WorksPhysics and Astronomy
X-ray Emission Study of TeV Blazars
Abstract
In this work, we present a systematic X-ray study of 85 TeV-detected blazars based on archival Swift-XRT and NuSTAR observations available up to May 2024. We derive unabsorbed X-ray fluxes and photon indices through spectral fitting, and investigate their light curves, spectral energy distributions, flux distributions, power spectral densities, and flux-index correlations. Using the fractional variability method, we identify 78 sources as variable in the soft X-ray band and 14 sources as variable in the hard X-ray band. For well-sampled soft X-ray light curves, the flux distributions are generally better described by log-normal functions, while the PSD slopes span –, with most values between 1.0 and 1.6, where the PSD is defined as . We find positive correlations between TeV and X-ray luminosities, supporting a leptonic emission scenario. We also find a correlation between X-ray and disk luminosities, whereas no significant correlation is found between X-ray luminosity and black hole mass.
Introduction
Blazars are a subclass of active galactic nuclei (AGNs) characterized by relativistic jets oriented at small angles with respect to the line of sight, resulting in extreme observational properties such as strong Doppler boosting, rapid variability, and high luminosities across the electromagnetic spectrum (e.g., (Urry & Padovani 1995; Hovatta & Lindfors 2019; Raiteri 2025). Based on the equivalent width (EW) of their optical emission lines, blazars are conventionally divided into two subclasses: BL Lacertae objects (BL Lacs) and flat-spectrum radio quasars (FSRQs). BL Lacs exhibit weak or absent emission lines (EW Å), whereas FSRQs show prominent broad emission lines with EW Å ((Stickel et al. 1991; Urry & Padovani 1995; Scarpa & Falomo 1997).
The broadband spectral energy distributions (SEDs) of blazars are characterized by two distinct components. The low-energy component, extending from radio to ultraviolet and in some cases X-ray bands, is widely attributed to synchrotron radiation from relativistic electrons in the jet. According to the synchrotron peak frequency, blazars are further classified as low-, intermediate-, and high-synchrotron-peaked sources, hereafter LSPs, ISPs, and HSPs, respectively (e.g., [1] (Fan et al. 2016; Yang et al. 2022). In leptonic scenarios, the high-energy component, peaking at X-ray to -ray energies, is generally interpreted as inverse Compton (IC) scattering of seed photons by the same electron population. This component can be produced through synchrotron self-Compton (SSC) emission when the seed photons are the synchrotron photons, or through external Compton (EC) emission when the seed photons originate outside the jet. Possible external photon fields include the accretion disk ((Dermer & Schlickeiser 1993),
Corresponding author: Hubing Xiao, Junhui Fan the broad-line region (BLR; Sikora et al. 1994; Fan et al. 2006), and the dusty torus (DT; [5] (Błażejowski et al. 2000; Sokolov & Marscher 2005)). Alternatively, the high-energy component can be attributed to proton synchrotron radiation and proton-induced secondary-particle cascades (e.g., [14] (Mücke & Protheroe 2001; Dimitrakoudis et al. 2012; Zheng & Kang 2013; Diltz et al. 2015; Cerruti et al. 2019; Ouyang et al. 2025; Xiao et al. 2026)).
The TeVCat 2.0 catalog2 contains 303 sources associated with TeV emission, among which 84 are confirmed as blazars. For FSRQs, TeV emission is often interpreted in terms of the EC process, with a substantial contribution from DT seed photons (Xiao et al. 2024). For BL Lacs, the TeV emission is commonly interpreted as SSC radiation. In such a scenario, X-ray emission takes the role of a seed photon to be scattered into the TeV band, and a consequent strong correlation between X-ray and TeV emission is expected. In hadronic scenarios, by contrast, TeV photons can be produced through proton synchrotron radiation, neutral pion decay, or cascade emission from secondary particles ([14] (Dermer et al. 2009; Xiao et al. 2026)). The TeV radiation may then depend on the proton distribution, magnetic field, target photon density, and interaction efficiency, leading to a looser correlation between X-ray and TeV emission.
TeV blazars have been widely investigated in previous works. For instance, Zhang et al. (2022) investigated the composition of TeV emission in blazars and separated it into the beamed and un-beamed emission; Liang et al. (2023) examined the differences between TeV and non-TeV BL Lacs in the mid-infrared band, suggesting that TeV BL Lacs are brighter and have a flatter spectrum than do the non-TeV BL Lacs in the mid-IR band. Ouyang et al. (2023) explored the blazar sequence in the TeV band and suggested that the blazar sequence in the TeV band could be explained by the Doppler beaming effect. Wierzcholska & Wagner (2016) analyzed the X-ray spectral properties of a blazar sample using data from the Neil Gehrels Swift Gamma-Ray Burst Observatory (Swift), accounting for different neutral hydrogen absorption corrections and separating synchrotron and IC components. Bhatta et al. [9] investigated the hard X-ray properties of blazars observed with the Nuclear Spectroscopic Telescope Array (NuSTAR), focusing on timing, spectral, and cross-correlation characteristics. Furthermore, comprehensive X-ray catalogs of blazars have been constructed, including a soft X-ray catalog by Giommi et al. (2021) and a hard X-ray catalog by Middei et al. (2022), which together provide valuable constraints on the spectral and variability behavior of these sources.
In this work, we present a systematic X-ray analysis of TeV blazars, using Swift and NuSTAR observations and aiming to characterize their X-ray spectral and variability properties and to explore the underlying physical processes responsible for their high-energy emission. The sample selection and X-ray data analysis are described in § 2. The results are presented in § 3. We discuss the implications of our findings in § 4, and our main conclusions are summarized in § 5. Throughout this work, we adopt the CDM cosmological parameters from Planck Collaboration et al. (2020), with km s Mpc and .
Data analysis
Sample description
Our sample is drawn primarily from the TeVCat catalog, and the source classifications are adopted from the Fermi-LAT 14-year Source Catalog (4FGL-DR4; [2]). It consists of 69 BL Lacs, 10 FSRQs, and five blazar candidates of uncertain type (BCUs). In addition, we include one BL Lac object, RGB J2042.1+2426 (4FGL J2042.1+2427), which is not listed in TeVCat but has been detected at TeV energies by the MAGIC telescope ([3]). In total, our sample comprises 85 TeV-detected blazars, including 70 BL Lacs, 10 FSRQs, and five BCUs, which are listed in Table 1.
| TeV Name (1) | 4FGL Name (2) | (3) | (4) | (5) | Index (6) | (7) | (8) | (9) | Index (10) | (11) | (12) |
| J0013-188 | J0013.9-1854 | 0.09485 | 10 | – | – | – | – | ||||
| J0033-193 | J0033.5-1921 | 0.61 | 6 | – | – | – | – | ||||
| J0035+598 | J0035.9+5950 | 0.086 | 157 | 1 | – | ||||||
| J0112+227 | J0112.1+2245 | 0.265 | 20 | 2 | N/A | ||||||
| J0136+391 | J0136.5+3906 | 0.75 | 10 | – | – | – | – | ||||
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
Table 1. Sample of 85 TeV blazars
All X-ray data for the sources in our sample, available up to May 2024, were retrieved from the public archive of the High Energy Astrophysics Archive Research Center (HEASARC)3.
