Mon. Not. R. . Soc. 324, 521±528 2001)

BeppoSAX observations of the Seyfert 2 galaxies NGC 7172 and ESO 103-G35

A. Akylas,1,2 I. Georgantopoulos1w andA. Comastri 3 1Institute of Astronomy & Astrophysics, National Observatory of Athens, I. Metaxa & B. Pavlou, Penteli, 15236, Athens, Greece 2Physics Department University of Athens, Panepistimiopolis, Zografos, 15783, Athens, Greece 3Osservatorio Astronomico di Bologna, Via Ranzani 1, I-40127 Bologna, Italy

Accepted2001 January 22. Received2001 January 8; in original form 2000 June 13

ABSTRACT We investigate the X-ray spectra of the type 2 Seyfert galaxies NGC 7172 andESO 103- G35, using BeppoSAX observations, separatedby approximately one year. We findthat the X-ray spectra of both NGC 7172 andESO 103-G35 can be well fittedusing a power-law model with an Fe Ka emission line at 6.4 keV. We didnot findany statistically significant evidence for the existence of a reflection component in the X-ray spectra of these two galaxies. The continuum flux has decreased by a factor of approximately 2 during this period, in both objects. However, the spectral index as well as the absorption column have remained constant. We find weak evidence for the decrease of the normalization of the Fe Ka emission line in a similar manner to the continuum in NGC 7172. We also report tentative evidence for a broad Fe Ka line in agreement with previous ASCA observations. In contrast, in the case of ESO 103-G35 the line flux does not change while its width remains unresolved. Key words: galaxies: active ± galaxies: individual: NGC 7172 ± galaxies: individual: ESO 103-G35 ± quasars: general ± galaxies: Seyfert ± X-rays: general.

studies can provide strong constraints in understanding the 1 INTRODUCTION geometry of the circumnuclear matter in these objects. For In recent years X-ray missions such as EXOSAT and ASCA example, the time lags between the power-law continuum andthe have demonstrated that Seyfert 2 galaxies have a very complex normalization of the Fe line couldconstrain the origin of the line. X-ray spectrum see Mushotzky, Done & Pounds 1993 for a recent In the case where the Fe line originates in an accretion disc as is review). In particular their spectrum contains at least the following the standard scenario for Seyfert 1 galaxies), the Fe line should features: i) a power-law continuum with slope G , 1:9 absorbed closely track the variations of the intrinsic power law. by a very large column .1023 cm22) e.g. Turner & Pounds In this paper we present an analysis of four observations of the 1988); this obscuring screen is probably associatedwith a Seyfert 2 galaxies NGC 7172 andESO 103-G35 from the molecular torus, i) an Fe K line at 6.4 keV ie originating from BeppoSAX archive. The two observations for each galaxy are neutral Fe, i) a spectral curvature hump) above 10 keV; this is separatedby an interval of one year. Our goal is to studyany long- generally believedto originate from reflection of the power-law term variations in the spectral features between the two epochs. component on coldmaterial e.g. Lightman & White 1988; The broadenergy bandof BeppoSAX, as well as its goodspectral George & Fabian 1991). resolution, provides the opportunity to place stringent constraints In contrast to the recent progress in understanding the spectral on the geometry of the nuclei of ESO 103-G35 andNGC 7172 and features of Seyfert 2 galaxies, our knowledge of their X-ray any surrounding gaseous media. variability properties remains scanty. Usually, this has been based on comparisons of the X-ray flux between observations from different missions spanning time intervals of several years. 2THESAMPLE Recently, Turner et al. 1997) have detected some evidence for 2.1 NGC 7172 short-term variability in a handful of Seyfert 2 galaxies using ASCA. The systematic monitoring of Seyfert 2 galaxies has become NGC 7172 is an obscured, almost edge-on galaxy. It belongs to the possible with the RXTE mission. RXTE observations of a few compact group HCG 90 andhas a redshiftof z ˆ 0:0086: It is objects revealedstrong variability in time-scales of hours as well classifiedas a type 2 Seyfert galaxy. Strong infraredemission as evidence for spectral variability Georgantopoulos et al. 1999, from the nucleus is observed, which is variable on time-scales of Georgantopoulos & Papadakis 2001). Indeed, time variability approximately three months Sharples et al. 1984). The first X-ray observation which was performedby EXOSAT showeda G ˆ 1:84 w 23 22 E-mail: [email protected] power law absorbedby NH , 10 cm Turner & Pounds 1989). q 2001 RAS 522 A. Akylas, I. Georgantopoulos and A. Comastri

