THE ASTROPHYSICAL JOURNAL, 476:677–684, 1997 February 20
© 1997. The American Astronomical Society. All rights reserved. Printed in U.S.A.

Keck Spectropolarimetry of Two High-z Radio Galaxies: Discerning the Components of the Alignment Effect 1

ANDREA CIMATTI, 2 ARJUN DEY, 3, 4 AND WIL VAN BREUGEL

Institute of Geophysics & Planetary Physics, Lawrence Livermore National Laboratory, Livermore, CA 94550; cimatti@igpp.llnl.gov, dey@igpp.llnl.gov, wil@igpp.llnl.gov

AND

TODD HURT AND ROBERT ANTONUCCI

Physics Department, University of California, Santa Barbara, CA 93106; hurt@chester.physics.ucsb.edu, ski@chester.physics.ucsb.edu

Received 1996 June 24; accepted 1996 September 6


ABSTRACT

     We present optical spectropolarimetric observations obtained with the W. M. Keck Telescope of two powerful, high-redshift radio galaxies which exhibit radio-optical alignments, 3C 13 (z = 1.351) and 3C 356 (z = 1.079). 3C 13 is fairly strongly polarized in the blue, with the electric vector oriented perpendicular to the major axis of UV continuum emission. 3C 356 is known to have two radio/optical components (labeled a and b) along the radio source axis, but it is unclear which of them is the nucleus of the radio source. Our observations show that both components a and b are polarized with the electric vectors in both cases oriented approximately orthogonal to the optical a-b axis. Component a also shows evidence for broad Mg II λ2800 emission both in polarized and total light, while the narrow forbidden lines are unpolarized. Our observations support the unified model of powerful radio-loud active galactic nuclei (AGNs) and allow us for the first time to quantify the contribution of the different radiative components to the alignment effect of a high-z radio galaxy (3C 356a): the nonstellar radiation (scattered and nebular continua) constitutes about 80% of the total UV continuum emission at 2800 Å, and an evolved stellar population with an age ∼1.5–2.0 Gyr can account for the remainder of the UV light. We also detect the stellar Ca II K absorption line in the spectra of both components. Although the present data do not clarify unambiguously whether a or b is the nucleus of 3C 356, they suggest that the scenario in which a contains the hidden quasar is energetically more favorable. If the nucleus is located in a, our observations show that electron scattering is plausible, and support the scenario in which 3C 356 is surrounded by an ionized intracluster medium, as suggested by ROSAT observations.

Subject headings: galaxies: active—galaxies: individual (3C 13, 3C 356)—polarization—quasars: general—radio continuum: galaxies—scattering—ultraviolet: galaxies

FOOTNOTES

     1 Based on observations made at the W. M. Keck Observatory.

     2 Present address: Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, I-50125, Firenze, Italy; cimatti@arcetri.astro.it.

     3 Also Astronomy Department, University of California at Berkeley, Berkeley, CA 94720.

     4 Present address: NOAO/KPNO, 950 N. Cherry Avenue, P.O. Box 26732, Tucson, AZ 85726.

§1. INTRODUCTION

     During the past few years, a considerable effort has been made to understand the origin of the mysterious UV-radio alignments (the “alignment effect;” see McCarthy 1993 for a recent review) that are observed in high-redshift radio galaxies (HzRGs). While the relative contributions of stellar and nonstellar radiation to the rest-frame UV emission from HzRGs is not yet known in detail, polarimetric observations of a number of radio galaxies have demonstrated that a substantial fraction of their UV light is due to scattering of anisotropic radiation from obscured quasar nuclei (di Serego Alighieri et al. 1989; see Cimatti 1996 for a recent review). The advent of the large-aperture W. M. Keck 10 m telescope has now rendered it possible to obtain high signal-to-noise ratio, spatially resolved spectropolarimetric data for faint, distant HzRGs and thereby probe their hidden active nuclei, separate the stellar and nonstellar components, and study their ambient interstellar media. For example, Keck spectropolarimetric observations of 3C 324 (z = 1.206) and 3C 256 (z = 1.824) have shown that scattered light can account for a large fraction of the total spatially extended UV continuum emission (Cimatti et al. 1996; Dey et al. 1996a). Furthermore, in 3C 324, polarized broad permitted Mg II λ2800 line emission is also detected clearly. The presence of such polarized, scattered continuum and broad-line radiation is expected in the AGN unification scenario of powerful radio sources, where FR II–type narrow-line radio galaxies harbor a quasar nucleus not directly visible because of obscuration and orientation effects (see Antonucci 1993 and Urry & Padovani 1995 for recent reviews).

