{"id":"67579581-5a31-4a89-8305-267885ac77b3","arxiv_id":"1909.02706","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Most Class 0 protostars in Perseus show chemical evidence of accretion bursts within the past 10,000 years, and the inferred burst frequency declines from the Class 0 to the Class I stage.","lead":"An ALMA survey of 39 young protostars maps the CO and water snowlines through N2H+ and HCO+ emission, finding that most embedded protostars show chemical traces of past accretion bursts. The data suggest bursts become less frequent as protostars evolve from Class 0 to Class I, a timeline that would constrain when disks fragment and feed their stars.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 2400-to-8000 yr burst-frequency decline rests on comparing HCO+ snowline radii with a no-disk model, and the paper's own caveat in §5.2.1 admits that a denser or larger Class I disk would erase the trend. The claim is conditional until disk-inclusive models are tested.","rationale":"The reader's weakest_assumption correctly identifies the model's no-disk treatment as load-bearing. The paper itself flags the disk-evolution alternative, and the numerical claim is derived from the tracer (HCO+) for which disk shielding is strongest. I see no internal inconsistency: the observations are honestly presented, the calibration on two burst-phase sources is explicitly labeled as rough, and the binding-energy and optical-depth caveats are discussed. The concern is not that the authors are wrong, but that the headline claim is conditional on an untested physical assumption. The proposed disk-inclusive rerun is the decisive check; it is feasible with existing VANDAM/MASSES continuum data and the MHW19 grid once it is made available. Until then, CONDITIONAL is the appropriate verdict, and the stress-test does not move it.","tokens_in":34909,"tokens_out":12606,"duration_ms":134597,"concrete_test":"Make the MHW19 model grid (including the disk variant) public, and for the 17 Class 0 and 8 Class I sources with measured HCO+ peak radii, rerun the snowline-radius comparison with disk parameters from VANDAM 8 mm and MASSES 1.2 mm continuum (Tobin et al. 2016; Stephens et al. 2018; Table 1). If, with realistic Class I disk masses and radii, the Class I post-burst fraction (Lburst > Lbol and Lburst > 10 L⊙) becomes comparable to the Class 0 fraction, then disk evolution explains the apparent decrease and the 2400-to-8000 yr claim is not supported. If the Class I fraction stays lower when disks are included, the burst-frequency decline survives.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central chronological claim (abstract; §5.2.1) is that the inter-burst interval grows from ~2400 yr in Class 0 to ~8000 yr in Class I, based on the HCO+ (water-snowline) post-burst fractions of 7/17 and 1/8. This inference requires that the same no-disk MHW19 model (§4.2) converts the observed HCO+ peak radii to burst luminosities for both classes. If a Class I disk is denser or larger than a Class 0 disk, it shields the envelope and shrinks the water-snowline radius for a fixed luminosity; the adopted no-disk model then underestimates Lburst (§4.2 and §4.4.2), so the small Class I peaks can be read as disk growth rather than the absence of a recent burst. The authors state this directly in §5.2.1: 'if a disk is significantly denser/larger at the Class I stage than that at the Class 0 stage, it might shrink the emission peak inward... our conclusion that the burst frequency decreases from the Class 0 to the Class I stage might in turn not be robust.' The N2H+ (CO-snowline) data alone do not establish the decline: with Lburst > 10 L⊙ the fractions are 56±24% (Class 0) and 50±25% (Class I), giving ~18,000 yr and ~20,000 yr intervals (§5.2.1). Thus the strong 2400-to-8000 yr statement rests on the HCO+ subset, which is exactly the tracer most affected by disk shielding. The MHW19 model is also unpublished, so its disk variants cannot yet be independently checked.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents ALMA observations of N2H+ (1−0) and HCO+ (3−2) toward 39 Class 0 and Class I protostars in Perseus. The observed emission-peak radii are compared with the no-disk MHW19 radiative/chemical model to map radii to luminosities; sources with inferred past-burst luminosity exceeding the current bolometric luminosity are classified as post-burst. The authors report that 18/18 Class 0 and 9/10 Class I sources show CO-snowline (N2H+) post-burst signatures, while 7/17 Class 0 and 1/8 Class I show H2O-snowline (HCO+) signatures when requiring Lburst > 10 L⊙. They conclude that the interval between bursts increases from ~2,400 yr in Class 0 to ~8,000 yr in Class I, with burst luminosities of ~10–100 L⊙ and burst-phase mass accretion rates of ~(7.6–16.2)×10−6 M⊙ yr−1, and use these to estimate the accumulated mass during the embedded phase. The paper explicitly acknowledges that the frequency decline could instead reflect disk growth and shielding if Class I disks are denser/larger.","tokens_in":35336,"tokens_out":7206,"duration_ms":68656,"significance":"If the chronological trend is robust, this survey would be the first statistical evidence that episodic accretion becomes less frequent as protostars evolve from Class 0 to Class I, directly informing models of disk gravitational instability and the mass assembly history of Sun-like stars. Strengths of the paper include the relatively large ALMA sample within a single cloud, the use of two chemical tracers with complementary refreeze-out timescales, explicit tabulation of per-source uncertainties, and unusually candid discussion of model caveats. The analysis is not machine-checked, but the data tables and figures are sufficiently detailed to reproduce the statistical statements. However, the central claim is currently conditional on the no-disk model and on the HCO+ subsample, and the paper's own caveats indicate that the evolutionary interpretation is one of two viable readings.","major_comments":[{"comment":"The headline interval increase (2,400 yr to 8,000 yr) is derived only from the HCO+ subsample with Lburst > 10 L⊙ (7/17 Class 0 vs 1/8 Class I), while the N2H+ subsample at the same threshold yields 56±24% vs 50±25%, i.e., no significant difference. The paper presents the HCO+-based decline as the central chronological conclusion without a statistical significance test against the null hypothesis of equal fractions, and the authors' own numbers imply the CO-snowline tracer does not corroborate the claim. Please perform an explicit two-sample test (e.g., Fisher's exact test) on the 2×2 tables and report the resulting p-values for both tracers, and discuss the implications if the N2H+ data are consistent with no evolution.","section":"§5.2.1, Fig. 8"},{"comment":"The paper's own caveat in §5.2.1 — that a denser/larger Class I disk would shrink the HCO+ peak and erase the burst-frequency decline — directly undermines the central claim because the adopted model is the no-disk MHW19 model (§4.2). Figure 6 shows a disk model shifts the Lburst–radius relation to higher luminosity, but the disk model is not quantified or applied to the sample. Without a quantitative demonstration that disk growth between Class 0 and Class I is negligible (e.g., by varying disk parameters in the model), the observed Class 0/I difference in HCO+ peak radii remains fully consistent with disk evolution rather than a change in burst frequency.","section":"§5.2.1 and §4.2"},{"comment":"The model's CO binding energy is set to 1307 K specifically to match the two currently bursting sources, Per-emb-27 and Per-emb-44, and the same model is then used to derive Lburst for every other source. Because the text also notes the binding energy is degenerate with the density structure, this calibration does not provide an independent test of the model; any resulting error in the temperature–radius mapping propagates directly into the post-burst classifications and the derived burst intervals. The fact that MHW19 is unpublished further prevents the reader from assessing the sensitivity of the conclusions to this calibration and to the disk variant.","section":"§4.4.2"},{"comment":"Sources whose measured peak radius falls below the half-beam size are classified as non-post-burst and assigned an upper-limit Lburst, and for Class I this includes most of the HCO+ sample (seven of eight non-post-burst sources). Since the inferred post-burst fraction depends entirely on the treatment of these unresolved sources, the derived 8,000-yr interval is not robust to the exact threshold used or to the placement of upper limits; a survival-analysis treatment that includes upper limits as censored data would provide a more defensible estimate of the Class I burst interval.","section":"§4.3 and §5.2.1"}],"minor_comments":[{"comment":"The abstract reports the HCO+ fractions 7/17 and 1/8 without stating that these correspond to the Lburst > 10 L⊙ criterion; §4.3 initially reports 10/17 and 2/10 using Lburst > Lbol. Please state the criterion explicitly in the abstract or in Table 3 to avoid ambiguity.","section":"Abstract and §4.3/§5.2.1"},{"comment":"The caption reads \"The dark and light blue area represent a range\"; it should be \"areas represent\".","section":"Figure 6 caption"},{"comment":"The sentence \"MHW19 have built a grid\" should read \"MHW19 has built\" (or \"Murillo et al., in prep., have built\").","section":"§4.2"},{"comment":"The phrase \"the intervals