Later observations with Ginga confirmedthe above results and previous observations. They further detect short-term variability also indicated that a reflection component may be important within a day) of about 30 per cent. Until 1995, before the Smith & Done 1996). Furthermore, the analysis of ASCA BeppoSAX observations, the flux of NGC 7172 hadremained observations Turner et al. 1997) showedthe presence of a 6.4- almost constant between ,3 and ,5  10211 erg cm22 s21: Since 136 keV emission line with an equivalent width EW) of 68235 eV: In May 1996 the source has reacheda very low flux state of contrast with the previous results they founda rather hardphoton ,1  10211 erg cm22 s21: The flux variability history of NGC index, G , 1:7; while the amount of the absorption column 7172 is summarizedin Fig. 1. remainedconstant. Guainazzi et al. 1998), using two ASCA observations separatedby one year, foundan even flatter photon index of G , 1:5 absorbedby the same column densityfoundin 2.2 ESO 103-G35 The HEAO A2 hardX-ray source 1H 1832 2653 was identified as the S0/Sa type galaxy ESO 103-G35, with a redshift of z ˆ 0:013; by Phillips et al. 1979). They classifiedthe active galaxy nucleus AGN, as a type-1.9 Seyfert galaxy because of the weakness of the Hb emission line andthe presence of a broadH a line. EXOSAT 10:34 observations showeda steep photon indexof G ˆ 1:9220:34 23 22 absorbedby a high column densityof N H . 10 cm ; which was foundto vary by within factor of 2 in a periodof 90 days, possibly as a result of the motion of material in the broadline region across the line of sight Warwick, Pounds & Turner 1988). Later observations obtainedby Ginga strongly suggestedthe presence of a narrow emission line at 6.4 keV andalso of a 10:08 reflection component with a photon index G ˆ 2:1920:08 Smith & Done 1996). ASCA observations Turner et al. 1997) confirmed the previous EXOSAT results andshowedstrong evidencefor the presence of a triplet Fe line 6.4, 6.68, 6.98 keV). They also found a decrease of a factor of two in the continuum flux as compared to Ginga. Finally, Forster, Leighly & Kay 1999), analyzing three ASCA observations spanning a periodof ,2yr 1994±1996†; founda decreasein the EW of the Fe line while the continuum flux increasedby a factor of 2. An edgeat 7.4 keV was also found in the 1996 observation. The long-term variability history of ESO Figure 1. Long term X-ray variability curve for the 2±10 keV absorbed 103-G35 is summarizedin Fig. 2. continuum flux of NGC 7172 including HEAO-1 filledcircles), EXOSAT filledsquares), Ginga filledtriangles), ASCA crosses), and BeppoSAX open circles) observations. 3 OBSERVATIONS The scientific instrumentation on boardthe Italian±Dutch X-ray BeppoSAX includes a medium-energy concentrator spectrometer MECS) which consists of three units Boella et al. 1997): a low-energy concentrator spectrometer LECS; Parmar et al. 1997), a high-pressure gas scintillation proportional counter HPGSPC; Manzo et al. 1997) anda Phoswich detectorsystem PDS; Frontera et al. 1997), all of which point in the same direction. The MECS instrument consists of a mirror unit plus a gas scintillation proportional counter andhas imaging capabilities. It covers the energy range between 1±10 keV with a spatial resolution of about 1.4 arcmin at 6 keV anda spectral resolution of 8 per cent at 6 keV. The LECS instrument is similar to the MECS andoperates in the 0 :1±10 keV bandwith the same spectral and spatial resolution of the MECS. The PDS is a direct-view detector with rocking collimators andextendsthe BeppoSAX bandpass to high energies 13±300 keV†: Its energy resolution is 15 per cent at 60 keV. In our analysis we use MECS andPDS dataonly. Our objects have not been detected by the HPGSPC detector. We did not use the LECS data as there are nearby contaminating soft X-ray sources, probably stars in the case of ESO 103-G35, while in the case of NGC 7172 the contamination is causedby the Figure 2. Long-term X-ray variability curve for the 2±10 keV absorbed presence of the Hickson compact group H90. These sources are continuum flux of ESO 103-G35 including HEAO-1 filledcircles), not present in the hardMECS images. Another reason for EXOSAT filledsquares), Ginga filledtriangles), ASCA crosses), and excluding the LECS data was the poor photon statistics see BeppoSAX open circles) observations. Tables 1 and2).

q 2001 RAS, MNRAS 324, 521±528 BeppoSAX observations of Seyfert 2galaxies 523

Table 1. BeppoSAX observations of NGC 7172.

Instrument Exposure time sec) count rate

1996 October 15±16

LECS 15490 0.038 ^ 0.002 MECS1,2,3) 39172 0.148 ^ 0.002 PDS 17277 0.431 ^ 0.051

1997 November 6±7

LECS 22896 0.022 ^ 0.001 MECS1,2) 49437 0.054 ^ 0.001 PDS 21146 0.292 ^ 0.047

Table 2. BeppoSAX observations of ESO 103-G35.

