{"id":"14d3a502-12a2-422b-bc57-1f0fb31d9875","arxiv_id":"1908.06114","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"NG-LAMP, a drone- or hand-carried platform with CLLBC detectors, localizes gamma-ray and neutron sources in 3D in real time, including shielded neutron sources.","lead":"A team at Lawrence Berkeley National Laboratory mounted a combined gamma-ray and neutron detector on a drone and a backpack, and showed it can locate radioactive sources in 3D in real time. The system finds neutron-emitting material even when its gamma rays are hidden by shielding, which matters for nuclear security and emergency response.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Neutron-selection purity is unquantified: with pulse-height-only discrimination, gamma pile-up from the strong 60Co source could leak into the neutron channel and produce the claimed neutron hotspot as an artifact.","rationale":"Good-faith reading: the paper is an engineering demonstration. The strongest claim is the first 3D real-time neutron localization in a strong gamma field. The textual evidence is plausible: the authors show a 60Co and a shielded Pu surrogate separated by 10 m, reconstruct both with the same parameters, and only particle discrimination differs. If the neutron channel is pure, the demonstration is credible. The weakest link is exactly the purity of that channel; the reader identified this. I partially disagree with the reader's phrasing: for most results the discrimination is pulse-height only, not pulse-shape, making gamma pile-up a more realistic failure mode than the reader implied. This is not an internal inconsistency; it is an unverified external condition that can be settled by a calibration run. The verdict remains CONDITIONAL because the paper needs this quantitative evidence before the 'first demonstration' claim is fully supported. I would not reject the paper: the qualitative images and the spatial separation of the two sources are consistent with the claim, and the authors are careful to say 'we show ... the ability,' not to report quantitative accuracy. However, the discrimination characterization is necessary to rule out the pile-up artifact and to validate the central claim.","tokens_in":8855,"tokens_out":7073,"duration_ms":73814,"concrete_test":"Perform a calibration run: expose NG-LAMP to a 500 μCi 60Co source at 1 m with no neutron source present; apply the identical pulse-height neutron selection used in Figure 3 and record accepted-event rate and the energy spectrum of accepted events as the source is moved from 0.5 m to 3 m. If the accepted rate scales with the 60Co counting rate and the accepted-energy spectrum shows pile-up sum peaks rather than only the CLLBC neutron capture peak, gamma leakage is significant. The claim would be supported if the leakage rate at 1 m is at least an order of magnitude below the true neutron rate from the 1 SQ surrogate under the same geometry; otherwise the Figure 3 neutron hotspot is ambiguous.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim requires that every event fed into the neutron MLEM is a genuine neutron event. The paper states in Methods: 'Pulse height discrimination is used as a matter of convenience for most results that follow, due to low fast neutron sensitivity in CLLBC.' Thus, for the key Figure 3 demonstration ('only particle discrimination' separating the two maps), the neutron/gamma separation is a pulse-height cut, not the pulse-shape discrimination that the system also provides. With a 500 μCi 60Co source at ~1 m standoff, the gamma interaction rate in the four CLLBC crystals is high; gamma pile-up can sum two or more 60Co pulses into the neutron-selection window, and no characterization of this leakage is provided. The 60Co and Pu surrogate are ~10 m apart, so a leakage fraction that is only a few percent near the truck could contribute counts comparable to the true neutron rate from the 1 SQ surrogate. If that happens, the 'neutron-only' MLEM hotspot in Figure 3 could be biased toward, or even be an artifact of, the gamma source. The paper reports no neutron-channel energy spectrum, no measured gamma rejection ratio, and no control measurement with the 60Co present and the neutron source absent, so the first-demonstration claim is not yet separated from a pile-up artifact.