{"id":"1cad4994-c2e5-4f76-b099-4a880f58100b","arxiv_id":"2412.12346","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":10,"one_line_summary":"A Geant4 and graph-neural-network feasibility study shows the EIC Zero Degree Calorimeter can reconstruct Lambda0 -> n pi0 decays with resolutions that meet the EIC Yellow Report requirements.","lead":"This simulation study shows that the neutral decay of a forward Lambda baryon (Lambda0 to neutron plus pi0) can be spotted and measured with the planned high-granularity Zero Degree Calorimeter at the future Electron-Ion Collider. If real detectors behave like the simulation, this would unlock kaon structure and spin measurements that the charged decay channel cannot reach at high energies.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Realistic noise and forward-background are absent from the simulation; the claimed 'comfortable margin' rests entirely on clean single-Lambda events.","rationale":"I read the paper as a carefully executed clean-simulation feasibility study, and the physics chain—generating Lambda0->n pi0, simulating the SiPM-on-tile ZDC, reconstructing with IDOLA or a GNN, and comparing to Yellow Report requirements—is coherent and internally consistent. The IDOLA method is a plausible iterative vertex finder, the test/training split for the GNN is reasonable, and the comparison to single-neutron performance correctly identifies the neutron as the resolution driver. The reader's weakest-assumption identification matches the most load-bearing concern I can find: the simulation includes no electronic noise, no beam-gas background, and no underlying event, while the feasibility margin is quantified from those clean numbers. I do not think this concern invalidates the study; for a first feasibility demonstration, a clean simulation is acceptable. But the abstract's claim of meeting Yellow Report requirements 'with a comfortable margin' goes beyond what the current evidence supports, because the margin could shrink or disappear once realistic ZDC occupancies are included. The concrete test I propose directly addresses this by overlaying noise and forward backgrounds and re-evaluating the same pipelines without retraining. The verdict should remain CONDITIONAL, so no change from the reader's verdict is needed.","tokens_in":17249,"tokens_out":4809,"duration_ms":57454,"concrete_test":"Generate an overlay sample: take the same 500k-event discrete-energy Lambda0->n pi0 and single-neutron samples, then add (a) SiPM dark-count and electronic noise at the expected EIC ASIC rate and (b) forward hadrons from beam-gas or underlying-event generators in the ZDC acceptance, using an ePIC forward-background model. Re-run HEXPLIT, topoclustering, IDOLA, and the published GNN models without retraining, and recompute the Lambda efficiency, neutron misidentification rate, and energy/angle resolutions of Figs. 11-13. If the 250-GeV GNN Lambda efficiency falls to the level of the conventional method, or if the energy resolution crosses the 50%/sqrt(E)+5% Yellow Report curve, the claimed comfortable margin is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim—that Lambda0->n pi0 reconstruction at the EIC ZDC meets Yellow Report requirements with a comfortable margin—is supported only by Geant4 events containing the Lambda decay products and nothing else. Section 2.2 states 'No noise was included in the simulation,' and Section 4.1 explicitly lists beam-gas interactions, SiPM noise, and non-DEMP backgrounds as unstudied. This omission is load-bearing because the two reconstruction paths have different sensitivities to extra activity. IDOLA requires exactly two clusters passing photon-shape cuts and at least one additional cluster; dark-count hits or beam-gas showers can create fake photon-like clusters, merge neutron and photon showers, or produce a third photon candidate, breaking the photon-pair constraint that drives the displaced-vertex estimate. The GNN was trained and tested only on single-Lambda and single-neutron events, so its quoted 43-70% efficiency and <1% neutron misidentification are upper bounds in a topology where the ZDC may see multiple forward particles. The summary itself conditions the feasibility margin on 'more realistic simulations, including backgrounds such as noise from SiPMs.' The paper is an honest clean-simulation feasibility study, but the specific quantified margin claimed in the abstract and summary is not yet established.