Dressed Light
Couple a photon to matter faster than either can lose the quantum and the exchange itself becomes two new particles — light with mass, collisions, and a thermodynamic future. Dressed light condensed in 2006, quantized its vortices in 2008, flowed past an obstacle without friction in 2009, and in 2025 crystallized while staying coherent — the full résumé this paper attributes to the vacuum, performed by a fluid that is half photon. A condensate that leaks carries a Goldstone mode that diffuses instead of propagating, and the sky shows no such window
The Debt, Called In
The photon-BEC chapter named polariton condensates the right place to learn what “equilibrium enough” means and left them waiting on “a chapter about dressed light in matter.” The comb chapter logged the deferral; the laser-cooling chapter renewed it with the door’s hinges oiled; and the supersolid chapter raised the interest rate — a supersolid made of light now exists, and a paper whose thesis is that the vacuum is a condensate carrying light had better have an account of light that condenses. What does light become when it is coupled to matter — and why do the children of that coupling behave, item for item, like this paper’s vacuum?
In the substrate lens there is no such thing as bare light. The modon is a counter-rotating circulation pattern of the substrate — a collective excitation of a medium, carrying that medium’s granularity as a floor and its cell size as a ring. Dressed light then is the only kind of light there has ever been. The polariton is what the dressing looks like when the seam is sewn coarsely enough to see.
The Fourth Regime of the One Coupling
The laser chapter reduced light–matter contact to one coupling run in three regimes: capture, cloning re-shed, spontaneous shed. All three show that the exchange completes. A boundary captures a modon and keeps it long enough to call the capture an event; the re-shed is a second event; between them the quantum has a definite owner. That assumption has a validity condition, and it is the memory ladder’s: the exchange must be slower than the ring-down. Couple weakly and each handoff finishes before the next begins — perturbation theory, Fermi’s golden rule, the entire weak-coupling canon.
Now break the condition on purpose. Put one sharp matter resonance — a quantum-well exciton, a bound electron–hole circulation with a clean transition — between two mirrors spaced so the photon that the exciton sheds is returned before the exciton forgets it. If the round-trip exchange rate \Omega_R beats every loss rate in the problem — the exciton’s dephasing, the mirror’s leak — then capture and re-shed stop being events and become an oscillation: the quantum is not in the light and not in the matter but in the alternation, traded back and forth coherently, and the question “which owns it” loses its meaning. Diagonalize that back-and-forth and the true normal modes are the symmetric and antisymmetric mixtures — the upper and lower polariton, split by \hbar\Omega_R, each a fixed blend of photon and exciton with the blend set by detuning (the Hopfield coefficients: \cos\theta\,|\text{light}\rangle + \sin\theta\,|\text{matter}\rangle, with a knob on \theta). Hopfield wrote the construction down for bulk crystals in 1958 (Phys. Rev. 112, 1555); Weisbuch and colleagues built the microcavity version and watched the two branches repel around their avoided crossing in 1992 (PRL 69, 3314). Where the bare photon line and the bare exciton line would cross, the dressed branches refuse — level repulsion, the oldest signature of two oscillators that have stopped being two.
The crystal-optics chapter already met the propagating version of this — the phonon-polariton, which it read as a modon partially dissolved into the crystal’s boundary structure. Strong coupling is that dissolution caught in a bottle and pushed past halfway: not a modon perturbed by boundaries as it passes, but a modon and one boundary locked in an exchange fast enough to deserve a new name. In substrate terms the criterion \Omega_R \tau > 1 reads: the note sounds before the string damps. The one coupling’s first three regimes were the boundary and the modon trading a quantum and parting; the fourth regime is the trade run so fast it becomes a standing bond — and the two polariton branches are the bonding and antibonding orbitals of a molecule whose two atoms are a photon and a piece of matter.
Mass from One Parent, Collisions from the Other
What makes the lower polariton more than a spectroscopic curiosity is its inheritance. From the photon parent it takes its dispersion: the cavity’s curvature hands it an effective mass four to five decades below the electron’s — within an order of magnitude of the Bonn cavity photon’s 6.7\times10^{-36} kg, and eight to nine decades below any atom that has ever been condensed. From the exciton parent it takes what the Bonn photon conspicuously lacked: collisions. Two polaritons meeting scatter through their exciton cores — a genuine interparticle interaction, light that pushes back on light. Recall what the photon BEC had to outsource: unable to thermalize on its own, the photon gas borrowed a dye reservoir as its thermometer. The polariton gas carries its thermalizer onboard. Mass and interactions — the two atom-like properties Einstein’s mechanism needs — acquired in one move, by marriage rather than by apparatus.
