Magnetic Filaments: Vast Glowing Threads Above the Galactic Centre

Where they live, and how absurd their shape is

The inner few hundred parsecs of the Milky Way - the Central Molecular Zone (CMZ) - is a violent, crowded place: roughly 3 x 10⁷ solar masses of dense molecular gas, three of the Galaxy's most massive young clusters (the Arches, the Quintuplet and the Central cluster), and the 4.3 x 10⁶ solar-mass black hole Sagittarius A* at 8.2 kpc from Earth. Threaded through all of that are the filaments.

Their geometry is what makes them extraordinary. The longest run for tens of parsecs - the record-holders approach 150 light-years, about 45 pc. Their widths are barely resolved even at the 4-arcsecond resolution of MeerKAT's Galactic Centre mosaic, which at 8.2 kpc corresponds to only 0.16 pc; true widths are of order 0.1-0.3 pc, less than a light-year. That is a length-to-width ratio of several hundred to one, held straight across a medium that is turbulent, clumpy and swept by bulk cloud motions of order 100 km/s.

They also come in bundles. The Radio Arc at Galactic longitude l ~ 0.18 degrees is a bank of a dozen or more parallel threads spanning roughly 40 pc, standing nearly perpendicular to the plane and bending at its northern end into the Galactic Centre Lobe. Filaments elsewhere are mostly isolated, but the grouped ones are strikingly evenly spaced, like the strings of an instrument.

The physics of the glow: synchrotron radiation as a magnetometer

Filaments are visible at radio wavelengths and essentially nowhere else, because they shine by synchrotron radiation. An electron with Lorentz factor gamma spiralling around a field B radiates near the critical frequency nu_c ~ gamma² x (eB / 2*pi*m_e*c), and the non-relativistic gyrofrequency eB/2*pi*m_e*c is 2.8 MHz per gauss. At MeerKAT's 1.28 GHz in a 100 microgauss field that gyrofrequency is only ~280 Hz, so gamma² ~ 3 x 10⁶ and gamma ~ 1,700: the radiating particles are electrons of about 1 GeV.

Three hard constraints follow:

• The spectrum encodes the particle population. Filaments have power-law flux densities S proportional to nu^alpha with alpha typically between -0.5 and -1. Since alpha = -(p-1)/2 for an electron distribution N(E) proportional to E^-p, this implies p ~ 2-3, the signature of shock or reconnection acceleration. Thermal plasma gives flat or rising spectra instead.
• The polarisation encodes the field direction. Synchrotron emission from an ordered field is linearly polarised with the electric vector perpendicular to the projected field. Once Faraday effects are removed, filaments show intrinsic polarisation reaching ~30-70% at 8-10 GHz, and the inferred field runs along the filament's long axis. This is the decisive measurement: they are not chance alignments of clouds, they are illuminated field lines.
• The cooling time sets a clock. The synchrotron lifetime is t = 3*m_e*c / (4*sigma_T*U_B*gamma). At 100 microgauss (U_B ~ 4 x 10^-10 erg/cm³) a GeV electron lives ~1 million years; at 1 milligauss the same 1.28 GHz emission comes from gamma ~ 550 electrons surviving only ~5 x 10⁴ years. Strong fields therefore demand local, ongoing injection along the whole length of every filament.

How strong is the field? Two arguments that disagree

Two ways of weighing the field have disagreed by an order of magnitude for thirty years.

Minimum-energy (equipartition) analysis assumes the energy densities in relativistic particles and in the field are comparable, minimising the total energy needed for the observed brightness. For typical filaments this gives B ~ 50-200 microgauss - the least assumption-laden estimate, but a lower bound in disguise, since a far stronger field would give the same emission from far fewer electrons.

The rigidity argument, pressed hardest by Mark Morris, is more demanding. A filament stays straight only if magnetic pressure resists the ram pressure of gas ploughing past it: B²/8*pi > rho*v²/2, so B > v*sqrt(4*pi*rho). For ambient gas of n ~ 10 cm^-3 at 100 km/s that gives B ~ 170 microgauss; for denser, faster flows near CMZ clouds (n ~ 100 cm^-3, 200 km/s) it climbs to ~1 milligauss. Zeeman measurements in dense CMZ clouds do find ~0.1-3 mG. Against this, LaRosa et al. (2005) argued from the diffuse non-thermal emission of the whole central region - blindingly bright if milligauss fields filled hundreds of parsecs - for a large-scale field nearer 10 microgauss, filaments being locally amplified flux tubes. The stakes are large: a milligauss field filling the inner 200 pc stores ~10⁵⁵ erg.

