{"id":"2ec8687c-30fe-433e-99fc-d9c94490ffd0","arxiv_id":"2501.14031","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"In TNG100-1, close pericentric passages enhance central sSFR by 1.6x, peaking ~0.1 Gyr later, with about 70% of the boost attributed to increased star formation efficiency rather than increased gas supply.","lead":"Using a large cosmological simulation, this paper tracks 18,534 galaxy encounters and finds that close passages boost central star formation by about 1.6 times, peaking about 0.1 billion years after closest approach. The authors break the boost down and conclude that higher star formation efficiency, not extra gas fuel, is the main driver.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Pericentre-time reconstruction systematics may bias the reported 0.1 Gyr peak delay and 75 kpc peak separation, given the snapshot-periodic structure in the Δt distribution.","rationale":"The reader's weakest_assumption identifies the pericentre-time reconstruction accuracy as the key assumption, and I agree. This is the single most load-bearing concern because it directly affects the quantitative timing claims (peak at ~0.1 Gyr after pericentre, mean separation 75 kpc) and could also bias the enhancement amplitude and the fuelled/efficiency split if the time axis is systematically shifted. The paper itself flags the periodic Δt distribution as systematic, yet no uncertainty is propagated from this into the reported peak time or the 70/30 split. A controlled recovery experiment is a definitive, feasible check: it uses simulations with known orbits to measure the bias of the interpolation method at the same snapshot cadence. If no substantial bias is found, the central claims survive; if bias is found, the conditional verdict should be strengthened. I did not identify a more load-bearing flaw: the efficiency decomposition is mathematically forced and the authors explicitly define SFEH, so it is not internally inconsistent, though its physical interpretation as 'efficiency' is somewhat definition-dependent. The pericentre-time concern, however, threatens the empirical anchor of the entire analysis and is therefore the most appropriate focus for a stress test.","tokens_in":25866,"tokens_out":11455,"duration_ms":100633,"concrete_test":"Run a controlled recovery experiment: take high-resolution merger simulations (e.g., Moreno et al. 2019 or simpler idealized orbits) with known true pericentre times, degrade the outputs to the TNG100-1 snapshot cadence (~0.15 Gyr), apply the same 6D kinematic interpolation used by Patton et al. (2024), and compare the recovered pericentre times and stacked sSFR curves against the truth. If the recovered peak time is systematically offset by more than 0.05 Gyr or the recovered Q(sSFR) amplitude changes by more than 10%, the reported 0.1 Gyr delay and 75 kpc separation should be revised or downgraded. Alternatively, split the 0-50 kpc sample into (a) encounters where the pericentre falls within 0.02 Gyr of a snapshot and (b) those where it falls between snapshots; if the stacked peak time or the fuelled/efficiency split differs between subsamples, interpolation bias is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The stacked analysis in Sections 3.2–3.5 hinges on assigning each snapshot a time relative to pericentre, Δt, using pericentre times from 6D kinematic interpolation between TNG100-1 snapshots that are ~0.15 Gyr apart. The authors themselves note in Section 2.2 and Figure 2 that the Δt distribution shows periodic variations at the snapshot cadence which they call 'likely systematic'. If the interpolation systematically pins pericentre times to preferred phases (e.g., near snapshots or midway between them), then the zero-point of Δt is biased, and the reported peak at +0.12 Gyr, the mean separation of 75 kpc at that peak, and even the smoothed enhancement amplitude Q(sSFR)=1.6 could be artifacts of the time-axis error rather than real orbital-phase behaviour. The boxcar smoothing with a 3000-galaxy bin width does not correct a systematic offset; it merely averages over noise. Because the central claim' s timing and the separation-dependent interpretation in Figure 5 depend directly on these Δt values, this is the most load-bearing assumption. The decomposition into fuelled vs efficiency is definition-dependent but transparently stated; the time-axis systematics, by contrast, affect the raw data placement and are acknowledged but unquantified in the error budget.