{"id":"e480053a-4c87-418e-b7bd-66816adc46d0","arxiv_id":"2608.00173","paper_version":2,"verdict":"ACCEPT","confidence":"HIGH","novelty_score":3.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A review of planetary ejection mechanisms and their predicted free-floating planet demographics, concluding that planet-planet scattering, cluster encounters, and binary instabilities likely dominate FFP production.","lead":"This paper reviews the four main ways planets get ejected from their home star systems: close encounters with neighboring planets, instabilities in binary star systems, flybys of passing stars, and aging of the host star. It compares predicted ejection rates and observable signatures, using the growing census of free-floating planets as a test bed.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The central inference that low-mass microlensing FFPs are ejected planets rests on the unbound fraction of the FFP census; the paper flags this ambiguity but does not quantify it, so the claimed rates remain conditional.","rationale":"The reader's weakest_assumption is exactly the load-bearing point: the empirical anchor for 'ejection is common' is the microlensing FFP census, and the paper's own Section 6.1 acknowledges that a large fraction of apparent FFPs could be wide-orbit bound planets. I agree with that identification. However, I weigh the paper's explicit caveat heavily: the authors state the assumption, cite the relevant bound-planet prediction, and call for host-star searches as a future test. For a review article, the central claim is appropriately hedged ('suggests', 'if truly unbound'), and the synthesis of the four ejection mechanisms remains valuable even if the absolute FFP rate or the precise ranking in Table 1 changes. The absence of propagated uncertainties in Table 1 is a limitation but not an internal inconsistency. Therefore the concern does not move the verdict: the paper should remain ACCEPT, and the proposed two-population fit or host-star follow-up would settle the ambiguity while the field waits for Roman and Earth 2.0 data.","tokens_in":58602,"tokens_out":3357,"duration_ms":34386,"concrete_test":"Take the sample of short-timescale microlensing events from Sumi et al. (2023) and Mróz et al. (2017-2020) and re-fit them with a two-population model: truly unbound planets plus wide-orbit bound planets with host masses drawn from a plausible IMF and projected separations below the detection threshold of current follow-up. Marginalize over host detection/non-detection to compute the posterior on the unbound fraction. If the median unbound fraction is below ~50% for the sub-Neptune mass range, the 'ejection is common' claim and the relative rates in Table 1 need to be revised downward. An observational alternative is to obtain deep high-resolution imaging of ~20 FFP events 5-10 years after peak, searching for the lens host at the predicted position; the fraction with detected hosts directly measures the contamination.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The strongest claim—that the low-mass FFP population is dominated by ejected planets—hinges on the completeness-corrected FFP abundance (Section 6.1, Eq. 14, ~20 FFPs/star from Sumi et al. 2023) being a census of truly unbound planets. Microlensing events are classified as FFPs when no host star is detected, but wide-orbit bound planets produce nearly identical light curves if the host is too faint or too far from the lens. Section 6.1 concedes that 'apparent FFPs observed through microlensing are not necessarily unbound' and cites Hadden & Wu (2025) predicting that roughly half of Neptune-mass apparent FFPs may be on wide bound orbits. If that fraction is large, the ~20/star normalization is not an ejection rate; Table 1's relative rankings ('planet-planet scattering, embedded cluster encounters, and binary instabilities are likely predominant') lose their empirical anchor. The paper does not propagate this uncertainty into Table 1 or into the summary statement, so the central claim is conditional on an unmeasured quantity. This is a genuine soft spot, but it is explicitly acknowledged and the review's qualitative synthesis does not depend on a precise value of the unbound fraction, so it does not change the verdict.