Critical Opalescence: The Moment a Clear Fluid Turns to Milk

What Actually Happens Inside the Tube

The experiment is deceptively simple. A thick-walled fused-silica tube is charged with carbon dioxide at exactly the critical density, 467.6 kg/m³, and sealed. Get the fill wrong and the demonstration fails: too heavy and the meniscus rises until liquid floods the tube; too light and it sinks until the last liquid evaporates. Only at the critical density does the interface hold its place while the fluid is heated, because liquid and vapour are then forced to converge on each other rather than one consuming the other.

Warm the tube slowly toward 30.98 °C. The internal pressure climbs the saturation curve to 7.38 MPa — 73.8 bar, 72.8 atmospheres — which is why the glass is thick. Three things then happen within the last degree, all from the same root cause:

• The meniscus flattens. Surface tension collapses, so the curved edge where the liquid climbs the wall straightens out.
• The interface starts to shimmer and ripple. Thermally excited capillary waves that are normally sub-nanometre become visible undulations, because the restoring force holding the surface flat has almost gone.
• The fluid goes milk-white, first with a bluish tint in scattered light and an orange cast in transmission, then a dense uniform white that hides the far wall of the tube.

Within a few millikelvin of T_c, the meniscus does not break or boil — it simply fades out, and the tube is left holding one homogeneous supercritical fluid. Cool back down and it all runs in reverse: the tube clears, a faint horizon reappears, and the meniscus condenses back into existence.

The Meniscus Dies: Order Parameter and Vanishing Surface Tension

Liquid and vapour are distinguishable only because their densities differ. That difference, Δρ = ρ_L − ρ_V, is the order parameter of the liquid–gas transition, exactly analogous to the magnetisation of a uniaxial ferromagnet. Approaching the critical point it closes as a power law in the reduced temperature t = (T − T_c)/T_c:

ρ_L − ρ_V ∝ |t|^β, with β ≈ 0.326

Van der Waals' 1873 mean-field theory predicted β = 1/2. In 1900 Jules-Émile Verschaffelt, in Kamerlingh Onnes' Leiden laboratory, measured about 0.35 for isopentane and said so; the discrepancy was ignored for half a century. Edward Guggenheim's 1945 corresponding-states plot collapsed eight fluids — neon, argon, krypton, xenon, nitrogen, oxygen, carbon monoxide, methane — onto a single curve far better fit by an exponent near 1/3. Modern conformal-bootstrap calculations of the 3D Ising model give β = 0.326419, and precision fluid data agree.

Surface tension dies with a different exponent. A liquid–vapour interface costs free energy because it disrupts correlations over a length ξ; dimensional analysis (hyperscaling) gives an interfacial free energy of roughly k_BT per area ξ², so

σ ≈ k_BT/ξ² ∝ |t|^2ν ≈ |t|^1.26

For CO₂, σ is modest even far from T_c — about 15 mN/m at −50 °C, 4.6 mN/m at 0 °C, barely 1.2 mN/m at 20 °C — and it falls under 1 µN/m within 10 mK of T_c, roughly 10⁵ times weaker than a soap film. The interface can no longer pay to stay flat against thermal kicks, its capillary waves grow to micron amplitude, and it dissolves into the fluctuating bulk. The latent heat vanishes with it (via Clapeyron, L ∝ Δρ roughly, so L → 0 as |t|^β): at the critical point the first-order transition has become second-order, which is precisely why nothing boils.

Why It Turns White: Compressibility, Correlation Length, and Light

The whiteness is a scattering effect, and it follows from a single divergence. The isothermal compressibility κ_T = −(1/V)(∂V/∂P)_T blows up as

κ_T ∝ |t|^−γ, γ ≈ 1.237

Statistical mechanics ties compressibility directly to density noise: ⟨(ΔN)²⟩/⟨N⟩ = ρk_BTκ_T. A diverging κ_T therefore means it costs almost nothing in free energy to make a region of fluid denser than its surroundings, so the fluid spontaneously fills itself with transient dense and rarefied patches. The patch size is the correlation length:

ξ = ξ₀|t|^−ν, ξ₀ ≈ 0.15 nm, ν ≈ 0.630

Run the numbers for CO₂ (T_c = 304.13 K). Three kelvin out, t ≈ 10⁻² and ξ ≈ 2.8 nm — under ten molecular diameters, far too small for visible light to notice. At 10 mK, ξ ≈ 100 nm; at 2 mK, ≈ 280 nm; at 1 mK, ≈ 430 nm. Over the last hundredth of a degree the blobs grow from molecular size into the visible band, 400–700 nm.

