An ultraviolet-selected galaxy redshift survey – II. The physical nature of star formation in an enlarged sample

We present further spectroscopic observations for a sample of galaxies selected in the vacuum ultraviolet (UV) at 2000 Å from the FOCA balloon-borne imaging camera of Milliard et al. This work represents an extension of the initial study by Treyer et al. Our enlarged catalogue contains 433 sources (≃3 times as many as in our earlier study) across two FOCA fields. 273 of these are galaxies, nearly all with redshifts z ≃0–0.4. Nebular emission-line measurements are available for 216 galaxies, allowing us to address issues of excitation, reddening and metallicity. The UV and H α luminosity functions strengthen our earlier assertions that the local volume-averaged star formation rate is higher than indicated from earlier surveys. Moreover, internally within our sample, we do not find a steep rise in the UV luminosity density with redshift over 0< z <0.4. Our data are more consistent with a modest evolutionary trend, as suggested by recent redshift survey results. Investigating the emission-line properties, we find no evidence for a significant number of AGN in our sample; most UV-selected sources to z ≃0.4 are intense star-forming galaxies. We find that the UV flux indicates a consistently higher mean star formation rate than that implied by the H α luminosity for typical constant or declining star formation histories. Following Glazebrook et al., we interpret this discrepancy in terms of a starburst model for our UV-luminous sources. We develop a simple algorithm which explores the scatter in the UV flux–H α relation in the context of various burst scenarios. Whilst we can explain most of our observations in this way, there remains a small population with extreme UV–optical colours which cannot be understood.     

Radio sources in the 2dF Galaxy Redshift Survey – II. Local radio luminosity functions for AGN and star-forming galaxies at 1.4 GHz

We have cross-matched the 1.4-GHz NRAO VLA Sky Survey (NVSS) with the first 210 fields observed in the 2dF Galaxy Redshift Survey (2dFGRS), covering an effective area of 325 deg² (about 20 per cent of the final 2dFGRS area). This yields a set of optical spectra of 912 candidate NVSS counterparts, of which we identify 757 as genuine radio identifications – the largest and most homogeneous set of radio source spectra ever obtained. The 2dFGRS radio sources span the redshift range     to 0.438, and are a mixture of active galaxies (60 per cent) and star-forming galaxies (40 per cent). About 25 per cent of the 2dFGRS radio sources are spatially resolved by NVSS, and the sample includes three giant radio galaxies with projected linear size greater than 1 Mpc. The high quality of the 2dF spectra means we can usually distinguish unambiguously between AGN and star-forming galaxies. We make a new determination of the local radio luminosity function at 1.4 GHz for both active and star-forming galaxies, and derive a local star formation density of         .     

Star formation rates and chemical abundances of emission-line galaxies in intermediate-redshift clusters

We examine the evolutionary status of luminous, star-forming galaxies in intermediate-redshift clusters by considering their star formation rates (SFRs) and the chemical and ionization properties of their interstellar emitting gas. Our sample consists of 17 massive, star-forming, mostly disc galaxies with   M_B ≲−20  , in clusters with redshifts in the range  0.31 ≲ z ≲ 0.59  , with a median of  〈 z 〉= 0.42  . We compare these galaxies with the identically selected and analysed intermediate-redshift field sample of Mouhcine et al., and with local galaxies from the Nearby Field Galaxy Survey of Jansen et al.
From our optical spectra, we measure the equivalent widths of  [O  ii ]λ3727, Hβ  and [O  iii ]λ5007 emission lines to determine diagnostic line ratios, oxygen abundances and extinction-corrected SFRs. The star-forming galaxies in intermediate-redshift clusters display emission-line equivalent widths which are, on average, significantly smaller than measured for field galaxies at comparable redshifts. However, a contrasting fraction of our cluster galaxies have equivalent widths similar to the highest observed in the field. This tentatively suggests a bimodality in the SFRs per unit luminosity for galaxies in distant clusters. We find no evidence for further bimodalities, or differences between our cluster and field samples, when examining additional diagnostics and the oxygen abundances of our galaxies. This maybe because no such differences exist, perhaps because the cluster galaxies which still display signs of star formation have recently arrived from the field. In order to examine this topic with more certainty, and to further investigate the way in which any disparity varies as a function of cluster properties, larger spectroscopic samples are needed.     

