- Email: [email protected]
The Receding Torus Model—evidence from emission-line luminosities and the quasar fraction
New Astronomy Reviews 47 (2003) 205–209 www.elsevier.com / locate / newastrev
The Receding Torus Model—evidence from emission-line luminosities and the quasar fraction Jennifer A. Grimes a , *, Steve Rawlings a , Chris J. Willott b a
b
Astrophysics, Denys Wilkinson Building, Keble Road, Oxford OX1 3 RH, UK Herzberg Institute of Astrophysics, National Research Council, 5071 West Saanich Road, Victoria, BC, V9 E 2 E7, Canada
Abstract Evidence for the Receding Torus Model for radio-loud AGN comes from two main sources: (i) the rise in quasar fraction with emission-line luminosity; and (ii) the difference in emission-line luminosity between radio galaxies and radio quasars. We have constructed a new type of luminosity function which shows that a two-population model, with a low-luminosity population of radio galaxies and a high-luminosity population composed of radio galaxies and radio quasars, can mimic the effects of a receding torus. The data are best explained, however, by a two-population model with a receding torus in both populations. 2003 Elsevier B.V. All rights reserved. Keywords: Galaxies: active; Radio continuum: galaxies
1. Introduction It has long been known that there is a strong positive correlation between the extended-radio luminosities and narrow-emission-line luminosities of 3C radio sources (Baum and Heckman, 1989; Rawlings et al., 1989), and this correlation has been extended to the 7C Redshift Survey (Willott et al., 1999), implying that it is not primarily a redshift effect. This suggests that the sources of narrow-line emission and radio jets are linked, possibly by accretion rate and / or black hole mass. Emission from the narrow-line region is believed to be independent of jet axis orientation, and thus for simple, arguably naive, unified schemes (where the opening angle of the obscuring torus is constant with ionising *Corresponding author. E-mail address: [email protected] (J.A. Grimes).
luminosity) we expect radio galaxies and radio quasars to have similar distributions of narrow-emission-line luminosities. However, it has been found that values of [OIII] emission lines are on average higher for radio quasars than radio galaxies in properly matched samples (Jackson and Browne, 1990). Because of the scatter in the relationship between the radio and emission-line luminosities, this is compatible with the Receding Torus Model (Simpson, 1998; Lawrence, 1991), where the opening angle of the torus increases with ionising luminosity. However, radio luminosity functions (RLFs) have used two-population models to best fit the data (Willott et al., 2001), and it is possible that a twopopulation model with a simple unified scheme in one population, combined with the effects of scatter, could mimic the effects of the Receding Torus Model in producing both emission line differences between radio galaxies and radio quasars and the
1387-6473 / 03 / $ – see front matter 2003 Elsevier B.V. All rights reserved. doi:10.1016 / S1387-6473(03)00026-5
206
J. A. Grimes et al. / New Astronomy Reviews 47 (2003) 205–209
gradual rise in quasar fraction with emission-line luminosity (Willott et al., 2000).
