How faithfully can astronomers map a star’s magnetic field?

Testing a map against a known answer

How much should astronomers trust a magnetic map of a distant star? Its surface usually cannot be resolved directly, so researchers work backwards from its spectrum: the distribution of light across wavelengths. A magnetic field splits spectral lines and changes their polarisation. Stellar rotation shifts the signals from different surface regions, providing clues to where magnetic structures lie.

The reconstruction technique, called Zeeman–Doppler imaging, or ZDI, must turn those overlapping clues into a map. Different maps can fit imperfect observations, so the calculation also favours simpler structures unless the data demand more detail.

O. Kochukhov, A. Al Kanale and J. Braithwaite tested how faithfully modern ZDI recovers known magnetic configurations. Their study is available as an arXiv preprint; the supplied manuscript records acceptance for publication in Astronomy & Astrophysics. It reports numerical experiments, not new telescope observations.

Building stars with known magnetic fields

The team used three surface maps from three-dimensional magnetohydrodynamic simulations, which model the interaction of magnetic fields with electrically conducting stellar material. These simulations produce stable configurations relevant to the long-lived, or fossil, fields of magnetic early-type stars.

One map resembled a distorted dipole, the broad two-pole pattern familiar from a bar magnet. Another had a strong toroidal contribution, describing field that wraps around within spherical surfaces. The third contained a more complicated arrangement of regions with alternating magnetic polarity.

All three maps were scaled to an average surface field of 5 kilogauss. The researchers calculated the spectra these model stars would produce, added random noise, and asked their imaging codes to reconstruct the fields without supplying the original maps.

The observing setup was deliberately favourable: 20 evenly spaced rotational phases, a viewing inclination of 60 degrees, moderate rotational broadening, and exactly known stellar parameters. These choices isolate difficulties caused by magnetic complexity rather than gaps in observations or mistaken assumptions about the star.

What extra polarisation reveals

The main comparison concerned how much information the reconstruction received. Stokes I describes light intensity, V circular polarisation, and Q and U two components of linear polarisation. The team compared inversions using all four with those using only I and V.

With all four parameters and three individual iron lines, mean errors in the magnetic vector components ranged from 1.8% to 6.9% across the reasonably well-observed surface. Crucially, those percentages are measured against each model’s maximum field strength, not against the local field at every point. Mean errors in field direction, measured relative to the surface normal, ranged from 1.9 to 8.1 degrees.

Removing linear polarisation made the most complicated map substantially harder to recover. Its radial and north–south field components had mean errors of about 15% of the maximum field, while its mean directional error reached about 17 degrees. The recovered field strength could remain fairly accurate even when the direction was misplaced. A convincing strength map therefore does not guarantee an equally reliable vector map.

Combining lines without losing the picture

The researchers also tested least-squares deconvolution, which combines information from many spectral lines into a higher-quality mean profile. Their iron-line mask included 968 lines. Importantly, they applied the same extraction procedure to the simulated observations and the spectra used in reconstruction, rather than treating the mean profile as an ordinary single line.

These inversions performed comparably overall. For the most complicated map using only I and V, the mean radial-field error fell from about 15% to 11%, and the directional error from about 17 to 12 degrees. But the combined profiles were not consistently better for simpler geometries.

Full-polarisation inversions also broadly recovered the division between poloidal and toroidal surface magnetic energy. The discrepancies reached 7.3 percentage points in the individual-line tests and 4.7 points with combined profiles. Imposing an additional relationship between poloidal components helped some weaker reconstructions, but was neutral or harmful elsewhere—even though the input maps satisfied that relationship.

A benchmark, not a universal guarantee

The reported accuracy depends on visibility: poorly seen latitudes had larger errors, and an unseen polar region cannot be directly constrained. The underlying magnetic simulations also omit rotation and stellar winds and do not predict absolute field strengths.

Nor did these tests reproduce every complication of real stars. Iron abundance was uniform, so small artificial abundance variations measured limited cross-talk in that setup, not success at recovering realistic chemical spots.

The result is a useful best-case benchmark for the tested methods and fossil-field geometries. It supports their ability to recover more than simple dipoles, while showing why complete polarisation measurements matter when magnetic structure becomes complicated.