Faraday rotation supplies a third handle, measuring the line-of-sight field weighted by electron density: RM = 0.812 x integral of n_e*B_parallel*dl (rad/m², cm^-3, microgauss, pc). Filaments show |RM| of hundreds to a few thousand rad/m², exceeded only by Sgr A* itself at about -5 x 10⁵ rad/m². Below ~5 GHz such rotation measures depolarise the filaments by smearing polarisation angles across the beam, which is why polarimetry is done at centimetre wavelengths.

From four filaments to a thousand: the observational history

The discovery paper is Yusef-Zadeh, Morris & Chance (1984, Nature 310, 557), 'Large, highly organized radio structures near the Galactic Centre'. Farhad Yusef-Zadeh was a graduate student working with new Very Large Array maps; out came the Radio Arc, unlike anything in the catalogue of Galactic radio sources. VLA polarimetry at 6 and 3.6 cm over the following decade established the non-thermal spectrum and the field-aligned polarisation, and added isolated filaments: G359.1-0.2, 'the Snake', a ~30 pc thread with two sharp kinks, and the long Sgr C filament (G359.45-0.06).

The population exploded in 2022. MeerKAT - the 64-dish, 13.5 m SARAO array in South Africa's Karoo - produced a 1.28 GHz mosaic of 6.5 square degrees from about 200 hours at 4-arcsecond resolution (Heywood et al. 2022, ApJ 925, 165); the same data had already revealed the ~430 pc bipolar Galactic Centre radio bubbles (Heywood et al. 2019, Nature). Running a filament-finder over it, Yusef-Zadeh and colleagues catalogued close to a thousand filaments - roughly a tenfold increase over the ~100 candidates known before - enabling real statistics: length and spectral-index distributions, field estimates near ~100 microgauss, and a tendency to appear in quasi-regularly spaced groups.

Individual objects earned models. The 'radio harps' - short filaments crossed by evenly spaced knots - were modelled by Thomas, Pfrommer & Ensslin (2020) as the track of a compact object moving across a bundle of field lines, injecting cosmic-ray electrons onto each in turn; the spreading of the knots then measures cosmic-ray transport directly. A few filaments have X-ray counterparts from Chandra and XMM-Newton, notably G0.13-0.11.

The second population: a fossil of an outflow?

MeerKAT did more than multiply the known filaments - it exposed a population that had been hiding. In Yusef-Zadeh et al. (2023, ApJL 949, L31), a distinct set was identified that is not vertical at all. These lie nearly parallel to the Galactic plane, are short (roughly 5-10 pc rather than tens of parsecs), point radially away from Sgr A*, and - the striking part - sit preferentially on one side of the plane, mostly to the south, within about 20 arcminutes (~50 pc) of the black hole.

They are still synchrotron sources with field-aligned magnetic fields, but their geometry is that of material combed outward from a central engine. The proposed reading is a fossil of a bipolar outflow or jet from Sgr A* a few million years ago - of order 6 Myr - that stretched ambient field lines into radial spokes and left relativistic electrons on them. The timescale is not arbitrary: it matches the estimated age of the Galactic Centre radio bubbles (~7 Myr) and, on far larger scales, the 10 kpc Fermi Bubbles found by Su, Slatyer & Finkbeiner (2010) and their X-ray counterparts mapped by eROSITA (Predehl et al. 2020). Sgr A* radiates at ~10^-9 of its Eddington luminosity today, but evidence that it was not always so keeps accumulating.

A related puzzle is the field's two-faced geometry. Far-infrared and submillimetre dust polarimetry (JCMT/SCUBA, more recently SOFIA's HAWC+) shows the field inside dense CMZ clouds lying mostly parallel to the plane - toroidal - while the low-density medium traced by the filaments is poloidal. The switch appears where gas pressure and gravity overwhelm magnetic tension.

What they are not: separating filaments from look-alikes

Almost every mistake made about these objects is a failure to distinguish a synchrotron thread from something else that merely looks linear on a radio map.