2. A new type of luminosity function Using a new type of luminosity function, described in detail in Grimes et al. (2003), we can constrain the scatter in the relationship between lowfrequency (151 MHz) radio luminosity L151 and emission-line luminosity L[OIII] , and determine quantitatively if the receding torus is more likely than a simple non-luminosity-dependent unified scheme and find the best-fit unified scheme parameters. We have constructed generalised luminosity functions (GLFs) which predict the space densities of radio galaxies and radio quasars related by a unified scheme at a given value of redshift, L151 and L[OIII] , constrained by the 3CRR, 6CE and 7CRS (parts I and II) complete low-radio-frequency-selected samples. These functions were constructed using a principal components analysis of L151 and L[OIII] to give first and second principal components a and b. The first principal component is essentially the product of L151 and L[OIII] and accounts for about 93% of the scatter in the data. The second principal component is essentially L151 /L[OIII] , and encodes the scatter. A model with a simple broken-power-law in a and Gaussian scatter in b produced the result that a receding torus model (model 1R) is favoured overwhelmingly over the simplest type of unified scheme (model 1S). A plot of the quasar fraction versus a is shown in Fig. 1. From the 3CRR, 6CE and 7CRS data, it is clear that the quasar fraction is zero at low a and rises at high a. It is immediately clear that model 1S cannot explain the quasar fraction data at low values of a, but that model 1R provides a reasonably good fit to the data. However due to the finite size of the error bars, it seems that a step function in a with enough scatter could mimic the effects of a receding torus model, causing a rise in quasar fraction with a. A step function in a could be generated by a two-population model. Historically, radio galaxies have been divided into two classes, FR I and FR II (Fanaroff and Riley, 1974), with a division at log 10 (L151 / WHz 21 sr 21 ) . 25.5. A two-population
Fig. 1. The quasar fraction as a function of the first principal component, a for models 2R (dashed line), 2S (dot-dashed line), 2RC (dotted line), 1R (solid line) and 1S (thick solid line). The ] squares are the binned 3CRR, 6CE and 7CRS data with ŒN errors, (the error for the first bin assumes it contains 1 quasar whereas it contains none).
scheme was used to construct RLFs in Willott et al. (2001), where the break is at log 10 (L[OIII] /W ) . 35 corresponding to log 10 (L151 / WHz 21 sr 21 ) . 26.5. This division is concerned with the optical properties of the system, and thus with the properties of the central engine. The less radio luminous population is composed of FR Is and FR IIs with weak or absent emission lines and has a quasar fraction of . 0.1 and the more radio luminous population consists of strong-line FR II radio galaxies and quasars with a quasar fraction of . 0.4. These populations seem to correspond to an ‘Eddington-tuned’ population and a ‘starved’ population, with the division implying that there are ‘starved’ sources with both FR I and FR II radio structure. There is also evidence (Blundell and Rawlings, 2001) that optically powerful quasars sometimes have FR I radio structures, indicating that the FR I / FR II division is not a fundamental division based on central engine physics. It was found (Chiaberge et al., 2002) that the FR II population is not homogeneous and only a fraction of them can be unified with quasars. They find a low-radiative-efficiency accretion, weak or absent broad-line emission and a lack of a significant nuclear absorbing structure for weak-jet, low-ionisation narrow-line galaxies. For broad line objects and obscured high-
J. A. Grimes et al. / New Astronomy Reviews 47 (2003) 205–209
207
Table 1 The unified schemes used in each population for different models Model
Population 1
Population 2
Model
One-population schemes
2S 2R 2RC
None None Receding torus
Simple Receding torus Receding torus
1S 1R
Simple Receding torus
ionisation narrow-line galaxies, they see or infer intense ionising emission, powerful jets and a toruslike absorber. We investigated the effects of two-population models where the division in population was made as a function of a, rather than radio or optical luminosity. This has the potential to yield interesting results for comparison with other methods of population division as the principal components analysis can identify physically meaningful parameters from various observational signatures, and it is likely that the parameter a represents some combination of accretion rate and black hole mass (Grimes et al., 2003). We have a low-a population composed entirely of radio galaxies (i.e. zero quasar fraction) and a high-a population containing radio galaxies and radio quasars related by a unified scheme, either a receding torus model 2R or a simple non-luminosity-dependent unified scheme 2S. We also looked at a two population model with a receding torus model in
Fig. 2. The 151 MHz radio luminosity function generated from the two-population GLFs using model 2RC (dotted lines) and Model B from Willott et al. (2001) (solid lines) for redshifts z 5 0, 0.5, 1.0, 2.0, 3.0. Models 2S and 2R have very similar RLFs to 2RC and are not shown.
both populations, 2RC (Table 1 lists the unified schemes used in each model). These models were also constrained by source counts and the local RLF (Sadler et al., 2002). Unlike the one-population models, the RLFs produced by these GLFs are very similar to the RLFs of Willott et al. (2001), see Fig. 2.
