A significant fraction of Seyfert galaxies show evidence for star formation in the circumnuclear regions (Wilson 1988). It appears that star formation is more important for Seyfert 2s than 1s (Mulchaey et al. 1994), although this may be a selection effect, since most Seyfert 2s were discovered in Markarian surveys, biased toward blue colours, such as would arise from ongoing star formation.
The narrow-line ionized gas in Seyferts has several extranuclear components. Firstly, the NLR close to the nucleus has either a smooth morphology centered on the nucleus or, more often, a complex filamentary structure with a bipolar or linear appearance. This gas is probably photoionized by the continuum photons from the AGN, as evidenced by the emission line ratios, and it sometimes exhibits large noncircular kinematics. The second component consists of the ENLR (Unger et al. 1987). The low velocity dispersion, orderly velocity field and high excitation indicate that the ENLR is ambient gas in the disc or halo of the galaxy photoionized by the (anisotropic) nuclear radiation field. The last component of ionized gas is knotty or clumpy and is usually associated with spiral arms or circumnuclear rings. The emission line ratios indicate photoionization by OB stars in giant H II regions and this gas is kinematically part of the rotating disk of the galaxy. In a minority of Seyferts, this component includes a circumnuclear starburst on scales from a few hundred pc to a few kpc. The colour maps presented in this paper relate mainly to this component.
Similarly, nonthermal radio sources in Seyferts can be divided into two morphologies. The majority have linear (double or triple) morphology straddling the optical nucleus and are probably fuelled by the AGN. The second, less frequent, class have diffuse or blob-like morphology. These are associated with the star forming regions seen in the optical and presumably represent the integrated emission from multiple SNRs or free-free emission (Miley et al. 1985; Edelson et al. 1987).
There is also diversity in the MIR and FIR spectra of Seyferts.
Many Seyfert 2s exhibit prominent dust features in the 
spectra similar to starbursts, whereas Seyfert 1s
generally lack them (Roche et al. 1991), except those with
circumnuclear starbursts, e.g. NGC 1365, NGC 7469 and NGC 7582.
Many Seyfert 2s have FIR colours similar to starburst galaxies,
whereas Seyfert 1s resemble quasars (Miley et al. 1985). The MIR
(
10 
m) emitting region in many Seyferts is spatially
extended over kpc scales, necessitating a distributed heating
source, probably young stars (Edelson
et al. 1987).
The only galaxy in our sample showing clear circumnuclear starburst activity is NGC 7469. The diffuse radio morphology, IRAS colours and radio spectral index are typical of starburst galaxies (Wilson et al. 1986; Condon & Broderick 1988). The emission line morphology is different from the aligned NLR often found in Seyferts (e.g. Haniff et al. 1988) and the emission line ratios all argue that much of the star formation occurs in the ring of diameter 3'' around the Seyfert nucleus. This ring is clearly resolved in our V-I colour map (Plate 10).
Can any of the other detected features in our sample be due to star formation? In NGC 1068, NGC 4151, Mrk 3 and Mrk 573, the radio structure is linear while the CO structure of NGC 1068 and NGC 4151 is diffuse and not peaked near the nucleus. For these galaxies, a more attractive explanation is scattering of the nuclear light along the radio axis (see Sect. 6.2). The radio structure of NGC 3227 is diffuse as in starbursts and the relatively good correspondence between the CO emission and the B-I map and the colours of the blue maxima could be due to star formation. However, there is no direct evidence of star formation occurring in the circumnuclear region of NGC 3227, while other evidence (see Schmidt & Miller 1985) strongly point towards scattering.
The colour images presented in this paper offer a new, independent method to test the unified models of AGN, and the technique should be applied to a larger sample. Unified models of AGN have become recently increasingly popular (e.g. Barthel 1989; Antonucci 1993). In these models, all Seyferts have the same basic structure but appear as either type 1 or 2 (or intermediate types) depending on obscuration and their orientation with respect to our line-of-sight.
Optical spectropolarimetry of NGC 1068 (Antonucci & Miller 1985) showed that this archetypal Seyfert 2 galaxy has polarised broad emission lines, characteristic of a Seyfert 1 nucleus obscured from our direct view by an optically and geometrically thick torus (see also Krolik & Begelman 1986). A fraction of the nuclear luminosity is scattered (polarised) into our line of sight by electrons in a tenuous, warm gas along the axis of the torus. The commonly accepted view now is that the distinction between broad and narrow lined AGN may simply be due to our viewing angle. In Seyfert 1s, we look along the axis of the obscuring torus, and directly see the nucleus and the BLR, whereas in Seyfert 2s the torus blocks our direct view of the nucleus, which can only be seen in reflected polarised light. Several other Seyfert 2 galaxies (e.g. Miller & Goodrich 1990; Tran et al. 1992; Kay 1994) indeed have broad lines in polarised light. In addition, the presence of broad NIR lines in many Seyfert 2s reveals highly obscured BLRs (e.g. Blanco et al. 1990; Nakajima et al. 1991; Goodrich et al. 1994; Ruiz et al. 1994).
