Interstellar Polarization and Magnetic Fields

You cannot photograph a magnetic field. Yet the Galaxy is threaded by a field of a few microgauss — a millionth of Earth's — and we have mapped its geometry across the whole sky. The trick is dust: tiny elongated grains of silicate and carbon, a fraction of a micron long. Left alone they tumble randomly, but the field spins them up and parks them with their short axis along $\mathbf{B}$ , and an aligned population behaves like a sky-sized wire-grid polarizer. Background starlight forced through it emerges faintly polarized — a percent or two — with its electric vector combed parallel to the projected field, because the perpendicular component was preferentially absorbed by the grains' long axes. The same warm grains glow in the far-infrared and submillimetre, and that thermal emission comes out polarized perpendicular to the field. Two channels, one field, ninety degrees apart. That single fact turns dust into a magnetometer. The direction of the polarization vectors traces where the plane-of-sky field points; the dispersion of those vectors about a smooth mean measures how badly turbulence has bent the field lines, and through the Davis–Chandrasekhar–Fermi relation that scatter converts into a field strength. How a grain gets aligned — Barnett magnetism, paramagnetic dissipation, and above all radiative torques — is why the observed polarization is never as strong as a perfect aligner would give. From the Serkowski law of optical starlight to Planck's all-sky submillimetre maps, aligned-grain polarimetry is now both the primary probe of Galactic magnetism and the dominant foreground between us and primordial gravitational waves in the cosmic microwave background.

Key takeaways

Elongated dust grains, spun up and aligned by the interstellar magnetic field so their short axis lies along $\mathbf{B}$ , act as a dichroic filter on background starlight.
Transmitted optical/near-IR starlight is polarized parallel to the projected field (Serkowski law, peaking near $\lambda_{\max}\approx0.55\,\mu\mathrm{m}$ ); thermal dust emission in the far-IR/submm is polarized perpendicular to it — the two differ by exactly $90^\circ$ .
Polarization efficiency is capped empirically at $p_V/A_V\lesssim3\%\ \mathrm{mag^{-1}}$ ; most sightlines fall well below because the field is tangled or tilted out of the plane of sky.
Modern alignment is understood through Radiative Alignment Torques (RATs): anisotropic starlight spins irregular grains suprathermally, where the classical Davis–Greenstein paramagnetic mechanism is far too slow.
The Davis–Chandrasekhar–Fermi method converts polarization-angle dispersion plus gas velocity dispersion and density into a plane-of-sky field strength — tens to hundreds of $\mu$ G in dense cores.
Key idea: magnetically aligned grains dichroically extinguish starlight (E-vector ∥ projected B) and thermally emit (E-vector ⊥ B); the polarization direction maps the field's plane-of-sky orientation, and the angle dispersion measures its strength and turbulence.

The core insight

Magnetically aligned, elongated grains dichroically extinguish background starlight (E-vector parallel to the projected field) and thermally emit (E-vector perpendicular to it); measuring the polarization direction maps the field's plane-of-sky orientation, while the angle dispersion measures its strength and turbulence.

Where this lives today

Aligned-grain polarimetry is now a mature, multi-instrument enterprise at the crossroads of star formation and cosmology. Planck's 353 GHz maps delivered the first all-sky picture of the plane-of-sky Galactic field, revealing that interstellar filaments align with the field — dense striations run perpendicular to $\mathbf{B}$ , diffuse ones parallel — a clue to how gas flows onto forming clouds. On smaller scales, SOFIA's HAWC+ camera and ground-based submillimetre polarimeters (JCMT/POL-2, ALMA) map fields down into individual protostellar cores and disks, feeding DCF and its refinements to test whether magnetic support or turbulence sets the pace of collapse. And the field has acquired a cosmological urgency: polarized dust emission is the most stubborn foreground standing between experiments like BICEP/Keck, the Simons Observatory, and LiteBIRD and a detection of primordial gravitational-wave B-modes in the CMB. The 2014 BICEP2 episode — an early B-mode claim later shown by the joint BICEP–Planck analysis to be largely aligned-grain dust — made the point permanently: to see the beginning of the Universe, we must first understand the microphysics of a spinning grain of interstellar silicate.

