The Harmonic Oscillator in Phase Space

In September 2021, physicists at Caltech and elsewhere celebrated a strange milestone: drums a few micrometres across had been cooled until they performed the quietest oscillation permitted by nature — a single quantum of vibration, then none at all. The theoretical scaffolding for that experiment is not exotic. It is the humble mass-on-a-spring, drawn not as a wiggle in time but as a curve in phase space, where its geometry becomes so clean that quantum mechanics can be read off it with a ruler. Viewed in the $(q,p)$ plane, the harmonic oscillator stops being a formula and becomes a shape. Energy conservation pins each motion to an ellipse; Hamilton's equations sweep the state clockwise around it; and — the oscillator's special miracle — every ellipse is traversed in exactly the same time $2\pi/\omega$ , whatever its size. Rescale the axes and the entire flow becomes a rigid rotation of the plane: the one system in mechanics whose phase portrait you can spin in your head. That rigidity makes the oscillator the perfect laboratory for the deep theorems of this chapter. Liouville's volume conservation, abstract in general, is here literally a rotating patch that never changes shape. The enclosed area $2\pi E/\omega$ is the action variable that survives into quantum mechanics as Planck's cells of size $h$ — which is why the oscillator's quantum levels are equally spaced. Everything from LIGO's squeezed light to the ground-state drums is bookkeeping on the ellipses drawn in this article.

Key takeaways

The oscillator's phase trajectories are the level curves of $H = p^2/2m + \tfrac12 m\omega^2 q^2$ : nested ellipses with semi-axes $q_{\max} = \sqrt{2E/m\omega^2}$ and $p_{\max} = \sqrt{2mE}$ , traversed clockwise.
In rescaled coordinates $Q = \sqrt{m\omega}\,q$ , $P = p/\sqrt{m\omega}$ the flow is a rigid rotation at angular rate $\omega$ — every state circles at the same rate, so the period $T = 2\pi/\omega$ is independent of amplitude (isochronism).
The enclosed area is the action, $J = \oint p\,\mathrm{d}q = 2\pi E/\omega = ET$ — linear in energy, with units of action, and an adiabatic invariant.
A rigid rotation preserves areas exactly: the oscillator is Liouville's theorem made visible, a patch of phase fluid rotating without distortion.
Bohr–Sommerfeld quantization selects the ellipses enclosing area $(n + \tfrac12)h$ , which spaces the quantum levels evenly: $E_n = (n + \tfrac12)\hbar\omega$ .
Key idea: the harmonic oscillator's phase flow is a clockwise rotation on nested energy ellipses of area $2\pi E/\omega$ — amplitude-independent in period, exactly area-preserving, and quantized by slicing that area into cells of $h$ .

Where this lives today

The oscillator ellipse is now an object laboratories manipulate directly. In LIGO, the light circulating in the interferometer is a harmonic oscillator mode whose quantum uncertainty occupies a minimum patch of phase space; since 2019 the detectors inject squeezed vacuum — states whose uncertainty patch is deliberately stretched along one axis and thinned along the other, area intact, exactly the unit-Jacobian games of equation (4) — to push measurement noise below the naive quantum limit. The gravitational-wave events announced each year are read through a reshaped oscillator ellipse. Down at the nanoscale, the ladder of the third figure is an engineering resource. Superconducting qubits are read out and coupled through microwave oscillators whose few lowest ellipses are individually addressable; mechanical drums and phonon modes have been cooled to $\bar n < 1$ (Worked Example 2 is the design arithmetic), entangled with each other, and used as quantum memories. Meanwhile the classical content of this article never left the workhorse toolkit: every accelerator bunch is described by its emittance — the area of its ellipse — and every molecular vibration in a spectroscopy table is a $\hbar\omega$ read off the same ladder. One quadratic potential, one geometry, from Panthéon-scale pendulums down to a half-cell of $h$ .

