Fermi-LAT 4FGL J1544.2-2554: Black Widow MSP

• Fermi-LAT Source 4FGL J1544.2–2554 is a point-like gamma-ray emitter confirmed as a black widow millisecond pulsar with distinct multiwavelength signals.
• Multiwavelength observations, including gamma-ray spectral curvature, optical light-curve modulation, and radio pulsations, have established key system parameters and orbital dynamics.
• The source serves as a benchmark for population studies, dark matter subhalo searches, and improvements in machine learning-based probabilistic classification in gamma-ray astronomy.

Fermi-LAT Source 4FGL J1544.2–2554 is a point-like gamma-ray emitter identified in the Fermi Large Area Telescope (LAT) Fourth Source Catalog and its subsequent data releases. Located at Galactic coordinates $l\approx344.76^\circ$, $b\approx22.59^\circ$, this source has attracted particular interest due to its multiwavelength phenomenology, its role in population statistics, its association with a compact binary millisecond pulsar, and its candidacy in dark matter subhalo searches. The following sections detail the characterization, cataloging methodology, discovery as a black widow millisecond pulsar, probabilistic classification, dark matter subhalo analyses, clustering context, and astrophysical implications.

1. Fermi-LAT Catalog Characterization

The Fermi-LAT cataloging process assigns precise source positions, fluxes, spectral parameters, and temporal properties to entries like 4FGL J1544.2–2554 (Collaboration, 2019, Ballet et al., 2023). Localization employs maximum-likelihood mapping of the log-likelihood surface, resulting in best-fit coordinates with error ellipses defined by the parameters Conf_95_SemiMajor, Conf_95_SemiMinor, and Conf_95_PosAng. Source fluxes are extracted bandwise across $50\,\mathrm{MeV}$–$1\,\mathrm{TeV}$ via weighted likelihood analysis, accounting for systematic uncertainties in the diffuse background.

Spectral modeling utilizes either a Power Law ($dN/dE=K\,(E/E_0)^{-\Gamma}$), a LogParabola [$dN/dE=K\,(E/E_0)^{-\alpha-\beta\,\ln(E/E_0)}$], or a Power Law with subexponential cutoff. The preferred model for a given source is chosen based on curvature significance (e.g., TS_curv, LP_SigCurv) and likelihood maximization. The catalog provides both total energy/phton fluxes and spectrally resolved energy distributions (SEDs) across eight or more energy bands.

Light curves capture temporal variability using one-year or two-month binning (DR3, DR4). Variability indices (TS_var) are computed to quantify non-steady emission, with transient detection frameworks able to flag sources that display notable emission only in limited intervals (Ballet et al., 2023).

The catalog’s association and identification process integrates Bayesian and likelihood-ratio cross-matches against multiwavelength reference catalogs (pulsars, AGNs including BLLs and FSRQs, SNRs, GCs, TeV sources), revising associations as new data become available and flagging sources that may still lack a robust counterpart.

2. Multiwavelength Identification: Black Widow Millisecond Pulsar Discovery

4FGL J1544.2–2554 has been robustly established as a black widow millisecond pulsar (BW MSP), PSR J1544–2555, through coordinated multiwavelength campaigns (Karpova et al., 25 Nov 2024, Diaz et al., 11 Sep 2025). Its confirmation follows gamma-ray identification (low variability, spectral curvature), optical ephemeris (ULTRACAM, $P_\mathrm{orb} = 2.7\,$hr), radio detection (MeerKAT, Parkes, Effelsberg, Nançay) of $2.4\,$ms pulsations synchronized to the binary orbit, and a timing solution spanning 16 years of LAT data.

The optical light curve, exhibiting a periodicity of 2.7 hours and amplitude $>2.5$ mag, is well modeled by direct irradiation (and later, spot asymmetry models using Icarus): inclination $\approx65^\circ$–$83^\circ$, companion mass $0.095$–$0.102\,M_\odot$, neutron star mass up to $1.96\,M_\odot$, “day-side” temperature $\sim6100$–$7200$ K, “night-side” $\sim2300$–$3000$ K, Roche-lobe filling factor $0.65$, and a system distance $2.1$ kpc (Karpova et al., 25 Nov 2024). X-ray eROSITA detections provide evidence for nonthermal emission, likely synchtrotron from an intra-binary shock. Notably, blue color near minima in the optical light curve further supports this.

The joint radio/gamma-ray/optical timing solution resolves orbital period variations and yields tightly phased gamma-ray pulsations (H-test up to 627.8) (Diaz et al., 11 Sep 2025). The model utilizes two-component sinusoids for the optical modulations:

$$A_0 + A_1\sin\left(\frac{2\pi(t-T_\mathrm{asc})}{P_\mathrm{orb}}+\pi\right) + A_2\cos\left(\frac{4\pi(t-T_\mathrm{asc})}{P_\mathrm{orb}}\right)$$

3. Probabilistic Source Classification and Catalog Context

Probabilistic classification frameworks leverage ML to assign class probabilities across the major gamma-ray source populations---pulsar, AGN, and "OTHER" (Bhat et al., 2021). For 4FGL J1544.2–2554, ensemble ML algorithms (RF, BDT, LR, NN) using 16 catalog features (spectral indices, curvature, SED parameters, hardness ratios, etc.) compute per-class likelihoods. High-confidence classification emerges when the sum of individual model probabilities exceeds empirically determined thresholds (e.g., sum $>7$ out of eight models). The three-class system improves reliability by explicitly modeling the "OTHER" category; this suppresses bias in population estimation (for instance, supernova remnants, globular clusters, and star-forming regions are not forcibly assigned to AGN/pulsar classes).