The physical picture

Picture a single elongated grain as a microscopic wire-grid polarizer. An electric field oscillating along the grain's long axis drives the loosely bound charges the full length of the grain and loses energy — that polarization state is strongly absorbed. A field along the short axis has little room to push charge and passes more freely. One grain does almost nothing, but if a whole population shares an alignment the transmitted beam is net-polarized along the low-absorption direction. What sets that direction is the magnetic field: aligned grains sit with their short axis along $\mathbf{B}$ , so the least-absorbed polarization — the one that survives to reach us — is the one whose electric vector lies parallel to the projected field. Why do the grains align at all? A spinning grain is a gyroscope carrying a built-in magnet. Rotation itself magnetizes it (the Barnett effect, the rotational analogue of Einstein–de Haas), and that magnetic moment makes the spin axis precess around $\mathbf{B}$ far faster than any collision can disturb it — on average the grain "knows about" only one direction, the field. Dissipative torques then bleed away the wobble, leaving the angular momentum, and statistically the short axis, locked along $\mathbf{B}$ . The first figure fixes the geometry everything else depends on: which way the grain points, and which way the two polarization channels come out.

Figure

An elongated grain aligns with its short axis along the field B. Unpolarized starlight passes through and emerges polarized parallel to the projected field, because the long-axis absorbs the other component; the grain's own thermal glow is polarized perpendicular to B — the two channels differ by 90° and the transmitted fraction obeys p_V/A_V ≲ 3% per magnitude

Radiative torques do the heavy lifting. An irregular grain in anisotropic starlight is a windmill with helically twisted blades: the photon flux spins it up to thousands of times its thermal rotation rate, freezing in an alignment the feeble paramagnetic mechanism could never secure for a $0.1\,\mu\mathrm{m}$ grain, and the suprathermal spin averages away the buffeting from gas atoms. This is why radiative alignment torques (RATs), not Davis–Greenstein dissipation, are the accepted modern theory. But alignment is fragile: deep inside a dense core where starlight is extinguished, the torque weakens and the polarization fraction collapses — the observed "polarization hole" is a direct test of RAT physics. Finally, because a single sightline pierces many grains sampling many patches of field, the observed polarization is a line-of-sight average. A field lying along the line of sight, or a tangled turbulent field, averages the vectors toward zero even when every grain is perfectly aligned. Low polarization therefore need not mean weak alignment or a weak field — it may simply mean disordered or unfavourably oriented geometry. The direction is robust; the fraction encodes the three-dimensional field structure. The second figure shows the quantitative signature that makes optical polarization a fingerprint of grain size: the Serkowski law.

Figure

The Serkowski law p(λ) = p_max exp(−K ln²(λmax/λ)) with λmax = 0.55 μm and K = 1.66 λmax = 0.913: the polarization peaks in the optical V band and falls off symmetrically in ln λ, dropping to 96% of peak at B (0.44 μm) and only 17% by K band (2.2 μm) — the peak at optical wavelengths pins the aligned grains near a ≈ 0.1 μm

The core equations

The empirical backbone of optical polarimetry is the Serkowski law — not derived from first principles but an extraordinarily good three-parameter fit to the wavelength dependence of starlight polarization, whose shape (a Gaussian in $\ln\lambda$ ) fingerprints the aligned grain-size distribution.

p(\lambda) = p_{\max}\,\exp\!\left[-K\,\ln^2\!\left(\frac{\lambda_{\max}}{\lambda}\right)\right] \qquad (1)

where: $p(\lambda)$ — fractional linear polarization at wavelength $\lambda$ ; $p_{\max}$ — peak polarization (typically a few percent); $\lambda_{\max}$ — wavelength of the peak, $\approx0.55\,\mu\mathrm{m}$ in the diffuse ISM; $K$ — width parameter, empirically $K\approx1.66\,\lambda_{\max}$ with $\lambda_{\max}$ in microns. Reading: the curve peaks at grain-sized wavelengths and is symmetric in $\ln\lambda$ , so bluer and redder light are both less polarized than the V band. Polarization is really a statement about the Stokes parameters — the observables a polarimeter measures. The fraction $p$ and the position angle $\theta$ together give the length and direction of the polarization vector we draw on the sky.

p = \frac{\sqrt{Q^2+U^2}}{I},\qquad \theta = \tfrac{1}{2}\arctan\!\frac{U}{Q} \qquad (2)

where: $I$ — total intensity; $Q,\,U$ — linear Stokes parameters; $p$ — polarization fraction; $\theta$ — position angle of the electric vector. Reading: the factor of $\tfrac12$ marks polarization as a "headless" spin-2 quantity — a $180^\circ$ rotation gives back the same state, so polarimetry yields an axis, never a sense of direction. The magnitude is set by dichroism, the optical-depth difference between the two grain axes. In the optically thin limit the polarization grows linearly with the dust column — the origin of the observed $p\propto A_V$ correlation — capped by an empirical efficiency ceiling.