The physical picture

Take the motion you already know from the simple-harmonic-motion article — $q(t) = A\cos(\omega t + \varphi)$ , an endless sinusoid — and fold time away. Plot the state $(q, p)$ instead: at a turning point the momentum vanishes, so the dot sits at $(A, 0)$ ; a quarter period later the mass barrels through equilibrium at top speed, the dot at $(0, -m\omega A)$ . Join the instants and the entire infinite wiggle closes into one finite curve, an ellipse the state orbits clockwise forever:

Figure

A sinusoid in time is a single clockwise ellipse in phase space — the whole history in one closed curve

Why an ellipse and not some other loop? Because the energy $\tfrac{p^2}{2m} + \tfrac12 m\omega^2 q^2 = E$ is a sum of two squares — the defining equation of an ellipse — and conservation forbids the dot from leaving its level curve. The two semi-axes have different units (metres and kilogram-metres-per-second), which is the general phase-space situation: the aspect ratio of the ellipse is an artifact of units, set by $m\omega$ . So remove the artifact. Stretch one axis and shrink the other by $\sqrt{m\omega}$ and the ellipses become circles — and, remarkably, the dynamics becomes a rigid rotation of the whole plane at rate $\omega$ . Rigid is a strong word and it earns its keep three times over. Every circle is completed in the same time $2\pi/\omega$ : isochronism, the property that made pendulum clocks conceivable and makes oscillators the universal timekeepers of physics. Neighbouring states never separate: no chaos, no shear, the tamest flow in mechanics. And any patch of states rotates without changing shape — Liouville's area conservation, which in general requires a theorem, here happens by inspection:

Figure

Rescale q and p and the flow is a rigid rotation at rate omega — the gentlest possible Liouville theorem

The last picture is the payoff. The area inside an energy- $E$ ellipse is $2\pi E/\omega$ — linear in energy. Old quantum theory's one commandment (Planck, Bohr, Sommerfeld) was that nature only permits orbits enclosing whole cells of area $h$ , plus the half-cell nature throws in for free. Slice a linear function of $E$ into equal slabs of area and you get equally spaced energies: the oscillator's famous ladder $E_n = (n + \tfrac12)\hbar\omega$ , read directly off classical geometry:

Figure

Equally spaced areas make equally spaced energies — the oscillator's quantum ladder read off classical ellipses

The core equations

The Hamiltonian is the energy written in phase-space variables,

H(q,p) = \frac{p^2}{2m} + \frac{1}{2} m\omega^2 q^2 \qquad (1)

where: $m$ — mass; $\omega = \sqrt{k/m}$ — angular frequency; $p = m\dot q$ — here the canonical momentum coincides with the kinetic one, since the Lagrangian is the standard $T - V$ . Both terms are non-negative, so the level sets are bounded for every $E > 0$ — the first hint that all motion is trapped and periodic. Hamilton's equations turn (1) into a flow:

\dot q = \frac{\partial H}{\partial p} = \frac{p}{m}, \qquad \dot p = -\frac{\partial H}{\partial q} = -m\omega^2 q \qquad (2)

where: the first equation is the definition of velocity, the second is Hooke's law. Check the circulation sense once and keep it forever: at $(q_{\max}, 0)$ equation (2) gives $\dot p < 0$ — the dot dives — so the motion is clockwise, as for every Hamiltonian orbit around a potential minimum. Setting $H = E$ and dividing gives the trajectory in standard form,

\frac{q^2}{2E/m\omega^2} + \frac{p^2}{2mE} = 1 \qquad (3)

where: the semi-axes are $q_{\max} = \sqrt{2E/m\omega^2}$ (the amplitude $A$ ) and $p_{\max} = \sqrt{2mE} = m\omega A$ (the peak momentum). Doubling the energy stretches both axes by $\sqrt2$ : the ellipses are concentric, similar, and never intersect — the foliation drawn in every figure above. The unit-fixing rescaling deserves display status because it exposes the flow's true character:

Q = \sqrt{m\omega}\,q, \qquad P = \frac{p}{\sqrt{m\omega}} \;\;\Rightarrow\;\; H = \frac{\omega}{2}\left(P^2 + Q^2\right) \qquad (4)

where: $(Q,P)$ — rescaled canonical coordinates (the transformation preserves areas, since the two factors cancel). In these variables the flow equations become $\dot Q = \omega P$ , $\dot P = -\omega Q$ : a rotation, solved in the first derivation below. The explicit solution in original variables is the one you know,

q(t) = A\cos(\omega t + \varphi), \qquad p(t) = -m\omega A \sin(\omega t + \varphi) \qquad (5)

where: $A, \varphi$ — amplitude and phase fixed by the initial point. Notice what is absent: the frequency contains no trace of $A$ . Every orbit, tiny or huge, closes in

