The Photon BEC
Condensation of light with no pump and no inversion: give a photon gas an effective mass, a low-frequency floor, and a conserved number, and it condenses by thermodynamics alone — the bench correction to the laser chapter’s closing image, run on a cutoff the vacuum has carried natively all along
The Splinter in the Closing Image
The laser chapter ended on the framework’s most quotable sentence — the universe is a laser that reached threshold once — and the sentence has a splinter in it. It earns its keep on two counts: macroscopic occupation of a single mode, and a spontaneously chosen U(1) phase, with the laser threshold formally a second-order phase transition. But a laser is pumped. Its coherence is a non-equilibrium steady state — inversion held open by an energy circuit, gain clamped to loss, the whole arrangement dead within microseconds of the wall plug going quiet. The boil was nothing like that: a medium cooling through a transition at equilibrium — no pump, no inversion, no circuit to hold open. If the laser were the bench’s only demonstration of macroscopic single-mode occupation, the framework’s founding event would have a driven analog for an undriven mechanism, and a careful reader should refuse the trade.
The splinter comes out because the bench has done better. In 2010, Klaers, Schmitt, Vewinger & Weitz condensed light itself (Nature 468, 545): a photon gas brought to genuine room-temperature equilibrium and pushed over the Bose–Einstein threshold by number, not by gain — no inversion anywhere in the apparatus. And the trick that made it possible — the thing that turns a photon gas from uncondensable to condensable — is a structure the framework’s vacuum has carried since the modon floor was derived: a low-frequency cutoff below which no compact mode exists. This chapter walks through the experiment, reads its machinery in substrate terms, and then makes the correction the laser chapter’s closing image owed: the universe is not a laser that reached threshold. It is a condensate that crossed its transition temperature.
Why Light Was Not Supposed to Condense
Einstein’s 1925 prediction — cool a gas of conserved bosons and, below a critical temperature, the ground state takes macroscopic occupation — was written for atoms, and for eighty-five years the standard account of why it skips light was a clean two-count obstruction:
- No conserved number. Blackbody photons carry zero chemical potential — Planck’s law has \mu=0 built in — because the walls create and destroy them freely. Cool the cavity and the photons do not crowd into the lowest mode; they leave, absorbed into the walls. The gas dims instead of condensing.
- No thermalizing contact that spares the number. Photons barely interact with each other, so a photon gas cannot reach equilibrium on its own; and the matter that could thermalize it is precisely the matter that eats it.
The Bonn experiment repairs both counts with one small object: a dye-filled microcavity, mirrors 1.46\;\mum apart. Three design moves, each of which the framework recognizes:
- A cutoff that gives light a mass. At that mirror spacing only one longitudinal mode order (q=7) survives inside the dye’s spectral range, which pins a minimum energy \hbar\omega_c\approx2.1 eV (cutoff wavelength 585 nm). For the surviving transverse motion the dispersion is E\simeq\hbar\omega_c+\hbar^2k_\parallel^2/2m_\text{ph} — a massive, two-dimensional, nonrelativistic particle, with m_\text{ph}=\hbar\omega_c/c^2\approx6.7\times10^{-36} kg, ten orders of magnitude lighter than the atoms of a conventional BEC. Critical temperature scales inversely with mass, which is why this condensate forms at 300 K rather than at nanokelvin. Below the cutoff there are simply no modes: the gas has a floor, and therefore a ground state to condense into.
- A trap. The mirrors are curved, so the cutoff energy rises gently off-axis: a harmonic potential (\Omega\approx2\pi\times41 GHz) for the massive photons. A uniform two-dimensional Bose gas does not condense at finite temperature; a trapped one does.
- Thermal contact that conserves the number. The dye absorbs and re-emits each photon many times before the cavity loses it. In the one-coupling table’s terms this is rows two and three run in closed cycle — capture, then re-shed — with the boundary’s vibrational bath taking and giving the small change in between. The Kennard–Stepanov relation (the ratio of a dye’s absorption to its emission at each frequency follows a Boltzmann factor at the solvent temperature) guarantees that every cycle re-prices the photon’s energy from the dye’s thermal deck while returning the photon itself. Run many cycles and the photon gas acquires the dye’s temperature at fixed mean photon number — light with a genuine, nonzero chemical potential.
