\nu^2-1)/(8\mu) + \cdots)\) (Abramowitz and Stegun, §9.7.1). Hence \(1 - R \approx 1/(4\kappa)\) and \(1 - R^2 = (1-R)(1+R) \approx 1/(2\kappa)\) for large \(\kappa\).
For the action variance: \(\mathrm{Var}(L_z) = \hbar^2\sum_m m^2 c_m ^2/\sum_m c_m ^2\) with \(c_m \propto I_m(\kappa)\). Using the identity \(\sum_m m^2 I_m(\kappa)^2 = (\kappa/2)\sum_m[I_{m-1}(\kappa)I_m(\kappa) + I_m(\kappa)I_{m+1}(\kappa)]\) (from the recurrence \(mI_m = (\kappa/2)(I_{m-1}-I_{m+1})\)), the leading asymptotic is \(\mathrm{Var}(L_z) \approx \hbar^2\kappa/2\) for large \(\kappa\).

Combining: \(\mathrm{Var}(L_z)(1-R^2)/[(\hbar^2/4)R^2] \approx (\hbar^2\kappa/2)(1/(2\kappa))/(\hbar^2/4) = 1\) as \(\kappa\to\infty\). The von Mises state asymptotically saturates the bound in Proposition 4.2.

Remark 4.5 (Entropic uncertainty: an always-nontrivial bound on the Fourier tradeoff). The variance-based bound of Proposition 4.2 becomes trivial when \(R=0\) (uniform angle distribution, as in an \(L_z\) eigenstate). Entropic measures avoid this limitation. For a state with angular-momentum probabilities \({|c_m|^2}\) and angular density \(p(\phi)=|\psi(\phi)|^2\), the Shannon entropy \(H(m)=-\sum|c_m|^2\ln|c_m|^2\) and differential entropy \(h(\phi)=-\int p(\phi)\ln p(\phi)\,d\phi\) satisfy the entropic uncertainty relation \(h(\phi)+H(m)\ge\ln(2\pi)\) (Bialynicki-Birula and Madajczyk, 1985). This bound is always nontrivial: for an \(L_z\) eigenstate, \(h(\phi)=\ln(2\pi)\) and \(H(m)=0\), saturating the bound; for any state with reduced angular entropy \(h(\phi)<\ln(2\pi)\), the angular-momentum entropy must compensate: \(H(m)\ge\ln(2\pi)-h(\phi)>0\). The mixed differential-discrete entropy formulation is valid because the Fourier series on the circle connects an \(L^2\) function on \([0,2\pi)\) to a sequence in \(\ell^2(\mathbb{Z})\), and the inequality follows from the Hausdorff–Young inequality on the circle group. This is the circular analog of the Hirschman–Beckner–Bialynicki-Birula–Mycielski bound \(h(x)+h(p)\ge 1+\ln\pi\) for position-momentum on the line (Beckner, 1975; Bialynicki-Birula and Mycielski, 1975), which is saturated by Gaussians — the coherent states of Example 6.1.

Remark 4.6 (Gabor limit: the time-frequency analog of action–angle indeterminacy). The circular Fourier tradeoff of Section 4 has a direct counterpart in signal processing: the Gabor limit. A signal cannot be simultaneously localized in time and frequency, with the uncertainty bound \(\Delta t\cdot\Delta f\ge 1/(4\pi)\) (equivalently \(\Delta t\cdot\Delta\omega\ge 1/2\) for angular frequency \(\omega=2\pi f\)). Gabor (1946) identified the “logon” — the elementary signal unit of minimum time-bandwidth product — as the Gaussian-windowed sinusoid \(\exp(-t^2/2\sigma^2)\cdot e^{2\pi i f_0 t}\), which is literally a coherent state of the harmonic oscillator (Example 6.1) under the identification \(t\leftrightarrow q\), \(f\leftrightarrow p/h\) [Gabor1946]. Gabor frames — the overcomplete set of time-frequency shifted Gaussians \({g(t-na)\cdot e^{2\pi ikbt}}\) — provide a resolution of identity when \(ab<1\) (overcomplete regime); at the critical density \(ab=1\) the system is a Riesz basis, but the Balian–Low theorem then forbids simultaneous good localization of the window in both time and frequency (Daubechies, 1992) [Daubechies1992]. The Fourier tradeoff of Section 4 is the circular analog of the Gabor limit: on \(S^1\), the entropic uncertainty of Remark 4.5 and the variance bound of Proposition 4.2 play the role of the Gabor limit on \(\mathbb{R}\), with the von Mises state (Remark 4.4) serving as the circular analog of the Gaussian window. The entropic inequality \(h(\phi)+H(m)\ge\ln(2\pi)\) of Remark 4.5 maps to the Hirschman–Beckner inequality for position-momentum on the line, while the circular concentration \(R\) of Proposition 4.2 is the counterpart of the spectral concentration studied in Slepian–Landau–Pollak theory (prolate spheroidal wave functions).

5. Foundations message: orbit pictures require packets/superpositions

This witness supports a simple interpretive guardrail for central-force intuition: a single stationary eigenstate (even when it carries classical-sounding quantum numbers) is typically not a localized classical orbit with a definite phase/orientation. Variables like the azimuthal phase \(\phi\) (and, in more structured integrable cases, other angle variables on the invariant torus) become localized only in coherent superpositions of many stationary modes.

In other words, “classical orbit pictures” correspond to semiclassical packets and stationary-phase concentration, not to exact eigenstates that are sharp in the conserved actions.

Remark 5.1 (Temporal coherence and quantum revivals). The superpositions that localize an angle variable also have a temporal constraint: for anharmonic spectra (\(d^2E/dm^2\neq 0\)), the packet disperses on a timescale \(t_{\mathrm{disp}}\sim\hbar/(|d^2E/dm^2|\,\Delta m)\) and reforms at the revival time \(t_{\mathrm{rev}}\sim 2\pi\hbar/|d^2E/dm^2|\). Only for a linear spectrum (\(d^2E/dm^2=0\)) does the packet rotate rigidly like a classical orbit for all time. Thus classical orbit pictures require not only spatial localization (many \(m\)-modes, Section 4) but also approximate spectral linearity for temporal coherence.

