Star formation is a continuous ongoing process occurring over the lifetime of our Galaxy and the universe. However understanding how stars form from their pre-natal clouds of gas and dust remains a mystery. During the last two decades we have made remarkable progress toward unraveling this mystery mainly due to advances in observational technology especially at infrared and millimeter wavelengths which allow direct observation of the sites of star birth. Such observations suggest that embedded clusters may be the fundamental units of star formation in molecular clouds. Low star formation efficiency and rapid gas dispersal make these clusters disperse to provide the field star population. Consequently embedded clusters provide important laboratories for investigating fundamental issues of star formation within our Galaxy.

Since the earliest stages of star formation occur deep in clouds of gas and dust, they are hidden from optical view. Nevertheless, infrared and millimeter wavelength observations over the last twenty years have enabled increasingly detailed studies of the processes involved in stellar birth. High resolution, aperture synthesis imaging of the millimeter-wave radiation from dust and molecular gas in star-forming clouds has proven particularly effective. On the other hand, there can be pitfalls to be avoided in the use of mm/sub-mm interferometry techniques. Here, we consider what has been learned from currently-operating mm interferometers and the potential of the next generation of arrays.

Understanding the evolution of cores into protostellar envelopes is important in the star formation study. The initial condition of star formation such as density structure, kinematics, chemical composition, and magnetic field of dense cores can be derived from high resolution molecular line and continuum images at millimeter/submillimeter wavelengths. Besides, most low-mass stars are formed in clusters although the study of protoclusters is still rudimentary because of lack of observations with sufficiently high resolution to resolve individual cores/envelopes. High resolution observations with the coming SMA and ALMA will be crucial to testing and improving theoretical models of low-mass star formation.

A brief summary is presented of our current knowledge of the structure of cold molecular cloud cores that do not contain protostars, sometimes known as starless cores. The most centrally condensed starless cores are known as pre-stellar cores. These cores probably represent observationally the initial conditions for protostellar collapse that must be input into all models of star formation. The current debate over the nature of core density profiles is summarised. A cautionary note is sounded over the use of such profiles to ascertain the equilibrium status of cores. The magnetic field structure of pre-stellar cores is also briefly discussed.

We summarize the current status of the turbulent model of star formation in turbulent molecular clouds. In this model, clouds, clumps and cores form a hierarchy of nested density fluctuations caused by the turbulence, and either collapse or re-expand. Cores that collapse can be either internally sub- or super-sonic. The former cannot further fragment, and can possibly be associated with the formation of a single or a few stars. The latter, instead, can undergo turbulent fragmentation during their collapse, and probably give rise to a cluster of bound objects. The star formation efficiency is low because only a small fraction of the density fluctuations proceed to collapse. Those that do not may constitute a class of “failed” cores that can be associated with the observed starless cores. “Synthetic” observations of cores in numerical simulations of non-magnetic turbulence show that a large fraction of them have subsonic internal velocity dispersions, can be fitted by Bonnor-Ebert column density profiles, and exhibit “coherence” (an apparent independence of linewidth with column density near the projected core centers), in agreement with observed properties of molecular cloud cores.

This review covers hot cores in the context of high-mass star formation. After giving an overview of chemical processes and diversity during high-mass star formation, it reviews the ‘warm envelope’ phase which probably precedes the formation of hot cores. Some recent determinations of the cosmic-ray ionization rate are discussed, as well as recent evidence for hot cores around low-mass stars. Routes for future hot core research are outlined.

Recent observations with high angular resolution revealed chemical differentiation in several prestellar cores; while N2H+ emission peaks at the core center, CO, CS and CCS emission peaks are offset from the center. Molecular abundances also vary among cores; some cores have high CCS abundance and low N2H+ abundance, but others have less CCS and more N2H+. Numerical calculations of a chemical-reaction network in contracting cores show that these differentiations and variations can be diagnostics of physical evolution of cores, because molecular abundances and distributions are determined by the balance between the dynamical and chemical time scales. In prestellar cores, low temperatures and high densities cause adsorption of molecules onto grains. Depletion time scale varies among species; early-phase species deplete first because of destruction via gas-phase reactions in addition to the adsorption. N2H+ is the last to deplete because of the low adsorption energy of its parent molecule N2 and depletion of main reactants such as CO. Molecular D/H ratio is another probe of core evolution, since it increases as the adsorption proceeds.

