On the H i content, dust-to-gas ratio and nature of Mg ii absorbers

We estimate the mean dust-to-gas ratio of Mg  ii absorbers as a function of rest equivalent width W ₀ and redshift over the range  0.5 < z < 1.4  . Using the expanded Sloan Digital Sky Survey/ Hubble Space Telescope sample of low-redshift Lyman-α absorbers, we first show the existence of a 8σ correlation between the mean hydrogen column density  〈 N _{H  i} 〉  and W ₀, an indicator of gas velocity dispersion. By combining these results with recent dust-reddening measurements, we show that the mean dust-to-gas ratio of Mg  ii absorbers does not appreciably depend on rest equivalent width. Assuming that, on average, dust-to-gas ratio is proportional to metallicity, we find its redshift evolution to be consistent with that of   L ^★  galaxies from   z = 0.5  to 1.4, and we show that our constraints disfavour dwarf galaxies as the origin of such absorbers. We discuss other scenarios and favour galactic outflows from  ∼ L ^★  galaxies as the origin of the majority of strong Mg  ii absorbers. Finally, we show that, once evolutionary effects are taken into account, the Bohlin et al. relation between A _V and N _H is also satisfied by strong Mg  ii systems down to lower column densities than those probed in our Galaxy.     

The radial Tully–Fisher relation for spiral galaxies – I

We find a new Tully–Fisher-like relation for spiral galaxies holding at different galactocentric radii. This radial Tully–Fisher relation allows us to investigate the distribution of matter in the optical regions of spiral galaxies. This relation, applied to three different samples of rotation curves of spiral galaxies, directly proves that: (i) the rotation velocity of spirals is a good measure of their gravitational potential and both the rotation curve's amplitudes and profiles are well predicted by galaxy luminosity, (ii) the existence of a dark component, less concentrated than the luminous one, and (iii) a scaling law, according to which, inside the disc optical size:   M _dark/ M _lum= 0.5( L _B /10¹¹ L _{B ⊙})^−0.7  .     

An 80-kpc Lyα halo around a high-redshift type-2 quasi-stellar object

We announce the discovery of an extended emission-line region associated with a high-redshift type-2 quasi-stellar object (QSO). The halo, which was discovered in our new wide-field narrow-band survey, resides at   z = 2.85  in the Spitzer First Look Survey region and is extended over ∼80 kpc. Deep very long baseline interferometry (VLBI) observations imply that approximately 50 per cent of the radio emission is extended on scales >200 pc. The inferred active galactic nuclei (AGN) luminosity is sufficient to ionize the extended halo, and the optical emission is consistent with being triggered coevally with the radio source. The Lyα halo is as luminous as those found around high-redshift radio galaxies; however, the active nucleus is several orders of magnitude less luminous at radio wavelengths than those Fanarof–Riley type II (FRIIs) more commonly associated with extended emission-line regions. AMS05 appears to be a high-redshift analogue to the radio-quiet quasar E1821+643 which is core dominated, but which also exhibits extended Fanarof–Riley type I (FRI)-like structure and contains an optically powerful AGN. We also find evidence for more quiescent kinematics in the Lyα emission line in the outer regions of the halo, reminiscent of the haloes around the more powerful FRIIs. The optical to mid-infrared spectral energy distribution is well described by a combination of an obscured QSO  ( L _bol∼ 3.4 ± 0.2 × 10¹³ L_⊙)  and a 1.4 Gyr old simple stellar population with mass  ∼3.9 ± 0.3 × 10¹¹ M_⊙  .     

The concentration–velocity dispersion relation in galaxy groups

Based on results from cold dark matter N -body simulations, we develop a dynamical model for the evolution of subhaloes within group-sized host haloes. Only subhaloes more massive than 5 × 10⁸ M_⊙ are considered, because they are massive enough to possibly host luminous galaxies. On their orbits within a growing host potential the subhaloes are subject to tidal stripping and dynamical friction. At the present time  ( z = 0)  , all model hosts have equal mass  ( M _vir= 3.9 × 10¹³ M_⊙)  but different concentrations associated with different formation times. We investigate the variation of subhalo (or satellite galaxy) velocity dispersion with host concentration and/or formation time. In agreement with the Jeans equation, the velocity dispersion of subhaloes increases with the host concentration. Between concentrations of ∼5 and ∼20, the subhalo velocity dispersions increase by a factor of ∼1.25. By applying a simplified tidal disruption criterion, that is, rejection of all subhaloes with a tidal truncation radius below 3  kpc at   z = 0  , the central velocity dispersion of the 'surviving' subhalo sample increases substantially for all concentrations. The enhanced central velocity dispersions in the surviving subhalo samples are caused by a lack of slow tangential motions. Additionally, we present a fitting formula for the anisotropy parameter which does not depend on concentration if the group-centric distances are scaled by r _s, the characteristic radius of the Navarro, Frenk & White profile. Since the expected loss of subhaloes and galaxies due to tidal disruption increases the velocity dispersion of surviving galaxies, the observed galaxy velocity dispersion can substantially overestimate the virial mass.     

