For more than a century, astronomers have scrutinized a nearby pair of K-dwarf stars, but the latest search for planets finds that the strongest short-term motions in the stars are coming from the stars themselves. New high-precision measurements of 70 Oph A and B show no statistically significant planetary signals, while a roughly 20-day pattern and its harmonics are consistent with rotation-driven stellar activity. At the same time, the study finds that a region around 70 Oph A where a planet could potentially have liquid water is dynamically stable, leaving Saturn-mass and smaller planets there among the possibilities not excluded by the current data.
The 70 Oph system lies about 5.1 parsecs, or roughly 16.7 light-years, from Earth and consists of two main-sequence K dwarfs. The primary, 70 Oph A, is classified as K0 V, while the fainter companion, 70 Oph B, is a K4–K5 V star. Their orbit takes about 88 years, and astronomers have accumulated radial-velocity and astrometric measurements of the system for more than a century.
The system also has an unusual place in the history of planet hunting. In 1855, an astronomer reported what was interpreted as a possible planet associated with 70 Oph, but later work showed that the claim was an artifact of 19th-century visual measurements. No planet has since been established in the system. The long record of observations nevertheless makes 70 Oph useful for looking for companions whose gravitational pull could produce small motions in either star.
For the new study, the researchers added observations from the Planet Finder Spectrograph on the 6.5-meter Magellan II Clay Telescope in Chile. Their campaign covered 48 nights between August 2023 and March 2025. The individual exposures reached radial-velocity uncertainties of about 0.7–0.8 meters per second, while nightly averaging produced a median precision of about 0.27 meters per second.
The new observations were combined with radial velocities from several other instruments, including the McDonald Observatory, Keck/HIRES, HARPS and the Automated Planet Finder, along with more than a century of relative astrometry and absolute astrometric information from Hipparcos and Gaia. The combined data allowed the researchers to refine the orbit of the binary before searching for additional motions that might betray planets.
The resulting orbital period is 88.126 ± 0.010 years, with an eccentricity of 0.50015 ± 0.00017 and an inclination of 121.144° ± 0.028°. The derived dynamical masses are 0.8827 ± 0.0043 solar masses for 70 Oph A and 0.7319 ± 0.0031 solar masses for 70 Oph B. The mass uncertainties are below 1 percent for both stars.
The improved measurements also substantially tighten the binary solution compared with an earlier pre-Gaia analysis. The uncertainty in the eccentricity is reduced by roughly a factor of 12, while the masses are more precise than the earlier values by factors of about 4.6 and 3.2 for the two stars, respectively.
The strongest signal follows the stars’ rotation
Once the gravitational effect of the binary companion was accounted for, the remaining radial-velocity variations did not produce a convincing planetary signal.
For 70 Oph A, the first periodogram analysis produced a prominent signal near 10 days. That period is approximately half the star’s independently measured rotation period of about 20 days. The fundamental roughly 20-day period also appears in the radial velocities.
The pattern becomes more persuasive when the researchers analyze the combined long-term data. After subtracting the binary orbit, the maximum-likelihood periodogram shows prominent signals near 9.7, 11.9, 20.3 and 59.2 days. The 20.3-day signal is strongest in the high-precision PFS and McDonald data and agrees closely with the known rotation period of 70 Oph A.
The other prominent periods also line up with fractions or multiples of that rotation period. A roughly 9.7-day signal corresponds to half the rotation period, while a roughly 5.4-day signal that emerges after removing the strongest periodic component corresponds to about one-quarter of it. Another analysis identifies signals at 20.3, 10.3 and 6.8 days, corresponding approximately to the rotation period and its first two harmonics.
The researchers interpret these repeated relationships as evidence that the radial-velocity variations arise from stellar activity rather than several orbiting planets.
The proposed physical picture is that active regions on the stellar surface rotate in and out of view. Such regions can alter the shapes of stellar spectral lines and thereby shift the measured radial velocity. Because the active regions evolve and are not distributed perfectly symmetrically across the stellar surface, their effects can produce a primary rotation signal along with harmonics. The authors describe this interpretation as consistent with the observed structure and the finite coherence of the signals.