Optical and infrared diagnostics of SDSS galaxies in the SWIRE survey

We present the rest-frame optical and infrared colours of a complete sample of  1114 z < 0.3  galaxies from the Spitzer Wide-Area Infrared Extragalactic (SWIRE) Legacy Survey and the Sloan Digital Sky Survey (SDSS). We discuss the optical and infrared colours of our sample and analyse in detail the contribution of dusty star-forming galaxies and active galactic nuclei (AGN) to optically selected red sequence galaxies.
We propose that the optical  ( g − r )  colour and infrared  log( L ₂₄/ L _3.6)  colour of galaxies in our sample are determined primarily by a bulge-to-disc ratio. The  ( g − r )  colour is found to be sensitive to the bulge-to-disc ratio for disc-dominated galaxies, whereas the  log( L ₂₄/ L _3.6)  colour is more sensitive for bulge-dominated systems.
We identify ∼18 per cent (195 sources) of our sample as having red optical colours and infrared excess. Typically, the infrared luminosities of these galaxies are found to be at the high end of star-forming galaxies with blue optical colours. Using emission-line diagnostic diagrams, 78 are found to have an AGN contribution and 117 are identified as star-forming systems. The red  ( g − r )  colour of the star-forming galaxies could be explained by extinction. However, their high optical luminosities cannot. We conclude that they have a significant bulge component.
The number densities of optically red star-forming galaxies are found to correspond to ∼13 per cent of the total number density of our sample. In addition, these systems contribute ∼13 per cent of the total optical luminosity density, and 28 per cent of the total infrared luminosity density of our SWIRE/SDSS sample. These objects may reduce the need for 'dry mergers'.     

Ultraviolet dust attenuation in star-forming galaxies – II. Calibrating the A(UV) versus LTIR/LUV relation

We investigate the dependence of the total-IR (TIR) to ultraviolet (UV) luminosity ratio method for calculating the UV dust attenuation   A (UV)  from the age of the underlying stellar populations by using a library of spectral energy distributions for galaxies with different star formation histories. Our analysis confirms that the TIR/UV versus   A (UV)  relation varies significantly with the age of the underlying stellar population: that is, for the same TIR/UV ratio, systems with low specific star formation rate (SSFR) suffer a lower UV attenuation than starbursts. Using a sample of nearby field and cluster spiral galaxies, we show that the use of a standard (i.e. age-independent) TIR/UV versus   A (UV)  relation leads to a systematic overestimate up to 2 mag of the amount of UV dust attenuation suffered by objects with low SSFR and in particular H  i -deficient star-forming cluster galaxies. This result points out that the age-independent TIR/UV versus   A (UV)  relation cannot be used to study the UV properties of large samples of galaxies including low star-forming systems and passive spirals. Therefore, we give some simple empirical relations from which the UV attenuation can be estimated taking into account its dependence on the age of the stellar populations, providing a less biased view of UV properties of galaxies.     

A sample of radio-loud active galactic nuclei in the Sloan Digital Sky Survey

A sample of 2712 radio-luminous galaxies is defined from the second data release of the Sloan Digital Sky Survey (SDSS) by cross-comparing the main spectroscopic galaxy sample with two radio surveys: the National Radio Astronomy Observatories (NRAO) Very Large Array (VLA) Sky Survey (NVSS) and the Faint Images of the Radio Sky at Twenty centimeters (FIRST) survey. The comparison is carried out in a multistage process and makes optimal use of both radio surveys by exploiting the sensitivity of the NVSS to extended and multicomponent radio sources in addition to the high angular resolution of the FIRST images. A radio source sample with 95 per cent completeness and 98.9 per cent reliability is achieved, far better than would be possible for this sample if only one of the surveys was used. The radio source sample is then divided into two classes: radio-loud active galactic nuclei (AGN) and galaxies in which the radio emission is dominated by star formation. The division is based on the location of a galaxy in the plane of 4000-Å break strength versus radio luminosity per unit stellar mass and provides a sample of 2215 radio-loud AGN and 497 star-forming galaxies brighter than 5 mJy at 1.4 GHz. A full catalogue of positions and radio properties is provided for these sources. The local radio luminosity function is then derived both for radio-loud AGN and for star-forming galaxies and is found to be in agreement with previous studies. By using the radio to far-infrared (FIR) correlation, the radio luminosity function of star-forming galaxies is also compared to the luminosity function derived in the FIR. It is found to agree well at high luminosities but less so at lower luminosities, confirming that the linearity of the radio to FIR correlation breaks down below about 10²² W Hz⁻¹ at 1.4 GHz.     