Observation and data reduction
All data reduction and analysis were performed using the HEASoft software package (v6.36) provided by HEASARC. The procedures adopted for each X-ray instrument are described below.
Swift observations
The Neil Gehrels Swift Observatory (Swift; Gehrels et al. 2004) was launched in 2004 and carries three instruments: the Ultraviolet and Optical Telescope (UVOT; Roming et al. 2005), the X-Ray Telescope (XRT; Burrows et al. [15]), and the Burst Alert Telescope (BAT; Barthelmy et al. [7]).
The Swift-XRT observations were obtained in Photon Counting (PC) and Windowed Timing (WT) modes. XRT spectra were generated using the Swift-XRT data products provided by the UK Swift Science Data Centre at the University of Leicester (Evans et al. 2009). Events with grades 0–12 for PC mode and 0–2 for WT mode were selected. Pile-up effects become significant when the count rate exceeds 0.5 counts in PC mode or 100 counts in WT mode, and pile-up corrections were applied following the standard procedures described by the UK Swift Science Data Centre. The spectra were grouped using the grppha tool to ensure at least one count per bin, allowing the use of Cash statistics (Cash [16]). Only observations with exposure times greater than 0.3 ks and data sets with total counts exceeding 60 were included in the analysis. In addition, WT-mode observations with count rates greater than 0.4 counts were retained. Before applying the selection criteria, the data contained 12381 Swift-XRT spectra, including 6491 PC-mode and 5890 WT-mode spectra. After filtering, 9833 spectra were retained for further analysis, of which 5694 were obtained in PC mode and 4139 in WT mode.
NuSTAR observations
The Nuclear Spectroscopic Telescope Array (NuSTAR; Harrison et al. 2013), launched in 2012, operates in the hard X-ray band from 3 to 79 keV and consists of two co-aligned Focal Plane Detector Modules (FPMA and FPMB). The raw data were processed using the NuSTAR Data Analysis Software (NuSTARDAS, v2.1.2) with calibration files from CALDB (version 20231017). Cleaned event files were produced using the nupipeline task. Source spectra were extracted from circular regions centered on the source, while background spectra were extracted from surrounding annular regions, with extraction radii chosen depending on the source brightness. We restricted the spectral analysis to the 3–60 keV energy range, where the instrumental response is optimal, and background contamination is minimized. The spectra were generated using the nuproducts task and grouped to at least one count per bin to allow for Cash fitting. Only observations with exposure times greater than 1.0 ks were included. A total of 146 NuSTAR spectra were initially available. After selection, 140 spectra were retained for further analysis.
Spectral fitting
Spectral fitting was performed using the XSPEC package (Arnaud [6]). Solar abundances were set to wilm (Wilms et al. 2000), and photoelectric absorption cross-sections were adopted from vern (Verner et al. 1996). Each spectrum was fitted using two models: a power-law and a log-parabolic model, both multiplied by the Galactic absorption model tbabs (Wilms et al. 2000). The Galactic total hydrogen column density was fixed to the value reported by H4PI Collaboration et al. (2016), as shown in Table 1.
The power-law model is defined as
where is the normalization and is the photon index. The log-parabolic model is given by
where is the pivot energy, fixed at 1 keV for Swift-XRT and at 5 keV for NuSTAR spectra, is the photon index at , and is the curvature parameter. For NuSTAR spectra, a cross-calibration constant was included for FPMB relative to FPMA, yielding values within a few percent in all cases, consistent with previous calibration studies (Madsen et al. 2015). Parameter uncertainties were estimated at the 68.3% confidence level, corresponding to 1 for one parameter of interest.
The power-law model is nested within the log-parabolic model, with the latter including one additional free parameter, the curvature parameter . We used the F-test to assess whether the improvement in the fit obtained by introducing the curvature parameter was statistically significant (e.g., Bevington & Robinson [8]). We adopted the log-parabolic model only when the F-test probability was below 0.01, corresponding to a 99% confidence level. Otherwise, the power-law model was selected.
Finally, for each source we derived the unabsorbed integrated fluxes in the 0.3–10 keV band for Swift-XRT and in the 3–60 keV band for NuSTAR, along with the corresponding best-fitting spectral parameters. For each source, we compiled a complete X-ray data table containing all Swift-XRT and NuSTAR observations used in this work, which is shown in Appendix § 7.
Results
Long-term light curves
For each TeV blazar in our sample, we derived the unabsorbed X-ray fluxes integrated over 0.3–10 keV for Swift-XRT and over 3–60 keV for NuSTAR observations, together with the corresponding photon indices from the best-fitting spectral models. The resulting long-term light curves are shown in Figure 1. For convenience, we combine the Swift-XRT measurements (PC and WT modes) and refer to them as soft X-ray observations, while NuSTAR measurements are referred to as hard X-ray observations. For each source, we computed the arithmetic mean flux and photon index separately for the soft and hard X-ray observations; these values are listed in Table 1.
Figure 1. Long-term X-ray light curves of TeV blazars. The black open circles denote Swift-XRT observations and the red open squares denote NuSTAR observations. For each source, the upper panel shows the unabsorbed flux and the lower panel shows the photon index. (Only four representative sources are shown as examples.)
Spectral energy distributions
Using the best-fitting spectral models obtained in § 2, we constructed X-ray spectral energy distributions (SEDs) for each TeV blazar to examine spectral variability and evolution. The X-ray SEDs are shown in Figure 2. The spectra are corrected for Galactic absorption in a manner consistent with our spectral fitting using the cross-sections from Morrison & McCammon (1983) and the Galactic hydrogen column densities listed in Table 1. For comparison, archival broadband SED data were retrieved from the Space Science Data Center (SSDC)4. For clarity, only FPMA spectra are shown for NuSTAR observations.
