{"id":"d0bfe398-5ce3-41ac-a666-cc9da337d2e6","arxiv_id":"2501.03506","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":12,"one_line_summary":"Adding binary and triple systems to microlensing population simulations predicts that 55% of OGLE-like events involve a multiple system, most of which masquerade as single-star events.","lead":"Binary and triple star systems may be involved in more than half of the microlensing events seen by surveys like OGLE, according to a new population simulation. Most of these events look like ordinary single-star events, and including them shifts the predicted event duration distribution toward the observed one.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The claimed tE shift (19.1→21.3 d) is driven by assigning binary lens events a tE based on total system mass, an approximation invalid for the ~50% of wide binaries with projected separation > θE; the apparent improvement over OGLE may be partly an artifact.","rationale":"The reader's weakest_assumption identifies exactly the total-system-mass treatment of binary-lens tE, and my independent reading converges on the same point. This assumption is load-bearing because the abstract's headline numbers (19.1→21.3 d shift, better agreement with OGLE) are the quantitative evidence that multiples are a missing ingredient. The paper's own statement that ~50% of binary lenses have projected separation exceeding θE makes the total-mass approximation internally inconsistent: the code extends its search radius to accommodate wide binaries but still assigns them a point-mass tE based on the sum of both masses. Since Table 6 shows binary-lens events have mean tE = 37.8 d versus 16.8 d for single-lens events, this approximation is precisely what creates the long-tE tail that improves the KS comparison. If that tail is an artifact, the central claim weakens to 'binaries are common and may be misclassified,' which is plausible but is not the same as demonstrating that binaries reconcile simulated and observed tE distributions. The paper has real strengths: the simulations are public, the lightcurve fitting with BAGLE is a reasonable attempt to mimic OGLE selection, and the fraction of obvious multi-peaked events (3.1%) matches observed binary detection rates. These support the qualitative conclusion and justify a conditional rather than a reject verdict. The concern does not change the reader's CONDITIONAL verdict, so UNCHANGED is appropriate. The proposed concrete test directly settles whether the tE shift is physical or an artifact.","tokens_in":28693,"tokens_out":6299,"duration_ms":58935,"concrete_test":"Recompute the single-peaked M Run tE distribution using only the primary lens mass (not M_sys) for all PSBL/BSBL events with projected separation a sin(i) > θE, while keeping all other cuts fixed. Then recompute the mean/median tE and the KS p-value against Mróz et al. (2017). If the mean tE drops back toward ~19 d and the KS p-value moves closer to the S Run value (3e-10), the claimed improvement is an artifact of the total-mass approximation. A stronger check is to fit the simulated binary-lens lightcurves with a proper binary-lens model and use the fitted tE; if the resulting M Run distribution no longer improves agreement with OGLE, the central quantitative claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central quantitative claim is that adding multiples shifts the mean Einstein crossing time from 19.1 d to 21.3 d and improves agreement with the OGLE tE distribution (Table 6, Fig. 6). In PopSyCLE, tE for any event with a multiple lens (PSBL, BSBL, triples) is computed from Eq. 7 using the total system mass M_sys (Sec. 2.2.2, Table 9). This is only valid when the binary separation is much smaller than the Einstein radius, so the system behaves as a point mass. The paper itself states that the projected binary separation is larger than θE approximately 50% of the time (Sec. 2.2.2), which is why the search radius is extended. For such wide binaries, the companion does not contribute to the magnification of the source passing near the primary; the event tE should be based on the lens mass that actually produces the event (approximately the primary), not the total system mass. Using M_sys inflates tE by a factor sqrt(1+q), up to sqrt(2) for equal masses. Table 6 shows that M Run Mult Lens events have mean tE = 37.8 d versus 16.8 d for M Run Sing Lens events, and this population drives the overall mean shift. The improved KS p-value (9.34e-4 vs 3.06e-10) and the stated better alignment at long tE may therefore be substantially an artifact of the total-mass approximation, not a physical consequence of binarity. The paper acknowledges this bias in Sec. 6.2 for 'massive, distant, and unlensed companions' but does not quantify it, and the abstract and conclusion present the tE shift as evidence that multiples are a missing ingredient. This is the most load-bearing weakness because the quantitative validation of the central claim rests on it. The qualitative statement that binaries are common and often masquerade as PSPL events is less affected, but the specific 55% and tE-shift numbers are model-dependent until this is corrected.