p \simeq \tfrac{1}{2}\,(\tau_\parallel-\tau_\perp),\qquad \frac{p_V}{A_V}\lesssim 3\%\ \mathrm{mag^{-1}} \qquad (3)

where: $\tau_\parallel,\,\tau_\perp$ — optical depths for light polarized along the grain's long and short axes; $p_V$ — V-band fractional polarization (percent); $A_V$ — visual extinction (magnitudes). Reading: the ceiling is the best case of perfect alignment and a maximally ordered field; real sightlines fall below through imperfect alignment, imperfect elongation, and line-of-sight field disorder. Alignment is a competition of torques on the grain's angular momentum, and a simple timescale argument shows why the classical paramagnetic mechanism fails for the grains that carry the polarization.

\frac{\mathrm{d}\mathbf{J}}{\mathrm{d}t} = \boldsymbol{\Gamma}_{\mathrm{rad}} + \boldsymbol{\Gamma}_{\mathrm{gas}} + \boldsymbol{\Gamma}_{\mathrm{DG}},\qquad t_{\mathrm{DG}}\propto \frac{\rho\,a^2\,\omega}{\chi''(\omega)\,B^2}\ \gg\ t_{\mathrm{gas}}\ (\text{large }a) \qquad (4)

where: $\mathbf{J}$ — grain angular momentum; $\boldsymbol{\Gamma}_{\mathrm{rad}}$ — radiative (RAT) torque from anisotropic starlight; $\boldsymbol{\Gamma}_{\mathrm{gas}}$ — randomizing gas-collision torque; $\boldsymbol{\Gamma}_{\mathrm{DG}}$ — paramagnetic Davis–Greenstein dissipation; $t_{\mathrm{DG}}$ — its alignment time; $t_{\mathrm{gas}}$ — gas-randomization time; $\rho$ — grain density; $a$ — grain radius; $\chi''$ — imaginary magnetic susceptibility; $B$ — field strength. Reading: since $t_{\mathrm{DG}}\propto a^2/B^2$ grows with grain size, canonical $0.1\,\mu\mathrm{m}$ grains in $\mu$ G fields randomize before they align — hence RATs are needed. The single most powerful diagnostic is the Davis–Chandrasekhar–Fermi relation, which turns polarization-angle scatter into a plane-of-sky field strength by equating turbulent kinetic energy to the energy of the bent field lines.

B_{\mathrm{pos}} \approx Q\,\sqrt{4\pi\rho}\;\frac{\sigma_v}{\sigma_\theta} \qquad (5)

where: $B_{\mathrm{pos}}$ — plane-of-sky field strength (Gaussian units); $\rho$ — gas mass density; $\sigma_v$ — line-of-sight velocity dispersion; $\sigma_\theta$ — dispersion of polarization angles (radians); $Q\approx0.5$ — a correction factor from MHD simulations. Reading: a small angle scatter for a given turbulent velocity means stiff, hard-to-bend field lines — a strong field; large scatter means a weak one. Converting a field strength into a statement about star formation requires the mass-to-flux ratio, the quantity that decides whether magnetic tension can hold a cloud up against its own gravity.

\mu \equiv \frac{M/\Phi_B}{(M/\Phi_B)_{\mathrm{crit}}},\qquad \left(\frac{M}{\Phi_B}\right)_{\mathrm{crit}} = \frac{1}{2\pi\sqrt{G}} \qquad (6)

where: $M$ — cloud mass; $\Phi_B$ — magnetic flux threading it; $\mu$ — normalized mass-to-flux ratio; $G$ — gravitational constant. Reading: $\mu<1$ (subcritical) means the field wins and collapse is forbidden until ambipolar diffusion leaks flux out; $\mu>1$ (supercritical) means gravity wins and the cloud can collapse.

Derivation ADichroic extinction and the direction of starlight polarization

The goal is to show quantitatively why a population of aligned elongated grains leaves transmitted starlight polarized parallel to the projected field, and why the polarization fraction grows in proportion to the dust column at low optical depth. We model the grains as a differential absorber acting on two orthogonal polarization states of an unpolarized beam, carry each state through the extinction law, and form the observable fraction — connecting microscopic grain geometry to the field direction on the sky.

Step 1. Resolve the incoming unpolarized starlight into two equal orthogonal components, one with its electric vector along the grain's long axis and one along the short axis, and assign each its own optical depth $\tau_\parallel$ and $\tau_\perp$ . Aligned elongated grains absorb and scatter more when the field drives charge along the long axis, so $\tau_\parallel>\tau_\perp$ . Why: dichroism is precisely the inequality of the two optical depths; if the grains were spherical or randomly oriented the two would be equal and no net polarization could arise.

