by Xavier Rottenberg, Ph.D.
Research Engineer, IMEC
Much of the future growth in the MEMS market is expected to come from products that are still in the early stages of development, or are yet to be invented. There is a considerable potential for new products in which innovative MEMS are integrated with highly miniaturized electronics. In small packages that miniaturize existing larger tools and extend their use and functionality, or that bring a totally new functionality that cannot be implemented otherwise. Some of the many domains where MEMS will become prevalent are sensing, localization, and visualization applications for personal consumer electronics, optical switching and processing chips, or sensors and actuators for medical and wellness applications.
In this article, we want to look at some design and fabrication solutions that may be the key to produce these future MEMS, with better and more predictable electrical and mechanical characteristics than we can today.
We’ll first present an innovative SiGe (silicon-germanium) technology for MEMS fabrication. This technology has been implemented in a versatile platform that has allowed us to make and test high-quality MEMS with interesting properties. The technology also that may serve as basis for reliable modeling and simulation.
And second, we’ll look at some of the solutions we’ve developed in designing MEMS, and in particular MEMS resonators. Resonating structures may be used in a wide variety of MEMS applications, such as timing devices of RF switches, but also in gas sensors or ultrasound transducers. This article will focus on some of the challenges associated with developing high-quality resonators and on the solutions that we propose. Examples are the use of inclusions and holes in the SiGe to engineer a range of acoustic metamaterials so that we can tune and achieve optimal performance characteristics or a method to enforce the frequency stability over a wide temperature range.
SiGe platform – basis for monolithic design and fabrication
To fabricate our designs, in prototypes and low-volume series, we have set up two MEMS platforms in IMEC’s 200mm fab. One for conventional silicon-on-insulator (SOI) processing and one for SiGe technology. The latter is CMOS-compatible and suited for monolithic integration of MEMS on top of CMOS.
The SiGe-MEMS technology that we have developed and deployed is based on a MEMS-last approach, where the MEMS are processed after and on top of the CMOS circuits. In this scheme, the MEMS processing and characteristics are made largely independent of the CMOS technology they are combined with. Further, integrating MEMS devices with their driving and readout electronics on the same die leads to a better performance compared to other integration schemes: there is a better signal-to-noise ratio through a reduced interconnect parasitic resistance and capacitance, a smaller die size and package, and lower power consumption. A monolithic approach is especially suited for those applications where performance and miniaturization are of key importance, or when many (think of millions) interconnections are needed between the MEMS (array) and the CMOS.
The SiGe-MEMS platform is versatile. It consists of standard and optional modules that can be processed at a low temperature budget above standard CMOS, with many possibilities to tune and optimize the modules. The standard modules provide e.g. a CMOS protection layer, MEMS via and poly-SiGe electrode, an anchor and poly-SiGe structural layer, and thin-film poly-SiGe packaging. Optional modules (e.g. optical, piezoresistive, probes) can be added depending on the application.
Figure 1: Cross-section of a MEMS device created with a SiGe-MEMS platform.
The platform’s flexible and modular approach allows application-specific tuning and optimization. An example is the thickness of the MEMS structural layer, which can vary between 300nm and 4µm. A 300nm-thick layer allows making optical MEMS, such as micromirrors and gratings. For such devices, the process is extended to add various coatings with specific reflective properties. A 4µm structural layer is used, for example, to create inertial sensors or actuators. Other possible applications of the technology are microphones, µspeakers, µsensors, probe-based memories and micropower generation.
On this platform, we’ve for example fabricated accelerometers, starting from both in-plane and out-of-plane low-g designs. After fabrication, measurements of the out-of-plane accelerometer show that this device can sense the gravitation projection to the main sensing axis with an average sensitivity of 0.5mV/°. This sensitivity is comparable to state-of-the-art, but with a greatly improved noise performance compared to state-of-the-art accelerometers. The accelerometers have been built with a 4µm thick SiGe structural layer to obtain an improved capacitive readout of the in-plane devices.
Figure 2: In-plane (left) and out-of-plane (right) SiGe accelerometer.
Another type of applications for which this platform is especially well suited are large arrays of MEMS structures, such as micromirror arrays for vision systems .
Resonators – designing metamaterials with exactly the right characteristics
MEMS resonators are recognized as key components for future sensing, wireless and communications applications. At imec, we’ve focused our R&D work on bulk acoustic-wave (BAW) resonators. These have excellent characteristics, e.g., a high Q-factor, high resonance frequency (up to hundreds of MHz), and low sensitivity to residual stress. In addition, the SiGe platform, with its submicron features, allows us to make resonators with efficient and low-voltage electrostatic excitation and sensing.
Figure 3: SEM picture of typical SiGe bar-type BAW resonator.
In a BAW resonator, the resonance occurs when an acoustic wave, trapped in the bulk of a parallelepipedic slab of material, bounces in phase on the opposite faces of the material. The resonance frequency depends on the material properties and the dimension of the bar. So in a given technology, with fixed material properties, the design freedom is strongly limited; to design a resonator with a specific frequency, the designer can only play with the dimensions of the bar.
However, it is also possible to modify the effective material properties available to produce bar resonators by design. This can be achieved by adding holes, or macro-pores, to the bulk material. The resulting macro-porous materials with effective acoustic velocities, Young’s modulus, densities and Poison’s ratio that are different from those of the original bulk material. The designer has a continuous family of designable effective materials at his disposal for enabling typical devices with atypical characteristics.
