Sunyaev-Zeldovich Effect

A shadow on the oldest light in the universe

The cosmic microwave background is a near-perfect blackbody at 2.725 K filling all of space — the relic glow of the hot early universe, last scattered when the cosmos was 380,000 years old. On its way to us across billions of light-years, a tiny fraction of those photons happen to pass through a galaxy cluster: the most massive gravitationally bound structures in existence, glued together by gas at a hundred million degrees. When a CMB photon crosses that gas, there is a small but non-zero chance it collides with a fast-moving electron and gets kicked to a slightly higher energy. Multiply that nudge over the entire population of CMB photons crossing the cluster and you get a measurable, frequency-dependent dent in the background — colder than the average sky below 217 GHz, hotter than average above it. That dent is the Sunyaev-Zeldovich effect.

It is one of the most useful tools in observational cosmology, for one strange reason. Almost everything we observe in the distant universe gets fainter with distance: light spreads out, surface brightness dims as (1+z)⁻⁴, faint things at high redshift slip below detection. The SZ effect does not. Because it is a fractional distortion of a background that fills the whole sky, the cluster's signal is the same surface brightness whether it sits next door or halfway across the observable universe. Point a millimetre-wave telescope at the sky and the most massive clusters announce themselves as cold spots, at any redshift, with essentially the same contrast. That redshift independence is what turned SZ surveys into cluster-finding machines.

How it works: inverse-Compton scattering

Ordinary Compton scattering is what happens when an energetic photon hits a slow electron and gives the electron some of its energy — the photon comes out softer. Inverse-Compton scattering is the mirror image: a low-energy photon hits a fast electron and the energy flows the other way, from electron to photon. The CMB photons are cold — typical energy a fraction of a milli-electron-volt — while the cluster's electrons are blazing hot, with thermal energies of several keV. So when a 2.7 K CMB photon scatters off a 10⁸ K electron, it almost always comes out with a tiny bit more energy than it went in with.

Crucially, scattering does not create or destroy photons — it just redistributes their energies. The total number of CMB photons crossing the cluster is conserved; what changes is that each one is, on average, boosted slightly upward in frequency. This shifts the entire blackbody spectrum a hair toward higher frequencies. Photons get robbed from the low-frequency Rayleigh-Jeans tail and deposited into the high-frequency Wien tail. So an observer sees:

• A decrement at low frequencies (radio / long-millimetre): fewer photons than the unperturbed CMB, so the cluster looks colder than the average sky.
• An increment at high frequencies (sub-millimetre): more photons, so the cluster looks hotter.
• A null at the crossover, near 217 GHz, where the two effects cancel and the cluster is invisible in temperature.

This distinctive spectral signature — a decrement, a null at a fixed frequency, and an increment — is the fingerprint that lets us separate the SZ effect from every other source of millimetre-wave emission. No other astrophysical signal has a guaranteed sign flip at exactly 217 GHz.

The Compton y-parameter

The amplitude of the thermal SZ effect is captured by a single dimensionless number, the Compton y-parameter:

y = ∫ n_e σ_T (k_B T_e / m_e c²) dl

Read it from the inside out. The factor k_B T_e / m_e c² is the average fractional energy a photon gains per scattering: it compares the electron's thermal energy to its rest energy m_e c² = 511 keV. For a 5 keV cluster that ratio is about 0.01. The Thomson cross-section σ_T = 6.65 × 10⁻²⁵ cm² is the probability of a scattering. Multiplying by the electron number density n_e and integrating dl along the line of sight through the cluster counts how many scatterings a photon undergoes on its way through. So y is, in effect, the average number of scatterings times the average fractional energy gain per scattering — the line-of-sight integral of the electron pressure n_e k_B T_e.

In the low-frequency Rayleigh-Jeans limit, the fractional brightness-temperature distortion is beautifully simple:

ΔT/T ≈ -2y          (Rayleigh-Jeans, h𝜈 ≪ k_B T_CMB)

The minus sign is the decrement: at low frequencies the cluster is colder than the background by twice the y-parameter. Across the full spectrum the distortion is ΔT/T = y · g(x), where x = h𝜈 / k_B T_CMB and g(x) = x coth(x/2) − 4, which equals −2 in the Rayleigh-Jeans limit (x → 0), passes through zero at x ≈ 3.83 (𝜈 ≈ 217 GHz), and becomes positive at higher frequencies. The y-parameter sets the size; the function g(x) sets the shape.