     In this paper, we present our observations of two more HzRGs: 3C 13 at a redshift z = 1.351, and 3C 356 at z = 1.079. Although both galaxies are good examples of the alignment effect, 3C 356 is of particular interest, since it exhibits the alignment effect at both optical and near-infrared wavelengths (Eales & Rawlings 1990, hereafter ER90; Lacy & Rawlings 1994, hereafter LR94). This has led to the suggestion that the alignment effect in 3C 356 is a result of star formation triggered by a jet rather than scattered light from a quasar. The observations we present call this model into question, and therefore we discuss 3C 356 in some detail. Throughout this paper we assume H0 = 50 km s-1 Mpc-1 q0 = 0 and define h50 = H0/50. For this cosmology, the projected linear scale is 11.2 h$\mathstrut{^{-1}_{50}}$ kpc arcsec-1 and 11.9 h$\mathstrut{^{-1}_{50}}$ kpc arcsec-1, respectively, for z = 1.079 (3C 356) and z = 1.351 (3C 13).

§2. OBSERVATIONS

     We observed 3C 13 and 3C 356 at the Cassegrain focus of the 10 m W. M. Keck telescope using the Low Resolution Imaging Spectrometer (LRIS) in polarimetric mode on UT 1995 June 30 and July 1, respectively. The 1&farcs;5 wide slit was oriented along the major axes of the UV continuum emission (P.A. = 145° and 144°, respectively, for 3C 13 and 3C 356). 3C 13 was observed in two sets: the first consisted of four exposures of 20 minutes each, with the half-wave plate rotated to 0°, 45°, 22&fdg;5, and 67&fdg;5; the second set consisted of three exposures of 15 minutes each (the 67&fdg;5 exposure was not completed). 3C 356 was observed in three sets of 4 × 20 minutes, 4 × 10 minutes, and 4 × 15 minutes. We also observed a faint star located 3&farcs;2 east and 15&farcs;6 south of 3C 356 in order to check for any Galactic polarization. Unpolarized/flux (BD +33°2642) and polarized (HD 155528) standard stars were observed to calibrate the data. Both the spectra of 3C 13 and 3C 356 were extracted using an aperture of 10 pixels (2&farcs;14). More details about the observations and data reduction can be found in our previous papers (Cimatti et al. 1996; Dey et al. 1996a).

§3. RESULTS

§3.1. 3C 13

     The radio galaxy 3C 13 (z = 1.351) shows a clumpy, rest-frame UV continuum structure that is aligned with the radio structure (PAUV = 155°, PAradio = 145°; Le Fèvre et al. 1988a; see Law-Green et al. 1995 for a radio map). 3C 13 is faint, and the low signal-to-noise ratios of our individual spectra do not allow a detailed sampling of P(λ) across the spectrum. However, usng large spectral bins we detect continuum polarization at ≥3 σ level (see Fig. 1). Pcont(λ) and &thetas;(λ) appear to be wavelength independent within the errors. The average value of &thetas;(λ) = 63° is orthogonal to the UV continuum axis rather than to the radio axis. The errors in the emission-line bins are very large, but we note that the bin including the Mg II λ2800 line does not show a decrease in P as seen for the other emission lines (e.g., C II] λ2326, [Ne IV] λ2424, [O II] λ3727) and that &thetas; does not show strong rotation, suggesting that Mg II λ2800 may be polarized. The perpendicular polarization implies that scattering of nuclear anisotropic radiation is occurring, in agreement with the results generally found on distant radio galaxies (e.g., Cimatti et al. 1993).

Fig. 1

     Le Fèvre et al. (1988a) have proposed that at least part of the aligned UV continuum of 3C 13 is due to gravitational lens amplification caused by a foreground galaxy at z = 0.48 located about 4&arcsec; south of 3C 13. Although our observations cannot rule out the possibility of gravitational amplification in this case, the detection of polarization suggests that the alignment in 3C 13, as in other HzRGs, is more likely caused by the scattering of anisotropically radiated nuclear emission.

§3.2. 3C 356

     Unlike most powerful z ∼ 1 HzRGs, 3C 356 (z = 1.079) exhibits the alignment effect at both optical and near-infrared wavelengths. Its optical and near-infrared morphology are double, with the two components separated by 4&farcs;8 (∼54 h$\mathstrut{^{-1}_{50}}$ kpc) at an angle of 17° relative to the axis of the radio source hot spots (LR94). Keeping with tradition (Le Fèvre, Hammer, & Jones 1988b; ER90; LR94), we will refer to these northern and southern components as a and b, respectively. Figures 2 and 3 show our observations of 3C 356a and 3C 356b, respectively. Component a shows very high continuum polarization, with P(λ) rising toward the blue up to P ∼ 15%. The electric vector position angle is constant across the spectrum (average &thetas; = 64°), and the polarized flux spectrum is rather flat. The Mg II λ2800 line is detected in polarized light, and its continuum-subtracted polarization is P = 6.8% ± 1.2% and &thetas; = 64° ± 5° (including the narrow component of the line, which is presumably unpolarized). On the other hand, the narrow emission lines are unpolarized (e.g., P[O II] = 1.8% ± 0.7%). The low or null polarization of the narrow emission lines suggests low reddening, both intrinsic to 3C 356a and due to our Galaxy, also in agreement with the low polarization of the nearby faint field star (Pstar ∼ 0.1%–0.3%).