Instrument Exposure time sec) count rate

1996 October 3±4

LECS 10238 0.076 ^ 0.003 MECS1,2,3) 50619 0.308 ^ 0.003 PDS 21029 0.746 ^ 0.047 Figure 3. Backgroundsubtractedlight curves of NGC 7172 observations 1996±upper panel; 1997±bottom panel) for the MECS instrument 1997 October 14±15 1:65±10 keV†: The binning time is 3 ks. LECS 3699 0.04 ^ 0.004 MECS1,2) 14347 0.123 ^ 0.003 PDS 5915 0.579 ^ 0.087

Here we analyse two observations of NGC 7172 andESO 103-G35, both obtainedwith BeppoSAX within an interval of about 1 yr. More specifically, NGC 7172 was first observedin 1996 from October 15th to 16th) andthen in 1997 from November 6th to 7th). ESO 103-G35 was first observedin 1996 from October 3rdto 4th) andthen in 1997 from October 14th to 15th). Tables 1 and2 list the exposures times andnet count rates backgroundsubtracted)for all four observations. The three different or two after 1997 May 7, when MECS1 failed) MECS units are mergedtogether in orderto increase the signal-to-noise ratio. This is feasible because the three MECS units show very similar performance andthe difference in the position of the optical axis in the three units is smaller than the scale on which the vignetting of the telescopes varies significantly .5 arcmin).

4TIMINGANALYSIS In Figs 3 and4 we present the light curves for the MECS instrument for each observation separately, using a binning time of Figure 4. Light curves backgroundsubtracted)of ESO 103-G35 3 ks the errors correspondto the 68 per cent confidence level). observations 1996±upper panel; 1997±bottom panel) for the MECS When we perform a constant fit, the resulting x 2 implies short- instrument 1:65±10 keV†: The binning time is 3 ks. term variability in both objects. In particular, for the first observation of NGC 7172 we findthat the best-fit average count about 30 per cent, on time-scales of hours in a 1996 ASCA rate 0:147 ^ 0:007† gives a x2 ˆ 36 for 24 degrees of freedom, observation of NGC 7172. Furthermore, Forster et al. 1999), dof), while for the second observation we find 0:052 ^ 0:005 with using ASCA data, found statistically significant short-term x2=dof ˆ 64=33: In the case of ESO 103-G35 the best fit constant variability in a 1996 observation of ESO 103-G35. is 0:305 ^ 0:004 corresponding to x2=dof ˆ 72=29 for the first The variability amplitude was estimated in all four observations ^ 2 observation while for the secondone we find0 :122 0:004 with by means of the excess variance srms for a definition of this x2=dof ˆ 10=8: The above x 2 suggests the presence of statisti- quantity see Nandra et al. 1997). In Table 3 we present the values . 2 cally significant  3s) short-term variability in the 1997 and of srms andthe unobscuredluminosity for each observation. 1996 observations of NGC 7172 andESO 103-G35 respectively. Unfortunately a comparison with the results foundby Nandraet al. Our results are compatible with previous ASCA findings. More 1997) for a sample of Seyfert 1 galaxies is not straightforward. specifically, Guainazzi et al. 1998) detected a flux variability of This is because of the different exposure and binning times we q 2001 RAS, MNRAS 324, 521±528 524 A. Akylas, I. Georgantopoulos and A. Comastri

2 Table 3. The values of the variability amplitude, srms; and the unobscuredluminosity for each observation.

2 21 Object Date srms Luminosity erg s ) 2±10 keV NGC 7172 1996 0.0045 3.5 Â 1042 NGC 7172 1997 0.018 1.7 Â 1042 ESO 103-G35 1996 0.0029 2.7 Â 1043 ESO 103-G35 1997 0.004 1.5 Â 1043 have used. For example, our large binning time has the effect to 2 suppress the srms value. Indeed our values are slightly lower than the values quotedin Nandraet al. 1997; see their fig. 4).