Worked example: a massive cluster

Take a rich cluster like Coma. The intracluster gas near the center has an electron density of order n_e ≈ 3 × 10⁻³ cm⁻³ and a temperature k_B T_e ≈ 8 keV, spread over a path length through the core of about L ≈ 1 Mpc ≈ 3 × 10²⁴ cm. Estimate y by treating the integral as a simple product:

y ≈ n_e σ_T (k_B T_e / m_e c²) L

  = (3×10⁻³ cm⁻³) × (6.65×10⁻²⁵ cm²) × (8 keV / 511 keV) × (3×10²⁴ cm)

  = (3×10⁻³) × (6.65×10⁻²⁵) × (0.0157) × (3×10²⁴)

  ≈ 9 × 10⁻⁵

So y ≈ 10⁻⁴ for a massive cluster — a value confirmed by real measurements of Coma. Now turn that into a temperature dip in the Rayleigh-Jeans regime:

ΔT/T ≈ -2y ≈ -1.8 × 10⁻⁴

ΔT ≈ -1.8 × 10⁻⁴ × 2.725 K ≈ -490 μK

The cluster dims the CMB by about half a milli-kelvin — roughly 0.5 mK — at low frequencies. That sounds tiny, and it is: it is comparable to the primary CMB anisotropies and far smaller than the 2.725 K mean. But it is a stable, predictable signature with a known spectral shape, and it does not weaken with distance. A cluster of the same gas pressure at z = 1 produces the same ≈ −490 μK decrement. Compare that to the cluster's X-ray surface brightness, which scales with n_e² and dims by (1+z)⁻⁴ = 16× by z = 1, vanishing into the noise. The SZ shadow stays put.

Why distance does not dim it

The redshift independence deserves a sharper statement because it is the whole reason the effect matters for surveys. Most extragalactic flux measurements fight cosmological surface-brightness dimming: a source of fixed physical surface brightness appears fainter by (1+z)⁻⁴ as you push it to higher redshift, the combination of energy redshift, time dilation, and solid-angle effects. This is why high-z galaxies and the X-ray emission of distant clusters fade so fast.

The SZ effect sidesteps this entirely because the cluster's signal is not its own emission — it is a fractional modulation of the CMB, which itself fills the sky and dims by exactly the same (1+z)⁻⁴. The CMB temperature at the cluster's epoch was hotter, T_CMB(z) = 2.725 (1+z) K, and the up-scattered photons redshift back down by the same factor on their way to us. The ratio ΔT/T survives untouched, set only by the integrated pressure y. So the observed SZ surface brightness of a cluster of given mass and gas pressure is the same at z = 0.1 and z = 1.5. A millimetre survey of fixed sensitivity therefore selects clusters in a way that is nearly mass-limited rather than distance-limited — the holy grail for counting the most massive halos across cosmic time.

Variants and regimes

The "SZ effect" is really a family of related distortions:

• Thermal SZ (tSZ). The main effect, driven by the random thermal motion of hot electrons. It has the characteristic decrement/increment with the 217 GHz null and amplitude set by y. This is what cluster surveys exploit.
• Kinematic SZ (kSZ). A separate, smaller distortion from the cluster's bulk peculiar velocity v_pec relative to the CMB rest frame. It is a pure Doppler shift that preserves the blackbody shape, so it has no spectral null — the same sign at all frequencies — with amplitude ΔT/T ≈ −(v_pec/c) τ, where τ = ∫ n_e σ_T dl is the optical depth. For typical v_pec ~ 300 km/s and τ ~ 10⁻³, kSZ is roughly an order of magnitude below tSZ, but it directly measures velocities and probes the missing-baryon problem and large-scale flows.
• Relativistic corrections (rSZ). In the hottest clusters (k_B T_e ≳ 10 keV) the electrons are mildly relativistic, and the simple non-relativistic formula picks up percent-level corrections that shift the null frequency slightly above 217 GHz and let multi-frequency data measure the gas temperature directly from the SZ spectrum.
• Polarized SZ. A much fainter effect from the quadrupole of the CMB as seen by the scattering electrons, of interest for future high-sensitivity experiments.

Observational status

SZ astronomy moved from heroic single detections to industrial cluster surveys within a few decades. Early single-dish attempts on the Coma cluster and Abell 2218 in the 1970s–80s were difficult and disputed at the microkelvin level. Interferometers at the Ryle Telescope and OVRO/BIMA produced clean imaging detections in the 1990s. The transformation came with dedicated wide-field millimetre experiments:

• Planck (2009–2013) mapped the whole sky and delivered a catalog of more than a thousand SZ clusters plus an all-sky map of the integrated y-parameter (the tSZ power spectrum), a primary probe of σ8.
• The South Pole Telescope (SPT) and the Atacama Cosmology Telescope (ACT) surveyed thousands of square degrees at arcminute resolution, producing SZ-selected catalogs of thousands of clusters out to z > 1.5 — the highest-redshift massive clusters known were largely found this way.
• The integrated SZ signal Y_SZ (the y-parameter integrated over the cluster's solid angle) is the cleanest available proxy for total cluster mass, with remarkably low scatter, anchoring cluster-cosmology constraints on Ω_m and σ8.