Fig. 2 Fig. 3

     Object b is much fainter than a, but using broad continuum spectral bins it is possible to detect low polarization (P = 4.0% ± 1.2% and 2.8% ± 0.8% in the two bluest continuum bins). The electric vector angle &thetas; is offset by about -30° ± 10° relative to that of a. The Mg II λ2800 and [O II] λ3727 lines are too faint to allow a significant measurement of their polarization.

     The position angle of the line joining a and b is 145° both in the optical and K bands, and the radio source is at PA = 162° (Le Fèvre et al. 1988b; ER90; LR94). Inspection of Hubble Space Telescope Wide-Field Planetary Camera (HST WFPC2) image of 3C 356 (1700 s in filter F622W; courtesy of P. Best) shows that component a consists of two components oriented at PA = 152°. Therefore, the average position angle of the E-vector in a is orthogonal to the optical HST structure of a (Δ&thetas; ∼ 88°), and moreover it is roughly perpendicular to both the optical a-b direction (Δ&thetas; ∼ 81°) and the radio source axis (Δ&thetas; ∼ 82°).

§4. IMPLICATIONS FOR 3C 356

     As noted previously, the optical and near-infrared properties of a and b and their alignment with respect to the radio axis has led to the suggestion that the alignment effect in 3C 356 is due to jet-induced star formation rather than scattering of light from a hidden active nucleus (ER90; LR94). Of the two components that are observed in the optical and near-infrared, the northern one (a) exhibits high-ionization emission lines and a blue continuum spectrum, whereas b shows a lower ionization spectrum and is very red. The main question in 3C 356 is the location of the nucleus. The arguments presented by LR94 in favor of b being the nucleus are (1) its core radio source has a flat radio spectrum at high frequencies &parl0;α$\mathstrut{^{5\,{\rm GHz}}_{8\,{\rm GHz}}}$∼0.1; Sν ∼ ν-α), (2) it lies directly on the line joining the primary radio hot spots (Fernini et al. 1993), and (3) its K-band magnitude follows the K-z relation of radio galaxies, while a lies at ∼2 σ from it. However, we note that arguments (2) and (3) are rather weak and also that the radio spectrum of a shows a flattening at lower frequencies (LR94), rendering argument (1) not completely unambiguous.

     Best, Longair, & Röttgering (1996), based on recent HST images of 3C 356, have proposed that the radio/optical aligned multiple component morphology of component a is evidence that this is the actual parent galaxy of the 3C 356 radio source. However, we note that, also in this case, this argument alone is not conclusive.

     Although the location of the nucleus is not clear, LR94 adopted b as the radio source core and concluded that the jet-induced star formation scenario can explain successfully the observed spectrum of a, with young stars (a few × 107 yr old) dominating the UV continuum emission and shocks (resulting from the interaction of the radio jet with the interstellar medium) producing the high-ionization emission lines. Scattering and photoionization by an AGN were considered implausible for the following reasons: (i) the required luminosity of the incident quasar light would be too high; (ii) the a-b alignment is also present in the near-IR which, at λrest ∼ 1 μm, is presumably representative of stellar light; and (iii) the emission-line ratios of a are reproduced better by a shock model that includes local photoionization (ER90; LR94).

§4.1. The Hidden Quasar and the Nature of the UV Continuum

§4.1.1. The Broad Mg II λ2800 Line and the Scattered Light

     The strongly polarized continuum and broad Mg II λ2800 emission line observed in a show, however, that there is a significant contribution of scattered light from a hidden active nucleus and broad-line region. Figure 4 suggests that a broad component of the Mg II λ2800 line is visible also in the total flux spectrum. In order to verify this, we followed the method outlined in Dey & Spinrad (1996) and modeled the spectral region of the Mg II λ2800 feature with a multicomponent model. The model consisted of a broad Gaussian (for the broad component of the Mg II λ2800 feature) and six narrow Gaussians corresponding to the narrow emission lines of He II λ2733, [Mg V] λ2783, Mg II λλ2796, 2805, He I λ2829, and O III λ2836 (note that the [Mg V] λ2783 and He I λ2829 features were identified incorrectly by Dey & Spinrad with [Mg VII] λ2786 and O III λ2826, respectively). During the fitting, the narrow emission lines were constrained to have all the same velocity width typical of the narrow lines in 3C 356a (∼800 km s-1). The fitting was performed using the software SPECFIT in IRAF (Kriss 1994). After the fit, the narrow-line component model was subtracted from the observed spectrum of 3C 356a. The result is shown in Figure 4, where the broad hump centered at ∼2800 Å due to the Mg II λ2800 line is now clearly evident. It flux is 1.3 × 10-16 ergs s-1 cm-2, corresponding to a luminosity of 1.6 × 1042 h$\mathstrut{^{-2}_{50}}$ ergs s-1. Its velocity width is FWHM = 9250 ± 700 km s-1, and the rest-frame equivalent width is 13.3 ± 3 Å, both typical of the broad-line regions observed in quasars. This brings even more evidence in favor of the idea of obscuration, beaming, and scattering of quasar light, independently on the location of the quasar in a or b. We recall also that 3C 265 (Dey & Spinrad 1996) and, possibly, 3C 324 (Cimatti et al. 1996) show broad Mg II λ2800 emission in their total flux spectra.

