The Isotropic 3-D Harmonic Oscillator

Two — and only two — central forces in the universe close every bounded orbit they create. Bertrand's theorem, which you have met, delivers the verdict: gravity's inverse square, and the linear restoring force $\vec F = -k\vec r$ of the isotropic three-dimensional harmonic oscillator. Newton knew both intimately — Book I of the Principia solves the spring-force orbit before it solves the planetary one — and the pairing is no accident of algebra. Each of the two closure champions harbours a hidden symmetry, a conservation law beyond energy and angular momentum: Kepler has its Laplace–Runge–Lenz vector, and the isotropic oscillator carries something richer still — a full conserved tensor, whose algebra closes into the group SU(3). The system itself could not be more physical. Every particle sitting near the bottom of any smooth, rotationally symmetric potential well — an atom in an optical tweezer, an ion in a trap, a nucleon in a light nucleus, an atom in a Bose–Einstein-condensate trap — is, to leading order, an isotropic 3-D harmonic oscillator. That ubiquity, plus its complete solvability by two different routes, makes it the perfect specimen: separate it in Cartesian coordinates and it falls apart into three identical one-dimensional oscillators whose synchrony closes the orbit; separate it in spherical coordinates and it becomes a central-force problem whose apsidal angle comes out at exactly $90°$ , every orbit an ellipse centred on the force centre — subtly and instructively different from Kepler's ellipse with the centre at a focus. This article works both routes to the end, quantifies the degeneracy that closure implies, and then opens the hood on the SU(3) machinery — the conserved quadrupole tensor, honestly derived, that turns "the orbit closes" into "the symmetry is bigger than rotation."

Key takeaways

The system: $H = \dfrac{p^2}{2m} + \dfrac12 m\omega^2 r^2$ — the same frequency $\omega$ in every direction; force $\vec F = -m\omega^2\vec r$ always aimed at the origin.
Cartesian route: separates into three independent 1-D oscillators, all at $\omega$ ; equal frequencies make every bounded orbit a closed ellipse centred at the origin, traversed in the amplitude-independent period $T = 2\pi/\omega$ .
Spherical route: central-force machinery gives turning points $r_\pm$ with $E = \frac12 m\omega^2(r_+^2 + r_-^2)$ , $L = m\omega\, r_+ r_-$ , and an apsidal angle of exactly $\pi/2$ : the radial oscillation runs at exactly twice the angular frequency, for every orbit.
Contrast with Kepler: ellipse centred at the force centre (not focus), two close approaches per revolution (not one), apsidal angle $90°$ (not $180°$ ) — the two Bertrand-closure potentials differ in geometry while sharing closure.
In action variables $E = \omega(J_x + J_y + J_z)$ : energy depends on one combination only — full degeneracy, the classical shadow of the quantum level degeneracy $(n+1)(n+2)/2$ .
The hidden symmetry made explicit: the tensor $A_{ij} = \dfrac{p_ip_j}{2m} + \dfrac{m\omega^2}{2}x_ix_j$ is conserved (two-line proof); its eigenvectors are the orbit's axes; its five traceless components plus $\vec L$ generate SU(3) — eight generators, superintegrability, orbit closure explained.
Key idea: equal frequencies in all three directions force every orbit to close into a centred ellipse; the closure is bankrolled by a conserved quadrupole tensor whose SU(3) algebra makes the isotropic oscillator, with Kepler, one of the two maximally hidden-symmetric systems of classical mechanics.

The core insight

The isotropic 3-D harmonic oscillator $H = p^2/2m + \tfrac12 m\omega^2r^2$ is exactly solvable both as three synchronized Cartesian oscillators and as a central-force problem: every bounded orbit is an origin-centred ellipse with amplitude-independent period $2\pi/\omega$ and apsidal angle $\pi/2$ , the radial-to-angular frequency ratio locks at 2:1 for all initial conditions, and the degeneracy is generated by the conserved symmetric tensor $A_{ij} = p_ip_j/2m + m\omega^2x_ix_j/2$ whose traceless part joins angular momentum to close the algebra of SU(3).

Where this lives today

The isotropic 3-D oscillator is the default Hamiltonian of trapped-particle physics. An optical tweezer holds a bead or a bacterium in a potential that is, to excellent approximation, $\tfrac12 m(\omega_\perp^2 r_\perp^2 + \omega_z^2 z^2)$ ; calibrating the trap means measuring those frequencies, and tweezer force spectroscopy — the tool behind single-molecule biophysics and the 2018 Nobel Prize — is applied harmonic-trap mechanics. Trapped ions, the leading quantum-computing platform, are oscillators in a Paul trap's effective potential: their shared motional modes, laser-coupled to internal states, are the quantum bus of every two-qubit gate, and "sideband cooling to the motional ground state" is this article's Hamiltonian pushed to $n = 0$ . Ultracold atomic gases live in harmonic confinement almost by definition: the celebrated breathing and quadrupole modes of Bose–Einstein condensates are collective versions of the 2:1-locked radial oscillation, and isotropy-engineering of the trap is a standard experimental art. In nuclear and particle physics the oscillator is structural. The nuclear shell model starts from a 3-D harmonic well: the oscillator's degenerate shells (2, 8, 20 nucleons cumulative — the first magic numbers) fall out of $(n+1)(n+2)/2$ with spin, and the heavier magic numbers emerge once spin-orbit coupling splits them; Elliott's SU(3) survives inside modern symmetry-adapted shell-model codes. Quark models used oscillator confinement for baryon spectroscopy for decades; molecular physicists expand every polyatomic vibration problem about equilibrium in coupled oscillators, with the isotropic case as the degenerate benchmark. And in the theory curriculum the system is the standing first example of superintegrability — the testbed where hidden-symmetry methods (Runge–Lenz-style operators, spectrum-generating algebras) are learned before being aimed at hydrogen, and the classical seed of the SU(3) technology that, in its flavor incarnation, reorganized particle physics.

