1. Institute for Astronomy, University of Hawaii, 2680 Woodlawn Drive, Honolulu, Hawaii 96822, USA
2. Astronomical Institute, Tohoku University, Aoba, Sendai 980-8578, Japan
3. ISO Science Operations Centre, Astrophysics Division of ESA, Villafranca, E-28080, Madrid, Spain
4. Institute of Space and Astronautical Science, Yoshinodai, Sagamihara, Kanagawa 229, Japan
5. Institute of Astronomy, University of Tokyo, 2-21-1 Osawa, Mitaka, Tokyo 181, Japan

Correspondence to: A. J. Barger¹ Correspondence and requests for materials should be addressed to A.J.B. (e-mail: Email: barger@ifa.hawaii.edu).

Abstract

Optical surveys of the global star-formation rate in high-redshift galaxies show a strong peak in activity at a redshift of z 1.5, which implies that most of the star formation¹ has already been seen. High-redshift galaxies may, however, emit most of their energy at submillimetre wavelengths, if they contain substantial amounts of dust that absorbs the starlight and reradiates it as far-infrared light. Here we report a deep survey of a blank region of sky, performed at submillimetre wavelengths (450 and 850 m). We detect luminous sources in the 850-m band which, if they have similar spectra to low-redshift ultraluminous infrared galaxies and are primarily powered by star formation, must each be converting more than 100 solar masses of gas per year into stars: this is larger than themaximum star-formation rates inferred for most optically selected galaxies². The total amount of star formation at high redshifts is essentially fixed by the level of background light, but where the peak activity occurs at submillimetre wavelengths is not yet well established. However, the background light inferred from the sources that we have detected is already comparable to that from the optically selected sources. Establishing the main epoch of star formation will therefore require a combination of optical and submillimetre studies.