Fig. 4

     Under the assumption that the broad Mg II λ2800 line and the scattered continuum emission have the same fractional polarization, the detection of the broad Mg II λ2800 emission in the total flux spectrum is crucial because it allows us to estimate the fraction of unpolarized direct light at 2800 Å, R2800 = Funpol/Ftot = 1 - (EWtot/EWpol), where Funpol and Ftot are the unpolarized and total continuum fluxes, and EWtot and EWpol are the equivalent widths of the broad Mg II λ2800 line in total flux and polarized flux, respectively (see di Serego Alighieri, Cimatti, & Fosbury 1994; Tran 1995; Cimatti et al. 1996).

     The rest-frame equivalent width of the polarized Mg II λ2800 (26.9 ± 4.8 Å) has been estimated by fitting a Gaussian curve to the points that sample the line in polarized flux. We note that the measured value of EWpol of the Mg II λ2800 line is similar to that of the average spectrum of radio-loud quasars obtained by Cristiani & Vio (1990) (see for instance the bottom panel of Fig. 2 and Fig. 5).

     Therefore, we estimate that the contribution of the unpolarized radiation to the total flux at 2800 Å is R2800 = 0.50 ± 0.15, implying that about half the total UV flux at this wavelength comes from scattered radiation. This allows us to estimate that the undiluted intrinsic degree of polarization of the continuum is P(2800) = 21% ± 7%. This level of polarization may be characteristic of either electron or dust scattering (Manzini & di Serego Alighieri 1996), and therefore it does not provide strong clues to the nature of the scattering particles.

§4.1.2. The Unpolarized Light

     The unpolarized UV light at 2800 Å can be due both to stellar light and nebular continuum emission. For example, Stockton, Ridgway, & Kellogg (1996) have found that in 3C 368 (z = 1.132) a large fraction of the UV continuum is due to nebular emission. In order to derive the contributions of these components, we have calculated first the nebular continuum emission according to the prescriptions of Dickson et al. (1995) and Manzini & di Serego Alighieri (1996). The intensity of the nebular continuum is largely determined by the Hβ intensity, and the dependence on the temperature and density is weak. We estimated its intensity using the observed Hδ flux and assuming an intensity ratio Hδ/Hβ = 0.26 typical of case B. Since dust reddening can affect the Hδ flux, we have estimated the possible extinction using the He II λ3206/He II λ2733 ratio. We measure F(He II λ3206)/F(He II λ2733) = 2.1, slightly higher than the case B value ∼1.9, and we estimate EB-V = 0.11 using the Galactic extinction curve (Mathis, Rumpl, & Nordsieck 1977). Our estimate is in reasonable agreement with that of LR94 based on the He II λ3206/He II λ1640 ratio (EB-V = 0.08). Finally, for case B and T = 15,000 K, ne = 100 cm-3, taking into account bremsstrahlung emission, recombination continuum emission, and emission due to the two-photon process, and correcting for the dust reddening, we estimate that the nebular continuum contributes to 26% ± 10% of the UV continuum at 2800 Å.

     We can conclude that 80% ± 18% of the total UV flux at 2800 Å is due to nonstellar light (scattered+nebular continua), and 20% ± 18% can be ascribed to stellar radiation. If we assume that the K-band flux is dominated by stellar light (K = 17.9; ER90) and adopt the synthetic spectra of Bruzual & Charlot (1993; instantaneous burst, solar metallicity, Salpeter IMF), we find that a stellar population about 1.5–2.0 Gyr old can account completely for the stellar light fraction at 2800 Å.

     To summarize, the results discussed in this section suggest strongly that the UV continuum of a is dominated not by young (∼107 yr) stars (as claimed by LR94), but by nonstellar radiation (scattered+nebular continua). The modeling of the continuum and the emission-line spectrum of a will be presented elsewhere, but we note here that our measured line ratios are consistent with those of LR94.

§4.2. Where is the Quasar Nucleus?

     In this section we discuss how our observations can help to determine the location of the hidden quasar. Reassessing all the available data, there are three possible explanations for the observed properties of components a and b: (1) as suggested by LR94, b is the host galaxy of the 3C 356 radio source, and a is a companion galaxy that has wandered into b's radiation cone and jet; (2) a is the host galaxy of the 3C 356 radio source, and b is a nearby flat spectrum radio galaxy in a's beam; and (3) a and b are two unrelated radio sources, with b being the old, red host of the 3C 356 radio source and a being a young, compact steep-spectrum radio source. The latter two explanations both rely on the coincidence that the two sources both happen to be radio galaxies; case 3 requires also that their observed polarization position angles be similar for no physical reason, and therefore it is very unappealing.

