A bigger outer region than expected
A preprint analysis of lensed quasars reports that the emitting region measured in the near-infrared is substantially larger than a thin-disk model predicts. The analysis reached out to 14 light-days and found that the region at that scale was about three times larger than the thin-disk prediction. It also found a wavelength-size pattern that did not match the profile expected from the theory. The paper is a preprint posted as arXiv:2608.26039v1 on 26 August 2026.
At rest-frame 5007 Å, the joint estimate for the characteristic source size was 13.3 light-days. The reported uncertainty was asymmetric, at +2.9 and −2.7 light-days. The 5007 Å reference is the wavelength of one of the narrow [O III] emission lines used in the near-infrared spectra.
A wavelength pattern that misses the benchmark
The comparison becomes more precise when size is tracked across wavelengths. To summarize that change, the analysis uses a wavelength-size slope, p—a number describing how the inferred source size changes with wavelength. The near-infrared narrow-line measurements gave p = 0.68 ± 0.23, while the thin-disk benchmark is p = 4/3. The authors report that the benchmark was rejected at 98% confidence, equivalent to 2.33σ.
The reported mismatch was not confined to a single infrared comparison. The authors state that microlensing-based sizes were significantly larger than thin-disk predictions from X-rays through 5000 Å, and that their wavelength scaling did not fit the thin-disk profile across that range.
How the estimate was made
The sample contained 21 lensed-quasar image-pairs from seven lens systems. The data were homogeneous HST/WFC3 near-infrared observations made with the F140W filter and the G102 and G141 grisms. The spectra included narrow [Ne III] lines at 3870 and 3969 Å and [O III] lines at 4960 and 5007 Å.
Single-epoch spectroscopy separated continuum emission prone to microlensing from narrow emission lines treated as unaffected. Compared with a reference image, the difference between the continuum and line measurements defined the microlensing amplitude. In effect, the narrow-line measurements supplied the no-microlensing baseline for the estimate.
To convert those amplitudes into source sizes, the analysis convolved Gaussian source profiles with simulated magnification maps to produce magnification distributions. Individual probabilities for the three image-pairs in each of the seven lenses were multiplied into a joint likelihood for the sample. The two [Ne III] measurements were then iteratively rescaled to the [O III] reference wavelength of 5007 Å using a power-law size-wavelength relation. The first iteration used p = 4/3, and later iterations used the previous estimate until the result converged.
Why the result is conditional
The absolute sizes are conditional on the assumptions used to build the microlensing maps. The simulations adopted a 10% stellar surface-mass fraction and microlenses with a mass of 1 solar mass, averaging five map realizations per image. The authors note a mass-size degeneracy: 0.2-solar-mass microlenses would match the thin-disk size at 5007 Å, while an average stellar mass of about 0.1 solar masses or less would be needed at the shorter comparison wavelength.
One technical detail also needs clarification: the method describes the short-wavelength reference as 1026 Å, while the mass-size discussion gives 1028 Å.
Paper data and sources
Original title: Accretion Disk Sizes and Temperature Profiles in Lensed Quasars: NIR Microlensing Challenges Thin Disk Theory
Authors: V. Motta, E. Mediavilla
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text