Figure 2. Spectral energy distributions of TeV blazars. The black-filled circles denote Swift-XRT spectra, the red-filled squares denote NuSTAR (FPMA) spectra, and the gray-filled circles denote archival data compiled by SSDC. (Only four representative sources are shown as examples.)
Flux distribution of TeV blazars
The X-ray flux distribution provides a useful statistical probe of the variability processes in blazars. However, the number of available X-ray observations varies significantly across the sources in our sample, and many TeV blazars are only sparsely observed. To ensure meaningful statistical characterization of flux distributions, we restrict the following analysis to sources with sufficiently well-sampled soft X-ray light curves, while still describing variability properties for the full sample elsewhere in this work. Specifically, we consider sources with more than 30 soft X-ray observations to have adequate sampling for constructing flux histograms and fitting probability density functions (PDFs). For these well-sampled sources, we constructed histograms of the soft X-ray flux measurements and fitted them with both normal and log-normal PDFs, as shown in Figure 3. We compared the goodness of fit using the reduced values to identify the preferred statistical description. We find that, for the majority of these sources, the soft X-ray flux distributions are better described by a log-normal PDF, while TeV J0232+202 is better described by a normal distribution. The best-fitting parameters for both models are summarized in Table 3.
Figure 3. Soft X-ray flux distributions of TeV blazars. Histograms are shown for all sources displayed, while normal and log-normal probability density functions are fitted only for sources with sufficiently well-sampled soft X-ray light curves. The red and black solid curves denote the best-fitting normal and log-normal PDFs, respectively. (Only four representative sources are shown as examples.)
| TeV Name | ||||||||
| J0035+598 | 145.83 | 99.51 | 10.95 | 4.77 | 0.67 | 1.35 | 1.40 | 6487.52 |
| J0214+517 | 17.25 | 3.94 | 1.81 | 2.82 | 0.22 | 1.26 | 1.10 | 5491.63 |
| J0222+430 | 5.56 | 3.41 | 1.61 | 0.41 | 42.00 | 1.10 | 5592.69 | |
| J0232+202 | 25.92 | 8.52 | 1.23 | 3.19 | 0.36 | 2.53 | 1.20 | 5695.75 |
| J0348-278 | 4.01 | 1.66 | 1.41 | 1.31 | 0.40 | 0.38 | 1.00 | 1378.24 |
| J0416+010 | 30.56 | 10.31 | 3.58 | 3.37 | 0.33 | 1.57 | 1.30 | 5519.23 |
| J0509+056 | 3.63 | 1.83 | 46.22 | 1.20 | 0.40 | 6.62 | 1.10 | 5281.77 |
| J0521+211 | 22.40 | 20.07 | 16.94 | 2.76 | 0.83 | 2.98 | 1.10 | 5184.79 |
| J0550-322 | 67.98 | 16.00 | 2.39 | 4.19 | 0.24 | 1.40 | 1.10 | 7026.99 |
| J0650+250 | 84.10 | 42.05 | 3.24 | 4.29 | 0.56 | 2.22 | 1.40 | 4916.07 |
| J0710+591 | 48.17 | 18.19 | 9.95 | 3.81 | 0.37 | 8.16 | 1.30 | 5532.80 |
| J0721+713 | 20.76 | 16.28 | 115.67 | 2.78 | 0.70 | 2.39 | 1.30 | 6514.46 |
| J0854+201 | 10.21 | 6.43 | 2.19 | 0.49 | 5.84 | 1.30 | 6914.66 | |
| J0904-575 | 7.28 | 2.46 | 3.56 | 1.93 | 0.32 | 1.96 | 0.80 | 2682.51 |
| J0958+655 | 10.97 | 7.08 | 4.65 | 2.20 | 0.64 | 1.52 | 1.20 | 6428.02 |
| J1015+494 | 48.23 | 40.89 | 10.51 | 3.51 | 0.90 | 2.14 | 1.60 | 6701.96 |
| J1031+508 | 40.37 | 13.85 | 2.87 | 3.64 | 0.35 | 1.19 | 0.20 | 5760.83 |
| J1103-234 | 79.79 | 16.29 | 3.20 | 4.36 | 0.20 | 2.42 | 1.20 | 6831.19 |
| J1104+382 | 982.50 | 659.85 | 6.67 | 0.70 | 1.97 | 1.20 | 6983.96 | |
| J1136+701 | 28.59 | 34.32 | 19.81 | 2.75 | 1.10 | 2.78 | 1.40 | 5319.45 |
| J1159+292 | 5.42 | 2.09 | 6.02 | 1.60 | 0.45 | 2.27 | 1.20 | 5741.68 |
| J1221+282 | 4.58 | 3.48 | 81.93 | 1.34 | 0.56 | 3.86 | 0.50 | 3085.06 |
| J1221+301 | 76.08 | 45.54 | 113.95 | 4.16 | 0.60 | 1.74 | 1.20 | 6483.06 |
| J1224+213 | 6.08 | 2.38 | 7.30 | 1.74 | 0.35 | 2.72 | 1.10 | 4045.92 |
| J1256-057 | 19.97 | 12.97 | 2.87 | 0.46 | 46.73 | 1.20 | 6671.25 | |
| J1310+323 | 4.45 | 1.91 | 11.61 | 1.41 | 0.41 | 0.79 | 1.20 | 6223.61 |
| J1422+323 | 4.71 | 1.66 | 2.61 | 1.49 | 0.35 | 2.04 | 0.10 | 1906.35 |
| J1427+238 | 15.70 | 14.24 | 21.36 | 2.43 | 0.75 | 6.66 | 1.40 | 4676.22 |
| J1428+426 | 60.10 | 16.39 | 3.60 | 4.06 | 0.27 | 1.14 | 1.20 | 6815.60 |
| J1512-091 | 12.89 | 3.53 | 4.66 | 2.52 | 0.27 | 1.83 | 1.10 | 6192.80 |
| J1555+111 | 75.04 | 47.89 | 14.09 | 4.10 | 0.70 | 2.65 | 1.40 | 6941.88 |
| J1653+397 | 206.93 | 127.01 | 127.34 | 5.16 | 0.60 | 3.18 | 1.40 | 6993.36 |
| J1725+118 | 13.56 | 8.52 | 12.60 | 2.47 | 0.49 | 2.96 | 0.90 | 5627.65 |
| J1728+502 | 52.81 | 35.18 | 15.65 | 3.75 | 0.67 | 3.46 | 1.50 | 4806.24 |
| J1751+096 | 8.79 | 5.33 | 3.96 | 2.00 | 0.58 | 1.50 | 1.30 | 5994.63 |
| J1959+651 | 453.42 | 237.70 | 31.91 | 5.98 | 0.55 | 1.24 | 1.30 | 6900.35 |
| J2009-488 | 54.04 | 72.38 | 31.72 | 3.58 | 0.75 | 13.03 | 1.60 | 5825.59 |
| J2039+523 | 12.41 | 7.42 | 4.23 | 2.33 | 0.64 | 2.05 | 1.30 | 1095.05 |
| J2158-302 | 92.89 | 63.67 | 103.28 | 4.32 | 0.67 | 0.94 | 1.20 | 6594.40 |
| J2202+422 | 27.35 | 25.38 | 3.09 | 0.62 | 381.34 | 1.20 | 6730.69 | |
| J2347+517 | 29.59 | 14.37 | 3.29 | 0.42 | 3.04 | 1.00 | 6842.01 | |
| J2359-306 | 37.50 | 15.67 | 3.86 | 3.54 | 0.41 | 2.36 | 1.60 | 3082.59 |
Table 3. Fitting parameters of soft X-ray flux distributions and power spectral density slopes for TeV blazars with well-sampled light curves. Note— Col. (1): the TeV blazar name; Col. (2): the mean value of the normal distribution; Col. (3): the standard deviation of the normal distribution; Col. (4): the reduced value of the normal distribution; Col. (5): the mean value of the log-normal distribution; Col. (6): the standard deviation of the log-normal distribution; Col. (7): the reduced value of the log-normal distribution; Col. (8): the PSD slope; Col. (9): the total duration, in days, of the light curve over which the PSD slope is obtained.
