Cleaner SPHEREx maps reveal interstellar dust chemistry

Why these infrared features matter

Polycyclic aromatic hydrocarbons, or PAHs, are carbon-bearing molecules whose infrared glow helps astronomers study the material between stars. Different emission features respond to molecular structure, size, electrical charge and the surrounding radiation field.

The study focuses on features near 3.3 and 3.4 micrometers. The first mainly traces aromatic carbon–hydrogen bonds associated with carbon rings; the second is associated with aliphatic bonds, including attached chemical groups. Comparing their brightness can reveal chemical variation, although interpreting that ratio depends on how overlapping emission is separated.

The researchers also map Paschen-alpha, an infrared hydrogen emission line that traces ionized gas. Together, these signals distinguish different parts of star-forming environments.

The problem with assembling the observations

NASA’s SPHEREx uses linear variable filters: the wavelength transmitted changes gradually across each detector. Measurements of the same sky position therefore arrive with different wavelength responses.

Simply grouping measurements into wavelength bands and averaging them can create stripes, especially around emission features. These patterns reflect the observing geometry rather than real celestial structures.

The team’s method, LVFit, instead predicts what each individual detector measurement should record from a model containing molecular emission, hydrogen emission where needed, and a smooth underlying continuum. It then adjusts the components’ strengths to fit the observations, accounting for each measurement’s spectral response. Outlying measurements receive less weight.

The analysis uses approximately 16.5 months of observations through September 7, 2026, reorganized into a searchable pixel-level database. Additional screening removes affected detector edges associated with scattered moonlight and streaks from moving objects.

Learning a better molecular model

Simple mathematical profiles for the PAH features still produced stripes. The researchers therefore started with higher-resolution James Webb Space Telescope spectra of the Orion Bar and learned smooth corrections from five SPHEREx fields. Five additional fields helped select the corrections’ settings.

This matters because an inaccurate assumed emission profile can itself generate stripes, even with forward modeling. The improvement is substantial, but not absolute: across the training fields and three bright comparison fields, the measured sampling-related pattern in combined 3.3- and 3.4-micrometer brightness fell from 0.9% to 0.6% of local brightness. Residual patterns remained statistically detectable in feature-ratio maps, particularly in faint clouds.

The output maps have an angular resolution of about 8.5 arcseconds on 6.4-arcsecond pixels. Their spectral resolving power near the PAH features is about 35, compared with roughly 2 for WISE’s broad-band measurement. That comparison concerns wavelength discrimination, not a corresponding improvement in image sharpness.

What the maps reveal

In the Rosette nebula, the maps resolve structures about 0.06 parsecs across at the adopted distance. Paschen-alpha traces ionized gas inside the surrounding PAH-emitting region.

The 3.4-to-3.3-micrometer ratio is lower inside the ionized cavity and higher around dense molecular material. The authors interpret this as consistent with ultraviolet radiation preferentially destroying aliphatic groups unless dense material shields them. It is an interpretation, not a direct measurement of the destruction process.

In Andromeda, the PAH maps recover structure in its rings. In Triangulum, the ratio of 3.3-micrometer emission to stellar-continuum-corrected Spitzer 8-micrometer emission rises outward. The reported slope is 0.048 ± 0.003 dex per kiloparsec, with a separate background-related systematic shift of about 0.01 dex per kiloparsec. Changes in chemical abundance or the radiation field could contribute; the study does not uniquely identify the cause.

Reconstructed gaps and other limits

Stars and poorly modeled pixels are masked, then filled using a statistical model learned from surrounding spatial patterns. These values are predictions, not recovered observations. The released masks identify them. Masked fractions range from 2.3% in Triangulum to 34.5% in crowded Cygnus X, where genuine compact emission can also be removed.

Unremoved starlight probably explains apparent PAH emission in Andromeda’s nucleus and its companion M110. Other limitations include fixed molecular templates, incomplete treatment of the instrument’s image-blurring response, background errors and correlated uncertainties. The team empirically recalibrates its reported errors rather than relying solely on formal fit uncertainties.

Sensitivity, scope and publication status

The best reported 3.3-micrometer sensitivity is 2.7 × 10⁻⁸ erg s⁻¹ cm⁻² sr⁻¹ per map pixel in the deeply observed north ecliptic pole field—not uniformly across the sky. At matched angular resolution, those polar maps already outperform WISE as a PAH tracer, while several other fields remain noisier.

A two-year mission extension with a more even scanning strategy could bring sensitivity close to WISE’s across the sky; that is a projection. This paper demonstrates selected regions, not a completed all-sky LVFit map. Maps and analysis code are released. The findings appear in an arXiv preprint posted November 9, 2026, and remain provisional.