Collapse and fragmentation of rotating magnetized clouds – I. Magnetic flux–spin relation

We discuss the evolution of the magnetic flux density and angular velocity in a molecular cloud core, on the basis of three-dimensional numerical simulations, in which a rotating magnetized cloud fragments and collapses to form a very dense optically thick core of  >5 × 10¹⁰ cm⁻³  . As the density increases towards the formation of the optically thick core, the magnetic flux density and angular velocity converge towards a single relationship between the two quantities. If the core is magnetically dominated its magnetic flux density approaches  1.5( n /5 × 10¹⁰ cm⁻³)^1/2 mG  , while if the core is rotationally dominated the angular velocity approaches  2.57 × 10⁻³ ( n /5 × 10¹⁰ cm⁻³)^1/2 yr⁻¹  , where n is the density of the gas. We also find that the ratio of the angular velocity to the magnetic flux density remains nearly constant until the density exceeds  5 × 10¹⁰ cm⁻³  . Fragmentation of the very dense core and emergence of outflows from fragments will be shown in the subsequent paper.     

H i and optical spectroscopy towards the M15 intermediate-velocity cloud

We present single-dish Arecibo 21-cm H  i observations, covering a 0675×0625 RA–Dec. grid, of the intermediate-velocity cloud (IVC) centred upon the M15 globular cluster. The velocity and positional structure of the IVC gas at V _LSR=70 km s⁻¹ are investigated; it is found to be clumpy and has a peak surface density N _{H  i} ∼8×10¹⁹ cm⁻². Additionally, we have performed a long H  i integration towards HD 203664, a Galactic halo star some 31 from M15, in which optical IVC absorption has previously been detected. No H  i with a velocity exceeding 60 km s⁻¹ was found to a brightness temperature limit of 0.05 K. However, additional pointings did detect IVC gas approximately mid-way between HD 203664 and M15. Finally, we present both Arecibo H  i pointings and low-resolution spectra in the Ca  ii H and K lines towards 15 field stars in the general field towards M15, in an attempt to obtain the distance to the IVC. Intermediate-velocity H  i is detected towards seven sightlines. Stellar spectral types are derived for 12 of the sample. Assuming that these stars lie on the main sequence, their distances are estimated to lie in the range 150≤ d ≤1350 pc. No Ca  ii absorption is observed, either because the IVC is further away than ∼1350 pc or more likely because the gas along these sightlines is of too low a density to be detected by the current observations.     

The SCUBA HAlf Degree Extragalactic Survey – VI. 350-μm mapping of submillimetre galaxies

A follow-up survey using the Submillimetre High-Angular Resolution Camera (SHARC-II) at 350 μm has been carried out to map the regions around several 850-μm-selected sources from the Submillimetre HAlf Degree Extragalactic Survey (SHADES). These observations probe the infrared (IR) luminosities and hence star formation rates in the largest existing, most robust sample of submillimetre galaxies (SMGs). We measure 350-μm flux densities for 24 850-μm sources, seven of which are detected at ≥2.5σ within a 10 arcsec search radius of the 850-μm positions. When results from the literature are included the total number of 350-μm flux density constraints of SHADES SMGs is 31, with 15 detections. We fit a modified blackbody to the far-IR (FIR) photometry of each SMG, and confirm that typical SMGs are dust-rich  ( M _dust≃ 9 × 10⁸ M_⊙)  , luminous  ( L _FIR≃ 2 × 10¹² L_⊙)  star-forming galaxies with intrinsic dust temperatures of ≃35 K and star formation rates of  ≃400 M_⊙ yr⁻¹  . We have measured the temperature distribution of SMGs and find that the underlying distribution is slightly broader than implied by the error bars, and that most SMGs are at 28 K with a few hotter. We also place new constraints on the 350-μm source counts, N ₃₅₀(>25 mJy) ∼ 200–500 deg⁻².     

