The main sequence of star-forming galaxies
The study of the present and past star formation in individual galaxies relies on detailed studies of the stellar populations within galaxies. Such studies are only feasible with galaxies in the Local Group (within 1.5 Mpc from us) due to observation limitations. However, we can study the dominant drivers of the changes in star formation of galaxies statistically.
Although galaxies show a broad diversity, many tight scaling relations have been found. Hubble tried to interpret the Hubble sequence as a galaxy evolution sequence where ellipticals formed first and spirals later6 (Hubble, 1926). With the advent of large surveys like SDSS, as mentioned previously, the galaxy color-magnitude diagram shows a distinctive bimodal distribution (Fig. 1.3-Left). The u-r color is a proxy of specific star formation rates (sSFRs © SFR/Mstar) and the magnitude can be converted into stellar masses. Therefore, the distribution of red and blue galaxies in the color-magnitude diagram indicates that stellar masses are regulating the star formation in galaxies. Indeed, star-forming galaxies show a tight correlation ( with a dispersion less than 0.3 dex, Sargent et al. (2012); Schreiber et al. (2015)) between their SFR and stellar masses over a large redshift range, from z ≥ 0 to z & 4, and over a broad stellar mass range from down to 109. M§ to up to 1011.4 M§ (Brinchmann et al., 2004; Elbaz et al., 2007; Daddi et al., 2007; Noeske et al., 2007; Rodighiero et al., 2011; Whitaker et al., 2012; Daddi et al., 2009; Pannella et al., 2009; Magdis et al., 2010; Heinis et al., 2013; Pannella et al., 2015; Steinhardt et al., 2014; Bouwens et al., 2012; Stark et al., 2009; Stark et al., 2013). This is the so-called « main sequence » correlation. A minority of galaxies (« starbursts », only a few percent of the total star-forming galaxies, Rodighiero et al. (2011); Sargent et al. (2012); Schreiber et al. (2015), see also Fig. 1.4-Right) have highly elevated SFRs above the main sequence (Elbaz et al., 2011) and there are also galaxies with deeply reduced SFRs below the main sequence. This is illustrated in Fig. 1.3-Right.
The main sequence evolves with redshift, as shown in Fig. 1.4-Left, which implies that at a specific stellar mass, a galaxy would form stars more rapidly at higher redshift, i.e., the sSFR evolves continuously with redshift. In another word, the SFR of a main sequence galaxy at high redshift would be associated to a starburst galaxy with the same stellar mass but at lower redshift. On the other hand, the dispersion of the main sequence barely evolves with redshift, which means that at a specific stellar mass and a specific redshift, ≥ 68% of star-forming galaxies form stars at a rate within a factor of ≥ 2 (Schreiber et al., 2015). If we define the galaxies on the main sequence as « normal » star-forming galaxies, then the normal and starburst galaxies represent the two main modes that control the growth of galaxies, because starbursts are not simply an extension of main-sequence galaxies, as shown by the double Gaussian decomposition of the starburstiness (RSB = SFR/SFRMS) distribution of star-forming galaxies in Fig. 1.3-Right. In fact, these two populations show different evolution paths. Starbursts are often interpreted as driven by mergers which can substantially boost SFR instantaneously, while normal galaxies form stars more steadily. The low percentage of starburst in star-forming galaxies indicates that galaxy mergers have very limited contribution to the global star formation history. Moreover, at 0 < z < 3, the fraction of starbursts is not evolving with redshift, which means that the « normal » mode dominates the star formation in most galaxies (Rodighiero et al., 2011; Sargent et al., 2012; Schreiber et al., 2015). The steadily evolving main sequence has been used to constrain the cosmic star formation histories.
The quest for resilient supply chains
These phenomena are of high concern for businesses, insurers, and in general for orga-nizations operating through such multi-tiered, complex web of intermediaries, such as humanitarian organizations. Managing supply chain risks faces specific challenges, some of which have been analyzed through the lens of resilience (Sec. 2.2.1). Although, compet-itiveness objectives may hamper the design of resilient supply chains, certain aspects of resilience, such as agility and permanent reorganization, are seen as being a way to better compete (Sec. 2.2.2).