3. Results The most likely model was found to be 2RC but only by a probability ratio of | 5. This can be seen by the better fit to the data in Fig. 2, the most crucial difference being a more gradual rise in quasar fraction than either 2R or 2S, meaning that there are some quasars at low values of a. The interpretation is that a rise in quasar fraction with emission-line luminosity (thus producing an average value of emission line luminosity which is greater for quasars than radio galaxies) is more likely to be caused by a receding-torus-type model, although it is likely that some sort of two-population model is needed to account for the source count data, and finding a reasonable fit to the RLF (Willott et al., 2001). Monte-Carlo simulations of L151 against L[OIII] from the 2RC and 2S models are compared with the data in Fig. 3. The main difference is that the 2RC model can reproduce the distribution of quasars more accurately than the 2S, especially at low values of L151 , mirroring the effect in quasar fraction at low values of a. This does not necessarily mean that the physics is different in the two populations. The difference, for instance, could be caused by ‘starved quasars’ sometimes accreting and having tori, but whenever tori are present, appearing as receding tori. The value of the constant opening angle for the unified scheme in the high-a population for model 2S was 558, and for the receding torus model in 2RC we find Q0 5 318, for the relation Qtrans 5 tan 21 [tan
208
J. A. Grimes et al. / New Astronomy Reviews 47 (2003) 205–209
Fig. 3. Left: The [OIII] emission line luminosity against the 151 MHz luminosity for the 3CRR (squares), 6CE (triangles) and 7CRS (circles) samples. The simulated L[OIII] vs. L151 for the 2RS model (centre) and 2S model (right) are also shown. Radio quasars are indicated by closed symbols and radio galaxies by open symbols, and the radio galaxies in the low-a population in 2S are shown by larger symbols.
Q0 (L[OIII] /L0 )1 / 2 ] assuming log 10 (L0 / W) 5 35.5. The scatter in the L151 –L[OIII] relationship is found to have a value of | 0.35. This value was previously taken to be 0.6 (Simpson, 1998, 2003). This is remarkably small considering how the observable properties come from such different processes on such different scales.
4. Concluding remarks This new approach to building GLFs produces
datasets which simulate all the features of the data, and things derived from the data: e.g. the RLF, the L151 –L[OIII] correlation, the quasar fraction and emission-line differences between radio galaxies and radio quasars. Smoother RLFs are produced compared to approaches using the same data but not incorporating the effects of scatter (Willott et al., 2001). In the two-population scenario, a Receding Torus Model seems to be slightly favoured over a simple unified scheme but the odds are not large enough to be entirely convincing. There is, however, other
J. A. Grimes et al. / New Astronomy Reviews 47 (2003) 205–209
evidence for the Receding Torus Model (Simpson, 2003) from sources other than the quasar fraction and emission line differences. The effect of radio quasars being brighter than radio galaxies is also seen at submillimetre wavelengths (Willott et al., 2002). If there is a close relationship between optical luminosity and submillimetre luminosity, this result is in agreement with the Receding Torus Model, implying that quasarheated dust dominates the submillimetre luminosity for powerful radio quasars at z | 1.5. We are currently extending these GLFs to include submillimetre luminosities, which should allow firmer conclusions about the necessity of a Receding Torus Model to be made. It was also noted (Willott et al., 2002) that there is an anti-correlation between submillimetre luminosity and radio source age, so that hyperluminous far-IR objects tend to be associated with young ( , 10 7 yr) radio sources, and that the processes controlling submillimetre emission are synchronised with the jet-triggering event. There is also evidence that there is an anti-correlation between the CIV absorbing strength and the projected linear size of steep-spectrum quasars (Baker et al., 2002). We are currently investigating the submillimetre emission from broadabsorption-line quasars, which may be the early stages of outflows leading eventually to the termination of star formation (Rawlings et al., 2003). We will ultimately construct a model which incorporates variations with time as well as orientation.