Stellar populations in the nuclear regions of nearby radio galaxies

We present optical spectra of the nuclei of seven luminous ( P _178 MHz≳10²⁵ W Hz⁻¹ Sr⁻¹) nearby ( z <0.08) radio galaxies, which mostly correspond to the FR II class. In two cases, Hydra A and 3C 285, the Balmer and λ 4000-Å break indices constrain the spectral types and luminosity classes of the stars involved, revealing that the blue spectra are dominated by blue supergiant and/or giant stars. The ages derived for the last burst of star formation in Hydra A are between 7 and 40 Myr, and in 3C 285 about 10 Myr. The rest of the narrow-line radio galaxies (four) have a λ 4000-Å break and metallic indices consistent with those of elliptical galaxies. The only broad-line radio galaxy in our sample, 3C 382, has a strong featureless blue continuum and broad emission lines that dilute the underlying blue stellar spectra. We are able to detect the Ca  ii triplet in absorption in the seven objects, with good quality data for only four of them. The strengths of the absorptions are similar to those found in normal elliptical galaxies, but these values are consistent both with stellar populations of roughly similar ages (as derived from the Balmer absorption and break strengths) and with mixed young+old populations.     

Near-infrared [Fe ii] emission from supernova remnants and the supernova rate of starburst galaxies

In an effort better to calibrate the supernova rate of starburst galaxies as determined from near-infrared [Fe  ii ] features, we report on a [Fe  ii ] λ 1.644 μm line-imaging survey of a sample of 42 optically selected supernova remnants (SNRs) in M33. A wide range of [Fe  ii ] luminosities are observed within our sample (from less than 6 to 695 L_⊙). Our data suggest that the bright [Fe  ii ] SNRs are entering the radiative phase and that the density of the local interstellar medium (ISM) largely controls the amount of [Fe  ii ] emission. We derive the following relation between the [Fe  ii ] λ 1.644 μm line luminosity of radiative SNRs and the electronic density of the post-shock gas, n _e: L _{[Fe  ii ]}     (cm⁻³). We also find a correlation in our data between L _{[Fe  ii ]} and the metallicity of the shock-heated gas, but the physical interpretation of this result remains inconclusive, as our data also show a correlation between the metallicity and n _e. The dramatically higher level of [Fe  ii ] emission from SNRs in the central regions of starburst galaxies is most likely due to their dense environments, although metallicity effects might also be important. The typical [Fe  ii ]-emitting lifetime of a SNR in the central regions of starburst galaxies is found to be of the order of 10⁴ yr. On the basis of these results, we provide a new empirical relation allowing the determination of the current supernova rate of starburst galaxies from their integrated near-infrared [Fe  ii ] luminosity.     

Radio galaxies in the 2SLAQ Luminous Red Galaxy survey – II. The stellar populations of radio-loud and radio-quiet LRGs

We present an analysis of the optical spectra of a volume-limited sample of 375 radio galaxies at redshift  0.4 < z < 0.7  from the 2dF-SDSS (Sloan Digital Sky Survey) Luminous Red Galaxy (LRG) and QSO (quasi-stellar object) (2SLAQ) redshift survey. We investigate the evolution of the stellar populations and emission-line properties of these galaxies. By constructing composite spectra and comparing with a matched sample of radio-quiet sources from the same survey, we also investigate the effect on the galaxy of the presence of an active nucleus.
The composite spectra, binned by redshift and radio luminosity, all require two components to describe them, which we interpret as an old and a younger population. We found no evolution with redshift of the age of the younger population in radio galaxies, nor were they different from the radio-quiet comparison sample. Similarly, there is no correlation with radio power, with the exception that the most powerful radio sources  ( P _1.4 > 10²⁶  W Hz⁻¹) have younger stars and stronger emission lines than the less powerful sources. This suggests that we have located the threshold in radio power where strong emission lines 'switch on', at radio powers of around 10²⁶ W Hz⁻¹. Except for the very powerful radio galaxies, the presence of a currently active radio active galactic nucleus (AGN) does not appear to be correlated with any change in the observed stellar population of a luminous red galaxy at   z ∼ 0.5  .     

