{"id":"14a5d25a-f004-4484-970a-f24292da0042","arxiv_id":"2502.01721","paper_version":2,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Major mergers contribute only 3-13% of galaxy stellar mass growth at z~3-9, implying in-situ star formation dominates over hierarchical merging in the early universe.","lead":"Using JWST's JADES survey, the authors measured how often galaxies at redshift 3-9 sit in close pairs and merge, and calculated that major mergers add only about 3-13% of the stellar mass that galaxies build up in this era. The result matters because it suggests early galaxies grew mostly by making stars in place, not by colliding and merging, a key question in how the first galaxies formed.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 3-13% ex-situ fraction is set by a single merger timescale with factor 2-3 published scatter; the paper does not propagate this into its headline range, so the quantitative claim is not as robust as stated.","rationale":"The reader's weakest-assumption analysis identified the merger observability timescale τ_P as the most fragile external input, and I agree. The central numerical claim is the 3-13% ex-situ fraction, and Eq. 32 makes this quantity directly proportional to 1/τ_P. The paper itself notes in Section 6.5 that timescale estimates differ by factors of 2-3, yet the quoted uncertainties come only from bootstrapping pair fractions and do not include any systematic term for τ_P. Therefore the stated range is not an honest uncertainty interval, and the abstract's 'approximately 3-13%' is not robust at that precision. This is a load-bearing concern for the numerical claim specifically. However, it is not a reason to reject the paper: the qualitative conclusion that in-situ star formation dominates direct mass growth appears robust even if τ_P is shortened by a factor of 3, because the resulting ex-situ fraction would still be below 50%. The appropriate response is to require the authors to recompute their headline range under the published τ_P alternatives and report the resulting spread, and to soften or broaden the quantitative statement in the abstract accordingly. Since this is exactly the conditional-acceptance concern the reader raised, the verdict should remain conditional and unchanged.","tokens_in":52247,"tokens_out":8591,"duration_ms":93726,"concrete_test":"Recompute the cumulative ex-situ fractions in Sections 6.2-6.3 replacing Eq. 33 with (i) τ_P multiplied by 2.5, (ii) τ_P divided by 2.5, and (iii) the Conselice et al. (2022) mass-ratio-dependent timescale of Eq. 34, keeping all other choices fixed. If the resulting f_ex-situ values spread well outside 3-13% (for example, to roughly 1-30%), then the claimed interval is not robust and the abstract should be revised to a qualitative statement or a much wider range.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The quantitative headline (3-13% ex-situ mass fraction, abstract and Sections 6.2-6.3) is set by converting pair fractions to merger rates via Eq. 32 using the single observability timescale τ_P(z)=2.4(1+z)^{-2} Gyr (Eq. 33, Snyder et al. 2017). The authors choose the shortest of the published timescales, and Section 6.5 concedes that current theoretical estimates differ by factors of 2-3. Because RM = f_P/τ_P and the cumulative merger mass is proportional to the integral of RM, a factor-3 uncertainty in τ_P directly changes the ex-situ fraction by roughly a factor of 3: from about 13% to about 4% if τ_P is longer, or to about 39% if τ_P is shorter. The bootstrap uncertainties in Table 3 and Figure 13 do not include this calibration uncertainty, so the reported '3-13%' is not a robust confidence interval. Even a factor-3 shorter timescale would still leave in-situ star formation as the majority channel, so the qualitative conclusion may survive; but the specific numerical claim that major mergers contribute 'approximately 3-13%' is not supported at that precision without a systematic error budget for τ_P.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper uses JADES NIRCam imaging and the full photometric-redshift posterior distributions to measure close-pair fractions and major merger rates for galaxies with log(M*/Msun) in [8,10] over z~3-9. The authors find that the low-mass pair fraction rises to a peak near z~5-6 and then turns over, while higher-mass bins show flatter or declining evolution. Converting pair fractions to merger rates with the Snyder et al. (2017) observability timescale, they report merger rates of 2-8 Gyr^-1 per galaxy that flatten beyond z~6. Integrating the mass accretion from major mergers and comparing with star-formation rates, they conclude that major mergers contribute roughly 3-13% of stellar mass growth over z~3-9, with in-situ star formation dominating. A simple numerical model