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper describes Neutron Gamma LAMP (NG-LAMP), an extension of the LAMP platform that combines four CLLBC scintillators with real-time 3D scene data fusion to localize gamma-ray and neutron sources simultaneously. Three demonstrations are reported: a UAS flight that simultaneously localized a 500 μCi 60Co source (gamma rays) and a 1 SQ Pu surrogate (neutrons) placed about 10 m apart; a handheld survey that spectroscopically localized three gamma-ray sources and one PuBe neutron source; and a UAS flight around a vehicle containing a heavily shielded Pu surrogate, where the gamma-ray reconstruction is said to be consistent with a near-uniform background and the neutron reconstruction localized the source. The central claims are that this is the first real-time 3D neutron-localization capability from a small UAS/handheld platform, and that neutron signatures alone can locate shielded special nuclear material.","tokens_in":9062,"tokens_out":3483,"duration_ms":38849,"significance":"If the claims hold, the work is significant for nuclear security and emergency response: it would be the first demonstrated real-time 3D dual-particle mapper from a small UAS and handheld configurations, extending prior LAMP gamma-ray work to neutron signatures. The paper benefits from field-scale demonstrations, independent scenarios, and the use of existing, published MLEM/SLAM components rather than any circular fitting to the results. The manuscript is generally clear about what was measured, and the reconstructions appear consistent with the described source placements. However, the central claims rest on an unquantified neutron-selection purity and on qualitative statements about localization and background uniformity, which are not yet supported by the evidence presented.","major_comments":[{"comment":"The neutron-selection purity is not quantified, and this is load-bearing for the headline claim of neutron localization in the presence of a strong gamma-ray source. The manuscript states in Methods that 'Pulse height discrimination is used as a matter of convenience for most results that follow, due to low fast neutron sensitivity in CLLBC,' and Figure 3's caption says the neutron map was produced with 'only particle discrimination' applied. If the Figure 3 discrimination is a pulse-height cut rather than pulse-shape discrimination, then gamma-ray pile-up from the 500 μCi 60Co source could leak into the neutron window, especially near the truck. The paper reports no neutron-window energy spectrum, no gamma rejection ratio, no count rates in the neutron channel as a function of position, and no control measurement with the 60Co present and the neutron source absent. Without these, the possibility that the apparent neutron hotspot is biased or even caused by gamma leakage is not excluded. Please report the measured gamma rejection into the neutron channel and the neutron-window count rates along the flight path, or otherwise demonstrate that the neutron-only map is not contaminated.","section":"Methods and Approach; Results A, Figure 3"},{"comment":"No quantitative localization accuracy or ground-truth comparison is provided. The text says sources were 'correctly localize[d]' and that the neutron source was localized 'to the vehicle' or 'in the back of a vehicle,' but no coordinates, distances to known source positions, or localization-error metrics are given. For a paper whose central claim is 3D real-time localization, at least a table of true versus reconstructed source positions (or a distance error per source) is needed to substantiate the claim. Without this, the demonstrations show that hot spots appear in plausible locations, but not that the system localizes with any quantified accuracy. Please add quantitative comparisons for each demonstration.","section":"Results A, B, and C (Figures 3, 4, 6)"},{"comment":"The claim that the gamma-ray reconstruction is 'consistent with a near uniform distribution' is unsupported. The manuscript provides no measure of uniformity (e.g., a chi-square statistic, a flatness metric, or a comparison against a background-only measurement), no expected background model, and no uncertainty. Since the final scenario explicitly relies on gamma-ray data being 'consistent with background,' this qualitative statement is load-bearing. Please provide a quantitative uniformity test or the reconstructed gamma-ray source distribution with error bars, and state the acceptance criterion used to judge it 'nearly uniform.'","section":"Results C, Figure 6"},{"comment":"The reconstruction parameters are underspecified, which limits reproducibility and the strength of the 'we did not alter gamma-ray or neutron reconstruction parameters' claim. The paper states that three MLEM iterations were used for the Figure 4 data, but does not report iteration counts, voxel sizes, time binning, or the angular-detection-efficiency model for the other reconstructions. Since MLEM iteration count and the efficiency model directly affect the reconstructed distributions and the