Miniaturization is a simple yet very powerful concept. Applied to devices, it opens to plethora of new applications. Microelectromechanical systems (MEMS) are the result of scaling down the dimensions of devices to the micron range or the nano range (NEMS). This approach, leads to objects that not only are smaller, but often perform better than their bigger counterparts. It is not then a surprise that MEMS have been object of intense research in several different fields. In particular, the MEMS communications market is consolidating as we observe an increasing trend that points at hyper-connectivity of people and objects. RF MEMS-based front-end have been proposed in order to create communication systems that have low power consumption, easy integration with CMOS circuits and small form factor. Piezoelectric MEMS resonators are the core elements of the RF front-end. Among other technologies, contour mode resonators (CMRs) allow to chose the operating frequency via design. In this thesis, techniques to improve the performance of CMRs are studied. Two resonator parameters, the quality factor (Q) and the electromechanical coupling (kt^2), are markers to assess the resonator performance and to compare to state-of-the-art. As a matter of fact, a simultaneous enhancement of both has a great impact on the performance of the filters and oscillators built using them. Since Q is design-depended whereas kt^2 is mostly material dependent, a single approach often fails to enhance both. The thesis is based on two milestones. The first is to combine design optimization techniques with a materials with enhanced piezoelectric response compared to state of the art (i.e. aluminum nitride, AlN) to foster a simultaneous Q-kt^2 enhancement. The second is to test if a Q-kt^2 trade-off exists and how much designers can circumvent it combining selected techniques for each of these two parameters. The first part of the work focuses on Q. A modified fabrication technique is developed to: (i) take care of the Q instability with respect to the release area dimension and simultaneously (ii) create an acoustic reflector. This resulted in a 5X improvement in Q stability and up to 70 % Q recovery. The second part focuses on improving kt^2 using aluminum scandium nitride (AlScN), with 17 % Sc concentration. Our results show that anchor losses are still a dominant dissipation mechanism and that using AlScN we find an improvement in CMRs kt^2 higher than 2X. The highest FoM=70 (Q = 1658, kt^2 = 4.1 %) is obtained for a resonator operating at 390MHz, showing a motional resistance of 64 Ohm. The extracted relevant piezoelectric coefficient for AlScN CMRs, d31, is -3,89 pm/V (2.25X improvement compared to AlN). Finally, 2-port CMRs, are used to build oscillators. A new measurement technique is developed: the 2-port CMRs are directly contacted via RF probes, allowing the measurement of the phase noise (PN) for a large number of resonators. The best PN for the AlScN-based oscillator, measured at the highest Q =1413, is -100 dBc/Hz at 1 kHz offset frequency. Interestingly, when compared to an AlN-based oscillator with nearly identical Q, the AlScN PN performance is better than the AlN PN performance by 10 dBc/Hz at 1 kHz offset frequency. Since the only thing changing in the close loop setup is the resonator, this finding indicates that, regardless of the resonator Q, there may be a difference in the intrinsic resonator flicker noise due to the piezoelectric layer used.
EPFL2019