5SPECTRALANALYSIS We usedan extraction radiusof 4 arcmin. The above area encircles more than 85 per cent of the photons in the 2±10 keV energy band. The spectrum of the backgroundwas estimatedfrom source-free regions of the image. In particular, we consider an annulus with inner radius 6 arcmin and outer radius 12 arcmin for the MECS instrument. We usedthe rise-time threshold,backgroundsub- tractedPDS spectra releasedfrom the BeppoSAX archive. We use data from the energy ranges 1:65±10 keV and15±50 keV for the MECS andPDS detectorsrespectively, where the response matrices are well calibrated. We do not consider the PDS data above 50 keV as in this energy range the signal-to-noise ratio is low. The MECS spectral files were linearly rebinnedto give a minimum of 20 counts per channel while the PDS spectral files were logarithmically binnedin 18 channels. The spectral fitting was carriedout using the xspec v10 software package. All errors Figure 5. Data, folded model simple absorbed power-law model) and quotedin the best-fit spectral parameters correspondto the 90 per residuals for both 1996 upper panel) and 1997 lower panel) observations cent confidence level. The MECS and PDS data were fitted of NGC 7172. simultaneously. A relative normalization factor was introduced between the PDS andMECS dataspectrum, because the assumedthat both the reflection component andthe power law are BeppoSAX instruments show some mismatches in the absolute absorbedby the same column density. We fix the pexrav flux calibration. We assumeda PDS-to-MECS normalization normalization to that of the power law. We also arbitrarily fix the factor of between 0.77 and0.95 Fiore, Guainazzi & Grandi inclination angle of the plane of the reflecting material) to 1999). i ˆ 608, because the shape of the reflection spectrum in our energy band is relatively independent of the inclination angle. The cut-off energy of the incident power law was set at 300 keV. Finally, the 5.1 NGC 7172 31996) Fe andlight element abundanceswere kept fixedat the First we fit a simple model consisting of a power law absorbed by abundance values. We find that a reflection component is not 2 a neutral column NH Fig. 5). This provides a good fit to the data statistically significant Dx , 1:5 for one additional parameter). x2 ˆ 145 for 146 dof). The addition of a broad Gaussian line However, the rather uncertain value of the reflection parameter 2 10:7 significantly improves the fit Dx ˆ 10†: The best-fit line energy R ˆ 1:220:9† andthe steeper spectral photon index of G ˆ 10:6 10:26 was 6:020:6 keV: As this value is consistent with 6.4 keV, which 1:7820:19; which is closer to the canonical AGN spectral index indicates a neutral or weakly ionized Fe e.g. Nandra et al. 1997), Nandra & Pounds 1994), suggest that a reflection component we fix it at 6.4 keV in the rest of the analysis. The line width is cannot be ruledout. The fraction of the intrinsic reflectedflux 10:3 sKa ˆ 0:820:4 keV; so it is marginally resolvedgiven the energy relative to the total in the 2±50 keV bandis 0.25. resolution of the MECS. When we fix the line width at 0.1 keV we obtain an increase of Dx2 ˆ 2:8: The best-fit parameters are presentedin Table 4. 5.2 NGC 7172 31997) In Fig. 5 residuals can be seen above 7 keV. We therefore attempted to include an absorption edge at ,7.2 keV. We found Similar to the 1996 observation of NGC 7172, an absorbedpower- that the addition of the edge significantly improves the fit Dx2 ˆ law model G , 1:5† provides a good fit Dx2 ˆ 112 for 109 dof; 3:4 for the addition of one parameter). The optical depth of the see Fig. 5). The addition of a Gaussian line improves the fit 10:11 10:12 2 10:26 edge is 0:1820:12: The rather flat spectral index G ˆ 1:6420:09 significantly Dx ˆ 6†: The line energy is 6:7420:42 keV, which is couldsuggest that a reflection continuum is present. Therefore, we consistent with neutral Fe, andtherefore we hereafter fix the line further tried to add a reflection component to our model. We used energy at 6.4 keV. The range of the line width is 0:08±1 keV: The the pexrav model in xspec Magdziarz & Zdziarski 1995). We best-fit parameters are shown in Table 4. An inspection of the

q 2001 RAS, MNRAS 324, 521±528 BeppoSAX observations of Seyfert 2galaxies 525

Table 4. Best-fitting parameters for the two observations of NGC 7172.

2 Date NH G Apl Energy s Aga EW x /dof 1022 cm22) 1023 cts s21 keV21) keV) keV) 1025 cts s21) eV)

10:9 10:12 11:2 10:3 12:5 1210 96-10-15 9:020:7 1:6420:09 4:121:1 6.4 0:820:4 5:622:4 2802160 135/144 10:70 10:08 10:5 10:68 11:1 1195 97-11-6 8:320:65 1:5320:16 1:720:5 6.4 0:3220:24 1:521:0 1752140 106/107