209
In summary, the Receding Torus Model has many attractions but more work needs to be done before the emission-line differences between radio galaxies and radio quasars and the rise in quasar fraction with luminosity can be cited as conclusive evidence for it.
References Baker, J.C., Hunstead, R.W., Athreya, R.M., Barthel, P.D., de Silva, E., Lehnert, M.D., Saunders, R.D.E., 2002. ApJ 568, 592. Baum, S.A., Heckman, T.M., 1989. ApJ 336, 702. Blundell, K.M., Rawlings, S., 2001. ApJ 562, L5. Chiaberge, M., Capetti, A., Celotti, A., 2002. A&A 394, 791. Fanaroff, B.L., Riley, J.M., 1974. MNRAS 167, 31. Grimes, J.A., Rawlings, S., Willott, C.J., MNRAS, submitted. Jackson, N., Browne, I.W.A., 1990. Nature 343, 43. Lawrence, A., 1991. MNRAS 252, 586. Rawlings, S., Saunders, R., Eales, S.A., Mackay, S.D., 1989. MNRAS 240, 701. Rawlings, S., Willott, C.J., Hill, G.J., Archibald, E.N., Dunlop, J.S., Hughes, D.H., 2003. MNRAS, submitted. Sadler, E.M. et al., 2002. MNRAS 329, 227. Simpson, C., 1998. MNRAS 297, 39. Simpson, C., 2003. NewAR 47, this issue [doi:10.1016 / S13876473(03)00027-7]. Willott, C.J., Rawlings, S., Blundell, K.M., Lacy, M., 1999. MNRAS 309, 1017. Willott, C.J., Rawlings, S., Blundell, K.M., Lacy, M., 2000. MNRAS 316, 449. Willott, C.J., Rawlings, S., Blundell, K.M., Lacy, M., Eales, S.A., 2001. MNRAS 322, 536. Willott, C.J., Rawlings, S., Archibald, E.N., Dunlop, J.S., 2002. MNRAS 331, 435.
The Receding Torus Model—evidence from emission-line luminosities and the quasar fraction Jennifer A. Grimes a , *, Steve Rawlings a , Chris J. Willott b a
b
Astrophysics, Denys Wilkinson Building, Keble Road, Oxford OX1 3 RH, UK Herzberg Institute of Astrophysics, National Research Council, 5071 West Saanich Road, Victoria, BC, V9 E 2 E7, Canada
Abstract Evidence for the Receding Torus Model for radio-loud AGN comes from two main sources: (i) the rise in quasar fraction with emission-line luminosity; and (ii) the difference in emission-line luminosity between radio galaxies and radio quasars. We have constructed a new type of luminosity function which shows that a two-population model, with a low-luminosity population of radio galaxies and a high-luminosity population composed of radio galaxies and radio quasars, can mimic the effects of a receding torus. The data are best explained, however, by a two-population model with a receding torus in both populations. 2003 Elsevier B.V. All rights reserved. Keywords: Galaxies: active; Radio continuum: galaxies
1. Introduction It has long been known that there is a strong positive correlation between the extended-radio luminosities and narrow-emission-line luminosities of 3C radio sources (Baum and Heckman, 1989; Rawlings et al., 1989), and this correlation has been extended to the 7C Redshift Survey (Willott et al., 1999), implying that it is not primarily a redshift effect. This suggests that the sources of narrow-line emission and radio jets are linked, possibly by accretion rate and / or black hole mass. Emission from the narrow-line region is believed to be independent of jet axis orientation, and thus for simple, arguably naive, unified schemes (where the opening angle of the obscuring torus is constant with ionising *Corresponding author. E-mail address: [email protected] (J.A. Grimes).