Lyα absorption in the nearby Universe: the sightline to Q1821+643

We present the first results of a survey designed to understand the origin of Lyα-forest absorption systems at low redshift. Using the WYFFOS and HYDRA multifibre spectrographs on the William Herschel and Wisconsin Indiana Yale NOAO (WIYN) telescopes, we have identified 51 galaxies brighter than b _j := :18.5 within 30 arcmin of the sightline of the QSO 1821+643. We find three galaxies within 500 h ⁻¹ kpc of the QSO sightline; the nearest galaxy is 104 h ⁻¹ kpc away from the line of sight, and is at the same redshift as a strong ( W _r :=:0.63Å) Lyα absorption line. The remaining two galaxies have no corresponding absorption to extremely low equivalent width limits (< 0.05 Å). Beyond 500 h ⁻¹ kpc, Lyα absorption lines are found at redshifts similar to those of several galaxies, but we show that these coincidences are likely to be accidental.   Half of the Lyα systems for which we could have found at least an L ^* galaxy have no galaxies at the redshifts of the absorbers. For the majority of the remainder, we show that any apparent association with galaxies is probably coincidental. These Lyα systems are characterized by their weak equivalent widths ( W _r :<:0.2Å), and we conclude that this population of absorbers is uncorrelated, or at best weakly correlated, with galaxies.     

Unresolved emission and ionized gas in the bulge of M31

We study the origin of unresolved X-ray emission from the bulge of M31 based on archival Chandra and XMM–Newton observations. We demonstrate that three different components are present. (i) Broad-band emission from a large number of faint sources – mainly accreting white dwarfs and active binaries, associated with the old stellar population, similar to the Galactic ridge X-ray emission of the Milky Way. The X-ray to K -band luminosity ratios are compatible with those for the Milky Way and for M32; in the 2–10 keV band, the ratio is  (3.6 ± 0.2) × 10²⁷ erg s⁻¹ L⁻¹_⊙  . (ii) Soft emission from ionized gas with a temperature of about ∼300 eV and a mass of  ∼2 × 10⁶ M_⊙  . The gas distribution is significantly extended along the minor axis of the galaxy, suggesting that it may be outflowing in the direction perpendicular to the galactic disc. The mass and energy supply from evolved stars and Type Ia supernovae is sufficient to sustain the outflow. We also detect a shadow cast on the gas emission by spiral arms and the 10-kpc star-forming ring, confirming significant extent of the gas in the 'vertical' direction. (iii) Hard extended emission from spiral arms, most likely associated with young stellar objects and young stars located in the star-forming regions. The   L _X/SFR  (star formation rate) ratio equals  ∼9 × 10³⁸ (erg s⁻¹)(M_⊙ yr⁻¹)⁻¹  , which is about ∼1/3 of the high-mass X-ray binary contribution, determined earlier from Chandra observations of other nearby galaxies.     

Multiwavelength study of the nuclei of a volume-limited sample of galaxies – I. X-ray observations

We discuss ROSAT HRI X-ray observations of 33 very nearby galaxies, sensitive to X-ray sources down to a luminosity of approximately 10³⁸ erg s⁻¹. The galaxies are selected from a complete, volume-limited sample of 46 galaxies with     for which we have extensive multiwavelength data. For an almost complete subsample with     (29/31 objects) we have HRI images. Contour maps and source lists are presented within the central region of each galaxy, together with nuclear upper limits where no nuclear source was detected. Nuclear X-ray sources are found to be very common, occurring in ∼35 per cent of the sample. Nuclear X-ray luminosity is statistically connected to host galaxy luminosity – there is not a tight correlation, but the probability of a nuclear source being detected increases strongly with galaxy luminosity, and the distribution of nuclear luminosities seems to show an upper envelope that is roughly proportional to galaxy luminosity. While these sources do seem to be a genuinely nuclear phenomenon rather than nuclear examples of the general X-ray source population, it is far from obvious that they are miniature Seyfert nuclei. The more luminous nuclei are very often spatially extended, and H  ii region nuclei are detected just as often as LINERs. Finally, we also note the presence of fairly common superluminous X-ray sources in the off-nuclear population – out of 29 galaxies we find nine sources with a luminosity greater than 10³⁹ erg s⁻¹. These show no particular preference for more luminous galaxies. One is already known to be a multiple SNR system, but most have no obvious optical counterpart and their nature remains a mystery.     

