Phase diagram of the J−Jd Heisenberg model on the maple leaf lattice: Neural networks and density matrix renormalization group

We microscopically analyze the nearest-neighbor Heisenberg model on the maple leaf lattice through neural quantum state (NQS) and infinite density matrix renormalization group (iDMRG) methods. Embarking to parameter regimes beyond the exact dimer singlet ground state with a dimer bond spin exchange coupling J₃ varied against the exchange strength J of all other bonds, the iDMRG (NQS) method finds a dimer state paramagnetic phase for J₃/J>1. 464 (J₃/J>1. 39) and a canted 120^ magnetic order for J₃/J<1. 419 (J₃/J<1. 23). Assessing training convergence inaccuracies of the NQS method and the influence of finite cylindric circumference in the iDMRG method, we discuss the possible existence of an intermediate phase between the magnet and the dimer paramagnet.