Where the quantum picture hands back the old one

Every other essay in this field has argued that the classical picture fails. This one asks the reverse question, and the answer is less tidy than it is usually made to sound.

The oscillations do not shrink. |ψ_n|² is 2sin²(nπx/L) for every n, so it touches zero and reaches twice the classical value however high the quantum number goes. The usual statement — that the quantum result approaches the classical one for large n — is false if “approaches” means what it normally means.

What actually converges #

What converges is the density averaged over a window, and it converges as 1/n.

Take any fixed interval of the box, of width w, and average the quantum density over it. The oscillations partially cancel, leaving a departure from the classical value bounded by about 1/(nπw). At n = 1 with a window an eighth of the box wide, that bound is useless — the window is less than one period, so the average is nearly the local value. At n = 16 it is about 0.16, and at n = 10⁶ it is 10⁻⁷.

So the classical answer is recovered not because the quantum wiggles die away but because any measurement with finite resolution averages over many of them. The classical limit is a statement about the observer’s resolution as much as about the system.

That distinction has a practical consequence. The site’s figure gate measures this claim, and the first version of the check measured the wrong thing: it took the mean absolute departure of the drawn density from the drawn classical line, found it constant to two decimals at every n, and was correct — the pointwise departure genuinely does not fall. The check was rewritten to slide a fixed window across the middle of the box and take the worst window average, which does fall, and by the factor the argument predicts.

High in the ladder the states crowd toward the classical distribution, and drawing the tenth, twentieth and thirtieth together shows the approach rather than asserting it: the oscillations get finer and their average tracks the classical probability more and more closely. What never happens is that the oscillations go away. They become too fine to resolve, which is a statement about the instrument and not about the state.

The other three things that also have to happen #

Large quantum numbers alone are not sufficient, and the places where they are not sufficient are where the interesting physics is.

The levels have to be unresolvably close. A box’s levels go as n², so the relative gap $(E_{n+1} - E_n)/E_n = (2n+1)/n^2$ falls as $2/n$. At n = 10²⁶ — a marble in a shoebox — the relative gap is 10⁻²⁶ and no measurement could see the discreteness. At n = 3 it is 0.78 and nothing could miss it.

The state has to be a wave packet, not an energy eigenstate. A single stationary state does not move, at any n. A classical particle bouncing between walls does. What corresponds to the classical motion is a superposition of many neighbouring levels, forming a localised packet that travels — and that packet is not an eigenstate of anything.

The packet has to stay localised. It does not, in general. A free packet spreads, and a packet in a box spreads, reflects, and after a while spreads over the whole box; only then, at very long times, does it partially reassemble. The classical trajectory is recovered for a while and not forever, and how long depends on the system.

Localising a particle enough to say where it is requires a spread of momenta, which is the trade of an earlier rung — and the spread then guarantees the packet will disperse. So a classical trajectory is not a limit of a stationary state at all. It is a limit of a moving bundle whose lifetime is finite, and asking a stationary state to look classical is asking the wrong object.

The numbers, for something ordinary #

It is worth putting a real object through the arithmetic, because the magnitudes are what make the discussion feel abstract and should not.

A marble of one gram in a box ten centimetres wide, rolling at a millimetre per second. Its kinetic energy is 5 × 10⁻¹⁰ joules; the box’s ground-state energy is 5.5 × 10⁻⁶⁴ joules; so its quantum number is n ≈ 3 × 10²⁶.

Three consequences follow, and each is a different reason the classical description is safe.

The gap to the next level is a relative change of 7 × 10⁻²⁷ in energy, which is 10⁻³⁶ joules. No measurement of energy has ever approached that resolution and none will.

The probability density oscillates 3 × 10²⁶ times across the box, so the spacing between its zeros is 3 × 10⁻²⁸ metres — eighteen orders of magnitude below a nuclear diameter. Any instrument averages over an unimaginable number of oscillations.

And the marble’s de Broglie wavelength is 6.6 × 10⁻²⁸ metres, so nothing in the universe is narrow enough to diffract it.

Everything with a mass has a wavelength, and for anything above molecular size that wavelength is many orders of magnitude below every length that exists. A grain of dust at a micrometre a second has a de Broglie wavelength around $10^{-19}$ m — a billion times smaller than a nucleus. The classical description of a macroscopic object is not an approximation that happens to be good; it is one whose corrections have no possible observable consequence.

Why “ħ → 0” is a bad way to say it #

The classical limit is often written as ħ → 0, and the phrase is convenient shorthand for something that cannot literally be done.

ħ is a dimensional constant with a fixed value. What can be small is a dimensionless ratio — the system’s characteristic action divided by ħ, or its de Broglie wavelength divided by the scale over which the potential varies. Saying “ħ → 0” is saying that ratio is large, and the ratio is a property of the system rather than of the universe.

The distinction matters because the same object can be in both regimes depending on what is being asked. An electron in a cathode-ray tube follows a trajectory to excellent accuracy — the wavelength is picometres and the apparatus is centimetres — while the same electron in an atom has no trajectory at all. Nothing about the electron changed.

