Considerable progress has been made in our knowledge of galaxy evolution from the faint galaxy studies made possible by HST deep imaging, ground-based spectroscopy, and wide faint galaxy surveys (e.g. Ellis 1998; Steidel et al. 1996; Schade et al. 1996). This progress has been led by the optical domain, but crucial information came also from the far-infrared and millimeter domains: the cosmic IR and submm background radiation discovered by COBE (e.g. Puget et al. 1996; Hauser et al. 1998) yields an insight on the global past star-formation of the Universe, and the sources discovered at high redshift in the millimeter continuum and lines yield information on the structure of the past starbursts (Smail et al. 1997; Guilloteau et al. 1999). From all these data, a global view of star formation as a function of lookback time has been derived (e.g. Madau et al. 1996; Glazebrook et al. 1999), which still is submitted to big uncertainties, especially at high redshift. In particular, it is possible that the optically derived star formation rate is under-estimated, due to dust obscuration, and that only infrared/submm surveys could give the correct information (Guiderdoni et al. 1997).
This review focus on the dust and molecular content of galaxies, as a way to trace the evolution of star formation, and to detect the location of starbursts as a function of redshift. First the present state of knowledge is detailed, concerning CO emission lines as well as dust continuum, and their interpretation is discussed (respective role of starburst and AGN for instance). Then perspectives are drawn concerning the future surveys that will be conducted with the next generation of millimeter instruments.