We report on long-term stability experiments on a novel MEMS radio frequency (RF) resonator fabricated in Aluminum
Nitride technology. The AlN fabrication process allows for the realization of resonators, filters, and resonant sensors
operating over the frequency range from 500 kHz to in excess of 10 GHz using CMOS compatible materials. The 100
MHz resonators used in these experiments were a ring design with 140-micron outer diameter and 100-micron inner
diameter. Electrodes on the top and bottom of this AlN ring enable measurement of resonance. Wafer sections were
stored in air and vacuum and tested daily. We observed a steady degradation in the resonant frequency (600 ppm over
the 800 hours) for the devices stored in a vacuum. Small degradation was observed in the air experiment (50 ppm over
1200 hours). Failure analysis using secondary emission microscopy (SEM) revealed no differences between control
devices and devices on test. However, subsequent investigation of blank wafer sections by Time-of-Flight secondary ion
mass spectrometry (ToF-SIMS) found small levels of silicone surface contamination from vacuum chamber exposure.
This contamination added enough mass to shift the resonant frequency. These experiments demonstrate the need for
clean environments for future wafer-level testing and also packaging for these small-mass resonators.
There is an increasing demand to build highly sensitive, low-G, microscale acceleration sensors with the ability to sense
accelerations in the nano-G (10-8 m/s2) regime. To achieve such sensitivities, these sensors require compliant mechanical
springs attached to large masses. The high sensitivities and the difficulty in integrating robust mechanical stops into
these designs make these parts inherently weak, lacking the robustness to survive even the low level accelerations
encountered in standard handling, from release processing, where supporting interlayers present during fabrication are
etched away, through packaging. Thus, the process of transforming a MEMS-based acceleration sensor from an
unreleased state to a protected functional state poses significant challenges. We summarize prior experiences with
packaging such devices and report on recent work in packaging and protecting a highly sensitive acceleration sensor that
optically senses displacement through the use of sub-wavelength nanogratings. We find that successful implementation
of such sensors requires starting with a clean and robust MEMS design, performing careful and controlled release
processing, and designing and executing a robust handling and packaging solution that keeps a fragile MEMS device
protected at all times.
Long-term reliability testing of Micro-Electro-Mechanical Systems (MEMS) is important to the acceptance of these devices for critical and high-impact applications. In order to make predictions on aging mechanisms, these validation experiments must be performed in controlled environments. Additionally, because the aging acceleration factors are not understood, the experiments can last for months. This paper describes the design and implementation of a long-term MEMS reliability test bed for accelerated life testing. The system is comprised of a small environmental chamber mounted on an electrodynamic shaker with a laser Doppler vibrometer (LDV) and digital camera for data collection. The humidity and temperature controlled chamber has capacity for 16 MEMS components in a 4x4 array. The shaker is used to dynamically excite the devices using broadband noise, chirp or any other programmed signal via the control software. Driving amplitudes can be varied to maintain the actuation of the test units at the desired level. The actuation is monitored optically via the LDV which can report the displacement or velocity information of the surface. A springmass accelerated aging experiment was started using a controlled environment of 5000 ppmv humidity (roughly 13% at room temperature), temperature of 29 °C, and ±80μm maximum displacement of the mass. During the first phase of the experiment, the resonant frequency was measured every 2 hours. From 114.5 to 450 hours under stress, measurements were taken every 12 hours and after that every 24 hours. Resonant frequency tracking indicates no changes in the structures for 4200 hours of testing.
In the fabrication of MEMS devices, what has come to be known as "release stiction" occurs when the device is removed from the liquid phase into the ambient air. One widely used method for dealing with stiction is to deposit a hydrophobic coating on the surface of the device before it is removed from the liquid phase. This method can produce coatings with inconsistent morphology and device yield. This is to be compared with a new coating deposition scheme developed at Sandia National Labs, termed VSAMS (vapor-deposited self-assembled monolayers) that employs supercritical CO2 drying and chemical vapor deposition to address many of the concerns associated with release stiction. VSAMS is attractive due to its process benefits, which include increased throughput, reduced waste, and most importantly, it can be easily scaled to full wafer production. It is also attractive because films produced by this method are uniform and very hydrophobic. The deposition step makes use of a class of compound that is particularly suited for vapor phase reactions, amino-functionalized silanes. The yield of microengine test devices coated with films made from amino-functionalized silanes was examined over an extended period. Their function was determined before and after the application of VSAMS. The advantage of using amino-functionalized silane precursors for VSAMS is related to the strength of the bond between the film and the polysilicon surface as evidenced by the fact that films made with these precursors are stable across the entire humidity scale.
