Can Lomb-Scargle analysis find asteroid rotation periods from sparse photometry?

Why sparse photometry needs a different tool

Asteroid brightness varies as an irregular body rotates, so a light curve — brightness over one full rotation — carries the spin period. The Minor Planet Center archive holds hundreds of millions of such measurements, but typically only three or four per object per night, with gaps between nights, oppositions and observatories. The paper puts the typical photometric uncertainty at about 0.1 mag and notes systematic biases that depend on the observatory, its instruments and the star catalog used.

Classical Fourier methods assume evenly sampled data, which these archives do not provide. The Lomb-Scargle method instead fits sinusoids directly to unevenly sampled points and returns a power spectrum whose peaks mark candidate frequencies. The paper's question is whether that spectrum is sharp enough, once many observatories are pooled, to identify a rotation period rather than an artifact of the observing schedule.

Two routes from the same archive

The study compares two ways of handling the same data. In the separate dataset analysis (S-method), each observatory-and-band series is analyzed on its own, and the final period is a weighted average of consistent candidate periods. The weights combine the peak power of the spectrum, the number of measurements and the quality of the light-curve fit. Datasets with fewer than 70 brightness measurements were excluded, a threshold the authors describe as a pragmatic filter rather than a physical criterion.

In the combined dataset analysis (C-method), all measurements are corrected for band, star-catalog and observatory biases, reduced to a common scale and merged before a single Lomb-Scargle analysis. Merging lengthens the total time baseline, which sharpens frequency resolution, raises signal-to-noise with the number of measurements, and fills gaps in the daily cycle because observatories at different longitudes observe at different times. The paper adopts the C-method period as its final value, with an uncertainty set by the period resolution over the baseline.

Known periods recovered within 0.05%

Four asteroids with periods already listed in the ALCDEF database — 1951, 1963, 2134 and 2150 — served as the validation set. Both methods agreed closely with the database values. The largest discrepancy was for asteroid 1951, at 0.04% for the S-method and 0.05% for the C-method; for the other three test asteroids the differences were 0.03% or smaller.

That comparison is the paper's strongest evidence that the pipeline is not simply fitting noise. It also shows the two routes converging: the separate and combined analyses, built on different weighting and correction steps, landed on the same periods for the test objects. The validation therefore supports using the merged analysis as the default while treating the separate analysis as a cross-check.

Daily aliases decide which peaks are real

Ground-based observing leaves a gap of roughly one day between nights, and the paper explains that this window function creates peaks at multiples of one cycle per sidereal day, plus finer structure from the yearly gaps between oppositions. A double-peaked light curve with period P varies at twice the rotation frequency, so two candidate periods are daily aliases when 48/P1 − 48/P2 is close to an integer number of cycles per day. Combining observatories at different longitudes weakens these aliases but does not remove them.

This is why the paper treats a strong peak as a candidate rather than an answer. For asteroid 2607, the combined data gave dominant peaks at 2.936 and 1.468 hours with nearly equal power; the 1.47-hour curve had only one maximum and one minimum, so it was discarded as the half-period, and the adopted value was 2.9356 ± 0.0001 hours. The authors also note that a period half the true value will always appear, because folding a light curve in half costs little fit accuracy.

Six new periods with different confidence

New periods were reported for asteroids 2968, 2971, 3081, 3473, 3716 and 4303, but they do not carry equal weight. Asteroid 4303 is the cleanest case: fourteen datasets gave periods between 6.1362 and 6.1377 hours, the weighted mean was 6.1367 ± 0.0001 hours, and the combined analysis of 4802 measurements returned 6.136 hours with daily aliases at or below 0.06 power. Seven of eight oppositions with at least 50 measurements agreed, with only the poorly sampled 2013 opposition differing.

Asteroid 3716 shows the limit of the approach. The most likely period is 10.474 hours, but its daily alias at 13.405 hours gave a higher significance in the S-method, so the alias cannot be excluded completely. Asteroid 2968 remains ambiguous because of daily aliasing, with the S-method offering alternative candidates near 4.176 and 3.835 hours. For 3173, the strongest peaks at 22.99 and 45.98 hours are probably aliases of a slower brightness variation; the paper adopts no rotation period, because a single- versus double-peaked light curve, the synodic-versus-sidereal difference over the 21-year baseline, and possible non-principal-axis rotation cannot be resolved with the present data.

Where the method earns its keep

The paper's own comparison puts the value of merging in the time baseline and phase coverage rather than in raw precision alone. Longer baselines improve frequency resolution and period accuracy, and the wider phase coverage matters most for slow rotators, whose light curves are hardest to sample within a single night. The 2607 result illustrates the payoff: the combined period of 2.9356 ± 0.0001 hours agrees with a recently published Kepler K2 value of 2.81 ± 0.22 hours within its error, while being far more precise.

The authors conclude that the approach is promising for automated searches for unknown asteroid rotation periods. The evidence supports that as a statement about this pipeline on these twelve objects, not as a guarantee for every archive entry. The 70-measurement threshold, the correction of band and catalog biases, and the explicit alias check are all part of what made the recovered periods credible.

The alias check is the real bottleneck

The decisive difference among these twelve asteroids is not whether Lomb-Scargle found a peak but whether the alias structure could be resolved. Where many datasets and oppositions agreed — as for 4303, with fourteen consistent datasets and seven of eight oppositions matching — the period is adopted with a resolution-level uncertainty. Where the daily alias competed with the physical period, as for 3716 and 2968, the result stays conditional. Where even the shape of the variation was unresolved, as for 3173, no period is claimed.

For readers weighing whether to trust a Lomb-Scargle period from sparse photometry, that ordering is the practical takeaway: the method reliably recovers periods when independent datasets converge and the alias peaks are demonstrably weaker, and the honest output in the ambiguous cases is a ranked set of candidates rather than a single number.

Sources used in this answer