juliensimon/fermi-3fhl-hard-gamma-ray

Fermi LAT Third Catalog of Hard Sources (3FHL)

Credit: NASA/DOE/Fermi LAT Collaboration

Part of a dataset collection on Hugging Face.

Dataset description

The 3FHL catalog (Ajello et al. 2017) contains sources detected by Fermi LAT in the 10 GeV - 2 TeV energy range using 7 years of Pass 8 data. This catalog bridges the gap between the GeV regime covered by the standard Fermi catalogs and the TeV regime covered by ground-based Cherenkov telescopes (H.E.S.S., MAGIC, VERITAS).

Sources include blazars, pulsar wind nebulae, supernova remnants, and unidentified gamma-ray emitters. The catalog is essential for planning observations with current and future TeV observatories like CTA.

The 10 GeV - 2 TeV energy range probed by 3FHL occupies a critical frontier in high-energy astrophysics. Below ~10 GeV, the Fermi LAT standard catalogs (3FGL, 4FGL) provide comprehensive coverage with large photon statistics. Above ~100 GeV, imaging atmospheric Cherenkov telescopes (IACTs) like H.E.S.S., MAGIC, and VERITAS achieve superior sensitivity but with limited fields of view and duty cycles. The 3FHL catalog bridges this gap using Fermi LAT's Pass 8 event reconstruction, which dramatically improved the instrument's effective area and point-spread function at high energies, enabling detection of hard-spectrum sources that were previously buried in background.

The source population in 3FHL is dominated by blazars — active galactic nuclei whose relativistic jets point close to our line of sight, producing Doppler-boosted emission that peaks in the GeV-TeV band. The catalog also contains Galactic sources such as pulsar wind nebulae (where ultra-relativistic electron-positron winds from young pulsars produce inverse-Compton emission), supernova remnants (candidate sites of cosmic-ray acceleration up to PeV energies), and a significant fraction of unidentified sources that may represent entirely new source classes. For extragalactic sources, the interaction of TeV photons with the extragalactic background light (EBL) via pair production imprints a redshift-dependent spectral cutoff, making this catalog a powerful probe of the EBL intensity and its evolution.

The 3FHL is the primary seed catalog for the Cherenkov Telescope Array (CTA), the next-generation ground-based gamma-ray observatory. Nearly every 3FHL source above the CTA sensitivity threshold is expected to be detected, and the catalog's uniform sky coverage helps define CTA's key science programs including surveys of the Galactic plane and extragalactic deep fields.

This dataset is suitable for tabular classification tasks.