The physical picture

Hang a mass from three mutually perpendicular ideal springs of equal stiffness, or — the modern realizations — hold an atom at the waist of a focused laser beam, an ion at the null of a trapping field, a nucleon in the mean field of a light nucleus. Near the equilibrium point, every such particle feels a pull proportional to displacement and aimed straight back at the centre: $\vec F = -m\omega^2\,\vec r$ . Isotropy is the crucial word — the same $\omega$ whichever way you displace. The generic smooth well is not isotropic (three different curvatures along three principal axes, giving a Lissajous tangle that never closes); isotropy is what symmetric traps engineer deliberately, and what makes this problem special rather than merely solvable. Watch what equal frequencies do to the orbit. Displace the particle and release it: 1-D oscillation through the centre. Give it a sideways kick as well: now $x(t)$ and $y(t)$ each oscillate at the same $\omega$ , and two perpendicular oscillations at one frequency compose into an ellipse — centred on the origin, traced once per period $T = 2\pi/\omega$ , forever. No precession, no rosette, no drift: after one period the particle returns exactly to its starting state whatever the amplitudes and phases. Compare the near-miss cousins: make the frequencies slightly unequal and the ellipse begins to tumble, filling a box densely (the generic quasiperiodic fate); make the force any other central power law (except $1/r^2$ ) and the ellipse precesses into a rosette. Closure for every initial condition is the rarest of properties — Bertrand's theorem certifies exactly two owners — and here you can see its mechanism with the naked eye: three clocks that tick at the same rate cannot drift out of step.

Figure

The two Bertrand ellipses — the oscillator's orbit is centred, Kepler's is focal, and the apsidal angles tell them apart

The closed orbit is a symptom; the disease (of the most benign kind) is degeneracy. In a generic central potential the radial and angular motions run at incommensurate frequencies that vary from orbit to orbit; here the ratio locks at exactly $2{:}1$ for every orbit — the radius completes two full breathing cycles per revolution, which is geometrically obvious once you see the centred ellipse: it has two near points and two far points per lap. Frequency-locking for all initial conditions is never an accident in mechanics. It advertises extra conserved quantities beyond the guaranteed $E$ and $\vec L$ — enough constants to pin the orbit's shape and orientation, not just its plane and size. The final section produces them explicitly: a conserved symmetric tensor whose eigen-directions are the ellipse's axes, doing for the oscillator exactly what the Laplace–Runge–Lenz vector does for Kepler, but with the richer algebra of SU(3) — a piece of classical mechanics that became, in Elliott's hands, the organizing symmetry of light-nucleus spectroscopy.

The math: one Hamiltonian, two exact solutions

The Hamiltonian and its defining split:

H = \frac{\vec p^{\,2}}{2m} + \frac12 m\omega^2 \vec r^{\,2} = \sum_{i=1}^{3}\left(\frac{p_i^2}{2m} + \frac12 m\omega^2 x_i^2\right) \qquad (1)

where: the second equality is the entire Cartesian miracle — the potential $\tfrac12 m\omega^2(x^2+y^2+z^2)$ carries no cross terms, so $H$ is a sum of three independent, identical one-dimensional harmonic oscillators. Nothing else in central-force physics does this: $-GMm/r = -GMm/\sqrt{x^2+y^2+z^2}$ mixes the coordinates hopelessly. The isotropic oscillator is the unique central potential that is simultaneously central and Cartesian-separable, which is why it alone gets two complete solution routes. The Cartesian solution and its orbit (Derivation A) can be summarized in one line each:

x_i(t) = A_i\cos(\omega t + \varphi_i), \qquad T = \frac{2\pi}{\omega}\ \text{for every orbit} \qquad (2)

where: the shared $\omega$ delivers spatial isochrony — the 3-D generalization of the 1-D oscillator's amplitude-independent period. All orbits, from tiny circles to grand eccentric ellipses, take identical time; in an ideal isotropic trap, a cloud of non-interacting atoms released off-centre re-converges to a point every half-period (the "breathing" of trapped gases, used daily to measure trap frequencies). The orbit these components trace is a centred ellipse with semi-axes $a \ge b$ set by initial conditions, and the conserved quantities have clean elliptical readings:

E = \frac12 m\omega^2\left(a^2 + b^2\right), \qquad L = m\,\omega\, a\, b \qquad (3)

where: $L$ is the angular momentum magnitude in the (automatically fixed) orbit plane. Two shapes with the same $a^2 + b^2$ share an energy shell; within it, $L$ ranges from $0$ (degenerate line orbit through the centre, $b = 0$ ) to the maximum $m\omega a^2$ at $a = b$ (circular orbit) — one energy, a continuum of shapes, the classical face of the degeneracy that SU(3) will organize. The spherical route treats (1) as a central force with effective potential

U_{\rm eff}(r) = \frac{L^2}{2mr^2} + \frac12 m\omega^2 r^2, \qquad r_\pm^2 = \frac{E \pm \sqrt{E^2 - \omega^2L^2}}{m\omega^2} \qquad (4)

where: the turning-point equation $E = U_{\rm eff}$ is a quadratic in $r^2$ — the technical reason everything about this system is elementary — with roots satisfying $r_+^2 + r_-^2 = 2E/m\omega^2$ and $r_+ r_- = L/m\omega$ , i.e. precisely the ellipse relations (3) with $r_+ = a$ , $r_- = b$ . The circular orbit sits at the minimum: $r_c^2 = L/m\omega$ , with energy $E_c = \omega L$ — the classical antecedent of the quantum oscillator's linear energy-angular-momentum bookkeeping. Derivation B extracts from (4) the apsidal angle $\Delta\varphi = \pi/2$ , exact for every $E$ and $L$ — the closure statement in spherical language.

Figure

The effective potential, computed — a quadratic in r² whose turning points are the ellipse's two semi-axes

Derivation ACartesian separation — three synchronized clocks and the centred ellipse

We are solving the isotropic oscillator completely in Cartesian coordinates and proving that every bounded orbit is an ellipse centred at the origin, traversed in the universal period $2\pi/\omega$ . The strategy: separate, solve each axis trivially, then reduce the three-component solution to a two-component one using angular-momentum conservation, and identify the resulting plane curve. Watch how little of the work is calculus — equal frequencies do everything.

