THE ASTROPHYSICAL JOURNAL, 532:883È894, 2000 April 1 ( 2000. The American Astronomical Society. All rights reserved. Printed in U.S.A.

HUBBL E SPACE T EL ESCOPE OBSERVATIONS AND PHOTOIONIZATION MODELING OF THE LINER GALAXY NGC 10521 J. R. GABEL,F.C.BRUHWEILER,D.M.CRENSHAW,S.B.KRAEMER, AND C. L. MISKEY Institute for Astrophysics and Computational Sciences, Department of Physics, The Catholic University of America, Washington, DC 20064; gabel=iacs.gsfc.nasa.gov Received 1999 May 20; accepted 1999 December 18

ABSTRACT We present a study of available Hubble Space Telescope (HST ) spectroscopic and imaging obser- vations of the low-ionization nuclear emission-line region (LINER) galaxy NGC 1052. The WFPC2 imagery clearly di†erentiates extended nebular Ha emission from that of the compact core. Faint Object Spectrograph (FOS) observations provide a full set of optical and UV data (1200È6800Ó). These spec- tral data sample the innermost region(0A.86] 0A.86D 82 ] 82 pc) and exclude the extended Ha emission seen in the WFPC2 image. The derived emission-line Ñuxes allow a detailed analysis of the physical \ conditions within the nucleus. The measured Ñux ratio for Ha/Hb,FHa/FHb 4.53, indicates substantial intrinsic reddening, E(B[V ) \ 0.42, for the nuclear nebular emission. This is the Ðrst Ðnding of a large extinction of the nuclear emission-line Ñuxes in NGC 1052. If the central ionizing continuum is assumed to be attenuated by a comparable amount, then the emission-line Ñuxes can be reproduced well by a simple photoionization model using a central power-law continuum source with a spectral index of a \[1.2 as deduced from the observed Ñux distribution. A multidensity, dusty gas gives the best Ðt to the observed emission-line spectrum. Our calculations show that the small contribution from a highly ionized gas observed in NGC 1052 can also be reproduced solely by photoionization modeling. The high gas covering factor determined from our model is consistent with the assumption that our line of sight to the central engine is obscured. Subject headings: galaxies: active È galaxies: individual (NGC 1052) È galaxies: ISM È galaxies: nuclei

1. INTRODUCTION luminous AGNs (Ferland & Netzer 1983; Ho, Filippenko, & Sargent 1993) and shock-heating (Heckman 1980; Low-ionization nuclear emission-line regions, (LINERs) Dopita & Sutherland 1995; Dopita 1996). LINERs might have been found to occur in nearly one-third of all bright, represent a heterogeneous class of objects since the pro- nearby galaxies (Ho et al. 1995). They share many of the posed ionization mechanisms can all explain the low appar- emission characteristics of active galactic nuclei (AGNs), ent luminosity of LINERs. Perhaps more than one of these but are distinguished by their low luminosities and rela- various mechanisms is at work in the same object. I tively strong, low-ionization emission lines (e.g., [O ] NGC 1052 is a nearby (z \ 0.0045; de Vaucouleurs et al. II II II jj6300, 6363, [O ] j3727, [N ] jj6548, 6584, [S ] 1991) elliptical galaxy (E4) with a strong nuclear emission- jj6716, 6731). A recent study by Ho (1999) has demon- line spectrum dominated by low ionization features. For strated that the UVÈX-ray spectral energy distribution example, the [O I] j6300 and [O II] j3727 Ñuxes exceed (SED) of LINERs is harder than in higher luminosity that of [O III] j5007, plus [N II] j6584/Ha[1 (Fosbury et AGNs. This is consistent with the general trend exhibited in al. 1978; Ho et al. 1993, 1997). Fosbury et al. (1978, hereafter AGNs of decreasingaox with decreasing luminosity (Avni & F78) showed that the emission-line gas is made up of a Tananbaum 1982). The physical relation between LINERs luminous central component (d ¹ 2@@) and a di†use extended and AGNs is unclear. They may be simply the low- region (d D 20@@). Each emission region contributes approx- luminosity extension of the AGN phenomenon, powered by imately equally in total Balmer line Ñux. Furthermore, the accretion onto a compact object or, alternatively, they may extended component emits many of the other prominent be the result of entirely di†erent mechanisms. optical lines observed in the spectrum of the central region, Although central-source photoionization appears to be such as [O II] j3727 and [O III] j5007 (F78). the dominant mechanism powering the nebular emission The observed optical-continuum emission is dominated seen in the high-luminosity Seyfert galaxies, it has not by the host galaxy, with no clear detection of any under- been thoroughly tested for galaxies containing LINERs. lying nonthermal emission (F78; Fosbury et al. 1981). Competing ionization mechanisms that have been used Radio studies, on the other hand, clearly show it to have to successfully explain at least some of the emission both compact and extended continuum emission com- characteristics of LINERs include photoionization by hot ponents. At D1.5 GHz, its radio emission is characterized stars (Terlevich & Melnick 1985; Filippenko & Terlevich by a Ñat, unresolved (¹1A) core and two weaker lobes 1985), photoionization by nonthermal emission typical of extending 10AÈ15A on either side of the compact source (Wrobel 1984). NGC 1052 has been shown to have an IR excess in the 5È20 km band within the inner 2A of the 1 Based on observations made with the NASA/ESA Hubble Space Telescope, obtained from the data archive at the Space Telescope Science nucleus (Becklin, Tokunaga, & Wynn-Williams 1982). A Institute. STScI is operated by the Association of Universities for Research recent analysis of ROSAT /HRI imagery shows some in Astronomy, Inc., under NASA contract NAS 5-26555. extended soft X-ray emission, but the low spatial resolution 883 884 GABEL ET AL. Vol. 532 at these energies prevents isolating any compact emission at form a comprehensive picture of the activity of NGC 1052. scales less than 10A (Guainazzi & Antonelli 1999). X-ray Finally, we summarize our results and present our conclu- observations with ASCA indicate a very Ñat spectrum for sions in ° 5. the hard X-ray emission (2È10 keV) of NGC 1052 OBSERVATIONS AND EMISSION LINE MEASUREMENTS (Guainazzi & Antonelli 1999). 2. - As with LINERs in general, there is much disagreement 2.1. FOS Observations on the nature of the excitation mechanism responsible for The spectral data used in this study were obtained with the nuclear activity in NGC 1052. Initial Ðndings have sug- the Faint Object Spectrograph aboard the Hubble Space gested that shock-heating could produce the observed line Telescope (HST /FOS) on 1997 January 13. They were emission (Koski & Osterbrock 1976; F78). This claim is retrieved from the Space Telescope Science Institute (STScI) based largely upon the high electron temperature (T D III data archives. All observations were of the active nucleus of 33,000 K) implied by the strong [O ] j4363 emission. NGC 1052 through the0A.86 pair square aperture (post- Furthermore, simple shock-excitation models have been COSTAR), covering 1200È6800Ó. Thus, the aperture shown to match the prominent optical emission features sampled the central 82 ] 82 pc of the nuclear region of (Koski & Osterbrock 1976; F78). Alternate interpretations \ \ ~1 NGC 1052, assuming D 19.6 Mpc forH0 75 km s by Ferland & Netzer (1983), and more recently by Ho et al. Mpc~1. The optical and NUV data were obtained with (1993), have shown that many of the strong emission lines high-resolution gratings giving j/*j D 1300, while the UV observed in NGC 1052 (and in LINERs in general) could be spectrum has a lower resolution given by j/*j D 250. The produced by photoionization from a nonthermal contin- individual spectra are described in Table 1 along with their uum typical of Seyfert galaxies, namely a power law with [ exposure times, grating settings, and spectral resolutions. spectral index, aB 1.5. The major di†erence in the Because of a guide-star lock failure, the data may have models between the typically more luminous Seyfert gal- slightly degraded spectral resolution and sample a slightly axies and the LINERs is that the ionizing Ñux is inferred to larger spatial region than the aperture size would indicate. be signiÐcantly weaker in LINERs. Subsequent detailed If the extended regions emit di†erently than the nucleus, photoionization modeling of NGC 1052 by Pe quignot this may have an e†ect on the observed emission-line ratios (1984, hereafter P84) has shown that many of the optical III and absolute Ñuxes. emission lines, including [O ] j4363, could be Ðtted well Using the maps of Burstein & Heiles (1982), we Ðnd a by photoionization of a multidensity component gas. His foreground Galactic reddening value of E(B[V ) \ 0.02 in study assumed a central photoionizing source consisting of the direction of NGC 1052. Figure 1 shows the spectrum in a blackbody continuum with an X-ray tail extending to the rest frame of NGC 1052, corrected for Galactic extinc- higher energies. Recently, a polarimetric study has shown tion using the extinction law of Savage & Mathis (1979). NGC 1052 to have a broad Ha emission-line component with FWHM D 5000 km s~1 (Barth et al. 1998). This pro- 2.1.1. UV Emission vides compelling evidence for the presence of a compact Examination of the UV spectrum below 2000Ó in Figure object and lends strong support to the nonstellar photoion- 1 shows a weak UV continuum compared to the optical ization hypothesis. continuum, although, there appears to be a slight rise This paper presents a detailed analysis of the nuclear toward lower wavelengths. This UV continuum shows no activity in the prototypical LINER galaxy NGC 1052. Spe- evidence for recent starburst activity in the nucleus of NGC ciÐcally, we use Hubble Space Telescope (HST ) observations 1052. SpeciÐcally, the common signatures of hot O and B of NGC 1052 to explore whether or not photoionization stars, namely broad wind features from C IV j1549, Si IV represents a viable option in explaining the emission-line j1400, and N V j1240 and narrower photospheric lines spectrum in this LINER. Section 2 describes the HST (Smith Neubig & Bruhweiler 1997, 1999), are absent within imagery and spectroscopic observations used in this study the low detection limits of the data. Although the results are and presents our measured emission-line Ñuxes and analysis not deÐnitive, the absence of stellar features is consistent of the WFPC2 Ha imagery. We examine the possible role of with the UV continuum being produced by a nonthermal extinction on the emission-line Ñuxes and ionizing contin- compact source. Numerous UV emission lines are detected uum. The methodology and results of our photoionization in the FOS spectrum, as compared with the IUE study in modeling are given in ° 3. In ° 4, we discuss the implications which only two lines, C III] j1909 and C II] j2326, were of our results and compare them with previous studies to unambiguously measured (Fosbury et al. 1981).