claimed 'same reconstruction parameters' comparison in Figure 3, please specify these parameters for every demonstration or state where they are defined in the cited LAMP references.","section":"Methods and Approach; Conclusion"}],"minor_comments":[{"comment":"The phrase '3.5% FWHM at 661.7 keV' would be clearer as 'energy resolution of about 3.5% FWHM at 661.7 keV.' Also, the sentence in the Introduction that begins 'This is accomplished with four monolithic radiation detectors...' would read better if split into two sentences.","section":"Abstract and Introduction"},{"comment":"The caption says 'Three iterations of MLEM were used to estimate source locations,' but the body text in Section B does not mention the iteration count for the other figures. Please state the iteration count and all reconstruction parameters in one place, or in each caption.","section":"Figure 4 caption"},{"comment":"The color coding is described in the body text as 'neutron source in red' and 'gamma-ray data... in blue,' but the caption in the provided text does not reproduce this color description. Please make the caption self-contained so the figure is interpretable without the body text.","section":"Figure 6 caption"},{"comment":"Reference [1] is to an arXiv preprint, but the published version or a more stable citation would be preferable. Reference [6] for RMD lists only the company name and city; a product page or datasheet would be more useful.","section":"References"},{"comment":"There are typographical inconsistencies, such as 'Simultaneous and Mapping (SLAM)' in the Methods section, which should be 'Simultaneous Localization and Mapping,' and 'K. Vetter' appearing with lowercase 'k' in Reference [1]. A careful proofreading pass is recommended.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a field-demonstration paper with a defensible central idea, but the quantitative support is currently too thin for the strength of the 'first demonstration' claims. The neutron-purity concern is the most serious: because the paper itself states that pulse-height discrimination was used for most results, the Figure 3 claim must be backed by measured gamma rejection and count-rate information. I would advise requiring the additional analysis (neutron-window spectra, rejection ratio, localization-error table, and a quantitative uniformity test) before publication, rather than treating these as merely cosmetic improvements."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nWhat you should know: this is the first believable demonstration of real-time 3D neutron localization from a small UAS and handheld platform, using the LAMP architecture they've built over the past few years. It extends their gamma-ray SDF work by swapping in CLLBC detectors that see both gammas and neutrons, and the field tests are real: a concealed-source exercise with 60Co and a shielded Pu surrogate, a four-room spectroscopic survey, and a shielded 1-SQ surrogate in a vehicle. The paper is an engineering report, not a physics breakthrough, but the integration and the field results look genuine.\n\nWhat's good: the multi-source spectroscopic localization (Figures 4/5) is a nice demonstration—photo-peak windowing on three gamma sources and neutron capture simultaneously, with ~2 Hz updates. The shielded-SNM scenario is the one that matters for the application, and the reconstruction showing a neutron hotspot with gamma consistent with background is exactly the right data product. I also appreciate that they state clearly that pulse-height discrimination was used for most results, rather than hiding that fact.\n\nWhere I'd push back: the evidence is qualitative. There are no localization error numbers, no ground-truth coordinates, no statistical comparison of reconstructed source positions to known positions. The \"near uniform\" gamma background is asserted, not quantified. And the biggest gap: the neutron/gamma separation quality is never characterized. With a 500 μCi 60Co source at ~1 m standoff, the gamma count rate in CLLBC is high. They used a pulse-height cut for Figure 3, and they don't report gamma rejection ratio or a control run with 60Co present and the neutron source absent. The stress-test scenario—pile-up leaking into the neutron window—is plausible in principle. But note: if that leakage were the dominant neutron signal, the reconstructed hotspot would sit on the 60Co truck, not on the SUV ten meters away where they claim the Pu surrogate. So I don't think the Figure 3 claim is likely an artifact; still, the missing control measurement makes the confidence lower than it should be.