{"id":"45703b5a-9d14-4560-90ad-330d8461ec3a","arxiv_id":"2508.19577","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":9,"one_line_summary":"Microlensing could in principle separate a pulsar origin from a dark matter origin of the Galactic Centre gamma-ray excess, but detecting it requires a 10- to 10,000-fold sensitivity increase over Fermi-LAT.","lead":"This paper asks whether gravitational microlensing by stars between Earth and the Galactic Centre could reveal whether the mysterious gamma-ray glow there comes from many tiny millisecond pulsars or from dark matter. It concludes the idea works in principle but would need gamma-ray telescopes 10 to 10,000 times more sensitive than Fermi-LAT, so it is not a test for current instruments.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Event-rate predictions rest on a factor-20 velocity error and an inconsistent lens population; quantitative claims need correction, though the qualitative conclusion is likely robust.","rationale":"The paper's central physical idea is sound: point-like MSP sources are subject to caustic-crossing magnifications that extended dark-matter annihilation emission cannot experience, so microlensing is in principle a distinguishing probe. The qualitative conclusion that considerably greater sensitivity than Fermi-LAT or the APT is required also appears robust, since the detection thresholds in Section 4.2 derive from injected lightcurves and are largely independent of the event-rate normalizations. However, the quantitative event-rate claims are not supported by the paper's own parameters. I verified two internal inconsistencies: (1) the stated lens speed of 0.28 RE/day does not correspond to 200 km/s for a solar-mass Einstein radius, being a factor of about 20 too large; (2) the lens number and region radius in Table 1 imply an optical depth of roughly 4 × 10^-11, not the adopted 3 × 10^-6. These are not external uncertainties but internal arithmetic errors, so the paper must correct them before its event frequencies and the associated statements in the abstract and conclusions can be used. This does not warrant rejection, because the central claim about the need for greatly improved sensitivity does not hinge on the exact event rates, but it does reinforce the reader's CONDITIONAL verdict. The explicit acknowledgements in Section 5 of intrinsic flaring and the poorly constrained source-count distribution further limit the practical utility of the test, yet these are additional caveats rather than the primary load-bearing flaw. I therefore recommend no change to the reader's verdict.","tokens_in":16196,"tokens_out":19845,"duration_ms":174633,"concrete_test":"Recompute Equations (15)-(17) using the correct v = 0.0143 RE/day (for 200 km/s with RE = 8.1 au) and a self-consistent lens population with τ = Nlens / (R_region/RE)^2 = 3 × 10^-6 for R_region = 4 × 10^7 RE (requiring Nlens ≈ 5 × 10^9 instead of 6.3 × 10^4), and check whether the reported CC event rates change by more than a factor of ten. If they do, the paper's quantitative feasibility claims need revision.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's quantitative event-rate predictions rest on internally inconsistent lens parameters. Table 1 sets the lens speed as 0.28 RE/day for an assumed 200 km/s, but with RE = 8.1 au (Eq. 3) this velocity is 0.0143 RE/day, a factor of about 20 too high. This propagates directly into Eq. (15) and Eq. (17), inflating the swept area and therefore Nevents by roughly an order of magnitude. Independently, Table 1 lists Nlens = 6.3 × 10^4 lenses in a region of radius 4 × 10^7 RE, which gives an optical depth τ ≈ Nlens / (R_region/RE)^2 ≈ 4 × 10^-11, not the stated τ = 3 × 10^-6; a self-consistent population would require Nlens ≈ 5 × 10^9 for that region. These inconsistencies undermine the reported event rates (e.g., 'one CC event every ~1500 years' for Ns = 500) and mean the simulated lens distribution is not a faithful realization of the assumed optical depth. While the detection thresholds in Section 4.2 are computed from injected lightcurves and are less sensitive to these errors, the claimed event frequencies and the abstract's statement that 'events would occur multiple times a year' for a large population are not reliable until the velocity and lens-number inputs are corrected.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper proposes a new observational test of the origin of the Galactic Centre gamma-ray excess (GCE). If the GCE arises from a population of millisecond pulsars, gravitational microlensing by stellar-mass lenses along the bulge line of sight will produce time-dependent magnifications, including rare caustic-crossing spikes with peak magnification about 8700; if the GCE arises from smooth dark-matter annihilation, the emission is extended on AU scales and effectively immune. The authors model three source populations (Ns = 1e5, 500, 5000), simulate lens configurations and Poisson photon light curves, and compute the detector sensitivity (scaled by a relative to Fermi-LAT) needed to identify microlensing events. They find that Fermi-LAT cannot