We present sensitive 850 \mu m SCUBA photometry of a statistically-complete sample of six of the most luminous ( M _ { B } < -27.5 ; { { { { \nu L _ { \nu } { ( 1450 \thinspace \AA ) } \mathrel { \mathchoice { \vbox { % \offinterlineskip \halign { \cr } $ \displaystyle > $ \cr$ \displaystyle \sim$ } } } { \vbox { % \offinterlineskip \halign { \cr } $ \textstyle > $ \cr$ \textstyle \sim$ } } } { \vbox { % \offinterlineskip \halign { \cr } $ \scriptstyle > $ \cr$ \scriptstyle \sim$ } } } { \vbox { % \offinterlineskip \halign { \cr } $ \scriptscriptstyle > $ \cr$ \scriptscriptstyle \sim$ } % } } } 10 ^ { 13 } \hbox { $ \thinspace L _ { \odot } $ } ) , z > 4 radio-quiet quasars , reaching noise levels ( 1 \sigma \sim 1.5 { \thinspace mJy } ) comparable with the deep blank sky surveys .
These observations probe the rest frame far infrared region ( \sim 150 \mu m ) , at luminosity levels for thermal sources comparable with those that IRAS permitted for low redshift quasars .
One quasar ( BR2237 - 0607 ; z = 4.55 ) is detected at 850 \mu m with a flux of 5.0 \pm 1.1mJy ( 4.5 \sigma ) , whilst a second ( BR0019 - 1522 ; z=4.52 ) has a detection at the 2 \sigma level .
When combined with our previous millimetre measurements of z > 4 quasars , we find that there is a large range ( 5–10 ) in far infrared ( FIR ) luminosity ( L _ { FIR } ) at fixed UV luminosity , and that the typical quasar has a L _ { FIR } and mass of cool ( 50 K ) dust similar to that of the archetyepal low redshift ( z=0.018 ) ultraluminous IRAS galaxy ( ULIRG ) Arp220 ( L _ { FIR } \sim 5 \times 10 ^ { 12 } \hbox { $ \thinspace L _ { \odot } $ } ; M _ { d } { ( cool ) } \sim 10 ^ { 8 } \hbox { $ \thinspace M _ { \odot } $ } ) .
If one assumes a fiducial FIR luminosity of 5 \times 10 ^ { 12 } \hbox { $ \thinspace L _ { \odot } $ } for for all quasars with M _ { B } < -23 , we find that { { { { \mathrel { \mathchoice { \vbox { \offinterlineskip \halign { \cr } $ \displaystyle > $ \cr$% \displaystyle \sim$ } } } { \vbox { \offinterlineskip \halign { \cr } $ \textstyle > $ \cr$% \textstyle \sim$ } } } { \vbox { \offinterlineskip \halign { \cr } $ \scriptstyle > $ \cr$% \scriptstyle \sim$ } } } { \vbox { \offinterlineskip \halign { \cr } $ \scriptscriptstyle > $% \cr$ \scriptscriptstyle \sim$ } } } } 15 per cent of the sources in the SCUBA deep surveys could be classical broad-lined radio-quiet AGN .
Thus if one considers the observed ratio of Seyfert II to Seyfert I galaxies at low redshift and any contribution from totally optically obscured AGN , a significant fraction of the SCUBA source population will harbour AGN and hence the inferred star formation rates from submm fluxes may be overestimated if the active nuclei are bolometrically dominant or the IMF is top heavy .