The consequence arrived in 2006. Kasprzak and colleagues pumped a CdTe microcavity at 5 K — incoherently, far above the polariton branches, no phase injected anywhere — and watched the polariton gas relax, thermalize, and pile into its ground state past a critical density (Nature 443, 409): a Bose–Einstein distribution above the spike, macroscopic occupation in it, long-range first-order coherence across the spot, and a spontaneously chosen polarization — the U(1) choice, photographed in Stokes parameters. Trapped condensates followed within a year (Balili et al., Science 316, 1007, 2007). And because critical temperature scales inversely with mass, the story refused to stay cryogenic: polariton lasing in GaN at room temperature (Christopoulos et al., PRL 98, 126405, 2007), then in an organic film (Kéna-Cohen & Forrest, Nature Photonics 4, 371, 2010), then full room-temperature condensation in a polymer microcavity (Plumhof et al., Nature Materials 13, 247, 2014), where the exciton binding and Rabi splitting are so large — hundreds of meV — that the dressing survives summer. Dressed light condenses on a lab bench at the temperature of this room. The occupation ledger the photon-BEC chapter opened gets its missing middle entry: between the pumped laser and the equilibrium photon gas sits a quantum that condenses because it interacts, exactly the way matter does.
A Fluid of Light
This website is called Light Fluid. The review that consolidated this field is called Quantum Fluids of Light (Carusotto & Ciuti, Rev. Mod. Phys. 85, 299, 2013). The convergence is not decorative — the polariton condensate is the laboratory object for which the paper’s title phrase is a literal description, and the decade after 2006 spent itself demonstrating, one hallmark at a time, that a gas which is half photon does everything this paper claims a light-carrying condensate can do.
Interactions reshape light’s own dispersion first. A condensed polariton fluid’s excitation spectrum bends from the single-particle parabola into the Bogoliubov form — linear at low momentum, sound in a gas of light (Utsunomiya et al., Nature Physics 4, 700, 2008). Where there is a sound speed there is a Landau criterion, and in 2009 the Landau test ran: a polariton fluid flowing past a structural defect below its critical speed closes its streamlines and stops scattering — the Rayleigh halo the defect throws at high speed extinguishes, and light flows around an obstacle without friction (Amo et al., Nature Physics 5, 805, 2009); pushed supersonic, the same fluid throws the Cherenkov cone a wake geometry demands. And the fluid’s circulation does what a condensate’s must: it quantizes — vortices with 2\pi phase winding, pinned in the cavity’s disorder landscape, imaged interferometrically (Lagoudakis et al., Nature Physics 4, 706, 2008). Condensation, phase rigidity, Bogoliubov sound, a critical velocity, quantized vorticity: the full superfluid credential sheet, held by an object that is half photon by constitution. A skeptic who accepts each bench result and still balks at “the vacuum is a quantum fluid that carries light” is no longer objecting that the phrase names an impossible kind of thing; the bench keeps one in a cryostat. The objection has been pushed back to scale and venue — where this paper wants every objection pushed.
The Middle of the Dial, and the Mode That Keeps Score
Everything above was earned against a handicap the photon BEC never faced, and the handicap is the polariton’s real gift to this paper. A polariton lives picoseconds: its photon half leaks through the mirror, so the condensate is not a sealed equilibrium gas but a driven-dissipative steady state — a reservoir pumped from above, losses balanced by replenishment, coherence maintained in a system that is dying and being reborn continuously. This is the Kirton–Keeling dial’s middle made flesh, and the photon-BEC chapter’s question — what does “equilibrium enough” mean? — gets a two-part laboratory answer here.
Part one: coherence is more robust than equilibrium. Condensation, vortices, superfluid flow, algebraic long-range order — all of it survives the leak, provided thermalization stays in the race; and as cavity lifetimes stretched from picoseconds toward nanoseconds the same gas walked measurably toward the textbook: equilibrium Bose–Einstein distributions with a genuine chemical potential (Sun et al., PRL 118, 016602, 2017), Berezinskii–Kosterlitz–Thouless order in place (Caputo et al., Nature Materials 17, 145, 2018). The dial is not a wall but a road, and the polariton platform has driven it in both directions.
Part two — and this is the part the framework values most: the leak cannot be hidden, because it keeps its own ledger, and the ledger is the Goldstone mode. In a closed condensate the phase mode propagates: \omega = c_s k, sound, down to k \to 0. In a driven-dissipative condensate, theory said (Szymańska, Keeling & Littlewood, PRL 96, 230602, 2006; Wouters & Carusotto, PRL 99, 140402, 2007), the long-wavelength phase mode goes diffusive: below a wavevector set by the loss rate against the sound speed, the Goldstone branch flattens — a phase twist at long wavelength does not travel, it relaxes, because the pump re-fixes the local density before the twist can push on it. In 2025 that fingerprint was measured directly in a polariton fluid — the flat, non-propagating plateau at small k, closing as the symmetry is explicitly broken (Claude et al., Nature Physics, 2025). Hold the result in one sentence: a condensate that leaks cannot carry a wave at long wavelength.