detect these events and that even the proposed APT (log10 a about 1) is insufficient; for a bright 500-source population, caustic-crossing events could be seen with log10 a about 1.4 but are extremely rare, and for a faint 1e5-source population detection would need log10 a about 4.3. The paper concludes that microlensing is not a near-term discriminator but could constrain point-source populations in future observations.","tokens_in":16538,"tokens_out":9135,"duration_ms":81698,"significance":"The proposed test is conceptually novel and forward-modeled from standard microlensing physics. The distinction between point-like MSP emission and extended emission via finite-source magnification is physically sound, and the conclusion that current and near-future instruments lack sensitivity is robust to the quantitative errors discussed below. The paper provides a clearly described simulation pipeline (source placement, binary-lens caustic-crossing light curves, Poisson significance tests) and gives explicit detection thresholds. It also candidly acknowledges limitations: identical-source assumption, unknown source-count distribution, intrinsic flaring, and look-elsewhere corrections. However, the quantitative event rates rest on internally inconsistent lens parameters, so the numerical predictions in Sections 4.2.1 and 4.2.2 and the abstract's 'multiple times a year' statement need correction; the central methodological claim is not invalidated.","major_comments":[{"comment":"The lens population in Table 1 is internally inconsistent with the stated optical depth. For a circular region of radius R_region = 4e7 R_E, the optical depth is tau = N_lens / (R_region/R_E)^2; with N_lens = 6.3e4 this gives tau about 4e-11, not 3e-6. Matching tau = 3e-6 requires N_lens about 4.8e9. The simulations therefore do not generate a lens distribution that realizes the assumed optical depth, and the event rates computed from Eq. (17) and the counts in Figure 7 are not grounded in the simulated lens field.","section":"Section 4.1, Table 1 and Eq. (17)"},{"comment":"The stated conversion of the lens velocity is wrong by a factor of about 20. With R_E = 8.1 au for a solar-mass lens, v = 200 km/s corresponds to 0.0143 R_E/day, not 0.28 R_E/day. This overestimates the swept-area factor in Eq. (15), which also appears to be evaluated as 74 rather than about 36 given the stated inputs. The error propagates into Eq. (17), so all event-rate claims—the CC rates of one per 1500 and 150 years in Section 4.2.1, the per-century color bars in Figure 7, and the abstract's statement that a large population would produce 'events multiple times a year'—must be recomputed.","section":"Section 2.2 and Table 1; Eq. (15)"},{"comment":"Even using the published Table 1 inputs, the quoted CC event rates do not follow from Eq. (17). For Ns = 500, y = 2, tau = 3e-6, v = 0.28 R_E/day, and DeltaT = 11 years, Eq. (17) gives about 2 events with y < 2 in the survey, or about 0.07 CC events after the 3.5% fraction; this corresponds to one CC per about 150 years, not one per 1500 years. The analogous discrepancy affects the Ns = 5000 rate. The rates need to be reconciled with the equations and with the corrected velocity and lens-number inputs; the detection thresholds in Figure 6 are less affected because they use injected events, but they do not rescue the event-rate claims.","section":"Section 4.2.1"}],"minor_comments":[{"comment":"The fiducial value 74 for the swept-area factor is not consistent with v = 200 km/s and R_E = 8.1 au; please state the actual numeric value and use it consistently throughout.","section":"Section 3, Eq. (15)"},{"comment":"The relationship between the 70-day in-caustic plateau magnification of 'an order of magnitude' and the caustic-crossing spikes with mu about 8700 should be clarified; the detection algorithm appears to use only the two peaks, while the rate statement about 4 photons in a 70-day period uses the plateau.","section":"Section 4.2.1"},{"comment":"Figure 1 would benefit from labeling which panel corresponds to the q = 1, s = 1 configuration used in all CC simulations, since the reader cannot otherwise identify the lightcurve shape adopted in Section 4.","section":"Figure 1"},{"comment":"The text refers to 'Figure 5' for the minimum log10(a) as a function of maximum magnification, but the actual figure is Figure 7; correct the cross-reference.","section":"Section 4.2.2, Case 1"},{"comment":"The stated threshold ymax = 4 R_E yields a maximum single-lens magnification of about 1.006, i.e. a 0.6% enhancement, not 'approximately 6%'; please correct the numerical value.","section":"Table 1 and Section 4.1, step 3"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within the scope of PASA and the central idea is worth publishing after correction. The problems are concentrated in the event-rate normalization: the lens number, the velocity conversion, and the quoted CC rates must be made mutually consistent, and the affected numbers in the abstract, Section 4.2, and Figure 7 must be recomputed. I see no grounds for rejection, but the quantitative claims should not appear in their present form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know. First, the core suggestion is new: use gravitational microlensing of the Galactic Centre as a temporal discriminator between a smooth dark-matter GCE and a population of point-like MSPs. That is not in the prior morphology/spectrum/statistics literature, and the paper surveys that literature fairly. Second, the printed quantitative predictions are not internally consistent, so treat the event rates as illustrative until corrected.