Isocurvature vs Adiabatic Perturbations: Two Ways to Seed Cosmic Structure

What the two perturbation types physically are

Before galaxies could form, the universe needed tiny departures from perfect uniformity — regions slightly denser than average that gravity could later amplify. There are exactly two independent ways to write down such a departure at the earliest times, and they are distinguished by what they perturb.

An adiabatic perturbation compresses or rarefies everything together. If a patch has 1% more photons, it also has 1% more baryons, 1% more dark matter, and 1% more neutrinos — the number-density ratios n_i/n_j stay fixed. Because the total energy density changes, so does the local spatial curvature; adiabatic modes are therefore also called curvature perturbations, quantified by the gauge-invariant variable ζ (or R).

An isocurvature perturbation (also "entropy perturbation") does the opposite: it shuffles species relative to one another so that the total density — and hence the curvature — stays flat. A patch might have more dark matter compensated by fewer photons, keeping δρ_total = 0. The relative entropy per species, S_i = δ(n_i/s)/(n_i/s), is nonzero. These are genuinely distinct initial conditions, not two descriptions of the same thing.

The mechanism: how each seeds sound waves differently

Both perturbation types eventually drive acoustic oscillations in the tightly coupled photon–baryon plasma before recombination, but they set the clock at opposite phases — this is the crux of the observational distinction.

• Adiabatic: an overdense patch already has excess photon pressure and sits in a gravitational potential well from the very start. Compression and gravity work in concert, so every Fourier mode begins at maximum displacement. The temperature oscillation follows a cosine: δT/T ∝ cos(k·c_s·η), where c_s ≈ c/√3 is the plasma sound speed and η is conformal time.
• Isocurvature: initially there is no curvature and no density contrast, so the fluid starts at rest with zero displacement. Pressure gradients only develop as the mode evolves, driving a sine oscillation: δT/T ∝ sin(k·c_s·η).

Sine and cosine are 90° out of phase. Because the CMB acoustic peaks fall at extrema of the oscillation at recombination, the peak positions shift. The purely adiabatic first peak sits at ℓ ≈ 220; a pure cold-dark-matter isocurvature spectrum pushes it to ℓ ≈ 330 and reshuffles the peak spacing.

Key quantities and a worked comparison

The strength of an isocurvature admixture is parameterized by the isocurvature fraction β_iso = P_iso / (P_adi + P_iso), the ratio of isocurvature to total primordial power at a reference scale (Planck uses k = 0.05 Mpc⁻¹, with results also quoted at k = 0.002 and 0.1 Mpc⁻¹).

• Peak phase shift: pure CDM isocurvature moves the first acoustic peak from ℓ ≈ 220 to ℓ ≈ 330 — a smoking-gun ~50% shift that the data rule out.
• Planck 2018 (TT,TE,EE+lowE+lensing): β_iso < 2.5% for cold-dark-matter density isocurvature (CDI), < 7.4% for neutrino density (NDI), and < 6.8% for neutrino velocity (NVI) isocurvature, all at 95% confidence.
• Non-adiabatic variance: in mixed correlated models the non-adiabatic contribution to the observed CMB temperature variance is bounded to roughly 1.3–1.7% (95% CL).

So the universe is adiabatic to better than a few percent — a remarkably clean result. For scale, the total primordial curvature power is A_s ≈ 2.1 × 10⁻⁹ at k = 0.05 Mpc⁻¹, with spectral tilt n_s ≈ 0.965; any isocurvature component must hide beneath ~2.5% of that.

How it is observed and detected

The decisive laboratory is the cosmic microwave background. Three fingerprints separate the modes:

• Peak positions and spacing in the temperature power spectrum — the phase shift described above. The first peak's firm location at ℓ ≈ 220 (measured by BOOMERANG in 2000, then WMAP and Planck) killed any dominant isocurvature model on its own.
• Temperature–polarization (TE) cross-correlation. Because adiabatic and isocurvature modes correlate temperature and E-mode polarization with opposite phase, the TE spectrum is an exquisitely sensitive discriminator — this is why Planck's polarization data tightened the limits so much.
• Large-scale (low-ℓ) plateau. Isocurvature modes boost the Sachs–Wolfe plateau by a factor of ~6 relative to adiabatic for equal metric perturbations, so they overproduce large-angle power.

The observational program ran from COBE (1992, first detection of anisotropies) through WMAP (2003–2013) to Planck (2013, 2015, 2018). Future targets include CMB-S4, LiteBIRD, and 21-cm and spectral-distortion probes (PIXIE-class) that could reach smaller scales where isocurvature might still lurk.

How it compares to related regimes and cousins

The adiabatic mode is not alone — isocurvature is really a family of modes, one per independent species pair. In the standard four-component universe there are four regular isocurvature modes plus the single adiabatic mode:

• CDI — cold dark matter density isocurvature (the most-studied, arises naturally in axion dark-matter and curvaton scenarios).
• BDI — baryon density isocurvature (observationally degenerate with CDI, since both perturb matter vs radiation).
• NDI — neutrino density isocurvature.
• NVI — neutrino velocity isocurvature (a relative-velocity mode with no density analog).

These differ from tensor perturbations (primordial gravitational waves, parameterized by the tensor-to-scalar ratio r), which are a separate spin-2 sector entirely. They also differ from the late-time, nonlinear density contrast δ = δρ/ρ that grows under gravity — perturbation type refers strictly to the initial conditions set during or before inflation, before any mode re-enters the horizon.

Significance, famous cases, and open questions

The near-perfect adiabaticity of the primordial spectrum is one of the strongest pieces of evidence for single-field inflation. When a single scalar field drives inflation, there is only one "clock," so every species inherits the same fluctuation and the result is automatically adiabatic. Detecting any isocurvature would immediately require additional physics: a second light field, a curvaton, or an axion.

• Axion dark matter is the classic case. If the QCD axion exists and inflation happened at high scale, quantum fluctuations of the axion field imprint a CDI signature. The tight β_iso limits therefore constrain the inflationary energy scale in axion models — a live link between cosmology and particle physics.
• Curvaton scenarios can produce correlated mixtures of adiabatic and isocurvature power, motivating the "correlated" fits Planck reports.

Open questions: Is there a small isocurvature component just below current sensitivity? Could scale-dependent or blue-tilted isocurvature evade large-scale limits while showing up in small-scale probes like spectral distortions or primordial black holes? And can next-generation CMB and 21-cm surveys push β_iso into the sub-percent regime, where realistic multi-field models start to live? For now, the verdict stands: the cosmos was seeded overwhelmingly adiabatically.

Adiabatic vs isocurvature perturbations: defining properties and observational signatures

Property	Adiabatic perturbation	Isocurvature (entropy) perturbation
What varies	Total energy density ρ (curvature ζ ≠ 0)	Species ratios n_i/n_j; total ρ ≈ 0 initially
Species relation	δ(n_i/n_j) = 0 for all species	δ(n_i/s) ≠ 0; e.g. δn_cdm balanced by δn_γ
Sound-wave phase	Cosine mode (starts at max compression)	Sine mode (starts at zero displacement)
First CMB peak	ℓ ≈ 220	ℓ ≈ 330 (peaks shifted, ~90° out of phase)
Physical origin	Single-field inflation (one clock)	Extra light fields (axion, curvaton, multi-field)
Observed amplitude	Dominant: ~100% of primordial power	Subdominant: β_iso < 2.5% (Planck, CDM mode)