Remark 5.2 (Decoherence selects the localized packets). Environment-induced decoherence provides the dynamical mechanism that selects the coherent packets of Section 4 over the sharp-action eigenstates of Section 3. For a harmonic oscillator coupled to a thermal bath through position, coherent states minimize the rate of entanglement with the environment and emerge as the preferred “pointer states” — the states most robust against decoherence [ZurekHabibPaz1993]. Fock states, by contrast, decohere rapidly: superpositions of well-separated number states lose coherence on timescales much shorter than the thermal relaxation time, because the position operator (through which the environment couples) does not commute with the number operator. Classical orbit pictures thus emerge not only from semiclassical wavepacket structure (Sections 4–5) but from the environment’s dynamical selection of those packets as the robust states.

Remark 5.3 (Energy–time uncertainty: the Mandelstam–Tamm temporal analog). The action–angle tradeoff of Sections 3–4 has a temporal counterpart that avoids the well-known difficulty of defining a self-adjoint time operator (Pauli’s theorem for bounded-below Hamiltonians). Mandelstam and Tamm (1945) define the “evolution time” of an observable \(A\) as \(\Delta t_A:=\Delta A/|d\langle A\rangle/dt|\) — the time for \(\langle A\rangle\) to change by one standard deviation. The Robertson inequality for \((H,A)\) then gives \(\Delta E\cdot\Delta t_A\ge\hbar/2\): sharp energy implies slow evolution of every observable, just as sharp \(L_z\) implies uniform \(\phi\) (Section 3). An energy eigenstate (\(\Delta E=0\)) has \(d\langle A\rangle/dt=0\) for all \(A\) — a completely static state, the temporal version of the azimuthally uniform \(L_z\) eigenstate. Conversely, rapid evolution requires a broad energy superposition, just as angular localization requires many \(m\)-modes (Section 4). For the coherent state \(|\alpha\rangle\) of Example 6.1 below, \(\Delta E = \hbar\omega|\alpha|\) (from the Poisson distribution with mean \(|\alpha|^2\)) and \(\Delta t_x = \Delta x/|d\langle x\rangle/dt| = \sqrt{\hbar/(2m\omega)}/(|\alpha|\omega\sqrt{\hbar/(2m\omega)}) = 1/(2\omega|\alpha|)\), giving \(\Delta E\cdot\Delta t_x = \hbar/2\): the bound is saturated, confirming the coherent state as a minimum-uncertainty state for both the spatial and temporal versions of the tradeoff.

Remark 5.4 (Aharonov–Bohm effect: the action variable shifted by a gauge potential). The action–angle framework extends naturally to charged particles in electromagnetic fields through the canonical momentum \(\mathbf p=m\mathbf v+e\mathbf A\). In the Aharonov–Bohm effect (Aharonov and Bohm, 1959), a charged particle encircling a solenoid acquires a phase shift \(\Delta\varphi=e\Phi/\hbar\), where \(\Phi=\oint\mathbf A\cdot d\boldsymbol\ell\) is the enclosed magnetic flux — even though the magnetic field vanishes everywhere along the particle’s path. This is a direct modification of the action integral: the EBK quantization condition of Remark 6.3 becomes \(\oint\mathbf p\cdot d\mathbf q=(n+\alpha/4)\hbar+e\Phi/(2\pi)\), shifting the spectrum by the enclosed flux. In a ring geometry, the energy levels are \(E_m\propto(m-\Phi/\Phi_0)^2\) with flux quantum \(\Phi_0=h/e\), periodic in \(\Phi\) — directly observed as conductance oscillations in mesoscopic rings. The AB phase is topological: it depends only on the enclosed flux, not on the path shape, making it a gauge-potential holonomy — the same geometric structure as the Berry phase (Berry, 1984), where a cyclic adiabatic evolution accumulates a geometric phase depending only on the solid angle subtended by the parameter path. Tonomura et al. (1986) [Tonomura1986] provided the definitive experimental demonstration of the AB effect using electron holography with completely shielded magnetic flux. Note that in a ring geometry, only the azimuthal action integral acquires the flux shift \(e\Phi/(2\pi)\); the radial and polar action integrals are unaffected.

Remark 5.5 (Heisenberg limit: the action–angle tradeoff as a metrological resource). The Fourier tradeoff of Section 4 has a direct metrological consequence: the precision with which an angle (phase) can be estimated is limited by the spread of the conjugate action variable. For \(N\) independent probe particles (e.g., photons in an interferometer), each acquiring phase \(\varphi\) independently, the quantum Cramér–Rao bound with quantum Fisher information \(\mathcal{F}_Q=4\,\mathrm{Var}(\hat J)\) — where \(\hat J\) is the generator of the phase shift \(e^{-i\varphi\hat J}\) — gives a minimum phase uncertainty \(\delta\varphi\ge 1/\sqrt{\mathcal{F}_Q}\). For a coherent state (Poisson action distribution, \(\mathrm{Var}(\hat J)\approx N\)), this yields the shot-noise limit \(\delta\varphi\ge 1/\sqrt{N}\) (Braunstein and Caves, 1994). Entangled states can do better: a NOON state \((|N,0\rangle+|0,N\rangle)/\sqrt{2}\) achieves \(\mathrm{Var}(\hat J)=N^2/4\), saturating the Heisenberg limit \(\delta\varphi\ge 1/N\) — the ultimate bound for linear phase-encoding unitaries (Giovannetti, Lloyd, and Maccone, 2006). The gap between \(1/\sqrt{N}\) (shot noise) and \(1/N\) (Heisenberg) is precisely the gap between classical and quantum exploitation of the action–angle tradeoff: the more action variance a state carries, the finer the angle resolution it permits. This is the Fourier tradeoff of Section 4 in its most operationally consequential form.