Figure 6. Fe line flux versus power-law normalization contours for the 1996 and1997 observations of NGC 7172 in the joint fits. Confidence levels are 68, 90 and99 per cent for two interesting parameters. residuals shows that there is no evidence for the presence of an absorption edge. However, we attempt to include an edge at 7.25 keV, basedon the previous 1996 results. We foundthat there is no improvement in the fit Dx2 , 1†: The optical depth of the 10:15 10:08 edge is 0:0820:08: The rather flat spectral index G ˆ 1:5320:16 couldagain suggest the presence of a reflection component. However, when we try to include such a component, we find that, although it does not improve the fit Dx2 ˆ 1:5 for one additional 10:60 parameter), the photon index becomes steeper G ˆ 1:8120:40† and 11:3 the reflection parameter is R ˆ 2:521:8: The unabsorbedreflected flux is 40 per cent of the total flux in the 2±50 keV energy band. Figure 7. Data, folded model simple absorbed power-law model) and Comparison between the spectral fits in the 1996 and1997 residuals for both 1996 upper panel) and 1997 lower panel) observations of ESO 103-G35. observations reveals that there is no change in the photon index or column density within the errors. Finally, we perform joint fits in order to investigate any possible 5.3 ESO 103-G35 31996) variability in the normalization of the different spectral compo- nents. We fit both NGC 7172 observations with an absorbedpower A poor fit x2 ˆ 240 for 173 dof) is obtained with a single power- 10:07 10:3 22 22 law model G ˆ 1:6220:09 andN H ˆ 9:120:6  10 cm † plus a law model, emphasizing the importance of an Fe emission line. Gaussian component at 6.4 keV to account for the Fe line. The This is again evident in Fig. 7, where we plot the absorbed power- 10:4 width of the line is 0:620:5 keV andtherefore remains uncon- law model and residuals. Indeed, when we include a Gaussian strained. Comparison of the two observations reveals weak line, the x 2 is reduced by 66. The Fe line has an energy of 10:09 evidence for spectral variability. The power-law normalization 6:3720:10 keV consistent with neutral Fe andtherefore hereafter 10:1 has decreased by about a factor of two between the 1996 and 1997 we fix it at 6.4 keV. Its width is s ˆ 0:320:1 keV andtherefore the observations. This may be followedby a similar change in the Fe line remains essentially unconstrained. The photon index, G ˆ 10:06 line normalization. This is evident in Fig. 6 where we present the 1:8720:09; is consistent with the typical value of 1.9 for Seyfert 68, 90 and99 per cent confidencecontours for the Fe line against galaxies Nandra & Pounds 1994). The best-fit spectral parameters the power-law normalization. The contours were extracted are presentedin Table 5. We further investigate the existence of 1140 using the above best fit model. The EW are 2802130 and excess Fe absorption near 7.5 keV following the claims of 1190 2332160 eV for the 1996 and1997 observations, respectively. Warwick et al. 1993). Although there is some hint for an Fe However, we note that when we fix the width of the line at 0.1 keV edge in Fig. 7, we find no significant improvement when we this increases x 2 by 2.5) we do not detect any significant include an edge component Dx2 ˆ 1:5 for two additional 10:65 variability in the normalization of the line. parameters). The edge is found at 7:3520:35 keV; with optical q 2001 RAS, MNRAS 324, 521±528 526 A. Akylas, I. Georgantopoulos and A. Comastri

Table 5. Best-fitting parameters for the two observations of ESO 103-G35.

2 Date NH G Apl Energy s Aga EW x /dof 1022 cm22) 1023 cts s21 keV21) keV) keV) 1025 cts s21) eV)

10:9 10:06 12:7 10:12 15:0 1150 96-10-3 18:621:0 1:8720:09 18:523:3 6.4 0:320:12 1524:0 2652100 174/171 11:4 10:09 12:6 10:24 17:0 1500 97-10-14 19:123:4 1:8120:30 9:624:5 6.4 0:1620:16 10:525:4 290290 56/69

consisting of an absorbedpower-law plus a Gaussian line at 10:09 6.4 keV. The best-fit photon index is 1:8320:09 andthe column 10:2 22 22 density is 18:520:8 Â 10 cm : The line is unconstrained, i.e. the 1120 range of the line width is 0.15±0.36 keV. The EWs are 220250 1170 and330 2140 eV for the 1996 and1997 observations respectively. The continuum flux is reduced approximately by a factor of 2. However, there is no evidence that this variation is followed by the line flux, in contrast to the case of NGC 7172 see Fig. 8 where we plot the 68, 90 and99 per cent joint contours for the power law andFe line normalization).