• The Arched Filaments are not filaments in this sense. Sitting right beside the Radio Arc, they are curved arcs and they are thermal: flat or slightly rising spectra, radio recombination lines, essentially unpolarised. They are the ionised skins of molecular clouds lit by the ~10⁴ solar-mass Arches cluster. Spectral index plus recombination lines settles it in one observation.
• Supernova remnant filaments such as those in Sgr A East or G359.1-0.5 are non-thermal too, but trace shock-compressed field in a closed, limb-brightened shell; curvature and shock association give them away.
• Infrared and submillimetre dust filaments mapped by Herschel and ALMA - the Brick, the CMZ's molecular streams - are cold gas seen in thermal dust continuum, not synchrotron.
• Solar filaments (prominences seen against the disc) share a name and nothing else: cool, chromospheric-temperature plasma suspended in the corona by magnetic fields, some 10¹⁰ times smaller.
• Exotic proposals have failed. Early suggestions of cosmic strings do not survive the polarisation data - a cosmic string would not produce a field-aligned GeV-electron synchrotron spectrum, and the filaments' interactions with molecular clouds are ordinary astrophysics.

One misconception worth killing: you will never see these optically. Visual extinction toward the Galactic Centre is roughly 30 magnitudes - a suppression of ~10¹² - so every image of the filaments is a radio image, colour-mapped for human eyes.

Open questions

Forty years on, the fundamental question is unsolved: are the filaments the field, or merely where the field is lit?

• Global versus local field. In the Morris-Serabyn picture a pervasive, milligauss, largely poloidal field - possibly a compressed relic of a primordial or protogalactic field - threads the whole CMZ, and filaments mark where cosmic-ray electrons happen to be injected. In the LaRosa-Shore picture the ambient field is weak (~10 microgauss) and each filament is a locally amplified flux tube, perhaps the magnetohydrodynamic wake of a molecular cloud moving through it. The two make opposite predictions for the diffuse low-frequency emission; deep metre-wavelength imaging with MeerKAT and eventually the SKA is the discriminating measurement.
• Where do the electrons come from? Short cooling times demand distributed injection. Candidates include shocks and winds from the Arches and Quintuplet clusters, magnetic reconnection at cloud-field interfaces (Serabyn & Morris 1994), pulsars feeding particles onto field lines (the radio harps), and past Sgr A* activity.
• Why here and nowhere else? No comparable population is known elsewhere in the Milky Way - whether because the nucleus has a genuinely unique field geometry, or because nowhere else combines high cosmic-ray density, an ordered field and this much observing time.
• What sets the spacing and the widths? Quasi-regular spacing within bundles and sub-parsec widths both demand a characteristic length scale - resistive, injection or instability - that no model has convincingly supplied.

The filaments remain the most direct picture anyone has of an ordered magnetic field in a galactic nucleus, and that we still cannot say whether they represent 10⁵² or 10⁵⁵ erg measures how much of galactic-centre magnetism is still guesswork.

Straight radio structures near the Galactic Centre and how they are told apart

Structure	Geometry	Radio signature	Physical interpretation
Vertical non-thermal filaments (NTFs) - e.g. the Radio Arc, the Snake, Sgr C	Straight, mostly perpendicular to the Galactic plane, 10-45 pc long, <0.3 pc wide	Steep synchrotron spectrum (alpha ~ -0.5 to -1); linear polarisation up to ~30-70% at 8-10 GHz; B along the filament	Relativistic electrons on an ordered, largely poloidal field threading the inner few hundred parsecs
Radial / near-horizontal filaments (Yusef-Zadeh et al. 2023)	Short (~5-10 pc), nearly parallel to the plane, pointing radially away from Sgr A*, one-sided	Synchrotron, polarised, B again along the filament, but distinct length and spectral statistics	Possible fossil of a bipolar outflow or jet from Sgr A* a few million years ago
Arched Filaments (G0.07+0.04)	Curved arcs beside the Radio Arc	Flat free-free spectrum (alpha ~ -0.1 when optically thin, rising toward +2 where optically thick), radio recombination lines, essentially unpolarised	Thermal ionised gas - molecular cloud surfaces photoionised by the Arches cluster
Supernova remnant filaments (Sgr A East, G359.1-0.5)	Closed or curved shells, tens of parsecs	Synchrotron with alpha ~ -0.5, limb-brightened shell morphology	Shock-compressed field at an expanding blast wave, not a field-aligned thread
Far-infrared / submillimetre dust filaments (Herschel, SOFIA)	Filamentary molecular clouds within the Central Molecular Zone	Thermal dust continuum, polarisation from aligned grains implying a field mostly parallel to the plane	Dense gas whose internal field has been sheared into a toroidal geometry