Step 1. Separate. With (1) a sum of single-axis Hamiltonians, Hamilton's equations decouple: $\ddot x_i = -\omega^2 x_i$ for each $i$ , solved by $x_i(t) = A_i\cos(\omega t + \varphi_i)$ . Each axis conserves its own energy $E_i = \tfrac12 m\omega^2 A_i^2$ , with $E = E_1 + E_2 + E_3$ . Why separation is legitimate dynamics and not just algebra: no force component ever depends on the other coordinates — the three oscillators genuinely do not know about each other, and the only thing they share is the clock rate $\omega$ .

Step 2. Reduce to a plane. The force is central, so $\vec L = m\,\vec r\times\dot{\vec r}$ is conserved and the motion stays in the plane perpendicular to $\vec L$ (for $\vec L = 0$ : a line through the centre — the degenerate case). Choose axes so the plane is $z = 0$ : the orbit is $x = A_1\cos(\omega t + \varphi_1)$ , $y = A_2\cos(\omega t + \varphi_2)$ . Why we may also simplify the phases: shifting the time origin and rotating axes within the plane are free; use the time shift to set $\varphi_1 = 0$ , and write $\delta = \varphi_2 - \varphi_1$ for the one physical phase.

Step 3. Identify the curve. From $x = A_1\cos\omega t$ and $y = A_2\cos(\omega t + \delta) = A_2(\cos\omega t\cos\delta - \sin\omega t\sin\delta)$ , eliminate $t$ : $\cos\omega t = x/A_1$ and $\sin\omega t = \frac{1}{\sin\delta}\left(\frac{x}{A_1}\cos\delta - \frac{y}{A_2}\right)$ ; imposing $\cos^2 + \sin^2 = 1$ gives $\dfrac{x^2}{A_1^2} - \dfrac{2xy\cos\delta}{A_1A_2} + \dfrac{y^2}{A_2^2} = \sin^2\delta$ — a quadratic curve with bounded orbits: an ellipse, centred at the origin, with orientation set by $\delta$ . Rotating to its principal axes puts it in the standard form $x'^2/a^2 + y'^2/b^2 = 1$ , i.e. $x' = a\cos\omega t$ , $y' = b\sin\omega t$ . Why centred: the curve is invariant under $(x,y) \to (-x,-y)$ — the trajectory at $t + T/2$ is the antipode of the trajectory at $t$ . A Kepler ellipse has no such symmetry about the force centre; this is the geometric fingerprint distinguishing the two Bertrand potentials.

Step 4. Extract the conserved quantities in axis form. With $x' = a\cos\omega t$ , $y' = b\sin\omega t$ : energy $E = \tfrac12 m(\dot x'^2 + \dot y'^2) + \tfrac12 m\omega^2(x'^2 + y'^2) = \tfrac12 m\omega^2(a^2 + b^2)$ ; angular momentum $L = m(x'\dot y' - y'\dot x') = m\omega ab(\cos^2 + \sin^2) = m\omega ab$ — equations (3), with $L$ constant in time term by term, as it must be. Period: $2\pi/\omega$ regardless of $a, b$ — spatial isochrony. Why the special cases line up: $b = 0$ gives the line oscillation ( $L = 0$ ); $a = b$ the circle, where $E = \omega L$ — matching the effective-potential minimum of (4); everything in between shares the same clock. *Note: substituting $r^2 = a^2\cos^2\omega t + b^2\sin^2\omega t$ confirms that $r$ oscillates with period $\pi/\omega$ — half the orbital period, i.e. two radial cycles per lap, consistent with the $\pi/2$ apsidal angle derived next; and the per-axis energy bookkeeping $E_x = \tfrac12 m\omega^2a^2$ , $E_y = \tfrac12 m\omega^2 b^2$ sums correctly to (3).*

Go deeper · Derivation B · 2 minThe spherical route — turning points, and the exactly-90° apsidal angle

Derivation BThe spherical route — turning points, and the exactly-90° apsidal angle

We are re-solving the same system with the central-force machinery — effective potential, turning points, apsidal-angle integral — to obtain the closure property in the form Bertrand's theorem speaks: the apsidal angle (the swept angle between $r_{\max}$ and the next $r_{\min}$ ) equals $\pi/2$ exactly, for all $E$ and $L$ . The strategy: the substitution $s = r^2$ converts every integral in sight into a standard form, because the potential is quadratic — watch that substitution do in two lines what takes elliptic functions in any neighbouring power law.

Step 1. Set up the radial problem. Planar motion in polar coordinates with conserved $L$ : $E = \tfrac12 m\dot r^2 + U_{\rm eff}(r)$ with $U_{\rm eff}$ from (4). Turning points solve $E = U_{\rm eff}$ : multiply by $2r^2/m$ : $\ \omega^2 r^4 - \dfrac{2E}{m}r^2 + \dfrac{L^2}{m^2} = 0$ — quadratic in $r^2$ , roots $r_\pm^2$ as in (4), real whenever $E \ge \omega L$ (equality = circular orbit). Why quadratic-in- $r^2$ is the golden property: for a potential $\propto r^n$ the turning-point equation has degree $n + 2$ in general position — only $n = 2$ (here) and the Kepler case reduce to quadratics, and that shared algebraic mercy is the computational face of Bertrand's pairing.

Step 2. Write the apsidal angle: $\Delta\varphi = \displaystyle\int_{r_-}^{r_+} \frac{L/r^2}{\sqrt{2m\left[E - U_{\rm eff}(r)\right]}}\, dr$ — the standard central-force formula (angle swept while $r$ runs from one apsis to the other). Why this is the closure diagnostic: the orbit closes iff $\Delta\varphi$ is a rational multiple of $\pi$ for all orbits — a rational multiple for one special orbit is luck; for all orbits, it is symmetry.