In the unified model, the ionising nuclear radiation is
collimated by the torus and escapes anisotropically along the
torus axis, resulting in two oppositely directed cones. The ENLR
in Seyferts is indeed preferentially aligned along the radio
source (torus) axis (see Wilson & Tsvetanov 1994). In several
cases, [OIII]/H
images directly show the conical or
bi-conical geometry of the high excitation gas (e.g. Pogge 1989;
Tadhunter & Tsvetanov 1989; Haniff et al.\
1991; Tsvetanov &
Walsh 1992) and the cone regions have spectra consistent with
photoionization by the nuclear continuum.
Other lines of evidence also support this model. The ionizing
photon flux required to produce the line emission inferred from
the ionization state and energy balance of the gas in the NLR
and ENLR is higher than that inferred from directly
observed continuum in Seyfert 2s (e.g.
cite[Wilson et al. 1988]104;
Kinney et al. 1991), indicating that the NLR gas sees a stronger
ionising source than observed from Earth. This collimated
ionising radiation could result either from dust obscuration of
an isotropic source or from an intrinsically anisotropic source
(Acosta-Pulido et al. 1990). Finally, hard X-ray emission and
the Fe K
X-ray line has been found in several Seyfert 2s
(e.g. Koyama et al. 1989; Warwick et al. 1989; Awaki et al. 1990).
The inferred absorption column densities are constant in time
and in the range 
for all
Seyfert 2s except NGC 1068, therefore, hard X-rays can pass
through the torus (Krolik & Lepp 1989). In NGC 1068, the column
density is much larger, and the detected hard X-rays are due to
a scattered nuclear component.
What is then the origin of the blue extensions and maxima visible in the colour maps, most clearly in the case of NGC 1068, NGC 3227, Mrk 3 and Mrk 573? They are unlikely to be intrinsically blue regions of star formation because there is no direct evidence for this in the continuum images (although see Macchetto et al. 1994 for probable star clusters in the near nuclear region of NGC 1068), and because the morphology of the regions is diffuse, unlike the sharp boundaries seen e.g. in the OB associations further out in the disk of NGC 1068. Optical synchrotron emission from the radio jets is also unlikely, because there is no detailed correspondence between the blue maxima and the radio hot spot structure and the extended blue continuum is more extended than the radio continuum. The most likely remaining alternative is scattering of light from the Seyfert nucleus by dust or electrons. In either case the scattered continuum would appear blue, but more so for dust scattering because of the wavelength dependence. For NGC 1068, electron scattering in the nucleus is preferred, because of the wavelength-independent polarisation in the UV (Code et al. 1993), but in the circumnuclear region dust also contributes (Miller et al. 1991).
The blue knots in NGC 1068 and Mrk 3 are significantly offset
from the nuclei, as are their [OIII] peaks. In Mrk 573, on the
other hand, the blue maxima are even further away and lie beyond
the inner radio continuum structure, whereas the [OIII] emission
peaks on the nucleus. Similarly, NGC 1068 and Mrk 3 are
classical examples of Seyfert 2 nuclei with polarised broad
lines revealed by spectropolarimetry indicating a hidden Seyfert
1 nucleus (Antonucci & Miller 1985; Miller & Goodrich 1991),
whereas Mrk 573 has only slightly polarized broad H
line
(Kay 1994) and it does not show prominent blue knots
very near the nucleus (although see the inner pair of blue arcs
in Pogge & De Robertis 1995). The off-center blue continuum
knots may thus be ``mirrors'' scattering the nuclear light. A
tentative size of the scattering region of 
has been derived for NGC 1068 from multiaperture
polarimetry (Antonucci et al. 1994). It may be that Mrk 573
lacks a strong scattering mirror, with most of the scattering
distributed over a larger region, producing correspondingly
weaker polarisation. In NGC 1068 and Mrk 3 there are
unobstructed clouds close enough to the nucleus to produce
strong polarisation. This is confirmed for NGC 1068, where there
is a strongly polarised region NE of the nucleus (Miller et al.\
1991) with the orientation of the polarisation vector
perpendicular to the line of sight to the nucleus, coinciding
with the extended blue feature seen in our colour map. Similar
imaging polarimetry and spectropolarimetry of Mrk 3 and Mrk 573
would be needed to resolve the question for them.
In the simplest scattering model, an optically thick molecular
torus surrounds the BLR to obscure the ionizing photons, and the
difference between objects with (Seyfert 1) or without (Seyfert
2) BLR depends solely on our line of sight to the torus. Is
this model valid for the blue knots seen in Seyfert type 1
galaxies NGC 4151 and NGC 3227, in which we see the BLR
directly? Emission line imaging of NGC 4151 by Evans et al.\
(1993) and Boksenberg et al. (1995) indicates that our line of
sight lies outside the ionisation cones, and the central BLR
should not be visible directly through the thick torus. Evans
et al. and Boksenberg et al. considered several possibilities to
reconcile the geometry of NGC 4151 with unified models, and
prefer an optically thick torus surrounded by lower density
atmosphere that collimates the ionising radiation but leaves the
nucleus relatively unobscured.