MEMS surface-micromachining fabrication requires the use of many different tools to deposit thin-films, precisely define patterns using typical photolithography, and perform etching processes. As with any fabrication process there is inherent variation, which is acceptable when controlled within suitable limits. The ability to monitor and respond to this variation is paramount in maintaining a viable fabrication process. Electrostatic comb-drive resonators are candidate test structures used to validate uniformity in the MEMS fabrication process. Although directly dependent on mass and spring constant, a measure of their resonant frequencies generally provides a good indicator of both process repeatability and geometric variation.
In this study, sets of five graduated comb-drive resonator structures, located at each die on a ¼ wafer, were stimulated to resonant frequency using the “blur envelope” technique. This technique facilitates fast, straightforward, and repeatable resonant frequency measurements usually with a resolution of approximately 50-100 Hz. Wafer maps of resonant frequency versus die position for a ¼ wafer reveal a pattern with comb-drive resonator devices exhibiting highest resonant frequencies at the center and lowest at the perimeter of the wafer. Using a numerical model, coupled with discrete geometric measurements, a method was developed which links resonant frequency to fabrication parameters.
Anodic oxidation can be a catastrophic failure mechanism for MEMS devices that operate in high humidity environments. Shea and coworkers have shown that positively charged polysilicon traces can fail through a progressive silicon oxidation reaction whose rate depends critically on the surface conductivity over the silicon nitride. We have found a related anodic oxidation-based failure mechanism: progressive delamination of Poly 0 electrodes from silicon nitride layers, which then mechanically interfere with device function well before the electrode is fully oxidized. To explain this effect, we propose that the silicon oxide which initially forms at the electrode edge has insufficient strength to hold the local Poly 0 / silicon nitride interface together. This low-density silicon oxide also creates a bilayer system, which curls the edge of the 300 nm thick Poly 0 electrode away from the nitride. As delamination progresses more nitride surface is exposed and more of the interface is then attacked. This process continues cyclically until the electrode edge pushes against other device components, catastrophically and irreversibly interfering with normal operation. Additionally, we observe that the delamination only starts at electrode edges directly under cantilevers, suggesting the oxidation rate also depends on the perpendicular electric field strength.
The main thrust in any reliability work is identifying failure modes and mechanisms. This is especially true for the new technology of MicroElectroMechanical Systems (MEMS). The methods are sometimes just as important as the results achieved. This paper will review some of the methods developed specifically for MEMS. Our methodology uses statistical characterization and testing of complex MEMS devices to help us identify dominant failure modes. We strive to determine the root cause of each failure mode and to gain a fundamental understanding of that mechanism. Test structures designed to be sensitive to a particular failure mechanism are typically used to gain understanding. The development of predictive models follows from this basic understanding.
This paper will focus on the failure mechanism of wear and how our methodology was exercised to provide a predictive model. The MEMS device stressed in these studies was a Sandia-developed microengine with orthogonal electrostatic linear actuators connected to a gear on a hub. The dominant failure mechanism was wear in the sliding/contacting regions. A sliding beam-on-post test structure was also used to measure friction coefficients and wear morphology for different surface coatings and environments. Results show that a predictive model of failure-time as a function of drive frequency based on wear fits the functional form of the reliability data quite well, and demonstrates the benefit of a fundamental understanding of wear. The results also show that while debris of similar chemistry and morphology was created in the two types of devices, the dependence of debris generation on the operating environment was entirely different. The differences are discussed in terms of wear maps for ceramics, and the mechanical and thermal contact conditions in each device.
This paper reports on a set of parametric monitors for Sandia National Laboratories SUMMiT V (TM) (Sandia Ultra-planar Multi-level MEMS `i' Technology) five-level polysilicon surface micromachining process. Parametric monitors are typically used to monitor changes in the process due to either process drifts or intentional process changes. These parametric monitors are one of three types: electrical, electromechanical, or stress. In this paper we report on the design and characterization of these devices as well as how these devices are used to quantify the characteristics of the SUMMiT VTM process. We will demonstrate the use of these parametric devices in defining a baseline process.
Failure analysis (FA) tools have been applied to analyze tungsten coated polysilicon microengines. These devices were stressed under accelerated conditions at ambient temperatures and pressure. Preliminary results illustrating the failure modes of microengines operated under variable humidity and ultra-high drive frequency will also be shown. Analysis os tungsten coated microengines revealed the absence of wear debris in microengines operated under ambient conditions. Plan view imagine of these microengines using scanning electron microscopy (SEM) revealed no accumulation of wear debris on the surface of the gears or ground plane on microengines operated under standard laboratory conditions. Friction bearing surfaces were exposed and analyzed using the focused ion beam (FIB). These cross sections revealed no accumulation of debris along friction bear surfaces. By using transmission electro microscopy (TEM) in conjunction with electron energy loss spectroscopy (EELS), we were able to identify the thickness, elemental analysis, and crystallographic properties of tungsten coated MEMS devices. Atomic force microscopy was also utilized to analyze the surface roughness of friction bearing surfaces.