Step 2. Attenuate each component by its own extinction, $I_\parallel=\tfrac12 I_0\,e^{-\tau_\parallel}$ and $I_\perp=\tfrac12 I_0\,e^{-\tau_\perp}$ . The beam that survives is richer in the short-axis polarization because that component was absorbed less. Why: this is just Beer–Lambert applied independently to each polarization state — the whole effect is that the two states see different media.

Step 3. Form the fractional polarization from the two survivors, $p=(I_\perp-I_\parallel)/(I_\perp+I_\parallel)=\tanh\!\big[(\tau_\parallel-\tau_\perp)/2\big]$ , exact for the two-state model and bounded by 1. Why: this ratio is the definition of linear polarization fraction, and the tanh shows it saturating at large differential optical depth.

Step 4. Take the optically thin limit of diffuse sightlines, $\Delta\tau=\tau_\parallel-\tau_\perp\ll1$ , so $p\approx\tfrac12(\tau_\parallel-\tau_\perp)$ . Both optical depths scale with the dust column, so their difference does too, giving $p\propto A_V$ with an envelope near $3\%\ \mathrm{mag^{-1}}$ . Why: this linearity is why polarization and reddening track each other, and why $p/A_V$ cleanly measures alignment efficiency.

Step 5. Fix the direction. Aligned grains sit with their short axis along $\mathbf{B}$ , so the least-attenuated polarization — the surviving E-vector — lies parallel to the projected field $\mathbf{B}_\perp$ . Rotating to thermal emission, the grains radiate most efficiently along their long axis, so emission comes out polarized perpendicular to $\mathbf{B}_\perp$ , $90^\circ$ from the absorption vector. Why: this is the payoff — a measured polarization angle is a direct readout of the plane-of-sky field orientation, with a fixed $90^\circ$ offset between the absorption and emission channels.

Go deeper · Derivation B · 2 minThe Davis–Chandrasekhar–Fermi field strength from angle dispersion

Derivation BThe Davis–Chandrasekhar–Fermi field strength from angle dispersion

The aim is to derive the relation that turns a map of polarization vectors into a magnetometer: an estimate of the plane-of-sky field strength from nothing but the angle scatter, the gas velocity dispersion, and the density. The idea, due to Chandrasekhar and Fermi in 1953, is that turbulent gas motions ripple the field lines like plucked strings, and the ripple amplitude — visible as the dispersion of polarization angles about the smooth mean field — is set by the balance between turbulent kinetic energy and the tension of the bent field. We make that balance quantitative and solve for $B$ .

Step 1. Equate the turbulent kinetic energy density to the energy density of the field perturbation, $\tfrac12\rho\,\delta v^2\simeq(\delta B)^2/8\pi$ . This is energy equipartition for Alfvénic turbulence, where the transverse gas motion and the transverse field ripple are two faces of the same wave. Why: equipartition is the single physical assumption that couples the gas kinematics to the field geometry; without it there is no bridge from velocities to field strength.

Step 2. Introduce the Alfvén speed $v_A=B/\sqrt{4\pi\rho}$ , the speed at which transverse field perturbations propagate, and rewrite the equipartition relation as $\delta v\simeq v_A\,(\delta B/B)$ . Why: the Alfvén speed is the natural velocity scale of a magnetized fluid, and it is exactly the combination of $B$ and $\rho$ that will survive to the final formula.

Step 3. Identify the field ripple with the observed polarization-angle scatter. A small transverse field perturbation tilts the local field — and the grains that trace it, and hence the polarization vectors — through an angle whose dispersion is the measured $\sigma_\theta$ , so $\delta B/B\approx\tan\sigma_\theta\approx\sigma_\theta$ for modest scatter. Why: this is the observational heart of the method — it says the geometry we can see (angle scatter) directly measures the fractional field perturbation we cannot.

Step 4. Combine the pieces and identify $\delta v$ with the measured line-of-sight velocity dispersion $\sigma_v$ , giving $B\approx\sqrt{4\pi\rho}\,(\sigma_v/\sigma_\theta)$ , and multiply by a dimensionless factor $Q\approx0.5$ calibrated on MHD simulations to correct for line-of-sight and beam averaging, which otherwise make the raw estimate overshoot. Why: the correction factor is what turns a heuristic order-of-magnitude balance into a quantitatively trustworthy field estimate, and it is why modern DCF work always quotes $Q\approx0.5$ rather than unity.

