In order to study the size and shape of the absorbers that lie in front of the QSOs, in particular the Ly α forest, we present an analysis of 785 absorption lines in the spectra of five QSOs in close groupings: a pair (LB9605: 1517+2357 at z = 1.834 and LB9612: 1517+2356 at z = 1.903, with a separation of 102 arcsec between them) and a triplet (KP 76: 1623+2651A at z = 2.467, KP 77:1623+2653 at z = 2.526, and KP 78: 1623+2651B at z = 2.605, with separations of 127, 147 and 177 arcsec between pairs 76:78, 76:77 and 77:78, respectively).Both of these QSO groups have been observed before, but these data represent a drastic increase in signal-to-noise ratio and/or wavelength coverage over earlier data, and provide a qualitatively different view of the nature of the absorbers.The pair samples a scale critical in determining the size upper bound of Ly α absorbers, with significant leverage in redshift compared to previous studies.In the case of the triplet, this represents the spatially densest sample of Ly α forest absorbers ever studied, and an almost ideally-suited probe of the shape of absorbers. We observe a significant number of Ly α lines in common between the triplet sightlines, for lines stronger than rest equivalent width W o > 0.4Å (and no detected metal lines) and velocity differences up to 200 km s −1 , corresponding to a two-point correlation function ξ = 1.88 +0.78 −0.50 on scales 0.5 to 0.8 h −1 Mpc with z = 2.14, and inconsistent at the 99.999% level with the absence of any clustering. These data also show that a significant fraction of the W o > 0.4Å Ly α forest absorbers span all three sightlines to the KP triplet, indicating that the strong-lined absorbers are consistent with nearly round shapes, chosen from a range of possible cylinders of different elongations. This may be inconsistent with results from hydrodynamic/gravitational simulations of H I in the early Universe indicating that the theoretical counterparts of Ly α forest clouds are
AnalysisContinua were calculated and lines detected, de-blended and assigned identifications according to Crotts (1989). Deblending was performed using multiple gaussian fits instead of Voigt profiles in nearly all cases, with the exception of resolved lines. We use a lower than usual S/N cutoff of 3.5σ for line detections. We are confident of this approach because we are able to check our results for KP 77 against a high S/N Keck HIRES spectrum of the object, the highest S/N part of a larger sample collected for the triplet (Crotts, Burles & Tytler 1997). At a 3.5σ cutoff for the KPNO data, no false detections are found over the 3900-5700Å overlap between the two data sets. In fact the KPNO 4m observations do a good job of detecting all obvious lines in the Keck spectrum (except for some very weak lines in the red KPNO 4m spectrum past 4700Å), while, of course, not resolving very close lines. We are fairly confident of our linelists, therefore, and expect approximately four lines out of our 785 to be false detections due to statistical fluctuations.
Absorption ...