Remark 5.6 (Quantum Zeno effect: dynamical enforcement of the action–angle tradeoff). The static statement of Section 3 — sharp action implies delocalized angle — has a dynamical counterpart in the quantum Zeno effect (Misra and Sudarshan, 1977). Frequent projective measurement of the action variable (\(\hat N\) or \(L_z\)) keeps the system in an action eigenstate: the short-time survival probability after each measurement interval \(\delta t=T/N\) is \(p(\delta t)\approx 1-(\delta t\,\Delta H/\hbar)^2\), so after \(N\) measurements the total survival probability \(P(T)\approx[1-(\Delta H)^2 T^2/(\hbar^2 N^2)]^N\to 1\) as \(N\to\infty\). The system is frozen in its action eigenstate, which has a uniformly delocalized angle distribution — not because the measurement “randomizes” the angle, but because it prevents the system from building up the superposition of different action eigenstates that would be required for angle localization (Section 4). This is complementary to Remark 5.2 (decoherence): position-coupled decoherence selects coherent states (localized angle, spread action), while frequent action measurement selects Fock/\(L_z\) eigenstates (localized action, delocalized angle) — two opposite dynamical mechanisms that each enforce one extreme of the action–angle tradeoff. The Zeno effect was first demonstrated experimentally by Itano et al. (1990) using repeated optical pulses on trapped ions to inhibit transitions between internal states.

Remark 5.7 (QND measurement of action variables). Quantum nondemolition (QND) measurement completes the measurement story of Remarks 5.2 (decoherence) and 5.6 (Zeno effect). A QND observable \(A\) satisfies two conditions: \([A,H_{\mathrm{sys}}]=0\) (the observable is a constant of the free motion) and \([A,H_{\mathrm{int}}]=0\) (the measurement interaction does not feed back into \(A\)). Together these ensure that repeated measurement of \(A\) yields the same result without disturbing \(A\)’s subsequent evolution — though the conjugate variable is randomized by each measurement (Braginsky, Vorontsov, and Thorne, 1980). For action variables (\(\hat N\) or \(L_z\)), the first condition is automatic: they commute with their respective system Hamiltonians by construction. The second condition constrains the measurement apparatus: the interaction must couple to \(A\) without generating terms that fail to commute with it. The canonical experimental realization is dispersive photon-number readout in cavity QED (Nogues et al., 1999): Rydberg atoms traverse a microwave cavity and acquire a phase shift \(\Delta\varphi=\chi\,n\,t_{\mathrm{int}}\) proportional to the photon number \(n\) via the dispersive interaction \(H_{\mathrm{int}}=\hbar\chi\,\hat n\,\sigma_z\). Ramsey interferometry on the atomic superposition state then reveals \(n\) — distinguishing \(n=0\) from \(n=1\) — without absorbing or emitting photons. QND measurement is conceptually distinct from the Zeno effect (Remark 5.6): Zeno freezing relies on rapid projective measurements that prevent any evolution, while QND exploits the fact that the measured observable is already conserved — the measurement reads the value without needing to freeze the dynamics. Decoherence (Remark 5.2) selects coherent states via position coupling; QND measurement selects Fock states via action coupling. The choice of coupling determines which extreme of the action–angle tradeoff is enforced.

6. A second witness: the harmonic oscillator

The same structure appears in the simplest one-dimensional integrable system.

Example 6.1 (Harmonic oscillator: Fock states vs coherent states). For a harmonic oscillator of frequency \(\omega\), define the classical action variable \(J=E/\omega\). The quantum Fock states \(|n\rangle\) are the action eigenstates (\(J_n=(n+\tfrac12)\hbar\)), and their phase-space (Husimi) distribution is a ring centered at the origin — the orbit phase \(\theta\) is uniformly delocalized, exactly as \(\phi\) is delocalized in an \(L_z\) eigenstate. Conversely, a coherent state \[ |\alpha\rangle =e^{-|\alpha|^2/2}\sum_{n=0}^{\infty}\frac{\alpha^n}{\sqrt{n!}}\,|n\rangle, \qquad \alpha=|\alpha|\,e^{i\theta_0}, \] is the closest quantum analog of a classical orbit with definite amplitude \(|\alpha|\) and phase \(\theta_0\). Its Fock-state weights follow a Poisson distribution with mean \(\bar n=|\alpha|^2\), so localizing the phase to width \(\Delta\theta\sim 1/|\alpha|\) requires spreading the action over \(\Delta n\sim |\alpha|\) modes. The tradeoff is the same as in Section 4: sharp action implies delocalized phase, and vice versa.

Example 6.2 (Hydrogen atom: three action–angle pairs). In the hydrogen atom, the \(n^2\)-fold degeneracy (\(E_n\) depending only on the principal quantum number \(n\)) reflects an enhanced \(SO(4)\) symmetry [Sakurai2020]. Semiclassically, the bound orbits lie on a three-torus with action integrals quantized by \((n,\ell,m)\). A stationary eigenstate \(|n,\ell,m\rangle\) is sharp in all three actions and therefore delocalized in all three conjugate angles: the azimuthal phase \(\phi\) is uniform (Section 3), the in-plane orbit orientation has no preferred direction (the Runge–Lenz vector has vanishing expectation value: \(\langle n,\ell,m|\mathbf{A}|n,\ell,m\rangle = 0\) because \(\mathbf{A}\) changes \(\ell\) by \(\pm 1\) within the \(n\)-shell, so the bra and ket are orthogonal), and the radial probability \(|R_{n\ell}(r)|^2\) is time-independent — the sharp radial action leaves the conjugate radial phase uniformly delocalized. A classical Keplerian ellipse with definite eccentricity, orientation, and timing requires a coherent superposition over ranges of \((n,\ell,m)\), just as a coherent state in Example 6.1 requires superposing many Fock states.

Remark 6.3 (EBK quantization on the invariant torus). For a classically integrable system with \(d\) degrees of freedom, the Arnold–Liouville theorem provides \(d\) action variables \(I_k=(2\pi)^{-1}\oint_{\gamma_k}p\cdot dq\), integrated around the independent cycles \(\gamma_k\) of the invariant \(d\)-torus. The EBK (Einstein–Brillouin–Keller) quantization condition requires \[ I_k=\left(n_k+\frac{\alpha_k}{4}\right)\hbar,\qquad n_k\in\mathbb{Z}_{\ge0}, \] where \(\alpha_k\) is the Maslov index of the \(k\)-th cycle (counting caustic/turning-point contributions). The integer quantum numbers \(n_k\) select the torus; the conjugate angle variables \(\theta_k\in[0,2\pi)\) are uniformly distributed on that torus and carry no quantum-number information. This is the semiclassical counterpart of the fully quantum statement: stationary eigenstates (sharp actions) have delocalized angles. Examples 6.1 and 6.2 are the exact quantum versions of this principle for the \(d=1\) and \(d=3\) cases.