Schema

Column	Type	Description	Sample	Null %
name	str	Official 3FHL source designation, e.g. '3FHL J0001.2-0748'; encodes the catalog prefix and J2000 coordinates of the source centroid; uniquely identifies each entry in the catalog	3FHL J0534.5+2201	0.1%
ra	float64	Right ascension of the source in degrees (J2000, ICRS); range 0-360; derived from a maximum-likelihood fit to the >10 GeV photon distribution	83.6348	0.1%
dec	float64	Declination of the source in degrees (J2000, ICRS); range -90 to +90; derived from a maximum-likelihood fit to the >10 GeV photon distribution	22.0192	0.1%
lii	float64	Galactic longitude in degrees; range 0-360; derived from ra and dec via the standard IAU coordinate transformation; useful for studying Galactic plane sources	184.5543	0.1%
bii	float64	Galactic latitude in degrees; range -90 to +90; sources near bii=0 lie in the Galactic plane where diffuse emission complicates detection	-5.7805	0.1%
error_radius	float64	95% confidence radius on the source position in degrees; smaller values indicate better localization; typically 0.01-0.1 deg for bright sources	0.008	3.2%
roi	float64	Region of interest number used in the analysis; the sky was divided into overlapping ROIs for the maximum-likelihood source detection	430.0	0.1%
significance	float64	Detection significance in the 10 GeV - 2 TeV band expressed as the square root of the test statistic (sqrt(TS)); higher values indicate more confident detections; catalog threshold is sqrt(TS) >= 4	168.641	0.1%
pivot_energy	float64	Pivot (decorrelation) energy in MeV at which the flux uncertainty is minimized; chosen to reduce correlation between spectral index and flux normalization	22.73	0.1%
flux_density	float64	Differential photon flux at the pivot energy in units of ph/cm2/s/MeV; the spectral normalization parameter used in the likelihood fit	1.7132e-10	0.1%
flux_density_error	float64	1-sigma statistical uncertainty on flux_density in ph/cm2/s/MeV	3.3886e-12	0.1%
flux	float64	Integrated photon flux in the 10 GeV - 2 TeV band in ph/cm2/s; obtained by integrating the best-fit spectral model over the full energy range	8.6589e-09	0.1%
flux_error	float64	1-sigma statistical uncertainty on the integrated photon flux in ph/cm2/s	1.7127e-10	0.1%
energy_flux	float64	Integrated energy flux in the 10 GeV - 2 TeV band in erg/cm2/s; accounts for the energy weighting of the spectrum; more physically meaningful than photon flux for comparing source luminosities	4.9149e-10	0.1%
energy_flux_error	float64	1-sigma statistical uncertainty on the energy flux in erg/cm2/s	1.6631e-11	0.1%
curve_significance	float64	Significance of spectral curvature with respect to a simple power-law model; high values (>3) indicate that a curved spectrum (log-parabola or exponential cutoff) is preferred over a power law	1.43	0.1%
spectrum_type	str	Best-fit spectral model type: 'PowerLaw', 'LogParabola', or 'PLExpCutoff'; selected based on the curve_significance test	PowerLaw	0.1%
spectral_index	float64	Photon spectral index of the best-fit model; for a power law dN/dE ~ E^(-index), typical values are 1.5-4; harder (lower index) sources emit relatively more high-energy photons	2.2853	0.1%
spectral_index_error	float64	1-sigma statistical uncertainty on the spectral index	0.0378	0.1%
beta	float64	Curvature parameter for LogParabola spectra: dN/dE ~ (E/E0)^(-(alpha + beta*ln(E/E0))); zero for pure power-law sources; larger beta indicates stronger curvature	-0.0469	0.1%
beta_error	float64	1-sigma statistical uncertainty on the beta curvature parameter; null for sources fit with a simple power law	0.0181	0.1%
powerlaw_index	float64	Photon index when the source is fit with a simple power law regardless of whether curvature is significant; useful for uniform comparisons across all sources	2.2202	0.1%
powerlaw_index_error	float64	1-sigma uncertainty on the power-law photon index	0.025	0.1%
flux_10_20_gev	float64	Photon flux in the 10-20 GeV sub-band in ph/cm2/s; first of five logarithmically spaced energy bins used to construct the broadband spectral energy distribution	5.1699e-09	0.1%
flux_10_20_gev_neg_err	float64	Lower (negative) 1-sigma error on the 10-20 GeV photon flux in ph/cm2/s; asymmetric errors reflect the Poisson nature of photon counting	-1.3339e-10	1.3%
flux_10_20_gev_pos_err	float64	Upper (positive) 1-sigma error on the 10-20 GeV photon flux in ph/cm2/s	1.3339e-10	0.1%
nufnu_10_20_gev	float64	Energy flux (nu*F_nu) in the 10-20 GeV band in erg/cm2/s; represents the spectral energy distribution value at the geometric mean energy of the bin	1.642e-10	0.1%
sqrt_ts_10_20_gev	float64	Detection significance (sqrt of test statistic) in the 10-20 GeV band; values below ~2 indicate the source is not significantly detected in this sub-band	125.1574	0.1%
flux_20_50_gev	float64	Photon flux in the 20-50 GeV sub-band in ph/cm2/s	2.245e-09	0.1%
flux_20_50_gev_neg_err	float64	Lower 1-sigma error on the 20-50 GeV photon flux in ph/cm2/s	-8.6719e-11	7.2%
flux_20_50_gev_pos_err	float64	Upper 1-sigma error on the 20-50 GeV photon flux in ph/cm2/s	8.6719e-11	0.1%