The evidence is particularly strong because the approximately 20-day feature appears in both components of the binary. A planetary companion orbiting one star would not be expected to produce the same orbital period in the other, widely separated star. The common signal therefore supports an activity-related interpretation.
For 70 Oph B, the radial-velocity data are noisier and more complicated. Peaks occur between roughly 20 and 30 days, while the activity indicator Hα also shows excess power over a similar range. A separate feature near 18 days appears in more sophisticated analyses, but it does not cross the study’s adopted significance thresholds. The researchers suggest it may be a rotational harmonic, but they do not treat it as a detected planetary signal.
Other features are similarly treated cautiously. A 59-day signal appears in the PFS data from both stars and lies near three times the roughly 20-day rotation period, so the researchers favor interpreting it as a rotational alias or harmonic. A roughly 11.9-day feature does not correspond to a simple harmonic and is not consistently recovered when the data are divided into separate periods, which argues against a stable Keplerian signal.
Modeling the stellar activity leaves no planet signal
The researchers then used Gaussian-process regression to model the changing stellar activity directly. This approach is designed to represent variability that is periodic but not perfectly stable, as expected when active regions evolve over time.
The analysis focused on 70 Oph A because the periodogram methods found no significant signal in 70 Oph B. It used 230 radial-velocity measurements from PFS, APF, McDonald and Keck/HIRES spanning 27 years. The binary orbit was removed first, while the individual instrumental offsets and additional noise were fitted simultaneously.
The activity-only model produced a substantially better statistical fit than models containing only multiple Keplerian signals. Its Bayesian information criterion was 1797.37, compared with 1847.85 for a two-Keplerian model and 1856.53 for a three-Keplerian model. The activity model also reduced the root-mean-square residual to 4.11 meters per second, compared with 11.59 meters per second for the two-Keplerian model.
The fitted activity model gives a characteristic rotation period of 19.76 days, with an activity amplitude of about 11.9 meters per second and an active-region evolution timescale of about 72 days. The coherence parameter indicates that the activity remains correlated over several rotations rather than behaving as a perfectly repeating signal.
The researchers also tested hybrid models in which stellar activity was modeled together with individual Keplerian signals at approximately 7, 10, 12 and 60 days. None improved the statistical description enough to justify the added planetary component. The activity-only model had the lowest BIC of all the tested models.
Taken together, the periodograms and Gaussian-process analysis provide no statistically significant evidence for a planet around either component of 70 Oph AB in the current radial-velocity data.
The observations rule out some giant planets
A nondetection does not mean that the system is empty. The researchers used injection-and-recovery experiments to determine which planets the available radial-velocity data would have been capable of detecting.
They injected synthetic circular planetary signals into the observed data over a wide range of orbital periods, masses and orbital phases. The analysis used the combined 27-year data set for 70 Oph A and the approximately two-year PFS data set for 70 Oph B. A synthetic planet was considered recovered when the search identified its injected period within the specified tolerance and passed the adopted false-alarm threshold.
For 70 Oph A, the resulting completeness analysis excludes planets with minimum masses of about 0.3 Jupiter masses at 1 astronomical unit and about 0.5 Jupiter masses at 2 astronomical units at the 50 percent detection-completeness level. Jupiter-mass planets interior to about 5 au are excluded by the long radial-velocity baseline, with planets near 1 au being detectable at substantially lower masses than Jupiter.
For 70 Oph B, the available radial-velocity coverage is much shorter. The sensitivity reaches about 0.5 au, where planets more massive than roughly 0.25–0.3 Jupiter masses are excluded at the stated completeness level.
These limits leave a substantial range of possible planets untouched. In particular, planets with masses comparable to Earth remain below the present radial-velocity sensitivity because stellar activity produces velocity variations larger than the signal expected from such low-mass companions. The nondetection therefore constrains the population of giant planets much more strongly than it constrains terrestrial planets.
