This dissertation details efforts to achieve high-frequency, low-power, stable microelectromechanical systems (MEMS) resonator-based oscillators. This endeavor quickly identifies the key resonator performance enablers to be high frequency, high quality factor (Q), and low motional resistance (Rx). Towards this end, we explore two micromechanical resonator transduction mechanisms: capacitive-gap and piezoelectric. The high Qs of capacitive-gap MEMS resonators have already enabled stable oscillators, with commercial examples of them now permeating the industry. They, however, remain at conventional frequencies in the tens of MHz, since their motional resistances increase with frequency and thereby lead to high oscillator power consumption. On the other hand, piezoelectric MEMS resonators can achieve low motional resistance even at GHz frequencies, and GHz oscillators referenced to such resonators have already been demonstrated with reasonable power consumption. However, the insufficient Q of piezoelectric MEMS resonators leads to mediocre oscillator phase noise and poorer (than capacitive-gap) long-term stability. This dissertation attempts to achieve low-power high-frequency stable oscillators using both types of MEMS resonators by lowering the Rx of high-frequency capacitive-gap resonators with the use of large-scale array-composites (Chapter 2), increasing the Q of both types of resonators by UV-Ozone treatment (Chapter 3), and increasing the Q of piezoelectric resonators by converting them into capacitive-piezoelectric (Cap-Piezo) resonators (Chapter 4).
Following Chapter 1 which introduces the background of oscillators, Chapter 2 demonstrates a Pierce oscillator referenced to a capacitive-gap 199.2-MHz polysilicon two-by-five radial-contour mode 13.4-um-radius disk array-composite resonator with 36.1-nm transducing gap and achieved phase noise marks of -104.7dBc/Hz at 1-kHz offset and -149.6dBc/Hz far from the carrier. The array-composite resonator employs electrode-less buffer disks at the end of the array to avoid electrode-resonator contact and ensure proper device functions. Extensive FEA suggests large-scale differential array-composite resonator dampens spurious modes. A dual-mode 200-MHz oscillator demonstrates the use of the spurious mode of the 13.6-nm gap array-composite reference resonator.
Pursuant to recovering contaminated MEMS resonators and also further improving resonator Q, Chapter 3 presents an ultraviolet (UV)-ozone treatment process capable of atomic-scale surface cleaning that has enabled the highest measured room temperature Q and frequency-Q product (f-Q) for on-chip polysilicon acoustic resonators to date while also demonstrating an ability to restore catastrophically contaminated devices to operational performance specs.
Finally, Chapter 4 presents the first demonstration of a closed-loop oscillator referenced to a 20.4-um-radius high-Q radial-contour mode Aluminum Nitride (AlN) Cap-Piezo resonator that posts a measured phase noise of -87 dBc/Hz at 1-kHz offset from its 167.3-MHz carrier while consuming only 173 uW. The key enabler of this demonstration is the improved fabrication process which improves the yield of the Cap-Piezo resonator. The fabricated resonators achieved a Q of 4,568 at 311MHz and 4,536 at 167MHz. The Q of these resonators is limited by the stress-gradient-induced resonator-electrode contact. The Q may further increase by optimizing the AlN sputtering to minimize its vertical stress gradient. AlN Cap-Piezo whispering-gallery-mode (WGM) disk resonators are fabricated and tested, but the disk-to-electrode contact lowers the Q and prevents the direct determination of the intrinsic Q of AlN.