The relation between Lyman α absorbers and gas-rich galaxies in the local Universe

We use high-resolution hydrodynamical simulations to investigate the spatial correlation between weak  ( N _{H  i} < 10¹⁵ cm⁻²)  Lyα absorbers and gas-rich galaxies in the local Universe. We confirm that Lyα absorbers are preferentially expected near gas-rich galaxies and that the degree of correlation increases with the column density of the absorber. The real-space galaxy auto-correlation is stronger than the cross-correlation (correlation lengths   r _0,gg= 3.1 ± 0.1 Mpc  h ⁻¹  and   r _0,ag= 1.4 ± 0.1 Mpc  h ⁻¹  , respectively), in contrast with the recent results of Ryan-Weber, and the auto-correlation of absorbers is very weak. These results are robust to the presence of strong galactic winds in the hydrodynamical simulations. In redshift space, a further mismatch arises since at small separations the distortion pattern of the simulated galaxy–absorber cross-correlation function is different from the one measured by Ryan-Weber. However, when sampling the intergalactic medium along a limited number of lines-of-sight, as in the real data, uncertainties in the cross-correlation estimates are large enough to account for these discrepancies. Our analysis suggests that the statistical significance of difference between the cross-correlation and auto-correlation signal in current data sets is ∼1σ only.     

Chandra unveils a binary active galactic nucleus in Mrk 463

We analyse Chandra , XMM–Newton and Hubble Space Telescope ( HST ) data of the double-nucleus Ultraluminous Infrared Galaxy (ULIRG), Mrk 463. The Chandra detection of two luminous  ( L _2–10 keV= 1.5 × 10⁴³  and  3.8 × 10⁴² erg cm⁻² s⁻¹)  , unresolved nuclei in Mrk 463 indicates that this galaxy hosts a binary active galactic nucleus (AGN), with a projected separation of ≃3.8 kpc (  3.83 ± 0.01  arcsec). While the East nucleus was already known to be a type 2 Seyfert (and this is further confirmed by our Chandra detection of a neutral iron line), this is the first unambiguous evidence in favour of the AGN nature of the West nucleus. Mrk 463 is therefore the clearest case so far for a binary AGN, after NGC 6240.     

Predictions of the extent of self-enrichment in oxygen of giant metal-poor H ii regions

In general, H  ii regions do not show clear signs of self-enrichment in products from massive stars  ( M ≥ 8 M_⊙)  . In order to explore why I modelled the contamination with Wolf–Rayet star ejecta of metal-poor  ( Z = 0.001)  H  ii regions, ionized either by a  10⁶ M_⊙  cluster of coeval stars (cluster 1) or by a cluster resulting from continuous star formation at a rate of  1 M_⊙ yr⁻¹  (cluster 2). The clusters have   Z = 0.001  and a Salpeter initial mass function from 0.1 to  120 M_⊙  . Independent one-dimensional constant density simulations of the emission-line spectra of unenriched H  ii regions were computed at the discrete ages 1, 2, 3, 4 and 5 Myr, with the photoionization code cloudy , using as input, radiative and mechanical stellar feedbacks predicted by the evolutionary synthesis code starburst99 . Each H  ii region was placed at the outer radius of the adiabatically expanding superbubble of Mac Low & McCray. For models with thermal and ionization balance time-scales of less than 1 Myr, and with oxygen emission-line ratios in agreement with observations, the volume of the superbubble and the H  ii region was uniformly and instantaneously polluted with stellar ejecta predicted by starburst99 . I obtained a maximum oxygen abundance enhancement of 0.025 dex, with cluster 1, at 4 Myr. It would be unobservable.     

The local star formation rate and radio luminosity density

We present a new determination of the local volume-averaged star formation rate from the 1.4-GHz luminosity function of star forming galaxies. Our sample, taken from the   B ≤12  Revised Shapley–Ames catalogue (231 normal spiral galaxies over an effective area of 7.1 sr) has ≃100 per cent complete radio detections and is insensitive to dust obscuration and cirrus contamination. After removal of known active galaxies, the best-fitting Schechter function has a faint-end slope of  −1.27±0.07  in agreement with the local H α luminosity function, characteristic luminosity   L ∗=(2.6±0.7)×10²² W Hz⁻¹  and density   φ ∗=(4.8±1.1)×10⁻⁴ Mpc⁻³.  The inferred local radio luminosity density of  (1.73±0.37±0.03)×10¹⁹ W Hz⁻¹ Mpc⁻³  (Poisson noise, large-scale structure fluctuations) implies a volume-averaged star formation rate ∼2 times larger than the Gallego et al. H α estimate, i.e.   ρ _1.4 GHz=(2.10±0.45±0.04)×10⁻² M_⊙ yr⁻¹ Mpc⁻³  for a Salpeter initial mass function from  0.1–125 M_⊙  and Hubble constant of 50 km s⁻¹ Mpc⁻¹. We demonstrate that the Balmer decrement is a highly unreliable extinction estimator, and argue that optical–ultraviolet (UV) star formation rates (SFRs) are easily underestimated, particularly at high redshift.     

