ABSTRACT




	The main topic of this thesis is the modeling of dynamic systems by Cellular Automata (CA). CA can model a system by defining a set of microscopic rules in order to obtain an adequate macroscopic behavior. One of the main fields where this methodology has been applied (and which is other of the main topics of this thesis) is the modeling of Anisotropic Wet Etching (AWE): a chemical process which allows the creation of three-dimensional silicon microstructures. This feature has made AWE to become an important microfabrication technique. 
	AWE is used for the micromachining of Micro-Electro-Mechanical Systems (MEMS). MEMS consists of the integration of mechanical elements, sensors, actuators and electronics on a common silicon substrate through microfabrication techniques. MEMS have great influence in the industry as devices manufactured by this technology are used intensively in various fields such as automotive security systems, motion sensors in consumer electronics or injectors in printing systems. 
	AWE is a complex process and the final result depends largely on various process parameters like the etchant type, process temperature or etch time. The usage of a simulator before performing  experiments could result in a great reduction of design time and material utilization. 
 	Existing AWE simulators based on CA have several limitations: very long simulation times due to high computational requirements of the CA, the small set of existing calibrations and the inability to simulate new etchants such as TMAH+Triton. The resolution of these limitations is addressed in various chapters of the thesis. In detail, this thesis makes the following contributions: 

* Analysis of the modeling technique based on CA, using as a case of study the equalization of a gamma ray detector based on a continuous scintillator crystal. The implementation of CA on FPGAs for accelerating CA-based simulations is also evaluated.
* Acceleration of the latest models of CA used in AWE simulations by using graphics processing units (GPUs).
* Definition of a new calibration methodology for AWE-CA based on evolutionary algorithms and its application to a wide range of AWE-based etchants such as KOH, TMAH, KOH+IPA and TMAH+Triton.
* Redefining the AWE-CA model in order to eliminate inherent errors in current implementations, based in constant time steps.
 
	This thesis provides solutions to the issues pointed out by defining new methodologies and proving their correctness with tests and comparisons with experimental data.