The relationship between star formation rate and radio synchrotron luminosity at 0 < z < 2

We probe the relationship between star formation rate (SFR) and radio synchrotron luminosity in galaxies at  0 < z < 2  within the northern Spitzer Wide-area Infrared Extragalactic survey (SWIRE) fields, in order to investigate some of the assumptions that go into calculating the star formation history of the Universe from deep radio observations. We present new 610-MHz Giant Metrewave Radio Telescope (GMRT) observations of the European Large-Area ISO Survey-North 2 (ELAIS-N2) field, and using this data, along with previous GMRT surveys carried out in the ELAIS-N1 (North 1) and Lockman Hole regions, we construct a sample of galaxies which have redshift and SFR information available from the SWIRE survey. We test whether the local relationship between SFR and radio luminosity is applicable to   z = 2  galaxies, and look for evolution in this relationship with both redshift and SFR in order to examine whether the physical processes which lead to synchrotron radiation have remained the same since the peak of star formation in the Universe. We find that the local calibration between radio luminosity and star formation can be successfully applied to radio-selected high-redshift, high-SFR galaxies, although we identify a small number of sources where this may not be the case; these sources show evidence for inaccurate estimations of their SFR, but there may also be some contribution from physical effects such as the recent onset of starburst activity, or suppression of the radio luminosity within these galaxies.     

Chemical abundances in a UV-selected sample of galaxies

We discuss the chemical properties of a sample of UV-selected intermediate-redshift  (0≲z≲0.4)  galaxies in the context of their physical nature and star-formation history. This work represents an extension of our previous studies of the rest-frame UV-luminosity function (Treyer et al.) and the star-formation properties of the same sample (Sullivan et al.) . We revisit the optical spectra of these galaxies and perform further emission-line measurements restricting the analysis to those spectra with the full set of emission lines required to derive chemical abundances. Our final sample consists of 68 galaxies with heavy-element abundance ratios and both UV and CCD B -band photometry. Diagnostics based on emission-line ratios show that all but one of the galaxies in our sample are powered by hot, young stars rather than by an AGN. Oxygen-to-hydrogen (O/H) and nitrogen-to-oxygen (N/O) abundance ratios are compared with those of various local and intermediate-redshift samples. Our UV-selected galaxies span a wide range of oxygen abundances, from ∼0.1 to 1 Z_⊙, intermediate between low-mass H  ii galaxies and massive starburst nuclei. For a given oxygen abundance, most have strikingly low N/O values. Moreover, UV-selected and H  ii galaxies systematically deviate from the usual metallicity–luminosity relation in the sense of being more luminous by  2–3 mag  . Adopting the 'delayed-release' chemical evolution model, we propose our UV-selected sources are observed at a special stage in their evolution, following a powerful starburst that enriched their ISM in oxygen and temporarily lowered their mass-to-light ratios. We discuss briefly the implications of our conclusions on the nature of similarly selected high-redshift galaxies.     

Hubble Space Telescope imaging survey of sub-mJy star-forming galaxies – I. Morphologies at z ∼ 0.2

We present the first results of our Hubble Space Telescope HST WFPC2 F814W snapshot imaging survey, targeting virtually all sub-mJy decimetric radio-selected star-forming galaxies. The radio selection at ∼1 GHz is free from extinction effects and the radio luminosities are largely unaffected by AGN contamination, making these galaxies ideal tracers of the cosmic star formation history. A subsample of four targets is presented here, selected at 1.4 GHz from the spectroscopically homogenous and complete samples of Benn et al. and Hopkins et al. The redshifts are confined to a narrow range around z ∼0.2, to avoid differential evolution, with a radio luminosity close to L ∗ where the galaxies dominate the comoving volume-averaged star formation rate. We find clearly disturbed morphologies resembling those of ultraluminous infrared galaxies, indicating that galaxy interactions may be the dominant mechanism for triggering star formation at these epochs. The morphologies are also clearly different from those of coeval quasars and radio galaxies, as found in star-forming galaxies selected at other wavelengths. This may prove challenging for models that propose direct causal links between AGN evolution and the cosmic star formation history at these epochs. The asymmetries are typically much larger than seen in the Canada–France Redshift Survey at similar redshifts, optical luminosities and H α -derived star formation rates, indicating the possible existence of an obscuration-related morphological bias in such samples.     