Go deeper · Derivation C · 2 minThe magnetic critical mass-to-flux ratio

Derivation CThe magnetic critical mass-to-flux ratio

The purpose is to derive the threshold that decides whether a magnetized cloud can ever collapse — the critical mass-to-flux ratio at which magnetic tension exactly balances self-gravity for a flux-frozen cloud. This dimensionless number, together with a DCF field strength, turns a polarization map into a verdict on star formation: is the field strong enough to hold the cloud up, or not. We compare the gravitational and magnetic energies of a uniform cloud and show the balance depends only on the ratio of mass to flux, not on the cloud's size.

Step 1. Write the flux and mass of a uniform spherical cloud of radius $R$ , density $\rho$ , threaded by a field $B$ : $\Phi_B=\pi R^2 B$ and $M=\tfrac43\pi R^3\rho$ . Flux is conserved as the cloud contracts because the field is frozen into the highly conducting gas. Why: flux freezing is what makes $M/\Phi_B$ a conserved label of a cloud — it does not change under compression, so a cloud born subcritical stays subcritical until diffusion intervenes.

Step 2. Compare the gravitational binding energy $E_{\mathrm{grav}}\sim GM^2/R$ with the magnetic energy $E_{\mathrm{mag}}\sim(B^2/8\pi)\cdot\tfrac43\pi R^3$ . Eliminate $B$ using $B=\Phi_B/\pi R^2$ , so that $E_{\mathrm{mag}}\sim\Phi_B^2/6\pi R$ . Why: substituting the flux for the field is the key algebraic move — it exposes that the magnetic energy, like the gravitational energy, scales as $1/R$ , so the two cannot be separated by changing the cloud's size.

Step 3. Set the two energies comparable. Because both scale as $1/R$ , the radius cancels and the balance fixes a critical ratio of mass to flux rather than a critical size or density — the field either can or cannot support the cloud regardless of how far it has contracted. Why: this cancellation is the deep result; it is why magnetic support is characterized by the scale-free number $\mu$ and not by a Jeans-like critical mass alone.

Step 4. A careful treatment for a uniform sphere or sheet (Mouschovias & Spitzer 1976) fixes the coefficient at $(M/\Phi_B)_{\mathrm{crit}}=(2\pi\sqrt{G})^{-1}$ . Clouds with $\mu>1$ are supercritical and collapse; clouds with $\mu<1$ are subcritical and magnetically supported, able to form stars only as ambipolar diffusion slowly leaks flux out of the neutral gas. Why: the exact coefficient is what lets an observed $B$ and column density be converted into a definite sub- or supercritical verdict, closing the loop from polarimetry to star formation.

Limiting cases

How the polarization behaves as the geometry and grain physics change:

Case	Result	Meaning
Perfectly aligned, ordered field in the plane of sky	$p_{\max}\propto(C_\parallel-C_\perp)\,N_{\mathrm{grain}}$	Polarization saturates at the intrinsic dichroism set by the cross-section difference; uniform, maximal E-vectors.
Field along the line of sight	$p\to0$	Aligned grains present circular projected cross-sections; extinction persists but there is no preferred axis, so no net linear polarization.
Turbulent or tangled field	$p\sim p_{\max}\langle\cos2\theta\rangle<p_{\max}$	Angle averaging along the sightline suppresses the fraction; the residual scatter $\sigma_\theta$ is what feeds the DCF estimate.
Optical absorption vs far-IR emission	$\theta_{\mathrm{em}}=\theta_{\mathrm{abs}}+90^\circ$	Emission E-vectors are rotated $90^\circ$ from absorption; both trace the same $\mathbf{B}_\perp$ once you correct for the offset.
Dense core interior (extinguished starlight)	RAT weakens, $p$ drops (polarization hole)	With no anisotropic radiation to spin grains, alignment fails; low $p$ here is a loss of alignment, not of field.

Common trapLow polarization means a weak magnetic field.

It usually does not. Three distinct effects drive the observed fraction down with no change in field strength. First, only the plane-of-sky component polarizes: a field tilted toward or along the line of sight gives little linear polarization even if $|\mathbf{B}|$ is large, because the projected grain cross-sections become circular. Second, any real sightline averages over many field patches, and a turbulent or tangled field cancels the vectors and shrinks $p$ while the local field is perfectly strong and aligning. Third, in the shielded interior of a dense core the starlight that powers radiative torques is extinguished, so grains lose alignment and $p$ collapses in a "polarization hole" unrelated to the field. Remember too that the two channels are orthogonal — emission polarizes perpendicular to $\mathbf{B}_\perp$ , absorption parallel — so rotate an emission map by $90^\circ$ before comparing it to a starlight map. And polarimetry carries an inescapable $180^\circ$ ambiguity: it gives an axis, never a direction.