Step 3. Substitute $s = r^2$ , $dr = ds/2\sqrt{s}$ : the radicand becomes $2mE - m^2\omega^2 s - L^2/s$ times... collecting under one square root, $\Delta\varphi = \dfrac{L}{2}\displaystyle\int_{s_-}^{s_+} \frac{ds}{s\,\sqrt{-m^2\omega^2 s^2 + 2mEs - L^2}}$ . This is the standard integral $\int \frac{ds}{s\sqrt{\alpha s^2 + \beta s + \gamma}}$ with $\alpha < 0$ , $\gamma = -L^2 < 0$ , whose value between the roots of the radicand is $\dfrac{\pi}{\sqrt{-\gamma}} = \dfrac{\pi}{L}$ . Therefore

\Delta\varphi = \frac{L}{2}\cdot\frac{\pi}{L} = \frac{\pi}{2} \qquad (5)

Why the independence from $E$ and $L$ is the whole point: every dependence cancelled — not approximately, not for special orbits. The particle sweeps a quarter turn between far point and near point, hence apsides at $0, \pi/2, \pi, 3\pi/2$ : the four vertex points of the centred ellipse, and closure after exactly one revolution. Equivalently $\omega_r = 2\omega_\varphi$ for every orbit — the universal 2:1 lock.

Step 4. Place the result in Bertrand's ledger. Near-circular orbits in a power-law potential $U \propto r^n$ have apsidal angle $\pi/\sqrt{n+2}$ : the values $\pi/2$ ( $n = 2$ ) and $\pi$ ( $n = -1$ ) are the only ones that survive the theorem's global analysis (rationality for all amplitudes, not just near-circular ones). The oscillator sits at the "two perihelia per lap" end; Kepler at "one per lap." Why a doubled radial frequency was guessable in advance: the centred ellipse's $r(t)$ has period- $\pi$ structure by the antipodal symmetry of Derivation A — geometry and the integral agree, as they must. *The standard integral can be evaluated directly: with roots $s_\pm$ of the radicand, $\int_{s_-}^{s_+}\frac{ds}{s\sqrt{(-\alpha)(s_+-s)(s-s_-)}} = \frac{\pi}{\sqrt{-\alpha\, s_+s_-}} = \frac{\pi}{\sqrt{-\gamma}}$ using $s_+s_- = \gamma/\alpha$ . With $\alpha = -m^2\omega^2$ and $\gamma = -L^2$ this gives $s_+s_- = L^2/m^2\omega^2$ , matching $r_+r_- = L/m\omega$ from (3), and the result is consistent with Derivation A's period- $\pi/\omega$ radial oscillation.*

Go deeper · Derivation C · 3 minThe hidden symmetry — a conserved tensor, superintegrability, and SU(3)

Derivation CThe hidden symmetry — a conserved tensor, superintegrability, and SU(3)

We are producing the conserved quantity that explains the universal closure: the symmetric tensor $A_{ij} = \dfrac{p_ip_j}{2m} + \dfrac{m\omega^2}{2}x_ix_j$ , proving its conservation directly, reading off its geometric meaning, counting the independent constants it adds (superintegrability), and sketching — honestly, at the level of what is computed versus what is quoted — how its components close the algebra of SU(3). Watch how short the conservation proof is: two lines, using nothing but the equations of motion.

Step 1. Differentiate along the motion, using $\dot x_i = p_i/m$ and $\dot p_i = -m\omega^2 x_i$ : $\dot A_{ij} = \dfrac{\dot p_i p_j + p_i\dot p_j}{2m} + \dfrac{m\omega^2}{2}(\dot x_i x_j + x_i\dot x_j) = -\dfrac{\omega^2}{2}(x_ip_j + p_ix_j) + \dfrac{\omega^2}{2}(p_ix_j + x_ip_j) = 0$ . Every component of $A_{ij}$ is a constant of the motion. Why such a thing can exist at all: each Cartesian oscillator conserves not just its energy but its full quadrature structure, and equal frequencies mean the CROSS-correlations between axes — $A_{xy}$ and its cousins — also beat at zero frequency. Detune the axes and $A_{xy}$ oscillates at the difference frequency: anisotropy kills the hidden symmetry exactly where the closed orbit dies. This tensor is the oscillator's analogue of Kepler's Laplace–Runge–Lenz vector — the conserved object beyond $E$ and $\vec L$ that the closure advertised.

Step 2. Read $A$ geometrically. It is symmetric ( $6$ components) with trace $A_{ii} = H$ . Evaluate it on the principal-axis orbit $x' = a\cos\omega t$ , $y' = b\sin\omega t$ : the diagonal entries are the axis energies $\left(\tfrac12 m\omega^2a^2,\ \tfrac12 m\omega^2b^2,\ 0\right)$ and the off-diagonal entry vanishes: $A_{x'y'} = \dfrac{p_{x'}p_{y'}}{2m} + \dfrac{m\omega^2}{2}x'y' = \tfrac{m\omega^2}{2}ab(-\sin\omega t\cos\omega t + \cos\omega t\sin\omega t) = 0$ . So the orbit's principal axes are $A$ 's eigenvectors, and its eigenvalues are the semi-axis energies: the conserved tensor stores the ellipse's orientation and shape — precisely the data that "energy + angular momentum" leave free, and precisely the data that must be frozen for the orbit not to precess. Why this is the closure explanation: a precessing orbit is one whose axes rotate; here the axes are eigenvectors of a conserved matrix and cannot.

Step 3. Count the independent constants. $A$ (6) plus $\vec L$ (3) plus nothing new from $H$ (the trace) makes 9 candidates, but they obey relations — notably $A_{ij}L_j = 0$ ( $\vec L$ is $A$ 's null eigenvector: check on the orbit above, where the third axis carries no motion) and $\det$ -type identities linking $|\vec L|$ to $A$ 's invariants (from (3): $\tfrac{2}{m}\left[A_{11}A_{22} - A_{12}^2 + \text{cyclic}\right] = \omega^2 L^2/2$ -type relations). The honest tally, standard for this system: 5 functionally independent constants — the maximum possible for bounded motion in 3 degrees of freedom ( $2n - 1 = 5$ ), since the orbit (a closed curve in 6-D phase space) is the intersection of 5 independent level sets. Systems achieving the maximum are maximally superintegrable; the complete classical list of central examples is exactly Bertrand's list. Why $2n - 1$ is the ceiling: more independent constants would cut the trajectory down to a point.