is presented as support for this conclusion.","tokens_in":52574,"tokens_out":4544,"duration_ms":49160,"significance":"If the qualitative result holds, this is an important constraint on early galaxy assembly: it directly quantifies, from JWST data, that hierarchical major merging is a secondary channel for stellar mass growth at z~3-9, complementing simulation-based studies. The paper's strengths include the probabilistic treatment of photo-z uncertainties, the use of a large spectroscopic redshift catalogue for validation, careful stellar-mass completeness corrections, and a transparent comparison with literature measurements. The central quantitative claim, however, rests on a single assumed merger observability timescale whose factor-2-3 uncertainty is acknowledged but not propagated, and the numerical-model consistency check is partially circular. The qualitative conclusion that in-situ star formation dominates appears robust, but the specific 3-13% ex-situ fraction is not supported at that precision.","major_comments":[{"comment":"The headline claim that major mergers contribute 'approximately 3-13%' to mass growth is not supported at that precision, because the conversion R_M = f_P/tau_P in Eq. (32) carries the factor-2-3 uncertainty in tau_P that the paper itself identifies in Section 6.5, yet no systematic error is propagated into Table 3, Figure 13, or the abstract. Since the cumulative merger mass in Eq. (42) is proportional to the time integral of R_M, varying tau_P within published values changes the ex-situ fraction by roughly the same factor (e.g., from about 13% to about 4% or about 39% at z~3), so the quoted 3-13% is conditional on one adopted timescale rather than a robust confidence interval. The authors should propagate this systematic uncertainty into the reported range, or explicitly state that the range is conditional on tau_P = 2.4(1+z)^-2 Gyr.","section":"Section 5.1 (Eq. 33) and Section 6.5"},{"comment":"The numerical model is presented as an independent estimate ('in agreement with our previous estimate'), but it is not independent: it injects the same fitted merger rate R_M(z) from Section 5.2 (Eq. 35) and the same SFMS sSFR (McClymont et al. 2025a) used in Section 6.2, and the quenching threshold at 10^10.8 M_sun is an ad hoc input. The agreement between the purple and blue curves is therefore partly by construction. This section should be reframed as a consistency check, and the sensitivity to the assumed SFMS, burstiness, quenching threshold, and merger-ratio distribution should be quantified as a spread rather than folded into a single 3-13% range.","section":"Section 6.3 (Figure 13)"},{"comment":"The claimed turnover and peak in the lowest stellar-mass bin at z~5-6 is driven by one subfield, GS-Medium, and the text acknowledges that the peak becomes less evident without it and that overdensities at z~5-6 may be responsible. Because this evolutionary trend is one of the paper's principal observational results and feeds into the interpretation of the z~6 flattening of the merger rate, the authors should present the pair fraction computed with GS-Medium excluded (and, if possible, an overdensity-corrected version) and quantify the significance of the peak, rather than relying on the pooled average alone.","section":"Section 4.1 and Figure 8"}],"minor_comments":[{"comment":"Equation (28) contains corrupted typesetting symbols that make the formula unreadable and should be regenerated.","section":"Section 3.6 (Eq. 28)"},{"comment":"The ex-situ fraction is quoted as 1-8% in Section 6.2, 3-13% in the abstract and Section 6.3, and 5-12% for the alternative SFMS; these ranges should be reconciled and their definitions stated explicitly.","section":"Section 6.2 vs 6.3"},{"comment":"The text states that sMAR = R_M * mu and later uses <M*,2/M*,1> = 0.474, but the connection between the two is not made explicit; clarify that using mu = 0.48 is an approximation and quantify its effect on the cumulative mass budget.","section":"Section 5.4 (Eqs. 37-41)"},{"comment":"The assumed stellar mass evolution M*(z) = M_peak * ((1+z)/(1+z_peak))^2, which assumes a constant mass-to-light ratio and ignores the M*-z covariance, is a strong approximation; it should be tested against a simple stellar-population model or explicitly listed among the systematic caveats.","section":"Section 3.3.3 (Eq. 12)"},{"comment":"The hatched regions for the five sSFR prescriptions are difficult to distinguish when overplotted with literature data; consider separate panels or a zoomed inset for the z>6 range.","section":"Figure 12"}],"recommendation":"major_revision","confidential_remarks":"This is a solid observational study with a defensible qualitative conclusion, but the quantitative 3-13% headline is not robust without a systematic error budget for the merger timescale, and the numerical-model agreement is partially circular. I recommend major revision rather than rejection, as the required additions (timescale sensitivity, GS-Medium-excluded measurements, and a reframed consistency check) are within the scope of the manuscript. The paper is otherwise thorough and well suited to the journal."