Long-Term X-Ray monitoring of NGC6251: Evidence for a jet-dominated radio galaxy
aa r X i v : . [ a s t r o - ph ] J a n Draft version October 26, 2018
Preprint typeset using L A TEX style emulateapj v. 11/12/01
LONG-TERM X-RAY MONITORING OF NGC 6251:EVIDENCE FOR A JET-DOMINATED RADIO GALAXY
M. Gliozzi
George Mason University, 4400 University Drive, Fairfax, VA 22030
I.E. Papadakis
Physics Department, University of Crete, 710 03 Heraklion, Crete, Greece
R.M. Sambruna
NASA Goddard Space Flight Center, Code 661, Greenbelt, MD 20771
Draft version October 26, 2018
ABSTRACTWe present the first X-ray monitoring observations of the X-ray bright FR I radio galaxy NGC 6251observed with
RXTE for 1 year. The primary goal of this study is to shed light on the origin of theX-rays, by investigating the spectral variability with model-independent methods coupled with time-resolved and flux-selected spectroscopy. The main results can be summarized as follows: 1) Throughoutthe monitoring campaign, NGC 6251 was in relatively high-flux state with an average 2–10 keV absorbedflux of the order of 4.5 × − erg cm − s − and a corresponding intrinsic luminosity of 6 × erg s − .2) The flux persistently changed with fluctuations of the order of ∼ α line is never statistically required, although the presence of a strong iron line cannot beruled out, due to the high upper limits on the line equivalent width. The inconsistency of the spectralvariability behavior of NGC 6251 with the typical trend observed in Seyfert galaxies and the similaritywith blazars lead support to a jet-dominated scenario during the RXTE monitoring campaign. However,a possible contribution from a disk-corona system cannot be ruled out. Subject headings:
Galaxies: active – Galaxies: jets – Galaxies: nuclei – X-rays: galaxies introduction Non-blazar radio-loud active galactic nuclei (AGNs) –objects with jets forming large viewing angles to the line ofsight– are traditionally divided into two classes: Fanaroff-Riley II (FR II), with 178 MHz powers > × W/Hzand edge-darkened radio morphologies, and FR I galaxieswith lower powers and more compact morphologies (Fa-naroff & Riley 1974). For the same host galaxy opticalmagnitude, FRIs produce about one order of magnitudeless optical line emission than FRIIs (Baum et al. 1995)and have fainter or negligible UV continuum fluxes (Zirbel& Baum 1995, Ho 1999).While earlier models for the origin of the FRI/II di-chotomy focused mainly on accounting for their large-scaleradio morphologies, more recently new ideas have emergedconcerning the nature of the central engine in the two typesof radio galaxies in an attempt to explain the nuclear prop-erties. One school of thought is that the nuclear X-rayproperties of FR I and FR II are related to a differentaccretion rate onto the central supermassive black hole,with FRII being dominated by relatively large values of
L/L
Edd , while FRI would be accreting at sub-Eddingtonrates,
L/L
Edd ≪ − (e.g., Reynolds et al. 1996; Ghis-ellini & Celotti 2001).A fundamental step to gain insight into the nature ofthe central engine in radio-loud AGNs is to understandwhether the X-ray radiation is produced by disk/coronasystems as in Seyfert galaxies or by jets as in blazars. While the time-averaged spectroscopy (due to spectral de-generacy) and the pure temporal analysis (due to the factthat both radio-quiet and radio-loud show strong variabil-ity) cannot firmly discriminate between the two competingscenarios, their combination, i.e., time-resolved spectralanalysis and energy-selected temporal analysis, offers inprinciple a better way to distinguish between accretion-dominated and jet-dominated systems. This conclusion issupported by the strikingly different spectral variabilitybehavior shown by Seyfert-like objects (e.g., Papadakis etal. 2002; Markowitz & Edelson 2001) and by blazars (e.g.,Zhang et al. 1999; Fossati et al. 2000; Gliozzi et al. 2006).Here, we concentrate on the X-ray nuclear properties ofthe nearby radio galaxy NGC 6251 ( z =0.024), which is agiant elliptical galaxy hosting a supermassive black holewith mass M BH ∼ − × M ⊙ (Ferrarese & Ford 1999).Based on its radio power at 178 MHz, NGC 6251 is clas-sified as an FR I (e.g., Owen & Laing 1989), whereas inthe optical, it is classified as a type-2 AGNs (e.g., Shuder& Osterbrock 1981).Despite the intensive study of this source at all wave-lengths, the nature of the accretion process in NGC 6251is still a matter of debate. Based on the radio-to-X-rayspectral energy distribution, Ho (1999) suggested that anAdvection-Dominated Accretion Flow (ADAF) is presentin the nucleus of NGC 6251. On the other hand, Fer-rarese & Ford (1999) and Melia et al. (2002) favored astandard accretion disk. Finally, Mukherjee et al. (2002)1 Gliozzi et al.and Chiaberge et al. (2003) advocate a jet origin for thebroad-band emission, based on the possible association ofNGC 6251 with the EGRET source 3EG J1621+8203 andon its spectral energy distribution, respectively.In the X-ray band, NGC 6251 has been previouslyobserved with various satellites. For example,
ROSAT showed the presence of an unresolved nuclear source em-bedded in a diffuse thermal emission with temperature kT ∼ . α line (EW ≃ ASCA in 1994(Turner et al. 1997; Sambruna et al. 1999) suggesteda standard Seyfert-like scenario with accretion-dominatedX-rays.