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper uses the TNG100-1 cosmological simulation to construct a catalogue of 18,534 distinct pericentric encounters from orbit reconstructions of Patton et al. (2024). Stacking galaxy properties relative to pericentre time, it reports that within R1/2 the sSFR of hosts in the r_peri <= 50 kpc bin is enhanced by 1.6 ± 0.1 relative to the pre-encounter value, peaking about 0.1 Gyr after pericentre at a mean pair separation of about 75 kpc. The same stacking gives fgas and SFEH enhancements of 1.2 ± 0.1 and 1.4 ± 0.1, and a log-space decomposition of sSFR = fgas x SFEH is used to attribute roughly 70 per cent of the peak sSFR enhancement to increased SFEH and 30 per cent to increased fgas. The paper also reports central metallicity dilution and an outer-to-inner time lag, interpreted as evidence of inflow of pristine gas toward the centre.","tokens_in":25941,"tokens_out":7871,"duration_ms":71209,"significance":"If the timing result is robust, the paper provides a valuable bridge between observational pair catalogues and high-resolution merger simulations: a large, cosmologically representative sample of interactions with properties measured as a function of orbital phase rather than projected separation. The paper has genuine strengths: a large sample size, transparent discussion of selection effects, dropout-controlled comparison of snapshot-limited subsamples (Fig. 6), a first-passage robustness check (Appendix A), and an explicit statement of the indirect nature of the inflow evidence. The central fuelling-versus-efficiency result, however, rests on two methodological choices that need tightening before the quantitative claim can be accepted.","major_comments":[{"comment":"The stacked time axis is built from pericentre times reconstructed by 6D kinematic interpolation between snapshots separated by ~0.15 Gyr. The right panel of Figure 2 shows periodic structure in the Delta-t distribution at this cadence, and the text states that these variations \"are likely systematic.\" Because the zero-point of Delta-t enters every stacked curve, a systematic phase error in the reconstructed pericentres would both dilute the peak amplitude and bias the reported peak time of +0.12 Gyr and the corresponding mean separation of 75 kpc in Figures 4, 5, and 7. This is not a cosmetic issue; it directly affects the headline timing result. Please quantify the sensitivity, for example by Monte Carlo perturbing each pericentre time by up to half a snapshot interval, by repeating the stacking with pericentre times fixed to snapshot times, or by splitting the sample according to where in the snapshot interval the pericentre falls, and report the resulting changes in Q(sSFR), Q(fgas), Q(SFEH), the peak time, and the peak separation.","section":"Section 2.2, Figure 2"},{"comment":"SFEH is not measured directly per galaxy; it is computed from the ratio of the sample-averaged sSFR to the sample-averaged fgas (Section 3.3). For individual galaxies sSFR_i = SFEH_i x fgas_i, but the mean of a product is not the product of the means, so this ratio is an fgas-weighted mean SFEH whose weighting can evolve during the encounter. The log-space decomposition in Eqs. (7)-(11) is then an exact identity for the averaged Q values rather than a statement about the average of per-galaxy efficiencies, and the reported 70 per cent / 30 per cent efficiency/fuelling split could change if the covariance between SFEH and fgas evolves with time. Please compute SFEH directly per galaxy (for example SFR/MH) and compare the average of per-galaxy Q(SFEH) with the ratio-derived Q(SFEH), or otherwise demonstrate that the weighting effect is negligible.","section":"Section 3.3, Eq. (4)"}],"minor_comments":[{"comment":"The bullet \"At least 4 snapshots between the selected pericentre, and either the next pericentre, the merger...\" is ambiguous; it should be reworded to make clear whether this is a required temporal gap before or after the selected pericentre.","section":"Section 2.2, bullet list"},{"comment":"The caption says \"mean pair separation bounded by the 25th and 75th percentiles,\" which conflates a mean with a percentile range; please clarify that the shaded band spans the interquartile range of separations around the mean.","section":"Figure 5 caption"},{"comment":"The definition of SFEH as (SFR/MH2) x (MH2/MH) is unnecessarily roundabout; it reduces identically to SFR/MH, and simplifying the definition would avoid the impression that molecular gas directly enters the quantity actually measured.","section":"Section 3.1, Eq. (3)"},{"comment":"The \"70 per cent\" attribution is obtained from a log-space decomposition (Eq. 