We present deep HST/WFPC2 , rest-frame U images of 17 \sim L ^ { \star } quasars at z \approx 1 and z \approx 2 ( V and I bands respectively ) , designed to explore the host galaxies .
We fit the images with simple axisymmetric galaxy models , including a point-source , in order to separate nuclear and host-galaxy emission .
We successfully model all of the host galaxies , with luminosities stable to within 0.3 mag .
Combining with our earlier NICMOS rest-frame optical study of the same sample , we provide the first rest-frame U - V colours for a sample of quasar host galaxies .
While the optical luminosities of their host galaxies indicate that they are drawn purely from the most massive ( \mathrel { \lower 2.58 pt \hbox { $ \buildrel { \textstyle > } \over { \scriptstyle \sim } $ } } L% ^ { \star } ) early-type galaxy population , their colours are systematically bluer than those of comparably massive galaxies at the same redshift .
The host galaxies of the radio-loud quasars ( RLQ ) in our sample are more luminous than their radio-quiet quasar ( RQQ ) counterparts at each epoch , but have indistinguishable colours , confirming that the RLQ ’ s are drawn from only the most massive galaxies ( 10 ^ { 11 } -10 ^ { 12 } M _ { \odot } even at z \approx 2 ) , while the RQQ ’ s are slightly less massive ( \sim 10 ^ { 11 } M _ { \odot } ) .
This is consistent with the well-known anti-correlation between radio-loudness and accretion rate .
Using simple stellar population “ frosting ” models we estimate mean star formation rates of \sim 350 M _ { \odot } yr ^ { -1 } for the RLQ ’ s and \sim 100 M _ { \odot } yr ^ { -1 } for the RQQ ’ s at z \approx 2 .
By z \approx 1 , these rates have fallen to \sim 150 M _ { \odot } yr ^ { -1 } for the RLQ ’ s and \sim 50 M _ { \odot } yr ^ { -1 } for the RQQ ’ s .
We conclude that while the host galaxies are extremely massive , they remain actively star-forming at , or close to , the epoch of the quasar .