Now point the sentence at the sky. This framework’s deepest structural commitment puts light in the vacuum’s phase sector — the modon rides the substrate’s coherence, and the two-Goldstone architecture assigns electromagnetism to precisely the branch that the leak, if there were one, would flatten. If the vacuum’s condensate were driven-dissipative at any rate \Gamma — pumped from outside, losing quanta to anywhere — then below some frequency of order \Gamma light would stop propagating and start relaxing: a diffusive window at the bottom of the electromagnetic spectrum, radio that arrives as a smear or not at all. The sky shows no such window. Pulsar radio rides coherently across kiloparsecs down to the $$10 MHz band — dispersion fully accounted by interstellar plasma — which alone pins any diffusive crossover below \sim4\times10^{-8} eV, five decades under the modon floor, with the decades below that screened from view only by the interstellar medium’s own kHz plasma cutoff. In dial language: the polariton sits mid-dial and its Goldstone mode says so; the vacuum’s Goldstone mode is ballistic in every octave anyone has ever checked, and that is a measurement of the dial’s needle — the vacuum sits at the closed end. This is the dynamical face of the flickering-condensate argument: there, no reservoir to flicker with, read in number statistics; here, no pump to lean on, read in dispersion. Same closed-mode claim, second independent ledger.
Condense into Darkness, Then Crystallize
The chapter’s last two exhibits come from a single platform in Lecce, and each lands on a different chapter of this paper.
First, the condensate chose the dark mode. A photonic-crystal waveguide supports bound states in the continuum — modes that sit inside the radiation continuum, free by energy accounting to radiate, and forbidden to by symmetry: their internal structure cancels their own coupling to every outgoing wave. In 2022 the Lecce group condensed polaritons into one (Ardizzone et al., Nature 605, 447): offered a spectrum of leaky modes and one symmetry-protected dark one, the driven gas piled its macroscopic occupation into the mode that cannot couple out — because the Kirton–Keeling race is easiest to win where the loss term is engineered to zero. Set that beside the superradiance chapter’s central claim: the vacuum is dark because the post-burst medium occupies the subradiant configuration — the arrangement whose emission amplitudes cancel. The bench now shows the selection happening: condensation seeks darkness, generically, because darkness is where a condensate keeps its quanta. The vacuum is the limit case of a preference the laboratory has caught in the act.
Then the dark condensate crystallized. Pump the bound-state-in-the-continuum condensate harder and it spontaneously breaks translational symmetry: satellite condensates lock on at finite \pm k, a density modulation appears with no template at its wavelength, and phase coherence holds across the modulated state — a supersolid whose medium is mostly photon (Trypogeorgos et al., Nature 639, 337, 2025). The supersolid chapter waited on this result for one specific reason, and it pays out double. First, constitution: the dipolar supersolid proved matter can hold both broken symmetries at once; this one proves the conjunction survives when the medium is predominantly light — and it survives drive and dissipation too, both broken symmetries standing in a leaking, pumped system. Second, and better: depth. The light supersolid’s density modulation is faint — resolved at a few parts in a thousand — so by Leggett’s bound its superfluid fraction is essentially untouched: a crystal acquired nearly for free. Line the three supersolids up on the one axis the register argument cares about, modulation depth: the dipolar droplet array crystallizes deep — order-unity density contrast, superfluid fraction down to tens of percent; the polariton supersolid crystallizes faint — 10^{-3}, coherence whole; and the vacuum, on this paper’s account, takes the sequence to its endpoint — density modulation exactly nil, the crystal moved wholly into the circulation register, where the sky cannot Bragg off it. The supersolid of light is the rung between the droplets and the stealth vacuum: one step from invisible, built out of the same stuff the vacuum carries.
Light Was Never Bare
Now collect the chapter into the identification it has been circling. Hopfield’s 1958 point was never about microcavities: in any medium, the true propagating quantum is not the bare photon plus corrections — it is the photon–polarization mixture, the polariton, all the way down; the bare photon is the fiction you start the calculation from, not the thing that travels. Every “photon” that has ever crossed a pane of glass was a polariton while it did. The framework’s move is to delete the exemption clause for vacuum. The modon is circulation of the substrate — the photon of empty space is the substrate’s own polariton, dressed so uniformly, so isotropically, and so losslessly that the dressing reads as nothing at all. What the vacuum hides by perfection, the microcavity shows by coarseness: sew the seam with a visible material and every consequence of dressing surfaces at once — a floor and a mass (the polariton’s, engineered; the modon’s, native), interactions and thermalization (the exciton core’s; the substrate quanta’s own), condensability (Kasprzak’s bottle; the boil), and now crystallizability (Lecce; the circulation-register lattice). The polariton is not an analogy for the substrate’s quanta. It is the same construction — light sharing itself with a medium until the two are one particle — executed with parts big enough to see.