Remark 6.4 (Husimi function: visualizing action–angle states in phase space). The Husimi \(Q\)-function \(Q(\alpha)=\langle\alpha|\hat\rho|\alpha\rangle/\pi\) assigns a non-negative quasiprobability to each phase-space point \(\alpha\), using coherent states as the reference frame. For a Fock state \(|n\rangle\), \(Q(\alpha)=e^{-|\alpha|^2}|\alpha|^{2n}/(\pi\,n!)\) — a ring at radius \(|\alpha|=\sqrt{n}\), uniform in the phase angle: sharp action, fully delocalized angle. For a coherent state \(|\alpha_0\rangle\), \(Q(\alpha)=e^{-|\alpha-\alpha_0|^2}/\pi\) — a Gaussian blob centered at \(\alpha_0\), simultaneously localizing both action and angle to uncertainty-limited width. The ring-versus-blob distinction is the phase-space portrait of the Fourier tradeoff in Section 4, with the Husimi function providing a literal (non-negative) probability picture that the Wigner function’s sign changes would obscure.

Remark 6.5 (Squeezed states: continuous interpolation between ring and blob). The Fock ring and coherent blob are not the only options; squeezed states of the form \(D(\alpha_0)S(\xi)|0\rangle\) (displacement followed by squeezing) produce an elliptical Husimi contour at \(\alpha_0\), with the squeeze parameter \(r=|\xi|\) controlling the eccentricity. When the ellipse is aligned radially (amplitude squeezing), action uncertainty is reduced at the expense of angle uncertainty — approaching Fock-ring character. When aligned tangentially (phase squeezing), angle uncertainty is reduced at the expense of action uncertainty — giving a better approximation to a classical orbit with definite phase. The full family, parametrized by \(r\) and the squeeze angle, interpolates continuously between the extremes of Examples 6.1 and 6.4 while saturating the Robertson bound \(\Delta X_1\,\Delta X_2=\hbar/2\) for the dimensionless quadratures \(X_1=(a+a^\dagger)/\sqrt{2}\), \(X_2=(a-a^\dagger)/(i\sqrt{2})\) at every point (minimum-uncertainty states for these quadratures).

Remark 6.6 (Wigner function: sub-Planck structure beneath the Husimi portrait). The Husimi function is the Wigner function convolved with a Gaussian of width \(\sqrt{\hbar}\): this smoothing guarantees non-negativity (Remark 6.4) but erases interference fringes. For a Fock state \(|n\rangle\), the Wigner function \(W_n(x,p)=((-1)^n/\pi\hbar)\,L_n(2H/\hbar\omega)\,e^{-2H/\hbar\omega}\) (where \(H=(p^2+\omega^2 x^2)/2\) and \(L_n\) is a Laguerre polynomial) exhibits alternating-sign rings — structure finer than the minimal uncertainty cell \(\Delta x\,\Delta p\sim\hbar\) that the non-negative Husimi ring completely hides. The Wigner portrait thus complements Remarks 6.4–6.5: Husimi shows where the state “is” in phase space, while Wigner reveals the quantum coherences that make “where” an incomplete description.

Remark 6.7 (Quantum state tomography: reconstructing the phase-space portrait). The Husimi and Wigner functions of Remarks 6.4–6.6 are not merely theoretical constructs: they can be reconstructed from experimental data via quantum state tomography. For an oscillator mode, measuring the rotated quadrature \(X_\theta=x\cos\theta+p\sin\theta\) at angle \(\theta\) yields a marginal distribution \(P_\theta(s)\) that is a Radon projection of the Wigner function. Collecting \(P_\theta\) for all \(\theta\in[0,\pi)\) and inverting the Radon transform (filtered back-projection) recovers the full \(W(x,p)\). This was first demonstrated by Smithey, Beck, Raymer, and Faridani (1993), who reconstructed the Wigner function of a squeezed vacuum state using optical homodyne detection, where the local oscillator phase selects the measured quadrature angle. The ring-versus-blob distinction of Remark 6.4 is thus directly observable: Fock-state tomograms show the uniform-phase ring, while coherent-state tomograms show the localized Gaussian blob — the action-angle tradeoff of Sections 3–4 made experimentally visible.

Remark 6.8 (Spin coherent states: the angular-momentum analog of coherent-state localization). The harmonic-oscillator coherent states of Example 6.1 have a direct angular-momentum counterpart. The spin coherent state \(|j,\hat n\rangle = R(\hat n)|j,j\rangle\) — obtained by rotating the highest-weight state to point along \(\hat n=(\theta_0,\phi_0)\) (Radcliffe, 1971; Perelomov, 1972) — has \(m\)-coefficients that follow a binomial distribution with mean \(\bar m=j\cos\theta_0\) and variance \(\mathrm{Var}(m)=(j/2)\sin^2\theta_0\). The expectation value of the angular-momentum vector is \(\langle\mathbf J\rangle=j\hbar\,\hat n\), with angular uncertainty \(\Delta\theta\sim 1/\sqrt{2j}\) that saturates the SU(2) Robertson bound \(\Delta J_1\,\Delta J_2\ge(\hbar/2)|\langle J_3\rangle|\) for the pair \((J_1, J_2)\) perpendicular to the mean spin direction \(\hat n\). As \(j\to\infty\), the spin coherent states become delta-concentrated on the classical phase space \(S^2\) (Lieb, 1973) — the angular-momentum analog of the \(|\alpha|\to\infty\) classical limit of Example 6.1. The ring-versus-blob picture of Remark 6.4 thus extends from the flat (oscillator) phase space to the sphere: \(|j,m\rangle\) eigenstates are azimuthal rings with uniform \(\phi\), while spin coherent states are directional blobs that spread over \(\sim\sqrt{j}\) magnetic sub-levels.