BeppoSAX observed NGC 6251 in July 2001 dur-ing a high-flux state (the 2–10 keV flux, F X = 4 . × − erg cm − s − , was ∼ ASCA α line (EW <
100 eV) or other re-processing features, Guainazzi et al. (2003) proposed ascenario with two main spectral components: a blazar-likespectrum dominating the high-flux state and a Seyfert-like spectrum emerging during the low-flux state. Morerecently, in March 2002, NGC 6251 was observed with
XMM-Newton when the 2–10 keV flux was ∼
15% lowerthan the
BeppoSAX value. Despite the relatively highflux, the
XMM-Newton spectral results seem to supportthe picture emerged from the
ASCA observation with thepresence of a prominent (EW ∼
220 eV) and possibly broad( σ > . α line (Gliozzi et al. 2004). However,the existence of a broad Fe K α line in the XMM-Newton spectrum is still a matter of debate (see Evans et al. 2005and Gonz´alez-Mart´ın et al. 2006 for a discording and asupporting view, respectively).The controversial results derived from previous X-raystudies highlight the impossibility of firmly determiningthe origin of the X-rays in NGC 6251 based solely ontime-averaged spectral results. Here, we present the re-sults from a systematic study of the long-term X-ray fluxand spectral variability of NGC 6251 using one year longproprietary Rossi X-ray Timing Explorer (
RXTE ) obser-vations in the 2-12 keV range. We use model-independentmethods and time-resolved spectroscopy to study the X-ray temporal and spectral properties of this source. Themain purpose of this analysis is to shed light on the originof the X-rays and in particular on the role played by ajet in the X-rays. Once (if) the jet contribution is prop-erly assessed, the physical parameters characterizing theaccretion process onto the supermassive black hole can bebetter constrained, and hence it is possible to discriminatebetween competing theoretical models for the accretionprocess.The outline of the paper is as follows. In § §
3. In § § § § observations and data reduction We use proprietary
RXTE data of NGC 6251 that wasregularly observed for ∼ RXTE . Here we will consider only PCA data, becausethe signal-to-noise of the HEXTE data is too low for ameaningful analysis.The PCA data were screened according to the follow-ing acceptance criteria: the satellite was out of the SouthAtlantic Anomaly (SAA) for at least 30 minutes, theEarth elevation angle was ≥ ◦ , the offset from thenominal optical position was ≤ ◦ .
02, and the parameterELECTRON-2 was ≤ .
1. The last criterion removes datawith high particle background rates in the ProportionalCounter Units (PCUs). The PCA background spectra andlight curves were determined using the L7 −
240 modeldeveloped at the
RXTE
Guest Observer Facility (GOF)and implemented by the program pcabackest v.3.0. Thismodel is appropriate for “faint” sources, i.e., those pro-ducing count rates less than 40 s − PCU − .All the above tasks were carried out using the FTOOLS v.6.2 software package and with the help of the
REX scriptprovided by the
RXTE
GOF. Data were initially ex-tracted with 16 s time resolution and subsequently re-binned at different bin widths depending on the applica-tion. The current temporal analysis is restricted to PCA,STANDARD-2 mode, 2–12.5 keV, Layer 1 data, becausethat is where the PCA is best calibrated and most sensi-tive. PCUs 0 and 2 were turned on throughout the mon-itoring campaign. However, since the propane layer onPCU0 was damaged in May 2000, causing a systematic in-crease of the background, we conservatively use only PCU2for our analysis (see below). All quoted count rates aretherefore for one PCU.The spectral analysis of PCA data was performed us-ing the
XSPEC v.12.3.1 software package (Arnaud 1996).We used PCA response matrices and effective area curvescreated specifically for the individual observations by theprogram pcarsp , taking into account the evolution of thedetector properties. All the spectra were re-binned so thateach bin contained enough counts for the χ statistic to bevalid. Fits were performed in the energy range 2.5–12.5keV, where the signal-to-noise ratio is the highest. the x-ray light curves Although NGC 6251 is generally considered an X-raybright source – its average flux is of the order of 4 × − erg cm − s − with a corresponding luminosity L −
10 keV ∼ × erg s − that is nearly a factor 4 larger than the typ-ical values observed in low-power radio galaxies (Donatoet al. 2004) – it is rather weak for the RXTE capabilities.Therefore, before starting a detailed analysis of temporalproperties, it is necessary to demonstrate that the vari-ability observed cannot be ascribed to uncertainties in the