\n\nWhat the paper does well: it gets the basic physics right (Einstein radius vs source size; caustic magnification limits ~8700), builds a plausible forward simulation with Poisson photon noise, and reports detection thresholds as a function of detector sensitivity rather than pretending current instruments can see anything. The qualitative conclusion—Fermi-LAT and even the APT are far from detecting this, with only the bright-population caustic-crossing case within an order of magnitude of a future instrument—is robust to the arithmetic problems below.\n\nWhere it gets soft. Table 1 says lenses move at 0.28 RE/day for 200 km/s; with RE = 8.1 au, 200 km/s is 0.0143 RE/day, a factor of about 20 smaller. That inflates the swept area in Equations (15)–(17) and therefore the reported event frequencies. Independently, Table 1 lists Nlens=6.3e4 in a region of radius 4e7 RE, which gives optical depth ~4e-11, not the stated 3e-6; a self-consistent population would need ~5e9 lenses. And the text says 6% binaries × 70% caustic-crossing = 3.5%, but that product is 4.2%. The paper quotes one CC event per ~1500 years for Ns=500; plugging their own y<2 criterion and v=0.28 RE/day into Equation (17) gives ~146 years, about a factor of 10 more frequent. So either the velocity, the lens count, or the quoted rates are wrong.\n\nThe detection-threshold curves in Section 4.2 are less sensitive to these errors because they are computed from injected single-event lightcurves, but the \"multiple times a year\" claim for 1e5 sources and the rate comparison between cases should not be quoted until the parameters are reconciled. The source-count model (identical, steady sources) is crude, but the authors acknowledge it and it is a reasonable first cut. No code or data are shipped, and the logistic thresholds lack confidence intervals; both would strengthen the paper.\n\nWho this is for: anyone working on the GCE origin or on microlensing applications to compact object populations. It deserves a serious referee; the idea is worth publishing even if the numbers need major revision. I would send it to review and ask for a corrected Table 1, re-derived event rates, code/data release, and confidence intervals on the threshold values.","headline":"A genuinely new temporal test of the GCE origin, with a sound qualitative conclusion but several load-bearing numerical inconsistencies that must be fixed before the event rates can be trusted.","tokens_in":17054,"tokens_out":5543,"would_cite":true,"duration_ms":46558,"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":"Gravitational microlensing can tell whether the Galactic Centre gamma-ray excess comes from point-like pulsars or smooth dark matter—provided future detectors gain enough sensitivity.","keywords":["galactic centre excess","gravitational microlensing","millisecond pulsars","dark matter annihilation","gamma-ray astronomy","caustic crossing","detector sensitivity","time variability"],"falsifier":"A concrete check: inject the predicted twin-peak caustic-crossing lightcurve into one resolution element of synthetic Fermi-LAT data with photon counts scaled by $10^{1.41}$ (26x sensitivity) and run the paper's 3-day sliding-window Poisson significance search; if the twin peaks are not recovered with p ≤ 0.01, the detection threshold is too optimistic. On the observational side, a single clean measurement of the ~8700-amplification double peak separated by ~70 days from the Galactic Centre would rule out smooth dark-matter annihilation as the source of that flux, while a decade-long null search with a 26x detector in the 500-source scenario would start to constrain the compact-source fraction.","tokens_in":15990,"feed_emoji":"🔭","tokens_out":12062,"duration_ms":97444,"temperature":0.7,"pith_summary":"This paper proposes a temporal test to decide whether the gamma-ray excess at the Galactic Centre comes from annihilating dark matter or from a population of unresolved millisecond pulsars. The test exploits gravitational microlensing by ordinary stars along the line of sight: a stellar-mass lens magnifies a point-like source by roughly the ratio of the Einstein radius to the source size, so a 10-km pulsar can flash with magnification factors up to about 8700 during caustic crossings, whereas dark-matter annihilation emission, smooth over au scales, is effectively unaltered. The authors simulate populations of 500, 5000, and 100,000 identical pulsars and find that detecting these flashes requires detector sensitivity improvements over Fermi-LAT of about 26x, 200x, and 18,000x, respectively, all beyond the planned Advanced Particle-physics