.............................................................. The far-ultraviolet signature of the ‘missing’ baryons in the Local Group of galaxies Fabrizio Nicastro*, Andreas Zezas*, Martin Elvis*, Smita Mathur†, Fabrizio Fiore‡, Cesare Cecchi-Pestellini*, Douglas Burke*, Jeremy Drake* & Piergiorgio Casella* * Harvard-Smithsonian Center for Astrophysics, 60 Garden Street, Cambridge, Massachusetts 02138, USA † Astronomy Department, The Ohio State University, Columbus, Ohio 43210, USA ‡ Osservatorio Astronomico di Monteporzio, Via Osservatorio Via Frascati 33, Monteporzio-Catone (RM), I-00040 Italy ............................................................................................................................................................................. The number of baryons detected in the low-redshift (z < 1) Universe is far smaller than the number detected in correspond- ing volumes at higher redshifts. Simulations 1–3 of the formation of structure in the Universe show that up to two-thirds of the ‘missing’ baryons may have escaped detection because of their high temperature and low density. One of the few ways to detect this matter directly is to look for its signature in the form of ultraviolet absorption lines in the spectra of background sources such as quasars. Here we show that the amplitude of the average velocity vector of ‘high velocity’ O VI (O 51 ) absorption clouds detected in a survey 4 of ultraviolet emission from active galactic nuclei decreases significantly when the vector is transformed to the frames of the Galactic Standard of Rest and the Local Group of galaxies. At least 82 per cent of these absorbers are not associated with any ‘high velocity’ atomic hydrogen complex in our Galaxy, and are therefore likely to result from a primordial warm–hot intergalactic medium pervading an extended corona around the Milky Way or the Local Group. The total mass of baryons in this medium is estimated to be up to ,10 12 solar masses, which is of the order of the mass required 5 to dynamically stabilize the Local Group. The first evidence for an absorption system due to the warm–hot intergalactic medium has been discovered in the X-ray and ultra- violet spectra of the blazar PKS21552304 (ref. 6), obtained with the Chandra low-energy transmission grating (LETG) 7 , and the Far- Ultraviolet Spectroscopic Explorer (FUSE) 8 . But one line of sight is inadequate to fully define this absorbing medium. FUSE obser- vations of active galactic nuclei show two types of ubiquitous z < 0 O VI absorbers, similar to those observed in PKS21552304: (1) low-velocity O VI clouds (LV-O VI , velocity with respect to the Local Standard of Rest jv LSR j , 100 km s 21 ) 9–13 ; and (2) high-velocity O VI clouds (HV-O VI , jv LSR j . 100 km s 21 ) 4,14–16 . To investigate the nature of these absorbers we have examined (see Methods section) a sample of active galactic nuclei using publicly available FUSE data. This has allowed us to clearly identify the HV-O VI component with diffuse gas in an extended Galactic corona or in the Local Group. The O VI velocity distribution in the LSR (Fig. 1, dashed histo- gram) shows a narrow peak between ^100 km s 21 (LV-O VI ) with a much broader, roughly symmetric distribution (Fig. 1, solid histo- gram) extending to ^550 km s 21 (HV-O VI ). The bimodality of this distribution suggests that LV- and HV-O VI systems belong to two different populations of absorbers, as previously pointed out in refs 9 and 14. Here we concentrate on the HV-O VI absorbers, and use the LV-O VI absorbers only as a comparison sample. A complete discussion of both LV and HV systems is deferred to a forthcoming paper (F.N., A.Z., M.E. & P. Casella, manuscript in preparation). A plot in Galactic coordinates (latitude, b; longitude, l) of the LSR velocity distributions for the HV (Fig. 2a) absorption systems shows that the hemisphere with 08# l # 1808 contains only HV-O VI lines with negative velocities, while the other hemisphere contains mostly HV-O VI absorbers with positive velocity. There are three exceptions: two of these have LSR velocities very close to the threshold velocity of jv LSR j¼ 100 km s 21 ; and so may well belong to the LV-O VI population. The third lies at very high latitude, where the concept of longitude becomes meaningless. For comparison, the population of LV-O VI absorbers show no velocity segregation in the LSR. The red circles in Fig. 2 mark the six HV-O VI absorbers (,18% of the sample) which are tentatively identified with 21-cm high- velocity H I clouds or complexes, either because of spatial and velocity coincidence (within broad ranges) with entries in the catalogues of refs 17 and 18, or because they are identified as such in the HV-O VI compilation of ref. 4 (in which ,20% of HV-O VI are tentatively identified with HV-H I ). We stress here that these are the only six HV-O VI absorbers, in either compilation (ours and ref. 4), that may be identified with HV-H I clouds with a known distance, either in our Galaxy’s inner halo (with distances less than ,10 kpc), or between us and the two Magellanic Clouds. To be conservative we do not consider further these six HV-O VI absorbers. The systematic LSR velocity distribution of HV-O VI is consistent with matter that is (1) counter-rotating, with respect to the Galaxy disk rotation, on orbits external to the Sun’s orbit, (2) at rest in the Galactic halo, or (3) at rest in the intergalactic space surround- ing the Galaxy. The range of radial LSR velocities of the HV-O VI ð100 , jv LSR j HV , 550 km s 21 Þ greatly exceeds the range of observed radial velocities in the Galactic disk or halo, suggesting that the Galaxy-related options (1) and (2) are unlikely. Moreover, some of the HV-O VI velocities exceed a plausible measure of the escape velocity from the Milky Way 19 . Finally, clouds in the Galaxy’s halo would probably be rotating on random orbits around the Galaxy’s centre, as globular clusters do. The peculiar velocities of these clouds along these orbits would tend to randomize the apparent symmetry induced in the LSR by the circular motion of the Sun in the Galaxy for matter effectively at rest in the halo, as observed in globular clusters. An intergalactic origin, with distances larger than ,100 kpc, is more consistent with the data. The LSR is not a rest frame system for the HV-O VI absorbers (Fig. 2a, Table 1). This is confirmed by translations of the velocity distribution to other convenient rest frames: the Galactic Standard Figure 1 The velocity range of HV-O VI absorbers greatly exceeds typical rotational velocities in the Galaxy. Histogram of the HV-O VI (solid line) and LV-O VI (dashed line) velocity distributions in the Local Standard of Rest (LSR). Among the 54 objects of our sample, 45 (83%) show at least one clear O VI absorption component at z < 0 at our detection thresholds (7 out of the remaining 9 objects have poor-quality FUSE spectra, with detection thresholds of equivalent width $200 mA ˚ ). Four show multiple LV and HV absorption. 38 lines of sight show LV-O VI absorption (70% of the sample), and 32 (59% of the sample) show HV-O VI components, with 22 objects showing both. Only four lines of sight show multiple LV or HV absorption. letters to nature NATURE | VOL 421 | 13 FEBRUARY 2003 | www.nature.com/nature 719 © 2003 Nature Publishing Group