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nTake-home: this is a careful measurement of close-pair fractions and major merger rates from JADES at z~3-9, and the central claim that major mergers contribute only a minor share of stellar mass growth is probably correct in direction. But the specific '3-13%' ex-situ fraction in the abstract is not a robust confidence interval. It is set by a single assumed merger observability timescale, tau_P(z)=2.4(1+z)^-2 Gyr from Snyder et al. (2017), and the authors themselves note in Section 6.5 that published timescales differ by factors of 2-3. Since R_M = f_P/tau_P, a factor-3 change in tau_P shifts the ex-situ fraction by roughly a factor of 3. The bootstrap errors in Table 3 do not include this calibration uncertainty. Even with a factor-3 shorter timescale, the ex-situ fraction would be ~39%, so in-situ star formation would still dominate, but the quoted range should be presented as an order-of-magnitude estimate, not a precise interval.\n\nWhat's genuinely new here: mass-resolved pair fractions and merger rates out to z~9 from JADES, the flattening of the merger rate beyond z~6, and the contrast with Duan et al. (2024a), who reported ~71% merger-driven growth. The authors plausibly trace that discrepancy to the sSFR prescription (Bagpipes vs. SFMS) rather than to pair counts, which is a useful clarification for the field.\n\nThe methodology is solid: full photo-z posteriors, spec-z substitution, completeness and OSR weights, bootstrapped uncertainties, and a number of cross-checks (eazy-py masses, virial-radius-based separation). The authors are also honest about field-to-field variation: they show that the low-mass pair fraction peak at z~5-6 is driven by GS-Medium and may be boosted by known overdensities. That is the right way to handle a suspicious feature.\n\nSoft spots worth flagging:\n\n1. The '3-13%' range comes from three estimators that are not fully independent. The numerical model in Section 6.3 takes the same fitted merger rate and the same SFMS as inputs as the integral in Section 6.2, so the agreement between them is partly by construction. The central claim is still grounded in the measured merger rate, but the consistency check is weaker than it looks.\n\n2. The low-mass peak is fragile. Without GS-Medium, it is much less evident, and this is the bin that drives the redshift turnover. That should be either down-weighted or made conditional on field selection.\n\n3. There is no correction for projection effects. The authors cite Huško et al. (2022) showing that roughly half of projected pairs at high mass are chance alignments, but they do not attempt to correct for this. It is at least worth a sensitivity test.\n\n4. The selection code is not released; the data availability statement says 'shared on reasonable request'. Given the complexity of the method, that is a barrier to reproducibility.\n\nWho is this for? Observers working on high-z mergers, and simulators who need empirical constraints on the merger rate at z>3. It is a solid paper that deserves a serious referee, but the revision should include a systematic error budget for the timescale, a clearer caveat on the low-mass peak, and a code release. I would take it to reading group and would cite it once the timescale systematics are properly quantified.\n\nMy recommendation: send to peer review. The core measurement is valuable and the qualitative conclusion is likely right; the numerical headline needs to be softened and the systematics expanded.","headline":"Solid JADES close-pair measurement; the qualitative conclusion holds, but the 3-13% ex-situ fraction is not as precise as advertised because the merger timescale is uncertain by factors of 2-3.","tokens_in":53209,"tokens_out":3787,"would_cite":true,"duration_ms":36861,"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":"At redshifts 3 to 9, major mergers deliver only about 3-13% of the stellar mass that galaxies build, so in-situ star formation dominates early mass assembly.","keywords":["galaxy mergers","close pairs","major merger rate","in-situ star formation","ex-situ stellar mass","JWST","JADES","high-redshift galaxies"],"falsifier":"Measure the real merger rate at $z=6$–$9$ in the same stellar mass range using a fully spectroscopic sample (or resolved kinematics from JWST/NIRSpec IFU) and compare the number of actual mergers per galaxy per Gyr with the photometric-pair-derived $2$–$8\\,\\mathrm{Gyr}^{-1}$. Alternatively, in