It also matters because some quantities have no classical limit. Tunnelling goes to zero as the ratio grows, so it is not a correction to a classical process — it is a process whose classical value is exactly nothing. Spin has no classical counterpart. The exclusion principle survives the limit and is not derivable from classical mechanics. A “limit” in which some quantities converge and others vanish is not a limit in the ordinary mathematical sense.

What decoherence supplies #

There is a fourth ingredient, and it is the one that explains why the world looks classical rather than merely why the numbers agree.

Everything above concerns an isolated system. Nothing in it explains why a large object is never found in a superposition of two positions — the mathematics permits such a state at any mass, and interference has been demonstrated with molecules of 25,000 atomic mass units.

What prevents it is the environment. Every air molecule, every thermal photon, every stray field interacts with the object and carries away a trace of where it is. Once those traces exist anywhere, the two branches of a superposition can no longer interfere — which is exactly the condition the double slit is sensitive to — and what remains behaves like an ordinary probabilistic mixture.

The rates involved are what makes the effect decisive rather than gradual. A dust grain a micrometre across, in air at room temperature, is struck often enough that a superposition of two positions a micrometre apart decoheres in around 10⁻³¹ seconds. In a good laboratory vacuum it is still microseconds. There is no realistic isolation under which a macroscopic superposition survives long enough to be observed, and the reason is a collision rate rather than a law.

So the classical world has two separate explanations and both are needed. Large quantum numbers explain why the values agree. Decoherence explains why the superpositions are missing. Neither substitutes for the other.

The principle as a design constraint #

Bohr introduced correspondence in 1913 not as an observation but as a requirement: whatever the new theory turned out to be, it had to reproduce the old one where the old one worked.

Used that way it is a powerful tool, because it fixes constants. The Rydberg constant’s value can be derived by requiring that the frequency of a transition between adjacent high-n levels equal the orbital frequency a classical electron would radiate at — which it does, exactly, in the limit — and that requirement pins the constant with no spectroscopy at all.

The same requirement operates in every subsequent theory. Special relativity has to reduce to Newtonian mechanics at low speed, and it does; general relativity has to reduce to Newtonian gravity in weak fields; quantum field theory has to give back quantum mechanics at low energies. A candidate theory that failed such a test would be refuted by centuries of successful measurements without any new experiment being done.

What survives the limit unchanged #

Correspondence is usually discussed as a matter of what disappears. The list of what does not disappear is shorter and more interesting, and every item on it is a macroscopic fact.

Chemistry. Every bond energy, every molecular geometry and every reaction rate is a quantum result, and none of it becomes classical for a large molecule. Scaling up a protein does not make its structure classical; it makes it a large quantum calculation.

The hardness of matter. The exclusion principle has no classical limit at all — there is no version of classical mechanics in which identical particles are forbidden a state — and it is what stops a hand passing through a table.

Magnetism. The Bohr–van Leeuwen theorem shows that classical statistical mechanics predicts zero magnetisation at thermal equilibrium for any system of classical charges. Every magnet is a quantum object, at any size.

Conductivity. Whether a material is a wire or a window is decided by where a band’s filling stops, which is a counting result about discrete states. A classical treatment gives no gap and no distinction.

The colour of anything. Spectral lines, blackbody radiation and the reflectivity of a metal are quantum results, and a room full of coloured objects is a room full of macroscopic quantum effects.

The pattern in that list is worth naming. What becomes classical is the dynamics of a single degree of freedom — where something is, how fast it is going. What stays quantum is anything that depends on discreteness, on identity, or on the structure of states. A world in which only the first became classical is exactly the world observed.

Where the correspondence is misused #

Three misreadings are common and each hides something real.

“Quantum effects are small.” They are not small; they are unresolvable in some regimes and dominant in others. Superconductivity, magnetism, the colour of gold and the hardness of a table are all macroscopic and all quantum. The correct statement is that certain quantum distinctions are unresolvable for large systems.

“Classical physics is an approximation to quantum physics.” In the sense of numerical agreement, yes. In the sense of conceptual containment, no: a classical trajectory is not a limit of a quantum state, because there is no family of quantum states converging to one. What converges are predictions about specific measurements, and only those.

“The correspondence principle explains the classical world.” It explains the agreement of numbers. The absence of macroscopic superpositions is a separate question with a separate answer, and conflating the two makes the measurement problem look solved when it is not.

The oscillator, where the classical answer is not flat #

The box is the simplest case and it is misleading in one respect, so the harmonic oscillator is worth putting beside it.

A classical oscillator moves fastest at the centre and stops at the turning points, so it spends most of its time near the ends of its swing. Its probability density is therefore lowest in the middle and diverges at the turning points — the opposite shape to the box’s flat line.

The quantum ground state has exactly the opposite shape again: a single hump centred on the middle, with its maximum where the classical particle is least likely to be. At n = 1 the two answers disagree about which end of the box is favoured, not merely by how much.

As n rises, the quantum density develops n + 1 humps whose envelope rises toward the turning points, and by n = 20 that envelope is a close match to the classical curve while the humps themselves are as pronounced as ever. Averaged over any window the two agree; pointwise they never do.