TABLE 1 FOS OBSERVATIONS OF NGC 1052

Resolution Exposure Data Set Grating (AŽ ) (s) Template Galaxy

Y3K80105T ...... G160La D6.4b 1390 . . . Y3K80106T ...... G160L . . . 2430 . . . Y3K80107T ...... G160L . . . 2430 . . . Y3K80108T ...... G270H D2.1c 2320 NGC 3608 Y3K80109T ...... G400H D3.1c 1200 NGC 3608 Y3K8010AT ...... G570H D4.4c 890 NGC 5845

a G160L spectra coadded, weighted by exposure time. b j/*j D 250. c j/*j D 1300. No. 2, 2000 PHOTOIONIZATION MODELING OF NGC 1052 885

FIG.1b

FIG.1a

FIG.1c FIG.1d FIG. 1.ÈFOS spectra of the nucleus of NGC 1052 from 1250 to 6800Ó, corrected for Galactic reddening. The plotted template spectra are scaled to match the absorption features of NGC 1052 and o†set in Ñux for clarity.

2.1.2. Emission-L ine Measurements of NGC 1052 in Figure 1. Since there is very little stellar Visual inspection of the optical and NUV data at contribution to the UV in elliptical galaxies and no evi- j[2000Ó in Figure 1 clearly shows that the host elliptical dence for hot stars in NGC 1052, no template was used for galaxy dominates the continuum. Strong stellar absorption the G160L grating. features are evident and may a†ect Ñux measurements of We then measured the emission-line Ñuxes using the sub- certain emission lines, especially the Balmer lines. In addi- tracted spectrum corrected for Galactic extinction. Flux tion, small wavelength variations in the composite stellar ratios relative to Hb are given in Table 2. To measure the spectrum of the host galaxy may mimic emission-line fea- line Ñuxes, uncontaminated continuum regions were selec- tures. To remove the spectral contribution of the host ted on either side of the features to be measured. A linear Ðt galaxy, we have used template spectra of elliptical galaxies was made to the continuum underlying the emission fea- from the HST /FOS archival database to approximate its tures and the lines were extracted. In several cases, it was integrated stellar spectrum. We selected a proper template difficult to distinguish between real emission lines and arti- spectrum for each of the three optical gratings (G270H, facts remaining from our attempt to remove the underlying G400H, and G570H) based on the following criteria. (1) We spectrum of the host galaxy. This is most notable in the He required the prominent absorption features that do not II j4686 emission. In Table 2, we have indicated where these coincide with emission features in NGC 1052 to match uncertainties are large. The emission-line widths are those in the host spectrum of NGC 1052 (namely, Mg I broader than typically observed in the narrow line regions j2852, Ca II K j3934, G Band j4301, Fe I j4440, Fe I j4532, of AGNs, corresponding to a range of velocities between Mg Ib j5176, Fe I j5333, Ca I/Mg H j5590, Na I doublet 900 km s~1 ¹ FWHM ¹ 1100 km s~1 (by comparison, the j5893, and Ca I/Fe I j6161; Pritchet & van den Bergh 1977; instrumental resolution for the high-dispersion gratings is Crenshaw & Peterson 1985). (2) We required the overall FWHM D 230 km s~1). galactic Ñux distribution to be similar to that in NGC 1052. In several cases, it was necessary to deblend line contrib- The template spectra were then scaled in magnitude so that utors before accurate emission-line Ñuxes could be derived. the strengths of the absorption features matched that of In all cases, the spectra were converted to velocity space to NGC 1052 and subtracted from it. Our template spectra are preserve the correct line widths. For example, Ha is heavily listed in Table 1 and plotted with the corresponding spectra blended with the [N II] jj6548, 6584 doublet. To extract 886 GABEL ET AL. Vol. 532

