Deep observations are revealing a growing number of young galaxies in the first billion year of cosmic time. Compared to typical galaxies at later times, they show more extreme emission-line properties, higher star formation rates, lower masses, and smaller sizes. However, their faintness precludes studies of their chemical abundances and ionization conditions, strongly limiting our understanding of the physics driving early galaxy build-up and metal enrichment. Here we study a rare population of UV-selected, sub$-L^{*}$(z=3) galaxies at redshift 2.4$<z<$3.5 that exhibit all the rest-frame properties expected from primeval galaxies. These low-mass, highly-compact systems are rapidly-forming galaxies able to double their stellar mass in only few tens million years. They are characterized by very blue UV spectra with weak absorption features and bright nebular emission lines, which imply hard radiation fields from young hot massive stars. Their highly-ionized gas phase has strongly sub-solar carbon and oxygen abundances, with metallicities more than a factor of two lower than that found in typical galaxies of similar mass and star formation rate at $z\lesssim$2.5. These young galaxies reveal an early and short stage in the assembly of their galactic structures and their chemical evolution, a vigorous phase which is likely to be dominated by the effects of gas-rich mergers, accretion of metal-poor gas and strong outflows.

The European Space Agency has invested heavily in two cornerstones missions; Herschel and Planck. The legacy data from these missions provides us with an unprecedented opportunity to study cosmic dust in galaxies so that we can answer fundamental questions about, for example: the origin of the chemical elements, physical processes in the interstellar medium (ISM), its effect on stellar radiation, its relation to star formation and how this relates to the cosmic far infrared background. In this paper we describe the DustPedia project, which is enabling us to develop tools and computer models that will help us relate observed cosmic dust emission to its physical properties (chemical composition, size distribution, temperature), to its origins (evolved stars, super novae, growth in the ISM) and the processes that destroy it (high energy collisions and shock heated gas). To carry out this research we will combine the Herschel/Planck data with that from other sources of data, providing observations at numerous wavelengths (< 41) across the spectral energy distribution, thus creating the DustPedia database. To maximise our spatial resolution and sensitivity to cosmic dust we limit our analysis to 4231 local galaxies (v < 3000 km/s) selected via their near infrared luminosity (stellar mass). To help us interpret the data we have developed a new physical model for dust (THEMIS), a new Bayesian method of fitting and interpreting spectral energy distributions (HerBIE) and a state-of-the-art Monte Carlo photon tracing radiative transfer model (SKIRT). In this the first of the DustPedia papers we describe the project objectives, data sets used and provide an insight into the new scientific methods we plan to implement.

We report the discovery of 31 low-luminosity (-14.5 > M_{AB}(B) > -18.8), extreme emission line galaxies (EELGs) at 0.2 < z < 0.9 identified by their unusually high rest-frame equivalent widths (100 < EW[OIII] < 1700 A) as part of the VIMOS Ultra Deep Survey (VUDS). VIMOS optical spectra of unprecedented sensitivity ($I_{AB}$ ~ 25 mag) along with multiwavelength photometry and HST imaging are used to investigate spectrophotometric properties of this unique sample and explore, for the first time, the very low stellar mass end (M* < 10^8 M$_{\odot}$) of the luminosity-metallicity (LZR) and mass-metallicity (MZR) relations at z < 1. Characterized by their extreme compactness (R50 < 1 kpc), low stellar mass and enhanced specific star formation rates (SFR/M* ~ 10^{-9} - 10^{-7} yr^{-1}), the VUDS EELGs are blue dwarf galaxies likely experiencing the first stages of a vigorous galaxy-wide starburst. Using T_e-sensitive direct and strong-line methods, we find that VUDS EELGs are low-metallicity (7.5 < 12+log(O/H) < 8.3) galaxies with high ionization conditions, including at least three EELGs showing HeII 4686A emission and four EELGs of extremely metal-poor (<10% solar) galaxies. The LZR and MZR followed by EELGs show relatively large scatter, being broadly consistent with the extrapolation toward low luminosity and mass from previous studies at similar redshift. However, we find evidences that galaxies with younger and more vigorous star formation -- as characterized by their larger EWs, ionization and sSFR -- tend to be more metal-poor at a given stellar mass.

Using new spectroscopic observations obtained as part of the VIMOS Ultra-Deep Survey (VUDS), we perform a systematic search for overdense environments in the early universe ($z>2$) and report here on the discovery of Cl J0227-0421, a massive protocluster at $z=3.29$. This protocluster is characterized by both the large overdensity of spectroscopically confirmed members, $\delta_{gal}=10.5\pm2.8$, and a significant overdensity in photometric redshift members. The halo mass of this protocluster is estimated, by a variety of methods, to be roughly $3\times10^{14}$ $\mathcal{M}_{\odot}$ at $z\sim3.3$, which, evolved to $z=0$ results in a halo mass rivaling or exceeding that of the Coma cluster. The properties of 19 spectroscopically confirmed member galaxies are compared with a large sample of VUDS/VVDS galaxies in lower density field environments at similar redshifts. We find tentative evidence for an excess of redder, brighter, and more massive galaxies within the confines of the protocluster relative to the field population, which suggests that we may be observing the beginning of environmentally-induced quenching. The properties of these galaxies are investigated, including a discussion of the brightest protocluster galaxy which appears to be undergoing vigorous coeval nuclear and starburst activity. The remaining member galaxies appear to have characteristics which are largely similar to the field population. Though we find weaker evidence of the suppression of the median star formation rates amongst and differences in stacked spectra of member galaxies with respect to the field, we defer any conclusions of these trends to future work with the ensemble of protostructures that are found in the full VUDS sample.