To efficiently manage this novel degree of freedom in the design, we have to be able and model the behavior of such meta-materials. We therefore studied therefore the relation between various patterns and sizes of pores and the ensuing characteristics – with the goal to come up with the technique to design meta-materials with predictable characteristics and use these to design and fabricate non-homogeneous bar resonators, with minor impact on the Q-factor compared to homogeneous bar resonators .
In a first phase we studied square lattices of uniformly distributed square holes. Materials with such hole distribution could be used in resonator applications that make use of longitudinal resonance modes. For these patters, we were able to come up with a good agreement between the theoretical (simulated) values and the effective material properties (especially the acoustic longitudinal velocity) as extracted from measurements. Next, we are also experimenting with non-uniform holes distributions, we work to include a tensorial description of the material properties that we can model to account for their anisotropy, and we look to model the impact of these holes distribution on the achievable Q-factor, its improvement and/or degradation.
Figure 4: Composite SEM picture of various SiGe BAW resonators implementing designed metamaterials.
Resonators for timing devices – designing frequency stability
Silicon-based resonators provide an attractive alternative to replace conventional frequency control and timing devices that are based on high quality factor (Q) resonators such as quartz crystal, piezoelectric ceramic and surface acoustic wave. specially using monolithic integration and thin-film capping, it is possible to make high-performance, low-cost and small-footprint timing devices.
One important challenge for MEMS resonator-based timing devices is the stability of the resonance frequency over a wide temperature range. MEMS devices have a relatively large frequency drift over temperature (in the order of tens of ppm/°C) in comparison to quartz. Especially for wireless applications, the requirements on temperature stability are very demanding: only around 1ppm variance is allowed in the temperature range of -40 to +85 degrees Celsius.
So what is needed is a frequency-correcting mechanism. One solution is to have a compensation mechanism steered by an ambient temperature sensor. A solution is to embed the MEMS oscillator in a fractional PLL, whose fractional frequency control word is steered by a temperature sensor. In such a system, the PLL provides a clock output, stable over temperature. However the output is polluted with fractional-PLL noise. Moreover, the feed-forward temperature compensation requires multipoint calibration for ppm-level stability over a wide temperature range, and requires a high-resolution, sufficiently linear temperature sensor.
The solution that we have opted for is an 'oven-controlled MEMS oscillator (OCMO). Such a device keeps the resonator at a fixed temperature by placing it in a micro-oven. This target temperature should be above the operating range of the MEMS, so all effects on the resonance frequency of changes in the ambient temperature will be minimized. Such a design has the potential to meet the stringent requirements of quartz-based temperature-compensated crystal oscillators for wireless applications.
We have designed a bar-type BAW resonator that we heat to 95°C . Key to build such an application is the multi-physics optimization of resonator body and tethers that interact to define not only the Q-factor, resonance frequency and motional impedance of the MEMS device but also for example its thermal behavior. The design uses T-type supports that allow for Joule heating through the supports. At the same time, these optimize the design in terms of anchor losses to the substrate, therefore maximizing the Q-factor. The T-type design is rigid in the direction of actuation, allowing the relatively high bias voltages needed to lower the motional resistance and ease the design of an oscillator system. In addition, we showed that the lower thermal conductivity of poly-SiGe compared to SOI provides a significant advantage in the amount of power needed to heat the resonator for similar devices manufactured in both technologies.
MEMS slowly find their way into industrial and consumer applications, a path that is destined to accelerate. Key for this trend are the growing strong understanding of MEMS devices and their failure modes as well as the progressive maturing of MEMS technologies in industrial environments and MEMS-oriented multi-physics simulation packages. This has allowed the successful developments and commercialization of MEMS tilting mirrors for DLP applications, MEMS accelerometers and gyroscopes for navigation purposes, MEMS resonators for timing devices replacing quartz and others.
In the near future, new MEMS devices and systems will emerge that no longer result from a standard translation of macro devices to the micro world. In addition, the scaling from MEMS to N(ano)EMS will follow an accelerating path with deep implications on the nano-engineered materials available for the designers. For both these trends, key will be a multi-level multi-physics CMOS-MEM system co-design, and on the other hand, a technology-aware design of these devices and systems.
It is in this context that our designers contribute to developing complex new generation microsystems, (e.g. inertial sensors, compass, timing devices, light shaping systems, etc.) by helping to steer the developments of proprietary technologies and leveraging their full potential.
 Murali Jayapala: MEMS based micromirror arrays for vision systems and display applications, in MEMS Investor Journal (http://www.memsinvestorjournal.com/2011/04/mems-based-micromirror-arrays-for-vision-systems-and-display-applications.html).
 Xavier Rottenberg et al.: Acoustic meta-materials in MEMS bar resonators - Applied Physics A: Materials Science and Processing, Vol 103, Issue 3, pp 869-875.
 R. Jansen et al.: A CMOS-compatible 24MHz poly-SiGe MEMS oscillator with low-power heating for frequency stabilization over temperature, IEEE International Frequency Control Symposium, May 2011.
Xavier Rottenberg is Team Leader at IMEC. He was born in Brussels, Belgium, in April 1976. He received the M.S. degree in Physics Engineering and a supplementary degree in Theoretical Physics in 1998 and 1999 from “Université Libre de Bruxelles”, Belgium. He received further his Ph.D. degree in Electrical Engineering in 2008 from K.U.Leuven, Belgium. He worked one year at the Royal Meteorological Institute of Belgium in the field of remote sensing from space. He joined the Interuniversity Micro Electronics Center (IMEC) in 2000 where he currently leads the research activities in microsystems design and modeling.
Copyright 2012 MEMS Investor Journal, Inc.