luminosity) we expect radio galaxies and radio quasars to have similar distributions of narrow-emission-line luminosities. However, it has been found that values of [OIII] emission lines are on average higher for radio quasars than radio galaxies in properly matched samples (Jackson and Browne, 1990). Because of the scatter in the relationship between the radio and emission-line luminosities, this is compatible with the Receding Torus Model (Simpson, 1998; Lawrence, 1991), where the opening angle of the torus increases with ionising luminosity. However, radio luminosity functions (RLFs) have used two-population models to best fit the data (Willott et al., 2001), and it is possible that a twopopulation model with a simple unified scheme in one population, combined with the effects of scatter, could mimic the effects of the Receding Torus Model in producing both emission line differences between radio galaxies and radio quasars and the
1387-6473 / 03 / $ – see front matter 2003 Elsevier B.V. All rights reserved. doi:10.1016 / S1387-6473(03)00026-5
206
J. A. Grimes et al. / New Astronomy Reviews 47 (2003) 205–209
gradual rise in quasar fraction with emission-line luminosity (Willott et al., 2000).
2. A new type of luminosity function Using a new type of luminosity function, described in detail in Grimes et al. (2003), we can constrain the scatter in the relationship between lowfrequency (151 MHz) radio luminosity L151 and emission-line luminosity L[OIII] , and determine quantitatively if the receding torus is more likely than a simple non-luminosity-dependent unified scheme and find the best-fit unified scheme parameters. We have constructed generalised luminosity functions (GLFs) which predict the space densities of radio galaxies and radio quasars related by a unified scheme at a given value of redshift, L151 and L[OIII] , constrained by the 3CRR, 6CE and 7CRS (parts I and II) complete low-radio-frequency-selected samples. These functions were constructed using a principal components analysis of L151 and L[OIII] to give first and second principal components a and b. The first principal component is essentially the product of L151 and L[OIII] and accounts for about 93% of the scatter in the data. The second principal component is essentially L151 /L[OIII] , and encodes the scatter. A model with a simple broken-power-law in a and Gaussian scatter in b produced the result that a receding torus model (model 1R) is favoured overwhelmingly over the simplest type of unified scheme (model 1S). A plot of the quasar fraction versus a is shown in Fig. 1. From the 3CRR, 6CE and 7CRS data, it is clear that the quasar fraction is zero at low a and rises at high a. It is immediately clear that model 1S cannot explain the quasar fraction data at low values of a, but that model 1R provides a reasonably good fit to the data. However due to the finite size of the error bars, it seems that a step function in a with enough scatter could mimic the effects of a receding torus model, causing a rise in quasar fraction with a. A step function in a could be generated by a two-population model. Historically, radio galaxies have been divided into two classes, FR I and FR II (Fanaroff and Riley, 1974), with a division at log 10 (L151 / WHz 21 sr 21 ) . 25.5. A two-population
Fig. 1. The quasar fraction as a function of the first principal component, a for models 2R (dashed line), 2S (dot-dashed line), 2RC (dotted line), 1R (solid line) and 1S (thick solid line). The ] squares are the binned 3CRR, 6CE and 7CRS data with ŒN errors, (the error for the first bin assumes it contains 1 quasar whereas it contains none).
