Cosmology
CMB Polarization E-Modes: The Curl-Free Pattern from Thomson Scattering
Across the entire sky, the 2.725 K microwave background is faintly polarized at the level of a few parts in a million — a few microkelvin — and that whisper of polarization carries a hidden geometry. When you decompose it, most of the signal falls into a curl-free component called the E-mode, whose alignment vectors point radially toward or tangentially around temperature hot and cold spots. This pattern is the direct fingerprint of Thomson scattering off free electrons in a flowing, compressed primordial plasma 380,000 years after the Big Bang.
E-mode polarization is not an exotic subtlety — it is a guaranteed, predicted consequence of the same acoustic oscillations that shape the CMB temperature map, and it has been measured to high precision by DASI, WMAP, Planck, and modern ground-based telescopes. It anchors the physics of recombination, pins down the epoch of reionization, and provides the clean "template" against which cosmologists hunt for the far rarer B-modes of primordial gravitational waves.
- RegimeCMB polarization, recombination (z≈1090) & reionization (z≈8)
- Key numberPeak EE amplitude ~few μK (∼1–40 μK²·ℓ(ℓ+1)/2π level, peak ~40 μK²)
- Driven byThomson scattering off a local temperature quadrupole
- First describedPredicted by Rees (1968); E/B decomposition by Seljak, Zaldarriaga, Kamionkowski, Kosowsky, Stebbins (1997)
- Observed withDASI (2002, first detection), WMAP, Planck, ACT, SPT, BICEP/Keck
- Matters forRecombination physics, optical depth τ≈0.054, B-mode gravitational-wave searches
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
What E-modes are and why they matter
The cosmic microwave background is not just a temperature map — it is weakly linearly polarized, at roughly the few-microkelvin level, about ten times smaller than the ∼100 μK temperature anisotropies. Polarization is a headless vector (an orientation, not an arrow) at every point on the sky, so it cannot be split into a simple gradient and curl the way a true vector field can. Instead cosmologists decompose it into two coordinate-independent, parity-distinguished parts: E-modes (curl-free, even parity) and B-modes (divergence-free, odd parity).
E-modes are the dominant, guaranteed component. Their polarization vectors line up either radially or tangentially around temperature spots, producing symmetric "+" and "×" patterns with no handedness. This matters enormously: E-modes are produced by ordinary scalar density perturbations — the same sound waves that make the temperature acoustic peaks — so they cross-correlate with temperature (the TE spectrum) and confirm our recombination-era physics with an independent observable. They also provide the essential, well-understood foil against which the elusive primordial B-mode of inflationary gravitational waves must be distinguished.
The mechanism: Thomson scattering off a local quadrupole
Polarization is generated by Thomson scattering — the elastic scattering of photons off free electrons. The cross-section, σ_T ≈ 6.65×10⁻²⁵ cm², has an angular dependence (∝ 1 + cos²θ) that acts like a polarizer: an electron re-radiates the incoming radiation with a polarization set by the incident intensity pattern. The crucial requirement is that the radiation hitting the electron must have a quadrupole anisotropy. Monopole (uniform) and dipole (Doppler) patterns scatter into no net polarization; only the ℓ=2 component leaves a linear polarization aligned with the cold axis of the quadrupole.
In the pre-recombination plasma the photon mean free path is tiny, so photons are tightly coupled and see an almost isotropic bath — no quadrupole, no polarization. Only as recombination proceeds and the mean free path grows does a quadrupole develop, sourced by the velocity gradients of the acoustic (sound) waves in the photon-baryon fluid. Because scalar perturbations produce a quadrupole with a specific symmetric geometry, the resulting polarization is curl-free — an E-mode. The signal is thus intrinsically tied to the fluid velocity, peaking where temperature (density) peaks vanish, giving the characteristic phase offset between TT and EE.