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This review synthesizes the dynamical mechanisms by which planets can be liberated from their host stars, organized into four channels: planet-planet scattering (Section 2), instabilities in binary and multi-star systems (Section 3), stellar and substellar flybys (Section 4), and post-main-sequence stellar evolution (Section 5). The paper develops the relevant analytic tools (Safronov number, angular momentum deficit, variable-mass two-body equations, impulse approximation), surveys the solar-system and extrasolar evidence for each channel, and compiles a comparative table (Table 1) listing the percentage of stars susceptible to each mechanism, the associated ejection efficiencies, the typical ejected planet masses, and predicted excess velocities. The central claims are that planetary ejection is common, that the four mechanisms can account for the completeness-corrected microlensing census of roughly 20 free-floating planets per star (Sumi et al. 2023, Eq. 14), and that planet-planet scattering, embedded-cluster encounters, and binary instabilities are likely the predominant FFP production channels. Section 6.1 transparently lists the assumptions behind Table 1 and explicitly acknowledges that apparent microlensing FFPs may include wide-orbit bound planets; however, the Abstract and Section 7 state the conclusions without carrying that conditionality through.","tokens_in":58954,"tokens_out":26581,"duration_ms":233434,"significance":"If the central claim holds, this is a valuable consolidation of a dispersed literature: it unifies four research communities around common quantitative diagnostics (velocity distributions, mass-function features, environmental signatures), updates the earlier Safronov-number analysis with the current exoplanet census (Figures 3 and 4), and offers falsifiable forecasts, including velocity peaks near 2 to 3 km/s for scattering, 8 to 12 km/s for circumbinary and globular-cluster channels, a predicted mass-function trough near 0.8 Earth masses, and a roughly factor-of-ten expansion of the FFP sample with Roman. The analytic kernels (Eqs. 6 to 9, 10, 25 to 30, and 36) are standard and correctly stated; I verified that the stated integral of Eq. (14) over 0.3 to 13 Jupiter masses indeed gives about 0.09 FFPs per star and that the 14.2 AU accretion-limit example from Eq. (7) checks out. The synthesis is not circular: the cited population-level results, including several co-authored by the present authors, are independent N-body simulations with stated initial conditions rather than free parameters fitted here, and the paper does not invent new entities.","major_comments":[{"comment":"The text states that the piecewise power-law IMF of Kroupa (2001) yields 63%, 29%, and 8% of stars in the mass bins 0.075-0.225, 0.225-0.675, and >0.675 solar masses, respectively, and these fractions feed the derived '9% of stars' (p-type) and '10% of stars' (s-type) binary-instability susceptibilities listed in Table 1. Direct integration of the standard Kroupa slopes (alpha = 1.3 for 0.08-0.5 solar masses and alpha = 2.3 above 0.5 solar masses), normalized over 0.075-120 solar masses, gives approximately 50/35/16 for these same bins, and I could not reproduce the quoted 63/29/8 regardless of whether the substellar (0.01-0.075 solar masses) regime is included. Please recompute the star-count fractions from the cited IMF and state the normalization convention and mass range explicitly, or cite the source from which the 63/29/8 values were taken; the affected Table 1 entries should then be regenerated.","section":"Section 6.1 / Table 1"},{"comment":"Section 6.1 transparently acknowledges that 'apparent FFPs observed through microlensing are not necessarily unbound' and cites Hadden and Wu (2025) predicting that roughly half of Neptune-mass apparent FFPs may reside on wide bound orbits, yet this ambiguity is not propagated to the load-bearing inference: the Abstract and the first bullet of Section 7 assert that the low-mass FFP census 'implies a high rate of planetary ejection' without the condition, and the cross-check in Section 2.4.1 (roughly one giant planet ejected per instability) together with the 'can account for all candidate FFPs' statement in Section 6.1 treat the Sumi et al. (2023) normalization of Eq. (14) as an unbound census. Because the required ejection efficiency scales linearly with the assumed unbound fraction, I recommend adding a one-sentence sensitivity analysis (for example, noting that a 50% bound-orbit fraction halves the required efficiency while leaving the qualitative ranking intact) and inserting the conditionality into the Abstract or Section 7 where the central claim is stated.","section":"Section 6.1; Abstract; Section 7"}],"minor_comments":[{"comment":"In the sentence 'for e0 = 0 and the limiting case cos f0 -> 0, more than half of the stellar mass must be lost instantaneously to eject a planet,' the qualifier 'cos f0 -> 0' is unnecessary and potentially confusing: for e0 = 0, Equation 36 reduces to beta < 1/2 for every value of f0, so the limiting-case phrase should be removed or reworded.","section":"Section 5.4"},{"comment":"The '% stars' entries