T = \frac{2\pi}{\omega} \quad (\text{independent of amplitude}) \qquad (6)

where: this isochronism is the exclusive property of the quadratic potential — the pendulum article shows how a $\cos\theta$ potential destroys it, slowing large swings until the period diverges at the separatrix. Finally the geometric invariant. The area enclosed by the energy- $E$ orbit is

J = \oint p\,\mathrm{d}q = \pi\, q_{\max} p_{\max} = \frac{2\pi E}{\omega} = ET \qquad (7)

where: $J$ — the action variable (units: energy × time). This single number is the oscillator's currency across physics: it is the adiabatic invariant that stays fixed under slow parameter changes, the emittance that accelerator physicists conserve, and — sliced into cells of $h$ — the origin of the quantum ladder, derived below.

Which equation answers which question

you want	use	notes
the trajectory through a state	eq. (3) with $E = H(q_0, p_0)$	ellipse, semi-axes $A$ and $m\omega A$
position and momentum at time $t$	eq. (5)	clockwise ellipse, parametrized
the period	eq. (6)	same for every amplitude
enclosed area / action	eq. (7)	$2\pi E/\omega$ , linear in $E$
clean geometry	rescale via eq. (4)	circles, rigid rotation at $\omega$
quantum levels	$J = (n + \tfrac12) h$ in eq. (7)	$E_n = (n + \tfrac12)\hbar\omega$

Derivation ASolving the flow: the phase plane rotates rigidly

We are solving Hamilton's equations (2) exactly and extracting the oscillator's three signature properties in one stroke: clockwise elliptical orbits, amplitude-independent period, and exact area preservation. The strategy is the rescaling (4), which converts the lopsided system into a manifestly antisymmetric one — the generator of rotations — after which the solution is a rotation matrix whose properties can be read off rather than computed. Watch the direction conventions: with $Q$ rightward and $P$ upward, the physical flow rotates clockwise, and a sign error here silently flips every portrait you ever draw.

Step 1. Rescale. With $Q = \sqrt{m\omega}\,q$ and $P = p/\sqrt{m\omega}$ , equations (2) become

\dot Q = \omega P, \qquad \dot P = -\omega Q

Why rescale first: the raw system mixes $1/m$ and $m\omega^2$ , hiding the symmetry; in $(Q,P)$ the coefficient matrix is $\omega$ times the antisymmetric matrix that generates rotations, and antisymmetry is the whole story.

Step 2. Solve by recognizing the rotation. Differentiating the first equation and substituting the second gives $\ddot Q = -\omega^2 Q$ , but more useful is the matrix solution:

\begin{pmatrix} Q(t) \\ P(t) \end{pmatrix} = \begin{pmatrix} \cos\omega t & \sin\omega t \\ -\sin\omega t & \cos\omega t \end{pmatrix} \begin{pmatrix} Q_0 \\ P_0 \end{pmatrix}

Why this is a clockwise rotation: at small $t > 0$ a state on the positive $Q$ -axis acquires $P < 0$ — it dives, matching the physical check after equation (2).

Step 3. Read off isochronism. The rotation matrix depends on time only through $\omega t$ and is periodic with period $2\pi/\omega$ — for every initial condition simultaneously. All orbits, whatever their radius $\sqrt{Q_0^2 + P_0^2} = \sqrt{2E/\omega}$ , close at the same instant: equation (6) without a single integral.

Step 4. Read off area preservation. The determinant of the rotation matrix is $\cos^2\omega t + \sin^2\omega t = 1$ , so the map from initial to final states scales all areas by exactly 1; and since the rescaling (4) also has unit Jacobian, areas are preserved in the original $(q,p)$ plane too. Any patch of phase fluid rotates rigidly, never stretching — Liouville's theorem, here with the stronger bonus that even the shape survives (general Hamiltonian flows preserve only the volume).

Step 5. Undo the rescaling to recover equation (5): $q(t) = A\cos(\omega t + \varphi)$ , $p(t) = -m\omega A\sin(\omega t + \varphi)$ , with $A$ and $\varphi$ the polar coordinates of the initial point. Substituting (5) back into (2) checks term by term; the unit determinant matches the divergence-free property $\partial\dot q/\partial q + \partial\dot p/\partial p = 0 - 0 = 0$ proved generally in the Liouville article.