TABLE 2 (Fig. 2a). Figure 2b shows the resulting Ha Ðt compared to MEASURED AND DEREDDENED EMISSION-LINE RATIOSa the Hb velocity proÐle and demonstrates the reliability of this technique. Line E(B[V ) \ 0.02 E(B[V ) \ 0.44 Several other line blends were treated using the same C IV j1549 ...... 0.14 (^0.04) 0.75 (^0.64) technique. It is important to deblend and provide accurate He II j1640 ...... 0.16 (^0.05) 0.80 (^0.66) Ñux ratios for the [S II] jj6716, 6731 lines because they [O III] j1663 ...... 0.07 (^0.02) 0.35 (^0.28) provide a low-density constraint for the emission-line N III] j1750 ...... 0.04 (^0.01) 0.20 (^0.16) region. To Ðt them, Hb proÐles were centered at the intrin- Si II ] [Ne III] jj1808È1820 ...... 0.10 (^0.03) . . . sic wavelengths of the two [S II] lines. The Ñux ratio was Si III] jj1882, 1892b ...... 0.14 (^0.05) 0.80 (^0.71) varied until residuals were minimized. For line blends of C III] j1909b ...... 0.35 (^0.08) 2.00 (^1.73) other hydrogen lines, namely, Hc ] [O III] j4363 and N II] j2140 ...... 0.06 (^0.01) . . . ] III II c ^ ^ Hv [Ne ] j3967, the Hb proÐle was used as a template C ] j2326 ...... 0.58 ( 0.08) 3.80 ( 3.51) for the Balmer line since it will have an identical intrinsic [Ne IV] j2424 ...... 0.02 (^0.01) 0.10 (^0.09) [O II] j2470 ...... 0.14 (^0.03) 0.60 (^0.43) shape. All lines that required deblending are noted in Fe II jj2558È2583 ...... 0.03 (^0.01) . . . Table 2. Fe II jj2601È2640 ...... 0.14 (^0.03) . . . 2.1.3. Uncertainties II ^ Fe jj2660È2684 ...... 0.04 ( 0.01) . . . The primary sources of uncertainty are continuum place- Fe II jj2711È2765 ...... 0.15 (^0.03) . . . Fe II jj2823È2844 ...... 0.04 (^0.01) . . . ment in extracting emission lines and the approximations Mg II j2800 ...... ::0.46d 1.2 associated with the deblending procedure. For the Ðrst case, He II j3204 ...... ::0.05 ::0.08 the uncertainty depends on the strength of the emission line [Ne V] j3346 ...... ::0.015 ::0.03 relative to the noise of the data and the existence of host [Ne V] j3426 ...... 0.07 (^0.02) 0.11 (^0.04) galaxy features. We experimented with continuum levels for [O II] j3727 ...... 2.77 (^0.31) 3.98 (^0.81) [Ne III] j3869 ...... 0.60 (^0.08) 0.82 (^0.16) [Ne III] j3967b ...... 0.20 (^0.03) 0.27 (^0.05) Hvj3970b ...... 0.14 (^0.03) 0.19 (^0.05) [S II] jj4069, 4076 ...... 0.68 (^0.10) 0.88 (^0.17) Hdj4102 ...... 0.24 (^0.03) 0.31 (^0.05) Hcj4340b ...... 0.44 (^0.08) 0.53 (^0.11) [O III] j4363b ...... 0.27 (^0.05) 0.32 (^0.06) He II j4686 ...... ::0.18 ::0.19 Hbj4861 ...... 1.00e 1.00f [O III] j4959 ...... 0.91 (^0.13) 0.88 (^0.13) [O III] j5007 ...... 2.85 (^0.32) 2.72 (^0.31) [N I] j5200 ...... 0.23 (^0.03) 0.21 (^0.03) [N II] j5755 ...... ::0.09 ::0.07 He I j5876...... ::0.14 ::0.10 [O I] j6300 ...... 3.32 (^0.37) 2.22 (^0.49) [O I] j6364 ...... 1.13 (^0.13) 0.74 (^0.17) [N II] j6548b ...... 1.73 (^0.26) 1.10 (^0.29) Haj6563b ...... 4.53 (^0.68) 2.86 (^0.77) IG [N II] j6583b ...... 5.14 (^0.77) 3.24 (^0.87) F .2a He I j6678...... ::0.07 ::0.04 [S II] j6716b ...... 2.56 (^0.38) 1.57 (^0.43) [S II] j6731b ...... 2.35 (^0.35) 1.45 (^0.40)

a Relative to Hb. b Line deblending procedure required for Ñux measurement. c Contains [O III] j2321 and Si II] j2335 emission. d Mg II j2800 is heavily absorbed; a Ðt to the doublet with Gaussian features gives Mg II j2800/Hb \ 0.80. e F \ 2.65(^0.13) ] 10~14 ergs s~1 cm~2. f Hb\ ^ ] ~13 ~1 ~2 FHb 1.1( 0.7) 10 ergs s cm . the Ha feature, the normalized line proÐle of Hb was used to make a template for both of the [N II] lines since no broad wings were detected on the Hb proÐle. This tech- nique proved more accurate than Ðtting with Gaussian pro- Ðles because it approximates the actual structure of the emission lines. The [N II] template was created by separat- ing the two proÐles according to their intrinsic transition FIG.2b wavelengths and assuming that the Ñux ratio corresponds \ IG ] II to that of their Einstein A-values (Aj6584/Aj6548 2.96). F . 2.È(a) Deblending procedure for Ha [N ]. The solid line is the The wavelength o†set and relative Ñux of the template were Ha ] [N II] line blend with the host galaxy template removed. The [N II] II jj6548, 6584 emission-line template (dot-dashed line) was created from the then varied until the residual of the [N ] contribution Hb proÐle and subtracted to give the Ha (dashed line) emission line. subtracted from the Ha ] [N II] line blend gave the best Ðt (b) Comparison of the shape of the deblended Ha proÐle to the Hb to Ha, namely, a single-emission feature on a Ñat continuum proÐle scaled to match in Ñux. No. 2, 2000 PHOTOIONIZATION MODELING OF NGC 1052 887

TABLE 3 Additional uncertainties are introduced for lines that INTRINSIC AND DEREDDENED BALMER require deblending. To reliably estimate this uncertainty, LINE RATIOSa the Ñux of the residual associated with the [N II] j6584 line ^ b c (see Fig. 2a) was measured and shown to be 10% of the Balmer Line Intrinsic Dereddened measured Ñux for the line. This represents an upper limit to Haj6563 ...... 2.86 2.86 (^0.77) the general deblending uncertainty because it was the Hcj4340 ...... 0.47 0.53 (^0.11) largest relative residual of all procedures. These two uncer- Hdj4102 ...... 0.26 0.31 (^0.05) tainties were then added in quadrature to obtain the total Hvj3970 ...... 0.16 0.19 (^0.05) error for the measured line ratios. The total uncertainty a Relative to Hb. estimate for each emission line is listed in Table 2. b From Osterbrock 1989. c [ \ [ ] [ \ 2.1.4. Reddening E(B V )T E(B V )G E(B V )I 0.44. To estimate the internal reddening of NGC 1052, we \ lines of di†erent strengths. For the optical and NUV data, compared our measured Ha/Hb Ñux ratio (FHa/FHb 4.53, where the host galaxy is dominant but spectral resolution is corrected for Galactic extinction) to an assumed intrinsic high, this leads to an estimate of the uncertainty for ratio(Ha/Hb) . ForN \ 104 cm~3 andT \ 10,000 K, and ^ I e e \ ““strong ÏÏ lines of 5% (deÐned as havingF º FHb), for assuming case B recombination,(Ha/Hb)I 2.86, which is ““moderate strength ÏÏ lines ^10% (0.2 ] F \ F \ F ), consistent with the typical ratio determined for AGNs ^ ] Hb Hb and for ““weak ÏÏ lines 20%(F ¹ 0.2 FHb). For wave- (Osterbrock 1989). Using the extinction curve of Savage lengths below D2000Ó, there is much less contribution & Mathis (1979), this implies a reddening of from the host galaxy; however, the signal-to-noise ratio is E(B[V ) \ 0.42 ^ .20, where the uncertainty was derived signiÐcantly deteriorated. Consequently, we Ðnd Ñux uncer- from the errors associated with the Ha and Hb Ñux mea- tainties of the moderate strength lines to be ^20% and the surements. The deduced intrinsic line ratios are listed in weak lines to be ^30% for wavelengths below 2000 Ó. Table 2. Uncertainties were estimated from propagation of