RXTE background or to other artifacts. To this end, wehave performed the following test: We have compared thebackground-subtracted light curves obtained using PCU2layer 1 and PCU2 layer 3. Since the genuine signal inlayer 3 is quite small, its light curve can be used as a-ray Spectral variability of NGC 6251 3proxy to check how well the background model works. Ifthe latter light curve is significantly variable with a pat-tern similar to the one produced using layer 1, then thevariability is simply due to un-modeled variations of thebackground. Conversely, if the PCU2 layer 3 light curvedoes not show any pronounced variability or if the fluxchanges are uncorrelated with those observed in the layer1 light curve, we can safely conclude that the variabilitydetected in NGC 6251 is real.The two light curves in the 2–10 keV range (where thebackground PCA model is better parameterized; see Ja-hoda et al. 2006 for more details) are shown in Figure 1.A visual inspection of this figure suggests that the vari-ability in the layer 1 light curve is much more pronouncedthan the one observed in layer 3. Indeed, the mean countrate of layer 3 is consistent with zero, indicating that onlong timescales the background model works adequately.However, statistically speaking both light curves are con-sidered variable: χ = 229.5 (layer 1) and 128.3 (layer3) for 78 degrees of freedom (hereafter dof). On theother hand, a formal analysis based on the excess variance(i.e. the variance corrected for statistical errors), indicatesthat variability associated with the layer 1 light curve ismore than one order of magnitude larger than that asso-ciated with layer 3 ( σ xs = 4 × − and 3 × − s − ,respectively). Further support to the fact that the vari-ability associated with layer 1 is genuine comes from acorrelation analysis: When layer 3 is plotted versus thelayer 1 count rate (see Figure 2), no correlation is ob-served, as confirmed by a least square linear fit analysis, y = 0 . ± . − . ± . x . We therefore con-clude that most of the count rate changes observed in layer1 are associated with genuine intrinsic variations of the X-ray source in NGC 6251.3.1. PCU0 versus PCU2
In order to maximize the signal-to-noise (S/N) ratio ofthe light curves, one can combine the 2 PCUs at workduring the monitoring campaign (i.e., PCU0 and PCU2),provided that they are consistent with each other. In par-ticular, it is necessary to test whether PCU0, partiallydamaged since May 2000, is compatible with PCU2. Fig-ure 3 shows the background-subtracted, 2–12 keV lightcurves of PCU2 (top panel), PCU0 (middle panel), eachwith superimposed the light curve of the other PCU (rep-resented by continuous lines in top and middle panels),and their ratio (bottom panel) with a time bin of 4 days.A visual inspection of this figure suggests that the PCU0and PCU2 light curves are not fully consistent with eachother. Indeed, a formal check based on a χ test indicatesthat the ratio PCU0/PCU2 is not consistent with the hy-pothesis of constancy ( χ = 160 . ∼ ∼
3) and very large uncertainties. Therefore, also forthe spectral analysis (see §
5) we use only PCU2 data.Using data from PCU2 only, we constructed backgroundsubtracted light curves in two energy bands, namely the“soft” (2.5–5 keV) and the “hard” (5–12 keV) band. Weshow them in Figure 4, together with the hardness ratio(hard/soft) plotted versus time. Time bins are 4 days.The soft and hard band light curves are significantly vari-able ( χ = 113 . /
78 and χ = 175 . /
78, respectively) andappear to be qualitatively similar. This is formally con-firmed by the hardness ratio light curve, which appears tobe constant ( χ = 70 . / spectral variability: model-independentanalysis For this study we use simple methods such as hardnessratio versus count rate plots and the fractional variabilityversus energy plots. These can provide useful informationwithout any a priori assumption regarding the shape ofthe X-ray continuum spectrum. Thus, the results fromthe study of these plots can be considered as “model-independent”.Figure 5 shows the Hard/Soft X-ray color (5–12keV)/(2.5–5 keV) plotted versus the total count rate (2.5–12 keV) for un-binned (smaller, gray symbols) and binneddata (larger, darker symbols), respectively. A visual in-spection of this figure suggests the presence of a positivetrend between HR and the count rate (i.e., the spectrumhardens as the flux increases), although within a large scat-ter. This positive trend is apparently confirmed by thelinear fit of the data, y = (0 . ± .
2) + (2 . ± . x with χ /dof = 43 . /
77, which suggests the presence of a posi-tive correlation at 4.6 σ confidence level. This analysis wasperformed using the routine fitexy (Press et al. 1997)that accounts for the errors not only on the y-axis butalong the x-axis as well.Better insight into the presence of correlations between HR and total count rate can be obtained by investigat-ing the binned data. To this end, we have binned thedata, shown in Fig. 5, using count rate bins of fixed size(0.05 counts/s), and computed the weighted mean (alongboth the y and x-axis) of all the points which fall into abin. The error on the weighted mean is computed follow-ing Bevington’s prescriptions (Bevington 1969). Only binswith at least 5 data points have been plotted in Figure 5and considered for the linear fit. In this case, the best lin-ear fit, y = (0 . ± .
2) + (1 . ± .