The mass ladder says it quantitatively. Dysprosium condenses at tens of nanokelvin at 2.7\times10^{-25} kg; the polariton and the trapped photon condense at 5–300 K at \sim10^{-35}–10^{-36} kg; the substrate’s dc1 quantum, at m_1c^2 \approx 2 meV — 3.6\times10^{-39} kg — is three decades lighter than anything on the bench’s list. Lighter quantum, warmer condensate: the laboratory’s own trend line, extended three decades leftward, meets a medium whose condensation temperature was crossed once, on the way down from the boil, and never re-approached. The lightest quantum on the ladder formed the first condensate and still holds it.
Honest Accounting
Five debts, in the house discipline.
First, nothing here corrects the field. Hopfield’s construction, the Weisbuch splitting, the Kasprzak condensate, the quantum-fluids-of-light program, the driven-dissipative Goldstone theory and its 2025 measurement, and the Lecce dark-mode and supersolid results are standard, quantitatively complete physics; every number above belongs to those groups. The contribution is identification: strong coupling as the one coupling’s fourth regime, the diffusive Goldstone mode as the leak’s ledger and its absence in the sky as a dial reading, condensation-into-darkness as the superradiance argument caught in the act, and the faint-modulation supersolid as the rung nearest stealth.
Second, no number transfers. Rabi splittings are meV set by oscillator strengths, lifetimes are picoseconds set by mirror quality, modulation periods are microns set by photonic-crystal pitch. One near-coincidence must be explicitly declined: semiconductor Rabi splittings ($$10–30 meV) sit in the numerical neighborhood of the modon floor’s 13 meV, and the proximity is an accident of exciton binding scales in III–VI semiconductors — the framework claims nothing from it.
Third, the bench borrows its matter; the vacuum is claimed to be its own. Every laboratory polariton is light dressed by a foreign material — a quantum well, a dye, a polymer — while the framework’s photon is dressed by the very medium it is an excitation of: self-dressing, with no seam between the rider and the ridden. No bench object exhibits that, and the extrapolation from “dressed by matter” to “dressed by the vacuum” is structural, not demonstrated. What the bench certifies is the consequence list of dressing, not the vacuum’s membership on it.
Fourth, the Goldstone discriminator is framework-conditional. Prediction 1’s sky half assumes light rides the vacuum condensate’s phase sector — the photon-modon identification plus the two-sector assignment. A reader who rejects that architecture loses the discriminator, not merely the chapter. And its empirical reach is bounded from below: the interstellar plasma’s kHz cutoff screens the lowest decades of the spectrum, so “no diffusive window” is verified above \sim10^{-8} eV and extrapolated beneath.
Fifth, the mix is not derived. The framework has not computed the modon’s Hopfield angle — what fraction of the vacuum photon’s energy rides the boundary versus the bulk — from substrate parameters. The temptation to read the two-fluid share \alpha_{mf}=\tan^2\theta_W as that coefficient is real and is here declined: \alpha_{mf} lives in the mutual-friction ledger, and equating the two without a calculation would be numerology. Deriving the vacuum’s Hopfield angle is a well-posed piece of homework the polariton literature now shows how to phrase, and it joins the queue in Open Problems.
Place in the Framework
The Light section’s ledger closes its last column. The modon floor read the lattice as a frequency; the scattering ring as a wavevector; Spectrum-Free Light read the emission mechanism; the laser read phase at one boundary, superradiance at N; the photon BEC opened occupation; laser cooling sent the modon out as a courier and came home with the substrate’s portrait. This chapter reads the column under all of them — constitution: what light is, examined closely enough that the medium shows through. The bench’s answer is that light examined closely is never alone — give its togetherness with matter a fast enough exchange and it acquires mass, collisions, a condensate, a superfluid’s credentials, a taste for dark modes, and lately a crystal it chose itself, held at a modulation one step from invisible. The paper’s answer has been on the title page the whole time. The framework’s vacuum is a quantum fluid of light — the phrase the field chose for its own review — and the Kirton–Keeling dial now has every stop occupied: the laser, driven; the polariton, leaking and thriving; the photon BEC, at equilibrium; and at the closed end, reading ballistic in every octave of its Goldstone mode, the oldest fluid of light there is, carrying the others’ photons to the instruments that measure them.