Telescope. The result turns the unresolved point-like-versus-extended debate into a question a future gamma-ray observatory could answer by watching the bulge for the predicted flicker.","feed_headline":"Microlensing could unmask the Milky Way's gamma-ray excess","feed_subtitle":"Pulsars would flicker under stellar lenses; dark matter would not. Catching that needs far sharper detectors.","key_machinery":"The load-bearing object is the Einstein radius in the source plane, $R_E = \\sqrt{\\frac{4GM}{c^2}\\frac{D_{os}D_{ls}}{D_{ol}}}$, about 8.1 au for a solar-mass lens toward the Galactic Bulge. The key identity is the point-source magnification $\\mu = \\frac{y^2+2}{y\\sqrt{y^2+4}}$, whose spike at a binary-lens caustic crossing is capped by the finite source radius as $\\mu_{\\max} \\sim 2R_E/r_s$; for a 10-km pulsar this yields the ~8700-fold flashes that the paper uses as its primary signature. The argument runs on the contrast between point sources and extended sources: a pulsar is eight orders of magnitude smaller than $R_E$ and therefore gets magnified, while dark-matter annihilation emission is smooth on au scales and is essentially unmodified. The caustic-crossing events come from the 6% of lenses that are binaries, 70% of which produce caustic crossings, and the paper converts the optical depth $\\tau \\sim 3\\times10^{-6}$ into event rates via the area swept out by moving lenses.","core_discovery":"The central claim is that gravitational microlensing supplies a new, template-independent discriminant between the two leading explanations of the Galactic Centre excess: if the excess is produced by millisecond pulsars, the gamma-ray flux from a small patch of the bulge will occasionally show sharp temporal magnifications, while a smooth dark-matter signal will be temporally flat. The physical basis is the size-dependence of the Einstein magnification: for a point source of radius $r_s$, the maximum magnification is about $2 R_E / r_s$, and with a 10-km pulsar and a stellar lens $R_E \\sim$ au, this reaches about $2.4\\times10^8$ in perfect alignment and about 8700 at the caustic crossing of a typical binary-lens event. On this basis the paper predicts two characteristic signatures—twin caustic-crossing peaks roughly 70 days apart, and single-lens bumps lasting weeks—and simulates Poisson photon arrivals to find the detector sensitivity needed to see them. The required improvement in effective area over Fermi-LAT is a factor of $10^{1.41}$ for a population of 500 bright sources, $10^{2.31}$ for 5000 sources, and $10^{4.26}$ for 100,000 faint sources. None of these thresholds is met by Fermi-LAT or the planned Advanced Particle-physics Telescope, so the authors present the test as a tool for future observations rather than a resolution of the current debate.","pith_inferences":["The same size-discrimination argument could be turned around: even before reaching the detection thresholds, a decade-long gamma-ray lightcurve of the bulge with a future detector could place an upper limit on the compact-source fraction of the GCE from the absence of flicker, making a null result physically informative.","The thresholds are sensitive to the assumed 3.5% caustic-crossing fraction and to the binary fraction of lenses; if the true binary fraction is closer to the 23% found in recent population fits, caustic-crossing events would be several times more common, lowering the sensitivity needed for the bright-population cases.","A matched-filter search over the full lightcurve, rather than a sliding 3-day Poisson window, might recover the faint caustic-crossing peaks at lower sensitivity than the paper's thresholds, since it would use the known twin-peak shape; this is a testable modelling improvement not explored in the paper.","The same method could be applied to unresolved source populations elsewhere in the Galaxy, such as globular clusters or the Galactic plane, where future detectors could use microlensing flicker to measure the compact-object content independently of spectral template fits."],"forward_implications":["If a small population of ~500 bright MSPs produces the excess, a detector with about 26 times Fermi-LAT's effective area could detect caustic-crossing flashes, but such events are expected only once every ~1500 years.","If the population is intermediate (~5000 sources), a ~200x sensitivity detector would see one microlensing flash every ~150 years, while a ~18,000x detector would see several per year for the 100,000-source case—though each flash from faint sources is very hard to catch.","A single clean detection of the predicted twin-peak lightcurve would break the degeneracy between dark matter and pulsar interpretations, and the observed event rate would constrain the number of MSPs contributing to the excess.","Because the test looks at time variability rather than sky templates, it avoids the main systematic that has plagued spatial and spectral studies: mismodelling of the diffuse background and of the source template."],"supporting_citations":[{"why":"introduced gravitational microlensing as a probe of halo dark matter and supplies