The Herschel bright sources sample

Far-infrared observations have detected dusty star-forming galaxies, a subset of galaxies which is extremely dust-extincted from the ultraviolet down to near-infrared colours. Recent studies show that this population of sources contributes significantly to the history of star formation, especially out to very high redshift. Recent surveys with the Herschel Space Observatory have uncovered around half a million of these sources, with the largest of these surveys, the H-ATLAS, covering 616 square degrees. One of the most exciting discoveries is the lensing nature of the brightest of these sources, where the gravitational potential of a foreground galaxy lenses and amplifies the signal. The applications of gravitational lensing range from studying individual sources down to unprecidented resolution at high redshift in sub-mm wavelengths with ALMA, to cosmological studies by analysing the distribution of groups of lenses. In this thesis, I explore the effect of applying a more inclusive selection criterion for lensed sources, and study the properties of the sources that are selected. Whereas the first attempts at finding lensed sources use a strict S500μm > 100 mJy flux density cut, the sample I study is selected with a flux cut at 80 mJy: The Herschel Bright Sources (HerBS) sample. A photometric redshift cut of zphot > 2 is also taken, as most lensing takes place out at higher redshift. This redshift is calculated by fitting a spectral template to the 250, 350 and 500 μm observations from the Herschel SPIRE instrument. I push down the selection flux in order to select more lensed sources from the sub-mm surveys, whilst potentially including several unlensed sources. These unlensed sources could be among the most intrinsically luminous and star-forming objects in the Universe. Only less than five of such objects are known to exist, while our HerBS sample could contain up to 35 of these sources, which could teach us about the upper-limits of star-formation and their contribution to forming the most massive galaxies in the Universe. I use 850 μm SCUBA-2 observations on the James Clerk Maxwell Telescope (JCMT) to remove blazar interlopers, which results in 209 sources in the HerBS sample, after removing 14 blazar sources. At the time I wrote the paper upon which Chapter 2 is based, 24 sources had a spectroscopic redshift. I use this sub-sample to fit a two-temperature modified blackbody, and find a cold-body temperature of 21.3 K, a warm-body temperature of 45.8 K, a mass ratio of 26.7 and a dustemissivity index of 1.8. These values do not challenge the current knowledge of sub-mm galaxies, but the quality of the fit suggests a large diversity among the galaxies in the sub-sample, and that they are poorly fitted by a single template. This diversity is also found by the spectroscopic observations with the IRAM 30m-telescope observations on eight of the highest-redshift (zphot > 4) sources of the HerBS sample. We found five spectroscopic redshifts, with one of the sources at the highest known HerBS redshift at zspec = 4.8. The spectrum fitted in Chapter 2 shows a poor agreement with the photometric data points. The spatial resolution of the SPIRE instrument on Herschel is not fine enough to resolve the structure of these high-redshift sources. Worse still, the beam width is so large, ranging from 18 to 36 arcseconds, that it is unsure whether we observe a single galaxy, or perhaps observe multiple galaxies together. The beam width of the SCUBA-2 instrument at 850 μm is only 13 arcseconds. In the case the sample would be dominated by blended sources, one would expect to resolve several of the sources into their individual components. This is not seen in any of the continuum images, although the blended sources might be blended on scales smaller than 13 arcseconds. The IRAM-observations of two sources have detected multiple, contradicting spectral lines, suggesting we might be observing multiple sources, instead of a single source, that are aligned along the line-of sight. Unfortunately, only single spectral lines have been observed per source, and we are awaiting more observations verifying the blending nature of these sources, which are still expected to lie at high redshift. The hypothesis that our sample consists for a significant portion out of blended sources is in contradiction with multi-wavelength observations. When I look at the positions of these sources at different wavelengths, I find that most sources have a counterpart in these multi-wavelength observations, also when chance-encounters are considered. Considering the high redshift nature of our sources, together with the possibility of lensing, these counterpart sources are most likely foreground, lensing galaxies. I compare the positions of the HerBS sources to both the Sloan Digital Sky Survey (SDSS), which covers 121 out of the 209 sources, and the VISTA Kilo-Degree Infrared Galaxy (VIKING) survey, which covers 98 HerBS sources. For the SDSS counterparts, I use the H-ATLAS catalogue of counterpart sources, which was done by using a statistical estimator. This statistical estimator assumes a certain angular distribution between the sources in the Herschel position, and the opticalor near-infrared observations. I expect the majority of my sources to be lensed, and therefore I adjust the original angular distribution by including the effect of gravitational lensing. The adjustment is based on 15 ALMA observations of lensed, bright H-ATLAS sources. The revised analysis finds 41 counterparts, instead of the 31 that were found by the initial analysis. This catalogue is not available for VIKING counterparts, and therefore I had to do the entire analysis for the VIKING counterparts, starting from the VIKING fields. I use the sextractor package to extract the potential counterparts, and then derive the necessary estimators for the statistical method. I find a significantly different angular distribution, even than the one derived from the 15 ALMA observations of lensed H-ATLAS sources. The angular distribution extends to much larger angular scales, potentially suggesting a stronger contribution to galaxy-cluster lensing, which produces larger angular offsets due to the larger masses and different mass profiles associated with galaxy clusters. In total, I find 60 counterparts with a reliability greater than 80% to the 98 HerBS sources covered by VIKING. Possibly, not all counterparts could be positively identified, as the analysis showed 88% of sources has a source within 10 arcseconds when taking chance encounters into account. This is mostly due to ambiguity between several nearby sources, which causes a low reliability of the counterpart identification, but it does allow us to state that an counterpart could be present. A cosmological model suggest that 76% of our sources are gravitationally lensed. This model assumes a certain distribution of halo masses, and lensing magnification based on mass density profiles. The validity of these models has been shown with the 15 ALMA observations of lensed H-ATLAS sources, and also agree with the SMA observations from Bussmann et al. (2013). The IRAM observations provide me with both line luminosities and line velocity widths. Larger galaxies are expected to be brighter, and have larger line velocity widths. The five sources with confirmed redshift (and therefore line luminosity) have a luminosity-to-velocity width ratio agreeing with a magnification of around 10, when compared to unlensed, and known lensed sources. I show that the SDSS is not deep enough to observe all the foreground galaxies, while the VIKING observations agree with the results from the simulation, with 60 sources actually cross-compared, and 88% of sources have a source nearby, when accounting for random chance