Go deeper · Derivation B · 1 minThe action integral and the quantum ladder

Derivation BThe action integral and the quantum ladder

The goal is the two-line formula with the longest reach in this article: the enclosed area $J = 2\pi E/\omega$ , and from it the quantum spectrum $E_n = (n + \tfrac12)\hbar\omega$ . The strategy is elementary geometry — an ellipse's area from its semi-axes — followed by the single physical postulate of old quantum theory: allowed orbits enclose half-integer multiples of Planck's cell $h$ . The thing to watch is what the linearity of $J$ in $E$ does: it is the entire reason the oscillator's levels are evenly spaced, and the reason any potential steeper or shallower than quadratic has a crowding or spreading spectrum.

Step 1. Compute the area from the semi-axes of (3):

J = \oint p\,\mathrm{d}q = \pi\,q_{\max}\,p_{\max} = \pi \sqrt{\frac{2E}{m\omega^2}}\,\sqrt{2mE} = \frac{2\pi E}{\omega}

Why $\pi ab$ suffices: $\oint p\,\mathrm{d}q$ taken clockwise is precisely the enclosed area of the loop, and the mass factors cancel between the two semi-axes — the area, unlike the axes, is unit-honest, measured in energy × time.

Step 2. Note the equivalent forms $J = ET = 2\pi E/\omega$ : an orbit's area equals its energy times its period. This is the quantity that stays constant when the spring stiffens slowly (adiabatic invariance): $E$ and $\omega$ both drift, their ratio does not.

Step 3. Impose Bohr–Sommerfeld quantization, $J = \left(n + \tfrac12\right) h$ with $n = 0, 1, 2, \dots$ :

\frac{2\pi E_n}{\omega} = \left(n + \tfrac12\right) h \;\;\Rightarrow\;\; E_n = \left(n + \tfrac12\right)\hbar\omega

Why the half: the naive rule $J = nh$ was Planck's; the extra $\tfrac12$ emerges from the WKB connection formulas at the turning points and matches the exact Schrödinger result — for the harmonic oscillator, semiclassical quantization happens to be exact.

Step 4. Read the structure. Because $J$ is linear in $E$ , equal slabs of area mean equal steps of energy: $\Delta E = \hbar\omega$ between neighbours, the evenly spaced ladder that makes the oscillator the prototype of the photon and the phonon. A quartic well, where $J \propto E^{3/4}$ , has levels that spread apart; a gravitational bouncer, crowding together. The spectrum's shape is the area function's shape. As a consistency check: $E_0 = \tfrac12\hbar\omega$ matches the exact quantum ground state, and the level spacing $\hbar\omega$ matches the photon energy of a mode at frequency $\omega$ .

Limiting cases

case	result	meaning
$E \to 0$	ellipse shrinks to the fixed point at the origin	the oscillator at rest; a centre, stable
$\omega \to 0$ (soft spring)	ellipses flatten toward horizontal lines	free-particle portrait recovered; $p$ nearly constant
$\omega \to \infty$ (stiff spring)	ellipses squeeze vertical	position pinned, momentum swings
weak damping added	ellipse becomes an inward spiral, area decays as $e^{-\gamma t}$	no longer Hamiltonian — see the damped-oscillator article
quantum regime	only areas $(n + \tfrac12)h$ allowed; ground state keeps area $h/2$	zero-point motion: the dot can never sit at the origin

Common trapBigger ellipses must take longer — and other readings the ellipse forbids

Two seductive misreadings. First: "a larger orbit is a longer path, so the period must grow with amplitude." False for the quadratic potential — the state moves faster in exact proportion to the extra distance (peak speed $\omega A$ scales with amplitude), and every ellipse closes in $T = 2\pi/\omega$ . Isochronism fails the moment the potential deviates from quadratic, which is why it is a special property, not a general one. Second: the ellipse is not a picture of the motion in real space. The mass moves on a straight line; the ellipse lives in state space, with axes of different physical units, and its "shape" (aspect ratio) can be changed at will by choosing different units or the rescaling (4). Only unit-honest statements survive that freedom: the enclosed area $2\pi E/\omega$ , the period, the topology of nested loops. Never read metric properties — lengths, angles, eccentricities — off a phase portrait.