for evolved open clusters (10%), globular clusters (0.5%), AGB-phase evolution (5%), and Type II supernovae (0.1-1%) are not derived anywhere in Section 6.1, which supplies stated assumptions only for the binary and cluster-encounter rows; please add a sentence or footnote tracing each of these percentages to its source or to an explicitly stated stellar-population model.","section":"Table 1 / Section 6.1"},{"comment":"P. Mroz, A. Udalski, J. Skowron, et al. (2017, Nature 548, 183) appears twice as an identical entry in the reference list; please deduplicate, and verify that the in-text citations in Sections 1 and 6.2 refer to the same paper or, if not, disambiguate them with distinct years or identifiers.","section":"References"},{"comment":"The sentence 'Roughly a dozen microlensing FFP candidates have been reported to date that are attributed to low-mass (super-Earth-mass or below) planets' is followed by a list of fourteen citations; please reconcile the 'roughly a dozen' count with the length of the list, or clarify which of the cited events are considered low-mass candidates.","section":"Section 6.2"},{"comment":"The phrase 'at of order ≈20 FFPs' in the first section combines 'of order' with 'approximately' and reads awkwardly; consider revising to 'of order 20 FFPs' or 'approximately 20 FFPs'.","section":"Section 1"}],"recommendation":"major_revision","confidential_remarks":"This is a strong review paper, and the two major comments are both actionable within the manuscript's scope. I verified several quantitative claims en route: the integral of Eq. (14) over 0.3-13 Jupiter masses reproduces the stated 0.09 FFPs per star, and the 14.2 AU Safronov-number accretion limit checks out; however, I could not reproduce the 63/29/8 IMF star-count fractions in Section 6.1 by direct integration of the cited Kroupa (2001) IMF, so the authors should be asked to supply their normalization or correct the numbers before the Table 1 percentages can be considered final. The heavy reliance on very recent (2025-2026) references, several co-authored by the present authors, is understandable in a fast-moving field, but the duplicate Mroz et al. (2017) entry suggests a final pass on the reference list is warranted. The paper fits the journal's scope and, once the conditionality of the microlensing anchor is propagated and the Table 1 provenance is completed, should be acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Malena Rice, Will DeRocco, and Sean Raymond have written the review that the field needed: a clear, comparative account of the four mechanisms that can eject planets (planet-planet scattering, binary instabilities, stellar flybys, post-main-sequence evolution). Its main contribution is Table 1, which puts predicted ejection rates and excess velocities side by side, and the accompanying discussion of which observations would discriminate between mechanisms. That table alone will be cited often. The updated Safronov-number plot (Figure 3) is a nice confirmation of Ford & Rasio (2008) with a larger sample, but it's not a new result. The analytic material—Safronov number, AMD, variable-mass two-body equations—is standard and correctly stated.\n\nThe soft spots are real but mostly acknowledged. The biggest is the empirical anchor: the claim that low-mass FFPs are mostly ejected planets rests on the microlensing census being truly unbound. The paper says (Section 6.1) that apparent FFPs may be wide-orbit bound planets, and cites Hadden & Wu (2025) suggesting about half of Neptune-mass candidates could be bound. But that uncertainty is not propagated into Table 1 or the summary, which states that planet-planet scattering, embedded cluster encounters, and binary instabilities are 'likely predominant.' If the unbound fraction is much lower, those relative rankings lose their basis. The authors should have either added a column of caveats or made the conditional wording more prominent. Still, they did flag it, and the qualitative conclusion—that ejection is common—is probably robust.\n\nA minor issue: Table 1 has no uncertainties, and some entries rely on formation assumptions (e.g., that planets form in unstable regions) that are stated in the text but not reflected in the table. This makes the rates look firmer than they are. But the text does qualify them.