the optical-depth framework used for event-rate estimates.","marker":"Paczynski, 1986"},{"why":"standard reference for the point-mass magnification formula $\\mu = (y^2+2)/(y\\sqrt{y^2+4})$ used for single-lens lightcurves.","marker":"Schneider et al., 1992"},{"why":"provides the complex binary-lens equation whose caustics produce the twin-peak magnification events at the heart of the detection strategy.","marker":"Witt & Mao, 1995"},{"why":"gives the scaling of maximum magnification with source radius used to estimate ~8700 for MSP caustic crossings.","marker":"Gaudi & Petters, 2002"},{"why":"source-count analysis implying ~500 bright point sources, the basis for the Ns=500 case and its photon rate.","marker":"Lee et al., 2016"},{"why":"source-count distribution integrated to set the Ns=10^5 faint-MSP scenario and its per-source detection rate.","marker":"Gautam et al., 2022"},{"why":"Fermi-LAT measurement of the GCE spectrum and morphology that defines the excess the paper seeks to explain.","marker":"Ajello et al., 2016"},{"why":"shows that smooth emission can be misattributed to point sources, motivating a template-independent temporal test.","marker":"Leane & Slatyer, 2019"},{"why":"specifies the Advanced Particle-physics Telescope sensitivity used as the benchmark for detectability.","marker":"Buckley et al., 2022"}],"fun_headline_variants":["Gamma-ray excess: pulsars flicker, dark matter doesn't","Microlensing test: can it tell pulsars from dark matter?","Flickering pulsars vs steady dark matter: the microlensing clue","Gravitational lenses could expose the true nature of the gamma-ray excess","Why dark matter wouldn't flicker: a new microlensing test"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the pulsar population behind the GCE can be modelled as a set of identical, steady point sources emitting at a single constant photon rate (Section 3, Eq. 11); in reality the source-count distribution is poorly constrained, and a bright tail of sources or intrinsic flaring—which the authors acknowledge as a concern in Section 5—would change both the event rates and the detector sensitivity needed to see the predicted flashes.","fun_headline_variants_meta":{"raw":{"variants":["Gamma-ray excess: pulsars flicker, dark matter doesn't","Microlensing test: can it tell pulsars from dark matter?","Flickering pulsars vs steady dark matter: the microlensing clue","Gravitational lenses could expose the true nature of the gamma-ray excess","Why dark matter wouldn't flicker: a new microlensing test"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001161,"raw_usage":{"total_tokens":4907,"prompt_tokens":1146,"completion_tokens":3761,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":762,"completion_tokens_details":{"reasoning_tokens":3665}},"tokens_in":762,"tokens_out":3761,"duration_ms":24779,"temperature":1.0,"reasoning_tokens":3665,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:51:56.997374+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete check: inject the predicted twin-peak caustic-crossing lightcurve into one resolution element of synthetic Fermi-LAT data with photon counts scaled by $10^{1.41}$ (26x sensitivity) and run the paper's 3-day sliding-window Poisson significance search; if the twin peaks are not recovered with p ≤ 0.01, the detection threshold is too optimistic. On the observational side, a single clean measurement of the ~8700-amplification double peak separated by ~70 days from the Galactic Centre would rule out smooth dark-matter annihilation as the source of that flux, while a decade-long null search with a 26x detector in the 500-source scenario would start to constrain the compact-source fraction.","supporting_citations":[{"cited_title":"1986, ApJ, 304, 1 —","cited_arxiv_id":null,"evidence_quote":"introduced gravitational microlensing as a probe of halo dark matter and supplies the optical-depth framework used for event-rate estimates."},{"cited_title":"J., & Mao, S","cited_arxiv_id":null,"evidence_quote":"provides the complex binary-lens equation whose caustics produce the twin-peak magnification events at the heart of the detection strategy."},{"cited_title":"S., & Petters, A","cited_arxiv_id":null,"evidence_quote":"gives the scaling of maximum magnification with source radius used to estimate ~8700 for MSP caustic crossings."},{"cited_title":"K., Lisanti, M., Safdi, B","cited_arxiv_id":null,"evidence_quote":"source-count analysis implying ~500 bright point sources, the basis for the Ns=500 case and its photon rate."},{"cited_title":"M., Ferrario, L., et al","cited_arxiv_id":null,"evidence_quote":"source-count distribution integrated to set the Ns=10^5 faint-MSP scenario and its per-source detection rate."},{"cited_title":"B., et al","cited_arxiv_id":null,"evidence_quote":"Fermi-LAT measurement of the GCE spectrum and morphology that defines the excess the paper seeks to explain."},{"cited_title":"2022, in 37th International Cosmic Ray Conference, 655","cited_arxiv_id":null,"evidence_quote":"specifies the Advanced Particle-physics Telescope sensitivity used as the benchmark for detectability."}],"review_version":2}