A persistent puzzle is the mild tension between σ8 inferred from SZ cluster counts and the value from the primary CMB power spectrum. The discrepancy hinges on cluster mass calibration — converting Y_SZ into mass requires hydrostatic or weak-lensing masses, and a systematic "hydrostatic mass bias" of ~20–30% can shift the inferred σ8 enough to reconcile or worsen the tension. Future kSZ measurements and lensing-calibrated masses aim to close this.

SZ versus other cluster probes

Probe	Physical observable	Scales with	Redshift dimming	Best for
Thermal SZ	Integrated electron pressure	∫ n_e T_e dl (y)	None — redshift-independent	Finding & weighing distant clusters
Kinematic SZ	Bulk peculiar velocity × τ	(v_pec/c) ∫ n_e dl	None	Velocities, missing baryons
X-ray surface brightness	Bremsstrahlung emission	∫ n_e² Λ(T) dl	(1+z)⁻⁴ — fades fast	Nearby clusters, gas density
Optical richness	Count of member galaxies	Number of galaxies	Dimming + projection	Cheap large samples, low z
Weak gravitational lensing	Total projected mass	Σ(R) shear	Source-density limited	Direct mass, no gas assumptions
Galaxy velocity dispersion	Member kinematics	σ_v of galaxies	Spectroscopy-limited	Dynamical mass, low z

The standout column is "redshift dimming." X-ray brightness goes as n_e² and dims by 16× at z = 1; the SZ effect goes as the first power of n_e (through pressure) and does not dim at all. That single contrast is why SZ surveys reach the highest-redshift clusters and why X-ray follow-up is reserved for the nearer, brighter ones.

Quantitative analysis: from scattering to spectrum

The spectral distortion of the CMB by a population of hot electrons is governed by the Kompaneets equation, which describes how Compton scattering redistributes photon energies in a thermal bath. In the limit of small optical depth and non-relativistic electrons, integrating the Kompaneets diffusion over the line of sight gives the SZ intensity change:

ΔI(x) / I_0 = y · (x⁴ eˣ / (eˣ − 1)²) · g(x)

with  x = h𝜈 / k_B T_CMB
      g(x) = x coth(x/2) − 4
      y = ∫ n_e σ_T (k_B T_e / m_e c²) dl

The spectral function g(x) controls the shape. Evaluate its limits:

x → 0   (low 𝜈, Rayleigh-Jeans):   g(x) → -2     →  ΔT/T = -2y   (decrement)
x = 3.83  (𝜈 ≈ 217 GHz):           g(x) = 0      →  null
x → ∞   (high 𝜈, Wien):            g(x) → x - 4  →  ΔT/T > 0    (increment)

The crossover x = 3.83 follows from solving x coth(x/2) = 4, which lands at 𝜈 = x · k_B T_CMB / h ≈ 3.83 × (56.78 GHz) ≈ 217 GHz for T_CMB = 2.725 K. This null is fixed by fundamental constants and the CMB temperature, not by the cluster — which is exactly why multi-frequency observations straddling 217 GHz cleanly isolate the SZ signal from foregrounds and the primary CMB.

One more consistency check on the redshift independence: the optical depth τ = ∫ n_e σ_T dl is dimensionless and depends only on the gas column, while the energy-gain factor k_B T_e / m_e c² depends only on the gas temperature. Neither carries any explicit factor of luminosity distance or (1+z). The y-parameter is a property of the cluster gas alone, so the observed ΔT/T is too. Distance enters only through the angular size of the cluster on the sky — a far gentler dependence than the (1+z)⁻⁴ that kills surface-brightness observables.

Common pitfalls and misconceptions

• "The SZ effect heats the CMB photons because the gas is hot." The gas being hot matters, but the mechanism is inverse-Compton scattering off individual fast electrons, not bulk thermal emission. The cluster does not add its own photons; it redistributes the CMB photons that pass through, conserving their total number.
• Confusing the decrement with absorption. The low-frequency dip is not the gas absorbing light. It is photons being scattered up in frequency, leaving a deficit at low frequencies and a surplus at high frequencies. Total photon number is unchanged.
• Thinking the null at 217 GHz is a property of the cluster. The null frequency is set by the CMB temperature and fundamental constants, identical for every (non-relativistic) cluster. Only relativistic corrections in the hottest clusters shift it slightly.
• Assuming SZ flux measures gas mass directly. y integrates pressure n_e T_e, not density alone. Two clusters with the same gas mass but different temperatures have different y. The integrated Y_SZ correlates with total mass, but through the pressure–mass relation, which must be calibrated (hence the hydrostatic-mass-bias debate).
• Treating tSZ and kSZ as the same thing. They have different physical origins, different spectral shapes (kSZ has no null), and different amplitudes. Disentangling them requires multi-frequency data.
• Expecting the signal to fade with distance like everything else. It does not. The whole point is that ΔT/T is preserved as the CMB and the scattered photons redshift together. This is counter-intuitive precisely because every other extragalactic observable dims.