FIG. 3.ÈWFPC2 F658N image of the inner region of NGC 1052. The size of the FOS0A.86 square aperture is plotted over the compact emission core for reference. North is the top of this image. 888 GABEL ET AL. Vol. 532 errors associated with the reddening and observed Ñux intrinsic emission-line spectrum. Discrepancies between our ratios. Since all previous studies of NGC 1052 determined model results and the observed spectrum then provide negligible internal reddening, it is important to examine this further insight into the actual physical conditions in NGC issue in more detail (Koski & Osterbrock 1976; F78; 1052. Fosbury et al. 1981; Ho et al. 1993, 1997). Photoionization models were computed using Version 90 The intrinsic He II j1640/He II j4686 ratio is theoretically of the computer code CLOUDY (Ferland 1996). These a better measure of reddening than Ha/Hb because of the models were parameterized by the gas densitynH, contin- large span in wavelength between the two lines (Kraemer et uum shape, and ionization parameter al. 1994). However, the large Ñux uncertainties associated with the He II lines greatly diminish the usefulness of this \ Q U 2 , diagnostic for our data set. Yet, we can still use it as a check 4nR nH c on our reddening deduced from the Balmer line ratio. We where Q is the number of hydrogen ionizing photons s~1 II II \ assume an intrinsic ratio of (He j1640/He j4686) 6.6, and R the distance from the inner face of the cloud to the \ 4 ~3 \ I corresponding toNe 10 cm andTe 10,000 K central continuum source. When a good Ðt was achieved, (Seaton 1978). This is consistent with the dereddened ratio observational constraints on the ionizing luminosity pro- II II of He j1640/He j4686 º 4.2 considering the large vided constraints on R. uncertainties associated with these line Ñuxes (Table 2). We adopted the following idealized properties of the We can also use the relative Ñux ratios of the higher order emission-line gas. In cases where more than one density hydrogen Balmer lines beyond Hb as a consistency check. A component was required, we assumed that each component comparison of the dereddened Balmer ratios with their was photoionized by unattenuated radiation from the intrinsic values is given in Table 3 (Osterbrock 1989). They central source. Thus we assume, a priori, relatively small are seen to be consistent within the quoted uncertainties. covering factors for any inner gas components. Physically, [ \ The colors of the host galaxy of NGC 1052, B V 0.91, this corresponds to a clumpy gas composed of radiation- [ \ U B 0.42, are in accord with typical colors found in bounded clouds. We assume that the observed radiation is normal elliptical galaxies of similar Hubble type, E4 (F78). emitted from the ““front ÏÏ faces of the clouds, that is, the side Thus, the large extinction found above is not consistent if of the clouds facing the central continuum source. Each applied to the entire galaxy. If the implied reddening is component was assumed to have constant density. We correct, it suggests that dust is localized to the nuclear began by omitting dust grains from the emission-line gas; region of NGC 1052. There is evidence that supports this. hence, in our initial model, we assume solar abundances [ [ [ The B R and B I images of Carter et al. (1983) and B R and that all reddening occurs external to the line-emitting image of Davies & Illingworth (1986) clearly show a di†use region. Thus, the abundances relative to hydrogen are: dust lane extending out from the central region to about He \ 1.00 ] 10~1,C\ 3.55 ] 10~4,N\ 9.33 ] 10~5, 20A. Furthermore, the IR measurements of Becklin et al. O \ 7.41 ] 10~4,Ne\ 1.17 ] 10~4,Mg\ 3.80 ] 10~5, (1982) indicate a large IR excess between 5 and 20 km con- Si \ 3.55 ] 10~5,S\ 1.62 ] 10~5, and Fe \ 3.24 ] 10~5 Ðned within 2A of the nucleus. The strength of the IR Ñux at (Grevesse & Anders 1989). The e†ects of grains mixed with these wavelengths is not consistent with stellar radiation the gas will be treated in a later section. emanating from cool stars and may be thermal radiation from heated dust. 3.1. Choice of Model Parameters We adopted a simple power-law ionizing continuum 2.2. Ha Imagery parameterized by spectral index a. Observations in the UV Images of the nuclear region of NGC 1052 were taken and X-ray were used to constrain its shape (Fig. 4). We with WFPC2/HST using the F658N Ðlter on 1996 June 16. assumed that the observed UV continuum Ñux was These data were obtained from the STScI data archives. reddened by the same amount as the emission-line Ñuxes, This Ðlter isolates any extended Ha emission. In Figure 3, E(B[V ) \ 0.42, as deduced from the Ha/Hb ratio. This we have combined two exposures (with 800 and 900 s expo- gives a projected intrinsic Ñux at the Lyman edge of ] ~27 ~1 ~2 ~1 sure times) and removed cosmic ray events. The size of the F1 ryd D 1.9 10 ergs s cm Hz . Results from aperture used for the FOS observations is shown over the ASCA and ROSAT show NGC 1052 to have a rather Ñat bright emission core for reference, although the precise spectrum between 2 and 6 keV, a steeper continuum position of the aperture is uncertain because of the guide- between both 6 and 10 keV and D1 and 2 keV, and a star lock failure. Strong Ðlamentary nebular Ha emission is Ñattening at energies below 1 keV (Guainazzi & Antonelli seen to extend about 1A around the compact nucleus, with a 1999). This indicates that the soft X-ray emission is likely more di†use halo extending to farther distances. At a posi- heavily absorbed, which is consistent with the modeling by tion angle of D235¡ is a narrow Ðlament of Ha emission. Weaver et al. (1999). If this absorption is caused by a dusty This is compared to the radio jet/lobe observed by Wrobel gas, then our previous assumption that the UV continuum et al. (1984) at 1.5 GHz that is centered at a position angle of is attenuated by a column of dust is justiÐed. We chose the D275¡. observed ASCA Ñux at 6 keV,F \1.4]10~30 ergs s~1 ~2 ~1 6 keV PHOTOIONIZATION MODELS cm Hz , as the best estimate of the intrinsic nuclear 3. X-ray Ñux, as shown in Figure 4. Interpolating between the The purpose of the following analysis is to determine estimated intrinsic UV continuum Ñux at 1 ryd and the whether or not the emission in the compact nuclear region X-ray Ñux at 6 keV gives a \[1.2 and a total intrinsic of NGC 1052 is consistent with photoionization by a simple luminosity for the ionizing continuum of L ;1 rydD2.0 nonthermal continuum, which is characteristic of AGNs. ]1042 ergs s~1. We note the large uncertainty in this value We seek the simplest model, making the fewest number of owing to the uncertainty in the amount of obscuration by assumptions, that gives a satisfactory Ðt to the dereddened dust. Previous studies have indicated that the continuum No. 2, 2000 PHOTOIONIZATION MODELING OF NGC 1052 889