Worked examples

1 · Weighing the field in a dense core with the DCF method (fully worked)

A submillimetre polarimeter maps a dense core with molecular hydrogen number density $n(\mathrm{H_2})=10^4\ \mathrm{cm^{-3}}$ . The polarization vectors scatter about a smooth mean with a dispersion $\sigma_\theta=12^\circ$ , and a spectral line gives a one-dimensional velocity dispersion $\sigma_v=0.30\ \mathrm{km\,s^{-1}}$ . Estimate the plane-of-sky field strength, check that the method is self-consistent, and decide whether the core is magnetically sub- or supercritical given a column $N(\mathrm{H_2})=10^{22}\ \mathrm{cm^{-2}}$ .

Step 1 — Mass density. Using a mean mass of $2.8\,m_{\mathrm H}$ per hydrogen molecule to account for helium, $\rho=2.8\times m_{\mathrm H}\times n(\mathrm{H_2})=2.8\,(1.67\times10^{-24}\,\mathrm{g})(10^4\,\mathrm{cm^{-3}})=\mathbf{4.68\times10^{-20}\ g\,cm^{-3}}$ . Then $\sqrt{4\pi\rho}=\sqrt{4\pi\,(4.68\times10^{-20})}=\mathbf{7.67\times10^{-10}}$ in cgs.

Step 2 — Put the observables in cgs. The angle dispersion in radians is $\sigma_\theta=12^\circ\times(\pi/180)=\mathbf{0.209\ rad}$ , and the velocity dispersion is $\sigma_v=0.30\ \mathrm{km\,s^{-1}}=\mathbf{3.0\times10^4\ cm\,s^{-1}}$ .

Step 3 — Apply DCF. With $Q=0.5$ ,

B_{\mathrm{pos}}=Q\sqrt{4\pi\rho}\,\frac{\sigma_v}{\sigma_\theta}=0.5\,(7.67\times10^{-10})\frac{3.0\times10^4}{0.209}=5.5\times10^{-5}\ \mathrm{G}.

So $B_{\mathrm{pos}}\approx\mathbf{55\ \mu G}$ — squarely in the tens-to-hundreds of $\mu$ G expected for dense cores.

Step 4 — Self-consistency (is DCF even valid?). The Alfvén speed is $v_A=B/\sqrt{4\pi\rho}=(5.5\times10^{-5})/(7.67\times10^{-10})=7.2\times10^4\ \mathrm{cm\,s^{-1}}=0.72\ \mathrm{km\,s^{-1}}$ , giving an Alfvén Mach number $M_A=\sigma_v/v_A=0.30/0.72\approx\mathbf{0.42}$ . The turbulence is sub-Alfvénic, which is exactly the regime where DCF is trustworthy — the small-angle approximation $\delta B/B\approx\sigma_\theta$ holds.

Step 5 — Sub- or supercritical? Using the standard relation $\mu=7.6\times10^{-21}\,N(\mathrm{H_2})/B[\mu\mathrm{G}]$ (which is just $2\pi\sqrt G\,m_{\mathrm{gas}}N/B$ in disguise),

\mu=\frac{7.6\times10^{-21}\times10^{22}}{55}=\frac{76}{55}\approx\mathbf{1.4}.

The core is marginally supercritical ( $\mu>1$ ): gravity narrowly wins, so the field slows but cannot prevent collapse — a typical result for star-forming cores, where fields regulate rather than veto star formation.

2 · Reading grain size off the Serkowski curve

Along a diffuse sightline the polarization peaks at $\lambda_{\max}=0.55\,\mu\mathrm{m}$ with $p_{\max}=1.5\%$ . Find the width parameter $K$ , predict the polarization in the B band (0.44 μm) and the K band (2.2 μm), and estimate the size of the aligned grains.

Step 1 — Width parameter. The empirical relation gives $K\approx1.66\,\lambda_{\max}=1.66\times0.55=\mathbf{0.913}$ .

Step 2 — B band. $\ln(\lambda_{\max}/\lambda)=\ln(0.55/0.44)=\ln1.25=0.223$ , so $p(\mathrm B)=1.5\%\times\exp[-0.913\,(0.223)^2]=1.5\%\times e^{-0.0455}=\mathbf{1.43\%}$ — only 4% below the peak, because B lies close to $\lambda_{\max}$ in $\ln\lambda$ .