Characteristic numbers, scales, and the key relation
The E-mode amplitude is small because it requires the quadrupole to survive only during the brief last-scattering window. Roughly, the polarization fraction scales as the ratio of the last-scattering thickness to the wavelength, Π ∼ (k·Δη_LSS), yielding a few-percent polarization of the temperature anisotropy — a few μK in absolute terms. The EE power spectrum, plotted as ℓ(ℓ+1)C_ℓ/2π, rises to peaks of order 1–40 μK² across ℓ ≈ 300–1500, with the acoustic peaks out of phase with the temperature peaks because polarization tracks velocity while temperature tracks density.
Two scales dominate. The recombination E-modes live at ℓ ≳ 100 (sub-degree), set by the sound horizon at z ≈ 1090, when T ≈ 3000 K. The reionization bump sits at ℓ ≲ 10 (tens of degrees), generated when reionized electrons (z ≈ 6–8) re-scatter CMB photons across the new horizon. Its amplitude fixes the Thomson optical depth to reionization, measured by Planck as τ ≈ 0.054 ± 0.007, corresponding to a reionization midpoint near z ≈ 7.7.
How E-modes are detected
Measuring a μK polarization on top of a 2.725 K background demands exquisitely stable, differential radiometry, usually with rotating half-wave plates or paired detectors sensitive to orthogonal polarizations. The Degree Angular Scale Interferometer (DASI) at the South Pole announced the first E-mode detection in 2002 (∼4.9σ, later 6.3σ), operating at 26–36 GHz over multipoles ℓ ≈ 200–800. WMAP first mapped the large-scale TE correlation and the reionization signal; Planck (30–353 GHz) delivered full-sky EE and TE spectra and the definitive low-ℓ τ measurement.
Modern high-resolution ground experiments — the Atacama Cosmology Telescope (ACT), the South Pole Telescope (SPT), and the BICEP/Keck array — measure EE and TE to cosmic-variance precision at ℓ up to several thousand, at ∼95–150 GHz where the CMB dominates over synchrotron and dust foregrounds. Because E-modes and B-modes mix under partial-sky masking, analysts use "pure" E/B estimators to prevent leakage, a critical step when the goal is isolating faint B-modes from the far brighter E signal.
Where E-modes operate and how they differ from related effects
E-modes are sourced wherever free electrons scatter a quadrupole: at the surface of last scattering (z ≈ 1090) and again during reionization (z ≈ 6–10). Both are scalar-dominated, so both produce E-modes; the two epochs separate cleanly in multipole space. This is distinct from the temperature anisotropy, which is generated by density, Doppler, and gravitational (Sachs–Wolfe) effects rather than scattering geometry.
The sharpest distinction is E versus B. Scalar density perturbations, by symmetry, can only generate E-modes at recombination — they produce no primordial B-modes. Tensor perturbations (gravitational waves) generate both E and B. Therefore any confirmed primordial B-mode is a smoking gun for inflationary gravitational waves. However, a second B-mode source exists: gravitational lensing by large-scale structure shears E-modes into B-modes at small scales (ℓ ≈ 1000), a confirmed, subtracted "contaminant" for primordial searches. E-modes are thus both the reference template and, through lensing, the origin of the lensing B-mode background.
Open questions and significance
E-mode cosmology is now precision science, but frontier questions remain. The low-ℓ reionization bump still carries the largest uncertainty on τ, which is degenerate with the amplitude of primordial fluctuations A_s and with the sum of neutrino masses; a cleaner τ from a future large-scale polarization satellite (e.g., LiteBIRD) would sharpen constraints on σ_8 and on Σm_ν. Some tension between low-ℓ and high-ℓ likelihoods, and hints in TE, keep systematic-error control an active concern.