where their ratios are determined solely by temperature. Conversely, in a higher density gas the nebular line emis- sions are suppressed because of collisional deexcitation and, consequently, the auroral line Ñuxes become a more signiÐ- cant source of cooling. Hence, a density stratiÐcation e†ect is mistaken for an indication of high temperature. Multi- component gases having a range of densities are a common feature in models for LINERs and Seyfert galaxies (Halpern & Steiner 1983; Filippenko 1985; Kraemer et al. 1994, 1998). Since we seek the simplest possible model, our method was to Ðnd the fewest number of discrete cloud components that would reproduce the observed emission. Although assuming a small number of distinct, isolated gas components might be an oversimpliÐcation, it provides insight into the physical conditions and structure of the emission-line gas. An alternate approach, the locally opti- mally emitting cloud model, has also been shown to e†ec- FIG. 4.ÈMultiwavelength spectrum of the nucleus of NGC 1052. The tively model the narrow line-region emission in AGNs radio data (plus signs) are the measurements of the compact emission (¹1A) (Ferguson et al. 1997). by Wrobel et al. 1984. The IR data points (asterisks) are within less than 2A Thus, as a Ðrst modiÐcation to the simple model, we (Becklin et al. 1982). The observed and dereddened UV continuum data at 1300Ó are from this study, D0A.86 (diamonds). ASCA and ROSAT data assumed a gas made up of two density components. Our (triangles) are from Guainazzi & Antonelli (1999) with a resolution of [email protected] method was to Ðt the nebular lines that have very low criti- is unclear how much of this X-ray emission is compact. The ionizing cal densities ([S II] jj6716, 6731, [N II] jj6548, 6584, and continuum used in our photoionization models is shown as the solid line. [O II] j3727) with a low-density gas, component 1, and the The dashed line is the attenuated continuum, assuming the central engine strongest auroral lines (i.e., those of S`,O`, and O` 2) with is viewed through the high-density gas, component 2. a higher density gas, component 2. Critical densities of various lines place constraints on the density for each com- emission in AGNs is typically not reddened as much as the ponent. If the [O II] j3727 line is a principal coolant for emission-line Ñux, thus, we may be overestimating the total component 1, the gas density must not be signiÐcantly \ ~3 luminosity (Costero & Osterbrock 1977). greater than its critical density ofnc 4500 cm In Figure 4, we have plotted the SED of the continuum (DeRobertis & Osterbrock 1984). A lower limit on the emission from the radio through hard X-ray energies of the density for component 1 is obtained from the [S II] central point source. In each energy band, we have plotted jj6716, 6731 doublet lines. The Ñux ratio [S II] j6716/[S II] \ ~3 only the emission that cannot be spatially resolved. Data j6731 1.1 indicates an electron densityne [ 300 cm for both the observed and reddening-corrected UV contin- (Aller 1984). For the high-density component, the [O III] III ] 5 ] 7 uum are included, illustrating the large e†ect that our j4363/[O ] j5007 ratio implies 6.5 10 ¹ne¹3.3 10 assumed extinction has on the energetics of the active cm~3 as determined by the critical densities of the two lines. nucleus. Additionally, the strong [O I] j6300 emission indicates ] 6 ~3 Simple model calculations using a gas with constant ne ¹ 1.8 10 cm (DeRobertis & Osterbrock 1984). density do not match the intrinsic emission-line spectrum of Since the spectrum of NGC 1052 is dominated by lines of NGC 1052 for any choice of density. One important low-ionization species, lines emitted by S`,N`,O0,O`, problem with these models is that they cannot match both and C`, a low-ionization parameter is required. Figure 1 of the auroral and nebular line Ñuxes simultaneously. When a Ferland & Netzer (1983), which shows the relationship 3.5 ~3 low-density gas is used,nH D 10 cm , the nebular lines between various line strengths and the ionization parameter are Ðt well (i.e., [O III] jj5007, 4959, [N II] jj6548, 6584, for a low-density gas, indicates U \ 10~3.5 for component [O II] j3727, and [S II] jj6716, 6731), but their auroral line 1. We Ðnd that a slightly higher U is needed for component counterparts are underestimated, namely, [O III] j4363, 2 to sufficiently populate the states responsible for the [O II] j2470, and [S II] jj4069, 4076. For higher densities, auroral line emission. Model results usingn \ 103.5 cm~3, \ ~3.8 \ 6.2 ~3 \ H ~3.1 the opposite is true. As noted earlier, it was the high tem- U1 10 andnH 10 cm ,U2 10 for com- perature implied by the [O III] j4363/[O III] jj4959, 5007 ponents 1 and 2, respectively, are presented in columns 3 ratio that led to the original interpretation of shock excita- and 4 of Table 4. The composite model in column 5 of Table tion for NGC 1052. Indeed, our [O III] line Ñux ratio, [O III] 4 was created by assuming equal contribution from each j4363/[O III] jj4959, 5007 D 0.089, implies T [ 100,000 K component to the total observed Hb Ñux. in the low-density limit (Osterbrock 1989), which is a factor D5 higher than we would expect in the O`` zone of a 3.2. T he E†ects of Grains in the Emission-L ine Gas photoionized nebulae. However, choosing the high density Inspection of Table 4 shows that the overall intrinsic that matches this ratio underestimates the other nebular emission-line spectrum of NGC 1052 is Ðtted very well by line Ñuxes because their critical densities are much lower our photoionization model, which has employed a than that of [O III] jj5007, 4959, which have critical minimum of assumptions. In particular, the strong auroral \ ] 5 ~3 III II II density,nc 6.5 10 cm . line Ñuxes of [O ] j4363, [O ] j2470, and [S ] The detailed analysis of P84 showed this could be reme- jj4069, 4076 are matched well, as are the [O III] j4363/ died by employing two separate gas components having [O III] j5007 and [O II] j2470/[O II] j3727 ratios, which di†erent densities. In this scenario, the nebular lines over- are often mistakenly used as temperature indicators. whelm their auroral counterparts in a low-density region However, several of the prominent low-ionization features 890 GABEL ET AL. Vol. 532

TABLE 4 LINE RATIOS FROM MODEL WITHOUT DUSTa

Line Observedb Comp 1c Comp 2d Compositee (1) (2) (3) (4) (5)

C IV j1549 ...... 0.75 (^0.64) . . . 0.28 0.14 He II j1640...... 0.80 (^0.66) 1.09 1.28 1.19 [O III] j1663 ...... 0.35 (^0.28) . . . 0.49 0.25 N III] j1750 ...... 0.20 (^0.16) . . . 0.20 0.10 Si III] jj1882, 1892 ...... 0.80 (^0.71) 0.01 1.12 0.57 C III] j1909 ...... 2.00 (^ 1.73) 0.21 3.50 1.86 C II] j2326g ...... 3.80 (^ 3.51) 1.00 2.56 1.78 [Ne IV] j2424 ...... 0.10 (^0.09) . . . 0.01 0.01 [O II] j2470 ...... 0.60 (^0.43) 0.30 1.11 0.71 Mg II j2800 ...... ::1.2 1.77 4.50 3.14 He II j3204...... ::0.08 0.07 0.08 0.08 [Ne V] j3346 ...... ::0.03 ...... \0.01 [Ne V] j3426 ...... 0.11 (^0.04) ...... \0.01 [O II] j3727 ...... 3.98 (^0.81) 7.07 0.09 3.58 [Ne III] j3869 ...... 0.82 (^0.16) 0.82 1.92 1.37 [Ne III] j3967 ...... 0.27 (^0.05) 0.25 0.60 0.43 Hvj3970 ...... 0.19 (^0.05) 0.17 0.16 0.17 [S II] jj4069, 4076 ...... 0.88 (^0.17) 0.49 0.92 0.71 Hdj4102 ...... 0.31 (^0.05) 0.27 0.27 0.27 Hcj4340 ...... 0.53 (^0.11) 0.47 0.47 0.47 [O III] j4363 ...... 0.32 (^0.06) . . . 0.59 0.30 He II j4686...... ::0.19 0.12 0.14 0.13 Hb ...... 1.00 1.00 1.00 1.00 [O III] j4959 ...... 0.88 (^0.13) 0.34 1.72 1.03 [O III] j5007 ...... 2.72 (^0.32) 0.97 4.98 2.98 [N I] j5200 ...... 0.21 (^0.03) 0.25 0.01 0.13 [N II] j5755 ...... ::0.07 0.05 0.17 0.11 He I j5876 ...... ::0.10 0.17 0.15 0.16 [O I] j6300 ...... 2.22 (^0.49) 1.23 2.19 1.71 [O I] j6364 ...... 0.74 (^0.17) 0.39 0.70 0.55 [N II] j6548 ...... 1.10 (^0.29) 1.36 0.29 0.83 Haj6563 ...... 2.86 (^0.77) 2.92 2.96 2.94 [N II] j6583 ...... 3.24 (^0.87) 4.03 0.85 2.44 He I j6678 ...... ::0.04 0.05 0.03 0.04 [S II] j6716 ...... 1.57 (^0.43) 1.47 0.06 0.77 [S II]j6731 ...... 1.45 (^0.40) 2.03 0.14 1.09 [Ca II] j7291 ...... 0.61 0.06 0.34 [Ca II] j7324 ...... 0.40 0.04 0.22

a Relative to Hb. b Corrected for Galactic and internal reddening, E(B[V ) \ 0.44. c U \ 10~3.8, N \ 103.5. d \ ~3.1 H \ 6.2 U 10 , NH 10 . e Composite of components 1 and 2, assuming equal contribution to Hb Ñux. g Contains [O III] j2321 and Si II] j2335 emission. are considerably underpredicted. For example, the observed structure since, of the available data for dust grains, physi- [N II] jj6548, 6584, [S II] jj6716, 6731, and [O I] jj6300, cal conditions in an H II region most closely approximate 6363 line Ñuxes are matched within a factor of 2 but are those of NGC 1052. Constraints on the dust-to-gas ratio in all somewhat underestimated. The C II] j2326 Ðt is even NGC 1052 come from the strengths of the Si III] worse. jj1882, 1892 and Mg II j2800 Ñuxes. The calculated emis- All these poorly Ðt lines emit predominately in the warm sion of these lines becomes too weak when the full depletion partially ionized zones, [S II] jj6716, 6731 and [N II] deduced for Orion is used in our models (Ferland 1996: jj6548, 6584 most strongly in component 1 and [O I] based on results of Baldwin et al. 1991; Rubin et al. 1991; jj6300, 6363 and C II] j2326 primarily in the high-density and Osterbrock, Tran, & Veilleux 1992). Therefore, we component 2. The characteristic emission in these regions is assumed half of the depletion and grain abundances, giving likely greatly modiÐed by the presence of dust because of the following elemental abundances relative to hydrogen: the heavy depletion of important coolants (particularly He \ 9.77 ] 10~2,C\ 3.31 ] 10~4,N\ 8.13 ] 10~5, magnesium and silicon) and an increase in gas temperature O \ 5.50 ] 10~4,Ne\ 8.51 ] 10~5,Mg\ 1.07 ] 10~5, caused by the selective removal of photons near the Lyman Si \ 1.20 ] 10~5,S\ 1.29 ] 10~5, and Fe \ 1.02 ] 10~5. limit by dust grains. To treat this, we have included grains Model results, using the same parameters as in our pre- typical of the Orion Nebula mixed with the emission-line vious model, are given in Table 5. We have included model gas (Baldwin et al. 1991). We selected this type of grain predictions of the [Ca II] jj7291, 7324 lines for both the No. 2, 2000 PHOTOIONIZATION MODELING OF NGC 1052 891