     Can the observed polarization properties answer the question of the location of the nucleus? One way to address the question is to estimate if a typical 3C quasar can account for the observed scattered radiation luminosity observed in a, depending on its location in a or in b.

§4.2.1. Quasar Located in b

     Under the assumption that the hidden quasar nucleus is located in b and illuminates a, we can estimate the nuclear luminosity with the following formula (see ER90 and Cimatti et al. 1996 for more details):



where Lscattered and Lquasar are, respectively, the luminosity of the scattered and incident radiation, Dp is the projected a-b distance, α is the angle between the line of sight and the a-b radius vector, &thetas; = 180° - α is the scattering angle, &phis;s(&thetas;) is the phase function of the scattering particles, and τs = &angl0;ns&angr0;σs 2r is the scattering optical depth, where r is the radius of the scattering cloud and &angl0;ns&angr0; and σs are, respectively, the average particle number density of the cloud and the scattering cross section of the scatterers. For the dust grains, we assume an average radius of the particles of 0.1 μm (typical of Galactic dust) and a scattering phase function &phis;dust(90°) = 0.05 (see Cimatti et al. 1996 for more details).

     A deep HST+WFPC2 image obtained with the filter F622W (Δλrest ∼ 2700–3200 Å; courtesy of P. Best) shows that component a is made by two subcomponents each ∼0&farcs;6 in size. It is important to recall here that the two subcomponents of a are oriented approximately along the a-b axis: we call them a1 and a2, where a1 is the component facing b. The HST image shows that a1 contributes to ∼60% of the total flux (a1 + a2) in filter F622W. We investigate two cases: in the first we assume that all the observed scattered light comes from a1 + a2, and in the second we assume that it comes only from a1.

     1. We adopt an a-b distance of 54 h$\mathstrut{^{-1}_{50}}$ kpc, a scattering angle α = 90°, and a total size 2r = 1&farcs;1 (∼13 h$\mathstrut{^{-1}_{50}}$ kpc) for the component a = a1 + a2. For the adopted values, equation (1) cannot be solved for τs, but it can provide a useful lower limit on the quasar luminosity in the optically thick regime (τs ≫ 1). If we adopt the observed luminosity of the broad scattered Mg λ2800 line as derived in § 4.1, Lscatt(Mg II λ2800) = 1.6 × 1042 ergs s-1 h$\mathstrut{^{-2}_{50}}$, we find that Lquasar (Mg II λ2800) > 4.6 × 1044 ergs s-1 h$\mathstrut{^{-2}_{50}}$ for electron scattering. If we apply the same calculation to the scattered continuum luminosity at λrest = 2800 Å, Lcont,scatt(2800 Å) ∼ 1.4 × 1029 ergs s-1 Hz-1 h$\mathstrut{^{-2}_{50}}$, we find that Lcont,quasar(2800 Å) > 4.1 × 1031 ergs s-1 Hz-1 h$\mathstrut{^{-2}_{50}}$ for electron scattering. In case of dust scattering, similar results are obtained.

     2. Instead of using the whole size and luminosity of a1 + a2, we can assume that all the scattered radiation comes only from a1 because it faces b. In this case, according to equation (1), since the size of a1 is 2r = 0&farcs;6 and its scattered radiation luminosity is 60% of the total, the required quasar luminosities are higher. For electron scattering, we obtain >9.3 × 1044 ergs s-1 h$\mathstrut{^{-2}_{50}}$ and >8.3 × 1031 ergs s-1 Hz-1 h$\mathstrut{^{-2}_{50}}$ for the Mg II λ2800 line and the 2800 Å continuum, respectively. In the case of dust scattering, the results are very similar. The maximum luminosities observed in 3C quasars at z ∼ 1–1.5 are Lmax,quasar ∼ 1.1 × 1045 ergs s-1 and ∼9.0 × 1031 ergs s-1 Hz-1, respectively, for the Mg II λ2800 and the continuum at 2800 Å (see Cimatti et al. 1996).

     Therefore, the main result is that, if the quasar is located in b, its required luminosity must be close to that of the most luminous 3C quasars. Although this is possible, we regard it as unlikely (see § 4.2.2). For example, if the scattered radiation comes mostly from a1, a 3C quasar located in b would require α = 90° because a small change of α would increase Lquasar and push it beyond the luminosity range of 3C quasars. In addition, based on photoionization calculations, LR94 concluded that the observed emission-line spectrum from a is unlikely to be powered by a photoionizing beam from a quasar nucleus located in b, since this would require a hidden quasar about 300 times more luminous than the typical 3C quasars. We note also that the lower limits that we derive for Lquasar are so close to the maximum luminosity of 3C quasars that this result may be considered not too different from that derived by LR94 with photoionization calculations. However, if the nucleus is located in b, then 3C 356 may be regarded as a high-z counterpart of nearby radio galaxies in which an extranuclear cloud interacts with the radio jet and scatters the radiation emitted by a hidden quasar nucleus (e.g., PKS 2152-69: di Serego Alighieri et al. 1988; 3C 277.3: van Breugel et al. 1985 Dey et al. 1996b; 3C 321: Cimatti & di Serego Alighieri 1995; Young et al. 1996; and PKS 1414-211: Dey et al. 1996b).

