New Methods in Ultra-Wideband Array Design and Finite-Difference Time-Domain Modeling of Memristive Devices

Open Access
Gregory, Micah Dennis
Graduate Program:
Electrical Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
June 14, 2013
Committee Members:
  • Douglas Henry Werner, Dissertation Advisor/Co-Advisor
  • Douglas Henry Werner, Committee Chair/Co-Chair
  • Ram Mohan Narayanan, Committee Member
  • Douglas Edward Wolfe, Committee Member
  • Jack Brenizer Jr., Committee Member
  • Victor P Pasko, Committee Member
  • Ultra-Wideband Array
  • CMA-ES
  • Optimization
  • Aperiodic Arrays
  • Memristor
  • Reconfigurable Antennas
  • Finite-Difference Time-Domain
This dissertation covers two electromagnetics topics, the first is the design of ultra-wideband antenna array layouts. The second is the design of reconfigurable radio frequency devices with a newly discovered circuit element, the memristor. The two concepts are seemingly unrelated, however, reconfigurable devices are often used in antennas with multi-band abilities, sometimes switching between frequencies that can be octaves apart. In order to efficiently use these reconfigurable antennas in phased arrays they must be placed in a proper layout, the focus of the first topic. The primary focus of the first topic is the elimination of grating lobes and minimization of peak sidelobe levels in the radiation pattern of antenna arrays. The occurrence of grating lobes in array factors is very similar to aliasing in a digital system when the sampling frequency is below twice the maximum frequency content (i.e. the Nyquist frequency). These lobes appear in the array factor or radiation pattern of periodic arrays when the distance between elements becomes greater than one wavelength (for an unsteered array, and less for a steered phased array), resulting in radiation and reception in undesired directions. Strong mutual coupling negatively affects antenna performance when elements are placed less than a half-wavelength apart, limiting the closest that common antenna elements can be placed. These two phenomena generally limit the usable frequency bandwidth of conventional periodic array systems to about 2:1. As many emerging radio frequency systems that utilize multiple frequencies and ultra-wide bandwidths require high directivity, the need for capable array layout designs becomes apparent. The goal of this portion of research is the creation of design techniques which are capable of readily producing array layouts which yield no grating lobes and low sidelobe levels over vary large frequency bandwidths. The second focus involves a new type of electronic device called the memristor. Their existence was speculated by Leon Chua in the 1970s as the fourth basic circuit element which relates flux to charge. The result is a passive component with a charge or flux dependent resistance, allowing for design of interesting reconfigurable electromagnetic devices without active circuit elements such as transistors. It has received significant interest after a research team at Hewlett-Packard labs successfully fabricated a device using titanium dioxide and platinum which exhibits properties that can be modeled by a memristor. Most excitement resides around their potential for computer memory applications, however, their properties are also useful for other devices. Many frequency selective surface and antenna structures can have adjustable performance characteristics based on the values of embedded lumped resistor elements. Replacing the resistors with memristors allows reconfigurability without the use of significant controlling circuitry that is usually necessary with other reconfigurable devices such as active parts (transistors, varactors, etc.) or micro-electrical mechanical switches (MEMS). A specially tailored finite-difference time-domain (FD-TD) simulation code has been developed to analyze and design radio frequency structures with memristive elements. The code incorporates a memristor model with non-linear dopant drift, an advancement in accuracy versus the conventional linear dopant drift models. The time-domain nature of FD-TD permits capturing the non-linear and transient behavior of memristors, where quasi-static approximations could only be possible with a frequency domain code.