Remark 6.9 (Bargmann representation: action–angle duality made algebraic). The Bargmann–Segal representation (Bargmann, 1961) maps the oscillator Hilbert space to the space of holomorphic functions on \(\mathbb{C}\) with Gaussian measure \(d\mu=\pi^{-1}e^{-|z|^2}d^2z\). Under this map, Fock states become monomials — \(|n\rangle\mapsto z^n/\sqrt{n!}\) — whose rotational symmetry \(|z^n|^2=|z|^{2n}\) reflects the uniform phase distribution of Remark 6.4’s ring. Coherent states become reproducing kernels \(K_\alpha(z)=e^{\bar\alpha z}\), which extract the value of a holomorphic function at \(\alpha\) — the algebraic expression of “evaluation at a phase-space point,” i.e., maximal localization. Creation and annihilation become multiplication and differentiation: \(\hat a^\dagger\mapsto z\), \(\hat a\mapsto d/dz\). The Husimi function of Remark 6.4 is then the squared modulus of the Bargmann representative (weighted by the Gaussian measure), so the ring-versus-blob distinction of Remarks 6.4–6.8 is literally monomials versus peaked exponentials. For spin-\(j\) systems (Schwinger’s oscillator construction), the Bargmann space truncates to polynomials of degree \(\le 2j\), and spin coherent states become evaluation points on \(\mathbb{CP}^1\cong S^2\) — the finite-dimensional counterpart of the full oscillator duality.

Remark 6.10 (Hamiltonian monodromy: topological obstruction to global action-angle coordinates). The EBK quantization of Remark 6.3 and the Bargmann representation of Remark 6.9 assume that action-angle coordinates exist globally. Duistermaat (1980) showed that this can fail: for a completely integrable system, the regular invariant tori form a torus bundle over the space of regular values of the energy-momentum map, and this bundle may be non-trivial. Hamiltonian monodromy — a non-trivial \(\mathrm{GL}(d,\mathbb{Z})\) transformation of the action variables upon transport around a loop enclosing a singular value (such as a focus-focus equilibrium) — obstructs the existence of globally smooth action variables. The quantum manifestation (Cushman and Duistermaat, 1988; V\~u Ng\d{o}c, 1999) is that the joint spectrum of commuting observables cannot be embedded in a regular integer lattice: the quantum number lattice has a topological defect around which the labeling undergoes the same \(\mathrm{GL}(d,\mathbb{Z})\) transformation as the classical actions. The spherical pendulum provides the canonical example, with a single focus-focus point producing a lattice defect visible in the \((E,L_z)\) spectrum. This is “action-angle indeterminacy” at the global-topological level: the action variables of Sections 3–4 are locally well-defined but may not extend consistently around singular fibers of the energy-momentum map.

7. Outlook (kept minimal)

The preceding witnesses illustrate the action–angle tradeoff in systems with one, two, and three action–angle pairs, and Remark 6.3 shows that EBK quantization makes the same structural point in general: the more sharply the actions are specified, the less information remains in the conjugate phases.

Remark 7.1 (Boundary at integrability breaking). The action–angle framework of Sections 3–6 presupposes the existence of global action variables, hence applies exactly to integrable systems. For nearly integrable Hamiltonians, the KAM theorem guarantees persistence of most invariant tori (those with sufficiently irrational frequency ratios), on which the framework remains valid. In fully chaotic systems, the absence of conserved actions replaces the structured Fourier tradeoff with eigenstate thermalization — a more drastic delocalization where individual energy eigenstates appear thermal for local observables, with no residual action–angle structure to organize the uncertainty.

Remark 7.2 (Spectral statistics: from Poisson to random-matrix universality). The action–angle structure of Sections 3–6 also organizes the energy spectrum. Berry and Tabor (1977) showed that for a generic integrable system (incommensurable classical frequencies), EBK quantization on independent tori produces energy levels whose nearest-neighbor spacings follow a Poisson distribution \(P(s)=e^{-s}\) — uncorrelated, with no level repulsion. Bohigas, Giannoni, and Schmit (1984) conjectured (and verified numerically for the Sinai billiard) that classically chaotic systems instead display the level repulsion characteristic of random matrix theory: GOE statistics for time-reversal invariant systems, GUE when time-reversal is broken. The spectral transition — Poisson to Wigner-Dyson — mirrors the phase-space transition of Remark 7.1: independent torus quantization gives way to the global eigenstate entanglement that produces both level repulsion and eigenstate thermalization.

Remark 7.3 (Ehrenfest time: temporal boundary of the classical orbit picture). The Ehrenfest theorem — \(d\langle x\rangle/dt=\langle p\rangle/m\), \(d\langle p\rangle/dt=-\langle V’(x)\rangle\) — gives exact classical dynamics only when the potential is at most quadratic (so that \(\langle V’(x)\rangle=V’(\langle x\rangle)\)). For anharmonic systems, the wavepacket’s finite width introduces corrections that grow with time, and the classical-orbit picture breaks down after the Ehrenfest time \(t_E\). For integrable systems, the characteristic quantum timescales — both the dispersal time and the revival time \(t_{\mathrm{rev}}\sim 2\pi\hbar/|d^2E/dm^2|\) of Remark 5.1 — are algebraic in \(\hbar\). For classically chaotic systems with maximal Lyapunov exponent \(\lambda\), exponential trajectory divergence compresses this to \(t_E\sim(1/\lambda)\ln(1/\hbar)\) — logarithmic in \(1/\hbar\), hence far shorter (Berman and Zaslavsky, 1978). This dramatic contrast — algebraic versus logarithmic — is the temporal counterpart of the phase-space and spectral transitions described in Remarks 7.1 and 7.2: the coherent packets of Section 5 remain classical for polynomially long times in integrable systems, but only logarithmically long in chaotic ones.

Remark 7.4 (Gutzwiller trace formula: periodic orbits as the non-integrable spectral bridge). The EBK quantization of Remark 6.3 connects eigenvalues to classical tori in integrable systems. For non-integrable (chaotic) systems, invariant tori do not exist, and the Gutzwiller trace formula (1971) replaces them with isolated periodic orbits: the oscillatory part of the density of states is \(g_{\mathrm{osc}}(E)\approx\sum_\gamma A_\gamma\cos(S_\gamma(E)/\hbar-\mu_\gamma\pi/2)\), where the sum runs over all periodic orbits \(\gamma\) (primitive and repeated), \(S_\gamma=\oint_\gamma p\cdot dq\) is the classical action, \(\mu_\gamma\) is the Maslov index, and the amplitude \(A_\gamma\propto T_\gamma/|\det(M_\gamma-I)|^{1/2}\) involves the orbit period \(T_\gamma\) and the stability determinant of the monodromy matrix \(M_\gamma\) restricted to the transverse directions. For marginally stable orbits (\(M_\gamma\) has eigenvalue 1), this amplitude diverges — these are the torus orbits of integrable systems, and the trace formula degenerates to EBK quantization. This formula also underlies the spectral statistics of Remark 7.2: the Poisson statistics of integrable systems follow from the incommensurate action phases on independent tori, while the random-matrix (GOE/GUE) statistics of chaotic systems emerge from the exponentially proliferating periodic orbits (their number grows as \(\sim e^{hT}/T\) with topological entropy \(h\)) through semiclassical sum rules (Berry, 1985). The Gutzwiller formula is generically asymptotic (\(\hbar\to 0\)), but for surfaces of constant negative curvature, the Selberg trace formula provides an exact identity between the Laplacian spectrum and closed geodesics.

