From a Different Star: 3I/ATLAS in the Context of the Ōtautahi–Oxford Interstellar Object Population Model

This Astrophysical Journal Letters paper asks how the newly discovered third interstellar object (ISO), 3I/ATLAS (“3I”), fits within a physically motivated Galactic population model of ISOs. The question matters because ISOs are rare, fast-moving remnants of planet formation in other stellar systems, and—unlike comets bound to the Solar System—sample the Milky Way’s disk-wide diversity of protoplanetary environments. If we can connect an individual ISO’s measured orbit to the properties of its likely parent stars (e.g., metallicity, age, and volatile content), then each discovery becomes a direct probe of planetesimal formation and subsequent dynamical processing across the Galaxy.

The authors use the ¯Otautahi–Oxford (¯O–O) interstellar object population model (from Hopkins et al. 2025). The model is built to predict ISO chemodynamics by combining (i) Gaia DR3 constraints on the local stellar population (within 200 pc), (ii) assumptions about how planetesimals are released from protoplanetary disks as a function of stellar mass and metallicity, (iii) a disk-scattering prescription that efficiently ejects ISOs early at low speeds (≲ 10 km s−1) from beyond the water ice line, and (iv) long-lived survival of ISOs through stellar evolution. A key modeling step is that the “solar neighbourhood sine morte” stellar population—i.e., the stellar population that would exist if stars did not die—is used as a proxy for the sources of ISOs currently passing through the Solar System. The model then reweights this stellar distribution in velocity, metallicity, and age into a predicted ISO distribution in velocity, composition, and age, under the assumption that stars and ISOs trace the same Galactic potential and source populations.

Methodologically, the paper performs a first assessment of 3I by comparing its measured hyperbolic orbit to the ¯O–O model’s predicted distributions. The orbit fit yields an asymptotic speed at infinity of $v_{\infty }=57.99\pm 0.10\mathrm{km}{\mathrm{s}}^{-1}$ and eccentricity $e=6.143\pm 0.003$. From this, the authors infer the Galactic velocity components relative to the Sun as $(U,V,W)=(-51.0,-19.2,18.5)\mathrm{km}{\mathrm{s}}^{-1}$. They emphasize that the orbit-fit covariance implies an uncertainty in the velocity of less than $\sim 1\mathrm{km}{\mathrm{s}}^{-1}$, small compared with the model’s non-Gaussian velocity-structure features.

The first set of results compares 3I’s speed and radiant (approach direction) to the model’s predictions for ISOs with perihelion $q<5\mathrm{au}$, a range chosen to match the expected perihelia of the upcoming LSST ISO sample and consistent with the perihelia of the first two ISOs. The model predicts a broad, non-Gaussian distribution of $v_{\infty }$ and radiants; 3I is described as consistent with the expected speed range but somewhat unusual in its approach direction. Specifically, the paper notes that radiants are mainly in the Northern hemisphere in the model, with a concentration near the equator, while 3I’s declination at discovery is $\sim -19^{\circ }$, i.e., relatively far south. In velocity space, the model’s in-plane structure is dominated by “moving groups” (overdensities in $U$–$V$) created by spiral-arm dynamical effects. The authors find that 3I’s large vertical velocity component $W$ places it outside the main moving-group structures, which are associated with stars having relatively low vertical motion. This kinematic placement implies that 3I likely originates from older and lower-metallicity stars—those on Galactic orbits with larger oscillations out of the Milky Way plane.

To translate kinematics into chemodynamics, the authors compute model-based posterior distributions for the parent-star metallicity, water mass fraction $f_{{\mathrm{H}}_2\mathrm{O}}$, and age for ISOs whose velocities match 3I within a $5\mathrm{km}{\mathrm{s}}^{-1}$ window. They report that ISOs with 3I’s velocity are generally more water-rich than the overall ISO population, and that this follows from the moving-group membership differences: velocities within moving groups correspond to higher-metallicity stars, while velocities outside them correspond to more metal-poor stars. The paper also finds that the age–velocity correlation is weak in the model: at 3I’s velocity, the predicted age distribution spans a wide range with no strong skew toward young or old ages. This motivates a more careful age inference.

For age estimation, the authors use a simplified version of the age–velocity dispersion relation (AVR) method of Almeida-Fernandes & Rocha-Pinto (2018). Because the full 3D velocity distribution is highly non-Gaussian (with gaps and overdensities in $U$–$V$), they avoid fitting a Gaussian in all components and instead use only the out-of-plane component $W$. They assume $W$ is distributed as a 1D Gaussian centered on the local standard of rest component $W_{\mathrm{LSR}}$, with dispersion ${\sigma }_W$ increasing with age according to a two-slope AVR. With a uniform age prior over $0.01$–$14\mathrm{Gyr}$, they compute likelihoods and obtain posterior age distributions.

The key quantitative age result is that the 68% confidence interval for 3I’s age is $7.6$–$14\mathrm{Gyr}$. The authors contrast this with the first two ISOs: for 1I the 68% interval is $0.01$–$7.2\mathrm{Gyr}$, and for 2I it is $0.36$–$8.8\mathrm{Gyr}$. They stress that these posteriors remain non-zero for nearly all ages, meaning that—under this method—an ISO’s age is intrinsically uncertain when inferred from velocity alone. Nonetheless, the only strong constraint for 3I is that the posterior approaches zero for ages below $\sim 1\mathrm{Gyr}$, yielding a robust qualitative conclusion that 3I is likely very old. They also argue that commonly used “point estimate” methods (equating the ISO’s velocity relative to the LSR to a velocity dispersion and reading off an age) bias ages low, which would be especially problematic for 1I and would undermine back-tracing to parent stars.

