Multi-Dimensional Numerical Simulation of the Charge Collection Process in Semiconductor Image Sensors.

Image sensors have been widely used in many imaging applications such as image capturing applications in cameras recently. In this thesis, Charge Coupled Device (CCD) and Charge Injection Device (CID) type image sensors have been investigated. In these devices, optical signals are converted to electrical signals first by a charge generation and collection process and then by a read-out process by an external circuit. Optimum device performance for a particular application depends on many geometrical and physical parameters. Achieving optimum device performance by fabricating the device with different sets of parameters is a very expensive method. An alternative method, numerical simulation of the device before fabricating the optimized device, has been emerged in recent years because it is a very cost-effective and time-saving method. The main objective of this thesis is to simulate the charge generation and collection process in the image sensor in one, two, and three dimensions. For this simulation, CID has been used as a vehicle. For the numerical discretization in 1-dimensional space, Scharfetter-Gummel (S-G) approach is used because this method handles the exponential type changes in the carrier concentrations. For discretization in time, Gear's method has been employed. Then the S-G method has been extended to two and three dimensions. In this simulation model, the time-dependent current -continuity equation is solved in multi dimensions. The electric field distribution in the device is assumed to be known, so Poisson's equation is not solved. Also, the Metal Oxide Semiconductor (MOS) theory is included in the model. By simulating several test problems, it has been shown that the model is consistent with the physical device operation of the device. The linearity of charge collection process has been investigated and the results in these test problems have been interpreted by explaining the underlying physics. Moreover, the simulation results have been compared to those from a commercial device simulation program (MEDICI) for two examples. It has been found that provided that same models, and grid structure are used in both programs, the results agree in 5-10% range. On the other hand, the thesis program runs up to 2.85 times faster than MEDICI.