TABLE 5 BEST-FIT MODEL:INCLUDES DUST GRAINSa

Line Observedb Comp 1c Comp 2d Compositee

C IV j1549 ...... 0.75 (^0.64) . . . 0.38 0.19 He II j1640...... 0.80 (^0.66) 0.84 1.28 1.06 [O III] j1663 ...... 0.35 (^0.28) 0.01 0.53 0.27 N III] j1750 ...... 0.20 (^0.16) . . . 0.25 0.13 Si III] jj1882, 1892 ...... 0.80 (^0.71) 0.02 0.54 0.28 C III] j1909 ...... 2.00 (^1.73) 0.35 4.53 2.44 C II] j2326f ...... 3.80 (^3.51) 1.45 2.86 2.16 [Ne IV] j2424 ...... 0.10 (^0.09) . . . 0.01 0.01 [O II] j2470 ...... 0.60 (^0.43) 0.36 1.02 0.69 Mg II j2800 ...... ::1.2 0.89 1.16 1.03 He II j3204...... ::0.08 0.05 0.08 0.07 [Ne V] j3346 ...... ::0.03 ...... \0.01 [Ne V] j3426 ...... 0.11 (^0.04) ...... \0.01 [O II] j3727 ...... 3.98 (^0.81) 7.35 0.09 3.72 [Ne III] j3869 ...... 0.82 (^0.16) 0.78 1.68 1.23 [Ne III] j3967 ...... 0.27 (^0.05) 0.23 0.51 0.37 Hvj3970 ...... 0.19 (^0.05) 0.17 0.16 0.16 [S II] jj4069, 4076 ...... 0.88 (^0.17) 0.55 1.06 0.81 Hdj4102 ...... 0.31 (^0.05) 0.26 0.27 0.27 Hcj4340 ...... 0.53 (^0.11) 0.47 0.47 0.47 [O III] j4363 ...... 0.32 (^0.06) 0.01 0.58 0.30 He II j4686...... ::0.19 0.09 0.14 0.12 Hbg ...... 1.00 1.00 1.00 1.00 [O III] j4959 ...... 0.88 (^0.13) 0.33 1.50 0.92 [O III] j5007 ...... 2.72 (^0.31) 0.96 4.34 2.45 [N I] j5200 ...... 0.21 (^0.03) 0.40 0.02 0.21 [N II] j5755 ...... ::0.07 0.06 0.19 0.13 He I j5876 ...... ::0.10 0.16 0.14 0.15 [O I] j6300 ...... 2.22 (^0.49) 1.40 2.49 1.95 [O I] j6364 ...... 0.74 (^0.17) 0.45 0.79 0.62 [N II] j6548 ...... 1.10 (^0.29) 1.44 0.36 0.90 Haj6563 ...... 2.86 (^0.77) 2.93 2.92 2.93 [N II] j6583 ...... 3.24 (^0.87) 4.26 1.05 2.66 He I j6678 ...... ::0.04 0.04 0.03 0.04 [S II] j6716 ...... 1.57 (^0.43) 1.72 0.10 0.91 [S II] j6731 ...... 1.45 (^0.40) 2.20 0.23 1.22 [Ca II] j7291 ...... 0.07 0.01 0.04 [Ca II] j7324 ...... 0.04 0.01 0.03