9); the abstract and conclusions should state this explicitly so that readers do not interpret it as a linear additive decomposition of the sSFR enhancement.","section":"Abstract and Section 3.5"},{"comment":"The shaded regions in Figure 11 are described as 2-sigma errors from a Jackknife technique, while Figures 4, 7, and 8 use the 2-sigma standard error in the mean; please specify the resampling unit and label the two error definitions consistently.","section":"Figure 11"},{"comment":"The data availability statement mentions only the public IllustrisTNG data; if possible, please also state whether the new 18,534-encounter catalogue will be released, as it would be a valuable community resource.","section":"Data Availability"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nQuick take on Faria et al.: the paper delivers the first pericentre-resolved, large-sample look at how galaxy encounters change star formation in a cosmological simulation. The authors stack 18,534 distinct encounters from reconstructed orbits, find that sSFR within R1/2 peaks at 1.6±0.1 times pre-encounter values about 0.1 Gyr after pericentre, and split that peak into fuel (fgas, 1.2×) and efficiency (SFEH, 1.4×) contributions. The measurement itself is careful: they do dropout tests for encounters of different lengths, check first-passage versus all-passage subsets, and openly discuss selection effects. The pericentre-stacking methodology is new relative to the earlier papers in the series, and the result that efficiency appears to dominate over fuelling is worth taking seriously.\n\nThe soft spots are real but not fatal. The most load-bearing is the time axis. The Δt distribution shows periodic structure at the snapshot cadence (~0.15 Gyr), and the authors themselves say it is 'likely systematic'. Their pericentre times come from 6D kinematic interpolation between snapshots, so if the interpolation pins pericentres to preferred phases between snapshots, the reported 0.1 Gyr delay and the 75 kpc mean separation at peak are directly affected. The enhancement amplitude is probably less sensitive, but a systematic offset would smear the stacked peak and could shift Q(sSFR). They don't quantify this in the error budget. It is a legitimate concern, and testable—bootstrap the pericentre times or validate against TNG50's finer cadence.\n\nThe fuel/efficiency decomposition is transparent but partly definitional: since sSFR = fgas × SFEH, the log-space split is a restatement of the measured curves. The 70/30 claim has no error bars, and the galaxy-by-galaxy version gives a much weaker 57/43 margin. That doesn't make the paper circular—the enhancement curves themselves are empirical from the simulation—but the headline 'efficiency dominates' should be read with the caveat that SFEH here includes the assembly of star-forming gas, not just the conversion efficiency.\n\nMinor: the encounter catalogue and code aren't released, so reproducibility is limited to the TNG public data. The self-citation chain is grounded in previously published, publicly based work, so that doesn't bother me.\n\nBottom line: the central measurement is a worthwhile contribution and deserves a serious referee. I'd ask the authors to quantify the pericentre-time uncertainty and soften the 70/30 claim accordingly. Bring it to the group if you want to discuss stacking methodology; I'd cite it for the pericentre-resolved enhancement amplitude.","headline":"Solid pericentre-stacked measurement of interaction-triggered star formation in TNG100-1, with a real but unquantified time-axis systematic and a definition-dependent fuel/efficiency split.","tokens_in":26655,"tokens_out":2802,"would_cite":true,"duration_ms":25464,"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":"In close galaxy flybys inside the TNG100-1 simulation, the central star formation burst is mostly driven by higher star formation efficiency (about 70 per cent of the boost), not by extra gas supply.","keywords":["galaxy interactions","pericentric encounters","star formation efficiency","gas fraction","specific star formation rate","IllustrisTNG","TNG100-1","galaxy pairs"],"falsifier":"Recompute pericentre times directly from the simulation's particle velocities instead of the interpolated orbits and re-stack the 18,534 encounters; if the $1.6 \\pm 0.1$ peak, the roughly 0.1 Gyr delay, and the 70/30 efficiency-fuel split shift by more than the quoted uncertainties, the signal is an artefact of the time axis. Running the same stack on the higher-time-resolution TNG50 run, which the paper lists as future work, would