Worked examples

1 · From spring data to the exact phase ellipse (confidence builder)

A block of mass $m = 0.20\ \mathrm{kg}$ on a spring of stiffness $k = 5.0\ \mathrm{N/m}$ is released from rest at $q = 0.10\ \mathrm{m}$ . Find the phase trajectory, the state a time $T/8$ after release, and the enclosed area.

Step 1. Frequency and energy: $\omega = \sqrt{k/m} = \sqrt{5.0/0.20} = 5.0\ \mathrm{rad/s}$ ; $E = \tfrac12 k A^2 = \tfrac12 (5.0)(0.10)^2 = 0.025\ \mathrm{J}$ .

Step 2. Semi-axes of the ellipse: $q_{\max} = A = 0.10\ \mathrm{m}$ and $p_{\max} = m\omega A = (0.20)(5.0)(0.10) = \boxed{0.10\ \mathrm{kg\,m/s}}$ — check via $\sqrt{2mE} = \sqrt{2(0.20)(0.025)} = 0.10$ . ✓

Step 3. State at $t = T/8$ , i.e. $\omega t = \pi/4$ . From (5) with $\varphi = 0$ : $q = A\cos(\pi/4) = 0.071\ \mathrm{m}$ , $p = -m\omega A \sin(\pi/4) = \boxed{-0.071\ \mathrm{kg\,m/s}}$ — the dot has slid one-eighth of the way clockwise, into the lower-right quadrant (moving toward equilibrium, speeding up).

Step 4. Enclosed area: $J = \pi q_{\max} p_{\max} = \pi(0.10)(0.10) = \boxed{0.031\ \mathrm{J\,s}}$ ; cross-check $2\pi E/\omega = 2\pi(0.025)/5.0 = 0.031\ \mathrm{J\,s}$ . ✓ In quantum units this orbit spans $J/h = 0.031/6.6\times10^{-34} \approx 4.7\times10^{31}$ cells — deep classical territory.

2 · How cold must a quantum drum get? (exam level)

A nanomechanical resonator vibrates at $f = 5.0\ \mathrm{GHz}$ . Locate its thermal state on the phase-space ladder at room temperature ( $T = 300\ \mathrm{K}$ ) and at a dilution-refrigerator temperature of $10\ \mathrm{mK}$ . Boltzmann constant $k_B = 1.38\times10^{-23}\ \mathrm{J/K}$ .

Step 1. Level spacing: $\omega = 2\pi f = 3.14\times10^{10}\ \mathrm{rad/s}$ , so $\hbar\omega = (1.055\times10^{-34})(3.14\times10^{10}) = \boxed{3.3\times10^{-24}\ \mathrm{J}}$ .

Step 2. Room temperature: thermal energy $k_B T = (1.38\times10^{-23})(300) = 4.1\times10^{-21}\ \mathrm{J}$ . Mean occupation $\bar n \approx k_B T/\hbar\omega = 4.1\times10^{-21}/3.3\times10^{-24} \approx \boxed{1250}$ : the thermal blob smears over the first thousand-plus ellipses of the ladder — classical physics is an excellent description.

Step 3. At $10\ \mathrm{mK}$ : $k_B T = 1.38\times10^{-25}\ \mathrm{J}$ , so $\hbar\omega/k_B T = 3.3\times10^{-24}/1.38\times10^{-25} \approx 24$ . The occupation is Bose-suppressed, $\bar n \approx e^{-24} \approx 4\times10^{-11}$ : the drum sits in the $n = 0$ ellipse — the ground state, its residual motion the irreducible half-cell $h/2$ of zero-point area.

Step 4. Interpretation on the portrait. Cooling shrinks the thermal patch down the ladder of ellipses until quantum mechanics calls a halt at area $h/2$ . This is precisely the regime reached by the drum experiments of the last decade — and the reason GHz resonators (large $\hbar\omega$ ) got there first, while a kHz mirror would need nanokelvins: the criterion is not temperature alone but the ratio $\hbar\omega/k_B T \gtrsim 1$ .

Step 5. Sanity check the crossover: equality $k_B T = \hbar\omega$ at $T = 3.3\times10^{-24}/1.38\times10^{-23} = 0.24\ \mathrm{K}$ — dilution refrigerators (base $\sim 10\ \mathrm{mK}$ ) pass this line with a factor of 24 to spare, matching step 3.