Step 4. Organize the algebra. Split $A$ into its trace ( $H$ ) and its 5 traceless components; together with the 3 components of $\vec L$ that makes $5 + 3 = 8$ conserved quantities beyond... precisely the dimension of SU(3)'s generator algebra. Computing the Poisson brackets (a finite, mechanical exercise: $\{L_i, L_j\} = \epsilon_{ijk}L_k$ ; $\{L_i, A_{jk}\}$ rotates the tensor's indices as a rank-2 object; $\{A_{ij}, A_{kl}\}$ closes back onto $\omega^2\times$ angular-momentum terms) shows the set closes: the traceless $A$ and $\vec L$ generate su(3) under Poisson brackets, with $H$ the commuting (Casimir-building) trace. We state the $\{A, A\}$ closure rather than grinding all components here; the two-line conservation proof and the eigenvector geometry above are the load-bearing computations. In the quantum oscillator the same algebra acts on each energy level, whose degeneracy $\binom{n+2}{2} = \frac{(n+1)(n+2)}{2}$ — 1, 3, 6, 10, … — is exactly the dimension ladder of SU(3) symmetric-tensor representations; Elliott's 1958 nuclear model runs on this machinery. Why Kepler gets SO(4) and the oscillator SU(3): Kepler's extra object is a vector (3 components joining 3 of $\vec L$ : the 6 generators of SO(4)); the oscillator's is a quadrupole tensor (5 joining 3: the 8 of SU(3)). Bigger hidden symmetry, richer degeneracy — the oscillator is, by this measure, the more symmetric of the two most symmetric systems. *Three consistency checks are worth noting explicitly: the null-eigenvector relation $A\vec L = 0$ holds on the principal-axis orbit by inspection, since all $z$ -components of $p$ and $x$ vanish there; and the degeneracy ladder $(n+1)(n+2)/2$ yields 1, 3, 6, 10 for $n = 0,1,2,3$ , matching the count of symmetric index combinations. The bracket closure and its history are treated in Jauch & Hill, Phys. Rev. 57, 641 (1940); Fradkin, Am. J. Phys. 33, 207 (1965).*

Which equation answers which question

you want	use	notes
trajectory from initial conditions	Cartesian (2): three independent SHOs	same $\omega$ ; then rotate to principal axes
orbit geometry	centred ellipse; $E = \tfrac12 m\omega^2(a^2{+}b^2)$ , $L = m\omega ab$ , eq. (3)	$r_+ = a$ , $r_- = b$
period	$T = 2\pi/\omega$ , always	amplitude- and shape-independent
turning points from $E, L$	quadratic in $r^2$ , eq. (4)	real iff $E \ge \omega L$ ; equality = circle
closure/precession question	apsidal angle $= \pi/2$ , eq. (5)	$\omega_r = 2\omega_\varphi$ for every orbit
orbit orientation & shape constants	tensor $A_{ij}$ , Derivation C	eigenvectors = axes; eigenvalues = axis energies
degeneracy bookkeeping	$E = \omega(J_x + J_y + J_z)$ ; quantum $(n{+}1)(n{+}2)/2$	SU(3) representation ladder

Limiting cases

Case	Result	Meaning
$L = 0$	line oscillation through the centre	the degenerate ellipse ( $b = 0$ ); still period $2\pi/\omega$
$E = \omega L$	circular orbit at $r_c = \sqrt{L/m\omega}$	turning points merge; $U_{\rm eff}$ minimum
$a = b$ vs $a \gg b$	circle vs needle — same period	spatial isochrony: shape costs no time
anisotropic $\omega_x \ne \omega_y$	Lissajous; closed only for rational ratios	hidden symmetry and $A_{xy}$ conservation destroyed
small $r$ in any smooth symmetric well	this system, to leading order	why traps and nuclei keep meeting it
Kepler comparison	apsidal $\pi$ , focal ellipse, SO(4), LRL vector	the other Bertrand champion
quantum echo	$E_n = \hbar\omega(n + \tfrac32)$ , degeneracy $\tfrac{(n+1)(n+2)}{2}$	SU(3) symmetric-tensor dimensions
2-D isotropic version	apsidal $\pi/2$ likewise; algebra su(2)	the "hidden angular momentum" of the 2-D trap

Common trapIt's a central-force ellipse, so the centre of force is at a focus

Kepler has trained this reflex into every physics student, and the isotropic oscillator is the standing counterexample: here the force centre sits at the ellipse's geometric centre, not at a focus — the orbit is symmetric under $\vec r \to -\vec r$ , visits its near point twice per lap, and sweeps only $90°$ between apsides where a Kepler orbit sweeps $180°$ . Confusing the two anatomies breaks real calculations: the vis-viva-style shortcuts, the "perihelion once per period" timing intuitions, and any formula with the eccentric anomaly all silently assume the focal geometry. The clean discriminators to carry: (i) apsidal angle — $\pi/2$ (oscillator) vs $\pi$ (Kepler); (ii) where the speed peaks — at the ends of the minor axis ( $r = r_-$ ) for the oscillator, at perihelion for Kepler (both from $L =$ const, but the near points sit in different places); (iii) the conserved orientation object — a quadrupole tensor whose eigenvectors are the axes here, versus the Laplace–Runge–Lenz vector pointing at perihelion there. And a second reflex to disarm: "harmonic oscillator" does not mean "small oscillations about the orbit" — every result in this article ((2)–(5), the tensor, SU(3)) is exact at all amplitudes; the oscillator is not an approximation to itself. The approximation enters only when some other well is Taylor-truncated to quadratic — at which point anharmonic corrections may reopen everything, including precession.

