Faraday Rotation

A compass written into starlight

Magnetic fields thread the entire universe — through the disk of the Milky Way, through the gas between galaxies, through the violent magnetospheres of pulsars and the hot atmospheres of galaxy clusters. They are dynamically important: they channel cosmic rays, regulate star formation, and collimate the jets of active galaxies. And yet they are nearly invisible. Magnetism emits no light of its own, and the fields are weak — a few microgauss in the interstellar medium, a hundred thousand times feebler than a refrigerator magnet. To map something so faint across thousands of light-years, you need a messenger that physically carries the field's imprint to your telescope. Faraday rotation is that messenger.

The idea is simple and beautiful. When a linearly polarized radio wave passes through a plasma that is threaded by a magnetic field, the plane of polarization slowly rotates as the wave travels. The amount of rotation depends on how much plasma the wave crossed and how strong the magnetic field was along the path. Measure how much the polarization has turned, and you have, in effect, read a running tally of the magnetic field integrated all the way back to the source. Michael Faraday discovered the laboratory version in 1845, twisting light in a block of glass with a magnet; a century later radio astronomers realized the diffuse cosmos is full of magnetized plasma doing the same thing on galactic scales.

How the polarization plane rotates

Start by splitting the incoming linearly polarized wave into two circularly polarized components of equal amplitude — one rotating clockwise, one counterclockwise. Added together they make a polarization vector that simply oscillates back and forth along a fixed line; that is what "linear polarization" means. In empty space both circular components travel at exactly the same speed, so the line never tilts.

Now introduce a magnetized plasma. The free electrons gyrate around the magnetic field lines at the cyclotron frequency, and that gyration has a definite handedness set by the field direction. A circularly polarized wave whose rotation matches the electron gyration couples to the plasma differently from one that rotates the opposite way. The two circular modes therefore acquire slightly different refractive indices — the plasma is circularly birefringent. They travel at slightly different phase speeds, and over a long path one mode falls progressively behind the other in phase. When you recombine the two circular components at the far end, their phase mismatch means the linear polarization vector they reconstruct is no longer along the original line: it has rotated. The whole effect is controlled by the projection of the magnetic field onto the line of sight, B_∥, because only that component sets a consistent handedness for the wave to interact with.

linear in  →  R + L (equal)  →  R and L slip in phase  →  linear out, plane rotated
rotation governed by B_∥ (line-of-sight field), n_e (electrons), and path length

The λ² law and the rotation measure

Working through the cold-plasma dispersion relation for the two circular modes, the net angle through which the polarization plane turns comes out as a remarkably clean expression:

Δχ = RM · λ²

Here λ is the observing wavelength and RM is the rotation measure, which collects all the plasma physics and the field geometry into a single number. For the densities, fields, and distances relevant to astrophysics, the rotation measure works out to

RM = 0.81 · ∫ n_e · B_∥ · dl   rad/m²

  n_e  = thermal electron density   (cm⁻³)
  B_∥  = line-of-sight magnetic field (μG)
  dl   = path increment             (parsecs)
  integral runs from the source to the observer

The factor 0.81 simply absorbs the fundamental constants (electron charge and mass, the speed of light, and the unit conversions) so that the integral can be evaluated in these convenient astronomical units. Two features make this equation the workhorse of cosmic magnetism. First, the λ² dependence: it means the rotation is enormous at radio wavelengths and utterly negligible at optical ones, which is why this is a radio technique and why you can extract RM cleanly by plotting polarization angle against λ² and reading off the slope. Second, RM is a signed path integral of the field weighted by electron density — positive when the average field points toward you, negative when it points away — so it senses both the magnitude and the direction of the line-of-sight field.

Worked example: weighing the Milky Way's field

Suppose you observe a polarized pulsar through the disk of the Galaxy and measure two quantities. The first is the dispersion measure, the integrated electron column the pulses cross, which for our example pulsar is DM = ∫ n_e·dl = 30 pc·cm⁻³. The second is the rotation measure, measured from the λ² slope of the polarization angle, RM = +30 rad/m². We want the average magnetic field along the path.