6 DISCUSSION 6.1 NGC 7172 The spectral analysis of both observations implies that NGC 7172 10:3 22 22 is absorbedby a column densityof N H ˆ 9:120:6  10 cm ; in goodagreement with previous EXOSAT Turner & Pounds 1989), Figure 8. Fe line flux versus power-law normalization contours for the Ginga Smith & Done 1996) and ASCA Turner et al. 1997) 1996 and1997 observations of ESO 103-G35 in the joint fits. Confidence observations. The data require the presence of an Fe line at levels are 68, 90 and99 per cent for two interesting parameters. 6.4 keV. In the 1996 BeppoSAX observation the line is marginally resolved, at less than the 2s confidence level, while in the 1997 10:08 depth 0:120:10: Finally, basedon the ASCA results Turner et al. observation where the photon statistics are poorer) the line width 1997) we triedto includetwo emission lines at 6.68 and6.96 keV is unconstrained. We note that the line has been resolved in ionizedFe) but the reductionin x 2 was less than 1. A reflection previous ASCA observations Guainazzi et al. 1998). There is component is not statistically significant because the reduction of some evidence Fig. 6) for a variation in the normalization of the x 2 is very low Dx2 ˆ 1:1 for one additional parameter). Note that line. This is in agreement with Guainnazi at al 1998) who the power law spectral index in the case of a reflection component presentedsimilar evidencefor a variation of the line basedon two 10:2 10:4 is G ˆ 2:020:2 while R ˆ 1:220:6: The intrinsic flux of the ASCA observations 1995 and1996). The line variations in reflection component corresponds to the 20 per cent of the total conjunction with the line width, has important implications for the flux in the 2±50 keV energy band. geometry of the circumnuclear matter in NGC 7172. Indeed, the Fe line must originate close to the nucleus, possibly in the outer regions of an accretion disc. This would simultaneously explain 5.4 ESO 103-G35 31997) the fact that the continuum flux variations are trackedby changes An acceptable fit is obtainedwhen we use the absorbedpower-law in the line flux, and the large width of the line. Indeed, the line model with x2 ˆ 70 for 71 dof see Fig. 7). An Fe K emission line broadening may be caused by the large velocities in the disc, is highly significant, yielding Dx2 ˆ 14: The Fe line energy is similar to that witnessedin MCG-6-30-15 by Tanaka et al. 1995). 10:10 6:520 15 keV andtherefore we fix it at 6.4 keV again. The line is The observedbroadeningin the case of NGC 7172 implies a : 2 unresolved: its width range is 0±0:4 keV: The photon index, G ˆ velocity of ,70 000 km s 1. We triedto investigate further the 10:09 1:8120:30; is consistent with the previous 1996 observation. The above hypothesis by including a line from a relativistic accretion results are presentedin Table 5. A reflection component cannot be disc for an accretion disc viewed at an intermediate inclination 12:0 ruledout as the reflection parameter is R ˆ 1:621:2: However, the angle with solar abundances). The fit was only marginally reduction in x 2 is very low Dx2 ˆ 1:7 for one additional improved Dx2 , 1†: We note that Georgantopoulos & Papadakis parameter). When the reflection component is included, the power 2001) findno significant variation of the normalization of the line 10:30 law spectral index steepens to G ˆ 1:9720:30: The ratio between in NGC 7172 using several RXTE observations spanning a period the unabsorbedreflectedflux andthe total flux in the 2±50 keV of arounda week. This wouldset a lower limit of about a light energy bandis 0.3. Next, we investigate the existence of an Fe week to the size of the line-emitting region. edge at 7.35 keV. However, the reduction in x 2 was less than 1 for The observedEW cannot be interpretedin terms of transmis- 10:10 an additional parameter. The optical depth of the edge is 0:1120:11: sion processes in the torus alone. Indeed, Leahy & Creighton In general, we detect no spectral variability between the 1996 and 1993), using Monte Carlo simulations, foundthat a column 22 22 1997 observations in the sense that both the absorption column density of NH , 9  10 cm couldproducean Fe line EW of andthe photon indexremain constant. only ,60 eV. This independently suggests that a reflection We perform a joint fit in the 1996 and1997 observations in component may be present. Indeed, in this case the EW of the order to detect any possible line flux variations. We used a model line shouldbe increasedby ,150 eV as a result of reflection on

