Hidden oscillations in plain sight: identification of seismically unresolved red-giant asteroseismic binary candidates

Abstract: Light curves of oscillating stars provide valuable insights into the stellar interiors. When oscillations from a pair of stars are captured within a single photometric aperture, they can be considered as potential asteroseismic binaries (ABs). If the two stars oscillate at similar frequency ranges, the superpositioned oscillation patterns appear as if from a single star, leading to inaccurate asteroseismic parameters. We investigate seismically unresolved AB candidates consisting of two red-giant stars observed by Kepler. We directly compare the power density spectra (PDSs) of blended and separated oscillations from both stars, and examine the impact of oscillations from two stars on asteroseismic and stellar parameters. We selected APOKASC3 stars with at least one neighboring source within 20 arcsec and show oscillations in similar frequency ranges. We focus on the systems where the light curves from each star in AB candidates are available or can be extracted with a custom mask. We identified 6 seismically unresolved AB candidates whose PDS morphologies change noticeably across light curves extracted with different apertures. Oscillations from two stars in a PDS cause inaccurate mode identification and bias the seismic parameters. These biases propagate into stellar properties: masses and radii for the 6 AB candidates differ by up to about 3 and 2 times relative to the individual stars, respectively. For the AB candidate with the most complex PDS, core properties become unreliable, with the coupling factor often being overestimated. We checked that all 6 AB candidates are chance alignments. Our results indicate that the inconsistencies in asteroseismic and stellar parameters across different studies can be explained by potential seismically unresolved ABs. We highlight the importance of identifying and accurately accounting for such systems in asteroseismic analysis.

• The paper identifies six seismically unresolved red-giant binary candidates in APOKASC3 by comparing blended and deblended Kepler power spectra, finding that all are likely chance alignments.
• Blending biases key measurements by up to 30% in νmax, 35% in Δν, 150% in mass, and 75% in radius, while potentially misclassifying evolutionary states and creating spurious low-mass red-clump stars.
• The study shows that complex spectra and inflated coupling factors require contamination checks because unresolved companions can mimic magnetic suppression, distort mixed-mode identification, and produce misleading stellar inferences.

Motivation and scope

When a Kepler photometric aperture encloses more than one oscillating red giant, the extracted light curve records a superposition of solar-like oscillation signals from both stars. If the two stars oscillate in similar frequency ranges, their power density spectra (PDSs) merge into a single power excess, and the system is easily mistaken for one star — a configuration the authors call a seismically unresolved asteroseismic binary (AB). Such blending biases $\nu_{\max}$ and $\Delta\nu$, which propagate through the asteroseismic scaling relations into masses, radii, evolutionary classifications, and mixed-mode diagnostics such as $\Delta\Pi_1$ and $q$. The paper identifies six seismically unresolved red-giant AB candidates within APOKASC3 (2604.24476), quantifies the resulting parameter biases by direct comparison with PDSs of the individual stars, and shows that all six are chance alignments rather than gravitationally bound systems.

The motivation builds on prior evidence that photometric contamination produces false positives: misidentified eclipsing binaries (Abdul-Masih et al., 2016), anomalous high-amplitude peaks in 168 red giants (Colman et al., 2017), and 909 dwarf/subgiant targets whose apparent red-giant oscillations originated from nearby giants (Hon et al., 2019). Population synthesis predicted at least 200 ABs in Kepler long-cadence data [2014ApJ...784L...3M], but an observational study of 40 resolved AB candidates found that most are chance alignments (Espinoza-Rojas et al., 16 Sep 2025). This paper extends the census to the harder case where the two power excesses overlap.

Target selection and light-curve preparation

Starting from APOKASC3 (Pinsonneault et al., 2024), the authors selected pairs of stars separated by less than 20 arcsec (about five Kepler pixels) with $10 < K_p \leq 17$ and $\nu_{\max}$ differences below 30% of the mean value — a conservative threshold given that envelope widths scale as approximately $0.66\nu_{\max}^{0.88}$, allowing overlap for separations up to roughly 40%. Forty candidate pairs were selected.

For each pair, the authors assembled PDSs from multiple pipelines: KEPSEISMIC (KADACS-corrected), KBonus-Background PSF photometry deblended per Gaia DR3 source (Martinez-Palomera et al., 2023), PDCSAP, and custom apertures built from combined target pixel files using a modified MOSAIC pipeline with non-overlapping sub-pixel masks. Eighteen candidates were rejected because the "background" star's oscillations actually originate from its brighter neighbor; seven stars in the crowded clusters NGC 6791 and NGC 6819 were excluded due to severe contamination (CROWDSAP < 0.4). After cross-comparison of $\nu_{\max}$, $\Delta\nu$, and PDS morphologies across apertures, six seismically unresolved AB candidates (AB1–AB6) remained, each showing clearly different mode patterns between blended and single-star PDSs.

Asteroseismic analysis

Global parameters were obtained with TACO via a Bayesian MCMC fit of three super-Lorentzian granulation components plus a Gaussian oscillation envelope. Peaks were detected automatically, fitted with Lorentzian or sinc² profiles selected by AIC, and assigned spherical degrees using the universal pattern; $\Delta\nu$ came from weighted linear fits to $\Delta\nu$0 frequencies, verified with échelle diagrams. Mixed-mode parameters ($\Delta\nu$1, $\Delta\nu$2, $\Delta\nu$3, $\Delta\nu$4) were inferred with the BOChaMM forward-modelling pipeline (Kuszlewicz et al., 2023) on stretched periods. Masses and radii used scaling relations with Asfgrid $\Delta\nu$5 corrections; notably, $\Delta\nu$6 was fixed to unity because blending itself would bias any empirical calibration — a reasonable but acknowledged simplification.