§4.2.2. Quasar Located in a

     If the quasar is located in a, all the energetic problems are largely alleviated, and equation (1) can be solved for τs in order to have an estimate of the density and total mass of the scattering particles. We adopt the average luminosity of the Mg II λ2800 observed in 3C quasars, 2.9 × 1044 ergs s-1 (Cimatti et al. 1996), assume that the hidden quasar is located between a1 and a2, and consider the scattered light coming from a1 (60% of the total scattered light). If we adopt a projected distance between a1 and the quasar Dp = 0&farcs;6 = 6.72 kpc, 2r = 0&farcs;6, and α = 90°, we derive that the optical depth of the electrons is τe = 0.06. This optical depth is consistent with that estimated by Crawford & Fabian (1993) based on ROSAT PSPC observations of 3C 356, where the density of the hot gas in the central core is derived to be τe > 0.01 if the density profile varies inversely with radius as in nearby cooling flows. Assuming a spherical and homogeneous distribution of the scatterers within a1, we find that ne = 4.5 cm-3 and that the ionized gas total mass is Mion ∼ 1.6 × 1010 M&sun;. A similar calculation for dust, scattering provides Mdust ∼ 3 × 105 M&sun;.

     These reasonable masses and the less extreme energetics suggest slightly that the quasar may be located more naturally in a. Another piece of evidence in favor of this scenario is that the electric vector of the polarized radiation is perpendicular to the a1-a2 axis rather than to the a-b or the radio axes (see § 3.2). Deep multifrequency Very Large Array (VLA) observations would be important to study in more detail the spectral characteristics of a and b in the radio and provide new constraints to dissolve the remaining ambiguity.

§4.3. Electron Scattering and the Massive Cooling Flow

     The question of the nature of the scattering particles is particularly important in 3C 356 because ROSAT HRI observations have suggested the presence of extended X-ray emission ascribed to the most distant intracluster medium known to date (Crawford & Fabian 1996), then rendering the possibility of electron scattering particularly interesting.

     Dust scattering cannot be neglected because we know, from the estimate of reddening (§ 4.1), that dust is present in the ISM of 3C 356 and in general in HzRGs (Cimatti 1996). Figure 5 shows that the polarized flux spectrum of component a, although noisy in the red region, can be reproduced reasonably with the average spectrum of radio-loud quasars (Cristiani & Vio 1990) scattered either by Galactic dust (model of Manzini & di Serego Alighieri 1996) or by electrons. However, in the case of dust scattering we note that EB-V = 0.05 is required in order to redden the scattered light, which otherwise would be too blue to match the observed polarized flux spectrum. However, we would like to stress here that the properties of dust scattering are poorly known, and that the validity of the available scattering models has been called into question recently (see Manzini & di Serego Alighieri 1996 for a recent discussion of this issue). In addition, the problem of dust scattering is complicated even more by our ignorance of the dust properties and spatial distribution in high-z galaxies. Although our result cannot exclude dust scattering, we emphasize that what is relevant here is that, if the quasar is located in a, the observed scattered light luminosity can be ascribed realistically to electron scattering because of the reasonable values of τe, ne, and Mion (see previous section), and because of the shape of the polarized flux spectrum.

Fig. 5

     Assuming that a hosts the hidden quasar, and that Thomson scattering is dominant, we can estimate the temperature of the electrons and investigate the physical state of the ionized (scattering) gas. If we consider the component a1, located at a distance Dp = R = 6.72 kpc from the quasar (assumed to be at the center of the cluster), we can estimate the temperature of the scattering electrons by comparing the observed FWHMobs of the scattered broad Mg II λ2800 line (9250 km s-1, see § 4.1) with that of a typical quasar (see Dey & Spinrad 1996; Cimatti et al. 1996). In particular, the maximum temperature can be estimated by using the minimum FWHM of the Mg II λ2800 line observed in quasars (FWHMmin ∼ 2000 km s-1; Baldwin, Wampler, & Gaskell 1989, Brotherton et al. 1994). With this method, we estimate that Te,max ≤ 0.00597 &parl0;FWHM$\mathstrut{^{2}_{{\rm obs}}}$-FWHM$\mathstrut{^{2}_{{\rm min}}}$&parr0;∼ 5 × 105 K. We note here that, if a cooling flow is present, a contribution by hot electron scattering must also occur if their Thomson optical depth is large enough (Fabian 1989). The effect of scattering by hot electrons would be to smear out partially the scattered broad Mg II λ2800 line and to reduce its equivalent width in the polarized flux spectrum.

