Simulation Of Nonlinear Optical Properties With Sum-Over-States (SOS) Model On GPU

In last a few years, people have developed the graphics processor(Graphics Processing Unit, GPU) rapidly. Because GPU has powerfulcomputation capability and large memory bandwidth, the performance ofthe CPU is far behind the GPU. Cooperated with CPU, GPU canefficiently complete large-scale tasks with intensive data. GPU iscost-effective as well.In these days, with the size of the material becoming larger, theamount of the computational time of the nonlinear optical propertiesincreases rapidly. With high prices and not excellent performance, theexisting CPU can not efficiently complete these calculations, thus theadvantages of GPU are obvious. Therefore, with the help of GPU usingCUDA (Computer Unifie Device Architecture), we rewrote two programsto calculate the polarizability and the first h yperpolarizability. CUDA is aplatform designed for GPU computation.The nonlinear optical materials are playing an important role in thefield of optoelectronics and electronics, and search of the nonlinearoptical material with excellent performance has been one of the hottestresearch topics. In this thesis, we designed six BN/C-doped single-wallednanotubes, and then we used the density functional theory based methodB3LYP with6-31G(d,p) basis set to optimize the structures of these sixnanotubes, and then used Sum-Over-States (SOS) model to calculate thefirst hyperpolarizabilities of these six nanotubes. We find that the firsthyperpolarizability of the single-walled boron nitride nanotube isrelatively low, but with the introduction of conjugated car bon rings,hybrid nanotubes show a higher first hyperpolarizabilit y, and the firsthyperpolarizabilit y is closely related to the hybridization of carbon rings.With the conjugated carbocyclic rings of the hybrid nanotubes saturatedby hydrogen atoms, the first hyperpolarizabilities of the nanotubesbecome low, down to the same magnitude of the single-walled boronnitride nanotube.