q 2001 RAS, MNRAS 324, 521±528 BeppoSAX observations of Seyfert 2galaxies 527 the accretion disc Reynolds, Fabian & Inoue 1995; George & originates possibly far away from the central region, the EW Fabian 1991). Alternatively, the observedEW may be explained varies accordingly to the variations of the continuum. Although a on the basis of a high-Fe abundance. Indeed, excess absorption comparison is not straightforward, it is possible that a large from Fe is present in the first observation of NGC 7172 at fraction of the line emission couldoriginate from transmission ,7.2 keV, confirming earlier reports by Warwick et al. 1993); processes. however, note that superior spectral resolution observations with Some residuals around 7.5 keV see Fig. 7) may indicate the the ASCA SIS Turner et al. 1997) do not corroborate this result. presence of an absorption edge due to ionized Fe. This was first The edge energy is consistent with a neutral or low ionization suggestedby Ginga results Warwick et al. 1993) andwas material surrounding the source Kallman & McCray 1982). The confirmedlater with ASCA observations Forster et al. 1999). The , 10:05 optical depth of the edge corresponds to an absorption column of power-law photon index has a steep value of 1:8520:05 which is 115:3 22 22 22:5214:2  10 cm : This large column density may suggest an consistent with the typical value 1.9 for Seyfert galaxies Nandra Fe abundance well above the solar value. If this is the case, the & Pounds 1994). Despite our large energy bandpass we do not observedEW couldbe explainedby transmission processes alone. findstrong evidencefor a reflection continuum in contrast to Finally, note that the cases where a strong Fe line is detected, previous Ginga Smith & Done 1996) and RXTE Georgantopou- despite a lack of strong evidence for a reflection component, are los & Papadakis 2001) observations who found evidence for such not uncommon e.g. Gilli et al. 2000; Reynolds et al. 1995). a component but in a smaller band 3±20 keV†: Higher signal-to- The power-law photon index has a rather flat value of ,1.6 noise observations are necessary in order to test the presence of while it remains constant in the two BeppoSAX observations. The such a component. value of the photon index is in agreement with previous ASCA observations Guainazzi et al. 1998; Ryde et al. 1997) who found a photon index of 1:47 ^ 0:15: However, it is clearly flatter than 6.3 Spectral variability in Seyfert 2 galaxies that foundin EXOSAT Turner et al. 1989) and Ginga Smith & Previous studies of spectral variability in Seyfert 2 galaxies have Done 1996) G , 1:8†: Therefore the above may be suggesting shown some interesting results. In particular, Iwasawa et al. long term spectral variability Ryde et al. 1997). For example, 1994) andGriffiths et al. 1998) analyzednon-simultaneous assuming that the X-ray emission comes from Comptonization of Ginga, ASCA, ROSAT and BBXRT observations to investigate the ultraviolet UV) photons from an accretion disc on a hot electron spectrum of the Seyfert 2 galaxy Mrk 3. Their analysis demon- corona Rybicki & Lightman 1979), a steepening of the spectral stratedthat the Fe line doesshow some variability on time-scales index may be attributed to a change of the temperature or optical of a few years. Systematic monitoring observation became depth of the corona. Of course it is difficult to make a detailed feasible with the RXTE mission. In particular, Georgantopoulos comparison because of the differences in the energy bandpass of et al. 1999), Georgantopoulos & Papadakis 2001) and Smith, the various instruments as well as in the spectral analysis: for Georgantopoulos & Warwick 2001) present monitoring observa- example, the addition of a reflection component would result in tions of several Seyfert 2 galaxies Mrk3, ESO 103-G35, IC 5063, the steepening of the spectral index. However, we note that there is NGC 4507 and7172, andMrk 348) spanning time-periodsfrom no correlation between the flux level andthe photon index about seven days to seven months. In most cases, they detect observedin the case of the NGC 7172 observations in the past spectral variability in the sense that the column density decreases 20 yr. Finally, we note that despite the flat photon index found with increasing flux. However, they findno evidencefor Fe line above, our data do not significantly support the existence of any variability. Here instead, we find tentative evidence for line reflection continuum. A recent analysis of the RXTE monitoring variability in NGC 7172 in a periodof 1 yr confirming previous observations of NGC 7172 Georgantopoulos & Papadakis 2001) ASCA results Guainazzi et al. 1998). To our knowledge this is one again shows no evidence for a reflection component. of the very few examples of Fe line variability in Seyfert 2 galaxies. Our finding requires that at least one of the line-emitting 6.2 ESO 103-G35 regions has dimensions of not much more than about a light year. It wouldseem more likely that this scale size correspondsto an The spectral analysis of the two observations of ESO 103-G35 accretion disc or inner-cloud structure rather than the inner extent implies that a simple power-law model absorbed by a column of the putative molecular torus. Indeed, the molecular torus is density of N , 2  1023 cm22 plus a Gaussian line at 6.4 keV H probably a much greater structure. At least in the case of NGC provides a good fit to our data. EXOSAT observations of ESO 103- 1068, observations of molecular hydrogen Tacconi et al. 1994) G35 Warwick et al. 1988) reveal a 50 per cent decrease in the showedthat the obscuring torus is locatedbetween 30 and300 pc. column density over a 90-d period. This variation was interpreted in terms of an X-ray absorbing screen composedof broad-line clouds moving around the X-ray source. Alternatively, the N H 7 CONCLUSIONS variation couldbe attributedto photoionization of the neutral column as