Divide RM by DM. The 0.81 prefactor and the unit conversions combine so that the density-weighted mean line-of-sight field is

<B_∥> = 1.232 · (RM / DM)   μG
      = 1.232 · (30 / 30)
      = 1.23 μG     (pointing toward us, since RM > 0)

That is exactly the microgauss-scale regular field that pervades the Milky Way's disk. Notice how powerful the pairing is: the dispersion measure supplies the electron column, the rotation measure supplies the field-weighted column, and their ratio cancels the unknown density to leave the field strength itself. This RM/DM trick, applied to hundreds of pulsars across the Galaxy, is how the large-scale Galactic magnetic field — its ~1.5 μG strength, its spiral-arm reversals, and its toward-versus-away geometry — was first mapped. For a source without a pulsar's dispersion measure, you instead difference its RM against a model of the electron distribution, but the logic is the same: RM converts a polarization angle into a magnetic field.

Variants and regimes

The clean Δχ = RM·λ² law assumes a single foreground "Faraday screen" — a magnetized plasma that rotates the polarization but does not itself emit. Real sightlines depart from this in instructive ways:

• External Faraday screen. Emission happens behind a separate rotating slab. This is the textbook case: χ(λ²) is a perfect straight line, and a single RM describes everything. Most extragalactic background-source RMs work this way.
• Internal (differential) Faraday rotation. The emitting region and the rotating plasma are co-spatial, so light emitted at the far edge is rotated more than light from the near edge. The polarization angle becomes a non-linear function of λ², and the source depolarizes at long wavelengths as the differently-rotated contributions cancel.
• Multiple screens / Faraday-thick media. Several magnetized regions at different Faraday depths overlay along one sightline. No single RM applies; you need the full Faraday dispersion function recovered by RM synthesis.
• Beam depolarization. If the telescope beam averages over patches of differing RM, the polarization vectors add incoherently and the net fractional polarization drops. This is itself a diagnostic of small-scale field tangling.

From a slope to a spectrum: RM synthesis

For a single Faraday screen, three or more polarization-angle measurements across wavelength give a straight λ² line whose slope is the RM — the classic method, limited by the nπ ambiguity that arises because polarization angle is only defined modulo 180°. Wide, continuously sampled radio bands removed that limitation. Brentjens and de Bruyn (2005) recognized that the observed complex polarization P(λ²) and the distribution of polarized emission as a function of Faraday depth φ form a Fourier-transform pair:

P(λ²) = ∫ F(φ) · e^{ 2 i φ λ² } dφ        (Faraday dispersion function F(φ))

Inverting this transform — RM synthesis — turns a broadband polarization spectrum into a one-dimensional map of how much polarized signal sits at each Faraday depth, cleanly separating the multiple components that a single-slope fit would smear together. The width of the resolving "RM beam" is set by the λ² coverage of the band, exactly analogous to how aperture sets angular resolution. Every modern broadband polarimeter — ASKAP, LOFAR, MeerKAT, and the SKA — is designed with RM synthesis as a primary data product.

Observational status and what RM has measured

Faraday rotation underpins essentially everything we know quantitatively about cosmic magnetism:

• The Galactic magnetic field. Pulsar RM/DM pairs and grids of extragalactic background-source RMs reveal a ~1.5 μG regular field that broadly follows the spiral arms, with field reversals between arms and a vertical/toroidal halo component. Typical extragalactic RMs are tens of rad/m² at the poles, hundreds toward the plane.
• Galaxy clusters. RMs of radio sources seen through the hot intracluster medium reach hundreds to thousands of rad/m², implying few-microgauss fields ordered coherently over tens of kiloparsecs — far stronger and more organized than once assumed.
• The intergalactic medium and primordial fields. The near-absence of RM through cosmic voids (residual RM < ~1 rad/m²) bounds any cosmological seed field to roughly the nanogauss level — among the tightest constraints on magnetogenesis in the early universe.
• Extreme magnetoionic environments. The magnetar PSR J1745−2900 near the Galactic Center shows RM ≈ −67,000 rad/m², among the largest in the Galaxy. Fast radio bursts now deliver RMs along cosmological sightlines, with one repeater reaching ~10⁵ rad/m² from an exceptionally magnetized host.
• The all-sky RM grid. Surveys such as POSSUM on ASKAP are assembling RMs for millions of sources, the raw material for reconstructing the three-dimensional magnetic structure of the Milky Way and the cosmic web. The SKA will increase this density by another order of magnitude.