We summarize recent progress of observational studies of infall in protostellar envelopes, with great emphasis on results obtained using millimeter and submillimeter interferometers. Interferometric observations allow us to spatially resolve kinematical structures of protostellar envelopes, enabling us to detect infalling motions in the envelope directly. High angular resolution observations of infalling envelopes having compact disks sufficiently bright in continuum show inverse P-Cygni profiles, which are the least ambiguous evidence for infall. Observations of infalling envelopes using the Submillimeter Array (SMA) may allow us to study the innermost infalling envelopes, where infalling motions most probably transform to Keplerian motions, leading to formation of Keplerian or protoplanetary disks around protostars.

Magnetic fields are believed to play an important role in the evolution of molecular clouds, from their large scale structure to dense cores, protostellar envelopes, and protoplanetary disks. How important is unclear, and whether magnetic fields are the dominant force driving star formation at any scale is also unclear. In this review we examine the observational data which address these questions, with particular emphasis on high angular resolution observations. Unfortunately the data do not clarify the situation. It is clear that the fields are important, but to what degree we don't yet know. Observations to date have been limited by the sensitivity of available telescopes and instrumentation. In the future ALMA and the SKA in particular should provide great advances in observational studies of magnetic fields, and we discuss which observations are most desirable when they become available.

We present recent JCMT and BIMA array polarimetry data of nearby star-forming regions in order to compare the core and cloud-scale magnetic field geometries in two regions of Orion. The similarity of the magnetic field geometry in these cores to that of their ambient clouds is contrasted with JCMT data toward the Barnard 1 dark cloud in Perseus, which reveal a different magnetic field orientation between the majority of the cores and the surrounding cloud; each of the cores exhibits a different mean polarization position angle. We conclude that the preservation of the magnetic field geometry is better in cores formed within clouds with ordered large scale structures. In Barnard 1, the cores may quickly exhibit a different polarization pattern if they have, for example, rotation which differs from the large scale cloud motions, or a weaker component of ordered fields. This could also explain why the cores exhibit such different geometries from each other in Barnard 1.

High angular resolution observations of nearby galaxies in the optical using ground-based and space-based telescopes have not only revealed the presence of young stellar clusters, but also allowed to study their properties in various dynamical environments. These studies have shown that young massive clusters (YMCs) have typical masses of a few 1000 M⊙ and sizes of a few parsec irrespective of their site of formation (such as bulges, spiral arms, starburst rings, or mergers). This points toward a universal formation mechanism for these stellar clusters.

Observations of the dust and gas content in high redshift galaxies allows one to study the reservoir for star formation in the early universe. These studies reveal extremely high star formation rates of a few 1000 M⊙ yr−1, while the distribution of the molecular gas still seems to be comparable to what is observed in the local universe. The detection of considerable amounts of molecular gas via its CO lines in the highest redshifted QSOs known today (up to z=6.4) indicates that star formation in the early universe has already produced considerable amounts of metals.

We review recent observations of molecular gas in nearby galaxies and their implications for the star formation law on large (> 1 kpc) scales. High-resolution data provided by millimetre interferometers are now adding to the basic understanding that has been provided by single-dish telescopes. In particular, they confirm the good correlation between star formation rate (SFR) and molecular gas surface densities, while at the same time revealing a greater degree of heterogeneity in the CO distribution. Galaxies classified as SAB or SB tend to show radial CO profiles that peak sharply in the inner ∼20″, indicative of bar-driven inflow. The observed Schmidt law index of ≍1.5 may result from a nearly linear relation between SFR and H2 mass coupled with a modest dependence of the molecular gas fraction on the total gas density. The normalisation of the Schmidt law, giving the characteristic timescale for star formation, may stem from the generic nature of interstellar turbulence.