We do not perform an analogous analysis for hard X-ray flux distributions because the number of hard X-ray observations is generally much smaller than that of soft X-ray observations across the sample. Even for the most extensively observed source in hard X-rays (TeV J1104+382), the available number of measurements () remains insufficient for robust distribution fitting. As a result, the flux distribution analysis presented here focuses exclusively on the soft X-ray band.
Flux–index correlations of TeV blazars
Correlations between X-ray spectral shape and flux provide a valuable diagnostic of spectral evolution and can offer insights into the underlying radiation processes. We examined the relationships between X-ray fluxes and the corresponding photon indices for the TeV blazars in our sample. Following the same sampling considerations as in § 3.3, we perform the soft X-ray flux–index correlation analysis only for sources with sufficiently well-sampled light curves. For the hard X-ray band, we require more than 10 NuSTAR observations before performing the correlation analysis.
For each eligible source, we quantified the strength of the linear relationship between flux and photon index using the Pearson correlation test and performed ordinary least-squares (OLS) regression following Isobe et al. (1990). We treat correlations as statistically significant when the Pearson test yields a probability of , under the null hypothesis of no correlation. The best-fitting regression parameters are summarized in Table 4.
| TeV Name | Soft X-rays | Hard X-rays | ||||||
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) |
| J0035+598 | – | – | – | – | ||||
| J0214+517 | – | – | – | – | ||||
| J0222+430 | – | – | – | – | ||||
| J0232+202 | – | – | – | – | ||||
| J0348-278 | – | – | – | – | ||||
| J0416+010 | – | – | – | – | ||||
| J0509+056 | ||||||||
| J0521+211 | – | – | – | – | ||||
| J0550-322 | – | – | – | – | ||||
| J0650+250 | – | – | – | – | ||||
| J0710+591 | – | – | – | – | ||||
| J0721+713 | – | – | – | – | ||||
| J0854+201 | – | – | – | – | ||||
| J0904-575 | – | – | – | – | ||||
| J0958+655 | – | – | – | – | ||||
| J1015+494 | – | – | – | – | ||||
| J1031+508 | – | – | – | – | ||||
| J1103-234 | – | – | – | – | ||||
| J1104+382 | ||||||||
| J1136+701 | – | – | – | – | ||||
| J1159+292 | – | – | – | – | ||||
| J1221+282 | – | – | – | – | ||||
| J1221+301 | – | – | – | – | ||||
| J1224+213 | – | – | – | – | ||||
| J1256-057 | – | – | – | – | ||||
| J1310+323 | – | – | – | – | ||||
| J1422+323 | – | – | – | – | ||||
| J1427+238 | – | – | – | – | ||||
| J1428+426 | – | – | – | – | ||||
| J1512-091 | – | – | – | – | ||||
| J1555+111 | – | – | – | – | ||||
| J1653+397 | ||||||||
| J1725+118 | – | – | – | – | ||||
| J1728+502 | – | – | – | – | ||||
| J1751+096 | – | – | – | – | ||||
| J1959+651 | – | – | – | – | ||||
| J2009-488 | – | – | – | – | ||||
| J2039+523 | – | – | – | – | ||||
| J2158-302 | ||||||||
| J2202+422 | – | – | – | – | ||||
| J2347+517 | – | – | – | – | ||||
| J2359-306 | – | – | – | – |
Table 4. Table 4. Slopes and intercepts of linear regressions between photon index and X-ray flux for TeV blazars with well-sampled X-ray light curves. Note—The linear regression is described as , where and are the variables, and and are the slope and intercept, respectively. Symbol “–” represents the null value.
The flux–index relations derived from soft X-ray measurements are shown in Figure 4, while those based on hard X-ray measurements are shown in Figure 5. Sources with fewer observations are not included in the regression analysis, as limited sampling can obscure intrinsic correlations and lead to unreliable slope estimates.
Figure 4. Soft X-ray flux versus photon index for TeV blazars. Data points are shown for all sources displayed, while ordinary least-squares (OLS) linear regression is performed only for sources with sufficiently well-sampled soft X-ray light curves. The blue solid line denotes the resulting regression. (Only four representative sources are shown as examples.)
Figure 5. Hard X-ray flux versus photon index for TeV blazars. Data points are shown for all sources displayed, while ordinary least-squares (OLS) linear regression is performed only for sources with sufficiently well-sampled hard X-ray light curves. The blue solid line denotes the resulting regression. (Only four representative sources are shown as examples.)