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports a Geant4-based feasibility study of reconstructing Lambda0 -> n pi0 decays at the EIC using a high-granularity SiPM-on-tile zero-degree calorimeter. For Lambda0 energies between 50 and 300 GeV, the authors propose an iterative displaced-vertex reconstruction (IDOLA) that uses the pi0 mass constraint to locate the decay vertex, and a graph neural network that directly regresses the Lambda0 kinematics and classifies Lambda0 vs neutron events. They find that energy and polar-angle resolutions meet the EIC Yellow Report requirements, that the GNN improves reconstruction efficiency over the conventional topocluster method, and that the neutral channel extends the measurable energy range relative to the charged channel. The paper includes geometric acceptance studies, mass resolution, polarization sensitivity estimates, and a discussion of the impact of beam spread on pT resolution.","tokens_in":17426,"tokens_out":5220,"duration_ms":44627,"significance":"If the reported performance holds under realistic conditions, this work would establish a new path for forward-Lambda measurements at the EIC, enabling kaon structure and spin studies that were previously deemed impractical with the ZDC alone. The analysis has notable strengths: the Geant4 setup is validated against CALICE test-beam data, the GNN training/test split on continuous vs discrete energies avoids memorization, and the code and trained models are publicly released. The principal weakness is that the quoted efficiencies and resolutions are obtained from clean single-particle events; the paper's own limitations section acknowledges that noise and beam-gas backgrounds are not included. The feasibility margin claimed in the summary is therefore conditional on follow-up studies with more realistic event samples.","major_comments":[{"comment":"The simulation includes no electronic noise, beam-gas background, or underlying event; Section 2.2 states 'No noise was included in the simulation,' and Section 4.1 acknowledges that 'Background from beam-gas interactions, or SiPM noise, could affect these numbers as they are not included.' IDOLA requires exactly two clusters passing photon-shape cuts and at least one additional cluster, while the GNN was trained and tested only on single-Lambda and single-neutron events. Consequently, the 43-70% Lambda efficiencies and the '<1%' neutron misidentification are clean-simulation values, and the 'comfortable margin' claimed in Section 5 is not yet established for realistic running conditions. I recommend either adding a simplified noise/background overlay study (e.g., SiPM dark-count hits and beam-gas photons) or explicitly qualifying the summary as a clean-simulation feasibility result.","section":"Section 2.2 and Section 4.1"},{"comment":"The photon-identification cuts (length < 100 mm, width < 12 mm, z_cl - z_front < 300 mm) are defined in Section 3.2 using the same simulated sample on which the Section 4.1 efficiencies are reported, with no separate optimization sample or scan over cut values. If these cuts are tuned to the specific event topologies of the clean sample, the quoted 4-8% (IDOLA) and 43-70% (GNN) efficiencies could be optimistic. Please validate the stability of the efficiencies against reasonable variations of the cuts, or evaluate them on a dedicated tuning sample.","section":"Section 3.2 and Section 4.1"}],"minor_comments":[{"comment":"There is a typo in the summary: 'distibution' should be 'distribution'.","section":"Section 5"},{"comment":"In the middle panel of Figure 6, the fit label reads 'fit: = 0.03' with the fitted parameter omitted; it should be 'fit: mu = 0.03' (or similar). Also, the right-panel legend 'fit: 0.13/sqrt(E)' for photons omits the negative sign, which is inconsistent with Eq. (9) where B_gamma = -0.13.","section":"Figure 6"},{"comment":"In the Code Availability section, 'The code ... are found at' should be 'The code ... is found at' (singular subject).","section":"Code Availability"},{"comment":"The sentence 'These cuts were inspired by those used in CALICE prototypes [17,18] and the anticipated capabilities of the EIC ASIC for SiPM readout' is slightly ambiguous about which cuts (ADC dynamic range, threshold, or time window) are being referred to; consider listing them explicitly.","section":"Section 2.2"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is appropriate for NIMA in scope. The main concern is the disconnect between the abstract/summary claim of a 'comfortable margin' and the clean-simulation nature of the study; I would advise the editor that a major revision is warranted, but the paper is not fatally flawed. The reliance on the authors' prior work (Ref. [14]) is reasonable and well cited."