Electrostatic discharge (ESD) and electrical overstress (EOS) damage of Micro-Electrical-Mechanical Systems (MEMS) has been identified as a new failure mode. This failure mode has not been previously recognized or addressed primarily due to the mechanical nature and functionality of these systems, as well as the physical failure signature that resembles stiction. Because many MEMS devices function by electrostatic actuation, the possibility of these devices not only being susceptible to ESD or EOS damage but also having a high probability of suffering catastrophic failure doe to ESD or EOS is very real. Results from previous experiments have shown stationary comb fingers adhered to the ground plane on MEMS devices tested in shock, vibration, and benign environments [1,2]. Using Sandia polysilicon microengines, we have conducted tests to establish and explain the EDS/EOS failure mechanism of MEMS devices. These devices were electronically and optically inspected prior to and after ESD and EOS testing. This paper will address the issues surrounding MEMS susceptibility to ESD and EOS damage as well as describe the experimental method and results found from EDS and EOS testing. The tests were conducting using conventional IC failure analysis and reliability assessment characterization tools. In this paper we will also present a thermal model to accurately depict the heat exchange between an electrostatic comb finger and the ground plane during an ESD event.
Characterization tools have been developed to study the performance characteristics and reliability of surface micromachined actuators. These tools include: (1) the ability to electrically stimulate or stress the actuator, (2) the capability to visually inspect the devices in operation, (3) a method for capturing operational information, and (4) a method to extract performance characteristics from the operational information. Additionally, a novel test structure has been developed to measure electrostatic forces developed by a comb drive actuator.
Failure analysis tools have been applied to analyze failing polysilicon microengines. These devices were stressed to failure under accelerated conditions in both oxidizing and non-oxidizing environments. The dominant failure mechanism of these microengines was identified as wear of rubbing surfaces. This often results in either seized microengines or microengines with broken pin joints. Analysis of these failed polysilicon devices found that wear debris was produced in both oxidizing and non-oxidizing environments. By varying the relative percent humidity (%RH), we observed an increase in the amount of wear debris with decreasing humidity. Plan view imaging under scanning electron microscopy revealed build-up of wear debris on the surface of microengines. Focused ion beam (FIB) cross sections revealed the location and build-up of wear debris within the microengine. Seized regions were also observed in the pin joint area using FIB processing. By using transmission electron microscopy in conjunction with energy dispersive x- ray spectroscopy and electron energy loss spectroscopy, we were able to identify wear debris produced in low (1.8% RH, medium and high (39% RH) humidities.
Using a microengine as the primary test vehicle, we have examined several aspects of characterization. Parametric measurements provide fabrication process information. Drive signal optimization is necessary for increased microengine performance. Finally, electrical characterization of resonant frequency and quality factor can be more accurate than visual techniques.
Using a microengine as the primary test vehicle, we have examined several aspects of characterization. Parametric measurements provide fabrication process information. Drive signal optimization is necessary for increased microengine performance. Finally, electrical characterization of resonant frequency and quality factor can be more accurate than visual techniques.
The reliability of microengines is a function of the design of the mechanical linkage used to connect the electrostatic actuator to the drive. We have completed a series of reliability stress tests on surface micromachined microengines driving an inertial load. In these experiments, we used microengines that had pin mechanisms with guides connecting the drive arms to the electrostatic actuators. Comparing this data to previous results using flexure linkages revealed that the pin linkage design was less reliable. The devices were stressed to failure at eight frequencies, both above and below the measured resonance frequency of the microengine. Significant amounts of wear debris were observed both around the hub and pin joint of the drive gear. Additionally, wear tracks were observed in the area where the moving shuttle rubbed against the guides of the pin linkage. At each frequency, we analyzed the statistical data yielding a lifetime (t50) for median cycles to failure and (sigma) , the shape parameter of the distribution. A model was developed to describe the failure data based on fundamental wear mechanisms and forces exhibited in mechanical resonant systems. The comparison to the model will be discussed.
The first-ever reliability stress test on surface micromachined microengines developed at Sandia National Laboratories has been completed. We stressed 41 microengines at 36,000 RPM and inspected the functionality at 60 RPM. We have observed an infant mortality region, a region of low failure rate, and no signs of wearout in the data. The reliability data are presented and interpreted using standard reliability methods. Failure analysis results on the stressed microengines are presented. In our effort to study the reliability of MEMS, we need to observe the failures of large numbers of parts to determine the failure modes. To facilitate testing of large numbers of micromachines, we designed and built an automated system that has the capability to simultaneously test 256 packaged micromachines. The Sandia high volume measurement of micromachine reliability system has computer controlled positioning and the capability to inspect moving parts. The development of this parallel testing system is discussed in detail.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.