Note what move 3 is in substrate language: nothing. No new coupling is required. Capture and triggered release were already one geometry run in two directions, with B_{12}=B_{21}; the Kennard–Stepanov factor is that same symmetry dressed in Boltzmann weights. An uninverted boundary bath is therefore not an amplifier but a thermometer — each contact drags the modon gas toward the bath’s temperature — and thermalization of light is the cloning coupling run to detailed balance instead of run past it. A laser is this coupling forced out of balance by a pump; a photon BEC is the same coupling left alone.
The result, measured: the in-cavity spectrum relaxes to a room-temperature Bose–Einstein distribution, and when the photon number is raised past
N_c \;=\; \frac{\pi^2}{3}\left(\frac{k_BT}{\hbar\Omega}\right)^{\!2} \;\approx\; 77{,}000 \qquad (\text{observed: } \sim6\times10^4),
a condensate spike appears at the trap centre, at the cutoff wavelength — macroscopic occupation of the ground mode, reached by crowding, not by gain. The gas’s caloric properties have since been measured outright (Damm et al. 2016: heat capacity through the transition — light, calorimetered), and the phenomenon is no one-off: photon condensates now run in fiber cavities and plasmonic lattices as well. Bose–Einstein condensation of photons was “impossible” only as long as light lacked a floor and a conserved number. Give it those two atom-like properties and light condenses like anything else.
Same Occupation, Opposite Bookkeeping
Laser light and condensed light can look identical to a spectrometer — one bright mode, one phase. The bookkeeping underneath is opposite at every line:
| Laser | Photon BEC | |
|---|---|---|
| population | inverted — a side door held open by the pump | Boltzmann — the dye stays uninverted |
| state of the light | driven steady state, far from equilibrium | equilibrium at 300 K |
| occupied mode chosen by | maximum gain | minimum energy (trap ground state at the cutoff) |
| occupation set by | gain clamping to loss | number and temperature, through \mu |
| photon number | not conserved (pump in, loss out) | conserved on average (capture ↔︎ re-shed) |
| statistics | Poissonian | grand-canonical — g^{(2)}(0)\to2 (below) |
| switched off by | cutting the pump | nothing — it has no switch, only a temperature |
The two are not separate devices so much as two ends of one dial. Kirton & Keeling showed (2013) that a single dye-cavity model interpolates continuously between them: what decides laser-versus-condensate is the race between thermalization and loss. When photons leak out faster than the dye can re-price them, coherence must be bought with a pump and the device is a laser; when thermalization wins, the gas equilibrates and condenses for free. A pump is what coherence costs when the quanta leak. Hold that sentence — the vacuum is about to cash it.
There is also a classification note the framework should make explicitly. The laser chapter’s 2×2 grid classifies emission events — what sets each modon’s energy, and whether the events fire in phase. Condensation is not an emission mechanism at all; it is something the gas does after emission and capture have gone to detailed balance. That is why the photon BEC gets a chapter rather than a cell in the grid — it is the first entry in a third bookkeeping the Light section had not yet opened: not frequency, not phase, but occupation.
The Vacuum Has the Trick Built In
Now read the microcavity’s enabling structure against the framework’s vacuum, and the mapping is uncomfortably exact.
The design element that makes light condensable is the cutoff: a frequency floor, below which no mode exists, that hands the gas an effective mass and a ground state. The laser chapter already identified the vacuum’s version of this — the vacuum is a cavity everywhere, and the modon floor is its fundamental:
\hbar\omega_c \approx 2.1\;\text{eV}\ \ (\text{engineered, Bonn}) \qquad\longleftrightarrow\qquad E_\text{min} = 2\pi\,m_1c^2 \approx 13\;\text{meV}\ \ (\text{native, every cell}).