Optical spectroscopy and photometry of SAX J1808.4−3658 in outburst

We present phase resolved optical spectroscopy and photometry of V4580 Sagittarii, the optical counterpart to the accretion powered millisecond pulsar SAX J1808.4−3658, obtained during the 2008 September/October outburst. Doppler tomography of the N  iii λ4640.64 Bowen blend emission line reveals a focused spot of emission at a location consistent with the secondary star. The velocity of this emission occurs at  324 ± 15 km s⁻¹  ; applying a ' K -correction', we find the velocity of the secondary star projected on to the line of sight to be  370 ± 40 km s⁻¹  . Based on existing pulse timing measurements, this constrains the mass ratio of the system to be  0.044^+0.005_−0.004  , and the mass function for the pulsar to be  0.44^+0.16_−0.13 M_⊙  . Combining this mass function with various inclination estimates from other authors, we find no evidence to suggest that the neutron star in SAX J1808.4−3658 is more massive than the canonical value of  1.4 M_⊙  . Our optical light curves exhibit a possible superhump modulation, expected for a system with such a low mass ratio. The equivalent width of the Ca  ii H and K interstellar absorption lines suggest that the distance to the source is ∼2.5 kpc. This is consistent with previous distance estimates based on type-I X-ray bursts which assume cosmic abundances of hydrogen, but lower than more recent estimates which assume helium-rich bursts.     

XMM–Newton observations of the Seyfert 1 AGN H 0557−385

We present XMM–Newton observations of the Seyfert 1 active galactic nucleus (AGN)  H 0557 − 385  . We have conducted a study into the warm absorber present in this source, and using high-resolution Reflection Grating Spectrometer (RGS) data we find that the absorption can be characterized by two phases: a phase with log ionization parameter ξ of 0.50 (where ξ is in units of erg cm s⁻¹) and a column of  0.2 × 10²¹ cm⁻²  , and a phase with log ξ of 1.62 and a column of  1.3 × 10²² cm⁻²  . An iron Kα line is detected. Neutral absorption is also present in the source, and we discuss possible origins for this. On the assumption that the ionized absorbers originate as an outflow from the inner edge of the torus, we use a new method for finding the volume filling factor. Both phases of  H 0557 − 385  have small volume filling factors (≤1 per cent). We also derive the volume filling factors for a sample of 23 AGN using this assumption and for the absorbers with  log ξ > 0.7  , we find reasonable agreement with the filling factors obtained through the alternative method of equating the momentum flow of the absorbers to the momentum loss of the radiation field. By comparing the filling factors obtained by the two methods, we infer that some absorbers with  log ξ < 0.7  occur at significantly larger distances from the nucleus than the inner edge of the torus.     

The CO content of the Local Group dwarf irregular galaxies IC 5152, UGCA 438 and the Phoenix dwarf

We present a search for  CO(1 → 0)  emission in three Local Group dwarf irregular galaxies: IC 5152, the Phoenix dwarf and UGCA 438, using the ATNF Mopra radio telescope. Our scans largely cover the optical extent of the galaxies and the stripped H  i cloud west of the Phoenix dwarf. Apart from a tentative but non-significant emission peak at one position in the Phoenix dwarf, no significant emission was detected in the CO spectra of these galaxies. For a velocity width of 6 km s⁻¹, we derive 4σ upper limits of 0.03, 0.04 and 0.06 K km s⁻¹ for IC 5152, the Phoenix dwarf and UGCA 438, respectively. This is an improvement of over a factor of 10 compared with previous observations of IC 5152; the other two galaxies had not yet been observed at millimetre wavelengths. Assuming a Galactic CO-to-H₂ conversion factor, we derive upper limits on the molecular gas mass of  6.2 × 10⁴, 3.7 × 10³  and  1.4 × 10⁵ M_⊙  for IC 5152, the Phoenix dwarf and UGCA 438, respectively. We investigate two possible causes for the lack of CO emission in these galaxies. On the one hand, there may be a genuine lack of molecular gas in these systems, in spite of the presence of large amounts of neutral gas. However, in the case of IC 5152 which is actively forming stars, molecular gas is at least expected to be present in the star-forming regions. On the other hand, there may be a large increase in the CO-to-H₂ conversion factor in very low-metallicity dwarfs  (−2 ≤[Fe/H]≤−1)  , making CO a poor tracer of the molecular gas content in dwarf galaxies.