Page 48 - Sami Franssila Introduction to Microfabrication

3

Simulation of Microfabrication Processes

Microfabrication processes consist of tens or hundreds diffusion, implantation and oxidation, can be analysed
of steps that take weeks or months to complete, and by solving the relevant diffusion equations.
therefore the learning cycles can easily become too Etching, polishing and deposition produce topogra-
long. Simulation is one way of shortening the learning phy on a wafer. This build-up of topography is difficult
cycles. Simulation accuracy is strongly dependent on to simulate because it involves multiphysics and chem-
the details of the process to be simulated, and even a istry – plasmas, fluid dynamics and surface chemical
simple simulator can be extremely valuable if it saves reactions. Film deposition simulators depend on atom
enough experimentation time and effort. Simulators can arrival angles that are not physical constants like dif-
provide meaningful trend data and comparisons between fusivities but are parameters sensitive to experimental
different process options, even though the accuracy conditions. Etching reactions are complex interactions
might be less than perfect. Simulators can be used to between the chemical contributions (spontaneous etch-
explore possibilities and narrow down options before ing, free energy considerations) and physical processes
the experimental work is begun. Simulation can provide (e.g., ion bombardment enhanced desorption). Topogra-
information that is not experimentally available or is phy process simulators are usually semiempirical: some
difficult to measure. Because there is no dopant profiling important model parameters are extracted from experi-
ments without fundamental physical validation.
method with sub-10 nm resolution in both vertical and
Even though simulation is fast, simulator building is
lateral directions, simulation is the de facto method for
a two-dimensional dopant distribution analysis. slow and tedious. It is not possible to build simulators
There are two breeds of process simulators: integrated for all possible new materials, processes and devices,
because the calibration data needs to be available,
packages that can be used to simulate the whole fabrica-
and it is readily available only for those materials,
tion process with many different steps in sequence and processes and devices that are widely studied and used.
dedicated simulators for specific process steps. Dedi-
In this sense, the predictive power of process simulation
cated simulators are available for almost all processes,
remains poor.
ranging from ion-implantation damage production to
lithography defect modelling, to crystal structure predic-
tion of deposited films. Dedicated simulators are more
3.1 TYPES OF SIMULATION
detailed, more accurate and more computation inten-
sive. A basic principles diffusion simulator would start Process simulation, device simulation and circuit simu-
with lattice parameters, interatomic potentials, vacancy lation together are termed TCAD, for technology CAD
production and annihilation rates and atom-defect inter- (Figure 3.1), in contrast to the more established ECAD,
actions, and provide diffusion profiles as the output. electronic simulations, which involve logic and sys-
Integrated packages use simpler models, for instance, tems simulations. Process simulation deals with physical
macroscopic phenomenological diffusion models based structures such as atoms and their distributions, device
on Fick’s equations, but they offer seamless stitching simulation deals with currents and potentials in devices,
of different process steps into whole processes. Bulk and circuit simulation is used to study larger circuit
silicon process steps, that is, high-temperature steps blocks. The dopant concentrations produced by a process
that affect dopant distribution inside silicon, epitaxy, simulator are used as an input for the device simulator,

Introduction to Microfabrication Sami Franssila
 2004 John Wiley & Sons, Ltd ISBNs: 0-470-85105-8 (HB); 0-470-85106-6 (PB)