A nova’s fading visible light could guide the search for gamma rays

Two kinds of light, one possible engine

Nova eruptions occur on white dwarfs in binary star systems. In a classical nova, hydrogen-rich material accumulated from a companion undergoes a thermonuclear runaway. The eruption expels material, creating conditions in which shocks can accelerate particles and generate high-energy radiation.

Fermi’s Large Area Telescope, or LAT, first detected gamma rays from a nova in 2010. One leading explanation involves faster ejecta catching slower material expelled earlier. In systems with an evolved companion, ejecta can also collide with the companion’s surrounding wind.

These shocks might contribute to visible light as well. In the proposed picture, thermal X-rays produced by shocks are absorbed and re-emitted at optical wavelengths, while gamma rays escape more directly. Corresponding changes in optical and gamma-ray brightness would therefore be expected, although they would not alone establish how much visible light comes from shocks.

Choosing an analysis window

Owen K. Henry and colleagues examined 26 eruptions on a list of Fermi-LAT-detected novae spanning August 2008 through June 2024. They analyzed photons with energies from 50 million to 300 billion electronvolts, modeling nearby gamma-ray sources and diffuse background emission.

For each nova, they tested integration windows ranging from half a day to 1,500 days, generally starting at the listed gamma-ray discovery time. Their preferred window was the one that maximized a statistical measure called the test statistic, which compares models with and without a source at the nova’s position.

This is an optimized detection interval, not necessarily the physical duration of gamma-ray production. Weak sources often had broad ranges of windows giving similar results, making their preferred times less sharply defined.

Background modeling also mattered. The team examined 11 novae near unidentified sources in Fermi’s catalog, checking whether those catalog sources remained significant outside the eruptions. They identified three potential nova associations—V1324 Sco, V5855 Sgr and V549 Vel—and recommended removing the corresponding catalog entries from the background model when analyzing those eruptions, rather than treating them as separate background sources.

An optical guide, with substantial scatter

The researchers compared the gamma-ray windows with visible-light observations from the American Association of Variable Star Observers. They developed an open-source tool, nova-times, with support from the LSST Discovery Alliance’s Project Dovetail. Its machine-learning fitting method estimates brightness evolution from irregularly sampled observations and measures optical decline times.

The main candidate guide was t3: the time needed for a nova’s V-band brightness to decline by three magnitudes from maximum. The optical fading measured across the optimized gamma-ray windows clustered around three magnitudes, with the middle half of the distribution extending approximately from two to four magnitudes.

That makes t3 a potentially useful starting point, not a precise prescription. The scatter was substantial, and a four-magnitude decline time performed similarly to t3 in the authors’ comparison of distances from a one-to-one relationship. A two-magnitude decline time performed less well.

The analysis also strengthened the reported gamma-ray signal from V679 Car, previously considered marginal. The authors report a detection above five sigma with their windowing procedure. That reported significance should not be read as a separately demonstrated, search-wide false-alarm probability accounting for the full window-selection procedure.

Timing clues and remaining limits

For a brighter subset, the team compared optical and gamma-ray light curves, including their rates of change. Most results were compatible with no lag, but uncertainties were considerable. Several approximately one-day offsets remained within the respective time-step uncertainties.

For the rapidly fading nova V1674 Her, one test suggested that optical emission lagged by 4.56 hours, while the other two showed no lag. The study therefore does not establish that delay consistently across its tests.

Optical sampling adds another limitation: observations may miss maximum brightness or have calibration problems, and nova-times does not reconstruct an unobserved peak. Fermi downtime and missing early optical observations also required exclusions from parts of the comparison.

The findings support using optical behavior to guide gamma-ray analysis, rather than applying one fixed duration to every nova. They do not prove that shocks dominate every nova’s visible emission. The supplied arXiv record lists the study as published in The Astrophysical Journal, volume 1010, article 45; its November 6, 2026 revision corrected a table typo without changing the main results.