the flux increases; in this case an ionizedFe edge We have analysedfour BeppoSAX observations of NGC 7172 and should be detected. However, in our analysis we detect no ESO 103-G35 two observations for each galaxy separatedby variation of the column within the errors. An Fe line at 6.4 keV is about 1 yr). The goal was to search for spectral variability in a strongly needed to fit our data. Unfortunately, the width of the line large energy bandandtherefore to attempt to constrain the remains unconstrained. However, ASCA SIS observations with geometry of the circumnuclear matter in these two galaxies. The better spectral resolution have probably resolvedthe line into three spectra of both galaxies are fittedby an absorbedpower law plus a components at 6.4, 6.68 and6.96 keV Turner et al. 1997). Monte neutral Fe line in agreement with previous Ginga and ASCA Carlo simulations Leahy & Creighton 1993) show that a column observations. A neutral Fe edge is probably detected in the 1996 23 22 density of NH , 2  10 cm couldproducean Fe line EW of observation of NGC 7172. Our data do not require a reflection ,170 eV. However, in the case of ESO 103-G35 where the line component at a statistically significant level. We detected a flux q 2001 RAS, MNRAS 324, 521±528 528 A. Akylas, I. Georgantopoulos and A. Comastri variability in both objects of a factor of about two. We findno Fiore F., Guainazzi M., Grandi P., 1999, Handbook for BeppoSAX NFI significant evidence for spectral variability in the sense that the spectral analysis. ftp://www.sdc.asi.it/pub/sax/docs/saxabc_v1.2.ps.gz power-law spectral index and the column density have remained Forster K., Leighly K. M., Kay L. E., 1999, ApJ, 523, 521 constant in both galaxies, implying a homogeneous obscuring Frontera F., Gosta E., dal Fiume D., Feroci M., Nicastro L., Orlandini M., screen and constant physical conditions in the accretion disc Palazzi E., Zavattini G., 1997, A&AS, 122, 357 Georgantopoulos I., Papadakis I., 2001, MNRAS, 322, 218 corona despite the continuum flux variations but see Warwick Georgantopoulos I., Papadakis I., Warwick R. S., Smith D. A., Stewart et al. 1988 andGeorgantopoulos & Papadakis2001). However, in G. C., Griffiths R. G., 1999, MNRAS, 307, 815 the case of NGC 7172 we detect weak criteria for a variation of George I. M., Fabian A. C., 1991, MNRAS, 249, 352 the Fe line flux by a factor of two following the power-law flux, Gilli R., Maiolino R., Risaliti G., Dadina M., Weaver K. A., Colbert confirming previous ASCA results. If true, this provides strong E. J. M., 2000, A&A, 355, 485 support to a scenario where the Fe line, at least in the case of NGC Griffiths R. G., Warwick R. S., Georgantopoulos I., Done C., Smith D. A., 7172, originates in a region close to the nucleus. In this case, the 1998, MNRAS, 298, 1159 region responsible for the line emission shouldlie within one light Guainazzi M., Matt G., Antonelli L. A., Fiore F., Piro L., Ueno S., 1998, year away from the nucleus. This region may be associatedwith MNRAS, 298, 824 the outer part of an accretion disc. Iwasawa K., Yaqoob T., Hisamitsu A., Yasushi O., 1994, PASJ, 46, L167 Kallman T. R., McCray R., 1982, ApJS, 50, 263 The above results clearly emphasize the strength of monitoring Leahy D. A., Creighton J., 1993, MNRAS, 263, 314 observations in probing the geometry of the central engine in Lightman A. P., White T. R., 1988, ApJ, 335, 57 AGN. High resolution spectroscopic observations with gratings Magdziarz P., Zdiarski A. A., 1995, MNRAS, 273, 837 with both Chandra and XMM are expectedto revolutionize our Manzo G., Giarrusso S., Santangelo A., Ciralli F., Fazio G., Piraino S., knowledge of the physical conditions of the circumnuclear matter Segreto A., 1997, A&AS, 122, 341 in AGN. These combinedwith future monitoring observations Mushotzky R. F., Done C., Pounds K. A., 1993, ARA&A, 31, 717 with large effective area instruments such as EPIC on board XMM Nandra K., Pounds K. A., 1994, MNRAS, 268, 405 are expected to provide for the first time a detailed mapping of the Nandra K., George I. M., Mushotzky R. F., Turner T. J., Yaquoob T., 1997, central regions in AGN. ApJ, 476, 70 Parmar A. N. et al., 1997, A&AS, 122, 309 Phillips M. M., Feldamn F. R., Marchall F. E., Wansteker W., 1979, A&A, NOTE ADDED IN PROOF 76, L14 Reynolds C. S., Fabian A. C., Inoue R., 1995, MNRAS, 276, 1311 After the present work was accepted, two preprints by Dadina et Rybicki G. B., Lightman A. P., 1979, Ratiative processes in astrophysics. al. 2001) andWilkes et al. 2001) came out that discussedour John Wiley andSons, New York data sets. Their results are in general consistent with ours. Ryde F., Poutanen J., Svensson R., Larsson S., Ueno S., 1997, A&A, 328, 69 Sharples R. M., Longmore A. J., Hawarden T. G., Carter D., 1984, MNRAS, 208, 15 ACKNOWLEDGMENTS Smith D. A., Done C., 1996, MNRAS, 280, 355 We thank the anonymous referee for many useful comments and Tacconi L. J., Genzel R., Blietz M., Harris A. I., Madden S., 1994, ApJ, suggestions. This project was funded by a Greek±Italian scientific 426, L77 Tanaka . et al., 1995, Nat, 375, 659 collaboration under the title `Observations of active galaxies with Turner T. J., George I. M., Nandra K., Mushotzky R. F., 1997, ApJS, 113, the Italian astrophysics mission BeppoSAX'. AA is grateful to the 23 Bologna observatory group for their warm hospitality. We are Turner T. J., Pounds K. A., 1989, MNRAS, 240, 833 grateful to M. Cappi, C. Vignali andR. Lazio for many useful Warwick R. S., Pounds K. A., Turner T. J., 1988, MNRAS, 231, 1145 discussions. Warwick R. S., Sembay S., Yaqoob T., Makishima K., Ohashi T., Tashiro M., Kohmura Y., 1993, MNRAS, 265, 412 Wilkes B., Mathur S., Fiore F., Autonelli A., Nicastro F., 2001, ApJ, 549, REFERENCES 248 Boella G. et al., 1997, A&AS, 122, 327 Dadina M. et al., 2001, A&A, 370, 70 This paper has been typeset from a TEX/LATEX file preparedby the author.

q 2001 RAS, MNRAS 324, 521±528