     It is important to notice that the estimate of the temperature of the scattering electrons is a sort of upper limit: since the FWHM of the broad component of the Mg II λ2800 line in total flux is about 9000 km s-1, and since the FWHM of this line in quasars can be as high as ∼10,000 km s-1 (Brotherton et al. 1994), this implies that the electron temperature can be lower than 5 × 105 K.

     In general, we would like to stress that our spectropolarimetry does not prove that electron scattering is dominant, but it suggests that scattering by electrons can be realistic based on the estimates of τe, ne, and MH II, which are more plausible than the ones usually inferred in other high-z radio galaxies (see, e.g., Cimatti et al. 1993; di Serego Alighieri et al. 1994).

     In the most massive cooling flows known at z < 0.5 (Fabian & Crawford 1995, and references therein), the typical densities of the hot gas at a distance R ∼ 10 kpc from the center of the cD galaxy are of the order of 0.06–0.20 cm-3, with temperatures T ∼ 1–5 × 107 K. Therefore, the conditions of the scattering ionized gas in 3C 356 at a radius R = 6.72 kpc (Te ≤ 5 × 105 K, ne = 4.5 cm-3) are different from those derived by X-ray observations of cooling flows at a comparable radial distance from the center. In particular, the electron density is larger than that estimated for the hot gas but smaller than that derived by optical observations of the inner (2–5 kpc) regions of cooling flows, where the average densities are of the order of ne ∼ 100 cm-3 (Heckman et al. 1989; Fabian 1994). This may be due to the fact that our estimate, based on the scattered light, samples a different physical and/or geometrical region. Furthermore, it is important to note that our evaluation of ne provides only a value of the density averaged within a sphere of radius 6.72 kpc, whereas the ionized gas may be distributed more realistically in smaller and denser clumps within that radius. What is important here is to note that our estimated optical depth τe is compatible with that derived by Crawford & Fabian (1993) for the specific case of 3C 356.

     Another argument in favor of a dense environment in the central parts of 3C 356 comes from the estimate of the pressure of the gas. The pressure, P = 3ne kT ≤ 9.3 × 10-10 dyn cm-2, estimated within a radius R = 6.72 kpc according to the electron scattering requirements of ne and Te, is very high and comparable to that expected and estimated in several massive cooling flows (Heckman et al. 1989; Fabian 1994) and requires a large cooling atmosphere around 3C 356. The emission-line spectra of cooling flows show usually low ionization (Fabian 1994). The very high ionization observed in the emission lines of 3C 356a suggests that the hidden quasar may be the additional source of ionizing radiation.

     If massive cooling flows are common around high-z radio galaxies, this could explain the apparent “mass paradox” present in the case of electron-scattered radiation in HzRGs (see Cimatti 1996, and references therein). In fact, the large masses of ionized gas required to explain the scattered radiation luminosities are in contrast with the idea that the host galaxies are evolved massive ellipticals with (by definition) a small amount of ISM available for scattering. The paradox may be solved assuming the presence of massive halos of hot gas around the host galaxies that can provide a reservoir from which the cool and dense clumps in the scattering medium can be formed.

§4.4. The Infrared Alignment of a and b

     The detection of stellar features can provide additional information about the nature of objects a and b. Our high-signal-to-noise ratio data confirm the presence of the Ca II K λ3933 absorption line (which was detected marginally by LR94) in the total flux spectrum of a (Fig. 2) and show that it is also present in the spectrum of b. The line has a rest-frame equivalent width of 3.8 ± 0.5 Å in a and 6.2 ± 1.0 Å in b. These are very similar to the Ca II K stellar absorption line observed in 3C 265 by Dey & Spinrad (1996).

     The equivalent widths of Ca II K suggest that the absorption line has a stellar origin; the interstellar Ca II K lines in nearby galaxies generally have equivalent widths of <1 Å (see Albert 1983, and references therein; Crawford 1992). In nearby normal elliptical galaxies, the Ca II K line has equivalent widths between 10 Å to 20 Å (e.g., Alloin, Arimoto, & Bica 1989), and the smaller width observed in a suggests a strong dilution of the starlight by the nonstellar scattered continuum, as well as a younger stellar population compared to nearby elliptical galaxies.

     Contrary to the claim of LR94, who concluded that a is much younger than b, it is more likely that a is an old galaxy, possibly coeval with b, its blue color and somewhat smaller Ca II K equivalent width due not to a younger age and jet-induced star formation, but to the nonstellar scattered component and nebular continuum emission.

     If we assume that a and b are members of a bound system, then given their observed relative velocity of 1325 ± 140 km s-1, the total mass of the two galaxies is ∼2.2 × 1013 M&sun;. If we assume also that both galaxies have the same mass-to-light ratio, and partition of the masses according to their K-band magnitudes (ER90), then a and b have total masses of roughly 6 and 16 × 1012 M&sun;, respectively.