That last point is the reason the oscillator is the standard example. For most systems the quantum and classical frequencies agree only asymptotically; for the harmonic oscillator they agree exactly, which is why Planck’s assumption about cavity modes could be made without disturbing anything classical about the modes themselves.

The atom large enough to watch it happen #

Everything above is an argument about what would happen at large quantum numbers. There is a system in which large quantum numbers are prepared deliberately and the limit is watched rather than reasoned about.

Excite an atom’s outer electron to a principal quantum number near a hundred and it becomes a Rydberg atom. The orbit radius goes as n², so at n = 100 the electron is half a micrometre from the nucleus — an atom the size of a small bacterium. The orbital period goes as n³, so instead of femtoseconds it is a hundred picoseconds or so, slow enough for a laser pulse to be short compared with it. And the level spacing goes as 1/n³, so a short pulse is spectrally broad enough to excite many neighbouring levels at once.

That last point is the one that matters, because this essay’s third requirement was that the state be a packet rather than an eigenstate. A pulse that populates a band of adjacent Rydberg levels with the right relative phases produces exactly that: a localised bundle of probability, at one place on the orbit, travelling round it. Such packets have been made and their motion followed, by firing a second pulse at an adjustable delay and seeing whether the electron is where it should be. The electron goes round, on schedule, for several orbits — a classical Kepler orbit, assembled out of stationary states none of which moves at all.

And then it stops behaving classically, in the way the third requirement predicted. The levels are not exactly evenly spaced, so the components drift out of step, the packet spreads round the whole orbit, and the classical description fails after a number of orbits that the anharmonicity fixes. The correspondence limit is not a state the atom reaches; it is a phase the atom passes through.

Where large quantum numbers are not enough #

There is a class of systems for which the argument of this essay fails outright, and the failure is not exotic: it is most systems.

A packet spreads, and how fast depends on the dynamics it is spreading in. In a regular system — a box, an oscillator, a Kepler orbit — the spreading is slow, going as a power of the time, so a packet built at a large quantum number tracks the classical trajectory for a correspondingly long while. In a chaotic system neighbouring trajectories separate exponentially, and the packet’s width does the same.

Exponential growth from a very small starting width reaches the size of the system after a time proportional to the logarithm of the ratio between them. That is the Ehrenfest time, and a logarithm is brutal here: taking the classical action to be $10^{20}$ times Planck’s constant, which is a thoroughly macroscopic object, buys only about forty-six divided by the Lyapunov exponent. Forty-six times the timescale on which the system forgets its initial conditions, and no more, however large it is made.

The consequence is a genuinely uncomfortable one. Hyperion, a moon of Saturn some three hundred kilometres across, tumbles chaotically with a characteristic time of a few tens of days. Running the estimate for it gives an Ehrenfest time of roughly twenty years — after which an isolated Hyperion would be in a quantum superposition of wildly different orientations, and would have no classical description at all. It is not a statement about a small object or a delicate one.

What rescues it is the fourth ingredient rather than the first. Hyperion is not isolated: it scatters sunlight, and each scattered photon carries away information about which way it is facing. Decoherence acts on it enormously faster than the spreading does, and the classical description survives because the moon is being measured continuously by its surroundings.

So the two explanations offered above are not parallel alternatives after all. For a regular system, large quantum numbers suffice for the values and decoherence is needed only to suppress superpositions. For a chaotic one — which is to say for most of the world — decoherence is doing both jobs, and there is no classical limit without it.

What the picture cannot show #

The opening figure draws a classical density as a flat dashed line, which is correct for a box and hides the fact that the classical density is generally not flat. For a harmonic oscillator it piles up at the turning points, where the particle moves slowest, and the quantum density at high n oscillates about that shape instead. The correspondence is to whatever the classical answer is, not to a uniform one.

The figure also draws stationary states, which as argued above are the wrong objects for the comparison. The honest picture of the classical limit would be an animation of a packet bouncing between the walls and slowly dispersing, and a static drawing cannot carry it.

And nothing here shows the environment, which is where half the answer lives. Every figure in this field draws an isolated system, and every real system is not one.

Where the ladder goes next #

The rungs from here: the Ehrenfest theorem, which shows that the expectation values of position and momentum obey Newton’s laws exactly, and the conditions under which that implies anything about a trajectory; the WKB approximation, which is the classical limit made into a calculational method; quantum revivals, where a spread packet reassembles after a long time; the classical limit of the harmonic oscillator, where the turning-point pile-up makes the correspondence visible in a way the box cannot; and decoherence theory in detail, with the timescales that decide which superpositions survive.

The claim to carry forward is the one the figure was rewritten to state properly. Nothing about the quantum answer becomes classical. The oscillations are full-sized at every quantum number, the levels are always discrete, and the wavefunction never stops being a wavefunction. What happens as a system grows is that the structure gets finer than anything can resolve and the environment removes what would have distinguished the branches — and those are two facts about measurement, not two facts about the system.