Remark 7.5 (Geometric quantization: the action-angle framework made systematic). The action-angle structure underlying Sections 2–6 has a systematic mathematical home in geometric quantization (Kostant and Souriau, 1970). Given a symplectic manifold \((M,\omega)\), the prequantization condition requires \([\omega/(2\pi\hbar)]\) to be an integral cohomology class — precisely the EBK quantization condition of Remark 6.3, expressed as integrality of the Chern class of a line bundle \(L\to M\). A polarization — a choice of Lagrangian foliation — then reduces the prequantum Hilbert space to the physical one. For integrable systems with action-angle coordinates \((I,\theta)\), the real (action) polarization gives wave functions depending on \(I\) alone, yielding the number/Fock states of Example 6.1 — the “rings” of Remark 6.4 — while the complex (Kähler) polarization gives holomorphic wave functions, recovering the Bargmann representation of Remark 6.9 — the “blobs.” The metaplectic correction (Blattner, 1977; Guillemin and Sternberg, 1977 [GuilleminSternberg1977]) accounts for the half-density structure of the cornerstone manuscript, adding the \(\tfrac12\hbar\omega\) zero-point energy that naive geometric quantization misses. For spin-\(j\) systems, geometric quantization of \(S^2\cong\mathbb{CP}^1\) with symplectic form proportional to the area form produces the spin coherent states of Remark 6.8, with the prequantization condition forcing \(2j\in\mathbb{Z}\) — the integrality of spin as a topological constraint.

Remark 7.5a (Representation dependence of the t^{-d/2} normalization). The composition law for the quantum propagator forces the normalization \(N(t)\sim t^{-d/2}\) in position space [PathIntegralNormalization], but this forcing is representation-dependent. In action-angle coordinates \((I,\theta)\), the Van Vleck determinant satisfies \(\det(-\partial^2 S_{\mathrm{cl}}/\partial I_j\partial\theta’k) = 1\) for \(S{\mathrm{cl}} = -H(I)t + I\cdot(\theta-\theta’)\), so the \(t^{-n/2}\) factor does not arise from the action-angle Van Vleck factor; it arises instead from the energy-surface curvature \(|\det(t\,\Omega)|^{-1/2}\), where \(\Omega_{jk}=\partial^2 H/\partial I_j\partial I_k\) is the frequency matrix. The \(t^{-d/2}\) forcing is therefore maximally transparent in position-momentum representation and hidden in action-angle representation — a representation-dependence observation, not a weakness of the forcing argument. For the full treatment, see [PathIntegralNormalization].

A separate composition-law constraint that is natural in action-angle coordinates: Bohr-Sommerfeld quantization. Single-valuedness of \(K_t\) under \(\theta\to\theta+2\pi\) forces \(I = \hbar\cdot m\) (\(m\in\mathbb{Z}^n\)), recovering EBK quantization (Remark 6.3) directly from the periodicity of the propagator.

Remark 7.6 (Coherent-state path integral: the action-angle structure as a path integral). The coherent states of Example 6.1 provide not only a phase-space portrait (Remarks 6.4–6.9) but also a path integral formulation that makes the action-angle structure dynamical. The resolution of identity \(1=\pi^{-1}\int|\alpha\rangle\langle\alpha|\,d^2\alpha\) allows writing the quantum propagator as a functional integral over phase-space trajectories: \(K=\int\mathcal D^2\alpha\;\exp(iS_{\mathrm{cs}}/\hbar)\), where the coherent-state action is \(S_{\mathrm{cs}}=\int dt\,[i\hbar\,\dot\alpha^\alpha - H(\alpha^,\alpha)]\). The kinetic term \(i\hbar\,\dot\alpha^\alpha\) is the symplectic one-form \(p\,dq\) written in the complex coordinate \(\alpha=(q+ip)/\sqrt{2\hbar}\) — the same geometric connection whose holonomy gives the Berry phase (Berry, 1984). In the semiclassical limit, stationary phase selects trajectories satisfying Hamilton’s equations in complex form \(i\hbar\,\dot\alpha=\partial H/\partial\alpha^\); fluctuations around these yield the semiclassical corrections that connect to the Gutzwiller trace formula of Remark 7.4. For spin systems, replacing oscillator coherent states by spin coherent states (Remark 6.8) gives a path integral on \(S^2\) with a Wess–Zumino topological term that encodes the spin quantum number \(j\) — the path-integral expression of the prequantization integrality condition in Remark 7.5.