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"Abrams et al. extend the SPISEA/PopSyCLE population-synthesis pipeline to include binary and triple systems, simulate mock OGLE-IV microlensing surveys, and analyze how multiples affect event statistics. They report that 55% of events passing OGLE-like cuts involve a multiple lens, source, or both, that most such events are single-peaked and well fit by point-source-point-lens models, and that including multiples shifts the mean Einstein crossing time from 19.1 days to 21.3 days, improving agreement with the OGLE tE distribution (KS p = 9.34e-4 vs 3.06e-10). The paper also examines biases in light-curve fitting, the selection of binary parameters by microlensing, and the impact on black-hole astrometric follow-up selection.","tokens_in":29070,"tokens_out":7025,"duration_ms":63046,"significance":"If the central results hold, this is a valuable contribution. The authors provide a public, extensible simulation tool; they use external calibrations for binary statistics rather than fitting to the target distribution; and they make a concrete, falsifiable prediction that multiples are ubiquitous in microlensing samples and largely masquerade as single events. The external comparison to OGLE is appropriate, and the KS test provides a useful summary statistic. The negative result that binaries do not affect black-hole candidate selection is also useful. The main caveat is that the quantitative tE shift and the 55% fraction rest on modeling choices—total-system-mass tE for binary lenses and solar-neighborhood multiplicity statistics—that are acknowledged but not tested for robustness.","major_comments":[{"comment":"The central tE-shift result depends on assigning binary-lens events a tE based on the total system mass, Msys = M1 + M2, for all systems. This is unphysical for the ~50% of binaries whose projected separation exceeds the Einstein radius, as the paper itself notes in Sec. 2.2.2. For such wide binaries the companion does not contribute to the magnification of a source passing near the primary, so the event timescale should be set by the primary mass (or by the component actually responsible for the lensing), not by the total mass. Using Msys inflates tE by a factor sqrt(1+q), and Table 6 shows that M Run Mult Lens events have mean tE = 37.8 d versus 16.8 d for M Run Sing Lens events, so this approximation drives the reported shift from 19.1 d to 21.3 d and the improved KS p-value. The caveat in Sec. 6.2 about a 'massive, distant, and unlensed companion' is not sufficient; the paper needs a quantitative estimate. I request a rerun or post-processing correction in which wide binaries (projected separation > thetaE) are assigned tE based on the primary mass, and the resulting mean tE, Fig. 6, and KS p-value be reported.","section":"Sec. 2.2.2, Eq. (7), Table 9, Table 6, Sec. 5.2"},{"comment":"The multiplicity fractions and companion statistics are taken from the solar neighborhood (Lu et al. 2013; Duchene & Kraus 2013) and applied to the Galactic bulge. The quantitative breakdown in Table 5 (14.5% PSBL, 31.7% BSPL, 8.8% BSBL) is therefore only as reliable as that extrapolation. The paper acknowledges that the parameters are uncertain but does not explore how the 55% total or the tE shift respond to plausible variations in the multiplicity-fraction normalization (A, alpha), the companion-star-fraction normalization (B, beta), or the mass-ratio index (gamma). Since 'over half of observable events involve a multiple system' is a headline claim, a sensitivity analysis over these input parameters is needed to establish robustness.","section":"Sec. 2.1, Table 5"},{"comment":"The treatment of triple systems is approximate: for triples, the paper simulates only the primary-companion pair with the largest Delta m, rather than the full triple lens/source configuration. This approximation can affect the single-peaked versus multi-peaked classification and thus the comparison of the simulated tE distribution to OGLE in Sec. 5.2. Appendix C provides a partial test, but it assumes that correctly