Discussion
Correlation between X-ray luminosity and TeV luminosity
The correlation between X-ray and TeV luminosities could probe whether the two bands are produced by physically connected emission processes. We therefore compare the X-ray luminosities derived in this work with the TeV luminosities compiled from Ouyang et al. (2023). The X-ray fluxes are obtained from the spectral fitting described in § 2. The X-ray luminosity is then calculated as
where is the unabsorbed integrated X-ray flux, is the luminosity distance derived from the source redshift, is the redshift, and is the -correction. For spectra fitted with a power-law model, we adopt , while for spectra fitted with a log-parabolic model, we adopt , where is the photon index at the reference energy. For each source, the X-ray luminosity was calculated for each individual observation using (3). We then computed the arithmetic mean of these luminosities and adopted it as the representative X-ray luminosity of the source. The mean soft X-ray luminosity was derived from all available Swift-XRT observations in the 0.3–10 keV band, while the mean hard X-ray luminosity was derived from all available NuSTAR observations in the 3–60 keV band. These mean X-ray luminosities were then used to investigate their correlations with the TeV luminosities.
Figure 6 shows the TeV luminosity as a function of the soft and hard X-ray luminosities. We use OLS regression and the Pearson test to quantify the correlations; the results are listed in Table 2. We find significant positive correlations between TeV luminosity and both soft and hard X-ray luminosities. This result supports a physical connection between the X-ray and TeV emissions in the sample. In the leptonic framework, such a connection is expected, as the X-ray and TeV emissions are produced by the same population of relativistic electrons. For many TeV blazars, especially HSPs, the X-ray is usually dominated by synchrotron radiation from high-energy electrons, while the TeV emission is mainly produced through inverse Compton scattering by the same electron population. For LSPs and some ISPs, however, the X-ray may instead originate from the low-energy part of the inverse Compton component or the transition region between the synchrotron and inverse Compton components (e.g., (Finke et al. 2008; Dermer et al. 2009). Even in these cases, a positive luminosity correlation can arise if both bands respond to the changes in physical conditions in the emitting region, such as particle density or Doppler boosting.
Figure 6. The TeV luminosity as a function of the soft (left) and hard (right) X-ray luminosities. The black line is the regression line.
| Relation | ||||
| (1) | (2) | (3) | (4) | (5) |
| vs | ||||
| vs | ||||
| vs | ||||
| vs | ||||
| vs | ||||
| vs |
Table 2. Linear regression and Pearson correlation results
By contrast, in hadronic scenarios, the TeV emission can originate from several processes, such as proton synchrotron radiation, neutral pion decay, and cascade emission from secondary particles. In these cases, the TeV luminosity is controlled not only by the electron population and leptonic radiative component responsible for the X-ray emission, but also by the proton distribution, magnetic field strength, target photon density, and efficiency of photohadronic interactions (e.g., [14] (Cerruti et al. 2019). As a result, a tight or direct correlation between X-ray and TeV luminosities is not necessarily expected.
Correlation between X-ray luminosity and central engine
The X-ray emission of blazars is mainly produced in relativistic jets, and the jet power of blazars is connected with the central engine, including the black hole and accretion system [12, 11]. Previous studies have also suggested a close connection between accretion power and jet power in blazars (e.g., Ghisellini et al. 2009, 2014; Xiao et al. 2022; Xie et al. 2024). Motivated by this disk–jet connection, we further examine whether the X-ray luminosity is related to the basic properties of the central engine of TeV blazars. Specifically, we compare the X-ray luminosity with the black hole mass, , and accretion disk luminosity, , whose values are collected from Xie et al. (2024).
Figure 7 shows the black hole mass and disk luminosity as functions of the soft and hard X-ray luminosities, with the corresponding correlation results listed in Table 2. No significant correlation is found between black hole mass and X-ray luminosity. This suggests that the X-ray luminosity of TeV blazars is unlikely to be governed solely by black hole mass. This lack of dependence is reasonable because the observed X-ray emission is dominated by the relativistic jet rather than by the disk–corona system commonly seen in radio-quiet AGNs. Relativistic beaming, viewing angle, magnetic field strength, and particle acceleration efficiency can all broaden or obscure any intrinsic relation with black hole mass. In contrast, significant positive correlations are found between disk luminosity and both soft and hard X-ray luminosities. These results suggest that sources with higher disk luminosities tend to have stronger X-ray emission, which may reflect an underlying disk–jet connection. A higher disk luminosity generally indicates a larger accretion power, which can provide more energy for jet launching and particle acceleration, producing more powerful non-thermal radiation.
Figure 7. The black hole mass (upper) and disk luminosity (lower) as a function of the soft (left) and hard (right) X-ray luminosities. The black line is the regression line.
Timing analysis
Variability
We quantified the variability in excess of the measurement uncertainties for each source using the fractional variability amplitude, , which is commonly used in AGN variability studies and is defined as Vaughan et al. (2003)
where is the number of data points, and is the -th flux measurement, and are the mean flux and the mean square flux uncertainty, respectively. The uncertainty on , denoted by , was calculated as (Vaughan et al. 2003; Poutanen et al. 2008)
The results are listed in Table 1. When the numerator in Equation (4) is negative, no positive fractional variability is obtained, indicating that the observed variability is weak relative to the measurement uncertainties.
We classify a source as variable when . With this criterion, we identify 78 sources as variable in soft X-rays and 14 sources as variable in hard X-rays. This confirms that X-ray variability is common in TeV blazars. However, we caution that X-ray monitoring is often observationally biased toward bright and/or flaring states. In addition, many X-ray light curves are irregularly sampled and contain substantial observational gaps, which can affect variability diagnostics and may cause some sources to appear less variable than they actually are.
Power spectral density
The long-term X-ray light curves of TeV blazars exhibit stochastic variability spanning a wide range of amplitudes, with some sources varying by orders of magnitude. To characterize variability as a function of temporal frequency, we constrain the underlying power spectral density (PSD) properties. Because reliable PSD estimation requires a sufficient number of measurements, we restrict this analysis to sources with well-sampled soft X-ray light curves, as in the same procedure in § 3.3.
The sparse and uneven sampling of the long-term X-ray light curves in our sample, as well as the effects of aliasing and red-noise leakage, can distort the intrinsic PSDs. We therefore characterized the PSD slopes using the multiple fragments variance function (MFVF; Kastendieck et al. 2011), combined with a forward-modeling procedure analogous to the power spectral response (PSRESP) method (Uttley et al. 2002). The MFVF characterizes the variability of the source at different timescales by recursively dividing the light curve into two equal segments and computing the variance of each segment until a predefined minimum timescale is reached. We assumed an intrinsic power-law PSD, , and tested PSD slopes over the range in steps of 0.1. For each trial value of , we generated 1000 artificial light curves using the method of Emmanoulopoulos et al. (2013), which reproduces both the assumed PSD and the best-fitting flux PDF derived in § 3.3. To account for red-noise leakage, each artificial light curve was initially generated with a duration 100 times longer than that of the observed light curve. Each simulated light curve was then resampled to match the temporal sampling of the corresponding observed light curve. For each simulated light curve, we calculated the MFVF in the same way as for the observed data. Thus, for a given trial slope and MFVF bin , the 1000 simulations yielded a distribution of simulated MFVF values. For each bin, we compared the observed MFVF value with the corresponding simulated distribution by evaluating the probability density of this distribution at the observed MFVF value, denoted as . A larger value of indicates that the observed MFVF value is more consistent with the MFVF values produced by the simulations for the trial PSD slope. The likelihood of each trial PSD slope was then evaluated by combining the probability densities over all MFVF bins, with the log-likelihood defined as
The value of that maximizes was adopted as the best-fitting PSD slope.