Step 3 — K band. $\ln(0.55/2.2)=\ln0.25=-1.386$ , so $p(\mathrm K)=1.5\%\times\exp[-0.913\,(1.386)^2]=1.5\%\times e^{-1.755}=1.5\%\times0.173=\mathbf{0.26\%}$ — the near-infrared polarization is only 17% of the peak, so infrared polarimetry needs far higher precision than optical.

Step 4 — Grain size. The polarization efficiency peaks when the size parameter $x=2\pi a/\lambda\sim1$ , i.e. $a\sim\lambda_{\max}/2\pi=0.55/6.28\approx\mathbf{0.09\ \mu m}$ . The aligned grains are the same $\sim0.1\,\mu\mathrm{m}$ population that dominates the optical extinction curve — polarization and reddening are two views of one grain population.

Physics of the portrait

The map of the Galactic magnetic field begins with a number that should not have existed. When starlight polarization was discovered in 1949, the natural reading was that dust could align — and if dust aligns, a magnetic field is doing the aligning. Two people saw at once what to do with that: use the scatter in the polarization angles to weigh the field itself.

Origins · 1 minSubrahmanyan Chandrasekhar and how this was discovered

Subrahmanyan Chandrasekhar

1910–1995 · Galactic magnetic fields and the DCF method

Subrahmanyan Chandrasekhar — "Chandra" — is best remembered for the white-dwarf mass limit he found at nineteen on the boat from India to Cambridge, but his reach spanned nearly every domain of theoretical astrophysics, each treated in a self-contained monograph before he moved on. In 1953, working with Enrico Fermi at Chicago, he took up a problem opened four years earlier by Hiltner and Hall's discovery of interstellar polarization: if starlight is polarized by magnetically aligned dust, how strong is the Galactic field? Their two 1953 papers argued from the alignment of polarization vectors along the spiral arms that the field runs largely along the arms at a few microgauss, and — decisively — that the small but real dispersion of those angles about the smooth mean measures the field's stiffness against turbulent bending. Equating turbulent kinetic energy to the energy of the rippled field lines, they wrote down the Davis–Chandrasekhar–Fermi relation, $B_{\mathrm{pos}}\approx Q\sqrt{4\pi\rho}\,\sigma_v/\sigma_\theta$ , still used seventy years later to weigh magnetic support in star-forming clouds. It was characteristic Chandra: take a messy observation, find the one energy balance that controls it, leave behind a clean quantitative tool. He shared the 1983 Nobel Prize for the stellar-structure work and gave his name to NASA's Chandra X-ray Observatory — but the DCF method is his quiet fingerprint on every modern polarization map of the interstellar field.

History

1949

— Hiltner and Hall independently discover the linear polarization of starlight

1951

— Davis and Greenstein propose paramagnetic alignment of spinning grains by the Galactic field

1953

— Chandrasekhar and Fermi weigh the field from polarization-angle dispersion

1975

— Serkowski, Mathewson and Ford fit the empirical wavelength law of interstellar polarization

2015

— Planck maps polarized dust emission across the entire sky

Going further

Texts. Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §12.1 (dust, extinction, polarization) · Binney & Tremaine, Galactic Dynamics, 2nd ed. (interstellar medium and magnetic fields) · Draine, Physics of the Interstellar and Intergalactic Medium, Ch. 21–23 (grain alignment and polarization) · Dodelson, Modern Cosmology (CMB polarization and foregrounds) · Ryden, Introduction to Cosmology, 2nd ed. (foreground context). The key numerical anchors in this article are: the Serkowski parameters $\lambda_{\max}\approx0.55\,\mu\mathrm{m}$ , $K\approx1.66\,\lambda_{\max}$ , and the efficiency ceiling $p_V/A_V\lesssim3\%\ \mathrm{mag^{-1}}$ ; the DCF calibration factor $Q\approx0.5$ from MHD simulations; the critical mass-to-flux ratio $(2\pi\sqrt{G})^{-1}$ and its observational form $\mu\approx7.6\times10^{-21}N(\mathrm{H_2})/B[\mu\mathrm{G}]$ ; the DCF worked result $B_{\mathrm{pos}}\approx55\,\mu\mathrm{G}$ for $n=10^4\,\mathrm{cm^{-3}}$ , $\sigma_v=0.3\,\mathrm{km\,s^{-1}}$ , $\sigma_\theta=12^\circ$ ; the Serkowski ratios $p(\mathrm{B})/p_{\max}=0.96$ and $p(\mathrm{K})/p_{\max}=0.17$ ; and $\mu\approx1.4$ for $N=10^{22}\,\mathrm{cm^{-2}}$ .