The challenges of mitigating risks in complex supply chains
Supply chain disruptions are rather frequent — at least one per year for 80% of the re-spondents the BCI survey (2014). Their negative impact on the financial performance of companies has been empirically confirmed (Hendricks and Singhal, 2003, 2005). Miti-gating the risk of supply chain disruptions widely diﬀers from the management of other operational risks. Supply chains may indeed bring to your door the risks taken by another firm far away, both its operational risks — e.g., an accident on a production line — and its environmental risks — e.g., a climate or geopolitical event.
Managers therefore need to increase their monitoring capacity. However, they often lack visibility over their supply chain. While firms usually know their direct suppliers, they often struggle to keep track of their sub-suppliers, also called tier-2 suppliers, and of entities further away in the chain (BCI, 2014; Wang et al., 2015). Half of the disruptions seem however to originate from this deeper segment (BCI, 2014). In addition, supply chains are fluctuating systems — e.g., suppliers change their contractors, firms go bankrupt, others enter the market — and are therefore hard to map in real time.
Inherent diﬃculties of interorganizational communication may also accentuate the propagation of supply disruptions. Jüttner et al. (2003) identify such network-related risks: unclear responsibility, lack of responsiveness or overreaction, distorted information and mistrust. A small fluctuation in demand at one point of the chain may be magnified as orders cascade up the chain, leading to excess inventory, production down time, and transportation peaks. This phenomenon, known as the bullwhip eﬀect (Lee et al., 1997), is well known by supply chain managers, empirical documented (e.g. Thun and Hoenig, 2011) and was has been experimentally tested for decades through the so-called beer game (Sterman, 1989).
Managing the risk of supply disruption has become a prominent topic of supply chain management (SCM)8 . The main pillar of SCM is the coordination of processes across organizational boundaries, in which managers need to engage to avoid these issues.
Gas velocity dispersion and the energy sources
Galaxies at different cosmic epochs show quite distinct properties. Compared to their high-redshift counterparts at similar stellar masses, local star-forming galaxies are larger, and have relatively lower gas fractions and lower SFRs (Leroy et al., 2005; Daddi et al., 2010a; Tacconi et al., 2010; Madau and Dickinson, 2014). They are also less likely to experience violent events such as major mergers and gas accretion (Baugh et al., 1996; Genzel et al., 2008; Robotham et al., 2014). Despite of all the various properties, galaxy discs at all epochs tend to be in a state of marginal gravitational stability, which can be characterized by the close to unity Toomre (1964) Q parameter, as Q = Ÿ‡/fiGΣ, where Ÿ is the epicyclic frequency of the galaxy’s rotation, ‡ the velocity dispersion, represents the effect of pressure, and Σ the mass surface density, represents the effect of gravity. However, many theoretical and observational studies suggest that gas in higher-z galaxies has larger random motions compared to gas in nearby galaxies. Galaxies at z > 1 have velocity dispersions in the range of 50–100 km s≠1 (Nesvadba et al., 2006; Lehnert et al., 2009; Lehnert et al., 2013; Förster Schreiber et al., 2009; Wisnioski et al., 2015) and show an almost linear correlation with the SFR, while local galaxies show typical velocity dispersions of < 50 km s≠1 (Varidel et al., 2016; Yu et al., 2019). On the other hand, both local and high-z galaxies show velocity dispersions higher than expected from only the thermal contribution of gas. The characteristic temperature of 104 K corresponds to a typical velocity dispersion of ≥9 km s≠1 for the ionized gas emitting at H– (Glazebrook, 2013).