Another model was proposed by Robinson et al.\
(1994), who suggested that ionising radiation escapes from the
nucleus in a broad cone (opening angle 
120
), inclined
close to the galactic disk. The ENLR is formed as the
intersection of the radiation cone with the disk, and the wide
opening angle accounts for the misalignment between the ENLR and
the nuclear radio source. In their model, the opaque torus
required to collimate the radiation field is composed of
bar-driven inflow of dust and gas toward the nucleus.
Mrk 78 and Mrk 348 exhibit only weakly resolved blue structures outside nucleus. Do all Seyfert 2s then have hidden Seyfert 1 nuclei? Miller & Goodrich (1990) calculated the expected polarisations from randomly oriented scattering cones, and found that polarisations up to over 50% should be common. While such high polarisations have not been observed, this may be understood in the unified model, if scattering occurs close to the nucleus, and the scattering region itself is largely obscured. Especially in highly inclined galaxies, the dust lanes in the host galaxy can block the view to the scattering region. The remaining observed polarisation would also be strongly diluted by starlight. Similarly, our failure to convincingly detect the scattering region in these ``good candidates for hidden Seyfert 1s'' may mean that the electron scattering region is not present, not resolved and/or does not have sufficient optical depth to produce notable colour gradients.
Plate 1. The B band image of NGC 1068. The size of the
image is 
(
).
Knots 2-4 are marked on the
map. In this and other plates, north is up, east to the left and
the nucleus of the galaxy in the exact center of the plate. (Add
the period to all captions)
Plate 2. The B-I map of NGC 1068. The size of the image is
(
).
In this and other colour maps, dark
shades indicate blue and light shades red emission. The B-I
colour coding is from 1.1 to 3.8
Plate 3. The B-I map of the innermost 
(
)
of NGC 1068, showing the blue elongation parallel to the
radio jet axis. The B-I colour coding is from 1.4 to 2.7
Plate 4. The B band image of NGC 3227. The scale of the
full image is 
(
)
Plate 5. The B-I map of NGC 3227. The scale of the full
image is 
(
).
The B-I colour coding is from
0.50 to 5.3. Note the complex structure
Plate 6. The B-I map of the innermost 
(
)
of NGC 3227. The B-I colour coding is from 0.87 to 4.6. Note
the two blue maxima straddling the nucleus
Plate 7. The B band image of NGC 4151. The size of the
image is 
Plate 8. The B-I map of the innermost 
(
)
of NGC 4151. The B-I colour coding is from 1.9 to 3.3. Note
the blue elongation parallel to the radio jet axis and two blue
maxima on either side of the nucleus
Plate 9. The V band image of NGC 7469. The size of the
image is 
Plate 10. The V-I map of the innermost 
(
) of NGC 7469. The V-I
colour coding is from 1.1 to 1.6.
Note the circumnuclear ring with a diameter of 3''
Plate 11. The B band image of Mrk 3. The size of the image
is 
Plate 12. The B-I map of the innermost 
(
) of Mrk 3. The B-I
colour coding is from 1.8 to 4.2. Note
the elongated blue region with two maxima
Plate 13. The V-R ``excitation'' map of the innermost
(
) of Mrk 3. The V-R
colour coding is from
0.38 to 1.1. Note the tightly wound spiral structure and the
two red maxima
Plate 14. The B band image of Mrk 78. The size of the
image is 
Plate 15. The B-I map of the innermost 
(
) of Mrk 78. The B-I
colour coding is from 1.2 to 4.0. Note
the blue arclike structure N of the nucleus
Plate 16. The B band image of Mrk 348. The size of the
image is 
Plate 17. The B-I map of the innermost 
(
) of Mrk 348. The B-I
colour coding is from 0.19 to 4.0
Plate 18. The B band image of Mrk 573. The size of the
image 
Plate 19. The B-I map of the innermost 
(
) of Mrk 573. The B-I colour coding is from 0.14
to 4.2. Note the blue double structure and the red bridge
perpendicular to them. There is also evidence for a faint blue
spiral structure further away from the nucleus
Plate 20. The B band image of NGC 1667. The size of the
image is 
(
)
Plate 21. The B-I map of the innermost 
(
) of NGC 1667. The B-I
colour coding is from 0.47 to 3.4.
The H II regions in the spiral arms and the dust geometry are
clearly resolved
Plate 22. The B-I map of the innermost 
(
) of NGC 1667. The B-I
colour coding is from 0.56 to 3.4.
Note the double structure of red material across the nucleus