\n\nBottom line: this is a solid, useful engineering demonstration that deserves peer review. I'd send it out, with the reasonable request that the authors add a small set of quantitative validations: table of reconstructed vs known source positions, a leakage measurement for the neutron channel under high gamma rate, and a statistical description of the background reconstruction. For those who work in applied nuclear detection, this is worth a read.\n\nRecommendation: accept for review; expect a conditional decision pending those additions.","headline":"A real engineering first—real-time 3D neutron mapping on a drone—with the usual field-demo soft spots: no error bars, no ground-truth table, and no quantitative look at how clean the neutron channel really is.","tokens_in":9678,"tokens_out":2388,"would_cite":true,"duration_ms":25882,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A compact detector platform called NG-LAMP establishes that neutron sources can be localized in three dimensions and in real time from a drone or in hand, even when a much stronger gamma-ray source is nearby and even when the neutron…","keywords":["3D radiation mapping","neutron localization","gamma-ray spectroscopy","CLLBC scintillator","unmanned aerial system","scene data fusion","special nuclear materials","real-time reconstruction"],"falsifier":"Run NG-LAMP with a strong gamma-ray source and no neutron source, process the data through the neutron-only reconstruction, and look for a false hotspot. A quantitative version would measure the gamma rejection ratio of the CLLBC neutron channel and require that the reconstructed neutron intensity at the gamma source stay within background noise before accepting the neutron-localization claim.","tokens_in":8632,"feed_emoji":"☢️","tokens_out":6590,"duration_ms":57680,"temperature":0.7,"pith_summary":"This paper reports a system that maps gamma-ray and neutron sources in three dimensions, in real time, from a small drone or a hand-carried unit. The authors integrate the existing LAMP scene-data-fusion pipeline with four CLLBC scintillator crystals that detect both gamma rays and neutrons, calling the result NG-LAMP. They claim first demonstrations: a neutron source localized while a much stronger gamma-ray source sat nearby; simultaneous spectroscopic localization of three gamma sources plus a neutron source; and localization of a shielded plutonium surrogate by neutron signal alone, with the gamma-ray reconstruction flat as background. If true, this extends 3D scene-data-fusion radiation mapping from gamma-only to dual-particle search, which matters for finding shielded special nuclear materials.","feed_headline":"Drone-borne system maps neutron and gamma sources in 3D live","feed_subtitle":"The system localizes shielded neutron sources in real time, even beside strong gamma-ray emitters.","key_machinery":"The load-bearing mechanism is the fusion of two established pieces: the scene-data-fusion pipeline, which couples a simultaneous localization and mapping (SLAM) engine with maximum-likelihood expectation-maximization (MLEM) over voxelized space, and the CLLBC scintillator, a crystal that responds to both gamma rays and thermalized neutrons with distinct pulse shapes and pulse heights. For each detected event, pulse-shape and pulse-height discrimination assigns it to a gamma-ray or neutron channel, and angular-dependent detection efficiencies per detector module enter the MLEM update. This is what lets the same reconstructed volume separate a neutron hotspot from a nearby gamma-ray hotspot using only particle discrimination, and lets photopeak windowing run separate 3D reconstructions for each isotopic gamma signature.","core_discovery":"The central claim is that a single lightweight platform, carrying four monolithic CLLBC (Cs2LiLa(Br,Cl)6:Ce) scintillators and an onboard computer, can perform 3D maximum-likelihood reconstruction of both gamma-ray and neutron sources in real time, using only particle-type or photopeak selection to separate the channels. The paper reports three demonstrations: an eight-minute UAS flight that localized a 500 microcurie cobalt-60 source and a one-significant-quantity plutonium surrogate, with the neutron source shielded by 14 cm of lead, in the same reconstruction volume; a walk-through of four rooms that localized three photopeak-windowed gamma sources and a PuBe neutron source at about 2 Hz update rate; and a UAS flight around a vehicle whose shielded plutonium surrogate emitted a gamma dose rate indistinguishable from background, where the neutron reconstruction peaked at the source while the gamma reconstruction stayed nearly uniform. The authors conclude that NG-LAMP localizes neutron point sources in 3D and in real time in measurements under ten