Origins · 1 minMax Planck and how this was discovered

Max Planck

1858–1947 · Berlin

Planck's 1900 act of desperation — forcing the oscillators in a glowing cavity to exchange energy only in lumps $E = h\nu$ — was, in phase-space language, the decree that an oscillator's ellipse may only enclose whole cells of a new fundamental area, $h = 6.6\times10^{-34}\ \mathrm{J\,s}$ . He called $h$ the quantum of action, and the name is exact: action is precisely the phase-space area of this article, and Planck had discovered that nature pixelates it. A reluctant revolutionary, he spent years trying to reabsorb $h$ into classical physics and failed — the constant instead swallowed physics whole. When Bohr and Sommerfeld built the old quantum theory, their quantization rule $\oint p\,\mathrm{d}q = nh$ was Planck's cell applied to every periodic system; the oscillator, whose area grows linearly with energy, rewarded them with the evenly spaced ladder his radiation law had demanded all along.

History

1678

— Hooke publishes his spring law as the anagram's solution "ut tensio, sic vis", fixing the quadratic potential.

1900

— Planck quantizes cavity oscillators in units of a new constant with the dimensions of action — of phase-space area.

1913

— Bohr quantizes the hydrogen atom; Ehrenfest soon identifies the invariant worth quantizing as the adiabatic action.

1915

— Wilson and Sommerfeld state the general rule: allowed orbits enclose phase-space area

nh

1926

— Schrödinger's equation confirms the oscillator ladder exactly, upgrading

nh

(n + \tfrac12)h

with the zero-point half-cell.

2010

— O'Connell and colleagues cool a mechanical resonator into its quantum ground state — the lowest ellipse, occupied in the lab.

Going further

References. Morin, Introduction to Classical Mechanics, Ch. 15 (Hamiltonian mechanics) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 8 · Landau & Lifshitz, Mechanics, 3rd ed., §21–§23 · Taylor, Classical Mechanics, Ch. 5 & 13 · Arnold, Mathematical Methods of Classical Mechanics, Ch. 3 The Hamiltonian and canonical equations follow from the Legendre transform of the SHM Lagrangian. The energy-ellipse semi-axes, action $J=2\pi E/\omega$ , and period $T=2\pi/\omega$ are derived above. The fixed point is a centre with eigenvalues $\lambda=\pm i\omega$ , confirmed by the trace-determinant method. Divergence-free flow and unit-Jacobian area preservation follow from Liouville's theorem; the ellipse area $\pi q_{\max}p_{\max}=2\pi E/\omega$ and the unit determinant of the rotation matrix are the key checkable quantities. The Bohr–Sommerfeld result $E_n=(n+\tfrac12)\hbar\omega$ is the standard semiclassical value, exact for the harmonic oscillator.

Builds on: The Hamiltonian Function · Hamilton's Canonical Equations · Generalized Momenta · Phase Space. Leads to: Phase Portraits and Trajectories · Fixed Points and Stability in Phase Space · The Pendulum Phase Portrait and Separatrix · Hamiltonian Flow and the Phase Fluid · Liouville's Theorem.

Is the phase-space ellipse related to the elliptical orbit of a planet?

Only by the word "ellipse" — the two live in different spaces and have different causes. A Kepler orbit is a curve in real space, its shape fixed by the inverse-square force. The oscillator's ellipse is a curve in state space, whose axes carry different units; it is an ellipse simply because the energy is a sum of two squares, and its aspect ratio changes if you change units. Amusingly, the correspondence runs deeper the other way: a 2D isotropic oscillator's real-space orbit is also an ellipse (centred on the force centre rather than focused on it), which is a separate classical gem about closed orbits, not a phase-space statement.

Why does the phase point move clockwise and not counterclockwise?

Check two arrows. At the right turning point $(A, 0)$ the spring pulls back, so $\dot p = -m\omega^2 q < 0$ : the dot moves down. At the bottom $(0, -p_{\max})$ , $\dot q = p/m < 0$ : it moves left. Down on the right, left at the bottom — clockwise, always, given the standard axes ( $q$ right, $p$ up). This is not oscillator-specific: every Hamiltonian orbit circles a potential minimum clockwise, because $\dot q$ shares $p$ 's sign and $\dot p$ opposes the potential gradient. A counterclockwise portrait means an axis or a sign is flipped.

Isochronism seems too good to be true. What exactly breaks it in a real system?