The best-fitting PSD parameters are summarized in Table 3, together with the total duration of each light curve. Figure 8 presents representative observed MFVFs and the simulated MFVF distributions for their best-fitting PSD slopes, together with the likelihood profiles. The resulting PSD slopes span a range of , with the majority of the sources having values between 1.0 and 1.6, as listed in Table 3.
Figure 8. Variability of the source at different timescales for TeV blazars in the soft X-ray band, obtained by the multiple fragments variance function (MFVF). The analysis is performed only for sources with sufficiently well-sampled soft X-ray light curves. Gray circles show the observed raw MFVF, and black circles show the binned observed MFVF. Red squares represent the mean MFVF values in each frequency bin derived from 1000 simulated light curves based on the best-fitting PSD slope obtained through the forward-modeling procedure, while the corresponding uncertainties are derived from the standard deviation of the distribution of simulated MFVF values. The inset figure shows the log-likelihood as a function of PSD slope calculated by Equation (6), and the vertical red line marks the best-fitting PSD slope. (Only four representative sources are shown as examples.)
The PSD provides a compact description of variability as a function of temporal frequency and can therefore offer qualitative clues to the underlying stochastic processes. In radio-quiet AGNs, X-ray PSDs commonly show red-noise behavior and are sometimes better described by bending power laws, with characteristic bend frequencies linked to black hole mass and accretion rate (e.g., (McHardy & Czerny 1987; Lawrence & Papadakis 1993; Uttley et al. 2002; Markowitz et al. 2003; McHardy et al. 2004; González-Martín & Vaughan 2012)). Blazars differ from radio-quiet AGNs because their observed X-ray emission is generally dominated by the relativistic jet. Their X-ray PSDs can therefore probe stochastic processes operating in the jet rather than accretion-flow variability alone. Mukherjee et al. (2019) suggested that the characteristic timescale associated with a break in the blazar PSD may reflect the typical time interval between successive shocks propagating along the jet. Observationally, Kataoka et al. (2001) found breaks in the PSDs of the TeV blazars Mrk 421, Mrk 501, and PKS 2155-304 at timescales of about one day or longer, and suggested that these timescales are related to the typical duration of individual flares. At timescales shorter than one day, the PSDs of these sources were steep, with . On longer timescales, flatter X-ray PSDs have been obtained for several blazars. For Mrk 421, Isobe et al. (2015) obtained from three years of Monitor of All Sky X-ray Image (MAXI) monitoring. Goyal (2020) found X-ray PSD slopes of for Mrk 421 and PKS 2155-304, consistent with our results of for both sources. A similarly broad range has been found in non-TeV blazars. The long-term RXTE X-ray light curve of 3C 279 obtained in Chatterjee et al. (2008). In contrast, Goyal et al. (2022) obtained the flatter X-ray slopes of from the Rossi X-ray Timing Explorer (RXTE) and from Swift-XRT for the same source. The PSD slopes derived for our TeV-blazar sample, predominantly , are consistent with those reported from long-term X-ray monitoring of well-studied TeV blazars, suggesting that the long-term X-ray variability is driven by stochastic processes. A few sources in our sample show relatively flat PSD slopes, with , 0.2, 0.5, 0.8, and 0.9 for TeV J1422+323, TeV J1031+508, TeV J1221+282, TeV J0904-575, and TeV J1725+118, respectively. These sources have light curves with a relatively limited number of data points compared with others, as listed in Table 1, which may affect the determination of their PSD slopes and account for the relatively flat values.
Flux distribution
A normal flux distribution is commonly associated with additive stochastic variability, whereas a log-normal distribution is typically interpreted as a signature of multiplicative variability (Uttley et al. 2005). For sources in our sample with sufficiently well-sampled soft X-ray light curves, we find that the soft X-ray flux distributions are generally better described by log-normal probability density functions, with the exception of TeV J0232+202, whose flux distribution is better described by a normal distribution (see § 3.3).
The predominance of log-normal soft X-ray flux distributions is broadly consistent with previous variability studies of blazars. Long-term X-ray observations of BL Lacertae provided an early clear detection of log-normal flux ((Giebels & Degrange 2009)). Subsequent X-ray analyses have shown that the flux distributions can be source-dependent; for example, long-term RXTE observations revealed log-normal behavior for 3C 273, whereas Mrk 421 and Mrk 501 showed more complex, potentially multi-component distributions ((Khatoon et al. 2020)). Log-normal behavior has also been reported at high and very-high -ray energies, including in long-term monitoring of PKS 2155-304 (H. E. S. S. Collaboration et al. 2017) and in sample studies of Fermi-LAT blazars ((Shah et al. 2018; Wang et al. 2024)). The normal distribution found for TeV J0232+202 in our sample demonstrates that log-normal variability is not common among TeV blazars and that the variability process may depend on the source and/or its activity state.
Log-normal variability is commonly associated with multiplicative stochastic processes in accreting systems (Uttley et al. 2005). In blazars, such behavior may arise if fluctuations generated in the accretion flow propagate into the relativistic jet (McHardy 2008). Alternatively, multiplicative variability may be generated within the jet itself. The observed X-ray flux depends on several coupled physical quantities, including particle injection and acceleration, radiative cooling, and Doppler boosting, so stochastic variations in these parameters may naturally produce lognormal-like flux distributions. In addition, scenarios involving the superposition of multiple emitting sub-regions (e.g., “mini-jets”; [10]) or fluctuations in characteristic particle acceleration timescales (Shah et al. 2018; Sinha et al. 2018) have also been proposed to produce log-normal behavior. By contrast, an approximately normal flux distribution may suggest that the observed variability is dominated by the linear superposition of many statistically independent emission events or emitting regions, rather than by multiplicative processes.