References

  1. [TongQMLectures] David Tong, “Quantum Mechanics” (lecture notes, University of Cambridge, 2023). URL: https://www.damtp.cam.ac.uk/user/tong/quantum.html (retrieved 2026-02-21). (Contains \(Y_{l,m}(\theta,\phi)=P_{l,m}(\cos\theta)e^{im\phi}\) as simultaneous eigenstates of \(L^2\) and \(L_z\).)
  2. [Sakurai2020] J. J. Sakurai and Jim Napolitano, Modern Quantum Mechanics, 3rd ed., Cambridge University Press, 2020. ISBN 978-1-108-47322-4. (Standard treatment of angular momentum, spherical harmonics, and quantum measurement.)
  3. [ZurekHabibPaz1993] W. H. Zurek, S. Habib, and J. P. Paz, “Coherent States via Decoherence,” Physical Review Letters 70 (1993), 1187–1190. DOI 10.1103/PhysRevLett.70.1187. (Shows coherent states minimize entropy production under environmental coupling, emerging as preferred pointer states.)
  4. [Bargmann1961] Valentine Bargmann, “On a Hilbert space of analytic functions and an associated integral transform, Part I,” Communications on Pure and Applied Mathematics 14 (1961), 187–214. DOI 10.1002/cpa.3160140303. (Bargmann–Segal holomorphic representation; used in Remark 6.9.)
  5. [Perelomov1972] A. M. Perelomov, “Coherent states for arbitrary Lie group,” Communications in Mathematical Physics 26 (1972), 222–236. DOI 10.1007/BF01645091. (Generalized coherent states including spin coherent states; used in Remark 6.8.)
  6. [MandelstamTamm1945] L. I. Mandelstam and I. E. Tamm, “The uncertainty relation between energy and time in non-relativistic quantum mechanics,” Journal of Physics (USSR) 9 (1945), 249–254. (Energy–time uncertainty relation; used in Remark 5.3.)
  7. [BerryTabor1977] M. V. Berry and M. Tabor, “Level clustering in the regular spectrum,” Proceedings of the Royal Society A 356 (1977), 375–394. DOI 10.1098/rspa.1977.0140. (Poisson statistics for integrable systems; used in Remark 7.2.)
  8. [SusskindGlogower1964] Leonard Susskind and Jonathan Glogower, “Quantum mechanical phase and time operator,” Physics 1 (1964), 49–61. (Number–phase operators; used in Remark 2.1.)
  9. [Radcliffe1971] J. M. Radcliffe, “Some properties of coherent spin states,” Journal of Physics A 4 (1971), 313–323. DOI 10.1088/0305-4470/4/3/009. (Spin coherent states on S²; used in Remark 6.8.)
  10. [Lieb1973] Elliott H. Lieb, “The classical limit of quantum spin systems,” Communications in Mathematical Physics 31 (1973), 327–340. DOI 10.1007/BF01646493. (Classical limit of spin coherent states; used in Remark 6.8.)
  11. [SmitheyBeckRaymerFaridani1993] D. T. Smithey, M. Beck, M. G. Raymer, and A. Faridani, “Measurement of the Wigner distribution and the density matrix of a light mode using optical homodyne tomography: Application to squeezed states and the vacuum,” Physical Review Letters 70 (1993), 1244–1247. DOI 10.1103/PhysRevLett.70.1244. (First optical homodyne tomography; used in Remark 6.7.)
  12. [BermanZaslavsky1978] G. P. Berman and G. M. Zaslavsky, “Condition of stochasticity in quantum non-linear systems,” Physica A 91 (1978), 76–94. DOI 10.1016/0378-4371(78)90059-5. (Ehrenfest time logarithmic scaling; used in Remark 7.3.)
  13. [Gutzwiller1971] Martin C. Gutzwiller, “Periodic orbits and classical quantization conditions,” Journal of Mathematical Physics 12 (1971), 343–358. DOI 10.1063/1.1665596. (Gutzwiller trace formula; used in Remark 7.4.)
  14. [Berry1985] M. V. Berry, “Semiclassical theory of spectral rigidity,” Proceedings of the Royal Society A 400 (1985), 229–251. DOI 10.1098/rspa.1985.0078. (Periodic-orbit spectral sum rules; used in Remark 7.4.)
  15. [BohigasGiannoniSchmit1984] O. Bohigas, M. J. Giannoni, and C. Schmit, “Characterization of chaotic quantum spectra and universality of level fluctuation laws,” Physical Review Letters 52 (1984), 1–4. DOI 10.1103/PhysRevLett.52.1. (BGS conjecture: chaotic spectra follow random-matrix statistics; used in Remark 7.2.)
  16. [KostantSouriau1970] Bertram Kostant, “Quantization and unitary representations,” Lecture Notes in Mathematics 170 (1970), 87–208, Springer; Jean-Marie Souriau, Structure des systèmes dynamiques, Dunod, 1970. (Geometric quantization foundations; used in Remark 7.5.)
  17. [Tonomura1986] A. Tonomura, N. Osakabe, T. Matsuda, T. Kawasaki, J. Endo, S. Yano, and H. Yamada, “Evidence for Aharonov-Bohm effect with magnetic field completely shielded from electron wave,” Physical Review Letters 56 (1986), 792–795. DOI 10.1103/PhysRevLett.56.792. (Definitive AB experimental demonstration; used in Remark 5.4.)
  18. [AharonovBohm1959] Y. Aharonov and D. Bohm, “Significance of electromagnetic potentials in the quantum theory,” Physical Review 115 (1959), 485–491. DOI 10.1103/PhysRev.115.485. (Original Aharonov–Bohm prediction; used in Remark 5.4.)
  19. [Duistermaat1980] J. J. Duistermaat, “On global action-angle coordinates,” Communications on Pure and Applied Mathematics 33 (1980), 687–706. DOI 10.1002/cpa.3160330602. (Hamiltonian monodromy; used in Remark 6.10.)
  20. [CushmanDuistermaat1988] Richard H. Cushman and J. J. Duistermaat, “The quantum mechanical spherical pendulum,” Bulletin of the American Mathematical Society 19 (1988), 475–479. DOI 10.1090/S0273-0979-1988-15705-9. (Quantum monodromy; used in Remark 6.10.)
  21. [VuNgoc1999] San V\~u Ng\d{o}c, “Quantum monodromy in integrable systems,” Communications in Mathematical Physics 203 (1999), 465–479. DOI 10.1007/s002200050621. (Quantum lattice defect from classical monodromy; used in Remark 6.10.)