Subarcsecond infrared and radio observations yield important information about the formation of super star clusters from their surrounding gas. We discuss the general properties of ionized and molecular gas near young, forming SSCs, as illustrated by the prototypical young, forming super star cluster nebula in the dwarf galaxy NGC 5253. This super star cluster appears to have a gravitationally bound nebula, and the lack of molecular gas suggests a very high star formation efficiency, consistent with the formation of a large, bound cluster.

Astrophysical masers are one of the most readily detected signposts of high-mass star formation. Their presence indicates special conditions, probably indicative of a specific evolutionary phase. Masers also represent the ultimate high-resolution probe of star formation with the potential to reveal information on the kinematics and physical conditions within the region at milli-arcsecond resolution. To date this potential has largely remained unfulfilled, however, recent advances suggest that this will soon change.

The key to unlocking the potential of masers lies in identifying where they fit within the star formation jigsaw puzzle. I will review recent high resolution observations of OH, water and methanol maser transitions and what they reveal. I also briefly discuss how multi-transition observations of OH and methanol masers are being used to constrain maser pumping models and through this estimate the physical conditions in the masing region.

The collapse of massive molecular clumps can produce high mass stars, but the evolution is not simply a scaled-up version of low mass star formation. Outflows and radiative effects strongly hinder the formation of massive stars via accretion. A necessary condition for accretion growth of a hydrostatic object up to high masses M ≳ 20M⊙ (rather than coalescence of optically thick objects) is the formation of, and accretion through, a circumstellar disk. Once the central object has accreted approximately 10 M⊙ it has already evolved to core hydrogen-burning; the resultant main sequence star continues to accrete material as it begins to photoevaporate its circumstellar disk (and any nearby disks) on a timescale of ∼105 yr, similar to the accretion timescale. Until the disk(s) is (are) completely photoevaporated, this configuration is observable as an ultra-compact Hii region (UCHii). The final mass of the central star (and any nearby neighboring systems) is determined by the interplay between radiation acceleration, UV photoevaporation, stellar winds and outflows, and the accretion through the disk.

Several aspects of this evolutionary sequence have been simulated numerically, resulting in a “proof of concept”. This scenario places strong constraints on the accretion rate necessary to produce high mass stars and offers an opportunity to test the accretion hypothesis.

High-mass star formation is not well understood chiefly because examples are deeply embedded, relatively distant, and crowded with sources of emission. Using VLA and VLBA observations of H2O and SiO maser emission, we have mapped in detail the structure and proper motion of material 20-500 AU from the closest high-mass YSO, radio source I in the Orion KL region. We observe streams of material driven in a rotating, wide angle, bipolar wind from the the surface of an edge-on accretion disk. The example of source I provides strong evidence that high-mass star formation proceeds via accretion.

In this paper I discuss our current understanding of the formation of intermediate mass stars and the evolution of their circumstellar material. The results of the search for clusters around Herbig Ae/Be stars are discussed in the context of the possible relationship of the formation of massive stars via coalescence process. Current surveys are not able to offer a definite answer on whether there is a physical relationship between the most massive intermediate mass stars and the presence of clusters. The present day stellar densities in these clusters is too low to allow for frequent stellar collisions, but do not exclude the possibility of coalescence at earlier cluster ages. I review the evidence for and properties of circumstellar disks around Herbig Ae stars, as well as recent observational evidence for dust evolution within these disks. Evidence is presented for the presence of evolved dust in some of these disks, with grain growth to sizes as large as a few centimeters. The status of the search for the more elusive disks around Herbig Be stars is also discussed. Current observations show that disks around Herbig Be stars progenitors may have been common and massive in the protostellar phase, but optically visible Herbig Be stars, albeit very young, are never found to be surrounded by conspicuous dusty disks, suggesting dispersion timescales as short as 1 Myr.

The understanding of the formation process of massive stars requires a detailed knowledge of the physical conditions of the cloud environment which is thought to play a critical role in determining the formation mechanism. In recent years there has been a rapid growth of observational and theoretical studies concerning the formation of massive stars. Here I review observational data gathered during the last few years which are providing key evidence concerning the physical processes that take place during the formation of massive stars.