minutes, without changing reconstruction parameters except energy or particle cuts.","pith_inferences":["Because the gamma-ray reconstruction is used as a null check when the source is heavily shielded, the same dual-channel data could be used to estimate shielding thickness or residual gamma attenuation by comparing the two reconstructions.","The demonstrated sensitivity is limited by neutron count rate and CLLBC's low fast-neutron sensitivity; a natural next test is counting-rate-limited detection-distance curves for one significant quantity of plutonium under realistic shielding.","The architecture should extend to distributed neutron sources such as contamination by keeping MLEM rather than point-source fitting, though the paper only demonstrates point-like neutron sources.","Neutron-only localization to a specific vehicle compartment suggests that wide-area aerial neutron search could be practical, but the false-alarm rate from gamma leakage into the neutron channel remains the key quantity to characterize."],"forward_implications":["NG-LAMP can localize a neutron source in the presence of a far more active gamma-ray source from a single sUAS flight of about eight minutes, using only particle discrimination between channels.","The same reconstruction parameters, with only energy or particle cuts changed, produce gamma-ray maps that remain background-like when the gamma signal is shielded, giving an internal consistency check on the neutron-only localization.","Spectroscopic photopeak windowing supports near-real-time (about 2 Hz update rate) simultaneous localization of multiple isotopic gamma sources plus a neutron source in a handheld survey.","Both fission-neutron (Pu surrogate) and alpha-n (PuBe) sources were localized, so the method is not tied to one neutron energy spectrum.","Switching between handheld, ground-vehicle, and sUAS mounts takes simple mechanical changes under five minutes, so the same system can serve search and response roles."],"supporting_citations":[{"why":"Establishes the LAMP platform and its real-time 3D gamma-ray mapping capability, which NG-LAMP extends to neutrons.","marker":"[1]"},{"why":"Supplies the scene-data-fusion and MLEM reconstruction method used to produce the 3D volumetric source maps.","marker":"[2]"},{"why":"Introduces the Cs2LiLaBr6-xClx:Ce scintillator material with gamma-ray and neutron sensitivity used in NG-LAMP.","marker":"[4]"},{"why":"Documents scintillation and pulse-shape properties of the CLLBC crystal variant used for particle discrimination.","marker":"[5]"},{"why":"Provides the real-time SLAM system that gives position and orientation estimates needed to fuse detector counts into 3D reconstructions.","marker":"[9]"},{"why":"Defines the significant-quantity unit used to specify the surrogate plutonium source strength.","marker":"[10]"}],"fun_headline_variants":["Drone and backpack map 3D neutron and gamma sources live","Real-time 3D neutron localization despite strong gamma rays","Shielded neutron sources exposed by real-time 3D mapping","Simultaneous 3D neutron-gamma mapping in real time","Live 3D mapping finds neutron sources hidden by shielding"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The claim depends on CLLBC's pulse-shape and pulse-height discrimination cleanly separating neutron events from gamma-ray events; if strong gamma-ray counts leak into the neutron channel, the apparent neutron hotspot could actually be the gamma source.","fun_headline_variants_meta":{"raw":{"variants":["Drone and backpack map 3D neutron and gamma sources live","Real-time 3D neutron localization despite strong gamma rays","Shielded neutron sources exposed by real-time 3D mapping","Simultaneous 3D neutron-gamma mapping in real time","Live 3D mapping finds neutron sources hidden by shielding"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000438,"raw_usage":{"total_tokens":2317,"prompt_tokens":1130,"completion_tokens":1187,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":746,"completion_tokens_details":{"reasoning_tokens":1100}},"tokens_in":746,"tokens_out":1187,"duration_ms":9690,"temperature":1.0,"reasoning_tokens":1100,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:54:43.101063+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run NG-LAMP with a strong gamma-ray source and no neutron source, process the data through the neutron-only reconstruction, and look for a false hotspot. A quantitative version would measure the gamma rejection ratio of the CLLBC neutron channel and require that the reconstructed neutron intensity at the gamma source stay within background noise before accepting the neutron-localization claim.","supporting_citations":[],"review_version":1}