6) with χ /dof = 1 . / σ uncertainties), indicates that,if a few outliers with large count rate (and error-bar) areneglected, the significance of a positive correlation reducesto 2.67 σ . This is slightly lower than the 3 σ level, whichis generally accepted for significant variations, but it stillimplies a positive correlation at a confidence level of ∼ F var . This is a commonly used mea-sure of the intrinsic variability amplitude relative to themean count rate, corrected for the effect of random errors,i.e., F var = ( σ − ∆ ) / h r i (1)where σ is the variance, h r i the unweighted mean countrate, and ∆ the mean square value of the uncertaintiesassociated with each individual count rate. The error on F var has been estimated following Vaughan et al. (2003).We computed F var on selected energy bands, with meancount rates similar and sufficiently high.The results are plotted in Figure 6 and suggest that, overthe 3–8 keV energy range, a weak positive trend seems tobe present. However, the large uncertainties on F var dueto the low count rate in the narrow energy-selected bandshamper this kind of analysis. Indeed, statistically speak-ing, the trend shown in Figure 6 is consistent with the hy-pothesis of F var being constant in the energy range probedby RXTE . If only two broader energy bands are used forthis analysis, the resulting fractional variability in the hardenergy band 5–12 keV, F var , hard = (20 ± F var , soft = (10 ± F var , hard − F var , soft =(10 ± σ level only.In summary, the results from the study of the HR − ct plots show evidence for a positive correlation betweenhardness ratio and total count rate. Similarly, an anal-ysis of the fractional variability suggests that F var is morepronounced in the hard energy band than in the soft one.However, due to the rather limited signal-to-noise ratio ofour data these results are significant at just the 2 . . σ level, respectively. spectral analysis Time-resolved Spectroscopy
Given that the data consist of short snapshots span-ning a long temporal baseline, they are in principle wellsuited for monitoring the spectral variability of NGC 6251.However, due to the limited S/N of the data, the spectralslope Γ, measured from spectra of individual observations,cannot be adequately constrained and hence the spectralvariability cannot be investigated in such a way. Indeed,if we plot the values of Γ versus time, no variations aredetected due to the large errors on Γ. This is fully con-sistent with the results from the HR -time plot describedin §
4. In order to increase the S/N and investigate thepresence of possible spectral variations, we use spectra av-eraged over two-month intervals. This choice is a trade-offbetween the necessity of accumulate sufficient counts for areliable spectral analysis and the need to use limited tem-poral intervals to minimize the effects of the slow drift inthe detector gain.We fitted each two-month spectrum with a simplepower-law (PL) model absorbed by Galactic N H (5 . × cm − ). The model fits all the data reasonably well,as indicated by Fig. 7 that shows a typical spectrum fit-ted with a simple power law. The best-fit results arelisted in Table 1 and can be summarized as follows. Asimple PL model provides an acceptable parametrization for all spectra. The photon indices are all rather steep(Γ ∼ .
5) and consistent with each other within the er-rors. In other words, our results suggest that the source’sspectrum does not vary significantly on timescales longerthan two months. The weighted spectral slope mean is2 . ± .
1. Adding a Gaussian line at 6.4 keV to the PLcontinuum model does not improve the fit significantly inany of the six spectra, but the 90% confidence upper limitson the equivalent width are relatively high (EW ∼ F −
10 keV = 4 . × − erg cm − s − , Γ = 2 . ± . EW <
212 eV duringthe first half, and F −
10 keV = 4 . × − erg cm − s − ,Γ = 2 . ± . EW <
144 eV during the second half –are fully consistent with each other.Since there are no indications for long-term spectralvariability, we have fitted the 6 two-month spectra to-gether. This yielded Γ = 2 . ± . <
104 eV. Inthe 2–10 keV energy band, we obtained an average ab-sorbed flux of 4.4 × − erg cm − s − , and a correspond-ing intrinsic luminosity of 6 . × erg s − , assuming H = 71 km s − Mpc − , Ω Λ = 0 .
73 and Ω M = 0 .
27 (Ben-net et al. 2003).5.2.
Flux-selected Spectroscopy
In order to verify the presence of a direct correlationbetween HR and count rate derived in §
4, we performed aflux-selected spectral analysis. To this end we divided the94 individual spectra into 5 bins according to their averagecount rate (namely, < . , , . − . , . − . , . − . , and > .
40 c/s). In each bin, the individual spec-tra were fitted simultaneously with a PL model (absorbedby Galactic N H ) with their photon indices linked togetherand their respective normalizations free to vary.The results are summarized in Table 2, where the re-ported errors on Γ and flux are respectively 1 σ and σ/ √ N ,with N being the number of individual spectra per bin.The upper limits on EW correspond to the 90% confi-dence level. Table 2 indicates that the 2.5–12.5 keV spec-tra harden as the average flux increases. This is clearlyshown in Figure 8, where the values of Γ for each countrate bin have been plotted against their respective fluxvalues. The dashed line represents the best linear fit, y = 3 . ± . − (0 . ± . x , which reveals that theinverse correlation is significant at 3- σ level.For comparison, in Fig. 8 we have also plotted the val-ues corresponding to the ASCA , BeppoSAX , and
XMM-Newton observations. To this end, we have used
PIMMS toconvert the observed flux into the
RXTE energy range,assuming the best-fit spectral parameters reported inGuainazzi et al. (2003) for
BeppoSAX and
ASCA , andGliozzi et al. (2004) for
XMM-Newton . The
ASCA , Bep-poSAX , and
XMM-Newton photon indices seem to followthe same inverse trend shown by
RXTE , becoming flat-ter as the flux increases. However, their values appearto be significantly smaller than those obtained from theflux-selected
RXTE spectral analysis. This apparent dis-crepancy may be probably reconciled (at least for the
Bep- -ray Spectral variability of NGC 6251 5 poSAX and
XMM-Newton data) by bearing in mind that:1) Past studies have shown that photon indices measuredwith the
RXTE
PCA are systematically steeper than thosemeasured by other X-ray satellites (e.g., Yaqoob 2003). 2)The error-bars for the