The substrate supports no compact quantum of light below 13 meV — the smallest modon is one cell wide — so the vacuum’s photon gas natively has the structure Klaers had to engineer: a floor, and above it a spectrum of countable quanta with a lowest rung. (The honest disanalogy is flagged now and paid in the accounting: below the microcavity cutoff there is nothing, while below the modon floor light persists as a stretched winding — the substrate’s floor is a transparent change of quantization character, not a wall. The identification is exact for compact, countable, condensable quanta, and only for them.)
And for the boil itself, the mapping gets simpler, not harder — because the species that condensed did not even need the trick. The two atom-like properties the Bonn experiment had to engineer into light — mass and conserved number — are properties the substrate’s own quanta have by construction: a dc1 quantum carries m_1\approx2 meV natively, and the quanta are the medium, so their number cannot leak anywhere. Kirton & Keeling’s dial reads the vacuum at its far end: loss rate zero, thermalization total. A medium of massive, conserved, mutually thermalizing bosons that cools must condense — no pump possible, none needed. The dye microcavity is the demonstration that even light condenses once granted those properties; the substrate’s quanta never lacked them.
One more line of the table lands. In a laser the occupied mode is chosen by maximum gain; in a BEC, by minimum energy. The framework has already located the substrate’s macroscopically occupied mode: it is the lattice’s softest mode — the lowest-lying collective option, not the fastest-amplifying one. The vacuum’s mode was selected the way a condensate’s is, not the way a laser’s is. The occupation, the selection rule, and the absence of a pump all point the same direction.
Not Threshold, but Temperature
So the closing image can be repaired rather than retired. What the laser analogy got right survives untouched: one mode, macroscopically occupied; a U(1) phase chosen spontaneously at a genuine phase transition; phase stiffness by occupation, the Ginzburg rigidity. What it got wrong is the mechanism of arrival, and the photon BEC supplies the correction term for term: not driven but cooled; not inversion but crowding; not gain clamping but chemical potential; not threshold crossed by pumping harder but transition crossed by the temperature falling through T_c — once, on the way down.
The occupation table gains its promised row, and the row is small in number and right in kind:
| System | Grain | Coherent unit | Occupation | Reached by |
|---|---|---|---|---|
| dye-microcavity photon BEC | photon | trap ground mode at the cutoff | N_c\approx8\times10^4 | equilibrium |
| mW He-Ne laser | photon | one cavity mode | \sim6\times10^8 | pumping |
| substrate | dc1 quantum | one effective quantum | \nu\approx8.3\times10^8 | equilibrium |
| superfluid ^3He | atom | one Cooper pair’s reach | \sim10^6 | equilibrium |
The He-Ne matches the substrate’s occupation to a decade and misses its mechanism; the photon BEC misses the occupation by four decades and matches the mechanism exactly. Both comparisons are worth keeping — one calibrates the size of the vacuum’s number, the other its provenance.
The superradiance chapter answered one question the laser image left open — why the medium has been dark ever since (the post-burst subradiant state). This chapter answers the other — why no pump was ever needed. Between them the founding event is now stated entirely in equilibrium and collective-phase language, with the driven, engineered laser demoted from model of the boil to what it always was: the one machine where humans reproduce the boil’s product without being able to afford its method.
The Flickering Condensate
The Bonn system delivered a second result the framework values as much as the first, because it turns an old bookkeeping scruple into a measurement. Textbook BEC is taught in ensembles where condensate number fluctuations are small; the grand-canonical ensemble formally predicts macroscopic fluctuations in the condensate — \delta n \sim \langle n\rangle, the “grand-canonical catastrophe” — a result usually waved off as unphysical because no real reservoir exchanges particles with a condensate. The dye is such a reservoir. Schmitt et al. (PRL 112, 030401, 2014) measured the condensate’s number statistics and found the catastrophe alive and well: condensate populations flickering by of order their own mean, g^{(2)}(0)\to2 — thermal-grade fluctuations — persisting deep into the condensed phase, crossing over toward canonical quiet only as the condensate grows large relative to the dye reservoir. A condensate that holds one mode and flickers like a lamp.