Testing accuracy in the two-class system approaches 97–98%, with three-class accuracy about 92–94%. For this source---now confirmed as a pulsar---ML probabilistic cataloging yields high pulsar probability, consistent with its observational assignment.

Neural simulation-based inference (SBI) methods (DeepSphere U-net, neural ratio estimation) achieve near-complete recovery of 4FGL sources with $$S>3\times10^{-10}\,\mathrm{cm}^{-2}\,\mathrm{s}^{-1}$$ (Eckner et al., 5 May 2025), efficiently anchoring the source-count distribution ($dN/dS$) at high Galactic latitudes and demonstrating that cataloged sources like 4FGL J1544.2–2554 comprise essentially all significant contributors above the completeness limit.

4. Extended Source and Clustering Context

Clustering studies using DBSCAN with $\varepsilon = 0.005$ radians and MinPts = 2 have investigated whether unidentified point sources might actually arise from extended emission regions (Cozzolongo et al., 4 Apr 2025). For 4FGL J1544.2–2554, explicit DBSCAN clustering analysis does not indicate spatial association with other sources in the tested radius, nor does extended source modeling yield statistically significant preference (no TS_ext $>$ 16). Thus, the source is best characterized as a spatially-isolated point source, distinct from extended gamma-ray structures such as those analyzed for HESS J1813–178. No multiwavelength evidence of extended TeV emission is reported for this source.

5. Dark Matter Subhalo Candidate Searches

Several studies have assessed 4FGL J1544.2–2554 as a possible dark matter (DM) subhalo candidate based on gamma-ray spectral fits and ML/Bayesian inference (Butter et al., 2023, Cheng et al., 25 Dec 2024). Bayesian neural networks trained on realistic DM subhalo spectra (CLUMPY, N-body simulation-based $J$-factors, $dN/dE$ via PLSuperExpCutoff) compare the measured SED (eight energy bins, $50\,\mathrm{MeV}$–$1\,\mathrm{TeV}$) against DM model predictions. Candidate selection applies the $\mu-\sigma \geq t$ rule for network output ($\mu=$ mean, $\sigma=$ uncertainty, $t=$ threshold, typically 0.5–0.8).

Unbinned maximum likelihood analyses (DMfitFunction, PPP4DMID for $b\overline{b}$ yield) find that for 4FGL J1544.2–2554, a DM model sometimes attains higher likelihood than empirical fits, with best-fit DM particle masses in the $30$–$500\,\mathrm{GeV}$ range and $J$-factors up to $5.8\times10^{20}\,\mathrm{GeV}^2\,\mathrm{cm}^{-5}$ (Cheng et al., 25 Dec 2024). However, the intrinsic degeneracy between DM spectra and pulsar exponential cutoff power law (ECPL) emissions implies that further multiwavelength confirmation (especially radio and optical) is required for robust DM subhalo assignment. Contemporary results now favor the MSP binary interpretation for this source.

6. Galaxy Merger Associations and Population Implications

A spatial cross-match using SDSS galaxy merger catalogs (CNN classifier $p_m>0.95$), with a conservative $4\sigma$ error ellipse and a Poisson-based random association threshold, identifies numerous Fermi-LAT matches (Manna et al., 5 Jul 2025). Five robust gamma-ray associations lack catalog classification; a plausible implication is that sources like 4FGL J1544.2–2554, were they unclassified, could be linked to merger-induced phenomena (AGN activation, cosmic ray acceleration in shocks, starburst activity). Confirmed pulsar status, however, precludes this scenario for J1544.2–2554 at present.

7. Astrophysical Insights and Future Directions

The confirmation of 4FGL J1544.2–2554 as the black widow MSP PSR J1544–2555 provides key insights into binary evolution, the population statistics of spider pulsars, and the physical processes governing wind irradiation and ablation in tight binaries. The system’s companion mass ($\sim$0.10 $M_\odot$) positions it near the gap between BW and RB companion masses, relevant for evolutionary studies (Karpova et al., 25 Nov 2024). Optical, radio, gamma-ray, and X-ray data enable detailed studies of wind-irradiated companion heating, intra-binary shocks, and the impact of orbital period variations.

Further research directions include refinement of the neutron star mass via deeper optical spectroscopy and radial velocity curves, improved thermal modeling of the companion atmosphere, detailed phase-resolved X-ray spectroscopy, and continued monitoring for orbital period variations or state transitions. The system exemplifies the utility of Fermi-LAT catalogs as both a source discovery tool and a cross-band population resource (Diaz et al., 11 Sep 2025).

For dark matter studies, the degeneracy between pulsar and DM subhalo spectra in gamma-rays underscores the necessity for joint multiwavelength campaigns and improved modeling of spectral shape indicators. Population synthesis using neural simulation-based inference validates catalog completeness for resolved high-latitude sources above the well-established flux threshold, with 4FGL J1544.2–2554 serving as an important anchor.

---

In summary, Fermi-LAT Source 4FGL J1544.2–2554 is now established as a black widow millisecond pulsar in a compact binary, with well-characterized multiwavelength properties and a key role in catalog validation and population analysis. Despite previous candidacy as a dark matter subhalo, current evidence solidly favors the spider pulsar interpretation, with ongoing studies further illuminating the system’s evolutionary and physical nature.