A Merentid a p p d for eelf-optimizing di&sion Monte Carlo calculation was propoeed in this paper, which is a new algorithm d i n i n g three techniques such as optimbhg, ditbion and correlation samphg. This method can be used to directly compute the energy difTerential between two system in the diffusionprocess, d a n g the statistid em^^ of calculation be reduced to Order d 10' hartree, and recover about more than 80% of the codation ene%y. We employed this approach to set up a potential eneqg surface of a molecule, used a "rigid move" model, and utilized Jacobi tran$ormation to make energy calculation for two con6gumtions of a molecule having good positive correlation. So, an ~ccurate energy d 8 d could be obtained, and the potential energy slllface with good qual~ty can be depicted. In calculation, a technique called "post-equilibrium remaining sample" was set up h d y , which can save about 50% of computation expense. 'Ihis novd algorithm was used to study the potenid enew c w e ofthe pund state x'T of molecule G, and it can also be applied to study other related fields such as molecular s m and the energy variation in chemical reactions. K e y w o ~I ~ surface Merential approach, quantum Monte Carlo method, correlation samphng, potential energy
Iutroduction
In p~vious paper, the self-optimizing diffusion Monte Carlo (SODMC) method was suggested which can recover about more than 80% of the correlation energy for the ground and excited states of atoms and molecules with a statistical e m r of 104 hartree. AS a matter of fact, energy differenti& between two systems such as energy change in chemical reaction, potential energy sdace ( P a ) , and so on, me often more i m p o ~t in chemistry. lhese energy differentials themselves often range in size hrn lo-* to loT3 hartree. It is thus necessary to develop a new algorithm on the basis of SODMC , so that one can use