Worked examples

1 · A complete orbit, both routes (confidence builder)

A particle of mass $m = 0.2$ kg moves in an isotropic trap with $\omega = 3.0$ rad/s. At $t = 0$ : $\vec r_0 = (0.50, 0, 0)$ m, $\vec v_0 = (0, 0.90, 0)$ m/s. Find the orbit, $E$ , $L$ , the turning radii, and all the periods.

Step 1. Cartesian solution: $x(t) = 0.50\cos 3t$ (starts at max, no $x$ -velocity); $y(t) = (0.90/3)\sin 3t = 0.30\sin 3t$ . Already principal-axis form: a centred ellipse, $a = \boxed{0.50}$ m, $b = \boxed{0.30}$ m, in the $xy$ -plane.

Step 2. Conserved quantities from (3): $E = \tfrac12(0.2)(9)(0.25 + 0.09) = \boxed{0.306\ \text{J}}$ ; $L = (0.2)(3)(0.5)(0.3) = \boxed{0.090\ \text{kg}\,\cdot\,\text{m}^2/\text{s}}$ . Direct check: $E = \tfrac12 mv_0^2 + \tfrac12 m\omega^2 r_0^2 = 0.081 + 0.225 = 0.306$ ✓; $L = m\,x_0v_{y0} = 0.2 \times 0.5 \times 0.9 = 0.090$ ✓.

Step 3. Spherical route, eq. (4): $r_\pm^2 = \dfrac{0.306 \pm \sqrt{0.306^2 - 9(0.09)^2/\,}\ }{0.2 \times 9}$ — compute the discriminant: $E^2 - \omega^2L^2 = 0.0936 - 9(0.0081) = 0.0207$ , $\sqrt{\ } = 0.144$ ; $r_+^2 = (0.306 + 0.144)/1.8 = 0.25$ , $r_-^2 = (0.306 - 0.144)/1.8 = 0.09$ : $r_+ = 0.50$ , $r_- = 0.30$ m ✓ — the semi-axes, as promised by $r_+ = a$ , $r_- = b$ (see the computed $U_{\rm eff}$ figure, drawn for exactly these numbers).

Step 4. Clocks: orbital period $T = 2\pi/3 = \boxed{2.09\ \text{s}}$ ; radial (breathing) period $T/2 = 1.05$ s — the radius touches $0.50$ m twice and $0.30$ m twice per lap, $90°$ apart. Circular-orbit benchmark at this $L$ : $r_c = \sqrt{L/m\omega} = \sqrt{0.15} = 0.39$ m with $E_c = \omega L = 0.27$ J $< E$ ✓ (our orbit has $0.036$ J of radial breathing above the circular floor).

2 · The degeneracy audit — same energy, every shape (exam level)

(a) At the energy of example 1, what orbits are available? (b) Verify the frequency lock $\omega_r = 2\omega_\varphi$ from the action-angle formula $E = \omega(2J_r + L)$ . (c) Count the quantum states at the corresponding level number and connect to SU(3).

Step 1. The energy shell: $E = \tfrac12 m\omega^2(a^2 + b^2)$ fixes only $a^2 + b^2 = 2E/m\omega^2 = 0.34$ m². Available: everything from the line orbit ( $a = 0.583, b = 0$ , $L = 0$ ) through our ellipse ( $0.5, 0.3$ ) to the circle ( $a = b = 0.412$ m, $L_{\max} = m\omega a^2 = 0.102$ ). One energy, a one-parameter family of shapes plus arbitrary orientation: that orientation-and-shape freedom at fixed $E$ is the degeneracy, and the conserved tensor $A$ is what labels the members.

Step 2. Actions: for the isotropic oscillator the radial action works out to $J_r = \tfrac{1}{2}\left(E/\omega - L\right)$ , i.e. $E = \omega(2J_r + L)$ . Frequencies: $\omega_r = \partial E/\partial J_r = 2\omega$ and $\omega_\varphi = \partial E/\partial L = \omega$ — the 2:1 lock, now read off by differentiation rather than integration, valid for every orbit because $E$ depends only on the single combination $2J_r + L$ . Check against example 1: radial period $2\pi/\omega_r = 1.05$ s ✓, angular period $2.09$ s ✓. (Kepler's analogue: $E$ depends only on $J_r + L$ — hence $\omega_r = \omega_\varphi$ and one perihelion per lap. The two Bertrand systems are the two ways a central-force $E$ can collapse onto a single action combination.)

Step 3. Quantum shadow: levels $E_n = \hbar\omega(n + \tfrac32)$ with degeneracy $(n+1)(n+2)/2$ : $n = 0, 1, 2, 3 \to 1, 3, 6, 10$ states. These are the dimensions of SU(3)'s symmetric representations, and the $n = 2$ sextet — $(200), (020), (002), (110), (101), (011)$ in occupation labels — is the classical shape-and-orientation family of Step 1, quantized. Elliott's SU(3) model builds rotational bands of light nuclei ( $^{20}$ Ne, $^{24}$ Mg) from exactly these degenerate oscillator shells — the hidden symmetry of this article, promoted to working nuclear spectroscopy.

Step 4. Extract the structural moral: degeneracy = "energy blind to shape" = extra conserved quantities = enlarged symmetry group. The chain runs identically for Kepler ( $E$ blind to eccentricity → LRL vector → SO(4) → the $n^2$ hydrogen degeneracy) — and this parallel, visible entirely within classical mechanics, is the standard modern reading of Bertrand's theorem: closure happens where symmetry is larger than geometry requires.

Origins · 2 minRobert Hooke and how this was discovered

Robert Hooke

1635–1703 · London

Hooke published the law that owns this article's force — ut tensio, sic vis, "as the extension, so the force" — in 1678, characteristically first as an anagram (ceiiinosssttuv, 1676) to stake priority while he kept experimenting. As Curator of Experiments at the Royal Society and Gresham Professor, he was the most prolific instrument-physicist of his century: the balance-spring watch (bitterly disputed with Huygens), the compound microscope of Micrographia, the universal joint, and the first serious programme — springs, arches, "springy bodies" of every kind — of elasticity as quantitative science. It was also Hooke, in correspondence with Newton in 1679–80, who proposed analysing orbital motion as a tangential inertia bent by a centripetal attraction, and who pressed the inverse-square hypothesis — the exchange that, by Newton's own later admission, set him on the road to the Principia (and to a priority grievance Hooke nursed to his grave; Newton, notoriously, had Hooke's portrait vanish from the Royal Society). The Principia repaid the spring-master in kind: Book I solves motion under a force proportional to distance — Hooke's law made celestial — and proves its orbits are ellipses centred on the force, the very geometry of this article, standing forever next to the inverse-square ellipse as the only other perfectly closed orbit in the central-force universe.