settle the question directly.","tokens_in":25510,"feed_emoji":"🌌","tokens_out":12465,"duration_ms":94662,"temperature":0.7,"pith_summary":"This paper asks what actually powers the burst of star formation when a galaxy swings close past a companion: fresh gas arriving at the centre, or the central gas becoming better at turning itself into stars. Using 18,534 distinct pericentric encounters reconstructed from the TNG100-1 cosmological simulation, it stacks galaxy properties on a common clock measured from closest approach and finds that for the closest encounters ($r_{\\mathrm{peri}} \\le 50$ kpc) the central specific star formation rate rises by a factor of $1.6 \\pm 0.1$, peaking roughly 0.1 Gyr after pericentre when the pair is about 75 kpc apart. Because $\\mathrm{sSFR} = \\mathrm{SFE_H} \\times f_{\\mathrm{gas}}$, the enhancement splits cleanly in log space into a fuelling term and an efficiency term, and the split attributes about 70 per cent of the peak boost to a rise in star formation efficiency and only 30 per cent to a rise in gas fraction. The question matters because fuelling versus efficiency is currently contested in observations and merger simulations, and this is a large-sample, cosmological-context answer with an explicit phase-resolved decomposition.","feed_headline":"Galaxy flybys boost star formation via efficiency, not fuel","feed_subtitle":"Central star formation rises 1.6x in close encounters, and about 70% of the boost comes from higher efficiency.","key_machinery":"The load-bearing device is the identity $\\mathrm{sSFR} = \\mathrm{SFE_H} \\times f_{\\mathrm{gas}}$, with $\\mathrm{SFE_H} \\equiv \\mathrm{SFR}/M_{\\mathrm{H}}$ defined as star formation per unit total hydrogen mass in all gas phases, chosen because TNG100-1 does not report molecular gas mass. Taking logarithms turns the enhancement ratio into a sum, $\\log Q(\\mathrm{sSFR}) = \\log Q(f_{\\mathrm{gas}}) + \\log Q(\\mathrm{SFE_H})$; normalising by $\\log Q(\\mathrm{sSFR})$ splits the fractional rise $\\delta\\mathrm{sSFR}/\\mathrm{sSFR_i}$ into a fuelled term and an efficiency term that sum to exactly one. Applied to stacked averages phased on reconstructed pericentre times from 6D kinematic interpolation of the pair orbits, this decomposition is what converts two correlated boosts into the quantitative 70/30 split.","core_discovery":"Within the TNG100-1 cosmological simulation, the paper stacks 18,534 distinct pericentric encounters of massive galaxies ($M_* > 10^{10}\\,M_\\odot$) with companions at stellar mass ratios of 0.1 to 10 and tracks the specific star formation rate, gas fraction, and star formation efficiency inside the central stellar half-mass radius as functions of time relative to closest approach. For the closest encounters ($r_{\\mathrm{peri}} \\le 50$ kpc), mean central sSFR rises to $1.6 \\pm 0.1$ times its pre-encounter value, peaking about 0.1 Gyr after pericentre when the pair is on average about 75 kpc apart; gas fraction peaks at $1.2 \\pm 0.1$ and star formation efficiency at $1.4 \\pm 0.1$ on the same timescale, while central gas metallicity drops, signalling inflow of metal-poor gas. Because $\\mathrm{sSFR} = \\mathrm{SFE_H} \\times f_{\\mathrm{gas}}$, the fractional boost decomposes additively in log space, and the decomposition shows that about 70 per cent of the peak central sSFR enhancement is attributable to the rise in efficiency and 30 per cent to the rise in gas fraction. In the outer shell the same enhancements are weaker and peak about 0.05 Gyr earlier, and in per-galaxy terms 57.1 per cent of galaxies at the peak are individually efficiency-driven.","pith_inferences":["If higher efficiency is the dominant channel, resolved molecular-gas observations of close pairs should show a shortened depletion time even where the total gas fraction is not yet elevated; this is a direct observational discriminator the paper does not run.","The 70/30 average may hide subpopulations in which the balance inverts: the per-galaxy split at the peak is only 57/43, so binning by stellar mass ratio or orbital eccentricity could plausibly flip the dominant driver for some subsets.","Because the paper itself flags the systematic periodicity in pericentre times at the snapshot spacing, re-running the stack on the higher-time-resolution TNG50 run would test whether the roughly 0.1 Gyr delay is physical or pinned to the snapshot grid.","The inward-propagating signal (outer shell peaking about 0.05 Gyr before the centre) and the roughly 0.4 Gyr metallicity-dilution timescale together imply a measurable central gas inflow that could