between bursts increases\" should be \"increase\" for subject-verb agreement.","section":"§5.2.1"},{"comment":"The refreeze-out timescale formula in §5.2.1 is not numbered; please number it so that it can be referenced in the text and in the discussion.","section":"Equation (1)"}],"recommendation":"major_revision","confidential_remarks":"The abstract presents the burst-interval increase as the main result, but the body shows it is one of two interpretations and rests on a small subsample with an unresolved model dependence. I recommend asking for the significance test and a quantitative disk-shielding assessment before publication; the paper is likely publishable after major revision. The unpublished MHW19 model is a key dependency, so the sensitivity to the disk variant should be demonstrated in this paper or in an accessible companion paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, this is a substantial ALMA survey that gives the cleanest statistical picture yet of snowline radii in embedded protostars: 39 sources in Perseus, two independent chemical clocks (N2H+ for CO, HCO+ for H2O). The new result is the claimed timeline: burst intervals increasing from ~2400 yr in Class 0 to ~8000 yr in Class I. If that holds, it is the first direct handle on how episodic accretion evolves across the embedded phase. The paper deserves a careful read.\n\nThe strongest part is the observational work. The maps are systematic, the masking of outflows is thoughtful, and they check HCO+ against CH3OH where available. The identification of post-burst sources from peak radii is a sensible extension of earlier single-source studies. I also credit the authors for spelling out the disk-evolution alternative in §5.2.1 instead of burying it.\n\nNow the soft spots, in proportion. The 2400-to-8000 yr statement comes from the HCO+ data only, where the post-burst fractions are 7/17 and 1/8. The N2H+ data alone, using the same Lburst>10 Lsun cutoff, give 56% vs 50%, i.e., no decline. So the evolutionary trend is carried by a small, overlapping sample. The model that converts peak radii to burst luminosities is unpublished (MHW19), and the CO binding energy was chosen to match the two burst-phase sources. That is circular in the sense that the absolute Lburst scale is partly fitted, even if the Class 0/I comparison is not. And the no-disk assumption bites exactly where the trend is: a larger or denser Class I disk would shrink the observed HCO+ peak for fixed luminosity, and the authors admit this would erase the conclusion. These are not fatal flaws, but they make the central claim provisional rather than established.\n\nThere are also minor statistical nits: the binomial uncertainties are large (58±24% vs 20±16%), and the refreeze-out timescales are assumed, not measured.\n\nWho is this for? The star formation community, especially people working on episodic accretion, disk instability, and protostellar luminosity distributions. It is a solid observational contribution with an honest but fragile central interpretation. A serious referee should engage, not desk-reject. My recommendation: accept with major revision, and require the authors to either incorporate a disk model or explicitly reframe the claim as conditional on disk evolution. In the meantime, I would cite it with the caveat.","headline":"A large, careful snowline survey that makes the first statistical claim of declining burst frequency from Class 0 to I, but the trend rests on a small HCO+ subsample and a no-disk model, with the authors' own caveat that disk growth would erase it.","tokens_in":35906,"tokens_out":2893,"would_cite":true,"duration_ms":29080,"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":"This paper argues that protostars accrete in bursts roughly every 2,400 years early on, slowing to every 8,000 years as they evolve.","keywords":["episodic accretion","protostars","snowlines","ALMA survey","Class 0 protostars","Class I protostars","N2H+","HCO+"],"falsifier":"Measure disk radii and masses in the same 39 Perseus targets with long-baseline ALMA continuum and recompute the predicted snowline radii using disk-inclusive models; if Class I disks turn out to be systematically denser or larger and the disk models bring the Class I post-burst fractions up to the Class 0 level without any change in burst frequency, the chronological claim fails.","tokens_in":34741,"feed_emoji":"⭐","tokens_out":8399,"duration_ms":80792,"temperature":0.7,"pith_summary":"This paper tries to establish a chronology for episodic accretion, the idea that young stars gain mass in violent bursts rather than steadily, by using the positions of carbon monoxide and water snowlines as chemical clocks in 39 protostars in the Perseus cloud. Because N2H+ is destroyed by gaseous CO and HCO+ by gaseous H2O, the radius at which each molecule's emission peaks marks where the relevant ice has recently sublimated; a snowline sitting much farther out than the current luminosity can produce is a fossil of a past accretion burst. The survey finds that nearly every Class 0 source shows such a fossil at the CO snowline (18 of 18) and most also at the shorter-lived water snowline (7 of 17), whereas the fractions fall to 9 of 10 and 1 of 8 in Class I sources. Interpreting the two snowline lifetimes as about 10,000 years for CO and about 1,000 years for H2O, the paper concludes that the average interval between bursts grows from about 2,400 years in the Class 0 phase to about 8,000 years in the Class I phase. If true, this means disk instability and fragmentation operate most vigorously in the earliest embedded phase, and a typical protostar experiences on the order of a hundred bursts that deliver much of its mass.","feed_headline":"Accretion bursts slow from every 2,400 to 8,000 years","feed_subtitle":"Fossil snowlines around 39 Perseus protostars show the earliest embedded phase is the most burst-active.","key_machinery":"The central machinery is the chemical snowline fossil: N2H+ (1-0) emission is destroyed by gas-phase CO and therefore peaks just outside the CO sublimation front, while HCO+ (3-2) is destroyed by gas-phase H2O and peaks just outside the H2O snowline. Comparing the observed peak radii with the peak radii predicted by the companion radiation-transfer and chemical model grid (MHW19) for a given central luminosity turns a spatial offset into an estimate of the luminosity of the last accretion burst. The two molecules act as complementary stopwatches because CO refreezes in about 10,000 years and H2O in about 1,000 years at envelope densities, so a source showing an HCO+ fossil indicates a burst within the last millennium, an N2H+ fossil only indicates one within the last 10,000 years, and neither brackets the time since the last burst. The analysis adopts a rotationally flattened envelope without a disk, which makes the estimated burst luminosities lower limits and is the factor most likely to bias the Class 0 versus Class I comparison. Two sources independently thought to be bursting now, Per-emb-27 and Per-emb-44, fall near the model curve and serve as rough calibrators.","core_discovery":"The central claim is that the burst frequency of episodic accretion declines as protostars evolve: the intervals between accretion bursts increase from about 2,400 years in the Class 0 stage to about 8,000 years in the Class I stage, under the assumption of refreeze-out timescales of 1,000 years for H2O and 10,000 years for CO. The evidence comes from a statistical comparison of measured N2H+ and HCO+ peak radii with radiative-transfer and chemical models of a rotationally flattened envelope at different luminosities: sources whose snowlines lie farther out than the current bolometric luminosity can explain are classified as post-burst. The survey identifies burst-fossil sources at rates of 100 percent (18 of 18) in Class 0 and 90 percent (9 of 10) in Class I from N2H+, and 58 percent (7 of 17) in Class 0 versus 20 percent (1 of 8) in Class I from HCO+. The authors explicitly note the rival interpretation that disk growth from Class 0 to Class I shrinks the inferred snowline radius, which would mimic a decline in burst frequency; they therefore frame the chronological result as conditional on comparable envelope structure across stages. They also derive burst luminosities of about 10 to 100 solar luminosities and estimate that bursts deliver a total of roughly 0.1 to 0.3 solar masses, leaving most of the final stellar mass to be accreted quiescently or in rare super-bursts.","pith_inferences":["If disk evolution rather than a lower burst frequency explains the Class I deficit, then high-resolution images of Class I disks should show systematically larger or denser disks than Class 0 disks; current ALMA surveys could test this directly by measuring disk radii and masses in the same sample.","The chronology implies a testable luminosity distribution: Class 0 samples should contain a higher fraction of currently bursting, high-luminosity objects than Class I samples, something wide-field time-domain surveys could check statistically.","Because the two tracers bracket the time since the last burst, tallying sources with an N2H+ fossil but no HCO+ fossil gives a direct histogram of burst intervals rather than just an average; extending this binning to a larger sample would sharpen the 2,400 versus 8,000 year numbers.","The result suggests that inner-disk material is repeatedly heated and cooled during the Class 0 phase, so chemical models