  22. [BialynickiBirulaMadajczyk1985] Iwo Bialynicki-Birula and Jerzy L. Madajczyk, “Entropic uncertainty relations for angular distributions,” Physics Letters A 108 (1985), 384–386. DOI 10.1016/0375-9601(85)90277-4. (Entropic uncertainty for angle-angular momentum; used in Remark 4.5.)
  23. [BialynickiBirulaMycielski1975] Iwo Bialynicki-Birula and Jerzy Mycielski, “Uncertainty relations for information entropy in wave mechanics,” Communications in Mathematical Physics 44 (1975), 129–132. DOI 10.1007/BF01608825. (Entropic uncertainty for position-momentum; used in Remark 4.5.)
  24. [Beckner1975] William Beckner, “Inequalities in Fourier analysis,” Annals of Mathematics 102 (1975), 159–182. DOI 10.2307/1970980. (Sharp Hausdorff–Young inequality; used in Remark 4.5.)
  25. [BraunsteinCaves1994] Samuel L. Braunstein and Carlton M. Caves, “Statistical distance and the geometry of quantum states,” Physical Review Letters 72 (1994), 3439–3443. DOI 10.1103/PhysRevLett.72.3439. (Quantum Fisher information and Cramér–Rao bound; used in Remark 5.5.)
  26. [GiovannettiLloydMaccone2006] Vittorio Giovannetti, Seth Lloyd, and Lorenzo Maccone, “Quantum metrology,” Physical Review Letters 96 (2006), 010401. DOI 10.1103/PhysRevLett.96.010401. (Heisenberg limit for phase estimation; used in Remark 5.5.)
  27. [Holevo1982] Alexander S. Holevo, Probabilistic and Statistical Aspects of Quantum Theory, North-Holland, 1982. ISBN 0-444-86333-8. (POVM framework for quantum measurement; phase POVM; used in Remark 2.2.)
  28. [MisraSudarshan1977] Baidyanath Misra and E. C. George Sudarshan, “The Zeno’s paradox in quantum theory,” Journal of Mathematical Physics 18 (1977), 756–763. DOI 10.1063/1.523304. (Quantum Zeno effect; used in Remark 5.6.)
  29. [Itano1990] Wayne M. Itano, Daniel J. Heinzen, John J. Bollinger, and David J. Wineland, “Quantum Zeno effect,” Physical Review A 41 (1990), 2295–2300. DOI 10.1103/PhysRevA.41.2295. (First experimental demonstration of Zeno effect; used in Remark 5.6.)
  30. [BraginskyVorontsovThorne1980] V. B. Braginsky, Yu. I. Vorontsov, and K. S. Thorne, “Quantum nondemolition measurements,” Science 209 (1980), 547–557. DOI 10.1126/science.209.4456.547. (QND measurement theory; used in Remark 5.7.)
  31. [Nogues1999] C. Nogues, A. Rauschenbeutel, S. Osnaghi, M. Brune, J. M. Raimond, and S. Haroche, “Seeing a single photon without destroying it,” Nature 400 (1999), 239–242. DOI 10.1038/22275. (Dispersive QND photon-number measurement; used in Remark 5.7.)
  32. [Allen1992] L. Allen, M. W. Beijersbergen, R. J. C. Spreeuw, and J. P. Woerdman, “Orbital angular momentum of light and the transformation of Laguerre-Gaussian laser modes,” Physical Review A 45 (1992), 8185–8189. DOI 10.1103/PhysRevA.45.8185. (OAM of Laguerre–Gauss beams; used in Remark 3.3.)
  33. [Mair2001] A. Mair, A. Vaziri, G. Weihs, and A. Zeilinger, “Entanglement of the orbital angular momentum states of photons,” Nature 412 (2001), 313–316. DOI 10.1038/35085529. (OAM entanglement; used in Remark 3.3.)
  34. [Gabor1946] Dennis Gabor, “Theory of communication,” Journal of the Institution of Electrical Engineers 93 (1946), 429–457. DOI 10.1049/ji-3-2.1946.0074. (Gabor limit, logons, time-frequency uncertainty; used in Remark 4.6.)
  35. [Daubechies1992] Ingrid Daubechies, Ten Lectures on Wavelets, SIAM, 1992. ISBN 0-89871-274-2. (Gabor frames, Balian–Low theorem, time-frequency analysis; used in Remark 4.6.)
  36. [PeggBarnett1989] D. T. Pegg and S. M. Barnett, “Phase properties of the quantized single-mode electromagnetic field,” Physical Review A 39 (1989), 1665–1681. DOI 10.1103/PhysRevA.39.1665. (Hermitian phase operator via truncation and limit; used in Remark 2.3.)
  37. [Feynman1955] Richard P. Feynman, “Application of quantum mechanics to liquid helium,” in Progress in Low Temperature Physics, Vol. 1, ed. C. J. Gorter, North-Holland, 1955, pp. 17–53. (Quantized circulation in superfluid helium; used in Remark 3.4.)
  38. [Matthews1999] M. R. Matthews, B. P. Anderson, P. C. Haljan, D. S. Hall, C. E. Wieman, and E. A. Cornell, “Vortices in a Bose-Einstein condensate,” Physical Review Letters 83 (1999), 2498–2501. DOI 10.1103/PhysRevLett.83.2498. (First direct imaging of BEC vortex; used in Remark 3.4.)
  39. [Judge1964] D. Judge, “On the uncertainty relation for angle variables,” Il Nuovo Cimento 31 (1964), 332–340. DOI 10.1007/BF02733639. (First circular uncertainty relation for angle-angular momentum; used in Proposition 4.2.)
  40. [CarruthersNieto1968] P. Carruthers and M. M. Nieto, “Phase and angle variables in quantum mechanics,” Reviews of Modern Physics 40 (1968), 411–440. DOI 10.1103/RevModPhys.40.411. (Systematic treatment of angle uncertainty via circular statistics; framework for Proposition 4.2.)
  41. [MardiaJupp2000] Kanti V. Mardia and Peter E. Jupp, Directional Statistics, 2nd ed., Wiley, 2000. ISBN 0-471-95333-4. (Von Mises distribution properties, Bessel function moments; used in Remark 4.4.)
  42. [GuilleminSternberg1977] Victor Guillemin and Shlomo Sternberg, Geometric Asymptotics, AMS Mathematical Surveys 14, 1977. ISBN 0-8218-1514-8. (Metaplectic correction and half-density structure in geometric quantization; used in Remark 7.5.)
  43. [Berry1984] M. V. Berry, “Quantal phase factors accompanying adiabatic changes,” Proceedings of the Royal Society A 392 (1984), 45–57. DOI 10.1098/rspa.1984.0023. (Geometric/Berry phase; used in Remarks 5.4 and 7.6.)
  44. [PathIntegralNormalization] A. Rivero and A.I.Scaffold, “Path-Integral Normalization: The d/2 Exponent as Composition Compatibility Datum,” companion satellite paper in this series (2026). (Comprehensive treatment of the t^{-d/2} forcing; used in Remark 7.5a.)