Knud Jahnke, Sebastian F. Sanchez, Lutz Wisotzki, Marco Barden, Steven V. W. Beckwith, Eric F. Bell, Andrea Borch, John A. R. Caldwell, Boris Häußler, Catherine Heymans, Shardha Jogee, Daniel H. McIntosh, Klaus Meisenheimer, Chen Y. Peng, Hans-Walter Rix, Rachel S. Somerville, C. Wolf

• We have performed Hubble Space Telescope imaging of a sample of 23 high-redshift (1.8<z<2.75) active galactic nuclei (AGNs), drawn from the COMBO-17 survey. The sample contains moderately luminous quasars (M(B)similar to-23). The data are part of the Galaxy Evolution from Morphologies and SEDs imaging survey that provides high-resolution optical images obtained with the Advanced Camera for Surveys in two bands (F606W and F850LP), sampling the rest-frame UV flux of the targets. To deblend the AGN images into nuclear and resolved (host galaxy) components, we use a point-spread function subtraction technique that is strictly conservative with respect to the flux of the host galaxy. We resolve the host galaxies in both filter bands in nine of the 23 AGNs, whereas the remaining 14 objects are considered nondetections, with upper limits of less than 5% of the nuclear flux. However, when we co-add the unresolved AGN images into a single high signal-to-noise ratio composite image, we find again an unambiguously resolved host galaxy. TheWe have performed Hubble Space Telescope imaging of a sample of 23 high-redshift (1.8<z<2.75) active galactic nuclei (AGNs), drawn from the COMBO-17 survey. The sample contains moderately luminous quasars (M(B)similar to-23). The data are part of the Galaxy Evolution from Morphologies and SEDs imaging survey that provides high-resolution optical images obtained with the Advanced Camera for Surveys in two bands (F606W and F850LP), sampling the rest-frame UV flux of the targets. To deblend the AGN images into nuclear and resolved (host galaxy) components, we use a point-spread function subtraction technique that is strictly conservative with respect to the flux of the host galaxy. We resolve the host galaxies in both filter bands in nine of the 23 AGNs, whereas the remaining 14 objects are considered nondetections, with upper limits of less than 5% of the nuclear flux. However, when we co-add the unresolved AGN images into a single high signal-to-noise ratio composite image, we find again an unambiguously resolved host galaxy. The recovered host galaxies have apparent magnitudes of 23.0<F606W<26.0 and 22.5<F850LP<24.5, with rest-frame UV colors in the range -0.2<(F606W-F850LP)(obs)<2.3. The rest-frame absolute magnitudes at 200 nm are -20.0<M-200 nm<-22.2. The photometric properties of the composite host are consistent with the individual resolved host galaxies. We find that the UV colors of all host galaxies are substantially bluer than expected from an old population of stars with formation redshift z<=5, independent of the assumed metallicities. These UV colors and luminosities range up to the values found for Lyman break galaxies (LBGs) at z=3. Our results suggest either a recent starburst of, e. g., a few percent of the total stellar mass at 100 Myr before observation, with mass fraction and age strongly degenerate, or the possibility that the detected UV emission may be due to young stars forming continuously. For the latter case we estimate star formation rates of typically &SIM;6 M&ODOT; yr(-1) (uncorrected for internal dust attenuation), which again lies in the range of rates implied from the UV flux of LBGs. Our results agree with the recent discovery of enhanced blue stellar light in AGN hosts at lower redshifts…