A far-infrared view of the Lockman hole from ISO 95-μm observations – II. Optical identifications and insights into the nature of the far-infrared sources

We present the optical identifications of a 95-μm ISOPHOT sample in the Lockman hole over an area of approximately half a deg². The Rodighiero et al. catalogue includes 36 sources, making up a complete flux-limited sample for   S _95 μm≥ 100 mJy  . Reliable sources were detected, with decreasing but well-controlled completeness, down to   S _95 μm≃ 20 mJy  . We have combined mid-infrared (IR) and radio catalogues in this area to identify the potential optical counterparts of the far-IR sources. We found 14 radio and 13 15-μm associations, 10 of which have both associations. For the 11 sources with spectroscopic redshift, we have performed a spectrophotometric analysis of the observed spectral energy distributions (SEDs). Four of these 95-μm sources have been classified as faint IR (FIR) galaxies  ( L _FIR < 1. e 11 L_⊙)  , six as luminous IR galaxies (LIRGs) and only one as an ultraluminous IR galaxy (ULIRG). We have discussed the redshift distribution of these objects, comparing our results with evolutionary model predictions 95 and 175 μm. Given their moderate distances (the bulk of the closest spectroscopically identified objects lying at   z < 0.2  ), their luminosities and star formation rates (SFR; median value  ∼ 10 M_⊙ yr⁻¹  ), the sources unveiled by ISOPHOT at 95 μm seem to correspond to the low redshift  ( z < 0.3)  FIRBACK 175-μm population, composed of dusty, star-forming galaxies with moderate SFRs. We computed and compared different SFR estimators, and found that the SF derived from the bolometric IR luminosity is well correlated with that computed from the radio and mid-IR fluxes.     

The 2dF Galaxy Redshift Survey: galaxy luminosity functions per spectral type

We calculate the optical b _J luminosity function (LF) of the 2dF Galaxy Redshift Survey (2dFGRS) for different subsets defined by their spectral properties. These spectrally selected subsets are defined using a new parameter, η , which is a linear combination of the first two projections derived from a Principal Component Analysis. This parameter η identifies the average emission- and absorption-line strength in the galaxy rest frame spectrum, and hence is a useful indicator of the present star formation. We use a total of 75 000 galaxies in our calculations, chosen from a sample of high signal-to-noise ratio, low-redshift galaxies observed before 2001 January. We find that there is a systematic steepening of the faint-end slope ( α ) as one moves from passive  ( α =-0.54)  to active  ( α =-1.50)  star-forming galaxies, and that there is also a corresponding faintening of the rest frame characteristic magnitude   M *-5 log₁₀( h )  (from −19.6 to −19.2). We also show that the Schechter function provides a poor fit to the quiescent (Type 1) LF for very faint galaxies  [ M _{b J}-5 log₁₀( h )  fainter than −16.0], perhaps suggesting the presence of a significant dwarf population. The LFs presented here give a precise confirmation of the trends seen previously in a much smaller preliminary 2dFGRS sample, and in other surveys. We also present a new procedure for determining self-consistent k -corrections, and investigate possible fibre-aperture biases.     

Optical and submillimetre observations of Bok globules – tracing the magnetic field from low to high density

Although the stellar initial mass function (IMF) has only been directly determined in star clusters, it has been manifoldly applied on galaxy-wide scales. But taking the clustered nature of star formation into account the galaxy-wide IMF is constructed by adding all IMFs of all young star clusters leading to an integrated galactic initial mass function (IGIMF). The IGIMF is top-light compared to the canonical IMF in star clusters and steepens with decreasing total star formation rate (SFR). This discrepancy is marginal for large disc galaxies but becomes significant for Small Magellanic Cloud type galaxies and less massive ones. We here construct IGIMF-based relations between the total far- and near-ultraviolet luminosities of galaxies and the underlying SFR. We make the prediction that the Hα luminosity of star-forming dwarf galaxies decreases faster with decreasing SFR than the ultraviolet (UV) luminosity. This turn-down of the Hα/UV-flux ratio should be evident below total SFRs of  10⁻² M_⊙ yr⁻¹  .     