Spectral behavior
Spectral evolution can provide insight into the balance between particle acceleration and radiative cooling in the emitting region. Assuming a power-law electron distribution at injection,
where and are the minimum and maximum injected Lorentz factors, is the electron spectral index, and is a normalization related to the total particle density, the optically thin synchrotron emissivity scales as
where (Chiang & Böttcher 2002). The synchrotron spectral index is therefore , and in the X-ray band for a power-law photon spectrum. Because synchrotron cooling is energy dependent (), changes in acceleration and cooling conditions can modify the high-energy electron population and thereby affect both the X-ray flux and photon index.
Using the flux–index correlations derived in § 3, we find distinct behaviors among HSPs, ISPs, and LSPs (Figure 4 and Figure 5; Table 4). In the discussion below, we focus on cases with as candidate correlations.
HSPs. All HSPs with significant correlations exhibit an anti-correlation between soft X-ray flux and photon index, indicating a harder-when-brighter trend. Since the soft X-ray emission of HSPs is typically dominated by the synchrotron component (Figure 2), episodes of enhanced acceleration and/or an increased high-energy cutoff in the electron distribution can harden the X-ray spectrum and increase the flux and harden the X-ray spectrum. This result is consistent with previous X-ray studies of TeV HSPs, in which harder spectra at higher fluxes were associated with changes in the high-energy end of electrons (Wierzcholska & Wagner 2016; Wang et al. 2018; Giommi et al. 2021). The absence of a significant correlation in some HSPs should not be interpreted as evidence against this interpretation, because the flux–index relation can be diluted by sparse sampling or a limited observed window.
LSPs. TeV J1310+323 shows an opposite trend with softer-when-brighter. In LSPs, X-rays often probe the rising part of the inverse Compton component and/or a regime where multiple components contribute (Figure 2). An increase in the relative synchrotron contribution in the soft X-ray band (e.g., due to changes in the electron distribution) can soften the observed X-ray spectrum, producing a softer-when-brighter trend. We also note that TeV J1256-057 shows harder-when-brighter behavior. This may reflect source-dependent component mixing (e.g., competition between non-thermal jet emission and thermal/reflective components) and/or changes in the relative dominance of spectral components across states (e.g., Giommi et al. 2021; Grandi & Palumbo 2004; Mondal et al. 2022).
ISPs. Most ISPs show softer-when-brighter behavior. This trend is consistent with changes in the relative dominance of synchrotron and inverse Compton components as the source brightens (e.g., Wierzcholska & Wagner 2016; Wierzcholska & Siejkowski 2016). Three ISPs (TeV J1427+238, TeV J1512-091, and TeV J2039+523) exhibit harder-when-brighter trends. For TeV J1512-091, the optical–UV accretion-disk contribution visible in the SED may bias the inferred synchrotron peak classification, and the X-rays appear consistent with sampling the rising inverse Compton component.
Wierzcholska & Wagner (2016) identified X-ray spectral upturns in five TeV blazars, consistent with a superposition of the high-energy synchrotron tail and the low-energy inverse Compton component. Their multi-epoch analysis showed that the synchrotron component was more variable than the inverse Compton component, while the upturn energy changed modestly with flux. They suggested that the spectral evolution could not be explained solely by changes in the cooling of the highest-energy electrons. This result may help explain the diverse flux–index trends observed in our sample. In HSPs, where the soft X-ray band is usually synchrotron dominated, harder-when-brighter behavior may plausibly reflect changes in particle acceleration and cooling. By contrast, in ISPs and LSPs, the X-ray band can cover the transition between synchrotron and inverse Compton emission, such that either harder-when-brighter or softer-when-brighter behavior may arise depending on the relative contribution of the two components.
In hard X-rays, two HSPs, TeV J1104+382 and TeV J1653+397, show strong anti-correlations between flux and photon index, again indicating harder-when-brighter behavior (see Figure 5 and Table 4). This is consistent with hard X-rays tracing the high-energy synchrotron emission in these sources. For sources without significant correlations, limited sampling may dilute intrinsic relationships between flux and spectral shape.
The physical mechanism responsible for particle acceleration in blazar jets remains an open question. Among the commonly invoked scenarios, the shock-in-jet model (Marscher & Gear 1985) provides a qualitative framework in which disturbances propagating along the jet can modify the energy distribution of radiating particles and thereby produce correlated flux and spectral variability. Such scenarios have been discussed in the context of X-ray and optical spectral evolution (e.g., Rani et al. 2010; Agarwal & Gupta [4]; Xiong et al. 2017; Wang & Jiang 2024), and may offer a plausible, though not unique, interpretation for the spectral trends identified in this work.
Conclusion
We have conducted a systematic X-ray study of 85 TeV-detected blazars using archival Swift-XRT and NuSTAR observations. The sample shows common X-ray variability, with 78 sources classified as variable in soft X-rays and 14 in hard X-rays. For well-sampled soft X-ray light curves, the flux distributions are generally log-normal, except for TeV J0232+202, and the PSD slopes are typically flat, with values of –1.6. The flux–index relations show harder-when-brighter behavior in HSPs and more diverse trends in ISPs and LSPs, consistent with changes in the dominant X-ray emission component. We find positive correlations between TeV luminosity and both soft and hard X-ray luminosities, suggesting a physical connection between the X-ray and TeV emissions, consistent with a leptonic framework. The X-ray luminosity also correlates with disk luminosity, but not significantly with black hole mass, supporting an accretion–jet connection while indicating that the observed X-ray is primarily regulated by jet conditions.
Appendix
Tables of X-ray Observations
The X-ray measurements used in this work are provided in machine-readable form. Because the Swift-XRT and NuSTAR data have different instrumental quantities, we provide separate tables for the two instruments. The Swift-XRT table includes the observation information, count rate, spectral parameters, unabsorbed 0.3–10 keV flux, fit statistic, readout mode, and derived luminosity. The NuSTAR table includes the FPMA/FPMB exposures and count rates, cross-normalization factor, spectral parameters, unabsorbed 3–60 keV flux, fit statistic, and derived luminosity. Examples of the table formats are shown in Table 5 and Table 6.