of complex organic molecule formation should incorporate burst frequencies that decline with time rather than a constant rate."],"forward_implications":["If the burst interval truly grows from about 2,400 years in Class 0 to about 8,000 years in Class I, a protostar experiences roughly 63 to 100 bursts in the Class 0 phase and 39 to 60 bursts in the Class I phase, implying disk fragmentation is far more vigorous in the earliest embedded stage.","The burst-phase accretion rate is about (7.6 to 16.2) x 10^-6 solar masses per year and appears roughly constant across stages, so the evolution is in burst frequency rather than burst strength.","Bursts deliver only about 0.1 to 0.3 solar masses total, which is less than the typical 0.3 to 0.5 solar mass star; the remainder must be accreted during quiescence or in rare FU-Orionis-type super-bursts.","The inferred burst duration of about 120 years (a 5 percent duty cycle at Class 0) matches the 100 to 200 year durations predicted by gravitational-instability simulations.","A declining burst frequency supports the picture in which gravitational instability is sustained by the high infall rates and cold disk conditions of the earliest embedded phase."],"supporting_citations":[{"why":"Supplies the refreeze-out timescale formula (10,000 years for CO, 1,000 years for H2O) used to convert snowline fossils into burst intervals.","marker":"Visser et al. 2015"},{"why":"Established the HCO+ ring around methanol emission as a water-snowline tracer of past outbursts.","marker":"Jørgensen et al. 2013"},{"why":"Earlier N2H+ CO-snowline survey of very low luminosity objects; provides three targets and the peak-radius comparison method.","marker":"Hsieh et al. 2018"},{"why":"Previous CO-snowline survey of embedded protostars that estimated burst intervals of 2 to 5 x 10^4 years, the baseline this study refines.","marker":"Frimann et al. 2017"},{"why":"Validates HCO+ as a water-snowline tracer through its anti-correlation with water emission in NGC1333-IRAS2A.","marker":"van ’t Hoff et al. 2018b"},{"why":"Provides the sample's bolometric luminosities and evolutionary classifications scaled to the Perseus distance.","marker":"Dunham et al. 2015"},{"why":"Gives the roughly 5 percent high-luminosity fraction used to estimate burst duty cycle and duration.","marker":"Enoch et al. 2009"},{"why":"Provides the gravitational-instability burst model used to interpret the declining frequency and to expect 100 to 200 year burst durations.","marker":"Vorobyov & Basu 2005"}],"fun_headline_variants":["Accretion bursts slow as protostars mature","ALMA reveals burst frequency drops from Class 0 to I","Fossil snowlines trace episodic accretion history","Protostars burst less often as they evolve","Snowline survey: bursts every 2,400 yrs in Class 0, 8,000 in I"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that Class 0 and Class I envelopes have similar density structure, so the smaller snowline radii in Class I sources reflect a longer time since the last burst rather than a larger or denser disk shielding the envelope from the central star's radiation.","fun_headline_variants_meta":{"raw":{"variants":["Accretion bursts slow as protostars mature","ALMA reveals burst frequency drops from Class 0 to I","Fossil snowlines trace episodic accretion history","Protostars burst less often as they evolve","Snowline survey: bursts every 2,400 yrs in Class 0, 8,000 in I"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001086,"raw_usage":{"total_tokens":4672,"prompt_tokens":1214,"completion_tokens":3458,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":830,"completion_tokens_details":{"reasoning_tokens":3380}},"tokens_in":830,"tokens_out":3458,"duration_ms":26292,"temperature":1.0,"reasoning_tokens":3380,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T04:42:48.908577+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure disk radii and masses in the same 39 Perseus targets with long-baseline ALMA continuum and recompute the predicted snowline radii using disk-inclusive models; if Class I disks turn out to be systematically denser or larger and the disk models bring the Class I post-burst fractions up to the Class 0 level without any change in burst frequency, the chronological claim fails.","supporting_citations":[{"cited_title":"A., & J rgensen, J","cited_arxiv_id":null,"evidence_quote":"Supplies the refreeze-out timescale formula (10,000 years for CO, 1,000 years for H2O) used to convert snowline fossils into burst intervals."},{"cited_title":"I., & Basu, S","cited_arxiv_id":null,"evidence_quote":"Provides the gravitational-instability burst model used to interpret the declining frequency and to expect 100 to 200 year burst durations."}],"review_version":1}