Galaxy groups at 0.3 ≤z≤ 0.55 – I. Group properties

The evolution of galaxies in groups may have important implications for the evolution of the star formation history of the Universe, since many processes which operate in groups may suppress star formation and the fraction of galaxies in bound groups grows rapidly between   z = 1  and the present day. In this paper, we present an investigation of the properties of galaxies in galaxy groups at intermediate redshift  ( z ∼ 0.4)  . The groups were selected from the Canadian Network for Observational Cosmology Redshift Survey (CNOC2) redshift survey as described by Carlberg et al., with further spectroscopic follow-up undertaken at the Magellan telescope in order to improve the completeness and depth of the sample. We present the data for the individual groups, and find no clear trend in the fraction of passive galaxies with group velocity dispersion and group concentration. We stack the galaxy groups in order to compare the properties of group galaxies with those of field galaxies at the same redshift. The groups contain a larger fraction of passive galaxies than the field, this trend being particularly clear for galaxies brighter than   M _{B J} < −20  in the higher velocity dispersion groups. In addition, we see evidence for an excess of bright passive galaxies in the groups relative to the field. In contrast, the luminosity functions of the star-forming galaxies in the groups and the field are consistent. These trends are qualitatively consistent with the differences between group and field galaxies seen in the local Universe.     

Multiwavelength study of X-ray selected star-forming galaxies within the Chandra Deep Field-South

We have combined multiwavelength observations of a selected sample of star-forming galaxies with galaxy evolution models in order to compare the results obtained for different star formation rate (SFR) tracers and to study the effect that the evolution of the star-forming regions has on them. We also aimed at obtaining a better understanding of the corrections due to extinction and nuclear activity on the derivation of the SFR. We selected the sample from Chandra data for the well studied region Chandra Deep Field -South (CDFS) and chose the objects that also have ultraviolet (UV) and infrared (IR) data from Galaxy Evolution Explorer ( GALEX ) and Great Observatories Origins Deep Survey (GOODS) Spitzer , respectively.
Our main finding is that there is good agreement between the extinction corrected SFR(UV) and the SFR(X), and we confirm the use of X-ray luminosities as a trustful tracer of recent star formation activity. Nevertheless, at SFR(UV) larger than about  5 M_⊙ yr⁻¹  there are several galaxies with an excess of SFR(X) suggesting the presence of an obscured active galactic nucleus (AGN) not detected in the optical spectra. We conclude that the IR luminosity is driven by recent star formation even in those galaxies where the SFR(X) is an order of magnitude higher than the SFR(UV) and therefore may harbour an AGN. One object shows SFR(X) much lower than expected based on the SFR(UV); this SFR(X) 'deficit' may be due to an early transient phase before most of the massive X-ray binaries were formed. An X-ray deficit could be used to select extremely young bursts in an early phase just after the explosion of the first supernovae associated with massive stars and before the onset of massive X-ray binaries.     

A complete sample of radio sources in the North Ecliptic Cap, selected at 38 MHz – III. Further imaging observations and the photometric properties of the sample

Further imaging observations of a sample of radio sources in the North Ecliptic Cap are presented and a number of new identifications are made. Using redshifts from spectroscopic data presented in a companion paper by Lacy et al., the photometric properties of the galaxies in the sample are discussed. It is shown that: (1) out to at least z ≈0.6 radio galaxies are good standard candles irrespective of radio luminosity; (2) for 0.6≲ z ≲1 a large fraction of the sample has magnitudes and colours consistent with a non-evolving giant elliptical, and (3) at higher redshifts, where the R -band samples the rest-frame UV flux, most objects have less UV luminosity than expected if they form their stellar populations at a constant rate from a high redshift to z ∼1 in unobscured star-forming regions (assuming an Einstein–de Sitter cosmology). The consequences of these observations are briefly discussed.     

Recent arrival of faint cluster galaxies on the red sequence: luminosity functions from 119 deg2 of CFHTLS

The global star formation rate has decreased significantly since   z ∼ 1  , for reasons that are not well understood. Red-sequence galaxies, dominating in galaxy clusters, represent the population that have had their star formation shut off, and may therefore be the key to this problem. In this work, we select 127 rich galaxy clusters at  0.17 ≤ z ≤ 0.36  , from 119 deg² of the Canada–France–Hawaii Telescope Legacy Survey (CFHTLS) optical imaging data, and construct the r '-band red-sequence luminosity functions (LFs). We show that the faint end of the LF is very sensitive to how red-sequence galaxies are selected, and an optimal way to minimize the contamination from the blue cloud is to mirror galaxies on the redder side of the colour–magnitude relation. The LFs of our sample have a significant inflexion centred at     , suggesting a mixture of two populations. Combining our survey with low-redshift samples constructed from the Sloan Digital Sky Survey, we show that there is no strong evolution of the faint end of the LF (or the red-sequence dwarf-to-giant ratio) over the redshift range  0.2 ≲ z ≲ 0.4  , but from   z ∼ 0.2  to ∼0 the relative number of red-sequence dwarf galaxies has increased by a factor of ∼3, implying a significant build-up of the faint end of the cluster red sequence over the last 2.5 Gyr.     