| Source | ObsID | Exposure | MJD | Count Rate | () | C-stat/DoF | Mode | ||
| TeV J0013-188 | 00031806002 | 3800.86 | 55449.12 | 0.2436 0.0080 | 291.12/303 | PC | |||
| TeV J0013-188 | 00031806003 | 4215.40 | 55450.09 | 0.2864 0.0083 | – | 309.43/333 | PC | ||
| TeV J0013-188 | 00031806004 | 3858.30 | 55451.12 | 0.2573 0.0082 | 288.66/304 | PC | |||
| TeV J0013-188 | 00031806005 | 2744.99 | 55452.08 | 0.2605 0.0098 | – | 252.51/281 | PC | ||
| TeV J0013-188 | 00031806006 | 1525.85 | 56925.33 | 0.1321 0.0094 | 116.28/135 | PC | |||
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
Table 5. Example of the Swift-XRT Observation Table
| Source | ObsID | Exposure A | Exposure B | MJD | Count Rate A | Count Rate B | Const | () | C-stat/DoF | ||
| TeV J0035+598 | 60160017002 | 23279.71 | 23049.70 | 59998.35 | 0.2498 0.0033 | 0.2382 0.0033 | – | 1083.53/1081 | |||
| TeV J0112+227 | 60801029002 | 57315.95 | 56277.54 | 59905.67 | 0.0288 0.0007 | 0.0274 0.0007 | – | 986.27/1030 | |||
| TeV J0112+227 | 60801029004 | 100473.60 | 99547.95 | 59908.18 | 0.0324 0.0006 | 0.0311 0.0006 | – | 1344.44/1331 | |||
| TeV J0214+517 | 60160094002 | 21439.83 | 21364.04 | 57417.61 | 0.1718 0.0029 | 0.1596 0.0028 | – | 1082.26/1128 | |||
| TeV J0222+430 | 60601022002 | 24608.71 | 24358.41 | 59143.25 | 0.0187 0.0009 | 0.0183 0.0009 | – | 476.85/558 | |||
| ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... | ... |
Table 6. Example of the NuSTAR Observation Table
Note.— Col. (1): the TeV blazar name; Col. (2): the 4FGL name; Col. (3): the source redshift; Col. (4): the Galactic hydrogen column density in units of from HI4PI Collaboration et al. (2016); Col. (5): the number of soft X-ray observations; Col. (6): the arithmetic mean photon index of soft X-ray observations; Col. (7): the arithmetic mean flux of soft X-ray observations in units of ; Col. (8): the fractional variability of soft X-ray observations; Col. (9): the number of hard X-ray observations; Col. (10): the arithmetic mean photon index of hard X-ray observations; Col. (11): the arithmetic mean flux of hard X-ray observations in units of ; Col. (12): the fractional variability of hard X-ray observations.
Note— Col. (1): the relation; Col. (2): the slope and its uncertainty; Col. (3): the intercept and its uncertainty; Col. (4): the Pearson correlation coefficient; Col. (5): the Pearson chance probability. The linear regression is described as: , where and are the variables, and are the slope and intercept, respectively.
We sincerely thank the anonymous reviewer for the careful review and constructive comments, which helped improve the quality and clarity of the manuscript. H.B.X. acknowledges the support from the National Natural Science Foundation of China (NSFC, grant No. 12203034), the Shanghai Science and Technology Fund (grant No. 22YF1431500), the China Manned Space Project (grant No. CMS-CSST-2025-A07), and the Shanghai Municipal Education Commission regarding artificial intelligence-empowered research. This work is supported by the National Natural Science Foundation of China (NSFC) grant 12261141691. M.M. acknowledges the Croatian Science Foundation (HrZZ) under the project number IP-2022-10-4595 and the University of Rijeka Project uniri-mzi-25-3 funded by the European Union - NextGenerationEU. JHFan’s work is partially supported by the National Natural Science Foundation of China (grant Nos.12433004) and the National Key Research and Development Program of China, Grant No. 2025YFA1614102. The Swift-XRT data are supplied by the UK Swift Science Data Centre at the University of Leicester.
Note— Only the first five rows are shown here. The complete Swift-XRT table is available in machine-readable form. Exposure is given in seconds, MJD is the effective observing time, count rate is in counts , and flux is in erg . Symbol “–” indicates that the parameter is not available.
Note— Only the first five rows are shown here. The complete NuSTAR table is available in machine-readable form. Exposure A and Exposure B are the FPMA and FPMB exposures in seconds, Count Rate A and Count Rate B are the corresponding net count rates in counts , Const is the cross-normalization factor of FPMB relative to FPMA, and is the unabsorbed 3–60 keV flux in erg . Symbol “–” represents the null value.
References
- [1]Abdo, A. A., Ackermann, M., Agudo, I., et al. 2010, ApJ, 716, 30, doi: 10.1088/0004-637X/716/1/30DOI
- [2]Abdollahi, S., Acero, F., Baldini, L., et al. 2022, ApJS, 260, 53, doi: 10.3847/1538-4357/ac6751
- [3]Acciari, V. A., Ansoldi, S., Antonelli, L. A., et al. 2020, ApJS, 247, 16, doi: 10.3847/1538-4365/ab5b98DOI
- [4]Agarwal, A., & Gupta, A. C. 2015, MNRAS, 450, 541, doi: 10.1093/mnras/stv625DOI
- [5]Arbeiter, C., Pohl, M., & Schlickeiser, R. 2002, A&A, 386, 415, doi: 10.1051/0004-6361:20020221DOI
- [6]Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Series, Vol. 101, Astronomical Data Analysis Software and Systems V, ed. G. H. Jacoby & J. Barnes, 17
- [7]Barthelmy, S. D., Barbier, L. M., Cummings, J. R., et al. 2005, SSRv, 120, 143, doi: 10.1007/s11214-005-5096-3DOI
- [8]Bevington, P. R., & Robinson, D. K. 2003, Data reduction and error analysis for the physical sciences
- [9]Bhatta, G., Mohorian, M., & Bilikmsky, I. 2018, A&A, 619, A93, doi: 10.1051/0004-6361/201833628DOI
- [10]Biteau, J., & Giebels, B. 2012, A&A, 548, A123, doi: 10.1051/0004-6361/201220056DOI
- [11]Blandford, R. D., & Payne, D. G. 1982, MNRAS, 199, 883, doi: 10.1093/mnras/199.4.883DOI
- [12]Blandford, R. D., & Znajek, R. L. 1977, MNRAS, 179, 433, doi: 10.1093/mnras/179.3.433DOI
- [14]Böttcher, M., Reimer, A., Sweeney, K., & Prakash, A. 2013, ApJ, 768, 54, doi: 10.1088/0004-637X/768/1/54DOI
- [15]Burrows, D. N., Hill, J. E., Nousek, J. A., et al. 2004, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 5165, X-Ray and Gamma-Ray Instrumentation for Astronomy XIII, ed. K. A. Flanagan & O. H. W. Siegmund, 201–216, doi: 10.1117/12.504868
- [16]Cash, W. 1979, ApJ, 228, 939, doi: 10.1086/156922DOI







