Cosmic Birefringence: A Rotated CMB Polarization Angle from Axion-Like Fields

What cosmic birefringence actually is

The CMB is not just a bath of microwaves — about 5–10% of it is linearly polarized, imprinted when free electrons Thomson-scattered the last photons at recombination, roughly 380,000 years after the Big Bang (redshift z ≈ 1090). That polarization direction is a fixed physical quantity set by the plasma flows of the early universe.

Cosmic birefringence is the claim that this polarization direction has been rotated uniformly across the whole sky by an angle β during the photons' journey to us. The word borrows from optics: a birefringent crystal like calcite splits light because it has two different refractive indices. Here the 'crystal' is the vacuum of the universe itself, made optically active by a parity-violating field.

• It is achromatic — the rotation is identical at 100 GHz and 217 GHz, unlike Faraday rotation.
• It is a parity-odd effect: it distinguishes left- from right-handed light, something the Standard Model of electromagnetism forbids.
• It rotates only photons that travelled cosmological distances, which is the key to separating it from instrument errors.

The mechanism: a Chern-Simons coupling to an axion-like field

The favored explanation adds a single term to electromagnetism: a coupling between a pseudoscalar (axion-like) field φ and the photon, written as a Chern-Simons interaction in the Lagrangian:

L ⊃ −(g/4) φ F_μν F̃^μν,

where F is the electromagnetic field-strength tensor, F̃ is its dual, and g is the coupling constant (dimensions of inverse energy). Because F·F̃ is parity-odd, a nonzero, time-varying φ makes the vacuum optically active.

Solving Maxwell's equations with this term shifts the dispersion relation of the two circular polarizations in opposite directions: ω_± ≈ k ∓ (g/2) φ̇. Left- and right-handed photons accumulate different phases, so the linear-polarization angle — the superposition of the two — steadily rotates. Integrating along the light path gives a beautifully simple result:

β = (g/2) · Δφ = (g/2)·[φ(observer) − φ(last scattering)].

The rotation depends only on how much the field changed between recombination and today, not on the distance traveled or the photon frequency — the hallmark of a rolling axion or an axion-like dark-energy field.

Key numbers and a worked example

The measured signal is tiny but not infinitesimal. The landmark re-analysis by Yuto Minami and Eiichiro Komatsu (2020) of Planck 2018 data found:

• β = 0.35° ± 0.14°, disfavoring β = 0 at about 2.4 sigma (a 99.2% chance of being nonzero).
• A later joint Planck + WMAP analysis tightened this to β = 0.342° ± 0.094°, a ~3.6 sigma hint.
• Independent Atacama Cosmology Telescope (ACT DR6) work reports a comparable β ≈ 0.20°–0.22° ± 0.07°.

Worked estimate of the coupling: if an axion-like field constitutes the dark energy and rolls by roughly Δφ ~ f (its decay constant) over cosmic time, then β = (g/2)Δφ ≈ 0.3° = 0.005 rad implies a coupling of order g·f ~ 0.01. For a field with f near the Planck scale (~10^18 GeV), this corresponds to g ~ 10^−20 GeV^−1 — deep in territory unreachable by any laboratory experiment, which is why the CMB is the only viable detector.

How it is observed: the EB cross-spectrum trick

CMB polarization is decomposed into parity-even E-modes and parity-odd B-modes. In standard cosmology the correlation between them, the EB cross-power spectrum, must vanish because it is parity-violating. A uniform rotation by β takes some of the abundant E-mode power and leaks it into B-modes, generating a nonzero EB (and TB) signal:

• C_ℓ^EB ∝ sin(4β) · (C_ℓ^EE − C_ℓ^BB) — so EB grows linearly with β for small angles.

The historic obstacle was degeneracy: a mis-set detector angle α produces the same EB signal, and α is typically known only to ~0.5°, larger than the effect. The Minami–Komatsu breakthrough was to use the Galactic foreground emission (dust and synchrotron) as a built-in calibrator. Foreground photons are local, so birefringence β does not rotate them, but instrumental α does. By fitting CMB and foreground EB spectra jointly, the two effects separate. Missions like LiteBIRD (JAXA-led, launch ~2032) are designed to push the sensitivity toward a definitive 5-sigma detection.

How it differs from related phenomena

Several effects rotate or split polarized light; birefringence is distinguished by being cosmological, parity-violating, and achromatic.

• Faraday rotation twists polarization via magnetic fields in a plasma, but scales as wavelength squared (λ²) — strongly chromatic — so multi-frequency data rules it out as the source.
• Gravitational lensing of the CMB also mixes E into B modes, but it is parity-even and produces no EB/TB correlation, leaving a clean separation.
• Isotropic vs. anisotropic birefringence: a spatially uniform rotation gives the single number β; a fluctuating axion field imprints a direction-dependent rotation map, currently only bounded by upper limits.
• Instrumental miscalibration (α) mimics β exactly in the CMB alone — the crucial difference is that α also rotates foregrounds, which β cannot.

This chain of distinguishing tests — frequency independence, EB versus lensing, foreground behavior — is what elevates the signal from a curiosity to a candidate discovery.

Significance, controversy, and open questions

If confirmed, cosmic birefringence would be the first laboratory-grade evidence of parity violation in the electromagnetic sector on cosmological scales, and a direct detection of a new light field — plausibly the axion long sought as dark matter, or a dynamical dark-energy component (quintessence). It would connect the CMB to string-theory 'axiverse' predictions of many ultralight pseudoscalars.

But it is not yet a discovery. The outstanding worries are:

• Foreground modeling: the method assumes Galactic dust has zero intrinsic EB. If dust filaments are misaligned with the magnetic field, they can produce a small real EB that biases β.
• Residual systematics: the effect is smaller than typical polarization-angle calibration uncertainties, so unknown instrumental leakage remains a concern.
• Statistical strength: at ~3.6 sigma it falls well short of the 5-sigma bar for a claimed detection.

The resolution will come from next-generation experiments — LiteBIRD, the Simons Observatory, and CMB-S4 — plus dedicated on-sky polarization-angle calibrators, which together should either cement β ≈ 0.3° as new physics or expose it as a subtle artifact within this decade.

Isotropic vs. anisotropic cosmic birefringence, contrasted with instrumental polarization miscalibration

Property	Isotropic birefringence (beta)	Anisotropic birefringence	Instrument miscalibration (alpha)
Physical origin	Uniform change in axion field value between last scattering and today	Spatial fluctuations of the axion field on the sky	Detector/telescope polarization-angle error
Sky pattern	Same rotation everywhere	Direction-dependent rotation map	Same rotation everywhere (instrumental)
Rotates foregrounds?	No — only CMB photons that traversed cosmic distances	No — only CMB	Yes — rotates both CMB and Galactic foregrounds
Signature spectrum	Nonzero EB and TB cross-power	Birefringence B-modes, curl-like patterns	Nonzero EB and TB (degenerate with beta unless broken)
Current status	Hint at ~3.6 sigma (beta ≈ 0.34°)	Only upper limits so far	Calibrated out using foregrounds as a reference
Frequency dependence	None (Chern-Simons term is achromatic)	None	Depends on optics; can be chromatic