The habitable zone remains dynamically possible
The absence of a detected planet is especially notable because the researchers find that 70 Oph A has a region where planets could remain in stable orbits despite the gravitational influence of 70 Oph B.
The binary’s closest stellar approach is about 11.6 au. Using the stellar properties and the binary orbit, the researchers calculate a conservative habitable zone around 70 Oph A extending from 0.711 to 1.273 au. Their broader optimistic habitable zone extends from 0.561 to 1.342 au.
They then tested the long-term stability of planetary orbits using the REBOUND N-body integrator. Each simulation placed an Earth-mass test planet on a circular orbit in the same plane as the binary and followed its motion for 10 million years. The simulations covered orbital distances from 0.5 to 10.5 au in steps of 0.005 au.
The calculations show that stable S-type orbits around 70 Oph A extend to roughly 2.5 au, with the analytic estimate giving a critical boundary near 2.88 au. Beyond roughly 2.5 au, the simulated planets became dynamically unstable. Within the habitable zone, however, the model planets could remain stable, with only moderate eccentricity excitation from the companion star.
The simulations also reveal regions of enhanced orbital eccentricity between the outer edge of the habitable zone and the stability boundary. These features correspond to mean-motion resonances with the binary orbit. Around 1.5 au, in particular, the simulations produced eccentricities as high as about 0.63.
The stability calculations assume coplanar, prograde planetary orbits. The authors note that strongly inclined planetary orbits could behave differently because of Kozai–Lidov oscillations, which can drive large changes in orbital eccentricity and inclination. The mutual inclination between a possible planetary orbit and the binary cannot currently be established.
Within the assumptions of the simulations, however, an Earth-mass planet on a near-circular, coplanar orbit inside the habitable zone is dynamically viable. The radial-velocity observations simply are not sensitive enough to rule out such a planet.
What the current data leave open
The study therefore separates two questions that are easy to conflate. The first is whether a planet has been detected. It has not. The second is whether planets could exist in parts of the system that the observations cannot yet probe. They could, according to the radial-velocity limits and dynamical simulations presented here.
For 70 Oph A, the habitable zone lies roughly 0.7–1.3 au from the star, inside the region where the dynamical calculations permit stable circumstellar orbits. Yet terrestrial-mass planets in that zone remain below the radial-velocity detection threshold. Even sub-Jupiter planets can remain viable depending on their mass and orbital distance.
The researchers also calculate that a Saturn-mass planet at about 1 au would produce an astrometric signal of roughly 70 microarcseconds, while a 0.5-Jupiter-mass planet would produce a signal of about 100 microarcseconds. An Earth-mass planet at 1 au would produce a much smaller signal of roughly 0.2–0.3 microarcseconds.
Those estimates illustrate why the radial-velocity search alone cannot settle whether small planets inhabit the system. The study discusses direct imaging and high-precision astrometry as complementary approaches, particularly because 70 Oph is nearby and its habitable zone has an angular separation of roughly 140–250 milliarcseconds from the star.
The researchers also note an unresolved question about the system’s age. Activity- and rotation-based estimates give ages of roughly 2.8–3.0 billion years, while earlier asteroseismic modeling of 70 Oph A suggested an age of about 6–7 billion years. They suggest that weakened magnetic braking or older stellar-model assumptions could contribute to the discrepancy, and note that a modern asteroseismic analysis could help resolve it.
For the planet search itself, however, the central result is more direct. After accounting for the 88-year binary orbit and the stars’ own changing activity, the radial velocities contain no statistically significant coherent planetary signal. The strongest recurring variations instead track the roughly 20-day rotation of 70 Oph A and its harmonics, while the sensitivity tests show that the observations can exclude a range of giant planets but leave terrestrial-mass planets in the habitable zone beyond the current detection threshold.
The study was published in The Astrophysical Journal.