a hydrodynamic simulation, compute how long a pair that is truly merging satisfies the $5$–$30$ kpc projected-separation and photometric-redshift proximity criteria, and test whether that timescale differs from $2.4\\,(1+z)^{-2}$ Gyr by more than a factor of two. If either test returns a substantially different timescale or merger rate, the $3$–$13\\%$ ex-situ fraction is not robust.","tokens_in":52019,"feed_emoji":"🌌","tokens_out":16036,"duration_ms":126820,"temperature":0.7,"pith_summary":"Using deep JWST/JADES imaging of the GOODS-South and GOODS-North fields, this paper counts close pairs of galaxies with stellar masses $\\log(M_\\star/\\mathrm{M}_\\odot)=[8,10]$ and mass ratios $\\geq 1/4$ over the redshift range $3 \\leq z \\leq 9$, and converts those counts into merger rates. The central result is that the major merger rate per galaxy flattens beyond $z \\sim 6$ at roughly $2$–$8\\,\\mathrm{Gyr}^{-1}$, and that the cumulative stellar mass accreted through major mergers is only about $3$–$13\\%$ of the mass built by star formation. The paper concludes that in-situ star formation, not hierarchical merging, is the dominant direct channel of stellar mass growth in the early universe. This matters because models of early galaxy formation often assume mergers are the primary route to mass assembly; if correct, these results redirect attention to gas accretion and star formation as the main engine, with mergers playing a secondary but possibly still important indirect role.","feed_headline":"Major mergers drive only 3-13% of early galaxy growth","feed_subtitle":"Pair counts in JWST JADES show star formation, not collisions, builds most stellar mass by z=3-9.","key_machinery":"The load-bearing tool is the pair probability function $\\mathrm{PPF}(z) = \\mathcal{Z}(z)\\,\\mathcal{M}_\\theta(z)\\,\\mathcal{M}_{\\mathrm{pair}}(z)$, where $\\mathcal{Z}(z)$ is a redshift probability function formed from the overlap of the two galaxies' full photometric-redshift posteriors (with spectroscopic redshifts inserted when available), $\\mathcal{M}_\\theta(z)$ is a binary angular-separation mask translating $5$–$30$ kpc projected separation, and $\\mathcal{M}_{\\mathrm{pair}}(z)$ enforces the $\\geq 1/4$ stellar mass ratio and completeness limits. Integrating this function over each redshift bin, with weights for mass incompleteness, photometric-redshift quality, and survey-boundary area, yields the close-pair fraction. The observability timescale $\\tau_P(z) = 2.4\\,(1+z)^{-2}\\,\\mathrm{Gyr}$ converts that fraction into a merger rate, and comparison of the integrated merger mass accretion with the star-forming main sequence, cross-checked with the 50,000-galaxy model, yields the ex-situ mass fraction.","core_discovery":"The paper's central claim is that in the first two billion years of cosmic history, major mergers are not the main way galaxies build their stellar mass. Using the probability-based close-pair method on JADES data, the authors find that the close-pair fraction peaks near $z\\sim5$–$6$ for galaxies of $10^{8}$–$10^{8.5}\\,M_\\odot$ and is flat or mildly declining for more massive galaxies; the inferred major merger rate rises to $z\\sim6$ and then plateaus at $2$–$8\\,\\mathrm{Gyr}^{-1}$ per galaxy. Integrating the merger mass accretion rate and comparing it with the star-forming main sequence gives an ex-situ stellar mass fraction—the share of stars acquired from mergers rather than formed in the galaxy—of roughly $3$–$13\\%$ over $z \\approx 3$–$9$. A simple Monte Carlo model of 50,000 galaxies driven by the measured merger rate and a star-forming main sequence specific star formation rate reproduces the same range. The conclusion is that direct mass assembly at these redshifts is dominated by in-situ star formation; major mergers are a significant but secondary channel, though they could still trigger star formation indirectly.","pith_inferences":["The adopted observability timescale is the shortest among published estimates, so the derived $3$–$13\\%$ ex-situ fraction is likely an upper bound; a longer timescale at $z>3$ would make major mergers even less important for direct mass growth.","The analysis excludes minor mergers with mass ratios below $1/4$, which are expected to be more numerous; adding them would raise the total ex-situ fraction but is unlikely to reverse the conclusion that in-situ star formation dominates for typical $10^8$–$10^{10}\\,M_\\odot$ galaxies.","Because the data are photometric, line-of-sight projection can still masquerade as a pair; the turnover seen at $z\\sim6$ could partly reflect this contamination, and a spectroscopically complete sample would be the cleanest test.","The paper leaves open an indirect role for mergers: if they trigger the starbursts that make in-situ star formation