Any anharmonicity — any term in the potential beyond quadratic. The pendulum's $-\cos\theta$ potential softens at large angle, so big swings lag (the period grows, diverging at the separatrix); a stiffening $q^4$ correction does the reverse. Quantitatively, a perturbation $\epsilon q^4$ shifts the period at order $\epsilon A^2$ : the failure grows with amplitude squared, which is why small-amplitude clocks kept good time even before escapements were clever. Isochronism also fails for damped or driven oscillators in the transient sense, though the linear damped oscillator keeps an amplitude-independent quasi-period — linearity, not energy conservation, is the true parent of isochronism.

What is special about the rescaled coordinates Q and P — aren't they just a units trick?

They are a canonical transformation: the product of the two scale factors is 1, so areas — the physically meaningful phase-space quantity — are untouched, and Hamilton's equations keep their form. That is what elevates the trick above bookkeeping: in $(Q,P)$ the flow's true nature (rigid rotation) becomes visible, and quantities computed there (areas, the action, occupation numbers) are automatically correct in any coordinates. This is the first and simplest example of the transformation freedom that later articles industrialize — action-angle variables are precisely the polar coordinates of this rotating picture.

How literally should I take the claim that the quantum oscillator "occupies an ellipse of area (n+1/2)h"?

As a semiclassical picture with exact energy output but fuzzy geometry. Quantum states are not curves; the $n$ -th eigenstate is better imagined as a ring-shaped blur straddling the classical ellipse of area $(n+\tfrac12)h$ , with the ground state a round blob of area $\sim h/2$ centred on the origin (its Wigner function makes this precise). What is exact for the oscillator is the spectrum $E_n = (n + \tfrac12)\hbar\omega$ that the area rule produces — a happy accident of the quadratic potential; for other potentials Bohr–Sommerfeld is only the leading term of a systematic (WKB) approximation. The picture earns its keep in the lab: coherent states are minimal blobs circling the ellipse exactly like classical dots.

Why is the action J the thing that stays constant under slow changes, rather than the energy?

Energy is not even a candidate: stiffen the spring and you do work on the system, so $E$ must change. The nontrivial fact — Ehrenfest's adiabatic theorem — is that when the change is slow compared with the period, the area of the orbit is what survives: the ellipse deforms (taller in $p$ , narrower in $q$ as $\omega$ grows) at exactly fixed enclosed area, so $E = J\omega/2\pi$ rises proportionally to $\omega$ . A hand-waving proof: over one period the work done by slow stiffening averages against the orbit's symmetric sampling of positions, and the bookkeeping residue is $\mathrm{d}(E/\omega) = 0$ . This invariance is why quantum numbers don't change under slow driving — nature quantizes the very quantity that slow changes cannot alter.

The damped oscillator spirals inward. Can I still use anything from this article?

The reading skills, yes; the invariants, no. With weak damping $\gamma \ll \omega$ the trajectory is almost the same ellipse, shrinking by the factor $e^{-\gamma T}$ per lap — a tidy spiral whose geometry the damped-oscillator article develops. But the enclosed area now decays, $J(t) = J(0)e^{-2\gamma t}$ (both semi-axes shrink like $e^{-\gamma t}$ ), so the action is no longer invariant, Liouville fails (the flow has negative divergence $-2\gamma$ ... precisely why an attractor at the origin can exist), and there is no quantum ladder to read off. Phase-space language survives dissipation; phase-space conservation laws do not.

Where does the factor π in the area formula come from — why not 2π?

From the geometry of an ellipse: area $= \pi \times$ (semi-axis) $\times$ (semi-axis) $= \pi q_{\max} p_{\max}$ . Substituting the semi-axes each contribute a $\sqrt{2E}$ -type factor: $\pi\sqrt{2E/m\omega^2}\sqrt{2mE} = \pi \cdot 2E/\omega = 2\pi E/\omega$ . So the $2\pi$ in the final formula is the ellipse's $\pi$ times the 2 from the pair of square roots. A slicker mnemonic: $J = ET$ — area equals energy times period — and since $T = 2\pi/\omega$ , the $2\pi$ is really the period's.

The Harmonic Oscillator in Phase Space

Where this lives today

The physical picture

The core equations

Which equation answers which question

Limiting cases

Worked examples

1 · From spring data to the exact phase ellipse (confidence builder)

2 · How cold must a quantum drum get? (exam level)

Going further

Before this

Where this leads