Fast Simulation of FETS

Mercury is an Atlas module optimized for the fast simulation of FETs. Mercury is physics-based and so can be used for the predictive simulation of devices. The short simulation times allow Mercury to be used to analyse trends in the design of FETs and to investigate manufacturing yields.

Mercury is a Quasi-2D FET simulator. A Quasi-2D simulator is two 1D solvers working in conjuntion. The first solver calculates Poisson’s equation perpendicular to the surface of the device over a range of surface conditions. This characterises a 1D channel from the source to the drain. The second solver calculates the transport equations along this channel to generate the terminal characteristics of the device.

Mercury is optimized for the simulation of epitaxial FETs, that is FETs with a constant physical structure in the x-direction. As Mercury can simulate the large signal behavior of these devices the FET being simulated is always embedded in an external circuit. In the initial part of a Mercury simulation, Poisson’s equation is solved in the y-direction (perpendicular to the top surface of the device) for a range of surface conditions. This data is used to generate a look-up table that characterize the channel, between the source and drain contacts, where the carriers will flow. In this figure the electron density as a function of depth is calculated for a range of gate voltages.
Once the Poisson look-up table has been generated the calculation of the carrier transport from the source to the drain becomes a 1D problem, significantly reducing the time necessary for a simulation. Mercury can quickly calculate the DC-IV curves or the small-signal AC behavior of the FET being investigated.
In addition to the small-signal AC simulation, Mercury can calculate the small-signal noise generated by the device. Mercury can use the Harmonic Balance method to simulate the large signal behavior of a FET. This allows the investigation of gain compression as a function of input power.


Trend and Yield Analysis

The speed of Mercury means it is ideal for investigating trends. This figure shows the transconductance as a function of gate bias (at Vd=0.6V) as the doping in the device is changed between 4e16 cm-3 to 1e17 cm-3. Mercury can also be used to investigate the expected yield of a device taking into account variations in the manufacturing of the device. This figure shows the variation in the drain current and the transconductance as the doping, recess depth, and gate length are randomly varied from their design values.

Rev 110113_03