Faraday rotation across the cosmos

Environment	Typical |RM| (rad/m²)	n_e (cm⁻³)	B_∥ (μG)	Path scale
Cosmic voids / IGM	< 1	~ 10⁻⁷	< 0.001 (nG seed)	~ Gpc
Diffuse ISM (high latitude)	10 – 50	~ 0.03	~ 1.5	~ kpc
Galactic plane sightline	100 – 500	~ 0.05	~ 2 – 4	several kpc
Galaxy cluster (ICM)	100 – 3,000	~ 10⁻³	~ 1 – 5	tens of kpc
Spiral-arm HII region	1,000+	~ 1 – 10	~ 5 – 10	~ 10s pc
Galactic-Center magnetar	~ 67,000	~ 10 – 100	~ mG locally	compact
Extreme FRB host	~ 10⁵	dense, ionized	strong, ordered	compact

The seven-decade span in RM, from below 1 to above 10⁵ rad/m², is precisely why the technique is so versatile: the same λ² law that registers a faint twist through a cosmic void also captures the violent rotation in a magnetar's circumstellar plasma.

Quantitative analysis: where 0.81 comes from

The difference in refractive index between the two circular modes in a cold magnetized plasma, in the high-frequency limit where the wave frequency far exceeds both the plasma and gyro frequencies, is

n_R − n_L ≈  (ω_p² · ω_c,∥) / ω³

  ω_p² = 4π n_e e² / m_e   (plasma frequency squared)
  ω_c,∥ = e B_∥ / (m_e c)   (electron gyrofrequency, line-of-sight projection)

The polarization rotation accumulated over a path is half the phase difference between the modes, integrated along the line of sight:

Δχ = (ω / 2c) ∫ (n_R − n_L) dl
   = (2π e³ / m_e² c² ω²) ∫ n_e B_∥ dl
   = (e³ λ² / 2π m_e² c⁴) ∫ n_e B_∥ dl     (using ω = 2πc/λ)

Everything except the integral and the λ² is a bundle of fundamental constants. Evaluate that bundle and convert n_e to cm⁻³, B_∥ to microgauss, and dl to parsecs, and the prefactor reduces to the famous 0.812 (rounded to 0.81):

Δχ = RM · λ²,        RM = 0.81 ∫ n_e B_∥ dl   rad/m²

A quick numerical check anchors the scale. For a high-latitude sightline with n_e ≈ 0.03 cm⁻³, B_∥ ≈ 1.5 μG, over a path of about 700 pc, the integral is ∫ n_e B_∥ dl ≈ 0.03 × 1.5 × 700 ≈ 31 μG·pc·cm⁻³, giving RM ≈ 0.81 × 31 ≈ 25 rad/m². At a wavelength of λ = 0.21 m (the 21 cm line) that is a rotation of Δχ = 25 × 0.21² ≈ 1.1 radians, about 63° — easily measured. At optical wavelengths (λ ≈ 5 × 10⁻⁷ m) the same RM gives Δχ ≈ 6 × 10⁻¹² radians, hopelessly unmeasurable. The λ² law is the whole reason cosmic magnetism is a radio science.

Common pitfalls and misconceptions

• Confusing Faraday rotation with dispersion. Both are plasma propagation effects, but dispersion delays the arrival time of a pulse (∝ λ², via DM = ∫ n_e dl) and needs no magnetic field, whereas Faraday rotation twists the polarization angle (∝ λ², via RM = 0.81∫ n_e B_∥ dl) and vanishes if B_∥ = 0. They are measured from different observables and carry complementary information.
• Reading RM as the magnetic field. RM is a density-weighted path integral of B_∥, not the field itself. Without an independent electron column (a DM, or an emission-measure model) you cannot convert RM into a field strength; a large RM can mean a strong field, a long path, or a dense plasma.
• Forgetting the sign and the projection. RM senses only the line-of-sight component B_∥; a strong field perpendicular to the sightline contributes zero. Field reversals along the path can partly cancel, so a small net RM does not imply a weak field.
• The nπ ambiguity. Because polarization angle is defined modulo 180°, RM derived from a few widely spaced wavelengths can be wrong by multiples of π/Δ(λ²). Densely sampled wideband data or RM synthesis is required to nail it down.
• Assuming χ(λ²) is always linear. A straight line only holds for an external Faraday screen. Internal rotation, multiple screens, and beam averaging all bend the relation and depolarize the source; treating them with a single-slope fit gives a meaningless RM.
• Ignoring the ionosphere and the Sun. Earth's ionosphere adds a time-variable RM of a few rad/m² that must be modeled and removed, and the solar wind imprints its own rotation on sources seen near the Sun — both are foregrounds, not signal.