The dominant energy source of the non-thermal turbulent motions is unclear. Numerous drivers have been proposed, including star formation feedback (Mac Low and Klessen, 2004; Krumholz and Matzner, 2009; Murray et al., 2010), radial transport of gas in discs due to gravitation (Krumholz and Burkhart, 2016; Krumholz et al., 2018), gas accretion from the intergalactic medium and minor mergers (Dekel et al., 2009; Glazebrook, 2013), galactic shear from the differential rotation in disc galaxies (Krumholz and Burkhart, 2015; Federrath et al., 2016; Federrath et al., 2017), etc. In Chapter 5, I studied the driver of turbulent motion of eight spatially-resolved nearby star-forming galaxies. The results show that star formation feedback is not the main energy source of the turbulent motions in galaxies with low SFR surface density. However, recent studies by Krumholz et al. (2018) and Varidel et al. (2020) on the global properties of galaxies found that the models taking into consideration of both the star formation feedback and the gravitational energy release from radial transport of gas can yield excellent agreement with the observations of galaxies with SFR ranging from 10≠4 M§ yr≠1 to 103 M§ yr≠1. The model predicts a transition from gravity-dominantly-driven turbulence in high-z galaxies to star-formation-driven turbulence in low-z galaxies, where SFR is lower. The distinct conclusions come from the different treatments of beam smearing effects, as will be explained below. This model also explains why galaxy bulges tend to form at high redshift and discs at lower redshift, and why galaxies tend to quench inside-out, because the gas accretion rate increases much faster with velocity dispersion, than SFR with velocity dispersion (M˙acc Ã ‡gas3 vs. SFR Ã ‡gas) and then masses are transported more inward to a bulge in high-z galaxies and remain in the outskirts to form a disk in low-z galaxies (Krumholz et al., 2018).
Integral field unit (IFU), datacube and beam smearing eﬀect
The integral field unit technique has been widely used in optical and near-IR to study galaxy kinematics. It allows us to obtain spatially resolved spectral information in the galaxies. In Chapter 5, I will present my work on the energy sources of the turbulent motions in local star-forming galaxies making use of the IFU survey, the SAMI Galaxy Survey (Croom et al., 2012; Bryant et al., 2015). The three dimensional information from such IFU observations are saved in the so-called datacubes (Figure 1.14). This is similar to the data product obtained from the interferometers at submillimeter to radio wavelengths.
Measurements of velocity dispersions from these datacubes are limited by the spectral and spatial resolution of the instruments. The observed emission line is broadened by the spectral resolution, but this can be addressed by convolving the line spread function into the emission line fitting. The limited spatial resolution blurs the spatial distribution of the intrinsic flux, the line of sight velocity profile, and the line of sight velocity dispersion within the smallest resolved area. As a result, the observed velocity dispersion is elevated with the unresolved velocity gradient. This is the so-called beam smearing effect. Several tools have been developed to account for beam-smearing effect, e.g., 3DBBAROLO12 (DiTeodoro and Fraternali, 2015),GBKFIT13 (Bekiaris et al., 2016), GALPAK3D 14(Bouché et al., 2015), and BLOBBY3D (Varidel et al., 2019). They construct a 3D modelled cube for the galaxy and then spatially convolve the cube per spectral slice to simulate the effect of beam smearing. The convolved cube is finally compared to the observed data.
Optically dark galaxies and their association with a proto-cluster in formation at z ≥ 3.5
This work has been published in Astronomy & Astrophysics (Zhou et al., 2020), and is presented in Chapter 3. I studied thoroughly the properties of the six optically dark galaxies detected in the GOODS-ALMA 1.1 mm continuum survey. None of them is listed in the deepest H-band based CANDELS catalog in the GOODS-South field. Five of them suffer from the confusion with bright neighboring galaxies even in the highly resolved optical to near-IR images. To retrieve information at these wavelengths, I performed a de-blending procedure with the method developed by one of the co-authors, C. Schreiber (Schreiber et al., 2018a, code). It turns out that after this deblending analysis two out of the six galaxies end up having H-band counterparts with magnitudes brighter than the detection limit determined by in the CANDELS team on their catalog. They were missed because they were considered as part of the neighboring galaxy because of the confusion limit as a drawback of the source extraction procedure. I fitted the optical-to-MIR SEDs of the optically dark galaxies and their neighbors respectively. The derived redshifts confirm the confusion due to a projection effect, meaning that the optically dark galaxies and their neighbors are at different redshifts.
As co-I, I contributed to a spectroscopic follow-up using ALMA (2018.1.01079.S, PI: M. Franco) to identify the exact redshifts of these optically dark galaxies. I analyzed the data with CASA and GILDAS and found that one emission line was detected in two of the galaxies.