treated triples would yield multi-peaked or unobservable events, and it does not propagate the resulting classification changes through the KS test. Given that triples contribute about 10% of events and about half of the tE > 30 d difference between M Runs and S Runs, a more rigorous treatment, or at least a bracketing calculation, would materially strengthen the main conclusion.","section":"Appendix B, Appendix C, Sec. 6.2"}],"minor_comments":[{"comment":"The KS p-value for M Runs (9.34e-4) is still very small, indicating that the simulated distribution remains statistically inconsistent with OGLE even after including multiples; the text should acknowledge this and not overstate the level of agreement.","section":"Sec. 5.2"},{"comment":"Typo: 'standard Keplarian distributions' should read 'standard Keplerian distributions.'","section":"Sec. 2.1"},{"comment":"The sentence 'we accept that the estimates we use for tE, which are based on system mass, are sufficient' seems to preempt the very concern raised in Sec. 6.2; this statement should be revised to reflect the quantitative caveat.","section":"Sec. 5.3"},{"comment":"The caption lists triple fractions for PSBL/BSPL/BSBL but does not reference Table 10; adding a cross-reference would improve clarity.","section":"Fig. 3 caption"}],"recommendation":"major_revision","confidential_remarks":"The key concern is the total-system-mass tE assignment for wide binary lenses. If the authors can show, after correcting for this approximation, that the tE shift and the improved KS p-value persist, the paper would be close to publishable. As it stands, the headline quantitative claim is not yet supported. The reliance on self-citations (PopSyCLE, Lu et al.) is legitimate in context, but the authors should be careful not to overstate the novelty of the binary-fraction result given the direct dependence on solar-neighborhood calibrations."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper adds binaries and triples to the SPISEA/PopSyCLE pipeline and asks how often they show up in OGLE-like microlensing samples. The main new numbers are the event fractions: 55% of mock-OGLE events involve a multiple lens, source, or both (14.5% PSBL, 31.7% BSPL, 8.8% BSBL), and most of those are single-peaked and would be fit well by a PSPL model. That qualitative result is credible and worth taking seriously. The code and simulated data are public, and the pipeline is described in enough detail to reproduce.\n\nThe soft spot is the Einstein crossing time shift that anchors the abstract and conclusions. They compute tE for any multiple-lens event from the total system mass, as if the companion were a point mass at the primary. But they also state that the projected binary separation is larger than theta_E about 50% of the time. For those wide binaries the companion is not part of the lensing, so the event tE should scale with the primary mass, not sqrt(M_primary+M_companion). Using M_sys inflates tE by up to sqrt(2) for equal masses and is exactly what produces the long-tE tail in the M Run multi-lens population (Table 6: mean 37.8 d vs 16.8 d for single lenses). The improved KS p-value against OGLE (1e-3 vs 3e-10) is therefore partly an artifact of this approximation. The paper acknowledges the bias in Sec 6.2 for \"massive, distant, and unlensed companions\" but does not quantify it, and the abstract and conclusion treat the tE shift as evidence. That is my main disagreement with the reader's softer framing: the qualitative \"multiples are missing\" message holds, but the specific tE validation is load-bearing and unreliable.\n\nOther concerns are lesser: the binary statistics come from local populations (Lu 2013, Duchene & Kraus 2013) and may not apply to the bulge; binaries are static (no orbital motion, no binary evolution); triples are handled approximately. The paper is honest about all of these. The field-by-field event rates scatter widely, but the averages are reasonable.