a Relative to Hb. b Corrected for Galactic and internal reddening, E(B[V ) \ 0.44. c U \ 10~3.8, N \ 103.5. d \ ~3.1 H \ 6.2 U 10 , NH 10 . e Composite of components 1 and 2, assuming equal contribution to Hb Ñux. f Contains [O III] j2321 and Si II] j2335 emission. g Obs \ ] 39 ~1 Comp1 \ ] 39 ~1 Comp2 \ ] 40 L Hb 5.0 10 ergs s ;L Hb 9.1 10 ergs s ;L Hb 1.0 10 ergs s~1. \ ] 19 ~2 \ ] 20 ~2 dusty and solar abundance models in Tables 5 and 4, (NH II 3.1 10 cm andNH II 3.6 10 cm , for respectively. Ferland (1993) has shown that the weakness of components 1 and 2, respectively). Thus, most of the these lines in the narrow line regions of some AGNs is a reddening occurs in a neutral gas obscuring our line of sight good indicator of the presence of grainsÈa sharp decrease to the emitting regions. If this obscuration is caused by gas in their strengths is expected with depletion. Indeed, com- component 1 of our model,n \ 103.5 cm~3, a cloud thick- 1 H parison of the results of our two models demonstrates this. ness of onlyD2 pc is required to produce the observed SpeciÐcally, the dusty model predicts a decrease in the extinction. strength of the lines by a factor of 10. Although the [Ca II] The polarimetry study by Barth, Filippenko, & Moran lines lie outside of the spectral region covered by our data, (1999) shows that [O I] j6300 is more polarized than the the near-infrared spectral data of Diaz, Terlevich, & Pagel other ““narrow ÏÏ emission lines between D4800 and 6800 Ó, (1985) clearly show that their strengths are overpredicted by with a polarization percentage more than twice as much as our dust-free model (see their Fig. 1). Thus, there is evidence any other line. This indicates that [O I] j6300 may be trans- that signiÐcant depletion exists in NGC 1052. mitted through more dust than the other lines. This e†ect is Our depletion percentage corresponds to a gas column- explained naturally, as in our model, if [O I] j6300 is [ \ ] 22 to-reddening scaling of roughly NH/E(B V ) 1 10 emitted predominately in an extended, dusty, partially cm~2 mag~1 (Knapp & Kerr 1974). The ionized column ionized zone (PIZ). If we are viewing the line emission pri- densities of our gas components are comparatively small marily from the ionized faces of the gas clouds, then the 892 GABEL ET AL. Vol. 532 lines emitted in the ““back side ÏÏ of the gas pass through a required U on either of the two gas components would much larger dust column before escaping. Our calculations severely degrade our Ðt to the prominent low-ionization show that the PIZ of component 2, where [O I] j6300 is features, giving results characteristic of Seyfert galaxies. most strongly produced, has a much larger column density Indeed, it is this lower ionization that distinguishes LINERs than the PIZ of component 1. Since this is the region where from other active galaxies. Thus, there is an indication that the [N II] jj6548, 6584 and [S II] jj6716, 6731 lines are a highly ionized gas component is present in NGC 1052, emitted, this explains why there is much less polarization of although its contribution is small. I IV ] 4 these lines than of the [O ] j6300 line. The critical density of [Ne ] j2424 limits ne ¹ 3 10 cm~3. However, for a gas with this low density and having a 3.3. Structure of the Emission-L ine Region high-ionization parameter, the [O III] jj5007, 4959 lines We now turn our attention toward the structure of the become the dominant coolants and their Ñuxes are far over- emission-line region as deduced from our model results. predicted. We Ðnd that an emitting gas that is matter Since we have an estimate of the luminosity of ionizing bounded can account for this. For illustrative purposes, we radiation, we can calculate the radial distances of the di†er- have calculated the emission for a gas without dust grains \ ~1.5 \ 4.2 ~3 ent gas components from the putative central engine, R havingU3 10 andnH 10 cm and a total P 2 1 \ 21 ~2 andR2, usingU L (z1 ryd)/R . This givesR1 D 100 pc and column density ofNH 10 cm . This corresponds to a R2 D 2 pc. The deduced location of component 1 lies distanceR3 D 3 pc of the highly ionized gas from the central outside the projected aperture by about a factor of 2; continuum source. A very low covering factor of only however, the uncertainty in the intrinsic luminosity could D0.05 is required to match the observed C IV j1549 lumi- account for this discrepancy. Furthermore, if the gas is nosity with this extra component; hence, our assumption spherically distributed, then much of it would be projected that all components receive unattenuated continuum radi- into the aperture. ation remains valid. This simple model for a highly ionized The broad widths of the emission lines component matches the observed [Ne V] j3426 emission (900 ¹ FWHM ¹ 1100 km s~1) give clues to the gas line very well and the [Ne IV] j2424 line within a factor of 2. dynamics in NGC 1052. We Ðnd no correlation between Furthermore, since the emissivity of C IV j1549 is very high line width and critical density. Thus, since the high- and under these conditions, there would be only a small contri- low-density components are at very di†erent distances from bution from this component and, therefore, no change in the putative central massive object, the velocities of the two the low-ionization line ratios predicted by a three- components cannot be explained as simple circular component model from those listed in Tables4&5. P ~1@2 Keplerian orbits, for which v r . DISCUSSION As mentioned earlier, our assumption that the line emit- 4. ting gas is made up of distinct components having constant We have shown that a simple photoionization model densities is a rough approximation, but qualitatively, the explains the full set of dereddened optical and UV emission higher density gas is seen to be closer to the center of lines observed in the FOS data of NGC 1052 quite well. We the nucleus. To match the calculated Hb luminosity with note that the reddening correction we have applied to the the observed luminosity, covering factors ofv \ 0.28 and UV continuum is highly uncertain and that this enhanced \ 1 v2 0.25 are required for components 1 and 2, respectively, luminosity is essential in reproducing the line luminosities. assuming that each component contributes equally to the Our best-Ðt model employs a dusty emission-line gas with a total Hb emission. Model predictions ofL Hb for each com- high covering factor. This is consistent with the apparent ponent are given in Table 5. Since the covering factors for absorption of the soft X-ray Ñux observed in the ROSAT each component were calculated independently of each data, indicating that the line of sight toward the AGN is other, the attenuation of the inner component was not obscured. The Ñat intrinsic ionizing continuum (a \[1.2) treated; hence,v1 represents a minimum covering factor for we have deduced for NGC 1052 is in accord with the recent the outer component. Therefore, a minimum total covering study by Ho (1999), which demonstrated that a sample of \ factor for NGC 1052 isvT 0.53, which represents the LINERs all have a higher X-ray/UV continuum ratio than probability that the line of sight to the central continuum is found in typical high-luminosity AGNs. obscured by the emission-line gas. This is considerably higher than that determined for the narrow line regions of 4.1. Extended Emission-L ine Region Seyfert 1 galaxies (Netzer & Laor 1993). Thus, our previous We examine whether or not the deduced ionizing contin- assumption that the central ionizing continuum is occulted uum Ñux is sufficient to power the extended emission-line by a dusty gas is vindicated. This is consistent with the region (EELR) emission described in F78. They found a study by Barth et al. (1998), which concluded that many di†use emission-line gas extending out to about 20A from LINERs may have bright compact UV sources that are the nucleus and having a total Balmer line Ñux similar to obscured by dust. that observed in the nuclear emission region. Using their Ñux measurements and assuming that all of the emission- 3.4. T he High-Ionization Component line Ñuxes experience the extinction we have deduced from ] 40 ~1 Although the spectrum of NGC 1052 is dominated by our analysis givesL Hb D 2.6 10 ergs s for the total low-ionization features, there are several high-ionization Hb luminosity in the extended plus compact emission com- features present, namely, C IV j1549, [Ne IV] j2424, and ponents. A simple photon-counting argument (Netzer & [Ne V] j3426. As anticipated from our choice of low- Laor 1993) using our deduced continuum shape and Lyman ionization parameters,U1 andU2, these lines are greatly edge Ñux demonstrates that a covering factor of D2is underestimated. In particular, the predicted C IV j1549 is required to produce the Hb luminosity. D4 times too weak and [Ne IV] j2424 and [Ne V] j3426 A covering factor greater than unity indicates that if all of are underpredicted by a factor of D10. Imposing the the line emission in NGC 1052 (extended ] compact) is No. 2, 2000 PHOTOIONIZATION MODELING OF NGC 1052 893 powered by photoionization from the putative nonthermal 5. CONCLUSIONS continuum, then our view of this continuum must be even We have analyzed archival HST /FOS data of the pro- more obscured than we have deduced previously. Further- totypical LINER galaxy NGC 1052. The aperture sampled more, the large covering factor of the inner gas regions the inner D82 ] 82 pc of the nuclear region, thereby calculated in our model suggests that the shape of the ion- excluding the extended emission region. One important izing continuum incident upon the extended region would result is that the deduced Ha/Hb ratio provides evidence for be greatly modiÐed. However, the attenuation of UV con- considerable extinction [E(B[V ) \ 0.42] of the emission- tinuum photons by the emission-line gas in the compact line Ñuxes in NGC 1052. Dust appears to be concentrated in region would result in a much harder SED incident upon or near the nucleus and may also be attenuating the non- the extended gas, and a much higher actual covering factor thermal continuum of a compact source, which supports the for the EELR is required. An alternate explanation is that Barth et al. (1998) conclusion that the nuclei of many the central source emission is beamed in NGC 1052 such LINERs are obscured by dust. that the continuum Ñux ““ seen ÏÏ by the emitting gas in the From our modeling results, we Ðnd that the nuclear emis- extended regions is enhanced. Another possibility is that sion observed in the inner region of NGC 1052 is explained some process other than photoionization by the central well by photoionization from a simple power-law contin- source drives the emission in the EELRÈeither stellar uum having a Ñat spectral index, a \[1.2. Moreover, the photoionization or shock excitation. A detailed analysis of ionizing continuum is consistent with our deduced the di†use emission regions is needed to fully test this. reddening and observed Ñux distribution. This is also con- sistent with the study by Ho (1999), which showed that a 4.2. E†ects of the Choice of Extinction Curve sample of LINERs all have a Ñat SED in the ionizing con- As noted in ° 3, one discrepancy in our model is the tinuum. underestimate of the C II] j2326 Ñux by a factor of D2. The Most of the dereddened line Ñuxes are matched by a C II] j2326 emission is strong in a gas with a low ionization model having two dusty gas components of di†erent den- parameter and high density. However, calculations show sities photoionized by a single power-law continuum. The that the Mg II j2800 Ñux is far overpredicted using these appearance of the C IV j1549, [Ne IV] j2424, and [Ne V] model conditions. Alternately, C II] j2326 is produced pre- j3426 lines in the spectrum indicates a third gas component dominantly in the PIZ of a photoionized gas, suggesting may be present, characterized by a high-ionization param- that a stronger X-ray continuum would strengthen the eter. The key exception to our good model Ðt is the promi- feature. However, the [O I] jj6300, 6363 line Ñuxes are nent C II] j2326 line, which is underpredicted by a factor of more sensitive to the high-energy Ñux and are greatly over- about 2. This Ñux is heavily a†ected by the strength of the estimated if the X-ray continuum is enhanced. 2200Ó absorption feature. Hence, the C II] j2326 line A poor choice of the extinction curve may account for would be overestimated if the 2200Ó bump of the Savage & this discrepancy. Several studies have shown that the UV Mathis extinction curve is not applicable to NGC 1052. dust extinction in extragalactic objects is not represented This would also help to explain the underestimate of the well by the Galactic Savage & Mathis curve (Fitzpatrick [Ne IV] j2424 Ñux. 1985; Calzetti et al. 1994). Often a weaker 2200Ó absorp- The large widths of the emission lines and lack of corre- tion feature and steeper rise in the far-UV are found. As an lation between width and critical density suggests non- ad hoc demonstration, we show how the UV line strengths gravitational motion of the emission-line gas. are a†ected by removing the 2200Ó bump from the extinc- We conclude that a pure central-source photoionization tion curve in Table 6. The [C II] j2326 and [Ne IV] j2424 model with the simplest nonthermal continuum (a simple line Ñuxes are both signiÐcantly reduced, illustrating that if power law) reproduces the emission-line Ñuxes in the inner an extinction law incorporating a strong 2200Ó feature region of NGC 1052 quite well. Other processes, such as were erroneously applied, it could lead to a substantial shocks or photoionization by stars, are not required to overestimate of the intrinsic strengths of these lines. produce the observed emission. However, we cannot rule