RXTE data in Fig. 8 are at 1 σ level. On the other hand, it appears more difficult to rec-oncile the ASCA value with the extrapolation of the
RXTE best-fit trend, suggesting that the source was in a differentphysical state (for example, it lacked a jet contribution).We have also added a Gaussian line to the PL continuum(see Table 2 last column). Only for the bin with the lowestflux, the line is very marginally significant (the value of χ decreases by 1.2 for one additional dof) with EW ∼
400 eVand large uncertainties. For bins with larger flux values,only upper limits on EW can be derived. These findingsindicate that a Fe K α line is never statistically required,although its presence cannot be completely ruled out whenthe source flux is at the lowest level measured by RXTE . discussion We have undertaken the first X-ray monitoring studyof the FR I galaxy NGC 6251, investigating the tempo-ral and spectral variability as well as time-averaged andflux-selected spectral results.By comparing the X-ray fluxes measured by differ-ent satellites over nearly a decade of observations, it isclear that NGC 6251 shows large flux changes on longtimescales. For example, in October 1994
ASCA mea-sured a flux of ∼ . × − erg cm − s − in the 2–10keV energy band, whereas BeppoSAX observed a flux of ∼ . × − erg cm − s − in July 2001, and XMM-Newton of ∼ × − erg cm − s − in March 2002. Thehigh throughput of the EPIC pn camera aboard XMM-Newton also revealed the presence of low-amplitude fluxchanges (of the order of ∼ RXTE observations is in agreement with the sparse evidence thathad been gathered in the previous years, and show con-clusively that NGC 6251 is a persistently variable sourcein X-rays. The frequent
RXTE observations, spread overa period of one year, indicate that this radio galaxy ischaracterized by persistent variability in the total (2.5–12keV), soft (2.5–5 keV), and hard (5–12 keV) energy bands.Throughout the observation, the flux appears to randomlychange by a factor of ∼ α line (e.g., Turneret al. 1997; Sambruna et al. 1999; Gliozzi et al. 2004),whereas in other occasions NGC 6251 seems to be moreconsistent with a blazar-like behavior, showing a feature-less X-ray spectrum (e.g., Guainazzi et al. 2003). Thedensely sampled, year-long RXTE observations and theinvestigation of the flux and spectral variability propertiesoffer an alternative and model-independent way to shedsome light on this issue.6.1.
Evidence from the flux variability properties
Unlike the brightest blazars frequently monitored by
RXTE (e.g., Kataoka et al. 2001, Cui 2004, Xue & Cui2005), NGC 6251 does not show any prominent flare on any observable timescale, nor does it show the large am-plitude variability typically observed in several Seyfert-like objects during yearly-long monitoring campaigns (e.g.,Markowitz & Edelson 2001).On one hand, the apparent inconsistency with the largevariability observed in Seyfert galaxies can be explainedby the lower values of black hole mass in the latter ob-jects, which is typically one order of magnitude lower thanNGC 6251. On the other hand, the lack of prominent flaresin the NGC6251 light curve, which instead characterize theblazar light curves, can be understood by keeping in mindthat the blazars monitored by RXTE are the brightestmembers of this AGN class and that the observations areoften triggered only during their flaring activity. Nonethe-less, blazar monitoring campaigns with baselines coveringseveral years reveal that also the brightest blazars alter-nate prominent flaring activity with “quiescent periods”that are characterized by moderate flux variations. In-deed, prolonged periods of moderate variability have beendetected in several blazars (e.g., B¨ottcher et al. 2005;Marscher 2006).A model-independent study of the
RXTE variabilityproperties of the prototypical blazar Mrk 501, similar tothe one performed on NGC 6251 in this work, offers thebest opportunity for a more quantitative comparison ofNGC 6251 with a typical blazar behavior (Gliozzi et al.2006). For instance, Mrk 501, which showed a large out-burst in 1997, underwent a progressive decrease of its ac-tivity in the following years, resulting in a lower meancount rate accompanied by lower variability. Specifically,in 1999, when Mrk 501 reached a minimal flux value, thelight curve was characterized by variations of the order of ∼ RXTE (including Mrk 501) are basically all High-peakedblazars (HBLs), with the synchrotron component peak-ing in the X-ray range and a second spectral component,generally attributed to inverse Compton scattering, peak-ing at TeV energies. On the other hand, in the blazarframework, the broadband SED of NGC 6251 is consistentLow-peaked blazars (LBLs), with the inverse Comptoncomponent peaking in the X-rays (Chiaberge et al. 2003;Guainazzi et al. 2003). As a consequence, a formal com-parison of the X-ray properties of NGC 6251 should be inprinciple performed using the TeV properties of Mrk 501.However, detailed studies of X-ray and TeV emissions inHBLs have demonstrated the existence of a tight correla-tion between these energy bands, indicating that the X-rayand TeV radiation follow the same variability trend (e.g.,Fossati et al. 2004; Gliozzi et al. 2006). Further sup-port to this conclusion comes from recent investigations ofthe TeV properties of Mrk 501 and Mrk 421 carried outwith the MAGIC and Whipple telescopes (Blazejowki etal. 2006; Albert et al. 2007). These studies demonstratethat also in the γ -ray energy band HBLs show the typi-cal blazar spectral variability behavior observed in X-rays,that is a spectral hardening when the source brightens anda fractional variability more pronounced at higher energies(e.g., Zhang et al. 1999; Fossati et al. 2000; Gliozzi et al.2006; Rebillot et al. 2006 and references therein). We Gliozzi et al.therefore conclude that it is appropriate to compare theX-ray properties of Mrk 501 with those of NGC 6251.Another relevant finding from the flux variability anal-ysis of NGC 6251 is that the fractional variability appearsto be more pronounced in the hard than in the soft en-ergy band: ∆ F var ≡ F var , hard − F var , soft = (10 ± RXTE for several months or years (Markowitz & Edelson 2004).Specifically, all the Seyfert galaxies with long
RXTE mon-itoring yield negative values of ∆ F var ranging between -3.5and -24.8, with a mean value of − . ± .