be checked against particle-level gas velocities in TNG100-1, a test left for future work."],"forward_implications":["Pair surveys that classify interactions by projected separation should expect the starburst to appear well after closest approach, at mean separations near 75 kpc rather than 30 kpc, which explains enhanced sSFR seen at wide separations.","Mean sSFR in the closest-encounter bin falls below the passively evolving sample by about 0.5 Gyr after pericentre, so the burst is followed by a phase of suppressed star formation consistent with gas exhaustion or feedback.","Enhancements in the outer shell are weaker and peak about 0.05 Gyr earlier than in the centre, supporting a picture in which the encounter's effect propagates inward.","Models that attribute encounter-triggered star formation purely to gas inflow miss the dominant channel: in TNG100-1 the central galaxies become more efficient at assembling and converting star-forming gas.","Restricting to first-passage encounters, 70 per cent of the sample, reproduces the full-dataset enhancements, so repeated passages are not responsible for the average result."],"supporting_citations":[{"why":"Supplies the reconstructed orbits and pericentre times that define the encounters dataset phased on closest approach.","marker":"Patton et al. (2024)"},{"why":"Provides the TNG100-1 host-companion pair catalogue and the separation-based sSFR results this study refines into pericentre-phase results.","marker":"Patton et al. (2020)"},{"why":"The merger-simulation benchmark that found fuel supply to dominate, against which the paper's efficiency-driven split is positioned.","marker":"Moreno et al. (2021)"},{"why":"The observational study that found gas enhancement to dominate over efficiency in close pairs; the paper's central counterpoint.","marker":"Pan et al. (2018)"},{"why":"The ALMa-QUEST study finding fuelling, efficiency, and mixed cases in roughly equal thirds; a direct comparison for the 70/30 split.","marker":"Thorp et al. (2022)"},{"why":"Merger simulation used to motivate post-pericentre SFR enhancement timing and the transport of low-metallicity gas to galaxy centres.","marker":"Torrey et al. (2012)"},{"why":"Provides the molecular-to-atomic hydrogen ratio estimates used to interpret what the total-hydrogen fgas and SFEH capture centrally.","marker":"Diemer et al. (2019)"},{"why":"The Schmidt-law relation that sets the subgrid star formation rate in TNG100-1 and therefore underlies what SFEH measures in the simulation.","marker":"Kennicutt (1998)"}],"fun_headline_variants":["Galaxy flybys boost star formation 1.6x: efficiency, then fuel","Close encounters lift star formation: 70% efficiency, 30% gas","Pericentric passages: 70% of star formation boost from efficiency","TNG100: galaxy flybys enhance star formation, efficiency-driven","In galaxy encounters, star formation boost is 70% efficiency, 30% fuel"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The stacked timeline rests on the reconstructed moments of closest approach being accurate to better than half the roughly 0.15 Gyr gap between simulation snapshots, and the paper itself notes a systematic periodic pattern at exactly that spacing in how those moments fall.","fun_headline_variants_meta":{"raw":{"variants":["Galaxy flybys boost star formation 1.6x: efficiency, then fuel","Close encounters lift star formation: 70% efficiency, 30% gas","Pericentric passages: 70% of star formation boost from efficiency","TNG100: galaxy flybys enhance star formation, efficiency-driven","In galaxy encounters, star formation boost is 70% efficiency, 30% fuel"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00066,"raw_usage":{"total_tokens":3189,"prompt_tokens":1289,"completion_tokens":1900,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":905,"completion_tokens_details":{"reasoning_tokens":1798}},"tokens_in":905,"tokens_out":1900,"duration_ms":13077,"temperature":1.0,"reasoning_tokens":1798,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T15:27:31.794365+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute pericentre times directly from the simulation's particle velocities instead of the interpolated orbits and re-stack the 18,534 encounters; if the $1.6 \\pm 0.1$ peak, the roughly 0.1 Gyr delay, and the 70/30 efficiency-fuel split shift by more than the quoted uncertainties, the signal is an artefact of the time axis. Running the same stack on the higher-time-resolution TNG50 run, which the paper lists as future work, would settle the question directly.","supporting_citations":[],"review_version":1}