5mm Fig. 4.1 shows the schematic cross section of two cells of a three-phase, n-channel CCD and the electron energy at the semiconductor-oxide interface for different applied clock voltages. The signal charge gathers underneath the gates with high positive voltage. The free charge is transferred from one gate to the next by three separate mechanisms: self-induced drift, thermal diffusion, and fringing field drift. Self-induced drift is a charge repulsion effect which is important for large signal-charge densities. Thermal diffusion is caused by the strong gradient of the signal-charge at the edge of the transferring electrodes. The fringing field is the electric field in the direction of charge flow caused by the different potentials at the clock electrodes.
Either electrons or holes can be used as charge carriers, but due to the higher mobility of electrons and better performance usually n-channel CCDs are used.
CCDs can be constructed as surface-channel devices (SCCD) or as buried-channel devices (BCCD). In SCCDs a potential well is formed at the semiconductor-oxide interface (see Fig. 4.2). The structure of BCCDs is similar to those of SCCDs except for an additional thin silicon layer whose doping type is the opposite to that of the substrate. It causes a potential well below the surface (see Fig. 4.3).
In BCCDs the transfer efficiency, which is the most important performance parameter of a CCD, is higher than in SCCDs and the noise is lower because there is no trapping by fast interface states. To alleviate this effect a ``fat-zero'' signal can be used in SCCDs which guarantees that the interface traps are always filled. This of course reduces the dynamic range of the signal charge. An important disadvantage of BCCDs is that the maximum signal charge is considerably lower and that the dark current is higher.