nufnu_20_50_gev	float64	Energy flux (nu*F_nu) in the 20-50 GeV band in erg/cm2/s	1.1808e-10	0.1%
sqrt_ts_20_50_gev	float64	Detection significance (sqrt(TS)) in the 20-50 GeV band	88.7153	0.1%
flux_50_150_gev	float64	Photon flux in the 50-150 GeV sub-band in ph/cm2/s	9.2432e-10	0.1%
flux_50_150_gev_neg_err	float64	Lower 1-sigma error on the 50-150 GeV photon flux in ph/cm2/s	-5.4975e-11	36.3%
flux_50_150_gev_pos_err	float64	Upper 1-sigma error on the 50-150 GeV photon flux in ph/cm2/s	5.4975e-11	0.1%
nufnu_50_150_gev	float64	Energy flux (nu*F_nu) in the 50-150 GeV band in erg/cm2/s	1.0869e-10	0.1%
sqrt_ts_50_150_gev	float64	Detection significance (sqrt(TS)) in the 50-150 GeV band	59.0875	0.1%
flux_150_500_gev	float64	Photon flux in the 150-500 GeV sub-band in ph/cm2/s	2.759e-10	0.1%
flux_150_500_gev_neg_err	float64	Lower 1-sigma error on the 150-500 GeV photon flux in ph/cm2/s	-2.9165e-11	77.8%
flux_150_500_gev_pos_err	float64	Upper 1-sigma error on the 150-500 GeV photon flux in ph/cm2/s	3.136e-11	0.1%
nufnu_150_500_gev	float64	Energy flux (nu*F_nu) in the 150-500 GeV band in erg/cm2/s	9.2302e-11	0.1%
sqrt_ts_150_500_gev	float64	Detection significance (sqrt(TS)) in the 150-500 GeV band	33.0762	0.1%
flux_0p5_2_tev	float64	Photon flux in the 0.5-2 TeV sub-band in ph/cm2/s; the highest energy bin, probing the TeV regime where ground-based Cherenkov telescopes provide complementary coverage	6.684e-11	0.1%
flux_0p5_2_tev_neg_err	float64	Lower 1-sigma error on the 0.5-2 TeV photon flux in ph/cm2/s	-1.463e-11	93.8%
flux_0p5_2_tev_pos_err	float64	Upper 1-sigma error on the 0.5-2 TeV photon flux in ph/cm2/s	1.6916e-11	0.1%
nufnu_0p5_2_tev	float64	Energy flux (nu*F_nu) in the 0.5-2 TeV band in erg/cm2/s	6.9012e-11	0.1%
sqrt_ts_0p5_2_tev	float64	Detection significance (sqrt(TS)) in the 0.5-2 TeV band	15.5731	0.1%
npred	float64	Total number of predicted photons from this source in the model; a measure of signal strength that accounts for exposure and PSF; higher npred means more photons attributed to this source	2602.9	0.1%
hep_energy	float64	Energy of the highest-energy photon associated with the source in GeV; constrained by the LAT effective area and source spectrum; can exceed 1 TeV for the hardest sources	1463.3	0.1%
hep_prob	float64	Probability that the highest-energy photon truly belongs to this source rather than a neighboring source or diffuse background; values near 1.0 give high confidence in the association	1.0	0.1%
num_bayesian_blocks	float64	Number of Bayesian blocks in the source light curve; a value of 1 indicates no significant variability detected; higher values suggest flux changes over the 7-year observation period	1.0	0.1%
extended_source_name	str	Name of the spatial template if the source is modeled as spatially extended (e.g. supernova remnants, pulsar wind nebulae); null for point sources which are the majority of the catalog	IC 443	96.9%
alt_gammaray_name	str	Alternative gamma-ray catalog designation from earlier Fermi catalogs (e.g. 3FGL, 2FHL) or other experiments; useful for cross-matching with the broader gamma-ray literature	3FGL J0534.5+2201i	13.5%
tev_assoc_flag	str	Flag indicating association with a known TeV source from ground-based observations; non-null values indicate overlap with sources detected by H.E.S.S., MAGIC, or VERITAS	P	0.1%
assoc_tevcat	str	Name of the associated TeV source in TeVCat; provides the link between Fermi LAT and ground-based Cherenkov telescope detections of the same astrophysical object	Crab Nebula	91.5%
source_class	str	Astrophysical classification of the source (e.g. 'bll' = BL Lac blazar, 'fsrq' = flat-spectrum radio quasar, 'psr' = pulsar, 'snr' = supernova remnant, 'pwn' = pulsar wind nebula, '' = unassociated)	PWN	11.5%
assoc_name_1	str	Primary counterpart name at other wavelengths from automated association methods; typically a radio or X-ray catalog designation used to identify the gamma-ray source	Crab Nebula	11.5%
assoc_name_2	str	Secondary counterpart name at other wavelengths; provides an alternative identification when multiple association methods yield different but plausible counterparts	Vela X field	98.3%
assoc_prob_bay	float64	Bayesian association probability for the primary counterpart; range 0-1; values above 0.8 indicate a highly confident multi-wavelength association	1.0	3.2%
assoc_prob_lr	float64	Likelihood-ratio association probability for the primary counterpart; range 0-1; an independent association metric complementing the Bayesian probability	0.9999	3.2%
redshift	float64	Spectroscopic redshift of the associated extragalactic counterpart; null for Galactic sources and unassociated sources; used to compute intrinsic luminosity and EBL absorption corrections	0.03	64.8%
nupeak_obs	float64	Observed synchrotron peak frequency in Hz (log10 scale) for blazar-type sources; classifies blazars as low-synchrotron-peaked (LSP, <10^14 Hz), intermediate (ISP), or high-synchrotron-peaked (HSP, >10^15 Hz); null for non-blazar sources	1.1482e+17	29.6%

