Only 6 of 102 asteroid families show two distinct populations of rock

Nine hundred and ninety-nine parts signal, one thousand parts noise. That is roughly the situation Murat Kaplan describes for asteroid albedo measurements from NASA's NEOWISE survey: across the asteroid families he studied, the median ratio of observed spread to measurement error came out at about 0.99 in logarithmic space. The scatter you see in the data is, on average, no larger than the scatter the instrument itself introduces.

Albedo is simply how much sunlight a surface reflects. Coal-dark carbonaceous asteroids reflect a few percent; stony ones reflect several times more. Because albedo tracks composition, planetary scientists lean on it heavily when studying asteroid families, the swarms of fragments left behind when a larger body was shattered by a collision. If a family contains two clearly separated albedo populations, one dark and one bright, that hints at something interesting: a parent body that was layered, or an interloper population mixed in, or two collisions mistaken for one.

The trouble is that a bump in a histogram can come from real structure or from noisy measurements smeared across a range. Kaplan set out to tell the difference, and to say quantitatively when the distinction can be made at all.

Checking the ruler before measuring the room

The analysis covers 102 asteroid families. Before hunting for structure, Kaplan checked NEOWISE against an independent infrared survey, AKARI, using 1,498 asteroids observed by both. The agreement is decent: a fitted slope of 1.03, an R-squared of 0.76, and a median absolute albedo difference of 0.015. There is a known systematic offset of about 11 percent that depends on an asteroid's spectral type, which Kaplan accounts for rather than dismisses. SDSS photometry provided a further consistency check.

One methodological choice does real work here. Kaplan analyzes albedos in logarithmic space (log10 of the visual albedo) rather than linear, on the grounds that dark-and-bright mixtures are lopsided: the dark population is squeezed into a narrow band near zero in linear units, while the bright one sprawls. Taking logarithms evens out that asymmetry. The choice turns out to matter for the results.

Applying conservative criteria, Kaplan finds convincing evidence of two distinct albedo populations in 6 of the 102 families, or 5.9 percent. Only two of those, Nysa-Polana and Juno, show up in both linear and logarithmic albedo; the other four appear only in log space. Nysa-Polana is a well-known case, long suspected of being two overlapping families rather than one.

Absence of evidence, or evidence of absence?

A low detection rate invites an obvious objection: maybe the data are simply too noisy to see the structure. Kaplan tested this directly with Monte Carlo simulations, injecting artificial dark-bright splits at the characteristic separation seen in real cases (about 0.64 dex, meaning roughly a factor of four in reflectivity) into synthetic families carrying the same intrinsic scatter as the observed ones. Those injected splits were recovered with high probability. So the method can see a real split when one is there. The author reads the 5.9 percent rate as genuine compositional homogeneity: most families really are made of one kind of rock, as you would expect if they came from a single, well-mixed parent body.

The second result is less comfortable. Many published analyses report correlations between asteroid size and albedo within families, and such trends get interpreted as physical, perhaps space weathering that darkens surfaces over time, or size-dependent fragmentation. Kaplan tested 63 families for these correlations while controlling for the fact that surveys are magnitude-limited: a small bright asteroid and a large dark one can look equally faint from Earth, so any brightness-limited catalog preferentially collects small bright objects and large dark ones, manufacturing a correlation from nothing. After imposing diameter-limited controls, only 2 of the 63 families survived.