Marian Smoluchowski proposed the fluctuation explanation in 1908; Albert Einstein made it quantitative in 1910, with a scattered intensity proportional to k_BTρ²κ_T(∂ε/∂ρ)²/λ⁴ — the same calculation that explains the blue sky. His formula diverges unphysically at T_c, and Leonard Ornstein and Frits Zernike supplied the fix in 1914: correlations decay as e^−r/ξ/r rather than extending forever, giving the Lorentzian structure factor

S(q) = S(0)/(1 + q²ξ²), with q = (4πn/λ)sin(θ/2)

That crossover produces the colour change. While ξ < λ/2πn ≈ 80 nm — roughly 15 mK out — scattering is Rayleigh-like: weak, nearly isotropic, λ⁻⁴, so the fluid looks faintly blue from the side while the transmitted beam reddens like a sunset. Beyond it qξ > 1, S(q) falls as q^−(2−η) with η ≈ 0.036, and the scattering turns sharply forward-peaked, nearly achromatic and three or four orders of magnitude stronger. The photon mean free path drops below the tube's centimetre width, every photon scatters several times, and all directional and spectral information is scrambled — which is exactly what makes milk, cloud and fog white.

How It Is Measured: Meniscus Watching, Light Beating, and Microgravity

The crude method still works. Watching the meniscus vanish locates T_c to about a millikelvin, and it is how Thomas Andrews, in his 1869 Bakerian Lecture On the Continuity of the Gaseous and Liquid States of Matter, fixed CO₂'s critical temperature at 30.92 °C — within 60 mK of the modern 30.978 °C. Andrews coined the term “critical point” and described the opalescence as “a peculiar appearance of moving or flickering striæ throughout its entire mass.” Charles Cagniard de la Tour had already lost the meniscus in 1822, sealing liquids in tubes with a rolling flint ball and hearing its sound change when the liquid ceased to exist.

Static light scattering measures ξ and γ directly: fit the angular dependence to the Ornstein–Zernike Lorentzian and ξ falls out of the curvature while κ_T falls out of the q → 0 intercept. Turbidity measurements — V. G. Puglielli and Norman Ford's 1970 work on SF₆ is the classic — track the same divergence through the extinction of a transmitted beam, with multiple-scattering corrections becoming the dominant systematic within millikelvin of T_c.

Dynamic light scattering exposes the other half of the story: critical slowing down. Kawasaki mode-coupling and the Hohenberg–Halperin “model H” classification give a decay rate Γ ≈ k_BTq²/(6πηξ), a Stokes–Einstein result in which a blob diffuses like a Brownian sphere of its own size. At the peak of the spectrum q ≈ 1/ξ, so Γ ≈ k_BT/(6πηξ³). Put in ξ = 300 nm, η ≈ 3×10⁻⁵ Pa·s and T = 304 K: the lifetime is ~4 milliseconds, against roughly a picosecond in ordinary fluid — a slowdown of 10⁹, and slow enough that the flickering striae are visible to the naked eye. The same Norman Ford, with George Benedek, saw the corresponding Rayleigh linewidth narrow in SF₆ in 1965; Harry Swinney and Herman Cummins measured the narrowing in CO₂ in 1968.

Gravity is the enemy. Because κ_T diverges, a near-critical fluid cannot hold itself uniform against its own weight: dρ/dz = −ρ²gκ_T. For CO₂ some 30 mK out that is of order a few per cent of ρ_c across a 1 cm cell, so top and bottom pass through the critical density at different times and the sample stops being a single thermodynamic state. Thermal diffusivity collapses too — D_T ≈ 2×10⁻¹¹ m²/s at ξ = 300 nm, weeks to equilibrate a centimetre by conduction, rescued only by the piston effect, the fast adiabatic bulk compression identified in 1990 by Zappoli, Onuki–Ferrell and Boukari's group. Orbit removes the stratification: CNES flew ALICE and ALICE-2 with SF₆ on Mir, NASA flew the Critical Viscosity of Xenon experiments on STS-85 (1997) and STS-107 (2003), and since 2009 the CNES/NASA DECLIC facility on the ISS has run SF₆ and supercritical water with thermostats stable to well under a millikelvin.

Not Boiling, Not Milk, Not Opal

Critical opalescence is routinely mistaken for three different things, and the distinctions are physically sharp.