scheme was used to construct RLFs in Willott et al. (2001), where the break is at log 10 (L[OIII] /W ) . 35 corresponding to log 10 (L151 / WHz 21 sr 21 ) . 26.5. This division is concerned with the optical properties of the system, and thus with the properties of the central engine. The less radio luminous population is composed of FR Is and FR IIs with weak or absent emission lines and has a quasar fraction of . 0.1 and the more radio luminous population consists of strong-line FR II radio galaxies and quasars with a quasar fraction of . 0.4. These populations seem to correspond to an ‘Eddington-tuned’ population and a ‘starved’ population, with the division implying that there are ‘starved’ sources with both FR I and FR II radio structure. There is also evidence (Blundell and Rawlings, 2001) that optically powerful quasars sometimes have FR I radio structures, indicating that the FR I / FR II division is not a fundamental division based on central engine physics. It was found (Chiaberge et al., 2002) that the FR II population is not homogeneous and only a fraction of them can be unified with quasars. They find a low-radiative-efficiency accretion, weak or absent broad-line emission and a lack of a significant nuclear absorbing structure for weak-jet, low-ionisation narrow-line galaxies. For broad line objects and obscured high-
J. A. Grimes et al. / New Astronomy Reviews 47 (2003) 205–209
207
Table 1 The unified schemes used in each population for different models Model
Population 1
Population 2
Model
One-population schemes
2S 2R 2RC
None None Receding torus
Simple Receding torus Receding torus
1S 1R
Simple Receding torus
ionisation narrow-line galaxies, they see or infer intense ionising emission, powerful jets and a toruslike absorber. We investigated the effects of two-population models where the division in population was made as a function of a, rather than radio or optical luminosity. This has the potential to yield interesting results for comparison with other methods of population division as the principal components analysis can identify physically meaningful parameters from various observational signatures, and it is likely that the parameter a represents some combination of accretion rate and black hole mass (Grimes et al., 2003). We have a low-a population composed entirely of radio galaxies (i.e. zero quasar fraction) and a high-a population containing radio galaxies and radio quasars related by a unified scheme, either a receding torus model 2R or a simple non-luminosity-dependent unified scheme 2S. We also looked at a two population model with a receding torus model in
Fig. 2. The 151 MHz radio luminosity function generated from the two-population GLFs using model 2RC (dotted lines) and Model B from Willott et al. (2001) (solid lines) for redshifts z 5 0, 0.5, 1.0, 2.0, 3.0. Models 2S and 2R have very similar RLFs to 2RC and are not shown.
both populations, 2RC (Table 1 lists the unified schemes used in each model). These models were also constrained by source counts and the local RLF (Sadler et al., 2002). Unlike the one-population models, the RLFs produced by these GLFs are very similar to the RLFs of Willott et al. (2001), see Fig. 2.
3. Results The most likely model was found to be 2RC but only by a probability ratio of | 5. This can be seen by the better fit to the data in Fig. 2, the most crucial difference being a more gradual rise in quasar fraction than either 2R or 2S, meaning that there are some quasars at low values of a. The interpretation is that a rise in quasar fraction with emission-line luminosity (thus producing an average value of emission line luminosity which is greater for quasars than radio galaxies) is more likely to be caused by a receding-torus-type model, although it is likely that some sort of two-population model is needed to account for the source count data, and finding a reasonable fit to the RLF (Willott et al., 2001). Monte-Carlo simulations of L151 against L[OIII] from the 2RC and 2S models are compared with the data in Fig. 3. The main difference is that the 2RC model can reproduce the distribution of quasars more accurately than the 2S, especially at low values of L151 , mirroring the effect in quasar fraction at low values of a. This does not necessarily mean that the physics is different in the two populations. The difference, for instance, could be caused by ‘starved quasars’ sometimes accreting and having tori, but whenever tori are present, appearing as receding tori. The value of the constant opening angle for the unified scheme in the high-a population for model 2S was 558, and for the receding torus model in 2RC we find Q0 5 318, for the relation Qtrans 5 tan 21 [tan
208
J. A. Grimes et al. / New Astronomy Reviews 47 (2003) 205–209
Fig. 3. Left: The [OIII] emission line luminosity against the 151 MHz luminosity for the 3CRR (squares), 6CE (triangles) and 7CRS (circles) samples. The simulated L[OIII] vs. L151 for the 2RS model (centre) and 2S model (right) are also shown. Radio quasars are indicated by closed symbols and radio galaxies by open symbols, and the radio galaxies in the low-a population in 2S are shown by larger symbols.
Q0 (L[OIII] /L0 )1 / 2 ] assuming log 10 (L0 / W) 5 35.5. The scatter in the L151 –L[OIII] relationship is found to have a value of | 0.35. This value was previously taken to be 0.6 (Simpson, 1998, 2003). This is remarkably small considering how the observable properties come from such different processes on such different scales.
