Separating faint light between galaxies from telescope-scattered glow

Diffuse starlight surrounding galaxies, groups and clusters offers clues to how these systems assembled through mergers and the stripping of stars from galaxies. Astronomers call this component intra-halo light, or IHL. Measuring it is difficult because light from bright stars and galaxies spreads into faint, extended wings that can masquerade as genuine diffuse emission.

Lucía P. Garate-Nuñez, Aaron S. G. Robotham, Sabine Bellstedt and Luke J. M. Davies present a correction for this problem alongside a method for measuring IHL. The study’s second arXiv revision records acceptance by Monthly Notices of the Royal Astronomical Society on 5 November 2026.

Correcting the scattered light

The point spread function, or PSF, describes how the atmosphere and telescope distribute a source’s light across an image. Its faint outer wings matter especially when the signal being measured is itself extremely dim. Simply masking bright objects does not necessarily remove their scattered light beyond the masks.

The researchers’ correction begins by identifying sources and assigning pixels to them. They temporarily set pixels outside those source regions to zero, then apply a PSF model to estimate how much additional light spreads beyond each region. Comparing the enclosed light before and after this operation gives an approximate correction factor for each source.

They use those factors to increase the light assigned to the detected sources. Applying the PSF model to this adjusted source image then estimates the scattered light contaminating surrounding pixels, which is subtracted from the original background regions.

This is not an exact reconstruction of the original image. Because the observed sources are already blurred, the correction factors are approximate. The procedure also does not explicitly model scattering from diffuse IHL outside the detected source regions.

To measure the remaining IHL, the team masks detected objects and fits a smooth exponential profile to the unmasked light using ProFit. Bayesian sampling estimates the model parameters and their fitting uncertainties. The fitted profile also estimates diffuse light projected beneath the galaxy masks, rather than discarding that contribution.

What the controlled tests show

The team built 5,440 mock images from 544 galaxy groups and clusters, combining real Hyper Suprime-Cam galaxy pixels with catalogue information from GAMA and WAVES. Stars were inserted analytically at their observed positions and brightnesses. Each system received ten versions of a circular exponential IHL component, spanning fractions from 1% to 50% of the combined group-galaxy and IHL light.

Without PSF correction, the faintest injected components produced particularly severe errors. At a true IHL fraction of 1%, the median fitted flux was about 84.4 times the injected flux, while the effective radius—the radius enclosing half the model’s light—was about 9.1 times the injected value.

After correction, both median ratios fell to approximately 0.89. In other words, the method greatly reduced contamination but underestimated brightness and size by about 11% in that regime.

At a 50% IHL fraction, the uncorrected flux ratio was approximately 0.87, but the effective radius remained about twice its true value. Correction brought the flux ratio to about 0.95 and the radius ratio to about 1.01. Thus, reasonably recovering total brightness did not guarantee an accurate size.

These are controlled results, not universal correction factors. The main tests fitted the same circular exponential form used to inject the IHL, deliberately reducing complications from mismatched models.

One observed group, several measurement choices

The researchers also analysed UltraDeep Hyper Suprime-Cam images of GAMA group G400138, which has six observed members and lies at a redshift of approximately 0.21.

They compared circular and elliptical exponential fits with a surface-brightness-threshold approach supplemented by an inward extrapolation beneath the central masks. Their preferred estimates also increased the galaxy-light denominator to account for faint members below the spectroscopic limit.

The reported combined IHL fractions were approximately 19% in the g band, 8% in r and 6% in i, with respective uncertainty ranges of +9/−1, +6/−2 and +4/−2 percentage points. These measurements describe one system, not a population average.

Remaining limits

The correction mildly oversubtracts diffuse light. Its output depends on source segmentation and cannot recover light already removed during earlier sky subtraction. Fitting uncertainties alone omit additional systematic errors.

The mock galaxies also retain existing observational blurring before another PSF convolution is applied. Their analytical stars and smooth IHL provide a more controlled setting than real observations, where irregular structures and imperfect source identification complicate measurements.

The result is a substantial improvement within those limits: scattered light must be addressed explicitly, but correcting it does not eliminate uncertainty in what astronomers count as intra-halo light.