\n\nOverall, this is a useful review, not a groundbreaking result. It deserves a serious referee and will be a standard reference for people working on free-floating planets and upcoming microlensing surveys. I'd bring it to a reading group to discuss the table and the observational discriminators. For my own work, I'd cite it as the current synthesis of ejection rates.","headline":"A useful, honest review that consolidates ejection mechanisms and gives observers a comparative table; the central claim is conditional but explicitly flagged.","tokens_in":59378,"tokens_out":2268,"would_cite":true,"duration_ms":20548,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Planetary ejection is a common, near-universal stage of planetary system evolution, and the observed census of free-floating planets is best explained by scattering, binary instabilities, and cluster flybys rather than by in-situ formation.","keywords":["free-floating planets","planetary ejection","planet-planet scattering","binary star instabilities","stellar flybys","post-main-sequence evolution","gravitational microlensing","exoplanet dynamics"],"falsifier":"Imaging the lens fields of a few dozen low-mass microlensing FFP candidates several years after the events, or searching archival precovery images, and finding host stars for a substantial fraction would show that many free-floating planets are actually bound, undercutting the census anchor. Conversely, a velocity sample from parallax measurements that clustered at 8–12 km/s rather than the 2–6 km/s predicted by planet-planet scattering would point to circumbinary or cluster mechanisms as dominant.","tokens_in":1691,"feed_emoji":"🪐","tokens_out":1972,"duration_ms":88371,"temperature":0.7,"pith_summary":"Free-floating planets are common—roughly twenty per star in the completeness-corrected microlensing census—and this review argues that most of the low-mass ones are not born isolated but are planets ejected from their birth systems. It organizes the ejection process into four dynamical channels: planet-planet scattering, instabilities in binary and multi-star systems, flybys in dense stellar environments, and post-main-sequence stellar mass loss. Each channel leaves distinguishable signatures in the masses, velocities, and environments of the ejected planets, and the review's synthesis concludes that scattering, embedded-cluster encounters, and binary instabilities are the likely predominant producers. If that synthesis is right, the present-day planetary census records a history of violence: eccentric giant planets, tightly packed systems, and the rarity of resonant chains are all vestiges of past ejections.","feed_headline":"Most rogue planets were ejected from their birth systems","feed_subtitle":"Microlensing counts about 20 free-floating planets per star; the review shows scattering, cluster encounters, and binaries can explain them.","key_machinery":"The carrying structure is a set of quantitative diagnostics, one per ejection channel, that convert a scenario into a testable prediction. The Safronov number $\\Theta \\equiv v_{\\rm esc,p}^2/v_{\\rm esc,*}^2$—the square of the ratio of a planet's surface escape velocity to the local escape velocity from its host star—decides whether close encounters end in collision or ejection. The angular momentum deficit (AMD) measures how far a planetary system is from circular, coplanar orbits and sets the threshold for secular chaos and orbit crossings. For binaries, empirical stability boundaries give the outermost stable orbit for a circumstellar planet and the innermost stable orbit for a circumbinary planet, while the mass-loss index $\\Psi$ separates adiabatic from impulsive ejection during stellar evolution. Each channel also predicts an excess-velocity distribution, so measured FFP velocities can fingerprint the responsible mechanism.","core_discovery":"This paper's central claim is that planetary ejection is not a rare accident but a standard phase of planetary system evolution: wherever planets form, some are later lost. The abstract states it directly, saying that the ubiquity of free-floating planets suggests that planetary ejection is common, and the summary sharpens it: planet-planet scattering, embedded cluster encounters, and binary instabilities are likely predominant in FFP production. The load-bearing evidence is the completeness-corrected microlensing census of roughly 20 free-floating planets per star for masses 0.33–6660 Earth masses, together with the direct-imaging limit that in-situ formation tapers off below about 3–5 Jupiter masses. The paper argues that the four reviewed mechanisms can account for all candidate FFPs, while acknowledging that some apparent FFPs may be wide-orbit bound planets rather than truly unbound ones.","pith_inferences":["The velocity-prediction table acts as a ready-made classification scheme: if Roman-era parallaxes yield a bimodal FFP velocity distribution, the peaks can be mapped directly onto ejection mechanisms—a test the paper describes but does not itself run.","The review's relative rates assume planets form in unstable regions such as circumbinary zones and wide