Massive　Multi-input　Multi-output　(MIMO)　technique　shows　great　potentials　in　improving　energy　and　spectral　efficiency　while　suffers　high　costs　of　the　requirement　for　large　amount　of　radio　frequency　(RF)　chains.　In　this　paper,　we　develop　an　l(1/2)-regularity　based　downlink　transmit　antenna　selection　scheme　for　massive　MIMO　systems　with　limited　RF　chains.　With　the　objective　to　minimize　the　transmission　power　at　the　given　number　of　RF-chains　and　transmission　quality,　we　formulate　and　decompose　the　original　l(1/2)-norm　optimization　problem　into　two　separated　problems　with　l(1/2)-sparsity　and　signal-interference-to-noise-ratio　(SINR)　requirements,　respectively.　An　iterative　algorithm　based　on　coordinate　descent　and　feasible　set　projection　are　then　developed.　This　scheme　provides　an　efficient　way　to　evaluate　the　minimum　required　RF　chains　for　the　deployment　of　a　massive　MIMO　system.　Simulation　results　show　that　the　proposed　scheme　achieves　higher　energy　efficiency　compared　with　the　l(p)(p＞1/2)-norm　based　antenna　selection　methods,　especially　for　large　numbers　of　correlated　antennas.