Two lessons, one per direction.
For the discriminator. The laser chapter’s first prediction leans on the claim that the floor shows in statistics, not gain — that photon statistics read mechanism where spectra cannot. The flickering condensate is the bench-grade proof of principle: here are two sources — laser and photon BEC — that can present the same bright single mode with opposite g^{(2)}, differing in nothing but their bookkeeping. What number statistics measure is not brightness or coherence but reservoir contact: who the mode exchanges quanta with, and on what terms. Statistics are precisely the observable you use when two mechanisms share a spectrum. That is the discriminator’s whole logic, demonstrated.
For the vacuum. Ask the Schmitt question of the substrate: what are the number statistics of the vacuum’s own condensate? The answer follows from the same rule. Grand-canonical flicker requires an external reservoir, and the substrate’s condensate has none — it is the whole medium; there is nothing outside the mode for the mode to trade quanta with. The vacuum sits at the reservoir-free, canonical extreme of the very crossover Schmitt measured: number-rigid, fluctuation-quiet — the ensemble-language face of the occupation-rigidity argument the framework already runs, and one more certificate on the stealth vacuum’s silence. The flicker is not a pathology of condensation; it is a property of contact. The substrate has no one to flicker with.
Honest Accounting
Four debts, in the house discipline.
First, nothing here corrects the Bonn results. Klaers, Schmitt, Damm, Weitz and the platforms that followed are standard, quantitatively complete physics; every number above is theirs. The contribution is identification: the enabling cutoff is a structure the vacuum carries natively as the modon floor; thermalization of light is the framework’s one cloning coupling run to detailed balance; and the correct bench miniature of the boil is this experiment, not the laser.
Second, the analogy is structural, not numerical. The Bonn condensate is two-dimensional, harmonically trapped, engineered, and made of photons at 300 K; the boil was three-dimensional, self-assembled, and made of dc1 quanta. No number transfers between them — the N_c\approx8\times10^4 row is a scale handle exactly as the He-Ne occupation coincidence is, and the chapter’s claims ride on the mechanism column of the table, never the occupation column.
Third, the floor is not a hard cutoff, and the chapter must not pretend otherwise. Below the microcavity cutoff no modes exist; below the modon floor light continues as a delocalized winding through a transparent edge. The floor-as-cutoff identification is exact only for compact, countable quanta. Whether the floor additionally functions as the pair-breaking gap that protects the vacuum’s darkness is the same open identification the superradiance chapter declined to claim, and it stays open here.
Fourth, the boil’s thermodynamics are still owed. The framework asserts that the substrate condensed and identifies the mechanism class — equilibrium BEC of massive, conserved quanta — but it has not derived the transition’s critical point from substrate parameters, nor exhibited the condensation dynamics. This chapter supplies the founding event’s correct bench twin; it does not supply its calculation. That debt sits upstream, with the boil and why-matter-won accounts, and this chapter narrows it only by naming which textbook chapter the eventual calculation belongs to.
Place in the Framework
The Light section now reads the substrate off its brightest excitation six ways. The modon floor reads the lattice as a frequency; the scattering ring as a wavevector; Spectrum-Free Light reads the emission mechanism; the laser reads phase at one boundary; superradiance reads phase at N boundaries. This chapter opens the remaining ledger — occupation: what fills a mode when nothing drives it. The answer is that nothing needs to drive it; a gas with a floor, a mass, and a conserved number fills its ground mode by arithmetic the moment it is cold enough, and light itself has now done so on a bench in Bonn, at room temperature, with the pump conspicuously absent. The laser chapter ended: every laser is a small vacuum, switched on. Here is the other half, and the truer half: the photon BEC is a small vacuum at equilibrium — never switched on, and condensed anyway. The universe did not reach threshold once. It reached its condensation temperature once, on the way down — and everything since is the ground mode, holding.