History

1678

— Hooke's De Potentia Restitutiva publishes "as the extension, so the force": the linear restoring law, with springs as the laboratory.

1687

— Newton's Principia, Book I, Prop. X: motion under a force proportional to distance traces ellipses centred on the force centre — the isotropic oscillator solved at the birth of mechanics.

1873

— Bertrand proves his theorem: among all central power laws, only

r^2

and

1/r

close every bounded orbit — the oscillator and Kepler certified as the two champions.

1940

— Jauch and Hill work out the dynamical symmetry of the isotropic oscillator: the conserved quadrupole tensor and the SU(n) algebra behind the degeneracy.

1958

— Elliott's SU(3) model turns oscillator-shell degeneracy into nuclear spectroscopy, generating the rotational bands of light nuclei from the hidden symmetry.

Going further

Texts. Goldstein, Poole & Safko, Classical Mechanics, §3.3, §3.6 (central-force treatment, Bertrand), §9.7–9.8 (symmetry groups of mechanical systems) · Landau & Lifshitz, Mechanics, §14 & §23 (central motion; oscillations) · Kleppner & Kolenkow, Ch. 10 · Jauch & Hill, Phys. Rev. 57, 641 (1940) — the dynamical-symmetry paper · Fradkin, "Three-dimensional isotropic harmonic oscillator and SU₃," Am. J. Phys. 33, 207 (1965) — the pedagogical classic · Elliott, Proc. R. Soc. A 245, 128 & 562 (1958) — SU(3) into nuclear physics. The key results to keep in view are: the orbit-curve elimination in Derivation A yields the $\sin^2\delta$ quadratic form; the relations $E = \tfrac12 m\omega^2(a^2+b^2)$ , $L = m\omega ab$ mirror the turning-point identities $r_+^2 + r_-^2 = 2E/m\omega^2$ , $r_+r_- = L/m\omega$ ; the apsidal integral via $s = r^2$ gives exactly $\pi/2$ , with the root-product check $\sqrt{-\alpha s_+s_-} = \sqrt{-\gamma}$ ; the conservation $\dot A_{ij} = 0$ follows in two lines; worked example 1 carries the numbers throughout ( $E = 0.306$ J both routes, $L = 0.090$ , discriminant $\sqrt{0.0207} = 0.144$ , $r_\pm = 0.50/0.30$ m, $r_c = 0.39$ m, $E_c = 0.27$ J — the computed $U_{\rm eff}$ figure uses these); the action formula $E = \omega(2J_r + L)$ encodes the 2:1 lock; and the degeneracy ladder runs 1, 3, 6, 10.

Builds on: Simple Harmonic Motion · Central Force Fundamentals · The Effective Potential · Reduction to the Radial Equation. Leads to: Bertrand's Theorem · Degeneracy and Commensurate Frequencies · (quantum tracks) the 3-D quantum oscillator and shell structure.

Why does the oscillator get a conserved TENSOR while Kepler gets a conserved VECTOR?

Because the two orbits store different orientation data, and the conserved object must carry exactly what the orbit refuses to forget. A Kepler ellipse has a genuinely directed axis: perihelion sits on one side of the force centre, aphelion on the other, so the memory is an arrow — the Laplace–Runge–Lenz vector, pointing at perihelion with magnitude set by the eccentricity. The oscillator's centred ellipse is symmetric under $\vec r \to -\vec r$ : its major axis is a line, not an arrow — "axis at $30°$ " and "axis at $210°$ " are the same statement. An object that encodes an unoriented axis (plus the energy split between axes) is precisely a symmetric rank-2 tensor, whose eigenvectors are defined only up to sign. The group theory follows the geometry: 3 components of $\vec L$ + 3 of the LRL vector close into SO(4) for Kepler; 3 of $\vec L$ + 5 of traceless $A$ close into SU(3) here. The same logic runs in the quantum versions — hydrogen's $n^2$ degeneracy is an SO(4) multiplet, the oscillator's $(n+1)(n+2)/2$ an SU(3) one — and it generalizes: whenever a degeneracy looks "accidental," the working hypothesis is that some tensorial object of the right symmetry type is being silently conserved.

Is the motion genuinely three-dimensional, or is this secretly a 2-D problem?

Each orbit is planar — the force is central, $\vec L$ is conserved, and the trajectory stays in the plane perpendicular to it (or on a line, if $\vec L = 0$ ) — but the problem is irreducibly three-dimensional in two respects worth separating. Dynamically, the plane itself is free data: unlike a 2-D oscillator, which lives in a prescribed plane, the 3-D system chooses its orbital plane from initial conditions, and the conserved quantities must be able to describe every choice — that is why $\vec L$ has three components here and why $A_{ij}$ is a full $3\times3$ tensor with a null eigenvector along $\vec L$ (the tensor knows the plane, then describes the ellipse within it). Structurally, the dimension count changes the symmetry: the 2-D isotropic oscillator's hidden algebra is su(2) (three extra constants, famously mappable to a fictitious spin), while 3-D gives su(3) — and the quantum degeneracies differ accordingly ( $n + 1$ in 2-D vs $(n+1)(n+2)/2$ in 3-D). Practically, the reduction to the plane is the right first move in any explicit calculation (Derivations A and B both make it), but statements about the symmetry group, the degeneracy, and the count of conserved quantities are statements about the full 3-D phase space and cannot be made plane-by-plane.

The period is amplitude-independent — is that the same isochrony as the 1-D oscillator, and does anything in 3-D break it?