4. Concluding remarks This new approach to building GLFs produces
datasets which simulate all the features of the data, and things derived from the data: e.g. the RLF, the L151 –L[OIII] correlation, the quasar fraction and emission-line differences between radio galaxies and radio quasars. Smoother RLFs are produced compared to approaches using the same data but not incorporating the effects of scatter (Willott et al., 2001). In the two-population scenario, a Receding Torus Model seems to be slightly favoured over a simple unified scheme but the odds are not large enough to be entirely convincing. There is, however, other
J. A. Grimes et al. / New Astronomy Reviews 47 (2003) 205–209
evidence for the Receding Torus Model (Simpson, 2003) from sources other than the quasar fraction and emission line differences. The effect of radio quasars being brighter than radio galaxies is also seen at submillimetre wavelengths (Willott et al., 2002). If there is a close relationship between optical luminosity and submillimetre luminosity, this result is in agreement with the Receding Torus Model, implying that quasarheated dust dominates the submillimetre luminosity for powerful radio quasars at z | 1.5. We are currently extending these GLFs to include submillimetre luminosities, which should allow firmer conclusions about the necessity of a Receding Torus Model to be made. It was also noted (Willott et al., 2002) that there is an anti-correlation between submillimetre luminosity and radio source age, so that hyperluminous far-IR objects tend to be associated with young ( , 10 7 yr) radio sources, and that the processes controlling submillimetre emission are synchronised with the jet-triggering event. There is also evidence that there is an anti-correlation between the CIV absorbing strength and the projected linear size of steep-spectrum quasars (Baker et al., 2002). We are currently investigating the submillimetre emission from broadabsorption-line quasars, which may be the early stages of outflows leading eventually to the termination of star formation (Rawlings et al., 2003). We will ultimately construct a model which incorporates variations with time as well as orientation.
209
In summary, the Receding Torus Model has many attractions but more work needs to be done before the emission-line differences between radio galaxies and radio quasars and the rise in quasar fraction with luminosity can be cited as conclusive evidence for it.
References Baker, J.C., Hunstead, R.W., Athreya, R.M., Barthel, P.D., de Silva, E., Lehnert, M.D., Saunders, R.D.E., 2002. ApJ 568, 592. Baum, S.A., Heckman, T.M., 1989. ApJ 336, 702. Blundell, K.M., Rawlings, S., 2001. ApJ 562, L5. Chiaberge, M., Capetti, A., Celotti, A., 2002. A&A 394, 791. Fanaroff, B.L., Riley, J.M., 1974. MNRAS 167, 31. Grimes, J.A., Rawlings, S., Willott, C.J., MNRAS, submitted. Jackson, N., Browne, I.W.A., 1990. Nature 343, 43. Lawrence, A., 1991. MNRAS 252, 586. Rawlings, S., Saunders, R., Eales, S.A., Mackay, S.D., 1989. MNRAS 240, 701. Rawlings, S., Willott, C.J., Hill, G.J., Archibald, E.N., Dunlop, J.S., Hughes, D.H., 2003. MNRAS, submitted. Sadler, E.M. et al., 2002. MNRAS 329, 227. Simpson, C., 1998. MNRAS 297, 39. Simpson, C., 2003. NewAR 47, this issue [doi:10.1016 / S13876473(03)00027-7]. Willott, C.J., Rawlings, S., Blundell, K.M., Lacy, M., 1999. MNRAS 309, 1017. Willott, C.J., Rawlings, S., Blundell, K.M., Lacy, M., 2000. MNRAS 316, 449. Willott, C.J., Rawlings, S., Blundell, K.M., Lacy, M., Eales, S.A., 2001. MNRAS 322, 536. Willott, C.J., Rawlings, S., Archibald, E.N., Dunlop, J.S., 2002. MNRAS 331, 435.