TABLE 6 THE EFFECT OF THE 2200Ó FEATURE IN THE EXTINCTION CURVE

ModiÐed Line S & M Curvea S & M Curvea,b Best-Fit Model

C IV j1549 ...... 0.75 (^0.64) 0.75 0.19 He II j1640...... 0.80 (^0.66) 0.80 1.06 [O III] j1663 ...... 0.35 (^0.28) 0.35 0.27 N III] j1750 ...... 0.20 (^0.16) 0.20 0.13 Si III] jj1882, 1892 ...... 0.80 (^0.71) 0.70 0.28 C III] j1909 ...... 2.00 (^1.73) 1.69 2.44 C II] j2326c ...... 3.80 (^3.51) 2.01 2.16 [Ne IV] j2424 ...... 0.10 (^0.09) 0.07 0.01 [O II] j2470 ...... 0.60 (^0.43) 0.39 0.69 a [ \ E(B V )T 0.44. b Savage & Mathis (1979) extinction curve with 2200Ó feature removed, [ \ E(B V )T 0.44. c Contains [O III] j2321 and Si II] j2335 emission. 894 GABEL ET AL. out the possibility that these mechanisms contribute to the and Space Administration. We also would like to extended nebular emission. acknowledge support for the research from NASA grants NAG 5-3378 to the IACS at The Catholic University of This research has made use of the SIMBAD database America. Also, facilities of the Laboratory for Astronomy operated at CDS, Strasbourg, France, and of the NASA/ and Solar Physics at NASA/GSFC were used in performing IPAC Extragalactic Database (NED), which is operated by part of the research. Finally, we are grateful to the referee, the Jet Propulsion Laboratory, California Institute of Gary Ferland, for his instructive comments and sugges- Technology, under contract with the National Aeronautics tions.

REFERENCES Aller, L. H. 1984, Physics of Thermal Gaseous Nebulae (Reidel: Fitzpatrick, E. L. 1985, ApJ, 294, 219 Dordrecht) Fosbury, R. A. E., Mebold, U., Goss, W. M., & Dopita, M. A. 1978, Avni, Y., & Tananbaum, H. 1982, ApJ, 262, L17 MNRAS, 183, 549 (F78) Baldwin, J. A., Ferland, G. J., Martin, P. G., Corbin, M. R., Cota, S. A., Fosbury, R. A. E., Snijders, M. A. J., Boksenberg, A., & Penston, M. V. Peterson, B. M., & Slettebak, A. 1991, ApJ, 374, 580 1981, MNRAS, 197, 235 Barth, A. J., Filippenko, A. V., & Moran, E. C. 1999, ApJ, 515, L61 Grevesse, N., & Anders, E. 1989, in AIP Conf. Proc. 183, Cosmic Abun- Barth, A. J., Ho, L. C., Filippenko, A. V., & Sargent, W. L. W. 1998, ApJ, dances of Matter, ed. C. J. Waddington (New York: AIP), 1 496, 133 Guainazzi, M., & Antonelli, L. A. 1999, MNRAS, 304, L15 Becklin, E. E., Tokunaga, A. T., & Wynn-Williams, C. G. 1982, ApJ, 263, Halpern, J. P., & Steiner, J. E. 1983, ApJ, 269, L37 624 Heckman, T. 1980, A&A, 87, 152 Burstein, D., & Heiles, C. 1982, AJ, 87, 1165 Ho, L. C. 1999, ApJ, 516, 672 Calzetti, D. A., Kinney, A. L., & Storchi-Bergmann, T. 1994, ApJ, 429, 582 Ho, L. C., Filippenko, A. V., & Sargent, W. L. W. 1993, ApJ, 417, 63 Carter, D., Jorden, P. R., Thorne, D. J., Wall, J. V., & Straede, J. C. 1983, ÈÈÈ. 1995, ApJS, 98, 477 MNRAS, 205, 377 ÈÈÈ. 1997, ApJS, 112, 315 Costero, R., & Osterbrock, D. E. 1977, ApJ, 211, 675 Knapp, G. R., & Kerr, F. J. 1974, A&A, 35, 361 Crenshaw, D. M., & Peterson, B. M. 1985, ApJ, 291, 677 Koski, A. T., & Osterbrock, D. E. 1976, ApJ, 203, L49 Davies, R. L., & Illingworth, G. D. 1986, ApJ, 302, 234 Kraemer, S. B., Crenshaw, D. M., Filippenko, A. V., & Peterson, B. M. DeRobertis, M. M., & Osterbrock, D. E. 1984, ApJ, 286, 171 1998, ApJ, 499, 719 de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., Buta, R. J., Kraemer, S. B., Wu, C.-C., Crenshaw, D. M., & Harrington, P. 1994, ApJ, Paturel, G., & Fouque, P. 1991, Third Reference Catalog of Bright 435, 171 Galaxies (New York: Springer) Netzer, H., & Laor, A. 1993, ApJ, 404, L51 Diaz, A. I., Terlevich, E., & Pagel, B. E. J. 1985, MNRAS, 214, 41 Osterbrock, D. E. 1989, Astrophysics of Gaseous Nebulae and Active Dopita, M. A. 1996, in ASP Conf. Proc. 103, The Physics of LINERs in Galactic Nuclei, (Mill Valley: University Science Books) View of Recent Observations, ed. M. Eracleous et al. (San Francisco: Osterbrock, D. E., Tran, H. D., & Veilleux, S. 1992, ApJ, 389, 305 ASP), 44 Pe quignot, D. 1984, A&A, 131, 159 (P84) Dopita, M. A., & Sutherland, R. S. 1995, ApJ, 455, 468 Pritchet, C., & van den Bergh, S. 1977, ApJS, 34, 101 Ferguson, J. W., Korista, K. T., Baldwin, J. A., & Ferland, G. J. 1997, ApJ, Rubin, R. H., Simpson, J. P., Haas, M. R., & Erickson, E. F. 1991, ApJ, 374, 487, 122 564 Ferland, G. J. 1993, in The Nearest Active Galaxies, ed. J. Beckman, Savage, B. D., & Mathis, J. S. 1979, ARA&A, 17, 73 L. Colina, & H. Netzer (Madrid: Consejo Superior de Investigaciones Seaton, M. J. 1978, MNRAS, 185, 5P CientiÐcas), 75 Smith Neubig, M., & Bruhweiler, F. C. 1997, AJ, 114, 1951 ÈÈÈ. 1996, Hazy, A Brief Introduction to Cloudy, Univ. of Kentucky, ÈÈÈ. 1999, AJ, 117, 2856 Dept. of Physics and Astronomy Internal Report Terlevich, R., & Melnick, J. 1985, MNRAS, 213, 841 Ferland, G. J., & Netzer, H. 1983, ApJ, 264, 105 Weaver, K. A., Wilson, A. S., Henkel, C., & Braatz, J. A. 1999, ApJ, 520, Filippenko, A. V. 1985, ApJ, 289, 475 130 Filippenko, A. V., & Terlevich, R. 1992, ApJ, 397, L79 Wrobel, J. M. 1984, ApJ, 284, 531