It is the same isochrony, inherited three times over: each Cartesian component is a 1-D oscillator with amplitude-independent period, and since all three share $\omega$ , the vector motion repeats after $2\pi/\omega$ regardless of the amplitudes or phases — hence "spatial isochrony": circles, needles, and everything between keep identical time. What breaks it is exactly what breaks the 1-D version, plus one new channel. The old channel: anharmonicity — real trap potentials are quadratic only near the bottom, and larger orbits sample quartic corrections, picking up amplitude-dependent frequency shifts (the canonical-perturbation-theory article computes precisely such shifts, e.g. the pendulum's $-\varphi_m^2/16$ ). The new channel: anisotropy — if $\omega_x \ne \omega_y$ , the period of the vector motion is the (possibly nonexistent) common period of the components, orbits become Lissajous figures that close only for rational frequency ratios, and isochrony survives per-axis but dies for the orbit as a whole. Experimentalists exploit the sharpness of both statements: the breathing mode of a cloud in a trap rings at $2\omega$ with amplitude-independent frequency exactly insofar as the trap is harmonic and isotropic, so frequency shifts of that mode are used as diagnostics of anharmonicity, anisotropy, and interaction effects.

Where exactly does the 2:1 frequency ratio come from? It seems too clean to be coincidence.

It isn't coincidence — it is the centred-ellipse geometry, expressible three equivalent ways, and seeing all three is a good workout. Geometrically: a centred ellipse has two "far" vertices and two "near" vertices per circuit, so the radius must complete two full breathing cycles per revolution — $\omega_r = 2\omega_\varphi$ by inspection of the orbit. Analytically: $r^2 = a^2\cos^2\omega t + b^2\sin^2\omega t = \frac{a^2+b^2}{2} + \frac{a^2-b^2}{2}\cos 2\omega t$ — the radius literally oscillates at $2\omega$ while the angle advances at average rate $\omega$ . Structurally (the deep version): in action-angle variables $E = \omega(2J_r + L)$ , so $\omega_r = \partial E/\partial J_r = 2\omega$ and $\omega_\varphi = \partial E/\partial L = \omega$ — the ratio is locked for all orbits because the energy depends on the single combination $2J_r + L$ ; that collapse of a two-action function onto one combination is the precise meaning of "degenerate" in the commensurate-frequencies article, and it is the property canonical perturbation theory flags as fragile (a generic perturbation makes $E$ depend on both actions, unlocking the ratio and setting the ellipse precessing — which is exactly how trap imperfections manifest in the lab). Kepler runs the same story with $E$ depending on $J_r + L$ (ratio 1:1, one perihelion per lap). Bertrand's theorem, in this language: only the 2:1 and 1:1 locks can be maintained at all amplitudes by a central power law.

How honest is the claim that A_ij and L generate SU(3)? What's actually being asserted?

The load-bearing computations are elementary and shown in Derivation C: $\dot A_{ij} = 0$ along the motion (two lines), the eigenvector-geometry of $A$ (evaluated explicitly on the orbit), and the counting to five independent constants. The algebraic claim on top is: define the eight functions {traceless $A_{ij}$ (5), $L_i$ (3)}; compute all pairwise Poisson brackets; the results are linear combinations of the same eight functions — the set closes, and the structure constants are those of su(3). Each individual bracket is a finite mechanical computation ( $\{L_i, L_j\} = \epsilon_{ijk}L_k$ you know; $\{L_i, A_{jk}\}$ just says $A$ rotates as a tensor; the genuinely informative ones are $\{A_{ij}, A_{kl}\} \propto \omega^2(\delta L)$ -type terms, where the frequency-matching of the isotropic case is used) — this article states the closure and cites Jauch–Hill/Fradkin rather than grinding all 28 brackets. Two honesty notes. First, the classical statement is about a Poisson algebra isomorphic to su(3); the cleaner and more famous incarnation is quantum, where $a_i^\dagger a_j$ (nine bilinears: eight traceless + number operator) manifestly generate U(3) commuting with $H = \hbar\omega(\hat N + \frac32)$ — a three-line verification, and the honest fastest route if you want proof-level assurance. Second, "SU(3)" here is the degeneracy group of the oscillator, mathematically the same group but physically distinct from flavor SU(3) of particle physics — same algebra, different arena; Elliott's nuclear model is the direct physical descendant of this one.

Real traps are never perfectly isotropic. How much of this survives small anisotropy?

There is a sharp part and a robust part, and knowing which is which is the practical skill. Sharp (destroyed by any anisotropy): exact orbit closure, the cross-correlation conservation laws ( $A_{xy}$ etc.), and the SU(3) structure — with $\omega_x \ne \omega_y$ , the off-diagonal tensor components oscillate at the difference frequency $\omega_x - \omega_y$ , the ellipse's axes precess at that same slow rate, and for irrational frequency ratios the orbit densely fills a box (Lissajous). The degradation is graceful, though: for fractional anisotropy $\epsilon = \Delta\omega/\omega$ , the orbit looks closed for $\sim 1/\epsilon$ periods before the precession is noticeable — a trap tuned to $10^{-3}$ anisotropy keeps "Bertrand physics" for a thousand laps. Robust (survives): per-axis isochrony, energy and (for axisymmetric $\omega_x = \omega_y \ne \omega_z$ ) the $L_z$ conservation, the separability itself, and the quantum shell structure in modified form — anisotropy splits the $(n+1)(n+2)/2$ degeneracy in calculable patterns, which is precisely how experiments measure trap anisotropy spectroscopically. The theory-side moral mirrors the KAM/perturbation story: degeneracy-borne structures (closure, hidden symmetry) are the fragile aristocracy of mechanics — infinitely refined, destroyed by generic perturbations — while the underlying integrable skeleton (separability, per-mode invariants) is sturdy. Treating small anisotropy with the canonical-perturbation machinery on the degenerate $H_0$ is a classic exercise — and a genuinely instructive one, because the harmonic oscillator is exactly the case where naive nondegenerate averaging fails and the resonant (secular) treatment is mandatory.