     If a and b are two nearby, possibly bound, galaxies, their IR alignment with the radio axis can be explained either by a chance orientation of the a-b axis, or within the scenario of anisotropic mergers and formation of cD galaxies at high z (West 1994). Also, the presence of a massive cooling flow (Crawford & Fabian 1996) strengthens the idea that this radio galaxy lies in a dense cluster environment. However, we note that 3C 356 must be regarded as an unusual case because most of the residual IR alignments observed in 3C galaxies are probably due to other (nonstellar) processes (Dunlop & Peacock 1993).

§5. CONCLUSIONS

     We have used the spectropolarimeter on the W. M. Keck Telescope to observe two z ∼ 1 radio galaxies, 3C 13 and 3C 356. Both galaxies show polarized rest-frame UV continuum emission, with the electric vector oriented perpendicular to the UV major axis. In particular, 3C 356a shows broad Mg II λ2800 emission both in total and polarized light, allowing us to estimate that the nonstellar radiation (scattered+nebular continua) contributes about 80% of the total flux at 2800 Å, while the rest may be due to stellar light coming from an evolved (∼1.5–2.0 Gyr old) host galaxy, implying that the putative starburst component cannot dominate the UV continuum.

     These observations provide strong support for the AGN unification hypothesis, which relies on the effects of projection, obscuration, and collimated jets and ionizing radiation to explain many of the observed properties of radio galaxies and quasars.

     Our data do not allow us to determine unambiguously the location of the hidden quasar in 3C 356. However, we argue that component a is more likely the location of the nucleus because the scenario in which b contains the hidden quasar is energetically unfavorable. Nevertheless, the possibility that b is the nucleus cannot be ruled out completely by the available data, and more observations are required to answer this question.

     If the nucleus is located in a, the luminosity of the quasar-scattered radiation can be explained entirely with electron scattering by the dense ionized gas expected to be present in the central regions of the intracluster medium detected by ROSAT. In this regard, our observations show for the first time how the central regions of cooling flows around AGNs may be explored by means of optical spectropolarimetry.

ACKNOWLEDGMENTS

     We are grateful to the anonymous referee for the useful suggestions and constructive criticism. We thank Sperello di Serego Alighieri, Marshall Cohen, Bob Fosbury, Michael Gregg, Gerard Kriss, Hy Spinrad, and Hien Tran for useful discussions. A. C. is grateful to Philip Best for providing the HST image of 3C 356, to Sperello di Serego Aligheri and Bob Dickson for the information about the nebular continuum, and to the Institute of Geophysics and Planetary Physics for providing the opportunity to pursue this work. The W. M. Keck Observatory is a scientific partnership between the University of California and the California Institute of Technology, made possible by the generous gift of the W. M. Keck Foundation. This work was performed at IGPP/LLNL under the auspices of the US Department of Energy under contract W-7405-ENG-48. R. A. acknowledges the NSF (AST 93-21441). T. H. acknowledges partial support from IGPP-LLNL.

REFERENCES

FIGURES


Full image (37kb) | Discussion in text
     FIG. 1.—The spectral and polarization properties of 3C 13. From top to bottom: the total flux spectrum, the percentage polarization, the position angle of the electric vector, and the polarized flux spectrum. Filled circles and crosses indicate continuum and emission lines with their underlying continuum, respectively. The discrepant polarization in the bin between [Ne V] λ3346 and [Ne V] λ3426 is likely due to skyline residuals.

Full image (42kb) | Discussion in text
     FIG. 2.—The spectral and polarization properties of 3C 356a. From top to bottom: the total flux spectrum, the percentage polarization, the position angle of the electric vector, and the polarized flux spectrum. Filled circles and crosses indicate continuum and emission lines with their underlying continuum, respectively. The discrepant bin at λobs ∼ 8400 Å is likely due to a skyline residual effect. The continuum line in the P × F plot is a radio-loud quasar average spectrum (Cristiani & Vio 1990) scattered by electrons.

Full image (36kb) | Discussion in text
     FIG. 3.—The spectral and polarization properties of 3C 356b. From top to bottom: the total flux spectrum, the percentage polarization, the position angle of the electric vector, and the polarized flux spectrum. Filled circles and crosses indicate continuum and emission lines with their underlying continuum, respectively.

Full image (36kb) | Discussion in text
     FIG. 4.—The broad component of Mg II λ2800 in the total flux spectrum. Top: The observed spectrum with the fitted model (see text). Bottom: The broad hump of Mg II λ2800 after the subtraction of the narrow-line model (see text).

Full image (29kb) | Discussion in text
     FIG. 5.—The polarized flux spectrum of 3C 356a. Top: A fit obtained with Galactic dust scattering plus reddening with EB-V = 0.05 (model of Manzini & di Serego Alighieri 1996). Bottom: The fit obtained in case of electron scattering. In both cases, it has been assumed that the incident spectrum is that of an average radio-loud quasar (Cristiani & Vio 1990).