5. On the otherhand, a similar study, carried out using
RXTE monitoringdata of the blazar Mrk 501 between 1997 and 2000, yieldedpositive values of ∆ F var ranging between 5 and 42, with amean value of 15 ± Evidence from the spectral properties
A comparative analysis of the spectral variability ofNGC 6251 and the typical behaviors observed in bothradio-quiet AGNs (where the X-rays are thought to beproduced by Comptonization in the corona that is closelyconnected with the accretion disk) and radio-loud jet-dominated AGNs (whose radiation over the entire energyrange is ascribed to jet emission) can help us understandwhether the X-ray radiation from NGC 6251 is dominatedby jet or accretion-related emission.A flux-selected spectral analysis (in agreement with the HR − ct plot) has shown that the X-ray spectrum ofNGC 6251 hardens as the source brightens, following thelinear correlation Γ ∝ − . ± . × F X . A direct com-parison of this result with a similar spectral study, carriedout on 4 Seyfert galaxies monitored with RXTE , indicatethat the latter have always positive slopes in the Γ − F X plane ranging between 0.05 and 0.15 with a mean valueof 0 . ± .
01 (Papadakis et al. 2002). As a consequence,the spectral variability behavior of NGC 6251 appears tobe inconsistent with the typical Seyfert-like trend. On theother hand, the Γ − F X slope measured for Mrk 501 duringthe weakly variable period of 2000, − . ± .
06, appearsto be fully consistent with the behavior of NGC 6251 dur-ing the
RXTE monitoring campaign.The existence of a Seyfert-like component suggested byprevious X-ray studies was essentially based on the pres-ence of a strong Fe K α line. Unfortunately, the low S/Nspectra obtained during the RXTE monitoring coupledwith the poor spectral resolution of the
RXTE
PCA ham-pers a detailed investigation of this issue. Indeed, if the2-month averaged PCA spectra are fitted with a modelincluding a power law and a Gaussian line with spectralparameters fixed at the best-fit values obtained during the
ASCA observation (we conservatively assumed the best-fitparameters reported in Guainazzi et al. 2003), the resultsare statistically indistinguishable from the fits obtainedusing a simple power law. This indicates that
RXTE is unable to confirm or refute the presence of a Fe K α line,and suggests that only a relatively long exposure of XMM-Newton with its superior capabilities is able to detect thepresence of the line when the average flux of NGC 6251 isrelatively high.6.3.
The origin of X-rays in NGC 6251
The primary goal of this work is to investigate the ori-gin of the X-rays in NGC 6251 and in particular to assessthe role played by the putative jet. At first sight, thepossibility that the jet may dominate the radiative out-put of a radio galaxy may be surprising, and even moreso for NGC6251, which has a Mpc radio jet (and henceappears forming a large viewing angle) and Seyfert-likeemission during the low flux state. However, under spe-cific circumstances, the jet-dominance hypothesis becomesplausible. This is the case when the base of the jet is notwell collimated and the X-rays are produced by a part ofthe outflowing material that points towards the observer.Indeed, this is the framework proposed to explain the TeVemission detected in another radio galaxy M87 (Aharonianet al. 2006). Alternatively, if the base of the jet is tiltedwith respect to the large scale jet and forms a small view-ing angle, the possible dominance of the jet in the 2–12keV range can be naturally explained. In fact, this is thescenario put forward by Jones & Wehrle (2002) to explainthe large jet/counterjet brightness ratio on parsec scalesinferred for NGC 6251 from VLBA observations.By comparing the spectral variability properties ofNGC 6251 with those of radio-quiet AGNs and blazarobjects, we find that they are certainly inconsistent witha typical Seyfert-like behavior but fully consistent withblazars. Combining pieces of information from the model-independent analysis with the findings from the flux-selected spectral analysis, we are led to the conclusionthat, during the
RXTE monitoring campaign, the bulkof the hard X-ray radiation from NGC 6251 was domi-nated by the emission from the unresolved base of the jet.Nonetheless, the presence of a disk-corona component, de-tected in previous observations with
ASCA and
XMM-Newton , cannot be ruled out: the upper limits measuredon the equivalent width of the Fe K α line are indeed fullyconsistent with the values measured by ASCA and