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nTwo things to know up front. First, this is the first detailed detector-simulation study of Lambda0 -> n pi0 at the EIC ZDC, and it is a genuine step forward. Second, the performance numbers are all clean-simulation values: no electronic noise, no beam-gas interactions, no underlying event. The authors say so honestly in Sec. 2.2 and Sec. 4.1, but the summary's \"comfortable margin\" over the Yellow Report requirements is a statement about a noiseless detector, not yet about the EIC.\n\nWhat is actually new: the IDOLA iterative vertex finder is simple and clever. It uses the pi0 mass constraint to locate a displaced vertex tens of meters downstream, and in simulation it works with negligible bias and roughly 0.6-0.9 m resolution. The GNN regression/classification achieves 43-70% Lambda efficiency with less than 1% neutron misidentification at the tight cut, well above the 4-8% from conventional clustering. The energy resolution, around 38%/sqrt(E), and the polar-angle resolution both meet the Yellow Report neutron requirements. The polarization projection, using the neutron direction in the Lambda rest frame, is a useful addition. The paper is honest about its limitations: Sec. 4.1 explicitly lists beam-gas, SiPM noise, and non-DEMP backgrounds as unstudied, and the summary calls for more realistic simulations. Code and trained models are posted, though without datasets or a pinned commit, full reproduction would take some effort.\n\nThe soft spots are real but proportionate. The no-noise/no-background assumption is the main one. IDOLA requires exactly two photon-like clusters plus at least one more cluster; fake photon-like clusters or merged showers from extra activity could break the pair constraint or bias the vertex. The GNN was trained only on single-Lambda and single-neutron events, so the quoted efficiencies are upper bounds in any topology where the ZDC sees multiple forward particles. I would want either a noise/background study or a clearly softened claim before the \"comfortable margin\" language stays. Two minor issues: the photon-ID cuts appear to be tuned on the same sample used to report efficiency, and the quoted resolutions carry no uncertainties. The calibration of energy corrections to simulated truth is standard detector practice, not circular reasoning.\n\nThe central feasibility claim holds as a clean-simulation statement. It does not yet hold as established EIC-level feasibility, which is exactly what peer review should probe.\n\nMy recommendation: this paper deserves a serious referee. It is clearly written, builds on test-beam-validated simulation work, and gives the EIC community a concrete target. I would send it out, asking for a background/noise study or a tempered abstract.\n\nBest,\n\n[Your name]","headline":"A clearly argued clean-simulation feasibility study of a genuinely unstudied channel; worth refereeing, but the 'comfortable margin' over Yellow Report requirements is not yet established because noise and backgrounds are absent.","tokens_in":18061,"tokens_out":1826,"would_cite":true,"duration_ms":18549,"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 the neutral decay $\\Lambda^0 \\to n\\pi^0$ can be reconstructed at the EIC using a high-granularity zero-degree calorimeter, meeting the Yellow Report requirements.","keywords":["Lambda baryon","zero-degree calorimeter","displaced vertex reconstruction","graph neural network","EIC forward physics","kaon structure","hyperon polarization","calorimeter clustering"],"falsifier":"Simulate the same $\\Lambda^0 \\to n\\pi^0$ events with realistic SiPM noise, beam-gas background, and underlying event added, then rerun IDOLA and the GNN: if the reconstructed $\\Lambda^0$ mass peak broadens or shifts, the GNN classification efficiency drops below the neutron rejection needed, or the $38\\%/\\sqrt{E}$ resolution degrades materially, the feasibility claim as stated fails. A faster proxy is to inject random low-energy hits into the existing simulated events and measure how the two-photon invariant-mass constraint and cluster separation respond.","tokens_in":2037,"feed_emoji":"⚛️","tokens_out":2173,"duration_ms":58044,"temperature":0.7,"pith_summary":"This paper argues that the neutral decay $\\Lambda^0 \\to n\\pi^0$, previously considered out of reach at forward angles, can be reconstructed at the future Electron-Ion Collider using a high-granularity zero-degree calorimeter placed 35 m from the interaction point. The central problem is that $\\Lambda^0$ baryons with energies around 100 GeV travel tens of meters before decaying, so the decay vertex is unknown and meters downstream. The paper introduces an iterative vertex-finding algorithm, IDOLA, that locates the displaced vertex by requiring the two photons to reconstruct the $\\pi^0$ mass, and