We make a critical analysis of numerical results for and against phase separation and stripe formation in the $t-J$ model. We argue that the frustrated phase separation mechanism for stripe formation requires phase separation at too high a doping for it to be consistent with existing numerical studies of the $t-J$ model. We compare variational energies for various methods, and conclude that the most accurate calculations for large systems appear to be from the density-matrix renormalization group. These calculations imply that the ground state of the doped $t-J$ model is striped, not phase separated.

• Received 31 August 1999

DOI:https://doi.org/10.1103/PhysRevB.61.6320