• It is not boiling. Boiling requires a first-order transition: two coexisting phases, an interface, nucleation, and latent heat. At the critical point the latent heat has gone to zero and the interfacial tension with it. There are no bubbles because there is no surface that could bound one. The fluid never separates; it stays one phase whose density is merely uncertain.
• It is not a Tyndall-scattering emulsion. Milk is white because it holds real, persistent fat globules and casein micelles whose refractive index differs from water's. A near-critical fluid is chemically pure and holds nothing at all: the scatterers are density fluctuations of one substance, with no boundaries and a lifetime of milliseconds. The test is diagnostic — a colloid's turbidity barely cares about temperature, whereas critical opalescence appears and vanishes reversibly over millikelvin, follows an exact power law, and scatters as a clean Ornstein–Zernike Lorentzian in q rather than a Mie resonance pattern.
• It has nothing to do with opal. The name comes from the milky look, but a gem opal's play-of-colour is Bragg diffraction from a periodic lattice of silica spheres — an ordered structure, not a fluctuating one.

A fourth confusion is with spinodal decomposition. Quench below T_c rather than approaching from above and you also get a cloudy sample, but there the cloudiness comes from real phase-separating domains that coarsen with time and never stop growing. Critical opalescence is a steady state: ξ is set by temperature alone and stays put.

Universality, Real-World Consequences, and Open Questions

The most remarkable fact about the numbers above is that they are not properties of carbon dioxide. β ≈ 0.326, γ ≈ 1.237 and ν ≈ 0.630 are the same for xenon, SF₆, water, a binary liquid mixture at its consolute point, and a uniaxial ferromagnet at its Curie temperature. All of them belong to the three-dimensional Ising universality class, defined only by dimensionality, a one-component order parameter and short-range interactions. Benjamin Widom's scaling hypothesis (1965), Leo Kadanoff's block-spin picture (1966) and Kenneth Wilson's renormalisation group (1971, Nobel Prize 1982) explained why: as ξ runs away to infinity the system forgets its microscopic details, and only the fixed point survives. Critical opalescence is the one place where you can watch that forgetting happen with your eyes.

The physics has practical reach. The huge compressibility that makes the fluid opalescent is exactly what makes supercritical CO₂ a tunable solvent: above 31 °C and 74 bar its density — and with it the solvent power — is set by pressure alone, which is the basis of Kurt Zosel's decaffeination process, patented around 1970. Industrially the fluid is run far denser than the critical point itself, typically 40–80 °C and 150–300 bar, to strip caffeine while leaving most flavour compounds in the bean. In biology, the MIT groups of Toyoichi Tanaka and George Benedek showed that the cold cataract — the reversible clouding of a young mammalian lens on cooling, reported by Tanaka, Ishimoto and Chylack in Science in 1977 — is critical opalescence in a crystallin-protein solution nearing its own consolute point: the lens goes opaque for precisely the reason the CO₂ tube does.

Open questions remain. The asymmetry of real fluids — the Yang–Yang anomaly, and the curvature correction to Cailletet and Mathias' 1886 law of rectilinear diameters — is still being pinned down. Near-critical fluctuations generate measurable critical Casimir forces, first observed directly by Hertlein and co-workers in 2008 near a water–lutidine critical point and now a tool for reversible colloidal assembly. The largest prize is nuclear: quantum chromodynamics is predicted to have a critical point in the temperature–baryon-density plane, which would announce itself through non-monotonic fluctuations of conserved charges — a critical opalescence of nuclear matter. The STAR experiment's Beam Energy Scan at RHIC has hunted that signature in net-proton cumulants since 2010, with a tantalising but unresolved structure near √s_NN ≈ 20 GeV.

Fluids used to show critical opalescence, and why each is chosen

Fluid	Critical point (T_c, P_c)	Critical density / composition	Why it is used
Carbon dioxide (CO₂)	30.98 °C, 7.38 MPa	467.6 kg/m³	The textbook case since Andrews (1869); T_c sits just above room temperature, but 74 bar demands a thick-walled tube
Sulfur hexafluoride (SF₆)	45.57 °C, 3.75 MPa	742 kg/m³	Lecture-demo favourite and the ISS working fluid: only ~38 bar, non-toxic, non-flammable, chemically inert
Xenon (Xe)	16.58 °C, 5.84 MPa	1110 kg/m³	Monatomic and simple, so it is the cleanest test of theory; used in the Shuttle-era CVX viscosity experiments
Water (H₂O)	373.95 °C, 22.06 MPa	322 kg/m³	Industrially critical (supercritical boilers, oxidation reactors), but 221 bar and 374 °C make optical work brutal
Isobutyric acid + water	26.1 °C (consolute), 1 atm	38.8 wt% acid	A binary mixture reaches an Ising critical point at ambient pressure — same physics, no pressure vessel