separately trains a graph neural network on the calorimeter hits. With either approach the simulated energy resolution reaches about $38\\%/\\sqrt{E}$ and the polar-angle resolution is well below the Yellow Report requirement, matching the performance previously reported for single neutrons. If these results hold under real conditions, the neutral channel would complement the charged decay $\\Lambda^0 \\to p\\pi^-$ and extend kaon-structure and spin measurements to energies the charged channel cannot reach.","feed_headline":"Neutral Lambda decay channel passes EIC detector test","feed_subtitle":"Simulations show Lambda0 -> n pi0 meets Yellow Report resolutions, opening higher-energy kaon studies.","key_machinery":"The load-bearing objects are IDOLA and a graph neural network. IDOLA (Iterative Decay Origin for Lambda Analysis) is a bisection-style algorithm: it starts with the decay vertex at the interaction point, computes the $\\Lambda^0$ momentum from the three daughter clusters, then moves the assumed vertex along that momentum direction, halving the step at each iteration, until the opening angle of the two photons reproduces the known $\\pi^0$ mass; this recovers the displaced vertex, which can lie tens of meters downstream, with about 0.6–0.9 m resolution. The GNN treats the calorimeter hits as a point cloud with energy and position features, regresses the $\\Lambda^0$ energy and angles, and classifies $\\Lambda^0$ events against neutron background without explicitly estimating the vertex. Supporting machinery includes the HEXPLIT subcell-splitting algorithm, which exploits staggered calorimeter layers to improve transverse granularity, and topological clustering to separate the neutron and two photon showers.","core_discovery":"On its own terms, the paper establishes that $\\Lambda^0 \\to n\\pi^0$ is a viable measurement channel for the EIC's zero-degree calorimeter. The authors simulate 50–300 GeV $\\Lambda^0$ decays, reconstruct the three daughter showers with HEXPLIT subcell splitting and topological clustering, and infer the displaced decay vertex either with IDOLA or implicitly with a graph neural network. They find that the $\\Lambda^0$ energy resolution, about $38\\%/\\sqrt{E}$, and polar-angle resolution are essentially the same as for single neutrons and satisfy the EIC Yellow Report requirements for neutron measurements; the geometric acceptance rises from about 2% at 100 GeV to about 35% at 250 GeV, in contrast to the roughly 1% acceptance quoted for the charged decay channel at the $18\\times 275$ GeV setting. The paper also estimates that the neutron direction in the $\\Lambda^0$ rest frame, the observable that carries $\\Lambda^0$ spin information, can be reconstructed with roughly 100–120 mrad polar and 50–60 mrad azimuthal resolution by the conventional method.","pith_inferences":["The IDOLA displaced-vertex technique is not limited to $\\Lambda^0$; any long-lived neutral particle decaying to a photon pair plus a visible hadron at a forward calorimeter could use the same mass-constraint bisection, so the method could be adapted to other hyperon and meson decays at EIC-class facilities.","The quoted noiseless simulation numbers are an upper bound on real performance: adding SiPM noise, beam-gas background, and underlying-event activity could merge the neutron and photon clusters that IDOLA and the GNN rely on to separate, which is the most direct stress test the paper leaves for future work.","The GNN's edge over the conventional method in $\\Lambda^0$ efficiency (43–70% versus 4–8%) suggests that, when noise eventually degrades topological clustering, a point-cloud model trained on noisy hits may retain more of the signal than an explicit vertex-finding chain; this is testable in the follow-up simulations the paper proposes.","If the acceptance-versus-energy trend holds, the same ZDC design with a wider angular coverage could serve the lower-energy EicC program, since the displaced-vertex problem becomes less severe at lower boost."],"forward_implications":["At 50–300 GeV, the neutral channel's geometric acceptance grows with energy, so it reaches the higher-energy, lower-$x$ and higher-$Q^2$ region where the charged channel $\\Lambda^0 \\to p\\pi^-$ is blocked by tracker and magnet acceptance.","Combining the roughly 2–35% acceptance with the GNN's 43–70% identification efficiency, the authors estimate about 36k reconstructed $\\Lambda^0 \\to n\\pi^0$ events at an integrated luminosity of $10\\,\\mathrm{fb}^{-1}$, enough to begin kaon-structure and spin analyses.","The measured neutron direction in the $\\Lambda^0$ rest frame can be used to extract $\\Lambda^0$ polarization through the angular