To study the environment where these galaxies reside in, I constructed the surface number density map of galaxies in the GOODS-South field based on the ZFOURGE catalog. We choose ZFOURGE rather than CANDELS because the ZFOURGE catalog is based on detections extracted from the near-IR Ks band images, where galaxies at higher redshift appear brighter than they are in the optical images. We present evidence that nearly 70 % of the optically dark galaxies belong to the same over-density of galaxies at z ≥ 3.5. We also found that the most massive one of the optically dark galaxies is also the most massive galaxy at z > 3 in the GOODS-ALMA field after excluding galaxies hosting a luminous AGN potentially responsible for an overestimation of their stellar mass. This galaxy, AGS24, falls at the very center of the peak of the galaxy surface density. This suggests that the surrounding over-density is a proto-cluster in the process of virialization and this massive galaxy is the candidate progenitor of the future Brightest Cluster Galaxy (BCG). These optically dark galaxies unveiled by ALMA are good tracers of such large-scale structures in the early Universe and they can serve to test current theories on the formation of the most massive galaxies during the first billin years of the Universe.
Table of contents :
1.1 Star formation: the driver of galaxy evolution
1.2 Star formation obscured by interstellar dust
1.2.1 Multi-wavelength study of galaxies
1.2.2 The main sequence of star-forming galaxies
1.2.3 Cosmic star formation history
1.2.4 Observing star-forming galaxies in the (sub)millimeter
1.2.5 Optically-dark galaxies
1.3 Star formation fueled by gas
1.3.1 Gas tracers
1.3.2 Star formation law
1.3.3 Gas content
1.4 Star formation as one of the energy sources of gas
1.4.1 Gas velocity dispersion and the energy sources
1.4.2 Integral field unit (IFU), datacube and beam smearing effect
2 Summary of the work done in this thesis
2.1.1 Optically dark galaxies and their association with a proto-cluster in formation at z ≥3.5
2.1.2 Data compilation
2.1.3 Nascent AGNs in GOODS-ALMA
2.1.4 Contribution as third author
2.2 Extremely metal-poor galaxies
2.2.1 Spatially resolved dust emission
2.2.2 Gas content in IZw18
2.3 The SAMI Integral field Units (IFU) Survey
2.3.1 Energy source of turbulence in star-forming galaxies
3 GOODS-ALMA: optically-dark ALMA galaxies shed light on a cluster in formation at z = 3.5
3.2 Data and observations
3.2.1 ALMA data and observations
3.2.2 Ancillary data
3.2.3 Origin of the redshifts and stellar masses
3.2.4 Derived parameters of the optically dark galaxies
3.3 Results of the ALMA spectroscopic follow-up
3.3.3 Upper limits of AGS11, AGS15 and AGS24
3.4 GOODS-ALMA optically-dark galaxies
3.4.1 AGS4, an extremely massive galaxy at z=3.556 and a case of blending in the Hubble H-band image
3.4.2 AGS25, the most distant optically-dark galaxy in GOODS-ALMA
3.5 An over-density at z ≥3.5 in GOODS-ALMA
3.5.1 Clustering properties of optically dark galaxies
3.5.2 A clear peak at z ≥3.5 in the redshift distribution
3.5.3 Optically-dark galaxies at z ≥3.5
3.5.4 Spatial distribution of galaxies at z ≥3.5 in the GOODS-ALMA field
3.5.5 Dynamical state of the proto-cluster at z ≥3.5
4 Extremely weak CO emission in IZw 18
4.3.1 CO J=2-1
4.3.2 1.3mm continuum
4.4.1 SED and Submilimetre excess
4.4.2 Infrared luminosity and SFR versus LÕCO
4.4.3 The structure of the interstellar medium
5 The SAMI Galaxy Survey: energy sources of the turbulent velocity dispersion in spatiallyresolved local star-forming galaxies
5.2 Sample and data analysis
5.2.1 Sample selection
5.2.2 Gas kinematic information
5.2.3 Spatial resolution
5.3.1 The spatial distribution of !SFR, vgas, and ‡gas
5.3.2 The ‡gas – !SFR relation in local and high redshift star-forming galaxies
5.4.1 Main driver(s) of velocity dispersion
6 Spatially resolved dust emission of extremely metal poor galaxies1
6.2 Sample, observations and data analysis
6.2.1 The sample
6.2.3 Photometric measurements
6.3 The far-IR SEDs
6.3.1 The color-color diagrams
6.3.2 Modified black-body fitting
6.3.3 Spatial variations of SEDs and dust heating Mechanism
6.4 Dust-to-stellar mass ratio
7 Conclusion and perspectives
7.2.1 More on the optically dark galaxies
7.2.2 AGN feedback on high-z star-forming galaxies