| Summary: | This thesis explores the physical origins of noise in semiconductors. A novel method of analysing the electron and hole carrier flow in the junction is developed, and the results are analysed with the objective of comparing the model against physical measurements made on actual semiconductor junctions. The current state of the art for commercial device fabrication is currently about 100nm device length, and new approaches to modelling device behaviour and noise mechanisms are required as the technology shrinks. In addition higher order transport models are needed due to smaller transistor sizes, and the need to account for current crowding effects in high current density situations. The various known forms of noise are reviewed, and prioritised in terms of their contribution to a pre-selected demonstration vehicle, in this case a BJT, and their contribution to the BJT's overall noise figure. The alternative methods of analysing noise are considered, and compared against the objective of reducing noise in this selected semiconductor structure. Many simulation techniques are available offering 1, 2 and pseudo three dimensional (21/2-D) approaches. The continuing trend to reduce the dimensions of the active devices requires more accurate models, and so more detailed physical correlation, especially at higher current densities. The models should be able to run as quickly and efficiently as possible with this increased complexity. The minority carrier based operation of the Bipolar Junction Transistor or BJT is selected, and this is used as an example in the model, with the intention of validating the model against a mature and very predictable technology. The model is also validated using well known and established simulation methods and then the model is applied to evaluate new structures. in order to propose a low noise BJT or LNBJT. The dominant noise in BJT devices is shown to be due to shot noise at mid to high frequencies. The noise mechanism known as shot noise. has been assumed in the past to be a fundamental noise form that limits the noise performance of bipolar structures. This work develops and demonstrates a method for modelling the carrier transport in theory of devices down to geometries of 15nm between the emitter and collector, using a true 3D simulation model called JAMES (Junction Atomistic Modelling of Extrinsic Semiconductors), based on a multi-carrier model, and able to model a large number of carriers in realistic simulation times. The thesis then goes on to develop the model for studying the nature of the individual carrier flow, and proposes a physically consistent explanation of electrical shot noise, and the carrier transport intrinsic in the BJT current flow. The work demonstrates the use of the model to redesign the bipolar junction transistor, and by modifying its doping structure and biasing, the spectral density of the shot noise Sv can be reduced by -26.4dB at frequencies approaching DC, and greater than -lO.75dB reduction at IGhz. The mechanism used to reduce the shot noise is to re-thermalise the minority carriers which have been grouped into "noise quanta", by introducing a semiconductor region to return the individual carriers to a classically chaotic state before they reach the collector. This new proposed structure is a dual base LNBJT capable of not only reducing the shot noise and total device noise, but also increasing the AC voltage gain, and increasing the transition frequency when it is used as a switch. A proposed new constant will be introduced, referred to as KSA, defined as the ratio of reduction in the spectral density as a result of adding a new structure. The work offers as one result, a design methodology to minimise the impact of noise in integrated and discrete BJT's. The noise attenuation mechanism proposed can also reduce other noise forms however, but to a smaller extent. The mechanism proposed here is effective in attenuating shot noise because the noise quanta given by Q = n.q is relatively small, where q is the carrier charge, and n is the number of charge carriers in the quanta. Low frequency noise forms can have a larger noise quanta, and so the attenuation of these forms is therefore smaller.
|
|---|