XMM-Newton , but the low S/N of the spectra hampers a morequantitative analysis.In conclusion, we can try to exploit the main resultsfrom this work (i.e., the jet dominance in the high fluxstate, with a possible contribution from a Seyfert-like com-ponent emerging at low flux values) to derive some con-straints on the accretion process at work in NGC 6251.Assuming that an accretion-related component is alwayspresent in NGC 6251 and dominates during the low fluxstate, we can use the average flux measured in the lowcount rate bin (see Table 2) to compare the correspond-ing 2–10 keV luminosity – L X = 6 . × erg s − –to the Eddington value readily derived from the blackhole mass estimate. The relatively high value derived, L X /L Edd > × − , confirms that NGC 6251 is a brightFR I galaxy that is close to the FR I/FR II dividing linein terms of power of the central engine. However, it is notpossible to derive any firm conclusion on the nature of theaccretion process given the unknown contribution of the-ray Spectral variability of NGC 6251 7jet in the low-flux state. summary and conclusions We have used data from a year-long
RXTE monitoringcampaign to study the spectral variability of NGC 6251following model-independent and spectral model fittingmethods. The main results can be summarized as follows: • Throughout the monitoring campaign and espe-cially during the last 4 months, NGC 6251 was ina relatively high-flux state, with values of the 2–10keV absorbed flux comparable to that observed by
BeppoSAX in 2001. • The light curves show persistent variations by a fac-tor of 2 on timescales weeks/months in the total(2.5–12 keV), soft (2.5–5 keV), and hard (5–12 keV)energy bands. • The fractional variability, computed over the softand hard energy bands, reveals an enhanced vari-ability at harder energies ( F var ∼ F var ∼ • There is evidence of a positive trend in the HR − ct plot (or, analogously, of a negative trend in the Γ- flux plot); in other words the spectrum hardens asthe flux increases. • The 2-month averaged spectra are well fitted by apower-law model, with h Γ i ≃ . α line. Combining all the 2-month av-eraged spectra yielded EW <
104 eV. Only for thelowest flux-selected spectrum, there is marginal ev-idence for a Fe K α line. However, the low S/N doesnot allows one to put reliable constraints on theputative Seyfert-like component.The inconsistency of the spectral variability behaviorof NGC 6251 with the typical trend observed in Seyfertgalaxies and the similarity with blazars lead support to ajet-dominated scenario. However, based on the RXTE ob-servations, a substantial contribution from a disk-coronasystem cannot be ruled out.We thank the referee for the comments and suggestionsthat improved the clarity of the paper. MG acknowledgessupport by the RXTE Guest Investigator Program underNASA grant 200858. Funds from the NASA LTSA grantNAG5-10708 are also gratefully acknowledged.
REFERENCESAharonian, F., et al. 2006, Science, 314, 1424Albert, J., et al. 2007, ApJ, 669, 862Arnaud, K. 1996, in ASP Conf. Ser. 101, Astronomical Data AnalysisSoftware and Systems V, ed. G. Jacoby & J. Barnes (SanFrancisco: ASP), 17Baum,S.A., Zirbel, E.L., & O’Dea, C.P. 1995, ApJ, 451, 88Bennet, C.L. et al. 2003, ApJS, 148, 1Bevington, P.R. 1969. Data Reduction and Error Analysis for thePhysical Sciences. McGraw-Hill, New YorkBirkinshaw, M., & Worrall, D.M. 1993, ApJ, 412, 568Blazejowski, M., et al. 2005, ApJ, 630, 130B¨ottcher, M., Harvey, J., Joshi, M. et al. 2005, ApJ, 631, 169Chiaberge, M., Gilli, R., Capetti, A., & Macchetto, F.D. 2003, ApJ,597, 166Cui, W. 2004, ApJ, 605, 662Donato, D., Sambruna, R.M., & Gliozzi, M. 2004, ApJ, 617, 915Evans, D.A., Hardcastle, M.J., Croston, J.H., Worrall, D.M., &Birkinshaw, M. 2005, MNRAS, 359, 363Fanaroff, B.L. & Riley, J.M. 1974, MNRAS, 167, 31Ferrarese, L., & Ford, H.C. 1999, ApJ, 515, 583Fossati, G., et al. 2000, ApJ, 541, 153Fossati, G., Buckley, J., Edelson, R.A., Horns, D., & Jordan, M.2004, NAR, 48, 419Ghisellini, G. & Celotti, A. 2001, A&A, 379, L1Gliozzi M., Sambruna, R.M., & Eracleous, M. 2003, ApJ, 584, 176Gliozzi M., Sambruna, R.M., Brandt, W.N., Mushotzky, R., &Eracleous, M. 2004, A&A, 413, 139Gliozzi M., et al. 2006, ApJ, 646, 61Gliozzi M., Sambruna, R.M., Eracleous, M., Yaqoob, T. 2007, ApJ,664, 88Gonz´alez-Mart´ın, O., Masegosa, J., M´arquez, I., Guerrero, M.A., &Dulzin-Hacyan, D. 2006 A&A, 460, 