PDS morphologies and mode identification failures

The six systems span the full range of predicted configurations from artificial-AB simulations (Choi et al., 2 Jun 2025). In AB1, aligned $\Delta\nu$7 pairs make the blend resemble a single star, while partial alignment causes an $\Delta\nu$8 mode of AB1-B to masquerade as an $\Delta\nu$9 mode near 94 μHz. AB6 — two CHeB stars with only ~8% difference in $\Delta\Pi_1$0 and misaligned modes — reproduces the most complex predicted morphology, and any mode identification performed on its blended spectrum is inherently incorrect. Three of the six candidates contain suppressed-dipole-mode stars (AB1-B, AB3-B, AB4-A); blending can artificially inflate dipole visibilities when peaks from the companion are misattributed, which matters because depressed dipole amplitudes are interpreted as signatures of strong internal magnetic fields (Fuller et al., 2015). Any magnetic-field inference from these three stars requires prior exclusion of unresolved blending.

Quantified biases in stellar parameters

The central quantitative result is that treating a blend as a single star biases derived properties severely:

Quantity	Maximum bias observed
$\Delta\Pi_1$1	~30%
$\Delta\Pi_1$2	~35%
Mass	up to ~150% (~3×)
Radius	up to ~75% (~2×)

The worst case is AB2-B, whose mass is overestimated by factors approaching 3 depending on which blended aperture is used. Critically, the authors show that small biases in mass and radius do not imply correct values: compensating errors in $\Delta\Pi_1$3 and $\Delta\Pi_1$4 can cancel in the scaling relations, as demonstrated for AB4-B and AB5-B. Hence absence of bias in fundamental parameters cannot be used to dismiss contamination.

Two further consequences stand out. First, evolutionary classification is corrupted: AB2-B, an RGB star, is classified as CHeB in both APOKASC3 and the Vrard catalog, plausibly because peaks from the CHeB companion dominate its blended spectrum. Second, AB5-B illustrates how blending can manufacture apparently non-canonical low-mass red clump stars: its blended-PDS mass of $\Delta\Pi_1$5 sits near the theoretical lower mass limit for RC stars at the age of the Universe, whereas the clean KBonus-based mass is $\Delta\Pi_1$6. Some reported sub-$\Delta\Pi_1$7 red giants previously attributed to binary interaction or merger histories may instead be blends.

Mixed-mode parameters

Period spacings proved comparatively robust — $\Delta\Pi_1$8 differences between blends and individual stars are minor — but the authors caution this reflects insensitivity rather than accuracy: for AB6, a single optimized $\Delta\Pi_1$9 fails entirely to reproduce the $q$0 pattern. Coupling factors are far more vulnerable. Blended analyses yield systematically overestimated $q$1: AB1 gives $q$2 for an RGB star, well above the expected evolutionary trend from 0.18 to 0.12 (Mosser et al., 2016), and AB3 similarly yields inflated $q$3. Since $q$4 probes the evanescent zone and core structure, published coupling factors for stars with complex PDSs should be treated as suspect unless blending has been excluded.

Nature of the systems

Applying wide-binary diagnostics based on Gaia astrometry (Gilmutdinov et al., 2021, Espinoza-Rojas et al., 16 Sep 2025), all six candidates are consistent with chance alignments within the Kepler apertures, with projected separations of roughly 6,000–19,000 AU. This aligns with the finding that most detectable ABs are unassociated pairs, contrary to early population-synthesis expectations that fewer than 10% would be chance alignments. One complication remains: AB3 involves three sources within the aperture, and it cannot be determined whether the second oscillating component is KIC 3122188 or Gaia DR3 2051928738864291072; TESS light curves show no signal above the long-cadence Nyquist frequency for either candidate, leaving the pairing open. KIC 3122188 and AB3-B may form a genuine wide binary, which — if confirmed — would make the system a testbed for tidal damping of mixed modes in a suppressed-dipole star.

Limitations

Several caveats bound the results. The sample is small (six systems, ~0.03% of APOKASC3) and restricted to pairs with extractable individual light curves, so the incidence rate of seismically unresolved ABs remains unconstrained. The 30% $q$5 threshold is empirical and conservative. Direct comparison with the predicted morphology taxonomy of Choi et al. is imperfect because real apertures produce flux ratios varying across the envelope, so individual systems exhibit multiple morphology categories simultaneously. Finally, fixing $q$6 means absolute masses carry the usual uncalibrated-scaling systematic, though this affects blended and single-star values equally and does not compromise the differential bias measurements.

Conclusion

This work demonstrates that seismically unresolved red-giant ABs, though rare, produce biases large enough to corrupt mode identification, evolutionary classification, masses and radii (up to factors of ~3 and ~2), and especially coupling factors, while leaving some scalars deceptively unbiased through error cancellation. All six identified cases are chance alignments. The practical implication is that complex PDS morphologies, anomalously low seismic masses, and unusually high $q$7 values in red giants warrant screening for nearby oscillating contaminants before invoking exotic internal physics or non-canonical evolution. An open question the paper leaves is the true occurrence rate of such systems across the full Kepler sample, and whether forthcoming Gaia releases can resolve the ambiguous triple configuration around AB3.