\n\nWho benefits: microlensing modelers interpreting OGLE, Roman, and Rubin events. The fractions and the finding that most multiples masquerade as singles are useful even if the tE shift is suspect. This deserves a serious referee; the approximation issue should be quantified or corrected before publication, but it is not a desk-reject.","headline":"Credible simulation result that most OGLE-like events involve a multiple system, but the headline tE shift is driven by a total-mass approximation the paper itself flags.","tokens_in":29806,"tokens_out":2862,"would_cite":true,"duration_ms":25279,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Most microlensing events involve binary stars, not single stars alone.","keywords":["microlensing","binary stars","multiple systems","Einstein crossing time","population synthesis","OGLE","gravitational lensing","compact objects"],"falsifier":"A direct test would be to compare the predicted fraction of multi-peaked events (3.1% of all events under OGLE cuts) with a large, completeness-corrected sample from OGLE or KMTNet that accounts for cadence gaps. If the observed multi-peaked fraction is significantly lower (or higher) than 3.1% after such corrections, or if the recovered $t_E$ distribution does not shift by the predicted 2.2 days when binaries are included, the assumed binary population parameters or the system-mass approximation would be ruled out.","tokens_in":1579,"feed_emoji":"🌟","tokens_out":2946,"duration_ms":78134,"temperature":0.7,"pith_summary":"This paper argues that binary and triple star systems are not a minor correction to microlensing surveys but the majority of what such surveys actually see, and that neglecting them systematically distorts the inferred populations. By adding realistic multiple systems to the SPISEA stellar population code and injecting them into the PopSyCLE microlensing simulation, the authors find that 55% of events passing OGLE-like selection criteria contain a multiple lens, a multiple source, or both. Most of these events have light curves that look single, so they are routinely misclassified as single-star events. Including the multiples shifts the mean Einstein crossing time from 19.1 days to 21.3 days, bringing the simulated distribution into better agreement with observed OGLE data. If correct, this means that current single-star interpretations of microlensing surveys underestimate the prevalence of binaries and bias the inferred mass and event-duration distributions.","feed_headline":"Most microlensing events involve binary stars","feed_subtitle":"Adding binaries to simulations shifts the typical event duration by 2.2 days, matching observed OGLE data.","key_machinery":"The central machinery is the injection of physically motivated multiple-star systems into a Milky Way microlensing simulation. The paper extends SPISEA (a stellar population synthesis code) to generate companions with semi-major axes, eccentricities, and orbital orientations drawn from empirical distributions (e.g., a broken power law for separation versus primary mass based on Duchêne & Kraus 2013), and then matches these systems onto stars in PopSyCLE (a microlensing survey simulator built on the Galaxia Milky Way model). Each event's light curve is then computed for the multiple system, using a point-source–point-lens model for single events but a binary light-curve treatment for multiples, and OGLE-like detection cuts are applied to determine which events would actually be observed. The key step is that the Einstein crossing time for a multiple lens is computed using the total system mass, so adding companions directly lengthens $t_E$; this is what produces the shift from 19.1 to 21.3 days.","core_discovery":"The paper's central claim is that, under OGLE-like observational cuts, over half (55%) of microlensing events involve a binary or triple system as the lens, source, or both, specifically 14.5% with a multiple lens and single source, 31.7% with a single lens and multiple source, and 8.8% with both multiple. The great majority (94.4%) of these multiple events have only a single observable peak in their light curves, so they are easily mistaken for ordinary point-source, point-lens events. When these single-peaked multiples are included in the simulated event population, the mean Einstein crossing time $t_E$ shifts from 19.1 days (singles only) to 21.3 days, and the distribution becomes significantly more consistent with the observed OGLE $t_E$ distribution (the Kolmogorov–Smirnov p-value improves from $3.06\\times10^{-10}$ to $9.34\\times10^{-4}$). The paper concludes that multiple systems are a substantial missing piece in microlensing population synthesis and that binary-source and binary-lens–binary-source models should be routinely included in event analysis.","pith_inferences":["If the 55% multiple fraction holds for deeper surveys such as the Vera Rubin Observatory or the Roman Space Telescope, the implied bias in the inferred mass distribution of lenses could be even larger, because these surveys probe fainter sources where binary companions contribute a smaller fraction of the light.","The paper's treatment of triples by selecting the two-body pair with the largest magnification may underestimate the fraction of multi-peaked events; a full three-body simulation could