(arXiv abridged abstract) The observed UV rest-frame spectra of distant galaxies are the result of their intrinsic emission combined with absorption along the line of sight produced by the inter-galactic medium (IGM). Here we analyse the evolution of the mean IGM transmission Tr(Ly_alpha) and its dispersion along the line of sight for 2127 galaxies with 2.5<z<5.5 in the VIMOS Ultra Deep Survey (VUDS). We fit model spectra combined with a range of IGM transmission to the galaxy spectra using the spectral fitting algorithm GOSSIP+. We use these fits to derive the mean IGM transmission towards each galaxy for several redshift slices from z=2.5 to z=5.5. We find that the mean IGM transmission defined as Tr(Ly_alpha)=e^{-tau} (with tau the HI optical depth) is 79%, 69%, 59%, 55% and 46% at redshifts 2.75, 3,22, 3.70, 4.23, 4.77, respectively. We compare these results to measurements obtained from quasars lines of sight and find that the IGM transmission towards galaxies is in excellent agreement with quasar values up to redshift z~4. We find tentative evidence for a higher IGM transmission at z>= 4 compared to results from QSOs, but a degeneracy between dust extinction and IGM prevents to draw firm conclusions if the internal dust extinction for star-forming galaxies at z>4 takes a mean value significantly in excess of E(B-V)>0.15. Most importantly, we find a large dispersion of IGM transmission along the lines of sight towards distant galaxies with 68% of the distribution within 10 to 17% of the median value in delta z=0.5 bins, similar to what is found on the LOS towards QSOs. We demonstrate the importance of taking into account this large range of IGM transmission when selecting high redshift galaxies based on their colour properties (e.g. LBG or photometric redshift selection) or otherwise face a significant incompleteness in selecting high redshift galaxy populations.

[Abridged] We characterise a massive proto-cluster at z=2.895 that we found in the COSMOS field using the spectroscopic sample of the VIMOS Ultra-Deep Survey (VUDS). This is one of the rare structures at z~3 not identified around AGNs or radio galaxies, so it is an ideal laboratory to study galaxy formation in dense environments. The structure comprises 12 galaxies with secure spectroscopic redshift in an area of 7'x8', in a z bin of Dz=0.016. The measured galaxy number overdensity is delta_g=12+/-2. This overdensity has total mass of M~8.1x10^(14)M_sun in a volume of 13x15x17 Mpc^3. Simulations indicate that such an overdensity at z~2.9 is a proto-cluster that will collapse in a cluster of total mass M~2.5x10^(15)M_sun at z=0. We compare the properties of the galaxies within the overdensity with a control sample at the same z but outside the overdensity. We did not find any statistically significant difference between the properties (stellar mass, SFR, sSFR, NUV-r, r-K) of the galaxies inside and outside the overdensity. The stacked spectrum of galaxies in the overdensity background shows a significant absorption feature at the wavelength of Lya redshifted at z=2.895 (lambda=4736 A), with a rest frame EW = 4+/- 1.4 A. Stacking only background galaxies without intervening sources at z~2.9 along their line of sight, we find that this absorption feature has a rest frame EW of 10.8+/-3.7 A, with a detection S/N of ~4. These EW values imply a high column density (N(HI)~3-20x10^(19)cm^(-2)), consistent with a scenario where such absorption is due to intervening cold gas streams, falling into the halo potential wells of the proto-cluster galaxies. However, we cannot exclude the hypothesis that this absorption is due to the diffuse gas within the overdensity.

We present the VIMOS Ultra Deep Survey (VUDS), a spectroscopic redshift survey of ~10.000 very faint galaxies to study the major phase of galaxy assembly 2<z<~6. The survey covers 1 deg^2 in 3 separate fields: COSMOS, ECDFS and VVDS-02h, with targets selection based on an inclusive combination of photometric redshifts and color properties. Spectra covering 3650<lambda<9350 A are obtained with VIMOS on the ESO-VLT with integration times of 14h. Here we present the survey strategy, the target selection, the data processing, as well as the redshift measurement process, emphasizing the specific methods adapted to this high redshift range. The spectra quality and redshift reliability are discussed, and we derive a completeness in redshift measurement of 91%, or 74% for the most reliable measurements, down to i_AB=25, and measurements are performed all the way down to i_AB=27. The redshift distribution of the main sample peaks at z=3-4 and extends over a large redshift range mainly in 2 < z < 6. At 3<z<5, the galaxies cover a large range of luminosities -23< M_U < -20.5, stellar mass 10^9 M_sun< M_star < 10^{11} M_sun, and star formation rates 1 M_sun/yr< SFR < 10^3 M_sun/yr. We discuss the spectral properties of galaxies using individual as well as stacked spectra. The comparison between spectroscopic and photometric redshifts as well as color selection demonstrate the effectiveness of our selection scheme. With ~6000 galaxies with reliable spectroscopic redshifts in 2<z<6 expected when complete, this survey is the largest at these redshifts and offers the opportunity for unprecedented studies of the star-forming galaxy population and its distribution in large scale structures during the major phase of galaxy assembly.