Context : Free-floating substellar candidates with estimated theoretical masses of as low as \sim 5 Jupiter masses have been found in the \sim 3 Myr old $ σ $ ~Orionis open cluster .
As the overlap with the planetary mass domain increases , the question of how these objects form becomes important .
The determination of their number density and whether a mass cut-off limit exists is crucial to understanding their formation .

Aims : We propose to search for objects of yet lower masses in the cluster and determine the shape of the mass function at low mass .

Methods : Using new- and ( re-analysed ) published IZJHK _ { s } [ 3.6 ] - [ 8.0 ] -band data of an area of 840 arcmin ^ { 2 } , we performed a search for LT-type cluster member candidates in the magnitude range J = 19.5 –21.5 mag , based on their expected magnitudes and colours .

Results : Besides recovering the T type object S Ori~70 and two other known objects , we find three new cluster member candidates , S Ori 72–74 , with J \approx 21 mag and within 12 arcmin of the cluster centre .
They have theoretical masses of 4 _ { -2 } ^ { +3 } M _ { Jup } and are among the least massive free-floating objects detected by direct imaging outside the Solar System .
The photometry in archival Spitzer [ 3.6 ] – [ 5.8 ] -band images infers that S Ori~72 is an L/T transition candidate and S Ori~73 a T-type candidate , following the expected cluster sequence in the mid-infrared .
Finally , the L-type candidate S Ori~74 with lower quality photometry is located at 11.8 arcsec ( \sim 4250 AU ) of a stellar member of \sigma Orionis and could be a companion .
After contaminant correction in the area complete to J = 21.1 mag , we estimate that there remain between zero and two cluster members in the mass interval 6–4 M _ { Jup } .

Conclusions : We present S Ori 73 , a new candidate T type and candidate \sigma Orionis member of a few Jupiter masses .
Our result suggests a possible turnover in the substellar mass spectrum below \sim 6 Jupiter masses , which could be investigated further by wider and deeper photometric surveys .