The history of mass assembly of faint red galaxies in 28 galaxy clusters since z= 1.3

We measure the relative evolution of the number of bright and faint (as faint as  0.05 L *)  red galaxies in a sample of 28 clusters, out of which 16 are at  0.50 ≤ z ≤ 1.27  , all observed through a pair of filters bracketing the 4000-Å break rest frame. The abundance of red galaxies, relative to bright ones, is constant over all the studied redshift range,  0 < z < 1.3  , and rules out a differential evolution between bright and faint red galaxies as large as claimed in some past works. Faint red galaxies are largely assembled and in place at   z = 1.3  and their abundance does not depend on cluster mass, parametrized by velocity dispersion or X-ray luminosity. Our analysis, with respect to the previous one, samples a wider redshift range, minimizes systematics and put a more attention to statistical issues, keeping at the same time a large number of clusters.     

The cluster galaxy luminosity function at z= 0.3: a recent origin for the faint-end upturn?

We derive deep luminosity functions (LFs) (to   M _z =−15  ) for galaxies in Abell 1835  ( z = 0.25)  and AC 114  ( z = 0.31)  , and compare these with the local z ' LF for 69 clusters. The data show that the faint-end upturn, the excess of galaxies above a single Schechter function at   M _z < −17  , does not exist in the higher redshift clusters. This suggests that the faint-end upturn galaxies have been created recently, by infall into clusters of star-forming field populations or via tidal disruption of brighter objects.     

Further constraints on the evolution of Ks-selected galaxies in the GOODS/CDFS field

We have selected and analysed the properties of a sample of  2905 K_s < 21.5  galaxies in  ∼131 arcmin²  of the Great Observatories Origins Deep Survey (GOODS) Chandra Deep Field South (CDFS), to obtain further constraints on the evolution of K_s -selected galaxies with respect to the results already obtained in previous studies. We made use of the public deep multiwavelength imaging from the optical B through the infrared (IR) 4.5-μm bands, in conjunction with available spectroscopic and COMBO17 data in the CDFS, to construct an optimized redshift catalogue for our galaxy sample. We computed the K_s -band luminosity function and determined that its characteristic magnitude has a substantial brightening and a decreasing total density from   z = 0  to  〈 z 〉= 2.5  . We also analysed the colours and number density evolution of galaxies with different stellar masses. Within our sample, and in contrast to what is observed for less massive systems, the vast majority (∼85–90 per cent) of the most massive  ( M > 2.5 × 10¹¹ M_⊙)  local galaxies appear to be in place before redshift   z ∼ 1  . Around 65–70 per cent of the total assemble between redshifts   z = 1  and 3 and most of them display extremely red colours, suggesting that plausible star formation in these very massive systems should mainly proceed in obscured, short-time-scale bursts. The remaining fraction (up to ∼20 per cent) could be in place at even higher redshifts   z = 3–4  , pushing the first epoch of formation of massive galaxies beyond the limits of current near-IR surveys.     

The ultraviolet excess in nearby powerful radio galaxies: evidence for a young stellar component

We present William Herschel Telescope spectropolarimetry observations of a complete RA-limited sample of nine low-redshift  (0.05< z <0.2)  3CR radio sources in order to investigate the nature of the ultraviolet (UV) excess in nearby powerful radio galaxies. Of the nine galaxies studied in detail from this sample, we find that four show a measurable UV excess following nebular continuum subtraction, but none of the sources shows significant polarization in the UV. One of the radio galaxies with a UV excess – 3C 184.1 – shows evidence for broad permitted lines and hence direct active galactic nucleus (AGN) light. In the remaining three galaxies we argue that the most likely contributor to the UV excess is a young stellar component. For these three galaxies we find that the best-fitting model for the optical/UV continuum consists of a combination of an old stellar population  (10–15 Gyr  old elliptical galaxy) plus a reddened young stellar population  (0.05–2 Gyr)  . The reddened young stellar component typically accounts for half of the total flux at 4780 Å, following nebular continuum subtraction, and   E ( B - V )  values of between 0.2 and 0.7 mag are required. However, for the majority of sources in our sample (six out of nine), continuum modelling provides no evidence for a significant young stellar component in the nuclear regions of the host galaxies. Our results are discussed in the context of far-infrared evidence for star formation activity.