binary orbits; if future disk surveys show that planet formation avoids these zones, the binary and tidal channels would shrink dramatically.","The same machinery applied to moons and planetesimals suggests that interstellar objects should share the velocity fingerprints of FFPs, so a joint census of interstellar objects and FFPs could test the universality of the ejection picture."],"forward_implications":["If the completeness-corrected census is real, the ejected mass budget of about 80 Earth masses per star implies that protoplanetary disks must be heavier than the classic minimum-mass nebula, or the observed FFPs cannot all be explained by core accretion.","Present-day giant-planet systems should carry vestiges of past ejection: high eccentricities, packed spacings near the stability limit, and a dearth of mean-motion resonances.","Upcoming microlensing surveys are projected to find hundreds of FFPs, and parallax-measured velocities could separate ejection channels by their typical speeds.","Directly imaged super-Jovian FFPs in star-forming regions are consistent with in-situ collapse, while low-mass microlensing FFPs are not, making the low-mass objects the crucial test bed for ejection models.","A fraction of apparent FFPs may be wide-orbit bound planets, so host-star follow-up of microlensing events will decide how much of the census is truly unbound."],"supporting_citations":[{"why":"It supplies the completeness-corrected abundance and mass function of about 20 FFPs per star that anchors the ejection-is-common claim.","marker":"T. Sumi et al. 2023"},{"why":"It provides the match between giant-exoplanet eccentricities and high instability rates that supports scattering as a common process.","marker":"M. Jurić & S. Tremaine 2008"},{"why":"It shows that 30–60% of wide-binary systems with four giant planets eject at least one planet, forming the basis for the galactic-tide channel.","marker":"N. A. Kaib et al. 2013"},{"why":"It measures embedded-cluster ejection rates up to tens of percent for wide giant planets, forming the basis for the cluster rows in Table 1.","marker":"R. J. Parker & S. P. Quanz 2012"},{"why":"It provides the population synthesis connecting single- and binary-star instabilities to the FFP mass function that the paper uses for its relative-rate estimates.","marker":"G. A. L. Coleman & W. DeRocco 2025"},{"why":"It supplies the empirical stability boundaries for circumstellar and circumbinary orbits that the paper uses to compute binary instability thresholds.","marker":"M. J. Holman & P. A. Wiegert 1999"},{"why":"It predicts that about half of Neptune-mass FFP candidates may be bound on wide orbits, the main caveat on the microlensing census.","marker":"S. Hadden & Y. Wu 2025"},{"why":"It defines the mass-loss index and the ejection criteria for adiabatic and impulsive mass loss, serving as the backbone of the post-main-sequence section.","marker":"D. Veras et al. 2011"}],"fun_headline_variants":["Planet ejection is common, not exceptional","Most rogue worlds were born and then cast out","Scattering and binaries eject planets from systems","Free-floating planets are the ejected offspring of systems"],"cache_read_input_tokens":61568,"weakest_assumption_plain":"The load-bearing premise is that most microlensing free-floating planet candidates are truly unbound, so the inferred twenty per star is a real ejection count rather than a count of planets still orbiting faint host stars on very wide orbits.","fun_headline_variants_meta":{"raw":{"variants":["Planet ejection is common, not exceptional","Most rogue worlds were born and then cast out","Scattering and binaries eject planets from systems","Free-floating planets are the ejected offspring of systems"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000493,"raw_usage":{"total_tokens":2360,"prompt_tokens":824,"completion_tokens":1536,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":440,"completion_tokens_details":{"reasoning_tokens":1478}},"tokens_in":440,"tokens_out":1536,"duration_ms":16062,"temperature":1.0,"reasoning_tokens":1478,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:02:56.305106+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Imaging the lens fields of a few dozen low-mass microlensing FFP candidates several years after the events, or searching archival precovery images, and finding host stars for a substantial fraction would show that many free-floating planets are actually bound, undercutting the census anchor. Conversely, a velocity sample from parallax measurements that clustered at 8–12 km/s rather than the 2–6 km/s predicted by planet-planet scattering would point to circumbinary or cluster mechanisms as dominant.","supporting_citations":[],"review_version":2}