The deepest stakes are in the B-mode search: because delensing requires an accurate E-mode map, ever-better EE measurements directly improve limits on the tensor-to-scalar ratio r (currently r < 0.03–0.036). E-modes also test whether polarization is purely curl-free as expected, probing for exotic parity-violating physics (cosmic birefringence), where a nonzero EB or TB correlation would signal new physics. As the guaranteed, best-understood polarization signal, E-modes remain the backbone of CMB cosmology — the calibrated ruler that lets every fainter effect be measured.
| Property | E-modes | B-modes |
|---|---|---|
| Parity / symmetry | Even (curl-free), unchanged by mirror flip | Odd (divergence-free), flips sign under reflection |
| Pattern geometry | Radial / tangential to hot & cold spots | Swirling / vortex (±45° pinwheels) |
| Dominant source | Scalar (density) perturbations via Thomson scattering | Tensor (grav. waves) at recombination; lensing of E at small scales |
| Peak amplitude | ~few μK; EE peaks at ℓ≈300–1000 | Primordial: ≲0.1 μK, r-dependent; lensing: sub-μK at ℓ≈1000 |
| Recombination signature | Acoustic peaks anticorrelated/correlated with TT (TE) | No scalar contribution — clean gravitational-wave channel |
| Reionization signature | Low-ℓ 'bump' at ℓ<10 setting τ | Faint low-ℓ tensor bump if r>0 |
Frequently asked questions
Why does Thomson scattering only produce polarization from a quadrupole?
The Thomson cross-section varies as 1 + cos²θ, so an electron re-emits with a polarization weighted by the incoming intensity pattern. A uniform (monopole) or dipole radiation field averages out to zero net polarization by symmetry. Only a quadrupole — hotter along one axis and colder along the perpendicular axis — leaves a residual linear polarization aligned with the cold direction. That is why the CMB polarization is small: a quadrupole only develops briefly during last scattering.
What is the difference between E-modes and B-modes?
Both describe the geometry of the polarization pattern. E-modes are curl-free (even parity) with polarization vectors radial or tangential to spots, and they look the same in a mirror. B-modes are divergence-free (odd parity) with a swirling, handed pattern that flips under reflection. Scalar density perturbations produce only E-modes at recombination, while gravitational waves and gravitational lensing produce B-modes — which is why B-modes are the prime target for detecting inflationary gravitational waves.
How big is the E-mode signal compared to the temperature anisotropy?
The CMB temperature anisotropies are about 100 μK (parts in 10⁵ of 2.725 K). E-mode polarization is roughly ten times smaller — a few μK — because it only arises from the fraction of the quadrupole present during the short last-scattering interval. In power-spectrum terms the EE peaks reach order 1–40 μK² (in ℓ(ℓ+1)C_ℓ/2π) at multipoles ℓ ≈ 300–1500.
What is the reionization bump and what does it tell us?
When the universe reionized around z ≈ 6–8, freed electrons re-scattered CMB photons, generating fresh large-scale polarization on the horizon scale at that epoch. This appears as a distinctive bump in the EE and TE spectra at low multipoles (ℓ < 10). Its amplitude measures the Thomson optical depth to reionization, τ ≈ 0.054 from Planck, which in turn constrains when the first stars and galaxies reionized the cosmos.
Why are the EE acoustic peaks out of phase with the temperature peaks?
Temperature anisotropy tracks the density (compression) of the photon-baryon fluid, peaking at the extrema of the acoustic oscillation. Polarization is sourced by the quadrupole, which is produced by fluid velocity gradients — and velocity is maximal when density is passing through zero. This 90-degree phase offset shifts the EE acoustic peaks relative to the TT peaks, a prediction beautifully confirmed by the data and a strong consistency check on the standard model.
Who first predicted and first detected CMB E-mode polarization?
Martin Rees pointed out in 1968 that an anisotropic radiation field would leave the CMB linearly polarized. The clean E/B (or gradient/curl) decomposition was formulated in 1997 by Uroš Seljak and Matias Zaldarriaga, and independently by Marc Kamionkowski, Arthur Kosowsky, and Albert Stebbins. The first detection came from the DASI interferometer at the South Pole in 2002 at about 4.9σ, later strengthened to 6.3σ with three years of data, followed by full-sky maps from WMAP and Planck.