so dominant, then merger activity still shapes galaxies even while contributing little direct mass."],"forward_implications":["If the central claim holds, major mergers add at most a $\\sim$10% contribution to the stellar mass of $10^8$–$10^{10}\\,M_\\odot$ galaxies between $z=3$ and $z=9$, so gas accretion and star formation must be the primary growth channel in the early universe.","The flattening of the merger rate at $z \\gtrsim 6$ to $2$–$8\\,\\mathrm{Gyr}^{-1}$ means the often-assumed monotonically rising merger rate with redshift does not continue into the first billion years; this is a direct constraint for galaxy formation models.","The pair-fraction turnover at high redshift is tied to the paucity of massive galaxies in the exponential tail of the stellar mass function, not necessarily to a change in merger physics, so future surveys at fixed mass need to account for this selection effect.","The methodology of propagating full photometric-redshift posteriors into the pair count can be applied to other deep JWST fields to extend the merger history to $z>9$ and to fainter galaxies."],"supporting_citations":[{"why":"It supplies the merger observability timescale that converts pair fractions into merger rates and sets the scale of the ex-situ fraction.","marker":"Snyder et al. (2017)"},{"why":"It establishes the probabilistic close-pair analysis with full photometric-redshift posteriors that the paper adapts to mass-selected samples.","marker":"López-Sanjuan et al. (2015)"},{"why":"It introduces the mass-selected version of the probabilistic close-pair method and the 5-30 kpc separation convention adopted here.","marker":"Mundy et al. (2017)"},{"why":"It extends the probabilistic close-pair method to high redshift and defines the specific mass accretion rate comparison used in Section 6.2.","marker":"Duncan et al. (2019)"},{"why":"It provides the closest recent JWST close-pair and merger-rate measurements at z=4.5-11.5 that the paper's results are compared against and explains the discrepancy in ex-situ fraction estimates.","marker":"Duan et al. (2024a)"},{"why":"It provides the star-forming main sequence specific star formation rate used as the in-situ star formation reference for the cumulative mass comparison and in the numerical model.","marker":"McClymont et al. (2025a)"},{"why":"It supplies the JWST stellar mass function used to compute completeness weights and merger-rate densities at intermediate redshifts.","marker":"Navarro-Carrera et al. (2024)"},{"why":"It extends the stellar mass function to higher redshift, supporting the high-redshift completeness corrections and the cosmic stellar mass density calculation.","marker":"Harvey et al. (2025)"}],"fun_headline_variants":["Mergers add only 3-13% to early galaxy mass","Star formation dominates early galaxy assembly","JADES: mergers minor in z=3-9 galaxy growth","Early galaxies build stars, not by merging","In-situ star formation wins over major mergers"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing assumption is that a galaxy pair remains visible as a close pair for a time $\\tau_P(z) = 2.4\\,(1+z)^{-2}$ Gyr; the paper adopts this single simulation-based timescale to convert every measured pair fraction into a merger rate and into the $3$–$13\\%$ ex-situ fraction, and published alternatives differ by factors of $2$–$3$.","fun_headline_variants_meta":{"raw":{"variants":["Mergers add only 3-13% to early galaxy mass","Star formation dominates early galaxy assembly","JADES: mergers minor in z=3-9 galaxy growth","Early galaxies build stars, not by merging","In-situ star formation wins over major mergers"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000333,"raw_usage":{"total_tokens":1952,"prompt_tokens":1150,"completion_tokens":802,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":766,"completion_tokens_details":{"reasoning_tokens":726}},"tokens_in":766,"tokens_out":802,"duration_ms":7936,"temperature":1.0,"reasoning_tokens":726,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-09T14:40:47.536914+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the real merger rate at $z=6$–$9$ in the same stellar mass range using a fully spectroscopic sample (or resolved kinematics from JWST/NIRSpec IFU) and compare the number of actual mergers per galaxy per Gyr with the photometric-pair-derived $2$–$8\\,\\mathrm{Gyr}^{-1}$. Alternatively, in a hydrodynamic simulation, compute how long a pair that is truly merging satisfies the $5$–$30$ kpc projected-separation and photometric-redshift proximity criteria, and test whether that timescale differs from $2.4\\,(1+z)^{-2}$ Gyr by more than a factor of two. If either test returns a substantially different timescale or merger rate, the $3$–$13\\%$ ex-situ fraction is not robust.","supporting_citations":[],"review_version":1}