The paper then evaluates whether 3I could share a common origin with either 1I or 2I. Using Bayes’ theorem, the authors compute the posterior probability that 3I is “related” to 1I or 2I given an assumed stream velocity dispersion ${\sigma }_s$. They adopt a prior $P(r\mid {\sigma }_s)=0.01$, motivated by Forbes et al. (2024) and noting that related associations are more likely for shared clusters than for shared single stars. For the likelihood, they assume the stream is centered on the velocity of 1I or 2I and model the stream as a tri-variate Gaussian with isotropic covariance ${\sigma }_s^{2}I$. For the unrelated case, they estimate the density of 3I’s velocity using a kernel density estimate (KDE) from $10^6$ samples drawn from the ¯O–O model. The resulting posterior probabilities are low: the probability that 3I is related to 1I or 2I remains below $1.4\%$ regardless of ${\sigma }_s$, and it never exceeds the prior for association with 2I. A complementary test statistic—minimum pairwise velocity separation among random ISO triplets—places the observed separations of (1I,2I,3I) in the middle of the model’s distribution, indicating no kinematic evidence beyond what the model already predicts from chance alignment among many streams.

Beyond the model comparison, the authors discuss unusual observational aspects of 3I. They report that 3I is quietly active at a heliocentric distance $r_h\sim 4.4\mathrm{au}$ with a compact coma, and cite early photometry giving an absolute magnitude upper limit $H_V=11.9\pm 0.6$. They note that later work (post-submission) estimates $H_V=13.7\pm 0.2$, corresponding to a radius of $5.6\pm 0.7\mathrm{km}$. In contrast, 1I and 2I are much smaller ($\sim 0.1$–$1\mathrm{km}$). This size difference is used to argue that the intrinsic ISO size-frequency distribution (SFD) may be shallower than inferred from only the first two detections. Using a rough number density estimate for objects larger than 3I, $n(H\le 12)\sim 10^{-3}\mathrm{a}{\mathrm{u}}^{-3}$, and comparing to earlier estimates for objects larger than 1I, $n(H\le 22)\sim 10^{-1}\mathrm{a}{\mathrm{u}}^{-3}$, they infer an empirical power-law slope in absolute magnitude of $\alpha \sim 0.2$, corresponding (at fixed albedo) to a diameter slope $q\sim 2$. They caution that this ignores survey selection effects and survivorship bias and should be treated as order-of-magnitude only, consistent with the need for multi-slope SFDs.

They also highlight an orbital-orientation peculiarity: due to solar motion, ISOs are biased to arrive from the solar apex direction, leading to observational biases favoring certain perihelion hemisphere and argument-of-perihelion ranges. LSST simulations in the ¯O–O framework suggest that perihelia in the Southern celestial hemisphere dominate the discovered population ($\sim 70\%$). For the subset of Northern-hemisphere perihelia, they report that orbits with $60^{\circ }\le \omega \le 150^{\circ }$ are strongly biased against discovery. The likelihood that an LSST-discovered ISO has an orbit similar to 3I’s ($125^{\circ }\le \omega \le 130^{\circ }$) is $\sim 1\%$. The paper interprets this as potentially a rare but plausible draw from the model, or possibly evidence for additional structure such as unmodeled streams or alternative orbital distributions.

Limitations are implicit in the modeling choices and explicitly discussed in the age inference. The ¯O–O model relies on assumptions about ISO release scaling with stellar mass and metallicity, efficient early scattering beyond the water ice line, and long ISO lifetimes. The proxy assumption that the current solar neighbourhood sine morte stellar population can represent the sources of ISOs passing through the Solar System is a major conceptual approximation. For age inference, the authors acknowledge that the velocity–age correlation is weak and that the AVR-based method yields broad posteriors; the method also depends on the assumption that $W$ can be treated with a Gaussian centered on $W_{\mathrm{LSR}}$ and an AVR-derived dispersion evolution, which may not capture all non-Gaussianities. For the common-origin test, the analysis simplifies the stream model by centering the stream on the observed ISO velocity and using an isotropic Gaussian covariance, and it adopts a fixed prior that may differ for shared-star versus shared-cluster scenarios.

Practically, the results suggest that 3I is consistent with being a typical member of the Galactic ISO population in speed but is chemically and dynamically informative: its kinematics point to an older, lower-metallicity origin and a higher water mass fraction, implying that water sublimation activity may become observable once 3I reaches within $\sim 3\mathrm{au}$ of the Sun. This is important for observational planning (spectroscopy and water searches) and for interpreting future ISO discoveries from LSST. The paper also provides an early test of the ¯O–O model before the expected 5–50 ISO detections, and it cautions that age estimates from velocity must be treated probabilistically rather than via biased point estimates.

Overall, the core contribution is the first integrated comparison of a newly discovered ISO (3I) to a Gaia-informed chemodynamical population model, yielding quantitative constraints: $v_{\infty }$ and radiants consistent with expectations, a 68% age interval of $7.6$–$14\mathrm{Gyr}$, a predicted tendency toward higher water content, and strong ($<1.4\%$) Bayesian evidence against a shared origin with either 1I or 2I.