turn some single-peaked triples into observable multi-peaked events, which would raise the 'obvious' binary fraction.","The systematic use of total system mass for $t_E$ in wide binaries is likely the weakest link; if that approximation is refined (e.g., computing $t_E$ from the primary's Einstein radius and adding a separate companion signal), the reported 2.2-day shift might shrink or change sign for extreme mass ratios."],"forward_implications":["Surveys that fit only single-lens, single-source models will systematically misclassify the majority of microlensing events, biasing measurements of event rates and durations.","The Einstein crossing time distribution, a key observable for inferring the mass function of compact objects, is significantly different when binaries are included, so population-level conclusions from OGLE and similar surveys need to be revisited.","Binary lenses are preferentially found with separations of 1–10 AU, so microlensing samples will underrepresent wide binaries; the same applies to source binaries, which are biased toward larger separations.","Black hole astrometric candidate selection using the criteria $t_E > 120$ days and $\\pi_E < 0.08$ is unaffected by the presence of multiples, so the search for isolated black holes is robust to this effect.","The fraction of obvious multi-peaked events (3.1% of all events) matches the 2–11% binary fraction reported by surveys, suggesting that the missing multiples are hiding in plain sight as single-peaked events."],"supporting_citations":[{"why":"Introduced the PopSyCLE simulation code that the paper extends to include multiple systems; provides the baseline single-star microlensing simulation and the OGLE-like selection criteria.","marker":"[PopSyCLE, Lam et al. 2020]"},{"why":"The stellar population synthesis code to which the paper adds resolved binary and triple systems, supplying the companion masses, separations, and orbital parameters.","marker":"[SPISEA, Hosek et al. 2020]"},{"why":"The observed OGLE Einstein crossing time distribution that the paper compares against; the key dataset that the multiples-inclusive simulation matches better than the singles-only simulation.","marker":"[Mróz et al. 2017]"},{"why":"Provides the empirical mass-dependent semi-major axis distribution and binary frequency that the paper uses to generate the multiple system populations.","marker":"[Duchêne & Kraus 2013]"},{"why":"Supplies the default multiplicity fraction, companion star frequency, and mass-ratio distribution used by SPISEA to assign companions to primaries.","marker":"[Lu et al. 2013]"},{"why":"Provides the OGLE event rate and efficiency data used to compare the simulated event rates per field and per star.","marker":"[Mróz et al. 2019]"}],"fun_headline_variants":["Hidden binaries drive 55% of microlensing events","Most binary microlensing events appear single","Binaries add 2.2 days to typical microlensing durations","Single-peaked lightcurves often hide binary lenses"],"cache_read_input_tokens":31488,"weakest_assumption_plain":"The paper assumes that the Einstein crossing time of an event with a multiple lens is accurately given by the total system mass, as if all the mass were concentrated at the primary's position, an assumption that the authors themselves note can bias events with a massive, distant, and unlensed companion.","fun_headline_variants_meta":{"raw":{"variants":["Hidden binaries drive 55% of microlensing events","Most binary microlensing events appear single","Binaries add 2.2 days to typical microlensing durations","Single-peaked lightcurves often hide binary lenses"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000366,"raw_usage":{"total_tokens":2048,"prompt_tokens":1105,"completion_tokens":943,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":721,"completion_tokens_details":{"reasoning_tokens":877}},"tokens_in":721,"tokens_out":943,"duration_ms":8868,"temperature":1.0,"reasoning_tokens":877,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T21:52:38.946125+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A direct test would be to compare the predicted fraction of multi-peaked events (3.1% of all events under OGLE cuts) with a large, completeness-corrected sample from OGLE or KMTNet that accounts for cadence gaps. If the observed multi-peaked fraction is significantly lower (or higher) than 3.1% after such corrections, or if the recovered $t_E$ distribution does not shift by the predicted 2.2 days when binaries are included, the assumed binary population parameters or the system-mass approximation would be ruled out.","supporting_citations":[],"review_version":1}