Prerequisites. Interstellar Dust Grains · Interstellar Extinction and Reddening · Thermal Dust Emission and Infrared Radiation · Phases of the Interstellar Medium. Leads to. Turbulence in the Interstellar Medium · Virial Theorem for Molecular Clouds · The Jeans Instability and Jeans Mass · Molecular Clouds and Giant Molecular Clouds.

Ask the tutor: Want to walk through the DCF calculation for a real cloud, or see exactly why aligned grains polarize starlight parallel to the field but emit perpendicular?

Why is transmitted starlight polarized parallel to the field but emission perpendicular?

Both come from the same grain geometry seen two ways. Aligned grains sit with their long axis perpendicular to $\mathbf{B}$ and absorb most strongly the light whose electric vector lies along that long axis. So the component perpendicular to $\mathbf{B}$ is preferentially removed, and what survives is polarized parallel to $\mathbf{B}$ — the absorption channel. The same long axis that absorbs best also emits best (Kirchhoff's law), so the grain's own far-infrared glow comes out polarized along the long axis, perpendicular to $\mathbf{B}$ . The two channels are always $90^\circ$ apart, so before comparing a submillimetre emission map to an optical starlight map you must rotate one by a right angle.

What actually aligns the grains — the magnetic force on the dust?

Not directly; the field is far too weak to torque a neutral grain into place by brute force. Alignment is two-stage. A spinning grain becomes slightly magnetized simply by rotating (the Barnett effect), and that moment makes its spin axis precess rapidly around $\mathbf{B}$ , so the field defines the reference direction. Then something must remove the grain's wobble so its angular momentum settles along that axis. The classical answer, paramagnetic Davis–Greenstein dissipation, is far too slow for real $0.1\,\mu\mathrm{m}$ grains. The modern answer is radiative alignment torques: anisotropic starlight spins the irregular grain to suprathermal rotation, stiffening it against collisions and locking its alignment to the field. The field sets the direction; starlight supplies the muscle.

If polarization only gives the plane-of-sky field, is the method even useful?

Very. Polarimetry is blind to the line-of-sight component and carries a $180^\circ$ ambiguity — it gives an axis, not a vector, and only the projected part. But the plane-of-sky orientation is exactly what maps field geometry across the sky: how filaments align with the field, how it wraps around expanding shells, where turbulence has tangled it. And the angle dispersion, through DCF, delivers a genuine field strength. Complementary probes fill the gaps — Zeeman splitting measures the line-of-sight field strength directly, Faraday rotation the line-of-sight field weighted by electron density. Polarimetry is the wide-field mapping tool in that toolkit.

How reliable is a DCF field strength — could it be off by a large factor?

The raw energy-balance estimate systematically overshoots, sometimes by factors of a few, which is why every serious application multiplies by the calibration factor $Q\approx0.5$ from MHD simulations. The method assumes the angle scatter is a single population of sub-Alfvénic turbulence on a uniform mean field, so the small-angle approximation holds. When the field is well ordered and the turbulence mild — as in the worked example, Alfvén Mach number near 0.4 — DCF is good to a factor of order two. It degrades where the mean field curves within the beam or where the angle dispersion is large. Modern variants (structure-function and Hildebrand methods) separate turbulent scatter from field curvature to tighten it.

Why does the polarization fraction drop inside dense cores?

This is the "polarization hole," a direct consequence of how grains are aligned. Radiative torques need anisotropic starlight to spin grains up. Deep inside a dense core that starlight is heavily extinguished — the very dust that makes the core opaque shields its interior from the radiation that would align its grains. With the torque switched off, grains lose their alignment to gas collisions and the polarization fraction falls even though the dust column, and the field, is highest there. The hole is therefore not a magnetic effect but a test of alignment physics: it is among the strongest evidence that radiative torques, not a radiation-independent mechanism, dominate real grain alignment.

Why is interstellar dust polarization a problem for cosmology?

Because it looks just enough like the signal cosmologists hunt for. Primordial gravitational waves from inflation would appear as a faint curl pattern — B-modes — in the polarization of the cosmic microwave background. But aligned grains also emit polarized light, producing their own B-mode pattern at exactly the microwave frequencies CMB experiments observe. Even in the cleanest sky patches the dust foreground is comparable to or larger than any plausible primordial signal — which is why the 2014 BICEP2 B-mode claim collapsed once Planck's dust maps were folded in. Removing it demands measuring the dust's frequency dependence and spatial pattern precisely, which requires the microphysics in this article. The spinning silicate grain is, improbably, a gatekeeper for observational cosmology.