distribution $dP/d\\Omega_n = 1 + \\alpha \\vec{P}_{\\Lambda^0}\\cdot\\hat{p}_n$, with the conventional reconstruction giving 100–120 mrad polar resolution.","Because the neutron carries 75–94% of the $\\Lambda^0$ energy, the $\\Lambda^0$ energy resolution is essentially the neutron energy resolution, so improvements in neutron calorimetry directly improve $\\Lambda^0$ measurements.","The methods also apply to $\\Sigma^0 \\to \\Lambda^0\\gamma \\to n\\pi^0\\gamma \\to n\\gamma\\gamma\\gamma$, providing a cross-check for kaon form-factor extractions, albeit with only 2.1% of $\\Sigma^0$ events passing the selected cuts."],"supporting_citations":[{"why":"Supplies the SiPM-on-tile ZDC geometry, simulation setup, and single-neutron GNN performance baselines that all $\\Lambda^0$ results are compared against.","marker":"[14]"},{"why":"Defines the EIC Yellow Report science requirements and prior kinematic study of $\\Lambda^0 \\to n\\pi^0$, and provides the neutron energy and angle resolution targets.","marker":"[4]"},{"why":"Suggested measuring this decay by combining ZDC neutron detection with B0-region electromagnetic calorimetry; this paper's clustered approach goes beyond that kinematic suggestion.","marker":"[7]"},{"why":"Establishes the charged-decay channel feasibility at low beam energies and motivates $\\Lambda^0$ polarization measurements that the neutral channel extends.","marker":"[10]"},{"why":"Provides the HEXPLIT subcell-splitting algorithm that gives the detector its effective transverse granularity for separating the three showers.","marker":"[23]"},{"why":"Supplies the topological cluster-finding algorithm used to group subcell hits into neutron and photon showers.","marker":"[24]"},{"why":"Provides the leading-$\\Lambda$ production cross-section estimate used to project the roughly 36k-event sample size.","marker":"[38]"}],"fun_headline_variants":["EIC calorimeter sim validates neutral Lambda decays","Lambda0 to neutron-pion0 passes EIC resolution test","Neutral Lambda channel extends EIC energy reach","Graph neural network aids Lambda0 detection at EIC","Zero-degree calorimeter enables Lambda0 spin studies"],"cache_read_input_tokens":20096,"weakest_assumption_plain":"The whole feasibility result rests on a simulation with no SiPM electronic noise, no beam-gas background, and no underlying-event activity, so the quoted efficiencies and resolutions are clean-simulation values.","fun_headline_variants_meta":{"raw":{"variants":["EIC calorimeter sim validates neutral Lambda decays","Lambda0 to neutron-pion0 passes EIC resolution test","Neutral Lambda channel extends EIC energy reach","Graph neural network aids Lambda0 detection at EIC","Zero-degree calorimeter enables Lambda0 spin studies"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000538,"raw_usage":{"total_tokens":2632,"prompt_tokens":1042,"completion_tokens":1590,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":658,"completion_tokens_details":{"reasoning_tokens":1525}},"tokens_in":658,"tokens_out":1590,"duration_ms":11691,"temperature":1.0,"reasoning_tokens":1525,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T14:10:41.343516+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Simulate the same $\\Lambda^0 \\to n\\pi^0$ events with realistic SiPM noise, beam-gas background, and underlying event added, then rerun IDOLA and the GNN: if the reconstructed $\\Lambda^0$ mass peak broadens or shifts, the GNN classification efficiency drops below the neutron rejection needed, or the $38\\%/\\sqrt{E}$ resolution degrades materially, the feasibility claim as stated fails. A faster proxy is to inject random low-energy hits into the existing simulated events and measure how the two-photon invariant-mass constraint and cluster separation respond.","supporting_citations":[{"cited_title":"Deep Exclusive Meson Production as a probe to the puzzle of $\\Lambda$ hyperon polarization","cited_arxiv_id":"2308.09127","evidence_quote":"Establishes the charged-decay channel feasibility at low beam energies and motivates $\\Lambda^0$ polarization measurements that the neutral channel extends."},{"cited_title":"Leveraging Staggered Tessellation for Enhanced Spatial Resolution in High-Granularity Calorimeters","cited_arxiv_id":"2308.06939","evidence_quote":"Provides the HEXPLIT subcell-splitting algorithm that gives the detector its effective transverse granularity for separating the three showers."},{"cited_title":"Leading $\\Lambda$ production in future electron-proton colliders","cited_arxiv_id":"2306.09813","evidence_quote":"Provides the leading-$\\Lambda$ production cross-section estimate used to project the roughly 36k-event sample size."}],"review_version":1}