45Guainazzi, M., Grandi, P., Comastri, A., & Matt, G. 2003, A&A,410, 131 Hardcastle, M.J., & Worrall, D.M. 2000, MNRAS, 314, 359Ho, L. 1999, ApJ 516, 672Jahoda, K., Swank, J., Giles, A.B., et al. 1996, Proc.SPIE, 2808, 59Jones, D.L. & Wehrle, A.E. 2002, ApJ, 580, 114Kataoka, J., et al. 2001, ApJ, 560, 659Markowitz, A. & Edelson, R. 2001, ApJ, 547, 684Markowitz, A. & Edelson, R. 2004, ApJ, 617, 939Marscher, A.P. 2006, ChJAS, 6, 262Melia, F., Liu, S., & Fatuzzo, M. 2002, ApJ 567, 811Mukherjee, R., Halpern, J., Mirabal, N. , & Gotthelf, E.V. 2002,ApJ, 574, 693Owen, F.N., & Laing R.A. 1989, MNRAS, 238, 357Papadakis, I.E., Petrucci, P.O., Maraschi, L., et al. 2002, ApJ, 573,92Press, W.H., Teukolsky, S.A., Vetterling, W.T., & Flannery, B.P.1997, Numerical Recipes (Cambridge: Cambridge Univ. Press)Rebillot, P.F., et al. 2006, ApJ, 641, 740Reynolds, C.S., Di Matteo, T., Fabian, A.C., Hwang, U., &Canizares, C.R. 1996, MNRAS, 283, 111;Rotschild, R.E., Blanco, P.R., Gruber, D.E., et al. 1998, ApJ, 496,538Sambruna, R.M., Eracleous, M., & Mushotzky, R. 1999, ApJ, 526,60Shuder,J.M. & Osterbrock, D.E. 1981, ApJ, 250, 55Turner, T.J., George, I.M., Nandra, K., & Mushotzky, R.F. 1997,ApJS, 113, 23Vaughan, S., Edelson, R., Warwick, R.S., & Uttley, P. 2003, MNRAS,345, 1271Xue, Y. & Cui, W. 2005, ApJ, 622, 160Yaqoob, T., et al. 2003, ApJ, 546, 759Zhang, Y.H., et al. 1999, ApJ, 527, 719Baum,S.A. & Zirbel, E.L. 1995, ApJ, 448, 521
Gliozzi et al.
Fig. 1.—
Top panel:
RXTE
PCA light curve in the 2–10 keV energy band, using PCU2 layer 1. Bottom panel: PCU2 layer 3 light curve.Time bins are 4 days. The dashed lines are the average count rate level.
Fig. 2.—
Layer 3 plotted versus the layer 1 count rate. The dashed line represents the best linear fit. -ray Spectral variability of NGC 6251 9
Fig. 3.—
Top panel: PCU2 light curve of NGC 6251 in the 2–10 keV range; the solid line represents the PCU0 light curve. Middle panel:PCU0 light curve of NGC 6251 in the 2–12 keV range; the solid line represents the PCU2 light curve. Bottom panel: PCU0/PCU2 lightcurve; the dashed line represents the average value of the ratio PCU0/PCU2. Time bins are 4 days.
Fig. 4.—
Top panel: Soft (2.5–5 keV) light curve of NGC 6251 using PCU2 data only. Middle panel: Hard (5–12 keV) light curve. Bottompanel: Hardness Hard/Soft ratio light curve; the solid line represents the average value of the hardness ratio. Time bins are 4 days. -ray Spectral variability of NGC 6251 11
Fig. 5.—
Hardness ratio (5–15 keV/2.5–5 keV) plotted versus the total count rate. The gray (light blue in color) small symbols correspondto the un-binned data points, whereas the darker, larger symbols refer to the weighted mean of data points in bins of fixed size of 0.05 c/salong the x-axis. The continuous line represents the best linear fit from the binned data, whereas the dashed lines are the ± σ departurefrom the best fit. Fig. 6.—
Fractional variability amplitude as a function of energy for NGC 6251. The error-bars along the x axis simply represent theenergy band width. The error bars along the y axis are computed following Vaughan et al. 2003. −3 no r m a li ze d c oun t s s − k e V − r a ti o Energy (keV)
Fig. 7.—
PCA spectrum of NGC 6251 during the third 2-month interval obtained using PCU2 data only. The model used is a simple powerlaw absorbed by Galactic N H . Fig. 8.—
Γ, obtained from the spectral analysis of flux-selected intervals, plotted against the average flux in the 2.5-12.5 keV band. Theerror-bars on Γ are 1- σ errors. The dashed line represents the best linear fit y = 3 . ± . − (0 . ± . x . For comparison, we have alsoplotted the values corresponding to the ASCA , BeppoSAX , and
XMM-Newton observations. To this end, we have used
PIMMS to convert theobserved flux into the
RXTE energy range, assuming the best-fit spectral parameters reported in Guainazzi et al. (2003) and Gliozzi et al.(2004). -ray Spectral variability of NGC 6251 13
Table 1Time-resolved Spectral results
Interval χ F −
10 keV
Γ EW
FeK α (10 − erg cm − s − ) (eV)1 0.60 4.1 2 . +0 . − . < . +0 . − . < . +0 . − . < . +0 . − . < . +0 . − . < . +0 . − . < Note:
All errors and upper limits refer to 90% confidence levels.
Table 2Flux-selected Spectral results
Individual Spectra Count Rate χ F . − . Γ EW
FeK α (s − ) (10 − erg cm − s − ) (eV)31 < .
25 0.53 2 . ± . . ± . < . − .
30 0.55 3 . ± . . ± . < . − .
35 0.50 3 . ± . . ± . < . − .
40 0.59 4 . ± . . ± . < > .
40 0.60 5 . ± . . ± . < Note:
The errors on Γ are 1 σσ