Quick stats

• 1,558 gamma-ray sources above 10 GeV
• 1,212 blazar-type sources (BL Lac, FSRQ, BCU)
• 548 with measured redshift
• Median detection significance: 7.0 sigma

Usage

from datasets import load_dataset

ds = load_dataset("juliensimon/fermi-3fhl-hard-gamma-ray", split="train")
df = ds.to_pandas()

from datasets import load_dataset
import matplotlib.pyplot as plt

ds = load_dataset("juliensimon/fermi-3fhl-hard-gamma-ray", split="train")
df = ds.to_pandas()

# Highest significance sources
print(df.sort_values("significance", ascending=False).head(10)[["name", "ra", "dec", "significance"]])

# Sky map in Galactic coordinates
fig, ax = plt.subplots(figsize=(12, 6), subplot_kw={"projection": "aitoff"})
import numpy as np
l = np.radians(df["lii"] - 180)  # center on Galactic center
b = np.radians(df["bii"])
ax.scatter(l, b, s=df["significance"] * 0.5, alpha=0.5, c=df["energy_flux"],
           cmap="inferno", norm=plt.matplotlib.colors.LogNorm())
ax.set_title("Fermi 3FHL Sources (>10 GeV) — Galactic Coordinates")
ax.grid(True)
plt.tight_layout()
plt.show()

# Spectral index distribution
plt.figure(figsize=(8, 5))
plt.hist(df["spectral_index"].dropna(), bins=40, edgecolor="black", alpha=0.7)
plt.xlabel("Photon Spectral Index")
plt.ylabel("Number of Sources")
plt.title("3FHL Spectral Index Distribution")
plt.tight_layout()
plt.show()

Data source

https://heasarc.gsfc.nasa.gov/W3Browse/fermi/fermi3fhl.html

Related datasets

• juliensimon/gamma-ray-bursts

• juliensimon/pulsar-catalog

If you find this dataset useful, please consider giving it a like on Hugging Face. It helps others discover it.

About the author

Created by Julien Simon — AI Operating Partner at Fortino Capital. Part of the Space Datasets collection.

Citation

@dataset{fermi_3fhl_hard_gamma_ray,
  title = {Fermi LAT Third Catalog of Hard Sources (3FHL)},
  author = {juliensimon},
  year = {2026},
  url = {https://huggingface.co/datasets/juliensimon/fermi-3fhl-hard-gamma-ray},
  publisher = {Hugging Face}
}

License

CC-BY-4.0